IC 4 ME 2. Proceedings. International Conference on Computer, Communication, Chemical, Materials & Electronic Engineering.

Transcription

1 Proceedings International Conference on Computer, Communication, Chemical, Materials & Electronic Engineering IC 4 ME 2 24~25 March 2016 Faculty of Engineering University of Rajshahi, Bangladesh

2 Proceedings

3 ISBN:

4 Foreword On behalf of the IC 4 ME , I warmly invite you to the International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering, IC 4 ME , at the green premise of the University of Rajshahi during 24~25 March, The IC 4 ME is the second of its kind hosted by the Faculty of Engineering of University of Rajshahi. The IC 4 ME continues the successful format from previous years ICMEIE While having the same overarching goal of presenting cutting-edge results, ideas, techniques, and theoretical advances in the mentioned theme, the IC 4 ME is separately tasked by focusing on emerging topics that complement the areas covered by the main technical program. The objective of organizing this conference is to bring together leading scientists, researchers and scholars to exchange and share their experiences and research results about all aspects of Electrical, Electronics, Communication, Chemical Engineering and IT, and discusses the practical challenges encountered and the solutions adopted. The international character of this meeting is illustrated by the participation of researchers from Fiji, Germany, India, Japan, Malaysia, Nepal, and Sri Lanka. Among the huge number of submissions only 56% have been accepted for publication. The conference includes 6 keynote speeches, 2 invited papers and 89 contributed papers distributed in 1 plenary session and 15 oral sessions. The information presented herein should help open up new avenues for research and provide researchers of the mentioned themes with new ideas to help them improve their production efficiency. We thank the reviewers from various countries who did this thankless job even in their busy schedule. The editorial team of this book deserves special thanks for their outstanding efforts in reviewing and preparing the abstracts for publication. Sponsorship for this meeting is an important feature of its success. On behalf of the organizing committee of IC 4 ME , we thank the University Grants Commission of Bangladesh for their support to promote the meeting. Finally, we would like to thank the presenters for their willingness to share their latest research and ideas. Without their efforts, this conference would not be possible. We hope you will enjoy your staying at very beautiful campus of the University of Rajshahi. Abu Bakar Md. Ismail, PhD Chair International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering, IC 4 ME Faculty of Engineering, University of Rajshahi Rajshahi 6205, Bangladesh

5 Contents Program Schedule 08 Organizing Committee 10 International Advisory Committee 10 Technical Program Committee 12 Reception Committee 14 Budget, Finance and Sponsor Committee 14 Registration Committee 15 Publication Committee 15 Decoration and Discipline Committee 16 Accommodation Committee 16 Social Event and Tour Committee 17 Food and Refreshment Committee 17 Paper ID Title Keynote Keynote Simulation of Normal Incidence Sound Absorption Coefficients of Perforated Panels with/without Glass Wool by Transmission Line Parameters in a two-port Network Takayoshi Nakai Settlement of Crystalline Structure of Group II-VI Semiconductor Nanoparticles by Profile Refinement Technique and Size Determination by Tight Binding Model Surendra K. Gautam Keynote Readerless RFID Transponder: A Concept 19 Mamun Bin Ibne Reaz Keynote Science, Technology and Innovation: For Engineering Our Future 19 Aravind Chinchure Keynote Progress of Low-Temperature Fabrication Technologies of Thin Film Transistors for Preservation of Global Environment Susumu Horita Keynote Factors that Facilitate and Impede Hitherto Untried R&D in Engineering 20 Invited Sisil Kumarawadu Water-in-oil Microemulsions as Nanoreactors to Prepare Nanoparticles with Tunable Electrical, Optical, and Antibacterial Properties Md. Abu Bin Hasan Susan, M. Muhibur Rahman and M. Yousuf A Mollah Invited Physics and Technology in Radiation Oncology and Imaging 21 Golam Abu Zakaria 5 Fabrication of Cardiac Biomarker Immunosensor based on Three Different Electrode Surfaces and Contrasting Their Efficiencies Payal Gulati, Prabhash Mishra and Safiul Islam 9 Simulation of the Electrical Activity of Cardiac Tissue by Finite Element Method 27 Tanzina Rahman and Md. Rezaul Islam 10 Baking of Ilmenite on Moistening with Sulfuric Acid followed by Leaching with Dilute Sulfuric Acid Solution Ranjit K. Biswas, Aneek K.Karmakar, Jinnatul Ara and Muhammad A. Gafur 11 Thermal Treatment of Ilmenite on Moistening with Concentrated HF followed by Leaching with Dilute Sulfuric Acid Solution Ranjit K. Biswas, Mohammad A. Habib, Aneek K. Karmakar and Mohammad J. Alam

6 Paper ID Title 12 Solvent Extraction of V(V) from Nitrate Medium by Tri-n-Octylamine Dissolved in Kerosene Ranjit K. Biswas, Aneek K. Karmakar and Mottakin 13 Kinetics of Extraction of Ti(IV) from SO 2-4 Medium by Cyanex 301 Dissolved in Kerosene Ranjit K. Biswas and Aneek K. Karmakar 16 Production and Improvement of Waste Tire Pyrolysis Oil to be Utilized as Biofuel in Diesel Engine Md. Nurul Islam and Md. Rafsan Nahian 24 Biological Evaluation of Radiotherapy Treatment Plan for Different Field Techniques in 3-Dimensional Conformal Radiotherapy (3DCRT). Kausar A, Azhari H A and Zakaria G A 25 Design of a Linearly Polarized Multi-Band Transmission Line Feed Microstrip Patch Antenna for Wireless Communications Sheikh Dobir Hossain, Md. Khalid Hossain and Rebeka Sultana 26 Design and Fabrication of an Unmanned Video Transmitting Tele-Bot using 3G GSM Network Md. Mamunoor Islam and Mehdi Hasan Chowdhury 29 Effect of Sintering Temperature on Nb+Nd Doped Bismuth Ferrite 63 Sadia Tasnim Mowri, M A Gafur, Quazi Delwar Hossain, Aninda Nafis Ahmed and Muhammad Shahriar Bashar 30 Silicon Nanocrystals Rich Lanthanum Fluoride Films for Future Electronic Devices 67 Ferdous Rahman, Sk. Rashel Al Ahmed, Md. Golam Saklayen and Abu Bakar Md. Ismail 41 Study on the Displacement Effect at Cylindrical Ionization Chamber with different Radii in High Energy Photon of Flat Beam and True Beams Kumaresh Chandra Paul, Guenther H Hartmann, Enamul Hoque and Golam Abu Zakaria 42 Electrical and Optical Properties of Cu-Nanoparticles- Doped α-fe 2 O 3 Thin Film Spin- Coated on Glass Substrate Sanjida Ferdous, Afroza Yasmin, Jinia Sultana and Abu Bakar Md. Ismail 43 Study on Morphological Properties of Cu-NPs Doped α-fe 2 O 3 Thin Film deposited on Glass Substrate Jinia Sultana, Afroza Yasmin, Sanjida Ferdous and Abu Bakar Md. Ismail 47 MRI Segmentation using Fuzzy C-Means Clustering and Bidimensional Empirical Mode Decomposition Gulam Sarwar Chuwdhury, Md. Khaliluzzaman and Md. Rashed-Al-Mahfuz 48 Wear and Morphological Behavior of Electron Beam Dose Irradiated Polyoxymethylene 86 Copolymer (POM-C) Md. Shahinur Rahman, Heon-Ju Lee, Muhammad Sifatul Alam Chowdhury and Konstantin Lyakhov 49 Study of Structural and Optical Properties of Pyrolised CuO Films 90 M. Majhar, S. Ahmed, M. Mozibur Rahman and M. K. R. Khan 52 Algorithm for Performance Appraisal using CAW Method 94 M. Z. Ahsan and Md. Mamun-Ur-Rashid Khandker 56 Friction and Morphological Properties of Ion Implanted Polyoxymethylene Copolymer (POM-C) Md. Shahinur Rahman, Md. Mehedi Hasan, Muhammad Sifatul Alam Chowdhury and Konstantin Lyakhov

7 Paper ID Title 57 Fabrication and Mechanical Characterization of Aluminium- Rice Husk Ash Composite by Stir Casting Method Adnan Adib Ahamed, Rashed Ahmed, Md. Benzir Hossain, Masum Billah 58 Analysis of Annual and Seasonal Precipitation Concentration Index (PCI) of North- Western Region of Bangladesh Ahsan Habib Rasel, Md Monirul Islam and Mumnunul Keramat 62 Bitwise Template Fusion of Noisy Images for Enhanced IRIS Recognition System 111 Sohel Ahammed and Biprodip Pal 65 Kinetics of Extraction of Ti(IV) from Sulfate Medium by Cyanex Ranjit K. Biswas and Aneek K. Karmakar 66 Autonomous Human Face and Tracking System with Variant Poses, Blur and Illumination 120 Md. Zweel Rana, Monimul Islam and Mohiuddin Ahmad 67 Electrochemical Corrosion Characterization of Artificially Aged Al-6Si-0.5Mg (-1Cu) Alloys in Sodium Chloride Solution Abul Hossain, M. A. Gafur, Fahmida Gulshan and A S W Kurny 69 Effects of Inclusions on the Mechanical Properties of Structural Steel Reinforced Bars 129 Abul Hossain, Fahmida Gulshan and A S W Kurny 72 Utilizing Solar Energy in the Filling Stations of Bangladesh: Technical and Economical 133 Representation Mohammad Jalal Uddin, Muhammad Sifatul Alam Chowdhury, Md. Ridwanul Karim, Md. Arman Uddin and Md. Bakiuzzaman 74 Conversion of Prawn Shell Waste into Value Added Products for Textile Finishes 137 Md. Mofakkharul Islam, Firoz Ahmed and Md. Ibrahim H. Mondal 75 Textile Performance of Functionalized Cotton Fibre with Silane Coupling Agents 141 Md. Khademul Islam, Md. Abdul Aziz and Md. Ibrahim H. Mondal 77 Synthesis and Characterization of Hydrogels from Cellulosic Materials for Green Adsorbent Products Md. Obaidul Haque, Md. Abu Sayeed and Md. Ibrahim H. Mondal 83 Study the Encryption Techniques for Multimedia 149 Md. Martuza Ahamad and Md. Ibrahim Abdullah 84 Influence of Deposition Temperature on the Deposition of SiO 2 Films from Reaction of Silicone Oil Vapor and Ozone Gas Arifuzzaman Rajib, Susumu Horita, Atowar Rahman and Abu Bakar Md. Ismail 85 An Improved Representation of Audio Signal in Time-Frequency Plane 158 Kazi Mahmudul Hassan, Ekramul Hamid and Takayoshi Nakai 86 On the Optimization of Number of Message Copies for Multi-Copy Routing Protocols in Scalable Delay-Tolerant Networks Md. Sharif Hossen and Muhammad Sajjadur Rahim 90 Emotional Bangla Speech Signals Classification using K-NN 168 Md. Tohidul Islam, Md. Ekramul Hamid and Somlal Das 92 Content Based Image Searching Using Multidimensional MSF 172 Saiful Islam, Ekramul Hamid and Emdadul Haque 93 Silicon Nanocrystals based Schottky Junction Solar Cell Fabrication and Characterization 177 A.T.M. Saiful Islam, Md. Enamul Karim, Arifuzzaman Rajib and Abu Bakar Md. Ismail 95 Fabrication and Characterization of α-fe 2 O 3 Homo-Junction Photocathode for Efficient Solar Water Splitting Arifuzzaman Rajib, Atowar Rahman, A.T.M. Saiful Islam and Abu Bakar Md. Ismail

8 Paper ID Title 97 A Practical Approach to Spectrum Analyzing Unit using RTL-SDR 185 Md. Habibur Rahman and Md. Mamunoor Islam 101 Fabrication of Bismuth Ferrite Multiferroic Perovskite Nanoparticles Using an Aqueous Organic Gel Route Mayeesha Haque, M. S. Parvez, M. S. Islam, M.A. Hakim and M. A. Gafur 105 Analysis of GLDAS data for Estimating and Distribution of Evapotranspiration and Rainfall over Bangladesh Md Ataur Rahman, Md Mainul Islam Mamun and Md Monirul Islam 113 Robustification of Logistic Classifer for Binary Classification in Microarray Gene 199 Expression Data Md. Shahjaman, Md. Mushfiqur Rahman, Md. Manir Hossain Mollah, Anjuman Ara Begum, S. M. Shahinul Islam and Md. Nurul Haque Mollah 115 Molecular Evolutionary Analysis of a-defensin Peptides in Vertebrates 203 Arafat Rahman, M Sahidul Islam, Otun Saha and Titon Chandra Saha 117 Microencapsulated Probiotic Bacteria Protect the Spoilage of Functional Foods 207 Md. Symoom Hossain, Md. Abdul Alim Al-Bari, Mir Imam Ibne Wahed and Md. Anwar Ul Islam 124 Preparation of Highly Cross-Linked Magnetic Polymer Composite Particles and Application in the Separation of Arsenic from Water M. K. Sharafat, H. Ahmad, M. A. Alam and M.M. Rahman 125 Preparation of Hydrophobic Poly((lauryl methacrylate)-coated Magnetic Nano- Composite Particles and their Application as Adsorbents for Organic Polutants Rukhsana Shabnam and Hasan Ahmad 126 Statistical Methods for Functional Analysis of Metagenomes 219 Zobaer Akond and Md. Nurul Haque Mollah 127 Simulation of Microalgae and CO 2 Flow Dynamics in a Tubular Photobioreactor and Consequent Effects on Microalgae Growth Saumen Barua, Mohammad Morshed Alam and Ujjwal Kumar Deb 137 Evaluation of PCA in spatial, frequency and wavelet domains for face recognition 229 Samsi Ara and M. Babul Islam 139 Time-Frequency Coherence Analysis of Multichannel Electroencephalography Signals 233 using Synchrosqueezing Transform Md. Sujan Ali, Mst. Jannatul Ferdous, Md. Ekramul Hamid and Md. Khademul Islam Molla 142 Experimental Study on Optical Characterization of Mono Crystalline Silicon Solar Cell 237 Nusrat Chowdhury, Md. Abdur Rafiq Akand and Zahid Hasan Mahmood 143 Canonical Correlation Analysis for SNP based Genome-Wide Association Studies 241 Atul Chandra Singha, Arafat Rahman, Md. Jahangir Alom and Md. Nurul Haque Mollah 148 Frequency Recognition of SSVEP for BCI Implementation using Canonical Correlation 245 Analysis with Adaptive Reference Signals Shalauddin Ahamad Shuza, Md. Rabiul Islam, Md. Kislu Noman and Md. Khademul Islam Molla 151 Expert Reviewer Detection from Online Experiential Product Reviews 249 Atiquer Rahman Sarkar 153 FPGA based Pulse Oximeter using VHDL 253 Farhana Binte Sufi, Md. Fahmidur Rahman and Md. Maruful Islam

9 Program Schedule Conference Kit Distribution 23 March, :00 am 02:00 pm 05:00 pm 08:00 pm 24 March, :00 am 08:50 am Inaugural Session 24 March, :00 am 9:45 am Keynote Session 24 March, :00 am 11:30 am 11:45 am 01:15 pm Invited Talk 24 March, :00 pm 04:00 pm Technical Session Conference Room Department of CSE 4 th Science Building University of Rajshahi Senate Bhaban University of Rajshahi Senate Bhaban University of Rajshahi Senate Bhaban University of Rajshahi Senate Bhaban University of Rajshahi 24 March, :30 pm 06:00 pm 4 th Science Building 25 March, :00 am 12:30 pm 4 th Science Building Keynote Session: 24 March, :00 am - 1:15 pm Room # Senate Bhaban Chair: Prof. M. Abdus Sobhan, University of Rajshahi, Bangladesh Title Simulation of Normal Incidence Sound Absorption Coefficients of Perforated Panels with/without Glass Wool by Transmission Line Parameters in a Two-Port Network Takayoshi Nakai Settlement of Crystalline Structure of Group II-VI Semiconductor Nanoparticles by Profile Refinement Technique and Size Determination by Tight Binding Model Surendra K. Gautam Readerless RFID Transponder: A Concept Mamun Bin Ibne Reaz Science, Technology and Innovation: For Engineering Our Future Aravind Chinchure Progress of Low-Temperature Fabrication Technologies of Thin Film Transistors for Preservation of Global Environment Susumu Horita Factors that Facilitate and Impede Hitherto Untried R&D in Engineering Sisil Kumarawadu 8

10 Invited Talk: 24 March, :00 pm - 4:00 pm Room # Senate Bhaban Chair: Prof. Mamun Bin Ibne Riaz, Kebangsaan Universiti, Malaysia Title Water-in-oil Microemulsions as Nanoreactors to Prepare Nanoparticles with Tunable Electrical, Optical, and Antibacterial Properties Md. Abu Bin Hasan Susan, M. Muhibur Rahman and M. Yousuf A Mollah Physics and Technology in Radiation Oncology and Imaging Golam Abu Zakaria Technical Session Session Chair Date & Time Paper ID Venue 1A 1B 1C 1D 2A 2B 2C 2D 3A 3B 3C 3D Prof. Khademul Islam Molla University of Rajshahi Dr. Surendra Kumar Gautam Tribhuvan University, Nepal Dr. Susumu Horita JAIST, Japan Prof. Abdur Razzak RUET Prof. Md. Rabiul Islam RUET Prof. C. M. Mostofa University of Rajshahi Dr. Riazul Islam University of Dhaka Prof. Abu Reza University of Rajshahi Prof. Golam Abu Zakaria Anhalt University of Applied Sciences, Germany Prof. Ranjit K Biswas University of Rajshahi Prof. Md. Abu Bin Hasan Susan University of Dhaka Prof. Mamnunul Keramat University of Rajshahi 24 March, :30 pm 6:00 pm 25 March, :00 am 10:30 am 25 March, :00 am 12:30 pm 62, 85, 90, 92, 137, 148 Room # th Science Building 10, 11, 16, 65, 74, Room # th Science Building Room # 122 3, 14, 23, 29, 30, 42 4 th Science Building 25, 26, 83, 97, 151, Room # th Science Building 45, 47, 52, 53, 66, Room # th Science Building 12, 13, 67, 76, 77, Room # th Science Building 43, 48, 49, 56, 69, Room # th Science Building 113, 126, 129, 143 Room # th Science Building 5, 9, 24, 41, 60, 139 Room # th Science Building 102, 115, 117, 125, 127, 157 Room # th Science Building 93, 95, 124, 156, Room # , th Science Building 58, 72, 86, 105, 142, Room # th Science Building 9

11 Organizing Committee Chair : Abu Bakar Md. Ismail, Ph.D Dean, Faculty of Engineering & Professor Dept. of Applied Physics & Electronic Engineering University of Rajshahi, Bangladesh Co-Chair : Professor Dr. Ranjit Kumar Biswas Dept. of Applied Chemistry & Chemical Engineering University of Rajshahi, Bangladesh Secretary : Professor Mamun Ur Rashid Khandker Dept. of Applied Physics & Electronic Engineering University of Rajshahi, Bangladesh Treasurer : Professor Dr. Dipankar Das Dept. of Information & Communication Engineering University of Rajshahi, Bangladesh Members: Chairman, Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Chairman, Dept. of Applied Chemistry & Chemical Engineering, University of Rajshahi Chairman, Dept. of Computer Science & Engineering, University of Rajshahi Chairman, Dept. of Information & Communication Engineering, University of Rajshahi Chairman, Dept. of Materials Science & Electronic Engineering, University of Rajshahi Professor Dr. M. Mozaffor Hossain, Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Professor Dr. M. Abdus Sobhan Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Professor Dr. Mumnunul Keramat Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Professor Dr. M. Rostom Ali Dept. of Applied Chemistry & Chemical Engineering, University of Rajshahi Professor Dr. M. Mamunur Rashid Talukder Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Professor Dr. M. Khademul Islam Molla Dept. of Computer Science & Engineering, University of Rajshahi Engr. S. M. Rezaul Kabir Principal, BCMC College of Engineering & Technology, Jessore International Advisory Committee Professor Dr. Surendra K. Gautam Tri-Chandra Campus Tribhuvan University, Nepal Professor Dr. Sisil Kumarawadu Dept. of Electrical Engineering University of Moratuwa, Sri Lanka Dr. D.M.G. Preethichandra Discipline Leader Electrical Engineering School of Engineering and Technology Central Queensland University, Australia 10

12 Professor Dr. Keikichi Hirose Dept. of Information and Communication Engineering The University of Tokyo, Japan Dr. Keiji Nagai Associate Professor Chemical Resource Laboratory Tokyo Institute of Technology, Japan Dr. Wu Ping Associate Professor Division of Engineering Product Development Singapore University of Technology and Design, Singapore Professor Dr. Md. Mamun Bin Ibne Reaz Dept. of Electrical, Electronic and Systems Engineering University Kebangsaan Malaysia, Malaysia Professor Dr. Hartmut Baerwolff Dept. of Analog & Optoelectronics Cologne University of Applied Sciences, Germany Professor Dr. Tomokazu Iyoda Chemical Resource Laboratory Tokyo Institute of Technology, Japan Professor Dr. Norihiko Kamata Dept. of Functional Materials Science Graduate School of Science and Engineering Saitama University, Japan Professor Yousuke Nakashima Plasma Research Center University of Tsukuba, Japan Dr. Susumu Horita School of Materials Science Japan Advanced Institute of Science & Technology, Japan Professor A B M Shawkat Ali Department of Computer Science and Information Technology Dean, School of Science and Technology The University of Fiji, Fiji Dr. Golam Zakaria Dept. of Medical Radiation Physics University of Cologne, Germany Anis Haque, Ph.D, P.Eng. Senior Instructor Fellow, Institute for Sustainable Energy, Environment and Economy (ISEEE) Associate Director of Students Dept. of Electrical and Computer Engineering University of Calgary, Canada Dr. Mohammad Abdul Fazal Senior Lecturer University of Malaya, Malaysia 11

13 Technical Program Committee Co-Chair : Professor Dr. C. M. Mostofa Dept. of Applied Chemistry & Chemical Engineering University of Rajshahi, Bangladesh Secretary : Professor Dr. Shamim Ahmad Dept. of Computer Science & Engineering University of Rajshahi, Bangladesh Members: Professor Dr. Md. Ariful Islam Nahid Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Professor Dr. M. Shameem Ahsan Dept. of Applied Chemistry & Chemical Engineering, University of Rajshahi Professor Dr. M. Taufiq Alam Dept. of Applied Chemistry & Chemical Engineering, University of Rajshahi Professor Dr. Ekramul Hamid Dept. of Computer Science & Engineering, University of Rajshahi Professor Dr. A. K. M. Akhter Hossain Dept. of Computer Science & Engineering, University of Rajshahi Professor Dr. Md. Nurul Haque Mollah Dept. of Statistics, University of Rajshahi Professor Dr. Md. Ziaur Rahman Khan Dept. of Electrical & Electronic Engineering, BUET Professor Dr. Zahid Hasan Mahmood Dept. of Electrical & Electronic Engineering, Dhaka University Professor Dr. M. M. A. Hashem Dept. of Computer Science & Engineering, KUET Professor Dr. Mohammad Abdul Goffar Khan Dept. of Electrical & Electronic Engineering, RUET Professor Dr. Md. Rafiqul Islam Dept. of Electrical & Electronic Engineering, KUET Professor Dr. Shahid Uz Zaman Dept. of Computer Science & Engineering, RUET Professor Dr. S. M. Abdur Razzak Dept. of Electrical & Electronic Engineering, RUET Professor Dr. Md. Anisur Rahman Dept. of Computer Science & Engineering, Khulna University Professor Dr. Mohammad Shahidur Rahman Dept. of Computer Science & Engineering, Shahjalal University of Science & Technology Professor Dr. Mohammad Iqbal Dept. of Industrial & Production Engineering, Shahjalal University of Science & Technology Professor Dr. Md. Mahbubur Rahman Dept. of Computer Science & Engineering, MIST Professor Dr. Mortuza Ali Dept. of Electrical & Electronic Engineering, Eastern University, Dhaka Professor Rezaul Karim Mazumder Dept. of Electronics & Telecommunication Engineering, ULAB, Dhaka 12

14 Professor Dr. Jugal Krishna Das Dept. of Computer Science & Engineering, Jahangirnagar University Professor Dr. Farid Ahmed Dept. of Physics, Jahangirnagar University Professor Dr. Md. Shahjahan Dept. of Applied Physics, Electronics & Communication Engineering, BSMRSTU Professor Dr. M. Mahbubur Rahman Dept. of Information & Communication Engineering, Islamic University Professor Dr. M. Maniruzzaman Dept. Applied Chemistry & Chemical Engineering, Islamic University Professor Dr. Momtazul Islam Dept. Applied Physics, Electronics & Communication Engineering, Islamic University Dr. Bilkis Ara Begum Head, Chemistry Division, BAEC, Dhaka Dr. Shamshad Begum Quraishi Chief Scientific Officer, Chemistry Division, BAEC, Dhaka Dr. Samia Tabassum Senior Scientific Officer, BCSIR, Dhaka Dr. M. Babul Islam Associate Professor, Dept. of Applied Physics & Electronic Engg., University of Rajshahi Dr. Md. Shafiqul Islam Associate Professor, Dept. of Nuclear Engineering, Dhaka University Dr. Md. Atowar Rahman Associate Professor, Dept. of Applied Physics & Electronic Engg., University of Rajshahi Dr. Md. Nur-Al-Safa Bhuiyan Associate Professor, Dept. of Information & Communication Engg., University of Rajshahi Dr. Sabbir Ahmed Associate Professor, Dept. of Information & Communication Engg., University of Rajshahi Dr. Md. Emdadul Haque Associate Professor, Dept. of Information & Communication Engg., University of Rajshahi Dr. Asadul Hoque Associate Professor, Dept. of Materials Science & Engineering, University of Rajshahi Dr. M. Anwarul Kabir Bhuiya Associate Professor, Dept. of Materials Science & Engineering, University of Rajshahi Dr. G. M. Shafiur Rahman Associate Professor, Dept. of Materials Science & Engineering, University of Rajshahi Mr. Subrata Pramanik Associate Professor, Dept. of Computer Science & Engineering, University of Rajshahi Mr. Muhammad Sajjadur Rahim Associate Professor, Dept. of Information & Communication Engg., University of Rajshahi Dr. N. A. Ruhul Azad Lecturer & Programme Leader for Engineering Foundation, LBIC, Brunel University, UK Dr. Md. Iqbal Aziz Khan Assistant Professor, Dept. of Computer Science & Engineering, University of Rajshahi Mr. Mahboob Qaosar Assistant Professor, Dept. of Computer Science & Engineering, University of Rajshahi Mr. A. F. M. Mahbubur Rahman Assistant Professor, Dept. of Computer Science & Engineering, University of Rajshahi Mr. M. Rashed Al Mahfuz 13

15 Assistant Professor, Dept. of Computer Science & Engineering, University of Rajshahi Reception Committee Convener : Professor Rostom Ali Dept. of Applied Chemistry& Chemical Engineering University of Rajshahi Members: Chairman Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Chairman Dept. of Applied Chemistry & Chemical Engineering, University of Rajshahi Chairman Dept. of Computer Science & Engineering, University of Rajshahi Chairman Dept. of Information & Communication Engineering, University of Rajshahi Chairman Dept. of Materials Science & Engineering, University of Rajshahi Chairman Dept. of Electrical & Electronic Engineering, University of Rajshahi Professor M. Mozaffor Hossain Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Professor Dr. M. Abdus Sobhan Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Professor Dr. Mumnunul Keramat Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Budget, Finance and Sponsor Committee Convener : Professor Abu Bakar Md. Ismail Dean, Faculty of Engineering, University of Rajshahi Members: Professor Md. Ekramul Hamid Dept. of Computer Science & Engineering, University of Rajshahi Professor M. Ahsan Habib Dept. of Applied Chemistry & Chemical Engineering, University of Rajshahi Dr. G. M. Shafiur Rahman Dept. of Materials Science & Engineering, University of Rajshahi Dr. Nur Al Safa Bhuyan Dept. of Information & Communication Engineering, University of Rajshahi 14

16 Registration Committee Convener : Professor Mamun Ur Rashid Khandker Dept. Applied Physics & Electronic Engineering, University of Rajshahi Members: Professor Shameem Ahsan Dept. of Applied Chemistry & Chemical Engineering, University of Rajshahi Dr. M. Babul Islam Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Muhammad Sajjadur Rahim Dept. of Information & Communication Engineering, University of Rajshahi Dr. Sabbir Ahmed Dept. of Information & Communication Engineering, University of Rajshahi Dr. Anwarul Kabir Bhuiya Dept. of Materials Science & Engineering, University of Rajshahi Dr. M. Iqbal Aziz Khan Dept. of Computer Science & Engineering, University of Rajshahi Mahboob Qaosar Dept. of Computer Science & Engineering, University of Rajshahi Publication Committee Convener : Professor Shamim Ahmad Dept. of Computer Science & Engineering, University of Rajshahi Members: Dr. M. Babul Islam Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Muhammad Sajjadur Rahim Dept. of Information & Communication Engineering, University of Rajshahi Dr. Md. Emdadul Haque Dept. of Information & Communication Engineering, University of Rajshahi Dr. Sabbir Ahmed Dept. of Information & Communication Engineering, University of Rajshahi Dr. M. Asadul Hoque Dept. of Materials Science & Engineering, University of Rajshahi Dr. Anik K Karmakar Dept. of Applied Chemistry & Chemical Engineering, University of Rajshahi Foez Ahmed Dept. of Information & Communication Engineering, University of Rajshahi A.F.M. Mahbubur Rahman Dept. of Computer Science & Engineering, University of Rajshahi Farhana Binte Sufi Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Sangeeta Biswas Dept. of Computer Science & Engineering, University of Rajshahi 15

17 Decoration and Discipline Committee Convener : Professor Md. Ekramul Hamid Dept. of Computer Science & Engineering, University of Rajshahi Members: Professor Rubaiyat Yasmin Dept. of Information & Communication Engineering, University of Rajshahi Dr. M. Abdul Matin Dept. of Materials Science & Engineering, University of Rajshahi Dr. Halida Homyara Dept. of Information & Communication Engineering, University of Rajshahi Dr. Sinthia Shabnam Mou Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Mahboob Qaosar Dept. of Computer Science & Engineering, University of Rajshahi Sanjoy Kumar Chakravarty Dept. of Computer Science & Engineering, University of Rajshahi Sangeeta Biswas Dept. of Computer Science & Engineering, University of Rajshahi Shammi Farhana Islam Dept. of Materials Science & Engineering, University of Rajshahi Saiful Islam Dept. of Applied Chemistry & Chemical Engineering, University of Rajshahi Mousumi Haque Dept. of Information & Communication Engineering, University of Rajshahi Accommodation Committee Convener : Dr. M. Babul Islam Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Members: Professor Shamim Ahmad Dept. of Computer Science & Engineering, University of Rajshahi Professor Md. Ekramul Hamid Dept. of Computer Science & Engineering, University of Rajshahi Mirza A. F. M. Rashidul Hasan Dept. of Information & Communication Engineering, University of Rajshahi Dr. Md. Asadul Haque Dept. of Materials Science & Engineering, University of Rajshahi Dr. Sabbir Ahmed Dept. of Information & Communication Engineering, University of Rajshahi Md. Morshedul Arefin Dept. of Computer Science & Engineering, University of Rajshahi Khairul Islam Dept. of Applied Chemistry & Chemical Engineering, University of Rajshahi Sajjadul Kabir Dept. of Computer Science & Engineering, University of Rajshahi 16

18 Social Event and Tour Committee Convener : Professor Shaikh Enayet Ullah Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Members: Professor Syed Mustafizur Rahman Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Dr. M. Atowar Rahman Dept. of Applied Physics & Electronic Engineering, University of Rajshahi Dr. Anwarul Kabir Bhuiya Dept. of Materials Science & Engineering, University of Rajshahi Utpalananda Chowdhury Dept. of Computer Science & Engineering, University of Rajshahi Sajjadul Kabir Dept. of Computer Science & Engineering, University of Rajshahi Food and Refreshment Committee Convener: Professor Md. Rezaul Islam Dept. Applied Physics & Electronic Engineering, University of Rajshahi Members: Professor Dr. M. Mamunur Rashid Talukder Dept. of Applied Physics & Electronic Engineering Professor Abu Zafor Muhammad Touhidul Islam Dept. of Information & Communication Engineering, University of Rajshahi Dr. Aurangzib Md. Abdur Rahman Dept. of Information & Communication Engineering, University of Rajshahi Dr. Anwarul Kabir Bhuiya Dept. of Materials Science & Electronic Engineering, University of Rajshahi Dr. M. Hasnat Kabir Dept. of Information & Communication Engineering, University of Rajshahi Md. Rokonuzzaman Dept. of Computer Science & Engineering, University of Rajshahi Abu Mohammad Dept. of Materials Science & Electronic Engineering, University of Rajshahi Dilip Kumar Sarker Dept. of Applied Chemistry & Chemical Engineering, University of Rajshahi 17

19 Simulation of Normal Incidence Sound Absorption Coefficients of Perforated Panels with/without Glass Wool by Transmission Line Parameters in a two-port Network Takayoshi Nakai Professor Department of Electrical and Electronic Engineering Shizuoka University, Japan ABSTRACT This paper describes simulation of normal incidence sound absorption coefficients of perforated panels by transmission line parameters in a two-port network. Maa and Sakagami have investigated micro perforated panels, MPP. But their theories can treat only near 1% perforation rates of perforated panels with back cavities. If sound propagates as a plane wave, sound propagation can be represented as transmission line parameters in a two-port network. Perforated panels, back cavities, and glass wool absorption materials are represented as matrix of transmission line parameters, respectively. Transmission line parameters of a perforated panel with a back cavity are calculated as multiplication of their matrices. An input impedance can be calculated from the transmission line parameters. A normal incident absorption coefficient is calculated from the input impedance. Holes of the perforated panels have losses of viscous friction and thermal conduction at their walls. Simulations are done in the condition of 0.25 mm to 5 mm diameters of holes, 0.25 % to 25 % perforation rates, 0.5 mm to 5 mm thickness of the perforated panels with back cavities in which there are or are not glass wool absorption materials. The results of these simulations are good agreements with the results of our measurements by transfer function method except in the condition of more than 1 mm diameter of holes. Settlement of Crystalline Structure of Group II-VI Semiconductor Nanoparticles by Profile Refinement Technique and Size Determination by Tight Binding Model Dr. Surendra Kumar Gautam Department of Chemistry Tribhuvan University, Nepal ABSTRACT Group II-VI semiconductor nanoparticles have gained huge interest both in fundamental research and technical applications due to their unique optical and electrical properties. They have extensive range of applications in the field of opto-electronic devices such as light emitting diodes (LEDs), bio-sensors, photo-detectors, solar cells etc. The properties of such semiconductor nanoparticles are dependent on their crystal structure and size. The settlement of crystalline structure of those nanoparticles is done by profile refinement of X-ray diffraction (XRD) pattern. The size quantization effect and blueshift resulting in the change of band gap with crystalline size are studied from optical absorption spectra. Particle sizes are calculated by Tight Binding (TB) model. The verification of particle size and the crystalline structure of as-synthesized nanocrystals are further carried out from Transmission Electron Microscopy (TEM) images and Selected Area Electron Diffraction (SAED) patterns. 18

20 Readerless RFID Transponder: A Concept Mamun Bin Ibne Reaz Professor Department of Electrical Electronic and Systems Engineering Universiti Kebangsaan Malaysia Bangi, Selangor Malaysia ABSTRACT RFID is a technology that enables the non-contact, automatic and unique identification of objects using radio waves. It is projected that the RFID market will be worth more than US $25 billion in However, RFID system suffers from limited address space, local mobility, security and privacy. Above all, it is a monopoly business with few vendors, which are trying to dominate the market with proprietary standard of RFID reader. In order to overcome this major issue, we are proposing an RFID tag communicating with the existing wireless network interface card (WNIC) instead of the reader. In this novel RFID transponder, IPv6 address will be used to provide a universal identification number to the objects with seamless address space, global mobility and data security. The existing EPC (Electronic Product Code) of RFID will now directly map into IPv6 address by using auto configuration method. This talk will mainly focus on different key issues related to this readerless RFID system and suggests the mechanism of reducing significant cost, physical location detection and usage of global unique address. Science, Technology and Innovation: For Engineering Our Future Aravind Chinchure Chair Professor Centre for Enterpreneurship & Innovation Symbiosis International University Pune, India ABSTRACT Every third person in this world today subsists with income less than 2 dollars a day. Majority of these people live in Asian countries. The rising socio-economic inequality and climate change are posing biggest risk to the world. The world is also moving towards knowledge-based competencies and industries where the emphasis is not on tangible assets, but on intangible knowledge assets, which is good news for the developing countries. Today, world s major knowledge industries are based on telecommunication, microelectronics, new materials, and information technology. This offers a great opportunity to researchers from academia and industry from Asian countries to appropriately engineer the future of the nation and society by developing innovative solutions to cater to the needs of the people for the sustainable and inclusive development. To reap the full benefit of the emergence of knowledge-based industry requires building a new culture in academia and industry that seamlessly connects science, technology and innovation with relevance. This talk provides several examples of academic institutions, professors, researchers who have been able to successfully leverage science, technology and innovations for national and social good. I will also present new ideas that are emerging and debated in the world on the cost-conscious frugal science and technology. 19

21 Progress of Low-Temperature Fabrication Technologies of Thin Film Transistors for Preservation of Global Environment Dr. Susumu Horita Associate Professor School of Materials Science Japan Advanced Institute of Science and Technology, Japan ABSTRACT Low-temperature fabrication technology can contribute enhancement of preservation of global environment due to reduction of not only power or natural source consumption but also greenhouse effect. So, as an example, we have been studying on lowtemperature fabrication of poly-or microcrystalline silicon (poly-si) films on temperature-sensitive and cheap substrates. The appropriate applications of poly Si are thin-film transistor (TFT) in an active matrix flat panel display, solar cells, and so on because of its higher reliability and higher mobility. For future application, industry requires shorter annealing time, lower annealing temperature, more uniform electrical property such as mobility, threshold voltage, and so on in a whole substrate area. For meeting these requirements, we have been investing solid-phase crystallization of Si film on a glass by using pulsed laser annealing(pla) with crystallization-induction (CI) layer of yttria-stabilized zirconia [(ZrO 2 ) 1-x (Y 2 O 3 ) x :YSZ]. PLA method can crystallize Si films effectively at room temperature because of its short pulse duration time less than 10 ns. The YSZ CI layer can transfer its crystalline information to crystallized Si on a glass so that we can obtain higher and more uniform crystalline quality of Si films. As another important technology for TFT, we have low-temperature fabrication of insulator, e.g, oxide film, in particular, for gate layer. Our group uses atmospheric pressure CVD (AP-CVD) with silicone oil and ozone gas to obtain Si oxide. As you know, silicone oil is a safety material for human body, chemically stable, and cheap, and ozone is an indispensable substance for global environment. By using this technology, we can produce a SiO 2 film at 200 to C. In this conference, our previous and current research results on the above low temperature technologies are presented, including new two-step method in PLA for much improving film quality and characteristics of Si TFTs fabricated by the above techniques. Factors that Facilitate and Impede Hitherto Untried R&D in Engineering Dr. Sisil Kumarawadu Professor Department of Electrical Engineering University of Moratuwa, Sri Lanka ABSTRACT Completing a research & development (R&D) or postgraduate research project requires stamina, determination, and willingness to stretch your intellectual and emotional capabilities. An unmistakable initial momentum is vital to avoid stall out at some point when working on a research project as it may lead to feel as if there is no light at the end of the tunnel, or encounter disappointments or unexpected setbacks, or even 20

22 embarrassments. Not to be confused with procrastination or self-doubt, which is a mental road-block, second-guessing at the conception of the R&D project is to guess, predict or anticipate what you will encounter during the project duration and what you will finally end up with. This is quite challenging as the research project opportunities that come along the way of a researcher are rarely repetitions of previous experiences (hitherto untried). Right second-guessing requires some quality time for an effective literature review process, peer consultation, reflection, mulling things over and sitting with your thoughts. The speaker intends to use his close to two decades long experience in academic and industrial research projects as well as at postgraduate and undergraduate levels to discuss, with some real world examples, the factors that facilitate and impede hitherto untried R&D efforts in engineering & technology. Water-in-oil Microemulsions as Nanoreactors to Prepare Nanoparticles with Tunable Electrical, Optical, and Antibacterial Properties Md. Abu Bin Hasan Susan Professor Department of Chemistry, University of Dhaka Dhaka 1000, Bangladesh ABSTRACT Nanomaterials with tunable electrical, optical and antibacterial properties have been a fascinating domain of current research for their promising application in diverse areas. In this work, we report preparation, characterization and applications of metal and metallic oxide nanoparticles, core@shell nanoparticles and polymer-nanocomposites from waterin oil (w/o) microemulsions as nanoreactors. Attempts have been made to tune electrical, optical and antibacterial properties of the nanoparticles and nanocomposites by controlling parameters for preparation of w/o microemulsions. Physics and Technology in Radiation Oncology and Imaging Golam Abu Zakaria Professor Department of Electrical, Mechanical and Industrial Engineering Anhalt University of Applied Sciences, Koethen, Germany ABSTRACT Medical Physics is the application of physics concepts, theories and methods to medicine or healthcare. On the other hand, biomedical engineering is the application of engineering principles and design concepts to medicine and biology for healthcare. Medical physicists and biomedical engineers play a vital and often leading role for any medical research team. Their activities cover some key areas such as cancer, heart diseases and mental illnesses. In cancer treatment, they primarily work together on issues involving imaging and the radiation oncology. Thus the medical physicist/biomedical engineer plays a mandatory role in every radiation oncology team. The capability of controlling the growth of any cancer with radiation dose is always associated with the unavoidable normal tissue damage. Accordingly, many physicaltechnical developments in radiotherapy facilities are aimed to give a maximum radiation 21

23 dose to tumour cells and at the same time - minimize the dose to the surrounding normal tissue. For that reason, 60-Co irradiation units were developed in the 50ties and medical linear accelerators in the following decades. Last but not least, neutrons, protons and even heavier ions have also been applied. At the same time, treatment calculation and delivery methods have continuously been improved from conventional multi-beam techniques to tumour shape conformal methods such as radio surgery, Intensity Modulated Radiotherapy (IMRT), Image Guided Radiotherapy (IGRT), Stereotactic Body Radiation Therapy (SBRT) tomotherapy and brachytherapy (BT). The concentration of dose to tumour requires precise information of the shape and the anatomical geometry of the tumour within the body. The techniques providing such pieces of Information in a visible form is summarized by the term Imaging. X-ray has played a dominant role almost from the time of its discovery in Up to now, the use of x-rays has been extended to tomographic imaging with Computer Tomography (CT) and other imaging modalities like Ultrasound (US), Magnetic Resonance Imaging (MRI) or Positron Emission Tomography (PET) which have been developed over the last decades. By their combined use, the required information level on the clinical tumour target volume for radiotherapy has been tremendously raised. The physical and technical development of radiation oncology and imaging are discussed in this talk covering aspects of biology as well. 22

24 Fabrication of cardiac biomarker Immunosensor based on three different electrode surfaces and contrasting their efficiencies Payal Gulati, Prabhash Mishra, S.S.Islam* Centre for Nanoscience and Nanotechnology Jamia Millia Islamia, New Delhi , India * sislam@jmi.ac.in Abstract Myoglobin is important biomarker for the detection of cardiac abnormalities like acute myocardial infarction because it is the first to release after the damage. Accordingly, a peculiar label-free electrochemical immunosensor is fabricated to detect myoglobin based on three different electrodes like glassy carbon, Indium tin oxide glass and porous silicon. All the electrodes were functionalized with GPMS resulting in highly reactive epoxy groups on their surface which covalently binds with the amino group of Monoclonal Anti-Myoglobin Human antibody. Finally, sensing of the electrodes was done with Ag-Mb in a linear range from 0.01 to 1.00µg/ml in PBS buffer (ph 7.4) by using cyclic voltammetry technique. All the electrodes responded to the stepwise changes done on their surface but out of these, glassy carbon electrode was found to be highly sensitive as it showed more current change with respect to small change in the antigen concentration. Keywords Electrochemical immunosensing, cyclic voltammetry, GC, PS and ITO. I. INTRODUCTION Cardiovascular disease (CVD) [1] is a life threatening disorder that affects heart and blood vessels. Coronary heart disease is a most common form of CVD, associated with two major clinical form that is heart attack (often known as Acute Myocardial Infarction) and angima. An Acute Myocardial Infarction (AMI) occurs when heart blood vessel is suddenly blocked, damaging the heart muscle and its function causing cell death. To predict CVD risk, cardiac biomarkers can be identified in the bloodstream which provides therapeutic value in medical applications. An ideal biomarker should have high sensitivity and specificity whereas on the other hand it should be quickly released in blood enabling early diagnosis. Myoglobin [2] (17.8KDa) is an iron- and oxygenbinding protein found in the skeletal muscles and heart of vertebrates. Levels of myoglobin start to rise within 2-3 hours of heart attack or other muscle injury reach their highest level within 8-12 hours and generally fall back to normal within one day. Normal level of myoglobin in serum is 10-65ng/ml and elevates to 200ng/ml after onset of AMI. Electrochemical biosensors [3] based detection offer sensitivity, selectivity and reliability, making them very attractive tools for biomarker protein detection. Due to their low cost, low power and ease of miniaturization, electrochemical biosensors hold great promise for sensing applications as compared to those tedious, time consuming multi-stage process used in hospitals like ELISA (Enzyme linked immuno sorbent assay), fluorescence, Radioimmunoassay. ITO (Indium Tin Oxide) glass provides certain attractive properties like excellent adhesion properties to the substrate, surface stability under harsh condition and it has good conductivity. GC is used for electrochemical sensing purpose because of its low electrochemical resistivity, high chemical resistance, good electrical conductivity and it is biocompatibility. PS is the etched form of silicon wafer, which has high surface area to volume ratio in order to increase adherence to the large molecules about micrometer range and it also exhibits high biocompatibility. For immobilization of Antibodies, Enzymes or DNA on the electrode surface requires functional interlayer of organic SAM (Self- assembled monolayer) like GPMS (3-Glycidoxypropyl trimethoxy Silane) [4]. In this work, we report a planar ITO, Glassy Carbon and Porous Silicon based immunosensor functionalized with GPMS silane whose exposed epoxy groups readily reacts with amino groups of the antibody. Finally sensing of myoglobin was done by cyclic voltammetry in a range from 0.01µg/ml to 1µg/ml. In order to compare the sensitivity of all the electrode surfaces, Anodic peak Current v/s Myoglobin Conc. plot was made. II. EXPERIMENTAL PROCEDURE A. Conditioning & development of electrode The selected electrodes were conditioned for the immunosensor fabrication. Due to highly conductive nature of Indium- tin oxide coated glass electrode, it was taken as a planar surface for the bio-electrode formation. It is a high quality glass because it is not affected by moisture. The ITO coated glass was cut into (5cm 0.8cm). Then electrodes were cleaned ultrasonically with acetone, ethanol and water for 10 min each respectively and dried. Further, they were immersed into 1M HCl for 10 min. Then wash with De-ionised water. Then immersed in solution (1:1:5) v/v H 2 O 2 (30%) / NH 4 OH (30%) /H 2 O (Pirhana 23

25 solution). Finally, rinse with DI water and Dried under the steam of N 2. Similarly, glassy carbon having high conductivity was also used as planar surface for the bio-electrode formation. The glassy carbon electrode (5mm in diameter), was cut into 1cm length pieces. Then electrodes were polished with rough and smooth sandpaper. Further, they were boiled in HNO 3 for 1 hour. Then wash with DI-water and air dried. In the same way, the p-si wafer [5,6] (100) with a resistivity: 1-10Ωcm, thickness: µm was anodized in Teflon cell using HF: DMF (1:3) electrolyte with current density: 30mA/cm 2 for 40 min etching time 6 to obtain porous silicon of high surface to volume ratio. B. Fabrication of immunosensor These activated electrodes were silanized with 10 ml of 5% (3-Glycidoxypropyl) trimethoxysilane (GPMS) in toluene for 18hr at room temperature to form self-assembled monolayer of it, acting as necessary linker molecules to immobilize biomolecules on solid electrode surface. After this treatment, electrode surface contains an active epoxy group which reacts with amino groups present on the antibody molecules. Further these functionalized electrodes were then immobilized with 100µg/ml monoclonal anti-myoglobin human antibody (produced in mouse, Sigma Aldrich) in PBS buffer ph=7.4 for 24 hours at 4 C. Wash these electrodes with PBS buffer to remove physically adsorbed antibodies and then incubate in a 1% Bovine serum albumin (BSA) in PBS (ph=7.4) for 1 hour at 37ᵒ C to block the non-binding sites present on the electrode surface. Again rinse the electrodes with PBS buffer. Finally, sensing of these bio-electrodes was done after they were incubated with different Myoglobin (from equine heart) concentration varying in a linear range: 0.01µg/ml, 0.05µg/ml, 0.10µg/ml, 0.05µg/ml, and 1.00µg/ml in PBS buffer (ph=7.4) for 1 hour at room temperature. C. Characterization Scanning Electron Microscopy (SEM) was used to study the surface morphology of the porous silicon i.e. porosity of the porous silicon using NOVA NANOSEM 450 model. Cyclic Voltammetry is a potentio-dynamic electrochemical method which is used to study the resulting current change of the modified electrode surface in an electrochemical cell. Cyclic voltammetry was conducted on Solatron 1280 C Potentiostat/Galvenostat with conventional three electrode system in which Ag/AgCl was used as a reference electrode, Platinum as counter electrode and ITO, Glassy Carbon, Porous Silicon as a working electrode. The alterations in the cathodic- anodic peak current and peak potentials of working electrode were studied using 5mM concentration of redox mediators K 3 Fe(CN) -3 6/ K 4 Fe (CN) -4 6 in PBS buffer (ph-7.4) with potential range from -0.9 to +0.9V at the scan rate of 50mV/s. III. RESULT & DISCUSSIONS Fig1 shows the formation of porous silicon having a pore size in a range of 4-5µm making it suitable for the immunosensor fabrication. The purpose to obtain highly porous sample is for the proper attachment of antibody on the electrode surface, resulting in uniform immobilization of antibody. Fig. 1: SEM image of PS Cyclic voltammetry measures the resulting current in an electrochemical cell by cycling the potential of the working electrode. As we can see in the fig1, 2, 3; (step a) well defined cathodic and anodic peaks (E pc & E pa ) was obtained because activated electrode surface contains OH groups. After silanization i.e. treating the electrode with GPMS (in step b), we found that there was significant decrease in current level indicating the SAM formation on the electrode surface rendering it insulating property. Fig. 2: CV spectra of Cleaned ITO (a), GPMS treated (b), monoclonal anti-myoglobin human antibody (c), BSA treated (d). Fig. 3: CV spectra of bare PS (a), GPMS treated (b), monoclonal anti-myoglobin human antibody (c), BSA treated (d). 24

26 Fig. 4: CV spectra of cleaned GC (a), GPMS treated (b), monoclonal anti-myoglobin human antibody (c), BSA treated (d). Further (in step c) immobilization of monoclonal anti-myoglobin human antibody whose amine groups were linked to the epoxy groups of GPMS; there was very slight increase in cathodic and anodic peak current (i pc & i pa ) due to presence of non-binding sites present on the electrode surface which allows electron to flow through the surface. Eventually, when non-binding sites present on the electrode surface was blocked by treating electrode with BSA solution (in step d), it was found that there was significant decrease in the cathodic current (i pc ) and peak to-peak separation was increased. This shows that remnant surface of electrode for electron transfer was occupied by the protein i.e. BSA. Fig. 7: GC electrode based CV spectra of different myoglobin concentration with myoglobin antibody: monoclonal anti-myoglobin human antibody (a), 0.01 µg/ml (b), 0.05 µg/ml (c), 0.10 µg/ml (d), 0.50 µg/ml (e), 1.00 µg/ml (f). It is observed from C-V studies that same pattern of result was observed on all substrates while studying different antigen (Myoglobin) concentration anodic peak current was different of every electrode with respect to different concentration. A comparison plot of efficiencies among all the three electrodes is made to determine the most sensitive material where sensitivity is calculated from the slope of the figs. It is also observed that as the concentration of antigen increases the current level decreases because of the immune-complex formation between antigenantibody. This spectra indicates as the concentration of antigen goes up, electron transfer is blocked on the surface. Fig. 5: ITO electrode based CV spectra of different myoglobin concentration with myoglobin antibody: monoclonal anti-myoglobin human antibody (a), 0.01 µg/ml (b), 0.05 µg/ml (c), 0.10 µg/ml (d), 0.50 µg/ml (e), 1.00 µg/ml (f). Fig. 6: PS electrode based CV spectra of different myoglobin concentration with myoglobin antibody: monoclonal anti-myoglobin human antibody (a), 0.01 µg/ml (b), 0.05 µg/ml (c), 0.10 µg/ml (d), 0.50 µg/ml (e), 1.00 µg/ml (f). Fig. 8: Current v/s Myoglobin Conc. plot achieved from cyclic voltammetry studies Fig.8 shows that GCE is highly sensitive in comparison to those materials as it is a stable, unreactive material and forms strong covalent binding upon functionalization. Being a carbon form it is highly conductive to electrons and has great electrochemical conductivity exhibiting good repeatability, reproducibility and fast electron transfer kinetics. The reason for PS being not as sensitive as GC is due to surface instability because of the presence of highly reactive Silicon Hydride species that are very reactive both in air and water. Thus PS substrate must be made stabilize by certain modification to use for biosensing purpose. Similarly, ITO glass being not as sensitive as GC because of the inhomogeneous deposition of In & Sn on the glass surface resulting in surface instability. 25

27 IV. CONCLUSION All the bio-electrodes developed are sensitive for the lowest antigen detection limit upto 10ng/ml. But glassy carbon electrode surface is highly sensitive in comparison to other two materials as it showed large change in current response with respect to small change in the myoglobin concentration. The advantage of these immunosensor is that they are very cheap and highly sensitive for abnormal myoglobin level detection in patient s serum with cardiac problems. Another advantage is that an untrained person can also use it. REFERENCES [1] Barry McDonnell, Stephen Hearty, Paul Leonard, Richard O'Kennedy Clinical Biochemistry 42 (2009) [2] Anjum Qureshi, Yasar Gurbuz, Javed H. Niazi Sensors and Actuators B (2012) [3] DorotheeGrieshaber, Robert MacKenzie, Janos Voros and Erik Reimhult Sensors 2008, 8, [4] Mamas I. Prodromidis Electrochimica Acta xxx (2009) xxx xxx. [5] Werner Kern J. Electrochem. Soc., Vol. 137, No. 6, June [6] Kumari, Vinita; Gulati, Payal; Mishra, Prabhash; Islam, S.S. Advanced Science Letters, Volume 20, Numbers 7-9, pp (4). 26

28 Simulation of the Electrical Activity of Cardiac Tissue by Finite Element Method Tanzina Rahman Dept. of Applied Physics and Electronic Engineering University of Rajshahi Rajshahi, Bangladesh Abstract The electrical activity is responsible for the periodic contraction and relaxation cycle of the human heart. Significant implications of simulating electrical activities are helpful to understand cardiac abnormalities like cardiac arrhythmias. Mathematical modeling of heart provides a better understanding for the complex biophysical phenomena related to electrical activity in the heart. Various reaction-diffusion models have been developed to study the proper function of human heart at different conditions. In this research work monodomain model which is coupled with the single cell FitzHugh-Nagumo model is used to simulation the electrical activities. Two dimensional monodomain model equations on a general domain with equal isotropy and no fiber orientation are considered. Finite element method which has been widely used as an analysis and design tool is used to deal with the complex monodomain model equations. For35 35 nodal elements current above the threshold values is applied to the middle node and less than threshold to the others. The outcome of the simulation shows variations of excitation propagation for uniform time steps. Keywords Electrical Activity of the Heart, Bidomain and Monodomain model, Finite lement method. I. INTRODUCTION Electrical activity of cardiac tissue is an important part of bio-medical science. Computer simulation is becoming an important tool in cardiovascular research. We are on the brink of a revolution in cardiac research, one in which computational modeling of proteins, cells, tissues, and the organ permit linking genomic and proteomic information to the integrated organ behavior, in the quest for a quantitative understanding of the functioning of the heart in health and diseases [1]. Mathematical model of cardiac electrical activity has been recognized as one of the significant approaches capable of revealing diagnostic information about the heart. The electrical activity of the heart is an important tool for the primary diagnosis of the heart diseases; it shows the electrophysiology of the heart and the ischemic changes may cause myocardial infection, conduction defects and arrhythmia [1]. The electrical activity of the cardiac as a whole is thus characterized by a complex multi scale structure, ranging from the microscopic activity of ion channels in the cellular membrane to the macroscopic properties of the anisotropic propagation of the excitation and recovery fronts in the whole cardiac. Mathematically, reaction diffusion systems take the form of semi-linear Md.Rezaul Islam Dept. of Applied Physics and Electronic Engineering University of Rajshahi Rajshahi, Bangladesh rima@ru.ac.bd parabolic partial differential equations. In order to describe electrical activity in the whole heart, a single cell model with the Partial differential equations model that describes how electricity flows across a network of cells [2]. The most complete model of such a complex setting is the anisotropic bidomain model that consists of a system of two degenerate parabolic reaction diffusion equations describing the intracellular and extracellular potentials in the cardiac muscle, coupled with a system of ordinary deferential equations describing the ionic currents flowing through the cellular membrane [3]. But this model requires a long simulation time, large computer memory. So in this research monodomain model is used. The technique of finite element method has been used to build a computer program to solving the phenomena propagation of excitations. The finite element method (FEM) has been widely used as an analysis and design tool in many engineering disciplines like structures and computational fluid mechanics. The FEM method is a powerful tool for solving differential equations. The method can easily deal with complex geometries and higher-order approximations of the solution [4]. A domain of interest is represented as an assembly of finite elements. II. BIDOMAIN MODEL Bidomain is a structure defining a model of heart tissue consisting of two interpenetrating domains representing cardiac cells and the surrounding space. It is one of the two differential equation based models for cardiac electrical activity [5]. The model is considered as the mathematical equations that have been used for simulating cardiac electrophysiological waves for years taking into account the non-linear dynamic nature of the cardiac signal and giving realistic simulation. This model gives the representation of the cardiac tissue at a macroscopic scale by relating the transmembrane potential, the extracellular potential and the ionic currents. Heart tissue can be classified into two groups: intracellular and extracellular as shown in fig. 1. It consists of a system of two non-linear partial differential equations coupled to a system of ordinary differential equations [6]. The set of bidomain equations is currently the most complete mathematical models for describing the spread of cardiac electrical activity, especially for simulating activity on the organ level. Conductivity is 27

29 usually greater along the fibers rather than across them [5, 6]. Fig. 1. Bidomain model The bidomain model of cardiac tissue is based on current flow, distribution of electrical potential, and the conservation of charge and current [6]. The description of each domain is based on a generalized version of Ohms law defining the relationship between the electric field E, derived from the potential (V), the current density J and the conductivity tensor D. So we have E (1) J DE D (2) Now in case of the intracellular and extracellular spaces we have J J i D i e D e i Where J i and J e are the intracellular and extracellular current densities, D i and D e are the corresponding conductivity of the tensors, respectively i and e are the electrical potential in the intracellular and extracellular spaces. In this paper we are considering the heart in isolation. Hence, a change in current density will be of equal magnitude in both domains but will have the opposite sign: Ji J e (3) ( J i Je ) 0 (4) Now Ji I m J e I m where I m is transmembrane current per unit volume which is composed of a capacitive component and an ionic component I ion. So we have Vm I m Cm ( ) Iion I appl (5) t e Here C m is the specified cell membrane capacitance, I appl is the stimulus current and V m is the transmembrane voltage which is given by Vm i e (6) Combing (4) and (5) we have Vm. Di ( Vm e) Cm( ) Iion Iappl (7) t (.(( DeDi ) e ).( DiVm ) (8) Equation (7) is the parabolic equation and (8) is the elliptic equation [61, 62]. Those are known as the governing equations of the biodomain model of cardiac tissue. III. MONODOMAIN MODEL The monodomain model is a simplification of the bidomain model that is easier to analyze. It is also notable that computational cost of using the monodomain model is about one-half to one-tenth the cost of using the bidomain model, depending on the complexity of the cell model used [7]. This model helps to understand the patterns of electrical conduction and propagation from the scale of a single tissue to whole heart. In this physical model the cell membrane is viewed as an electrical network with the fibers of myocardial cells constituting a cable [7]. In this analysis, it is assumed that the anisotropy of the intracellular and extracellular spaces is the same, i.e. that the conductivity in the extracellular space is proportional to the intracellular conductivity. D (9) e D i Here is a scalar, which representing the ratio between the conductivity of the intercellular and extracellular spaces. The choice of the value of can determines physiological accuracy, but it is important to select a suitable value that gives the satisfactory results [1, 7]. Substituting (9) into (8) we get.(( De Di ) e ).( DiVm ) (10) 1.( Di e ) ( ).( DiVm ) 1 (11) Substituting (11) into (7) we have 1 Vm ( ).( DiVm ) Cm ( ) I 1 t ion I appl (12) Since an effective conductivity = then by the monodomain model for cardiac tissue is given by V m.( DVm ) Cm ( ) Iion I (13) appl t The conductivity of the tensor D, in the above equation is defined largely by the orientation of the tissue. Cardiac cells are grouped into muscle fibers, 28

30 and the muscle fibers are grouped into sheets of fibers. The structure of the heart influences the flow of electricity. Conductivity is usually greater along the fibers rather than across them [1, 7]. IV. SOLUTION OF MONODOMAIN MODEL BY FEM In this research work the FEM is used to solve the monodomain model. FEM has been widely used as an analysis and design tool in many engineering disciplines. The method can easily deal with complex geometries and higher-order approximations of the solution. Galerkin method is used to discretize the monodomian model. Galerkin methods are a class of methods for converting a continuous operator problem (such as a differential equation) to a discrete problem [4]. In principle, it is the equivalent of applying the method of variation of parameters to a function space, by converting the equation to a weak formulation. Typically one then applies some constraints on the function space to characterize the space with a finite set of basis functions. Finite Element Equations The monodomain equation which is coupled the modified FitzHugh-Nagumo model is given below Vm. D( D i V m ) Cm ( ) Iion I appl (14) t From the above equation we can write C m ( V m ) I ion I appl D t ( x 2 V 2 m V ) D ( m ) 2 y x y 2 Although the implementation supports both two dimensional and three-dimensional problems, for the simplicity only two dimensional equations are used. The Intel dual core processor computer with 4GB RAM also used for simulation. Table 1 shows the parameters are used for simulation purpose and corresponding values. TABLE I. TABLE MONODOAMIN MODEL SIMUATION PARAMETERS. Parameters Description Value Unit I appl Applied current ma d y d x Diffusion coefficient in the horizontal axis Diffusion coefficient in the vertical axis e-004 cm cm C m Capacitance 1 μf/cm 2 dt Time ms A. Applied current at the middle node for35 35 nodal elements In simulation purpose cardiac tissue is considered as a uniform mesh for triangle elements. For this condition a MATLAB code is built for 1225 nodes and 2312 elements. The diffusion co-efficient d x along the fibers while the diffusion d y perpendicular to the fibers not in the plane. We can observe the propagation of excitations in the heart tissue by the Fig. 2. We can apply the Euler method in the above equation for time derivative and the application of Galerkin method to the diffusion term only over the entire domain of equation. Now we can write V 2 V 2 V C m m m m W d W ( I I d W D D d t ion appl ) [ x ( ) y ( )] x 2 y 2 where, W is the weighting function. Now applying the Euler method for time derivative the resulting algebraic equation in matrix form is following equation Cm n1 Cn n ( M k) v Mv MI ion MI appl t t Here M is the FEM lumped mass matrix and K is stiffens matrix. V. SIMULATION RESULT AND DISCUSSION The simulation result is obtained by the MATLAB code to generate the uniform mesh for the heart tissue as a triangular element. The first step in the finite element method is to divide the structure or solution region into subdivisions or elements. Hence, the structure is to be modeled with suitable finite elements. The number, type, size, and arrangement of the elements are to be decided. (a) (b) (c) 29

31 (d) (e) Fig. 2. Surface plot for the propagation of excitations in the heart tissue at a uniform time step for nodal element The surface plots show the propagation of excitations for nodal numbers. The current above the threshold value is applied to the middle node of the square mesh grid less than threshold value to other nodes. The diffusion coefficient d x along the fibers and d y perpendicular to the fibers but not in the plane. The value of d x is more than d y. The excitations from the middle nodes propagated to the nearby horizontal and vertical nodes. The propagation of excitations along the horizontal axis is more than the vertical. It is because value of d x is more than d y. The potential of those nodes changes consequently. After a uniform time interval the excitations reached at the end of the nodes. Then the process started from the initial condition. Fig. 2 consequently shows the process. VI. CONCLUSION In this research work propagation of excitation in the cardiac tissues is simulated, based on reactiondiffusion monodomain model. This work has been able to create some insights about the electrical behavior of human heart, revealing the nature of the excitation, propagation pattern in the cardiac tissue. The electro-physical mathematical models are based on ordinary differential equations and partial differential that describe the behavior of this electrical activity. Generating an efficient numerical solution of these models is a challenging task, and in fact the physiological accuracy of tissue-scale models is often limited by the efficiency of the numerical solution process. The two dimensional monodomain equations solved to study the behavior of excitation propagation. The surface plots showed how excitation propagated in the cardiac tissue. For nodal elements current above the threshold is applied to the middle node and less than threshold to the others. Excitation propagated from the middle nodes to nearby horizontal and vertical nodes of the rectangular mesh grid. Then the nearby nodes also excited. In this way excitation propagated to the corner nodes. The propagation rate in the horizontal axis is more than vertical axis. The simulation worked is done in a small portion of the SA node. The model is solved for two dimensional equations for less complicity. If we want to increase more elements and nodes number then the simulation process becomes difficult. Because the simulation process need more time, computer memory and computational power. In future the models can be updated with patient-specific geometry and body surface potentials. At the same time more detailed cell models can be used and eventually the model can be improved by coupling with other electro-mechanical and blood flow models. ACKNOWLEDGMENT The authors are thankful to department of Applied Physics and Electronic Engineering, University of Rajshahi, Rajshahi-6205, Bangladesh. REFERENCES [1] J. Sundnes, G. T. Lines, X. Cai, B. F. Nielsen, K.-A. Mardal, and A. Tveito, Computing the electrical activity in the heart,springer-verlag, Berlin, [2] A. J. Pullan, M. L. Buist, and L. K. Cheng, Mathematically modelling the electrical activity of the heart: from cell to body surface and back again, World Scientific, New Jersey, [3] C. S. Henriquez, Simulating the electrical behavior of cardiac tissue using the bidomain model,crit. Rev. Biomed. Eng.,vol. 21, pp. 1 77, [4] S. C. Brenner and L. R. Scott, The mathematical theory of finite element methods,springer,berlin, [5] Jairo Villegas G and Andrus Giraldo M, The electrical activity of cardiac tissue via finite element method, Adv. Studies Theor. Phys., vol. 6, PP , [6] J.M.Rogers and A.D. McCulloch, A collacation- Galerkin finite element model of cardiac action potential propagation, IEEE Trans. Biomedical Engineering.41, PP , [7] Shuaiby M. Shuaiby, M. A. Hassan, Abdel-Badie Sharkawy and Abdel-Rasoul M.M.Gad, A finite Element Model for the electrical activity in human cardiac tissues, J. Eco. Heal. Env.vol. 1, PP.25-33,

32 Baking of Ilmenite on Koistening with Sulfuric Acid followed by Leaching with Dilute Sulfuric Acid Solution Ranjit K. Biswas, Aneek K. Karmakar and Jinnatul Ara Dept. of Applied Chemistry and Chemical Engineering Rajshahi University Rajshahi-6205, Bangladesh Abstract It is very difficult to dissolve ilmenite quantitatively; and as a result, hundreds of papers and patents are available on its dissolution using various ways and techniques. But none of these claims for cent percent dissolution. The objective of ilmenite dissolution is to prepare pigment grade TiO 2, or at least, the Chloride feed grade TiO 2. A new, easy and attractive technique for ilmenite dissolution is described in this work. In this work, ilmenite has been moistened with conc. H 2 SO 4, baked (heated) at a temperature near the boiling point of H 2 SO 4 and then leached with very dilute solution of H 2 SO 4. The optimized baking temperature and time are found to be 300 o C and 30 min (for ilmenite to H 2 SO 4 wt. ratio of 1), respectively. The single stage baked mass can be leached by 0.10 mol/l H 2 SO 4 solution at its boiling temperature under reflux and at ilmenite to liquid ratio of 0.01 g/ml for 20 min to extract ~86% Fe and only 1.1% Ti from ilmenite. On the other hand, the 3-stage baking and 1-stage leaching remove as much as 97% Fe and only 2.4% Ti. The residue left after 3-stage baking and 1-stage leaching is almost white and identified as rutile, which can be regarded as a very good quality Chloride feed for manufacture of pigment grade TO 2. Reasonable low temperature and time required in baking, together with the requirements of very low concentration of H 2 SO 4 and time in leaching will attract technologists of this field to adopt this technique for almost complete removal of Fe from ilmenite to produce Chloride feed grade TiO 2. Index Terms Ilmenite, H 2 SO 4, Baking, Leaching, Chloride feed INTRODUCTION Pure TiO 2 is widely used in paint, plastic, paper, textile, rubber and ceramic industries. Its sources are rutile (natural occurring TiO 2 ) and ilmenite (TiFeO 3 ). Rutile is generally used to prepare pigment grade TiO 2 using the well-known Chloride process. But the reserve of rutile in nature is almost exhausted by this time. Consequently, the TiO 2 industry now depends solely on ilmenite (mixed oxides of ferrous and titanium). Ilmenite can be processed either for synthetic rutile by eliminating Fe from it, or for pigment grade TiO 2 on its complete dissolution (Sulfate) method. A considerable amount of heavy mineral deposits containing ilmenite, rutile, magnetite, zircon, monazite, garnet, leucoxene, kyanite, epidot, amphobiles, biotite, apatite etc. have been discovered in the Sitakundu-Chittagong-Banskhali-Cox s Bazar- Inani-Teknuf beach and off-shore islands of Moheshkhali, Matabari, Kutubdia, Nijhum Dwip etc. Muhammad A. Gafur Pilot Plant and Project Development(PPPD) Division BCSIR Laboratories, Dr. Kudrat-E-Khuda Road, Dhanmondi Dhaka-1205, Bangladesh d_r_magafur@yahoo.com in Bangladesh. This (sp. gr.>2.88 g/cc) accounts for 10% of the total sand and contains about 27% ilmenite on average [1, 2]. In processing ilmenite to pigment grade TiO 2, it is necessary to dissolve either both Fe and Ti or at least one component from ilmenite completely. This step still remains unsolved [1, 3-8] causing low recovery percentage in the existing manufacturing technologies of TiO 2. However, methods have been developed for separating Fe(II)/Fe(III) from Ti(IV) existing in the leach solution obtained by partial leaching of both Ti and Fe from ilmenite [9-13]. As Bangladesh has considerable reserve of ilmenite, it is very important to work on processing of ilmenite to get either pigment grade TiO 2, or at least, rutile enriched feed for the Chloride process. Previously [14], it was observed that copper dust on baking with 98% H 2 SO 4 followed by leaching with 0.05 mol/l H 2 SO 4 resulted in preferential leaching of copper leaving almost all iron in the residue. Being inspired with this result and as the mineral acid baking of ilmenite is not yet reported, the present study has been carried out with an aim to its complete dissolution, or at least, preferential leaching of either Fe or Ti from it. EXPEIMENTAL A. Materials Ilmenite was collected in a single lot from the Pilot Plant of Beach Sand Exploitation Centre at Kalatali (Cox s Bazar) of Bangladesh. Collected ilmenite was dry-ground and sieved to collect <53 µm sized particles to use in the most baking studies. Reagent grade chemicals of either Merck, BDH or Loba chemie were used in this study without further purification. B. Analytical The XRD patterns of powdered samples were taken from the PPPD Center of the BCSIR Laboratories, Dhaka. The SEM images were taken from the Glass and Ceramics Engineering Department, BUET, Dhaka. Routine analyses of Ti(IV) and Fe(III) in aqueous solutions were carried out by the H 2 SO 4 H 2 O 2 method at 410 nm and HNO 3 NH 4 SCN method at 480 nm, respectively, using a visible spectrophotometer (SP 1105, China). The HNO 3 NH 4 SCN method estimated the amount of Fe(III) present in a solution; whereas, prior oxidation of the solution by boiling for 5 min with 2 31

33 ml conc. HNO 3 followed by the application of the HNO 3 NH 4 SCN method yielded the amount of Fe(II) plus Fe(III) in the solution. The difference gave the amount of Fe(II). C. Work Plan The plan of work is schematically depicted below (Fig. 1) which is self-explanatory: Ilmenite (1g) Conc. acid Quartz boat/ Platinum Crusible Left overnight and placed in furnace Furnace Coiled plastic tube for acid condensation To vent Fig. 2. X-ray diffraction pattern ( ), : ilmenite (), : Fe 9TiO 15 Small air pump Residue Fig. 1. Flow sheet of the work plan D. Procedure for baking An aliquot of 1 g sized ilmenite was taken on a quartz boat and moistened with 1 ml conc. H 2 SO 4 and left overnight. With the aid of a wire attached at one end of the boat, the container was then placed in a tubular furnace (Carbolite CTF 12/65/550 type, UK) at a certain thermostated temperature ( o C). The system was closed. Using a small air pump, air was passed through the furnace and allowed to flow from the other end via a plastic tube. Exit air carried the acid vapor for condensation in the tube and allowed to vent. After baking, the material with the container was taken out from the furnace and put on a flat ceramic sheet for cooling. It was then quantitatively scratched over to a glossy paper for leaching. E. Procedure for leaching An aliquot of 100 ml of 0.10 mol/l H 2 SO 4 solution was taken in a 500 ml quick-fit conical flask fitted with a 1 m long glass tube acting as a condenser. The flask was set on a magnetic hot plate for heating as well as for pulp agitation using a magnetic capsule of 2 cm in length. When the boiling of the leaching agent started, the baked mass was added and switched on for time counting. When the ambient temperature was greater than 25 o C, the glass tube acting as condenser, was wrapped with moist cloth or cotton to facilitate condensation. After being leached for a specific time, the leached slurry was filtered and the filtrate was analyzed for its Ti(IV), Fe(II) and Fe(III) contents. RESULTS AND DISCUSSION A. Characterization of Ilmenite Used in This Study The collected ilmenite fraction of size <53 µm was used for powder XRD, SEM-XRF, chemical analyses and baking studies. The XRD pattern as taken by Cu Kα radiation (40 kv, 30 ma, scan speed 0.02 o /s and step time 0.8 s) is depicted in Fig. 2, in order to identify the main phases existing in the sample. Main identical phases are: FeTiO 3 and Fe 9 TiO 15 (TiO 2. 4Fe 2 O 3.FeO). Other minor phases detected are SiO 2, Fe 2 O 3, TiO 2, (Fe, Mg) Ti 2 O 5 and FeCr 2 O 4. Baked mass Leaching assemble Filter Filtrate Analysis for metal ions Fig. 3. SEM image of Ilmenite The SEM image of the sample as shown in Fig. 3, shows the crystallinity of different phases in the sample. The almost complete dissolution of the sample in molten KHSO 4 for about 6 h, subsequent solublization in 15% H 2 SO 4 solution and the colorimetric analysis of the resultant solution indicates that it contains 32.4% Fe (in both -ous and -ic states) and 27.8% Ti(IV). In the dissolution process, ~5% material remains undissolved. The other minor ingredients such as Si, Mn, Cr, Mg etc. are not quantified. On considering that the insoluble part does not contain any Fe and Ti, the extent of leaching (thermal treatment as well) has been monitored on the basis of 32.4% total Fe and 27.8% Ti existing in the sample. B. Optimization of Baking Temperature The progress in baking is monitored by the extents (%s) of titanium and iron dissolved in succeeding leaching experiments conducted using 0.10 mol/l H 2 SO 4 solution at its boiling temperature under reflux. The effect of baking temperature is investigated for a constant baking time of 30 min. The effect of baking temperature on dissolution of ilmenite, within o C, is provided in Fig. 4, as the weight percentages of Ti(IV), Fe(II), Fe(III) and total Fe dissolved on subsequent leaching versus baking temperature plots. It is seen from this figure that the % dissolution of Ti(IV) is very low (< 4%), irrespective of baking temperature. On the other hand, Fe(II) and Fe(III) dissolutions under identical conditions are notable. A 33.1% Fe(II) dissolution at 100 o C is increased to 36.2% at 200 o C; and then continuously decreased with the rise of baking temperature. At about 600 o C, it becomes negligible. On the other hand, 29.5% Fe(III) dissolution is increased to 44% at 400 o C and then, gradually decreased to 41.3% at 500 o C followed by rapid fall within 500 o C to 600 o C. At 800 o C, Fe(III) dissolution is only 0.6%. Total weight percentage of Fe dissolved from the baked mass at 300 o C is the maximum (73.6%) At both lower and higher temperatures, total Fe dissolutions are lower, but considerable within baking temperatures of 100 o C o C. Over 500 o C, total Fe dissolution is sharply decreased to only ~1% at 800 o C. The combined baking and leaching results indicate that near the boiling point of H 2 SO 4 (316 o C), ilmenite sample 32

34 used in this study is principally broken down to TiO 2 and mixture of FeSO 4 and Fe 2 (SO 4 ) 3, during baking, according to the following reactions: TiFeO 3 + H 2 SO 4 TiO 2 + FeSO 4 + H 2 O (1) Fe 9 TiO H 2 SO 4 TiO 2 + FeSO 4 + 4Fe 2 (SO 4 ) H 2 O (2) followed by dissolutions of only FeSO 4 and Fe 2 (SO 4 ) 3 in very dilute sulfuric acid solution, leaving intact TiO 2 as a solid phase. One more point is notable here. With the increase of baking temperature, the reason for the decreased Fe(II) dissolution and increased Fe(III) dissolution at higher temperature regions (up to ~ 500 o C) is the enhanced oxidation of Fe(II) to Fe(III), in air atmosphere as per the following reaction: 4FeSO 4 + 2H 2 SO 4 + O 2 2Fe 2 (SO 4 ) 3 + 2H 2 O (3) At temperature above ~ 500 o C, the decomposition of Fe 2 (SO 4 ) 3 starts possibly according to following reaction: Fe 2 (SO 4 ) 3 Fe 2 O 3 + 3SO 3 (4) followed by decomposition of SO 3 in the atmosphere of low partial pressure of O 2 as shown below: 2SO 3 2SO 2 + O 2 (5) Ferric oxide, formed in this way, is not possibly leached out by dilute sulfuric acid solution used in this study. This result opens a way for selective leaching out of iron from ilmenite using conc. H 2 SO 4 in baking and dil. H 2 SO 4 solution in leaching. Wt% of metals dissolved (based on amounts present in ilmenite) Baking temperature, C Fig. 4: Effect of baking temperature. Baking time = 30 min, leaching time = 1 h. ( ), Ti 4+ ; (), Fe 3+ ; (), Fe 2+ ; ( ), Total Fe (Fe 3+ + Fe 2+ ). C. Optimization of Baking Time At a constant baking temperature of 300 o C, the effect of baking time (5 min interval within 50 min) is investigated keeping other conditions stated unchanged. Results are presented in Fig. 5, as plots of wt.% metal ions dissolved versus baking time (min). The figure indicates that the baking for 30 min gives the best dissolution of Fe from ilmenite. For Ti(IV) dissolution, the percentage is gradually decreased with time of baking (3.8% at 5 min to 0.9% at 50 min). The decreased Fe dissolution for baked masses over 30 min may be due Eqs. (3) and (4) occurring for longer time in a high temperature of 300 o C. It is concluded that the best partial leaching of Fe from ilmenite can be achieved at a baking temperature of 300 o C for 30 min. Wt.% of metals dissolved Baking time, min Fig. 5. Effect of baking time on dissolution. Baking temperature = 300 o C and leaching time = 1 h. ( ), Ti 4+ ; (), Fe 3+ ; (), Fe 2+ ; ( ), Total Fe (Fe 3+ + Fe 2+ ). D. Optimization of Leaching Time The baked mass of H 2 SO 4 moistened ilmenite (wt. ratio of ilmenite/conc. acid = 1, baking temperature = 300 o C and baking time = 30 min) is leached with 100 ml of 0.10 mol/l H 2 SO 4 at boiling temperature under reflux for variable leaching time ranging from 2 60 min to examine the effect of leaching time in the system. Results are given in Fig. 6. It is seen that the Ti dissolution is limited within less than 1.7% up to leaching time of 40 min which increased to ~3.7% at 60 min. The Fe(III) dissolution percentage of 44.1% at 2 min is increased to ~ 61.5% at 20 min followed by a decrease to 43.6% at 60 min. On the other hand, the dissolution percentage of Fe(II) is gradually increased from 13.2% at 2 min to 30% at 60 min. The maximum total iron dissolution of ~ 86% is obtained at leaching time of 20 min. The reason for decrease in dissolution percentage with increasing leaching time is not known. Wt.% of metals dissolved Leaching time, min Fig. 6. Effect of leaching time, following baking, on dissolution. Baking time = 30 min and baking temperature = 300 o C. ( ), Ti 4+ ; (), Fe 3+ ; (), Fe 2+ ; ( ), Total Fe (Fe 3+ + Fe 2+ ). E. Stage-wise Baking The results of stage wise baking (the mass obtained in the 1 st stage of baking is re-baked on addition of 1 ml conc. H 2 SO 4 and so on) but single stage leaching is provided in Fig. 7. The stagewise baking is carried out at baking temperatures of 200 o C and 300 o C. Figure 7 shows that increase in baking - stage number, operated at 200 o C, has little influence on Ti(IV) dissolution. In contrast, 36% Fe(II) dissolution from single - stage baked mass is decreased to ~ 26% from three - stage baked mass. On the other hand, ~ 35.5% Fe(III) dissolution from singlestage baked mass is increased, extensively, to ~ 70.8% from threestage baked mass. Eventually, total Fedissolution is increased if stagebaking is 33

35 practiced. Total iron dissolution of ~ 71.7% from single - stage baked mass is increased to 90% from two stage baked mass and to 96.8% from the threestage baked mass. Almost similar results are obtained for baking at 300 o C. In these cases, Ti(IV)dissolution is more decreased. About 3.75% Ti(IV)dissolution obtained from single - stage baked mass is decreased to ~ 1.8% Ti(IV) - dissolution from three - stage baked mass. Although Fe(II)dissolution curve shows a maximum, Fe(III)- and total Fe dissolution percentages are increased with increase in stage number. ~ 86% total Fedissolution from single - stage baked mass is increased to 98.4% from two - stage baked mass and to 98.7% from three - stage baked mass. Wt.% of metals dissolved Number of baking stage Fig. 7. Effect of stage-wise baking. Wt. of H 2SO 4 added in each stage = 1 g, baking time =30 min, baking temperature = 200 o C (open symbols) and 300 o C (closed symbols), leaching time = 20 min. (, ), Ti 4+ ; (, ), Fe 3+ ; (,), Fe 2+ ; (,), Total Fe (Fe 3+ + Fe 2+ ). F. Characterization of baked mass and leached residue The X-ray diffraction pattern (not shown) of three stage baked mass (at 300C for 30 min) indicates that it is semi-crystalline. Comparison of this pattern with that in Fig. 2, clearly indicates that the crystal lattices of components present in ilmenite are completely destroyed on baking. The very weak new peaks appeared are probably due to the presence of Fe(II) and Fe(III) compounds of sulfate (however, not assigned) which could be leached out by dilute sulfuric acid solution and even mostly by water. Probably, the baked products do not get enough time during baking for large crystal growths. On the other hand, XRD pattern of baked (3-stages) leached residue at optimized conditions (not shown) differ from those of ilmenite and baked ilmenite. It is seen that the peaks of ilmenite (Fig. 2) and of baked mass are not present in the XRD pattern of residue. The residual product appears almost as amorphous; but some diffraction peaks for rutile is obtained. The SEM images of baked (3-stages) mass and baked leached residue are depicted in Figs. 8. The comparison of these images with that given in Fig. 3 demonstrates the changes occurring in baking and leaching. (a) (b) It can be demonstrated that, from 3-staged baked mass at optimized condition, 36.8% total Fe (32% Fe(III) and 4.8% Fe(II)) can be dissolved by even distilled water; while, 96.8% total Fe can be dissolved by 0.10 mol/l H 2 SO 4 solution. This supports the formations of FeSO 4 and Fe 2 (SO 4 ) 3 during baking. CONCLUSION Ilmenite, available in Bangladesh, containing TiFeO 3 and Fe 9 TiO 15 as main phases, can be baked with conc. H 2 SO 4 (1:1 by wt.) at 300 C for 30 min to produce a mass from which only iron can be leached out efficiently. A two stage baking at optimized condition (300 o C, 30 min, ilmenite / conc. H 2 SO 4 wt. ratio of 1) followed by single stage leaching by 0.10 mol/l H 2 SO 4 solution at boiling temperature under reflux for 20 min can extract more than 98% Fe from ilmenite. On the other hand, for baking at 200 o C, 90% iron is dissolved from a two stage baked mass; whereas, ~97% iron is dissolved from a three stage baked mass on leaching under identical condition. The changes during baking and leaching are demonstrated by XRD and SEM. The technique can be considered as a break through phenomena in titanium technology; as the method uses smaller amount of acid, comparatively lower temperature and comparatively little times in baking and leaching. The ultimate residue can be regarded as a rich Chloride feed for pigment grade TiO 2. REFERENCES [1] R. K. Biswas and M. G. K. Mondal, A study on the dissolution of ilmenite sand, Hydrometallurgy, vol. 17, pp , [2] M. A. B. Biswas, Lithology of recent titaniumzirconium places in the Bay of Bengal beach sands in the region of Cox s Bazar (Bangladesh): Dissertation for the degree of Doctor of Philosophy, Voronezh State University, Voronezh, USSR, pp , [3] R. K. Biswas, M. F. Islam and M. A. Habib, Dissolution of Ilmenite by roasting with LPGpyrolysed products and subsequent leaching, Ind. J. Eng. Mat. Sci., vol. 1, pp , [4] C. Sasikumar, D. S. Rao, S. Srikant, B. Ravikumar, N. K. Mukhopadhyay and S. P. Melhotra, Effect of mechanical activation on the kinetics of sulfuric acid leaching of beach sand ilmenite from Orissa, India, Hydrometallurgy, vol. 75, pp , [5] F. Islam, R. K. Biswas and C. M. Mustafa, Solvent extraction of Ti(IV), Fe(III) and Fe(II) from acidic sulphate media with HDTP benzene hexan 1 ol system: A separation and mechanism study, Hydrometallurgy, vol. 13, pp , [6] R. K. Biswas, M. A. Habib and N. C. Dafadar, A study on the recovery of titanium from hydrofluoric acid leach solution of ilmenite, Hydrometallurgy, vol. 28, pp , [7] R. K. Biswas and R. K. Jana, Crude electrorefining electrolyte obtained from ICC(Ghatsila, India) copper dust, Min. Proc. Extr. Metall, vol. 113, pp. C45-C52, Fig. 8. SEM images of (a) H 2SO 4 - baked mass and (b) residue obtained on leaching of H 2SO 4 - baked mass. 34

36 Thermal Treatment of Ilmenite on Moistening with Concentrated HF followed by Leaching with Dilute Sulfuric Acid Solution Ranjit K. Biswas, Mohammad A. Habib, Aneek K. Karmakar and Mohammad J. Alam Dept. of Applied Chemistry and Chemical Engineering Rajshahi University Rajshahi-6205, Bangladesh Abstract As it is very difficult to dissolve ilmenite (FeTiO 3 ) quantitatively, enormous studies on ilmenite dissolution have been patented and published. The objective of this dissolution is to prepare pigment grade TiO 2, or at least, the chloride feed grade TiO 2. A new technique for Fe dissolution from ilmenite is described in this work. Ilmenite has been moistened with conc. HF, heated (baked) at a higher temperature and leached with dilute H 2 SO 4 solution. The optimized baking temperature and time are found to be >170 C and 20 min for ilmenite to HF wt. ratio of 1, respectively. The baked mass can be leached by 0.5 mol/l H 2 SO 4 solution at its boiling temperature under reflux and at ilmenite to liquid ratio of 0.01 g/ml for 1 h to extract ~37% Fe (total) and 7% Ti from ilmenite sample. About 37% Fe-dissolution from the 1 st stage baked mass is increased to ~96% Fedissolution from the 6 th stage baked mass. Comparisons of the XRD pattern and SEM image of mother ilmenite with those of the residue obtained on 96% Fe removal indicates that ilmenite crystal is completely destroyed to form an amorphous-almost white product containing most of titanium present in ilmenite. As the residue contains ~4% Fe only, it can be regarded as a feed material for the chloride process of pigment grade TiO 2 manufacture. Keywords Ilmenite, HF, H 2 SO 4, Baking, Leaching, Chloride feed I. INTRODUCTION Titanium, being light and possessing high strength, is used in aircrafts and its oxide is extensively used in paint, plastic, paper, textile, rubber and ceramic industries. Their source materials are rutile and ilmenite. Previously, rutile was used to prepare pigment grade TiO 2 using the well-known chloride process in which TiO 2 -graphite mixture was heated in presence of Cl 2 gas to distill out TiCl 4 with other chlorides. Then TiCl 4 was decomposed by air to form TiO 2 and to regenerate Cl 2. But as the reserve of rutile in nature is depleted off, the TiO 2 -industry now depends solely on ilmenite. It can be processed either for synthetic rutile by eliminating Fe from it, or, for the pigment grade TiO 2 on its complete dissolution, solution purification by crystallization and solvent extraction, thermo-hydrolytic precipitation and ignition. There is a considerable of reserve of ilmenite in the beach sand of Bangladesh. The pilot plant beach sand exploitation centre of BAEC at Kalatali, Cox s Bazar has successfully fractionated beach sand into ilmenite, rutile, zircon, monazite, garnet etc. In processing ilmenite for production of pigment grade TiO 2, it is necessary to dissolve it completely, or at least one component (Fe or Ti) preferentially. This step still remains unsolved [1-7] resulting in low yield in the existing manufacturing technologies of TiO 2. However, methods have been developed for separating Fe(II)/Fe(III) from Ti(IV) existing in the leach solution obtained by partial leaching of both Ti and Fe from ilmenite [8-12]. Sulfuric and hydrochloric acid have been widely used for dissolving ilmenite [13-18]. But no method demands complete dissolution of ilmenite or of one component from ilmenite. There are a few reports on partial dissolution of ilmenite by HF [19-21]. As Bangladesh possesses a considerable reserve of ilmenite, it is worthy to work on processing of ilmenite to get either pigment grade TiO 2, or at least, rutile enriched feed for the chloride process. Previously [22], it was observed that copper dust on baking with 98% H 2 SO 4, followed by leaching with 0.05 mol/l H 2 SO 4, resulted in preferential leaching of Cu leaving almost all iron in the residue. Being inspired by this work and as the mineral acid baking of ilmenite is not yet reported, the present study on baking of ilmenite after moistening with conc. HF, followed by H 2 SO 4 leaching has been undertaken with an aim to its complete dissolution, or at least, preferential leaching of either Fe or Ti from it. II. EXPEIMENTAL A. Materials Ilmenite was collected from BAEC s Pilot Plant of Beach Sand Exploitation Centre at Kalatali (Cox s Bazar). The sample was found to contain 27.8% Ti and 32.4% Fe (both in Fe 2+ and Fe 3+ states). It was dry-ground and sieved to collect < 53 m sized particles to use in the most baking studies. The chemicals were A. R. grade E. Merck BDH products and used without further purification. B. Analytical The XRD patterns were taken from the PPPD center of BCSIR laboratories, Dhaka. The SEM images were taken from the Glass and Ceramic Engineering Department of BUET, Dhaka. Routine 35

37 analyses of Ti(IV) and Fe(III) in aqueous solutions were carried out by the H 2 SO 4 -H 2 O 2 method at 420 nm and HNO 3 -NH 4 SCN method at 480 nm, respectively, using a visible spectrophotometer (SP 1105, China). The HNO 3 -NH 4 SCN method estimated the amount of Fe(III) present in a solution; whereas, prior oxidation by boiling for 5 min with 2 ml conc. HNO 3 followed by the application of the HNO 3 - NH 4 SCN method yielded the amount of Fe(II) + Fe(III) in the solution. The difference gave the amount of Fe(II). C. Work Plan The plan of work is schematically depicted below (Fig. 1) which is self-explanatory: Ilmenite (1g) Conc. acid Quartz boat/ Platinum Crusible Left overnight and placed in furnace Small air pump Furnace Fig. 1. Flow sheet of the work plan D. Procedure for baking An aliquot of 1 g ilmenite (< 53 m size) was taken in a Pt-crucible and moistened with 1 g conc. HF and left overnight. With the help of a wire, tightened cross-sectionally at the middle, the container was placed in a tubular furnace (Carbolite CTF 12/65/550 type, UK) at a certain thermostated temperature ( C). The system was closed. Using a small air pump, air was passed through the furnace and allowed to flow from the other end via a plastic tube. Exit air carried the acid vapor for condensation in the tube and allowed to vent. After baking, the material with the container was taken out from the furnace, cooled and then quantitatively scratched over to a glossy paper for leaching. E. Procedure for leaching An aliquot of 100 ml of 0.50 mol/l H 2 SO 4 solution was taken in a 500 ml quick-fit conical flask fitted with a 1 m long glass tube acting as a condenser. The flask was set on a magnetic hot plate for heating as well as for pulp agitation using a magnetic capsule of 2 cm in length. When the boiling of the leaching agent started, the baked mass was added and switched on for time counting. When the ambient temperature was greater than 25 o C, the glass tube acting as condenser, was wrapped with moist cloth or cotton to facilitate condensation. After being leached for a specific time, the leached slurry was filtered and the filtrate was analyzed for metal ions contents. III. RESULTS AND DISCUSSION A. Characterization of Ilmenite Used in This Study The collected sized (<53 µm) ilmenite has been used for powder XRD, SEM, chemical analyses and baking studies. The XRD pattern as taken by Cu Kα radiation (40 kv, 30 ma, scan speed 0.02 o /s and step time 0.8 s) in Fig. 5 (i) indicates the presence of Coiled plastic tube for acid condensation Residue Baked mass Leaching assemble Filter To vent Filtrate Analysis for metal ions FeTiO 3 and Fe 9 TiO 15 along with some un-identified phases. However, a Japanese analysis indicates the presence of SiO 2, Fe 2 O 3, TiO 2, (Fe, Mg) Ti 2 O 5 and FeCr 2 O 4 as minor phases. The SEM image shows the crystallinity of different phases in the sample. The almost complete dissolution of the sample in molten KHSO 4 for about 6 h, subsequent solublization in 15% H 2 SO 4 solution and the colorimetric analysis of the resultant solution indicates that it contains 32.4% Fe (in both -ous and -ic states) and 27.8% Ti(IV). In the dissolution process, ~5% material remains undissolved. The other minor ingredients such as Si, Mn, Cr, Mg etc. are not quantified. On considering that the insoluble part does not contain any Fe and Ti, the extent of leaching (thermal treatment as well) has been monitored on the basis of 32.4% total Fe and 27.8% Ti existing in the sample. B. Optimization of Baking Temperature The progress in baking is monitored by the extents (%s) of Ti and Fe dissolved in succeeding leaching experiments conducted using 0.50 mol/l H 2 SO 4 solution at its boiling temperature under reflux. The effect of baking temperature is investigated for a constant baking time of 20 min. The effect of baking temperature on dissolution of ilmenite, within C, is given in Fig. 2, as weight %s of metal ions dissolved on subsequent leaching versus baking temperature plots. It is seen from this figure that the dissolution percentages of Ti(IV) and Fe(II) are little varied, but those of Fe(III) and total Fe are increased considerably, within C. The Ti(IV) dissolution percentage from baked mass up to 200 o C is within (7 )% and this is decreased to ~ 4.3% for baked mass at 225 o C. Although Fe(II)-dissolution percentages lie within ( ) C for baking temperature variation of o C, the Fe(III)- dissolution percentage is increased from ~ 11% for baked mass at 150 C to ~ 21% for baked mass at 225 C. Total Fe-dissolution percentage of 28 % for baked mass at 150 C is increased to 39% for baked mass at 175 o C and this value is little varied within C. This result predicts the preferential leaching of Fe over Ti(IV) from ilmenite. Ilmenite matrix is broken down at least partially by the treatment with HF at a temperature of ~ 170 C within only 20 min. C. Optimization of Baking Time The effect of baking time on subsequent leaching has been investigated. The HF moistened ilmenite has been baked at 200 o C for either 10, 20 or 30 min and the resultant masses are subjected to H 2 SO 4 acid solution leaching under identical condition to find out the leaching percentages of metal ions. The results from the above experiments are given as wt. % of metals dissolved versus time in min plots in Fig. 3. All metal ion dissolution %s show maxima at baking time of 20 min. From a baked mass at 200 o C for 20 min, a total Fe dissolution of 37% with < 7% dissolution of Ti(IV) are achieved. 36

38 Wt.% of metals dissolved Wt.% of metals dissolved (based on amounts present in ilmenite) Wt.% of metals dissoved International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering ( ) Total Fe; () Fe 3+ ; () Fe 2+ ; ( ) Ti 4+ chlorination process in order to obtain pigment grade TiO ( ) Total Fe () Fe () Fe ( ) Ti Baking temperature, o C Fig. 2. Effect of baking temperature of concentrated HF moistened ilmenite. Wt. of ilmenite sample = 1 g, particle size = < 53 µm, amount of concentrated HF added to moisten ilmenite = 1 g, baking time = 20 min, leaching agent = 100 ml 0.5 M H 2SO 4, leaching temperature = boiling temperature, leaching time = 1 hour. At around 150 C and above, ilmenite sample used in this study is principally broken down to TiO 2, TiOF 2, FeF 2, FeF 3 etc. during baking, according to the following reactions: TiFeO HF = TiO 2 + FeF 2 + H 2 O (1) TiO HF = TiOF 2 + H 2 O (2) Fe 9 TiO HF = TiO 2 + FeF 2 + 8FeF H 2 O (3) followed by dissolutions of FeF 2, FeF 3 and TiOF 2 in dilute H 2 SO 4 solution, leaving intact TiO 2 as a solid phase. The presence of rutile (TiO 2 ) and titanium oxide fluoride (TiOF 2 ) in the baked mass has been identified by XRD (see later). D. Stage wise baking Interesting result is obtained from the stage-wise baking study. An aliquot of 1 g ilmenite is just moistened with conc. HF (left over-night) and baked at optimized baking temperature of 200 o C for 20 min. The leaching of this baked mass was carried out as usual to calculate dissolution %s of metal ions for gathering single- stage baking results. To obtain the second stage baking result, the baked mass obtainable from the first stage was again moistened with conc. HF, baked and leached as usual. In the similar way, for six - stage leaching, addition of HF to ilmenite and baking was continued for 6 times. The results obtained on stage-wise baking followed by leaching at identical conditions are shown as the dissolution wt. % of metal ions versus baking stage number plots in Fig. 4. The Ti(IV)-dissolution of ~7% from singlestage baked mass is gradually decreased to only 1% dissolution from the sixth - stage baked mass. On the other hand, though Fe(III) dissolution percentage is decreased to some extent with the increase in baking stage number, Fe(II) and total Fe-dissolutions are remarkably increased with the increase in baking stage number. The Fe(II)-dissolution of 16% from the single - stage baked mass is increased to 82% from the sixth - stage baked mass. Likewise, total irondissolution of 37% from the single - stage baked mass is increased to 96% from the sixth - stage baked mass. Therefore, a procedure involving six-stage baking followed by single-stage leaching with 0.50 mol/l sulphuric acid solution produces a mass from which 96% iron can be leached out leaving 99% titanium in the solid state. The residual product is almost white and can be used as a raw material for Baking time, min Fig. 3. Effect of baking time of concentrated HF moistened ilmenite. Wt. of ilmenit sample = 1 g, particle size = < 53 µm, amount of concentrated HF added to moisten ilmenite = 1 g, baking temperature = 200 o C, leaching agent = 100 ml 0.5 M H 2SO 4, leaching temperature = boiling temperature, leaching time = 1 hour ( ) Total Fe () Fe 3+ () Fe 2+ ( ) Ti Number of baking stage Fig. 4. Effect of stage-wise baking on metals dissolution from ilmenite. Wt. of ilmenite sample = 1 g, particle size = < 53 µm, amount of concentrated HF added to moisten ilmenite = 1 g in each stage, baking temperature = 200 o C, leaching agent = 100 ml 0.5 M H 2SO 4, leaching temperature = boiling temperature, leaching time = 1 hour. E. XRD patterns and SEM images The XRD patterns of ilmenite sample used in this study and of ilmenite-hf baked mass of sixth stage are shown in Fig. 5, for comparison. It is seen that the diffraction peaks for TiFeO 3 and Fe 9 TiO 15 are almost disappeared in the XRD pattern of baked mass with the appearance of diffraction peaks for rutile (TiO 2 ) and titanium oxide fluoride (TiOF 2 ). Therefore, Eqs. (1)-(3) suggested above for baking reactions appear true. The incomplete baking in single stage might be due to the volatility of HF at the baking temperature region and also due to higher stoichiometric HF acid requirement. The XRD pattern of the baked-leached residue appears as of amorphous substance. The SEM images of ilmenite sample used in this study and of baked leached residue are given in Fig. 6. The crystals of species present in the sample disappear to form amorphous lumps in baked-leached residue. IV. CONCLUSION The baking of ilmenite after moistening with conc. HF at 200 C for 20 min produced a mass from which iron can be leached out efficiently. A six stage baking at optimized condition (200 C, 20 min, ilmenite moistened with conc. HF) followed by single stage leaching by 0.50 mol/l H 2 SO 4 solution at 100 o C under reflux for 1 h can extract ~96% Fe from 37

39 ilmenite. The changes during baking and leaching are demonstrated by XRD and SEM. This technique can be regarded as another break through phenomena in titanium technology; as the method uses smaller amount of acid, comparatively lower temperature and little times in baking and leaching. The ultimate residue can be regarded as a rich chloride feed for pigment grade TiO 2. (i) ( ), : ilmenite; (), : Fe9TiO15 Fig. 5. X-ray diffraction pattern of (i) ilmenite (ii) ilmenite - hydrofluric acid baked mass. Fig. 6. SEM images of (a) Ilmenite sample and (b) residue left after leaching of HF - baked mass. REFERENCES ( ), : Rutile; (), : Titanium Oxide Fluoride (a) (ii) [1] R. K. Biswas and M. G. K. Mondal, A study on the dissolution of ilmenite sand, Hydrometallurgy, vol. 17, pp , [2] M. A. B. Biswas, Lithology of recent titaniumzirconium places in the Bay of Bengal beach sands in the region of Cox s Bazar (Bangladesh): Dissertation for the degree of Doctor of Philosophy, Voronezh State University, Voronezh, USSR, pp , [3] R. K. Biswas, M. F. Islam and M. A. Habib, Dissolution of Ilmenite by roasting with LPGpyrolysed products and subsequent leaching, Ind. J. Eng. Mat. Sci., vol. 1, pp , [4] R. K. Biswas, M. F. Islam and M. A. Habib, Processing of ilmenite by roasting with the reformed product of LPG-water mixture in presence of nickel powder and subsequent leaching, Bang. J. Sci. Ind. Res., vol. 33, pp , [5] R. K. Biswas, M. F. Islam and M. A. Habib, Processing of ilmenite through sodium carbonatewater vapor roasting and leaching, Bang. J. Sci. Ind. Res., vol. 32, pp , [6] R. K. Biswas, M. F. Islam and M. A. Habib, Processing of ilmenite through salt-water vapour roasting and leaching, Hydrometallurgy, vol. 42, pp , (b) [7] M. A. Habib, R. K. Biswas, P. K. Sarkar and M. Ahmed, Leaching of ilmenite by mixed solvents and their kinetics, Bang. J. Sci. Ind. Res., vol. 38, pp. 1-12, [8] C. Sasikumar, D. S. Rao, S. Srikant, B. Ravikumar, N. K. Mukhopadhyay and S. P. Melhotra, Effect of mechanical activation on the kinetics of sulfuric acid leaching of beach sand ilmenite from Orissa, India, Hydrometallurgy, vol. 75, pp , [9] F. Islam, M. Ali and S. Akhter, Separation and recovery of titanium from iron bearing leach liquors by solvent extraction with di-2-ethyl hexyl phosphoric acid-tributyl phosphate-thiocyanate system, Bang. J. Sci. Ind. Res., vol. 13, pp , [10] F. Islam and Z. Kawnin, Separation and recovery of titanium from iron bearing leach liquors by solvent extraction with di-2-ethyl hexyl phosphoric acid, Bang. J. Sci. Ind. Res., vol. 13, pp , [11] F. Islam, H. Rahman and M. Ali, Solvent extraction separation study of Ti(IV) Fe(III) and Fe(II) from aqueous solutions with di-2-ethyl hexyl phosphoric acid in benzene, J. Inorg. Nucl. Chem., vol. 41, pp , [12] F. Islam, R. K. Biswas and C. M. Mustafa, Solvent extraction of Ti(IV), Fe(III) and Fe(II) from acidic sulphate media with HDTP benzene hexan 1 ol system: A separation and mechanism study, Hydrometallurgy, vol. 13, pp , [13] R. K. Biswas, M. A. Habib and N. C. Dafadar, A study on the recovery of titanium from hydrofluoric acid leach solution of ilmenite, Hydrometallurgy, vol. 28, pp , [14] A. A. Baba, S. Swaroopa, M. K. Ghosh and F. A. Adekola, Mineralogical characterization and leaching behavior of Nigerian ilmenite ore, Trans. Nonfer. Met. Soc. China, vol. 23, pp , [15] L. Jia, B. Liang, L. Lu, S. Yuan, L. Zheng, X. Wang and C. Li, Beneficiation of titania by sulfuric acid pressure leaching of Panzhihua ilmenite, Hydrometallurgy, vol. 150, pp , [16] M. H. H. Mahmood, A. A. Afifi and I. A. Ibrahim, Reductive leaching of ilmenite ore in hydrochloric acid for preparation of synthetic rutile, Hydrometallurgy, vol. 73, pp , [17] T. Ogasawara and R. V. Veloso de Araujo, Hydrochloric acid leaching of a pre-reduced Brazilian ilmenite concentrate in an autoclave, Hydrometallurgy, vol. 56, pp , [18] V. S. Gireesh, V. P. Vinod, S. K. Nair and G. Ninan, Catalytic leaching of ilmenite using hydrochloric acid: A kinetic approach, Int. J. Miner. Processing, vol. 134, pp , [19] R. K. Biswas and M. G. K. Mondal, A study on the dissolution of ilmenite sand, Hydrometallurgy, vol. 17, pp , [20] K. Nagasubramanian and K. Liu, Recovery of TiO 2 from ilmenite type ore by a membrane based electrodialysis process, US Patent No. 4,107,264, [21] D. A. Hansen and D. E. Traut, The kinetics of leaching rock ilmenite with hydrofluoric acid, J. Metals, vol. 41, pp , [22] R. K. Biswas and R. K. Jana, Crude electrorefining electrolyte obtained from ICC(Ghatsila, India) copper dust, Min. Proc. Extr. Metall, vol. 113C, pp ,

40 Solvent Extraction of V(V) from Nitrate Medium by Tri-n- Octylamine Dissolved in Kerosene Ranjit K. Biswas, Aneek K. Karmakar and Mottakin Dept. of Applied Chemistry and Chemical Engineering Rajshahi University Rajshahi-6205, Bangladesh Abstract The liquid-liquid extraction of V(V) from nitrate medium by tri-n-octylamine (TOA) dissolved in distilled colorless kerosene has been investigated as functions of various experimental parameters. Equilibration time is less than 10 min. HNO 3 is found to be extracted by TOA (HNO 3 + (C 8 H 17 ) 3 N (C 8 H 17 ) 3 NH.NO 3 ). The nature of species extracted into the organic phase depends on the existing aqueous species prevailed at a certain ph. At lower ph regions, the extraction of VO + 2 occurs via cation (H + ) exchange of (C 8 H 17 ) 3 NH.NO 3 as follows: VO + 2 +NO 3 -H.TOA VO 2 NO 3.TOA + H +. On the other hand, at higher ph region, anionic V(V) species such as V 10 O 26 (OH) 4-2, V 10 O 27 (OH) 5-6-, V 10 O 28 etc. are extracted by solvated 4- ion-pair formation mechanism (eg. V 10 O 26 (OH) 2 +2H + +TOA H 2 V 10 O 26.TOA + 2H 2 O). The TOA concentration dependence depends on ph region of extraction. It is seen that the extraction ratio increases with the increase in V(V) concentration in the aqueous phase which is possibly due to the formation of extraction favorable V(V) species with increasing its concentration in the aqueous phase. The extraction is also found to be favored by the rise of nitrate concentration in the aqueous phase; and the variation may be quantitatively expressed by: log D = const. + log (1+1.82[NO - 3 ]); where, const. depends on ph and [TOA] used. Temperature has a pronounced effect with H < 55 kj/mol. The extracted species can be stripped by 0.75 mol/l NH 4 OH solution to the extent of 71%; but stagewise stripping is not so effective. A very high loading of ~2.3 mol V(V) per mol/l TOA is observed. Keywords Extraction, stripping, V(V), TOA, loading, NO 3 - -medium I. INTRODUCTION Vanadium is extensively used as an alloying element for high strength steel used in the manufacture of pipe lines, rail lines, axles and crankshafts for motor vehicles, high speed tool steel, jet engines etc. Its non-steel application fields include welding, nuclear engineering, superconductor, catalyst, fuel cell etc. In 2012, 74,000 tons of vanadium was produced worldwide [1]. However, there is no rich ore of V; and consequently, it is manufactured generally from petroleum fly ashes, tar sand fly ashes, black shape desulphurization waste catalyst, carnolite etc. using the hydrometallurgical route involving roasting, leaching, extraction, stripping and precipitation. On leaching, vanadium dissolves either as V(IV) or V(V). Vanadium (IV) can be easily oxidized to V(V) by boiling with HNO 3 ; and V(V) can be easily reduced by Na 2 SO 3 (SO 2 ) solution. The extraction of both V(IV) and V(V) by various extractants have been investigated substantially [2-15]. As species variation of V(IV) is limited, mechanistic studies on V(IV) extraction are simple in nature. On the other hand, V(V) exists as VO 2 + a ph < 1 and starts to hydrolyze and polymerization with increasing ph to form species such as VO 7 (OH) 3, V 10 O 26 (OH) 2 4-, V 10 O 27 OH 5-, V 10 O 28 6-, V 3 O 9 3-, etc. within acidic ph region. It appears, therefore, that the extraction of V(V) in ph region 1-6 by various extractants are little investigated from the mechanistic point by view, though there are several reports on separation of V(V) from other metal ions using various extractants. Consequently, the extraction of V(V) from nitrate medium by TOA is reported in the paper from mechanistic point of view. II. EXPEIMENTAL A. Materials Tri-n-octylamine (TOA) was collected from Tokyo Kaie Ind (90%) and was used without further purification. It was diluted with distilled colorless kerosene to constitute the organic phase. All other chemicals were of reagent grade and used without further purifications. Kerosene was bought from the local market and distilled to collect colorless fraction distilling over C. Requisite amount of NH 4 VO 3 was dissolved in 170 ml 6 mol/l HNO 3 solution and diluted to 500 ml to obtain a stock solution containing 5 g/l V(V) and 1.92 mol/l HNO 3. In most cases, this solution was 10 times diluted to obtain test solutions containing 0.5 g/l V(V) and mol/l HNO 3 or NO - 3. B. Analytical A double buffer calibrating ph meter (mettle toledo 220) was used for ph measurement and adjustment of aqueous phase by the addition of anhydrous Na 2 CO 3. Aqueous V(V) concentration was determined by the H 2 O 2 colorimetric method [16] at 450 nm using a visible spectrophotometer () C. Procedure for extraction All studies excepting the temperature dependence were carried out at (298±1) K using a thermostatic water bath. Equal volumes (20 ml) of aqueous and organic phases were taken in a 100 ml reagent bottle, stoppered and shaken well, in a thermostatic water bath for predetermined time of 10 min usually for equilibration. On equilibration, phases were settled and disengaged. Aqueous phase was analyzed for its 39

41 log D log D log D International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering V(V) content and equilibrium ph. The extraction ratio D was then calculated using the following relation: D = (C V(V) O at equilibrium)/(c V(V) A at equilibrium) = (C V(V) (ini) A C V(V) (eq) A )/(C V(V) A at equilibrium) (1) D. Procedure for loading An aliquot of the organic phase (20 ml 0.10 mol/l TOA) was repeatedly contacted for 5 min in each stage with fresh aqueous solution containing 2 g/l V(V) at ph 1 and 298 K. After each stage extracted V(V) concentration was estimated to calculate cumulative concentration of V(V) in the organic phase. E. Procedure for stripping This was virtually similar to the procedure for extraction. In this case, the organic phase containing V(V) as solvated ion pair with TOA was equilibrated with equal volume of NH 4 OH solution for 10 min. On equilibration, phases were separated on settling and the aqueous phase was analyzed for V(V) content. Then stripping percentage (S, %) was calculated as follows: S, % = ((C V(V) (eq) A ) 100)/(C V(V) (ini) O ) (2) III. RESULTS AND DISCUSSION Preliminary experiment showed that the extraction process under consideration, was quite fast (5 min required for equilibration). However, subsequent experiments were carried out with equilibration time of 10 min to ensure the attainment of equilibration at different conditions. The effect of [V(V)] on extraction ratio is depicted in Fig. 1 as log D vs. log {[V(V)], mol/l} plots for ph (ini) of 1 but at [TOA] of 0.01, 0.05 and 0.10 mol/l. It is found that the extraction ratio is extensively increased with increasing [V(V)] in the aqueous phase. The broken line represents the data at ph (eq) of 2.85 for 0.10 mol/l TOA system. In this case, also D is increased with increasing [V(V)] with a slope of ~2.5. In general D value should remain unchanged with the variation of metal ion concentration if extractable species variation is absent there. The result is the indicative of species variation with increasing [V(V)]. Not all V(V) species is extractable by TOA; rather particular [V(V)]-species is susceptible to be extracted by TOA and the concentration of this species is increased with increasing [V(V)] in the aqueous phase. The variations of extraction of V(V) with aqueous ph are displayed in Fig. 2, as log D vs. ph (ini) and in Fig. 3, as log D vs. ph (eq) plots for [TOA] of 0.03, 0.05 and 0.10 mol/l. Initial ph is varied significantly on extraction; and so the isotherms given in Fig. 2 are not representative to actual systems. On the other hand, the isotherms given in Fig. 3 are the better representative to the system (some error still exists owing to [TOA] variation on different extents of extractions). For cation exchange reactions, the phdependency (slope of log D vs. ph plot) should be positive due to liberation of H + by extraction reaction. The negative ph-dependency (negative slope of log D vs. ph plot) is indicative of the association of H + with existing V(V) species to form the extractable species log [V(V)], mol/l Fig. 1. Effect of [V(V)] on extraction. ph (ini) = 1, [TOA] = 0.01 (), 0.05 () and 0.10 ( ) mol/l, Temp. = 298 K, O/A = 1, time = 10 min. ph (eq) = (), 1.3±0.18; (), 2.3±0.20; ( ), 2.85±0.6. Dashed line ( ), ph (eq) = 2.85, [TOA] = 0.10 mol/l ph (ini) Fig. 2. Effect of initial ph on extraction. [V(V)] (ini) = 0.50 g/l, temp. = 300 K, O/A = 1, time = 10 min, ( ), [TOA] = 0.10 mol/l; ( ), [TOA] = 0.05 mol/l and (), [TOA] = 0.03 mol/l ph (eq) Fig. 3. Effect of equilibrium ph on extraction. [V(V)] (ini) = 0.50 g/l, temp. = 300 K, O/A = 1, time = 10 min, ( ), [TOA] = 0.10 mol/l; ( ), [TOA] = 0.05 mol/l and (), [TOA] = 0.03 mol/l. According to Zeng and Yong Cheng [17], yellow colored VO + 2 exists within ph 1-2; whereas, within ph 2-6.5, orange-red V 10 O exists. However, it is also reported [7] that with the gradual of increase of ph from zero, VO + 2 is gradually transformed to VO(OH) 3 (to a small extent), V 10 O 26 (OH) 4-2, V 10 O 27 (OH) 5-, V 10 O 6-28, V 3 O 3-9, V 4 O , VO 2 (OH) 2 4- etc. V 10 O 26 (OH) 2 is virtually doubly protonated V 10 O 6-28 ; whereas, V 10 O 27 OH 5- is single protonated V 10 O The distribution of these species as function of ph are also available [7] as shown in Table I. 40

42 log D log D log D International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering Table I. Distribution of V(V)-species at different ph values [7]. ph % (approximate) Theoretical ph dependence + VO 2 VO(OH) 3 4- V 10H 26(OH) 2 V 10H 27OH 5-6- V 10O V 3O 9 - VO 2(OH) 2 4- V 4O TOA is found to extent HNO 3 just by acid-base reaction as shown below: (C 8 H 17 ) 3 N + HNO 3 (C 8 H 17 ) 3 NH.NO 3 (3) triocttylammonium salt of nitrate The product may extract low ph existing predominant species (VO 2 + ) according to the following reaction: VO (C 8 H 17 ) 3 NH.NO 3 (C 8 H 17 ) 3 NVO 2.NO 3 + H + (cation exchange) (4) S = 2.6 S = 1.2 S = 0.8 S = 0.5 At higher ph values, extraction occurs via the following reaction: V 10 O 26 (OH) H + + ntoa 4H 2 V 10 O 26.nTOA +2H 2 O (ion pair solvation) (5) V 10 O 27 (OH) H + + ntoa H 4 V 10 O 27.nTOA + H 2 O (ion pair solvation) (6) V 10 O H + + ntoa H 6 V 10 O 28.nTOA (7) (ion pair solvation) Rer V(V), the extractant dependence for reaction represented by Eqs. (4), (5), (6) and (7) are +1, -0.4, and -0.6, respectively. It is, therefore, possible to calculate the theoretical ph dependence at a given equilibrium ph (eg. at ph (eq) = 2.50, ph dependence = (-0.4) = ; and at ph (eq) = 3.0, ph dependence = (-0.4)+0.15 (-0.5) = ). The theoretical ph dependences at various ph (eq) values are shown in the last column of Table I. Experimental results on ph (eq) dependence are of similar nature; positive sloped to zero sloped to negative sloped curves with increasing ph (eq) value are obtained consequently, above extraction reactions satisfy well the experimental ph-dependence. The extractant dependence curve is shown in Fig. 4 for four sets of experimental parameters. In all cases, straight lines with positive slopes are obtained. In each case, though initial ph was kept constant, equilibrium ph was varied. No correction is made here. The positive extractant dependence supports the solvation of ion-pair by TOA, but the number of TOA per ion-pair varies with ph (eq). - The effect of NO 3 on extraction is presented in Fig. 5. The log D vs. log {[NO - 3 ], mol/l} plot is a curve which can be presented by Eqn: log D = log (1+1.82[NO - 3 ]). Therefore, the extraction ratio is independent of [NO - 3 ] in its lower concentration region, but it is directly proportional to [NO - 3 ] in its higher concentration region. The results support the extraction of HNO 3 by TOA log {[TOA], mol/l} Fig. 4. Effect of [TOA] on extraction. [V(V)] (ini) = 0.50 g/l, temp. = 298 K, O/A = 1, time = 10 min, ( ), ph (ini) = 1, ph (eq) = 2.3±1.1; ( ), ph (ini) = 2, ph (eq) = 4.1±1.0; (), ph (ini) = 3, ph (eq) = 4.25 ± 0.55; ( ), ph (ini) = 5, ph (eq) = 4.25± log {[NO - ], mol/l 3 Fig. 5. Effect of [NO - 3 ] on extraction. [V(V)] (ini) = 0.50 g/l, [TOA] = 0.01 mol/l, ph (ini) = 1, Temp. = 300 K, Curve is theoretical: log D = log (1+1.82[NO - 3 ]); whereas, points are experimental S = 3.25 x 10 3 H = 62 kj/mol S = 2.8 x 10 3 H = 54 kj/mol (1/T)x10 3, K -1 Fig. 6. Effect of temperature. [V(V)] (ini) = 0.5 g/l, O/A = 1, time = 10 min, [NO - 3 ] = mol/l, ph (ini) = 1; ( ), [TOA] = 0.10 mol/l, ph (eq) = 2.96; ( ), [TOA] = 0.05 mol/l, ph (eq) = Figure 6 represents the vant Hoff s plots for the extraction system. It is seen that with the rise of 41

43 Cummulative [V(V)] (org), g/l International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering temperature, extraction ratio is decreased. The extraction process is, therefore, exothermic with H value of about -58 kj/mol Loading of TOA by V(V) is shown in Fig. 7. It is seen that the organic phase is almost saturated with V(V) at about 12 th contact when 0.10 mol/l TOA solution is repeatedly contacted with 2 g/l V(V) solution at ph (ini) = 1. The cumulative [V(V)] increases gradually with contact number and almost levels off at around 12 th contact. At O/A = 1 in each contact, 0.10 mol/l TOA can extract as much as g V(V)/L; consequently, the loading capacity is g V(V)/mol TOA. It is exceedingly a very high value demanding commercialization. Stripping of the extracted V(V) can be done by 0.75 mol/l NH 4 OH solution to the extent of 72% in single stage. Stage-wise stripping is not so impressive. IV. CONCLUSION Vanadium (V) can be well extracted by tri-noctylamine within 5 min contact. Extraction ratio is found to increase with increasing [V(V)], [NO 3 - ] and [H + ] in the aqueous phase and with increasing [TOA] in the organic phase. More than 90% V(V) can be extracted at ph (eq) = 2 by 0.10 mol/l TOA solution. Mechanism of extraction is suggested. Rise of temperature decreases extraction. 1 mole TOA can extract as much as 2.3 mol V(V) at saturated loading. Extracted species can be stripped by 0.75 mol/l NH 4 OH solution to the extent of 72% in single stage; but stage-stripping is not fruitful Contact number Fig. 7. Loading of V(V) in the organic phase. [V(V)] = 2 g/l, [TOA] = 0.1 mol/l, ph (ini) = 1, Temp. = 298 K. REFERENCES [1] [2] F. Islam and R. K. Biswas, The solvent extraction of vanadium(iv) from acidic sulhate-acetato solutions with HDEHP in benzene and kerosene, J. Inorg. Nucl. Chem., vol. 42, pp , [3] R. K. Biswas and M. G. K. Mondal, Kinetics of VO 2+ extraction by D2EHPA, Hydrometallurgy, vol. 69, pp , [4] M. A. Hughes and R. K. Biswas, Kinetics of vanadium(iv) extraction in acidic sulphate-d2ehpan-hexane system using Rotating Diffusion Cell Technique, Hydrometallurgy, vol. 26, pp , [5] K. O. Ipinmoroti and M. A. Hughes, Mechanism of V(IV) extraction in a chemical kinetic controlled regime, Hydrometallurgy, vol. 24, pp , [6] P. Nekovar and D. Schrotterova, Extraction of V(V), Mo(VI) and W(VI) polynuclear species by primene JMT, Chem. Eng. J., vol. 79, pp , [7] M.A. Olazabal, M. M. Orive, L. A. Fernandez and J. M. Madariaga, Selective extraction of V(V) from solutions containing molybdenum (VI) by ammonium salts dissolved in toluene, Solvent Ex. Ion Exch., vol. 10, pp , [8] P. Zhang, K. Inoue, K. Yoshizuka and H. Tsuyama, Extraction and selective stripping of molybdenum (VI) and vanadium (IV) from sulfuric acid solution containing aluminum (III), cobalt (II), nickel (II) and iron (III) by LIX 63 N in Exxsol D80, Hydrometallurgy, vol. 41, pp , [9] S. Jayadas and M. L. Reddy, Solvent extraction separation of V(V) from multivalent metal chloride solutions using 2-ethylhexyl phosphonic acid mono-2- ethylhexyl ester, J. Chem. Tech. Biotech., vol. 77, pp , [10] R. K. Biswas and A. K. Karmakar, Liquid liquid extraction of V(IV) from sulfate medium by Cyanex 301 dissolved in kerosene, Int. J. Nonfer. Met., vol. 2, pp , [11] T. Sato, S. Ikoma and T. Nakamura, Solvent extraction of vanadium (IV) from hydrochloric acid solutions by neutral organophosphorous compounds, Hydrometallurgy, vol. 6, pp , [12] V. L. Bykhovtsov and G. N. Melixhova, Extraction of quinquevalent vanadium with a technical mixture of trialkylphosphine oxides, Zh. Prikl. Khim., vol. 43, pp , [13] S. Kopacz and L. Paidovski, Extraction of vanadium (V) from sulphuric acid solutions by aliphatic alcohols, Russ. J. Inorg. Chem., vol. 16, pp , [14] Y. Anjaneyulu, B. S. R. Sarma and V. P. R. Rao, Studies on the influence of some inorganic amines on the extraction of vanadium (V)-oxine complex, Ind. Chem. Soc., vol. 4, pp , [15] A. Saily and S. N. Tandon, Liquid-liquid extraction behavior of V(IV) using phosphinic acids as extractants, Fresenius J. Anal. Chem., vol. 360, pp , [16] J. Bassette, R. C. Denny, G. H. Jeffery and J. Mendham, Vogel s Textbook of Quantitative Inorganic Analysis, 4 th Edn., ELBS, London., p. 752, [17] L. Zeng and C. Y. Cheng, A literature review of recovery of molybdenum and vanadium from spent hydrodemlphusisation catalysts; Part I and II, Hydrometallurgy, vol. 98, pp. 1-20,

44 Kinetics of Extraction of Ti(IV) from SO 4 2- Medium by Cyanex 301 Dissolved in Kerosene Ranjit K. Biswas and Aneek K. Karmakar Dept. of Applied Chemistry and Chemical Engineering Rajshahi University Rajshahi-6205, Bangladesh rkbiswas694@gmail.com Abstract The kinetics of Ti(IV)-extraction by Cyanex 301 (HA) have been investigated by measuring initial flux of Ti(IV)-transfer using a Lewis cell operated at 3 Hz. The empirical flux eqation at 298 K is found to be: F (kmol/m 2 s) = [Ti(IV)](1+447[H + ]) - 1 [HA] (o) (1+1.18[SO 4 2- ]) -1. The activation energy, E a has been measured to be kj/mol depending on experimental parameters and temperature region. The S ± value is always highly negative. Analysis of the flux equation has been done, at various parametric conditions, to elucidate the mechanism of extraction. The rate determining chemical reaction step, at all parametric conditions appears as: TiO 2+ A - TiOA + ; and this step occurs via and S N 2 mechanism. Keywords Kinetics, extraction, Ti(IV), Cyanex 301, Lewis cell I. INTRODUCTION Bangladesh has a considerable reserve of ilmenite in the beach sand, which is now considered as the principal raw material for Ti-technology. Ilmenite can be dissolved in sulfuric acid using various techniques to yield solution containing Ti(IV), Fe(II) and Fe(III), principally [1-4]. The separation of Ti(IV) from Fe(II) and Fe(III) can effectively be carried out through extraction by organophosphorous extractants [5-8]. Recently [9], the extraction behavior of Ti(IV) by Cyanex 301 in kerosene-5% heptan-1-ol has been reported from the equilibrium point of view. This paper reports the kinetics of this process using Lewis cell technique to elucidate the rate parameters and the mechanism of extraction. II. EXPEIMENTAL A. Materials Cyanex 301 was supplied by Cytec Canada Inc. and used without further purification. Locally available kerosene was distilled to collect colorless fraction obtained within K. Other chemicals were of A. R. grade (E. Merck-BDH) products and used without further purification. B. Analytical The aqueous Ti(IV) concentration was estimated by the H 2 SO 4 -H 2 O 2 method at 420 nm [10] using a uvvisible spectrophotometer (UV-1650 PC, Shimadzu, Japan). A Mettler Toledo ph meter was used for ph measurement. For ph adjustment, either anhydrous Na 2 CO 3 or dilute H 2 SO 4 was used. C. Preparation of solutions of Ti(IV) from TiO 2 The stock solution of Ti(IV) was prepared by digesting 50 g TiO 2 in 50 ml conc. H 2 SO 4 for 2 h with constant stirring at a low heat ( K) followed by its partial lixiviation in 15% (v/v) H 2 SO 4 solution. The insoluble part was filtered out to obtain 1 L solution containing g Ti(IV) and 3.45 mol SO On the other hand, the standard solution of Ti(IV) was prepared on fusion of 1 g TiO 2 with 10 g KHSO 4 (Pt- crucible, 2 h) followed by cooling and complete dissolution in 15% (v/v) H 2 SO 4 solution to obtain 1 L solution (1 ml = 0.60 mg Ti(IV)). The former solution on proper dilution, sulfate addition and ph adjustment was used in flux measurement; whereas the later solution was used for construction of calibration curve for colorimetric analysis of Ti(IV). D. Preparation of organic phase Unlike other acidic organophosphorous acids, Cyanex 301 is monomeric [11]; and on this basis, ml of Cyanex 301 (Mol. wt. 322 and density 0.95 g/ml) was diluted by distilled kerosene (containing 5% (v/v) heptan-1-ol) to 250 ml for preparing 2 mol/l Cyanex 301 stock solution. This solution was properly diluted by distilled kerosene containing 5% (v/v) heptan-1-ol (de-emulsifier) for use in flux measurement. E. Cell and operating procedure The cell consists of a jacketed glass container facilitating temperature control. An aliquot (100 ml) of aqueous solution is taken in this container. The container has an air-tight lid having three bores: one for inserting the shaft of an electric stirrer, another for introducing a funnel having bend-tail directed towards the wall of the beaker, to pour down 100 ml organic phase without much disturbance of the interface, and the rest for introducing a glass tube to take out aqueous solution from the middle section of the aqueous phase with the aid of a polythene tube and a syringe. An electric stirrer having stirring blades at two levels (1 cm long two blades at each level) rotating at 3 Hz enables phase agitation without interface waving. After certain interval, 2 ml aqueous phase is taken out for analysis of its Ti(IV)- content. The interfacial area is kept at m 2, which can be changed by setting the circular plastic rings within the container where interface being formed. 43

45 log (F, kmol/m 2 s) log (F, kmol/m 2 s) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering F. Treatment of experimental data The flux (F) of Ti(IV)- transfer can be calculated from the relation: F (kmol/m 2 s) = /A i.t (1) At a constant temperature, F is related to concentration terms as follows: F (kmol/m 2 s) = k f [Ti(IV)] a [H + ] b [HA] (o) c [SO 4 2- ] d (2) On taking logarithm of both sides of Eq. (2), one gets: log(f, kmol/m 2 s) = logk f + a log[ti(iv)] + b log[h + ] + clog [HA] (o) + d log [SO 4 2- ] (3) Equation (3) states that when [H + ], [HA] (o) and [SO 4 2- ] are kept constant at (h), (ha) (o) and (so 4 2- ), respectively, then the plot of log F vs. log [Ti(IV)] should yield a straight line with s equaling to a and I equaling to log k f (h) b (ha) (o) c (so 4 2- ) d from which the value of k f can be evaluated after finding the values of b, c and d. These values can be obtained from slopes of the log F vs. log [H + ] (or, ph for which s = -b), log F vs. log [HA] (o) and log F vs. log [SO 4 2- ] plots, respectively. The effect of temperature can be treated by the Arrhenius equation: log F = constant E a /2.303 R T (4) Moreover, these can also be treated by the Activated Complex Theory of reaction rate: log(fh/kt) = -H ± /2.303 RT + S ± /2.303 R + logf (R) (5) III. RESULTS AND DISCUSSION The effects of A i, [Ti(IV)], [H + ], [HA] and [SO 4 2- ] on F have been measured. It is seen that flux is independent of A; which means that the cell of any dimension can be used for F measurement. The log F vs. log [Ti(IV)] plots are shown in Fig. 1. For each parameter, pairs of straight line intersecting of [Ti(IV)] of ~1.25 g/l is obtained. At lcr, s 1 i.e. a in Eq. (2) is unity. At hcr, s = a = a negative value is contrary to the general principle of chemical kinetics. It is possibly due to formation of non-extractable polymerized aqueous Ti(IV) species that starts formation at [Ti(IV)] near to 1.2 g/l. The log F vs. ph plots are not straight lines; rather curves are obtained which can be fitted to Eq.: log F = constant - log ( [H + ]). This equation indicates that log F vs. log ( [H + ]) plot would be a straight line with unity slope. Such plots are shown in Fig. 2. It is therefore concluded that F ( [H + ]) -1 ; i.e. rate of extraction is directly proportional to ph in its lower region; whereas, it is independent of ph in its higher region. The plots in Fig. 3 represents the variation of flux with [HA] (o). In each case, log F vs. log [HA] (o) plot is a straight line with unity slope. Therefore, the rate of extraction is directly proportional to the extractant concentration in the organic phase ( ), ph = 2.50, [HA](o) = 0.06 mol/l, Temp. = 293 K; s = 0.98, I = ( ), ph = 2.50, [HA](o) = 0.40 mol/l, Temp. = 293 K; s = 0.95, I = (), ph = 1.60, [HA](o) = 0.40 mol/l, Temp. = 293 K; s = 0.98, I = ( ), ph = 2.50, [HA](o) = 0.40 mol/l, Temp. = 318 K; s = 0.95, I = log {[Ti(IV)], mol/l} Fig. 1. Effect of [Ti(IV)] on flux. [SO 4 2- ] = 0.50 mol/l, A i = m 2, o/a = 1 (o = 100 ml) ( ), [Ti(IV)](ini) = 1000 mg/l, [HA](o) = 0.06 mol/l, Temp. = 293 K; I = ( ), [Ti(IV)](ini) = 1000 mg/l, [HA](o) = 0.40 mol/l, Temp. = 293 K; I = (), [Ti(IV)](ini) = 1000 mg/l, [HA](o) = 0.06 mol/l, Temp. = 318 K; I = ( ), [Ti(IV)](ini) = 100 mg/l, [HA](o) = 0.40 mol/l, Temp. = 318 K; I = log (1+447x10 -ph ) Fig. 2. Effect of ph on flux. [SO 4 2- ] = 0.50 mol/l, Ai = m 2, o/a = 1 (o = 100 ml). The log F vs. log [SO 4 2- ] plots are curves which can be fitted to Eq.: log F = constant log ( [SO 4 2- ]). This equation indicates that log F vs. log ( [SO 4 2- ]) plot should be a straight line with unity slope. Such plots are shown in Fig. 4. It is, therefore, concluded that F ( [SO 4 2- ]) -1 ; i.e. rate of extraction is inversely proportional to [SO 4 2- ] in its hcr and independent of [SO 4 2- ] in its lcr. The Arrhenius i.e. log F vs. 1/T plots are shown in Fig. 5. In all cases, straight lines are obtained. E a values are measured to be within 37 to 60 kj/mol. Treatments of temperature dependence data by the Activated Complex Theory (log (Fh/kT) vs. 1/T plot) yield straight lines (not shown). The estimated H ± value varies within 36 to 60 kj/mol; and the S ± value varies within -127 to -205 J/mol K depending on experimental paremeters. From the intercepts of lines in Figs. 1-4, the values of k f are calculated to be and at 293 K and 318 K, respectively, yielding E a value of 50.4 kj/mol. From the above measurements, the flux equation, at 293 K, for Ti(IV) transfer can be represented by: F = [Ti(IV)] ( [H + ]) -1 [HA] (o) (1+1.8[SO 4 2- ] -1 (6) 44

46 log (F, kmol/m 2 s) log (F, kmol/m 2 s) log (F, kmol/m 2 s) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering (), ph = 1.6, [Ti(IV)](ini) = 1000 mg/l, Temp. = 318 K; s = 0.99, I = ( ), ph = 2.50, [Ti(IV)](ini) = 1000 mg/l, Temp. = 293 K; s = 1.04, I = (), -1.2 ph = 2.50, [Ti(IV)](ini) -1.0 = mg/l, Temp. -0.6= 318 K; -0.4 s = 1.01, I = Fig. 3. Effect of [HA] (o) on flux. [SO 4 2- ] = 0.50 mol/l, Ai = m 2, o/a = 1 (o = 100 ml) log {[HA], mol/l} Eq. (12) alone is slow; or, this step along with diffusion of at least a reactant to the reaction site or a product from the reaction site are equally slow to be rate determining. High negative S ± values suggest that Eq. (12) occurs via an S N 2 mechanism i.e. the attack of 1 st anionic ligand (A - ) on hydrated TiO 2+ to form higher co-ordinated activated complex [TiO(H 2 O) n.a] + is slower than the dehydration step to form normal co-ordinated [TiO(H 2 O) n-1 A] + and also the addition of 2 nd A - to form [TiOA 2 ]. Similarly, on consideration of the prevailing Ti(IV) species at the conditions cited, Eqs. (8), (9) and (10) can be modified, respectively, to: F = ( /1.8) P o/i K a -1 [TiO 2+ ] [A - ] (13) ( ), ph = 1.80, [HA](o) = 0.40 mol/l, [Ti(IV)](ini) = 1000 mg/l ( ), ph = 2.50, [HA](o) = 0.20 mol/l, [Ti(IV)](ini) = 100 mg/l log (1+1.8x[SO 2-4 ] Fig. 4. Effect of [SO 4 2- ] on flux. Ai = m 2, o/a = 1 (o = 100 ml), Temp = 293 K. whence, [Ti(IV)] < 1.20 g/l. This equation yields the following four limiting cases: Case I: at lcr of H + and SO 4 2- F = [Ti(IV)] [HA] (o) (7) Case II: at lcr of H + but hcr of SO 4 2- F = ( /1.8) [Ti(IV)] [HA] (o) [SO 4 2- ] -1 (8) Case III: at hcr of H + but lcr of SO 4 2- F = ( /447) [Ti(IV)] [HA] (o) [H + ] -1 (9) Case IV: at hcr of H + and SO 4 2- F = ( /( )) [Ti(IV)] [HA] (o) [H + ] -1 [SO 4 2- ] -1 (10) At lcr of H + (higher ph region), Ti(IV) exists most likely as TiO(OH) + ; so that [TiO(OH) + ] = β TiO(OH) [TiO 2+ ] [OH - ]. HA partition between organic phase and interface (P o/i = [HA] (o) /[HA] (i) ) and at interface it is ionized (K a = [H + ] (i) [A - ] (i) /[HA] (i) ). The H + and A - are then partitioned between interface and aqueous phase ( = [H + ] (i) /[H + ] (a) ; = [A - ] (i) /[A - -1 ] (a). In such a case, [HA] (o) = P o/i K a [H + ] [A - ]; so that Eq. (7) takes the form: F = P o/i K a [TiO 2+ ] [A - ] (11) Equation (11) suggests that the rate determining step in this process is the following chemical reaction: TiO 2+ + A - TiOA + (12) The E a value of kj/mol at lcr of H + indicates that the process is either intermediate or chemical control. Depending on experimental condition, either ( ), ph = 2.50, [Ti(IV)](ini) = 1000 mg/l, [HA](o) = 0.06 mol/l; s = -2100, Ea = kj/mol ( ), ph = 1.60, [Ti(IV)](ini) = 1000 mg/l, [HA](o) = 0.40 mol/l; s = -3100, Ea = kj/mol (), ph = 2.50, [Ti(IV)](ini) = 1000 mg/l, [HA](o) = 0.40 mol/l; s = -1920, Ea = kj/mol ( ), ph = 2.50, [Ti(IV)](ini) = 100 mg/l, [HA](o) = 0.40 mol/l; s = -2700, Ea = kj/mol (1/T)x10 3, K -1 Fig. 5. Effect of temperature on flux (Arrhenius plots). [SO 4 2- ] = 0.50 mol/l, Ai = m 2, o/a = 1 (o = 100 ml). F = ( /447) P o/i K a [TiO 2+ ] [A - ] (14) F = ( /( )) -1 P o/i K a [TiO 2+ ] [A - ] (15) Equations (13) (15) are similar to Eq. (11); the difference is only with the constant terms. So same mechanism holds in cases II, III and IV. IV. CONCLUSION The initial rates per unit area of Ti(IV) transfer in the system: Ti(IV) SO 4 (H +, Na + ) Cyanex 301 kerosene 5% (v/v) heptan-1-ol has been measured at different experimental parameters to establish a rate equation. This equation has been analyzed to provide mechanism of extraction. Whatever be the experimental condition the addition of the first anionic ligand of the extractant (A - ) to TiO 2+ to form TiOA + is the slowest step which is supported by high activation energy. This rate determining chemical reaction step occurs via S N 2 mechanism; i.e. addition of A - to TiO(H 2 O)x 2+ in forming higher co-ordinated activated complex [TiO(H 2 O)xA] + is the slowest step. V. ACKNOWLEDGEMENT Authors are grateful to Cytec Canada Inc. for supplying Cyanex 301 as gift. 45

47 NOTATIONS AND ABBREVIATIONS HA = Cyanex 301 (bis-2,4,4- trimethylpentyl dithiophosphinic acid) = Amount of Ti(IV) transferred in 100 ml organic phase (if concentration is expressed in mg/l, then it equals to c/ ), kmol k f = Forward extraction rate constant E a = Activation energy, kj/mol H ± = Enthalpy change on activation, kj/mol S ± = Entropy change on activation, J/mol K F = Flux of Ti(IV) transfer, kmol/m 2 s A i = Interfacial area, m 2 t = Time of phase contact, s β = Stability or formation constant a, b, c, d = Reaction order with respect to (wrt) Ti(IV), H + 2-, HA and SO 4 concentrations, respectively h = Planck s constant, J s k = Boltzman constant, J/K T = Absolute temperature, K I = Intercept s = Slope (ti), (h), (ha), (so 4 2- ) = Constant concentrations of Ti(IV), H +, HA and SO 4 2-, respectively P = Partition coefficient R = Molar gas constant, J/K mol = Acid dissociation constant of HA f(r) = Function of reactants, (ti) a (h) b (ha) c (o) (so 2-4 ) d Subscript (o) = Organic phase Subscript (i) = Interface Subscript (a) = Aqueous phase wrt = With respect to lcr = Lower concentration region hcr = Higher concentration region S N 2 = Substitution neucleophilic bimolecular REFERENCES [1] M. Mozammel and A. Mohammadzadeh, The influence of pre-oxidation and leaching parameters on Iranian ilmenite concentrate leaching efficiency: Optimization and measurement, Measurement, vol. 66, pp , [2] A. A. Baba, S. Swaroopa, M. K. Ghosh and F. A. Adekola, Mineralogical characterization and leaching behavior of Nigerian ilmenite ore, Trans. Nonferrous Met. Soc. China, vol. 23, pp , [3] R. K. Biswas, M. F. Islam and M. A. Habib, Processing of ilmenite by roasting with reformed product of LPG-water mixture in presence of nickel powder and subsequent leaching, Bangladesh J. Sci. Ind. Res., vol. 33, pp , [4] R. K. Biswas, M. F. Islam and M. A. Habib, Processing of ilmenite through salt-water vapor roasting and leaching, Hydrometallurgy, vol. 42, pp , [5] F. Islam, M. Ali and S. Akhter, Separation and recovery of titanium from iron bearing leach liquors by solvent extraction with di-2-ethylhexyl phosphoric acid tributyl phosphate -thiocyanate system, Bangladesh J. Sci. Ind. Res., vol. 13, pp , [6] F. Islam, R. K. Biswas and C. M. Mustafa, Solvent extraction of Ti(IV), Fe(III) and Fe(II) from acidic sulphate medium with di-o-tolylphosphoric acid benzene-hexan-1-ol system: A separation and mechanism study, Hydrometallurgy, vol. 13, pp , [7] K. C. Sole, Recovery of titanium from the leach liquor of titaniferrous magnetites by solvent extraction. Part I, II and III, Hydrometallurgy, vol. 51, pp , [8] F. Islam and R. K. Biswas, The solvent extraction of Ti(IV), V(IV), Fe(III), Cr(III) and Mn(II) from acidic sulphate-acetate media with bis(2-ethyl hexyl) phosphoric acid in benzene: A theoretical separation study, J. Bang. Acad. Sci., vol. 5, pp , [9] R. K. Biswas and A. K. Karmakar, Solvent extraction of Ti(IV) from acidic sulphate medium by Cyanex 301 dissolved in kerosene, Sep. Sci. Technol., vol. 49, pp , [10] J. Bassette, R. C. Denny, G. H. Jeffery and J. Mendham, Vogel s Textbook of Quantitative Inorganic Analysis, 4th ed.; ELBS: London, p. 750, [11] B. K. Tait, Cobalt-nickel separation: the extraction of cobalt(ii) and nickel(ii) by Cyanex 301, Cyanex 302 and Cyanex 272. Hydrometallurgy, vol. 32, pp ,

48 Production and Improvement of Waste Tire pyrolysis Oil to be Utilized with Diesel in CI Engine Md. Nurul Islam Department of Mechanical Engineering Rajshahi University of Engineering & Technology Rajshahi, Bangladesh. Abstract The standard of living, quality of life and development of a nation depend on its per capita energy consumption. Global energy supply that mainly depends on fossil fuel is decreasing day by day. It is estimated that the energy demand will be increased five times by the year 2021 from present scenario. Due to the fossil fuel crisis, the development of alternative fuel technologies are drawn more attraction to deliver the replacement of fossil fuel. Pyrolysis is one of the promising alternative fuel technology that produces valuable oil, char and gas product from organic waste. Early investigations report that tire pyrolysis oil extracted from vacuum pyrolysis method seemed to have properties similar to diesel fuel. The main concern of this paper is to produce and improve the properties of crude tire pyrolysis oil by desulfurizing, distilling and utilize it with diesel in CI engine to analyze the efficiency for various compositions. Keywords solid tire waste; pyrolysis; crude pyrolysis oil; improvement; alternative fuel. I. INTRODUCTION Approximately 1.5 billion tires are produced each year which will eventually enter the waste stream representing a major potential waste and environmental problem [1]. In Bangladesh, total waste tire generation of each year is about tons [2]. Vehicle tires contain long chain polymer (butadiene, isoprene and styrene- butadiene) which cross-linked with sulpher thus have excessive resistant to degradation. On the other hand, as a result of burning of these tires excessive damage to human health caused by the pollutant emissions such as polyaromatic hydrocarbons (PAHs), benzene, styrene, butadiene and phenol-like substances [3]. Conversion of these waste tires to energy through pyrolysis is one of the recent technology in minimizing not only the waste disposal but also to be utilized as an alternative fuel for internal combustion engines. Pyrolysis is generally described as the thermal decomposition of the organic wastes in the absence of oxygen at mediate temperature about 450 o C [4]. The advantage of pyrolysis process is its ability to handle waste tire. It was reported that pyrolysis oil of automobile tires contain 85.54% C, 11.28% H, 1.92% O, 0.84% S and 0.42% N components [5]. Pyrolysis is also non toxic or environmental harmful emission unlike incineration [6]. Tire pyrolysis oil has been found to have a high gross calorific value of around MJ/Kg. It would encourage their use as replacement for diesel fuel if it Md. Rafsan Nahian Department of Mechanical Engineering Rajshahi University of Engineering & Technology Rajshahi, Bangladesh. samsrafsan@gmail.com is properly distilled [7]. Therefore, these waste tires should be utilized by converting to new and clean energies. II. PRODUCTION OF CRUDE TIRE PYROLYSIS OIL (TPO) At first, automobile tires are cut into a number of pieces and the bead, steel wires and fabrics are removed. The tire chips are washed, dried and fed in a mild steel fixed bed reactor unit. Heat Solid waste Heat Fig. 1. Steps of pyrolysis process The feedstock is externally heated up in the reactor in absence of oxygen. The pyrolysis reactor design for the experiment is a cylindrical chamber of inner diameter 110 mm, outer diameter 115 mm, height 300 mm and fully insulated. 2kW of power is supplied to the reactor for external heating. The temperature of the reactor is controlled by a temperature controller. The process is carried out between ( ) o C. The heating rate is maintained at 5 K/min. The residence time of the feedstock in the reactor is 120 minutes. The products of pyrolysis in the form of vapour are sent to a water cooled condenser and the condensed liquid is collected as fuel. Three products are obtained in the pyrolysis namely; Tire Pyrolysis Oil (TPO), Pyro gas and Char. 1.9 kg of feedstock is used to produce 1 kg of Tire Prolysis Oil. The heat energy required for pyrolysis process per kg of TPO produced is around 6 MJ/kg [8]. The percentages of pyrolysis products are given below in TABLE I. Pyrolysis products Percentage (%) TABLE I. PERCENTAGE OF PYROLYSIS PRODUCTS Tire Pyrolysis Oil Moisture Thermodynamic decomposition Gases (CO2+CO+CH4)+ Volatile Condenser Liquid Char Non condensable gases Char Pyro gas Moisture

49 supplied to carry out producer gas from the reactor to the condenser and also create inert environment to the reactor. 80% of TPO is distilled in the distillation whereas 5% of TPO is left out as pyro gas and 15% is found as sludge. Fig. 2. Crude tire pyrolysis oil from waste tire III. IMPROVEMENT OF CRUDE TIRE PYROLYSIS OIL (TPO) The improvement of crude TPO involves three stages, A. Removal of moisture. B. Desulphurization. C. Distillation. A. Removal of moisture Initially crude TPO is heated up to 100 o C, in a cylindrical vessel for a particular period to remove the moisture, before subjecting it to any further chemical treatment. B. Desulphurization The moisture free crude TPO contains impurities, carbon particles and sulpher particles. A known volume of concentric hydrosulphuric acid (8%) is mixed with the crude TPO and stirred well. The mixture is kept for about 40 hours. After 40 hours, the mixture is found to be in two layers. The top layer is a thin mixture and the bottom layer is thick sludge. The top layer is taken for atmospheric distillation and the sludge is removed and disposed off. C. Distillation Distillation is a commonly used method for purifying liquids and separating mixtures of liquids into their individual components. The distillation process is shown in Fig. 3. Heating Evaporating Condensing Fig. 3. Sequence of distillation process Atmospheric distillation process is carried out to separate the lighter and heavier fraction of hydrocarbon oil. A known sample of chemically treated crude TPO is taken for vacuum distillation process. The sample is externally heated in a closed chamber by electric heater of 1.5 KW. The vapour leaving the chamber is condensed in a water condenser and the distilled tire pyrolysis oil (DTPO) is collected separately. Non condensable volatile vapours are left to the atmosphere. The distillation is carried out between ( ) o C. Nitrogen gas is Fig. 4. Experimental setup of the distillation plant The DTPO has irritating odor like acid smell. The odor can be reduced with the help of adding some masking agents or odor removal agents. Fig. 5 shows the physical view of distilled tire pyrolysis oil (DTPO). Fig. 5. Distilled tire pyrolysis oil The properties of tire pyrolysis oil (TPO), distilled tire pyrolysis oil (DTPO) and diesel fuel are shown in TABLE II. TABLE II. Properties Density at 15 o C, kg/l Kinematic Viscosity at 40 o C, cst Pour point, o C Flash point, o C Gross Calorific Value, MJ/kg PROPERTIES OF TPO, DTPO AND DIESEL FUEL Tire pyrolysis oil Distilled tire pyrolysis oil Diesel fuel Below to Below Above

50 IV. PERFORMANCE TEST OF DISTILLED TIRE PYROLYSIS OIL (DTPO) Engine performance indicates the effect of a fuel in the engine. It shows the trend and possibility of using distilled tire pyrolysis oil to replace diesel fuel without any engine modifications [9]. It is necessary to determine engine brake power, brake specific fuel consumption and brake thermal efficiency. The performance parameters can be calculated by following equations [10]. A. Engine Brake power Engine brake power (P) is delivered by engine and absorbed load. It is the product of torque and angular engine speed where P is engine brake power in KW; N is angular speed of the engine in RPM as: P= B. Brake specific fuel consumption Brake specific fuel consumption (BSFC) is the comparison of engine to show the efficiency of the engine against with fuel consumption of the engine in kg/kwhr where (m f ) is the fuel consumption rate in kg/hr as: BSFC= C. Brake thermal efficiency The percentage of brake thermal efficiency of the engine is related to engine brake power and total energy input to the engine. The quality of the blended DTPO with diesel fuel is tested in Beco diesel engine. The engine is kept fixed at 27% part load. The specifications of the engine are shown in TABLE III. TABLE III. Engine type Cooling Speed Rated Power ENGINE SPECIFICATIONS 4 Stroke CI engine Water cooling 1500 RPM 7.5 HP DTPO has about 7% higher heating value than crude TPO. This is due to the elimination of the impurities, moisture, carbon particles, sulpher and sediments. Four test fuels have been taken for the performance test. These are 100% diesel fuel, 75% diesel with 25% distilled pyrolysis oil (DTPO 25), 50% diesel fuel with 50% distilled pyrolysis oil (DTPO 50) and 25% diesel with 75% distilled pyrolysis oil (DTPO 75). TABLE IV. Fuel 100% Diesel 75% Diesel+ 25% DTPO 50% Diesel+ 50% DTPO 25% Diesel+ 75% DTPO PERFORMANCE RESULTS FOR DIFFERENT BLENDED FUELS Brake Power (KW) Efficiency (%) Brake specific fuel consumption, kg/kwhr The graphical representation of Performance of the engine with neat diesel and DTPO blends are described below in Fig. 6 and 7. Efficiency (%) Efficiency Vs Brake Power Brake Power (KW) 100% Diesel % Diesel % DTPO % 9.3 Diesel % DTPO Fig. 6. Variation of efficiency with respect to brake power at 27% part load Fig. 6 shows the comparison of the brake thermal efficiency with brake power for the tested fuels at 27% part load. It is observed from the figure that at 0.45 KW, the thermal efficiency is 9.5% for diesel fuel (DF) whereas for blending of 25% DTPO, 50% DTPO and 75% DTPO with diesel are 9.498%, 9.398% and 9.304% respectively. The thermal efficiencies of DTPO-DF blends are lower compared to diesel fuel. Reduction in thermal efficiencies by about %, 1.07% and 2.06% for blending of 25% DTPO, 50% DTPO and 75% DTPO with diesel compared to diesel fuel. 49

51 BSFC (kg/kwhr) Brake specific fuel consumption (BSFC) Vs Brake Power % Diesel % Diesel % DTPO % Diesel+ 75% DTPO Brake Power (KW) Fig. 7. Variation of BSFC with respect to brake power at 27% part load Fig. 7 shows the comparison of the break specific fuel consumption (BSFC) with brake power for the tested fuels at 27% part load. It is observed from the figure that BSFC increases with increase in the concentration of DTPO in DTPO-DF blend. At 0.45 KW, the BSFC is for diesel fuel whereas for blending of 25% DTPO, 50% DTPO and 75% DTPO with diesel are 0.852, and respectively. The BSFC of DTPO-DF blends are higher compared to diesel fuel. Increase in BSFC by about 0.046%, 1.17% and 1.64% for blending of 25% DTPO, 50% DTPO and 75% DTPO with diesel compared to diesel fuel. V. CONCLUSION 75% Diesel+ 25% DTPO In the presented study, it is found that the distilled tire pyrolysis oil is similar to diesel fuel and able to replace diesel fuel in small engine. Blends of DTPO 25 gives better results than DTPO 50 and DTPO 75. Following are the conclusions based on the experimental results obtained while operating single cylinder diesel engine with DTPO blends. I. DTPO 25 blends can be directly utilized in diesel engine without any engine modification. II. The brake thermal efficiency of DTPO 25 is slidely lower than diesel fuel. But for DTPO 50 and DTPO 75 it is much lower with compared to diesel fuel. III. Brake specific fuel consumption of DTPO 25 blend is very close to the specific fuel consumption of diesel. But for DTPO 50 and DTPO 75 it is slidely higher. So it is advisable not to use DTPO 50 and DTPO 75 in CI engines. REFERENCES [1] P.T Williams, Pyrolysis of waste tire: a review Aug;33(8): doi: /j.wasman [2] Bangladesh Bureau of statistics, Government of Peoples Republic of Bangladesh, Statistical Year book of Bangladesh 2008, 24 th editon. [3] Reisman JI. Air emissions from scrap tire combustion, EPA- 600/R ; 1997 [4] P.T Williams, S Besler, Environmental Science and Technology Fuel 74(9) , 1995 [5] Mastral AM, Murillo R, Callen MS, Garcia T. Optimisation of scrap automotive tires recycling into valuable liquid fuels. Resour conserv Recycl 2000;29: [6] J. Scheirs and W. Kaminsky, Feedstock Recycling and Pyrolysis of Waste Pastic: Converting Waste Plastics into Diesel and Other Fuels, John Wiley & Sons Ltd, Chichester, doi: / [7] M.R. Islam, H. HANIU, R.A. Beg, Liquid fuels and chemicals for pyrolysis of motorcycle tire waste: product yield, compositions and related properties Elsevier, Fuel, Vol: 87, pp , 2008 [8] Isabel de Marco Rodriguez, et al Pyrolysis of Scrap Tires. Fuel processing technology; 72:9-22, 2001 [9] C. Wongkhorsub, N. Chindaprasert, A comparison of the use of pyrolysis oils in diesel engine. Enrgy and power engineering, 2013,5, pp [10] O. Arpa, R. Yumrutas and Z. Argumhan, Experimental investigation of the effect of diesel-like fuel obtained from waste lubrication oil on engine performance and exhaust emissions. Fuel process technology, Vol. 91, 2010, pp

52 Biological Evaluation of Radiotherapy Treatment Plan for Different Field Techniques in 3-Dimensional Conformal Radiotherapy (3DCRT) Kausar A*, Azhari H A, Chaudhury S, Bhuiyan M A, Zakaria G A Department of Medical Physics and Biomedical Engineering. Gono University Dhaka, Bangladesh * abukausar79@yahoo.com Abstract The purpose of this study is to evaluate the 3D radiation treatment plan variants considering biological parameters in external beam radiotherapy. The biological parameters was calculated for 4- fields, 6-fields, 7-fields and 9-fields beams for cervical and prostate carcinoma. In biological parameters, the tumor control probability (TCP) and normal tissue complication probability (NTCP) calculation done by the Poisson statistics model and Lyman-Kutcher- Burman model respectively. In cervical carcinoma, the TCP of four different fields satisfied the biological criteria (TCP 0.5). However, only the NTCP of bladder for all fields comply with the protocol (NTCP 0.05) where the rectum and the right & left femour go beyond the tolerance limit. Similarly, in prostate carcinoma, the TCP provides the agreeable results for all fields. On the other hand, NTCP of bladder for 4-fields and 7-fields, the rectum and the right & left femour exceeds the acceptable limit where only the 6-fields and 9-fields of rectum assure the protocol. It was observed that the no treatment plan complies the biological criteria in both cases, thus it is important to evaluate the treatment plan biologically to achieve the better outcomes. Hence the biological evaluations of treatment plan need to be practiced by the radiation oncologist as well as the medical physicist. Moreover, TPS should integrate the biological evaluation tools and should be used appropriate parameters for biologically selecting the best treatment plan. Keywords: 3DCRT, TCP, NTCP, Treatment plan. I. INTRODUCTION The goal of radiation therapy (RT) is to deliver a therapeutic dose of radiation to target tissues while minimizing the risks of normal tissue complications. Until recently, the quality of a RT plan has been judged by physical quantities, i.e., dose and dosevolume (DV) parameters, thought to correlate with biological response rather than by estimates of the biological outcome itself [1]. There are many tools used to evaluate the treatment plan biologically. Nowadays, the evaluation of treatment plans is usually done by analysis of dose volume histogram (DVH) as well as two-dimensional and threedimensional spatial dose distributions [2]. In the early days of radiation oncology, the biological consequences of treatment were judged mainly by the dose absorbed in the tumor and surrounding normal tissues, with experience driven accounting for overall treatment time and fractionation. To correct the later two factors nominal standard dose (NSD), cumulative radiation effect (CRE), and time dose fractionation (TDF) formalisms were developed [3] [4] [5]. Different TCP model can be used for evaluating the treatment plan radio-biologically such as; Poisson statistics model, Zaider-Minerbo TCP model etc. The roots of normal tissue complication probability (NTCP) modeling lie in attempts to quantify the dependency of the tolerance dose for a certain radiation effect on the size of the treated region [6]. NTCP modeling gained more attention with the advent of three-dimensional conformal radiation therapy (3DCRT). Different NTCP models may be used for calculation such as, the Lyman-Kutcher- Burman (LKB) model, Relative Seriality model (Kallman K-and-S model), sigmoidal dose response NTCP model, and Critical volume NTCP model. In this study, the Poisson statistics model has been used for TCP and the LKB model used for NTCP calculation. The biological evaluation has been done on two cancer cases i.e. Cervical and Prostate carcinoma. II. MATERIALS AND METHODS For evaluating a treatment plan biologically, two common cancer cases were selected. A case of cervical carcinoma having 2b stage; size of the tumor (T2), the number of node involvement (N1) and the metastasis (M0) had been taken. This patient undergone radiotherapy treatment as follows; EBRT dose 50Gy (Cervix with Nodal pelvis) is delivered in 25 fractions in 3DCRT. Further, the brachytherapy: 21Gy is delivered in 3 fractions. A case of prostate carcinoma having the 2b stage (adenocarcinoma), size of the tumor T3a, number of nodes involved (N0) and the metastasis (M0) had been taken. This patient undergone radiotherapy treatment as follows; total EBRT radiation dose delivered by 76Gy (prostate with nodal pelvis) and the nodal pelvis irradiated by 50 Gy by 25 fractions in 51

53 25 days and other 26 Gy is delivered only as the prostate boost in 3DCRT technique. Table 1. Tumor data for TCP calculation [7] [8] [9]. Organ D 50 [Gy] 50 α/β[gy] Prostate-T Prostate, T0 - T Prostate, T Prostate, T Prostate, T0 and T Cervix For biological evaluation, TCP was calculated by TCP based Poisson statistics model and the NTCP calculation was done by the Lyman-Kutcher-Burman (LKB) model. TCP models generally rely on the assumption that tumor control requires the killing of all tumor clonogens. Poisson statistics predict that the probability of this occurring is:... (1) Where, N=initial number of clonogens P s (D) =cell survival fraction after a dose D. If it is assumed that cell survival can be described by single-hit mechanics,... (2) The expression in Eq. (11) can be rewritten in terms of the two parameters describing the dose and normalized slope at the point of 50% probability of control, D 50 and γ 50 and:... (3) Using the assumption of independent subvolumes, for the case of heterogeneous irradiation, the overall probability of tumor control is the product of the probabilities of killing all clonogens in each tumor subvolume described by the DDVH: Thus, for a given DDVH {Di,vi}, the TCP can be calculated using the following two-parameter TCP formula: The above formula originates from an attempt to predict the TCP for an individual patient from a mechanistic perspective [2]. The most widely used phenomenological approach is the LKB (Lyman-Kutcher-Burman) model to account the probability of risk in normal tissues induced by radiation. A mathematically equivalent but more conceptually transparent formulation of the LKB model was first proposed by Mohan et al. (1992) [1]. According to this model, NTCP is calculated using the following equations: Where, and ( ) D eff = is the dose that, if given uniformly to the entire volume, will lead to the same NTCP as the actual non-uniform dose distribution which is conceptually identical to the geud. i = fractional organ volume receiving a dose Di TD 50 risk. m n = dose where NTCP i = 50% for the organ of = measure of the slope of the sigmoid-curve. = the volume effect parameter. NTCP is in the range of: ( ) Table 2. Normal tissue end points and tolerance parameters [7]. Organ Bladder Rectum Femoral Head and Neck Fit Parameters V ref n m TD 50 Whole organ Whole organ Whole organ End Point Symptomatic bladder contracture and volume loss Severe proctitis/necrosis /stenosis/ festula Necrosis...(4) 52

54 III. RESULTS AND DISCUSSIONS A. Comparison of Biological parameters for four different field Techniques in cervical carcinoma. In TCP, the calculated result for four different fields shows the agreeable values (TCP 0.5) [10]. Hence, all plan techniques TCP results were satisfactory. Table 3 Comparison of Biological parameters for four different field techniques in cervical carcinoma. 4-field 4-field DVH Parameters 4f 6 f 7f 9f Biological TCP NTCP Bladder NTCP Rectum NTCP Lt. femour NTCP Rt. femour *f indicate field. 6-field 6-field DVH In (NTCP), the calculated results for the bladder of different fields satisfied the tolerance limit (NTCP 0.05) [10]. Moreover, NTCP of the rectum and right & left femour for all fields are beams do not accept the NTCP tolerance limit. Hence, according to the biological estimation among the four different fields, only the TCP provides the pleasurable value for all fields but the NTCP is very high for all risk organs excluding the bladder. B. Comparison of Biological parameters for four different field techniques in Prostate carcinoma. In TCP, the calculated results for four different fields provided the agreeable values (TCP 0.5) [10]. Hence, the all plan techniques TCP provided very good results. Parameters 4f 6f 7f 9f Biological TCP NTCP Bladder NTCP Rectum NTCP Lt.femour NTCP Rt.femour Table 4. Comparison of Biological parameters for four different field techniques in prostate carcinoma. In NTCP, only the bladder of 6-field and 9-field are satisfied the criteria (NTCP 0.05) [39]. Moreover, NTCP of rectum and right & left femour for all fields were very much higher than the tolerance limits (NTCP 0.05). Hence, according to the biological estimation among the four different fields, only the TCP fulfill the biological criteria for all fields but the NTCP is very high for all risk organs excluding bladder. 7-field 9-field 7-field DVH 9-field DVH Fig 1. An Example of Field arrangements and DVH of four different fields in Prostate Carcinoma. Although, biological criteria includes TCP and NTCP parameters but in cervical carcinoma, no treatment fields provide the acceptable value according to the biological evaluation. Finally, no plans were not accepted according to the biological evaluation of the treatment plan and organs at risk (OAR) doses are very much higher than the tolerance limit. Therefore, the physicist should evaluate the treatment plan very carefully. If the treatment plan can be done in IMRT technique or through optimization between target dose and OARs than NTCP may be minimized. So that, OAR have to be contoured accurately. IV. CONCLUSION In this study, there was a significant difference biological treatment plan evaluation, therefore, the 53

55 biological evaluation of the treatment plan need to be introduced in TPS parallel to physical treatment planning. This study shown how to make out the physical evaluation and make more accessible current radiobiological modelling knowledge, and may serve as a useful aid in the prospective and retrospective analysis of radiotherapy treatment plans. Thus, it is important to evaluate the treatment plan considering biological criteria then the treatment plan may achieve better outcomes. Therefore, the medical doctors and medical physicists should be aware to integrate biological information into treatment plans for better selection in next generation of Radiotherapy. [9] P. Okunieff, D. Morgan, A. Niemierko, and H.D. Suit, Radiation dose-response of human tumors, Int J RadiatOncolBiolPhys, vol. 32(4), pp [10] E. B. Podgorsak, Basic Radiobiology, Review of Radiation Oncology Physics: A Handbook for Teachers and Students, International Atomic Energy Agency: Vienna, Austria, pp: 408, May ACKNOWDGMENT I would like to show my greatest appreciation to Professor Dr. Golam Abu Zakaria, Department of Medical Physics and Biomedical Engineering, Gono Bishwabidyalay, Dhaka, Bangladesh and Professor and Chief Medical Physicist, Department of Medical Radiation Physics, Academic Teaching Hospital of the University of Cologne, Gummersbach, Germany. I would like to gratefully acknowledge the enthusiastic supervision of Dr. Hasin Anupama Azhari, Department of Medical Physics and Biomedical Engineering, Gono Bishwabidyalay I would like to express my gratitude towards my family, friends and other colleagues for their understanding, endless patience and encouragement REFERENCES [1] L.Allen, A. Markus, D. Joseph, J. Kyung Wook, M. Mary, M. Charles. "The use and QA of biologically related models for treatment planning: Short report of the TG-166 of the therapy physics committee of the AAPM," vol. 39 (3), pp. 1387, [2] S. Arun, Oinam, L. Singh, "Dose volume histogram analysis and comparison of different radiobiological models using inhouse developed software," Journal of Medical Physics, vol. 36(4), pp , [3] M. Strandqvist, Studien uber die Kumulative Wirkung der Rontgenstrahlen bei Fraktionierung, Acta Radiol Suppl, vol. 55, pp.1 300, [German]. [4] J.Kirk, W. Gray, and E. R. Watson, Cumulative radiation effect. I. Fractionated treatment regimes, Clin Radiol, vol. 22(2), pp , [5] F Ellis. Dose, time and fractionation: A clinical hypothesis. Clin Radiol, vol.20 (1), pp.1 7, [6] T.E. Schultheiss, C. G. Orton, and R. A. Peck. Models in radiotherapy: Volume effects, Med Phys, vol. 10(4), pp , [7] C. Burman, G.J. Kutcher, B. Emami, and M. Goitein, Fitting of normal tissue tolerance data to an analytic function, Int. J RadiatOncolBiolPhys, vol. 21 (1), pp , [8] P. Källman, B.K. Lind, and A. Brahme An algorithm for maximizing the probability of complication free tumor control in radiation therapy, Phys Med Biol., vol. 37 (4), pp ,

56 Design of a Linearly Polarized Multi-band Transmission Line Feed Microstrip Patch Antenna for Wireless Communications Sheikh Dobir Hossain 1* Department of Physics Jessore University of Science and Technology Jessore-7408, Bangladesh *dobir.aece@gmail.com Md. Khalid Hossain 2 Institute of Electronics Atomic Energy Research Establishment Savar, Dhaka, Bangladesh khalid.baec@yahoo.com Rebeka Sultana 3 Department of Computer Science & Engineering University of Rajshahi Rajshahi 6205, Bangladesh rebeka19sultana@gmail.com Abstract A linear polarized transmission line feeding dual band rectangular micro-strip patch antenna is designed for the application in wireless communication. The microstrip antenna contains a rectangular patch on the upper layer of the dielectric material with dielectric constant of 2.4 and there is a ground plane below the dielectric material. Here the introduced of cavity model with transmission line feed has the favor of low profile, high gain and wide bandwidth of the antenna. The antenna has overall size of 46.9 mm by mm and gives bandwidth of about 90 MHz at resonance frequency of 2.45 GHz and that of 115 MHz at 4.1 GHz frequency with Defected Ground Structure (DGS) which are found to be favorable for wireless communications. introduced two I slot on the patch with length and width are 14.2 mm and 1.4 mm respectively to obatain the dual band antenna operating at the resonance frequencies of 2.45 GHz and 4.1 GHz respectively. Keywords- Micro-strip Antenna; Smith Chart; Cavity Model; Dual Band Antenna; Transmission Line Feed. I. INTRODUCTION The electronic circuit miniaturization increases the importance of wireless communication systems. In commercial and government communication systems, it is required to develop. This technological trend has focused on the development of micro-strip antennas (MSA) with low cost, minimal weight and low profile antennas that are capable of maintaining high performance over a large spectrum of frequencies. The disadvantageous features of MSA such lower value of efficiency, higher value of Q factor, poor polarization purity and spurious feed radiation have diminished their versatility. However it is possible to improve the bandwidth (as long as 90 percent) and efficiency (up to about 35 percent) of a microstrip antenna by increasing the thickness of the dielectric materials with cavity model [1]-[3]. II. EXPERIMENTAL MODEL The proposed microstrip antenna is obtained using copper (annealed) rectangular patch of length mm, width 46.9 mm and thickness 0.1 mm on the upper layer of the substrate whose thickness is 2.40 mm. Also there is a ground plane on the lower side of substrate with the thickness of 0.01 mm. Here we have Fig.1. Geometry of dual band MSA The very popular and practical approximate for different parameters of rectangular microstrip patch antenna are [7]-[9] the width that s leads to good radiation efficiencies is- Here, (1) =Velocity of light= Resonant frequency =Dielectric constant of the substrate The value of effective dielectric constant is (2) 55

57 Where, h is the thickness of the substrate and must be in mm unit the normalized extension of the length is ( ) ( ) (3) The effective length of the patch is (4) Where actual length is (5) III. RESULTS The studied parameters are return loss (RL), voltage standing wave ratio (VSWR), smith chart, directivity and gain. The RL indicates how amount of power is lost in the load and does not return as a reflection. Our proposed antenna shows the return losses of db and db at the resonance frequencies of 2.45 GHz and 4.1 GHz, respectively. As it is lower than the acceptable value of return loss i.e. -10 db so the designed antenna is perfectly matched and the power loss is minimum.the parameter VSWR determines how well the antenna is matched. Our designed antenna shows the VSWR are of at 2.45 GHz and that of at 4.1 GHz frequency which is below 2 (desired value for good antenna). Hence our designed antenna is perfectly matched with minimum loss. The study of smith chart is very importan during the design of a MSA. Using smith chart it is possible to obtain proper impedance matching between antenna and transmission line feeding. The impedance of our designed dual band antenna is 49 ohms which is approximately equal to the desired value of 50 ohms, indicating the minimum power loss. The directivities of our designed antenna are 7.2 dbi in positive Z direction with angular width of 75.7 deg and 6.3dBi at an angle 49.0deg from positive Z direction with angular width of 68.5 deg at the resonance frequencies of 2.45 GHz and 4.1 GHz respectively and the gains are db and db at that frequencies, respectively which agrees well with the previous results [3]-[6]. Fig.2. Simulated return losses of dual band MSA at 2.45 GHz and 4.1 GHz (a) (b) Fig.3. The VSWR of dual band MSA at (a) 2.45 GHz; (b) 4.1 GHz 56

58 (b) Fig.4. Smith Chart of dual band MSA at 2.45 GHz and 4.1 GHz Fig.6. Gain of dual band MSA at (a) 2.45GHz; (b) 4.1GHZ IV. SUMMARY OF THE RESULTS The results of the dual band rectangular microstrip antenna operating at 2.45 GHz and 4.1GHz are summarized in the following table I. (a) TABLE I. OUTPUT PARAMETERS OF THE DUAL BAND MSA RESONATING AT 2.45GHZ AND 4.1 GHZ FREQUENCIES Resonating frequency fr (GHz) Return loss (db) Bandwidth (MHz) VSWR Directivity (db) Gain (db) (b) Fig.5. Directivity of dual band MSA at (a) 2.45GHz; (b) 4.1GHZ V. CONCLUSIONS In this paper, firstly we have looked on the design and simulation of single band antenna and then extend it to dual band antenna. The various parameters like return loss, VSWR, smith chart, directivity, gain, bandwidth and operating frequency are studied and also the effects of physical parametric on the performance of the designed antenna are studied. The designed antenna shows good impedance matching of approximately 49 ohm s also it provides good gain and efficiency at the resonant frequencies of 2.45 GHz and 4.41 GHZ which indicate that the designed antenna can be used for various applications like RADAR, Bluetooth, Biomedical instruments etc. ACKNOWLEDGMENT We would like to thank all concerned with the Department of Applied Physics, Electronics & Communication Engineering, Islamic University, Kushtia 7003, Bangladesh for their all-out effort to support us for completing this research. (a) REFERENCES [1] Shagun Maheshwari, Priyanka Jain and Archana Agarwal, CPW-fed Wideband Antenna with U-shaped Ground Plane, I.J. Wireless and Microwave Technologies (IJWMT). Volume 5, November

59 [2] Ali A. Saleh and Abdulkareem S. Abdullah, A Novel Design of Patch Antenna Loaded with Complementary Split-Ring Resonator and L- Shape Slot for (WiMAX/WLAN) Applications, I.J. Wireless and Microwave Technologies (IJWMT), Volume 3, October [3] C.A. Balanis, "Antenna Theory: Analysis and Design," Third Edition, ISBN X, Copyright 2005 John Wiley & Sons, Inc. [4] D.M. Pozar, Microstrip Antennas, Proc. IEEE, Vol. 80, No. 1, pp , January [5] C.M. Krowne, Cylindrical-Rectangular Microstrip Antenna, IEEE Trans. Antennas Propagat., Vol. AP-31, No. 1, pp , January [6] I. Lier and K. R. Jakobsen, Rectangular Microstrip Patch Antennas with Infinite and Finite Ground-Plane Dimensions, IEEE Trans. Antennas Propagat., Vol. AP-31, No. 6, pp , November [7] Y.X. Guo, K.M. Luk and K.F. Lee, U-slot circular patch antennas with L-probe feeding, IEE Electronics Letters, Vol.35, No.20, pp , [8] Carver, Keith R. and James Mink, Microstrip antenna technology, Antennas and Propagation, IEEE Transactions, pp 2-24, Feb [9] C.A. Balanis, "Advanced Engineering Electromagnetics," JohnWiley & Sons, New York,

60 Design and Fabrication of an Unmanned Video Transmitting Tele-bot using 3G GSM Network Md. Mamunoor slam Department of Electrical and Electronic Engineering Chittagong University of Engineering & Technology Chittagong-4349, Bangladesh Mehdi Hasan Chowdhury Department of Electrical and Electronic Engineering Chittagong University of Engineering & Technology Chittagong-4349, Bangladesh Abstract Appropriate robotic systems can play a vital role where human has limitations to work. Several systems can be used to control a robot from distance (such as Bluetooth, RF communication, Zigbee etc.) which may have the limitation of working range. At present, the use of GSM network has increased significantly as it has a massive coverage area throughout the globe. This paper demonstrates a technical method of construction of an unmanned robotic unit using 3G GSM network which can be controlled by a cell phone from any part of the world along with the provision of wireless video transmission. An experimental study was conducted with suitable conditions to test the feasibility and effectiveness of the implemented system. The outcome of experimental result is thoroughly examined in this paper. Keywords Unmanned Robot; Wireless Video Transmission; GSM; 3G Technology; Edge Avoiding Technology I. INTRODUCTION Robotic unit is an artificial representative which is usually an electro-mechanical machine guided by computer, cell phone or electronic programming to perform tasks on its own [1]. There has been an immense advancement of robotics in this modern engineering world as researchers all over the world have been working enormously in the field of robotics. As interest in robotics continues to rise every day, robots have been increasingly assimilated in our practical life [2]. To accomplish work in a remote place, a dangerous environment, and macro-space, an unmanned robot will be a dynamic replacement of human. However, to operate the robot that will exist in such an environment remotely, the high-speed communication protocol and wire range are necessary and indispensable [3]. It has been more than twenty years or so that scientists, engineers are working on long distance controller systems [4]. At present the cell phone is an essential part in everyday life and any application based on this device has a wide acceptance due to the wide coverage of mobile networks (GSM, GPRS, 3G, 4G, and UMTS etc.) [5]. Nowadays, it is important to control and acquire information from anyplace where the GSM communication network step in with its large coverage area. An unmanned robot based on GSM network generally uses the signals generated by the DTMF (Dual-Tone Multi-Frequency) systems through a simple telephone call. The DTMF systems, which have been invented in the 50s, were developed by Bell Labs and they have been applied since then for the communication between telephonic equipment [5]. Nowadays, DTMF has become a medium for the implementation of guiding robots from distance. Generally robots use RF, Bluetooth circuits for the feature of wireless control, which have the imperfections of confined working region and inadequate control. Use of a smart phone for monitoring can overcome these abridgements as it provides the advantages of enormous working range and interference free controlling scheme [6]. So GSM network has been used in various applications for its enhanced performance. In 2013, Md. N. Chowdhury (et al.) has developed a concept of GSM network for robots using DTMF technology [7]. A cell phone based vehicle remote control system has also been developed by B. B. Pathik (et al.) in 2014 [8]. Another practical approach of microcontroller based smart phone operated robot for emergency rescue system has been proposed using DTMF technology by A. S. M. Z. Shifat (et al.) too [6]. So this paper presents the design and fabrication of an unmanned robot which will be controlled by 3G GSM network. Different keys in the dial pad of a smart phone will be utilized which produce DTMF tones. The 3G technology of cellular communication has been added with the robotic system for wireless video transmission using the video call technology. Moreover, Edge Avoiding Technology (EAT) has also been equipped within the system for avoiding sharp edges which can damage the system. II. SYSTEM ARCHITECTURE A. DTMF Technology DTMF is a universal communication term for touch tone (a Registered Trademark of AT&T). These tones are produced when pressing the different digits of dial pad of cell phone [9]. Pressing any digit generates a unique tone which is convolution of two different frequencies [10]. Generally there is always a chance that an arbitrary sound will be on the same frequency which will trip up the system. So if two tones are used to denote a digit, the likelihood of a false signal occurring is ruled out. This is the reason to use dual tone in DTMF communication [9]. Each of the tone is composed of two sine waves of the low and high frequencies superimposed on each other. These two frequencies explicitly represent one of the digits on the cell keypad. Thus generated signal can be expressed mathematically as follows: ( ) 59

61 where A H, A L are the amplitudes and f H, f L are the frequencies of high & low frequency range [10]. TABLE I. FREQUENCIES OF EACH DTMF SIGNAL Frequency (Hz) A B C 941 * 0 # D detect and decode all 16 DTMF tone-pairs into a 4-bit code. The output bits for the different keys are given below in Table II. The circuit diagram of DTMF decoder has also been shown in Fig. 2. Key TABLE II. Low Frequency (Hz) High Frequency (Hz) OUTPUT FOR DIFFERENT KEYS Q 1 Q 2 Q 3 Q * # Fig.1. DTMF signal for key "1" [11]. Fig. 1 displays the DTMF signal for touchtone of key "1". The upper subplot illustrates the two underlying frequencies and the bottom subplot displays the signal attained by averaging the sine waves with those frequencies [11]. B. 3G Technology The 3G technology, known as third generation of cellular communications technology is based on a set of standards used for cellular devices. Cellular communications use services and networks that comply with the International Mobile Telecommunications-2000 (IMT-2000) specifications by the International Telecommunication Union [12]. The 3G technology comes with enhancements over earlier wireless technologies, like high-speed transmission, advanced multimedia access, global roaming and internet access at speeds up to 2 megabits per second (2Mbps) [13]. 3G is mostly used in cell phone as a means for making voice and video calls, to download and upload data and surfing the net [2]. So in this prototype of robot, video call technology (one of the applications of 3G) has been equipped to control the robot remotely and to transmit the wireless video simultaneously. C. DTMF Decoder : MT8870 To decode DTMF signal, decoder IC MT8870 has been used. The MT8870 is a complete DTMF receiver assimilating both the band split filter and digital decoder functions. The filter unit uses switched capacitor techniques for high and low group filters; the decoder uses digital counting techniques to Fig.2. D. Motor Driver: L293D The circuit diagram of decoding operation. L293D is a dual H-bridge motor driver integrated IC. Motor drivers act as current amplifiers since they take a low current signal and provide a higher current signal which drives the motors. This driver IC contains two inbuilt H-bridge driver circuits. In its common mode operation, two dc motors can be driven simultaneously, both in forward and reverse direction. Some specified data of this motor driver IC has been shown below in Table III. E. Edge Avoiding Technology (EAT) EAT has been a dynamic solution equipped in this robotic system for subsiding the damage of the system. At the time of operation, if any sharp edge in the ground comes in front of the robot, it will stop instantly in that position. Then the user will be notified so that proper measures can be taken. An ultrasound motion sensor (GS-311) has been used for the operation of EAT. The sensing range of this sensor is 2mm-3m which is handy for operating the robot in rugged areas. 60

62 III. METHODS OF WORKING The proposed system is based on DTMF technology as previously described. The 3G technology has been added for the wireless control and video transmission at the same time. Two 3G enabled phone is required for the implementated system. One of them should be eqquipped with the robot and the other will act as a controller unit. The controller unit will be operated by the user. At first, the user have to give a video call to the cell phone attached with the robotic unit. After receiving Fig.4. The motor driver controlling circuit. Open Dial pad Of Cell phone Video Call To the Robot Video Call Receive Fig.3. Block diagram of GSM controlled robot. the call, if users presses any key of dial pad of controller unit then corresponding generated DTMF signal will be transmitted via GSM network. This signal will be received by the cell phone of the robotic unit. This signal will be relayed to the decoder circuit via universal 3.5mm audio jack connected to that cell phone. The DTMF decoder circuit filters the signal and decodes it into a 4-bit output binary code (Q1, Q2, Q3 and Q4) shown in Table II. These decoded bits are sent to an 8-bit microcontroller which is preprogrammed to take up decisions for the corresponding input (Pressed key) and sends the decision to motor driver IC (L293D). Finally the motor driver drives the motors according to the commands sent by the users. This whole procedure has been presented in a block diagram at Fig. 4. In is worthwhile to mention that, this process is enable to run simultaneously during video call. So the wireless video will be transmitted by the robot to the users with provision of controlling the robot from any distance in the world as its coverage area is the wide-spread 3G GSM network. The circuit diagram of motor driver has been given in Fig. 5 below. I. IMPLEMENTATION AND ANALYSIS The proposed GSM controlled robotic system has been implemented practically to observe the performance of the system. Several keys have been used for this implementation which is given below in Table III. No Pressed Key Video Displayed on Cell phone Fig.5. Robot Turned ON? Yes Apply Instructions Robot Operated based on Instructions TABLE III. End Video Call Binary DTMF Code Yes Flow Chart of Overall System. IMPLEMENTED COMMANDS Command Move Forward Turn Left Turn Right Move Backward Stop * Turn Camera to Left Turn Camera in Front # Turn Camera to Right The commands (shown in Table III) have been applied to practically control the robot in different No End Video Call 61

63 direction. The performance of the robot is very decent as it requires a small time to respond accordingly. An observation of wireless video transmission has also been arranged to study the performance of the system s capability to transmit video to the users which is illustrated in Fig. 6. Fig.6. Name of the system Implementation of wireless video transmission. TABLE IV. Effective range COMPARATIVE STUDY Additional features Bluetooth based robot 50 meters - [14] Zigbee based robot [15] 1000 meters - DTMF based robot GSM network - [5,6,7] The Proposed Robot GSM network Wireless Video Transmission, EAT From the comparative study between the proposed robotic system and existing systems (Table IV), it is apparent that existing Bluetooth and Zigbee based system have limitations in effective range. Though the DTMF based systems [5, 6, and 7] have eliminated this limitation but the proposed system based on 3G technology is capable of both wireless control with an additional feature of wireless video transmission. So this implemented system can be more preferable in practical applications. II. FUTURE WORK AND CONCLUSION The video transmission efficiency can be enhanced significantly by using 4G and LTE technology. This feature may be incorporated in future. At present the system can be used only for observing wireless video. Moveable arm, gripper, smoke and gas detecting sensor etc. can be added to use this robot in other applications. As the cellular network is wide-spread across the globe, the proposed robotic unit can be monitored from any part of the world. With the wireless video transmission and EAT technology added in this robot, the proposed system must play a significant role in practical applications where human has restraints to operate. Advanced Engineering, Volume 3, Special Issue 2, January 2013, Page [2] Ashish Jadhav, Mahesh Kumbhar, and Meenakshi Pawar, Cell Phone Controlled Ground Combat Vehicle, International Journal of Computer and Communication Engineering, Vol. 1, No. 2, July 2012, Page [3] Dong-ying Ju, Rui Zhong, Masaya Takahashi, Development of Remote Control and Monitor System for Autonomous Mobile Robot Based on Virtual Cell Phone, Proceedings of ICICIC 2007, Kumamoto, Japan. [4] Md. Shahinoor Mannan, Md. Nazmus Sakib, GSM Based Remote Device Controller using SIM548C, 5th ICCCNT , July 11-13, 2014, Hefei, China. [5] Artal J.S., Caraballo J. and Dufo R., DTMF Technology applied to the Identification and Control of a Small Mobile Robot, Tecnologias Aplicadas a la Ensenanza de la Electronica (Technologies Applied to Electronics Teaching) (TAEE), 2014 XI. [6] A. S. M. Zadid Shifat, Mohammad Shaifur Rahman, Md. Fahim-Al-Fattah, Md. Asadur Rahman, A Practical Approach to Microcontroller Based Smart Phone Operated Robotic System at Emergency Rescue Scheme, The 9th International Forum on Strategic Technology (IFOST), October 21-23, 2014, Cox s Bazar, Bangladesh. [7] M d. Nasimuzzaman Chowdhury, Md. Khaled Hossain, M2M: GSM Network for Robots using DTMF, Global Journal of Researches in Engineering Electrical and Electronics Engineering, Volume 13, Issue 10, Version 1.0, Year [8] Bishwajit Banik Pathik, A.S.M. Ashraf Ahmed, Labina Alamgir, Abu Nayeem, Development of a Cell Phone Based Vehicle Remote Control System, International Conference on Intelligent Green Building and Smart Grid, Taipei, [9] Tuljappa M Ladwa, Sanjay M Ladwa, R Sudharshan Kaarthik, Alok Ranjan Dhara, Nayan Dalei, Control of Remote Domestic System Using DTMF,ICICI-BME 2009 Bandung, Indonesia, page [10] Roshan Ghosh DTMF Based Controller for Efficiency Improvement of a PV Cell & Relay Operation Control, (IJERA), Vol. 2, Issue 3, May-Jun 2012, pp [11] Cleve Moler, Touch-Tone Telephone Dialing, (Retrieved on August 2015). [12] Clint Smith, Daniel Collins. "3G Wireless Networks", page [13] Saliyah Kahar, Riza Sulaiman, Anton Satria Prabuwono, Mohd Fahmi Mohammad Amran, Suziyanti Marjudi, Comparative Study of Data Transfer for Mobile Robot Controller via Mobile Technology, 2011 International Conference on Electrical Engineering and Informatics, July 2011, Bandung, Indonesia. [14] Ritika Pahuja, Narender Kumar, Android Mobile Phone Controlled Bluetooth Robot Using 8051 Microcontroller, International Journal of Scientific Engineering and Research (IJSER), Volume 2, Issue 7, July 2014, Page [15] B. Bharathi, B. Suchitha Samuel, Design and Construction of Rescue Robot and Pipeline Inspection Using Zigbee, International Journal of Scientific Engineering and Research (IJSER), Volume 1, Issue 1, September 2013, Page REFERENCES [1] Dhiraj Singh Patel, Dheeraj Mishra, Devendra Pandey, Ankit Sumele, Asst. Prof. Ishwar Rathod, Mobile Operated Spy Robot, International Journal of Emerging Technology and 62

64 Effect of Sintering Temperature on Nb+Nd Doped Bismuth Ferrite Sadia Tasnim Mowri 1, Quazi Delwar Hossain 1 1 Electrical and Electronic Engineering Department Chittagong University of Engineering and Technology Chittagong, Bangladesh sdtasnim03@gmail.com M A Gafur 2, Aninda Nafis Ahmed 2, Muhammad Shahriar Bashar 3, 2 Pilot Plant & Process Development Centre 3 Institute of Fuel Research & Development Bangladesh Council of Scientific and Industrial Research Dhaka, Bangladesh Abstract (Bi 2 O 3 Fe 2 O 3 ) 0.4 (Nb 2 O 5 Nd 2 O 3 ) 0.6 was developed by employing solid state ceramic method to study the effect of different sintering temperature. XRD (X-ray diffraction) analysis implies that three phases were obtained for sample which was sintered at 850 C and 925 C. SEM (Scanning Electron Microscopy) image of the samples suggested that grain size increases with the increasing sintering temperature. Dielectric property reveals that, with the increase of frequency, dielectric constant and loss tangent of the sample decreases. Metal to insulator transition region is found from DC resistivity analysis. Keywords-XRD; SEM; Dielectric Property; DC resistivity; Bismuth Ferrite I. INTRODUCTION Multiferroic materials proclaim both the magnetic and ferroelectric ordering. These materials are well known to the researcher due to its effective application to the new magneto-electric devices. It is also effective in analyzing the basic science behind the coupling mechanism between the electronic and magnetic order parameters.because of high-curie temperature at 1043K and G-type antiferromagnetic ordering temperature at 655K, single-phase multiferroics, BiFeO 3 (BFO) is most broadly studied[1]. The room temperature phase of BiFeO 3 is rohombohedral with perovskite structure and belongs to R3c space group. Multiferroics have widely been used due to their identical behavior of coupling among ferroelectricity, ferromagnetism and ferroelasticity [2].Various state memory devices and modern operational sensnor designing methodology is becoming more flexible because of the coupling ability of either the magnetic polarization or electric polarization[3]. Several types of multiferroics have been broadly explored due to technological development[4]. This material can also be used in sensing, actuation, and devices in spintronics [5,6]. Multiferroic materials are famous for their applications in multiple state memory elements, electronic field controlled ferromagnetic resonance devices and transducers [7,8]. II. EXPERIMENTAL DETAILS Samples were prepared by the standard solid state reaction method. 20% Bismuth Oxide (Bi 2 O 3 ), 20% Iron(II) Oxide(Fe 2 O 3 ), 30% Niobium Oxide (Nb2O5) and 30% Neodymium Oxide (Nd2O3) were used as raw materials. Raw materials were weighed and mixed completely in an agate mortar for 1 hour. After that samples were again mixed thoroughly using Y 2 O 3 stabilized ZrO 2 balls in ethanol. After 24 hour milling in ethanol medium the solution was dried at 100 C for 24 hour. Dried samples were again pasted with binder (Polyvinyl Alcohol, 4%). These powders were then compacted into discs with 10 mm diameter and 4mm thickness. Powders were pressed under 254 MPa pressure by using a Hot Press (P/O/WEBER, PO 40H, Germany). The final sintering was done at 850 C and 925 C for 2 hour with 1hour holding time at 600 C for the elimination of binder. For sintering, heating rate was 5 C/min and cooling rate was 3 C/min. III. CHARACTERIZATION Structural properties were investigated by X-ray Diffraction (D8 Advance, BRUKER, Germany) with CuKα (λ= 1.54 A ) radiation and 2ϴ ranges from 10 to 70. Scanning Electron Microscopy (EVO 18 Research: ZEISS, Germany) image of the samples were taken for the analysis of surface morphology. For measuring the size of grain from SEM image, a software named ImageJ was used. Dielectric properties of the samples were measured by an Impedance analyzer (Wayne Kerr 6500B, UK), for the measurement of dielectric property, both side of the polished pellet was painted by Ag paste. DC resistivity of samples was studied by an Electrometer (6517B Electrometer/High Resistance Meter, Keithley, Germany). A.XRD XRD analysis implies that three phases were obtained for samples which were sintered at 850 C and 925 C. One phase (NdFeO3) is found at both samples. Bi 1.34 Fe 0.66 Nb 1.34 O 6.35 is found for samples sintered at 850 C, whether Bi δ Fe Nb O 7 is found for samples sintered at 925 C (Fig-1) (JCPDS card no: ) with Face centered cubic lattice and Fd-3m (227) space group confirmed the existence of Bi δ Fe Nb O 7. Bi 1.34 Fe 0.66 Nb 1.34 O 6.35 (JCPDS card no: ) also has Face centered cubic lattice and Fd-3m (227) space group. Lattice of NdFeO 3 (JCPDS card no: 00-63

65 ) is orthorohombic and space group is Pbnm (62). the increase of frequency. This decrease of dielectric constant with the increase of frequency can be happened due to dipole relaxation phenomenon, where the dipoles are able to follow the frequency of the applied field at low frequencies [10]. Fig-1: XRD pattern (Bi 2O 3Fe 2O 3) 0.4(Nb 2O 5 Nd 2O 3) C and 925 C sintered at Fig-3: Effect of frequency on Dielectric Constant of (Bi 2O 3Fe 2O 3) 0.4(Nb 2O 5 Nd 2O 3) 0.6 Fig-2: SEM image of (Bi 2O 3Fe 2O 3) 0.4(Nb 2O 5 Nd 2O 3) 0.6 sintered at a) 850 C and b)925 C B. SEM Fig.2 reveals the SEM images of (Bi 2 O 3 Fe 2 O 3 ) 0.4 (Nb 2 O 5 Nd 2 O 3 ) 0.6 sintered at 850 C and 925 C. For the sample sintered at 850 C sintering temperature, average size of grain is 0.44μm and average size of grain is 0.66 μm when sintering temperature is 925 C. Consequently it can be said that size of grain increases when sintering temperature increases. It can be explained by the following way. Unstable grain structure is an inherent nature of grain. Heating at higher temperature, migrate boundaries in the direction of their center of curvature due to unbalanced forces and rise of grain growth occurred as a result of the free energy change [9]. C. Dielectric Property Variation of dielectric constant and dielectric loss with the variation of frequency for (Bi 2O 3Fe 2O 3) 0.4(Nb 2O 5 Nd 2O 3) 0.6 are shown in Fig-3 and 4, respectively. For the samples sintered at 850 C and 925 C the dielectric constant are 35 and 16,respectively. It is observed from the Fig.3 and 4, that the dielectric constant and dielectric loss decreases rapidly with the increase of frequency. But at high frequency, dielectric constant slowly decreases with Fig-4: Effect of frequency on Dielectric Loss of (Bi 2O 3Fe 2O 3) 0.4(Nb 2O 5 Nd 2O 3) 0.6 Fig-5 and 6 represent the effect of temperature on dielectric constant and loss tangent. The dielectric constant is remained constant upto 300 C whereas dielectric loss is constant upto 600 C. Dielectric constant increases moderately after 600 C and drastic increase is observed at 800 C measuring temperature for the both samples sintered at 850 C and 925 C. Though dielectric loss starts to increase after 600 C, but drastic increase of dielectric loss is also observed at 800 C.This rise of dielectric constant and loss tangent reveals that an anti-ferromagnetism to paramagnetism transition [11] has occurred. So, it can be said that for this sample Neel temperature (T N ) will be nearly at 800 C. 64

66 Fig-5: Effect of Temperature on Dielectric Constant of (Bi 2O 3Fe 2O 3) 0.4(Nb 2O 5 Nd 2O 3) 0.6 Fig-7: Effect of Temperature on DC resistivity of (Bi 2O 3Fe 2O 3) 0.4(Nb 2O 5 Nd 2O 3) 0.6 sintered at 850 C Fig-6: Effect of Temperature on Dielectric Constant of (Bi 2O 3Fe 2O 3) 0.4(Nb 2O 5 Nd 2O 3) 0.6 D. DC Electrical Property Change of resistivity with temperature for (Bi 2 O 3 Fe 2 O 3 ) 0.4 (Nb 2 O 5 Nd 2 O 3 ) 0.6 at 850 C sintering temperature with different voltages are shown in Fig- 7. A metal to insulator transition region occurred in this sample and metal to insulator transition temperature is varying with the variation of voltages. Upto 200 C, metallic region is displayed by the sample and after that semiconducting nature is found upto 300 C. After 300 C, temperature has no effect on resistivity. Change of resistivity with temperature for (Bi2O3Fe2O3)0.4(Nb2O5Nd2O3)0.6 at 925 C sintering temperature with different voltages are shown in Fig-8. Metal to Insulator transition temperature of is 150 C for this sample. Though an anomaly is observed near about 200 C, overall an semiconducting behavior [12] is observed after 150 C measuring temperature. Fig-8: Effect of Temperature on DC resistivity of (Bi 2O 3Fe 2O 3) 0.4(Nb 2O 5 Nd 2O 3) 0.6 sintered at 925 C. IV.CONCLUSION Effect of sintering temperature on differernt properies of Nb+Nd doped bismuth ferrite are effectively performed here. Structural analysis revealed that three phases were found. Increasing grain size is found with the increase of sintering temperature from SEM image analysis. Dielectric properties exhibited the similar response of previous research and can be explained by preceding literature. Metal to insulator transition region is found from DC resistivity analysis. Phase found in this research are widely used for the application of gas sensor, though it is CO and HCs sensitive and also has high catalytic activities. ACKNOWLEDGMENT The author of this paper is grateful to Chittagong University of Engineering and Technology, Chittagong-4349 for their financial support and very much thankful to Pilot Plant & Process Development 65

67 Centre, Bangladesh Council of Scientific and Industrial Research, Dhaka-1207, Bangladesh for their technical support REFERENCES [1] A.Z Simoes, Filiberto Gonzalez Garcia,C.S Riccardi Piezoresponse behavior of niobium doped bismuth ferrite thin films grown by chemical method Journal of Alloys Compounds 493(2010) [2] Hemant Singh and K L Yadav Dielectreic, magnetic and megnetoelectric properties of La and Nb codoped bismuth ferrite J. Phys: Condens Matter 232 (2011) (6pp) [3] Hill N A 2000 Why are there so few Magnetic Ferroelectrics j.phys.chem.b [4] Youn-Ki Juna, Won-Taek Moona, Chae-Myung Changa, Hyun-Su Kima,Hyun Sam Ryua, Jae Wook Kimb, Kee Hoon Kimb, Seong-Hyeon Honga,* Effect of Nb doping on electric and magnetic properties in Multi-ferroic Bi FeO 3 ceramics Solid State Communications, 135 (2005) [5] Eerenstein W, Mathur N D and Scott Multiferroic and magnetoelectric materials J F 2006 Nature [6] Catalan J F and Scott Physics and Applications of Bismuth Ferrite 2009 Adv.Mater.21. [7] Ryu, S Priya, K Uchino, HE. Kim, "Magnetoelectric Laminate Composites of Piezoelectric and Magnetostrictive Materials", J Electroceram vol. 8, pp. 107,2002. [8] T Kanai, SI Ohkoshi, A Nakajima, T Watanabe, K Hashimoto, " A ferroelectric ferromagnet composed of (PLZT)x(BiFe0 3 )I-x solid solution" Adv Mater, vol.l3, pp.487, 2001 [9] C.Barry Carter, M.Grant Norton Ceramic Materials: Science and Engineering [10] Manoj Kumara and K. L. Yadav Rapid liquid phase sintered Mn doped BiFeO3 ceramics with enhanced polarization and weak magnetization, Applied Physics Letters 91, [11] Chandrashekhar P. Bhole Antiferromagnetic to paramagnetic phase transitions in bismuth ferrite (BiFeO3) ceramics by solid state reaction Ceramics- Silikáty 2012;56(2) [12] Ameer Azma, Ali Jawab, Arham S.Ahmedb, M. Chamanb, A.H Naqvib Structural, optical and transport properties of Al3+ doped BiFeO3 nanopowder synthesized by solution combustion method Journal of Alloys and Compounds 509 (2011)

68 Silicon Nanocrystals Rich Lanthanum Fluoride Films for Future Electronic Devices Md. Ferdous Rahman* 1 1 Department of Electronics & Telecommunication Engineerin, Begum Rokeya University Rangpur-5400, Bangladesh ferdousapee@gmail.com Sk. Rashel Al Ahmed 2 2 Department of Electronics & Telecommunication Engineering Pabna University of Science & Technology, Bangladesh. Md. Golam Saklayen 3 and Abu Bakar Md. Ismail 3 3 Department of Applied Physics & Electronic Engineering University of Rajshahi, Rajshahi 6205, Bangladesh Abstract Investigation on Silicon nanocrystals (Si-NCs) rich Lanthanum Fluoride (LaF 3 ) film fabricated using a novel one-step chemical method has been reported here. Colloidal solution of Si-NCs in hydrofluoric acid (HF) was prepared from meso-porous silicon by ultrasonic vibration (sonication). On a silicon (Si) substrate LaCl 3 solution in HCL is allowed to react with the colloidal solution of prepared Si-NCs. LaCl 3 reacts with HF of Si- NCs solution and produces LaF 3 crystals that deposits on the silicon substrate as a film embedding Si-NCs. This is a novel single step chemical way of depositing LaF 3 insulating layer embedding Si-NCs (LaF3:Si-NCs). The XRD and EDX analysis of the deposited film show a polycrystalline and non-stoichiometric nature of LaF 3. The presence of Si-NCs was confirmed by SEM and FTIR. Application of this material has been tested for low-voltage operating non-volatile memory (NVM) and Schottky junction solar cells. The Al/LaF 3 :Si-NCs/Al structure as non-volatile memory (NVM) offered a memory window of 525 mv at a programming and erasing bias of 2V. LaF 3 :Si-NCs films showed strong light absorption. Current-Voltage (I-V) characteristics of the Schottky device in ITO/LaF 3 :Si-NCs/Al structure showed a dependency on the incident light intensity where current changed under various light illumination. Experimental results show a lot of promise of Si-NCsrich LaF 3 film to be used as an insulating film in nonvolatile memory as well as a photoactive material in Schottkey junction solar cell. Keywords Silicon Nanocrystal, Nonvolatile Memory, Schottky Junction Solar Cell. I. INTRODUCTION Recently researchers have been considering nanocrystal-based memory devices as a solution to ultra-large scale integration of electronic nonvolatile memories. One major barrier to such integration of NVMs is the local defect related leakage. Using discrete nanocrystals instead of the conventional continuous floating gate as charge storage nodes, local defect-related leakage can be reduced efficiently to improve data retention [1]. In this regard, discrete-trap type semiconductor storage materials such as Si nanocrystals (Si-NCs) embedded in a dielectric matrix have been demonstrated as potential candidates for the fabrication of high-speed, high-density, low power-consuming, and nonvolatile memories [2-6]. Therefore, the poly-silicon oxide nitride oxide silicon (SONOS)-type structure memories including nanocrystal memories have recently attracted much attention for the application in the next-generation nonvolatile memories [7] [14] because of their great potential for achieving high program/erase (P/E) speed, low programming voltage and low power performance. For conventional SONOS, erase saturation and vertical stored charge migration [13-14] are the major drawbacks; while for nanocrystal memories good enough charge keeping capability of the discrete storage nodes and the formation of nanocrystals with constant size, high density and uniform distribution are the extremely challenging issues. LaF 3 has been chosen as alternative candidates for gate insulator because of their large band gap, high dielectric constant, and large refractive index. Moreover, the lanthanide fluorides (LaF 3 ) show good characteristics without pre-formed interfacial layer, and regarded as high-k dielectrics [15]. In this work our goal has been to create Si-NCs-rich LaF 3 film to be used as an insulating film in non-volatile memory as well as a photoactive material in Schottky junction solar cell. II. EXPERIMENTAL DETAIL Colloidal suspensions of silicon nanocrystals were fabricated from porous silicon in hydrofluoric (HF) acid by sonication (Ultrasonic Vibra cell, VCX 130) at a frequency of 20 khz. Porous silicon samples were prepared by standard electrochemical etching in a homemade double-tank cell [Figure-1] of p-type <100> Si wafers in a HF (48%): Ethanol = 2.5: 1 solution at different current densities for 30 minutes under room light illumination. Figure 1: Complete double-tank cell setup for PS fabrication. 67

69 After the etching process, the PS samples were immersed in hydrofluoric acid (HF) in a small plastic container and this container was immersed in water. Ultra sonic wave was applied to the water through the ultrasonic probe to vibrate the PS layer for about 60 minutes. The sonication was done at ultrasonic power 40 watt (amplitude~80%). After formation the colloidal suspensions of silicon nanocrystals in HF and then it will be allowed to react with Lanthanum chloride (LaCl 3 ) solution in hydrochloric acid (HCl) at room temperature. The chemical reaction between the LaCl 3 solution and colloidal solution of Si-NCs in HF can be given as LaCl 3 + HF LaF 3 + HCl Then the colloidal solution of LaF 3 with Si-NCs were made acid-free by centrifuging process using ultra cooling Sigma Laboratory Centrifuges Machine. LaF 3 with Si-NCs thin films were spin-cast onto indium tin oxide (ITO)-coated glass substrates at different rpm for different rotating times by VTC 100 Vacuum Spin Coater. Before applying colloidal solution of LaF 3 with Si-NCs on ITO coated flat glass substrate we have programmed VTC 100, it starts to spin with the specified parameters. III. RESULT AND DISCUSSION The X-ray diffraction of the deposited layer shows a polycrystalline LaF 3 deposition on silicon. The presence of Si -NCs was confirmed by Scanning Electron Microscopy (SEM) [Figure-4] as well. Figure 4: SEM on the LaE 3 surface. Scanning Electron Microscopy (SEM) has been performed by JEOL-JSM-6490LA with different magnification. SEM has been also investigated fabricated device of cross sectional view [Figure-5]. FTIR spectroscopy of the deposited LaF 3 powder also confirmed the presence of Si-NCs. Figure 2: Schematic diagram of a fabricated device for I-V characterization. For I-V characterization of the LaF 3 with Si-NCs deposited on ITO coated glass sample, Aluminium (Al) film was deposited onto the LaF 3 films with Si- NCs. Then the copper wires were connected onto the Al and ITO layer with silver paste. The arrangement for I-V characterization is shown in [Figure-2]. Figure 3: Schematic diagram of a fabricated device. Figure 5: SEM image of cross sectional view. The memory measurement has been performed by an impedance analyzer. Capacitance-voltage (C- V) study of the MIS [Al/LaF 3 :Si-NCs/Al] structure reveals that resonant tunneling of electron and charge storage was there when the MIS was biased from accumulation to inversion, which created a memory window. This type of memory window is called hysteresis. The MIS structure showed hysteresis for forward and reverse bias scan, enabling the structure to be used a non volatile memory. The Capacitance- Voltage (C-V) curves observed of the LaF 3 layer deposited nonvolatile memory device for various fryquencies and various bias voltage and comparative these (C-V) curves. The C-V characteristics were measured over a voltage range [ -2V to 2V and back to -2V ], [ -4V to 4V and back to -4V ], [ -6V to 6V and back to -6V ] at a room temperature with frequencies 500Hz, 1MHz, 2MHz, 3MHz, 4MHz and 68

70 5MHz. At 1MHz frequency and over a voltage range [ -2V to 2V and back to -2V ] hysteresis voltage difference (memory window) (525mV) is so good [Figure-6]. So, Finally observed that to study the C- V characteristics of MIS device, it was shown [Figure-6] that a memory window of about 525 mv is achievable at a bias voltage of (-2V to +2V), indicating of the stucture used as a non-volatile memory device. Figure 8(a): Light dependent I-V response of ITO/ LaF 3 :Si- NCs /Al Figure 6: C-V response of Al/ LaF 3 :Si-NCs /Al All experimental results of optical and electrical are discussed here for 120nm of postannealed LaF 3 films with Si-NCs. There were three [LaF 3 :SiNCs/ITO/Glass] structures fabricated for optical study and each had similar performance. The optical absorption spectrum of LaF 3 films with and without Si-NCs on ITO coated glass substrate were measured and are presented in [Figure-7]. The LaF 3 films with Si-NCs showed strong absorption. Figure 7: Absorption of LaF 3 film with Si-NCs and LaF 3 film.` The I-V characteristics of the film showed a dependency on the incident light intensity where current changed under various light illumination [Figure-8(a) & 8(b)]. Experimental results shows a lot of promise of Si-NCs embedded LaF 3 layer to be used as an insulating layer in MIS devices as well as a photoactive material in Schottkey junction solar cell. Figure 8(b): Light dependent I-V response of ITO/ LaF 3 :Si- NCs /Al VI. CONCLUSION In this work, The Al/LaF 3 :Si-NCs/Al structure was tested as NVM and a memory window of 525 mv was obtained at a programming and erasing bias of 2V when. The LaF 3 : Si-NCs films showed strong absorption.i-v characteristics of ITO/LaF 3 /Si/Al structure showed a dependency on the incident light intensity where current changed under various light illumination. Experimental results show a lot of promise of Si-NCs-rich LaF 3 film to be used as an insulating film in non-volatile memory as well as a photoactive material in Schottkey junction solar cell. REFERENCES [1] Aaron VY, Leburton JP: Flash memory: towards single-electronics. IEEE Potentials, pp.21:35, [2] Tiwari S, Rana F, Hanafi H, Hartstein A, Crabbe EF, Chan K: A silicon nanocrystals based memory. Appl. Phys. Lett., 68: [3] Hanafi HI, Tiwari S, Khan I: Fast and long retention-time nano-crystal memory. IEEE Trans Electron Devices, 43:1553, [4] King YC, King TJ, Hu C: Charge-trap memory device fabricated by oxidation of Si1-x Gex. IEEE Trans Electron Devices, 48:696, [5] Ng CY, Chen TP, Ding L, Fung S: Memory characteristics of MOSFETs with densely stacked silicon nanocrystal layers in the gate oxide synthesized by low-energy ion beam. IEEE Trans Electron Device Lett, 27:231, [6] Kapetanakis E, Normand P, Tsoukalas D, Beltsios K, Stoemenos J, Zhang S, van der Berg J: Charge storage and interface states effects in Sinanocrystal 69

71 memory obtained using low-energy Si+ implantation and annealing. Appl.Phys.Lett., 77:3450, [7] R. Lankhorst, S. Ketelaars, and M. Wolter. Nature Materials, 4:347, [8] S. Tiwari, F. Rana, H. Hanafi, A. Hartstein, E. F. Crabb e, and K. Chan. Appl.Phys.Lett. 68:1377,1996. [9] R. Muralidhar, R. F. Steimle, M. Sadd, and R. Rao.IEDM Technical Digest, [10] L. J. Guo, E. Leobandung, and Stephen Y. Chou.Science, 275:649, [11] S. M. Sze. Modern semiconductor device physics. John Wiley & Sons,INC, [12] Y. Shi, K. Saito, H. Ishikuro, and T. Hiramotob. J. Appl. Phys., 84:2358, [13] S. Tiwari, J. A. Wahl, H. Silva, F. Rana, and J. J. Welser. Appl. Phys. A, 71:403, [14] Y. Yu and M. Cardona. Fundamentals of semiconductors: physics and materials properties. Spring Press, Berlin, [15] A.J. Steckl, J.Xu, H.C. Mogul and S.M. Prokes, J. Elecrochem. Soc, 142(5), pp ,

72 Study on the Displacement Effect at Cylindrical Ionization Chamber with Different Radii in High Energy Photon of Flat Beam and True Beams Kumaresh Chandra Paul Dept. of Medical Physics and Biomedical Engineering, Gono Bishwabidyalay Dhaka, Bangladesh Guenther H. Hartmann Dept. of Medical Physics in Radiation Oncology, German Cancer Research Center, Im Neuenheimer Feld 280, Heidelberg, Germany Enamul Hoque Dept. of Physics, Jahangirnagar University, Savar, Dhaka, Bangladesh Golam Abu Zakaria Dept. of Medical Radiation Physics, Gummersbach Hospital, Academic Teaching Hospital of the University of Cologne, Germany Abstract Absorbed dose to water determination in the clinical practice protocols require the placement of the air-filled ionization chambers in the water phantom, which introduces several fluence perturbations in high-energy photon and electron beams. Displacement perturbation is one of them, which is to be considered in dosimetry. It is possible to correct the displacement effect by introducing the chamber-specific quality correction factor (k Q ) or by introducing the concept of effective point of measurement (EPOM). The EPOM is the point in the chamber at which the measured dose would be the same as in the measuring depth in the absence of the radiation sensitive device or dosimeter. The aim of this study was to measure the displacement effect at cylindrical ionization chambers in 6 and 10 MV true and flat photon beams. Linear accelerator, cylindrical chambers, Semiflex chamber, the Roos chamber and water phantom were used in the study. Percentage of depth doses (PDDs) were considered for determining the shift of EPOM with respect to the well established Roos chamber. The displacement effect obtained a range of 0.25 to 0.57 times r (radius of the chamber) both in true and flat photon beams, which disagreed with the TRS-398 protocol recommended constant value of 0.6 r. Keywords cylindrical ionization chamber, displacement effect, true and flat photon beams. INTRODUCTION The effective point of measurement plays a important role in relative megavoltage photon beam dosimetry [1]. The determination of the dose to water from the detector signal is based on the cavity theory [2, 3]. Using of ionization chamber a problem arises: it will replace water equal to the volume to the chamber. This replacement has a certain influence on the dose determination which is referred to a replacement effect. This displacement effect requires special consideration in particular at cylindrical ionization chambers. The correction of displacement is possible to be compensated by two alternative ways. Other than multiplying the displacement perturbation factor P dis [4], it can be used to put the center of the chamber on the depth of interest on the central axis of the beam [5]. The shift of the chamber can be used to do the correction. There are two concepts to take into account the displacement effect: (i) can be compensated by setting an EPOM at the measuring depth, or (ii) by introducing a displacement correction factor [6, 7]. The displacement effect is the shift of EPOM divided by the chamber radius. The x-ray beam is coming out from the linac head after interacting with scattering foil (filter) is called flat beam and the beam coming out without the interaction is called true beam. In true beam the dose rate is higher than that of flat beam. The objective of the study was to measure the displacement effect at cylindrical chamber in 6 and 10 MV true and flat beams. MATERIALS Linear accelerator (Elekta Versa HD,) Roos chamber, six specially designed cylindrical chambers, Semiflex chamber and water phantom (IBa blue phantom) were used for the study. 6 and 10 MV true beam and flat photon beams were used in this experimental study. Measurement was performed with six Farmer chambers with inner radius, ranging from 1.0 mm to 6.0 mm. These chambers were water proof and types were TM , , 30013, , and , PTW Freiburg, Germany. For all six chambers the cavity length was 23 mm, the wall was made up of PMMA (poly-methyl methacrylate) with a thickness of mm and an additional thin graphite layer of 0.09 mm, the central electrode was consist of aluminum with a diameter of 1.15 mm. The chambers were referred as 71

73

74 PDD PDD International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering Percentage Depth Dose (PDD) Curves in Photon Beams 6 MV Photon Beams 6 MV FFF Photon Beams MV Photon Beams (c) (a) 6 MV-Roos R1-6 MV R2-6 MV R3-6 MV R4-6 MV R5-6 MV R6-6 MV Semiflex-6 MV Roos-10 MV R1-10 MV R2-10 MV R3-10 MV R4-10 MV R5-10 MV R6-10 MV Semiflex-10 MV PDD PDD MV FFF Photon Beams (d) (b) Roos-6 MV FFF R1-6 MV FFF R2-6 MV FFF R3-6 MV FFF R4-6 MV FFF R5-6 MV FFF R6-6 MV FFF Semi-6 MV FFF 10 MV FFF R1 10 MV FFF R2 10 MV FFF R3 10 MV FFF R4 10 MV FFF R5 10 MV FFF R6 10 MV FFF Semi 10 MV FFF was found that the more accurate positioning of EPOM was given by a distance of 1.5 mm above from the front surface of the chamber. The well established Roos chamber was carefully set as -1.5 mm from the front surface of the chamber and was taken as the reference measurement depth for the study. The comparison of displacement effects between IAEA, TRS-398 protocol [10] and measured results were shown in the figure-3. The graph shows that the experimental results are below the TRS-398 recommended protocol (red line) in all the energies for true and flat beams Depth (cm) Depth (cm) CONCLUSION displacement/radius Graphs: PDD curves in: (a) 6 MV flat beam, (b) 6 MV true beam, (c) 10 MV flat beam and (d) 10 MV true beam Displacement effect comparing with TRS MV Photon 6 MV (FFF) Photon 1.0 Measured 6 MV TRS 398 TRS measured 6 MV FFF The experimental displacement effect appeared to be dependent on beam energy, the chamber cavity and independent on depth after the depth of dose maximum. A recommended constant value of displacement correction factor of the protocol does not satisfy with this experimental study. Monte Carlo simulated study can be the alternative solution for estimating the chamber specific displacement effect displacement/radius TRS MV Chamber radius (mm) TRS MV FFF Figure-2: PDD in 6 and 10 MV flat and true photon beams. DISCUSSION In the plane parallel chamber, the position of the EPOM is at the inner surface of the entrance window at the center of the opening window in electron and photon beams. The positioning the Roos chamber was placed -1.5 mm (- sign indicates toward the radiation source) so that the EPOM was on the measuring depth in the water phantom. The meaning of measuring depth at z is equal to z m + d f - d eq, where z m is measuring depth, d f is the thickness of the entrance window and d eq is it s water equivalent thickness using the ratio of the densities of water and entrance window materials. Figure-3: Comparison of displacement effect between measured value and TRS-398 recommendation. A few numbers of studies aiming at the reduction of the uncertainty related to the exact positioning of the EPOM [8, 9]. With respect to the Roos chamber it ACKNOWLEDGEMENT It was gratefully acknowledged the PTW Freiburg for designing and building the non standard Farmer chambers for the study. It was acknowledged the support of the Department of Radiation Oncology, University Clinic Mannheim; University of Heidelberg, Germany and the financial assistance by DAAD for the study. It was specially acknowledged the Department of Medical Physics and Biomedical Engineering (MPBME), Gono Bishwabidyalay (GB) for selecting the author as Ph. D researcher under the collaboration with Heidelberg University, Germany and MPBME, GB. REFERENCES [1] Kawrakow I, The effective point of measurement in megavoltage photon beams, Med. Phy. 33: , [2]. Spencer LV, Attix FH. A theory of cavity ionization. Radiat Res. 1955; 3(3): [3]. Spencer LV, Attix FH. A cavity ionization theory including the effects of energetic secondary electrons. Radiology 1955; 64(1):113. [4]. Andreo O, Burns DT, Hohlfeld K, Huq MS Kanai T, Laitano F, Smyth V, Vynckier S. Absorbed dose determination in external beam radiotherapy. An international code of practice for dosimetry based on standards of dose to water. Technical report Series TRS- 73

75 398, Vienna. International Atomic Energy Agency [5]. Skaggs LS. Depth dose of electrons from the betatron. Radiology 1949; 53 (6): [6]. Ervin, B. Podgorsak. Review of Radiation oncology Physics: A Handbook for Teachers Students, International Atomic Energy Agency. 2003; pp-179. [7]. Mayles P and Nahum A; Hand book of radiotherapy physics. Theory and Practice, Taylor & Francis, New York, London, 2007: 90. [8]. Das IJ, Mc Neeley SW, Cheng CW. Ionization chamber shift correction and surface dose measurements in electron beams. Phy Med Biol 1998; [9] Shimono T, Nanbu H, Koshida K, Kikuchi Y. Analysis of the effective point of measurement of a thimble chamber dosimeter set parallel to the X-ray beam axis. Igaku Butsuri 2007; [10]. Andreo O, Burns DT, Hohlfeld K, Huq MS Kanai T, Laitano F, Smyth V, Vynckier S. Absorbed dose determination in external beam radiotherapy. An international code of practice for dosimetry based on standards of dose to water. Technical Report Series TRS- 398, Vienna. International Atomic Energy Agency

76 Electrical and Optical Properties of Cu-Nanoparticles- Doped α-fe 2 O 3 Thin Film Spin-Coated on Glass Substrate Sanjida Ferdous *, Afroza Yasmin, Jinia Sultana and Abu Bakar Md. Ismail Department of Applied Physics and Electronic Engineering University of Rajshahi Rajshahi-6205, Bangladesh *sfj.sanju11@gmail.com Abstract Copper nanoparticles (Cu-NPs) doped α- Fe 2 O 3 films were prepared by spin coating method on glass substrate. Cu-NPs were prepared by chemical reduction method using hydrazine hydrate as reducing agent and was confirmed by UV-VIS spectroscopic signature. Study of absorbance and transmittance were done by UV-visible spectrophotometer. The optical band gap of the films was determined using absorbance data. From the optical measurement, the indirect band gap of the thin film was estimated to be around 1.67 ev. The resistivity of Cu-NPs doped α-fe 2 O 3 films were measured by Van-der-paew s method. Cu-NPs were found to influence resistivity of Fe 2 O 3 thin films.the electrical and optical property of Fe 2 O 3 were engineered by doping it with Cu-NPs. Keywords Oxide semiconductor, photocatalysis, copper nanoparticles, resistivity, optical absorption. I. INTRODUCTION Recently, considerable interest has been focused on α-fe 2 O 3 to be used for photoelectrochemical water electrolysis half reaction [1-4] because of its special characteristics and low cost. It has an appropriate band gap ( 2 ~ 2.2 ev ) [5] which is good for water splitting. Solar water splitting consists of two parts where one is anode and another is cathode. Hydrogen is produced in cathode and oxygen is produced in anode section. In this work, the electrical and absorption properties of copper nanoparticles doped iron oxide (Cu-NPs doped α-fe 2 O 3 ) was investigated for further use to fabricate photoanode. Again, α-fe 2 O 3 has short hole diffusion length. It has low conductivity and poor absorbance in solar radiation. To overcome these problems we used Cu- NPs as doping material. Dopants including Mg, Cu, Zn ( p-type ) [6] Ti, Sn, Zr, Si, Ge (n-type) [6] are well known for improving photoelectrochemical properties of iron oxide. Many researchers used costly materials like platinum [7] or gold [8] nanoparticles for doping with Fe 2 O 3 for the enhancement of electrical conductivity and absorbance of Fe 2 O 3. We searched for a low cost material. That s why we used copper nanoparticles for doping with Fe 2 O 3. At first fabrication of Fe 2 O 3 thin film on glass substrate has been done and its properties were studied. Then Cu-NPs doped Fe 2 O 3 was fabricated and studied its resistivity and absorbance. II. EXPERIMENTAL A. COPPER NANOPARTICLES SYNTHESIS AND EXTRACTION Copper nanoparticles were prepared in aqueous solution by reducing Cu 2+ ions with Sodium borohydrate, Hydrazine hydrate. The metallic copper produced was immediately capped by SDS (Sodium Dodycyle Sulphate) to prevent further growth. The detailed procedure of copper nanoparticles was as follows: The particles were prepared in aqueous phase by chemical reduction of copper sulfate using sodium borohydred of hydrazine hydrate in the presence of capping agent SDS. For a typical set, ml of CuSO 4 (0.005M-0.1M) was added with a small amount(0.5ml-1ml) of 0.5M NaOH and stir for a time period of 20 minutes ml of aqueous solution of SDS was then added dropwise to it under constant stirring. After 10 minutes 1-2ml Hydrazine hydrate was added to the solution dropwise while the solution was at constant stirring. The solution color changed to dark brown on complete addition of reducing agent indicating the formation of nanoparticles. The Cu-NPs was confirmed by UV- VIS spectroscopic signature[9]. Then the particles were extracted by centrifugation. CuSO 4 + NaOH SDS Hydrazine Hydrate Formation of dark Brown Nanoparticles B. CU-NPS DOPED Fe 2 O 3 SYNTHESIS The extracted Cu-NPs were mixed with Fe 2 O 3 solution. Thus we synthesised Cu-NPs Doped Fe 2 O 3. For changing the doping concentration, we changed the concentration of CuSO 4. We used CuSO 4 of 0.005M, M, 0.01M to change the doping concentration. C. CHARACTERIZATION Electrical resistivity of the deposited films were measured by using Van der Pauw s method. Optical 75

77 Absorbance (a.u) Resistivity (ohm-m) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering spectra has been measured in wavelength ranges 360 <λ< 1000 nm using a T60 UV-visible spectrophotometer. The optical band gap is being evaluated using absorbance data. A. RESISTIVITY III.RESULTS AND DISCUSSION Electrical resistivity of metal and semiconductor film of any shape may be measured by using Van der Pauw s technique. The resistivity of film having any arbitrary shape can be uniquely determined by using this method. Figure-1 shows the Van-der-Pauw s specimen with four small contacts A, B, C, and D in order; 1, 2, 3 and 4 indicate the terminals of the electrometer for the measurement of currents and voltage, respectively. Through the commutative switches the connections are made between the film and the meter terminals. decreased by doping it with Cu-NPS. That means the conductivity is increased Concentration(M) Fig-2: Variation of resistivity with various concentration of CuSO 4. It was found that the resistivity of Fe 2 O 3 films increased with the increasing concentration of CuSO 4. Conversion rate of CuSO 4 increases with the increase of molar ratio of reducing agent (hydrazine hydrate) to CuSO 4 [10], and why increase in concentration of CuSO 4 keeping the concentration of reducing agent hydrazine hydred constant resulted in less conversion of CuSO 4, which in turn resulted in higher resistivity of Fe 2 O 3 film. Fig-1: Schematic diagram for the measurement of film resistivity using Van-der-Pauw s method; 1, 2, 3 & 4 are meter terminals and A, B, C,& D are the film terminals. Using reciprocal theorem Van der pauw s (1958) showed that R t AB, CD R 2 BC, DA R f R AB, CD BC, AD m (1) The correction factor f has been calculate by Van-der -pauw and is equal to unity when R AB, CD R BC,AD and then.266t( R R ) m 2 AB, CD BC,DA where, t is the thickness of film in meter. (2) From this equation we calculated the resistivity of the films. Fe 2 O 3 has a resistivity of about 10 Ω-m. Figure-2 shows that the resistivity of Fe 2 O 3 is B. ABSORBANCE Fig-3 shows the optical absorbance of the Cu-NPs doped Fe 2 O 3 thin films deposited on glass substrate. The absorption spectrum of Cu-NPs doped Fe 2 O 3 thin film exhibit strong peaks at lower visible wavelength (< 600 nm). The critical wavelength treated as 500 nm. Above 500 nm, the absorbance of Cu-NPs doped Fe 2 O 3 thin film decreased L(1).005M L(2).0075M L(3).01M Wavelength (nm) Fig-3: Absorbance of Cu-NPs doped Fe 2 O 3 for various concentration of CuSO 4. Fig-4 shows the absorbance of Fe 2 O 3 deposited thin film. Comparing Fig-3 and Fig-4, it shows that the 76

78 sqrt(absorbance*energy) Absorbance(a.u) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering absorbance of Fe 2 O 3 is enhanced by doping it with Cu-NPs Wavelength(nm) Fig-4: Absorbance of Fe 2 O 3. In order to determine the clear fundamental absorption edge, we have to calculate the energy gap E g.. The energy band gap is determined from the curve of sqrt( absorbance*energy ) verses energy for the Cu- NPs doped Fe 2 O 3 thin film. The curve has a good transient line fit over higher energy range above the absorption edge. It indicates an indirect optical band gap. Based on Figure-5, the indirect band gap of Cu- NPs dopped Fe 2 O 3 film has been calculated and the band gap is 1.67 ev IV.CONCLUSION In this work, we used spin coating method for Cu-NPs doped Fe 2 O 3 film deposition. The prepared films showed the enhancement of electrical conductivity compared to the only Fe 2 O 3. The absorption spectrum indicated the strong absorption ability of Cu-NPs doped Fe 2 O 3 films. It has high absorption peaks at lower visible wavelengths. From the optical measurement, the indirect band gap F g is estimated to be 1.67eV. The band gap indicated that the absoption band of Fe 2 O 3 was increased by doping it with Cu-NPs.. REFERENCES [1] Steinfeld, A.; Weimer, A. W, Thermochemical Production of Fuels with Concentrated Solar Energy. Optics Express, vol.18, pp.a100-a111, April [2] Christopher perkins, Solar thermochemical production of renewable hydrogen. AlChE Journal, vol.55, pp , February [3] Kanan, M. W.; Nocera, D. G, In situ formation of an oxygen-evolving catalyst in neutral water containing phosphate and Co2+. Science, vol.321, pp , August [4] Brimblecombe, R.; Dismukes, G. C.; Swiegers, G. F.; Spiccia, L, Molecular water-oxidation catalysts for photoelectrochemical cells. Dalton Transactions, vol.43, pp , [5] M.F,Al, M. Saleem, S.M.A Durrani, Optical properties of iron oxide ; Journal of Alloys and Compounds, Vol.521, pp , April [6] Bard, A. J.; Faulkner, L. R, Electrochemical methods : fundamentals and applications. 2 ed.; John Wiley & Sons: New York, energy (ev) Fig-5: The curve of sqrt(absorbance*energy) vs energy for the Cu-NPs doped Fe 2 O 3 film. [7] Ofer Neufeld and Maytal Caspary Toroker, Platinum- Doped α-fe 2 O 3 for Enhanced Water Splitting Efficiency ; The Journal of Physical Chemistry C, vol.119, pp , February [8] Elijah Thimsen, Florian Le Formal, Influence of Plasmonic Au Nanoparticles on the Photoactivity of Fe 2 O 3 electrodes for Water Splitting ; Nano Letters, vol.11, pp.35-43, December [9] Om ParkashSiwach.P.Sen, Synthesis and study of fluorescence properties of Cu nanoparticles. J Nanopart Res, vol. 10, pp , March [10] ZHANG Qiu-li et al, Preparation of copper nanoparticles by chemical reduction method using potassium borohydride, Trans. Nonferrous Met. Soc. China, vol.20 pp.s240 s244, December

79 Study on Morphological Properties of Cu-NPs Doped α-fe 2 O 3 Thin Film Deposited on Glass Substrate Jinia Sultana*, Afroza Yasmin, Sanjida Ferdous and Abu Bakar Md. Ismail Department of Applied Physics and Electronic Engineering University of Rajshahi Rajshahi 6205, Bangladesh jinia54@gmail.com Abstract Morphological properties of Copper Nanoparticles (Cu-NPs) doped Iron Oxide (α-fe 2 O 3 ) have been investigated. For higher catalytic effect Cu- NPs was mixed with Fe 2 O 3 and the structural and morphological properties of synthesized material have been investigated and compared with Fe 2 O 3 without Cu- NPs. The presence of Cu-NPs inside the prepared Fe 2 O 3 was confirmed by FTIR analysis and its spectroscopic signature determined by UV-VIS spectroscopy. Atomic Force Microscope has been used to study the surface morphology of Cu-NPs-doped Fe 2 O 3 films. The mean grain size was started decreasing with increasing layers for the same concentration of CuSO 4. The mean grain size and the average surface roughness were increasing with the film thickness and with increasing concentration of CuSO 4. Index Terms: Photo-catalysis, Oxide Semiconductor, Cu-NPs, Surface morphology. I.INTRODUCTION Intense researches have been going on worldwide to find alternative solutions for present energy crisis. Thin films are expected to play an increasingly important role in the studies of a variety of solid-state phenomena of basic and practical interest. Iron-oxide (Fe 2 O 3 ) is promising earth abundant materials for photoelectron chemical water electrolysis half-reaction. Fe 2 O 3 has an appropriate band gap (2~2.2 ev) [1] which is good for water splitting. It has also corrosion free characteristics. But the biggest problem of Fe 2 O 3 is its short hole diffusion length; it is known that hole diffusion length is only about 2~20 nm. It shows lower current and poor absorbance. One possible solution to this problem is to dope Fe 2 O 3 with appropriate material. Doping Fe 2 O 3 with other material might introduce a change in the morphology of the thin-film of Fe 2 O 3, that in return might influence the catalytic activity of the photoelectrode fabricated using Fe 2 O 3. Dopants including Mg, Cu, Zn (p-type) [2], Ti, Sn, Zr, Si, Ge (n-type) [2] are well known for improving photoelectron chemical properties of iron-oxide by substituting Fe site and donating carriers. Many researchers used costly materials like platinum [3] or gold [3] nanoparticles for doping with Fe 2 O 3 for the enhancement of electrical conductivity. Under this background low-cost Cu-NPs have been used in this study to investigate the influence of Cu-NPs on the morphology of Cu-NPs-doped Fe 2 O 3. α-fe 2 O 3 was synthesized by mixing 1M iron chloride (FeCl 2 ) solution with 0.1M urea (NH 2 CONH 2 ); where ethanol is used as solvent. The Cu-NPs was prepared in aqueous solution of copper sulfate (CuSO 4 ) with sodium borohydrate and hydrazine hydrate. To prevent further growth, the metallic copper produced was immediately capped by SDS (Sodium Dodycyle Sulfate). The presence of Cu-NPs is verified by testing UV-Visible absorption spectrum [4]. Cu-NPs were extracted by centrifuge the solution at rpm. Then the extracted Cu-NPs were dried at 80 0 C for 10 minutes. The doping concentration is changed by changing the concentration of CuSO 4 solution. Fourier Transform Infrared (FTIR) Spectroscopy has been used to know the composition of the synthesized Cu-NPs-doped Fe 2 O 3 films. The main purpose is to enhance the optical properties of earth abundant α-fe 2 O 3 to be used in solar water splitting. II.EXPERIMENTAL The Cu-NPs were prepared in aqueous phase by chemical reduction of CuSO 4 using sodium borohydrate of hydrazine hydrate in the presence of capping agent SDS. For a typical set, ml of CuSO 4 (0.005M-0.1M) was added with a small amount (0.5ml-1ml) of 0.5M NaOH and stir for a time period of 20 minutes ml of aqueous solution of SDS was then added drop wise to it under constant stirring. After 10 minutes 1-2ml hydrazine hydrate was added to the solution drop wise while the solution was at constant stirring. The solution color changed to dark brown on complete addition of reducing agent indicating the formation of nanoparticles. The particles were then extracted by centrifugation. CuSO 4 +NaOH SDS Hydrazine hydrate Formation of dark brown nanoparticles. For iron-oxide (Fe 2 O 3 ) film deposition FeCl 2, Urea (NH 2 CONH 2 ) and ethanol is used as solvent solution for spin coating method. 78

80 IRON-OXIDE SYNTHESIS 1M FeCl 2 Solution Preparation Distilled Water FeCl 2 25 ml g 0.1 M NH 2 CONH 2 Solution Preparation Distilled Water 12.5 ml NH 2CONH g Then the two solutions were mixed and stirred it for 8 minute. After that iron oxide solution is prepared. CU-NPS DOPED FE 2 O 3 SYNTHESIS The extracted Cu-NPs were mixed with Fe 2 O 3 solution. Thus Cu-NPs doped Fe 2 O 3 were synthesized. For changing the doping concentration, the concentrations of CuSO 4 were changed.30ml CuSO 4 of 0.005M, M and 0.01M were used to change the doping concentration. SPIN COATING This procedure is used to apply uniform thin films on to flat substrate. After preparing the solution of iron oxide with Cu-NPs, deposition of the films by spin coating were started. At first, the spin coater was set for two intervals of time with corresponding speeds. Vacuum spin coater (Model, VTC-100) has been used in this process. When the spin coater started rotating, the solution was drop cast on the slide. Rotational Speed & Time for Film Preparation SP1 T1 SP2 T2 500 rpm 3s 7000 rpm 30s Grain size, r = 14.4 / (ΔΕ.ε) (3) Here r is expressed in angstroms when E is in ev, is the dielectric constant and its value is taken as 3, halfway between those of vacuum and glass and e is the electronic charge. The grain size is also thickness dependent. The grain size increases with the film thickness. Neugebauer and Webb have shown that the values of r are extended from 8 Aº for the thinnest and 26 Aº for the thickness film. ROUGHNESS MEASUREMENT It is a measurement of surface irregularity or unevenness along out of plane of a thin film. They play a key role in transparent & optical properties like by inducing surface states, traps, scattering sites etc to name a few. There are two roughness, one is average roughness and other is rms roughness. They represent the deviation of hillocks and valleys (or pits) on the film surface from a reference plane. Avg. roughness is simply the average of positive (hillocks deviation) and negative (valley deviation) values from the reference plane. Here average roughness has been calculated. The roughness has been measured from the AFM image. The roughness was measured by the horizontal and vertical line of the XEI software, then the average value has been taken.. STUDY WITH ATOMIC FORCE MICROSCOPE (AFM) After the spin coater had stopped, the slide was taken off. The film had been dried in the air. The solution had been drop cast again on the dried films. This process was repeated several times to prepare films of different layers as well as different thickness. ANNEALING After the deposition, the samples were annealed in a thermal annealing furnace (carbolite CWF 12/13) in air. At first the deposited films placed in a heater. Then the temperature increased slowly up to desired level. Then the prepared film is annealed at c for 4 hours. Then iron oxide is converted to α-fe 2 O 3. III.RESULTS AND DISCUSSION GRAIN SIZE MEASUREMENT The activation energy may be expressed in terms of r, e and ε according to Neugebaur and Webb [5]. ΔΕ = e 2 / (ε.r) (1) Or, r = e 2 / (ΔΕ.ε) (2) Fig-1: AFM image of 0.005M molar concentration at layers 3 of Cu-NPs:Fe 2O 3 film. The surface morphology of Cu-doped-α-Fe 2 O 3 films were studied by AFM. The grain size of the deposited films has been measured from the AFM image using XEI software. Figure 1 shows the AFM image of 0.005M molar concentration of CuSO 4 at layers 3 of Cu-NPs:Fe 2 O 3 film. The variation of grain size of the films with different concentration of CuSO 4 at different layers are given below. 79

81 Grain size(µm²) Grain size(µm²) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering Concentration of CuSO4 At layers 3 At layers M (µm²) 2.332(µm²) M (µm²) 3.15 (µm²).01m (µm²) (µm²) The roughness of the deposited films has been measured from the AFM image. The variation of the roughness of the films with different concentration of CuSO 4. at different layers are given below. Frequency [cm -1 ] Associated Species Reference 399 Fe-O(Bonds) [6] 611 Fe-O asymmetric stretching [6] 1384 N-O bending [6] 1619 H 2O stretching [6] 3411 O-H stretching [6] Variation of grain size with different concentration of CuSO 4 at layers 3. molar Concentration of CuSO 4. At layers 3 At layers M (nm) (nm) M (nm) (nm) 0.01M (nm) (nm) STUDY WITH FTIR Fourier Transform Infrared (FTIR) spectroscopy measurements of Cu-doped-Fe 2 O 3 sample (Powder sample) have been performed by using Perkin-Elmer Spectrum 100 spectrometer, using pressed KBr pallets. The spectrum for the prepared sample has been recorded in the range from 225 to 4000 cm Concentration(M) Fig-3: Variation of grain size with different concentration of CuSO 4 at layers 3. Variation of grain size with different molar concentration of CuSO 4 at layers Fig-2: FTIR transmittance spectra of Fe 2O 3 with embedded Cu-NPs sample. Figure 2 shows the FTIR transmittance spectra of Cudoped-Fe 2 O 3. In the FTIR spectra, the absorption bands at around cm -1 and cm -1 are attributed to O-H stretching and bending vibration of H 2 O respectively. The absorption bands at around cm -1 is attributed to N-O stretching and bending vibration respectively. It is seen from figure 2 that the samples have a characteristics absorption band at cm -1 of the asymmetric stretch mode of the Fe-O bond. It is also seen that absorption band at cm -1 and at cm -1, of the stretching mode of Fe-O bonds. This proves that there is significant amount of Fe 2 O 3 in the sample Concentration(M) Fig-4: Variation of grain size with different concentration of CuSO 4 at layers 5. Identification of the IR modes observed in the Fe 2 O 3 with embedded Cu- NPs powder sample. 80

82 Roughness(nm) Roughness(nm) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering Variation of roughness with different molar concentration of CuSO 4 at layers Concentration(M) Fig-5: Variation of roughness with different concentration of CuSO 4 at layers 3. Variation of roughness with different molar concentration of CuSO 4 at layers IV.CONCLUSION It can be observed that the film thickness increased linearly with the increasing number of deposition layers. The mean grain size decreases with increasing layers for the same molar concentration of CuSO 4 while it increases with increasing layers for different molar concentration of CuSO 4. The roughnesses of the films are considerably high. The average roughness increases with increasing layers for different molar concentration of CuSO 4. REFERENCES 1. M.F, Al, M. Saleem, S.M.A Durrani, Optical properties of iron oxide, Journal of Alloys and Compounds, vol. 521, pp , April Bard, A. J., Faulkner, L. R., Electrochemical methods : fundamentals and applications, 2nd ed., John Wiley & Sons: New York, Brimblecombe R., Dismukes G. C., Swiegers G. F., Spiccia, L., Molecular water-oxidation catalysts for photoelectrochemical cells, Dalton Transactions, vol. 43, pp , November Om Parkash Siwach, P. Sen, Synthesis and study of fluorescence properties of Cu nanoparticles, J Nanopart Res, vol. 10, pp , C.A. Neugebauer and M.B. Webb, Electric conductivity and structure of discontinuous metal films on dielectric, J. Appl. Phys.33, vol. 7, pp , May Wojcieszak.R., Ghazzal.M.N., Gaigneaux.E.M., Ruiz. P., Low temperature oxidation of methanol to methyl formate over Pd nanoparticles supported on γ-fe 2 O 3, Catalysis Science & Technology,vol.4,pp , Concentration(M) Fig-6: Variation of roughness with different concentration of CuSO 4 at layers 5. 81

83 MRI Segmentation using Fuzzy C-Means Clustering and Bidimensional Empirical Mode Decomposition Gulam Sarwar Chuwdhury 1 Md. Khaliluzzaman 2* Dept. of Computer Science & Engineering Dept. of Computer Science & Engineering International Islamic University Chittagong International Islamic University Chittagong Chittagong, Bangladesh Chittagong, Bangladesh Md. Rashed-Al-Mahfuz 3 Dept. of Computer Science & Engineering University of Rajshahi, Rajshahi, Bangladesh Abstract - Image segmentation is a vital step in medical image processing. Magnetic resonance imaging (MRI) is used for brain tissues extraction in white and gray matter. These tissues extraction help in image segmentation applications such as radiotherapy planning, clinical diagnosis, treatment planning. This paper presents utilization of fuzzy C-means (FCM) clustering by using wavelet and bidimensional empirical mode decomposition (BEMD) to improve the quality of noisy MR images. The signal to noise ratio (SNR) value is calculated from FCM clustering data to examine the best segmentation technique. The experiment with synthetic Brain Web images has demonstrated the efficiency and robustness of the appropriate approach in segmenting medical MRI. Keywords - Image segmentation; fuzzy C-means; magnetic resonance imaging; wavelet, BEMD; SNR I. INTRODUCTION The joint space-spatial frequency representations have received spatial attention in the field of vision, image processing, and pattern recognition. Image segmentation from Magnetic Resonance (MR) image is such a challenging and significant step in medical image analysis. This step is essential in many medical image applications. Magnetic Resonance Imaging (MRI) is a noninvasive method for imaging internal tissues and organs. As MR images have difficult nature, also have no linear features. MR images performance is affected by many issues as partial volumes effects (PVE) which means a pixel contains more than one tissue, this leads to misclassification as a result of blurred boundary between artifact intensities of the same tissue are not constant over the image spatial domain and geometric deformation [1]. To satisfy the increasing requirement of image segmentation from MR images a variety of segmentation methods have been developed over past several years. The problems of image segmentation i.e. identification of tissues and organs of MR images have been described extensively in the literature and many algorithms have been developed in an attempt to solve the problems. One of the goals of segmentation of MR images is to determine the volumes of organs, tissues and lesions present in a given patient. These volumes and the changes in these volumes over time, may aid in the diagnosis, prognosis and treatment planning of patients under investigation. Segmentation process also helps to find region of interest in a particular image. The key goal is to make image more simple and meaningful. There are different classifications of medical image segmentation techniques, however, no standard classification technique. The most commonly used segmentation techniques can be classified into many approaches, such as region based segmentation techniques that look for regions satisfying a given homogeneity criterion [2], edge-based segmentation techniques that look for edges between regions with different characteristics [3], artificial neural networks base method [4], data fusion based methods [5], Markov random field based methods [6] and hybrid based Methods [7]. Fuzzy C-means (FCM) is being employed for quite some time. A method given in [8] proposes an MRI segmentation using neural network based FCM clustering algorithm. The authors have performed experiment on one channel MR data, however, whereas MR images are multi-spectral and provides additional information; due to noise and in homogeneity this algorithm fails to work. The FCM based techniques in [9] employ Gaussian smoothing to produce a more homogeneous and low-noise subject to work with. This approach is, however, limited by the equal feature weights of the standard FCM. In [10], a modified FCM algorithm is given for MRI brain image segmentation using both local and non-local spatial constraints. This technique takes into account local and non-local spatial information using a variation index instead of typical distance metric. This approach has some limitation for its applicability to large 3D data and is computationally expensive. In [11], authors employ the FCM segmentation of MRI in a technique. By using the Gullied filter for pre-processing to remove in homogeneity in the images. A method proposed for image segmentation using wavelet based multi-resolution Expectation Maximum (EM) algorithm in [12]. The drawback in EM is that it is based on identical and independent distribution of pixel intensities which may not be the case with noisy medical images. In [13], represent a method of Fuzzy C-Means clustering by using wavelet decomposition technique for features extraction. This method performs better for smooth MR images but not satisfactory for the noisy MR images. This method is non-adaptive to characterize textures by filter responses directly. In [14], gives an overview of the state-of-the-art methods of EMD to decompose an image into a number of IMFs and a residue image with a minimum number of extrema points. In [15], authors developed an algorithm based on BEMD to extract features at 82

84 multiple scales or spatial frequencies. This method derived data from image that are fully unsupervised and permits to analyze non-linear and non-stationary data as texture images. The BEMD technique is a fully data driven method, it does not use any pre-determine filter or wavelet functions. For MRI segmentation clustering is one of the most usable or utilizable technique, where it classifies pixels into classes, without knowing previous information or training. It classifies pixels with highest probability into the same class. It may find unclassified pixels which do not belong to any class probability. Clustering techniques training is done by using pixel features with properties of each class [16]. This paper has compared two image decomposition techniques i.e. wavelet and BEMD to determine the best technique which is used before the image clustering technique FCM. For that purpose, first apply FCM on decomposition image then select the best segmented technique from the two segmented approach through the value of SNR. The paper is organized as follows. In Section II proposed method is described. In the next section experimental result is explained. The paper is concluded in Section IV. II. PROPOSED METHOD The proposed method is an efficient approach to segment the noisy MRI brain images. Two major stages are involved in proposed methodology i.e. feature extraction and clustering. Feature extraction process is performed by using 2D wavelet decomposition and BEMD. The wavelet decomposition outputs are low pass that is approximation component and high pass that is detailed components at horizontal, vertical, and diagonal. To obtain the wavelet features, here Dubechies-1(DAUB1) wavelet is applied to the images. Feature extraction from wavelet decomposition and BEMD are given to Fuzzy C-Means (FCM), FCM applied on the feature vector obtained from previous step for clustering. The output image will be segmented into two classes i.e. white matter and gray matter MRI images. Then calculate the SNR for segmented outputs. Finally, provide the output which gives the best SNR for MR images. The workflow of the proposed method is as shown in Fig. 1. Fig. 1. Workflow of the proposed method. A. Wavelet Decomposition Although the fourier transform has been mainstay of transform-based image processing, a more recent transformation, called the wavelet transform, is now making it even easier to compress, transmit and analyze MR images. Characteristics of Wavelet Transforms (WT) families share properties of their basis functions, primarily the finite support for the frequency and original domains, as well as the scalability. Wavelet families include Haar, Daubechies, Symlets, Coiflets, Biorthogonal, and Reverse biorthogonal, whereas in this paper Daubechies family is rather used. This family is one of the most used wavelet families in image and signal processing applications such as compression, de-noising, classification, and segmentation. The input image, wavelet decomposition and wavelet reconstruction is as shown in Fig. 2(a), Fig. 2 (b) and Fig. 2(c) respectively. (a) (b) (c) Fig. 2. a) Input MRI, b) Wavelet decomposition MRI, and c) wavelet reconstructed MRI. B. Bidimensional Empirical Mode Decomposition(BEMD) Empirical mode decomposition (EMD), originally developed by Huang et al. [17], is a data driven signal processing algorithm that has been established to be able to perfectly analyze adaptive, nonlinear and nonstationary data by obtaining local features and timefrequency distribution of the data. The first step of this method is decomposes the data/signal into its characteristic intrinsic mode functions (IMFs), while the second step finds the time frequency distribution of the data from each IMF by utilizing the concepts of Hilbert transform and instantaneous frequency. The complete process is also known as the Hilbert-Huang transform (HHT) [17]. This decomposition technique has also been extended to analyze two-dimensional (2D) data/images, which is known as bidimensional EMD (BEMD). In this paper BEMD has been used to de-noising and decomposition for MR images based on IMFs weighted threshold. After the image decomposed by BEMD method, it was concerned that the images noise mainly distributed in the high and intermediate frequency. These frequency components obtained from the image by applying an algorithm called sifting process. The sifting procedure decomposes a sampled signal by means of the EMD. The sifting procedure is based on two constraints. Firstly, each IMF has the same number of zero crossings and extrema. Secondly, each IMF is symmetric with respect to the local mean. Furthermore, it assumes that has at least two extrema. Finally, reconstruct the original image to achieve the effect of de-noising. During MR segmentation, it was concerned that the smoothness of the data will influence the segmented quality. The processing example of the BEMD on MRI image is as shown in Fig. 3. In this processing example considered the 48 iteration for IMF function to get the best response. Fig.3. (a) Input MRI (b) IMF (1,...,6,,12,,20,,48),, residue image (c) reconstructed Image 83

85 C. Fuzzy-C Means Fuzzy C-means (FCM) technique is one of the unsupervised clustering techniques used in image segmentation. FCM idea depends on clustering data into two or more classes only by known number of classes. Algorithm is based on minimization of the following objective function: ( ) (1) Here, u is between 0 and 1; C i is the centroids of cluster I; d ij is the Euclidean distance between i th cancroids and j th data point; m [1, ] is a weighting function. Fuzzy portioning of known data sample is carried out through an iterative optimization of the objective function: ( ) This iteration will stop when ( ) (2) (3) { ( ) ( ) } (4) where, is a termination criterion between 0 and 1, k is the iteration steps. This procedure converges to a local minimum or a saddle point of J m. Fig. 7. Processing example of noisy MRI image for sample 1 (a) original image (b) segmented image (c) Gray Matter (d) White Matter. Fig. 8. Processing example of smooth MRI image for sample 2 (a) original image (b) segmented image (c) Gray Matter (d) White Matter. (a) (b) (c) Fig. 4. a) Input MRI, b) segmented image using Wavelet & FCM, and c) segmented image using BEMD & FCM. III. EXPERIMENTAL RESULT The proposed method is implemented in MATLAB environment and tested on MRI brain web database. The brain web images are simulated MR images generated by the Brain Web simulator with different level of noise 0%, 1%, 3%, 5%, 7%, 9% and with different level of INU 0%, 20% and 40%. These images are obtained from Brain Web Database at the McConnell Brain Imaging Centre of the Montreal Neurological Institute, McGill University. An example of images is as shown in Fig. 5(a), 5(b) and 5(c). Fig. 9. Processing example of noisy MRI image for sample 2 (a) original image (b) segmented image (c) Gray Matter (d) White Matter. (a) (b) (c) Fig. 5. (a), (b) and (c) are T1 simulated brain web images. Fig. 10. Processing example of smooth MRI image for sample 3 (a) original image (b) segmented image (c) Gray Matter (d) White Matter. Fig. 6. Processing example of smooth MRI image for sample 1:(a) original image (b) segmented image (c) Gray Matter (d) White Matter. Fig. 11. Processing example of noisy MRI image for sample 3 (a) original image (b) segmented image (c) Gray Matter (d) White Matter 84

86 Fig. 6, Fig. 8 and Fig. 10 shows the processing example of sample smooth MRI images applying wavelet and BEMD with FCM clustering. Fig. 7, Fig. 9 and Fig. 11 shows the processing example of sample noisy MRI images applying wavelet and BEMD with FCM clustering. The evaluation of the segmentation performance in this paper is measured by the Eq. (1). ( ) ( ) (1) where ( ) the input is image and ( ) is the segmented image. Sample MRI image TABLE 1: SNR OF THE INPUT IMAGES Input MRI with SNR Smooth & Adding Wavelet & Gaussian noise FCM BEMD & FCM Smooth mean= 0 and variance = Smooth mean=0 and variance = Smooth mean= 0 and variance = The SNR value of image sample 1, sample 2 and sample 3 for smooth and adding Gaussian noise is as shown in Table I both for Wavelet and BEMD decomposition method and FCM. The SNR of the segmented image generated from BEMD and FCM gives the higher value both for smooth and noisy MR Image. So, it should be noted that BEMD decomposition approach is better than the wavelet algorithm for the use of decomposition before applying the segmentation algorithm such as FCM. IV. CONCLUSION In this paper, presents a robust and efficient approach for the segmentation of medical MR images. For that, wavelet and BEMD decomposition method have been applied in MR images to extract the features. After decompositions, the FCM is use for clustering the image data i.e. generated the segmented image. Then, the best approach is selected by calculating the SNR value. The appropriate approach has been found robust against various medical MR images. The experiments with synthetic Brain Web images have demonstrated the efficiency and robustness of the appropriate approach in segmenting smooth and noisy medical MR images. In future, this works will be improved by increasing overall segmentation performance using another version of FCM for different types of noisy (white noise) MR images. REFERENCES [1] H. Shamsi, H. Seyedarabi, & S. Erfani, MRI image segmentation based on new fuzzy c-means algorithm, International Journal on Computer Science and Engineering, 3(8), 3115,2011. [2] A. Kouhi, H. Seydarabi, A. Aghagolzadeh, A Modified FCM Algorithm for MRI Brain Image Segmentation. Machine Vision and Image Processing (MVIP), 2011 (7th Iranian Digital Object Identifier: /IranianMVIP , pp [3] G. B. Aboutanos, J. Nikanne, N. Watkins and B. Dawant, Model Creation and Deformation for the Automatic Segmentation of the Brain in MR Images. IEEE Transactions on Biomedical Engineering, 46(11) [4] D.A Karras and B.G. Mertzios. On Edge Detection in Mri Using the Wavelet Transform and Unsupervised Neural Networks, EC-VIP-MC th EURASIP Conference focused on Video I Image Processing and Multimedia Communications, 2-5 July 2003, Zagreb, Croatia. [5] L. Gui, R. Lisowski, T. Faundez, P.S. Huppi, F. Lazeyras and M. Kocher. Automatic Segmentation of Newborn Brain Mri Using Mathematical morphology /11/$ IEEE. [6] J. Tohka, I. D. Dinov, D.W. Shattuck, A.W. Toga. Brain MRI tissue classification based on local Markov random fields Magnetic Resonance Imaging, Volume 28, Issue 4, May 2010, pp [7] B.S. Anami, P.H. Unki. A combined fuzzy and level sets based approach for brain MRI image segmentation Computer Vision, Pattern Recognition, Image Processing and Graphics (NCVPRIPG), 2013 Fourth National Conference on Digital Object Identifier: /NCVPRIPG pp [8] P.M. Birgani, M. Ashtiyani and S. Asadi, MRI Segmentation Using Fuzzy C-means Clustering Algorithm Basis Neural Network, In Proc. ICTTA rd International Conference on Information and Communication Technologies: From Theory to Applications, pp: 1-5,2008. [9] K. Xiao, S. H. Ho and A. Bargiela, Automatic Brain MRI Segmentation Scheme Based on Feature Weighting Factors Selection on Fuzzy C-means Clustering Algorithms with Gaussian Smoothing, International Journal of Computational Intelligence in Bioinformatics and Systems Biology, 1 (3): , [10] J. Wang,, J. Kong, Y. Lu, M. Qi and B. Zhang, A Modified FCM algorithm for MRI Brain Image Segmentation Using Both Local and Nonlocal Spatial Constraints, Computerized Medical Imaging and Graphics, 32 (8): , [11] Ardizzone, E., R. Pirrone and O.G. Orazio, Fuzzy C-Means Segmentation on Brain MR Slices Corrupted by RF- Inhomogeneity, In Proc. the 7 th international workshop on Fuzzy Logic and Applications: Applications of Fuzzy Sets Theory (Berlin, Heidelberg, 2007), WILF 07, SpringerVerlag, pp: , [12] M.G. Mostafa, M.F. Tolba, T.F. Gharib and M.A. Mageed, Medical Image Segmentation Using Wavelet Based Mutiresolution EM Algorithm. [13] I. Khalifa, A. Youssif & H. Youssry, MRI brain image segmentation based on wavelet and FCM algorithm. International Journal of Computer Applications, 47(16), [14] A. Linderhed, Image empirical mode decomposition: A new tool for image processing, Advances in Adaptive Data Analysis, 1(02), , [15] J. C. Nunes, Y. Bouaoune,E. Delechelle, O. Niang, & P. Bunel, Image analysis by bidimensional empirical mode decomposition, Image and vision computing, 21(12), , [16] S. Luo, "Automated Medical Image Segmentation Using a New Deformable Surface Model," IJCSNS International Journal of Computer Science and Network Security, vol. 6, pp , [17] N. E. Huang, The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis, Proc. R. Soc. Lond. A 454 (1998), pp

87 Wear and Morphological Behavior of Electron Beam Dose Irradiated Polyoxymethylene Copolymer (POM-C) Md. Shahinur Rahman, Heon- Ju Lee, Konstantin Lyakhov Department of Nuclear and Energy Engineering Jeju National University Jeju , South Korea Lamia Sultana Department of Chemical and Biological Engineering Jeju National University Jeju , South Korea Muhammad Athar Uddin, Muhammad Sifatul Alam Chowdhury Department of Electrical and Electronic Engineering International Islamic University Chittagong Chittagong, Bangladesh Abstract Polyoxymethylene copolymer (POM-C) is the most prominent engineering thermoplastic consisting of repeating carbon-oxygen bonds in the form of oxymethylene group (OCH 2 ). It is widely used to make small gear wheels, ball bearings, precision parts, automotive and consumer electronics. In this study, the POM-C round blocks were irradiated with 165 KeV electron beam energy in five doses (100, 200, 300, 500 and 700 kgy) in vacuum condition at room temperature. The wear and morphological properties of electron beam dose irradiated POM-C have been analyzed using optical microscopy, Raman spectroscopy, 3D nano surface profiler and scanning electron microscopy (SEM). The electron beam dose irradiation transferred the wear of unirradiated POM- C sample from the abrasive wear (plough and cracks), adhesive wear (grooving/striation, micropitting) and scraping to mild scraping and striation for the 100 kgy dose irradiated POM-C sample due to cross-linking (macroscopic networks), chemical free radicals formations and partial physical modification (smoothness), which can be concluded from optical microscopic, SEM and Raman spectroscopic observations. The level of tribological (wear and morphology) attribute improvement relies on the electron beam irradiation condition (energy and dose rate) depending on chemical and physical factors of polymeric materials. Keywords Electron beam; Optical microscopy; Raman spectroscopy; Surface roughness; SEM; Wear I. INTRODUCTION Polyoxymethylene copolymer (POM-C) is a superior engineering thermoplastic for its own outstanding chemical, physical, self lubricating and mechanical properties [1]. However, poor wear resistance and severe crack formation are real drawbacks of POM-C in applications of high tech areas. Radiation induced surface modification of polymeric materials is a prominent and cutting edge new method to solve this intricate problem [2-5]. Electron beam irradiation on pure polymeric material results in the formation of three-dimensionally bonded networks through the chemical free radicals generation [2]. The generated free radicals can initiate complex reactions which can lead to cross-linking or chain scission of polymer chains. As a result, radiation changes the chemical, physical, electrical, mechanical, optical and morphological properties of irradiated polymeric materials. The radiation initiated change can occur in crystalline and amorphous region of polymeric materials through cross-linking and chain scission depending on chemical and physical factors of irradiated materials [6-8]. Recently electron beam irradiation method is becoming an optimistic approach to optimize the physical, mechanical and chemical properties of polymeric materials. Basically energetic electrons intercept the polymeric surface materials and, it can change in molecular structural arrangement through ionization, atomic displacement, carbonization and free radicals generation [2, 8]. The EB dose radiation not only changes the chemical and physical factors of polymeric materials, rather it can also increase the trapped charges or make some defects in the polymeric materials. The EB dose irradiation initiated well suited cross-linking can develop the physical and morphological properties of POM-C in molecular scale, which can reduce the wear and enhance the morphological properties in optimum level. The optimum improvement of wear loss in industrial applications of POM-C can save huge amount of energy and GDP of any highly industrialized nation. The general objective of this study is to reduce the abrasive and adhesive wear with good correlation among chemical, physical and mechanical properties at optimum EB irradiation condition. II. MTHODOLOGY The POM-C (1 mm thickness and 25 mm diameter) used was supplied from Korea Engineering Plastics Co. (Seoul, Korea). The density g /cm 3 (ISO 1183), melt flow rate - 3 g/10 min and melting point C are the typical properties of used POM-C blocks. EB irradiation was conducted under vacuum condition with an ELV-12 (coreless DC and 86

88 induction coupling) electron accelerator at a voltage of 165 KeV in KAPRA (Gangwon-do, South Korea). The general diagram of ELV-12 EB accelerator is shown in Fig. 1. The POM-C blocks (numbered as a- f) were irradiated to doses of 100, 200, 300, 500 and 700 kgy, respectively. The EB current was 300 ma and the dose rate was 25 kgy per pass. The worn surfaces morphology (abrasive and adhesive wear scars with crack formation) at nanoindentation oriented sliding tracks of all unirradiated and irradiated POM-C blocks was observed with an Olympus BX51M computerized HR digital optical microscope. The surface morphology and chemical structure were observed using scanning electron microscope and Raman spectroscopy, respectively. The surface roughness of all unirradiated and irradiated POM-C blocks was measured using 3D nano surface profiler (WT- 250) at 20x objective. formation perpendicular to sliding track are also noticed shown in Fig. 2-a. In Fig. 2-b, the worn surface of 100 kgy EB dose irradiated POM-C appears smooth with shallower and milder scuffing. The carbonization, free radicals formation and stack of polymer deformation debris along sliding track were observed, which were also ascertained from SEM [2-11]. In Fig. 2 (c-f), abrasive wear (plough and surface crack), polymer deformation, grooving, scuffing, flaws, huge fatigue cracks and ribbon tracks were observed from kgy EB dose irradiated POM-C worn surfaces. From the Optical microscopic, SEM and 3D Nano surface profiler observations, it can be concluded that 1 MeV, 100 kgy EB dose irradiation transferred the wear behavior of POM-C block from the abrasive wear, adhesive wear and scraping for the unirradiated surface to shallow and mild scraping for the irradiated surface due to well suited cross-linking, partial physical degradation, carbonization and formation of a small molecular substance [12-14]. So, the 100 kgy EB dose irradiated POM-C block surface treated as self- lubricating surface during dry sliding condition. Fig. 1. General diagram of ELV-12 (KAPRA, South Korea). III. RESULTS AND DISCUSSION A. Optical microscopic image analysis The Optical microscopic images at x20 resolution in Fig. 2 (a-f) show the worn surfaces to analyze the wear mechanism at sliding tracks on POM-C blocks with and without EB dose irradiation. The worn surface of unirradiated POM-C block experienced huge plastic ruffles and ripples flow with polymer deformation along the sliding track. Shrouded with grooving, scuffing, friction traces and crack Fig. 2. Optical microscopic images of the worn surfaces at sliding tracks on POM-C surfaces with x20 resolution: (a) unirradiated POM-C; (b) 100 kgy; (c) 200 kgy; (d) 300 kgy; (e) 500 kgy and (f) 700 kgy electron beam dose irradiated surfaces. B. SEM analysis The EB dose irradiated and unirradiated POM-C samples were studied using scanning electron microscope to get better insights about surface morphology. Changes of surface morphology in different EB dose irradiation are noticed in Fig. 3 (a - f). In Fig. 3-a, some surface tracks and unevenness are observed on unirradiated POM-C sample. The 100 kgy dose irradiated sample is smoother and fine compared to all unirradiated and irradiated surface shown in Fig. 3-b, which is due to well suited cross- 87

89 linking and physical interactions between energetic electrons-target material atoms collisions [7 14]. A threshold value of cross-linking can initiate free radicals formation on the top surface by the absorption of radiated electron beam energy, which can easily improve the mechanical, morphological and chemical properties depending on chemical and physical factors of irradiated polymeric surface [2, 7 9]. The threshold value of cross-linking mostly depends on the polymeric surface EB dose delivered. Cracking and ribbon traces are observed from 200 and 300 kgy EB dose irradiated POM-C surfaces shown in Fig. 3 (c-d). These cracking and ribbon traces can initiate through chain scission and, it can damage the polymeric surface through gas evolution, breaking polymeric molecular bonds and displacing the atomic arrangement. Severe cracking and high unevenness are observed in 500 and 700 kgy EB dose irradiated POM-C surfaces, which may be due to severe chain scission and gas evolution [7 9]. According to aforementioned observations, the 100 kgy dose irradiation is well suited to get better morphology oriented surface. vibrational peak assignments: 538 cm -1 (δ, C-O-C), 917 cm -1 (υ g, COC), 1095 cm -1 (υ a, C-O-C), 1333 cm -1 (t, CH 2 ) and 1489 cm -1 (δ, CH 2 ) shown in Fig. 4-a. The 100 kgy EB dose irradiation highly increased the vibrational band strength and peak intensity of C-O-C symmetrical and asymmetrical stretching at 921 and 1104 cm -1, respectively. It also strengthens the vibrational band strength of CH 2 bending at 1486 cm - 1. These increased vibrational bands strength can protect form different chemical structural degradation due to well suited cross-linking shown in Fig. 4-b [16-17]. The other EB doses irradiation reduced the vibrational bands strength and peak intensity of all fundamental vibrational peak assignments gradually shown in Fig. 4 (c - f) due to increasing chain scission. So, 100 kgy dose irradiation is well suited for the occurrence of cross-linking, which can give desired chemical structural modification. Fig. 4. Raman spectra demonstrating the chemical structural modification due to electron beam dose irradiation on POM-C surfaces: (a) unirradiated; (b) 100 kgy; (c) 200 kgy; (d) 300 kgy; (e) 500 kgy and (f) 700 kgy. Fig. 3. SEM images at 20 KeV and 1000x magnification: (a) unirradiated; (b) 100 kgy; (c) 200 kgy; (d) 300 kgy; (e) 500 kgy and (f) 700 kgy EB dose irradiated POM-C samples. Scale bar 500 µm in each case. C. Chemical structure analysis The chemical structural modification analysis due to EB dose irradiation was studied using Raman spectroscopy (LABRAM HR EV) shown in Fig. 4 (af). Raman spectroscopy is also a viable tool to study about rotational and vibrational transitions in atomic level, which can give good insights for the degree of cross-linking as non-destructive technique [15-17]. The unirradiated POM-C sample gave the following D. Surface roughness analysis EB induced surface roughness changes are shown in Fig. 5 (a - f). The average surface roughnesses (Ra) of unirradiated and 100 kgy EB dose irradiated POM-C surfaces are nm and nm, respectively. Significant surface roughness reduction is noticed by 100 kgy EB dose irradiation, which can be the reason of energetic electrons- target atoms well controlled collisions and threshold value of cross-linking [2, 6 8]. The threshold value of cross-linking can reduce surface roughness, when aggregation and atomic displacement are suppressed at desired level. In Fig. 5 (c -f), the 200, 300, 500 and 700 kgy EB dose irradiation increased the R a to , , and nm, respectively. The gradually increase of R a might be the reason of gradually increase chain scission depending on surface delivered EB dose [16-17]. 88

90 Fig. 5. 3D surface profiler images on POM-C samples at 20x objective: (a) unirradiated; (b) 100 kgy; (c) 200 kgy; (d) 300 kgy; (e) 500 kgy and (f) 700 kgy. IV. CONCLUSIONS From the Optical microscopic, SEM, Raman spectroscopic and 3D nano surface profiler observations, it can be concluded that 165 KeV, 100 kgy EB dose irradiation transferred the wear behavior of POM-C block from the abrasive wear, adhesive wear and scraping for the unirradiated surface to shallow and mild scraping for the irradiated surface due to well suited cross-linking, partial physical degradation, carbonization and formation of a small molecular substance. It also reduced the average surface roughness with higher vibrational bands strength, which is ascertained from 3D nano surface profiler and Raman spectroscopic observations. Moreover, the 100 kgy EB dose irradiated POM-C surface play as a self-lubricating surface due to chemical free radicals formation. After all, the electron beam irradiation condition (energy and dose) is the main parameter to reduce wear loss depending on chemical and physical factors of POM-C surface. REFERENCES [1] Duan, Y., Li, H., Ye, L., & Liu, X. (2006). Study on the thermal degradation of polyoxymethylene by thermogravimetry Fourier transform infrared spectroscopy (TG FTIR). Journal of applied polymer science, 99(6), [2] Raghu, S., Kilarkaje, S., Sanjeev, G., Nagaraja, G. K., & Devendrappa, H. (2014). Effect of electron beam irradiation on polymer electrolytes: Change in morphology, crystallinity, dielectric constant and AC conductivity with dose.radiation Physics and Chemistry, 98, [3] Turos, A., Jagielski, J., Piątkowska, A., Bieliński, D., Ślusarski, L., & Madi, N. K. (2003). Ion beam modification of surface properties of polyethylene.vacuum, 70(2), [4] Kofanova, O. A., Mommaerts, K., & Betsou, F. (2015). Tube Polypropylene: A Neglected Critical Parameter for Protein Adsorption During Biospecimen Storage. Biopreservation and biobanking, 13(4), [5] Ramola, R. C., Chandra, S., Negi, A., Rana, J. M. S., Annapoorni, S., Sonkawade, R. G.,... & Srivastava, A. (2009). Study of optical band gap, carbonaceous clusters and structuring in CR-39 and PET polymers irradiated by 100MeV O 7+ ions. Physica B: Condensed Matter, 404(1), [6] Reichmanis, E., Frank, C. W., & O'Donnell, J. H. (1993). Irradiation of polymeric materials(processes, mechanisms, and applications). In A. C. S. symposium series. American Chemical Society. [7] O'Donnell, J. H., & Reichmanis, E. (1989). The effects of radiation on high-technology polymers. American Chemical Society. [8] Gopal, R., Zuwei, M., Kaur, S., & Ramakrishna, S. (2007). Surface modification and application of functionalized polymer nanofibers. InMolecular Building Blocks for Nanotechnology (pp ). Springer New York. [9] Lapin, S. C., & Davenport, I. A. Modification of Polymer Substrates Using Electron Beam-Induced Graft Copolymerization. [10] Gupta, M. C., & Deshmukh, V. G. (1983). Radiation effects on poly (lactic acid). Polymer, 24(7), [11] Deepalaxmi, R., & Rajini, V. (2014). Gamma and electron beam irradiation effects on SiR-EPDM blends. Journal of Radiation Research and Applied Sciences, 7(3), [12] Tian, N., Li, T., Liu, X., & Liu, W. (2001). Effect of radiation on the friction wear properties of polyetherketone with a cardo group. Journal of applied polymer science, 82(4), [13] Pei, X., & Wang, Q. (2007). Effect of electron radiation on the tribological properties of polyimide. Tribology transactions, 50(2), [14] Murray, K. A., Kennedy, J. E., McEvoy, B., Vrain, O., Ryan, D., Cowman, R., & Higginbo, C. L. (2013). Characterisation of the Surface and Structural Properties of Gamma Ray and Electron Beam Irradiated Low Density Polyethylene. International Journal of Material Science. [15] Nguyen, T. T., Gobinet, C., Feru, J., Pasco, S. B., Manfait, M., & Piot, O. (2012). Characterization of type I and IV collagens by Raman microspectroscopy: Identification of spectral markers of the dermo-epidermal junction. Journal of Spectroscopy, 27(5-6), [16] Hirschl, C., Biebl Rydlo, M., DeBiasio, M., Mühleisen, W., Neumaier, L., Scherf, W.,... & Kraft, M. (2013). Determining the degree of crosslinking of ethylene vinyl acetate photovoltaic module encapsulants A comparative study. Solar Energy Materials and Solar Cells, 116, [17] Vašková, H., Maňas, D., Ovsík, M., Maňas, M., & Staněk, M. (2013). Microhardness of polyamide 12 after crosslinking due to beta radiation.international Journal of Mathematical Models and Methods in Applied Sciences. 89

91 Study of Structural and Optical Properties of Pyrolised CuO Films M. Majhar Department of Materials Science and Engineering Rajshahi University Rajshahi -6205, Bangladesh. Abstract Cupric oxide (CuO) thin films were fabricated onto glass substrate at temperature 350 C using spray pyrolysis technique. The source material was the monohydrate cupric acetate, [(CH 3 COO) 2 Cu. H 2 O] with different (0.1M, 0.2M and 0.3 M) molar concentrations. Structural, surface morphology and optical properties of deposited films were studied by XRD, AFM and UV-VIS spectrophotometer. The films are monophasic polycrystalline in nature. The preferred orientation of the deposited films was found to be (111) plane and the crystal structure were identified as monoclinic. The films were found to be a good absorber of visible EM radiation. The optical absorbance is ~96% in the range ( ) nm after which it decreases with increasing wave length. The optical band gap of the deposited film was found to be dependent on film thickness which varies from ev. Keywords Spray pyrolysis technique, Cupric oxide, Thin film, and Band gap energy. INTRODUCTION Transparent conducting semiconductor oxide thin film exhibits numerous properties which attract a lot of research interest because of the variety of applications such as light transparent electrodes, solar cells, thin film photovoltaic and many other optoelectronic devices [1-4]. Cupric oxide (CuO) thin film has been reported to exhibit p-type semiconductor having monoclinic crystal structure. CuO is an important metal oxide semiconductor has band gap in the range of ev [5] which depends on the grown method along with film thickness. There are various methods to grow CuO thin films. These are spray pyrolysis technique [6], radio frequency magnetron sputtering [7], molecular beam epitaxy [8], sol-gel [9], electrodeposition [10], metalorganic chemical vapor deposition [11], chemical bath method [12], etc. Among these methods, spray pyrolysis offers numerous advantages such as low cost, easy to handle, less harmful and nano-structured film preparations. In this paper, we report the preparation and structural, electro-optical characteristic of CuO thin films for the utilization in electro-optical devices. EXPERIMENTAL DETAIL The CuO thin films were prepared by spray pyrolysis technique. An aqueous solution of 0.1 M, S. Ahmed, M. Mozibur Rahmn and M. K. R. Khan * Department of Physics Rajshahi University Rajshahi-6205, Bangladesh *fkrkhan@yahoo.co.uk 0.2M and 0.3 M [(CH 3 COO) 2 Cu.H 2 O] were used as precursor. The solution was pumped into the air stream in the spray nozzle at a rate of 0.33 ml/min by means of a syringe pump. The CuO film was deposited on glass substrate at substrate temperature ~350ºC.The constant distance between the tip of the nozzle and upper surface of the substrate was 21 cm. The temperature of the glass substrate was controlled using a copper-constantan thermocouple. The possible chemical reaction that takes place on the heated glass substrate to form CuO films may be as follows: (CH 3 COO) 2 Cu.H 2 O + H 2 O Heat CuO+CO 2 +CH 4 +steam Newton s rings method was used to measure the thickness of the films. The x-ray diffraction study was performed by using Cukα monochromatic radiation of wavelength, λ= nm. For morphological investigations, AFM images were recorded by noncontact mode atomic force microscopy using NCHRtip (AFM-MODEL: XE70 PARK SYSTEMS). The UV-visible spectrophotometer (UV-1601 PC SHIJMADZU) was used to obtain optical data. RESULT AND DISCUSSION The XRD patterns of the deposited films are shown in Fig.1 which confirms that the CuO films are polycrystalline in nature and could be indexed with monoclinic structure. The characteristic peaks were identified from (JCPDS) card and these are at 2θ = , , and corresponding (hkl) values (111 - ), (111), and (020) respectively. All peak positions are compared with the standard values of cupric oxide (from JCPDS card). No peak corresponding to Cu 2 O phase has appeared in the XRD pattern that indicates the formation of pure phase of CuO films. The 2D and 3D AFM images of films of thicknesses 215 nm and 170 nm are shown in Figs.2 and 3, respectively. It is seen that the surface morphology and surface roughness changed with the variation of molar concentration. The roughness of film thickness 215 nm is nm while in roughness of film thickness 170 nm is o.449 nm. 90

92 Fig.1: XRD patterns of CuO films with different molar concentrations. Fig. 2(b): The 3D AFM image of 215 nm CuO film. Fig. 2(a): Surface roughness with 2D image for 215 nm CuO film by AFM. Fig. 3(a): Surface roughness with 2D image for 170 nm CuO film by AFM. 91

93 Transmittance, T (%) Absorption Coefficient, cm -1 ) x 10 7 Reflectance, R (%) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering 9 8 t =110 nm t =170 nm t = 235 nm Wavelength, (nm) Fig.5: Variation of Reflectance with wavelength. film. Fig. 3(b): 3D AFM image of 170 nm CuO 5 4 t=110nm t=170nm t=215nm The optical transmittance and reflectance as a function of wavelength of incident radiation for CuO films are shown in Fig.4 and Fig.5. The optical absorbance is ~96% in the range ( ) nm after which it decreases with increasing wave length. Variation of absorption coefficient with photon energy is shown in Fig.6. A plot of (αhν) 2 vs. hν for CuO films of different thicknesses are shown in Fig.7. The direct band gap of CuO films have been obtained from the intercepts on the energy axis after extrapolation of the straight-line of (αhν) 2 vs. hν curves. Depending on the film thickness (110 nm, 170 nm and 215 nm) of the CuO films, the optical band gap of the deposited film was found to be 1.72 ev for film thickness 170 nm, 1.72 ev for 215 nm and 1.80 ev for 110 nm respectively Photon energy,hev) Fig. 6: Variation of absorption coefficient with photon energy for CuO thin film t=110 nm t=170 nm t=215 nm h cm -1 ev) 2 x t=110 nm t=170 nm t=215 nm Photon energy, h(ev) Wavelength, (nm) Fig. 7: Variation of (αhν) 2 with photon energy for CuO films for allowed direct transition. Fig. 4: Variation of Transmittance with wavelength. 92

94 CONCLUSION A crystalline CuO film is successfully prepared by simple SP method. The surface roughness of the film changes with the film thickness. The direct band gap of the films was found to vary with film thickness. ACKNOWLEDGEMENTS Authors gratefully acknowledged Bangladesh Atomic energy Commission (AEC) for providing XRD; Dept. of Physics, University of Rajshahi for UV-visible spectrophotometer; Central Science Lab., University of Rajshahi for AFM. REFERENCES [1] K.T. Ramakrishna Reddy, G.M. Shanthini, d. Johnston, R. w. Miles, Thin Solid Films 427(2003) [2] S. Ilican, Y. Caglar, M. Kundakci, A. Ates. Int. J. Hydrog, Energy 34(2009) [3] C.H. Champness, Z. Xu, Appl. Surf. Sci., (1998) [4] R.A. Ismai,O.A. Abdulrazaq, Sol. Energy Mater. Sol. Cells 91(2007) [5] Balamurugan, B. & Mehta, B.R. 2001,Thin Solid Films 396(1-2) : [6] D. Sivalingam, J.B. Gopalakrishnan, J.B.B. Rayappan, Mater. Lett. 77(2012) [7] Subramanyan,T.K.; uthanna, S.; Srinivasulu Naidu, B. Physica Scripta vaccum 36(1986) [8] L.Li, Z. zuo, J.H. Lim, J.L. Lin, Appl. Surf. Sci. 256(2010) [9] E. Mosquera, I. Pozo, M. Morel, J. Solid State Chem. 206(2013) [10] M. Tortosa, M. Mollar, B. Mari, J. Cryst. Growth 304(2007) [11] Zhao, Z.; Morel, D.L.; Ferekides, C.S. Thin Solid Films 2002, 413, [12] Z. Liang, R. Gao, Jo-Lin Lan, O. Wiranwetchayan, Qifeng Zhang, Chundong Li, Guozhong Cao, Solar Energy Materials & solar Cells 117(2013)

95 Algorithm for Performance Appraisal using CAW Method M. Z. Ahsan* Physics Dept, Military Institute of Science and Technology Mirpur Cantonment Dhaka-1216, Bangladesh Professor Dr. Mamunur Rashid Khandkar Department of Applied Physics and Electronic Engineering University of Rajshahi Rajshahi, Bangladesh Abstract Performance appraisal is an HR process, which plays important role to drive employees for achieving organizational goal. Simple Additive Weighing (SAW) method is popularly used as a tool to calculate individual performance score (PS). Reflective factors (opportunity factor and optimum contribution factor) are not accounted in this method. Consequently, the assessed PS found to be less representative and biased. In the proposed Cumulative Average Weighing (CAW) method, those factors have been taken into consideration to make PS more representative and unbiased. A case studied for 5 years on 3 employees of the same status. Overall Performance Index (OPI) and Aptitude Index (API) have been calculated using performance score (PS) obtained in both the methods by Computer Based Performance Appraisal System (CB-PAS) software, developed in VB. The analysis, using statistical tools (SD, MAD and AD) reflects that the PS as calculated by CAW method is more representative than that of calculated by the SAW method. Rate of change of API as calculated from the aptitude score facilitate the organization to talent management. Besides, Graphical Model for Score Interpretation (GMSI) used as an alternative tool for screening out and selecting the best option using data obtained from the CB-PAS. Keywords Performance Score (PS), Overall Performance Index (OPI), Aptitude Index (API), Reflective factors. I. INTRODUCTION Performance Appraisal is an HR process, which critically linked with selection, retention, promotion, layoffs, compensation, utilization of talent, and training. As such, it plays key role in human resource management to drive employees for keeping them aligned with the organizational goal and desired outcome. Accordingly, a variety of tools has already been developed, and researches are still being continued in varying context of organizations. Till to date, Simple Additive Weighting (SAW) method is being popularly used as a tool to calculate individual performance in many organizations. But in this method, reflective factors (opportunity factor and optimum contribution factor) are not accounted. Accordingly, SAW method found to be less representative because of biasing effect due to personal liking-disliking, lack in the level flat condition related to opportunity, unexpected gap in assessed scores due to varying assessors and perspectives, inherent lack in taking cognizance of individual contribution over the years, and finally lack in scope of utilizing individual expertise related to job specifications. To address the issues for making the PS more representative and unbiased, a new mathematical model, called Cumulative Average Weighing Method (CAW) proposed, where reflective factors in accounted. Three variables (aptitude, attitude and ability) have been postulated to derive PS formula. Each variable evaluated by weighted values of corresponding set of attributes using well known 9-point likert scale. This formula realized by software, named Computer Based Personal Appraisal System (CB-PAS), which has been developed in VB at the front and MySql at the back end in bottom up approach as shown its tree algorithm. It has the options to calculate PS by both the methods. A case has been studied to supplement the proposition. The findings are analyzed using statistical tools (SD, MAD and AD). The CAW method found to be more representative due to accounted reflective factors in PS calculation. Hence, the objective of this paper is to introducing algorithm of the CAW method, comparison with SAW method and GMSI approach for screening out and selecting the best option. II. MATHEMATICAL APPROACH A. Postulates The postulates to develop the mathematical model of CAW method are given below: a. Aptitude, attitude and ability are variables to determine performance score. These have been considered as independent variables for the purpose, though they are complexly related to each other. b. Each variable has a set of attributes i, which will be weighted using 9 point Likert Scale. c. Each variable is 1 - degree function of attributes, when they are weighted. d. Performance Score (PS) over a year is the summation of the averages of these variables. e. Opportunity factor has been calculated based on logical assessment and thereby accounted in aptitude variable by translation ii as to bring balance in effects. f. Optimum Contribution iii factor has been accounted by translation as cumulative effect in annual assessment i.e. in the assessment of 94

96 current year to remove affects of over or under assessment. g. Performance score of current year will be additive as following in the SAW method. h. Performance Score, at quiescent year will be the average of last year performance score, (is called here optimum contribution factor) and performance score of current year, which is termed here as cumulative average iv. j. Performance appraisal will have to be done by Team Based Appraisal (TAB) technique. k. Arithmetic Mean (AM) is used for determination of performance indices. l. GMSI approach used for trend analysis of individual performances over the years. Trend lines may also be used to determine performance indices as an alternative to verify correctness. B. Formula In CAW method, the formula, derived for performance score calculation based on the postulates, is given in the sequence below: The formula for the quiescent year [ ] (1) Where, = The performance score of current year. = The performance score at quiescent year. = Performance score of last year for taking account of optimum contribution of individual employee. The formula for the performance score of the current year (excluding optimum contribution) (2) Where, = Overall average of aptitude score (opportunity factor integrated), done by j assessors. = Overall average of attitude score, done by j assessors. = Overall average of ability score, done by j assessors. C. Performance Indices There are two significant parameters namely Overall Performance Index (OPI) and Aptitude index (API) in Performance Indices. They can be used as tools to select candidates or employees for appointment, promotion and even for retention. OPI and API are mathematically defined here as the arithmetic mean (AM) of series of respective scores, assessed over the years. As such their mathematical formula will be as: and [where m is the number of assessed years.] Notable here, that they may also be determined by the value at Q point v (as shown in figure-1) on the respective trend line from the graph in GMSI approach as an alternative method. These two parameters will give true and unbiased reflection of performance on individual scores. Standard Deviation vi (SD) may be used to verify their representativeness as and when required. However, OPI will be compared with the expected standard of the position as set by the organization to determine the range of selection for the purpose and thereby it may be termed as screening factor. API, on the other hand, will be used as tool to select the best option in the selected range and therefore, may be termed as selection factor. Besides, the slop of trend line, drawn for aptitude variable, will give the rate of change in aptitude of the subject employee over the years, which facilitates to have insight of talent management vii. Similarly, trend lines for other variables, will help to identify the area of weakness of the subject employee. Moreover Absolute Deviation (AD) as calculated by the formula may help to pin point the individual lapses in performance as well as error occurred during assessment. III. ALGORITHMIC APPROACH Algorithm is common language for nature, human and computer. It replaces mathematical model in a logical sequence to solve any problem that to a computational model. Let us now proceed to develop algorithm for score calculation in sequence as described below: A. Computational Model In order to derive computational model, let us consider the attributes of aptitude, attitude and ability are a i, b i and c i respectively where, i indicate the number of attributes. All these attributes will be weighted using 9 point Likert Scale by the assessors. Now, if j is the number of assessors for performance appraisal using Team Based Appraisal (TBA) technique. Then the computation model for evaluating of each variable will be as in the following sequence: In 1 st step: Weighting attributes by j assessors. - ix attributes for aptitude. - i x attributes for attitude. - i x attributes for ability In 2 nd step: Averaging of attributes over i attributes. - i x attributes for aptitude. - i x attributes for aptitude. - i x attributes for aptitude. In 3 rd step: Averaging of attributes over j assessors. - j x assessors. - j x assessors. - j x assessors. In 4 th step: Averaging of each variable over current year. appraisal. - n x no. of 95

97 - n x no. of appraisal. - n x no. of appraisal. Within close limit taking Opportunity factor. Where, is the opportunity factor as determined from the logical weighting (1 or 0) to the attributes of opportunity. Here 1 means yes and 0 means No. Accordingly, this is defined as the ratio of number of 1s to the total number of attributes, k, postulated for opportunity. Mathematically, it can be expressed as:. This is one of the reflective factors, which has been brought in the above formula to ensure level flat conditions and perspectives in the working environment for the employee under assessed. In 5 th step: Averaging of variables for final score of current year. ( ) - 3 x variables (aptitude, attitude and ability). In Final step: Cumulated average. -This is the performance score for the quiescent year. Note that it will be taken as the performance score of last year for subsequent year s assessment. B. Tree Algorithm Now, to convert this computational model to tree algorithm, let us consider an arbitrary case. Suppose, an employee is going to be assessed by a committee of 4 members and 9 in number attributes for each variable is chosen as convenience though this number may differ from organization to organization. Likert scale has been used to weighting these attributes. Hence the tree algorithm to show sequential steps in bottom-up approach will be as: attributes (10 x personal traits and 25 x demonstrated traits) have been weighted using 9- point likert scale against each employee over 05 (2010 to 2014) years. Individual PS has been calculated by both CAW and SAW methods using CB-PAS software, which has been developed basing on the above algorithm. The analysis, using statistical tools (SD, MAD and AD) reflects that the PS as calculated by CAW method is more representative than that of calculated by the SAW method. The findings, based on SD and MAD, are presented in table-1. Absolute deviation (AD) has also been used here to find the difference between consecutive PS in order to ascertain its correctness and to have the insight of causes if such deviation (+ or -) greater than. Because, representativeness in value of PS refers to the open limit {AD: } that to confirm correctness in assessment, yielded it from the concept of expressing any error in %. But its value pin points to individual lapses as like for X in the year of 2010 and 2012 as shown in table-2, which helps to focus subsequent corrective measures. IV. DISCUSSION A. Representativeness A case has been studied in small scale for verifying representativeness of PS and to supplement the proposition in this paper. In that three (X, Y and Z) employees of same position and having almost same length of service are considered to evaluate PS at the quiescent year. 35 B. Score Interpretation In performance appraisal, Graphical Model for Score Interpretation (GMSI) may be used to visualize performance trend of employees over the years for making decision on retention, promotion, requirement of training and talent management. This GMSI is just a graph or bar chart over years fitted with trend lines as shown in figure-1(using data from the case study). These trend lines lead to the following interpretations: a. If the trend line goes up that means performance of the subject employee increases over the times. So, he is still having capability to render service and effective as well. The observations of associated trend lines may determine the state of aptitude and attitude, 96

98 and also to find components having more effects on the performance score. This fact will help to ascertain talents and contribute to the talent management. b. If the trend line goes down that means performance of the subject employee decreases over the times. So his capability is in question. Close observation will enable to find the domain of weakness of the subject employee. other two attributes as well as set policy with other tracing factors (if any). The slope (fist derivative with respect to time if the line is non-linear) of trend lines for API and OPI will measure corresponding rate of changes. These rates will vary from employee to employee. So, the slope for API may be used as tools for talent assessment and ranking. The formula and technique for slope determination is described here using GMSI for API only as shown in figure-3: c. If the trend line be constant over reasonable times, then the subject employee may be retained but may not have chance to go up position or promotion. Now, the formula for slope determination will be as: However, these interpretations should be compared with the expected Standard for the position. Notable here that the expected standard of performance indices for each position may be determined either by corresponding average of performance indices (API and OPI) over assessed employees at the quiescent year [The formula is as: and, where w is the number of assessed employees of the same status] or by promulgated guideline basing on job description suited to achieve the organizational goal. A bar chart of API and OPI along with corresponding set standard of performance indices is presented in figure-2 (using data from the case study). It leads the selection committee to screen out the non prospective candidates or employees in one hand; on the other hand, comparing API of the prospective candidates or employees helps to select the best option using their judgment focusing on Slope for API (rate of change of aptitude) = per year. A comparison of this figure (0.055) may be used to identify the employee having potentiality to drive the organization to meet the challenges in the days to come. Similarly, the slope for OPI (not shown) may lead to sort out employees who need counseling and training to increase their standards. By thus, these slopes may contribute to the talent management. C. Significances The significances of CAW method are therefore as under: a. The proposed model as well as formula for PS calculation is more representative. But its accuracy in performance reflection mostly depends on identified variables and attributes, suited to the subject organization. b. Opportunity of working situation and contributions to the organization of individual over the years, by rendering services, has been taken into account as reflective factors. c. The maintenance of level flat condition in performance score, done by translation of opportunity factor, in aptitude score, within close limit 0 1. d. The contributions rendered by the individual to the organization over the years have been taken in effect in the performance score, by translation of optimum contribution factor, (last year s PS). e. Overall Performance Index, OPI and Aptitude Index, API, determined from the arithmetic mean (AM), ensures unbiased (balance between under and 97

99 over assessment) performance of individual at the quiescent year. f. Moreover, it sets limits not to deviate much from the range of selection in case of promotions, appointments and retentions as applicable. g. The slop of trend line will give the rate of change in aptitude per year, which will be useful in talent management for the organization. h. Close observations and analysis of trend lines will enable the concern to find the area of weakness of the subject employee and thus be able to determine whether training may compensate these lapses or weakness in his performance or he or she should not be retained. V. CONCLUSION The analysis to develop this mathematical model for performance appraisal is done on sample data from a case study. The variables and attributes are postulated from logical thoughts out of experiences, and studying relevant documents. The formula for performance score, calculation is based on simple arithmetical averaging concept and Simple Additive Weighting (SAW) method. Opportunity factor and optimum contribution factor have been considered here as reflective factors basing on observations and practical experiences. These factors have been brought in performance score calculation formula and also taken into account by translation in SAW method using Taylor s philosophy. This mathematical model is proposed here as Cumulative Average Weighting (CAW) method. Notable here that the PS, calculated by this method is more representative than that of SAW method as revealed from the SD and MAD measurements as stated above. Besides AD measurement facilitates to find the errors or lapses in performance sore for subsequent corrective measures. However, there are scopes for testing the said model taking more real data by surveying any suitable organization. Moreover, the postulated variables and corresponding attributes may not be same for all categories, rather depend on the perspectives and objectives of the organizations. As such, flexibility of choosing variables and attributes remains on hands of the organization. They can choose variables and attributes by job analysis and also set standard for each position in the hierarchy of the organization. Accordingly, the accuracy in performance score and organizational success mostly depends on these variables and attributes. Besides, there are rooms to put thoughts regarding optimum and opportunity factors to take in account for performance appraisal. Thereby, the concerns are encouraged to do more research in these aspects. ii Translation is one type transformation of function manipulation iii The immediate last year performance score is accounted for optimum contribution factor. iv The term Cumulative is used here to take effect of previous value to the newly calculated value in average. v The value of performance indices, determined from trend lines may vary from that of AM at quiescent year, but it ensures removal of errors at each year to help the assessors. vi SD is a statistical parameter to measure the dispersion from the mean or average value. Its 0 value indicates 100% accurate, which is not practically possible. But its value less than 1 indicates that the determined mean or average value is in agreement and acceptable. So the mean or average value will be representative as much as SD s value closer to 0. vii A new concept of HR functions. References [1] Anisseh, M., J. Dodangeh, et al. (2007). 360 Degree Personnel Performance Appraisal Using the MADM Models and Presenting a Model for overall Ranking. IEEE International Conference on Industrial Engineering and Engineering Management. [2] Brutus, S. (2010). Words versus numbers: A theoretical exploration of giving and receiving narrative comments in performance appraisal. HumanResource Management Review, 20: [3] Chang, D. Y. (1996). Applications of the extent analysis method on fuzzy AHP. European Journal of Operational Research, 95(3): [4] Chenhall, R. H. and K. Langfield-Smith (2007). Multiple Perspectives of Performance Measures. European Management Journal 25(4): [5] D Andres. R.J.L-Lapresta, et al. (2010a). Performance appraisal based on distance function methods. European Journal of Operational Research, 207: [6] D Andres. R.J.L -Lapresta, et al. (2010b). Multigranular linguistic performance appraisal model. Soft Computing, 14: [7] Ferris, G. R., T. P. Munyon, et al. (2008). The performance evaluation context: Social, emotional, cognitive, political and relationship components. Human Resource Management Review, 18: [8] Fisher, C., L. F. Schoenfeldt, et al. (2006). Human Resource Management. Boston, Houghton Mifflin. [9] Fletcher, C. (2001). Performance appraisal and management: The developing research agenda. Journal of Occupational and OrganizationalPsychology, 74: [10] Hwang, C. L. and K. Yoon (1981). Multiple attribute decision methods and applications. Berlin, Heidelberg, Springer. [11] Jones, J. E. and W. E. Bearley (1996). 360-Degree Feedback: strategies, tactics, and techniques for developing leaders, Human Resource Development Press. Note: CB-PAS Design concept is shown at the end. End Notes: i Attributes may vary from organization to organization, which will have to be determined by the concern beforehand. 98

100 Friction and Morphological Properties of Ion Implanted Polyoxymethylene Copolymer (POM-C) Md. Shahinur Rahman, Konstantin Lyakhov Department of Nuclear and Energy Engineering Jeju National University Jeju, South Korea Muhammad Athar Uddin, Muhammad Sifatul Alam Chowdhury Department of Electrical and Electronic Engineering International Islamic University Chittagong Chittagong, Bangladesh Lamia Sultana Department Of Chemical and Biological Engineering Jeju National University Jeju, South Korea Md. Mehedi Hasan Division of Energy Systems Research Ajou University Ajou, South Korea Abstract Polyoxymethylene copolymer (POM-C) round block was implanted with 120 KeV ions of He to doses of 5 x and 1 x ions cm -2. It was also implanted with 120 KeV ions of Ar + He and He + Ne to dose of 1 x ions cm -2, respectively. The friction behavior of both implanted and unimplanted POM-C blocks was investigated using a pin on disk tribometer against steel ball. The morphological properties of ion beam implanted POM-C blocks have been characterized by scanning electron microscopy (SEM), 3D nano surface profiler and Raman spectroscopic methods. The friction coefficient of He ion implantation at a dose of 5 x ions cm -2 is lower compared to other ion doses implanted POM-C blocks and, it is also lower than the unimplanted one. It also shows the moderate surface texturing (atomic rearrangement), lower surface roughness and good chemical structural behavior compared to both unimplanted and other ion doses implanted POM-C blocks due to some extent of carbonization, cross-linking and ions-target atoms collisions, which is ascertained from SEM, Raman spectroscopic and 3D nano surface profiler observations. The other ion doses implanted POM-C blocks demonstrate the higher friction coefficient and surface roughness with polymer surface deformation (crazing, cracking, pitting and gas evolution, bond breaking) due to severe chain scission, surface dose delivered atomic displacements and chemical structural degradation. It is concluded that the variation of friction coefficient behavior of POM-C blocks resulted from its structural response for ion beam implantation on the surface. In contrast, further investigation is required to get the best tribological attributes at optimum ion beam surface dose delivered. Keywords POM-C; SEM; Surface roughness; Friction coefficient; Raman spectroscopy I. INTRODUCTION Polymers have been widely used in many sophisticated fields due to low density, mold ability, low manufacturing cost, and good mechanical and electrical / electronic properties. Polyoxymethylene copolymer (POM-C) is an attractive copolymer to make small gear wheels, bearings, seals, precision parts, automotive and consumer electronics [1-3]. However, its use in extreme operating condition is sometimes limited due to its undesired properties of the surface (friction, wear and chemical resistivity). Therefore, it s very crucial issue to modify the surface mechanical and morphological properties in a controlled manner to expand POM-C applications in low and high tech areas with good processibilities. Ion implantation is an established powerful method for modification the physical and chemical properties of polymers up to several micrometers thickness [4-6]. This established method operates at low temperature in vacuum condition without any thermal degradation of pure material. It s a very effective technique in surface modification due to its higher cross section for prompt ionization and larger linear energy transfer capability. Ion implantation can easily improve the surface mechanical properties (friction, hardness, surface roughness and wettability) through precise crosslinking. Cross-linking can easily initiate strong chemically bonded networks, rigidity of backbone frame, anchoring joints for the molecular chain and restraining of atomic displacement [1-7]. Moreover, the judicious choice of ion implantation condition can give precise modification of surface properties depending upon chemical and physical factors of target polymeric material. The key objective of this study is to reduce the friction coefficient phenomena up to desired level through ion beam induced physical and chemical properties modifications at molecular scale. II. METHODOLOGY The POM-C (Natural) blocks used were supplied from DYNEX engineering plastic company (South Korea). The dimension of the POM-C samples was d 30 mm x 10 mm. The samples were ground and polished to the roughness Ra = 0.38 µm before ion 99

101 implantation. The friction coefficient was measured on a J&L tech tribometer using a ball (SUJ 52100) on disk method at linear speed 100 mm/s. The load was 1 N and the sliding distance was 100 mm over a period of 16 minutes. The morphology of all unimplanted and ions doses implanted POM-C blocks was observed using scanning electron microscopy (SEM) at 500x magnification. The surface roughness was measured using 3D nano surface profiler at 20x objective. The chemical structural analysis was analyzed using Raman spectroscopy. The ion implantation was performed on the KE04-01 (Cockcroft-Walton) low energy ion implanter with the energy of 120 KeV. The ion beam current was 2 µa. The following ions doses conditions were selected during for our study: bonds) at the time of energy transfer from an incident ion to target polymeric atom shown in Fig. 2-b [1,7 9]. Other ions doses implanted POM-C surfaces demonstrate the huge crazing, cracking, hillocks, deformation and atomic displacements due to energy transfer in binary nuclear collisions and huge gas evolution shown in Fig. 2 (c - e) [1, 6]. (a) He ions to doses of 5x1016 and 1x1016 ions cm-2 (b) Ar+He and He+Ne to doses of 1x1016 ions cm-2. III. RESULTS AND DISCUSSION A. Friction co-efficient In this study, friction coefficient reduction at intricate operating condition is taken into consideration, where conventional lubricating technique is not applicable. The friction coefficient phenomena for all unimplanted and ions doses implanted POM-C blocks is shown in Fig. 1 (a -e). The friction coefficient of He ion implantation at a dose of 5 x ions cm -2 is lowest among all unimplanted and implanted POM- C blocks shown in Fig. 1-b. It might be the reason of well suited cross-linking, chemical free radicals formation and physical structural modification (atomic rearrangement due to collisions between ions and target atoms of exposed polymer) [1, 4 9]. The threshold value of cross-linking can easily initiate free radicals formation on the top surface of target material, which can play role as solid lubricant. This formed solid lubricant layer can reduce the friction coefficient during dry sliding condition. In Fig. 1 (c - e), other ions doses implantation increased friction coefficient due to huge chain scission, surface dose delivered atomic displacements and polymeric bond breakage depending on physical and chemical factors [6 9]. So, He ion implantation at a dose of 5 x ions cm -2 is well suited to reduce friction coefficient depending on ion bombardment initiated chain reactions. B. SEM analysis Morphological analysis of different ions beam doses implanted and unimplanted POM-C blocks was investigated by SEM technique at 5 KeV accelerating voltage and 500x magnification. The unimplanted POM-C demonstrates the scuffing, ribbon tracks, waves and flaws on the surface shown in Fig. 2-a. The He ion implantation at a dose of 5 x ions cm - 2 makes the surface smooth with mild scuffing and ribbon tracks due to elastic nuclear collisions and degassing (carbon ratio enhancement in polymeric Fig. 1. Friction coefficient of POM-C blocks against SUJ ball for : (a) unimplanted, (b) He 5 x 10 16, (c) He 1 x 10 16, (d) Ar + He 1 x and (e) He + Ne 1 x ions cm -2 implanted surfaces (load : 1 N, sliding distance: 100 m). Fig. 2. SEM images on POM-C surfaces with x500 resolution: (a) unimplanted, (b) He 5 x 10 16, (c) He 1 x 10 16, (d) Ar + He 1 x and (e) He + Ne 1 x ions cm -2 implanted surfaces. Scale bar 10 µm in each case. C. Chemical structure Raman spectroscopy is a well-established nondestructive powerful technique to analyze rotational and vibrational transitions in molecular scale for solid surface, which can determine the degree of crosslinking. The chemical structural modification induced by ion beam implantation on POM-C blocks was observed using Raman spectroscopy shown in Fig. 3 (a-e). In Fig. 3-a, the unimplanted POM-C polymer block demonstrates the following peak vibrational assignments: 994 cm-1 (r, CH2), 1109 cm-1 (υa, C- O-C), 1328 cm-1 (t, CH2) and 1484 cm-1 (δ, CH2). In Fig. 3-b, the He ion implantation at a dose of 5 x 1016 ions cm-2 significantly increased the peak intensity and vibrational band strength of 100

102 asymmetrical stretching at wave numbers 1109 cm-1 due to suitable cross-linking, polymerization and chemical free radicals formation in polymeric molecular bonds, which can prevent from any kinds of molecular degradation [7, 8]. In Fig. 3 (c - e), other ions doses implantation broke the polymeric bonds with new types of bond evolution, which can be the reason of chain scission and ion induced inelastic collisions between ions and target atoms [7 9]. Fig. 4. 3D nano surface profiler images on POM-C blocks at 20x objective: (a) unimplanted, (b) He 5 x 10 16, (c) He 1 x 10 16, (d) Ar + He 1 x and (e) He + Ne 1 x ions cm -2 implanted surfaces. Fig. 3. Raman spectra demonstrating the effects of ion beam implantation on POM-C blocks: (a) unimplanted, (b) He 5 x 10 16, (c) He 1 x 10 16, (d) Ar + He 1 x and (e) He + Ne 1 x ions cm -2 implanted surfaces. D. Surface roughness Ion bombardment initiated surface roughness changes are demonstrated in Fig. 4 (a e). In Fig. 4 (a - b), the average surface roughness (Ra) of unimplanted POM-C block was nm and, the Ra of He ion implanted POM-C block at a dose of 5 x 1016 ions cm-2 was nm. The He ion implantation at a dose of 5 x 1016 ions cm-2 reduced the average surface roughness significantly, which might be the reason of threshold value of cross-linking and well controlled physical sputtering effects, when atomic displacements and aggregation are suppressed [1-9]. The threshold value of cross-linking and controlling of atomic displacements depend on chemical and physical factors of ion bombarded polymeric material. In Fig. 4 (c - e), other ions implanted doses increase the surface roughness to 0.78, 1.00 and 1.25 µm, respectively. This might be the reason of huge atomic sputtering and electronic excitation effects (inelastic scattering) [8 10]. So, He ion implantation at a dose of 5 x 1016 ions cm-2 is better than other ions doses implantation to reduce surface roughness due to well suited elastic scattering and cross-linking effects CONCLUSION From aforementioned systematic study of tribometer testing, Raman spectroscopic, SEM and 3D nano profiler observations, it is concluded that the He ion implantation at a dose of 5 x ions cm -2 reduced friction coefficient, surface roughness and atomic displacements significantly due to the well suited degassing, 3 dimensionally formed macroscopic networks, free radicals formation and elastic scattering. After all, friction coefficient reduction depends on ion fluence condition, chemical and physical factors of target material. REFERENCES [1] Popok VN, Ion implantation of polymers: formation of nanoparticulate materials, Reviews on Advanced Materials Science, 30(1), 1-26, [2] Turos A, Jagielski J, Piątkowska A, Bieliński D, Ślusarski, Madi NK, Ion beam modification of surface properties of polyethylene,vacuum, 70(2), , [3] Ramola RC., Chandra S., Negi A., Rana JMS., Annapoorn S., Sonkawade RG., A. Study of optical band gap, carbonaceous clusters and structuring in CR-39 and PET polymers irradiated by 100MeV O 7+ ions, Physica B: Condensed Matter, 404(1), 26-30, [4] Jagielski J., Friction properties of ion-beam modified materials: Where can we search for practical applications of ion implantation?, Vacuum, 78(2), , [5] Calcagno L., Compagnini G., Foti G., Structural modification of polymer films by ion irradiation, Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, 65(1-4), , [6] Cleland M. R., Parks L. A., Medium and high-energy electron beam radiation processing equipment for commercial applications, Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, 208, 74-89, [7] Reichmanis E., Frank C W., O'Donnell JH, Irradiation of polymeric materials(processes, mechanisms, and applications), In ACS. symposium series, American Chemical Society, PP 1-8,

103 [8] O Donnell JH, Radiation Chemistry of Polymers, In: High Energy Chem, ACS, pp 1 13, [9] Gopal R, Zuwei M, Kaur S, Ramakrishna S., Surface Modification and Application of Functionalized Polymer Nanofibers, In: Mol. Build. Blocks Nanotechnol,Springer New York, New York, NY, pp 72 91, [10] Stephen C. Lapin, Modification of Polymer Substrates using Electron Beam Induced Graft Copolymerization, Radtech Conf.,

104 Fabrication and Mechanical Characterization of Aluminium- Rice Husk Ash Composite by Stir Casting Method Adnan Adib Ahamed*, Rashed Ahmed, Md. Benzir Hossain, Masum Billah Department of Materials and Metallurgical Engineering, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh Abstract Metal matrix composites constitute an important category of design and weight-efficient materials. Composites containing low density and low cost reinforcements has produced an increasing interest. Rice husk ash (RHA) is an agricultural byproduct containing large amount of silica. In our present work, an attempt is made to fabricate RHA particle reinforced Aluminium matrix composite. RHA particles were incorporated into the matrix melt by stir casting by a mechanical stirrer. Magnesium (~1%) is used as a wetting agent. RHA was added in 3, 6 and 9 wt. % to the matrix. The microstructural features of the composites and particle distribution were observed by means of scanning electron microscope. The mechanical properties such as density, tensile strength, percentage elongation and hardness were studied. The results reveal that with increasing reinforcement, the density of the composites decreased, while the ultimate tensile strength and hardness of the composites increased compared to the unreinforced aluminium. Keywords MMC, Aluminium, RHA, Stir casting, Mechanical properties. I. INTRODUCTION Composites are materials made from two or more constituent materials of significant properties, both physical and chemical that, when combined, produce a material with a whole new characteristics better than the individual characteristics of each constituents. However, within the composite the different materials can be identified easily as they do not dissolve or blend into each other. A typical composite has two parts: a strong, stiff reinforcement, that provides the strength and rigidity which is distributed in a second material, called the matrix, which serves to bind and protect the reinforcements [1]. Composites can be classified as dispersion strengthened, fiber reinforced and particulate reinforced composites based on the reinforcement type and as metal matrix, ceramic matrix and polymer matrix composites based on the matrix type [2]. Metal Matrix Composites (MMCs) are composites where metals or metallic alloys are used as matrix and ceramics or other materials are used as reinforcements. The ductility of the metal matrix with the stiffness and rigidity of the ceramics reinforcements are used which makes MMCs fascinating. The MMCs are lightweight and resist wear and thermal distortion better than their metal counterparts, so it is mainly used in the automobile industry since here the emphasis is on weight reduction and engine efficiency to improve fuel economy [3]. Aluminum is popular as a matrix due to its high strength to weight ratio, ductility and malleability, machinability, durability. Rice husk, a low cost and low density material, exhibits superior physical and mechanical properties. Thus RHA can be utilized more effectively in the development of composite materials for various applications [4]. Presence of silica is an additional advantage in comparison to other byproduct materials which makes RHA an important material for a wide range of manufacturing and application oriented processes [5] Stir casting technique is currently the most common and cost effective commercial method for composite making. Mechanical mixing of the reinforcement particulate into a molten metal bath is the basic principle of this technique. The molten metal matrix is stirred vigorously to form a vortex at the surface of the melt and the reinforcement material is introduced at the side of the vortex. The stirring is continued for a few minutes before the slurry is cast. During stir casting for the synthesis of composites, stirring helps in transferring particles into the liquid metal and maintaining the particles in a state of suspension [6]. Different research works has been conducted on the utilization of RHA as particulate reinforcements in different Al alloy matrix to prepare composites. A study investigating the effects of 2, 4, 6 and 8% wt. RHA particulate reinforcement in A356.2 alloy produced by stir casting technique showing the reduction in density from about 2.7gm/cm 3 to 2.55 gm/cm 3 while increasing the hardness from around 65 to 82 BHN [7]. The trends in the rate of increase in tensile strength, hardness and the rate of decrease in density was similar in the case of AlSi10Mg alloy reinforced with up to 12% RHA particles and this study showed that the rate of increasing the tensile strength dropped when the percentage of reinforcement was increased to around 9% and then the tensile strength began to drop with further increasing the amount of reinforcement [8]. Up to 15% RHA was incorporated in Al2.8Si0.8Fe alloy to obtain increased tensile strength (115MPa- 103

105 126MPa) and hardness (55BHN-80BHN) while decreasing density (2.8gm/cm 3-2.6gm/cm 3 ), the higher percentage of RHA retention being due to the employment of double stir casting [9]. An attempt to make a hybrid composite of Al (6061) alloy, RHA and Cu by stir casting technique retained RHA particles up to 32% by weight of matrix material with 3% Cu and the hardness of the composites were measured to increase with increasing reinforcement concentration [10]. In this work, Rice Husk Ash (RHA) was used as the reinforcement. The main idea was to utilize the agro-industrial waste to reduce its negative environmental impact and at the same time to develop a composite material economically with superior properties. A. Material Selection II. EXPERIMENTAL PROCEDURES The matrix material used in this study was commercially pure aluminium of 99.3% purity and the reinforcement material was rice husk ash prepared from the rice husks collected from Shibganj, Chapainawabganj, Bangladesh. Magnesium at about 1% of the weight of the matrix was used as the wetting agent for the better adhesion of the reinforcements with the matrix. B. Rice Husk Ash Preparation The rice husks were washed thoroughly with water to remove the dusts and dried at room temperature for 24 hours. They were then burnt at open atmosphere to remove the moisture and organic constituents. The color of the husk changed from yellowish to black at this stage due to charring of the organic matter. This was followed by heating the husks at 650 o C in a Blue M furnace for two hours to remove the carbonaceous constituents leaving the grayish white silica rich ash to be used as the reinforcement of the composite. The ash was subjected to chemical analysis by adding drops of HF and heating a 0.1gm of RHA so that the silica content is evaporated as SiF 4 and weighing the residue afterwards to determine the percentage of silica content in the ash. C. Composite Making The composite was made by stir casting method. Initially 1.2kg of Al was charged into graphite crucible and melted to 810 o C in a pit furnace. As the metal melted about 15gm of magnesium was added to the melt as a wetting agent between matrix and reinforcement since it reduces the casting fluidity as well as the surface tension of the molten Al [11]. The molten Al was then transferred to the holding furnace of the stir casting apparatus. RHA particles were preheated to 700 o C for 1 hour so that the silica content remains amorphous below 800 o C. The graphite stirrer was lowered into the melt and stirring was commenced at 500rpm to create a vortex. The preheated RHA (~ µm) was added slowly into the melt at the side of the vortex to ensure a uniform distribution. The stirring was continued for another 7-8 minutes after adding the reinforcement so that the particle distribution and the mixing is done properly. The mixture was poured at a temperature of 732 o C in to a mild steel permanent mold that was preheated earlier to obtain uniform solidification of the casting. In this process 3, 6 and 9% RHA reinforced Al matrix composites were prepared. III. RESULTS AND DISCUSSIONS A. Rice Husk Ash Analysis 948gm of rice husk ash was obtained after burning which was about 21% of the 4.5kg rice husk. About 27% ignition loss was calculated after furnace heating of the ash to obtain silica rich final reinforcement materials. The silica content of the RHA was obtained at 90% by means of chemical analysis with HF. Upon sieving the RHA particles, the particles in the size range of μm were taken to reinforce the Al matrix. B. Microstructural Features The microstructure of the unreinforced matrix and the reinforced composites are shown in Fig. 1. From the SEM micrographs it can be seen that RHA particles were well distributed in the matrix for 3% and 6% by weight reinforcements but was not as satisfactorily for the 9% wt. RHA reinforcement. The particles did not retain their original shape due to the vigorous stirring during mixing and during handling prior to mixing. The chemical analysis was performed on each of the 3%, 6% and 9% reinforced composite samples. The analysis revealed increasing presence of Silicon materials which were due to the addition of RHA which confirmed the new phases in the micrograph to be SiO 2 rich RHA. C. Density Measurement The density was measured by measuring mass of the prepared samples and employing the Archimedean principle by immersing them into a labelled measuring cylinder partially filled with water hence calculating the displacement of the water level indicating the volume of the sample. The variation in the densities of the base metal and the reinforced composites are displayed in Fig. 2. It is found that with increasing weight percentage of RHA, the density of the composites decreased. This is due to the fact that amorphous silica that constitutes most of the reinforcement, has much lower density than aluminium. Some porosity were also apparent from the microstructures which arose during casting of the composites and they also attributed to the reduction of density. 104

106 Fig. 2. Density of pure Al, Al+3%RHA, Al+6% RHA and Al+9% RHA Fig. 1. SEM micrographs of (a) pure Al, (b) Al-3% RHA, (c) Al- 6% RHA and (d) Al-9% RHA composites D. Tensile Strength The tensile properties of the composites were measured by preparing tensile samples of pure aluminium and the reinforced composites according to ASTM E8/E8M-09 sub-size standard. The variation in the yield strength and ultimate tensile strength from pure aluminium to 3, 6 and 9% RHA reinforced composites increased gradually as shown in Fig. 3 and Fig. 4. Both the yield and ultimate tensile strength of the composites increased with increasing amount of reinforcement due to the presence of stiffer RHA particles which produced mismatch in the matrix reinforcement interface to impede the motion of dislocations. The strength of the Al-9% RHA reinforced composite was the highest but the rate of increase of strength was significantly smaller than the previous composites which was due to nonhomogeneous distribution of the reinforcing particles into the matrix as was apparent from the micrograph. E. Hardness Brinell hardness was measured for the base metal and the composites. 4.9KN load was applied by using a steel ball of diameter 10mm. The hardness of the composites was increased in comparison to the unreinforced Al and the increase was gradual with increasing percentage of reinforcement. The variation in the Brinell hardness number of the composites is shown in Fig. 5. The reinforcing particles, being harder and stiffer than the matrix, resisted the plastic deformation caused by the indenter and contributed to the increasing hardness of the composites with increasing reinforcement percentage. Fig. 3. Yield strength of pure Al, Al+3% RHA, Al+6% RHA and Al+9% RHA Fig. 4. Ultimate tensile strength of pure Al, Al+3% RHA, Al+6% RHA and Al+9% RHA 105

107 application of materials with lighter weight but superior strength. ACKNOWLEDGMENT The authors would like to acknowledge the support of Department of Materials and Metallurgical Engineering, BUET for the experimental supports. Fig. 5. Brinell hardness of pure Al, Al+3% RHA, Al+6% RHA and Al+9% RHA To sum up, the variation in density, yield strength, ultimate tensile strength and hardness from pure Al to composites with 3%, 6% and 9% RHA reinforcements can be tabulated as in Table 1. The results found show that the yield strength, ultimate tensile strength and hardness of the composites increased gradually with increasing amount of RHA reinforcement into the matrix while decreasing the density. These are in consistency with different research on the utilization of RHA particles in Al alloy matrix [7,8,9,10]. TABLE I. VARIATION IN DENSITY, YIELD STRENGTH, ULTIMATE TENSILE STRENGTH AND BRINELL HARDNESS OF AL, AL+3%RHA, AL+6%RHA AND AL+9%RHA COMPOSITES Samples Density (gm/cm 3 ) Yield Strength, MPa Properties Ultimate Tensile Strength, MPa Brinell Hardness, BHN Al Al+3%RHA Al+6%RHA Al+9%RHA IV. CONCLUSION From our work we can conclude that RHA can be successfully incorporated into pure aluminium matrix for the production of composites. This can also solve the problem of storage and disposal of RHA and utilization of an agricultural waste. Addition of up to 9% by weight RHA to aluminium was done by stir casting route to produce composites. Addition of magnesium improved the wettability of RHA with aluminium melt and thus increased the retention of the RHA in the composite. Hardness of the composites is increased from 22BHN to 33 BHN with addition of RHA and magnesium compared to unreinforced condition. The Ultimate tensile strength and the yield strength has improved with increase in RHA content while decreasing the density ensuring the potential REFERENCES [1] K. Y. Lin, Composite Materials: Materials, Manufacturing, Analysis, Design and Repair, William E. Boeing Department of Aeronautics and Astronautics, University of Washington, 2014, pp.1-4. [2] D. R. Askeland, P. P. Fulay, and W. J. Wright, The Science and Engineering of Materials, 6th ed., Cengage Learning Inc., 2010, pp [3] D. Hull, and T. W. Clyne, An Introduction to Composite Materials, Cambridge University Press, [4] R. Keshavamurthy, V. Prabhu, S. Pai, V. Kumar, and H.N. Vinay, Development and characterization of industrial waste reinforced metal matrix composite, Project REF. No. 37S0401, Dayananda Sagar College of Engineering, Bangalore. [5] A. Kumar, K. Mohanta, D. Kumar, and Om Prakash, Properties and industrial applications of rice husk: a review, International Journal of Emerging Technology and Advanced Engineering, vol. 2(10), pp.86-90, [6] A. A. Adebisi, M. A. Maleque, and M. M. Rahman, Metal matrix composite brake rotor: historical development and product life cycle analysis, International Journal of Advanced Mechanical Engineering, vol. 4, pp , [7] S. Prasad, and A. R. Krishna, Fabrication and characterization of A356.2-rice husk ash composite using stir casting technique, International Journal of Engineering Science and Technology, vol. 2(12), pp , [8] S. D. Saravanan, and M. S. Kumar, Effect of mechanical properties on rice husk ash reinforced aluminium alloy (AlSi10Mg) matrix composites, in International Conference on Design and Manufacturing, Procedia Engineering, vol. 64, pp , [9] V. S Aigbodin, Development of Al-Si-Fe/rice husk ash particulate composites synthesis by double stir casting method, USAK University Journal of Material Sciences, vol. 2, pp , [10] A. Mittal, and R.Muni, Fabrication and characterization of mechanical properties of Al-RHA- Cu hybrid metal matrix composites, International Journal of Current Engineering and Technology, vol. 3(5), pp , [11] A. M. Korolkov, Casting Properties of Metals and Alloys, Consultant Bureau, New York, 1963, pp

108 Analysis of Annual and Seasonal Precipitation Concentration Index of North-Western Region of Bangladesh Ahsan Habib Rasel Dept. of Applied Physics & Electronic Engineering, University of Rajshahi, Rajshahi-6205, Bangladesh Md Monirul Islam Dept. of Applied Physics & Electronic Engineering, University of Rajshahi, Rajshahi-6205, Bangladesh Mumnunul Keramat Dept. of Applied Physics & Electronic Engineering, University of Rajshahi, Rajshahi-6205, Bangladesh Abstract--The precipitation concentration index (PCI) of north-western region of Bangladesh is computed. The PCI is estimated on Seasonal and Annual distributions, variations and trends. The seasonal estimation were based on two seasons dry season (November to April) and wet season (May to October). Precipitation Concentration Index is analyzed at annual and seasonal scale to identify the pattern of rainfall in the study area for the period of The contribution of wet seasonal rainfall is more than (90%) and the dry seasonal rainfall is less than (10%) of the study area. In wet season PCI value is 10.4 to 10.9 that mean uniform precipitation distribution and in dry season PCI value varies from 25.0 to 30.7 that mean strongly irregular distribution of rainfall. Annual PCI valuewhich shows irregular distribution of rainfall with values in the range of 17.6 to 18.8.The PCI value at Syedpur, Rangpur, Dinajpur and Bograshows increasedand in the rest of the stations Rajshahi and Ishwardi, the PCI value shows decreasing trend. Key words: Precipitation Concentration Index; Rainfall; Variability. I. Introduction An attempt has been made to determine the trend of rainfall characteristics in the study region. It deals with the analysis of annual and seasonal variation for individual station and in average. Stationbase and distribution of rainfall characteristics have been estimated. Precipitation Concentration Index (PCI) is very useful to evaluate the degree of seasonal concentration of precipitation. It provides information to compare different climates in terms of seasonality of precipitation regime. Precipitation Concentration Index is an index or descriptor of rainfall variability, which in essence means that the index can provide information on water availability within an environment. This information can be used for a wide variety of hydrological, water resources and environmental management program. Furthermore, the index can be used as a warning tool for disaster preparedness in relation to flooding and erosion within the area [1]. The more concentrated is precipitation, the more difficult is water management, irrigation control, soil erosion prevention and agriculture [2]. Similarly, information derived through a good understanding of the spatial and temporal characteristics of rainfall is also very important for agricultural planning, flood frequency analysis, hydrological modelling, water resources assessments, assessing and understanding climate change impacts and other environmental assessments [3]. In order to study the inter-annual aspects of the rainfall distribution, PCI for each station is calculated. II. Materials and Methods A. Study Site North-Western region of Bangladesh is just north to the tropic of cancer. Geography, global air and sea currents, tree cover, global temperatures and other factors influence the climate of an area, which causes the local weather [4]. The climate of the greater Rajshahi division i.e., the north western region of Bangladesh is dominated by tropical monsoons. It is characterized by high temperature, moderate rainfall with often excessive humidity and fairly marked seasonal variations [5]. Figure 1: Location Map of the study area. Bangladesh occupies an area of 143,998 km2 and has a subtropical humid climate. The study regions are Rajshahi and Rangpur divisions located in the north western part of Bangladesh and extends from ' to ' N latitude and from ' to ' E longitude. Figure 1 shows the study area. Study area consists of 16 districts and 128 Upazilas. Except the Barind Tract, most of this region is a low-laying plain land. Six weather stations in all, one from each greater district were selected for this study. The stations are scattered around the division. The area of the region km 2 and the population is 33,99,4000 (preliminary figures) according to census 2011 of which 80% people live in rural area and directly or indirectly depend on agriculture. B. Data The daily rainfall data for twelve years ( ) are collected from Bangladesh Meteorological Department (BMD) of six stations. The nature of data collected is depth in millimeter (mm). For analysis, the daily data were computed to give an total N 107

109 monthly, total annual and stationwise total and average value. The continuity of the daily rainfall records was disrupted by some missing data which was checked carefully and estimated those by averaging from the available data for a particular month, year or station. Usual procedure for estimating the missing records (i.e., weighted average interpolation) was not possible because of single point consideration. C. Methodology In an attempt to define temporal aspects of the rainfall distribution within a year, the Precipitation Concentration Index (PCI) proposed by Oliver (1980) [6] is used. The PCI was estimated on an annual and seasonal scale, the seasons into two categories wet season(may to October) and dry season(november to April). The number 100 in the formula for the annual PCI (Equation-1) represents 12 months of the year signifying 100%. The number 50 in (Equations-2) represent the number of months in each season as a percentage of 12 months of the year. It is expressed as percent (%) in according to the following formula: Where, p = precipitation of i-th month and P = annual precipitation. (1) (2) Where, p = precipitation of i-th month and P = Total Precipitation of these months. Table-1 : Precipitation Concentration Index (PCI) classification by Oliver. PCI value Distribution of precipitation PCI < 10 Uniform precipitation distribution Moderate precipitation concentration Irregular distribution PCI>20 Strongly irregular distribution III. Spatial Distribution of Precipitation Concentration Index(PCI) The Precipitation Concentration Index values for the six stations of Rajshahi and Rangpur division are calculated based on the equations given by Oliver (1980) [6]. The maximum PCI value of 18.8 is obtained in Syedpur and minimum PCI value of 17.6 in Ishwardi. The results are presented in figure-2and compare with table-1 which indicate irregular distribution of rainfall is occurring in all the six stations. The spatial distribution of PCI is shown in figure-3. With reference to the PCI value 2000 its change. The changed of PCI for the year 2005 and 2010 have been estimated and presented. Also the present situation in average in the region has studied. Figure-3(a) shows the PCI distribution of the region. In the south-western the index value is maximum and in the south-eastern corners its minimum. But the distribution has been changed over study period. During the year 2005 the maximum PCI is obtained in the southern region and minimum has changed to the north-eastern corner. This situation is not continued with time. During next five years the distribution pattern has been changed and in the year 2010 (figure-3(d)) it is observed that the PCI has been decreased in the southern region of the study area. These trend is in agree with the average PCI variation over the region as shown in 3(a). From this analysis it can be said that the PCI is decreasing over the region of study Precipitation Concentration Index Station Figure 2: Precipitation Concentration Index (PCI) Values of of the stations. I. Trend of Precipitation Concentration Index Annual and seasonal rainfall variability, the PCI are estimated and presented in table-2. Thus, accordingly, the PCI of rainfall in the study area generally shows irregularity.the table-2 shows that the rainfall is very much seasonal. The contribution of wet seasonal rainfall is more than (90%) and the dry seasonal (November to April) rainfall is less than (10%) for all six stations from table-2. In wet season PCI value is 10.4 to 10.9 that mean uniform precipitation distribution in wet season. In dry season PCI value varies from 25.0 to 30.7 and compare with table-1 it is conclude that strongly irregular distribution of rainfall in dry period. Annual PCI varies from 17.6 to 18.8 indicates irregular distribution of rainfall of the study area. In this case most of the stations show an increase in PCI values, which indicates an increase in the variability of the distribution of monthly rainfall. The PCI value at Syedpur, Rangpur, Dinajpur and Bogra has increased over time shown in table-3.in the rest of the stations Rajshahi and Ishwardi, the PCI value shows decreasing trend. Therefore, an increase in PCI indicates an increase in the seasonality of monthly rainfall. From table-3 wet seasonal PCI ofsyedpur, Dinajpur, Rajshahi and Ishwardi stations are decreasing and rest of the stations Rangpur and Bogra, PCI show increasing trend. The PCI value at Syedpur, Rangpur, Dinajpur and Ishwardi has decreased trend and Bogra and Rajshahi station show increasing trend. 108

110 N (a) (b) (c) (d) Figure 3: Spatial distribution of Precipitation Concentration Index, (a) average, (b) in the year 2000, (c) 2005 and (d) 2010 of study region. Table-3 :Seasonal and annual trends (per year) of PCI of six station. Seasonal PCI trend Annual Station Wet Dry PCI season season trend Syedpur Rangpur Dinajpur Bogra Rajshahi Ishwardi Wet seasona PCI y = x R² = Year Figure 4: Wet seasonal Precipitation Concentration Index (PCI) and its trend of the study area. Dry seasonal PCI y = x R² = Year Figure 5: Dry seasonal Precipitation Concentration Index (PCI) and its trend of the study area. Precipitation Concentration Index Figure 6: Annual Precipitation Concentration Index (PCI) and its trend of the north-western region of Bangladesh. II. Conclusion Hence it can be concluded that the PCI is complex and affect the local precipitating conditions and is also related to the global atmospheric features in the region. The trends and consequently the variability in rainfall both annual and seasonal are analyzed for annual and seasonal data. From these PCI trend analysis two types of observation are observed. In this study area the trend is increasing and decreasing. From the table-2 it is clear that in the northern region the trend is increasing and in the southern region its decreasing. It is also observed that the overall precipitation duration is decreasing. The effect of this trend may have negative effect to sustain the present environmental condition of the study area. References y = x R² = Year [1] Adegun, O., Balogun, I. and Adeaga, O, 2012, Precipitation Concentration Changes in Owerri and Enugu, Special Publication of the Nigerian Association of Hydrological Sciences, [2] Shahid, S. and Khairulmaini, O.S., 2009, Spatio- Temporal variability of Rainfall over Bangladesh 109

111 during the time period , Asia-Pacific Journal of Atmospheric Sciences Vol.45 (3), pp [3] Ngongondo, C., Xu, C.Y., Gottschalk, L., and Alemaw, B. (2011).Evaluation of spatial and temporal characteristics of rainfall in Malawi: a case of data scarce region.theoretical and Applied Climatology, Vol.106, pp [4] Pender, J.S., 2008, What Is Climate Change? And How It Will Effect Bangladesh. Briefing Paper.(3rd Draft).Rajshahi, Bangladesh: Church of Bangladesh Social Development Programme. [5] Rashid, H. E., 1991, Geography of Bangladesh. Dhaka University Press Limited. [6] Oliver, J.E., 1980, Monthly precipitation distribution: a comparative index. Professional Geogr, Vol.32, pp

112 Bitwise Template Fusion of Noisy Images for Enhanced IRIS Recognition System Md. Sohel Ahammed Department of Computer Science & Engineering, Rajshahi University of Engineering & Technology (RUET), Rajshahi-6204, Bangladesh Birprodip Pal Department of Computer Science & Engineering, Rajshahi University of Engineering & Technology (RUET), Rajshahi-6204, Bangladesh Abstract Iris recognition is one of the reliable and accurate biometric identification system available today. However, many results have been published under favorable conditions. The work presented in this paper involves developing an iris recognition system considering noisy data (noisy images). The performance of many existing systems gradually decreases for noisy images. An image can be noisy in different ways like at the time of capturing image, eyelids, poor segmentation or any other ways. The median filter was used for removal. After segmentation and feature encoding, several templates were found for a single person. Finally modified majority voting system (MVS) was used to generate a final template of the same person by considering common feature. Hamming distance was used for the determination of performance of the system. However, trivial hamming distance was able to carry out an accuracy of 94.49%, on the other hand, the proposed approached was able to accomplish an accuracy of 96% for noisy data (noisy images). The performance increases up-to 99.3% for the standard images (noise free image) also. The main feature of our proposed IRIS recognition system was the enhancement of performance for noisy image. Bitwise template based fusion approach helped us to achieve this goal. Keywords Iris Recognition, Noisy image, Template fusion, Majority voting system, Hamming distance. I. INTRODUCTION Biometrics refers to automatic identification of a person on a basis of his or her unique physiological or behavioral characteristics. Behavioral biometrics include signatures, voice recognition, gait measurement, and even keystroke recognition.physiological biometrics include facial recognition, fingerprinting, hand profiling, iris recognition, retinal scanning, and DNA testing. Behavioral methods tend to be less reliable because they are easier to duplicate. Biometric methods based on physiolog-ical attributes are more trusted. Among those method, iris rec-ognition is gaining much attention as an accurate and reliable one. To improve accuracy, most of the biometric authentication systems store multiple templates per user to account for varia-tions in biometric data. Therefore, these systems suffer from storage space and computational overheads. In order to address these issues, there is need to optimize the computational and storage complexities by creating a reliable specimen iris template per user rather than maintaining multiple templates. In this paper we focus on the issues how to create a final template that contains the common property of several templates of a particular person. II. PREVIOUS RESARCH Human iris possesses genetic independence and contains extremely information-rich physical structure and unique texture pattern which makes it highly complex enough to be used as a biometric signature. Statistical analysis reveals that the iris is the most mathematically unique feature of the human body because of the hundreds of degrees of freedom it gives with the ability to accurately measure its texture [1]. Reliable biometric verification and identification techniques based upon iris patterns have been presented by Johnston [2], Daugman [3,4], Wildes et al. [5 7], Boles [8,9]. Other known iris recognition systems have been introduced by Zhu et al. [10], Lim et al. [11], Noh et al. [12], Tisse et al. [13] and Ma et al. [14]. Motivated by these works, several researchers worked on enhancing the performance of iris recognition systems. Some researches focus on improving the image acquisition systems [15,16], some deals with enhancing the segmentation algorithms [17,18], others are devoted to improving the features extraction and encoding process [19]. In biometrics in general, it has been found that using multiple images for enrollment and comparing the probe to multiple gallery samples will result in improved performance.several papers show that this is also true for iris recognition. Du performed experiments using one, two, and three images to enroll a given iris. The resulting recognition rates are 98.5%, 99.5%, and 99.8%, respectively. Liu and Xie presented an algorithm that uses direct linear discriminant analysis. Their results using 1200 images showed that recognition performance increases dramatically in going from two images per iris to four images, and then incrementally from 4 to 8, and 8 to 10. Algorithms that use multiple training samples to enroll an image must decide how to combine the scores from multiple comparisons. Ma et al. suggested analyzing multiple images and keeping the best-quality image. The same author, reported that the average of a three scores is taken as the final matching distance when matching an input feature vector with three templates of a class. Krichen et al. represent each class in the gallery with three images, so that for each person and for each test image, they kept the minimum value of its similarity measure to the three images. The use of the min operation to fuse a set of similarity scores is generally more appropriate. Considering multiple scans of an iris, 111

113 Schmid et al. used the average Hamming distance of multi-sample matching. This is compared to using a log-likelihood ratio, and it is found that, in many cases, the log-likelihood ratio outperforms the average Hamming distance. Hollingsworth et al. acquire multiple iris codes from the same eye and evaluate which bits are the most consistent bits in the iris code. They suggest masking the inconsistent bits in the iris code to improve performance. III. IRIS RECONGITION SYSTEM The iris recognition process consists of five major steps. The first step is the image acquisition of a person s eye at enroll-ment time or check time. The second step is to segment the iris out of the image containing the eye and part of the face, which localizes the iris pattern. Step three is the normalization; here the iris pattern will be extracted and scaled to a predefined size. Step four is the template generation; here the details of the iris are filtered, extracted and represented in an iris code. The last step is the matching phase, where two iris codes will be compared and a similarity score is computed. These steps are shown schematically in Fig. 1. when the integro-differential operator attains its maximum. The iris boundary was de-scribed with three parameters: the radius r, and the coordinates of the center of the circle, x0 and y0. More recently, Daugman proposed alternative segmentation techniques to better model the iris boundaries taking into account that the pupil-lary and limbic boundaries are often not perfectly circular and the eyelids or eyelashes occlusion. B. Normalization Different iris images may not be all of the same size, either due to the change in distance from the camera or due to the changes in illumination which can cause the iris to dilate or contract. To compensate for the different size of each iris input image, Daugman resampled the segmented iris region to the fixed-size rectangular image by mapping the extracted iris region into a normalized coordinate system. To accomplish this normalization, every location on the iris image was defined by two coordinates ( r, h), where 0 < r 1 and 0 θ360 regardless of the overall size of the image. This normalization assumes that the iris stretches linearly when the pupil dilates and contracts. Although this approximation is good, it does not perfectly match the actual deformation of an iris.fig. 3 shows the normalized iris segmented above. Fig. 2 Iris segmentation Fig. 1 Block diagram of iris recognition system In order to generate the base templates, we use Masek and Kovesi s algorithm with appropriate modifications to handle images from CASIA V.3 database. The algorithm is based primarily on the methods given by Daugman [3] and is out-lined in the next subsections. Finally our proposed methodologies were applied to generate the final template. A. Segmentation A good segmentation algorithm should involve two procedures: iris localization and noise reduction. The iris localization process takes the acquired image and find both the boundary between the pupil and iris, and the boundary between the iris and the sclera. The noise reduction process refers to localizing the iris from the noise (non-iris parts) in the image. These noises include the pupil, sclera, eyelids, eyelashes, and artifacts. Fig. 2 depicts the iris segmentation step. Typical iris segmentation methods include Daugman s integro-differential operator [3] and edge detection using the circular Hough transform [7]. Daugman s method, which is used in this work, assumes the pupillary and limbic boundaries of the eye as circles and an integro-differential operator is utilized to detect the iris boundary by searching the parameter space. The circular boundary is detected Fig. 3 Normalized iris Fig. 4 Feature encoding C. Feature encoding In the feature encoding step, a template representing iris pat-tern information is created using a Gabor filter, log-gabor filter, or zero-crossing of the wavelet transform [9]. The differences in lighting between two different images causes error when directly comparing the pixel intensity of two different iris images. To alleviate this difficulty, Duagman extracted the features from the normalized iris image by using convolution with 2-D Gabor filters. In that system, the filters are multiplied by the raw image pixel data and integrated over their domain of support to generate coefficients which describe, ex-tract, and encode image texture information. A noise mask associated with the feature template is generated to mark the corrupted bits in the template, Fig

114 D. Matching The goal of matching is to evaluate the similarity of two iris representations. Created templates are compared using the Hamming distance or Euclidean distance. The normalized Hamming distance used by Daugman measures the fraction of bits for which two iris codes disagree. A low normalized Hamming distance implies strong similarity of the iris codes. If parts of the irises are occluded, the normalized Hamming distance is the fraction of bits that disagree in the areas that are not occluded on either image. To account for rotation, comparison between a pair of images involves computing the normalized Hamming distance for several different orientations that correspond to circular permutations of the code in the angular coordinate. The minimum computed normalized Hamming distance is assumed to correspond to the correct alignment of the two images. IV. PROPOSED IRIS RECOGNITION SYSTEM Fig. 6 Proposed voting approach for template fusion In fig. 6 the proposed voting approach was shown. Final template was produced by considering the probability of common feature of same person. Several MxN dimensional templates were considered for voting system. Each pixel value of final template was calculated by considering the bitwise probability of having same value in each pixel of MxN templates of that particular person. According to this modified biometric template fusion strategy, a final template was generated considering the uniqueness of those templates. Fig. 5 Proposed iris recognition system Iris recognition is one of the reliable and accurate biometric identification system available. However, many results have been published under favorable conditions. The work presented in this paper involves developing an iris recognition system considering noisy data (noisy images). The performance of many existing system falls for the noisy images. An image can be noisy in different ways like at the time of capturing, eyelids, poor segmentation or any other ways. In figure 5, the proposed iris recognition system was shown. First of all images were taken from database. Then median filter was used for noise(salt & pepper) removal. After segmentation and Feature encoding, several templates were found for a single person. Finally modified majority voting system (MVS) was used to generate a final template from several template of the same person by considering common feature. Then the modified Hamming distance was applied for the classification. V. TEMPLATE FUSION The idea of image fusion is used in pattern recognition and is generally applied in two different ways. The first involves seg-menting the image then the segmented feature objects and the original image are fused to improve the rate of object recognition. The fusion strategy adapted here uses the majority rule in a plain voting system to combine different base templates. Fig. 7 Mechanism of voting system The fusion process is illustrated in Fig. 7 using three assumed templates. Given three features templates produced from three images of the same eye, the assumed templates are shown in Fig. 7a. The three templates are fused into one final template, Fig. 5d, by electing the major bit value at each entry. The reliability of the elected bit value at each entry is determined by the maximum probability of having same values on those bits position. VI. EXPERIMENTAL RESULTS In this paper, CASIA V.3 database was used. Basically two database were used for the experiment. One of them was standard images dataset and another was noisy image dataset. Noisy database means the addition of some noise like salt & pepper with the standard images. A. Database of standard iris images We have used CASIA Database version 3 as the standard dataset. 10 different people s dataset was taken for the experiment. Each dataset contains 10 different template of each person. Total Dataset was 100 in number. Each eye image 320x280 in dimension. Their format was in JPEG. 113

115 Dataset Table 1. Data set considered No of person No of total eye images Possible comparison CASIA V B. Database of noisy iris images We have also used CASIA Database version 4 as the noisy dataset by adding some noise like salt and pepper. We have took 10 different people s dataset for experiment. Each dataset contains 10 different template of each person. Total Dataset was 100 in number. Each eye image 320x280 in dimension. Their format was in JPEG. Table 2. Noisy dataset considered Dataset No of No of total Possible person eye images comparison CASIA V C. Hardware & software specification To evaluate the proposed algorithms, we implement our algorithms described in Chapter four using MATLAB 2010 software. The environment where the experiments are performed in is Acer ASPIRE 4749Z PC, Dual Core Intel Pentium Processor (2.2 GHz), with 2GB RAM and Windows 8 operating system. In our experiments, Free public iris' databases (CASIA v4) was used. In order to evaluate our algorithms in the whole iris recognition system, some functions from Masek s iris recognition algorithms are used in the stages of normalization and parts of feature extraction. We have developed source code in MATLAB for Fusion Strategy and made modification the code of Labor Masek as needed. VII. COMPARISON Existing Proposed system Dataset system HD (%) HD(%) FHD(%) Standard Noisy In this paper we have discussed about our proposed methods as well as existing methods. Analysis of the developed iris recognition system has revealed a number of interesting conclusions. The implantation and performance of our proposed system was also discussed step by step thoroughly throughout the entire paper. The Main difference between our proposed and existing approach is that the performance for noisy image falls gradually in existing system but our proposed system is able to show better performance in this case. REFERENCES [1] Jain AK, Bolle R, Pankanti S. Biometrics: personal identification network society. Kluwer Academic Publishers; [2] Johnston R. Can iris patterns be used to identify people? Chemical and Laser Sciences Division LA PR, Los Alamos National Laboratory, Los Alamos, California, Tech. rep.; [3] Daugman John. High confidence visual recognition of persons by a test of statistical independence. IEEE Trans Pattern Anal Mach Intell 1993;15(11): [4] Daugman JG. Statistical richness of visual phase information:update on recognizing persons by iris patterns. Int J Comput Vision 2001;45(1): [5] Wildes R, Asmuth JC, Green GL, Hsu SC, Kolczynski RJ, Matey JR, et al. A system for automated iris recognition. In: Proceedings of the IEEE workshop on applications of computer vision; 1994.p [6] Wildes RP. Iris recognition: an emerging biometric technology. Proc IEEE 1997;85(9): [7] Wildes R, Asmuth J, Green G, Hsu S, Kolczynski R, Matey J,et al.. A machine-vision system for iris recognition. Mach Vision Appl 1996;9:1 8. [8] Boles WW. A wavelet transform based technique for the recognition of the human iris. In: Proceedings of the international symposium on signal processing and its application, ISSPA, Gold Coast, Australia; August p [9] Boles WW, Boashash B. Human identification technique using images of the iris and wavelet transform. IEEE Trans Signal Process 1998;46(4): [10] Zhu Y, Tan T, Wang Y. Biometric personal identification based on iris pattern. In: Proceedings of 15th international conference on pattern recognition, vol. 2; p [11] Lim S, Lee K, Byeon O, Kim T. Efficient iris recognition through improvement of feature vector and classifier. ETRI J 2001;23(2): [12] Noh S, Pae K, Lee C, Kim J. Multiresolution independent component analysis for iris identification. In: The 2002 international technical conference on circuits/systems, computers and communications, Phuket, Thailand; [13] Tisse C, M artin L, Torres L, Robert M. Person identification tec-hnique using human iris recognition. Proc Vision Interf 2002: [14] Ma L, Wang Y, Tan T. Iris recognition using circular symmetric filters. National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences; [15] Matey JR, Naroditsky O, Hanna K, Kolczynski R, LoIacono D,Mangru S, et al.. Iris on the Move: acquisition of images for iris recognition in less constrained environments. Proc IEEE 2006;94(11): [16] Park Kang Ryoung, Kim Jaihie. A real-time focusing algorithm for iris recognition camera. IEEE Trans Syst Man Cybern 2005;35(3): [17] Sung Hanho, Lim Jaekyung, Park Ji-hyun, Lee Yillbyung. Iris recognition using collarette boundary localization. In: International conference on pattern recognition; pp. IV: [18] Proenc a Hugo, Alexandre Luıs A. Iris segmentation methodology for non-cooperative recognition. In: IEE proceedings on vision,image and signal processing, vol. 153; p [19] Yao Peng, Li Jun, Ye Xueyi, Zhuang Zhenquan, Li Bin. Iris recognition algorithm using modified log-gabor filters. In: International conference on pattern recognition; p

116 Kinetics of Extraction of Ti(IV) from Sulphate Medium by Cyanex 302 Ranjit K. Biswas Dept. of Applied Chemistry and Chemical Engineering Rajshahi University Rajshahi-6205, Bangladesh Aneek K. Karmakar Dept. of Applied Chemistry and Chemical Engineering Rajshahi University Rajshahi-6205, Bangladesh Abstract The kinetics of Ti(IV)-extraction by Cyanex 302 (H 2 A 2 ) has been investigated by measuring initial Ti(IV)-transfer flux using a constant interfacial area stirred cell operated at 3 Hz. The empirical flux equation, at 293 K, is: F = [Ti(IV)] (1+233[H + ]) -1 [H 2 A 2 ] (o) 0.5 (1+3.2[SO 4 2- ]) -1. The activation energy, E a is measured to be kj/mol depending on experimental conditions. The enthalpy change on activation, S± is always highly negative. Analysis of the flux equation has been done, at various concentration regions of H + and SO 4 2-, to elucidate the mechanism of extraction. The rate determining chemical reaction step, irrespective of extraction condition, appears as: TiO 2+ + A - TiOA + ; and this step occurs via an S N 2 mechanism. Keywords Kinetics, Ti(IV), Cyanex 302, Lewis cell, Mechanism I. INTRODUCTION Ti(IV), obtainable from leaching of ilmenite (available world-wide in beach sands including Bangladesh), can be extracted by acidic organophosphorous extractants like di-2-ethylhexyl phosphoric acid (D2EHPA) [1, 2], di-tolylphosphoric acid (HDTP) [3], di-2-ethylhexylphosphonic acid mono 2-ethylhexyl ester [4, 5], bis-(2,4,4- trimethlypentyl) phosphinic acid (Cyanex 272) [6], bis-(2,4,4-trimethylpentyl) dithiophosphinic acid (Cyanex 301) [7], bis-(2,4,4-trimethylpentyl) monothiophosphinic acid (Cyanex 302) [8]. Previously, the kinetics of Ti(IV) - transfer from sulfate medium [9] and also from chloride medium [10] by D2EHPA have been reported. The same by other organophosphoric acids, excepting HDTP [11] are lacking. It is therefore worthy to work on the kinetics of Ti(IV) extraction by Cyanex 302 to know about the insights of this extraction process. A Lewis cell has been used to measure the kinetics of the considered system. II. EXPEIMENTAL A. Materials Cyanex 302 was received as a gift from Cytec Canada Inc. and used without further purification. Kerosene was bought from local market and distilled to collect colorless fraction within K. Other chemicals were of A. R. grade (E. Merck BDH) products and used without further purifications. B. Analytical Aqueous Ti(IV) concentration was estimated by the H 2 SO 4 H 2 O 2 method at 420 nm [12] using a UVvisible spectrophotometer (UV-1650 PC, Shimadzu, Japan). A Mettler Toledo ph-meter was used for ph measurement. Wherever necessary, either anhydrous Na 2 CO 3 or dilute H 2 SO 4 solution was used for ph adjustment of the aqueous phase. C. Preparation of solutions of Ti(IV) from TiO 2 The standard solution for spectrophotometric estimation of Ti(IV) was prepared by fusion of accurately weighed 1 g TiO 2 with 10 g KHSO 4 in a platinum crucible followed by dissolving cooled fused mass in 15% (v/v) H 2 SO 4 solution to obtain 1 L standard solution (1 ml 0.6 mg Ti). On the other hand, the stock solution of Ti(IV) for kinetic measurement was prepared by digestion of 50 g TiO 2 in 50 ml conc. H 2 SO 4 for 2 h under constant stirring by glass rod at K, followed by its partial dissolution in 15% (v/v) H 2 SO 4 solution. The insoluble part was filtered out to obtain 1 L solution containing g Ti(IV) and 3.45 mol SO D. Preparation of organic phase As Cyanex 302 (mol. wt. 306) is dimeric [13] having density of 0.93 g/ml, ml of Cyanex 302 was diluted to 250 ml by kerosene containing 5% (v/v) hexan-1-ol as de-emulsifier for preparing 1 mol/l dimeric Cyanex 302 solution. This solution was properly diluted by kerosene containing 5% (v/v) hexan-1-ol for being used in kinetic measurements. E. Cell and operating procedure The construction of Lewis cell and operating procedure to work with this cell are given elsewhere (Biswas et al., 1998). Experiments were carried out with 100 ml of each phase. The interfacial area formed in the cell was m 2. The cell was operated at 3 Hz. Most experiment were carried out at 293 K; otherwise, stated. F. Treatment of experimental data The Ti(IV)-transfer flux (F) in a Lewis cell can be calculated as follows: F (kmol/m 2 s) = /A t (1) were, = (c 100)/( ) and in this case, c is concentration change due to 115

117 log (F, kmol/m 2 s) log (F, kmol/m 2 s) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering extraction in mg/l, 100 is phase volume in ml, 1 st 1000 stands for 1 L = 1000 ml and 2 nd 1000 is for conversion of mg to g, 48 stands for converting to mole and last 1000 is for conversion of mol to kmol. According to rate law, F is related to concentration terms as follows: F = k f [Ti(IV)] a [H + ] b [H 2 A 2 ] (o) c [SO 4 2- ] d (2) where, unit of k f is dependent of reaction orders: a, b, c and d. Eq. (2) can be rewritten as: Log F = log k f + a log [Ti(IV)] + b lo g[h + ] + c log [H 2 A 2 ] (o) + d log [SO 4 2- ] (3) Equation (3) indicates that when [H + ], [H 2 A 2 ] (o) and [SO 4 2- ] are kept constant at (h), (h 2 a 2 ) (o) and (so 4 2- ), respectively, then the plot of log F vs. log Ti(IV) should yield a straight line with s equaling to a and I equaling to log k f (h) (h 2 a 2 ) (o) (so 4 2- ); from which the value of k f can be evaluated after finding out the values of b, c and d. These values can be obtained from the slopes of log F vs. log [H + ] (or, ph for which s = -b), log F vs. log [H 2 A 2 ] (o) and log F vs. log [SO 4 2- ] plots, respectively. Temperature dependence of rate can be treated by Arrhenius equation (log F = constant E a /2.303 RT), as well as, by the Activated Complex Theory (log (Fh/kT) = -(H ± /2.303 RT) + (S ± /2.303 R) + log f(r)) to calculate the values of S ± and H ±. III. RESULTS AND DISCUSSION The dependences of A, [Ti(IV)], [H + ], [HA] and [SO 2-4 ] on flux of Ti(IV)-transfer have been determined. F is independent of A, as is found to be directly proportional to A. This means that the cell of any cross-sectional area can be used for flux measurement. The rate measurements are confined to less than 5% material transfer, and consequently, the measured flux can be regarded as the initial flux of Ti(IV)-transfer in the investigated system. Figure 1, represents the log F vs. log [Ti(IV)] plots. For each set of experimental parameters, straight line of positive unity slope up to [Ti(IV)] of ~1.2 g/l is obtained, and at more [Ti(IV)] region, the levels off and then ultimately shows negative slope. The rate of Ti(IV)-transfer is, therefore, directly proportional to [Ti(IV)] in its lcr. The decrease in rate of extraction with increasing [Ti(IV)] in its hcr is contrary to the general rate law. The negative slope at hcr is possibly due to the formation of polymerized Ti(IV) species in the aqueous solution which is not extractable. The log F vs. ph plots are curves which can be fitted to Eqn.: log F = constant log ( ph ); where constant is dependent of experimental parameters and rate constant. Consequently, log F vs. -log ( ph ) plots would be straight lines. Such plots, at four sets of experimental parameters, are shown in Fig log ([Ti(IV)], mol/l) Fig. 1. Effect of [Ti(IV)] on flux. [SO 2-4 ] = 0.10 mol/l, O/A = 1 (O = 100 ml), Stirring speed = 3 Hz, A i = m 2, [Hexanol] (o) = 5% (v/v). ( ), ph = 1.50, [H 2A 2] (o) = 0.20 mol/l, Temp. = 293 K; s = 1.02, I = ; ( ), ph = 2.50, [H 2A 2] (o) = 0.20 mol/l, Temp. = 293 K; s = 1.01, I = ; (), ph = 2.50, [H 2A 2] (o) = 0.05 mol/l, Temp. = 293 K; s = 1.01, I = ; ( ), ph = 2.00, [H 2A 2] (o) = 0.10 mol/l, Temp. = 318 K; s = 1.01, I = log (1+233x10 -ph ) Fig. 2. log F vs. -log ( ph ) plots. [SO 4 2- ] =0.10 mol/l, O/A = 1 (O = 100 ml), Stirring speed = 3 Hz, A i = m 2, [Hexanol] (o) = 5% (v/v). ( ), [Ti(IV)] (ini) = 1.00 g/l, [H 2A 2] (o) = 0.10 mol/l, Temp. = 293 K; ( ), [Ti(IV)] (ini) = 0.20 g/l, [H 2A 2] (o) = 0.05 mol/l, Temp. = 293 K; (), [Ti(IV)] (ini) = 0.20 g/l, [H 2A 2] (o) = 0.05 mol/l, Temp. = 318 K; ( ), [Ti(IV)] (ini) = 1.00 g/l, [H 2A 2] (o) = 0.40 mol/l, Temp. = 293 K. Figure 3 represents the dependence of flux on extractant concentration. The log F vs. log [H 2 A 2 ] (o) plots for all sets of experimental parameters are straight lines with slopes equaling to The rate of extraction of Ti(IV) by Cyanex 302 is, therefore, directly proportional to the square root of Cyanex 302 concentration. The variation of flux with the variation of [SO 4 2- ] is shown in Fig. 4, as log F vs. -log (1+3.2[SO 4 2- ]) plots. Virtually, the log F vs. log [SO 4 2- ] plots are curves; which can be fitted to Eq.: log F = constant log (1+3.2[SO 4 2- ]). Consequently, the experimental points in Fig. 4 fall on unity sloped straight line for a particular set of experimental parameters. Therefore, the rate of extraction in the investigated system is independent of [SO 4 2- ] in its lcr; but is inversely proportional to [SO 4 2- ] in its hcr. The effect of temperature on flux of Ti(IV) transfer by Cyanex 302 is shown in Fig. 5, as Arrhenius plots for 4 sets of experimental parameters. In all cases, straight lines are obtained; and from the slopes of these lines, E a -values have been calculated. At low ph values, E a of >50 kj/mol is obtained; whilst, at high ph values, E a value is ~40 kj/mol. 116

118 log (F, kmol/m 2 s) log (F, kmol/m 2 s) log (F, kmol/m 2 s) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering Fig. 3. Dependence of flux on Cyanex 302 concentration. [SO 2-4 ] = 0.10 mol/l, O/A = 1 (O = 100 ml), Stirring speed = 3 Hz, A i = m 2, [Hexanol] (o) = 5% (v/v). (), [Ti(IV)] (ini) = 1.0 g/l, ph = 2.0, Temp. = 293 K; s = 0.50, I = ; ( ), [Ti(IV)] (ini) = 1.0 g/l, ph = 1.5, Temp. = 293 K; s = 0.50, I = ; (), [Ti(IV)] (ini) = 0.2 g/l, ph = 1.8, Temp. = 318 K; s = 0.49, I = log ([Cyanex 302], mol/l) F = ( /3.2) [Ti(IV)] [H 2 A 2 ] (o) 0.5 [SO 4 2- ] -1 (6) Case III. At hcr of H + and lcr of SO 4 2- : F = ( /233) [Ti(IV)] [H 2 A 2 ] (o) 0.5 [H + ] -1 (7) Case IV. At hcr of H + and SO 4 2- : F = ( /( )) [Ti(IV)] [H 2 A 2 ] (o) 0.5 [H + ] -1 [SO 4 2- ] -1 (8) Cyanex 302 in dimeric form in the organic phase cannot be partitioned to the aqueous phase due to its high hydrophobicity (C 32 ); rather its monomeric form at equilibrium with the dimeric form is likely to be distributed to the aqueous phase where it can be ionized. Therefore, the monomeric model of Cyanex 302 can be represented as: log (1+3.2 [SO 2- ]) 4 Fig. 4. Dependence of flux on [SO 2-4 ]. O/A = 1 (O = 100 ml), Stirring speed = 3 Hz, A i = m 2, [Hexanol] (o) = 5% (v/v). (), [Ti(IV)] (ini) = 0.20 g/l, ph = 2.00, [H 2A 2] (o) = 0.10 mol/l, Temp. = 293 K; I = -7.76; ( ), [Ti(IV)] (ini) = 1.0 g/l, ph = 1.60, [H 2A 2] (o) = 0.10 mol/l, Temp. = 293 K; I = ; (), [Ti(IV)] (ini) = 0.20 g/l, ph = 1.80, [H 2A 2] (o) = 0.10 mol/l, Temp. = 318 K; I = The temperature dependence data have also been treated by the Activated Complex Theory. The log (Fh/kT) vs. 1/T plots (not shown) are also straight lines. From the slopes of these lines, H ± values of kj/mol (value decreases with increasing ph) are obtained. From the intercepts of these lines and values of f(r), S ± are calculated to be - ( ) J/mol K (value decreases with increasing ph). From the intercepts of lines in Figs. 1-4, the value of rate constant, k f has been calculated to be , with standard deviation of log k f being , at 293 K; and , with standard deviation of log k f being , at 318 K. These two rate constants at two temperatures yield E a = 53 kj/mol, which is similar to that pointed out earlier. Based on above results, the rate in the investigated system at 293 K can be expressed as: F = [Ti(IV)] (1+233 [H + ]) -1 [H 2 A 2 ] 0.5 (o) (1+3.2 [SO 2-4 ]) -1 (4) provided initial [Ti(IV)] is kept below 1.2 g/l. Equation (4) yields four limiting cases: Case I. At lcr of H + and SO 2-4 : F = [Ti(IV)] [H 2 A 2 ] (o) (5) Case II. At lcr of H + and hcr of SO 2-4 : (1/T)x10 3, K -1 Fig. 5. Arrhenius plot. [SO 4 2- ] = 0.10 mol/l, O/A = 1 (O = 100 ml), stirring speed = 3 Hz, A i = m 2, [Hexanol] (o) = 5% (v/v). ( ), [Ti(IV)] (ini) = 1.00 g/l, ph = 1.40, [H 2A 2] (o) = 0.10 mol/l; s = -3.04, E a = 58.2 kj/mol; ( ), [Ti(IV)] (ini) = 0.20 g/l, ph = 2.20, [H 2A 2] (o) = 0.40 mol/l; s = -2.62, E a = 50.2 kj/mol; (), [Ti(IV)] (ini) = 0.20 g/l, ph = 2.50, [H 2A 2] (o) = 0.10 mol/l; s = , E a = 41.0 kj/mol; ( ), [Ti(IV)] (ini) = 0.50 g/l, ph = 3.00, [H 2A 2] (o) = 0.05 mol/l; s = -1.87, E a = 35.8 kj/mol. 2 [H 2 A 2 ] (o) = K 2 P HA [A - ] 2 [H + ] 2 (9) At lcr of H + (i.e. at high ph) and SO 2-4, Ti(IV) is likely to be present mostly as TiOOH + ; and so the [Ti(IV)] terms in Eq. (5) is virtually [TiOOH + ]; which equals to [TiO 2+ ] [OH - ]. This relation together with Eq. (9) converts Eq. (5) to: F = K P HA [TiO 2+ ] [A - ] (10) In case II, Ti(IV) is likely to be present mostly as TiO.OH.SO - 4 ; and so the [Ti(IV)] term in Eq. (6) is virtually [TiO.OH.SO - 4 ]. This concentration term equals to [TiO 2+ ] [OH - ] [SO 2-4 ]. This relation together with Eq. (9) convert Eq. (6) to: F = ( /3.2) [TiO 2+ ] [A - ] (11) For case III, Ti(IV) presents as [TiO 2+ ]; and so Eq. (7) takes the form (with the help of Eq. (9): F = ( /233) K P HA [TiO 2+ ] [A - ] (12) Finally, for the last case, Ti(IV) is likely to be present as TiOSO 4 - ; and so the [Ti(IV)] term in Eq. (8) is virtually [TiOSO 4 ], which equals to [TiO 2+ ] [SO 4 2- ]. This relation is conjunction with Eq. (9) convert Eq. (8) to: F = ( / ) K P HA [TiO 2+ ] [A - ] (13) 117

119 Equations (10) - (13) are of similar form, differing only in constant terms; therefore, same extraction mechanism holds good. As the flux equations contain [TiO 2+ ] and [A - ] terms only, the addition of a monomeric anionic ligand of the extractant to a TiO 2+ in the aqueous phase to form TiOA + species is rate determining. It appears therefore that the locale of rate determining chemical reaction step is the bulk aqueous phase. The high activation energy (>50 kj/mol) in low ph region supports above mechanism. However, in high ph region, E a <48 kj/mol indicate that the cited reaction step and the diffusion of a reactant to the reaction site or a product from the reaction site are equally slow. The negative S ± indicates that the rate determining chemical reaction step occurs via an S N 2 mechanism. In this mechanism, A - co-ordinates to hydrated TiO 2+ species to form a higher co-ordinated activated metal ion species; and this step is slower than the rate of elimination of a coordinated water molecule from the activated complex to form normal co-ordinated species; and also that the rate addition of second A - ligand to TiOA + in forming extractable species, TiOA 2. IV. CONCLUSION The initial rate of Ti(IV)- transfer in the Ti(IV)- SO 4 (H +, Na + ) Cyanex 302 kerosene 5% (v/v) hexan-1-ol system can be measured by a Lewis cell operated at 3 Hz. The reaction orders with respect to [Ti(IV)], [H + ], [H 2 A 2 ] (o) and [SO 2-4 ] have been determined. The rate constant at 293 K and 318 K are determined to be and , respectively. With these values, the empirical flux equation is obtained. Moreover, from the temperature dependence data of flux, values of E a, S ± and H ± have been determined. Analysis of flux equation leads to the fact that the addition of first A - to TiO 2+ in forming TiOA + within the bulk aqueous phase is slow or rate determining. The E a value supports this statement in lower ph-conditions; however, it suggests the extraction process to be intermediate controlled at higher ph region. The high negative S ± - value suggest that the rate controlling chemical reaction step occurs via an S N 2 mechanism, i.e. the addition of first A - to [TiO(H 2 O) n ] 2+ in forming higher co-ordinated activated complex, [TiO(H 2 O) n A] + is the slowest step. V. ACKNOWLEDGEMENT Authors are grateful to Cytec Canada Inc. for supplying Cyanex 301 as gift. NOTATIONS AND ABBREVIATIONS H 2 A 2 = Dimeric bis-(2,4,4- trimethylpentyl) monothiophosphinic acid (Cyanex 302) HA = Monomeric Cyanex 302 = Amount of Ti(IV) transferred in 100 ml organic phase, kmol k f = Forward extraction rate constant E a = Activation energy, kj/mol H ± = Enthalpy change on activation, kj/mol S ± = Entropy change on activation, J/mol K F = Flux of Ti(IV) transfer, kmol/m 2 s A = Interfacial area, m 2 wrt = With respect to a, b, c, d = Reaction order wrt [Ti(IV)], [H + ], [H 2 A 2 ] (o) and [SO 2-4 ], respectively h = Planck s constant = kj s k = Boltzman constant = J/K T = Absolute temperature, K s = Slope I = Intercept (ti), (h), (h 2 a 2 ) (o), (so 4 2- ) = Constant concentration of Ti(IV), H +, H 2 A 2 (o) and SO 4 2-, respectively, which are kept constant at a set of experiment K 2 = Dimerization constant of HA, kmol/m 3 P HA = Partition coefficient of HA = Acid dissociation constant of HA, kmol/m 3 R = Molar gas constant = J/mol K f(r) = Function of reactant = (ti) a (h) b (h 2 a 2 ) c (o) (so 2-4 ) d lcr = Lower concentration region hcr = Higher concentration region S N 2 = Substitution neucleophilic bimolecular Subscript (o) = Organic phase REFERENCES [1] R. K. Biswas and D. A. Begum, Solvent extraction of tetravalent titanium from chloride solution by D2EHPA in kerosene, Hydrometallurgy, vol. 49, pp , [2] R. K. Biswas, M. R. Zaman and M. N. Islam, Extraction of TiO 2+ from 1 M (Na +, H + 2- ) SO 4 by D2EHPA, Hydrometallurgy, vol.63, pp , [3] R. K. Biswas and M. R. Ali, Solvent extraction of Ti(IV) from acidic sulfate solutions by HDTPbenzene-i-butanol system, Rajshahi University Studies (Press & Publication Department of Rajshahi University), vol. 15B, pp , [4] D. Fontana, P. Kulkarni and L. Pietrelli, Extraction of titanium(iv) from acidic media by 2-ethylhexyl phosphonic acid mono-2-ethylhexyl ester, Hydrometallurgy, vol. 77, pp , [5] J. Saji and M. L. P. Reddy, Selective extraction and separation of titanium(iv) from multivalent metal chloride solutions using 2-ethylhexyl phosphonic acid mono 2-ethylhexyl ester, Sep. Sci. Technol., vol. 38, pp , [6] J. Saji, J. K. Saji and M. L. P. Reddy, Liquid-liquid extraction of tetravalent titanium from acidic chloride 118

120 solutions by bis(2,4,4-trimethylpentyl)phosphinic acid, Solvent Extr. Ion Exch., vol. 18, pp , [7] R. K. Biswas and A. K. Karmakar, Solvent extraction of Ti(IV) from acidic sulphate medium by Cyanex 301 dissolved in kerosene, Sep. Sci. Technol., vol. 49, pp. 1-12, [8] R. K. Biswas and A.K. Karmakar, Solvent extraction of Ti(IV) in the Ti(IV)-SO 4 2- (H +, Na + )- Cyanex 302 kerosene 5% (v/v) hexan-1-ol system, Hydrometallurgy, vol , pp. 1-10, [9] F. Islam and R. K. Biswas, Kinetics and mechanism of solvent extraction of Ti(IV) from acidic aqueous solutions with HDEHP in benzene, J. Inorg. Nucl. Chem., vol. 40, pp , [10] R. K. Biswas and D. A. Begum, Solvent extraction of tetravalent titanium from chloride solution by D2EHPA in kerosene, Hydrometallurgy, vol. 49, pp , [11] R. K. Biswas and M. R. Ali, Kinetics of extraction of Ti(IV) from acidic sulphate solution by HDTPisobutanol-benzene system, Indian J. Chem., vol. 28A, pp , [12] J. Bassette, R. C. Denny, G. H. Jeffery and J. Mendham, Vogel s Textbook of Quantitative Inorganic Analysis, 4th ed.; ELBS: London, p. 750, [13] B. K. Tait, Cobalt-nickel separation: the extraction of cobalt(ii) and nickel(ii) by Cyanex 301, Cyanex 302 and Cyanex 272, Hydrometallurgy, vol. 32, pp , [14] R. K. Biswas, M. A. Habib, M. R. Ali and M. Z. Haque, Kinetics of Mn2+ extraction in the acidic chloride-d2ehpa-kerosene system using the constant interfacial area stirred cell technique, Pak. J. Sci. Ind. Res., vol. 41, pp ,

121 Autonomous Human Face Detection and Tracking System with Variant Poses, Blur and Illumination Md. Zweel Rana 1, Monimul Islam 2,,Mohiuddin Ahmad 3 Department of Electrical and Electronic Engineering (EEE) Khulna University of Engineering and Technology (KUET), Khulna-9203 zweelrana@hotmail.com, i.monimul@hotmail.com, mohiuddin.ahmad@gmail.com Abstract This paper approaches the technique for real time variant human face detection and tracking using a modified version of Viola-Jones algorithm. However, the detection of a face in real time environment having uneven illumination, pose and nonuniform motion blur is quiet difficult. In this paper, we propose three level approaches to detect faces and tracking. The low, intermediate, and high level processing involve the improvement of image quality, feature extraction, pattern detection and further improvement of the quality of the image. Then face tracking for moving camera is performed via servo and Arduino. This implementation is robust and efficient to detect and track the face in case of non-uniform motion, blur, illumination and variant pose. This paper focuses on automatic face detection and tracking on video streams for surveillance in public places. This system can be used for human interaction robot and security purpose. Keywords Face detection, AdaBoost, pose, illumination, Blur, Arduino I. INTRODUCTION In recent year face detection and tracking system is a technology in computer vision that determines the movement of a face in arbitrary images. Many computer applications use face detection algorithm to detect and tracking faces [3]. Recently face detection and tracking has become an important research subject in the field of pattern recognition and computer vision research, which has an importance in the field of autonomous face recognition, video conference, intelligent video surveillance, advance human-computer interaction, medical diagnosis and so on. It detects facial feature and ignores anything else, such as trees, buildings, objects and even bodies. However, many faces can t detect the correct face due to the movement of human face from left to right or right to left, tilt in certain angles and sometimes it rotates and blur due to camera or object movement. The research purpose of computer vision aims to simulate the behavior of human eyes directly by using computer. Computer vision is such type of research field which aims to percept and represent the 3D information for world objects, 3D objects surface reconstruction and representation not only provide theoretical outcomes, but also are required by numerous applications. Robust face detection algorithm is very essential to the application purpose that applies this face detection for other process, such as facial expression recognition or personal identification system or human interaction robot. Therefore, the detection must be accurate to get a desired tracking result. Face detection is the base for facial recognition and tracking, whose results directly affect the process and accuracy of facial recognition. Several techniques have been proposed to detect a face from real time environment or still image or image sequence. The most popular face detection techniques are that the facial image or image sequence is obtained in a controlled environment with uniformed illumination and a simple background. When capturing a face image door, normally the camera distance and illumination settings are carefully controlled and constantly recalibrated to ensure that the settings are identical across subjects [3]. This situation arises in a complex environment where a cluttered background and sometimes occlusions can occur. Illumination variation is a major problem in uncontrolled environment. It can result in huge variations for the same person due to the change in illumination. The problem is mainly due to the 3D shape of human faces under lighting in different directions [1]. II. METHODOLOGY There is many research has been done in the area of face detection and tracking but only few approach have combine detection and tracking. There are several work face detection across illumination [6] or pose or non-uniform motion blurring [7], [8]. Zahir and Samad [5] developed a method for real time pose face detection and tracking system that detect human face across variant pose. The system was based on segmentation the skin color and background color and contour detection. The most famous and successful real time face detection method is Viola s frame work which was initially proposed by Viola and Jones [1], [2] and improved by Lienhart [4]. This frame work used a set of Haar-like features in which each feature was described by the template. A face detection and tracking system using pyramidal and Lucas Kanaden algorithm by N. Sethi and A. Aggerwal [2]. In this method at the stage of face detection, Shi and Thosmasi algorithm was used to extract feature points and pyramid al Lucas-Kanade algorithm used to track detect features. W. Kao, C. Chiu and Y. Yang present a new face image restoration approach based on Wiener filter and pattern recognition techniques [8]. 120

122 A. Haar-like Feature and AdaBoost Algorithm This algorithm s framework was introduced by Viola and Jones which applied to face detection. In this method Viola and Jones select a large number of very efficiently computable features. In every week classifier implements a simple threshold function on one of the feature. By using a large set of week classifier, AdaBoost learning is used to small number of weak classifiers and to combine them into a classifier deciding whether an image is a face or a non-face. In AdaBoost, most non-face patches were quickly rejected by the early nodes. Cascade detectors have demonstrated impressive detection speed and high detection rates. This algorithm is very fast and effective for object detection. The framework of this method includes following steps [1]: Use of haar like features and computation of haar features using integral image. Selection of rectangle feature (weak classifier) to form a strong classifier. Cascade framework and use of AdaBoost to train cascade nodes. The algorithm of AdaBoost works as follows [1]: AdaBoost starts with uniform distribution of "weights" over training examples. The weight tells the learning algorithm the importance of the example. Obtain a weak classifier from the weak learning algorithm. Increase the weights on the training examples that were misclassified. Repeat. At the end, carefully make a linear combination of the weak classifiers obtained at all iteration. B. Wiener Filter for Deblurring Images: An input blurred image can be mathematically expressed as Eq. (1) if the impulse response is stationary across the image and object field.,,,, g x y h x y f x y q x y (1) where the variables x and y give the coordinate of a pixel. The functions g, h, f, and q represent the recorded image, the ideal image, 2-D impulse response, and noise contribution respectively. The discrete Fourier transform of Eq. (1) can be used to yield the frequency domain model as Eq. (2).,,,, G u v H u v F u v Q u v (2) where H(u,v) gives the samples of the frequency response of the blurring system, and u and v are the discrete horizontal and vertical spatial frequency variables. Directly finding the inverse function of H(u,v) usually suffers from some problems since blind deconvolution is an ill-posed problem. Wiener filter uses the power spectrum of both image and the noise to prevent excessive noise amplification. The frequency response of the filter is chosen to minimize the mean squared restoration error and the solution to this minimization problem which is given by Eq. (3). H ( u, v) 1 H( u, v) Hw( u, v) (, ) (, ) (, ) * H u v K H u v H u v K (3) where H * (u,v) gives the complex conjugate of H(u,v), and K is a regularization parameter. By using Wiener filter for image deconvolution, the problem becomes how to determine the radius of circle of confusion (COC) in the defocused image. The uniform circular PSF model stated in Eq. (4) is typically used as an approximation to these effects: h( x, y) 1 2, r 0, 2 2 x y r otherwise (4) In brief, Wiener filter can give better restoration results with ringing artifacts attenuated. However, the deblurring radius r as well as the parameter K is still required to be fine-tuned. Since the ringing artifacts while deblurring face images have similar features, it is possible to develop an autonomous approach for determining the quantities and preventing the reconstructed image from generating undesired artifacts. C. Kalman Filter for Face Pose Detection Algorithm: The initial face image and its pose in the first frame captured by the web camera, the task of finding the face location and the human facial pose in subsequent frames can be represented as simultaneous 3D human facial pose tracking and face detection. Given the 2D face model obtained from the initialization, the current facial pose parameters X t = (ω t, φ t, κ t, λ t ) and the current face image location (x t, y t ), the Kalman Filtering based face pose tracking consists of the steps which are given below 1) Combined Prediction Method : Assuming the state vector at time t be represented as X t = (ω t φ t κ t λ t x t y t ) t According to the theory of Kalman Filtering, the system can be presented as Eq. (5) Xt 1 Xt wt (5) where Φ is the state transition matrix, and w t system perturbation. Given the system model, X t +1, the state 121

123 vector at t + 1, can be predicted by Eq. (6) along with its covariance matrix Σ t+1 to determine its uncertainty. X X w t 1 t t (6) Kalman Filtering assumes smooth face movement. The prediction will be wrong significantly if head undergoes a sudden rapid movement. To deal with this problem, we propose to approximate the face movement with eyes movement since eyes can be reliably tracked in each frame. Assuming the predicted face pose vector at t+1 based on eyes motion be X t p +1. Then the final predicted facial pose should be on the basis of combining the one from Kalman with the one from eyes as Eq. (7), i.e. * ( p X t X t 1 t 1 t 1 ) t X X (7) The simultaneous use of Kalman Filtering and eyes motion allows to track accurately facial pose prediction even under significant and rapid head movements. We can then have a new covariance * * matrix X t 1 for X t 1 using the above equation to specify the uncertainty. 2) Detection Technique: * Given the predicted state vector X t 1 at time t+1 and the prediction uncertainty Σ t+1, we can perform a facial pose elimination by menace of face detection * verification. Specifically, X t 1 and Σ t+1 form a local pose space at the moment of t+1. The pose search in this local pose space will bring towards the measured pose z t+1. The search can be formulated as a minimization problem to be detailed later. Let the measurement model in the form needed by the Kalman filtering be given by Eq. (8). z HX m (8) t1 t1 t1 where m t+1 states measurement uncertainty, normally distributed as m t+1 N(0, S), where S is measurement error covariance matrix. The matrix H is related with the state X t+1 to the measurement z t+1. 3) Updating Face Value: Based on the given the predicted face poses X * t+1, its covariance matrix Σ t+1 and the measured face pose z t+1, face pose value updating can be performed to bring out the final pose X t+1 as Eq. (9). * * Xt 1 X t 1 t 1( z HX ) (9) t1 t1 update the face model dynamically on account for the significant changes under different face orientations or facial expressions. D. Illumination Detection: The illumination conditions vary greatly, many times causing shadows on parts of the face. Poor illumination conditions may mean that color information is degraded and so techniques that rely upon color information may fail. One of the most popular techniques for correcting poor lighting condition is Light compensation. The brightest regions of the image are defined to be the reference white, and all other color channels are then scaled so that these regions become true white. Here we have proposed a simple and effective method for light compensation. We lighten the whole image through the reference white approach with proper brightness. The works show that our method can make it possible to enhance the brightness of the dark regions and meanwhile can preserve the details of the bright regions. However, the detail enhancing method usually enhances the noise too but it can be removed in the later using noise removal technique. III. PROPOSED MODEL Here we proposed an algorithm that combine AdaBoost algorithm face detection, Wiener Filter for deblurring, Kalman Filter for face detection and an approach for remove illumination. Then by using Arduino we set camera position at the center of the human face. In this research we use a classifier is trained so that it can be detected in a specific region of interest of an input image captured from the real time environment. The classifier will give output 1 if it detects human face and 0 if it not. Haar-like Feature and AdaBoost Algorithm smoothly detects face from the captured real time environment. Furthermore, to cope up with the various illumination situation Haarlike Feature and AdaBoost Algorithm gives more flexibility. For the worst conditions, to remove blur Weiner Filter is used so that from the reconstructed frame the system can detect human face if it is present in the blurry image. To enhance skill of this system Kalman Filter is used to detect human face in variant pose and track them in real time environment. The search window can move across the whole image with the help of classifier. The whole process is shown by the following flow chart in Fig. 1. where K t+1 is the Kalman gain matrix. 4) Face Model Update: If current face pose aspect significantly varies from the face model aspect, the face model should be updated. The face model of frame t+2 is updated based on successful face pose estimation for the frame t+1. Our work shows that it is important to 122

124 START Scan Input Image Removing Blur No Scan Human Face YES No Scan Human Face YES Face detect considering pose and illumination Face Capture Detect Face Position (x,y) Initiate Servo Set Servo Position END Figure 1: Flowchart of proposed model IV. IMPLEMENTATION For implanting the proposed model, the images sequence is captured by web-cam. Then from the captured image detect the human faces with poses and track using Arduino and servo motor. If no face presence at the frame, then it deblurred and again search faces for tracking. First we generate a rectangle class which keeps track of the face coordinates. The OpenCV serial library is necessary for communicating with the open source hardware platform Arduino. There after adjusting Screen Size Parameters on contrast/brightness values we convert the real time image coming from webcam to greyscale format. Logical operation initiated to make out decision if any faces are detected. If a human face is tracked, the midpoint of the first face in the frame is detected. By manipulating these values, we find the midpoint of the rectangle. Finding out if the Y component of the face is below the middle of the screen, the tilt position variable to lower the tilt servo is updated. Moving forward we find out if the X component of the face is to the left of the middle of the screen. So the pan position variable to move the servo to the left side is updated. If the X component of the face is to the right of the middle of the screen the pan position variable is updated to move the servo to the right. The servo positions are updated by sending the serial command to the Arduino. The pan & tilt position of the servo motor linked with web camera is directly proportional to the serial command of the coordinates to the Arduino of the X & Y components of the human face from midpoint of the rectangle which was detected earlier. The generalized hardware diagram of the proposed research is shown in Fig. 2. Figure 2: Hardware implementation of proposed model. V. RESULTS AND DISCUSSIONS This work has been done in such way that it can detect human face illumination sensitivity, blurred image and various pose. Real rime results are obtained by detecting several faces in a single frame simultaneously. Figure 3: Face tracking in indoor or normal environment Fig. 3 shows the testing results of the face detection. Human faces are detected successfully from all most every image. We consider both even and uneven background as well as indoor and outdoor in real time scenario. Whereas if background color is complex having similar objects that are similar to the human subject s face the detection technique is able to detect the human face. This system can also track multiple human faces in single frame simultaneously along with having different skin color of human face. Figure 4: Face detection considering dark and bright illumination Fig. 4 shows the face detection considering the illumination factor. Our proposed model is able to cope pace with the uneven illumination. But if the illumination is too bright or too dark then the program fails to detect a human face. Referring to Fig. 5, we can verify that this system can detect human face even if the human object is moving. Moreover, the proposed system can detect human face from blur image. So the system has more 123

125 immunity along with the ability to track the human continuously. Figure 5: Face detection considering blurry image Figure 6: Face detection considering variant poses Fig. 6 shows the collection of images which are taken in various poses. With the change of angle human face can be detected by the system. By using this proposed model, it is found that 10~15 frames are processed to detect face by 1 second or less which means that this model is suitable for real time environment. The detection efficiency was greatly improved by using open source OpenCV. Figure 7: Face tracking using servo via web camera The result of this project was capture by external camera and screen shot from computer. When camera got any human faces then it detects and indicates by triangle. If face (single) move at any direction, then the webcam also moving at same direction by power of servo motor. This tracking system via servo is shown in Fig. 7. Table I shows the performance of proposed work for normal environment, light variance, blur and pose. From these result, it is obvious that our work can robustly detect human face considering various parameters. VI. CONCLUSIONS This paper proposed the real time environment human face tracking system using a web camera having different illumination, variant pose, blurry image and tracking human face object continuously. This system was developed with the help of contour detection technique to track the shapes of various human faces. The performance of this system was enhanced by the use of open source OpenCV. From the above discussed result it can be said that this system can be used robustly for both indoor and outdoor environment considering various parameters. This system can be implanted in various purpose like surveillance system in military base, museum as well as other important and local areas. In near future we will improve this work for too dark environment REFERENCES [1] P. Viola and M. Jones, Robust real-time face detection, The international Journal of Compute Vision, vol. 57(2), pp , [2] N. Sethi and A. Aggarwal, Robust face detection and tracking using pyramidial Lucas Kanade tracker algorithm, International Journal of Computer Technology and Applications, vol. 2(5), pp , [3] Y. Ming-Hsung, D. J. Kriegman and N. Ahuja, Detecting faces in images: a survey Electronic Education, vol. 30, pp , [4] R. Lienhart and J. Maydt, Haar-like features for rapid object detection, in Proc. of IEEE International Conference on Image Processing, p , [5] N. B. Zahir, R. Samad, M. Mustafa, Initial Experimental Results of Real-Time Variant Pose Detection and Tracking System, in Proc. The IEEE International Conference on Signal and Image Processing Applications (ICSIPA), p. 1-5, [6] J. Wang, H. Wang, A Survey Preprocessing Methods for Illumination Insensitive Face Recognition Journal of Electrical and Electronic Education, vol. 30, pp , [7] A. Danielyan, V. Katkovnik, and K Egiazarian, BM3D Frames and Variational Image Deblurring IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 21, NO. 4, pp , [8] W. Kao, C. Chiu, Y. Yang, Automatic Ringing Artifact Detection in Restoring Blurred Face Images IEEE International Midwest Symposium on Circuits and Systems (MWSCAS), p , TABLE I. SUMMARY OF FACE DETECTION DATA Condition No. of face used Detected Missed Precision (%) Normal Light Variance Blur Pose

126 Electrochemical Corrosion Characterization of Artificially Aged Al-6Si-0.5Mg (-1Cu) Alloys in Sodium Chloride Solution Abul Hossain 1*, M. A. Gafur 2, Fahmida Gulshan 3, and ASW Kurny 3 1 MSTE Plant (KTL), Sylhet Gas Fields Ltd., Golapgonj, Sylhet Pilot Plant & Process Development Centre, BCSIR, Dhaka 3 Department of Materials and Metallurgical Engineering, BUET, Dhaka * ah_buetmmesgfl@live.com Abstract In this study, corrosion behavior of artificially peakaged Al-6Si-0.5Mg (-1Cu) alloys in 0.1M NaCl solution has been investigated using potentiodynamic polarization and electrochemical impedance spectroscopy (EIS) techniques. The potentiodynamic polarization curves reveal that 1 wt%cu content alloy is less prone to corrosion than the Cu free alloy. The EIS test results also showed that corrosion resistance or charge transfer resistance (Rct) increases with the addition of 1wt%Cu into Al-6Si- 0.5Mg alloy. Higher charge transfer resistance (R ct ) was obtained with the addition of 1 wt% Cu and lower R ct value was for Cu free Al-6Si-0.5Mg alloy. Due to addition of Cu and thermal modification, the magnitude of open circuit potential (OCP), corrosion potential (Ecorr) and pitting corrosion potential (E pit ) of Al-6Si- 0.5Mg alloy in NaCl solution were shifted to the more noble direction. Keywords Al-Si alloy, potentiodynamic polarization, electrochemical impedance spectroscopy. I. INTRODUCTION Owing to their light weight, suitable strength and strong resistance to corrosion, aluminum alloys are used in a broad spectrum of engineering applications. The corrosion resistance of aluminum is attributed to an exceptionally stable oxide film that forms on its surface. This film is resistant to attack from water and oxygen in a wide range of temperatures and ph levels, making aluminum alloys useful in a variety of environments (G. D. Claycomb et al. 2002). Corrosion behavior of aluminum alloys is significantly affected by the presence of particles in the matrix. Particles that contain Al, Cu and Mg tend to be anodic relative to the alloy matrix, while those that contain Al, Cu, Fe and Mn tend to be cathodic relative to the matrix (R. P. Wei et al. 1998). The composition of an alloy and its thermal treatment are important do determine the susceptibility of the alloy to corrosion (M. Czechowski, 2007 and M. Abdulwahab et al. 2011). Over the years a number of studies have been carried out to assess the effect of Cu content and the distribution of second phase intermetallic particles on the corrosion behavior of Al alloys. The distribution of Cu in the microstructure affects the susceptibility to localized corrosion. Intergranular corrosion (IGC) is generally believed to be associated with Cu containing grain boundary precipitates and the precipitates free zones (PFZ) along grain boundaries (G. Svenningsen et al. 2006, G. Svenningsen et al and M. H. Larsen et al. 2006). In heat treatable Al-Si-Mg(-Cu) series alloys the susceptibility to localized corrosion [pitting and / or intergranular (IGC)] and the extent of attack are mainly controlled by the type, amount and distribution of the precipitates which form in the alloy during any thermal or thermomechanical treatment performed during manufacturing processes (G. Svenningsen et al. 2006, G. Svenningsen et al. 2006, M. H. Larsen et al. 2006, G. Svenningsen et al and G. Svenningsen et al. 2006). Depending on the composition of the alloy and parameters of the heat treatment process, these precipitates form in the bulk of the grain, or in the bulk as well as grain boundaries. As indicated by several authors, the precipitates formed by heat treatment in Al-Si-Mg alloys containing Cu are the θ (Al 2 Cu) Q-phase (Al 4 Mg 8 Si 7 Cu 2 ), β-phase (Mg 2 Si) and free Si if Si content in the alloy exceeds the Mg 2 Si stoichiometry (S. Zor et al. 2010, H. Zhan et al and M. H. Larsen et al. 2010) It has demonstrated that in Al-Cu-Si alloys a more finely and homogeneously distributed Al 2 Cu and needle-like Si particles in the ternary eutectic mixture, tend to improve the corrosion resistance mainly due to the galvanic protection of both Al 2 Cu and Si phases (M. Park, 2005). Although it has also been reported (M. Park, 2005 and W. R. Osório, et al. 2011) that fine Si particles tends to decrease the corrosion resistance of binary Al Si alloys when associated with the Al 2 Cu intermetallic phase, a better galvanic protection is provided for finer Al Cu Si alloy microstructures. It was also reported that the ternary eutectic mixture consisting of Al + Al 2 Cu + Si phases is nobler than the Al-matrix and Al-phase in the eutectic mixture (W.R. Osório, et al. 2007). The present study is an attempt to investigate the corrosion behavior of Al-6Si-0.5Mg alloy containing 1wt%Cu in 0.1M NaCl solution and examined corroded surfaces by optical and scanning electron microscopy. 125

127 II. EXPERIMENTAL Materials Preparation- The Al-6Si-0.5Mg(-1Cu) alloys were prepared by melting Al-7Si-0.3Mg (A356) alloys and adding Al and Cu into the melt. After solidification the alloys were homogenised (500 o C for 24hr), solution treated (540 o C for 2hr) and finally artificially aged (225 o C for 1hr). Then the rectangular samples (30mm x 10mm x 5mm) were prepared for metallographic observation and subsequent electrochemical test. Deionized water and analytical reagent grade sodium chloride (NaCl) were used for the preparation of 0.1M solution. All measurements were carried out at room temperature. Potentiodynamic Polarization Measurements-A computer-controlled Gamry Framework TM Series G 300 and Series G 750 Potentiostat/ Galvanostat/ZRA was used for the electrochemical measurements. There were used saturated calomel as reference, platinum as counter and the sample (10mm x 5mm) as working electrode. Only one 10mm x 5mm surface was exposed to the test solution, the other surfaces being covered with Teflon tape. The system was allowed to establish a steady-state open circuit potential (OCP). The potential range selected was -1 to +1V and measurements were made at a scan rate of 0.50 mv/s. The corrosion current (icorr), corrosion potential (E corr ), pitting corrosion potential (E pit ) and corrosion rate (mpy) were calculated from Tafel curve. The corroded samples were cleaned in distilled water and examined under optical light and scanning electron microscope. Electrochemical Impedance Measurements (EIS)- As in potentiodynamic polarization test, three electrode cell arrangements were also used in EIS. In EIS test, a frequency range of 100 khz to 0.2 Hz using a 5mV amplitude sinusoidal voltage. The impedance spectra were collected, fitting the experimental results to an equivalent circuit using the Echem Analyst TM data analysis software and evaluating the solution resistance (R s ), charge transfer resistance or polarization resistance(r ct ) and double layer capacitance(c p ) of the alloys. steady state. The steady state OCP of Cu free alloy (Alloy-1) is V and it is the higher negative OCP value between the alloys under investigation. The OCP values mainly depend on the chemical compositions and thermal history of the alloys. The data obtained were simulated and the equivalent circuit that best fitted to the experimental data is shown in Figure 1. Rs, Rct and Cp are the solution resistance, charge transfer resistance and electrical double layer capacitance respectively. Figure 2 shows the Nyquist diagrams of the Al-6Si- 0.5Mg (-1Cu) alloys in 0.1M NaCl in de-mineralized (DM) water. In Nyquist diagrams, the imaginary component of the impedance (Z") against real part (Z') is obtained in the form of capacitive-resistive semicircle for each sample. Figure 1. Electrical equivalent circuit used for fitting of the impedance data of Al-6Si 0.5Mg (-1Cu) alloys. At high frequencies, only the very mobile ions in solution are excited so that the solution resistance (R s ) can be assessed. At lower intermediate frequencies, capacitive charging of the solid-liquid interface occurs. The capacitive value C p can provide very important information about oxide properties when passivation or thicker oxides are formed on the surface. At low frequency, the capacitive charging disappears because the charge transfer of electrochemical reaction can occur and this measured value of the resistance corresponds directly to the corrosion rate. For this reason, this low frequency impedance value is referred to as polarization or charge transfer resistance (R ct ). Impedance Measurements III. RESULTS AND DISCUSSION Table I. Impedance test results Alloy Compositions R s(ω) R ct(kω) C p(µf) OCP (V/SCE) Al-6Si-0.5Mg Al-6Si-0.5Mg- 1Cu The open circuit potential (OCP) with exposure time of aged Al-6Si-0.5Mg(-1Cu) alloys in 0.1M NaCl solution is shown in Table1. Large fluctuations in open circuit potential for the alloys were seen during the time of 100s exposure. After a period of exposure the OCP fluctuation decreased and reached Figure 2. Nyquist plots for the peakaged Al-6Si- 0.5Mg (-1Cu) alloys in 0.1M NaCl solution. The solution resistance (R s ) of the alloys varies from 40-44Ω (Table I) and these values are very similar to each other. The R s values are negligible with respect to R ct and the electrolyte behaves as a 126

128 good ionic conductor. Impedance measurements showed that adding 1wt% Cu in the Al-6Si-0.5Mg alloy increases the charge transfer resistance (R ct ). For the Al-6Si-0.5Mg alloy, the charge transfer resistance (R ct ) value is 15.57kΩ, and this is increased to 27.13kΩ with the addition of 1wt% Cu to the Al-6Si-0.5Mg alloy. The increase in the charge transfer resistance indicates an increase in the corrosion resistance of the alloys with Cu addition. The double layer capacitance (C p ) of the Al-6Si- 0.5Mg alloy is 1.259µF which is lower than the Al- 6Si-0.5Mg-1Cu(3.219 µf) alloy. Potentiodynamic Polarization Measurements Potentiodynamic polarization curves of Al-6Si- 0.5Mg(-1Cu) alloys in 0.1M NaCl solution are shown in Figure 3. Anodic current density of Al-6Si-0.5Mg alloy decreased with Cu addition. This is caused by the slowing of the anodic reaction of Al-6Si-0.5Mg- 1Cu alloy. The addition of Cu caused the formation of micro-galvanic cells in α-aluminum matrix. The different intermetallic compounds (like Mg 2 Si, Al 2 Cu etc.) can lead to the formation of micro-galvanic cells because of the difference of corrosion potential between intermetallics and α-aluminum matrix. 6Si-0.5Mg alloy and the corresponding corrosion rate decreases for the alloy (2.474mpy). IV. MICROSTRUCTURAL INVESTIGATION There was evidence of corrosion products and pits in the examined microstructures (Figures 4-5). It is probable that the pits are formed by the intermetallics dropping out from the surface due to the dissolution of the surrounding matrix. However, it is also possible that the pits are caused by selective dissolution of the intermetallic/or particles of the second phase precipitates. The forms of corrosion in the studied Al-6Si-0.5Mg (-1Cu) alloys are slightly uniform and predominantly pitting corrosion. Table II. Potentiodynamic polarization test results Alloy Compositions Al-6Si- 0.5Mg Al-6Si- 0.5Mg-1Cu I corr E corr E pit (µa) (mv) (mv) Corrosion rate(mpy) a Figure 3. Potentiodynamic polarization curves of aged Al-6Si-0.5Mg (-1Cu) alloys in 0.1M NaCl solution. The addition of Cu increased the corrosion potential of a number of Al-Si-Cu alloys (M. Park 2005). For the Cu free Al-6Si-0.5Mg alloy corrosion potential is -764mV, which is the higher negative potential between the investigated alloys. With increasing Cu, the corrosion potential of the alloy shifted towards more positive values. Pitting potential (Epit) of Cu content alloy also shifted towards more positive value (-480mV to -370mV). Potentiodynamic tests showed that adding Cu in the Al-6Si-0.5Mg alloy decreases the corrosion current (I corr ). For Al-6Si-0.5Mg alloy, the I corr value in 0.1M NaCl solution is 6.3µA, and this decreased to 2.950µA with the addition of 1wt% Cu into the Al- b Figure 4. (a)olm and (b) SEM images show ascorroded T6 aged Al-6Si-0.5Mg alloy in 0.1M NaCl solution. The peakaged Al-6Si-0.5Mg alloy exhibited pits on their surface (Figure 4), which apparently had nucleated randomly. Conversely, the exposed surface of the alloys exhibited a corrosion product covering the surface after polarization. There are more and 127

129 severe pits in Al-6Si-0.5Mg alloy compared to Al- 6Si-0.5Mg-1Cu alloy. All the micrographs (Figures 4-5) also showed that there was no corrosion in the fragmented and modified Al-Si eutectics. a b Figure 5. (a)olm and (b) SEM images show ascorroded T6 aged Al-6Si-0.5Mg-1Cu alloy in 0.1M NaCl solution V. CONCLUSIONS The EIS tests have shown that the addition of 1wt%Cu into Al-6Si-0.5Mg alloy increase the corrosion resistance in NaCl. The linear polarization and Tafel extrapolation plot show that the corrosion current (I corr ) and corrosion rate (mpy) decrease with the addition of 1wt%Cu into Al-6Si-0.5Mg alloy. The forms of corrosion in the studied Al-6Si-0.5Mg (-1Cu) alloys are pitting corrosion as obtained from the microstructures study with pits observations. 128

130 IC 4 ME , 24~25 March, 2016 Effects of inclusions on the mechanical properties of structural steel reinforced bars Abul Hossain 1*, Fahmida Gulshan 2, and ASW Kurny 2 1 MSTE Plant (KTL), Sylhet Gas Fields Ltd., Golapgonj, Sylhet Department of Materials and Metallurgical Engineering, BUET, Dhaka * ah_buetmmesgfl@live.com Abstract Inclusions content of pencil ingots and continuously cast billets (both ladle refined and unrefined) produced from induction melted liquid steel have been determined. It has been seen that billets produced from unrefined or improperly refined melts contain higher amounts of slag, inclusion and inhomogeneity in the microstructure while refining in ladle refining furnace of induction melted assorted scrap gives fairly clean and refined liquid steel. Metallographic study of the reinforcing bars produced from properly refined continuously cast billets show uniform grain size, no heterogeneity in the microstructure, little or no slag and little inclusions and better mechanical properties. Keywords Inclusions, reinforced bars, ladle refining, tensile properties. I. INTRODUCTION Most of the steel bars used for reinforcement of concrete are produced by melting iron and steel scrap in induction furnaces. Because of the nature of induction heating, the melt in the furnace is continuously stirred and complete separation of inclusions cannot take place. An induction furnace is a melting unit. Very little, if any, refining takes place in an induction furnace. Thus the quality of steel produced in an induction furnace is directly related to the quality of scrap and other raw material. Only limited number of steelmaking units use ladle refining furnaces for enhancing the quality of steel produced. Ladle metallurgy incorporates inert gas stirring for homogenization of temperature and composition of the steel in the ladle as well as for floatation of inclusions. For continuous casting, inert gas stirring of steel in the ladle has become a standard practice to improve the quality of steel (M. S. Millman, 1999 and C. Amit et al., 1997) In recent years, new clean and ultra-clean steels have been developed and commercialized by steel producers around the world, thereby responding to the current and future market demands of steel having significantly improved mechanical properties (N. Ånmark et al., 2015). Steel cleanness is an important and growing research area driven by the demands to produce high quality steel. Inclusion content in steel is an important criterion to assess clean steel. Inclusions are generally removed by reacting with slag. This is primarily achieved by optimizing the process conditions to promote contact and reaction between the inclusion and slag (B. Deo et al., 1993) Non-metallic inclusions found in steel can affect its performance characteristics. Their impact depends not only on their quality, but also, among others, on their size and distribution in the steel volume (T. Lipiński et al., 2014). The quantity and quality of non-metallic inclusions is determined mostly by the steel melting technology. Out-of furnace treatment regimes are also introduced to minimize the quantity of non-metallic inclusions. The quantity of non-metallic inclusions in steel is relatively low, nevertheless, they have a significant impact on the structure, technological and strength parameters of the resulting alloy. The distribution of inclusions is an equally important factor (T. Lis, 1999, J. Wypartowicz et al., 2006, T. Lis, 2002 and M. Fernandes et al., 2003). As inclusions inside the material act like stress raisers in the matrix the crack initiation and propagation during internal fracture is affected by the inclusion size and the applied stress amplitude. The size of inclusions does undoubtedly play an important role; the detrimental effect of inclusions depends not only on their size but also on their chemical composition. Their results show for instance that titanium nitride inclusions are about as harmful as inclusions containing aluminum oxide, although the latter are several times larger. Generally, inclusions of all types become more detrimental with increasing size. One reason why the types of the inclusions, that is their chemical composition, also influence the degree of harmfulness (J. Monnot et al., 1988) For efficient removal from the steel, the inclusions must attach to and dissolve in the slag phase. Research on inclusion removal in steel refining is principally divided into categories of flotation of inclusion to the steel/slag interface, modification to improve reactivity/separation with the slag phase and dissolution in the slag phase (B. J. Monaghan et al., 2005, Y. Miki et al., 1992, K. H. Sandhage et al., 1991, M. Valdez et al., 2001, B. J. Monaghan et al and M. Valdez et al. 2002) This work was aimed at determining the extent of variation of inclusion content with the liquid metal processing route and the effects of inclusions on the mechanical properties of structural steel. II. MATERIALS AND EXPERIMENTAL PROCEDURE This study is based on analysis of samples (Table I) collected from the process stream of a steel plant in 129

131 IC 4 ME , 24~25 March, 2016 Bangladesh. Care was always exercised to collect a representative sample. The plant concerned produces different grades of deformed bars from ingots produced in metal moulds (pencil ingots) and also from billets produced by a continuous casting machine (Figure 1). The plant has a ladle refining furnace and samples of continuously cast billets, both ladles refined and not refined, were collected. Ladle refining Molten Steel Continuous Casting Continuous Casting Ingot Mould Quality Standard Billet (QB) Normal Standard Billet (NB) Pencil Ingot (PI) Hot Rolling Hot Rolling Hot Rolling Reinforced Bar Reinforced Bar Reinforced Bar Figure 1. Flow diagram of different routes of processing reinforcing bars. In this presentation, the term PI - refers to the ingots produced by bottom pouring of vertical metal moulds. The molten steel is tapped directly into a tundish inserted between the furnace and the ingot mould to ensure a uniform metal stream. The tundish has its own nozzle to regulate the flow. The tundish is filled to a certain depth which is maintained at the same level throughout the teeming period to eliminate the flow variation due to the varying ferro-static head in the tundish. NB - Normal standard billets refer to the billets produced in the continuous casting machine from molten steel that has not been refined in the ladle furnace after induction melting and ferroalloy addition. The liquid metal was poured in a ladle and was taken up to the continuous casting machine for casting 100 mm x 100 mm x 6000 mm steel billets. QB - quality standard billets - refer to the billets produced in the continuous casting machine from molten steel that has been subjected to ladle refining. In the ladle refining furnace nitrogen gas was blown through a porous plug in the bottom of the furnace. Bottom blowing accelerates slag metal reactions, improves yield of iron and allows higher retention of manganese in steel. More efficient slag removal is achieved, non-metallic inclusions of a certain size are removed and a uniform distribution of inclusions is obtained. For two heats QB-1, QB-2 purging pressure and time were 3.30, 3.00 bars and 33, 35 minutes respectively. A number of pencil ingots of different heats, normal standard billets and quality standard billets were chemically analyzed. The chemical compositions of the samples were examined and the heats chosen for further work was based on carbon equivalent factor. Inclusion content was estimated by the method of direct counting and measurement of inclusions on metallographic samples (N. Ånmark et al., 2015, B. Deo et al., 1993 and T. Lipiński et al., 2014). The specimens for metallographic assessment of inclusions were prepared by using standard techniques. Extreme care was used to ensure that the inclusions are retained. It was not always possible to retain all the inclusions. However, this did not affect the result because the cavities in the samples were counted as inclusions (T. Lis, 1999). Both longitudinal and transverse sections were examined. The locations of inclusion analysis were selected randomly. Standard tensile test specimens were tested with a universal tensile testing machine to obtain data on yield strength(ys), ultimate tensile strength (UTS), percentage of elongation (%EL), and percentage of reduction in area (%RA). Table I. Chemical compositions of the ingots/billets Heat No Casting Type %C %Mn Carbon Equivalent PI-1 Pencil PI-2 Ingot NB-1 Normal Standard NB-2 Billet QB-1 Quality Standard QB-2 Billet III. RESULTS AND DISCUSSION The results of inclusion contents in two pencil ingots (mid-height), two normal standard billets (middle casting) and two quality standard billets (middle casting) are listed in Table II. Heat No Table II. Inclusion distributions in ingots/billets Inclusions size distribution (µm) >40 Total No. Inclusions /cm 2 PI PI NB NB QB QB IV. MECHANICAL PROPERTIES Tensile properties yield strength, ultimate tensile strength, %elongation and %reduction in area of the reinforced bars are presented in Figure 2 and Figure 3.Yield strength and ultimate tensile strength of the reinforcing bars produced from unrefined pencil ingot is lower than normal and quality standard billets. The reinforcing bars from quality standard billet shows highest yield strength due to the lower number and smaller sizes inclusions content in the steel bars. 130

132 IC 4 ME , 24~25 March, 2016 During ladle refining operation nitrogen gas was purged through the purging plug at the bottom of the ladle and the molten metal picked up nitrogen. In the case of pencil ingot or normal standard billet processing, the maximum nitrogen content was 0.02 per cent but in quality standard billet nitrogen content increased to 0.03 percent. This higher content of nitrogen is also another reason for higher yield strength and ultimate tensile strength a a b b Figure. 3 (a) Percentage of elongation and (b) reduction in area of the reinforcing bars. Figure 2. Variation of (a) yield strength and (b) ultimate tensile strength of the reinforcing bars. All the reinforcing bars from pencil ingot show a low value of %elongation and %reduction in area (Figure 3). These may be due to the presence of slags and larger size in higher number of non-metallic inclusions. It can be seen that the strength of the reinforcing bars produced from quality standard billets is higher than other two types of bars. It was experimentally seen that lesser amount of slags and inclusions are present in quality standard billets. Some inclusions contained in pencil ingots were larger than 100µ. In the normal standard billets the size of most of the inclusions are lower than those in pencil ingots but larger than those in quality standard billets. For this reason the reinforcing bars from quality standard billets show better %elongation and %reduction in area and also better toughness than the bars produced from pencil ingots or normal standard billets. During rolling the plastic inclusions change shape and therefore their size, whilst the harder types of inclusions are not affected by reduction. Inclusions in the form of thin films located on grain boundaries are especially dangerous for steel quality. These are usually low-melting oxysulphide inclusions precipitated in liquid state during steel solidification. They weaken the intergranular bonds, especially at elevated temperatures (red shortness). Inclusion particles with sharp edges may be quite dangerous; these are usually high melting inclusions (J. Wypartowicz et al., 2006, T. Lis, 2002, M. Fernandes et al. 2003). Rounded off inclusion particles are considered less harmful. They are formed by substances which have low melting point and are poorly wettable by the metal. A low concentration of inclusions in steel is not, by itself, the guarantee of high quality because the inclusions may be concentrated in particular places of an ingot or billet. V. CONCLUSIONS The following conclusions can be drawn from this study. Pencil ingots contain different (higher in number and larger in sized) amounts of inclusions and give uncertain physical and mechanical properties. Normal standard billets were unrefined and contain slags, inclusions and inhomogeneities in structure. Final products from these billets give inferior physical properties. 131

133 IC 4 ME , 24~25 March, 2016 Quality standard billets produced through proper refining give clean and refined liquid steels or a little slags and inclusions. Finished products from these billets have better mechanical properties. 132

134 Utilizing Solar Energy in the Filling Stations of Bangladesh: Technical and Economical Representation Mohammad Jalal Uddin, Muhammad Sifatul Alam Chowdhury *, Md. Ridwanul Karim, Md. Arman Uddin and Md. Bakiuzzaman Department of Electrical and Electronic Engineering International Islamic University Chittagong Chittagong, Bangladesh * m.sifatul.alam@gmail.com Abstract To achieve high economic growth as envisaged in the 6th five-year plan, Government of Bangladesh has prioritized power sector development in a sustained manner. Equally, it is realized that energy is the most key ingredient to alleviate poverty and to improve socio-economic development and uplift human lifestyle. By this time solar technology is convenient, proven and well accepted in Bangladesh. It is environment friendly and price of solar panel is decreasing day by day. Natural gas reserve in the country is decreasing gradually. While establishing new power plant it is equally important to consider the enviromental issues but bangladesh is neglecting this due to meet high demand of power. To produce power, developed countries are focusing on increased use of renewable technologies. Sun energy directly converted into electrical energy in case of solar power system technology and to facilate the high demand of power it is an impressive technology for a developing country like Bangladesh. This research considered a fuel station named M/s Alhas younus & Co. filling station to represent the detail technical and economical calculation while utilizing solar energy as main power source. Keywords Solar energy; Photovoltaic panel; Renewable energy; I. INTRODUCTION Initially the power sector of Bangladesh were dominated by the natural gas having the participation of more than 90% for the year The domination of natural gas continues to the consecutive years and it was 80.37% for the year 2012.Seems that the participation of natural gas reduced 9.63% but 80.37% was also a huge dependency [1]. Government of Bangladesh is planning to reduce the participation of natural gas in power generation and estimated to 52% within 2016 and 20% within Government is forced to do this because of the depletion of gas reserve. Besides this the diversification of primary fuel sources for power generation are equally done and the primary fuel source is shifting to coal and liquid oil by reducing the dependency on natural gas. However the generation cost of power from liquid fuel powered power plant is higher than the average cost of generation. Though the difference is decreasing day by day but the cost of renewable energy system is still more than that of conventional system [2]. Bangladesh is a recipient of sufficient sunshine round the year. Solar energy can be produced in all parts of the country. It is environment friendly and price of solar panel is decreasing day by day. Under the circumstances, Bangladesh Government has under taken 500 MW solar mission. So, this research proposed solar powered fuel station initially in Chittagong to achieve the goal of 500 MW solar mission and to overcome the power crisis in the country. Main reason behind selecting fuel stations is that, fuel stations maintain large space especially in the rooftop and the open area of the station to move vehicle where photovoltaic panel can be set up in a large scale. Besides this, the economic stability of the fuel stations are typically better than general consumers which made them one of the most convenient consumer of the solar power system. Large scale implementation of solar power system in the filling station of Bangladesh will greatly reduce dependency on natural resources specifically on natural gas. This research briefly discussed the technical and economic condition if solar power system is implemented in the fuel stations and to conduct the research in a realistic way a fuel station from Chittagong city is considered. II. MATHEMATICAL MODELLING This research considered a filling station from Chittagong city named M/s Alhas younus & Co. filling station having, Total area in the rooftop= m 2 Total area including rooftop = m 2 We can set up 103 solar panel of 250wp on the roof top area and 268 solar panel on the total area in this filling station. But we are designing for the filling station that power consumption. The filling station consumption power is 50 KWh in December. But we 133

135 are calculating for 65 KWh in a month due to December is the winter season. Power produced by any panel per month = ( ) III. RADIATION, CLEARNESS INDEX & SUN SHINE Yearly solar radiation of Chittagong when Latitude: 22 29'53.52" & Longitude: 91 43'10.56" TABLE I. CLEARNESS INDEX, SOLAR RADIATION & SUN SHINE OF CHITTAGONG [3] Month CI Radiation Daily Bright Sunshine(hr) January February March April May June July August September October November December Average IV. TECHNICAL REPRESENTATION Design of solar panel Per day power consumption=2.17kwh Energy to load (KWh/day) = KW p * Radiation * System efficiency (ɳ p ) Or 2.17= KW p* 4.2*0.55 Or KW p = KW p or 940W p This research considered 250w p solar panel available in the market. So required number of solar panel= Calculation of battery Battery sizing watt hour rating:- DOA* Energy to load (Wh) = Purchase capacity of battery (Wh)* DOD Or 3*2170=Wh*0.7 Or Wh=9300 Wh capacity=ah capacity*terminal voltage Or, 9300 =Ah capacity*12 Or, Ah capacity= 775 This research considered 130Ah solar battery from market. So number of battery= Design of Inverter Power rating of Inverter=power required by the load /(Load factor*inverter efficiency) Or Inverter power rating =1000/(0.8*0.9) Or Inverter power rating VA Charge controller design Cost calculation I=P/V I=2170/12 I=180.83Amp Cost of solar panel=4*250*62=62,000 Taka Cost of battery=14000*6=84,000*5= Taka Cost of charge controller design=5000*25=125000taka Cost of inverter=15000*25=375000taka Cost of miscellaneous equipment s=20000 Taka Total cost=10, 02,000Taka A. Technical representation for rooftop area To design grid connection now we analysis total roof top area. There are 103 solar panel with 250w p rating can be set up in the rooftop of this filling station. So power produce from solar panel=103*250=25.75kw Design of Inverter Inverter power rating =power required by the load /(Load factor*inverter efficiency) Or Inverter power rating =25750/(0.8*0.9) Or Inverter power rating = VA Cost calculation Cost of solar panel=103*250*62= Taka Cost of Inverter=40000*25= Taka Cost of Wiring=5000 Taka Cost of structure=10000 Taka Cost of miscellaneous equipment s=20000taka Total cost= taka B. Technical representation for total area To design grid connection now we analysis for total area. There are 268 solar panel with 250w p rating can be set up in the total area of this filling station. So power produce from solar panels =268*250=67KW Design of inverter Inverter Power rating =power required by the load /(Load factor*inverter efficiency) Or Inverter power rating =67000/(0.8*0.9) Or Inverter power rating = VA 134

136 Cost calculation Solar panel=268*250*62= Taka Inverter=60000*25= Taka Wiring=10000Taka Structure=500000Taka Miscellaneous=30000Taka Total cost= taka Annual income = Taka Payback period = cost of the system/annual income = / =2.77 years B. Economical representation for total area V. ECONOMICAL REPRESENTATION A. Economical representation for rooftop area Fig. 3. Daily generation statistics Fig. 1. Daily generation statitics Fig. 2. Income Monthly generation statistics While considering, 1 KWh=12 Taka Expected income/year =12* = Taka While considering, 1 KWh=15 Taka Expected income/year =15* = Taka Payback period Considering 1 kwh = 12 Taka Total cost of the system = Taka Annual income = Taka Payback period = cost of the system/annual income = / =3.45 years Considering 1 kwh = 15Taka Total cost of the system = Taka Income Fig. 4. Monthly generation statistics While considering, 1 KWh=12 Taka Expected income/year =12* = Taka While considering, 1 KWh=15 Taka Expected income/year =15* = Taka Payback period While considering 1 kwh =12 Taka Approximate cost to set up the system = Taka Annual income = Taka Payback time = cost of the system/annual income = / =3.13years While considering 1 kwh = 15 Taka Approximate cost to set up the system = Taka Annual income = Taka Payback time = cost of the system/annual income = / =2.5 years VI. CURRENT ENERGY SECTOR OF BANGLADESH Among the various natural resources natural gas is dominating in case of power generation due to the limited availability of the other natural resources. According to the Fig. 5, participation of imported power is higher than the only hydro power plant also 135

137 referred as renewable power plant which is 500 MW with 4.12% share of total installed capacity. Heavy fuel oil participating in the power sector by adding 2565 MW which is equal to 21.15% of total installed capacity. High speed diesel contributing in the power sector by adding 956 MW with 7.88% of total installed capacity. This indicates that liquid fuel is playing a vital role in the power sector of Bangladesh by contributing 29.03% of total installed capacity. Except the power from hydro power plant, all other primary source used to produce power consists negative economic and environmental impact. Fig. 5. Total installed capacity of Bangladesh [8] The generation cost of power from the liquid fuel based power plant is higher than the other natural resources and renewable power. Liquid fuel fuels are mainly used in the quick rental power plants (QRPP) and rental power plants (RPP) and the participation of liquid fuel in power sector of Bangladesh is pretty impressive which 29.78% of total derated capacity (including both HFO and HSD). The per unit fixed cost in the case of gas based generation units for the QRPPs is nearly 2.5 times higher than that of the rentals but, for the furnace oil based units the per unit cost of the rentals is more than 2 times that of the QRPPs [9]. As Bangladesh is not enriched with the natural resources like liquid fuel so it s not a sustainable way to increase the use of liquid fuel in the power generation. Large scale implementation of renewable energy technologies more specifically solar energy technology can play significant role to reduce the dependency on the natural resources based power plants. CONCLUSION Solar energy is getting popular both in city and rural areas due its low maintenance cost and this technology can fed national grid if the all the fuel station can be empowered by solar power system. By implementing solar power system both the owner of the fuel stations and the government will be benefited as government do not requires to invest. In this research a brief projection of economic perspective of implementing solar power system are represented. As the installation will be done by the owners so no extra investment is requires from the government. Besides this, a large number of liquid fuel and natural gas will be saved and people will get power from grid with a lower cost. Besides the economic projection, the technical obstacle and perspective are focused deeply including the design of the solar panel, battery calculation, inverter design, and charge controller design. The major advantage of fuel stations as mentioned earlier is that they maintain large area in the rooftop. Except the rooftop area, the overall area of the fuel stations are more than any conventional infrastructure. Also the financial condition of this fuel stations are comparatively better than regular consumer. After extensive study and research, fuel stations are found the most convenient medium to implement the solar power system in large scale. REFERENCES [1] Power generation ststistics can be accesesd at: t&view=article&id=150&itemid=16 [2] Details about 500 MW solar program can be accessed at: 7d42b92a-8f a0a8- b38c d/ [3] Islam, M. M., Das, N. K., Ghosh, S., & Dey, M. Design and implementation of cost effective smart solar charge station. In Strategic Technology (IFOST), th International Forum on (pp ). IEEE. [4] Angelis-Dimakis, A., Biberacher, M., Dominguez, J., Fiorese, G., Gadocha, S., Gnansounou, E.,... & Robba, M. Methods and tools to evaluate the availability of renewable energy sources. Renewable and Sustainable Energy Reviews, 15(2), [5] Details about solar electricity cost can be accesed at: [6] James, P., & Dunlop, P. E. (1997). Batteries and Charge Control in Stand-Alone Photovoltaic Systems. Florida Solar Energy Center. [7] Mateus, T., & Oliveira, A. C. Energy and economic analysis of an integrated solar absorption cooling and heating system in different building types and climates. Applied Energy, 86(6), [8] Detail about installed capacity and derated capacity can be accessed at: t&view=article&id=150&itemid=16 [9] Mujeri, M. K., & Chowdhury, T. T. Quick rental power plants in Bangladesh: an economic appraisal. [10] Nieuwenhout, F. D. J., Van Dijk, A., Lasschuit, P. E., Van Roekel, G., Van Dijk, V. A. P., Hirsch, D.,... & Wade, H. Experience with solar home systems in developing countries: a review. Progress in Photovoltaics: Research and Applications, 9(6),

138 Conversion of Prawn Shell Waste into Value Added Products for Textile Finishes Md. Mofakkharul Islam and Md. Ibrahim H. Mondal Polymer and Textile Research Lab, Department of Applied Chemistry and Chemical Engineering, Rajshahi University, Rajshahi 6205, Bangladesh. Abstract-- In this paper, synthesis, characterization and application of chitosan (Ch) and its functional derivatives from prawn shell waste to enhance the effectiveness of the natural biopolymer for intensified textile and other uses, for the eco-friendly modified cotton fibres, to avoid chemical modifier has been reported. The N-octyl chitosan (NOCh) and carboxymethyl chitosan (CMCh) was prepared through reductive amination and carboxymethylation of chitosan respectively. The molecular weight, degree of deacetylation (DDA) and ash content of prepared chitosan were 1,39,958 Da, 85% and 2.33% respectively. The moisture content, water holding capacity and total nitrogen content were above 10%, 450% and 6.5% respectively. CMCh had an average degree of substitution The reductive amination, carboxymethylation and modification of cotton were confirmed by FTIR spectroscopy. The thermal behavior of chitin, chitosan, cotton and, NOCh and CMCh treated cotton were investigated by thermogravimetric analysis. Surface morphology of the modified fibres were carried out by SEM. As the modified fibres showed good dyeability and colour fastness as well as other properties, the chitosan derivatives as textile modifier would be helpful to avoid synthetic petroleumbased chemical modifier. Keywords-- Prawn shell waste, N-octyl chitosan, Carboxymethyl chitosan, Cotton fibre, Modifier, Textile finish. 1. INTRODUCTION Seafood processing waste such as prawn shell waste, crabs, krill etc. are the potentially rich source of several useful products including chitin [1] and has long been generated in large tonnages worldwide. Chitin, N-acetyl glucosamine is economical and is the second most abundant bio-waste material after cellulose. Literature shows the annual worldwide chitin production from arthropods (e.g. crustaceans and insects), molluscs (e.g. squid and cuttlefish), fungi etc. is estimated at about tons [2]. A steady supply of chitinous waste materials from the seafood processing industry has been the major source of commercial products such as chitin, chitosan [3] and its ecofriendly valuable functional derivatives for textiles. Chitosan is the deacetylated product of chitin, soluble in dilute acetic acid and insoluble in water, organic solvents and aqueous bases. Due to the limited solubility of chitosan, NOCh and CMCh derivative were prepared, to enhance the usability of the bio-polymer. A few researchers have been attempted to deal with this solubility problem by preparing water-soluble derivatives of chitosan, which are safe and suitable for human use [4]. Cotton is the natural fibre have certain unfavorable properties such as, stiffness, low elasticity, susceptibility toward sunlight and microbial attacks etc, which causes the hindrances on their use. For this, modification has been done using chemicals [5], but there are lack of sincerity about ecofriendliness and waste management. The purpose was to alter the petroleum based chemical modifier with bio-degradable and bio-compatible natural modifier for improve the resultant physical properties of textile products. A few researchers have also prepared chitosan from prawns crustaceans and crabs [6]. The strategy explored in this research project was the synthesis of chitosan, NOCh and CMCh from prawn shell wastes for their ecofriendly and biodegradability along with to use as value added products in textile sector. 2. EXPERIMENTALS 2.1 Materials: Prawn shells were collected from Mongla, Bangladesh those are wastes from prawn processing area. Cotton fibres were collected from Keya Spinning Mill Ltd., Dhaka, Bangladesh. Acetic acid, sodium hydroxide, ethanol, hydrochloric acid, acetone, octanal, sodium borohydride etc. were purchased from Merck (Germany), sodium chlorite and sodium carbonate were bought from BDH (England). The detergent was bought from Kohinoor Chemical Co. Ltd., Bangladesh. Reactive dyes were collected from Sigma Chemical Co. USA. 2.2 Methods Processing of Prawn shell waste to chitin and conversion of chitosan: The collected prawn shell was washed with hot water and dried at 105 o C for 72h in an oven, and the dried shell was ground to different meshes. The deproteinization and demineralization of raw chitin was carried out using 1M NaOH and 1M HCl by maintaining solid:liquor ratio of 1:16 at 100 o C for 4h respectively. Prepared chitin is an intermediate product of chitosan. Chitosan was obtained through deacetylation of chitin using NaOH in presence of ethanol (chitin:naoh=1:30 w/w) at suitable temperature for 4h in a reflux system. The resultant solid was washed and dried in an oven at 50 o C for 20 h that was known as chitosan Preparation of N-Octyl chitosan: NOCh was prepared by introducing an octyl group to NH 2 on C2 of glucosamine unit through reductive amination of 137

139 chitosan in presence of alkaline sodium borohydride according to the method of Bobu et al. [7] Preparation and purification of CMCh: The procedure for the carboxymethylation of chitosan and purification was carried out according to Caraschi and Campana-Filho [8] Preparation of fibre: Cotton fibres were treated with 0.2% Na 2 CO 3 solution at 75 o C for 30 min in the ratio of 1:50 [5]. The fibre was then thoroughly washed with distilled water, dried at 105 o C Treatment of cotton fibre with chitosan and N-Octyl chitosan: The washed cotton fibres were dipped into dissolved chitosan in aq. acetic acid of different concentration at 60 o C for 1h. The ph of the solution was maintained at by using 0.2M acetic acid [9]. The treated fibres were washed with distilled water and subsequently dried in hot air at 60 C till constant weight Treatment of cotton fibres with CMCh: 12.5g fibres were soaked into 500 ml solution containing 20 mg CMCh/g fibre and 0.05M CaCl 2 at ph 7-8 at temperature 80 o C for 2h. After this treatment the fibres were washed with deionized water and ionexchanged to their H-form or Na-form as described by Fras et al. [10] Exhaust dyeing of modified cotton fibres: Dyeing was carried out using 0.3% reactive dye (on the weight of fibre) in a dyeing machine (DYSIN, Taiwan, China) included plastic stoppered conical flasks Determination of the amount of residual dye in the exhausted baths: The concentration of dye in the exhausted dye bath was calculated with the help of corresponding calibration curve. The optical density of the exhausted liquors was measured by a colorimeter (Type-S104, No-221, Spectrophotometer, WPA Linton Cambridge, UK). The exhaustion of dye was determined by using the following equation: Exhaustion of dye, % = Where, Do and De are the original and exhausted dye bath concentration respectively. 2.3 Characterization of chitosan and its derivative, raw and modified fibres Moleculer weight of chitosan: Four different concentrations 0.2%, 0.4%, 0.6% and 0.8% solution of chitosan in 0.1M acetic acid and 0.2M NaCl solution (1:1, v/v) were prepared and the molecular weight was determined by Ostwald viscometer. From the intrinsic viscosity, the molecular weight of chitosan was calculated by using Mark- Houwink equation [11]: [η]=k(mw) a Where K and a are constant for given solute-solvent system and temperature. The values of K and a were and 0.93 respectively Degree of deacetylation of chitosan: The degree of deacetylation was determined from the carbon/ nitrogen ratio (C/N) according to elemental analysis [12]. It was therefore calculated according to the following equation: DDA = {1-(C/N-50145)/( )} Degree of substitution of CMCh: The degree of substitution of CMCh was determined according to the following formula: DS ; A=V NaOH C NaOH Where, V NaOH and C NaOH are the volume and molarity of aqueous NaOH, respectively, m CMCh is the mass of CMCh and 161 and 58 are the respective molecular weights of glucosamine (chitosan skeleton unit) and a carboxymethyl group Infrared Spectroscopy: FTIR spectra of chitin, chitosan, NOCh, cotton, chitosan treated cotton and NOCh treated cotton were recorded with KBr pellets on Shimadzu IR-8900 spectrophotometer (Shimadzu Kyoto Japan) Thermal Analysis: The experiments were performed using a Seiko-Exstar-6000, TG/DTA-6300 (Seiko Instruments Inc. Japan). The tests were conducted between C under an inert atmosphere (argon). The heating rate and the air flow rate were 10 C/min and 200 ml/min respectively Scanning Electron Microscopy: Scanning electron microscopy (SEM) was performed using a scanning electron microscope (FEI Quanta Inspect, Model: S50, Netherlands) to observe the microstructure and the surface morphology of the treated as well as untreated fibres. 3. RESULTS AND DISCUSSION Prawn shell waste is tightly associated with protein, lipid, pigment and calcium deposits. Preparation of chitosan using prawn shell waste involved three main steps: Deproteinization to remove proteins, demineralization to remove calcium carbonate and calcium phosphate and deacetylation to remove acetyl groups. 3.1 Molecular weight: The physicochemical properties of chitosan vary significantly as a function of its molecular weight and molecular weight distribution. The molecular weight of chitosan was determined by viscometric method and the value was 1,39,958 Da. 3.2 Degree of deacetylation: DDA is the quality measurer factor of chitosan and indicates the 138

140 removed quauntity of acetyl groups from the molecular chain of chitin. The DDA mainly depends on the method of purification and reaction conditions such as temperature, alkaline concentration and reaction time. Obtained DDA was 86%. 3.3 Moisture and ash content: The moisture content of chitosan was determined by gravimetric method and it is 10% due to formation of hydrogen bond among hydroxyl and amino group of chitosan with water molecules. Ash measurement is an indicator of effectiveness of the demineralization step for removal of calcium carbonate and the ash of the prepared chitosan is 2.33%. 3.4 Water holding capacity: The water holding capacity was obtained 450% due to presence of more amino (-NH 2 ) group which was deacetylated part and indication of the higher DDA. 3.5 Nitrogen content: The nitrogen content of prepared chitosan is 6.5% which is greater than chitin. Higher value indicates the higher degree of conversion from chitin to chitosan. 3.6 Infrared spectroscopy: Fig. 1 shows the consequent conversion and spectral evidence of chitin to chitosan, chitosan to NOCh and CMCh, and modification of cotton with Ch, NOCh and CMCh through the arrow mark characteristic peaks respectively. 3.8 Surface morphology: Fig. 3(a-d) shows the SEM micrograph which represents the microporous surface of the untreated cotton fibre. But, the chitosan, NOCh and CMCh treated cotton fibres exhibits smoother surface due to the absorption of chitosan, NOCh and CMCh on the fibre. Fig. 2. Thermograph of (a) chitin (b) chitosan (c) cotton fibre (d) chitosan treated cotton fibre (e) NOCh treated cotton fibre and (f) CMCh treated cotton fibre a b c d Fig. 1. FTIR spectra of (a) Chitin, (b) Chitosan, (c) NOCh, (d) CMCh (e) Cotton, (f) Chitosan treated cotton and (g) NOCh treated cotton (h) CMCh treated cotton 3.7 Thermal analysis: Thermal behaviour of chitin, Ch, cotton, Ch treated cotton, NOCh treated cotton and CMCh treated cotton fibres were examined by thermograms (Fig. 2). From this study thermal stability of that compounds and fibres follows the order, cotton chitin chitosan chitosan treated cotton CMCh treated cotton NOCh treated cotton, which may be happen due to the incorporation of chitosan and its derivative with the cotton fibres. Fig. 3. SEM of (a) Raw cotton fibre; (b) Chitosan treated cotton fibre; (c) NOCh treated cotton fibre and (d) CMCh treated cotton fibre. 3.9 Dyebath exhaustion: Three reactive dyes such as, Reactive Orange 14, Reactive Brown 10 and Reactron Red were applied on unmodified and modified cotton fibre and they shows different exhaustion behavior in Fig. 4. NOCh treated cotton fibre exhibits comparatively lower dye absorption, 139

141 because of some amino (NH 2 ) group were engaged with octyl groups to form chitosan derivative. Fig. 4. Effect of dye absorption on dyeing of (a) unmodified, (b) chitosan modified (c) NOCh modified and (d). CMCh modified fibre. 4. CONCLUSION Abandoned prawn shell was utilized to prepare chitosan and its derivatives and used as the biodegradable, natural modifiers for the cotton fibres. Expected uniform modifications were confirmed by used instrumental techniques. Due to modification, chitosan and its derivative plays as the bridges between the fibre surfaces and dye molecules and resulting higher reactive dye absorption. They are alternative of chemical modifier and to convert prawn waste as value added products, would enhance the textile performances of the modified cotton. 5. ACKNOWLEDGEMENT The authors would like to acknowledge the Ministry of Education in Bangladesh for funding the project as Higher Education Research Grant in 2014 (Project Ref. No.: (38)/6-35). [5] Md. I. H. Mondal, F.I. Farouqui, R. K. Sheikh, and Md. A. Hoque, Physico-chemical characteristics of jute fiber grafted with nitrile monomer, Cellulose Chemistry and Technology, 41, 23-28, [6] A. Gopalakannan, G. Indra Jasmine, S. A. Shanmugam, G. Sugumar, Application of proteolytic enzyme, papain for the production of chitin and chitosan from shrimp waste, J. Mar. Biol. Assoc. Ind. 42, , [7] E. Bobu, R. Nicu, M. Lupei, F. L. Ciolacu, J. Desbrieres, Synthesis and characterization of N- alkyl chitosan for paper making applications. Cellul. Chem. Technol., 45(9-10), , [8] J. C. Caraschi and S. P. Campana-Filho, Influencia do Grau de Substituicao e da Distribuicao de Substituintessobre as Propriedades de Equilı brio de Carboximetil celuloseem Solucao Aquosa. Polı meros, Ciencia e Tecnologia, 9, 70 77, [9] L.Gerald. and A. Witucki, Silane Primer: Chemistry and application of alkoxysilanes, Journal of Coatings Technology, 65, 57-60, [10] L. Fras, L. S. Johansson, P. Stenius, J. Laine, K. Stana-Kleinschek and V. Ribitsch, Colloids Surfaces A, Phys. Chem. Asp. 260, 101, [11] H. K. No., S. H. Lee, N. Y. Park, S. P. Meyers, Comparison of physicochemical, binding and antibacterial properties of chitosan prepared without and with deproteinization process, J. Agricultural and Food Chemistry 51, , [12] M. R. Kasaai, J. Arul and G. Charlet, Intrinsic Viscosity Molecular Weight Relationship for Chitosan, Journal of Polymer Science Part B Polymer Physics 38, , REFERENCES [1] R. J. L. Meanwell and G. Shama, Production of strep-tomycin from chitin using Streptomyces griseus in bioreactors of different configuration, Bioresource technology, 99(13), , [2] R.N. Tharanathan and F.S. Kittur, Chitin - The undisputed biomolecule of great potential. Critical Reviews in Food Science and Nutrition, 43(1) 61-87, [3] A. F. Hayes, PROCESS: A versatile computational tool for observed variable mediation, moderation, and conditional process modeling [White paper]. Retrieved from public/ process [4] W. S. Perkins, Functional finishes and high performance textiles. Text. Chem. Colorist & Am. Dyest. Rep., 32(4), 24-27,

142 Textile Performance of Functionalized Cotton Fibre with Silane Coupling Agents Md. Khademul Islam and Md. Ibrahim H. Mondal* Polymer and Textile Research Lab, Department of Applied Chemistry and Chemical Engineering, Rajshahi University, Rajshahi 6205, Bangladesh Abstract - Modification of cotton fibre was studied by condensation polymerization with functionalized silane coupling agents like vinyltrimethoxysilane (VTMS) and 3- Glycidoxypropyltriethoxysilane (GTS) in ethanol/water medium. The use of sodium dodecyl sulfate as a surfactant in ethanol/water mixture accelerate the dispersion of VTMS and GTS in ethanol/water mixture which accelerates the hydrolysis, condensation as well as the fibre surface functionalization. The optimized condition of modification for VTMS and GTS was 600% and 400% respectively in ethanol/water mixture (60:40) containing surfactant by maintaining ph 3.5 at 40 o C in the fibre-liquor ratio of 1:40. The modified cotton fibre was subjected to evaluate some of the properties like swelling behavior in different solvents and moisture absorption. It was observed that swelling behavior and moisture absorption of modified cotton fibres were decreased in polar solvents whereas increased in nonpolar solvents. Fourier transform infrared spectroscopy was used to identify the incorporation of silicon containing species, Energy Disperse X-ray analysis determines the quantities of atomic silicon which directly reflects it s valence bond with organic moieties, Scanning electron microscopy and Thermo gravimetric analysis were used to investigate the surface morphology and thermal behavior of the modified fibre respectively. The modification of cotton fibre enhanced the tensile propperties, water repellency, wrinkle recovery and flexibility due to Si-O-Si bond in the chain of trisilanol pendent branch attached through Si-O-C bond between trisilanol and cellulose substrate. 1. INTRODUCTION Cotton, a natural cellulosic fibre, is cheap, renewable, biodegradable and is the most abundant organic raw materials in the world [1,2]. It can be easily transformed into multifarious products affecting every phases of our daily life because of its widespread applications. It is the backbone of the world s textile trade. Cotton contains approximately 95-98% cellulose, which contains three hydroxyl groups per each repeating unit which directly affect durability of cotton fibre. It is important therefore, to improve the physicochemical properties to a view of making the fibre more comfortable with different modes. Many physicochemical modification steps have already done to overcome these properties such as alkaline treatment, acetylation, benzoylation, acrylation, oxidation and isocynation of the natural fibre [3]. The modification of cotton fibre by silane coupling agent has been receiving considerable attention recently. The modified cotton fibre exihibits versatile physico-chemical properties include improved tensile strength, elasticity, swelling properties, wrinkle recovery properties, fastness and thermal stability properties etc. In this study, we present the chemical modification of cotton fibres treated with two organosilane coupling agents, such as vinyltrimethoxysilane (VTMS) and 3- glycidoxypropyltriethoxysilane (GTS) in an ethanol/water system. The effect of various grafting parameters and chemical structure of organosilane on the grafting quantity and treated fibres were characterized using various experimental techniques such as Scanning electron microscopy (SEM), Thermogravimetric analysis (TGA) and Fourier transform infrared spectroscopy (FTIR). The moisture absorption, swelling behavior and tensile strength were also studied for physical characteristics of modified cotton fibre. 2. EXPERIMENTAL 2.1 Materials Cotton fibres were collected from Keya spinning mills Ltd., Dhaka, Bangladesh. Ethanol, sodium carbonate and acetic acid were purchased from Merck Germany. The silane coupling agents vinyltrimethoxysilane (VTMS) and glycidoxypropyltriethoxysilane (GTS) was collected from Aldrich, USA. All other reagents and solvents were commercial products of high purity. 2.2 Methods The fibres were at first washed throughly with 0.2% Na 2 CO 3 solution at 75 C for 30 min in the ratio of 1: 50 [4]. The pretreated fibres (washed) were dipped in ethanol-water mixture (60:40 v/v) containing VTMS and GTS for 1h. The ph of the solution was maintained 3.5 to 4 by using 0.2M acetic acid [5]. The treated fibres were washed with distilled water and subsequently dried in hot air at 60C. 2.3 Evaluation of Physical and Chemical properties Moisture Absorption The moisture absorption of the silane treated and raw cotton fibres was performed at a constant humidity level and calculated using the following formula [4]. 141

143 Moisture absorption, % = Wf Wi 100 Wi Where W i and W f are the weight of the dried samples and the final weight of the sample taken out from the humidity chamber. Swelling Behavior Swelling behavior of the modified and raw cotton fibres was determined immersing in 100 ml of solvent at room temperature for 72 h. The samples were filtered and the excess solvent was removed with the help of filter paper, then the final weight W f was measured. The percent swelling was calculated [4] as: Swelling, % = Wf Wi 100 Wi Where W i and W f are indicates the initial and final weight of the fibres respectively. 2.4 Characterization of Raw and Surface Modified Cotton Fibre Infrared Spectroscopy FTIR spectra of the silane treated and raw fibres were recorded with KBr pellets on shimadzu IR-8900 spectrophotometer (Shimadzu, Kyoto, Japan). Thermal Analysis The experiments were performed using a Seiko- Extar-TG/DTA-6300 (Seiko, Japan). The tests were conducted at a temperature between C under an inert atmosphere (argon). The heating rate and the air flow rate were 10C/min and 200 ml/min. Scanning Electron Microscopy Scanning electron microscopy (SEM) was performed using a scanning electron microscope (FEI Quanta Inspect, Model: S50, Kyoto, Japan) to observe the micro structure, the surface morphology of the treated as well as untreated fibres. 3. RESULTS AND DISCUSSION The percent graft yield was increased with an increase of silane concentration upto 600% for VTMS and 400% for GTS on the basis of fibre. This was happened due to the higher crosslinking reaction between the cellulosic OH group and OH group of the silanol at higher concentration. The rate of conversion of VTMS and GTS into reactive hydroxyl group by hydrolysis of the silanes is directly related to the ethanol/water ratio, ph value and reaction temperature. At 60:40 ratio of ethanol/water, the maximum amount of silane coupling agents are hydrolyzed. The formation of silanol was enhanced by the protonation of alkoxy (OCH 3 and OCH 2 CH 3 ) groups in acidic ph value and that was 3.5 both for VTMS and GTS respectively. At ph value lower than 3.5, the formation of silanol group was insufficient [6]. Temperature is an important parameter for this reaction mechanism because it decreased the activation energy of the reactant which increase molecular collision between the reactant molecules. The graft yield increased with the increase of reaction temperature upto 35 o C and 30C for VTMS and GTS respectively [7]. 3.1 FTIR Studies The FTIR spectra of raw, VTMS and GTS modified cotton fibres are shown in Fig. 1(a-c) respectively. The spectra of the modified cotton fibre are more or less similar to the raw cotton fibre, except the peaks at 761 cm -1 and 897 cm -1 for Si-O-Si symmetry stretch and Si-O-C bond for VTMS modified cotton fibres and at 860 cm -1 and 1207 cm -1 for Si-OH and Si-O-CH 3 [8-10] which confirmed the presence of silicon containing species on the modified cellulosic fibre. Fig. 1. IR spectra of (a) raw cotton, (b) VTMS treated and (c) GTS treated cotton fibres 3.2 Surface Morphology Fig. 2-4 shows the SEM micrograph of unmodified, VTMS modified and GTS modified cotton fibres Fig. 2. SEM of raw cotton fibre 142

144 Fig. 3: SEM of VTMS modified cotton fibre Fig. 5. TGA and DTG of raw cotton fibre Fig. 6. TGA and DTG of GTS modified cotton fibre Fig. 4. SEM of GTS modified cotton fibre respectively. The untreated cotton fibre shows the presence of the large amount of micro pores on its surface. After VTMS and GTS treatment, the cotton fibre surfaces are coated with an outer layer of silane monomer [11] which represents in the Fig. 3 and Fig. 4. The rupture surface of the modified fibre indicates the excess deposition of the silane layer on the cotton fibre surface 3.3. Thermal Analysis Thermal behavior of raw, VTMS and GTS modified cotton fibres were examined by a study of TGA thermogram. Each of the figures represents two thermogram curves namely TGA and DTG. From the Fig. 5-7, it can be seen that the loss in weight is around 69.3% at 388.5C for raw cotton, 39.5.% at o C for VTMS and 55.7% at 384 o C for GTS modified cotton, respectively. From the DTG curve, the rate of decomposition of raw cotton fibre is higher than that of VTMS and GTS modified cotton fibres. Thus, the thermal stability of VTMS and GTS modified fibres are higher than that of unmodified fibres, which may be happen due to the incorporation of silane coupling agents with the cellulosic fibres. Fig. 7. TGA and DTG of VTMS modified cotton fibre 3.4 Physical Properties of Raw and Silane Modified Cotton Fibres Table 1 shows the swelling behavior of raw cotton as well as VTMS and GTS modified cotton fibres in both polar and nonpolar solvents. Swelling ability reflects the relationship between void structures in backbone polymer and size of solvents molecule [12,13]. The raw cotton fibres were exhibited maximum swelling with polar solvents like water and methanol and least swelling with nonpolar solvents like CCI 4. After treating with silane coupling agents, there was a decrease of the swelling in the polar solvents whereas increased in the nonpolar 143

145 solvent because of decreasing the hydrophilic character of raw cotton fibre. The tensile strength of modified cotton fibre was higher than that of raw cotton fibre and these were due to the modification of cotton fibre with VTMS and GTS [14]. The wrinkle recovery angle of modified cotton fabric was higher than that of unmodified cotton fabric for warp and weft directions respectively. Because, the presence of Si-O bond in the functionalized fabric shows high flexibility that recover the wrinkle which exerted on the fabric surface by silane loading [15]. The moisture absorption sites are blocked after incorporation of silane chain through surface modification due to which the less affinity for moisture is observed than the original fibre. 4. CONCLUSION In this work, we have presented the chemical modification of cotton fibres with silane coupling agents. Table 1. Swelling behavior, tensile strength and moisture absorption properties of raw and modified cotton fibres Fibre type Raw cotton Swelling behaviour % H2O CH3OH CCl4 Breaking load, Kg/yarn Tensile strength Tenacity, g/count Elongation % Moisture absorption % Wrinkle recovery degree VTMS modified cotton degree GTS 8.77 modifed degree cotton Maximum weight gain percent is obtained at optimized value of the reaction parameters, such as silane concentration, ph, ethanol-water ratio and temperature. The chemical attachment between silanol and hydroxyl group of cotton fibres were evaluated by FTIR analysis. The modified fibres showed improved physicochemical properties such as tensile properties, moisture absorption, elongation, wrinkle recovery and thermal stability properties than that of the unmodified cotton fibres. This new type of cotton was obtained through modification with silane coupling agents which would enhance the application of garment products, textiles etc. [3] F. Sadov, M. Korchagin and A. Matetsky, Chemical Technology of Fibrous Materials, Mir Publishers, Moscow, 1973, pp. 14, [4] A. S. Singha and V. K.Thakur, Synthesis and characterization of silane treated grewiaoptivafibres, International Journal of Polymer Analysis and Characterization, 14, , 2009a. [5] L. Gerald and A. Witucki, Silane Primer: Chemistry and application of alkoxysilanes, Journal of Coatings Technology, 65, 57-60, [6] C. J. Brinker, Hydrolysis and condensation of silicates: Effects on structure, Journal of Non- Crystalline Solids, 100, 31-50, [7] Md. I. H. Mondal, Grafting of Methyl Acrylate and Methyl methacrylate onto jute fibre: physic- chemical characteristics of the grafted jute, Journal of Engineered Fibres and Fabrics, 8(3), 42-50, [8] B. Smith, Infrared Spectral Interpretation, a systematic approach, CRC Press, Boca Raton, FL,1999. [9] N. A. Rangel-Vazquez and T. Leal-Garcia, Spectroscopy Analysis of Chemical Modification of Cellulose Fibers, Journal of Mexican Chemical Society, 54(4), , [10] J. M. He and Y. D. Huang, Effect of silane coupling agents on interfacial properties of CF/PI composites. Journal of Applied Polymer Science, 106, , [11] A. Rashidi, H.Moussavipourgharbi and M. Mirjalili, Effect of low-temperature plasma treatment modification of cotton and polyester fabrics, Indian Journal of Fibre &Textile Research, 29, 74-78, [12] A. S. Singha, A. Shama and V. K. Thakur, Pressure Induced Graft Co-polymerization of Acrylonitrile onto SaccharumcilliareFibre and Evaluation of some Properties of Grafted Fibres, Journal Bulletin of Material Science, 31(1), 7-13, [13] Singha, A. S., and V. K. Thakur, Morphological, Thermal and Physico-chemical Characterizations of Surface Modified PinusFibres, International Journal of Polymer and Analysis and Characterization, 14(3), , 2009b. [14] International Standard ISO (E). Textilewoven fabrics. Determination of breaking strength and elongation (Strip Method). International Organization for Standardization, Switzerland, 1977 [15] N. Abidi, E. Hequet and S. Tarimala, Functionalization of cotton fabric with vinyltri-methoxysilane, Textile Research Journal, 77, , REFERENCES [1] K. Mohanty, A. Wibowo, M. A. Misra, and L. T. Drzal, Effect of process engineering on the performance of natural fiber reinforced cellulose acetate biocomposites. Composites Part A, 35, , [2] B. Van voorn,h. H. G. Smit, R. J. Sinke and B. de Klerk, Natural fibre reinforced sheet moulding compound, Composites Part A: Applied Science and Manufacturing, 32, ,

146 Synthesis and Characterization of Hydrogels from Cellulosic Materials for Green Adsorbent Md. Obaidul Haque and Md. Ibrahim H. Mondal Polymer and Textile Research Lab, Department of Applied Chemistry and Chemical Engineering, Rajshahi University, Rajshahi-6205, Bangladesh. Abstract - Hydrogels with biodegradable property is highly appreciable and basic expectation at present time. Copolymerization treatment of cellulose and cellulose derivatives can play a vital role in producing green hydrogels and improving the absorption performance of many absorbent products. Hydrogels are polymeric materials and due to hydrophilic functional groups in their structure capable of holding large amount of water compared to its body mass. Crosslinking is one of the simplest reactions used to improve the physical properties of cellulose and cellulose derivatives. In this work an innovative cellulose based environment friendly hydrogel was experimentally synthesized as an alternative to acrylate based synthetic hydrogels for personal care product and other absorption purposes. The cellulose-based hydrogels were prepared by free radical graft copolymerization reaction of cotton with acrylic acid and acrylamide using N,N-methylne-bisacrylamide as a crosslinker in the presence of initiator potassium persulphate system. The maximum water absorption capacity of the prepared hydrogels were 50 g/g in deionized water. Preparation condition was optimized depending on monomer concentration, temperature and crosslinker concentration. ph dependency on water absorbency and time for equilibrium water absorption were also investigated. For structural characterization FTIR spectroscopy, SEM analysis, TGA and XRD technique were performed. The result reveals that prepared hydrogel can be effectively used in personal absorbent products, tissue engineering, biomedical application etc. as an absorbent material and in polluted water treatment. Keywords: Hydrogels, Cotton, Copolymerization, Crosslinker, Absorbent. 1. INTRODUCTION According to definition hydrogels are such crosslinked polymeric substances that remain insoluble and can hold large amount of fluid compared to its body mass. Super absorbents (SAPs) are commercial member of hydrogel family. Due to higher water absorbency of hydrogels it can be used in personal healthcare, agriculture, biomedical applications [1] healthcare [2,3], construction [4], and other industrial application [5,6]. Recent trend of polymeric materials is to be biodegradable so that they will be harmless for the environment. In this regard scientists have gained enough success by incorporating cellulose in polymeric chains [7]. Polymers with functional groups in its structure exhibits substantial response towards natural stimuli (ph, temp, ionic yield, conc., presence of enzyme etc.) and swell or shrink respectively. This swellshrink behaviour has profound effect on the application and can be controlled by altering methods of preparation. Among different methods of hydrogel preparation radical copolymerization technique is the most widely used due to its simple reaction path and less expensive instruments. In radical copolymerization monomers are copolymerized in presence of initiators and crosslinker and grafted on cellulosic materials. In this study hydrogel was synthesized following techniques described in [8] by copolymerization reaction of cotton with acrylic acid (AA) and acrylamide (AM) using N,N-methylne-bisacrylamide (MBA) as a crosslinker in the presence of initiator potassium persulphate (K 2 S 2 O 8 ).This study also deals with effects of different process parameters on the water absorption of cotton hydrogel to determine the optimum conditions for preparation and characterization of the internal structure of the prepared sample using modern analytical tools-sem, FTIR, XRD and AAS. 2. EXPERIMENTAL 2.1 Materials: Cotton was collected from local market and other purchased chemicals were of analytical grade. 2.2 Preparation of Cotton-g-Poly(Acrylamide- Acrylic acid) Hydrogels: Cotton-g-poly (acrylamide-acrylic acid) hydrogels were synthesized by free radical graft copolymerization of cotton with AA and AM in presence of an initiator K 2 S 2 O 8 and a crosslinking agent (MBA). In a typical sample preparation 0.5g of cotton pulp immersed in 15 ml of distilled water in three necked flask fitted with magnetic stirrer, a reflux condenser and a nitrogen line at 60 o C for bubbling oxygen free nitrogen for 30 min before starting reaction. Initially 0.05g of initiator (10% w/w, based on dry cotton weight) was dissolved and added to the cotton solution with stirring for 10 min to create radicals. The total volume of the solution was controlled to 20 ml. Thereafter 0.6 g of AM (120% w/w, based on dry cotton weight) and 0.56 g of AA (120% w/w, based on dry cotton weight) were added and the mixture was stirred for 30 min. Then the reaction was proceeded for 2h more. At the end of the reaction the 145

147 prepared hydrogels were removed carefully and washed with distilled water for several days. Again the hydrogels was washed several times with pure ethanol for dewatering and immersed in NaOH solution for 24h. Finally the hydrogel was washed with distilled water and dried in oven for 24 h at 60 o C. 2.3 Characterization Water Absorbency of Hydrogels: To measure water absorbency pre-weighed sample was immersed in distilled (about 12 h) water and weighed again to remove surface water by gently dabbing with tissue paper. The water absorbency was calculated by the equation: Q eq =(M eq M o ) / M o (1) Where M o is weight of dry hydrogel, M eq is the weight of wet hydrogel at equilibrium and Q eq is the water absorbency in g/g FTIR Analysis: FTIR Analysis of the pure cotton, hydrolyzed cotton and hydrogels was conducted by(model:ftir-8900, Shimadju, Japan) for the observation of functional groups present in samples within the frequency range from 400 to 4000 cm -1 by the method of transmission using KBr technique Morphology and Crystallinity of Hydrogels: Surface morphology and crystallinity of the prepared hydrogels was observed by SEM and XRD analysis. S 2 O C N 2 2SO 4 - ; Cellulose -OH+SO Effect of ph on the Swelling Behavior: Buffer solution at different ph was used to determine the water absorbency of hydrogels at different ph using Eq Absorption Experiments of Metal Ions: Amount of adsorbed metal ions were calculated by immersing certain amount of synthesized nonhydrolyzed hydrogels in separate solutions (100 ml) containing 200 ppm of Cu 2+ and Cr 3+ with stirring at 100 rpm for 3h. After filtration the concentration of the filtrate was determined by AAS [9]. 3. RESULTS AND DISCUSSION The final product hydrogels were prepared by graft copolymerization of acrylic acid and acrylamide onto hydrolysed cotton in presence of K 2 S 2 O 8 as a free radical initiator and MBA as a cross linking agent. The suggested mechanism for hydrogel preparation is in Fig. 1. According to the proposed mechanism K 2 S 2 O 8 produces sulfate anion radicals which reacts with hydroxl groups of cotton to form more active macro radicals. The macro radicals attacks the monomers (AA and AM) to propagate a polymeric chain and thereafter become free radical donor to neighboring molecule to enlarge grafted chain. The polymeric chain simultaneously react with MBA and crosslinked structure is formed. The mechanism of hydrogels preparation is similar with those proposed in the literature [6]. 50 C N 2 Cellulose -O+SO 4 - Cellulose -O + nch = CH 2 + mch = CH 2 COOH CO NH 2 50 o C N 2 Cellulose -O H H 2 C C CH CH 2 COOH CO n NH 2 m Cellulose -O H C H 2 C CH CH 2 COOH CO n NH 2 + NH NH H 2 C O O m MBA CH 2 MBA Hydrogels Fig. 1. Suggested mechanism for the preparation of Cot-g-p(AA-co-AM) hydrogels. 3.1 Effect of Reaction Conditions on Water Absorption Figure 2(a-d) shows the effects of different reaction conditions for synthesis of hydrogels and water absorption. The graft yield of the hydrogel at different temperature was displayed in Fig. 2a. The graft yield was higher at 50 o C. Hydrophilic nature of cotton was increased by grafting with hydrophilic monomer AA and AM on the backbone of cotton. As a result Fig. 2b. shows water absorbency increases up to certain concentration of monomer and again water absorbency decreases due homopolymerization of monomer rather than grafting and also enhances the chance of chain transfer of monomer molecules. The optimum concentration of monomer was 120 % of cotton weight. Figure 2c shows the effect of ph on water absorption of hydrogels in aqueous medium. Water absorption is less at lower and higher ph values. At ph value of neutral medium water absorption is the highest. Another important character of hydrogels is its equilibrium water absorbency showed in Fig. 2d. It shows that almost 80% of absorbed water was taken by hydrogel sample within 1h which indicates that the material can be used in personal absorbent product. 146

148 Grafting (%) Temp. ( c) a showed that grafting reaction reduced the crystallinity of cotton. Water absorption (g/g) Wt.% of monomer(aa+am) b Fig. 3. FTIR spectra of (a) raw cotton, (b) Hydrolysed cotton and (c) Hydrogels Water absorption (g/g) Water absorption (g/g) ph Time (h) c d Fig. 4. XRD spectra of (a) raw cotton, and (b) Hydrogels a a b Fig. 2. The effects of different reaction conditions for synthesis of hydrogels and water absorption 3.2 Characterization of the Synthesized Hydrogels Among three spectra a new peak is present in Fig. 3c. A peak at 1667 cm -1 is a shift of C=O in stretching vibration, caused by superposition of C=O in amide I and C=O in COO -. The new peak at 1570 cm -1 corresponds to the asymmetric -COO - stretching vibration. When comparing the X-ray diffraction patterns of raw cotton and hydrogels from Fig. 4 it is observed that broad peaks at 15 0 and 22 0 diffraction angles were shorten; the smaller intensity of the peak b Fig. 5. SEM images of (a) raw cotton and (b) Hydrogels 147

149 Figure 5 shows the SEM images of raw cotton and hydrogels. The raw cotton has a smooth and continuous surface while hydrogels have rough and discontinuous. The porous surfaces of hydrogels helps water diffusion in the polymeric network, thereby producing high swelling capacity in the final hydrogels. 3.3 Metal Ions Uptake Certain concentration of Cu 2+ and Cr 3+ ions containing solutions were prepared. Then a weighted small amount of dry hydrogel sample was put into the beaker containing the ionic solutions. After stirring for 3 h the sample got coloured, which indicates that the hydrogels can absorb ions from solution. 4. CONCLUSIONS The new green absorbent (hydrogels) were synthesized by grafting of acrylic acid and acrylamide onto backbone of hydrolysed cotton in presence of a crosslinker N,N-methylene-bisacrylamide using free radical graft copolymerization. The optimal conditions for hydrogels synthesis were 10% potassium per sulphate, 5% MBA, and a monomer concentration of 120% AA and AM (all chemicals are calculated based on hyydrolysed cotton weight) and the hydrogel showed maximum water absorbency which is 50 g/g for water. Further studies is required for the kinetic study of absorption of metal ions by the hydrogel which will provide useful information for the treatment of waste water. [6] N. Kashyap, N. Kumar and M. Kumar, Hydrogels for Pharmaceutica and Biomedical Applications, Critical Review in Therapeutic Drug Carrier Systems, 22, 107, [7] M. A. Hubbe, A. Ayoub, J. S. Daystar, R. A. Venditti and J. Pawlak, Enhancement absorbent products incorporating celllose and its derivatives: A review, Bio Resources 8(4), , [8] X. T., Liang, Z.Q. Huang, Y.J. Zhang, and Liu Synthesis and properties of novel superabsorbent hydrogels n with mechanically activated sugarcane bagasse and acrylic acid, Polym.Bull.70 (6), (2013b), [9] M. K. Krusic, N. Milosavlievic, A. Debelikovic, O. B. Uzum, and E. Karadag, Removal of Pb 2+ ions from water by poly(acryl amide-co-sodium methacrylate) hydrogels, Water air Soil Pollut. 223(7), (2012) REFERENCES [1] V. Thomas, Murali Mohan Yallapu, B. Sreedhar, S.K. Bajpai, A versatile strategy to fabricate hydrogel silver nanocomposites and investigation of their antimicrobial activity, Journal of Colloid and Interface Science, 315, 389, [2] M. Sadeghi and H. Zadeh, Synthesis of Starch Poly (Sodium Acrylate-co-Acrylamide) Superabsorbent Hydrogel with Salt and ph- Responsiveness Properties as a Drug Delivery System, Journal of Bioactive and Compatible Polymers, 23(4), , [3] K. Maithili and M. Ram, Disposable diapers: A hygienic alternative, Indian Journal of Pediatrics, 70(11), , [4] F. X. Song, J. F. Wei and T. S. He, A method to repair concrete leakage through cracks by synthesizing super-absorbent resin in situ, Concrete and Bilding Materials, 23(11), , (2009). [5] Pourjavadi, Sh.Barzegar, and F. Zeidabadi, Synthesis and properties of biodegradable hydrogels of K-carrageenam grafted acrylic acidco-2-acryl amido-2-methyl propane sulfonic acid as canditaes for drug delivery systems, Reac. Funct. Polym., 67(7), , (2007). 148

150 Study the Encryption Techniques for Multimedia Md. Martuza Ahamad Dept. of Computer Science & Engineering Islamic University, Kushtia, Bangladesh Abstract Cryptographic techniques play crucial role when users exchange information. Multimedia plays an important role in learning and sharing experiences. When multimedia contents are shared among the users, it faces security threats. Usually multimedia contents take much space. Encryption technique should be time efficient. In this work we consider four encryption techniques: Blowfish, AES, XOR and RSA and two types of media content: text and image. Simulation shows that AES is time efficient than others. Comparing between symmetric and asymmetric cryptography, symmetric cryptographic techniques take less time than asymmetric technique. Keywords Cryptography, Security, Encryption, Decryption, RSA, Blowfish, AES, XOR. I. INTRODUCTION During recent years the telecommunication industry has made tremendous progress in their development of systems that offer more bandwidth to the end user. Users share experiences, learning, and observation within their community using computer network. The information usually contains various media such as text, image video, audio or animation. When these media contents transmit in computer network, it faces many security threats. For example if an adversary change/edit an map when arm force plan to attack enemy or branch office represent monthly sales report through online and some one change data. To protect the information from unauthorized user it should be encrypted. Usually multimedia content file size is large. Thus encryption algorithm should be time efficient. Encryption techniques can be classified as Symmetric key algorithm and Asymmetric key algorithm. Symmetric-key algorithms also known as single-key and asymmetric algorithm uses two different keys Private key and a Public key to execute encryption /decryption process [1]. In this work we consider four most used encryption algorithms: Blowfish, AES, XOR and RSA. First three algorithms are symmetric and RSA is asymmetric. For encryption and decryption we use two types of multimedia contents: text and image. Our objective is to find the suitable encryption technique that is less time complex. The rest of the paper organized as follows: section II describe about the used encryption techniques. In section III we discuss the related work. In section IV, we describe the simulation in detail and finally conclude in section V. Md. Ibrahim Abdullah Dept. of Computer Science & Engineering Islamic University, Kushtia, Bangladesh ibrahim25si@yahoo.com II. CRYPTOGRAPHIC ALGORITHMS This section provides information about the cryptographic algorithms used in this work. There are two general categories of cryptographic keys: symmetric key and asymmetric key systems. The symmetric key systems use a single key. The single key is used both to encrypt and decrypt the information. Both sides of the transmission need to keep the key in secret from. The security of the transmission will depend on how well the key is protected. The biggest difficulty with this approach, of course, is the distribution of the key. Secret key cryptography schemes are generally categorized as being either stream ciphers or block ciphers. Most popular symmetric key algorithms are DES, Triple DES, AES, IDEA, TEA, Blowfish etc. [1][2]. Asymmetric or public key cryptosystem uses two different keys to encrypt and decrypt. Both keys are mathematically related. If A encrypts a message with his private key then B, the recipient of the message can decrypt it with A's public key. Similarly anyone who knows A's public key can send him a message by encrypting it with his public key. A will then decrypt it with his private key. Public key cryptography was developed in 1977 by Rivest, Shamir and Adleman ("RSA") in the US. This kind of cryptography is more efficient than the symmetric key cryptography because each user has only one key to encrypt and decrypt all the messages that he or she sends or receives [2]. Some of examples for asymmetric key cryptosystem are RSA, ELGAMAL, and ECC etc. Security of asymmetric algorithms depend on key length, key generation techniques. Asymmetric key encryption is slow compared to symmetric encryption [2] [3]. Each algorithm has own advantages and limitations. Since multimedia contents take much storage than others. We consider four algorithms AES, Blowfish, XOR and RSA [4] [5] suitable for large file. AES: AES stands for Advanced Encryption Standard also known as Rijndael. It is symmetric block cipher algorithm. In 2001 two Belgian cryptographers Joan Daemen and Vincent Rijmen first develop this algorithm at National Institute of Standards and Technology (NIST). It supports 128 bits fixed length block size and variable length key size of 128, 192 and 256 bits [6]. 149

151 Blowfish: Blowfish is one of most used encryption algorithm. In 1993 Bruce Schneier designed this algorithm. It is symmetric block cipher algorithm takes variable lengthkey from 32 bits to 448 bits. It was the nice alternatives of DES or IDEA [3]. XOR: XOR is simply bitwise exclusive OR operation. Where a key stream XORed with a plain text stream. It is symmetric variable key length stream cipher algorithm. In XOR operation, if the two bits are same then output will be true else output will be false. It is also known as modulus 2 addition. Same operations are held on encryption and decryption [7]. RSA: RSA is one of the widely used, secure and applicable algorithm. In 1977 three mathematician Ron Rivest, Adi Shamir and Leonard Adleman publicly describe the algorithm. It is first practical asymmetric public key algorithm where two key encryption key and decryption key are exists, the encryption key is public and decryption key is secret which is differ from first one [2][3]. III. PREVIOUS WORKS In early most of research work carried on comparison between symmetric algorithms or asymmetric algorithms or both [7-12]. In paper [9], DES, 3DES, AES, and Blowfish and compare them on the basis of rounds block size, key size and encryption/decryption time and shown that, Blowfish is better than other algorithm. In paper [10] Comparative Performance Analysis of Cryptographic Algorithms takes five common algorithms: Twofish, Blowfish, IB_mRSA, RSA and RC with the parameter of rounds block size, key size and encryption/decryption time and shown that IB_mRSA is the better than other algorithm. In [11] the authors consider two algorithms AES and Blowfish and compare basis of ARM implementation and shown that Blowfish is best algorithm. In [12] image files are encrypted using different key length and compare them. IV. EXPERIMENTAL ANALYSIS In our work, we develop a simulator using java programming language where all of implementations of the algorithms for text and image files are takes place. Using that simulator we measure the execution time of the program for various input size, then analyze the performance of the algorithms. Our PC setup was: HP 4 th Gen. Probook 450, Intel Core i5-4200m 2.50 GHz, 4 GB RAM with Ubuntu LTS (Precise Pangolin) 64-bit Operating System. For our experiment we use two types of multimedia contents text and image. At the time of simulation we use 1KB to 1MB total 23 samples of text data, 10KB to 2200KB total 14 samples of BMP images for measure both encryption and decryption time. When we develop our simulation program we use 128bit key size for all of algorithms. Table 1: Encryption times for different text files Time (ms) KB KB KB KB KB KB KB KB KB KB KB KB KB Blowfish AES XOR RSA Time (ms) 1 KB 10 KB 50 KB Table 2: Decryption times for different text files 100 KB 200 KB 300 KB Blowfish AES XOR KB RSA KB 600 KB 700 KB 800 KB 900 KB 1024 KB 150

152 Fig. 1: File size vs encryption time for text Fig. 2: File size vs decryption time for text Table -1 and table -2 show the encryption and decryption time for different size of text files. Fig. -1 and fig. -2 illustrate the encryption and decryption time. It is found that AES algorithm take less time than others. Table -3 and table -4 show the encryption and decryption time for different size of image files. Fig. - 3 and fig. -4 draw the corresponding curves. AES takes less time to encrypt and decrypt. Table 3: Encryption times for different image files Time (ms) 10 KB 50 KB 100 KB 200 KB 400 KB 600 KB 800 KB 1000 KB 1200 KB 1400 KB 1600 KB 1800 KB 2000 KB 2200 KB Blowfish AES XOR RSA Table 4: Decryption times for different image files Time (ms) 10 KB 50 KB 100 KB 200 KB 400 KB 600 KB 800 KB Blowfish AES XOR RSA KB 1200 KB 1400 KB 1600 KB 1800 KB 2000 KB 2200 KB Fig. 3: File size vs encryption time for image Fig. 4: File size vs decryption time for image 151

153 V. CONCLUSION Now it is seen that, AES is the best performed algorithm than other most used algorithms and RSA is the most time consuming algorithm. Blowfish and XOR has average rate of performance. In AES for a certain range of data the taken time is same. Blowfish is the second best performed algorithm and its taken time is linearly increased with load. In XOR, it perform nice when load is small but when load is large its performance will be poor. So it is clear that, the taken time is always increased when the load is increase. Symmetric key algorithms are faster than asymmetric key algorithms but asymmetric key algorithms are more secure than symmetric key. So we can suggest that, for data transmission symmetric key algorithms should be use and for key distribution asymmetric key algorithms should be use. REFERENCES [1] Behrouz A. Forouzan, Cryptography and Network Security, 2 nd Edition, Tata McGraw Hill, [2] W. Stallings, Cryptography and Network Security: Principles and Practice, 6th Edition, Pearson, [3] Bruce Schneier, Applied Cryptography, Second Edition, John Wiley & Sons, [4] algorithms.html (Last Access 28/02/2016) [5] (Last Access 28/02/2016) [6] Ritu Pahal, Vikas kumar, Efficient Implementation of AES, International Journal of Advanced Research in Computer Science and Software Engineering, Volume 3, Issue 7, July 2013, ISSN: X [7] Anuj Kumar, Sapna Sinha, Rahul Chaudhary, A comparative Analysis of Encryption Algorithms for Better Utilization, International Journal of Computer Application ( ), Volume 71-No. 14, May 2013 [8] Ali E. Taki El_Deen, El-Sayed A. El-Badawy, Sameh N. Gobran, Digital Image Encryption Based on RSA Algorithm IOSR Journal of Electronics and Communication Engineering, Volume 9, Issue 1, e- ISSN: [9] Ravali. S.V.K, Neelima.P, Sai Dileep.P Manasa.B, Implementation of Blowfish Algorithm for Efficient Data Hiding in Audio, International Journal of Computer Science and Information Technologies, Vol. 5 (1), 2014, [10] Lalit Singh, Dr. R.K. Bharti, Comparative Performance Analysis of Cryptographic Algorithms, International Journal of Advanced Research in Computer Science and Software Engineering, Volume 3, Issue 11, November 2013, ISSN: X [11] Pallavi H. Dixit, Dr. Uttam L. Bombale, Vinayak B. Patil, Comparative Implementation of Cryptographic Algorithms on ARM Platform, International Journal of Innovation Research in Science, Engineering and Technology, Vol. 2, Issue 10, October 2013, ISO 3297 [12] Ranjeet Masram, Vivek Shahare, Jibi Abraham, Ranji Moona, Analysis and Comparison of Symmertic Key Cryptographic Algortihms Based on various File Features International Journal of Network Security & Its Applications (IJNSA) Vol. 6, No.4 July

154 Influence of Deposition Temperature on the Deposition of SiO 2 Films from Reaction of Silicone Oil Vapor and Ozone Gas Arifuzzaman Rajib Department of Applied Physics, Electronics and Communication Engineering University of Bangabandhu Sheikh Mujibur Rahman Science and Technology Gopalganj, 8100, Bangladesh rajib.apee.38@gmail.com Susumu Horita School of Material Science, Japan Advanced Institute of Science and Technology (JAIST) 1-1 Asahidai, Nomi, Ishikawa Prefecture , Japan. Atowar Rahman and Abu Bakar Md. Ismail Department of Applied Physics and Electronic Engineering University of Rajshahi, Rajshahi 6205, Bangladesh Abstract This work report, the deposition of SiO 2 films on silicon substrate by using chemical reaction of silicone oil vapor and ozone gas at low temperature. An organic solution as a catalyst at atmospheric pressure has been used to enhance the deposition rate of SiO 2. The deposition rate of SiO 2 films was found to vary with the variation of the concentration (1.525x 10-5 moles/cm 3, 3.05x10-5 moles/cm 3, x10-5 moles/cm 3 and 6.10x10-5 moles/cm 3 ) of the catalyst, catalyst solution temperature (21.9 C, 28.1 C & 34.1 C) and deposition temperature (160 C ~ 260 C). The chemical structures of the asdeposited SiO 2 films were studied by Fourier transform infrared (FT-IR) spectroscopy. The thickness and refractive index of the as-deposited films were measured by the laser ellipsometry. FT-IR spectra of the asdeposited films are very much similar to those of SiO 2 films found in literature. In addition, two weak absorption peaks were found at about 950 cm -1 and 3350 cm -1, which were related to Si-OH bonds in the samples. The average deposition rate was found to be 13.2 nm per minute when the organic catalyst was used, whereas it was found around 4.2 nm per minute without the organic catalyst. So, organic catalyst can be able to enhance the deposition over the three times. Index Terms Low-Temperature SiO 2 deposition, APCVD, FT-IR spectroscopy(keywords) I. INTRODUCTION Advanced very large scale integrated (VLSI) devices require high aspect ratio topography on semiconductor substrates. SiO 2 films have been widely applied in the production of electronic devices, integrated devices, optical thin film devices, sensors, and so on [1]. Moreover, low temperature deposition of SiO 2 film on a silicon substrate is desired for high quality gate oxide films to obtain high performance thin film transistors (TFTs) and to form interlayer dielectric (ILD) to suppress the disconnection of interconnect metal [2]. Recently special attention has been paid to thin-film transistor (TFT) technology because it is used for large area flat panel displays such as liquid crystal displays (LCD). TFT performance strongly depends on the SiO 2 / Si interface properties. Conventionally, oxidation, Plasma Enhanced Chemical Vapor Deposition (PECVD), Low Pressure CVD (LPCVD) and Atmospheric Pressure CVD (APCVD) are used for deposition of SiO 2 on silicon substrate. In oxidation method, silicon dangling bonds are formed due to lattice mismatch between Si/SiO 2 interfaces which degrade the device performance [3]. LPCVD is used to deposit a wide range of possible film compositions with good conformal step coverage but an extra low pressure system is required to attain the pressure of Pa (the available pressure LPCVD). PECVD method has the advantages of a low deposition temperature as well as controlling the deposition rate quickly and accurately to obtain high-quality SiO 2 film [4]. However, the setup of this system is very complicated and also, sometimes the deposited films damaged due to plasma. In this study, APCVD methods have been used to deposit SiO 2 films at low temperature with taking advantage of their easy set-up and lower cost (no extra setup is required for pressure control). In this method, silicone oil has been used as starting materials. It is a promising materials because of their low cost, high thermal stability and high safety [5, 6], compared with tetraethylorthosilicate [TEOS:Si(OC 2 H 5 ) 4 ] [7, 8], a widely used as a common source materials to deposit SiO 2 films. TEOS is also toxic to human eyes and throats. Moreover, deposition of SiO 2 films in controlled ambient requires temperature in excess of 600 C. In this work, we have been able to deposit SiO 2 films at relatively lower temperature by adding a more aggressive oxidant into silicone oil. In particular, we have used ozone (O 3 ) as aggressive oxidant. In addition, an organic solution has been added to the system to enhance the dehydration reaction at the final stage of deposition, which in turn enhances the deposition rate of SiO 2 films. II. EXPERIMENTAL METHODOLOGY Figure 1 shows a schematic diagram of the atmospheric pressure chemical vapor deposition (AP) CVD system used for the deposition of silicon oxide films in this study. Here the conventional APCVD system has been slightly modified by introducing a new line part which facilitates to add an organic solution to the main stream of the system. The chemically cleaned substrates were loaded into the chamber, which was uniformly heated by heater at a deposition temperature. The silicone oil (TSF-405) 153

155 was heated to about 50 ºC and vaporized directly by bubbling with N 2 gas at a flow rate of 0.35 standard liter per minute (slm) through a stainless tube heated to about 55 C to avoid the condensation of SO vapor. The organic solution was mixed with silicon oil vapor and N 2 gas flow rate, and then flown into the chamber. This organic solution was used for enhancing the dehydration reaction. The ozone was generated by an ozonizer from O 2 gas at the flow rate of 0.5 slm and then introduced into the chamber together with the silicone oil vapor. The substrates were n-type single crystals with a resistivity of 5-15 Ω-cm. The crystallographic orientation of the substrate was [111]. The dopant of the wafer is phosphorus. The thickness of the substrate = 525 ± 25 µm. They were chemically cleaned by hot ammoniacal solution (NH 4 OH : H 2 O 2 : H 2 O = 1 : 1 : 4) and hot acid solution (HCl:H 2 O 2 : H 2 O= 1 : 1 : 4). After each chemical solution treatment, the chemical oxide was removed by 1% HF solution. Fig.1: Schematic diagram of APCVD method. In our experimental process,o 3 is decomposed to O 2 and reactive O. This decomposition is dependent on deposition temperature [9]. Oliver R.Wulf& Richard C. Tolman specified the range of decomposition temperature that is 148 ºC to 179 ºC [10]. Since the O is very chemically active, the overall reaction of the process of reactive oxygen and SO. In the first stage, the chemically reactive oxygen atoms are reacted with the methyl (-CH 3 ) bond of the SO in the gas phase and broken the methyl bond to form hydroxyl (-OH) bond then intermediate products of decahydroxycyclopentasilaxane (precursors) are formed together with by-products of H 2 O and CO 2. In other words, -CH 3 side groups are replaced with hydroxyl (-OH) and silanol bonds of Si-OH cover the sides of the siloxane chains. In the second stage, the OH groups on the surface are eliminated the by the dehydration reaction with the OH groups on the precursors. The organic solution concentration has been determined as follows: C org = N org /F org (1) where C org is the mole concentration of organic solution, N org is number of moles adding at the samples surface per minute, which is calculated from the equation (1), F org is the flow rate of nitrogen for organic solution. N org = (F org P)/(R T org ) (2) wherep is vapor pressure, T org is the organic solution temperature, and R is ideal gas constant = 8.318KP a /mol.k. From equation 1and 2 that, C org = P/(R T org ) (3) Thickness and Refractive index were measured by laser ellipsometry (ULvac ESM 1A) at a wavelength nm, spot size of about 1 nm, angle of incidence 70 º, sampling time 20ms as well as measured 10 times at the same location. Chemical structures of as-deposited films were studied by Fourier Transform Infrared Spectroscopy (Perkin Elmer (Spectrum 100) FT-IR Spectrometer) of wave number range 500 to When the deposited films are investigated by FT-IR spectrophotometer, the deposited films is kept into the holder and wait 5 to 6 minutes for removing unwanted H 2 O and CO 2, which are absorbed in the films. III. RESULTS AND DISCUSSIONS Figure 2 shows the FT-IR spectra obtained from the as-deposition films with different deposition time as indicated on the figure, in the range between 500 and 4000 cm -1. For the sake of simplicity, spectra are divided into two frequency regions: low frequency region between 500 and 2000 cm -1 and high frequency region between 2000 and 4000 cm -1. In the low frequency region, they consist of three broad bumps in the range between 500 and 1250 cm -1, showing quite different shape and width, but the comparable absorption amplitude. The line width and the peaks position are same for all the deposition films. The broad absorption band for Si-O-Si bonds are cm -1 and cm -1, which are attributed to the Si-O-Si stretching and bending vibrations, respectively [11-13]. Thus, The peaks at the wave numbers of 1075 cm -1 and 810 cm -1 are labeled as Si-O(1) and Si-O(2), respectively. Another feature is observed in the range between 1500 to 1800 cm -1, which is due to the absorbed H 2 O in the deposited films [14-16].On the other hand, in high frequency region of the FT-IR spectra a broad absorption band between cm -1 are observed. A weak band at about 2300 cm -1 is attributed as absorbed CO 2 in the films [14-16]. Hydrogenated silicon oxide thin films are characterized by the absorption bands at cm -1 and cm -1, which are attributed to the Si-OH stretching vibrations [16, 17].It is observed 154

156 Deposition rate (nm/min) Absorbance (arb. unit) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering from the figure that the amplitude of these two peaks is increased with the deposition time. The maximum peaks at the wave numbers of 950 and 3350 cm -1 are labeled as Si-OH(1) and Si-OH(2), respectively. The FT-IR spectra showed that the films deposited from 160 to 260 ºC are almost stoichiometric Si oxides without signals related to carbon, although small signals of Si-OH Absorbed CO 2 Si-OH(2) Si-O(1) Flow rate of nitrogen for oil FN2=0.35 slm, Substrate temp. TS=200 C, Flow rate of oxygen, FO2=0.5 slm, Deposition time 15 Minutes 10 Minutes 5 Minutes Absorbed H2O Wavenumber (cm -1 ) Si-OH(1) Si-O(2) Fig.2: FT-IR spectra of silicon oxide films at different deposition time & without using organic solution. It can be seen in the Figure 3 that the deposition rate increases with the certain deposition temperature and then decreases. It is also observed from the figure that the peak of the deposition rate shifted to left with increasing the organic solution temperature. The deposition rate increase with the deposition temperature which occurs in low deposition temperature is a common behavior as a thermally active reaction process. At lower temperature, molecular energy of the reagents increases with increasing the deposition temperature which in turn overcome the activation energy of the reaction, as result, increase the reaction rate as well as deposition rate [16, 19, 17, 19]. But, the saturation of the deposition rate and decline with the higher temperature suggest that a different chemical reaction occurs in the pathway. It is supposed that, the decrease in deposition rate at higher temperatures may be a consequence of increased in the upstream gas phase reaction, leading to products which deplete the deposition precursors. Therefore, the deposition rate is reduced on the substrate surface [20, 21].In another word, increase of deposition temperature increases the dehydration reaction on the substrate, and also the heat radiation enhances it in space. So, not only SiO 2 film is deposited on the substrate, but also a small fraction of SiO 2 is formed in space and flown away. Throughout most of this study, the total flow rates of the gases (silicone oil (SO), ozone and organic solution) were kept constant at N 2 = 0.35 standard liters per minute (slm) for silicon oil, O 2 = 0.5 slm for ozone, and N 2 = 0.1 slm for organic solution. In equation 2, the ratio between P and T org is an important factor. The temperature of the organic solution at about 21.9 ºC, 28.1 ºC and 34.1 ºC, these ratios are 0.25, 0.35 and 0.46 respectively. Now, considering equation 2, it is proposed that there is a relation between these ratio (P/T org ) and maximum positions of the deposition rate. The ratio P/T increases, the maximum position of the deposition rate decreases. When the organic solution temperature is low, the ratio of P/T org is also low, so that the amount of organic solution per min is low. Therefore, the side reaction is also relatively small. As a result, the maximum peak appears in the higher temperature region. But, when the ratio of P/T org increase, N org increases. As a result all the precursors are reacted in the relatively low temperature. Thus, the maximum positions of the deposition rate are shifted from right to left Org. soln. Temp. T org = 34.1 ºC T org = 28.1 ºC T org =21.9 ºC Deposition time = 10 min, N2 flow rate for oil = 0.35 slm, N2flow rate for org. soln.= 0.1 slm, O2flow rate = 0.5 slm Deposition temperature ( o C) Fig. 3: Deposition rate vs deposition temperature curve for different organic solution temperature at the catalyst concentration of 3.05x10-5 moles/cm 3. It can be observed from Figure 4 that the refractive index decreases with the deposition temperature and then increases. It is also observed the peak position of the refractive index is shifted to left with the increase of organic solution temperature. The relative absorption peak intensity is plotted as a function of deposition temperature and is shown Figure 5. The peak ratios are required for qualitative discussion of refractive index. The natures of Si- OH(2) absorption peak and refractive index are reversed. When the amount of Si-OH(2) increases, the porosity of the film increases, because Si-OH(2) is the bond of SiO 2 and water. As a result decrease in density as porosity is increased in the structure. Now, from the Lorentz-Lorentz model [22-25], there is a physical relation between density and refractive index (the refractive index is directly proportional to the density of the film), So that, refractive index is inversely proportional to the Si-OH(2) bond. 155

157 Deposition rate (nm/min) Refractive index International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering Deposition time = 10 min, N 2 flow rate for oil = 0.35 slm, N 2 flow rate for org. soln.= 0.1 slm, O 2 flow rate = 0.5 slm When the partial pressure is decreased, the molecules are diluted, which lead to a lower deposition rate on the substrate surface [26] Org. soln. Temp. T org = 34.1 ºC T org = 28.1 ºC T org =21.9 ºC Deposition temperature ( o C) Fig. 4: Refractive index vs deposition temperature curve for the different organic solution temperature at the constant deposition time of 10 minutes. It can be seen from Figure 5 that the ratio of Si-OH(1)/Si-O(1) decreases with the deposition temperature. As mentioned in experimental, -Si-O-Siis produced from precursors. Dehydration reaction introduces production of O-Si-O- from the precursors. It is known that, Dehydration is very active reaction. At the low deposition temperature, some of precursors are not reacted. But, when the deposition temperature increases the dehydration rate also increased, more precursors are reacted with their OH bond. So the ratio of Si-OH(1)/Si-O(1) decreases with deposition temperature. Fig. 5: Peak ratio vs deposition temperature curve for the different organic solution temperature at the constant deposition time of 10 minutes. In Figure 6 shows the variation of deposition rate as a function of organic solution concentration. It is observed from Figure 6 that the deposition rate increase with the organic solution concentration as it is expected. Increasing the organic solution concentrations, the dehydration reactions are enhanced, so that the deposition rate increases. But, the deposition rate decreases when the organic solution concentration is larger than x 10-5 moles/cm 3 except the deposition time 15 minutes (3.05 x 10-5 moles/cm 3 ) this corresponding the fact that the deposition rate is decreased due to a high flow of nitrogen for organic solution which in-terms decreasing organic solution partial pressure [26] Deposition time 15 minutes 10 minutes 5 minutes Org. Soln. Concentration (x 10-5 moles/cm 3 ) Fig. 6: Deposition rate vs organic solution concentration curve for the different deposition time at the constant deposition temperature 200 ºC. IV. CONCLUSIONS The SiO 2 films were deposited on a silicon substrate at atmospheric pressure and temperature between 160 ºC and 260 ºC by using the chemical reaction between silicone oil vapor and ozone gas as well as an organic solution is used as a catalyst. The FT-IR spectra showed that the films deposited from 160 ºC to 260 ºC were almost stoichiometric. Although weak peaks from Si-OH related bond was found in the FT-IR spectra no peak related to Carbon was observed. The presence of Si-OH bond indicated the presence of small amount of pores in the deposited film that was responsible for the variation in refractive index of the deposited films. From the experimental data it can be concluded that the deposition rate of SiO 2 from the chemical reaction between silicone oil vapor and ozone gas can be enhanced by optimizing the concentration of organic catalyst, catalyst-solution temperature and the deposition temperature. Future prospects of this work is to find an optimum deposition rate of SiO 2 films by varying other deposition parameters, likes, ozone concentrations, silicone oil concentrations, the position of the substrate, etc. REFERENCES [1] N. Boumaiza, S. Achour, M.E. Tayar, thin solid film journal, 261, 1-2, 352 (1995). [2] S. Higashi, D. Abe, S. Inoue, and T. shimoda: Jpn. J. Appl. Phys., 40, 4171 (2001). [3] M. Okoshi, M. Kuramatsu, and N. Inoue:Jpn. J. Appl. Phys.40 L41 (2001). [4] H. Takao, M. Okoshi, and N. Inoue:Jpn. J. Appl. Phys.42 L461 (2003). [5] Castellarin A, Grigorian R, Bhagat N, etal.vitrectomy with silicone oil infusion in severe diabetic retinopathy. Br J Ophthalmol; 87: (2003). [6] D. McLeod Br J Ophthalmol, 87(10): (2003) Oct. 156

158 [7] H. Nakashima, K. Omae, T. Takebayashi, C. Ishizuka and T. Uemura: J.Occup. Health 40, 270 (1998). [8] Moretto, Hans-Heinrich; Schulze, Manfred; Wagner, Gebhard "Silicones". Ullmann's Encyclopedia of Industrial Chemistry. Weinheim: Wiley-VCH. doi: / a24_057 (2005). [9] Susumu Horita, Koichi Toriyabe and Kensuke Nishioka, Japanese Journal of Applied Physics 48, (2009). [10] Oliver R. Wulf* And Richard C. Tolman, Gates Chemical Laboratory, California Institute of Technology, Pnas. Org. 13, 272 (1927). [11] H. Rinnert, M. Vergnat, Journal of Non-Crystalline Solids, 320, (2003). [12] J. A Theil, D.V. Tsu, M. W. Watkins, S.S Kim, and G. Lucovsky, J. Vac. Sci. Technol. A8(3) 1374 (1990). [13] H. Juarez, M Pacio, T Diaz, E Rosendo, G Garcia, A Garcia, F. Mora and G Escalante, Journal of Physics, 167, (2009). [14] H. Rinnert, M. Vergnat, Journal of Non-Crystalline Solids, 320, (2003). [15] H. Juarez, M Pacio, T Diaz, E Rosendo, G Garcia, A Garcia, F. Mora and G Escalante, Journal of Physics, 167, (2009). [16] Josep Arno, Zheng Yuan, and Shawn Murphy, Journal of the Electrochemical Society, 146 (1), (1999). [17] W Rzodkiewicz and A Panas, Journal of Physics, 181, (2009). [18] W.A. Pliskin,J. Vac. Sci. Technol.14, 1064 (1977). [19] Tang Longjuan, Zhu Yinfang, Yang Jinling, Li Yan, Zhou Wei, Xie Jing, Liu Yunfei, and Yang Fuhua, Journal of Semiconductor, 30 (9), (2009). [20] G. Lucovsky, J. Yang, S.S Chao, J.E. Tyler, W. Czubatyj, Phys. Rev. B (1984). [21] G. Lucovsky, J.D. Jaonnopoulos (Eds.), The Physics of Hydrogenated Amorphous Silicon II, Springer, Berlin, p. 235 (1984). [22] Young Bae Park and Shi Woo Rhee, Journal of Applied Physics American Institute of Physics, 66, 3477 (1995). [23] L. Banyai, P. Gartner,Phys. Rev. B 29, 728 (1984). [24] Oughstun K E, Cartwright N A (2003). [25] K. Vedam, P. Limsuwan,J. Chem. Phys.69, 4772 (1978). [26] Young Bae Park and Shi Woo Rhee, Journal of Applied Physics American Institute of Physics, 86(3), 1346 (1999). 157

159 An Improved Representation of Audio Signal in Time-Frequency Plane Kazi Mahmudul Hassan Dept. of Computer Science & Engineering Varendra University Rajshahi, Bangladesh Ekramul Hamid Dept. of Computer Science & Engineering Rajshahi University Rajshahi, Bangladesh Takayoshi Nakai Dept. of Electric & Electronic Engineering Shizuoka University Hamamatsu-shi, Japan Abstract To analyze non-stationary signal like audio, time-frequency representation is an important aspect. In case of representing audio signal in timefrequency-energy distribution, hilbert spectrum is a popular way which has several advantages than other methods like STFT, WT etc. Hilbert-Huang Transform is a prominent method consists of both Empirical Mode Decomposition (EMD) and Hilbert Spectral Analysis (HSA). An enhancement of EMD called Turning Tangent empirical mode decomposition (2T-EMD) has recently developed to overcome some limitations like cubic spline problems, sifting stopping condition etc. 2T-EMD based hilbert spectrum of audio signal encountered some issues due to the generation of too many IMFs in the process. In this work, a mutual implementation of 2T-EMD & classical EMD is proposed which enhances hilbert spectrum representation of audio signals. This refinement of hilbert spectrum not only contributes to the future work of source separation problem but also many other applications in audio signal processing. Keywords time frequency representation; HHT; hilbert spectrum; EMD; 2T-EMD; filter bank; fractional Gaussian noise; I. INTRODUCTION Audio signals are information rich nonstationary signals which play an important role in our day-to-day communication, perception of environment, and entertainment. Because of its non-stationary nature of audio signal, time or frequency only approaches are inadequate in analyzing these signals. A joint timefrequency (TF) representation approach would be a better choice which provides some temporal and spatial information simultaneously. Due to the several limitations of existing timefrequency methods like Fourier Transform, Short- Time Fourier Transform, Wavelet Transform etc., in 1998 Huang et al proposed a new method known as Hilbert Huang Transform (HHT) which includes Empirical Mode Decomposition (EMD) and Hilbert Transform (HT) to analysis non-linear and nonstationary signals in time, frequency and energy domain [1].The outcome of HHT process is instantaneous frequency, amplitude and phase which are used to generate Hilbert Spectrum (HS), a new representation of a nonstationary signal like audio in 3D graph where time-frequency-energy is represented by x, y, and z axis respectively. Using the EMD method, any complicated data set can be decomposed into a finite and often small number of components called Intrinsic Mode Functions (IMF). EMD process suffers from some mathematical problem like prediction problem, spline problems, optimization problems etc. [2]. Also it produces less number of IMFs to make a successful decomposition of an audio signal in time-frequency domain [3]. Turning Tangent Empirical Mode Decomposition (2T-EMD) is a novel EMD algorithm which differs from other approaches of EMD like EEMD, BEMD etc. by its computational lightness and algorithmic simplicity, is developed to overcome some limitations of EMD like boundary conditions, sifting stopping condition etc. [4]. But 2T-EMD method decomposes a signal with so many redundant IMFs which makes the Hilbert Spectrum representation of an audio signal clumsy and unfruitful for relevant applications. In this paper work, a mutual implementation of 2T-EMD and classical EMD method is proposed based on the filter bank property of both methods, analyzing with respect to fractional Gaussian noise to enhance the hilbert spectrum representation in case of audio signal. Fractional Gaussian Noise (fgn) is a generalization of ordinary white noise which is a versatile model of homogeneously spreading broadband noise without any dominant frequency [5]. EMD acts essentially as a dyadic filter bank is stochastic situation which is observed in numerical experiments based on broadband noise like fgn [6]. This paper try to figure out 2T-EMD property in stochastic situation using similar experiment used in case of EMD and proposed a method based on the outcome. In this method, initially a nonstationary signal like audio is decomposed with 2T-EMD method until it follows dyadic filter bank properties. A threshold point is needed to be selected based on the property which is in between IMF index 10 to 14 in case of audio signal that can vary according to other applications. This decomposition generates a IMFs set and residue signal. The residue signal is then further 158

160 decomposed with classical EMD method which produces another set of IMFs. Finally those two IMFs set are concatenated to form final set of IMFs which later on hilbert transformed and plotted into timefrequency plane. This proposed method produces a much better hilbert spectrum representation in timefrequency plane especially in case of audio signal rather than pure 2T-EMD based hilbert spectrum, which can contributes other relevant application like source separation, speech recognition etc. significantly. II. TURNING- TANGENT EMD BASICS The hilbert spectrum is a relatively new joint timefrequency representation introduced in [1]. Two phases are required to generate the HS. In the first phase, EMD is employed, which is an adaptive decomposition method [6]. Then discrete hilbert transform (DHT) is employed in the second phase. HS is generated by the combination of EMD and DHT. EMD focuses on the level of local oscillations and decomposes the signal into a finite set of AM-FM oscillating components which are bases of the decomposition. The principle of the EMD technique is to decompose a signal s(t) into a sum of the bandlimited functions α m(t) or bases called intrinsic mode functions (IMFs). Each IMF satisfies two basic conditions: (i) in the whole data set, the number of extrema and the number of zero crossings must be the same or differ at most by one, (ii) at any point, the mean value of the envelope defined by the local maxima and the envelope defined by the local minima is zero [1]. To successfully decompose an audio signal, the classical EMD method produces a few number of IMFs which is insufficient for different application like source separation, frequency analysis etc. Turning Tangent Empirical Mode Decomposition also called 2T-EMD is another approach of EMD algorithm which differs from other by its lightness of computation and simplicity [4]. The 2T-EMD algorithm redefines the signal mean envelope and offers the possibility to decompose multivariate signals without any projection. There are several improvements is done in 2T-EMD over EMD. Firstly, a robust computation of the mean trend is preferably obtained for 2T-EMD by averaging two envelops: A first envelop interpolates the even indexed barycenters which include signal borders, and a second envelop interpolates the odd indexed barycenters which also include signal borders. i.e. let h even (t) and h odd (t) are even indexed and odd indexed upper envelop respectively. Similarly l even (t) and l odd (t) are even and odd indexed envelop respectively. So the ultimate local mean µ 1 (t) will be calculated as µ 1 (t) = ((h even (t)+ h odd (t))/2 +(l even (t)+ l odd (t))/2)/2 (1) which is actually an average of two local mean µ even (t) (even indexed) and µ odd (t) (odd indexed)[4]. Secondly, the interpolation is performed using cubic splines with classical boundary conditions as for the classical EMD, where the signal border is directly added to the list of estimated oscillation barycenters. Fig. 1. Block diagram of proposed method Thirdly, the sifting process is stopped using a Cauchy-like criterion. In case of audio signal, using 2T-EMD instead of EMD produces a great number of IMFs. At first sight seems that it solves the matter of limited number of IMFs but actually produces many redundant IMF especially in the lower frequency band of any audio signal. Because of this, the hilbert spectrum representation of any audio signal using 2T-EMD method instead of classical EMD becomes quite meaningless especially in the lower frequency band [3]. III. PROPOSED METHOD To mitigate this problem, a method of combined implementation of 2T-EMD and classical EMD is proposed which is given below: St. 0: Choose a threshold point based on filter bank properties for selecting number of IMFs from 2T-EMD method using fgn. St. 1: Decompose the signal with 2T-EMD method to get IMFs set and a residue signal. St. 2: Decompose residue signal with classical EMD method and get new IMFs set. St. 3: Concatenate the 2T-EMD IMFs set with EMD IMFs set to make the final set of IMFs. St. 4: Hilbert Transform the IMFs and generates Hilbert Spectrum. The block diagram of proposed method is shown in Fig

161 IV. ANALYSIS To select a threshold point as stated in the proposed method we need to analysis 2T-EMD methods filter bank property based on fractional Gaussian noise (fgn). In order to better understand the way 2T-EMD behaves in stochastic situation involving broadband noise, we report here on numerical experiment based on fgn. In case of EMD such an experiment emits that it acts essentially as a dyadic filter bank resembling those involved in wavelet decompositions [7]. The fractional Gaussian noise (fgn) is used here as a reference signal which is generalization of ordinary white noise. It is a versatile model of homogeneously spreading broadband noise without any dominant frequency band [8]. Recall that fgn is defined as the increment process of fractional Brownian motion (fbn) [9]. The statistical properties fgn are entirely determined by its second-order structure, which depends solely upon one single parameter, Hurst exponent [10]. In case of EMD it follows the filter bank property where H can vary from 0.1 to 0.9. The data length has been typically set to N=4096 and, for each value of H, 5000 individual sample paths generated via the Wood and Chan s algorithm [11]. In this experiment we set H=0.6 based on preliminary analysis. (a) (a) (b) (c) Fig. 2. Apply EMD on fgn, the average spectra of the 7 first IMFs (a) Log 2(energy) vs. normalized frequency (b) Log 2(zero crossing) vs. IMF index (c) Log 2(energy) vs. IMF index For a given signal x(t), EMD and 2T-EMD both ends up with a representation of the form: Where m k (t) stands for a residual trend and the modes {d k (t), k=1,... K} are constrained to be zeromean AM-FM waveforms. (b) Fig. 3. Apply 2T-EMD on fgn, the average spectra of first 25 IMFs (a) Log 2(energy) vs. normalized frequency (b) Log 2(zero crossing) vs. IMF index Applying EMD on each of the sample path of fgn generates separate IMFs set. In Fig. 2 (a) the average spectra of the 7 first IMFs are plotted as a function of normalized frequency. Fig. 2(b) shows the logarithm (base 2) of the average number of IMFs zero-crossing which is decreasing linearly as the increase of IMF index [6]. And finally Fig. 2(c) shows the logarithm (base 2) of the average energy of IMFs which are x(t) = m k (t) + k (t) similarly decreasing like zero-crossing as the increase of (2) IMF index. 160

162 International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering IC4ME2-2016, 24~25 March, 2016 depending on application) to stop 2T-EMD methods IMF generation process. V. RESULTS & DISCUSSIONS Fig. 4. Apply 2T-EMD on fgn s each sample path, Log2(energy) of each IMF set vs. IMF index Apply the same experiment using 2T-EMD on fgn with that same parameter, the following result is observed shown in Fig. 3. It shows that in case of 2TEMD, the relation of normalized frequency bin index and zero-crossing of IMFs with IMFs index is not linear like EMD. Also it is visible that the frequency bin overlapping issues in Fig. 3(a) due to the excessive number of IMFs. Plotting individual IMF sets log (base 2) energy with respect to IMF index, we get the result (shown in Fig. 4) where a huge number of IMFs energy is increasing after first few and as the IMF index increased, this rate is increasing too which is contradictory with previous study [12]. Taking histogram of first energy increment IMF index of each sample fgn path IMFs set, the outcome is following (see Fig. 5): The efficiency of the proposed method is tested by representing real world audio signal from their mixture. The sample audio mixture length is one second which is sampled at 16 khz sampling rate and 16-bit amplitude resolution, containing flute sound and male voice pronouncing 1, 2, 3 and 4 with some short interval between each number. As discussed earlier, pure 2T-EMD method produces so many redundant IMFs especially in lower frequency band of audio signal which makes the hilbert spectrum representation clumsy. Experiment result shows that the rate of changes in IMF s zerocrossing decreasing slowly with the increment of IMF index which can be observed from Fig. 3(b) and due to this, higher index IMF causes frequency bin overlapping problem in hilbert spectrum (see Fig. 7(a)). Fig. 7(b) shows that proposed method mitigates frequency bin overlapping problem by reducing number of IMFs and with the increment of rate of change in zero-crossing of IMFs (see Fig. 8(a)) and also the average log2 (energy) of IMFs are continuously decreasing with the increment of IMF index (see Fig. 8(b)). The only energy increment is in the concatenation point which is because of different Fig. 5. Histogram of fgn IMFs energy increment (1st) (Apply 2T-EMD on the fgn) And according to the average log2(zero-crossing) of IMFs, the rate of changes in normalized frequency bin index decreased inconsistently after first few(10 to 14) IMFs (see Fig. 6). (a) (b) Fig. 6. IMFs Log2(norm. freq. bin index) vs. IMF index (Apply 2T-EMD on the fgn) Experiment result exhibits inconsistency of maintaining filter bank property by 2T-EMD method after first few IMFs (10 to 14 IMF index approx.) generation. A threshold point has to be selected (between IMF indexes 10 to 14 which can varies (c) Fig. 7. Hilbert spectrum of sample audio signal using (a) pure 2T-EMD method, (b) proposed method and (c) reconstruction error signal mean definition of EMD and 2T-EMD methods. Hilbert spectrum of sample audio signal using pure 161

163 2T-EMD method and proposed method are shown in Fig. 7(a) and 7(b) respectively. [1] N.E. Huang, Z. Shen, S.R. Long, M.L. Wu, H.H. Shih, Q. Zheng, N.C.Yen, C.C. Tung and H.H. Liu, The empirical mode decomposition and hilbert spectrum for nonlinear and non-stationary time series analysis, Proc. Roy. Soc. London A, Vol. 454, pp , [2] Peel, M.C., G.E. Amirthanathan, G.G.S. Pegram, T.A. McMahon and F.H.S. Chiew Issues with the application of empirical mode decomposition analysis, EGU Session CL21, Peel et al (a) [3] Md. Ekramul Hamid, Md. Khademul Islam Molla, Md. Iqbal Aziz Khan, Takayoshi Nakai. Speech enhancement using hilbert spectrum and wavelet packet based softthresholding, Science Journal of Circuits, Systems and Signal Processing. Vol. 1, No. 1, 2015, pp doi: /j.cssp [4] Julien Fleureau, Jean-Claude Nunes, Amar Kachenoura, Laurent Albera, and Lotfi Senhadji, Turning Tangent Empirical Mode Decomposition: a framework for mono- and multivariate signals, IEEE Transactions On Signal Processing, Vol. 59, No. 3, March 2011 (b) Fig. 8. Apply proposed method on fgn, the average spectra of first 18 IMFs (a) Log 2(zero crossing) vs. IMF index (b) ) Log 2(energy) vs. IMF index It s quite visible from both 2T-EMD and proposed method based hilbert spectrum representations that the later one has more distinguishable energy density which represents the presents of information with respect to the time and frequency domain. Fig. 7(c) also shows the reconstruction error signal that has maximum magnitude 8x10-16 which is acceptable in most applications. VI. CONCLUSION In this paper, a new method is introduced which is a combined approach of 2T-EMD and EMD methods, emphasizing on the improvement of pure 2T-EMD based hilbert spectrum. This proposed method can much more efficiently decompose an audio signal into sufficient number of IMFs which is not only enhances hilbert spectrum representation but also contributes several kind of audio signal processing work. This work reveals the nature of 2T-EMD method in stochastic situation which analyzes with respect to the fractional Gaussian noise. This enhancement of hilbert spectrum can now represents significant amount of information in both time and frequency domain with energy distribution. [5] G. Rilling, P. Flandrin, and P. Goncalves, Empirical mode decomposition, fractional Gaussian noise and Hurst exponent estimation, in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process., 2005, vol. 4, pp [6] P. Flandrin, G. Rilling and P. Gonqalves, Empirical mode decomposition as a filter bank, IEEE Signal Processing Letters, vol. 11, no. 2, (2004). [7] Z. Wu and N.E. Huang, A study of the characteristics of white noise using the Empirical Mode Decomposition method, submitted to Proc. Roy. Soc. London A, Dec [8] Md. Khademul Islam Molla, Md.Rabiul Islam, Toshihisa Tanaka, Tomasz M. Rutkowski, Artifact suppression from EEG signals using data adaptive time domain filtering, Article In Neurocomputing 97: November 2012 [9] B.B. Mandelbrot and J.W. van Ness, Fractional Brownian motions, fractional noises and applications, SIAM Rev., Vol. 10, pp ,1968. [10] G. Rilling, P. Flandrin, and P. Goncalves, Empirical mode decomposition, fractional Gaussian noise and Hurst exponent estimation, in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process., 2005, vol. 4, pp [11] A.T. Wood and G. Chan, Simulation of stationary processes in [0,1] d, J. Comp. Graph. Stat., Vol. 3, pp , [12] Md. Ekramul Hamid, Md. Khademul Islam Molla, Xin Dang, Takayoshi Nakai, Single channel speech enhancement using adaptive soft-thresholding with bivariate EMD, Hindawi Publishing Corporation, ISRN Signal Processing, Volume 2013,Article ID , 9 pages ACKNOWLEDGMENT The author would like to thank Dr. M. Khademul Islam Molla, Professor, Dept. of Computer Science & Engineering, University of Rajshahi, Bangladesh for providing the sample audio mixture signals and valuable guidance to accomplish this work. REFERENCES 162

164 On the Optimization of Number of Message Copies for Multi-Copy Routing Protocols in Scalable Delay-Tolerant Networks Md. Sharif Hossen and Muhammad Sajjadur Rahim Department of Information and Communication Engineering University of Rajshahi Rajshahi-6205, Bangladesh Abstract Delay-Tolerant Networks (DTNs) are such kind of networks, where there is no direct path from source to destination. Such networks are characterized by long/variable delays, asymmetric data rates, limited resources and high error rates. In such network, when the number of nodes increases but memory in buffer is fixed for each node, then there is a need for optimizing the number of message copies in multi-copy DTN routing protocols, such as, Spray-and-Wait (SNW) and Spray-and-Focus (SNF) routing to evaluate the percentage of number of message copies (indicated as L) that will show better delivery, lower delay, and lower overhead under scalable DTN scenario. Opportunistic Network Environment (ONE) simulator is used for this investigation. The investigation results demonstrate that SNF routing shows better performance using only 2% of L copies but in case of SNW it will be 10% of L copies. Hence, using the same performance metrics and simulation setting, SNF exhibits the better delivery with considerable lower latency and lower overhead with only 2% of L copies. Keywords delay-tolerant networks; multi-copy routing; simulation; performance metrics; ONE simulator I. INTRODUCTION Delay-Tolerant Networks (DTNs) are kinds of mobile ad-hoc networks, where there is no persistent route from source to destination. It is an intermittent and sparsely connected mobile ad-hoc network due to limited transmission range and mobility model. This network can be characterized by lack of connectivity, long/variable propagation delay, and high bit error rates [1 2]. In such challenging network, popular adhoc routing protocols such as Ad-hoc On-Demand Distance Vector (AODV) [3] and Dynamic Source Routing (DSR) [4] cannot be implemented for routing data successfully, as they require a continuous end-toend path between source and destination. DTNs are also referred to as Intermittently Connected Mobile Networks (ICMN) [5], which is featured by intermittent connectivity and temporarily broken links [6]. DTN uses Store and Forward strategy for routing messages where message is successively moved from the source to the next available node and stored in the buffer which is forwarded to other nodes in hops towards the destination [7 9] as illustrated in Figure 1. DTNs have been extensively used in many areas including interplanetary networks [10], underwater networks [11], satellite communication [12], wildlife tracking sensor networks [8], military networks, and vehicular ad-hoc networks [13], etc. Fig. 1. Store and forward strategy The rest of this paper is organized as follows: Section II describes the technique of spray-and-wait routing. Section III gives the description of spray-andfocus routing. Section IV provides the details of the simulator and simulation setup. Section V discusses the obtained results in terms of three performance metrics. Section VI provides the concluding remarks and future works about this research endeavor. II. SPRAY AND WAIT ROUTING Spray-and-Wait (SNW) [14] routing protocol is a class of replication-based schemes that attempts to find a good delivery ratio by limiting the number of replicas of a given message while keeping resource utilization low as in forwarding-based routing. SNW achieves resource efficiency by setting a fixed upper bound on the number of copies per message allowed in the network. The SNW protocol consists of the following two phases: (i) Spray Phase: For every message originating at a source node, L message copies are initially spread forwarded by the source and possibly other nodes receiving a copy to L distinct relays. (ii) Wait Phase: If the destination is not found in the spraying phase, each of the L nodes carrying a 163

165 message copy performs direct transmission (i.e. will forward the message only to its destination). For example, in the spray phase, for each message generated at the source, L copies are distributed to L distinct relays, as shown in Figure 2 (a). If the destination is not reached during the first phase, each of the L relays spreads in turn the message to their neighbors until the attainment of the destination, which is the task of the wait phase, as depicted in Figure 2 (b). The parameter L is selected depending on the density of the network and the desired average time. Fig. 3. Vanilla and binary version of SNW routing In case of Binary version as shown in the lower part of Figure 3, here we consider L = 4. In this case, S divides the existing copies by 2 and now remaining copies are 2 (4/2 = 2). Now S forwards the remaining copies to the first two nodes a, and b. Nodes a and b again divide their copies by 2 and hence only have a single copy. This copy is not forwarded to any other nodes except destination node D. Fig. 2. SNW routing There are two main versions of the SNW routing protocol, respectively known as Vanilla and Binary. The two versions differ in the mechanism employed to spray the L copies of a message. To achieve this, a simplest way, called Vanilla, is to transmit a single copy of the message from the source to the first L distinct nodes it encounters after the message is generated. The second version, referred to as Binary, works as follows: the source node starts with L copies of the message. It transfers L/2 of the copies to the first node it encounters. Each node then transfers half of its copies to future nodes they meet that have no copy of the message. When a node eventually gives away all of its copies, except for one, it switches to the wait phase where it waits for a direct transmission opportunity with the final destination of the message. The advantage of the Binary version is that messages are disseminated much faster than in the Vanilla version. In this investigation, we have considered only the Binary Spray-and-Wait (B-SNW). In Vanilla version, if we consider that the number of copies is L = 3, then source, S, forwards these three copies to first three nodes a, b, and c as shown in the upper part of Figure 3. Since these nodes are not the destinations, so they are now in waiting phase and try to communicate directly to the destination. III. SPRAY AND FOCUS ROUTING Spray-and-Focus (SNF) [15] routing protocol overcomes the shortcomings of simple spraying algorithms. Existing spraying scheme, i.e., Spray and Wait scheme [5, 16], consists of two phases: in the first phase it distributes a fixed number of copies to the first few relays encountered, and in the second phase each of these relays waits until it encounters the destination itself (i.e. Direct Transmission ). It is easy to see that, here, this scheme would spread all its copies quickly to the node s immediate neighborhood, but then few if any of the nodes carrying a copy might ever see the destination [17]. What is more, if the network is not too sparse, there might exist partial paths over which a message copy could be transmitted fast to a node closer to the destination. Yet, in schemes like Spray-and-Wait, a relay with a copy will naively wait until it moves within range of the destination itself. This problem could be solved if a sophisticated single copy scheme [18-19] is used to further route a copy after its handover to a relay, a scheme that takes advantage of transmissions (unlike Direct Transmission). Thus, in second phase ( focus phase) rather than waiting for the destination to be encountered, each relay can forward its copy to a potentially more appropriate relay, using a carefully designed utility-based scheme. Thus, when a relay for a given message has only one forwarding token left for that message, it switches to the Focus phase. Unlike SNW, where in the Wait phase messages are routed using Direct Transmission (i.e. forwarded only to their destination) [6, 16], in the Focus phase, a message can be forwarded to a different relay according to a given forwarding criterion: 164

166 (i) Age of last encounter timers with transitivity: Let us assume that each node maintains a timer for every other node in the network, which records the time elapsed since the two nodes last encountered each other as follows: initially set and, ; whenever i encounters j, set ; at every clock tick, increase each timer by 1. (ii) Single-copy utility-based routing: Let every node i maintain a utility value for every other node j in the network. Then, a node A forwards to another node B a message destined to a node D, if and only if, where (utility threshold) is a parameter of the algorithm. For example, as shown in Figure 4, unlike sprayand-wait, in second phase (focus phase), nodes a and b can forward their single copy to the next nodes c and d, respectively using single-copy utility-based routing algorithm if the following conditions are satisfied: & Fig. 5. Opportunistic networks scenario of Helsinki city area B. Simulation Environment Setup Simulation configuration for varying percentage of L copies in a scalable network is shown in table I. Two groups of nodes have been considered. These are pedestrians and cars. Both groups have equal number of nodes. For example, for total number of nodes equal to 50, pedestrians has 25 nodes and car also has 25. TABLE I. SIMULATION ENVIRONMENT PARAMETERS IV. Fig. 4. SNF routing SIMULATION TOOLS AND SETUP Opportunistic Network Environment (ONE) simulator with program version of is used in this paper. This section explains the ONE simulator with Graphical User Interface (GUI), and the environment modeling parameters. A. The ONE Simulator ONE is an agent-based discrete event simulation engine designed for evaluating the performance of DTN routing protocols. At each simulation step, ONE combines mobility modeling, inter node contacts, DTN routing, message handling and visualization in one package that provides a rich set of reporting and analyzing modules. A detailed description of the simulator is available in [20] and the ONE simulator project page [21] where the source code of the simulator is also available. Source codes of ONE simulator are written using java programming language. Figure 5 shows the GUI of the ONE simulator where node locations, current paths, connections between nodes, number of messages carried by a node, etc. are all visualized in the main window. Parameters Value Simulation Time 12 hours Number of Nodes 50, 100, 150, 200 pedestrians Speed m/s Group id p cars Speed m/s Group id c Interface Bluetooth Interface Interface Type Simple Broadcast Interface Transmit Speed 250 kbps Transmit Range 10 m Buffer Size 5 MB Message Generation Rate 2, i.e., one message in seconds Message TTL Movement model Message Size Number or percentage of L copies in SNF routing Simulation Area Size 300 minutes Shortest Path Map Based 500 KB 1 MB 2, 5, 10, 20, m 3400 m V. SIMULATION RESULTS AND DISCUSSION This section discusses obtained results by running the simulations on the following performance metrics, i.e., delivery probability, average latency, and overhead ratio. A. Performance Analysis on Delivery Probability Delivery probability is the ratio of the total number of messages delivered to the destination over the total number of messages created at the source. 165

167 Fig. 6. Delivery probability with varying number of nodes for L copies Hence we see that SNF routing shows higher delivery when L = 2. On the other hand, in case of SNW routing, we see the better delivery for L = 10. So we can say that SNF. B. Performance Analysis on Average Latency Average latency is the measure of average time between messages generated and messages received to destination. only 2 copies but SNW requires 10 copies to obtain its optimization. C. Performance Analysis on Overhead Ratio The overhead ratio defines how many redundant packets are relayed to deliver a single packet. It simply reflects the cost of transmission in a network. With increase of number of message copies, overhead ration increases for SNW routing. Although SNW routing shows lower overhead for L = 2 and L = 5 than L = 10, we have considered L = 10 as a standard considering its delivery probability. Hence SNW shows lower overhead for L = 10 compared to L = 20 and 50. On the other hand, SNF routing shows very lower (approximately zero) overhead ratio for L = 2 copies. So we can say that SNF routing shows lower overhead with only 2 % of L copies. Fig. 8. Overhead ratio with varying number of nodes for L copies Fig. 7. Average Latency with varying number of nodes for L copies Average latency increases gradually in accordance with the increase of number of nodes. Hence in case of SNF routing, average latency approximately same for varying percentage of L copies. But here in case of SNW routing, we see that it shows lower latency with increase of L copies. Since L = 10 shows lower latency than L = 2 and L = 5. So we can give a standard with its higher delivery probability. Therefore, SNF routing gives lowest latency using VI. CONCLUSION AND FUTURE WORKS In this paper, we have investigated the optimization of number of message copies for multicopy routing protocols, such as, Spray-and-Wait (SNW) and Spray-and-Focus (SNF) in scalable delaytolerant network. The evaluation demonstrates that SNF routing shows better performance using only 2% of L copies but in case of SNW it will be 10% of L copies. Hence, using the same performance metrics and simulation setting, SNF exhibits the better delivery with considerable lower latency and lower overhead with only 2% of L copies. Further investigation can be done in future to analyze the above mentioned routing protocols in terms of energy consumptions and congestion control in the network. 166

168 REFERENCES [1] K. Fall, A delay-tolerant network architecture for challenged internets, in Proc. of ACM SIGCOMM, Karlsruhe, Germany, Aug. 2003, pp [2] S. Jain, K. Fall, and R. Patra, Routing in a delay-tolerant network, in Proc. of ACM SIGCOMM, Portland, USA, Oct. 2004, pp [3] C. E. Perkins, and E. M. Royer, Ad-hoc on-demand distance vector routing, 2nd IEEE Work. on Mob. Comp. Sys. and App., New Orleans, LA, USA, Feb. 1999, pp [4] D. B. Johnson, and D. A. Maltz, Dynamic source routing in ad hoc wireless networks, Mobile Com., Kluwer Academic Publishers, Feb. 1996, ch.5, pp [5] L. Pelusi, A. Passarella, and M. Conti, Opportunistic networking: data forwarding in disconnected mobile ad hoc networks, IEEE Comm. Mag., vol. 44, no. 11, Nov. 2006, pp [6] Z. Zhang, Routing in intermittently connected mobile ad hoc networks and delay tolerant networks: overview and challenges, IEEE Comm. Sur. and Tut., vol. 8, no. 1, Jan. 2006, pp [7] J. Burgess, B. Gallagher, D. Jensen, and B. N. Levine, Maxprop: routing for vehicle-based disruption-tolerant networks, in Proc. of IEEE INFOCOM, Barcelona, Spain, Apr. 2006, pp [8] P. Juang, H. Oki, Y. Wang, M. Martonosi, L. S. Peh, and D. Rubenstein, Energy-efficient computing for wildlife tracking: design tradeoffs and early experiences with zebranet, in Proc. of ACM ASPLOS, San Jose, CA, USA, Dec. 2002, pp [9] A. Chaintreau, P. Hui, J. Crowcroft, C. Diot, R. Gass, and J. Scott, Impact of human mobility on opportunistic forwarding algorithms, IEEE Tran. on Mob. Com., pp , Jun [10] S. Burleigh, A. Hooke, L. Torgerson, K. Fall, V. Cerf, B. Durst, K. Scott, and H. Weiss, Delay-Tolerant Networking: An approach to interplanetary internet, IEEE Comm. Mag, vol. 41, 2003, pp [11] J. Partan, J. Kurose, and B. N. Levine, A survey of practical issues in underwater networks, 1 st ACM Int. Works. on Underwater Net. in Con. with ACM MobiCom, Los Angeles, California, USA, Sep. 25, 2006, pp [12] G. E. Prescott, S. A. Smith, and K. Moe, Real-time information system technology challenges for NASA s earth science enterprise, in Proc. of the 20th IEEE Real-Time Sys. Sym., Phoenix, AZ, USA, Dec [13] J.Ott, and D. Kutscher, A disconnection-tolerant transport for drive-thru internet environments, in Proc. of IEEE INFOCOM, Miami, FL, USA, Mar. 2005, vol. 3, pp [14] T. Spyropoulos, K. Psounis, and C. S. Raghavendra, Spray and wait: an efficient routing scheme for intermittently connected mobile networks, in Proc. of ACM WDTN, Philadelphia, USA, Aug.2005, pp [15] T. Spyropoulos, K. Psounis, and C. S. Raghavendra, Spray and focus: efficient mobility-assisted routing for heterogeneous and correlated mobility, in Proc. of IEEE PerCom, White Plains, NY, USA, Mar. 2007, pp [16] J. Leguay, T. Friedman, and V. Conan, Evaluating Mobility Pattern Space Routing for DTNs, in Proc. of IEEE INFOCOM, Barcelona, Spain, Apr [17] S. Grasic, E. Davies, A. Lindgren, and A. Doria, The Evolution of a DTN Routing Protocol PRoPHETv2, in Proc. of ACM MobiCom Workshop on Challenged Networks (CHANTS 2011), Las Vegas, Nevada, USA, Sep [18] T. Small, and Z. Haas, Resource and performance tradeoffs in delay-tolerant wireless networks, in Proc. of ACM WDTN, Philadelphia, PA, USA, Aug. 2005, pp [19] A. Chaintreau, P. Hui, J. Crowcroft, C. Diot, R. Gass, and J. Scott, Impact of human mobility on the design of opportunistic forwarding algorithms, in Proc. of IEEE INFOCOM, Barcelona, Spain, Apr [20] A. Keränen, J. Ott, and T. Kärkkäinen, The ONE simulator for DTN protocol evaluation, in Proc. of 2 nd Int. Conf. on Sim. Tools and Tech., Rome, Italy, Mar [21] Project page of the ONE simulator. [Online]. Available: [Accessed Nov 2014]. 167

169 Emotional Bangla Speech Signals Classification using K-NN Md. Tohidul Islam 1*, Somlal Das 2 and Md. Ekramul Hamid 3 Department of Computer Science and Engineering University of Rajshahi Rajshahi, Bangladesh *tohid@ru.ac.bd 1, somlal_ru@yahoo.com 2, ekram_hamid@yahoo.com 3 Abstract Human speech carries various kinds of emotions. The detection of emotional state of a speaker from his or her speech is crucial to provide feedback information in many applications. In this paper, we will explore how a Bangla speech emotion classification system with Mel Frequency Cepstral Coefficients (MFCC) be used to detect the emotional state of a speaker. First of all an emotional Bangla speech database is created with the help of the three actors. MFCCs features are then calculated for every frame of each of the speech signals from the emotional Bangla speech database. The features of each speech signals are pass through some aggregate functions to reduce the feature dimension. Finally, K-Nearest Neighbor (K-NN) classifier is employed to classify the emotion. The outcome of the classification system is very goods. We got 98% accuracy for anger. The overall accuracy is about 72%. Keywords Emotion Recognition, Bangla Emotional Speech, MFCC, KNN. I. INTRODUCTION Emotion detection from speech is the identification of emotional or physical state of a human being from his or her voice. Although emotional state does not alter the linguistic content, it is an important factor in human communication because it provides feedback information in many applications such as ticket reservation system, call center. Some other applications of speech emotion recognition include psychiatric diagnosis, intelligent toys, learning environment, lie detection, educational software and safe driving. Emotion detection from speech is a relatively new field of research. Researchers are still debating what features influences the recognition of emotion in speech. As mentioned in [1], there are three different categories of emotional speech: acted speech, elicited speech, and spontaneous speech. In acted speech recording, actors are invited to record utterances, where each utterance needs to be spoken with multiple emotions. This method is adopted by most researches because it can get large amount of data in a short time and the data is undistorted. In elicited speech recording, the Wizard-of-Oz (WOZ) is used. The WOZ means using a program that interacts with the speaker and drives him into a specific emotion situation and then records his voice. So designing a good program that can induce the participator to say something in the expected emotional state is a major challenge. In spontaneous speech recording, the realworld utterances that express emotions are recorded. Data got from this method has the best naturalness but it is the most difficult because we need to follow the speaker. In this case, it is required to hide the recording device to make the speaker without any pressure to present his real emotion. So the method is generally infeasible. In the recent years, a great deal of research has been done to recognize human emotions using speech information [2], [3]. Most of them were done on English language. Besides, some works also done one German, Mandarin, Telugu language. No work is done yet on Bangla emotional speech. The most common challenge is that there is no emotional Bangla Speech database. In this research work we constructed a Bangla speech database. There are a number of acoustic and prosodic features that can be extracted from speech. In this work, we have calculated the MFCC coefficients for each frame of input speech signal. The first 13 MFCC coefficients along all frames were then used to calculate the mean, variance, maximum and minimum. With the help of K-NN classification algorithms emotional speech were employed to classify any speech signal. II. CORPUS OF EMOTIONAL SPEECH DATA According to Banse & Scherer s study of vocal emotional expression [4], there are fourteen distinct emotional categories namely Disgust, Panic, Anxiety, Hot Anger, Cold Anger, Despair, Sadness, Elation, Happy, Interest, Boredom, Shame, Pride and Contempt. In this work, we used only the most traditional emotions: happy, sad, angry and neutral (no emotion). In this work, we had chosen the acted speech method to record the utterances from the speaker. To construct the Bangla database, we had chosen 62 sentences that are semantically neutral statement such as numbers and dates. The average length of the sentences are 3. For recording emotional speech, we asked 3 persons to pronounce the selected sentences in each emotional mood. The speaker tried their best to simulate each emotion. After several attempts, they were able to deliver the best emotions. Thus we finally obtained 744 emotional speech sentences (248 sentences from each person). The recording system used here was AVS Audio Editor 7.1. The sampling rate was 16 KHz and encoded in single channel 16bit PCM. All recording were transcribed by hand. After the recording procedure, a listening test was held to evaluate these recorded sentences and the result was found quite acceptable. 168

170 III. FEATURE EXTRACTION The first step in speech emotion recognition is to extract features i.e. identify the components of the audio signal that are good for identifying the emotion of the speaker. MFCCs are the most widely used spectral representation of speech in many applications, including speech and speaker recognition. They were introduced by Davis and Mermelstein in the 1980 s, and have been state-of-the-art even since. Kim et al. argued that statistics relating to MFCCs also carry emotional information [5]. Figure 1 shows the MFCC feature extraction process. As shown in Fig.1 feature extraction process contains the following steps: Framing: The speech signal is segmented into small duration blocks of 25ms known as frames. This means the frame length for a 16 KHz signal is 0.025*16000=400 samples. Frame step is usually 10ms (160 samples). Framing is required as speech is a time varying signal but when it is examined over a sufficiently short period of time, its properties are fairly stationary. Windowing: Each of the above frames are multiplied with a hamming window in order to keep continuity of the signal. Basically the spectral distortion is minimized by using window to taper the voice sample to zero at both beginning and end of each frame. FFT: Fast Fourier Transform (FFT) is a process of converting each frame of N samples from time domain into frequency domain. FFT is ideally used for evaluating the frequency spectrum of speech. Mel Filterbank and Frequency Wrapping: The Mel filter bank consists of overlapping triangular filters with the cutoff frequencies determined by the center frequencies of the two adjacent filters. The filters have linearly spaced center frequencies and fixed bandwidth on the Mel scale. Take Logarithm: The logarithm has the effect of changing multiplication into addition. Therefore, this step simply converts the multiplication of the magnitude in the Fourier transform into addition. Take discrete cosine transform: We apply DCT on the 20 log energy E k obtained from the triangular band pass filters to have L Mel-scale Cepstral coefficients. DCT transforms the frequency domain into a time-like domain called quefrency domain. These features are referred to as the Melscale Cepstral coefficients. Emotion Speech Framing Windowing FFT In our case, the output of MFCC is a 13 coefficients for each 25ms frame. So if there are 100 frames in a speech sentence, we have 100x13=1300 coefficients. For each coefficient we calculated the mean, variance, maximum and minimum across all frames. That means, for each speech signal we have 13 mean values, 13 variance values, 13 maximum values and 13 minimum values. We also calculated the mean, variance, maximum and minimum of the mean of each coefficient. This results a feature vector of length 56. Figure 2 shows the variation in the first three MFCCs for a speaker uttering Pohela September in four different emotional states. It is evident from the figure that the mean of the first coefficient is higher Magnitude Spectrum Figure 2: Variation in MFCCs for different emotional states MFCC DCT Log Mel Spectrum Log Mel Spectrum Figure 1: MFCC feature extraction Mel Filterbank and frequency wrapping when Pohela September is uttered in anger mode rather than any other modes. It is also true for the second coefficient. However in third coefficient, anger has the lowest mean. 169

171 IV. EMOTION CLASSIFICATION K-NN is a non-parametric method used for classification and regression. In both cases, the input consists of the k closest training examples in the feature space. In K-NN classification, the output is a class membership. An object is classified by a majority vote of its neighbors, with the object being assigned to the class most common among its k nearest neighbors (k is a positive integer, typically small). If k = 1, then the object is simply assigned to the class of that single nearest neighbor. In emotion classification, the speech signals are classified by using the extracted features from the feature extraction stage. The extracted features from the speech signals are given to the K-NN classifier. Figure 3 shows the block diagram of our emotion recognition system. We calculated MFCC as the emotional feature from each input data. Then, the speech was classified by pattern classification method (K-NN). Training Data Testing Data Feature Extraction (MFCC) Feature Extraction (MFCC) Classific ation & Decision Making (K-NN) Classified Result features, we also observe that anger has the highest value of mean, variance and maximum in the 1 st MFCC coefficient. Again anger has the highest value of mean, variance and maximum of the mean along all Figure 4(ii): Sample speech (noy shoto baro) in anger Figure 3: Block diagram of emotion recognition V. EXPERIMENTAL RESULTS The proposed Bangla Speech emotion classification system was implemented in the working platform of MATLAB (version ) on Windows 8.1 platform with Intel Core i3, 3.30 GHz and 8GB of RAM. Figure 4 shows the sample input in four different emotions for the utterance noy shoto baro. The corresponding extracted features relating to MFCCs is shown in fig 5. We observe that the Figure 4(iii): Sample speech (noy shoto baro) in neutral Figure 4(i): Sample speech (noy shoto baro) in happy amplitude of original anger speech signal has the highest value. If we closely look at the extracted the coefficients. Figure 4(iv): Sample speech (noy shoto baro) in sad 170

172 accuracy. Table 1 shows the confusion matrix of the emotion recognition system from Bangla speech. The rows and columns represent original and recognized emotion categories, respectively. For example, first row says that 96 sentences that belong to angry were recognized as angry, 00 sentence as happy, 00 sentence as sad and 02 sentence as neutral. So, the recognized accuracy of anger is 98%. We can see that our system do better in recognizing anger. The mean recognizing rate is 71.43%. Table 1: Confusion matrix of our System Angry Happy Sad Neutral Accuracy Angry % Happy % Sad % Neutral % VI. CONCLUSION & FUTURE WORK In this paper, we proposed a Bangla speech emotion classification system using K-NN classifier. The classification process was made by extracting features relating to MFCC from input speech signals. During the testing process, if a set of speech emotional signal is given as input it classifies the speech signals based on the features. In the future we will continue to get more emotion category into our database and improve the recognition accuracy to make it complete so that we can use it to some real world applications such as Ticket reservation system, call center. REFERENCES [1] Raquel Tato, Rocio Santos, Ralf Kompe, Emotional Space Improves Emotion Recognition, Man Machine Interface Lab, Advance Technology Center Struttgart, Sony International (Europe) GmbH. [2] Cowie, R., Douglas-Cowie, E., Tsapatsoulis, N., Votsis, G., Kollias, S., Fellenz, W., and Taylor, J. G., Emotion recognition in human-computer interaction, IEEE Signal Processing magazine, Vol. 18, No. 1, 32-80, Jan [3] D. Ververidis, and C. Kotropoulos, Automatic speech classification to five emotional states based on gender information, Proceedings of the EUSIPCO2004 Conference, Austria, , Sept Figure 5: Corresponding extracted features (noy shoto baro) The performance of the proposed Bangla emotion classification system was evaluated with different person s emotion speech and found quiet good [4] R.Banse, K.R.Scherer, Acoustic profiles in vocal emotion expression, Journal of Personality and Social Psychology, Vol.70, , 1996 [5] S.Kim, P.Georgiou, S.Lee, S.Narayanan. Real-time emotion detection system using speech: Multi-modal fusion of different timescale features, Proceedings of IEEE Multimedia Signal Processing Workshop, Chania, Greece,

173 Content Based Image Searching Using Multidimensional MSF Md. Saiful Islam 1 *, Md. Emdadul Haque 2, and Md. Ekramul Hamid 1 1 Department of Computer Science and Engineering 2 Department of Information and Communication Engineering University of Rajshahi Rajshahi, Bangladesh * msiscse@mail1.ru.ac.bd Abstract Markov Stationary Features (MSF) not only considers the distribution of colors like the histogram method, also characterizes the spatial cooccurrence of histogram patterns. However, handling large scale database of images, simple MSF method is not sufficient to discriminate the images for desired level. In this paper, we have proposed a robust content based image searching algorithm that extends the original MSF to multidimensional MSF (MMSF) by generating multiple co-occurrence matrices with different quantization levels of an image. Publicly, available Corel1000 database containing 1000 with 10 categories of images variation in color, lighting, size, and orientation, is used to evaluate the performance of the proposed algorithm. The experimental result justifies the effectiveness of the proposed method. Keywords Markov stationary feature; Markov chain; MMSF; CBIR; Content based image searching; I. INTRODUCTION With the proliferation of the world-wide-web, and the rapid increase of the multimedia data such as image and video, there is a strong demand for developing efficient techniques for storing, browsing, retrieval and indexing to exploit the full benefit of explosive growth [1-4]. In content-based image searching, images from database are automatically indexed by summarizing their low level visual contents such as color, texture, shape or spatial relationship according to user's visual requirement. Color feature is one of the most flexible and reliable visual features used in image searching methods. It is almost independent of image size and orientation and robust to its background complication. There are several methods to represent colors of an image like color histogram, color coherence vector(ccv), color auto correlogram (CAC), Markov stationary feature(msf), and so on. Color histogram [5] is an effective way to represent the colors of an image which describes the global color distribution in the image. However, the histogram comparison saturates the discrimination if the database contains a large number of images. Thus, it was inevitable to integrate spatial information of an image with its color information. In CCV [6], each histogram bins is divided into two types. It is coherent, if it belongs to a large color region or incoherent if it does not. The CCV method has better performance than color histograms if the images in the database have mostly uniform colors and the image is texture dominated. CAC method [7] is used to characterize both the color distribution of pixels and the spatial correlation of pairs of colors. Instead of the indirect use of spatial information, the CAC method encodes local spatial structure information directly into histograms. However, CAC method only take cares the information of between histograms. An effective content-based image feature called Markov Stationary Feature (MSF) was introduced [8] for image indexing, searching and classification in the context of large scale image databases. It characterizes the spatial co-occurrence of histogram patterns by Markov chain models. In the context of MSF characterization, images in the database are divided into four categories depending upon the discrimination capability of histogram analysis: histogram-level distinguishable, intra-bin distinguishable, extra-bin distinguishable and histogram undistinguishable images. To form the so-called MSF, initial and stationary distributions of the homogeneous Markov chain are combined to encode the intra-bin and extra-bin relationship of histogram, respectively. In practice, the MSF method generally outperforms the corresponding earlier content based methods. Subsequently, some research such as Directed Markov Stationary Feature (DMSF) [9], Multi-direction Markov Stationary Feature MDMSF [10], Markov Stationary Features and Vector Quantization Histogram (MSFHQ) [11] and some other methods [12-14] based on the same model have been demonstrated to enhance the performance of the original MSF. Though the MSF method and its extended versions are far better than the earlier methods, they still suffers difficulty if the underlying image database is heterogeneous. To overcome the problem, we propose multidimensional Markov Stationary Feature (MMSF) model that extends the current MSF by populating more spatial information by computing multidimensional co-occurrence matrices of an image with multiple numbers of histogram bins depending on the image quantization levels. 172

174 II. MARKOV STATIONARY FEATURE (MSF) The MSF extends the histogram features by characterizing the spatial co-occurrence of histogram patterns utilizing the Markov chain models and improves the distinguishable capability to extra-bin distinguishable level. A brief discussion of MSF is given bellow. Suppose, an image I is quantized into K levels, thus the set of histogram bins of the image is,,...,. The co-occurrence matrix containing spatial information is defined as with each element where d indicates distance between two (adjacent if d=1, that is in our case) pixels and, and accumulates the number of co-occurrence for bin and. When the pattern and have large spatial co-occurrence, the possibility that transit to is high. Note here, the co-occurrence matrix C is a nonnegative symmetric matrix and can be interpreted from a statistical view [8]. Markov chain model is adopted to characterize the spatial relationship of histogram bins, which treats bins as states in that model. The transition matrix, essential component of the chain, statistically derived from the spatial cooccurrence matrix is defined as, where The is precisely the element of the transition matrix P. The matrix P must satisfies the basic properties of a Markov chain, namely, and (3) (4) It should be noted that, every individual image can be represented by a Markov chain what should be modeled as an individual transition matrix. Thus, comparing two Markov chains (i.e. two images) means comparing the two corresponding transition matrices. However, the transition matrix is space expensive because it requires M(K 2 ) spaces for K states i.e. bins. At the same time, time complexity of comparing two transition matrices is O(K 2 ) what is impractical for a large image database. Thus, it is desirable to build up a compact yet robust feature representation from the transition matrix rather than to become the required feature itself (i.e. the full transition matrix). Based on the basic properties of the transition matrix and considering the Markov chain's two potential conditions [15], namely, irreducibility and aperiodicity, Li el al [8] formulates a space efficient (2K instead of K 2 in feature dimension) yet robust feature, exploiting the concept of Chapman- Kolmogorov equation. The so called feature representation is known as Markov stationary features [, π], which combines two K-dimensional vectors: initial distribution denoted by and stationary distribution denoted by π. The initial distribution, also known as auto-correlogram, encodes intra-bin transitions of histogram bins of the underlying image, can be obtained as, Whereas the stationary distribution encodes extra-bin transitions. With a regular Markov chain, the stationary distribution of the transition matrix denoted as, satisfying However, when the chain is irregular, there may not exist unique solution to (6). Thus, for a general case (i.e. for both regular and irregular chain), the fundamental limit theorem of Markov chain [16] would be a better solution as To mitigate the approximation error due to small n, it is a good idea to average the rows of the matrix A further as below, III. PROPOSED MULTIDIMENSIONAL MSF In the original MSF scheme discussed above, image (with K level quantization) pixels in all directions (i.e. 8-neighborhood) are counted to generate a single dimensional co-occurrence matrix (K-by-K in size). The two components (i.e. initial and stationary distribution) of the MSF are then derived based on the co-occurrence matrix. Practically, the MSF has a better searching performance on a large scale image database ranging from homogeneity to limited heterogeneity of image storages. However when the underlying image database contains moderate or high degree heterogeneous images (i.e. histogram-level distinguishable, intra-bin distinguishable, extra-bin distinguishable and histogram undistinguishable images with different size, orientation, color and light condition) the single dimensional MSF method does not show the desirable performance. It is because, with a original (i.e. single dimensional) MSF, it is possible to capture only limited spatial information what is not sufficient to discriminate the images. 173

175 In this paper we propose the so called Multidimensional Markov Stationary Features (MMSF) algorithm to address the problem by populating more spatial information in the feature components. In this method, an image I is quantized into nk (e.g. K=10) levels, thus the set of histogram bins are. Here, N indicates the dimension of our proposed MMSF method. The corresponding cooccurrence matrices are calculated as (if, K=10), ( ) Where n=1,2,..n. As example, if N=3, there will be three different co-occurrence matrices of sizes 10-by-10, 20-by-20 and 30-by-30, respectively. After computing the N different co-occurrence matrices, the Markov chain model (discussed in the previous section) is adopted for every co-occurrence matrix to characterize the spatial co-occurrence relation of corresponding set of histogram bins. Corresponding initial distribution (using (5)) and stationary distribution (using (6)) are then calculated and combined to form one dimensional MSF. The resultant MSFs for every co-occurrence matrices contribute to form our proposed MMSF feature space (K=10, in our case), as case) where and,,, when N=3 (in our each matching is then stored in an array in order to display the top matches from the database according to the ranking of similarity. V. RESULT AND DISCUSSION Our proposed method is tested on a worldwide recognized database named Corel1000 [18]. Because of the size and heterogeneity, it is axiomatic that the Corel database meets all the requirements to measure the performance of any content based image searching system. The images of the database are pre-classified into 10 different categories each with 100 in sizes. The Fig.1 shows the some sample images of each categories of the Corel1000 database. The image searching result with our proposed method is shown in Fig.2, where the top left image is a query image and the rest are retrieved relevant images. IV. SIMILARITY MEASUREMENT Assuming that no textual captions or other manual annotations of the images are given, the proper representation of the visual features of an image are then will be the proper description of the image content, such MMSF. To find images that are visually similar to the given query, it should have a measure that can determine how similar or dissimilar the different images are from the query [17]. For matching the MMSF features between the query and database images, we have selected Chi square distance measure in our simulation results. For any two feature vectors, and of the images and, respectively, the chi square distance can be defined as, ( ) The above distance formula will be used for initial and stationary distribution individually, which are then be summed for totaling. The similarity result for Fig.1. Sample images of Corel 1000 database for different categories: African, Beach, Building, Bus, Dinosaur, Elephant, Flower, Horse, Mountain, and Food. Fiq.2. Searching Result of the proposed method obtained for the query image-47. To validate the general performance of histogram, Color Auto Correlogram (CAC), original MSF method and our proposed MMSF method, we have 174

176 used two standard metrics, namely, recall and precision, which are defined as, The comparative searching results of the original histogram, color auto correlogram, original MSF methods and our MMSF technique in terms of precision-recall curves for a single query image (e.g. African ) are shown in Fig.3, respectively. The average precision-recall curves of histogram, color auto correlogram, original MSF and the proposed MMSF methods for some 25 randomly selected query images from the Corel 1000 database are summarized in Fig.4. The Fig.3 illustrates the better performance of the proposed MMSF method compare to the other existing techniques. From the average precision-recall curves in Fig.4, it is clear that the proposed method obviously outperforms the other techniques, which validates the effectiveness of our method, since the spatial co-occurrence information of the different number of color quantization of the same image is exploited. ACKNOWLEDGMENT This work is supported by the Information and Communication Technology Division, Ministry of Posts, Telecommunication and Information Technology of Bangladesh, under the ICT-fellowship program. Fig.3. Precision-recall curves for Histogram, CAC, original MSF, and proposed MMSF method for the query image-47. Fig.4. Average Precision-recall curves of Histogram, CAC, original MSF, and proposed MMSF for some 20 randomly selected images from Corel 1000 database. REFERENCES [1] A. N. Tikle, C. Vaidya and P. Dahiwale, "A Survey of Indexing Techniques For Large Scale Content-Based Image Retrieval,"Electrical, Electronics, Signals, Communication and Optimization (EESCO)," International Conference on, Visakhapatnam, 2015, pp [2] K. Juneja, A. Verma, S. Goel and S. Goel, "A Survey on Recent Image Indexing and Retrieval Techniques for Low- Level Feature Extraction in CBIR Systems,"Computational Intelligence & Communication Technology (CICT)," IEEE International Conference on, Ghaziabad, 2015, pp [3] Keyuri M. Zinzuvadia1, Bhavesh A. Tanawala, Keyur N. Brahmbhatt, "A Survey on Feature Based Image Retrieval Using Classification and Relevance Feedback Techniques", International Journal of Innovative Research in Computer and Communication Engineering, January 2015, vol. 3, issue 1. [4] Mussarat Yasmin, Sajjad Mohsin, Muhammad Sharif, "Intelligent Image Retrieval Techniques: A Survey," Journal of Applied Research and Technology, 2014, vol. 12, pp [5] G. Pass, R Zabih, "Histogram Refinement for Content-Based Image Retrieval," IEEE Workshop on. Application of Computer Vision, 1996, pp [6] G. Pass, R Zabih, and J. Miller, "Comparing Images Using Color Coherence Vectors," Proc. ACM Multimedia, 1997, pages [7] J. Huang, S. Kumar, and et al, "Image Indexing Using Color Correlograms," Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 1997, p [8] J. Li, W. Wu, T. Wang and Y. Zhang, "One Step Beyond Histograms: Image Representation Using Markov Stationary Features," Proc. CVPR. IEEE, [9] B. Ni, S. Yan, and A. Kassim, "Directed Markov Stationary Features for Visual Classification," Proc. of IEEE International Conference on Acoustics, Speech, and Signal Processing, [10] F. Lee, K. Kotani, Q, Chen, and T. Ohmi, "Face Recognition Algorithm Using Multi-directional Markov Stationary Features and Adjacent Pixel Intensity Difference Quantization Histogram," Proc. of the 7th International Conference on Systems and Networks Communications, [11] Qiu Chen, Koji Kotani, Feifei Lee, Tadahiro Ohmi, "Face Recognition Using Markov Stationary Features and Vector Quantization Histogram," Proc. of IEEE 17th International Conference on Computational Science and Engineering, 2014, pp

177 [12] Y. Song, Xiao Chen, S, Qu, "Content based Image Retrieval with Color Invariants," Proceedings of the 2nd International Conference on Computer Science and Electronics Engineering, China, 2013, pp [13] C. Zhang, J. Liu, H. Lu and S. Ma, "Web image mining using concept sensitive Markov stationary features," Multimedia and Expo, ICME IEEE International Conference on, New York, NY, 2009, pp [14] Feifei Lee, K. Kotani, Chen Qiu, T. Ohmi, "A Robust Face Recognition Algorithm Using Markov Stationary Features and Adjacent Pixel Intensity Difference Quantization Histogram," Seventh IEEE International Conference on Signal-Image Technology and Internet-Based Systems (SITIS), 2011, pp [15] Olle Häggström, "Finite Markov Chains and Algorithmic Applications," Handbook, Cambridge University Press, [16] Karl Sigman, "Stationary Marked Point Processes: An Intuitive Approach (Stochastic Modeling Series)," Handbook, Published by Taylor Francis Ltd, United States (1995) [17] Yossi Rubner, Carlo Tomasi, "Perceptual Metrics for Image Database Navigation," e-book, 1st edition, ISBN , [18] Corel 1000 image databases for research comparison. Link: 176

178 Silicon Nanocrystals based Schottky Junction Solar Cell Fabrication and Characterization A.T.M. Saiful Islam 1, Arifuzzaman Rajib 3 Dept. of Applied Physics, Electronics and Communication Engineering Bangabandhu Sheikh Mujibur Rahman Science and Technology University Gopalganj, Bangladesh saifulslm6@gmail.com 1 Abstract A photovoltaic cell can be developed from the Schottky junction between a semiconductor and a metal, with or without an insulating layer between them. This work reporting a single-layer Schottky photovoltaic device that was fabricated by drop-casting silicon nano-crystals (Si-NCs) from colloidal solution.an optoelectronic Schottky Junction solar cell with Aluminium/Silicon Nano-crystals/Indium-Tin-Oxide structure on glass substrate has been investigated with and without lanthanum fluoride insulating layer. The effect of photoactive layer (Si-NCs) thickness on absorption and photo-luminance has been studied. The surface morphology of Si-NCs layer was investigated by Scanning Electron Microscope (SEM).The currentvoltage (I-V) characterization has been studied with and without LaF3 at different parameters. Keywords Schottky junction; silicon nanocrystal;electrode; photovoltaic; I. INTRODUCTION Historically, conventional solar cells were built from inorganic materials such as silicon. Although the efficiency of such conventional solar cells is high, very expensive materials and energy intensive processing techniques are required. Current solar cells are expected to last for years at temperatures between 20 C and 90 C, which organic solar cells may not be able to achieve [1]. Compared to inorganic semiconductors, carrier mobility in organic solar cells is still low. New ways of manufacturing solar cells that can scale up to large volumes and low cost are required. One of the inorganic solar cell is silicon nanocrystal based schottky junction solar cell that can be alternative for conventional silicon solar cell. Lead selenide (PbSe) [2] and lead sulfide (PbS) [3] nanocrystals were used in Schottky junction solar cells. These devices are infrared harvesting and can utilize the broad solar spectrum. The efficiency of the nano-crystals based schottky junction solar cell can be increased by using buffer layer. In this work a thin layer of ITO was fabricated on the glass substrate using electron beam evaporation technique. After that a layer of Si-NCs was developed on that ITO coated glass by drop casting deposition of Si-NCs. Then a thin layer of aluminium was developed over the Si-NCs layer by using electron beam evaporation technique. The aluminium layer MD. Enamul Karim 2, Abu Bakar Md. Ismail 4 Dept. of Applied Physics and Electronic Engineering Rajshahi University Rajshahi, Bangladesh was used as cathode and ITO layer was used as the transparent anode. The effect of photoactive layer thickness on the absorption was studied by using U-V spectrometer.laf3 was also studied as buffer layer of the fabricated schottky junction solar cell. II. EXPERIMENTAL The experimental procedures to fabricate this solar cell are divided in following steps A. Cleaning of the glass substrate. B. Deposition of thin films by E-beam. C. Preparation of colloidal solution of Si-NCs with DCB. D. Final device fabrication and encapsulation. E. Device characterization A. Cleaning of the glass substrate Before using the glass substrates to deposit the layers onto it, they were cleaned by the following steps: a) Initially the glass substrate were dipped and vibrated in acetone by Ultrasonic Vibrator for 10 minutes to remove organic residues. b) Then the substrates were cleaned with de-ionized (DI) water. c) Again the substrate were dipped and sonicated in acetone by Ultrasonic Vibrator for 10 minutes. d) Again the glass substrates were cleaned with deionized water. e) And finally the substrates were dried in hot air. B. Deposition of films ITO thin films were deposited on a glass substrate at room temperature by electron beam evaporation technique using an Edwards E 306.The various quantities such as source to substrate height, deposition rate, beam current, chamber pressure, quality of substrate, substrate temperature, size of atomized particles etc. affect the film properties. Very thin layer of ITO (~ 100 nm) was deposited onto a glass substrate by electron-beam evaporation process. After ITO deposition, the samples were annealed in a furnace (Carbolite CWF 12/13) in air at 600 C for 10 min [4]. In the same manner a layer of LaF3 of about 5nm, followed by an aluminum layer of thickness 177

179 about 50μm was deposited by electron-beam evaporation process. C. Preparation of colloidal solution of Si-NCs Silicon Nanocrystals can be fabricated through a variety of techniques including ion implantation [5 6], aerosol synthesis [7 8], ion beam co-sputtering [9 10]. In this work silicon nano-crystals were fabricated using electrochemical etching of silicon wafer. Si-NCs were found to form stable suspensions in 1, 2-dichlorobenzene (DCB). For the formation of Si-NC colloids, the following protocol was used: DCB was dried with molecular sieves and degassed by bubbling nitrogen for 60 min in a Schlenk line to reduce the oxygen and moisture level. Si-NCs were dispersed in DCB (ACROS) to form a cloudy, yet stable solution. The solution concentration was at 4 mg Si-NCs/ml DCB. To reduce Si-NC agglomeration in the solution, the cloudy solution was sonicated for different time such as 30 min, 40 min, 50 min, and 60 min. D. Device fabrication A schematic of the layered structure of fabricated Schottky junction solar cells (with and without buffer layer) are shown in Fig. 1. (a) annealed at 80 C for 10 min to dry the layer. Then the samples were cooled down to room temperature. In the next step, a very thin layer of LaF3 is deposited over the Si-NCs layer by E-beam evaporation technique. After that a layer of aluminum (thickness about 50μm) was deposited as the cathode of the final device. Finally the samples were annealed in a thermal annealing furnace (carbolite CWF 12/13) in air at 150 C for 20 min to make the good connections at different layers of the fabricated device. After annealing they were left to be cooled naturally to the room temperature. E. Device characterization Hitachi S-3400N scanning electron microscopy (SEM) imaging was used to study the surface morphology of the drop casted Si-NCs layers. The dependency of the absorption on layer thickness of the Si-NCs was studied using UV spectrophotometer of model SHIMADZU UV-1650PC. Photo-current versus voltage (I-V) measurement was performed with a Keithley model 2400 source meter and a solar simulator system. III. RESULT AND DISCUSSION The effect of Si-NCs size along with ITO layer on absorption was studied using U-V spectrophotometer with the wavelength range of nm. Fig. 2 shows the effect of absorption over wavelength for Si-NCs layer with different sonication time (which is correspond to the nano-crystals size). The absorption spectra of thin films obtained by drop-casting a colloidal solution of Si-NCs with DCB showed a significant change when the sonication time is changed. From this figure it is clear that the absorbance of the Si-NCs layer increases with increasing of exility of the Si-NCs. (b) Fig. 1. Schematic of the layered structure of schottky junction solar cells (a) without LaF3 buffer layer, (b) with LaF3 buffer layer. The Fabrication of Schottky junction solar cells starts with an indium tin oxide (ITO) film coated on a glass substrate. Traditionally indium tin oxide (ITO) film has been used as a transparent conducting electrode in Schottky junction solar cell [2, 11].After the deposition of ITO layer the samples were annealed at 600 C for 10 min to reduce the surface roughness and to increase the transparency. In next stage the colloidal solution of Si-NC and DCB is applied on ITO coated glass by drop-casting method with micropipette. After drop-casting, the Si-NCs layer with DCB on ITO coated glass substrate was Fig. 2 Absorption versus wavelength curve of Si-NCs layer at different sonication time (correspond to different particle sizes). As mention earlier the surface morphology of the Si-NCs deposited over the ITO coated glass was study by the Scanning Electron Microscope (SEM). Fig. 3 shows the surface structure of the Si-NCs on 178

180 Current (ma) Current (ma) Current (ma) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering ITO coated glass having sonication time 60 min with different magnification. From the above figure, the NCs are clearly detected. It was also observed that the NCs sizes were in the diameter range from 50nm-130 nm at the 60 min sonication time. (around 15 times compared to without LaF3 structure) in the photo-generation and collection that is reflected in the enhancement of the reverse current of the device. Both I-V characteristics curve indicates poor short circuit current. This is because of the recombination of photo generated charges before they are reached at collectors. 4 3 Data Point 2 (a) (b) (c) Fig. 3 Surface morphology of Si-NCs layer without LaF3 at sonication time 60 min with magnification (a) 500 (b) 5000 (c) and (d) EDX studies were used to analyze the elemental composition of the Si-NCs deposited over ITO coated glass substrate. Fig. 4 indicates the strong presence of silicon with minor presence of some other compositional particles. (d) Voltage (V) Fig. 5 The I-V characteristics of the fabricated schottky junction solar cell in absence of light Voltage (V) Fig. 6 The I-V characteristics of the fabricated schottky junction solar cell under 1.5 AM light without LaF3 buffer layer Fig. 4 EDX spectra for Si-NCs layer deposited on ITO coated glass. The I-V characteristics of the designed Al/Si- NCs/ITO without LaF3 were studies under forward and reverse bias condition. The dark I-V characteristic of the fabricated schottky junction solar cell (Fig. 5) shows a typical rectifying junction behavior with threshold voltage of 0.3V. The reverse I-V characteristics under 1.5AM simulated light illumination were investigated without LaF3 (Fig. 6) and with LaF3 (Fig. 7) buffer layer to find the influence of the buffer layer on the performance of fabricated schottky junctionphotovoltaic cell. The I-V characteristic with LaF3 buffer layer clearly indicates an enhancement Voltage (V) Fig. 7 The I-V characteristics of the fabricated schottky junction solar cell under 1.5 AM light with LaF3 buffer layer. IV. CONCLUSIONS Silicon Nano Crystals can be fabricated at low temperature from solution processing without any vacuum equipment or high-temperature processing. Compared to Schottky junction solar cells reported 179

181 previously, the whole fabrication method (in this work) for preparing Schottky junction photovoltaic is simple, cheap and fast.the performance parameters such as, JS C, VO C, FF and conversion efficiency was low for our fabricated cells. This was due to fabrication process used to fabricate the solar cell, processing environment, quality of the Si-NC etc. ACKNOWLEDGMENT This work was funded by the Department of Applied Physics and Electronic Engineering, University of Rajshahi, Rajshahi 6205, Bangladesh. We acknowledge the help of Bangladesh atomic energy commission, Dhaka for the SEM and EDX characterization. We also acknowledge the help of central science laboratory of University of Rajshahi for UV characterization support. REFERENCES [1] Antonio A, HegedusS(eds.), editor. Handbook of Photovoltaic Science and Engineering. Wiley, Chichester; [2] Luther, J. M., Law, M., Beard, M. C., Song, Q., Reese, M. O., Ellingson, R. J., Nozik, A. J., Schottky Solar Cells Based on Colloidal Nanocrystal Films, Nano Letters,Volume 8, pp , [3] Keith W. Johnston, Andras G. Pattantyus-Abraham, Jason P. Clifford, Stefan H. Myrskog, Dean D. MacNeil, Larissa Levina, and Edward H. Sargent, Schottky-quantum dot photovoltaics for efficient infrared power conversion, Applied Physics Letters, Volume 92, pp , April [4] A.T.M. Saiful Islam, Mushtaq Ahmed Sobhan and Abu Bakar Md. Ismail, The effect of introducing Macroporous Silicon as the Cathode of Bulk Heterojunction Hybrid Solar Cell, International Conference on Materials, Electronics & Information Engineering (ICMEIE), June-2015, Rajshahi, Bangladesh. [5] T. Shimizu-Iwayama, K. Fujita, S. Nakao, K. Saitoh, T. Fujita, and N. Itoh, Visible photoluminescence in Si+ - implanted thermal oxide films on crystalline Si, Journal of Applied Physics, Volume 75, pp , [6] K. S. Min, K. V. Shcheglov, C. M. Yang, Harry A. Atwater, M. L. Brongersma and A. Polman, Defect-related versus excitonic visible light emission from ion beam synthesized Si nanocrystals in SiO 2, Applied Physics Letters, Volume 69, pp 2033, [7] K. A. Littau, P. J. Szajowski, A. J. Muller, A. R. Kortan, and L. E. Brus, A luminescent silicon nanocrystal colloid via a high-temperature aerosol reaction, Journal of Physical Chemistry,Volume 97, pp , [8] D. M. Holunga, R. C. Flagan, and H. A. Atwater, A Scalable Turbulent Mixing Aerosol Reactor for Oxide-Coated Silicon Nanoparticles, Industrial & Engineering Chemistry Research,Volume 44, pp , [9] Q. Zhang, S. C. Bayliss, and D. A. Hutt, Blue Photoluminescence and Local-Structure of Silicon Nanostructures Embedded in Sio 2 Matrices, Applied Physics Letters, Volume 66, pp , [10] J. U. Schmidt and B. Schmidt, Investigation of Si nanocluster formation in sputter-deposited silicon sub-oxides for nanocluster memory structures, Materials Science and Engineering, B, Volume 101, pp28-33, [11] J. Tang, X. Wang, L. Brzozowski, D. A. R. Barkhouse, R. Debnath, L. Levina and E. H. Sargent, "Schottky quantum dot solar cells stable in air under solar illumination," Advanced Materials, volume 22, pp ,

182 Fabrication and Characterization of α-fe 2 O 3 Homo-junction Photocathode for Efficient Solar Water Splitting Arifuzzaman Rajib Department of Applied Physics, Electronics and Communication Engineering University of Bangabandhu Sheikh Mujibur Rahman Science and Technology Gopalganj, 8100, Bangladesh rajib.apee.38@gmail.com A.T.M. Saiful Islam Department of Applied Physics, Electronics and Communication Engineering University of Bangabandhu Sheikh Mujibur Rahman Science and Technology Gopalganj, 8100, Bangladesh Atowar Rahman and Abu Bakar Md. Ismail Department of Applied Physics and Electronic Engineering University of Rajshahi, Rajshahi 6205, Bangladesh Abstract This work report, the fabrication and characterization of homo-junction photocathode from a scientifically attractive material hematite (α- Fe 2 O 3 ) with the aim of efficient solar water splitting by using it as a familiar photo-electrochemical cell. To make a homo-junction of α-fe 2 O 3, which is naturally, n- type, Zn was doped into α-fe 2 O 3 (Zn: α-fe 2 O 3 ) to turn it into p-type by the simple spin coating method. The semiconductor type of α-fe 2 O 3 and Zn: Fe 2 O 3 were confirmed by Hall measurement. X-ray diffraction (XRD) analysis of Zn:α-Fe 2 O 3 filmsrevealed the presenceof Zn in α-fe 2 O 3. The doping concentration was also tuned for the long time stability of material properties like semiconductor type and also found that lightly doped Zn: α-fe 2 O 3 retained its p-type conduction for comparatively long time. The optical band gap of α- Fe 2 O 3 and Zn:α-Fe 2 O 3 were also studied with the help of spectroscopic measurement to find the optimum condition for homo-junction photocathode. The thickness of α-fe 2 O 3 and Zn: α-fe 2 O 3 was also tuned for the optimum output. The fabricated photocathode was tested in a home-made experimental set-up with twoelectrode system. In this experimental set-up, platinum electrode used as a counter electrode and 1M alkaline solution used as electrolyte. I-v characteristics of photocathode clearly show a rectifying behavior with and without water. Index Terms Photo-catalysis, Metal-oxide semiconductor, Homo-junction photocathode (keywords) I. INTRODUCTION Production of H 2 by water splitting using sun light has become of great interest among the researchers [1-3]. Currently H 2 is produced by the use of fossil fuels as the main feedstock. The main drawback of this process is pointed out as the main cause of CO 2 emission and serious global warming [4]. Therefore, solar water splitting is considered as an ideal technology to harvest and store abundant solar energy as clean and easily transportable hydrogen fuel. In comparison with photo-catalytic water splitting using homogeneous semiconductor, photoelectrochemical (PEC) water splitting possesses great advantages as follows: (i) the external or self-bias voltage can be suppressed by recombination of photogenerated charge carriers and as a result, improve the charge separation as well as transfer rate; (ii) H 2 and O 2 can be easily separated via collection at different photo-electrodes; (iii) last, but not the least, semiconductor films are deposited on the conductive substrates, which favors scale up for industrial application in the future. In practice, the performance of solar water splitting is dominated by the properties of the semiconductor photo-catalysts which harvest solar energy for the production of H 2 fuels. In the case of splitting of water using sunlight into H 2 and O 2, a theoretical minimum potential of 1.23 ev is required. But, in practice the semiconductor having band gap of 1.23 ev is not sufficient to perform the reaction due to entropic losses, over-potentials caused by non-ideal catalysis and other parasitic losses. In reality, therefore, the band gap of about 2 ev is required for the splitting of water process using only one semiconductor [5]. Few semiconducting materials having high band gaps are actually able to split water. In this tandem configurations, H 2 is formed into the cathode by the reduction phenomena through the hydrogen evolution reaction (HER) while oxidation of water occurs at the anode to form O 2 through the oxygen evolution reaction (OER). Some semiconductors such as Si, Cu 2 O, TiO 2, Fe 2 O 3, have been shown to be good candidates for running photo-electrochemical HER. Owing to its low band gap of about ev, α-fe 2 O 3 have been used to fabricate homo-junction photo diode. Moreover, α-fe 2 O 3 is able to absorb all UV light and the range of visible light from 295 to 565 nm, which comprises 38% of the photons of sunlight at air mass (AM) 1.5 [6, 7]. Even though the band gap of Fe 2 O 3 is suitable to allow absorption of about 38% of sunlight, its photo-response is not favorable, mainly due to its low conductivity and consequent recombination of photo-generated carriers. Besides, α-fe 2 O 3 is naturally abundant in the earth s crust and is therefore, a low cost material. It can be shown also corrosion-resistant in acidic and alkaline medium [8]. On account of these remarkable characteristics, many researchers have studied the potential of using hematite for the formation of 181

183 Intensity (arb. unit) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering homo-junction photo-cathode to the purpose of conversion of solar energy into H 2 fuels are prepared by different routes, e.g. pressing the material into pellets [9-11], sputtering [12], spray pyrolysis [8, 13-15], chemical vapor deposition [16], laser assisted CVD [17], sol-gel [18] and spin coating [19-21]. In this paper, we report a promising method to fabricate α-fe 2 O 3 and Zn: Fe 2 O 3 films and finally α-fe 2 O 3 homo-junction photo-cathode with great reproducibility and homogeneous structure, deposited by simple conventional spin coating method. II. EXPERIMENTAL METHODOLOGY The p-n α-fe 2 O 3 homo-junctions were prepared by a simple two-step spin coating method on ultrasonically cleaned fluorine doped tin oxide (FTO) (300 nm thick with a sheet resistance of ~8Ω/sq.) coated glass. In the first step, 1.0M α-fe 2 O 3 (prepared from FeCl 2 + urea + ethanol) was deposited on FTO substrate with rotational speed of 7000 rpm, then annealed for two-hours at 450 ºC in air. In second step, 0.005M Zn doped Fe 2 O 3 (prepared from FeCl 2 + zinc acetate + ethanol) was deposited on the α-fe 2 O 3 films with rotational speed of 7000 rpm, and finally annealed for four-hours at 450 ºC in air. The fabricated photocathode was tested in a home-made experimental set-up with two electrode system (Fig-1) consisting of a α-fe 2 O 3 /Zn: α-fe 2 O 3 sample as the working electrode and a Pt plate (~1 1 cm 2 ) as the counter electrode, and 1M NaOH as the aqueous electrolyte. The crystal structure of α-fe 2 O 3 and Zn: α- Fe 2 O 3 were studied by X-ray diffractometer (XRD, Shimadzu, XRD-600 X) with CuKα (λ = Å) radiation in the 2θ range of 10-70º (scan speed 2º/min). The X-ray tube was operated at 60KV/30 ma. The carrier concentration and conduction type of α-fe 2 O 3 and Zn:α-Fe 2 O 3 were studied by Hall measurement with the magnetic field of 1T and the fixed source voltage of 10V. The film was placed in the middle of two circular magnetic bar with the help of stand. The use of sunlight to drive the solar fuels production of our fabricated homo-junction photocathode was studied by Keithley 2400 source meter with 100 points, the source voltage at -1 V~ +2V and sweep delay at 1000ms. Figure 1: Experimental setup for water splitting. III. RESULTS AND DISCUSSIONS Figure 2 and 3 shows the XRD spectra of the α-fe 2 O 3 and Zn:α-Fe 2 O 3 films, respectively. As it is observed from Figure 2 that there are six sharp peaks with different crystal orientation. According to literatures [20-25], six major diffraction signals are identified as the (102), (104), (110), (113), (024) and (116) planes of pure α-fe 2 O 3. Among six peaks, the peak from (104) plane is identified as the strongest. The grain size can be found with the help of Scherrer s equation D = 0.9λ/βcosθ (1) where D is the mean grain size, λ is the X- ray wavelength, β is the full width at half maximum (FWHM), and θ is the Bragg angle. From this equation, the mean grain size calculated from the (104) peak was found to be 26 nm <102> <104> <110> <113> - Fe 2 O 3 <024> <116> (degree) Figure 2: XRD spectra of un-doped α-fe 2O 3 thin films prepared by simple spin coating method. From the XRD spectra of Zn:α-Fe 2 O 3 films in Figure 3, it is observed that there are eleven sharp upward XRD peaks with different crystal orientation. 182

184 Current (ma) Intensity (arb. unit) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering As it is mentioned above, all the peaks marked as black triangle can be identified as the different planes of pure α-fe 2 O 3. Previous report suggests that all the other XRD peaks marked as red triangle are from the different planes of ZnFe 2 O 4 structure. The peak of α- Fe 2 O 3 and Zn:α-Fe 2 O 3 are stronger in lower Bragg angle Fe 2 O 3 ZnFe 2 O (degree) Figure 3: XRD spectra of zinc doped α-fe 2O 3 thin films prepared by simple spin coating method. TABLE 1: VARIOUS PARAMETERS OF UN-DOPED AND ZINC DOPED HEMATITE THIN FILMS. Film Condition Layer 1, Un-doped Zn Layer 3, Un-doped Zn Layer 5, Un-doped Zn Layer 1, M- Zn Layer 3, M- Zn Layer 5, M- Zn Layer 1, 0.005M- Zn Layer 3, 0.005M- Zn Layer 5, 0.005M- Zn Layer 1, 0.01M-Zn Layer 3, 0.01M-Zn Layer 5, 0.01M-Zn Carrier Concentr ation (x /cm 3 ) at 550ºC Carrier Concentr ation (x /cm 3 ) at 450ºC Resistiv ity (Ω-m) at 550ºC Resisti vity (Ω-m) at 450ºC Semicon ductor Type N-type N-type N-type P-type P-type P-type P-type P-type P-type P-type P-type P-type Table 1 shows data obtained from Hall measurement of different α-fe 2 O 3 and Zn:α-Fe 2 O 3 samples. Different parameters like carrier concentration, conduction type and resistivity were also studied by the Hall measurement. Un-doped α- Fe 2 O 3 in all thickness was found to be n-type whereas Zn: α-fe 2 O 3 was found to be p-type. Higher carrier concentration is found in the sample prepared at comparatively lower annealing temperature of 450ºC; as a result, the lower annealing temperature of 450 ºC was taken to fabricate homo-junction photo-cathode. As well as it is also found from the experiment that lightly doped Zn: α-fe 2 O 3 retained its p-type conduction for comparatively long time and the carrier concentration was also high. As a result zincacetate concentration of M was taken to fabricate homo-junction photocathode. It is also observed from experimental data of Hall Effect that conduction conformity as well as carrier concentration was lower in low thickness. This is the fact that layer 1 film looks like one dimensional, the dimensional increases with the thickness. Consequently, the mean free path of the deposited film is also increased. As a result, the conformity of conduction type of p-type semiconductor is also upgraded. For the taken of these advantages, 5 layer of α-fe 2 O 3 /Zn: α-fe 2 O 3 was used to fabricate homojunction photocathode for the solar water splitting Dark current (without water) Light current (without water) Dark current (with water) Light current (with water) Bias voltage (V) Figure 4: I-V characteristics for 5 layer of n/p into the water and without water. Figure 4 shows the I-V characteristics of homo-junction photo-cathode in both with and without water. I-V characteristics of photocathode clearly show a rectifying behavior for both with and without water. The dotted line was ascribed the dark current and light current of without water. It is observed from Figure 4 that there is no physical difference of dark and light I-V curve for without water. It may be due to the high sheet resistance of n and n + region. n + region are the n type FTO coated glass. The combined sheet resistance of n and n + 183

185 protect the illumination of light. As a result, in our experimental case, the I-V characteristics of dark and light current without water are obeyed same manner. It can be observed from Figure 4 that in the matter of with water, the I-V characteristics are not same for dark current and light current. The knee voltage of homo-junction photocathode for light conditions was comparatively smaller than the dark condition. In light with water, there are two dominant factor of solar light as well as charges come from electrolyte that can be able to accelerate the charges, and which in turn decreases the knee voltage of fabricated homo-junction photo-cathode. It is observed from the I-V characteristic of the junction that the knee voltage of homo-junction photocathode in the water was smaller than without water. Reduction of knee voltage in the water can be interpreted by the fact that some of the charges came from electrolyte were accumulated on the junction, and which in turn decreases the knee voltage. IV. CONCLUSIONS Naturally n-type α-fe 2 O 3 have successfully converted into p-type α-fe 2 O 3 by doing Zn. XRD spectra of Zn-doped α-fe 2 O 3 films confirm the presence of Zn in α-fe 2 O 3, whereas the evidence of p- type conversion has been confirmed by Hall measurement. The experimental results suggest that the α-fe 2 O 3 /Zn: α-fe 2 O 3 homo-junction structure can be fabricated and successfully used as photocathode for the efficient solar water splitting. Future prospects of this work is to find a more efficient homo-junction photo-cathode for solar water splitting by varying other parameters, likes, ph value of electrolyte, converting the formation of homojunction photo-cathode like Zn: α-fe 2 O 3 / α-fe 2 O 3, etc. REFERENCES [8] Saroj Kumari, Chanakya Tripathi, Aadesh P. Singh, Diwakar Chauhan, RohitShrivastav, SahebDass, and Vibha R. Satsangi, Journal of Current Science, 91(8), , (2006). [9] Leygraf, C., Hendewerk, M. and Somorjai, G. A, J. Catal, 78, , (1982). [10] Gurunathan, K. and Maruthamuthu, P., J. Hydrogen Energy, 20, , (1995). [11] Aroutiounian, V. M., Arakelyan, V. M., Shahnazaryan, G. E., Stepanyan, G. M., Turner, J.A. and Khaselev, O., Int. J. Hydrogen Energy, 27, 33 38, (2002). [12] Virtanen, S., Schmuki, P., Böhni, H., Vuoristo, P. and T., J. Electrochem. Soc., 142, , [1995]. [13] S. S. Shinde, R. A. Bansode, C. H. Bhosale, and K. Y. Rajpure, Journal of Semiconductor, 32(1), (1-8), (2011). [14] William B. Ingler Jr. John P. Baltrus, and Shahed U. M. Khan, J. AM. CHEM. SOC. 126, , ( 2004). [15] K. Ravichandran, K. Subha, N. Dineshbabu, and A. Manivasaham, Journal of Alloys and Compounds, 656, (2016) [16] FlavioL.Souza a, KirianP.Lopes a, PedroA.P.Nascente b, EdsonR.Leite, journal of Solar Energy Materials &SolarCells 93, ,(2009). [17] Kan-Rong Lee, Ya-Ping Hsu, Jeng-Kuei Chang, Sheng-Wei Lee, Chung-Jen Tseng, Jason Shian-Ching Jang, Int. J. Electrochem. Sci., , (2014). [18]H.Miyake,H.Kozuka,J.Phys.Chem.B, 109, , (2005). [19] Chun-Lin Kuo, Yu-Kuei Hsu, and Yan-Gu Lin, Solar Hydrogen and Nanotechnology IX, Vol. 9176, 91760N (2014). [20] FlavioL.Souza, KirianP.Lopes, PedroA.P.Nascente, EdsonR.Leite, Solar EnergyMaterials&SolarCells 93, , (2009). [21] Kan-Rong Lee, Ya-Ping Hsu, Jeng-Kuei Chang, Sheng-Wei Lee, Chung-Jen Tseng,, Jason Shian-Ching Jang, Int. J. Electrochem. Sci., 9, , (2014). [22] Asif Ali Tahir, K. G. Upul Wijayantha, Sina Saremi- Yarahmadi, Muhammad Mazhar, and Vickie McKee, Chem. Mater., 21, , (2009). [23] Alan Kleiman-Shwarsctein, Muhammad N. Huda, Aron Walsh, Yanfa Yan, Galen D. Stucky, Yong-Sheng Hu, Mowafak M. Al-Jassim, and Eric W. McFarland, Chem. Mater, 22, , (2010). [24] Li Fu, Hongmei Yu, Yongkun Li, Changkun Zhang, Xunying Wang, Zhigang Shao and Baolian Yia, Phys.Chem.Chem.Phys., 16, 4284, (2014). [25] M. Mohapatra and S. Anand, International Journal of Engineering, Science and Technology, Vol. 2, No. 8, pp , (2010). [1] Mauro Malizia, Brian Seger, IbChorkendorff and Peter C. K. Vesborg, J.Mater.Chem.A, 2, (2014). [2] Xiaoping Chen, Zhixiang Zhang, Lina Chi, AathiraKrishnadas Nair, WenfengShanggua, Zheng Jiang, Journal of Nano-Micro Lett, DOI /s , (2015). [3] Rui Liu, ZhiZheng, Joshua Spurgeon, and Xiaogang Yang, Journal of Energy Environ. Sci. 7, , (2014). [4] Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, ed. Core Writing Team, R. K. Pachauri and A. Reisinger, IPCC,Geneva, Switzerland, p. 104, (2008). [5] M. G. Walter, E. L. Warren, J. R. McKone, S. W. Boettcher, Q. Mi, E. A. Santori and N. S. Lewis, Chem. Rev. 110, , (2010). [6] William B. Ingler Jr., John P. Baltrus, and Shahed U. M. Khan, J. AM. CHEM. SOC., 126, , (2004). [7] (a) Komp, R. J. Practical PhotoVoltaics: Electricity From Solar Cells, 3rd ed.; Aatec Publications: Ann Arbor, MI, (b) Wehrli, C. Extraterrestrial Solar Spectrum; Physikalisch- MeterologischesObservatorium World Radiation Center (PMO/WRC), (c) Bird, R. E.; Hulstrom, R. L.; Leis, L. J. Sol. Energy, 30, 563(1983). 184

186 A Practical Approach to Spectrum Analyzing Unit using RTL-SDR Md. Habibur Rahman, Md. Mamunoor Islam Department of Electrical and Electronic Engineering Chittagong University of Engineering & Technology, Chittagong-4349, Bangladesh Abstract In the present scenario, there has been an immense advancement in the field of wireless communication in this modern engineering world. Nowa-days Software Defined Radio (SDR) technology is an indisputable emerging technology and presents new challenges for communications engineers. The advancement of SDR system has made significant progress in recent years which makes it as a serious substitute of traditional hardware radio architectures where the mathematical procedures are obligatory to decode and process radio signals using analogue circuitry. Recently, computers have turned out to be powerful enough to do the required mathematical calculations using software. So aim of this paper is to demonstrate a RTL-SDR based spectrum analyzer which can be used proficiently as an alternative of existing hardware spectrum analyzer. This approach will lessen the complexity of analogue hardware system with the higher tractability of software based filtering and demodulation techniques. As RTL-SDR devices are quite cheap (Approximately 20$) and small sized, this system also offers cost effectiveness with provision of portability. An experimental study was conducted with suitable conditions to examine the feasibility and efficiency of the proposed system. The outcome of experimental result is thoroughly examined in this paper. Keywords Software defined radio (SDR); RTL-SDR dongle; Spectrum analyzer; Tuner IC; Waterfall display I. INTRODUCTION Traditionally spectrum analyzing activities are mainly performed by governmental agencies where expensive specialized hardware setups are used [1]. With these radio architectures, a special receiver for almost each radio communication standard is needed. Moreover in many devices, the radio hardware and the decoder hardware are amalgamated on the same board or are not intended to work individually. As a result, using existing radio hardware for an unintended purpose turns out to be problematic or even impossible, which means high costs for specified radio hardware, e.g. for research purposes [2]. In recent years, SDR technology has turn out to be a revolution by bringing much functionality as software with the reduction of the cost of hardware maintenance and up-gradation [3]. It is an extremely low-priced software defined radio based on DVB-T TV (Digital HD TV) USB receiving dongles which has RTL2832U chip in it. In March 2010, Eric Fry, Antti Palosaari and the Osmocom team first discovered this device who were developing their own SDR at that time [4]. From then, several approaches have been made by the researchers all over the world employing this device on their research works. In 2014, a software-defined sensor architecture for large-scale wideband spectrum monitoring system has been proposed (Damian Pfammatter (et al.) where distributed data collection have been done in real-time over the Internet using RTL-SDR [5]. Another approach have been made by Ken Tapping et al. who have presented SDR technology as an alternative to switched radiometers for continuum radio astronomy [6]. A User-Friendly Android-Based Tool for Spectrum-Analysis based on RTL-SDR have been approached by Jens Saalmüller et al. too [7]. In early 2015, a concept of wireless spectrum analyzer in pocket has also been developed using RTL-SDR (Tan Zhang et al.) [8]. One of the main feature of RTL-SDR is the transfer of a complete signal spectrum in a selected frequency range with a defined sample rate to the computer. This means that all received data is available in a raw format and can be used without the restrictions and information losses of traditional radio hardware, e.g. caused by a fixed filter bandwidth or signal demodulation. Therefore, one single device can work as a receiver for very different types of signals. This leads to advanced radios that previously required complicated analogue hardware now being able to be implemented easily in software [2]. So in this paper, RTL-SDR has been demonstrated as an alternative approach of a spectrum analyzing unit with advanced radio capabilities such as wideband tuning and waterfall displays. II. SYSTEM ARCHITECTURE The projected system is based on the RTL-SDR device, a multi-purpose wide band radio scanning unit consisting of economical hardware entity for signal reception and a software portion for signal processing. The hardware part which is available in the form of DVB-T USB dongle, consists of an antenna connected to a tuner chip which is connected 185

187 to the RTL2832U chip via I2C [7]. The tuner IC has been used for receiving the analog signal and filtering out the desired frequency. Then it transforms this frequency down to an intermediate frequency (IF) generating in-phase and quadrature components (I/Q signals) and sending them into the RTL2832U chip. This chip then samples the signal with a maximum sampling rate of 3.2 MS/s with 8 bit I/Q samples output. These samples are then sent to the computer via USB. The software part finally processes the raw samples data and illustrates the signal spectrum with waterfall display. frequency range of the dongle. There are two commonly used tuners such as R820T and E4000 chips. There are also the less common FC0013 and FC0012. Recently there is also the R828D and FC2580 which are even less common. TABLE.I FREQUENCY RANGE OF TUNER CHIPS Tuner Minimum Frequency (MHz) Maximum Frequency (MHz) R820T E FC FC Each tuner offers different frequency ranges, gains, amplifiers and filters. But they support a frequency range of at least 60 to 1100 MHz in common which is very wide spread range. Fig.1. Basic Oparational Method A. The RTL2832U Demodulator The RTL2832U is a baseband demodulator which is precisely designed for receiving DVB-T and radio broadcasting. But the application of this demodulator is not limited to these operations. The RTL2832U supports Zero-IF and low IF frequency and has a maximum sample rate of 3.2 MS/s. It receives the IF I/Q signals from the analog tuner IC and outputs the 8 bit I/Q samples [7]. The RTL2832U contains ADC (Analog-to-Digital Converter) and DSP (Digital Signal Processor). It performs DDC (Digital Down-Conversion) via I/Q mixers (phase is 90 degrees apart), digital low-pass filtering, me /Q resampling, and sends 8-bit I/Q data via the USB port [9]. The RTL2832U contains USB 2.0 interface which supports full and high speed modes. This interface has been used to transfer the samples via bulk transfer to the connected host and also to configure the chip through control transfer messages. Another feature of this interface is that it can act as a repeater for the I2C bus. If the repeater is enabled, control messages which is received over USB are forwarded to the I2C bus as well as messages received on the I2C bus are forwarded to the USB port. This mode permits configuration of the tuner chip through USB interface, as the tuner chip is connected to the RTL2832U via the I2C interface. B. Tuner Chip Almost any DVB-T dongle with the RTL2832U chip can be used with the RTL-SDR drivers. However, one must pay attention to the tuner chip used in the dongle. The tuner chip defines the Fig.2. General signal processing inside a tuner IC [7] Fig. 2 illustrates an overview of the signal processing inside the tuner IC. At first, the received RF signal is passed into a low-noise amplifier (LNA) where the signal is amplified either automatically or by a manually configurable gain. Next, a certain frequency range is filtered out according to the selected frequency band (VHF II, VHF III, UHF or L-band). After that the mixer transforms the signal into a low frequency IF or Zero-IF and transfers it to the intermediary frequency filter section and gain section where the frequency range is narrowed down to extract the preferred frequency and bandwidth [7]. C. Software More than hundred software are used on RTL SDR for different purpose in different platform [10]. The most commonly use and available package are given below. 1) Windows based: a) Free: SDR#, HDSDR, SDR- RADIO.COM V2, Linrad,CubicSDR, cusdr, PowerSDR,QtRadio, SeeDer. b) Paid or trail: Matlab, Studio1, Sodira. 2) Linux baseband: a) Free: GNU Radio,Linrad, GQRX, QtRadio, CubicSDR,ShinySDR (web based), WebRadio, MultimodeSdrangelove. 3) Mac baseband: a) Free: Linrad, GQRX, CubicSDR. 4) Android-Based: a) Free: RFAnalyzer. 186

188 b) Trail or Paid: SDR Touch, Wavesink Plus. From those package, SDR# is chosen for several facility. At present it is the most popular windows based free RTL-SDR compatible software. Set up procedure is relatively easy with respect to other one [11]. It has abundant amount of GUI which make it easy to use. It has some advanced features such as different plugins. Though most of plugins are in 3 rd party, those are effective. MATLAB also released the RTL SDR plugging on their R2013b version [12]. With this support package, MATLAB can interface with the RTL-SDR and digital signal processing algorithm can then be written in MATLAB. GNU Radio is another powerful tool for SDR technology [13]. But complexity may arise at the time of installation. Both two software are most powerful in research sector. In recent version of LAB View, there have also scope for interfacing with the RTL-SDR. III. IMPLEMENTATION AND RESULT ANALYSIS As there are numerous allocations, uses of these bands have been increasing enormously for shortrange and low power communications systems in recent years. Cordless phones, Bluetooth devices, near field communication (NFC) devices, and wireless computer networks all use frequencies allocated for ISM bands [15]. TABLE.II ALLOCATIONS OF DIFFERENT ISM BANDS [16] Frequency Range (MHz) Bandwidth (KHz) Availability Subject to local acceptance Worldwide Worldwide Worldwide Subject to local acceptance In this experimental approach, the ISM band 434 MHz has been tested as an example by a RF frequency generator and the signal has been received by the RTL-SDR system which is illustrated in Fig.5 below. A. GSM Signal The GSM signal has also been received by this spectrum analyzing system. In Bangladesh, GSM 900 and GSM 1800 bands have been used for mobile communication [14]. Banglalink, a local mobile operator in Bangladesh uses MHz frequency band for uplink and MHz frequency band for downlink [14]. This uplink and downlink frequencies received by RTL-SDR are illustrated in Fig.3 and Fig.4. Fig.4. FFT Spectrum and Waterfall Display of ISM Band C. Walkie Talkie A walkie-talkie known as a handheld transceiver (HT) is a hand-held, portable and two-way radio transceiver. The FFT spectrum of Walkie Talkie by this RTL-SDR device is illustrated in Fig. 7. Fig.3. FFT Spectrum of Uplink Frequency with Waterfall Display B. ISM Band The ISM bands, known as industrial, scientific and medical bands are the radio bands which is reserved internationally for the purpose of industrial, scientific and medical uses [15]. Generally communication equipment functioning in these bands have to endure any interference generated by ISM equipment, and users have no regulatory shield from ISM device activities. Fig.5. FFT Spectrum of Walkie Talkie with Waterfall Display 187

189 D. FM Broadcasting Signal FM broadcasting, a VHF broadcasting technology uses frequency modulation (FM) technique to provide high-quality sound over broadcast radio. The FM signal can also be received by this spectrum analyzing unit. A local Bangladeshi FM broadcasting signal at 89.6 MHz is illustrated in Fig. 9 below. Fig.6. Signal FFT Spectrum of FM Broadcasting Signal with Waterfall Display TABLE.III EXPERIMENTED DATA Measured Approximate Bandwidth ( KHz) SNR ( Signal to Noise Ratio) GSM Uplink FM Signal Walkie Talkie ISM Band In the Table III, We have experimented some of the signals and measured aprroximate bandwidth and SNR. The experimented data are not accurate but close to the real ones. For TV broadcasting signal, the bandwidth can not be measured beacause the bandwidth exceeds the measuring range of this device. As the signals are received from a place far from base stations, the SNRs are less. IV. FUTURE WORK In future, we make it as a wide band portable spectrum analyzer which may be applied on integrated single board pc. Hence the complexity of using laptop will be removed.tinny screen will help to see both FFT and waterfall diagram by which SNR and bandwidth will be calculated automatically. To make this device as a vector analyzer is also possible. Decrypting of different types of signal and message is another great scope by this device. V. CONCLUSION In this paper, it is shown that the RTL-SDR device can be used as an alternative for spectrum analyzing purposes by which the complexity in signal analysis can be lessened easily. Though the performance of this system is not fully apposite, the system is both user-friendly and cost effective compared to the traditional hardware system. So it can be used as a modern spectrum analyzing tool which has wide range of frequency tuning conveniences for analyzing the signal more proficiently. REFERENCES [1] Ana Nika, Zengbin Zhang, Xia Zhou, Ben Y. Zhao and Haitao Zheng, Towards Commoditized Real-time Spectrum Monitoring, HotWireless 14, September 11, 2014, Maui, Hawaii, USA. [2] Thomas Rudolph, Analyzing Security-related Signals Using Software defined Radio, Bachelor s Thesis. February 12, [3] Govarthanam K S, Abirami M, Kaushik J, Economical Antenna Reception Design for Software Defined Radio using RTL-SDR, Proceedings of the Intl. Conf. on Innovative trends in Electronics Communication and Applications 2014, page [4] Carl Laufer, The Hobbyist's Guide to the RTL-SDR: Really Cheap Software Defined Radio. Kindle Edition, Published on May 14, [5] Damian Pfammatter, Domenico Giustiniano, Vincent Lenders, A Software-defined Sensor Architecture for Large-scale Wideband Spectrum Monitoring, 14th International Conference on Information Processing in Sensor Networks (IPSN '15), April 14 16, 2015, Seattle, WA, USA. [6] Ken Tapping, Marcus Leech, RTLSDR-based, Software Defined Radio Alternative to Switched Radiometers for Continuum Radio Astronomy, (Retrieved on August 2015). [7] Jens Saalmüller, Matthias Kuba, Andreas Oeder, A User- Friendly Android-Based Tool for 868 MHz RF Trafficand Spectrum-Analysis, embedded world 2015, February 24-26, 2015, Nuremberg, Germany. [8] Tan Zhang, Ashish Patro, Ning Leng, Suman Banerjee, A Wireless Spectrum Analyzer in Your Pocket, HotMobile 15, February 12 13, 2015, Santa Fe, New Mexico, USA. [9] Dr. Phil, Realtek RTL2832U: The mystery chip at the heart of RTL-SDR, version 1, published on [10] (Retrieved on August 2015). [11] (Retrieved on August 2015). [12] Communications System Toolbox Support Package for RTL-SDR Radio, communications-system-toolbox-support-package-for-rtlsdr-radio (Retrieved on August 2015). [13] (Retrieved on August 2015). [14] ies3.html#bangladesh (Retrieved on August 2015). [15] "ARTICLE 1 - Terms and Definitions". life.itu.ch. International Telecommunication Union. 19 October "industrial, scientific and medical (ISM) applications (of radio frequency energy): Operation of equipment or appliances designed to generate and use locally radio frequency energy for industrial, scientific, medical, domestic or similar purposes, excluding applications in the field of telecommunications." [16] "Radio Regulations, Edition of 2012". ITU-R (Retrieved ). 188

190 Fabrication of Bismuth Ferrite Mulriferroic Perovskite Nanoparticles Using and Aqueous Organic Gel Route Mayeesha M. Haque*, M. S. Parvez, M. S. Islam Department of Materials Science and Engineering, University of Rajshahi, Rajshahi-6205, Bangladesh * mayeesha009@yahoo.com M. A. Hakim 2 Department of Glass and Ceramic Engineering, Bangladesh University of Engineering and Technology (BUET), Dhaka- 1000, Bangladesh M. A. Gafur Bangladesh Council of Scientific and Industrial Research (BCSIR), Dhaka- 1000, Bangladesh Abstract The potential of multiferroic materials in advanced technological applications is the inspiration behind the attempt to synthesize bismuth ferrite (BiFeO 3 or BFO) nanoparticles in the present investigation. Nano-sized single-phase BFO particles within the size range of nm have been synthesized by an aqueous organic gel route in this experiment. Exploiting FT-IR (Fourier Transform Infrared) spectroscopy, XRD (X-ray Diffraction) and SEM (Scanning Electron Microscopy), the vibrational, structural, morphological and thermal properties of the prepared BFO powders have been studied. The FT-IR study preliminarily indicates the formation of FeO 6 and BiO 6 octahedra and XRD results confirm the formation of BiFeO 3 phase in all of the fabricated samples and a single phase BFO is obtained at a temperature of 500 o C and higher. The crystallite size of the prepared powder samples has been calculated using Scherrer s formula and remained within the range of nm which is below the spin cycloid of BFO (64 nm) and increases with the increase in calcination temperature. SEM study reveals that the observed particle size of the BFO samples is within the range of nm which in well agreement with the calculated particle size obtained from XRD line broadening. Keywords Multiferroic, Perovskite, Nanoparticles, Sol-gel. X-ray Diffraction, Scanning Electron Microscopy. I.INTRODUCTION The foundation of the technological world is based on functional materials like semiconductors, magnets, ferroelectrics and more. The new era of civilization, however, demands the development of new generation of smart devices and to avail that, scientists and researchers are dedicating themselves in exploring more sophisticated classes of functional materials. Multiferroic materials exhibit both a magnetization and dielectric polarization in a single phase that can result in the magnetoelectric effect due to the coupling of the magnetic and dielectric ordering [1]. A single phase multiferroic with strong coupling between ferroelectric and ferromagnetic order parameters holds immense prospect for development of the ultimate device functionality that allows simple control over the magnetic nature of the material with an applied electric field at room temperature and/or vice versa and provides an additional degree of freedom in the design of sensors, actuator, magnetic storage media, spintronic devices and so on [2, 3]. Even though multiferroic materials are one of the most well sought materials in the modern age, there is a scarcity of materials that exhibit multiferroism in a single phase. There are a number of factors that contribute in development of multiferroism in a material including symmetry, electronic properties and chemistry. There are only 13 point groups that can give rise to multiferroic behavior. Additionally, there exists a seeming contradiction between the conventional mechanism off-centering in a ferroelectric and the formation of magnetic order. Because ferroelectrics by definition are insulators while itinerant ferromagnets need conduction electrons, ferroelectricity requires B-site ions with d 0 electrons whereas magnetism requires partially filled d j electrons [4]. In the arena of multiferroic research, one of the most successful stories is BiFeO 3 (BFO). BFO adopts the perovskite structure and not the ferrite structure, despite its nomenclature. Bulk BFO can be described as rhombohedrally distorted ferroelectric perovskite with the space group R3C [5, 6]. Fabrication techniques are fundamentally important factors that determine the properties of the material. Tailoring the desired properties in nanomaterials depend on various parameters such as route of fabrication, composition of the precursor material, mixing approach, temperature variation, grinding methods and so on. Recently, wet chemical methods have received great attention for the synthesis of BFO materials [7]. In this experiment, sol-gel method has been employed. The details of the formation technique to obtain single phase BFO and the characterization methods of the samples are delineated in this paper. II.EXPERIMENTAL PROCEDURE A. Preparation Sol-gel process or Pechini method was employed in preparing BFO powder samples. High purity bismuth nitrate pentahydrate (Bi(NO 3 ) 3.5H 2 O, 99.5%) as a source of Bi 3+ ions, ferric nitrate 189

191 nonahydrate (Fe(NO 3 ) 3.9H 2 O, 98%) as a source of Fe 3+ ions, tartaric acid (C 4 H 6 O 6, 99%) as chelating agent and nitric acid (HNO 3, 65%) were used as raw materials to prepare a stock solution. First, 500ml 0.025M stock solution was prepared by mixing 6.063g Bi(NO 3 ) 3.5H 2 O and g Fe(NO 3 ) 3.9H 2 O with stirring to form a hydrolyzed colloid of Bi- and Fe- species. The colloidal solution was treated with 15 ml of 65% HNO 3 added gradually to form a clear water soluble complex solution. It was then transferred into a 500ml volumetric flask and deionized water was added to the mark to form the transparent stock solution. After that, convenient amount of stock solution was stirred at C and 0.5 M tartaric acid was added until the solution turned a pale yellow color which was heated under vigorous stirring at C in a silicone oil bath for 5 hours or until the liquid components were evaporated. During stirring and heating, the color of the solution darkens very slowly until it attains a reddish-brown color with viscosity akin to glue. The viscous solution was heat treated in a natural convection oven at C for 3-5 hours to obtain fluffy yellow colored gel. The gel was further dried in the convection oven at C to remove remaining moisture and other volatile impurities. The dried gel was roughly ground and the resultant complex solid was used as the precursor powder. The precursor powder was divided into several sections and calcined at C, C, C, C, C and C in a furnace for 2 hours. The calcined product was finely ground to obtain bright orange-red colored powder. The formation mechanism can be expressed as chemical reactions as follows. Bi(NO 3 ) 3.5H 2 O + H 2 O Bi(OH) 3 + 3HNO 3 + 3H 2 O (a) Fe(NO 3 ) 3.9H 2 O + H 2 O Fe(OH) 3 + 3HNO 3 +7H 2 O (b) Bi(OH) 3 + Fe(OH) 3 BiFeO 3 + 3H 2 O (c) (at supercritical condition) B. Characterization techniques The vibrational properties of the prepared samples were examined using a IR spectrometer (Perkin Elmer Spectrum-100 FTIR Spectrometer). Structural properties and phase identification of the prepared powder samples were studied using X-ray diffraction patterns taken with an X-ray diffractometer (Bruker D8 Advance X-ray Diffractometer) using CuKα (λ= A o ) radiation. The scanning drive axis was taken as 2θ and count data was taken between degrees. The values of Miller Indices (hkl) were identified by adopting JCPDS file (reference code: ). SEM analysis (JSM 7600f series FESEM instrument from JEOL) was performed to observe the particle size and morphological properties of the prepared samples. The thermal properties of the prepared samples were studied using TG/DTA analysis (EXStar6000 TG/DTA Thermo- Gravimetric/Differential Thermal Analyzer from Seiko Instruments Inc.) with an aim to delineate as accurately as possible the various temperatures associated with the thermal behavior of the samples, i.e. temperature of decomposition, stability range of an intermediate compound and the temperature at which the reaction got completed. Fig. 1. Flow chart for formation of BFO powder from precursor solution. III. RESULTS AND DISCUSSIONS The obtained BFO powders calcined at different temperatures such as C, C, C, C, C and C were introduced to room temperature FTIR spectroscopy to get an insight about the vibrational spectra. The vibrational spectra of the samples are seen to be dominated by broad and multi-component bands that are related to the internal vibrations of the chemical bonds present. In the spectrum, there are some characteristic peaks easily identifiable. Fig. 2 represents IR spectra of BFO powder calcined at different temperatures for 2 hours. The spectra exhibits characteristic absorption bands around cm -1, cm -1, cm -1, 1384 cm -1, 1096 cm -1, 847 cm -1 and cm -1. The band around cm -1 appears due to the overlapping of O-H stretching of H 2 O [8]. The small absorption bands visible around cm - 1 region represents O=C=O absorption modes. The CO 2 absorption modes might have appeared due to the CO 2 present in the air ambiance of the IR environment. The slightly sharper band observed between cm -1 represents the stretching 190

192 vibrations of C=O and N-O [9]. The band around 1384 cm -1 appeared due to the presence of trapped nitrates [10]. The absorption bands visible around 1096 cm -1 and 847 cm -1 attribute to the presence of small carbonate phases [11]. Strong absorptive bonds are noticed between cm -1. The band around 444 cm -1 represents Fe-O stretching and bending vibrations, while the one around 531cm -1 is the attribute of BiO 6 octahedral structure unit [12, 13, 14]. The presence of both of these characteristic bands is a significant indication of formation of BFO phase. Fig. 3. XRD pattern for the obtained samples of BFO powder samples calcined at different temperatures. Fig. 2. FTIR spectral lines for the obtained samples of BFO powder calcined at different temperatures. From the IR study of the BFO samples, it is clear that the integrated peak area around 3400 cm -1 corresponds to the vibrational properties of H 2 O, which broadens and decreases in intensity with the increase in temperature. This implies the removal of H 2 O and residual moisture as calcination temperature was raised. Also, around 1384 cm -1, the sharp peak corresponding to nitrate ions is present in all of the spectra. From the wide bands between the frequency range cm -1 corresponding to Fe-O and Bi-O octahedra, it can be assumed that the metal oxides have formed in all of these samples. It is also observed that these bands have considerably sharpened at higher calcination temperatures. Therefore, it can be assumed that the purity of the samples increased with the increase in temperature. However, the IR spectra only provides enough information about the band structures. Better understanding of phase purity and crystallinity is observed by studying the XRD patterns of the samples. XRD patterns for prepared samples calcined at different temperatures are shown in fig. 3. The diffraction pattern for the sample calcined at C shows some clear reflections from 212), (104), (110), (202), (511), (402), (122), (116), (223), (621), (302) and (018) planes. Among these reflections, (104), (110), (202) and (122) are characteristic BFO peaks. Reflections corresponding to the presence of Bi 2 O 3 are visible at (212), (511) and (223) and reflections from planes (110) and (116) correspond to hematite (Fe 2 O 3 ). Generally, Bi 2 O 3 and mullite Bi 2 Fe 4O 9 secondary phases are frequently reported during synthesis of BFO and different reasons are proposed for the appearance of these phases. The low temperature stability of impurity phases, metastable and off-stoichiometric nature of BFO are thought to be the prime causes for the formation of secondary phases [15, 16]. Significant improvement in crystallinity of BFO phase with the increase in temperature can be observed in the case of the sample calcined at C. Reflections from the planes (012), (104), (110), (202), (024), (122), (214), (208), (208) and (128) show much higher intensity than the previous XRD pattern. Also, the reflections from (104) and (110) planes now appear to be a clearly doublet peak, which is an important trait of the intensity ratios of BFO orthorhombic phase. This particular doublet peak is a sign of crystalline BFO phase. The intensities of reflections representing Bi 2 O 3 observed in the previous pattern have considerably reduced. On the other hand, no significant characteristic peak for 191

193 Fe 2 O 3 can be observed in this pattern. From the observations, one can assume that, at 450 o C calcination temperature, the crystallinity of BFO has significantly increased. The IR analysis performed for the same sample is in agreement with this. It can be noted that at higher calcination temperatures, the samples exhibited gradual enhancement in crystallinity with the impurity peaks decreasing in intensity. The samples calcined at C and C don t exhibit the impurity peaks, indicating that pure phase BFO have been obtained around C and higher. The average crystallite size of the calcined powder was determined from X-ray line broadening using Scherrer s formula as expressed below, T= 0.9 λ/ Bcosθ (1) Here, T is the particle size, λ is the wavelength of the radiation (in this case λ= A o for CuKα radiation), θ is Bragg s angle and B is FWHM (full width at half maximum). The calculated particles sizes are distributed in Table 1. Since the XRD study suggested an Rhombohedral crystal structure, by knowing the interplaner spacing d hkl and Miller indices h,k and l, it is possible to calculate the lattice constants and cell volumes of the samples. Table. 2 represents the lattice constants and cell volumes of BFO perovskite (rhombohedral R3C crystal system) particles obtained in this experiment at different calcination temperatures. According to JCPDS data, the values of a and c are and respectively, that yields a cell volume of (A o ) 3. The experimental samples have approximately equal cell parameter values compared to the standard data. SEM images of the BFO samples are presented in fig.5. From the SEM images, it can be concluded that nanoparticles ranging between about nm in size were formed in the experiment, which is in well agreement with XRD studies. There are some abnormally large agglomerates, which can be attributed to the Van der Waals forces. To reduce the surface energy, the primary particles have a tendency to form nearly spherical agglomerates, in a minimum surface to volume ratio [17]. This type of grain structure is common in oxide, ferrite and titanate ceramics which is a result of an abnormal/discontinuous grain growth, also called an exaggerated grain growth [18, 19]. In abnormal growth, some grains grow faster than other grains with increasing sintering temperature. Abnormal grain growth may be the result of: (1) the existence of second phase precipitates or impurities, (2) materials with high anisotropy in interfacial energy and (3) materials with high chemical equilibrium. Table 1: Average particle size of obtained BFO samples as a function of calcination temperature. Temperature o C Average particle size nm Fig. 4. Change in particle size of obtained BFO samples with the change in calcination temperature. Fig. 5. SEM micrograph of BFO samples calcined at (a) C, (b) C, (c) C. 192

194 Table 2: Lattice parameters a, b and c of BFO samples obtained at different calcination temperatures Variation of Temp. Miller Indices Lattice parameters Angles between the sides (degrees) ( o C) (hkl) a b c α β γ Cell volume V (A o ) (012), (221), (104), (110), (202) (012), (221), (104), (110), (202) (012), (221), (104), (110), (202) Fig. 6 TG/DTA analysis curve for BFO precursor powder. Fig. 6 illustrates the TG/DTA curves of the stoichiometric precursor powder or dried gel of BFO. The TG curve indicates that the weight of the dried gel decreased in a gradual step-like manner upto a temperature of C. There are three distinct steps observed in the weight-loss pattern. The first decrease in weight was detected at about C where about 10% of the material was lost which corresponds to the evaporation of physisorbed water [20]. The second step, where 73.4% of the sample remained, was observed close to C, which was due to the reduction of hydroxyl groups. At C a sudden plunge in the curve is observed along with a sharp exothermic DTA peak around the same temperature. This is due to the metal tartarate complexes that formed during the reaction along with other organic species [21]. Maximum amount of weight loss happens at this stage (about 44%) and it can be assumed that the large amount of energy and gases like CO 2, NO 2 were released at this stage along with residual carbon. The releasing of these gas components were due to the production of CO 2 at some lower temperature that adsorbed on the surface of the material. Other possibilities for CO 2 release can be due to the oxidation of pyrolytically formed carbon by traces of oxygen in the atomosphere. It could also be an event related to the decomposition of Bi- complexes. The third step involves a plateau region which is observed from C to C. This is related to the decomposition of the complex polymeric phases and formation for metal oxide. This plateau region is slightly inclined towards weight loss until it reaches C and the quality of the sample remained significantly unchanged. It can be said that the decomposition of the organics was done around C and the formation of metal oxide phase begun. There is a clear exothermic elevation at C which is related to the formation of pure phase BFO. It can be assumed that in this experiment, it is possible for perovskite BFO to start forming around C [22]. From the almost linear TG curve starting at ~300 0 C and corresponding DTA curve it can be said that the calcination temperature for BFO phase formation safely lies between the temperature range of C ~ C in this experiment. There was no clear schemes in the DTA curve that corresponds to BFO crystal formation. IV. CONCLUSIONS In this experiment BFO ceramic nanoparticles were synthesized successfully at variable calcination temperatures via an aqueous organic gel route. The soft chemical route provided an easy and reproducible approach to synthesizing nanoparticles of desired morphologies. The feasibility of this experiment in forming BFO nanoparticles has been verified by the characterization techniques IR, XRD, SEM and TG-DTA. Nanocrystalline BFO is a promising candidate in the fields of multiferroics and smart materials. This experiment can be considered to be the cornerstone to an expanded multi-dimensional research work that could be developed in the future. REFERENCES [1] L. D. Landau, E. M. Lifshitz, Electrodynamics of continuous media. Fizmatgiz, Moscow. (1959) [2] N. Hur, S. Park, P.A. Sharma, J.S. Ahn, S. Guha, S.W. Cheong. Electric polarization reversal and memory in a multiferroic material induced by magnetic fields, Nature 429, ( ), (2004) [3] G. Catalan, JF. Scott, Physics and applications of bismuth ferrite. Advanced Materials 21, (24), (2009) [4] Hill, N. A., Why Are There so Few Magnetic Ferroelectrics? Journal of Physics and Chemistry B, 104, (29), (2000). [5] J. M. Moreau, C. Michel, R. Gerson, W. J. James. Ferroelectric BiFeO 3 X-ray and neutron diffraction 193

195 study. Journal of Physics and Chemistry of Solids, 32, (1971) [6] Dinesh Varshney, Poorva Sharma, S. Satapathy, P.K. Gupta. Structural, electrical and magnetic properties of Bi0.825Pb0.175FeO3, and Bi0.725La0.1Pb0.175FeO3 multiferroics, Materials Research Bulletin 49, , (2014) [7] J. Silva, A. Reyes, H. Esparza, H. Camacho, L. Fuentes. BiFeO3: a review on synthesis, doping and crystal structure, Integrated Ferroelectrics 126, 47 59, (2011) [8] G. K. L. Goh, F. F. Lange, S. M. Haile, C. G. Levi, Hydrothermal synthesis of KNbO 3 and NaNbO 3 powders. Journal of Materials Research. 18, 2, 338. (2003) [9] A. Z. Simoes, B. D. Stojanovic, M. A. Ramirez, A. A. Cavalheiro, E. Longo, J. A. Varela. Lanthanum-doped Bi 4 Ti 3 O 12 prepared by the soft chemical method: Rietveld analysis and piezoelectric properties. Ceramics International, 34, 2, 257. (2008) [10] A.V. Zalesskii, A. A. Forlov, T. A. Khimich, A. A. Bush. Composition-induced transition of spinmodulated structure into the uniform antiferrolagnetic state in a Ba 1-x LA x FeO 3 system studied using 57Fe NMR. Physics of the Solid State, 45, 134. (2003) [11] B. Sreedhar, C.S. Vani, D.K. Devi, M.V.B. Rao, C. Rambabu. Shape controlled synthesis of barium carbonate microclusters and nanoparticles using natural polysachharide gum acacia. American Journal of Materials Science, 2, 5. (2012) [12] J. Yang, X. Li, J. Zhou, Y. Tang, Y. Zhang, Y. Li, Factors controlling pure phase magnetic BiFeO 3 powders synthesized by solution combustion synthesis. Journal of Alloys and Compounds. 509, (2011) [13] G. V. Subba Rao, C. N. R. Rao, J. R. Ferraro. Infrared and electronic spectra of rare earth perovskites: ortho chromites, magntites and ferrites. Applied Spcectroscopy, 24, 436. (1970) [14] B. Bhusan, A. Basumallick, S. K. Bandopadhyay, N. Y. Vasanthacharya, J. Das. Effect of alkaline earth metal doping on thermal, optical, magnetic and dielectric properties of BiFeO 3 nanoparticles. Journal of Physics D: Applied Physics, 42, (2009) [15] Archna Sagdeo, Puspen Mondal, Anuj Upadhyay, A.K. Sinha, A.K. Srivastava, S. M. Gupta, P. Chowdhury, Tapas Ganguli, S.K. Deb, Correlation of microstructural and physical properties in bulk BiFeO3 prepared by rapid liquid-phase sintering, Solid State Sciences 18, 1 9, (2013) [16] Matjaz Valant, Anna-Karin Axelsson, Neil Alford. Peculiarities of a solid-state synthesis of multiferroic polycrystalline BiFeO3, Chemistry of Materials, 19, , (2007) [17] Y.B. Khollam, A.S. Deshpande, A.J. Patil, H.S. Potdar, S.B. Deshpande, S. Date, Microwavehydrothermal synthesis of equi-axed and submicronsized BaTiO3 powders. Materials Chemistry and Physics, 71, (2001) [18] A.Z. Simões, E.C. Aguiar, A.H.M. Gonzalez, J. Andrés, E. Longo, J. A. Varela, Strain behavior of lanthanum modified BiFeO3 thin films prepared via soft chemical method. Journal of Applied Physics. 104, (2008) [19] A.Z. Simões, L.S. Cavalcante, C.S. Riccardi, J.A. Varela, E. Longo, Improvement of fatigue resistance on La modified BiFeO3 thin films. Current Applied Physics, 9, 520. (2009) [20] G. Biastto, A. Z. Simoes, C. R. Foschini, S. G. Antonio, M. A. Zaghete, J. A. Varela, et al., A novel synthesis of perovskite bismuth ferrite nanoparticles. Processing and Application of Ceramics. 5, 171. (2011) [21] S. Godara, N. Sinha, G. Ray, B. Kumar, Combined structural, electrical, magnetic and optical characterization of bismuth ferrite nanoparticles synthesized by auto-combustion route. Journal of Asian Ceramic Societies. 2, 4, 416. (2014) [22] C. Fu, M. Huo, W. Cai, X. Deng, Preparation of Bismuth Ferrite nanopowders at different calcinations temperatures. Journal of Ceramic Processing Research. 13, 5, 561. (2012) 194

196 Analysis of GLDAS data for estimating and distribution of evapotranspiration and rainfall over Bangladesh Md Ataur Rahman 1, Md Mainul Islam Mamun 1, Md Monirul Islam 1 * 1 Department of Applied Physics & Electronic Engineering, University of Rajshahi, Bangladesh * Corresponding Author rajib_apee@yahoo.com Abstract Bangladesh is one of the most climate vulnerable country. Nowadays climate is setting more stress on agricultural productivity. The climate change may lead to change in evapotranspiration (ET). The present work investigates the spatio-temporal variations in ET and rainfall under climate change in Bangladesh with the help of Global Land Data Assimilation System (GLDAS). ET and rainfall are analyzed at annual and seasonal scale to find out the pattern of ET and rainfall in the study area during the period The results show that under current climate situation eastern region reported the highest ET and rainfall in comparison with other regions where western region reported the lowest ET and rainfall over Bangladesh. In the period 2004 to 2013, ET and rainfall are increasing over Bangladesh but in north-western region surprisingly both of them are decreasing except Rangpur. Spatio-temporal variation of ET and rainfall show that higher ET and rainfall are observed in eastern part while lower in western part of Bangladesh. As a result of decreasing ET and rainfall in northwestern region of Bangladesh specially Rajshahi and Ishwardi will require more irrigation and vegetation. Keywords Evapotranspiration, Rainfall, GLDAS, Bangladesh. I. INTRODUCTION Weather is the condition of the atmosphere at a particular place over a short period of time. Weather is the state of the atmosphere, to the degree that it is hot or cold, wet or dry, calm or stormy, clear or cloudy. Seen from an anthropological perspective, weather is something all human in the world constantly experience through their senses, at least while being outside. On Earth's surface, temperatures usually range ±40 C ( 40 F to 100 F) annually. Over thousands of years, changes in Earth's orbit can affect the amount and distribution of solar energy received by the Earth, thus influencing long-term climate and global climate change [1]. Evapotranspiration (ET) and rainfall may change due to the change of climate as well as weather. Climatic variables like temperature, relative humidity, sunshine hour, wind speed and rainfall are important parameters for the agricultural and regional water resources planning [2]. Evapotranspiration (ET) is the sum of evaporation and plant transpiration from the Earth's land and ocean surface to the atmosphere. Evaporation accounts for the movement of water to the air from sources such as the soil, canopy interception, and waterbodies. Transpiration accounts for the movement of water within a plant and the subsequent loss of water as vapor through stomata in its leaves. Evapotranspiration is an important part of the water cycle. The rate of evapotranspiration at any location on the Earth's surface is controlled by several factors: Energy availability, Humidity, Wind speed, Water availability, Physical attributes of the vegetation, Plants regulate transpiration and Soil characteristics [3]. The evapotranspiration pattern of western part of Bangladesh is showed lower values while the eastern part is showed higher values. Seasonal analysis investigate that in summer rate of ET is high while in other seasons rate of ET is considerablely lower. In meteorology, precipitation is any product of the condensation of atmospheric water vapor that falls under gravity. The main forms of precipitation include drizzle, rain, sleet, snow, graupel and hail. Precipitation occurs when a portion of the atmosphere becomes saturated with water vapor, so that the water condenses and "precipitates". Precipitation is one of the most important factors of Bangladesh where the economy strongly based on agricultural. About 80% people of Bangladesh live in rural area and directly or indirectly depend on agriculture [4]. The erratic rainfall and their associated extreme events may affect ecosystems, productivity of land, agriculture, food security, water availability and quality, health and livelihood of the common people of Bangladesh. Therefore, a better understanding of precipitation variations has important implications for the economy and society of Bangladesh [5]. The aim of this work is to study the time-series analysis, the spatio-temporal distribution and the seasonal variations of evapotranspiration and rainfall over Bangladesh by using GLDAS satellite data. II. METHODOLOGY A. Climate condition of Bangladesh Weather describes the state of atmospheric conditions at a particular place and time. Weather may change from day to day, but climate changes only over years. Many animals and plants need a certain kind of climate to survive. The variation in weather conditions in Bangladesh can be easily observed. For instance, today it may be raining in 195

197 Sylhet but the weather in Dhaka might be sunny. In this way, the variation in weather patterns with respect to location and time can be observed. However, climate is the pattern of weather that we expect to see in a particular place. The climate of Bangladesh is mainly sub-tropical monsoon, i.e. warm and humid. The Bengali calendar is divided into six seasons: Grishmo (summer), Borsha (rainy), Shorot (autumn), Hemanta (late autumn), Sheet (winter), Boshonto (spring). For practical purposes, however, three seasons are distinguishable: summer, rainy and winter. Sarcastically saying, thanks to global warming that we are experiencing prolonged seasons with extremities in forms of unusual rise in temperatures during summer, abnormal rainfall patterns during rainy season and freezing cold temperature during winter. B. Geography of Bangladesh Bangladesh is a low-lying, riverine country located in South Asia with a largely marshy jungle coastline of 710 km on the northern littoral of the Bay of Bengal. Bangladesh has a tropical monsoon-type climate, with a hot and rainy summer and a dry winter [6]. January is the coolest month with temperatures averaging near 26 deg C (78 deg F) and April the warmest with temperatures from 33 to 36 deg C (91 to 96 deg F). The climate is one of the wettest in the world. Most places receive more than 1,525 mm of rain a year, and areas near the hills receive 5,080 mm). Most rains occur during the monsoon (June-September) and little in winter (November-February) [7]. The total area of Bangladesh is 147,570 square kilometers and only the land area is 133,910 square kilometers where arable land 67%, forest and woodland 16%, permanent crops 2%, meadows and pastures 4%, others 11%. C. GLDAS Satellite The Global Land Data Assimilation System (GLDAS) is generating a series of land surface state (e.g., soil moisture and surface temperature) and flux (e.g., evaporation and sensible heat flux) products simulated by four land surface models (CLM, Mosaic, Noah and VIC). These products are now accessible at the Hydrology Data and Information Services Center (HDISC), a component of the NASA Goddard Earth Sciences Data and Information Services Center (GES DISC). Current data holdings include a set of 1.0-degree resolution data products from the four models, covering 1979 to the present; and a 0.25-degree data product from the Noah model, covering 2000 to the present. The products are in Gridded Binary (GRIB) format and can be accessed through a number of interfaces. Users can search the products through keywords and perform on-the-fly spatial and parameter subsetting and format conversion of selected data. More advanced visualization, access and analysis capabilities will be available in the future [8]. GLDAS will include four components: Calibration and Validation, Land Surface Observation, Land Surface Data Assimilation and Land Modeling. III. RESULTS AND DISCUSSIONS A. Annual mean distribution of ET and rainfall Earlier works showed that there was a significant increasing temporal trend in surface temperature and a decreasing trend in evapotranspiration (ET) in India which was mainly caused by a significant increase in the relative humidity and a consistent significant decrease in the wind speed throughout the country [9]. Over Bangladesh an increasing trend in evapotranspiration is observed during the last decade ( ) as shown in Fig.1 where rainfall shows decreasing trend as shown in Fig.2. Fig. 1. Trend in evapotranspiration over Bangladesh for the period of Fig. 2. Trend in rainfall rate over Bangladesh for the period of The eastern part of Bangladesh has showed a very high evapotranspiration in last 10 years and the western part of Bangladesh has showed very low evapotranspiration, as shown in Fig.3 while Fig.4 shows that the eastern part of Bangladesh has showed very high rainfall rate in last 10 years and the western part of Bangladesh has showed very low rainfall rate. The north-eastern parts including Chittagong, Rangamati, Sylhet, Kishorganj, Bhairab Bazar, Brahmanbaria, Mymaensingh, Jamalpur and small part of Narsinghdi of Bangladesh are shown very high ET values (Fig.3). 196

198 Fig. 3. Annual-mean spatial distribution of ET over Bangladesh for the period of Fig. 4. Annual-mean spatial distribution of rainfall rate over Bangladesh for the period of The lower values of ET are shown in the western parts including Rajshahi, Nawabganj, Ishwardi, Pabna, Kustia and Naogaon of Bangladesh. Both the GLDAS satellite models are shown the same very high values of rainfall rate in northern region (Sylhet division) of Bangladesh, which is shown in Fig.4. B. Yearly distribution of ET The spatial annual distribution images of ET of GLDAS_NOAH model over Bangladesh for each year was analyzed. Yearly distribution of ET shows the result that the high ET area over Bangladesh has came to light rapidly and the high ET area was in eastern part of Bangladesh. The western part is always shown a very much cleaner area with lower ET. High ET values were observed in the northeastern parts including Chittagong, Rangamati, Sylhet, Kishorganj, Bhairab Bazar, Brahmanbaria, Mymaensingh, Jamalpur and small part of Narsinghdi of Bangladesh. C. Yearly distribution of rainfall The spatial annual distribution images of rainfall rate of GLDAS_NOAH model over Bangladesh for each year was analyzed. Thus the rainfall rate of each year shows the significant result that the high rainfall area over Bangladesh has came to lower and seen that the high rainfall area was in eastern part (especially Sylhet division) of Bangladesh. The western part is always shown a very much lower rainfall rate. High rainfall rates were observed in the north-eastern parts including Sylhet division of Bangladesh. It is also seen that the lower rainfall area was in north-western parts including Chapainawabganj, Rajshahi, Ishwardi, Natore, Pabna, Kustia, Naogaon of Bangladesh. The western part of Bangladesh is warmer and more humidity than eastern part. The annual average relative humidity for the period was found to increase over Rajshahi by % [10]. Again Rajshahi division is the hottest division in Bangladesh. Rainfall is also insufficient in this area especially in Natore and Rajshahi. Also droughts are occurred regularly in north-western part of Bangladesh [11]. D. Seasonal variations of ET and rainfall The seasonal variations of ET and rainfall rate over Bangladesh are studied by GLDAS satellite data. The spatial distribution of seasonal mean ET and rainfall rate over Bangladesh in only the year 2013 were analyzed for four representative seasons namely winter (December-February), pre-monsoon (March- May), monsoon (June-September), post-monsoon (October-November) and which reveals high spatial variability. ET and rainfall rate over Bangladesh for summer season are shown in Fig.5 and Fig.6 respectively. The seasonality of ET exhibits the patterns of higher values during summer seasons whereas lower values during winter, monsoon & post-monsoon seasons. On the other hand, it is also seen that the seasonality of rainfall rate exhibits the patterns of higher values during summer season whereas lower values during winter, monsoon and post-monsoon seasons. Note that some area in north-western Rangpur and some parts of Rajshahi division shows extremely low ET distributions in winter. During the winter evaporative energy is low and lower temperature may be the reason for this. About 20% of annual rainfall is distributed over Bangladesh unevenly throughout the winter season because of lower temperature and high solar radiation. The north-eastern part shows higher ET distributions and western part shows lower ET distributions during summer season (Fig.4). It is found that the most of the drought primarily occurred in summer (March-May) season. It is also seen from Fig.6 that the north-eastern part shows higher rainfall rate and western part including Rangpur and Rajshahi divisions shows lower rainfall rate during premonsoon season. About 80% of annual rainfall occurs over Bangladesh during the hot summer season because of high temperatures and low solar radiation. 197

199 higher values while western region of Bangladesh shows lower values. Seasonal analyses showed that the highest ET and rainfall rate were found in summer season. Minimum ET and rainfall rate were found in winter because during the winter evaporative energy is low and lower temperature may be the reason for this. As a result of decreasing trend of ET and rainfall, western region of Bangladesh will require more irrigation and vegetation. ACKNOWLEDGMENT The present study was supported by the GLDAS teams at NASA for the provision of satellite data. REFERENCES Fig. 5. ET distribution pattern over Bangladesh during summer, Fig. 6. Rainfall distribution pattern over Bangladesh during summer, In monsoon season, The ET concentration is decreasing from high to low from north-eastern part to southwestern part of the country. In post-monsoon, spatial distribution shows comparatively higher values of ET and rainfall than the other seasons over Bangladesh. It is therefore recommending that further more researches are needed to achieve a better understanding of spatio-temporal variations in ET and rainfall and their various climatic impacts. IV. CONCLUTION The ET and rainfall patterns of the recent decade ( ) the eastern region of Bangladesh shows [1] S. Afsar, N. Abbas, and B. Jan, Comparative study of temperature and rainfall fluctuation in Hunza-Nagar district, Journal Basic & Applied Siences, Vol. 9, pp , [2] R. Ayub, and M. M. Miah, Effects of change in temperature on reference crop evapotranspiration in the northwest region of Bangladesh, 4 th Annual Paper Meet and 1 st Civil Engineering congress, Deember 22-24, 2011, Dhaka, Bnagladesh, ISBN: [3] H.V. Knapp, Evaporation and Transpiration. In Handbook of Applied Meteorology, ed. D.D. Houghton, pp , New York: John Wiley and Sons, Inc., [4] S. Shahid, Impact of climate change on irrigation water demand of dry season Boro rice in northwest Bangladesh, Climatic Change, DOI /s , [5] S. Shahid, Trends in extreme rainfall events of Bangladesh, Theoretical and Applied Climatology, Vol. 104(3-4), pp , [6] A. Ali, Climate change impacts and adaptation assessment in Bangladesh, Climate Research, Vol. 12, pp , [7] S. Karmakar and M. L. Shrestha, Recent climate change in Bangladesh, Report No. 4, SAARC Meteorological Research Centre (SMRC), Dhaka, Bangladesh, [8] H. K. Beaudoing, W. L. Teng, M. Rodell, H. Fang, and B. E. Vollmer, Global Land Data Assimilation System (GLDAS) Products, Services and Application from NASA Hydrology Data and Information Services Center (HDISC), ASPRS Annual Conference, Baltimore, MD United States, 8-31 Mar [9] A. Bandyopadhyay, A. Bhadra, N. Raghuwanshi, and R. Singh, Temporal Trends in Estimates of Reference Evapotranspiration over India, Journal of Hydrologic Engineering Vol. 14(5), pp , [10] M. G. Ferdous and M. A. Baten, Climatic Variables of 50 Years and their Trends over Rajshahi and Rangpur Division, Journal of Environmental Science & Natural Resources, Vol. 4(2), pp , [11] H. Murad and A. K. M. S. Islam, Drought assessment using remote sensing and gis in north-west region of bangladesh, 3rd International Conference on Water & Flood Management (ICWFM),

200

201 informative genes/features to train the logistic classifier. II. FORMULATION OF LOGISTIC CLASSIFIER FOR BINARY CLASSIFICATION Suppose we have a training gene expression dataset with n 1 control and n 2 test samples, respectively for p genes, where x ij,(g) consist of expressions of ith gene in gth group (i =1, 2,, p; j =1, 2,, n g ; g=1,2). Here the problem is to classify a new patient having the vector x = (x 1,x 2,..., x p ) T of expressions with p genes into one of two groups (control/test) corresponding to their populations Π 1 and Π 2, respectively. Then the logit model is to fit ( ) ( ( ) ( ) ) (1) Inverting the relationship in (1) we get ( ) where, ( ) ( ) ( ) and ( ) ( ) Now define a response variable Y that identifies the population to which X belongs, { The conditional likelihood function is given by ( ) ( ( )) (1- ( )) where, ( ) ( ) Then the conditional log-likelihood is ( ) * ( ) ( ) ( ( ))+ { ( ) ( )} (2) The Maximum likelihood (ML) estimates, ( ) of ( ) is obtained by differentiating (2) and setting the derivatives equal to zero ( ) ( ) ( ) ( ) Then iteratively we obtain * ( )+=0 ( )* ( )+ ( ) ( ) ( ( )) ( ) (3) Where, the derivatives are evaluated at ( ) We observe that the equation (3) needs to compute the inverse. In microarray technology minimum 10 thousands of genes/variables (p) and 3 patients/samples (n) expression is generated simultaneously as discussed earlier. So, to solve the problem of matrix inversion the number of genes/variables p should be smaller than number of patients/samples. In spite of being popular of logistic classifier, it is very difficult to classify patients/samples for binary classification using all genes as feature variables. In order to solve this problem, we select few important (DE) features/genes to train the logistic classifier. Many literatures have been discussed about the selection of feature variable for binary classification [7-9]. However, they did not consider the problems of outliers in the dataset. So existing popular classifiers produces misleading results when the training data set is contaminated by outliers. Therefore, in this paper we consider outlier modification and significant gene selection for binary classification with lower misclassification error rate and higher accuracy using logistic classifier will be discussed in the next section. III. OUTLIER MODIFICATION AND IMPORTANT (DE) GENE SELECTION USING LOGISTIC CLASSIFIER Microarray gene expression datasets are often contaminated by outliers due to several steps involve in the data generating process from hybridization to image analysis. There are mainly two types of statistical approaches for analyzing data that are contaminated with outliers [10]; 1) robust approaches, those are used to detect outliers and to provide resistant (stable) results in the presence of outliers and 2) outlier diagnostics methods, which first detect outliers and remove them from the dataset or replace (modify) the outliers with the appropriate values to get the reduced/modified dataset. In this paper, we discussed the later approach by replacing outlying gene expression with the appropriate values and the application of classical t-test to select top differentially expressed (DE) genes for the logistic classifier as follows: a) If an observation does not fall in to the interval [MED i,(g) -k MAD i,(g),med i,(g) - k MAD i,(g) ] then we classify it to an outlier. Where, MED i,(g) =median(x ij,(g) ; i =1, 2,, p; j =1, 2,, n g ; g=1,2) is the median expressions of ith gene in gth group and MAD i,(g) = median( x ij,(g) - MED i,(g) ) is the median absolute deviation of ith gene in gth group. We used k=3 as the default value of threshold. 200

202 b) Check the existence outliers for each gene from both groups (test/control) separately using a). If outliers exist, replace outliers by their respective group medians. c) Apply t-test in the modified training dataset to identify differentially expressed (DE) genes. Then arrange the genes from top DE genes by ranking the p-values of t- test. d) Select top k < max(n 1, n 2 ) genes out of p genes from both patterns of DE genes (upregulated/downregulated) and estimate the logistic classifier using the expressions of these top k genes methods (LDA, SVM, NB, Logistic and Proposed). It is obvious from this figure that all the methods produce the lower MER in absence of outliers. Figure 3 (a2, b2 and c2 ) represent the average training MER against the number of top DE genes in presence of outliers with μ =0.5, 1.0 and 2.0, respectively. It is also clear from this figure that in presence of outliers our proposed methods performs better than the others methods for different values of μ as it produces MER almost zero(0) for every cases. IV. SIMULATION AND REAL GENE EXPRESSION DATA RESULTS Simulated Dataset To investigate the performance of the proposed method in a comparison of the other methods in both absence and presence of outliers, we generate several training and test datasets from the following data generating model Gene Expression Patterns (No. of Genes) Control (n 1 ) Test (n 2 ) Pattern 1(p 1 ) ( ) ( ) Pattern 2(p 2 ) ( ) ( ) Pattern 3(p 3 ) ( ) ( ) Fig. 2. Simulated gene expression data generating model with three patterns of genes p 1, p 2 and p 3 respectively. We generated three dataset from Fig.2. with μ = 0.5, 1.0 and 2.0 respectively and common variance σ 2 = 1. Each dataset contains p =100 genes. Of which pattern 1 contains, p 1 =10 DE genes, pattern 2 contains, p 2 =10 DE genes and pattern 3 contains, p 3 = 80 EE (Equally expressed) genes. Each gene is generated with N =440 sample expressions of which N 1 =220 expressions are generated from control/normal patients and N 2 =220 expressions are generated from test/cancer patients. We construct training and test datasets from each dataset by randomly choosing n 1 =N 1 /2=110 samples from N 1 =220 control/normal samples and n 2 =N 2 /2=110 samples from N 2 =220 test/cancer patients. The rest of the patients belong to the test dataset. After constructing the training and test dataset we contaminated 10%-20% patients with 20% genes of training data by outliers. Then we computed both training and test misclassification error rate (MER) for every methods with respect to the increasing number of top DE genes as feature variables. We repeated this procedure 100 times and calculate the average of training and test MER. Figure 3 (a1, b1 and c1) represent the average training MER against the number of top DE genes in absence of outliers with μ =0.5, 1.0 and 2.0, respectively for different Fig. 3. Plots of average MER against the number of top DE genes in absence and presence of 10%-20% patient samples contaminated by outliers for randomly selected 20% genes in each of training datasets. (a1-a2) Average MER produced by different methods in absence of outliers and in presence of outliers respectively, with μ=0.5 and σ 2 =1.(b1-b2) Average MER produced by different methods in absence of outliers and in presence of outliers respectively, with μ=1.0 and σ 2 =1. (c1-c2) Average MER produced by different methods in absence of outliers and in presence of outliers respectively, with μ=2.0 and σ 2 =1. TABLE I. PERFORMANCE EVALUATION BASED ON SIMULATED DATASET IN PRESENCE OF OUTLIERS. Performance index Methods AUC pauc LDA SVM NB Logistic Proposed

203 We also observed that classical logistic did better performance compare to the other classical methods (LDA, SVM and NB). So in our next real data study we compare our proposed method with classical logistic classifier. The table 1 shows the performance evaluation based on simulated dataset generated from fig.2 in presence of outliers. It is noticeable from the table 1 that in presence of outliers our proposed method produces higher area under the curve (AUC) and partial area under the curve (pauc). So we can say that in presence of outliers our proposed method performs better than the other methods. Real Dataset To examine the performance of the proposed method with the classical logistic classifier we consider in this study the mouse head and neck cancer gene expression dataset. This dataset consist 12,625 genes from 22 tumor and 22 normal patients, respectively. We downloaded this dataset from We construct training dataset by randomly choosing n 1 =N 1 /2=11 samples from N 1 =22 normal samples and n 2 =N 2 /2=11 samples from N 2 =22 tumor patients. The rest of the patients belong to the test dataset. We select top 20 genes using t- (a) Head & neck cancer data (b) Top DE gene selection using t-test (c) Training data (e) Classical Logistic Fig. 4.(a) Head & neck cancer data with 12,625 genes and 44 samples, (b) p-value against abs(t-statistic) to select the top DE genes (red color), (c) contaminated training dataset,(d) test dataset,(e) classified test data by logistic classifier,(f) classifier test data by proposed method. (d) Test data (f) Proposed shows the heatmap of head & neck cancer dataset and p-value against t-statistic to select the top DE genes, respectively. We contaminated the training data as mentioned earlier by outliers to assess the performance of the proposed method. Fig.4(c-d) shows contaminated training data and the test dataset, respectively. Fig.4(e-f) shows the classified test data set by logistic classifier and proposed classifier, respectively. We can observe from the figure that our proposed classifier has classified the test dataset properly in presence of outliers whereas classical logistic classifier fails to classify the test data. V. CONCLUSSION Classifying patients/samples in to one of two classes (normal/cancer) using gene expression data has become popular from few decades. Many classifier exists in this regards. But they are not robust against outliers. In this paper we propose an outlier detection and modification technique to robustify the classical logistic classifier. Our simulated and real dataset results show that our proposed classifiers performs better than the other classifiers both in absence and presence of outliers ACKNOWLEDGMENT This study is supported by HEQEP Sub-Project (CP-3603, W2, R3), Bioinformatics Lab., Department of statistics, University of Rajshahi, Rajshahi-6205, Bangladesh. REFERENCES [1] A. Sharma, and K.K. Paliwal. Cancer classification by gradient LDA technique using microarray gene expression data. Data Knowl. Eng., vol. 66, pp , [2] S. Dudoit, J f Fridlyand, T. P Speed. Comparison of Discrimination Methods for the Classification of Tumors Using Gene Expression Data. Journal of the American Statistical Association, vol. 97, No. 457, pp , Mar [3] T.R. Golub, D.K. Slonim, P. Tamayo, M. Gaasenbeek C. Huard, J.P. Mesirov, H. Coller, M.Loh, J.R. Downing, M.A. Caligiuri, C.D. loomfield, and E.S. Lander. Molecular classification of cancer: Class discovery and class prediction by gene expression monitoring. Science, pages , Oct [4] Kun-Huang Chen, Kung-Jeng Wang, Min-Lung Tsai, Kung- Min Wang, Angelia Melani Adrian, Wei-Chung Cheng, Tzu-Sen Yang, Nai-Chia Teng, Kuo-Pin Tan and Ku-Shang Chang. Gene selection for cancer identification: a decision tree model empowered by particle swarm optimization algorithms. BMC Bioinformatics, 15:49, [5] Desheng Huang, Yu Quan, Miao He and Baosen Zhou. Comparison of linear discriminant analysis methods for the classification of cancer based on gene expression data. Journal of Experimental & Clinical Cancer Research, 28:149, 2009 [6] Ubharup Guha, Yuan Ji and Veerabhadran Baladandayuthapani. Bayesian Disease Classification Using Copy Number Data. Cancer Informatics, vol. 13 (S2), pp , [7] Sharma, A., C.H. Koh, S. Imoto and S. Miyano. Strategy of finding optimal number of features on gene expression data. Elect. Lett., vol. 47, pp , 2011a. [8] H.A.L. Thi, V.V. Nguyen and S. Ouchani. Gene selection for cancer classification using DCA. Adv. Data Min. Appli., vol. 5139, pp , [9] H. Rattikorn, K. Phongphun, Tumor classification ranking from microarray data. BMC genomics journal, vol. 9, pp. s21, September [10] P. J. Huber. Robust Statistics. John Wiley and Sons: New York,

204 Molecular Evolutionary Analysis of α-defensin Peptide in Vertebrates Arafat Rahman *, M Shahidul Islam, Otun Shaha, Titon Chandra Shaha Department of Microbiology Noakhali Science and Technology University Noakhali, Bangladesh * arafat@nstu.edu.bd Abstract α-defensin is a group of polypeptides with antimicrobial activity found in the host-defense system and it is widely distributed in but mammalian epithelial cells and phagocytes. These molecules protect the organism from a diverse spectrum of bacteria, viruses, fungi, and protozoan parasites. Different studies revealed wide sequence variation within α- defensin sequences, but the underlying evolutionary cause is not well-studied. In this study, the α-defensin gene from 25 vertebrate species has been comprehensively collected and computationally analyzed. Gene and nucleotide databases of NCBI were accessed to extract meta-information of α-defensin gene's 101 defensin domain and 85 leader propeptide sequences, and full CDS sequences downloaded from its nucleotide database by splitting out intron sequence. MEGA software used to construct phylogenetic tree using Neighbor-Joining method which indicates that α- defensin gene evolution does not matches with species evolution and copy number of this gene differs among species. Selection analysis was carried out using DataMonkey web-server's FEL, SLAC, IFEL, MEME, TOGGLE and REL program on both propeptide and defensin super-family codon-aligned sequences to test different hypothesises. Positively selected sites found on both propeptide and defensin domain, but the effect of negative selection pressure is higher on leader sequences. Phyre2 web-server was used for homology modeling of selected α-defensin genes. Structural variation observed on α-defensin proteins which may indicate heterogenous structure-function relationship between species that reflects its interaction with diverse pathogens. This study provides a new perspective on the relationships among α-defensin gene repertoires which will help to infer its evolution. Keywords α-defensin, antimicrobial peptides, vertebrates, evolutionary analysis, selection pressure I. INTRODUCTION Antimicrobial peptides, which are polypeptides of fewer than 100 amino acids, commonly found in animal defense systems and defensins are antimicrobial peptides of innate immunity [1]. In mammalian animals, defensins and cathelicidins are main two peptide family, but defensins are particularly prominent in human [2]. Structurally, defensins are a family of evolutionary related vertebrate antimicrobial peptides with a characteristic β-sheet-rich fold and a framework of six disulphide-linked cysteines that can be divided into two major subfamilies, namely α- and β-defensins, which differ in the length of peptide segments between the six cysteines and the pairing of the cysteines that are connected by disulphide bonds [1]. Defensins form pores in the cell membrane of bacteria by carpet-wormhole model of action using its amphipathic properties and protect the host. Defensins are a very diverged group of molecules. What drives the evolution of defensin is a basic question and there are several studies to explore the nature and effect of selective pressure on β-defensins [3][4]. But there is no such study on α-defensins. To inspect this research gap, this study conducted a selection analysis approach. The ground of selection analysis is based on the occurrences of random mutations which can be fixed eventually in a population. In general, advantageous mutations are extremely rare because proteins already function quite well, so the chance that any change to them is an improvement is very low. Similarly, a deleterious mutation will have little chance because they will be selected against and will not rise in frequency. Based on that, it can be inferred most mutations are neutral in nature and will not change an amino acid [5]. This relationship can be expressed using Ka/Ks ratio, where, Ka/Ks = non-synonymous mutation rate per non-synonymous site / synonymous mutation rate per synonymous site [6][7]. This metric enables to measure the effect of evolutionary pressure on sequence level. Recently, many statistical methods have been implemented in datamonkey webserver ( that are based on the calculation of Ka/Ks ratio and further optimized for different scenario [8]. For example, fixed effect likelihood (FEL) is an overall method in terms of the tradeoff between statistical performance and computational expense [9]. IFEL (Internal FEL) used to test for site-wise selection on internal branches of the tree [10]. Meanwhile, single likelihood ancestor counting (SLAC) is used as the most conservative method to detect selection pressure [9]. Mixed effect fixed evolution (MEME) is an extension of FEL which combines fixed effects at the level of a site with random probability and used to detect episodic diversifying selection affecting individual codon sites [11]. On the other hand, TOGGLE analysis can identify sites which toggle between a wild-type and escaped amino acid state [12]. Finally, the method REL can be used because it allows synonymous rate variation [9]. In this study, these methods are employed to explore selection pressure on α-defensin sequences, and phylogenetic and structural analysis 203

205 were conducted to infer about the effect of selection process. II. MATERIALS AND METHODS A. Dataset formation Nucleotide sequences of α-defensin were searched in the NCBI Gene database. Each hit was accessed, intron part of the sequences was cleaved out by using meta-data of GenBank format file, subsequences combined to form coding sequences (CDS) and downloaded. PROSITE and SMART used for checking the presence and verification of defensin domain and signal sequence [12][13]. Two datasets in fasta format were formed - one for signal propeptide and another for defensin domain sequences. 85 sequences from 25 mammalian species were used to form propeptide dataset; while 101 sequences of 25 mammalian species were used to form defensin domain dataset (Table I). B. Selection analysis Both datasets were aligned by codon using ClustalW algorithm in MEGA5.2. Datamonkey.org web server was used for evolutionary selection analysis by using following six models: FEL, SLAC, MEME, REL, TOGGLE and IFEL [15]. Before analysis, model selection was carried out. For each analysis, default parameter setting was kept and p = 0.1 was used as threshold. C. Phylogenetic analysis Both datasets were used to reconstruct phylogenetics using MEGA5.2. Neighbor-Joining method used with a bootstrap test of phylogeny was used with 1000 replication. Kimura 2-parametric model was used in both cases as substitution model with Gamma distributed rates among sites. Both trees were visualized in FigTree software [16]. D. Homology modelling To predict and analyze structure and characteristic protein folding of α-defensin peptides Phyre2 ( =index) was used in intensive modelling mode with the default parameter of the web server, homology model of α-defensin peptides was constructed and visualized using PyMol software [17]. E. Sequence logo visualization Sequence logo of defensin domain dataset was visualized using WebLogo web server ( after translating in amino acid sequences to demonstrate the consensus amino acid at various positions of the sequence [18]. TABLE I. Propeptide dataset Aotus nancymaae Chrysochloris asiatica Colobus angolensis Callithrix jacchus Chlorocebus sabaeus Cricetulus griseus Cercocebus atys Dipodomys ordii Equus caballus Gorilla gorilla Homo sapiens Jaculus jaculus Macaca mulatta Macaca nemestrina Microtus ochrogaster Macaca fascicularis Microcebus murinus Mus musculus Nomascus leucogenys Pan paniscus Pan troglodytes Papio Anubis Pongo abelii Peromyscus maniculatus Rattus norvegicus DESCRIPTION OF DATASET USED IN THIS STUDY List of Species III. RESULT Defensin domain dataset Aotus nancymaae Cricetulus griseus Cercocebus atys Colobus angolensis Chrysochloris asiatica Chlorocebus sabaeus Callithrix jacchus Dipodomys ordii Equus caballus Gorilla gorilla Homo sapiens Jaculus jaculus Macaca mulatta Macaca fascicularis Mus musculus Microtus ochrogaster Microcebus murinus Nomascus leucogenys Pan troglodytes Papio anubis Pongo abelii Pan paniscus Peromyscus maniculatus Rattus norvegicus Selection Analysis: After model selection, general reversible model (REV) and Hasegawa, Kishino and Yano (HKY85) was found as the best substitution model for defensin domain dataset and leader propeptide dataset respectively. For defensin domain, FEL found 10 positively and 9 negatively selected codon. For propeptide domain, FEL found 5 positively and 13 negatively selected codon. For defensin domain, IFEL found 6 positively and 7 negatively selected codon. For propeptide domain, it found 5 positively and 12 negatively selected codon. For defensin domain, SLAC found 5 positively and 9 negatively selected codon. For propeptide domain, this method found 5 positively and 7 negatively selected codon. MEME found 10 sites with evidence of episodic diversifying selection on both datasets. In TOGGLE analysis, it was observed that toggling occur more on defensin domain comparatively leader propeptide domain. These results are summarized in table II. 204

206 TABLE II. SELECTION ANALYSIS RESULTS ON DEFENSIN DOMAIN AND PROPEPTIDE DATASET Method Defensin Domain Propeptide +ve Selection FEL 7, 8, 9, 11, 12, 13, 17, 24, 26, 28 IFEL 9, 12, 13, 17, 24, 28 SLAC 7, 9, 11, 13, 19 MEME -ve Selection 1, 4, 5, 10, 14, 18, 20, 33, 34 1, 4, 5, 10, 14, 20, 34 1, 4, 5, 10, 14, 18, 20, 33, 34 +ve Selection 19, 33, 44, 48, 55 19, 24, 33, 37, 55 11, 31, 42, 48, 55 -ve Selection 3, 4, 7, 11, 13, 16, 20, 23, 27, 38, 39, 54 3, 9, 11, 13, 16, 23, 25, 27, 31, 38, 39, 43 3, 4, 7, 13, 16, 20, 24 Evidence of Episodic Diversifying Selection 7, 9, 11, 12, 13, 17, 19, 21, 24, 26 4, 19, 32, 33, 37, 42, 44, 55, 56, 61. Phylogenetic Analysis: Phylogenetic analysis reveals, in general, species are clustered in the same clade, but in some cases, the occurrences of distinct species within different clade were observed (tree of propeptide not shown). This occurrence is particularly higher in defensin domain sequence set (Figure 1). In both cases, the α-defensin gene propeptide and defensin domain sequences did not maintain speciesevolution pattern. Fig. 1. Phylogenetic reconsturction of α defensin defensin domain sequences. Homology Modelling: Thirty computational models of α-defensin were constructed. In the α- defensin peptides, the overall structure composed of 2-3 β-strands, where antiparallel β-sheet is prominent. But 3 predicated (Macaca mulatta ; Mus musculus-68009; Equus caballus ) model did not contain any β-strands, although contained α- helix coil. These models suggest that there are differences in the arrangement of disulfide bond and folding. Sequence Logo: In the defensin domain amino acid sequence logo, six cysteine bases were consensus at position 1, 4, 10, 20, 33, 34 which indicates the characteristics conserve amino acid sequence of α- defensin (Figure 2). Besides, there are major consensus of glutamic acid and glycine at 14 and 18 positions respectively. Fig. 2. Amino acid sequence logo visualization of defensin domain sequences showing conserved and diverged base-positions. 205

207 IV. DISCUSSION In general, positive selection on condon position 7, 9, 11, 13, 17 and negative selection on codon position 1, 4, 5, 10, 14, 18, 20, 34 was persistently reported by different methods on defensin domain dataset. On the other hand, positive selection found on codon 19, 33, 55 while negative selection on codon position 3, 4, 7, 13, 16, 20 was reported by these methods on propeptide domain dataset. Mixed effect selection was also reported on these positions. The implication of these results is that relatively positive selection pressure is greater on defensin domain while negative selection is greater on propeptide sequences. On the sequence logo of defensin dataset, the positively selected positions show diverged basevariance while negatively selected positions shows conserve nature. This phenomenon is also reflected on the reconstructed phylogenetic tree. In the tree based on propeptide sequences, same species are clustered within a clade in general. But in the case of defensin domain dataset tree, interleaving nature of species between different clade was more frequent. In homology modeling study, structural variations of defensin protein were evident. This study examined the effect of natural selection on α-defensin gene. The conclusion of the study is that although both positive and negative selection is acting on both propeptide and defensin domain of the gene, there is a dominant role of positive and negative selection on two distinct part of gene negative selection is dominant on propeptide and positive selection is dominant on defensin domain CDS. The explanation of this partial dominance can be hypothesized that function of propeptide is to carry premature defensin peptide to the specific location of a cell so that further maturation of it can take place, and this function, in general, did not change drastically throughout various species in different tissues. But defensin domain, which mainly take part in host-pathogen interaction has to face a wide range of circumstance in which pathogen is continuously changing and trying to evade host immune system [19]. To cope with, defensin peptides need to change itself which might be the cause behind the dominance of positive selection in this region. Further computational and experimental analysis is required to understand this observation. REFERENCES [1] T. Ganz, "Defensins: antimicrobial peptides of innate immunity", Nat Rev Immunol, vol. 3, no. 9, pp , [2] T. Ganz, "Defensins and Other Antimicrobial Peptides: A Historical Perspective and an Update", Combinatorial Chemistry & High Throughput Screening, vol. 8, no. 3, pp , [3] D. Li, J. Tu, D. Li, Q. Li, L. Zhang, Q. Zhu, U. Gaur, X. Fan, H. Xu, Y. Yao, X. Zhao and M. Yang, "Molecular Evolutionary Analysis of β-defensin Peptides in Vertebrates", Evolutionary Bioinformatics, p. 105, [4] M. Boniotto, A. Tossi, M. DelPero, S. Sgubin, N. Antcheva, D. Santon, J. Masters and S. Crovella, "Evolution of the beta defensin 2 gene in primates", Genes and Immunity, vol. 4, no. 4, pp , [5] M. KIMURA, "Preponderance of synonymous changes as evidence for the neutral theory of molecular evolution", Nature, vol. 267, no. 5608, pp , [6] M. KIMURA, "Preponderance of synonymous changes as evidence for the neutral theory of molecular evolution", Nature, vol. 267, no. 5608, pp , [7] M. Kimura, "A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences", J Mol Evol, vol. 16, no. 2, pp , [8] W. Delport, A. Poon, S. Frost and S. Kosakovsky Pond, "Datamonkey 2010: a suite of phylogenetic analysis tools for evolutionary biology", Bioinformatics, vol. 26, no. 19, pp , [9] S. Kosakovsky Pond, "Not So Different After All: A Comparison of Methods for Detecting Amino Acid Sites Under Selection", Molecular Biology and Evolution, vol. 22, no. 5, pp , [10] S. Kosakovsky Pond, S. Frost, Z. Grossman, M. Gravenor, D. Richman and A. Brown, "Adaptation to Different Human Populations by HIV-1 Revealed by Codon-Based Analyses", PLoS Comp Biol, vol. 2, no. 6, p. e62, [11] B. Murrell, J. Wertheim, S. Moola, T. Weighill, K. Scheffler and S. Kosakovsky Pond, "Detecting Individual Sites Subject to Episodic Diversifying Selection", PLoS Genetics, vol. 8, no. 7, p. e , [12] W. Delport, K. Scheffler and C. Seoighe, "Frequent Toggling between Alternative Amino Acids Is Driven by Selection in HIV-1", PLoS Pathog, vol. 4, no. 12, p. e , [13] C. Sigrist, E. de Castro, L. Cerutti, B. Cuche, N. Hulo, A. Bridge, L. Bougueleret and I. Xenarios, "New and continuing developments at PROSITE", Nucleic Acids Research, vol. 41, no. 1, pp. D344-D347, [14] I. Letunic, T. Doerks and P. Bork, "SMART: recent updates, new developments and status in 2015", Nucleic Acids Research, vol. 43, no. 1, pp. D257-D260, [15] K. Tamura, D. Peterson, N. Peterson, G. Stecher, M. Nei and S. Kumar, "MEGA5: Molecular Evolutionary Genetics Analysis Using Maximum Likelihood, Evolutionary Distance, and Maximum Parsimony Methods", Molecular Biology and Evolution, vol. 28, no. 10, pp , [16] "FigTree", Tree.bio.ed.ac.uk, [Online]. Available: [Accessed: 01- Mar- 2016]. [17] L. Kelley, S. Mezulis, C. Yates, M. Wass and M. Sternberg, "The Phyre2 web portal for protein modeling, prediction and analysis", Nat Protoc, vol. 10, no. 6, pp , [18] T. Schneider and R. Stephens, "Sequence logos: a new way to display consensus sequences", Nucl Acids Res, vol. 18, no. 20, pp , [19] P. Raj and A. Dentino, "Current status of defensins and their role in innate and adaptive immunity", FEMS Microbiology Letters, vol. 206, no. 1, pp. 9-18,

208 Microencapsulated Probiotics Protect Spoilage of Functional Foods Md. Shahid Hossain, Md. Abdul Alim Al-Bari*, and Mir Imam Ibne Wahed Department of Pharmacy, University of Rajshahi, Rajshahi-6205, Bangladesh * alimalbari347@ru.ac.bd Abstract-The live microorganisms, probiotics confer nutritional- and health-promoting benefits. Recent demand for natural processes to maintaining health is the inclusion of probiotics in food and other health promoting products. Fruit juices represent an auspicious vehicle for probiotic bacteria; however, there are some encouraging challenges such as the survival of probiotics throughout storage and finally consumer acceptance. Here, we identified probiotics, lactic acid bacteria (LAB) and screened their associated resistance towards different classes of antibiotics. This study investigated the effect of microencapsulation on the survival of lactobacillus acidophyllus, Lactobacillus bulgaricus, Lactococcus lactis and Biffidobacterium biffidum and their acidification roles in fruit juices like orange juice (OJ), lemon juice (LJ) at 35 C for three days and at 4 C over five weeks of storage. Free form probiotics was found to have significant reduction in viability in OJ and LJ at both temperatures. Microencapsulation of these bacteria showed preventive role to decrease the survivability as well as the reduction in acidification at 35 C and 4 C. The loss of food potency and food spoilage associated with pathogenic microbial growth are serious problem in tropical countries, like Bangladesh. The probiotic OJ may become an important functional food due to its extension of shelf life, market reputation, profits and innate natural tastes. Keywords- Probiotics, microencapsulation, biopreservation, functional food. I. INTRODUCTION Probiotics are defined as live microbial feed supplements that have beneficial effects on the host by improving its intestinal microbial balance, or as live microorganisms that, when administered in adequate amounts, confer a health benefit on the host [1]. Probiotic bacteria have been paid considerable attention to scientific researchers for treating several diseases including inflammatory bowel diseases, colorectal cancer, metabolic syndromes, autoimmune diseases like rheumatoid arthritis etc., as well as in diagnostic purposes such as production of biomarkers e.g., detection of cancers and diabetes [2,3]. To achieve beneficial effects for human health, probiotic bacteria must have good technological properties, survive gastrointestinal passage and be able to function in the gut environment [4]. When probiotic bacteria administered orally, they must be protected from the stomach acidic condition and can be denatured by bile acids, antimicrobial compounds and degradative enzymes before reaching the target sites [4]. These obstacles limit their arrival to the gut in alive. The microencapsulation technique protects them from these unfavorable conditions. In microencapsulation, the live probiotic bacteria and/or their bioactive compounds are trapped to protect from the harsh conditions and to deliver them with improved survival rate. Microencapsulation is defined as the technology for packaging solids, liquids or gaseous materials in miniature, sealed capsules that can release their contents at controlled rates under specific conditions [5]. Microcapsule coating protects the active content from environmental stresses such as acidity, oxygen and gastric conditions, and can be used to help the content pass through the stomach. Besides enhancing the viability of bacteria, microencapsulation facilitates handling of cells and allows a controlled dosage. Microencapsulation technique of food processing is an innocuous and ecological approach for food preservation. To harmonize consumer demands with the necessary safety standards, traditional means of controlling microbial spoilage and safety hazards in foods are being replaced by innovative technologies including biological antimicrobial systems. This trend has favored consumption of foods enriched with physiologically active components including probiotic bacteria [6]. In biopreservation technology, probiotics especially lactic acid bacteria (LAB) and/or their natural antimicrobial products are used to prolong the shelf life and regulate the growth and proliferation of endogenous pathogenic bacteria in foods [1, 2]. LAB are also associated with the formation of fermented products and thus have generally recognized as safe (GRAS) status granted by the US Food and Drug Administration (USFDA) [7]. We generally designate a food as functional if it gives beneficial effects on body functions, in addition to the classical nutritional value. For examples, functional foods are those prepared with/or containing bioactive compounds, such as dietary fiber and active friendly probiotic bacteria that promote the homeostasis of intestinal bacterial strains. In addition to the functional ingredients, such as vitamins and minerals, the inclusion of probiotics in fruit juices like orange juice (OJ) can provide an effective means for the generation of healthy food with the increase in life expectancy [8]. Thus, development of foods with LAB is one of the key research priorities of the food industry. Modern technologies in food processing have reduced but not completely eliminated the food- 207

209 related illness and product spoilage by pathogens in industrialized countries. In order to achieve improved food safety against such pathogens, food industry makes use of chemical preservatives or physical treatments (e.g. high temperature) which have proven toxicity (e.g., chemical preservatives). So, there are increasing demands of consumers for products that are safe enough but minimally processed with no addition of preservatives [9]. To harmonize the consumer demands with necessary safety standards, the traditional means of controlling microbial spoilage and safety hazards in foods are being replaced by several innovative technologies such as biopreservation of foods with probiotics. The microencapsulation of probiotics is one of the processes for improving their viability in functional foods [10]. Here, we have designed our protocol to compare the preservative activities of isolated LAB. We have also designed the effect of microencapsulated LAB on the potency and functional characteristics of OJ. II. METHODS AND MATERIALS A. Sample collection and processing Probiotic lactic acid bacteria (Lactobacillus acidophyllus, L. bulgaricus, L. lactis, L. casei, L. delbrueckii, L. plantarum, L. helveticus, Biffidobacterium biffidum) were identified in our previous study on the basis of carbohydrate fermentation profiles, catalase activities and other biochemical tests were used in this study [11, 12]. All strains were stocked in 15% glycerol and stored at 80 C. The fruits, lemon and orange were purchased from the local market. The selected fruits were washed thoroughly with running tap water, rinsed with distilled water and blotted dry. Then good quality lemons and oranges were sliced and the seeds were separated manually from the pulp. The juices were then extracted by hand squeezing and straining the above prepared material through double fold muslin cloth. The OJ and LJ were deposited in sterile containers individually and kept at 4 C prior to use for further studies. Individual probiotic bacteria was cultured on the de Man, Rogosa and Sharpe (MRS) media (Difco, USA) and incubated at 25ºC and 37ºC temperatures as well as several ph (4.5 and 6.5) under anaerobic and aerobic conditions for 72 hrs. Before inoculation into O J a n d L J, the bacterial cells were washed in sterile saline to remove any residual MRS broth medium. Aliquots (10 ml) of CJ were dispensed into culture tubes with caps and were statically incubated at 30 C after inoculation. Viable cell counts as colony forming units (CFU/ml) of the inoculum were determined by the standard plate method with Lactobacilli MRS medium after 72 h of incubation at 30 C. B. Microencapsulation of probiotic bacteria For microencapsulation, individual LAB were cultured statically in MRS broth and mixed separately with an equal volume of a 3% w/v sterile sodium alginate solution at 21 C. The alginateorganism suspension was dispensed slowly into a beaker containing 600 ml of vegetable oil and 1 gm of Tween 80 to form an emulsion. The emulsion was shaken at 200 rpm and calcium chloride (0.01M) was added gently to the side of the beaker. After 60 min, the aqueous phase (remaining amount of calcium chloride) was removed by decantation and the beads were stored at 4 C until utilized. The enumeration of free form of LAB and microencapsulated organisms were performed using methods described by Hossain [11, 12]. Enumeration of the organisms in OJ and LJ were performed on a weekly basis over a period of six weeks, using MRS medium at 37 C for 72 hrs under anaerobic conditions. C. Biopreservation activities of probiotics To examine the biopreservative activities of probiotic bacteria, bacterial cell viability in OJ and LJ (25 ml) were fermented for 72 h at 30 C and then stored at +4 C for up to six weeks. The juices were taken at weekly intervals, and the viability of probiotic cultures in probiotic OJ and LJ containing free and encapsulated bacteria were determined and expressed as CFU/ml after culturing LAB on MRS anaerobically at 30 C for two days. III. RESULTS A. Cultural characteristics of lactic acid bacteria Lactic acid bacteria (LAB) were isolated and recovered from different samples of curds and probiotic products of different pharmaceutical companies on MRS media and their number was enumerated. The recovered bacteria have different colony morphologies from which four distinctive bacteria were selected and identified [11, 12]. In order to obtain a universal co-culture medium that would support to grow well and survive of all the LAB strains, a number of culture conditions were investigated such as different culture media, liquid or solid cultures, anaerobic or aerobic phases, different temperatures and a wide range of phs. MRS medium was found to support the growth of each of the test strains, and based on these studies; MRS broth medium was selected for future all assays. The cocultures of Lactobacilli strains displayed tolerance to broad range of phs, however, bacterial growth was better when ph ranged from 3 to 4 and maximum at ph 3.5, in 48-72h at 3 0 C. The physical appearance of fermented LJ before inoculation of LAB were very light green, flavour before inoculation of LAB, pleasant, initial phs before inoculation of LAB, and storage temperature 4 C. However, after inoculation of LAB, physical appearance of LJ was light green, unpleasant flavor at the end of the shelf life, ph increased up to 7 at 4 C. There were high similarities in growth of mono-culture for individual strains of bacteria, and were relatively constant in 48-72h at 30 C. 208

210 B. Survival rates of free and microencasulated probiotic bacteria in OJ and LJ The survival rates of free and microencapsulated LAB inoculated into OJ and LJ were shown in Fig. 1 and 2. Unencapsulated LAB in co-cultures declined rapidly in the OJ and LJ within the four weeks and there were no viable bacteria remaining by the fifth week (Fig. 1). In free form, all the probiotic bacteria in monocultures and cocultures showed similar loss of viability. However, a synergistic action occurred when LAB1-4 were given in combination and greater viable numbers were found than all the other monocultures and co-cultures. Free m u l t i p l e probiotic bacteria lost their viability slowly when compared to single cultured probiotic bacteria in the OJ and LJ within the four weeks. S i n c e OJ a n d L J w e r e acidic and the survival rate of multiple co-cultured probiotic bacteria was enhanced than any monocultured bacteria. The encapsulated probiotic bacteria in O J a n d LJ did not lose their viability as rapidly as the free probiotic bacteria and >10 6 CFU/mL were still present after six weeks of storage (Fig. 2). C. Biopreservation and stability of OJ and LJ The phs of OJ and LJ were adjusted to 3.94 using citrate buffer and the results are shown in Fig. 3 which indicates the ph changes in OJ a nd LJ containing free and microencapsulated bacteria and their effects during a storage period of 4-6 weeks. A similar trend in the decline in ph was seen in both free and microencapsulated OJ with LAB. However, the final ph at the end of the six weeks storage period of OJ and LJ with encapsulated bacteria was higher than that inoculated with free form. The average ph decreased from 3.94 to 2.71 in OJ and LJ containing free bacteria after four weeks of storage (Fig. 3), whereas the ph declined to only 3.3 in the juice containing encapsulated bacteria during the storage period (Fig. 3). This result suggests that bacteria in microencapsulated state had a more stable survival environment. The preservative activity of LAB had been observed due to the lowering of ph to below 4 through acid production and the inhibition of growth of pathogenic microorganisms which can cause food spoilage, food poisoning and disease [13]. By doing this, the shelf life of fermented food is prolonged. In order to investigate the role of LAB in OJ and LJ preservation, we checked ph changes juice only; juice containing pathogenic bacteria (PB); juice containing LAB and juice containing chemical preservatives (PR, propyl paraben and methyl paraben, ratio 5:1). The LAB containing juice decreased in ph from 3.94 to 3.3 (Fig. 3). After 72 hrs, ph increased in juice only due to contamination of microorganisms, and in juice containing PB. Comparing to LAB in juice, the juice had almost stabled ph when chemical preservative was used (Fig. 3). IV. DISCUSSIONS Microcapsules provide a more favorable anaerobic environment for the sensitive probiotic bacteria, as well as a physical barrier from the harsh acidic conditions of the OJ [14, 15]). The fascinating results suggested that the higher survival rate of microencapsulated multiple bacteria were found when inoculated into OJ than the free form (Fig. 1 and 2). It is important to have a significant number of viable lactic acid bacteria present in the probiotic products for maximum health benefits. Several factors could affect the cell viability of lactic acid cultures in probiotic food products. Several reports have indicated differences among strains of probiotic bacteria with respect to their survival in acidic environment [16]. Free form of bacteria m ight utilize carbohydrates and produced s e v e r a l organic acids, thus lowered the ph of the product during the storage. Many of the free bacteria were not viable at later stages of storage; although the dead probiotic cells could release enzymes for hydrolyzing sugars in the fruit juice and thus lowering the ph [15, 17]. These results demonstrated that microencapsulated bacteria would make a more stable product over a longer storage period. Several s tudies have also shown that encapsulated probiotic bacteria c a n make more stable functional food products [17]. Probiotic bacterial cultures are commonly used in the dairy industry; and some products produced during lactic acid fermentation such as lactic acid, diacetyl, and acetaldehyde. These products, which are produced during lactic acid fermentation, could be associated with the decrease in ph [18]. In this study, the LAB survived in the fermented OJ with high acidity and low phs. A rapid decrease in ph in the beginning of fermentation enhanced the quality of the end product as well as minimized the influence of spoilage bacteria. The inhibitory action of probiotic bacteria against the pathogens might help the accumulation of metabolites such as lactic acid as well as antimicrobial compounds such as H 2 O 2 and antibiotics, bacteriocins. Although the production levels and proportions among these compounds depend on the strain, medium compounds and physical parameters. The problem of food spoilage is prevailing especially in the developing countries [19]. In most of the regions, low temperature is used to increase the shelf life of food. In many countries including Bangladesh ambient temperature is high in most part of the year. In the present scenario, disrupted and expensive supply of electricity provision of low 209

211 temperature for storage purpose has become difficult for most of the people. In such situations food become more susceptible to bacterial and fungal attack. Although antimicrobial agents are available in wide varieties, the legal preservatives and marketing must be taken into considerations for prolonged food shelf life and effective means of antimicrobial agents ensuring food safety [20]. Thus, the preservative activity of LAB containing juices has become prominent in fermented products due to lower ph below 4. It inhibits the growth of pathogenic microorganisms which can cause food spoilage, food poisoning and diseases. V. CONCLUSION Since probiotication of fruit juices used no preservatives and long shelf-lives were obtained, it is important for the juice safety and potency. Although the encapsulation method increases the survival of probiotics in the juice, the effect of probiotic beads on the sensory characteristic and consumers conception should be analyzed further. log10(cfu/ml) Fig. 1. Viability of free form probiotic bacteria in juices log10 (CFU/ml) Initial week 1 week 2 week 3 Lemon juice Orange juice LAB 1 LAB 1,2 LAB 1-4 LAB 1 LAB 1,2 LAB 1-4 Initial week 1 week 2 week 3 week 4 week 5 12 Lemon juice Orange juice Fig. 2. Viability of microencapsulated probiotic bacteria in juices ph change LAB 1 LAB 1,2 LAB 1-4 LAB 1 LAB 1,3 LAB LJ LJ + PB Before LJ + LAB LJ + PR After OJ OJ + PB OJ + LAB OJ + PR Fig. 3. Biopreservation role of probiotic bacteria. Where lemon juice (LJ), orange juice (OJ), pathogenic bacteria (PB), lactic acid bacteria (LAB) and preservative (PR). REFERENCES [1] J.M. Wells, and A. Mercenier, Mucosal delivery of therapeutic and prophylactic molecules using lactic acid bacteria, Nat. Rev. Microbiol. vol. 6, pp , [2] E.R. Farnworth, The evidence to support health claims for probiotics, J. Nutrition vol. 138, pp. 1250S 4S, [3] T. Danino, A. Prindle, G.A. Kwong, M. Skalak, H. Li, K. Allen, et al. Programmable probiotics for detection of cancer in urine, Sci. Translational Med. vol. 7 (289), pp. 289ra84, [4] M.A. Islam, Y. Cheol-Heui, C. Yun-Jaie, and C. Chong-Su, Microencapsulation of live probiotic bacteria, J. Microbiol. Biotechnol. vol. 20(10), pp , [5] S. Rokka and P. Rantamäki, Protecting probiotic bacteria by microencapsulation: challenges for industrial applications, Eur. Food Res. Technol. vol. 231, pp. 1 12, [6] W.K. Ding and N.P. Shah, An improved method of microencapsulation of probiotic bacteria for their stability in acidic and bile conditions during storage, J. Food Sci., vol. 74, pp. M53 M61, [7] E. Stykova, R. Nemcova, and I. Valocky, Adherence of bacteria to mucus collected from different parts of the reproductive tract of heifers and cows, Can. J. Microbiol. vol. 59, pp , [8] D. Granato, G.F. Branco, F. Nazzaro, A.G. Cruz, J.A.F. Faria, Functional foods and nondairy probiotic food development: Trends, concepts, and products, Compr. Rev. Food Sci. Food Safety, vol. 9, pp , [9] G. Pereira-Caro, C.M. Oliver, R. Weerakkody, T. Singh, M. Conlon, G. Borges, et al. Chronic administration of a microencapsulated probiotic enhances the bioavailability of orange juice flavanones in humans, Free Rad. Biol. Med. vol. 84, pp , [10] J.M. Rolain, Food and human gut as reservoirs of transferable antibiotic resistance encoding genes, Frontiers in Microbiol. vol. 4, pp. 173, [11] M.K. Islam, M.A.A. Al-Bari, A. Khan, M.K. Zahan and M.A. Islam, Identification and characterization of Lactobacilli from Rajshahi traditional curds. British Microbiol. Res. J. vol. 6(1), pp , [12] M.S. Hossain, M.A.A. Al-Bari, and M.I.I. Wahed, Biochemical characterization of probiotics available in Bangladesh, J. Sci. Res. Vol. 8, pp , [13] S. Nath, S. Chowdhury, K.C. Dora and S. Sarkar, Role of biopreservation in improving food safety and storage. Intl J. Enginer. Res. Applic. Vol.4, pp , [14] G. Mohan, P. Guhankumar, V. Kiruththica, N. Santhiya and S. Anita, Probiotication of fruit juices by Lactobacillus acidophilus, Intl J. Adv. Biotechnol. Res.vol. 4, pp , [15] K. Kailasapathy, Survival of free and encapsulated probiotic bacteria and their effect on the sensory properties of yoghurt, LWT - Food Sci. Technol. vol. 39, pp , [16] A.C. Ford, E.M. Quigley, B.E. Lacy, A.J. Lembo, Y.A. Saito, L.R. Schiller, et. al. Efficacy of prebiotics, probiotics, and synbiotics in irritable bowel syndrome and chronic idiopathic constipation, Am. J. Gastroenterol. vol. 109(10), pp , [17] M. Saarela, H.L. Alakomi, J. Mättö, A.M. Ahonen and S. Tynkkynen, Acid tolerant mutants of Bifidobacterium animalis subsp. lactis with improved stability in fruit juice, Food Sci. Technol. vol. 44, pp , [18] K.R. Aneja, R. Dhiman, K.N. Aggarwal and A. Aneja, Emerging preservation techniques for controlling spoilage and pathogenic microorganisms in fruit juices, Intl J. Microbiol. vol. 2014,

212 [19] F. Jabeen and J.I.Qazi, Potential of bacterial chitinases and exopolysaccharides for enhancing shelf life of food commodities at varying conditions. Intl Res. J. Environ. Sci. vol. 3, pp , [20] C. Pérez-Pérez, C. Regalado-González, C.A. Rodríguez- Rodríguez, J.R. Barbosa-Rodríguez and F. Villaseñor-Ortega, Incorporation of antimicrobial agents in foodpackaging films and coatings, Adv. Agri. Food Biotechnol. vol. 37, pp ,

213 Preparation of Highly Cross-linked Magnetic Polymer Composite Particles and Application in the Separation of Arsenic from Water M. K. Sharafat*, H. Ahmad, M. A. Alam, M. M. Rahman Abstract - Iron oxide magnetic particles have become a promising research field in separation technology because of their easy separation by external magnetic field and can be applied for the removal of toxic metals from waste water. Highly cross-linked Fe 3 O 4 /P(S-DVB) particles were prepared in this research by suspension polymerization of styrene (S) and divinylbenzene (DVB) in presence of nanosized Fe 3 O 4 particles. At first Fe 3 O 4 nanoparticles were prepared by coprecipitation of Fe 2+ and Fe 3+ from their alkaline aqueous solution. To stabilize the magnetic particles, the surface of the particles was modified with oleic acid. The morphology and surface structure were characterized by FTIR, TEM, SEM, and light microscope. Keywords - Styrene, Divinylbenzene, Suspension polymerization, Highly cross-linked polymer. Depratment of Chemistry, Rajshahi University, Rajshahi-6205 * mksharafat10@gmail.com I. INTRODUCTION The contamination of daily used water by toxic metal ions is a worldwide environmental problem. Since the degradation of toxic metal ions does not occur naturally, they have a tendency to accumulate in organisms and enter the food chains [1]. Among the different toxic metals, arsenic (As) is a common and toxic pollutants which is mobilized by natural weathering reactions, biological activity, geochemical reactions, volcanic emissions and other anthropogenic activities. However, mining activities, combustion of fossil fuels, use of As pesticides, herbicides, and crop desiccants and use of As additives to livestock feed create additional impact. As shows different oxidation states 3, 0, +3 and +5 [2]. Inorganic forms of As present in water is more toxic than organic form [3]. 3 ) Arsenite (AsO 3 and arsenate (AsO 3 4 ), referred to as As(III) and As(V) are two common form of As in natural water. Pentavalent species predominate and are stable in oxygen rich aerobic environments whereas trivalent species in moderately reducing anaerobic environments such as groundwater [4]. Drinking As contaminated water over a long period causes skin, lung, bladder, and kidney cancer as well as pigmentation changes, skin thickening (hyperkeratosis) neurological disorders, muscular weakness, loss of appetite, and nausea [5,6]. Since exposure to toxic metals, even at trace level, is believed to be a risk for human beings [7-9], thus, removing of undesirable metals from water system has become a challenging task. Recently adsorption technique is used in large scale which offers flexibility in design and operation, and in many cases it will generate high-quality treated effluent [10]. Among different types of adsorbents nano-sized metal oxides (NMOs), including ferric oxides, manganese oxides, aluminum oxides, titanium oxides, magnesium oxides and cerium oxides, are highly used for removing toxic metals from aqueous systems [11-13] where iron oxide particles show recycling advantages because of its magnetic nature [14]. In this research iron oxide (Fe 3 O 4 ) nanoparticles are coated with divinylbenzene (DVB) cross-linked polystyrene (PS), the magnetic NMOs-based composite adsorbents also allow easy isolation by an external magnetic field from aqueous solutions for recycling or regeneration [15]. 212

214 II. EXPERIMENTAL A. Chemicals and instruments Styrene of monomer grade, purchased from Fluka, Chemika, Switzerland, was distilled under reduced pressure. Crosslinking agent divinylbenzene (DVB) from Sigma-Aldrich, Chemie (80% grade) was purified with aqueous 10% NaOH solution and subsequently dehydratd by stirring with anhydrous CaCl 2. Benzoyl peroxide (BPO) from BDH Chemicals Ltd. UK was recrystallized from methanol and preserved in the refrigerator before use. Polyvinyl alcohol (PVA ) from Thomas Baker (Chemicals) Limited, India, of molecular weight gmol -1 was used as polymeric stabilizer. As 2 O 3 from May & Baker, UK, ferric chloride (FeCl 3 ), ferrous sulphate heptahydrate (FeSO 4. 7H 2 O), NH 4 OH, Oleic acid and other chemicals were of analytical grade. Deionized water was distilled using a glass (Pyrex) distillation apparatus. TEM was performed with a JSEM-1230 microscope (JEOL, Tokyo, Japan) operating at a voltage of 100 kv, and optical image was obtained by a fluorescence microscope (IX71 Olympus, Japan) to see the particle morphology and size distribution. IR spectrophotometer, FTIR (Perkin Elmer, FTIR-100, UK), Double beam UV-visible spectrophotometer (Shimadzu, UV-1650pc), and microprocessor ph meter from Mettler Toledo (MP 220, Switzerland) Instruments were used in this study. Homogenizer (T 18 Digital Ultra Turrax, IKA, Germany) is used to homogenized the particles. B. Preparation of Fe 3 O 4 Nanoparticles Nano-sized Fe 3 O 4 particles were produced by co-precipitation of Fe 2+ and Fe 3+ from their aqueous solutions (molar ratio1:2) containing 25% NH 4 OH. The reaction was carried out in a three necked round flask under a nitrogen atmosphere for 2h. Oleic acid was used as a stabilizer. C. Preparation of reference P(S-DVB) particles Styrene (5.3 g), DVB (0.4 g), PVA (1.25 g), toluene (1.50 g) were taken in 50 ml distilled water, homogenized at 10,000 rpm sonication for 3 min in ice-water bath to form monomer droplets. The homogenized mixture was polymerized in presence of BPO at 75 C. D. Preparation of Fe 3 O 4 /P(S-DVB) composite particle Styrene (5.3 g), DVB (0.4 g), toluene (1.5 g) and 0.88 g (65% solid in octane) magnetic fluid were mixed to form the oil phase. PVA (1.25 g) and NaCl (1.5 g) were dissolved into 50 ml deionized water to form the water phase. Then the two phases were mixed together with ultrasonic treatment at 5,000 rpm sonication for 3 min prior to the polymerization in presence of BPO at 75 C. E. Adsorption behavior of Arsenic The absorption behavior of arsenic on both polymer and composite particles were measured using UV-Vis spectrophotometer. For each measurement, a mixture of 30 ml of 40 ppm As(III) solution was prepared from g purified particle solution (0.1g solid) and 1000 ppm arsenic stock solution. The absorbance was measured at the wavelength of 535 nm by a UV-visible spectrometer. The amount of arsenic adsorbed was calculated by subtracting the concentration in the medium from that of initial concentration. Calibration curve was used for this purpose. III. RESULTS AND DISCUSSIONS The particle size and particle distribution of Fe 3 O 4 nanoparticles are illustrated in illustrated in TEM images (Fig. 1). 213

215 a Fig. 1: TEM image of Fe 3 O 4 nanoparticles. Magnification scale represents 50 nm. Before the observation magnetic particles were separated by means of an external magnetic field from the dispersion and washed repeatedly using the same technique using water. The advantage of this magnetic separation is to separate only magnetic particles. The average size of Fe 3 O 4 is about 6-8 nm. The SEM photographs of Fe 3 O 4 nanoparticles and Fe 3 O 4 /P(S-DVB) composite polymer particles prepared at respectively 5,000 rpm sonication speed are illustrated in Fig. 2. The average diameter is µm for Fe 3 O 4 /P(S-DVB) composite particles. The particle size distribution is pretty broad as indicated from high coefficient of variation (45%). The composite polymer particles had microporous structure as evident from the magnified image (not shown). The particle size of the composite particles in the dispersion state characterized by light microcopy is nearer to the dry state taken by TEM. That suggests the high cross-linked and hydrophobic nature of composite particles. FTIR is a useful tool to characterize the structural composition of the colloidal particles [16]. A comparative plot of FTIR spectra of P(S-DVB), Fe 3 O 4 and Fe 3 O 4 /P(S- DVB) composite Particles is shown in Fig. 3. Fig. 2: SEM image of P(S-DVB)/Fe 3 O 4 composite particles prepared at sonication 5000 rpm. Magnification scale represents 5 μm. Fig. 3: FTIR spectra of a) P(S-DVB) particles, b) Fe 3 O 4 and c) Fe 3 O 4 /P(S-DVB) composite particles. Fe O bonds appear at 583 and 382 cm -1 in Fe 3 O 4 particles weakened in Fe 3 O 4 /P(S- DVB) composite particles. The new peaks that appear in the region cm -1 represent aliphatic CH stretching vibrations of aliphatic chain. In the spectrum of Fe 3 O 4 /P(S-DVB) composite particles C-H and C=C stretching vibrations of the aromatic ring appear at around 3000 cm -1. These results suggest that most of the Fe 3 O 4 particles have successfully been encapsulated. The adsorption magnitude of As in mass per unit area of the particle surface is shown in Fig. 4. The higher adsorption capacity on Fe 3 O 4 /P(S-DVB) composite polymer particles is due to both porosity and interaction with magnetic nanoparticles dispersed in polymer matrix. 214

216 Fig. 4: Magnitude of adsorption of arsenic (As) on P(S- DVB) polymer particles and Fe 3 O 4 /P(S-DVB) composite particles both prepared at 5,000 rpm sonication at 25⁰C and ph 9. IV. CONCLUSION Highly cross-linked magnetic polymer composite particles were prepared by a novel suspension polymerization of styrene and DVB in presence of nanosized Fe 3 O 4 particles. The removal of As by magnetic composite particles was driven by both porosity and interaction with magnetic nanoparticles. REFERENCES [1] J. Wang, C. Cheng, X. Yang, C. Chen, A. Li, A new porous chelating fiber: preparation, characterization, and adsorption behavior of Pb(II), Ind. Eng. Chem. Res., vol. 52, pp , [2] P.L. Smedley, H.B. Nicolli, D.M.J. Macdonald, A.J. Barros, J.O. Tullio, Hydrogeochemistry of arsenic and other inorganic constituents in ground waters from La Pampa, Argentina, Appl. Geochem. vol. 17, pp , [3] I. Bodek, W.J. Lyman, W.F. Reehl, D.H. Rosenblatt, Environmental Inorganic Chemistry: Properties, Processes and Estimation Methods, Pergamon Press, USA, [4] N.N. Greenwood, A. Earnshaw, Chemistry of Elements, Pergamon Press,Oxford, 1984, Chapter 13. [5] C.K. Jain, I. Ali, Arsenic: occurrence, toxicity and speciation techniques, Water Res. vol. 34, pp , [6] M.D. Kiping, J. Lenihan, W.W. Fletcher (Eds.), Arsenic, the Chemical Environment, Environment and Man, vol. 6, pp [7] S. Khan, Q. Cao, Y.M. Zheng, Y.Z. Huang, Y.G. Zhu, Health risks of heavy metals in contaminated soils and food crops irrigated with wastewater in Beijing, China, Environ. Pollut., vol. 152, pp , [8] A. Singh, R.K. Sharma, M. Agrawal, F.M. Marshall, Health risk assessment of heavy metals via dietary intake of foodstuffs from the wastewater irrigated site of a dry tropical area of India, Food Chem. Toxicol., vol. 48, pp , [9] M. Jamil, M.S. Zia, M. Qasim, Contamination of agro-ecosystem and human health hazards from wastewater used for irrigation, J. Chem. Soc. Pak., vol. 32, pp , [10] B.J. Pan, B.C. Pan, W.M. Zhang, L. Lv, Q.X. Zhang, S.R. Zheng, Development of polymeric and polymer-based hybrid adsorbents for pollutants removal from waters, Chem. Eng. J., vol. 151, pp , [11] J.E. Vanbenschoten, B.E. Reed, M.R. Matsumoto, P.J. McGarvey, Metal removal by soil washing for an iron oxide coated sandy soil, Water Environ. Res., vol. 66, pp , [12] J.A. Coston, C.C. Fuller, J.A. Davis, Pb 2+ and Zn 2+ adsorption by a natural aluminum- and iron-bearing surface coating on an aquifer sand, Geochim. Cosmochim. Acta., vol. 59, pp , [13] A. Agrawal, K.K. Sahu, Kinetic and isotherm studies of cadmium adsorption on manganese nodule residue, J. Hazard. Mater., vol. 137, pp , [14] A.R. Mandavian, M.A.S. Mirrahimi, Efficient separation of heavy metal cations by anchoring polyacrylic acid on superparamagnetic magnetite nanoparticles through surface modification, Chem. Engg. J., vol. 159, pp , [15] X. Zhao, L. Lv, B. Pan, W. Zhang, S. Zhang, Q. Zhang, Polymer-supported nanocomposites for environmental application: A review, Chem. Eng. J., vol. 170, pp , [16] X. Hu, Y. Wang, B. Peng, Chitosan-capped mesoporous silica nanoparticles as ph-responsive nanocarriers for controlled drug release, Chem. Asian J., vol. 9, pp ,

217 Preparation of Hydrophobic Poly (lauryl methacrylate-divinyl benzene) Coated Magnetic Nano-Composite Particles and their Application as Adsorbents for Organic Pollutants Rukhsana Shabnam* and Hasan Ahmad Department of Chemistry, University of Rajshahi, Rajshahi-6205, Bangladesh * samarhass@yahoo.com Abstract - This work introduces a method for the preparation of magnetic nano-composite particles coated with highly crosslinked poly(lauryl methacrylate) (PLMA), a hydrophobic polymer containing long chain alkyl groups for application in waste water treatment. The produced magnetic composite particles named Fe 3 O 4 /SiO 2 /P(LMA-DVB) were characterized by Fourier Transform IR (FTIR), transmission electron microscopy (TEM) and X-ray diffractometer (XRD) analyses. Then the prepared composite particles were used for the removal of organic pollutants from water. Keywords - lauryl methacrylate (LMA), suspension polymerization, hydrophobic, nano-composite, organic pollutant I. INTRODUCTION In the modern era, a chemist or an industrialist is going with a great challenge to face the contamination of the underground sources in irreversible way with recalcitrant and nonbiodegradable organics [1-5]. An excess of a few parts per billion (ppb) of some organic pollutants may cause serious health problem [6]. Due to the improved dispersibility in water/salt and compatibility with the organic pollutant, the polymer-coated magnetic nanoparticles offer advantages on removal of pollutants. Resins for chromatographic separation, water purification, oil absorbency agents, viscosity modifiers, oil-soluble drag reducers etc are the well known applications of hydrophobic latex particles [7, 8]. The coating of hydrophobic polymer on Fe 3 O 4 particles is expected to increase adsorption ability, particularly of organic pollutant. Lauryl methacrylate (LMA) is a well-known industrially important hydrophobic monomer because it has long chain alkyl group, which in addition to its hydrophobicity also gives polymers with high flexibility (low glass transition temperature, (T g ). In this investigation we prepared highly cross-linked PLMA coated magnetic nano-composite particles by restricting the polymerization of LMA and divinylbenzene (DVB) within the droplets containing Fe 3 O 4 /SiO 2 particles using oil-soluble initiator. II. A. Chemicals and instruments EXPERIMENTAL LMA from Fluka, Chemika (Switzerland) was washed with 10%NaOH aqueous solution to remove any inhibitor and finally passed through activated basic alumina by column chromatography. Crosslinking agent DVB from Sigma-Aldrich, Chemie (80% grade) was purified with aqueous 10% NaOH solution and subsequently dehydrated by stirring with anhydrous CaCl 2. Benzoyl peroxide (BPO) from BDH Chemicals Ltd., UK was recrystallized from methanol and preserved in the refrigerator before use. Tetraethylorthosilicate (TEOS) from Sigma Aldrich, USA, was preserved in the refrigerator and used without purification. Other chemicals were of analytical grade. Deionized water was distilled using a glass (Pyrex) distillation apparatus. TEM was performed with a JSEM-1230 microscope (JEOL, Tokyo, Japan) operating at a voltage of 100 kv, and optical image was obtained by a fluorescence microscope (IX71 Olympus, Japan) to see the particle morphology and size distribution. FTIR (Perkin Elmer, FTIR-100, USA), X-ray diffractometer (Bruker D8 Advance, Germany) were used for the characterization of the latex particles. B. Preparation of Fe 3 O 4 /SiO 2 /P(LMA-DVB) particle Magnetic (Fe 3 O 4 ) nano particles were prepared by co-precipitation of Fe 2+ and Fe 3+ from their alkaline aqueous solutions which was encapsulated with SiO 2 using TEOS. Then precipitation copolymerization was carried out using LMA and DVB in presence of Fe 3 O 4 /SiO 2 particles within stable droplets containing hexadecane toluene mixture (4:1 mixture HD-T) as porogen using BPO as an initiator. The produced magnetic composite particles named Fe 3 O 4 /SiO 2 /P(LMA-DVB) were washed by repeated replacement of continuous phase with distilled water. C. Adsorption behavior of organic pollutants on Fe 3 O 4 /SiO 2 /P(LMA-DVB) particle For adsorption measurement, purified Fe 3 O 4 /SiO 2 /P(LMA-DVB) composite particles were mixed with aqueous solution of gramaxone, phenol, salicylic acid and congo red respectively. Allowing the dispersion pollutant mixture to stand for 2 h and then the amount of adsorption was measured using UVvisible spectrophotometer. III. RESULTS AND DISCUSSION TEM images of Fe 3 O 4 /SiO 2 and Fe 3 O 4 /SiO 2 /P(LMA-DVB) composite particles are shown in Fig. 1. The average diameters are 43 nm and 52.6 nm for Fe 3 O 4 /SiO 2 and Fe 3 O 4 /SiO 2 /P(LMA-DVB) composite particles. The small increase is attributed to 216

218 the coverage of Fe 3 O 4 /SiO 2 particles with cross-linked PLMA layer. Due to nano dimension it is not possible to confirm the core-shell structure. The composite particles are bit coagulated while drying during sample preparation. a b characteristic peaks for Fe 3 O 4 nanoparticles at 30.3, 35.6, 43.5, 54.0, 57.4 and 63.0, which can be assigned to (220), (311), (400), (422), (511) and (440), respectively. These signals matched well with the database of magnetite in the JCPDS-International Center (JCPDS Card: ) [9]. The peak positions remained the same during different steps of modification indicating that the crystalline structure of the magnetite is essentially maintained. The amorphous character of SiO 2 and copolymer is represented by the broad diffraction signal averages at µm 2 µm Fig.1. TEM images of a) Fe 3 O 4 /SiO 2 and b) Fe 3 O 4 /SiO 2 /P(LMA-DVB) composite particles. The analysis of FTIR spectra polymer particles is well known for the determination of surface composition. For Fe 3 O 4 /SiO 2 nanoparticles, the strong bands at 543 cm -1 and 376 cm -1 are due to Fe O stretching vibrations of the magnetic nanoparticles. The strong band at 1059 cm -1 that corresponds to Si O Si bonds indicates the bonding of SiO 2 to Fe 3 O 4. The appearance of the signals at 3400 and 1617 cm -1 changed a bit from those of Fe 3 O 4 particles and can be explained by taking into account the stretching and bending vibrations of Si O H. Compared to this the intense bands due to Si O Si and Fe O bonds in Fe 3 O 4 /SiO 2 particles weakened or almost disappeared in Fe 3 O 4 /SiO 2 /P(LMA-DVB) composite particles. The new peaks that appear in the region cm -1 represent aliphatic CH stretching vibrations of aliphatic chain. The small signal at 1710 cm -1 corresponds to the stretching vibration of ester carbonyl group derived from LMA. The shoulder peak at around 3000 cm -1 may correspond to the aromatic CH stretching vibration of DVB crosslinker. These results confirmed the formation of trilayered magnetic Fe 3 O 4 /SiO 2 /P(LMA-DVB) composite particles. Fig.3. XRD spectra of Fe 3 O 4 /SiO 2 /P(LMA-DVB) composite particles. The efficiency of Fe 3 O 4 /SiO 2 /P(LMA-DVB) composite particles as an adsorbent for organic pollutants is shown in Fig. 4. It is likely that adsorption of pollutants on composite particles is driven by the hydrophobic interaction as the composite particles are coated with hydrophobic P(LMA-DVB) layer. It is evident that the magnitude of adsorption is dependent on the nature as well as molecular size of the pollutants. The adsorption efficiency reached maximum for hydrophobic large molecular weight congo red followed by phenol. Composite particles had minimum loading efficiency for herbicide. It is most likely that the higher water solubility of the pollutant is responsible for poor adsorption efficiency. phenol salicylic acid congo red gramaxo ne gramaxone phenol salicylic acid congo red Fig.2. FTIR spectra of a) Fe 3 O 4, b) Fe 3 O 4 /SiO 2 and c) Fe 3 O 4 /SiO 2 /P(LMA-DVB) composite particles. XRD pattern of polymer modified Fe 3 O 4 /SiO 2 composite particles shown in Fig. 3 exhibits six Fig. 4. Adsorption behavior of pollutants on magnetic Fe 3 O 4 /SiO 2 /P(LMA-DVB) com-posite particles measured at room temperature. IV. CONCLUSION Magnetic hydrophobic polymer-coated Fe 3 O 4 /SiO 2 /P(LMA-DVB) nano-composite particles were prepared via suspension polymerization in mixed 217

219 monomer-solvent (hexadecane-toluene) droplets containing Fe 3 O 4 /SiO 2 nanoparticles. The composite polymer particles had a three-layered structureas confirmed by FTIR and XRD analyses. The magnetic composite particles had a nano-dimension in the dried state as observed from the electron micrograph. The removal of different organic pollutants by magnetic composite particles is driven by hydrophobic interaction. REFERENCES [1] Y. Ding, Y. Zhao, X.Tao, Y.-Z. Zheng, J.-F. Chen, Assembled alginate/chitosan microshells for removal of organic pollutants, Polymer, vol. 50, pp , [2] F. Aloulou, S. Boufi, J. Labidi, Modified cellulose fibres for adsorption of organic compound in aqueous solution, Sep. Purif. Technol, vol. 52, pp , [3] R. Aravindhan, J.R. Rao, B.U. Nair, J. Environ. Application of a chemically modified green macro alga as a biosorbent for phenol removal, J. Environ. Manag., vol. 90, pp , [4] G. Bayramoglu, M.Y. Arica, Enzymatic removal of phenol and p-chlorophenol in enzyme reactor: Horseradish peroxidase immobilized on magnetic beads, J. Hazard. Mater. vol. 156, pp , [5] H. Ahmad, M. Nurunnabi, M.M. Rahman, K. Kumar, K. Tauer, H.Minami, M.A. Gafur, Magnetically doped multi stimuli-responsive hydrogel microspheres with IPN structure and application in dye removal, Colloids Surf. A: Physicochem. Eng. Asp., vol. 459, pp , [6] O. Kammona, E. Dini, C. Kiparissides, R. Allabashi, Synthesis of polymeric microparticles for water purification, Microporous and Mesoporous Mater., vol. 110, pp , [7] W.J. Xu, X.L. Zhu, Z.P. Cheng, J.Y. Atom transfer radical poly- merization of lauryl methacrylate, Appl. Polym. Sci. vol. 90, pp , [8] Y.Y. Xu, H. Becker, J.Y. Yuan, M. Burkhardt, Y. Zhang, A. Walther, S. Bolisetty, M. Ballauff, A.H.E. Muller, Double-grafted cylindrical brushes: synthesis and characterization of poly(lauryl methacrylate) brushes, Macromol. Chem. Phys., vol. 208, pp , [9] X. Zheng, Q. Wang, Y. Jiang, J. Gao, Biomimetic synthesis of magnetic composite particles for laccase immobilization, Ind. Eng. Chem. vol. 51, pp ,

220 Statistical Methods for Functional Analysis of Metagenomes Zobaer Akond Agricultural Statistics and ICT Division Bangladesh Agricultural Research Institute (BARI) Abstract Metagenomics is one of the fastest advancing fields of Bioinformatics discipline. It enables to understand the genetic profile of microbial and viral communities. The high-throughput and exponentially increasing amount of data generated from DNA sequencing of microbes necessitates the application of suitable statistical and computational tools to be able to identify the key differences in the functional roles and taxonomic makeup between microbial communities. The application of three different statistical tools studied the metabolic functions of 212 microbial metagenomes within and between 10 environments. Random Forests provided a robust and valuable description of both the classification of metagenomes and the metabolic processes that were important in separating microbial communities from different environments. Analyses showed that the presence of phage genes within the microbial community was a predictor of whether the microbial community was host-associated or free-living. The dominant metabolism that separated the aquatic samples was photosynthesis. Keywords: environment, functional role; linear discriminant analysis; metagenomes; multiple dimensional scaling; random forest I. INTRODUCTION Most life on this planet is microbial that help maintain to a large extent a balanced ecological place through their direct and indirect interactions with biotic and abiotic components of our environment. Metagenomics has recently started contributing to reveal an impressive biodiversity of this microbial life. It is however the technique of extracting and sequencing the DNA and RNA (metatranscriptomics) of microbial communities collected directly from any kind of samples e.g., human or plant/animalassociated, environmental, industrial, food sources and then using high performance computational and statistical analysis to associate function to each sequence[13]. Due to rapid advancement in IT as well as the continued and dynamic development of faster next-generation sequencing (NGS) technologies with various platforms, it has provided a powerful way to sequence multiple samples with tens or hundreds of millions of short DNA fragments or reads in a single run to study the multiple microorganisms lived in environmental communities without the need of isolating and culturing individual microbial species in a laboratory. More than 99% of millions microbial species on Earth cannot be cultured in a laboratory [1]. Annotation of a metagenome is conducted by comparing the sample DNA to that available in Md. Nurul Haque Mollah Bioinformatics Lab, Department of Statistics University of Rajshahi various databases, such as NCBI, SEED, MG-RAST, or COG [2, 3]. The number of sequences similar to each protein is identified; therefore a metagenome provides information on the taxonomic makeup and metabolic potential of a microbial community. Most of the focus in metagenomics has been on single environments such as coral atolls [4, 5], cow intestine [6], ocean water [7], and microbialites[8]. Early work compared extremely different environments, like soil microbes compared to water microbes [8]. More recently, the Human Microbiome Project has expanded our understanding of the microbes inhabiting our own bodies, comparing samples from the same site among and between individuals [11]. These studies reflect the dynamic and expanding field of metagenomics which has been reviewed elsewhere [12]. Metagenomics provides a complete analysis of the microbial activity in terms of how the microbial community taxa or metabolic potential vary between sampling locations or time points. In this paper statistical analysis using a large sample has been done to describe the abilities of metagenomes as well as to explain the metabolic profile of microbial communities that involves the analysis and visualization of large amounts of multivariate data. II. DATA RESOURCES To perform the statistical analysis, 212 metagenomes were selected from the set of publicly available data 1. They were classified into 10 different environments depending on the description provided by the researcher that collected the samples. The metagenomes spanned a range of sequencing technologies, and most environments were represented by two or more sequencing technologies [13]. III. METAGENOMIC FUNCTIONAL VARIABLES As more environmental measurements are collected at the time of metagenome sampling, the two data types: environmental and genomic can be analyzed simultaneously to provide direct evidence of how microbial communities differ across environmental gradients. Therefore the analyses used the percent of sequences in each metabolic or functional group as the data. The metabolic group is the response variable and the metagenomes as the observations. The 27 functional hierarchies used in the analysis were: Amino Acids and Derivatives; Carbohydrates; Cell Division and Cell Cycle; 219

221 Cell Wall and Capsule; Cofactors, Vitamins, Prosthetic Groups, and Pigments; DNA Metabolism; Dormancy and Sporulation; Fatty Acids, Lipids, and Isoprenoids; Membrane Transport; Metabolism of Aromatic Compounds; Miscellaneous; Motility and Chemotaxis; Nitrogen Metabolism; Nucleosides and Nucleotides; Phages, Prophages, and Transposable Elements; Phosphorus Metabolism; Photosynthesis; Plasmids; Potassium Metabolism; Protein Metabolism; Regulation and Cell Signaling; Respiration; RNA Metabolism; Secondary Metabolism; Stress Response; Sulfur Metabolism; Virulence[14] III. STATISTICAL AND GRAPHICAL METHODS The data however consisted of 10 classifications (the environments), 27 response variables (the functional metabolic groups), and 212 observations (the metagenomes). There are many statistical tools that can be used to explore multivariate data as provided by metagenomes. Here an attempt has been made using three different widely used multivariate statistical techniques: random forests (RF), multidimensional scaling (MDS) and linear discriminant analysis (LDA) to show how metagenomes vary between and within environments and identify the metabolic processes that are important in driving the separations. B. Multidimensional Scaling Multidimensional scaling (MDS) is a visualization technique that directly scales objects based on either similarity or dissimilarity matrices [16]. MDS takes for its input an n n dissimilarity matrix S for n metagenomes, constructed by some other statistical technique, such as random forest. C. Linear Discriminant Analysis Linear discriminant analysis is a robust supervised statistical technique that aims to separate the data into groups based on hyper planes and describe the differences between groups by a linear classification criterion that identifies decision boundaries between groups. IV. STATISTICAL SOFTWARE The statistical and graphical methods discussed here are implemented using open source Statistical Language Programming R Fig. 1. Variable importance determined by random forest analysis using mean decrease in Accuracy and Gini A. Random Forests The random forest [15] is a robust analysis tool and is typically used to classify the data either in supervised or unsupervised manner. It is a rapid classification technique that is less susceptible to over-fitting data and can be run in a bootstrap fashion [13]. In addition, the random forest provides a measure of the importance of each variable that can be used in other analyses. V. RESULTS AND DISCUSSIONS RF generates a measure of the importance of each variable calculated by either the mean decrease in accuracy or the mean decrease in the Gini.These two values indicate which variables contributed the most to generating strong trees and can be used in MDS and LDA analyses. A subset of the data and variables is used to generate the trees, and thus the approach can predict the environment to which a 220

222 metagenome belongs. For both Accuracy and Gini in Fig.1, the photosynthesis got highest position with score and respectively as well as the phage groups with second highest score and for both procedures were the most important response variables in separating the data sets, and in the both cases a break occurred between these two variables and the remaining variables, suggesting that just these two measures could be used to grossly classify the metagenomes. The next break occurred after the eighth variable. These eight variables were thus chosen for the following MDS and LDA analyses. MDS projects the proximity measures of the metagenomes as determined by RF to a lowerdimensional space (e.g., 2-dimensional space for plotting on xy-axis). For the RF, the similarity was measured as the number of times two metagenomes appeared on the same leaf in the trees (proximity), and is represented by the distance between two samples on the MDS plot. The MDS plots have been shown in Fig. 2 with the 10 predefined environments. In this analysis, the visualization highlights the separation of the microbes from human/animal hosts from other samples along the first dimension and the separation of the aquatic and mat communities along the second dimension. In Fig. 3, the LDA overall 27 metabolic variables separated the data and showed that the human and terrestrial animal associated metagenomes separated from a cluster consisting of all of the aquatic samples except the hypersaline community. The mat samples separated distinctly from the other cluster. Fig. 2. Multiple dimensional scale plots of the distances calculated from unsupervised random forest. The distances are the number of times the samples appear on the same leaf of the tree, and the MDS has scaled them so that they plot projects those distances into two dimensions. Plotted by the original environments the sample came from. 221

223 mmm: microbial mat community hh: human habitat ah: animal habitat pz: pelazic zone hw: hypersaline water cw: coastal water fb: fresh water biome az: aphotic zone hs: hot spring cr: coral reef Microbial mat community Human & Animal associated sample Aquatic samples Fig. 3. Linear discriminant analysis showing the position of the metagenomes in two-dimensional space from the 10 environments VI. CONCLUSION The analyses separated the microbial samples into three broad groups: the human and animal associated samples, the microbial mats, and the aquatic samples. The RF technique showed that phage activity is a major separator of host-associated microbial communities and free-living, suggesting that the phages are playing different ecological roles within each environment. The MDS and LDA techniques suggest that mat communities separated from both the animal associated metagenomes and the aquatic samples by the vitamin and cofactor metabolism, suggesting a role for secondary metabolism associated with growth in extreme environments. The dominant metabolism that separated the aquatic samples was photosynthesis. Metabolic variables TABLE 1. VARIABLE IMPORTANCE MEASURE WITH CORRESPONDING SCORE Mean decrease in Accuracy Mean decrease in Gini Metabolic variables Mean decrease in Accuracy Mean decrease in Gini Amino Acids and Derivatives Phages, Prophages, and Transposable Elements Carbohydrates Phosphorus Metabolism Cell Division and Cell Cycle Photosynthesis Cell Wall and Capsule Plasmids Cofactors, Vitamins, Prosthetic Groups, and Pigments Potassium Metabolism DNA Metabolism Protein Metabolism Dormancy and Sporulation Regulation and Cell Signaling Fatty Acids, Lipids, and Isoprenoids Respiration Membrane Transport RNA Metabolism Metabolism of Aromatic Compounds Secondary Metabolism Motility and Chemotaxis Stress Response Nitrogen Metabolism Sulfur Metabolism Nucleosides and Nucleotides Virulence

224 REFERENCES [1] D. H. Huson, A. F. Auch, J. Qi, and S. C. Schuster, MEGAN analysis of metagenomic data, Genome Res,vol. 17, pp , [2] R. K. Aziz, D. Bartels,A.A.Best, M. Dejongh, T. Disz, R.A.Edwards, et al., The RAST Server: rapid annotations using subsystem technology, BMC Genomics pp.9-75, doi: / [3] J. C. Wooley, A. Godzik and I. Friedberg, A primer on metagenomics, PLoS Comput. Biol. 6:e ,doi: /journal.pcbi [4] L. Wegley, R. A. Edwards, B. Rodriguez-Brito, H. Liu, and F. Rohwer, Metagenomic analysis of the microbial community associated with the coral Porites astreoides, Environ. Microbiol., vol. 9, pp ,2007. [5] E.A. Dinsdale, et al., Functional metagenomic profiling of nine biomes, Nature, vol. 452, pp. U628-U629, 2008a [6] J. M. Brulc, et. Gene-centric metagenomics of the fiber adherent bovine rumen Microbiome reveals fro age specific glycoside hydrolases, Proc. Natl. Sci.,U.S.A.,vol. 106, pp ,2009. [7] F. Angly, et al., The marine viromes of four oceanic regions, PLoS Biol.4:e368,doi: /journal.pbio , [7] M. Breibart et al., Metagenomic and stable isotopic analyses of modern freshwater microbialities in Cuatro CiEnegas, Mexico, Eviron. Microbiol. Vol.11 pp.16-34, [9] S. G. Tringe, C. Von Mering, A. Kobayashi, A. A. Salamov, K. Chen, H. W. Chang, et al., Comparative metagenomics of microbial communities, Science, vol. 308, pp , [10] K. Kurokawa, T. Ttoh, T. Kuwahara, K. Oshima, H. Toh, A. Toyoda, et al., Comparative metagenomics revealed commonly enriched gene sets in human gut microbiomes, DAN Res., vol. 14, pp , [11] P. J. Turnbaugh, et al., The human microbiome project, Nature, vol. 449, pp , [12] J.C.Wooley, A.Godzik,and I.Friedberg, A primer on metagenomics, PLoS Comput. Biol., 6:e doi: /journal.pcbi ,2010. [13] E. A. Dinsdale, et al., Multivariate analysis of functional metagenomes, Frontiers in Genetics,vol.4,pp.1-24,2013. [14] R. K. Aziz, et al., The RAST server: rapid annotation using subsystems technology, BMC Genomics, vol.9, pp.75, [15] L. Brieiman, Random forests, Mach. Learn., vol.45,pp.5-32,2001. [16] G. P. Quinn, and M. J. Keough, Experimental Design and Data Analysis for Biologists, Cambridge University Press,

225 Simulation of Microalgae and CO 2 flow dynamics in a Tubular photobioreactor and consequent effects on Microalgae growth Saumen Barua Department of Mathematics Sir Ashutosh Govt. College Chittagong, Bangladesh baruasaumen@yahoo.com Mohammad Morshed Alam Department of Mathematics Chittagong College Chittagong, Bangladesh morshed.math@gmail.com Ujjwal Kumar Deb Department of mathematics CUET Chittagong, Bangladesh ukdeb03@hotmail.com Abstract In this study, a two phase flow for CO 2 and Microalgae suspension is considered to understand fluid dynamics phenomena after injecting CO 2 gas inside a tubular Photobioreactor (PBR).The growth rate of the microalgae cell is taken as a function of available sun light at Chittagong University of Engineering & Technology (CUET). The tubular PBR is considered in our study have the radius of 0.025m while the entire length is 20.94m. To observe the growth of microalgae cell we selected the 21 st June for a bright sunny and the longest day of a year. From the simulation after day seven we observed a very slow growth for the microalgae culture and the growth related to concentration of microalgae is increased by day length with respect to continuous sunlight. A small fluctuation of shear rate around U-loop area is also found in our simulation which may be caused to reduce the volumetric production due to cell fragility. Keywords Microalgae, Biofuel, Tubular Reactor, CFD, Simulation. I. INTRODUCTION Continued uses of fossil fuel as an energy source has been unsustainable because of rapid depletion of fossil fuel reserves. Also global climate change and environmental degradation have engaged scientists, researchers and other concerned to find alternate energy sources. Biofuel as a renewable energy is widely considered to be most sustainable alternatives to fossil fuel and a feasible means for environmental and economic sustainability [1]. At Bangladesh where fuel such as oil, gas, and coil is too expensive day by day and to ensure a degradation free environment, alternative source of fuel (Biodiesel) is the time demanding decision. Biofuel can be classified into the 1 st, 2 nd and 3 rd generation biofuels. Those produced from organic matter like starch, sugars, animal fats and vegetables oils are the 1 st generation biofuels. The 2 nd generation biofuels are from cellulose products such as wood, straw, tall perennial grasses or wastes from the wood processing industry. Using Hydrogen as the primary source of energy, the 3 rd generation biofuels are microorganisms (yeast, fungi) biofuels and algae-based fuels like vegetable oils, bio-oil, jetfuels, biohydrogen, biodiesel, renewable diesel and many others. Now Microalgae is the main raw material from which such biofuels can be produced at high efficiency levels and at low investment. The 1 st and 2 nd generation biofuel have several drawbacks. In 1 st generation biofuel potatoes, sugar cane, soybean and rapeseed are used as raw material, shows that if too much fuel is produced from these may increase food price drastically. On the other hand 2 nd generation biofuel still not popular due to the high cost of production [2, 3]. Microalgae are photosynthetic microorganisms that convert sunlight, water and carbon dioxide to algal biomass [4]. Microalgae can be grown with minimal inputs: land, sunlight, water, some macro- and micronutrients and carbon dioxide (CO 2 ). The land need not be fertile, productive land; the ability to grow algae in wasteland regions means that the technology does not compete directly with food cropping. Similarly, low quality water is also applicable. The commercial-scale production of algal biofuels is a major challenge. Most of the currently used harvesting techniques have several drawbacks, such as high cost, non-feasibility of scale-up or flocculants toxicity, which impact the cost and quality of products. Substantial amounts of research and development initiatives are needed to develop a cost and energy-effective process for the dewatering of algae since harvesting cost may itself contribute up to one-third of the biomass production cost [5]. Microalgae can be grown in suspension or attached on solid surface. Each system has its own advantages and disadvantages. Currently, suspend-based open ponds and enclosed photobioreactors are commonly used for algal-biofuel production. In general, an open pond is simply a series of outdoor raceways, while a photobioreactor is sophisticated reactor design that can be placed indoors (greenhouse) or outdoors. Tubular Photo-bioreactors (PBR) are widely known as the most efficient choice compared with other closed methods including annular, flat plate, spiral, helical, torus, stirred tank, vertical column, plastic bags etc. of outdoor microalgae cultivation because of its wide illumination area for light penetration inside the culture, fairly good biomass productivity and relatively cheaper maintenance cost [6]. The main optimal factors for microalgae cultivations are light, CO 2, temperature and ph. Among the four prime factors, CO 2 is the main factors for microalgae production because CO 2 is the main carbon source for photosynthetic culture of microalgae [7]. During the cultivation time CO 2 is injected in the tubular PBR. But CO 2 creates bubbles inside the PBR which 224

226 impacts flow patterns and ultimately microalgae production is affected, which is not negligible in the case of production processes. In this study, a two phase flow, which means to understand fluid dynamics phenomena after injecting CO 2 gas in the tubular PBR together with microalgae suspension, is investigated. The consequent effects on the cell concentration due to the sunlight at CUET and some fluid dynamics phenomena including Share rate, Pressure and velocity profile are also investigated. II. MATHEMATICAL MODEL In this study an airlift driven Horizontal 3-Loop Tubular Photo bioreactor is considered. A uniform mixture of microalgae suspension and CO 2 were injected inside the Tubular photobioreactor. This mixture is considered as an incompressible two phase Newtonian fluid and the flow problem is assumed to be laminar in this simulation. II. A COMPUTATIONAL DOMAIN AND MESH DESIGN The photobioreactor considered in our study is showed in the Figure-1 with radius of m and length of 20.94m. The surface area and volume of the photobioreactor are 3.279m 2 and m 3 respectively. A coarse mesh design is considered for our simulation with 1, 25,691 elements and 10, 55,747 degrees of freedom. where u denotes the velocity of the mixture, ρ and η are its density and viscosity respectively, p is the pressure, g is the gravity, I is the identity matrix and F st is the surface tension force. The separation of the two-phase flow is described by the Cahn-Hilliard advection-diffusion equation [8]: u.. f (3) 2 t e p where is the dimensionless phase field variable, e p is a parameter controlling interface thickness, γ is the mobility, λ he mixing energy density. The function f is given by following equations: 2 e 2 2 p f. ep ( 1) 3e p 8 (4) (5) 2 ep (6) where the term denotes the phi-derivative of external free energy, ζ is the surface tension coefficient and χ is the mobility tuning parameter. The density and viscosity of the mixture are functions of volume fraction of microalgae suspension V l. The volume fraction of microalgae suspension is V ( 1) / 2 and the volume fraction of CO 2 gas l isv ( 1) / 2. For the two phase flow model, the g density and viscosity are defined to vary smoothly over the interface according to ( ) V (7) ( ) V g g l l g g l l (8) Figure 1: Computational domain and Mesh design for Horizontal 3- Loop Tubular Photobioreactor II.B GOVERNING EQUATIONS Since the two phase flow for microalgae suspension and CO 2 mixture is considered as an incompressible Newtonian fluid and is laminar, the governing equations are the continuity equation and the Navier-Stokes equations are as follows: u 0 u T ( u ) u [ p I ( t) ( u ( u) )] g Fst t (1) (2) In the equations above subscripts l and g are used for the algae suspension and CO 2 gas, respectively. The surface tension force in (2) is defined as F st G (9) where G is the chemical potential (Jm 3 ) given by G f (10) 2 e p The viscosity η l in (8) is given by η l = η o η r (t), where the relative viscosity(η r ) to be a ratio between microalgae suspension viscosity(η l ) and water viscosity(η o). Occurrence of microalgae cell proliferation changes the concentration and subsequently the viscosity of the algal suspension. A microalgae cell is assumed to be a small sphere in our study [9]. Then relative viscosity relating to concentration is determined by Einstein s relative viscosity equation as follows: ( t) 1C( t) r (11) 225

227 where ε is the Einstein s coefficient [10]. Based on the experimental data obtained by Hon-nami and Kunito [11], the cell concentration C (t) in (11) depending on the growth rate μ can be expressed by the following logistic function a C( t) C0 (12) t 1 be where C 0 is the initial concentration of the suspension and a and b are constant. As availability of light is an important limiting factor for biofuel production so we consider the specific growth rate of microalgae depends on average light irradiance according to E. Molina s study [7] which is given by the following equation: max Iav (13) Ik Iav where I k is a constant depending upon microalgae culture condition and μ max is the maximum growth rate of microalgae. If we ignore the dynamical and physiological properties of algae cell, the average irradiance (I av ) depends mainly on incident irradiance (I 0 ) available on the surface of the photobioreactor and is given by I0 ( DK ac0 ) Iav [1 e ] (14) DK C a 0 where K a is the extinction coefficient of the biomass, d D, d is the diameter of the photobioreactor cos tube and θ is the angle of incidence of direct radiation depending on a function of five parameters including the declination(δ), solar hour (sh), geographic latitude ( ), surface slope (β), and surface azimuth angle (η ), and the hour angle (ω)[12]. cos sin sin cos sin cos sin cos cos cos cos cos cos cos sin cos cos cos sin sin sin (15) morning hours and positive for afternoon hours. Thus an hour angle ω can be determined by 15( sh12) (18) In our simulation the geographical location for the Tubular photobioreactor is Chittagong University of Engineering and Technology, Chittagong, Bangladesh where the value for the geographical latitude is II.C BOUNDARY AND INITIAL CONDITIONS In this simulation the inlet initial velocity for CO 2 is 0.5ms -1 and the fluid flow is uniform. The volume frictions of CO 2 and microalgae suspension are 0.05 and Also we considered no-slip boundary condition on the wall i.e. u 0 and zero normal stress at the outlet of the domain which are given by T [ pi ( t)( u ( u) )] n o III NUMERICAL RESULTS In our study, COMSOL Multiphysics has been used to run simulation. The simulation is carried out on the seventh day of culture. The model parameter Maximum growth rate( μ max ) = s -1, Constant( I k )=114.67μmolm -2 s -1, Incident Irradiance( I 0 )= 1630 μmolm -2 s -1, Einstein coefficient(ε) = 2500m 3 kg -1, Initial concentration(c o )=0.55kgm -3,Constant(a)=1 Constant(b)=200, Extinction coefficient( K a )= 36.9 m 2 kg -1, CO 2 viscosity( η g)= Pa s, Water viscosity( η 0)=, Pa s, CO 2 density( ρ c, )= kgm -3,Microalgae density( ρ m )= 1020 kgm -3, Day of the year( N)=172, Phi-derivative of external free energy( f )=0.01, According to Grima et. al. s study [7], we found that the horizontally placed tube absorbs higher irradiance with respect to change in solar hour. Thus, the surface slope β is set to zero degree [18], which provides the following simplest form of (15), i.e, cos sin sin cos cos cos (16) where the declination δ is defined by sin[ (284 N)] 365 (17) where N is the day of the year [12]. To calculate an hour angle ω, we follow the concept of Duffie and Beckman [12]. They considered that the angular displacement is 15 degree per hour for earth rotation from east to west, and the value is negative for Figure-2: Velocity magnitude of the three cross section of the second U-loop area. 226

228 Figure-2 represent the velocity magnitude of the twophase flow along three cross-sections the second U- loop of the tubular photobioreactor and at which are the beginning (S1), the middle (S2) and the end (S3) respectively. The results show that the velocity magnitude is generally high at the middle of the tube. Comparing the magnitude of the velocity on the three planes, it is found that there is no significant different in the velocity magnitude. The highest flow speed at the middle plane (S2) of the second U-loop is higher. It is ms 1 whereas ms 1 in the beginning (S1) and end (S3) of the U loop. It indicates that mass transfer is increased due to bubble effect of CO 2 in the current simulation. Figure-3: Shear rate distribution along the entire computational domain. We know that fluid movement receives the shear stress in the domain boundary. So we gave extra attention to the shear rate distribution for the straight and U-loop portion of the domain. Figure-3 demonstrate the shear rate is uniform in the straight portion and it fluctuates positively in the curved (Uloop) portion. Since the shear stress is higher in the U loop area so than straight part so it is one of the reason for cell damage and ultimate microalgae production. Figure-4: Pressure profile the entire computational domain from inlet to outlet. In figure-4 we observed a uniform pressure drop from inlet to outlet in the entire domain and which means higher concentration provides lower pressure profile inside the photobioreactor. Figure-5: The cell concentration of microalgae culture with time A graph of cell concentration against time is represented in figure-5. The cell concentration of microalgae culture on the seventh day from morning (06:00) to the evening (18:00) increases about kg/m 3 which is very slow. We can interpret from this result that the growth related to concentration of microalgae is not fixed but increased with day length with respect to continuous light. IV CONCLUSION In our study, a two phase flow model for microalgae suspension and CO 2 is developed to understand the flow dynamics inside a photobioreactor. A general parabolic shape of velocity profile is observed at different cross-sections inside the photobioreactor. A very slow growth of microalgae is found due to light irradiance after the day seven of our simulation. In case of shear rate distributions we found irregular shape around U-loop area. ACKNOWLEDGMENT The authors gratefully acknowledge for technical supports to the Centre of Excellence in Mathematics, Mahidol University, Bangkok 10400, Thailand. REFERENCES [1] Feroz Alam, Saleh Mobin, Harun Chowdhury, Third generation biofuel from Algae 6th BSME International Conference on Thermal Engineering (ICTE 2014) [2] Magdalena Frac, Stefania Jezierska-Tys2 and Jerzy Tys, Microalgae for biofuels production and environmental applications : A review African Journal of Biotechnology Vol. 9 (54), pp , 27 December, 2010 [3] Marc Veillette, Mostafa Chamoumi, Josiane Nikiema, Nathalie Faucheux and Michèle Heitz, Production of Biodiesel from Microalgae Chemical Engineering and Biotechnological Engineering Department, Université de Sherbrooke,Canada [4] Yusuf Chisti, Biodiesel from microalgae beats bioethanol [5] Manjinder Singh, Rekha Shukla, and Keshav Das Harvesting of Microalgal Biomass Biorefining and Carbon Cycling Program, College of Engineering,The University of Georgia Athens, Georgia [6] Deb, U.K. Chayantrakom, K. Lenbury Y. Comparison of Single-phase and Two-phase Flow 227

229 Dynamics in the HLTP for Microalgae Culture, International Journal of Mathematics and Computers in Simulation. Issue 5, Vol 6, 2012, pp [7] E. Molina Grima, F.G.A. Fernandez, F. Garcia Camacho, ans Y. Christ, photo-bioreactors: light regime, mass transfer and scaleup, Journal of biotechnology; vol. 70, pp , Nov [8] J.W. Cahn, J.E. Hilliard, Free energy of a non-uniform system. I. Interfacial energy, Journal of Chemical Physics 28(2) (1958), pp [9] X. Wu, J. C. Merchuk, Simulation of algae growth in a Bench scale internal loop airlift reactor Chemical Engineering Science, 59 (2004), pp M. [10] Young,The Technical Writer's Handbook.Mill Valley, University Science, [11] A. Einstein, Ann. de. phys., 19, (1906), pp [12] K. Hon-nami, S. Kunito, Microalgae cultivation in a tubular bioreactor and utilization of their cells, China Journal of Oceanology and Limnology, 16 (1998), pp [13] A. Duffie, W. A. Beckman, Solar Engineering of Thermal Process, John Wiley and Sons, Newyork, (1980). 228

230 Evaluation of PCA in Spatial, Frequency and Wavelet Domains for Face Recognition Samsi Ara Department of Applied Physics and Electronic Engineering University of Rajshahi Rajshahi 6205, Bangladesh Abstract Face recognition is one of the challenging matter because it s the most important and complex multidimensional structure of human body and need best techniques for recognition. In this paper, the performance of PCA in spatial, frequency and wavelet domains has been evaluated for face recognition. In this evaluation process the ORL face database has been used which contains 400 face images of 40 persons, i.e, 10 for each person. PCA is a linear statistical approach of dimensionality reduction which transforms data in the high dimensional space to low dimensional space. In this dimension reduction process the PCA retrain the most of the useful information while reducing the noise and other undesirable artifacts. In frequency domain PCA (FPCA), the Fourier magnitudes of the intensity values of pixels of the images are used for feature extraction. In wavelet domain PCA, the 2-D DWT decompose an image into subbands of lower resolution which in turn reduce the computational complexity of estimation of eigenfaces. The recognition accuracy of spatial, frequency and wavelet domain PCAs are 87.5%, 96.5% and 92% respectively which clear that the recognition accuracy of FPCA always outperforms spatial and wavelet domain PCAs. Keywords Face recognition, PCA, Wavelet transform, ORL face database. I. INTRODUCTION Over the last decays face recognition has become a popular area of research in computer vision and one of the most successful applications of image analysis, processing and understanding. The main target of human face recognition systems is to determine as to which face image belongs to which person. It also an interesting computer vision problem with many commercial and law enforcement applications such as portal control, bank account, passport, credit card, ATM etc. If an effective face recognition system can be implemented user verification and user access control, crowd surveillance, enhanced human computer interaction all become possible. The applications of automatic face recognition systems usually work in controlled environments and recognition algorithms can take advantage of the environmental constraints to obtain high recognition accuracy. However, next generation face recognition systems are going to have widespread application in M. Babul Islam Department of Applied Physics and Electronic Engineering University of Rajshahi Rajshahi 6205, Bangladesh babul.apee@ru.ac.bd smart environments where computers and machines are more like helpful assistants. Ideally, the face recognition technique must be invariant to internal and external environment factors that reduce the performance of the recognition systems. Internal characteristics that affect face recognition can be identified as pose, face texture, and expression variations. External characteristics produce unreliable results due to uncontrolled environment under which the image is obtained. The uncontrolled environment factors include varying illumination, occlusion, and similar external factors. Therefore, it has a great importance to develop more sophisticated and effective techniques for face recognition under variety of environmental conditions. There are many face recognition techniques, such as PCA [1], ICA [2], Gabor filter [3], Linear Discriminant Analysis (LDA) [4, 5] etc and their performances are quite impressive. The performance of PCA in spatial domain has been found to be satisfactory. But its implementation in frequency and wavelet domains has hardly been tested. Therefore, in this paper, the PCA techniques for face recognition will be implemented in spatial, frequency and wavelet domains and its performance will be evaluated on ORL face database [6]. II. PRINCIPAL COMPONENT ANALYSIS The Principal Component Analysis (PCA) is one of the most successful techniques that have been used in image recognition and compression. PCA is a statistical method to reduce the large dimensionality of the data space (observed variables) to the smaller intrinsic dimensionality of feature space (independent variables), which are needed to describe the data economically. This is the case where there is a strong correlation between observed variables. A. Spatial Domain PCA In spatial domain PCA, eigenfaces are a set of Eigen vectors used in the computer vision problem of human face recognition. The eigenfaces are principal components of a distribution of faces, or equivalently, the eigenvectors of the covariance matrix of the set of the face images, where an image is considered as a two dimensional space. In face recognition, each 229

231 training image is transformed into a vector by row concatenation. Let I be a 2-D image of size transformed to 1-D image vector as I ( 1,1), I(1,2),, I( n, n) T n n and (1) Now, for M concatenated training images the mean image shown in Fig. 1(a) is computed as 1 M M i 1 i (2) The mean-centered image for each image is i i (3) Now, the covariance matrix of the mean-centered images is given by T C AA (4) where, 2,, A 1 M and the size of C is 2 2 n n. 2 Therefore, calculating n eigenvectors and eigenvalues correspond to C needs a tremendous computational effort. To reduce the computational cost A can be used instead of C as in [1] to obtain M eigenvectors and eigenvalues. Now, the i-th eigenface [Fig. 1(b)] can be obtained as A T T T U V V ( ), i 1,2, M (5) i i i i i, where V are the eigenvectors of C. i It should be noted that the P (< M) significant eigenvectors correspond to P highest eigenvalues instead of M eigenvectors can be used. (a) (b) Fig. 1: (a) Mean face, (b) Eigenface Now, a face image Γ can be projected into the eigenface space as ( ), k 1,2, P (6) k Uk T, In the recognition process, the Euclidean distance given in Eq. (7) of each face of training dataset from test faces used to determine the belonging class of the test face. 2 k k 2 (7) where Ω is a vector representing the test image in the eigenface space. A face is classified as belonging to class k when the minimum k is below c, otherwise the face is classified as unknown face. where c is the largest distance between any two face images given by c max j, k j k ; j, k 1,... M (8) Now, the reconstructed image vector by f is obtained f U k (9) The distance between the original test image and the reconstructed image is. 2 f f (10) Now, the classification of faces is done as follows: If min k 0, then 1, otherwise min k. If c then input image is not even a face image and not recognized. If c and k for all k then input image is a face image but it is recognized as unknown face. If c and k then the input images are the individual face image associated with the class vector. k B. Frequency Domain PCA Conventional PCA technique is affected by intraclass pose variations. In an effort to overcome this problem, the frequency domain PCA is used in [7]. In frequency domain PCA technique, the images are converted from spatial domain to spectral domain using 2D Fourier transform. Now, for a 2-D digital image I (m, n) of size m n, the Fourier transform is given by, I FD M 1 N 1 u v Im, n m0 n 0 um vn 2 j M N, e (11) The magnitude of the Fourier coefficient is 2 2 I u, v Im I u v I FD ( u, v) Re FD FD, (12) Figure 2 shows the Fourier magnitude spectrum of an image. Fig. 2: Fourier magnitude spectrum of an image. Therefore, the Fourier magnitudes of the intensity values of pixels of the images are used to obtain the 230

232 eigenface space following the same process as like in the spatial domain PCA. C. Wavelet Domain PCA In image processing and computer vision research, wavelet transform becomes one of the popular tools because of its ability to capture localized timefrequency information of image [8]. It has been found that the wavelet analysis is able to extract features from face images with distortions caused by the variations in illumination, facial expressions and poses [8]. The 2-D DWT decomposes an image into subbands of lower resolution which in turn reduces the computational complexity of estimation of eigenfaces. In this paper, the Daubechies 4 (db4) wavelet [9] has been used to decompose the face image into 4 subbands, such as, approximate, horizontal, vertical, and diagonal components. Then the approximate components are used in the PCA algorithm to obtain the eigenface space. III. EXPERIMENTAL RESULT In this paper the Olivetti Research Laboratory (ORL) face database [7] has been used. There are 10 different images of 40 persons. All the images are taken against a dark homogeneous background and the subjects are in up-right, frontal position and the size of each image is112 92, 8-bit grey levels as shown in table 1. Table 1: Experimental database Database ORL Face database No. of person 40 No. of face image per 10 person Image size We select 200 samples (5 for each individual) for training. The remaining 200 samples (rest 5 for each person) are used. In this paper, we observed recognition accuracy for spatial domain PCA, frequency domain PCA and wavelet domain PCA by varying the number of principal component as shown in table 2. The optimum result is obtained for the number of principal components 50. Table 2: Effect of no. of principal component on recognition accuracy No. of Recognition Accuracy(%) principal component PCA FPCA WPCA It has been also found in all cases FPCA outperforms PCA and WPCA. IV. CONCLUSION In this paper the performance of PCA has been observed in spatial, frequency and wavelet domains for face recognition using ORL face database. The optimum recognition rate is obtained by using 50 principle components. The highest recognition rate is 96% which is obtained for frequency domain PCA (FPCA). For wavelet domain PCA (WPCA) the recognition rate is 92% which is slightly lower than that of FPCA but much better than that of PCA. Accuracy (%) PCA FPCA WPCA No. of principal component Fig.3: A comparative recognition accuracy among PCA, FPCA and WPCA for 25, 50 and 75 principal components. REFERENCES [1] M. Turk and A. Pentland, Eigenfaces for recognition, Journal of Cognitive Neuroscience, vol. 3, no. 1, pp , [2] M. S. Bartlett, J. R. Movellan, and T. J. Sejnowski, Face recognition by independent component analysis, IEEE Trans. Neural Networks, vol. 13, no. 6, pp , [3] C.J. Liu, H. Wechsler, Gabor feature based classification using the enhanced Fisher linear discriminant model for face recognition, IEEE Transactions on Image Processing, vol. 11, no. 4, pp , [4] P. N. Belhumeur, J. P. Hespanha and D. J. Kriegman, Eigenfaces vs. Fisherfaces: recognition using class specific linear projection, IEEE Trans PAMI, vol. 19, no. 7, pp , [5] Juwei Lu, Kostantinos N. Plataniotis and Anastasios N. Venetsanopoulos, Face Recognition Using LDA-Based Algorithms, 231

233 IEEE Trans. Neural Networks, vol. 14, no. 1, pp. 4 20, [6] F. Samaria and A. Harter, "Parameterisation of a stochastic model for human face identification, 2nd IEEE Workshop on Applications of Computer Vision, Sarasota (Florida), December [7] M. Savvides, B.V.K. Vijaya Kumar and P.K. Khosla, Eigenphases vs. Eigenfaces. Proceedings of IEEE International Conference on Pattern Recognition, pp. III-810 -III-813, UK, [8] W. Zhao et al, Face recognition: a literature survey, Technical Report, University of Maryland, [9] I. Daubechies, The wavelet transform, timefrequency localization and signal analysis, IEEE Trans. Information Theory, vol. 36, no. 5, pp ,

234 Time-Frequency Coherence Analysis of Multichannel Brain Signals Using Synchrosqueezing Transform Md. Sujan Ali 1 and Mst. Jannatul Ferdous 1 Department of Computer Science and Engineering Jatiya Kabi Kazi Nazrul Islam University Mymensingh, Bangladesh 1 msujanali@gmail.com Abstract This paper presents a novel method of implementing time-frequency coherence between electrophysiological signals for brain-computer interfacing (BCI) paradigm. The neural synchronization mostly depends on both time and frequency. The timefrequency coherence is used to measure neural interdependencies. The short-time Fourier transform (STFT) and wavelet transform are generally used to measure the time-frequency coherence. The limitations of these approaches are resolved using synchrosqueezing transform (SST). Due to data adaptability and frequency reassignment properties, the SST produces a well-defined time-frequency representation. Time-frequency coherence between two synthetic signals and real electroencephalography (EEG) data are illustrated based on the STFT and SST. The experimental results show that the SST based timefrequency coherence executes enhanced result than STFT based approach. Keywords brain computer interface; synchrosqueezing transform; time-frequency coherence 1. INTRODUCTION Electroencephalography (EEG) is cost effective and easier way to implement brain computer interface (BCI). It is captured by spatially distributed EEG sensors of the scalp. The connectivity of different parts of brain is an interesting study to the BCI research community. Recently, the measure of coherence between the signals obtained by different sensors is quantified by coherence analysis [1, 2, 3]. It is usually implemented by spectral estimation with Fourier or wavelet transform.the neural data is usually nonstationary in nature and hence it is a great challenge to implement coherence based analysis. The short time Fourier transform (STFT) is considered to solve such problem, it is not entirely resolved due to the following reasons: i) within each short-time period the stationarity of neural data cannot be assured, ii) the resolution of time frequency representation is restricted by Heisenberg uncertainty principle [4]. Although wavelet transform is considered as data adaptive signal analysis method, it uses basis function called mother wavelet for signal decomposition and faces time-frequency resolution problem i.e. lower frequency resolution at high frequencies and higher at low frequencies. Wavelet analysis also depends on the selection of mother wavelet. The arbitrary selection of mother wavelet Md. Ekramul Hamid and Md. Khademul Islam Molla* Department of Computer Science and Engineering University of Rajshahi, Rajshahi, Bangladesh * khademul.cse@ru.ac.bd without matching with the analyzing signal is the cause of erroneous and non-reversible decomposition.the synchrosqueezing transform (SST) [9, 10] is one of the techniques based on the continuous wavelet transform (CWT) that generates highly localized time-frequency representations of nonlinear and nonstationary signals. Synchrosqueezing transform overcomes the limitations of linear projection based time-frequency algorithms, such as the short-time Fourier transforms (STFT) and continuous wavelet transforms. The synchrosqueezing transform reassigns the energies of STFT and CWT, such that the resulting energies of coefficients are concentrated around the instantaneous frequency curves of the modulated oscillations [11]. The frequency reassignment method in time-frequency representation [5, 6, 7] develops the meaningful localization of signal components in time-frequency space [8]. In this paper, the TF representation of EEG signals is implemented by SST and then TF coherence between two synthetic signals. The similar analysis is performed with the short-time Fourier transform in place of SST. After validating of the TF coherence paradigm with synthetic signals, the method is applied to real electroencephalography (EEG). It is clearly observed that in both synthetic and real data the SST based TF coherence performs better than STFT based method. The paper is organized as follows Section 2 discusses time-frequency representation methods including STFT and SST, the coherence in TF domain is explained in section 3, the experimental results are illustrated in section 4 and the section 5 includes some concluding remarks. 2. TIME-FREQUENCY REPRESENTATION Time-frequency representation (TFR) of any signal describes the energy as a function of both time and frequency. It maps a one dimensional signal of time, S(t), into a two dimensional function of time and frequency, T S (t, f). The value of the TFR surface provides idea as to which spectral components are present at what time. The TFR is useful to analyse and synthesize non-stationary or time-varying signals. 2.1 SHORT TIME FOURIER TRANSFORM Short-Time Fourier Transform (STFT) is a timefrequency analysis technique suited to non-stationary 233

235 signals. The STFT provides information about changes in frequency over time. It represents a sort of compromise between the time and frequency of a signal. Also, it gives some information about both when and at what frequencies a signal event occurs. During STFT, the signal is separated into small portions, where these portions of the signal can be assumed to be stationary. For this purpose, a window function w is chosen. The width of this window must be equal to the portion of the signal where its stationarity is valid. The STFT for a non-stationary signal y(t) is defined as ( t, f ) [ y( t). w ( t t' )]. e dt (1) * 2ft Where * is the complex conjugate, w(t) is the window function. The STFT of the signal is the Fourier transform of the signal multiplied by a window function. In STFT, there is a problem in resolution. Narrow window gives good time resolution, but poor frequency resolution. Wide window gives good frequency resolution, but poor time resolution. In addition, wide windows may violate the condition of stationarity. Consequently, choosing a window function is a big challenge. 2.2 SYNCHROSQUEEZING TRANSFORM Synchrosqueezing Transform (SST) is a method functional to the Continuous Wavelet Transform (CWT). The SST is used to make localized TF representations of non-stationary signals. The CWT is capable to create a TF representation of a signal that shows very good TF localization. The CWT algorithm recognizes oscillatory components of a signal through a series of time-frequency filters known as wavelets. To separate a continuous-time function into wavelets the CWT is used. A mother wavelet Φ(t) is a finite oscillatory function which convolved with a signal s(t) in the following form 1 t q Z ( p, q) ( ) s( t) dt (2) 1/ 2 p p Where Z(p, q) is the wavelet coefficients for each scale-time pair ( p, q). A set of bandpass filter produces the wavelet coefficients. The signal property depends on the scale factor p. The presentation of the signal is more complete when the scale factor is low. On the other hand, high scale factor expands the signal and hence its presentation shows less detail.the scale factors can changes the band width of the bandpass filters. Consequently, the energy of the wavelet transform of a signal will be increased and at original frequency ω r it will be spread out around the scale factor, where is the center frequency of a wavelet. Therefore, the original frequency ω r and the estimated frequency in those scales are same. As a result, for each scale-time pair ( p, q) the instantaneous frequency ( p, q) can be estimated as s 1 Z( p, q) s( p, q) iz( p, q) (3) q The TF representation maps the information from the time-scale plane to the time-frequency plane. In the synchrosqueezing operation, every point ( q, p) is converted to ( q, s ( p, q)) [9]. Because p and q are discrete values, we can have a scaling step pk pk1 pk for any pk where s( p, q) is computed. During mapping from the time-scale plane to the time-frequency plane ( q, p) ( q, w ( p, q)), inst the SST (, q) is calculated [11] only at the l centers l of the frequency range [ l / 2, l / 2], with l l1 : ( l, q ) Z( pk, q) p pk (4) p : ( p, q) /2 k s k l 3/2 The equation (4) shows that the TF representation of the signal is synchrosqueezed along the frequency (or scale) axis only [12]. In the SST, the coefficients of the CWT are reallocated to get a concentrated image over the time-frequency plane, from which the instantaneous frequencies are then extracted [13]. 3. TIME-FREQUENCY COHERENCE ANALYSIS The Time-Frequency (TF) coherence is a measure used to observe the linear correlation between two signals or data sets. In brain computer interfacing motor imagery paradigm the synchronization of neural activity has been measured using the TF coherence. Consider two stationary signals x(t) and y(t). For the signals the standard coherence function is defined as [14]: J x C, y ( f ) x, y ( f ) (5) J x, x ( f ) J y, y ( f ) Where is the cross spectrum between signal x and y, and and are the auto spectrums of the signal x and y respectively. The standard coherence function is not enough for EEG like nonstationary signals. Alternatively, a time-frequency extension approach measures the linear correlation between two signals in time-frequency plane [15]. The TF coherence of two non-stationary signals x and y is defined as J x, ( t, f ) C y x, y ( t, f ) (6) J x, x ( t, f ) J y, y ( t, f ) where signal partitioned into T segments and is the discrete frequency. The cross and auto spectrums are calculated as 234

236 Jx, y( t, f ) X ( t, f ). Y ( t, f ) (, ) 2 (, ), (, ) 2 Jx, x t f X t f J y, y t f Y ( t, f ) (7) where X ( t, f ) and Y ( t, f ) are the TF transform coefficients of the signal x and y respectively and Y ( t, f ) is the complex conjugate of Y ( t, f ). To measure time-frequency coherence between two signals based on synchosqueezing transform can be performed using the following algorithm: i) Select two signals or signals from two channels ii) Apply the SST on individual signal/channel to obtain the SST coefficients iii) Compute cross spectrum and auto spectrum based on the SST coefficients iv) Finally compute time-frequency coherence using the cross and auto spectrum using Eq. (6). 4. EXPERIMENTAL RESULTS The TF coherence between two EEG channels is estimated in this work based on the STFT as well as SST and compared. The performance of the proposed SST based time-frequency coherence is evaluated on both synthetic signals and real data. The experimental results show that the SST based TF coherence illustrates enhanced resolution than STFT. Two synthetic signals X=[sin(2πf 1 t), sin(2πf 2 t)], Y=[sin(2πf 1 t), sin(2πf 2 t)] and their TF coherence are shown in Fig 1, where f 1 =5Hz and f 2 =8Hz. There are different time alignment of two sinusoids to generate X and Y. It is observed that the proposed method is more efficient than STFT for localization of frequency components with higher resolution in coherence domain. The SST based method represents a sharper coherence frequency definition at 5Hz and 8Hz than that of STFT. In Fig 1 (c), the coherence between signals X and Y (5 Hz and 8 Hz frequency) is overlapped each other, whereas, in Fig 1 (d), the coherence between the same pair of signals is well separated. The phenomenon is clearly illustrated in Fig 2 which represents the marginal coherences of two methods. The marginal coherence is defined as ~ T 2 Cx, y( f ) t 1 Cx, y( t, f ), for f=1,2,,f. With STFT, the energies in coherence domain of closer frequencies are overlapped and that with SST sharply represents the contribution of individual frequencies. It is observed that the STFT based time-frequency coherence exhibits poor resolution than SST based method. The real electroencephalography (EEG) data collected from the Brain Computer Interface (BCI) Competition IV dataset are also used to evaluate the performance of the proposed method. The data is recorded from healthy subjects. In the whole session motor imagery is performed without feedback. For each subject two classes of motor imagery are selected from left hand, right hand, and foot. The calibration data ds1a of the BCI competition IV are continuous signals of left hand and foot movement. The data contains 59 EEG channels, total 200 trials with four second each. The sampling rate of the data is 100 Hz. As a pre-processing, the data offset has been removed from the EEG signals. Then the signal is passed through a band pass filter of the range between 8Hz and 12Hz to obtain alpha frequency band. Two channels T7 and T8 are chosen to measure the coherence in this experiment. Figure 1: TF coherence analysis (a) synthetic signal X, (b) synthetic signal Y, (c) STFT based TF coherence between and (d) SST based TF coherence between signal X and Y Figure 2: Marginal coherences of STFT based coherence (dashed red line) and of SST based coherence (solid black line) between the synthetic signal X and Y. Figure 3: First row is the raw EEG signal of channel T7 and T8 of left hand movement, second row is the filtered EEG signal and the third row is the spectrum of the filtered component. The raw EEG signal (first row), the filtered alpha component (second row) and the spectrum of the alpha component (third row) of channels T7 and T8 of left hand movement are shown in Fig 3. The STFT based time-frequency coherence between channel T7 and T8 for left hand movement motor imagery is shown in Fig 4(a). The time-frequency coherence based on SST between channel T7 and T8 of left hand movement is represented in Fig 4(b). The energy corresponding to the marginal coherence is illustrated in Fig 5 for left hand data. The marginal coherence based on STFT represents poor 235

237 localization of frequency components, whereas, SST based method illustrates sharp localization of each component within very narrow band of frequencies. In Fig 5, the frequencies of 8.5Hz and 11Hz are well separated with SST based marginal coherence but it is unable to separate those frequencies in STFT based approach. Hence the SST based time-frequency coherence is superior to that of using STFT. The underlying reason is that, the energy in STFT spreads over a wide range of frequency due to the use of window function with overlapping which introduces cross-spectral energy. applied to real EEG signals of different motor imagery. The performance is compared with STFT based measure of TF coherency. It is observed that SST based method is more efficient than STFT for localization of frequency components with higher resolution in coherence domain. ACKNOWLEDGMENTS This research work was supported by the Information and Communication Technology (ICT) division of the ministry of Post, Telecommunication and Information Technology, Bangladesh. REFERENCES Figure 4: (a) STFT based TF coherence and (b) SST based TF coherence between channel T7 and T8 of left hand movement data. Figure 5: Marginal coherences of STFT based coherence (dashed red line) and of SST based coherence (solid black line) between channel T7 and T8 of left hand movement data. 5. CONCLUSION A novel method to analysis the time-frequency (TF) coherence between a pair of signals is introduced in this paper. The cross and auto spectrums are calculated for the given signals in time frequency domain. The TF coherence is estimated with the spectra for synthetic signals with STFT and SST based time-frequency representation. The proposed SST based coherence estimation method is [1] G. G. Gregoriou, S. J. Gotts, H. Zhou, and R. Desimone, High Frequency, long-range coupling between prefrontal and visual cortex during attention, Science, vol. 324, no. 5931, pp , [2] H. Liang, S. L. Bressler, M. Ding, R. Desimone, and P. Fries, Temporal dynamics of attention-modulated neuronal synchronization inmacaque V4, Neurocomputing, vol , pp , [3] A. Brovelli, M. Ding, A. Ledberg, Y. Chen, R. Nakamura, and S. L. Bressler, Beta oscillations in a large-scale sensorimotor cortical network: directional influences revealed by Granger causality, Proceedings of the National Academy of Sciences ofthe United States of America, vol. 101, no. 26, pp , [4] S. Mallat, A Wavelet Tour of Signal Processing, Academic Press,New York, NY, USA, [5] K. Kodera, R. Gendrin, C. Villedary, Analysis of timevarying signals with small BT values, IEEE Trans. Acoust. Speech Signal Process.26 (1), 64 76, [6] F. Auger, P. Flandrin, Improving the readability of timefrequency and time-scale representations by the reassignment method, IEEE Trans. Signal Process.43(5), , [7] F. Auger, P. Flandrin, Y.- T. Lin, S. McLaughlin, S. Meignen, T. Oberlin,H.-T. Wu, Time-frequency reassignment and synchrosqueezing: an overview, IEEE Signal Process. Mag. 30(6), 32 41, [8] R. Carmona, W.-L. Hwang, B. Torrésani, Practical Time- Frequency Analysis, Academic Press, [9] I. Daubechies, J.Lu, H.-T.Wu, Synchrosqueezed wavelet transforms: an empirical mode decomposition-like tool, Appl. Comput. Harmo- nic Anal.30(2), , [10] I. Daubechies, S. Maes, A nonlinear squeezing of the continuous wavelet transform based on auditory nerve models, in: Wavelets in Medicine and Biology, CRC Press, pp , [11] A. Ahrabian, D. Looney, L. Stankovic, D. P. Mandic, Synchrosqueezing-based time-frequency analysis of multivariate data, Signal Processing, 106, , [12] Li, C., and M. Liang, A generalized synchrosqueezing transform for enhancing signal time-frequency representation: Signal Processing, 92, , [13] Wu, H.-T., P. Flandrin, and I. Daubechies, One or two frequencies? The synchrosqueezing answers: Advances in Adaptive Data Analysis, 3, no. 2, 29 39, [14] K. Saranyasoontorn, L. Manuel, P. S. Veers, A comparison of standard coherence models for inflow turbulence with estimates from field measurements, Journal of Solar Energy Engineering, vol. 126, , [15] L. B. White, B. Boashash, Cross spectral analysis of non-stationary processes, IEEE Transactions on Information Theory, vol.36, ,

238 Experimental Study on Optical Characterization of Mono Crystalline Silicon Solar Cell Nusrat Chowdhury 1, Zahid Hasan Mahmood 3 Institute of Energy 1, Department of Electrical and Electronics Engineering 3 University of Dhaka, Bangladesh nusrat_105@yahoo.com 1 Abstract The experimental works have been done on the optical characterization of mono crystalline solar cell by Surface Photo Voltage (SPV) measurement for fabricated solar cell by observing minority carrier diffusion length & life time. The core objective of this research was to reduce the cost of solar cell and increase the efficiency by analysis the optical characterization. A simple computer-controlled, normal incidence measurement system was designed for SPV measurements of minority carrier diffusion length and lifetime of Si-solar cell. Measurement system is based on a mini monochromator driven with a steeper motor to vary wavelengths in nm spectral range. Light induced surface photo voltage is measured as a function of the wavelength. SPV is measured using a Standard Research 510 lock in amplifier. A LabVIEW interface is used for system control and data acquisition. After calculating the experimental data obtained from mono crystalline silicon solar cells measurement, minority carrier diffusion length and life time were calculated. By using solar simulator s (Sun Simulator K3000 LAB55) platform at 25 o C efficiency was measured. Keywords Surface Photo Voltage, minority carrier diffusion length, carrier life time I. INTRODUCTION There is a direct relation between the minority carrier lifetime and solar cell efficiency for evaluating the performance of solar cell. The method of SPV is a contactless system that helps to analysis the optical characterization of semiconductors. It measures the diffusion length of minority carriers in the region of essential light absorption inside solar cells and wafers [1]. The minority carrier diffusion length, L is an important factor for cell efficiency and spectral response of the mono crystalline silicon solar cell. It is also essential for evaluation of the p-type silicon wafer. In this experiment the surface photo voltage (SPV) of fabricated solar cells have been analyzed and observed minority carrier flow by Light Current Voltage (LIV) tester as it represented the quality of fabrication process and efficiency of the solar cell. The minority carrier lifetime and efficiency are very important parameter to explain the quality of solar cell. We observed that diffusion length sometimes small valued. It happened due to recombination. Recombination occurs in many reasons. Doping causes defects and more recombination. In this paper two types of optical characterization have been done one is SPV and another is efficiency. Md. Abdur Rafiq Akand 2 Solar Cell Fabrication & Research Division, Institute of Electronics, Bangladesh Atomic Energy Commission, Savar, Dhaka, Bangladesh In the first part of this paper, we describe an experimental setup for measuring this two analysis and then we discuss about the result of our sampled cell. II. MINORITY CARRIER DIFFUSION LENGTH AND LIFE TIME The minority carrier diffusion length, L is essential for evaluation of the quality and transport properties of the material. In the base region, the diffusion length is a critical factor which impacts the conversion efficiency and spectral response of the cell [2]. The surface photo voltage is produced when some of the minority carriers that drift around in the bulk reach the surface. The numerical distance that carriers travel in the bulk before they recombine is the diffusion length [3]. Since, some of the minority carriers recombine before they reach the surface, therefore, the shorter the diffusion length, the smaller the SPV signal because of the high recombination losses. The diffusion length, L, is approximately related to the minority carrier lifetime by (1) Where, D is the diffusion coefficient. Note that the diffusion length, L, is independent of any built-in fields in contrast to the drift behavior of the electronhole pairs. Another important factor to note is that while the photo-generated majority carriers also diffuse towards the surface, however, their number, as a fraction of the thermally generated majority carrier density in a moderately doped semiconductor, is far too small to create any significant, measurable photo voltage. Both types of photo-generated carriers also diffuse towards the rear surface where their collection can introduce errors in data interpretation particularly when the diffusion lengths are larger than the wafer thickness. In real semiconductors, the measured diffusion length, (2) III. EFFICIENCY MEASUREMENT AND I-V CHARACTERISTIC Solar cells are characterized by their ability to convert sunlight into electricity. This solar cell efficiency measurement by sun simulator system find out 237

239 fundamental device characteristics including short circuit current (Isc), open circuit voltage (Voc), fill factor (FF), and maximum power (Pmax). Those collected results, can be used to determine the efficiency of solar cell. Incident intensity is controllable using absorptive metallic filters. System is independently calibrated with pre-qualified solar cells. Efficiency is the ratio of output to input; in case of solar cell energy conversion we call this value η. This is the percentage of solar energy to which the cell is exposed that is readily being converted to electrical energy. The value of η can be calculated by: Solar Cell Efficiency, η =FF*V OC *I SC /(E*A) (3) Where E is the input light energy in W/m 2, and A is the surface area of the solar cell in m 2. Generally this I-V characteristic would fall under electrical characterization, however since the measurement is usually taken by sun or sun-simulator i.e. optically induced; this is also an optical characterization. IV. EXPERIMENTAL SETUP A. SPV Setup To measure the minority carrier lifetime of Si, a simple, computer-controlled, normal incidence measurement system was designed. This system is based on a mini monochromator driven with a stepper motor to vary wavelengths in ~ nm spectral range. Light-induced surface photovoltage (SPV) is measured as a function of the wavelength. SPV is measured by a Stanford Research 510 lock-in amplifier. This system can be modified to measure SPV as a function of wavelength as well as a function of flux density at fixed wavelength. A LabVIEW interface is used for system control and data acquisition [2]. The fig.1 describes detailed system schematics of the minority carrier diffusion length and lifetime measurement system and fig.2 shows the experimental setup while measuring SPV. Fig. 2. Experimental setup for measuring SPV Prior to data acquisition, settings on the lock-in amplifier are manually set. This is done by looking at the SPV signal at wavelength of ~600 nm(red light). Following lock-in settings are fairly standard for these measurements: (a) chopper frequency in ~ HZ range, (b) time constant ~ ms, (c) all input filters engaged, (d) dynamic resolution off, (e) offset off, and voltage scale is set anywhere between ~ 0.2 mv to 2 mv. B. Efficiency Measurement The sun simulator is essentially a solar cell I-V measurement system that consists of multiple components such as the K201 solar simulator, K101 photovoltaic power meter, K401 solar simulator power supply, K202 auto controller, K901 sample mounting jig and K730 the software that measures the IV characteristics. This system, shown in fig.3, is generally meant for the IV measurement of thin films, however due to the flexible mounting jig measurement of our sample i.e. simple wafer is also possible. For the ideal output a K801 reference cell is used. Fig. 3. Complete setup of the system [5] Fig. 1. Basic block diagram of the SPV measurement For conducting the experimental study, we have choose a reference cell from India of 156 cm 2 monocrystalline silicon solar cell, which has been placed on the jig of solar simulator s platform at 25 o C. Xenon lamp of sun simulator produced light of input power 100 mwcm -2. The calibration has been done with respect to a reference cell before collecting 238

240 Reciprocal of SPV (1/Vspv) International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering the testing data. The vertical position of the reference cell has been adjusted such that the main performance parameters of the cell obtained from the simulator match accurately to its specification sheet [4]. Fig.4 shows the experimental setup while measuring cell efficiency. Fig. 5.The LabVIEW Image 120 Fig. 4. Experimental setup for measuring efficiency In order to measure solar cell performance, the positive and negative probes are connected to an electronically-controlled resistive system. As the voltage across the solar cell is varied, current flow across the load resistor goes from zero to a maximum value (short-circuit). Maximum power is achieved at a point often referred to as the maximum power point (MPP) at which the product P=V*I=V m *I m is at its maximum value. The ratio of V m *I m to V OC *I SC is defined as the solar cell fill factor (FF). FF is a complicated function of many variables including resistance, minority carrier lifetime, surface recombination; it usually in the ~ range, higher FF represents higher efficiency solar cell. Peak solar cell incident solar radiation is considered to be 100 mw/cm 2. This is also referred to as AM 1.5 solar insolation; all terrestrial efficiency measurements refer to AM 1.5. Therefore, for AM 1.5 illumination, on can rewrite equation (3) as follows: Solar Cell Efficiency, (η) =FF*V OC *I SC /(0.1*A) (4) In eq. (4), A is solar cell area is in cm2, V OC is in V, and I SC is in A. V. EXPERIMENTAL RESULTS A. SPV Measurements From the testing of SPV we get the following graphs shown in fig. 5. This graph shows wavelength vs amplitude curve for 400 nm to 1200 nm wavelength. From this figure we can say that from 750 nm to 850 nm we get maximum amplitude Reciprocal of Penetration depth (1/alpha) Fig. 6. 1/alpha v s 1/Vspv represents minority carrier diffusion length This fig. 6 is plotted in MATLAB. This figure is used to find the value of diffusion length that is found by its x intercept and the value is 92 μm. Equation of minority carrier diffusion length: Where this Ln is the diffusion length, D is the diffusivity (for Si D= 27 cm 2 /s), and τ is the lifetime. Minority carrier lifetime becomes B. For Efficiency Measurement By using sun simulator, we have measured the efficiency for different areas of same solar cell and analysis the changes of data for different areas of solar cell which indicate that the whole area of solar cell is almost uniformed. From sun simulator, we have obtained the following results which are presented in the table 1. We draw the I-V curve and P-V curve for test 1 data which shown in fig.7. From the table 1 we observed that the efficiency for different area is almost same and the cell is well fabricated. A LAB VIEW result shows the SPV results in arbitrary unit. From the testing data we plot 1/alpha v s 1/Vspv which represents minority carrier diffusion length. Fig.6 shows 1/alpha v s 1/Vspv curve. 239

241 Test Area cm TABLE I. RESULTS FROM SUN SIMULATOR Voc (V) Isc (ma) Jsc (ma/cm 2 ) Pmax (mw) Vmax (V) Imax (ma) Current (ma) & Power (mw) Fill Factor (%) Current (ma) Power (mw) Voltage (V) Fig. 7. I-V Characteristics Curve of Sample Solar Cell VI. CONCLUSION Efficiency (%) The SPV data was plotted as a function of 1/α. The minority carrier diffusion length for cell surface was determined to be 92 μm. For Si minority carrier diffusion coefficient, D, to be 27 cm2/sec; the effective lifetime for sample cell is 3.135μsec. By using sun simulator, we have measured the efficiency for different areas of same solar cell and get efficiency above 16%. VII. DISCUSSION The main challenge of SPV measurement was to calibrate the system. At first the light emitted from the monochromator must focused on the mirror properly and the second challenge was to set the sensitivity of lock in amplifier so that the peak of the curve doesn t cut and get the full curve in range of selected wavelength In sun simulator, reference cell and it s placement on the jig contact is very important because efficiency depends on the vertical position of the cell on the platform and it could change with the change of vertical position. Although the theoretical and experimental study of the performance parameters are almost to the same but it has been observed that some discrepancies exist between these efficiencies. In this model, neglect the effect of thickness in short circuit current density equation which is principally responsible for the observed discrepancy. This model would be useful to evaluate the silicon solar cell performance parameter if thickness is incorporated with short circuit current density equation. ACKNOWLEDGEMENTS I am obliged to Ms. Nahid Akter for her support throughout the entire work and also thanks to Khairul Bashar and Md. Kamrul Islam for SPV instrument and Mr. Shahriar Bashar for solar simulator. Laboratory Solar Cell Fabrication Laboratory, Institute of Electronics, Atomic Energy Research Establishment, Savar, Dhaka. Solar Energy technology, IFRD, Bangladesh Council for Scientific and Industrial Research (BCSIR), Dhaka. REFERENCES [1] L. Votoček and J. Toušek Surface Photovoltaic Effect and Its Applications to Si Wafers and Monocrystalline Si Solar Cells Diagnostics, WDS'05 Proceedings of Contributed Papers, Part III, , ISBN [2] Md. Abdur Rafiq Akand, Md. Rakibul Hasan, Mohammad Khairul Basher, and Mahbubul Hoq, Study of Surface Photo voltage for monocrystalline silicon solar cell fabricated at BAEC solar cell lab., International Journal of Innovation and Applied Studies ISSN Vol. 7 No. 4 Aug. 2014, pp [3] Suhaila Sepeai, Saeem H. Zaidi, S.L.Cheow, M.Y.Sulaiman, K.Sopian, Evaluation of Surface Photovoltage (SPV) in AlBack Surface Fields Bifacial Solar Cell, Proceedings IEEE PVSC 39, [4] Abu Kowsar, Abdullah Yousuf Imam, Mashudur Rahaman1, Muhammad Shahriar Bashar, Md. Saidul Islam, Sumona Islam, Nowrin Akter Surovi and Zahid Hasan Mahmood, Comparative study on the efficiencies of silicon solar cell, IOSR Journal of Applied Physics (IOSR-JAP) e-issn: Volume 6, Issue 6 Ver. IV (Nov.-Dec. 2014), PP [5] K3000 LAB Solar Cell I-V Measurement System. (n.d.). Retrieved December 20,2014, from mcscience.com/bbs/bbs/board.phpbo_table= en_product_01&wr_id=7. 240

242 Canonical Correlation Analysis for SNP based Genome-Wide Association Studies Atul Chandra Singha Bioinformatics Lab, Department of Statistics, University of Rajshahi, Bangladesh Department of Statistics, Noakhali Science and Technology University (NSTU), Bangladesh Arafat Rahman Department of Microbiology, Noakhali Science and Technology University (NSTU), Bangladesh Jahangir Alom Bioinformatics Lab, Department of Statistics, University of Rajshahi, Bangladesh Md. Nurul Haque Mollah Bioinformatics Lab, Department of Statistics, University of Rajshahi, Bangladesh Abstract-- Genome-wide association studies (GWAS) are powerful tools for measuring the association between genotype phenotype pairs in bioinformatics. Most of the human diseases and traits have a strong genetic architecture. GWAS is successful in identifying common genetic variants underlying complex traits or diseases like cancer, type-ii diabetes, cardiovascular disease, schizophrenia and quantitative traits such as lipid levels and metabolomics. Now an important approach to GWAS is to test the association between multiple single nucleotide polymorphisms (SNPs) against multiple quantitative phenotypes. Canonical Correlation Analysis (CCA) is one of the most popular multivariate statistical techniques to test the association between multiple SNPs against multiple quantitative phenotypes. However, it is not robust against phenotypic contaminations. To overcome this problem, in this paper an attempt is made to robustify the CCA. To robustify the CCA, we consider some popular robust analyzers like Minimum Covariance Determinant (MCD), Minimum Variance Ellipsoid (MVE), Orthogonalized Gnanadesikan-Kettering (OGK) estimators including the Minimum β-divergence estimator. Using simulated data analysis, we observed that CCA based on Minimum β-divergence method (proposed) shows better performance than classical CCA as well as robust CCA based on MCD, MVE and OGK estimators in presence of outliers. Otherwise proposed method keeps equal performance to the classical CCA as well as robust CCA based on MCD, MVE and OGK estimators. Keywords-- SNPs, GWAS, CCA, Quantitative traits, Outliers, Minimum β-divergence method. I. INTRODUCTION Recently, most of the human disease and traits are strongly related to the genetic components. Genomewide association studies (GWAS) are successful in identifying genetic components underlying complex traits or disease like cardiovascular disease [1], schizophrenia [2], type 2 diabetes[3], quantitative traits like lipid levels[4,5] and metabolomics[6,7]. The measure of association between multiple genotype and multiple phenotype provides precise result[8] Therefore, some complex genotype-phenotype correlations can be detected when testing several genetic components simultaneously[9] Recently, a multivariate statistical technique CCA has been successfully applied to GWAS [8, 9, 10,11] for identifying linear relationships between multiple SNPs against multiple phenotypes. CCA, first described by Harold Hotelling (1936), measures the association between two sets of multidimentional variables by maximizing the correlation between their linear combinations. However, it is not robust against phenotypic contaminations. To overcome this problem, in this paper an attempt is made to robustify the CCA. To robustify the CCA, we consider some popular robust analyzers like Minimum Covariance Determinant (MCD), Minimum Variance Ellipsoid (MVE), Orthogonalized Gnanadesikan-Kettering (OGK) estimators including the Minimum β- divergence estimator. II. METHODS Let X and Y denote genotype and phenotype matrices of dimensions N G and N P respectively where N the number of samples, G and P the number of genotypic and phenotypic variables respectively. CCA [12] provides a convenient statistical framework to simultaneously detect linear relationships between NG NP two groups of variables X and YR where X and Y represent two different views of the same objects. The objective is to find maximally correlated linear combinations of columns of each G matrix. This corresponds to finding vectors a R P and br that maximise T (Xa) (Yb) r = = Xa Yb a T a T XX The maximised correlation r is called canonical correlation between X and Y. Finally, for G a XY b b T YY b 241

243 genotypes and P phenotypes the j=min(g,p) canonical correlations are then calculated as the square root of the j eigenvalues of the canonical 1 1 correlation matrix where, XX and YY XX XY YY YX are the G G and P P covariance matrices for genotypes and phenotypes respectively while XY and YX are the between G P (or P G) covariance matrices. III. COVARIANCE MATRIX DETERMINATION USING ROBUST METHOD 3.1 Orthoganalized Gnanadesikan/ Kettenring (OGK) Gnanadesikan and Kettenring (1972) was proposed positive definite and approximately affine equivariant robust scatter matrices starting from any pair wise robust scatter matrix and then applied for robust covariance estimate the resulting of multivariate location and scatter estimates are called OGK. 3.2 Minimum Volume Ellipsoid Estimator (MVE) The Minimum Volume Ellipsoid (MVE) estimator, first proposed by Rousseeuw [13] has been studied extensively and frequently used in the detection of multivariate outliers. It seeks to find the ellipsoid of minimum volume that covers a subset of at least h data points. Subsets of size h are called half sets because h is often chosen to be just more than half of the data points. The location estimator is the geometrical center of the ellipsoid and the estimator of the variance covariance matrix is the matrix defining the ellipsoid itself, [13] have discussed the minimum volume ellipsoid estimator searches for the ellipsoid of minimal volume containing at least half of the points in the data set X. Then the location estimate is defined as the center of this ellipsoid and the covariance estimate is provided by its shape. The search for the approximate solution is made over ellipsoids determined by the covariance matrix of p+1 of the data points and by applying a simple but effective improvement of the sub-sampling procedure as described in [13] The MVE was the first popular high breakdown point estimator of location and scatter but latter it was replaced by the MCD, mainly because of the availability of an efficient algorithm for its computations [13]. Recently the MVE gained importance as initial estimator for S estimation. 3.3 Minimum Covariance Determination Estimators (MCD) Rousseeuw [13] introduced the minimum covariance determinant estimator (MCD) method to estimate the mean vector and covariance matrix along with detection of outliers in multidimensional data. The multivariate location and diffusion estimation in high breakdown principles is based on the determinant of the covariance matrix. If the covariance matrix positive semi-definite matrix, Eigen values are positive, the determinant of covariance matrix equals the product of Eigen values. Thus, a small value in the determinant reflects some linear patterns in the data. Consider all subsets, and compute the determinant of the covariance matrix for each subset. The subset with smallest determinant is used to calculate the usual mean vector, and corresponding covariance matrix, these estimators are called minimum covariance determinant estimators. 3.4 Minimum β- Divergence method (Proposed) To estimate the robust covariance matrix using maximum -likelihood estimator [15] we used the maximum -likelihood estimators for the mean vector μ and the covariance matrix obtained iteratively as follows: and t+ 1 = n j= 1 j t μ t t+ 1 (x /, )(x = j n 1 j= 1 n j= 1 n j= 1 )(x ( 1 + ) (x /, ) j t t (x /, )x j j t (x /, ) j t t t ) Where, T 1 ( x ; μ, ) exp{ ( x μ ) ( x μ )}, be the 2 β-weight function ( Mollah et al., 2007). It produces almost zero weight for contaminated data points. The notations μ t+1 and t 1 are the update of μ t and t in the (t+1)-th iteration respectively. IV. DATA SIMULATION To test the performance of CCA, we simulated SNP data for a hypothetical gene with the coalescent-based simulator GENOME [14]. We assumed a mutation rate of 10-8 per generation per base pair and an effective population size of 10,000. A study population of 2,000 individuals was formed by randomly drawing with replacement 1,000 pairs of chromosomes from a pool of 500 simulated haplotypes. Next, we simulated seven normallydistributed traits under seven models. In model 1, each of the three independent QTL individually explained 0.3% of the variance of trait 1, i.e. the QTL had no impact on the variance of the other four traits. In model 2 to 7, the three QTL contributed to the T t t j 242

244 variance of a progressively larger number of traits: 2, 3, 4, 5, 6 and all 7 traits, respectively. In all seven models, shared sources of variation between traits other than the QTL were also simulated so that the residual phenotypic correlation between pairs of traits was ~0.4. V. SIMULATION RESULTS To investigate the performance of proposed method in comparison to classical CCA method as well as robust methods like MVE, MCD and OGK we apply our simulated dataset. The analysis results (Table-2) shows that the proposed method and all other robust methods including classical method performs are almost same in absence of contamination (Fig. 1) Again, To examine more robustness of the proposed method in comparison to all other robust methods including classical method, we contaminate (5% only) the datasets. Then we apply the classical method and other robust methods including proposed method. The analysis results (Table-1) shows that our proposed method performs better than classical method as well as robust methods (Fig-2). Table-1: Canonical Correlation Analysis (CCA) Correlation Classical pairs MVE r 1 (p.value) r 2 (p.value) r 3 (p.value) r 4 (p.value) r 5 (p.value) r 6 (p.value) r 7 (p.value) 0.34 (0.000) 0.13 (0.568) 0.12 (0.756) 0.11 (0.853) 0.10 (0.951) 0.08 (0.985) (0.979) MCD OGK Proposed (RCCA) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.287) (0.287) (0.000) 0.13 ( (0.568) (0.293) (0.453) (0.653) (0.503) (0.762) (0.832) (0.782) (0.803) (0.872) (0.840) 0.31 (0.000) 0.23 (0.000) 0.19 (0.000) 0.15 (0.039) 0.12 (0.460) (0.876) 0.07 (0.865) VI. CONCLUSSIONS Canonical correlation analysis (CCA) is an efficient and powerful tool for measuring the association between genotypes and phenotypes in GWAS studies. This paper discusses an highly robust CCA approach using minimum β-divergence based covariance matrix. To investigate the performance of the proposed method in comparison to other robust method including classical method, we generated two synthetic datasets (eg.- Contaminated and uncontaminated). The analysis results show that the classical method, MVE, MCD, OGK and proposed method performs equally in absence of contamination. But in presence of contamination the proposed method performs better than the classical method as well as MVE, MCD and OGK method. Finally, we hope that our work helps extending the application area of CCA in the field of both genetics and outside genetics. Table-2: Canonical Correlation Analysis (CCA) Correlatio n pairs r1 (p.value) r2 (p.value) r3 (p.value) r4 (p.value) r5 (p.value) r6 (p.value) Classica l 0.31 (0.000) 0.23 (0.000) 0.18 (0.000) 0.15 (0.005) 0.11 (0.552) 0.08 (0.812) MVE MCD OGK Proposed (RCCA) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.453) (0.653) (0.503) (0.762) (0.832) (0.782) 0.31 (0.000) 0.23 (0.000) 0.20 (0.000) 0.15 (0.009) 0.12 (0.549) (0.802) r7 (p.value) (0.831) (0.803) (0.872) (0.840) 0.08 (0.819) 243

245 Fig. 1: Comparison of canonical correlations by different methods Fig. 2: Comparison of canonical correlations by different methods REFERENCES [1] Deloukas,P. et al. (2013) Large-scale association analysis identi_es new risk loci for coronary artery disease. Nature Genetics, 45, [2] Schizophrenia Working Group of the Psychiatric Genomics Consortium (2014) Biological insights from 108 schizophrenia-associated genetic loci. Nature,511, [3] Mahajan,A. et al. (2014) Genome-wide transancestry meta-analysis provides insight into the genetic architecture of type 2 diabetes susceptibility. Nature Genetics, 46, [4] Global Lipids Genetics Consortium (2013) Discovery and renement of loci associated with lipid levels.nature Genetics, 45, [5] Surakka,I. et al. (2015) The impact of low frequency and rare variants on lipid levels. Nature Genetics,47, [6] Kettunen,J. et al. (2012) Genome-wide association study identi_es multiple loci inuencing human serum metabolite levels. Nature Genetics, 44, [7] Shin,S.Y. et al. (2014) An atlas of genetic inuences on human blood metabolites. Nature Genetics, 46, [8] Inouye,M. et al. (2012) Novel Loci for metabolic networks and multi-tissue expression studies reveal genes for atherosclerosis. PLoS Genetics, 8, e [9] Marttinen,P. et al. (2014) Assessing multivariate gene-metabolome associations with rare variants using Bayesian reduced rank regression. Bioinformatics, 30, [10] Ferreira,M.A., Purcell,S.M. (2009) A multivariate test of association. Bioinformatics, 25, [11] Tang,C.S., Ferreira,M.A. (2012) A gene-based test of association using canonical correlation analysis.bioinformatics, 28, [12] Hotelling,H. (1936) Relations between two sets of variates. Biometrika, 28, [13] Rousseeuw, P.J. (1985): Multivariate estimation with high breakdown point, Mathematical Statistics and Applications, Vol. B (Grossmann et al., eds.), , Reidel, Dordrecht. [14] Liang, L., Zollner, S. and Abecasis, G.R. (2007) GENOME: a rapid coalescent-based whole genome simulator, Bioinformatics, 23, [15] Mollah, M. N. H., Sultana, N., Minami, M., and Eguchi, S. (2010): Robust extraction of local structures by the minimum beta-divergence method. Neural Network, 23:

246 Frequency Recognition of SSVEP for BCI implementation using canonical correlation analysis with adpative reference signal Shalauddin Ahamad Shuza, Md. Rabiul Islam and Md. Kislu Noman Department of Computer Science and Engineering Pabna University of Science and Tech. Pabna, Bangladesh Md. Khademul Islam Molla Department of Computer Science and Engineering University of Rajshahi, Rajshahi, Bangladesh Abstract This paper presents a data adaptive approach to frequency recognition for steady-state visual evoked potential (SSVEP) based brain-computer interface (BCI) implementation. The brain signals corresponding to SSVEP is captured by electroencephalography (EEG). Standard canonical correlation analysis (CCA) is used with adaptive reference signals to recognize the frequency of the target SSVEP signal. The adaptive reference signals are the combination of artificial sinusoids as well as the processed SSVEP signals. The real SSVEP signal has more features than artificial sinusoidal signals and hence higher recognition rate is achieved. The experimental results in terms of frequency recognition and information transfer rate (ITR) show that the proposed method performs better than that of the standard CCA. Keywords brain computer interface, canonical correlation analysis, visual evoked potential I. INTRODUCTION Brain-computer interface (BCI) is a communication pathway between brain and external device without any muscular movement [1]. At present BCI has received increasing attention from researchers in neuroscience, signal processing, machine learning and clinical rehabilitee. Before human execute any task brain have to think about it and it is called brain activity. BCI takes signals generated by the brain activity form the scalp or within brain and converts it into machine readable command. The frequency components of same task may differ in different time but the signal generated by retina excitement is similar to the visual flickering [2]. Based on this technique steady-state visual evoked potential (SSVEP) system is developed. At present SSVEP is intensively accepted in BCI to bridge human brain with extern device like computer. Most of the SSVEP-based BCI systems use electroencephalography (EEG) because of its excellent signal-to-noise ratio [2] and relative immunity to artifacts. EEG also costs very little, comfortable and portable, user just have to wear a special helmet in which electrodes are placed. Wang et al. demonstrated the feasibility of using a mobile SSVEP-based BCI platform to make a phone call [3]. Detecting accurately of the frequency-target plays an important role in SSVEP-based BCIs. Flickering of LCD (liquid crystal display) with a constant frequency is used to generate SSVEP-based command in BCI implementation. In SSVEP-based BCI, users are asked to gaze at one of the repetitive visual stimuli at different frequencies. Frequency analysis method such as Fourier transform [4] can be used to identify the target stimuli at which the user is gazing. Canonical correlation analysis (CCA) [5] has been widely used in the frequency detection of SSVEP-based BCI. CCA use artificially generated sine-cosine reference signal set to identify the target frequency. It finds the maximum canonical correlation value between the SSVEP and the stimulating signals to identify the frequency of SSVEP signal. But artificially generated sine-cosine reference signal set has fewer features than real SSVEP signal. Therefore, use of real SSVEP signal as reference signal will enhance the frequency identification accuracy and system will achieve better performance. In this paper, we proposed a method where the SSVEP signals with recognized frequency are successively included together with the artificial sinusoids to obtain the adaptive reference signals set. Then the CCA is employed to recognize the frequency of the upcoming SSVEP signal using thus obtained reference signals. II. METHODS AND DATA The CCA has been widely used in the frequency detection of the SSVEP in BCI systems. The standard CCA method, which uses artificially generated sinusoidal signals as reference signals, was first proposed for SSVEP detection without calibration. We are proposing a new method based on standard CCA method, which uses both artificially generated sinusoidal signals and real SSVEP signals as reference signals. 245

247 The data set and single-trial test data are denoted K C N M by Xˆ CN R and X R respectively; where, K is the number of stimuli, C is the number of channel, N is the number of sampling point, M is the number of trial. The trial in SSVEP means capturing of electroencephalography (EEG) for a single guess of visual stimuli for a specific stimulus frequency. Our aim is to identify the target signal. Take the target CN signal as X R and adding it to one of the K classes Z k, where k 1,2,, K. Any class Z k corresponds to the stimulus frequency f f, f, 2,. Unsupervised and supervised k 1 f K methods can be used to calculate k as f X, f X, respectively, and k Y k k Y k HN where Y k R 2 is an artificially generated reference signal for the k th stimuli generated by sin 2f kn cos 2f kn Y 1 2 k, l,,, Fs Fs sin 2f k Hn cos2f k Hn Z N Fs (1) here, F s is the sampling rate and H is the number of harmonics. The target Z can be identified by the following role arg max k k, k 1,2,..., K (2) In SSVEP based BCI, those methods goal is to find the maximum k to optimize the accuracy of target identification. A. Standard CCA method CCA is a statistical way to measure the underlying correlation between two sets of multidimensional variables. Let us consider X and Y are two x X T w and multidimensional variables. x y Y T w y are their linear combinations. CCA finds the weight vectors w x and w y such that x and y will be maximum by solving the following problem T E xy x, y max wx, wy T T Exx Eyy (3) To recognize the frequency of the SSVEP signal, CCA calculates the canonical correlation k between the multi-channel EEG test signal X and the reference signals at each stimulus frequency Y k. The frequency with the maximal correlation will be selected as the frequency of the SSVEP signal. B. Proposed method The set of reference sinusoidal signals Y Y, Y2,, including H harmonics Y T 1 K corresponding to the stimulus frequencies f 1, f2,, f K is used in CCA, where Y k is generated using (1). If X m is the m th trial of SSVEP signal for k th Artificial sinusoidal reference signals Fig. 1. Block diagram of the proposed method stimuli, CCA(X m,y) recognize the frequency as f k. In online BCI, the trials X1, X 2,, X m are available 1 during the recognition of stimulus frequency of the trial X m. In the proposed method called data adaptive CCA (AdCCA), an improved reference signal is derived using real SSVEP data defined as T Yˆ Y, X X T 1,, to recognize current trial m1 X, m where 1 C X X c i c. Fig. 1 shows the block 1 C diagram of the proposed method. The artificial sinusoidal signals are used to recognize the frequency of the 1 st trial of SSVEP. The average over the channels of the first trial is added to the reference set. Hence the reference signals set is updated which contains both artificial and real SSVEP reference signals to recognize the frequency of the next or upcoming trial of SSVEP signal. After recognizing any trial, it is included in the reference set to produce the data adaptive reference set to be used in CCA for frequency recognition of SSVEP. C. Data description Test trial CCA Channel average Classified successfully SSVEP Execute respective command We are using open access data punished by M. Nakanishi [6]. There were 12-target visual stimuli (6 6 cm each). They were presented on a 27-inch LCD monitor (ASUS VG278) with refresh rate 60Hz and resolution pixels. The stimuli were arranged in 4 3 matrix as shown in figure 2 as a virtual keypad of phone and tagged with different frequencies ( f Hz, f 0.5Hz). The subjects were seated in a comfort chair at a distance of 60cm in front of the monitor in a dim room. For each subject, the experiment consists of 15 blocks. In each block, subjects were asked to gaze at one of the visual stimuli indicated by the stimulus program in a random order for 4s and complete 12 trials corresponding to all 12 targets. A red square will appear for 1s at the position of the target stimulus before each trail (Fig 1A). Subjects were asked to shift their gaze to the target within the same 1s duration. After that, all stimuli started to flicker simultaneously for 4s on the monitor. Subjects were asked to movement artifacts. Data epochs comprising eight-channel 246

248 shows the individual subject ITR comparison between standard CCA Fig 2: Stimulus designed of the 12-target BCI system. (A) user interface of a virtual keypad for a phonedialing program. (B) Frequency and phase values specified for each target. The red square in (A) is the virtual cue indicating a target symbol 5 in the experiment [6] SSVEPs were extracted according to event triggers generated by the stimulus program. All data epochs were down-sampled to 256Hz and then band-pass filtered from 6-80Hz with an infinite impulse response (IIR) filter. Zero-phase forward and reverse IIR filtering was implemented using the filtfilt() function in MATLAB. Considering a latency delay in the visual system, the data epochs were extracted in [0.135 s d s], where the time 0 indicated stimulus onset and d indicated data length used in the offline analysis. The 135-ms delay was selected towards the highest classification accuracy. Fig. 3(a). Accuracy comparison between standard CCA and AdCCA method for subject 1 Fig. 3(c). Accuracy comparison between standard CCA and AdCCA method for subject 3 Fig. 3(b). Accuracy comparison between standard CCA and AdCCA method for subject 2 Fig. 3(d). Accuracy comparison between standard CCA and AdCCA method for subject 4 III. RESULTS In this study, the proposed method AdCCA is compared with standard CCA method to validate its effectiveness for SSVEP frequency recognition. For CCA and AdCCA method, all trials are recognized one after another. For CCA only artificially generated sine-cosine reference signals are used but in AdCCA both artificially generated sine-cosine reference signals and channel average of previously recognized SSVEP signals are used to recognize the present test SSVEP signal. This procedure is repeated for 15 times such that each run serves as one validation. The number of harmonics in the reference signal is 2. The recognition accuracy at various time window length (0.5s to 4s) for each subject is shown in Fig. 3 and Fig. 3(f) shows the average accuracy of all subject comparison. We can see that AdCCA achieved better recognition accuracy for each individual subject and also for the average accuracy of all five subjects. Information Transfer Rate (ITR) is a method of evaluating BCI system performance. Here is the equation of ITR [7] Fig. 3(e). Accuracy comparison between standard CCA and AdCCA method for subject 5 Fig. 3(f). Average of all subject accuracy comparison between standard CCA and AdCCA method and our proposed method AdCCA. Fig. 4(f) also shows the average of all subject ITR comparison between CCA and AdCCA. We can see that (Fig. 4) our proposed method AdCCA achieved better ITR than standard CCA method which indicates that our proposed method AdCCA is performing better than standard CCA method. 1 P 60 ITR log 2 K P log 2 P 1 Plog 2 K 1 T (4) where, P is the classification accuracy, K is the number of class and T is the average time for a selection (seconds/selection). In this study, classification performance is calculated using different T (Target gazing time: 0.5s to 4s with an interval of 0.5s, gaze shifting time: 0.5s). Fig. 4 Fig. 4(a). ITR comparison between standard CCA and AdCCA method for subject 1 Fig. 4(b). ITR comparison between standard CCA and AdCCA method for subject 2 247

249 In both comparisons at each time window AdCCA method achieves higher performance than standard CCA method. Fig. 4(c). ITR comparison between standard CCA and AdCCA method for subject 3 Fig. 4(d). ITR comparison between standard CCA and AdCCA method for subject 4 V. CONCLUSIONS It is observed that the performance of the proposed method is better than that of the standard CCA for a wide range of the length of SSVEP signals. The underlying reason of improved performance of AdCCA is that the real references contain more frequency components than artificial signals. Such property of the adaptive reference signals (combination of artificial and real SSVEP) produces higher correlation while recognizing using CCA. The processing time of single trial SSVEP of the proposed method is short enough to implement online BCI. Fig. 4(e). ITR comparison between standard CCA and AdCCA method for subject 5 IV. DISCUSSION Fig. 4(f). Average of all subject ITR comparison between standard CCA and AdCCA method Artificial sinusoidal signal has less feature than real SSVEP signal. Standard CCA method use only artificial sinusoidal reference signal but AdCCA use both artificial and real SSVEP signal as reference signal. Therefore, the proposed method AdCCA produces better recognition accuracy than standard CCA. Fig. 4 shows that the proposed method exhibits higher ITR than standard CCA yielding that AdCCA method offers better performance than standard CCA. Averaging is one of the most common method of noise reducing [8]. Before adding the recognized trial, we average the channel and it will reduce the interchannel noise; which provide better recognition result. The comparisons in terms of recognition accuracy and ITR between standard CCA and proposed method AdCCA of SSVEP frequency of all five subjects obtained with different time window from 0.5s to 4s. REFERENCES [1] J. B. V. Erp, F. Lotter and M. Tangermann, Brain-computer interfaces: beyond medical applications, Computer-IEEE Computer Society, vol. 45, issue: 4, April [2] F. Beverina, G. Palmas, S. Silvoni, F. Piccione and S. Giove, User adaptive BCIs: SSVEP and P300 based interfaces, PsychNology, vol. 1, [3] Y. T. Wang, Y. Wang, and T. P. Jung, A cell-phone-based brain computer interface for communication in daily life, J. Neural Eng., vol. 8, no. 2, p , Apr [4] H. Bakardjian, T. Tanaka, and A. Cichocki, Optimization of SSVEP brain responses with application to eight command brain computer interface, Neuroscience Letters, pp , Nov [5] Y. Zhang, G. Zhou, J. Jin and X. Wang, A. Cichocki, Frequency recognition in SSVEP-based BCI using multiset canonical correlation analysis, Nrueal System, vol. 0, 24 Jan, [6] M. Nakanishi, Y. Wang, Y. -T. Wang, T. P. Jung, A comparison study of canonical correlationi analysis based methods for detecting steady-state visual evoked potentials,plos ONE, vol.10, no.10, e140703, 19 Oct, [7] J. R. Wolpaw, N. Birbaumer, D. J. McFarland, G. Pfurtscheller and T. M. Vaughan, Brain-computer interfaces for communication and control, Clin Neurophusiol, vol. 113, issue: 6, Jun [8] U. Hassan, M.S. Anwar, Reducing noise by repetition: introduction to signal averaging, Eur. J. Phys. Vol. 31, 10 Mar,

250 Expert Reviewer Detection from Online Experiential Product Reviews Atiquer Rahman Sarkar Dept. of Computer Science & Engineering Varendra University Rajshahi, Bangladesh Abstract Consumer reviews have emerged as one of the most influential factors on a person s purchase behavior. Popular products have thousands of user generated reviews. It is important to detect expert reviewers from the reviewer community so that users get to know the credible expert opinions. The usual approaches of detecting experts/top reviewers are based on helpfulness vote. The usual approaches suffer from some limitations. Also techniques that solely rely on helpfulness votes can be manipulated. In this paper we address this problem by developing a frameworks for expert detection using a combination of the traditional techniques. We examine the existing and the proposed expert detection methods using a large movie review dataset from amazon.com and find that our proposed method improves on minimizing the deviation from true rating. Keywords Expert Reviewer Detection, Product Reviews, Experiential Product.. I. INTRODUCTION Customers are uncertain about the true quality of the products they want to buy and this uncertainty is greater for experiential products like movies, books, songs etc. because repeated purchase of the same product is quite uncommon for such products. People spend hours going through numerous reviews to make an informed decision and to build confidence on the product they are planning to buy. Several studies have shown that people cite product reviews as a top influence in their purchase behavior. In their study, Zhao et al. [1] has found evidence of stronger learning from product reviews than learning from own experience and this is more so to the cases of experiential products. McGlohon et al. [2] mentioned that consumer recommendations are the most credible form of advertising among 78% of survey responders. They also mentioned a BIGresearch survey (2009) which indicated that 43.7% of consumer electronics purchases are affected by word-of-mouth. Chen et al. [3] mentioned a Wall Street Journal report (2006) that 71% of online U.S. adults use consumer reviews for their purchases and 42% of them trust such a source. But reviews are not without problems of their own. An article in the Wall Street Journal [2] publicized that the average rating for top review sites is an astoundingly positive 4.3 out of 5 stars. Several studies have empirically shown that positive reviews are associated with higher sales whereas negative reviews tend to hurt sales of experiential products like This work was partially supported by the MoICT Fellowship under Grants No books and movies [4]. As a result, there are people who try to manipulate/post fake reviews to influence the customers. Zhao et al. [1] have shown that fake reviews create consumer uncertainty. They also showed that the effects of more positive reviews and more numerous reviews on consumer choice are lower on online retailing platforms which have fake product reviews. Sénécal and Nantel have found evidence that Consumers have been known to rely significantly more on recommendations for experiential products than other types of products (as cited in [1]). As such, credible reviews by the expert reviewers are likely to have greater impact on the consumer s choice probabilities. This raises the need to detect the expert reviewers from the reviewing community, highlight their reviews, and give consumers the chance to make a more informed decision. Our goal in this work is to address this issue and compare the performance of traditional expert finding techniques against our proposed technique for determining the experts. Specifically, we detail an algorithm for generating composite scores of the reviewers taking a link analytic approach. We then use the scores to rank the reviewers and compare the performance of the different algorithms against the Amazon s movie reviews data set. The rest of the paper is organized as follows: section 2 presents a brief description of the review system, section 3 discusses the traditional methods of finding experts. In section 4, we present our approach of detecting expert reviewers. In section 5, we verify the effectiveness of the proposed approach on a large dataset consists of movie reviews from Amazon.com. The paper is concluded in section 6. II. A BRIEF DESCRIPTION OF RATING SCALE USED ONLINE A rating scale is a set of categories designed to elicit information about a quantitative or a qualitative attribute of an object. Common examples of rating scales used online are the Likert scales and 1-5 rating scales in which a person selects the number which is considered to reflect the perceived quality of a product. Usually online rating scales only allow one rating per user per product. However some websites allow users to rate products in relation to several qualities. Most online rating facilities provide no or few qualitative descriptions of the rating categories. 249

251 However, there are exceptions to this such as on IMDB which describes only the top and bottom category or such as Yahoo! Movies, which labels each of the categories. There are two important points to be noted about these rating/review systems. First, it is assumed that, in these rating systems, the distance between categories are equal i.e., the distance between category 1 and 2 is the same as between category 3 and 4. And second, a good rating/reviewing system will present symmetry about a midpoint. In such symmetric scaling, equidistant quality of attributes will easily be inferred. While a Likert scale is indeed ordinal, if we can ensure that the scale being used is symmetric and equidistant, then it will behave like an interval-level measurement [5,6]. We can gain additional valuable information from these ratings if the two properties are ensured. And we can perform the kinds of analysis appropriate for interval-level measurements. A review usually consists of the following attributes [7]: (1) reviewer_id, (2) product_id, (3) given_rating, (4) helpfulness, (5) description. The research question then is: Given a set of reviews with the following attributes: (1) reviewer_id, (2) product_id, (3) product_rating, (e.g., 4 star out of 5) (4) helpfulness, (e.g., 5 thumb-ups, 2 thumbdowns), (5) description, how can we identify the expert reviewer III. TRADIOTIONAL EXPERT FINDING TECHNIQUES Given a user with n1 thumbs-ups and n2 thumbsdowns, we usually see the following techniques employed all over the web to detect the top reviewers (experts). A. Baseline/Popularity: The most used method uses the total number of thumb-ups as the score of the reviewer. A reviewer is as good as his total number of acquired thumbs-ups. A problem with this technique is that it completely ignores the thumbs-downs. For example: a reviewer with 1000 thumbs-ups but 500 of thumbs-downs could be ranked higher than a user with 999 thumbsups but 100 thumbs-downs. (example: top reviews by flipkart.com) B. Difference The second method is to use the difference between the number of thumb-ups and the number of thumb-downs as the score of the reviewer. Although it takes into consideration the thumb-downs, it introduces a problem of its own. For example: a user with 200 thumb-ups and 100 thumb-downs (difference- 100) will be ranked lower than a user with 2000 thumb-ups and 1800 thumb-downs (difference- 200). This does not seem right because the first user has twice the thumb-ups than thumb downs, while the other user has only slightly more thumb-ups than thumb-downs. C. Ratio/Proportion This method rank users by the ratio of thumb-ups to total number of thumb-ups and thumb-downs. This method works well when all the reviewers have got large number of votes. But it doesn t work well when the numbers of votes are small. For example: a reviewer with 999 thumb-ups and 1 thumb-down (ratio=0.999) is ranked lower than a reviewer with a single thumb-up but no thumb-downs (ratio = 1.0). D. Proprietary techniques Amazon rank reviewers based on their secret algorithm. But they state that the rank of a reviewer is determined by the overall helpfulness of all their reviews and the number of reviews they have written. However, to ensure that the reviewers remain active, they include a penalty term that controls the contribution of a review to the total score of the reviewer. The more recently a review is written, the greater is its impact on the rank. The problem with this technique is- it is not open source. We have observed that baseline and similar traditional approaches suffer from some limitations. In the next section, we are going to present an algorithms that combines the traditional approaches addressing the existing limitations. IV. PROPOSED MODEL We propose the following simple four step algorithm for detecting the experts. Input: All the product reviews. Output: A list of the required number of experts. BEGIN 1. Consider only the reviewers who have contributed sufficient numbers of reviews. (at least more than a minimum discard threshold) 2. Sort the remaining reviewers by their total thumbups to thumb-down differences. Select the top N1 reviewers (or those whose differences are above certain threshold). 3. Sort the remaining reviewers from step 2 by their helpfulness-proportion. Select the top N2 reviewers (or those whose helpfulness-proportion are above a certain percentage). 4. Sort the remaining reviewers from step 3 by their respective Mean Squared Error (MSE). Select the top required-number of reviewers (or those whose MSE are below a certain threshold). These reviewers are the experts. END * The 3 rd and 4 th steps can be interchanged. The order of execution of the third and fourth steps reflects which quality of the reviewers we give final preference to for being an expert: their accuracy (MSE) or the clarity of response (helpfulness-proportion). 250

252 When we talk about finding expert reviewers, we consider reviewers with three distinct qualities: (1) they have made a considerable amount of reviews, (2) their reviews are found helpful by a large number of users, and (3) less deviation of their product ratings from the true product ratings. The first step of the algorithm ensures that reviewers with number of reviews less than minimum discard threshold are not considered in the process. We can interpret this as a measure of imposing commitment towards the reviewing task. Also less than the minimum number of reviews is unlikely to be fruitful to capture the expertise of the reviewers. Helpfulness vote is primarily the most important factor in determining expertise. The more customers find a review helpful, the better the review is. Helpfulness votes are usually indicator of how well the review reflects the product quality as perceived by the users. So in the second step, we select the users who got the most number of helpful votes. Also a large amount of helpfulness votes usually means that the reviewer is posting quality reviews for considerable amount of time. Larger helpfulness proportion confirms the reviewer s ability to efficiently and clearly express his domain knowledge. Helpfulness proportion also indirectly measures the domain knowledge of the reviewer. Less Mean Squared Error reflects the reviewer s ability of predicting the true product quality. V. EVALUATION Since there is no ground truth for the true quality of the products, and nor of the reviewers, deciding how to evaluate the proposed models is another interesting problem. We cannot exactly answer the question Can we rank the reviewers correctly, according to expertise? Therefore, we propose an approximation to answer a related question, which is Given ratings of the products, how well do the algorithms perform to detect reviewers and minimize the cost function? To answer this, we have to do three things: 1. Finding the true rating of the products based on user ratings (reviews). 2. Finding the experts using our proposed approaches. 3. Comparing the performance against the traditional approaches. A. True Product Rating To find the true rating of the product, three popular methods are widely used in practice, each of which has their own limitation. The methods are 1. Baseline (Average) using only ratings * ( )+ 3. Median B. Dataset The dataset for this study was collected from Stanford University s Stanford Network Analysis Project. The dataset consists of movie reviews from Amazon. Amazon is one of the largest online retailers. Also it has one of the most active reviewing communities online. The data span a period of more than 10 years up to October Amazon s movie review consists of the following: the product id, the reviewer id, the review body, the time of the posting of the review, a helpfulness count (thumb-ups/total), and a rating of the movie on a discrete 5-point scale. There are in total reviewers, movies, and almost eight million ( ) reviews on the dataset. C. The Experiment In our dataset, there are reviewers. We limit out search for experts among those reviewers who have contributed more than 50 reviews each. Our goal is to find 200 experts from them which is roughly 1.2% of the eligible reviewers (16341 reviewers with more than 50 reviews). We have experimented with two variations (exchanging step 3 and 4) of our proposed expert detection algorithm with two different type of true product rating estimation technique (1 and 2). The flowchart of the variations is presented here: 2. Baseline (Average) using ratings and helpfulness votes 251

253 Table 1. Performance comparison between the different traditional methods and proposed approaches Method of detection No. of revie ws made No. of movie s covere d True Product Rating using only reviews Avg. Erro r Avg. MS E True Product Rating considering helpfulness votes Avg. Erro r Avg. MS E Popularity Difference Ratio Proposed1.1 Proposed1.2 Proposed2.1 Proposed D. Discussion Both of the proposed variation of our algorithm significantly outperform the traditional alternatives. On average, our algorithm achieves almost 40% reduction in MSE and 20% reduction in average-error compared to the traditional approaches. We note that our experts covered almost 40% less movies compared to the baseline approach. They also made fewer comments compared to the baseline approach (1:4). A likely explanation is that expert reviewers are choosy about which products to review. They don t review every products out there. The traditional perfect-ratio experts made the fewest of reviews. This was expected because perfect proportion of helpfulness is possible when number of reviews is few. The proposed approach without interchanging between step 3 and 4 shows lower average-error and lower MSE. This is expected because, in the last step of proposed approach, the 200 experts are selected from the 500 experts based on their sorted MSE whereas this sorting is applied on the alternative version (step 3 and 4 are exchanged) a step prior to the last step. VI. CONCLUSION Online retailers have already established themselves as a strong competitor of the traditional brick and mortar stores. These retailers are supporting their potential customers with product information by building online communities to provide product reviews to customers. However, credibility of these reviews is a decisive factor when purchasing a product. In the real life too, we tend to attach different weights to different sources according to their socialties and expertise level. Finding and highlighting the experts will be a great help to consumers in this regard. In summary, we presented an adaptation of the traditional techniques for detecting experts in review communities. We performed a large scale empirical evaluation of this method, demonstrating its effectiveness for detecting experts. We found that the detected experts made significant improvement on minimizing the deviation from the true rating. We have used traditional approaches to get the true rating of the products for calculating MSE. In future, more sophisticated and accurate product rating systems may be used. For example: we may use heuristics found in the literature to detect potentially stuff the ballot box reviews and discard them. We also used simple heuristic (such as at least 50 contributions) to limit our search-domain for experts. It may be useful to combine more features such as typical vs. outlier reviews, order of reviews (which reviews were already live) etc. in the process. ACKNOWLEDGMENT We thank Dr. Shamim Ahmad for valuable discussions. REFERENCES [1] Y. Zhao, and S. Yang, V. Narayan, and Y. Zhao, Modeling Consumer Learning from Online Product Reviews, Marketing Science, [2] M. McGlohon, N. Glance, and Z. Reiter, Star Quality: Aggregating Reviews to Rank Products and Merchants, in ICSWM 10 4 th International AAAI conference on web and social media, [3] Y. Chen, S. Fay, and Q. Wang, The Role of Marketing in Social Media: How Online Consumer Reviews Evolve, Available at [4] J. Chevalier, and D. Mayzlin, The effect of word of mouth on sales: Online book reviews, Journal of Marketing Research 43: , [5] S. Jamieson, Likert Scales: How to (Ab)use Them, Medical Education, Vol. 38(12), pp , 2004 [6] M. R. Harwell, G. G. Gatti, Rescaling ordinal data to interval data in educational research, Review of Educational Research, 71, , [7] L. Akoglu, R. Chandy, C. Faloutsos, Opinion fraud detection in online reviews by network effects, in International AAAI conference on web and social media (ICSWM,2013). 252

254 FPGA Based Pulse Oximeter using VHDL Farhana Binte Sufi Dept. of Applied Physics and Electronic Engineering University of Rajshahi Rajshahi, Bangladesh Abstract This paper describes a Field Programmable Gate Arrays (FPGA) based Pulse Oximeter using VHDL as the hardware description language. Oximetry refers to the determination of the percentage of oxygen saturation of the circulating arterial blood. A commercially available pulse oximetry probe is used to obtain the Photoplethymysography (PPG) signals. The probe is controlled and the signals processed with the FPGA. The calculated percentage of oxygen saturation will be then displayed. The project is ongoing and for experimentation a Spartan3E FPGA is used. Further scope includes heart-beat rate measurement. FPGAs can be used for wearable sensor systems. A pulse oximeter based on FPGA has the future scope of combined use with similar other systems (e.g. spirometers, etc.), thus providing a cheap, portable, non-invasive health monitoring system. Keywords FPGA, VHDL, Pulse Oximeter, Wearable Sensors, Photoplethymysography (PPG), Non-invasive Health Monitoring System I. INTRODUCTION This paper describes a Field Programmable Gate Arrays (FPGA) based Pulse Oximeter using VHDL as the Hardware Description Language. Oximetry refers to the determination of the percentage of oxygen saturation of the circulating arterial blood. Oxygen saturation, SpO 2 = [ ] [ ] [ ] (1) Where, [HbO 2 ] is the oxygenated hemoglobin concentration and [Hb] is the deoxygenated or reduced hemoglobin concentration [1]. Photoplethymysography (PPG) is a non-invasive method for the detection of cardiovascular pulse waves propagating across the human body. Pulse oximetry is based on the concept that arterial oxygen saturation can be made using two wavelengths. The Beer-Lambart law or the spectrophotometric technique applicable to hemolyzed blood yields best results for Red (650 nm) and Infrared (805 nm) wavelengths [1,2]. Red LED Finger Figure 1: Pulse Oximeter Probe [3] IR LED Photo sensor Md. Maruful Islam Dept. of Applied Physics and Electronic Engineering University of Rajshahi Rajshahi, Bangladesh Reduced hemoglobin (Hb) has higher optical extinction in the red (R) region (650 nm) of spectrum than oxygenated hemoglobin (HbO 2 ) and lower optical absorption in the near infrared (IR) region (805 nm) [4]. These differences in the extinction coefficients can be used for the determination of the light absorbed by Hb and HbO 2 by the normalized ratio, R [5]: R = ( ) (2) AC R is the ac signal for the Red wavelength and DC R the dc signal for the R wavelength. And AC IR and DC IR are the ac signal and the dc signal for the IR wavelength respectively. The ac signals are due to the pulsing of arterial blood while the dc signal is due to all the non-pulsing absorbers in the tissue [1][2]. Equation 2 correlates to SpO 2 [5] by SpO 2 = R ( ) II. DESIGN FLOW (3) The oximeter probe bought from the local market uses Red LED, IR LED and a photo sensor. The proposed idea for the FPGA based pulse oximeter that the FPGA is used to control the LEDs and also to do all necessary digital signal processing and finally display the output (Fig.2). IR and Red LED-Sensor probe FPGA Display Figure 2: Block diagram of the FPGA based Pulse Oximeter A finger is placed between the LEDs and the photo sensor for SpO 2 measurement. Absorbed light for HbO2 and Hb through tissue is detected with the photo sensor and a light to frequency converter produces pulsating signals. The outputs given by the probe are pulses of the two different wavelengths for the R and IR. The FPGA is used to control the LEDs on and off period through a demultiplexer. First the Red LED is turned on and off repeatedly for 1-3 seconds. When a finger is placed between the probe the pulse widths vary. The pulsating output from the sensor for the Red LED is then analyzed. And the whole process is repeated again for the IR LED. 253