SYNTHESIS AND INVESTIGATION OF NOVEL NANOMATERIALS FOR IMPROVED PHOTOCATALYSIS

Transcription

1 SYNTHESIS AND INVESTIGATION OF NOVEL NANOMATERIALS FOR IMPROVED PHOTOCATALYSIS by XIAOBO CHEN Submitted in partial fulfillment of the requirements For the degree of Doctor of Philosophy Dissertation Advisor: Dr. Clemens Burda Department of Chemistry CASE WESTERN RESERVE UNIVERSITY August, 2005

2 CASE WESTERN RESERVE UNIVERSITY SCHOOL OF GRADUATE STUDIES We hereby approve the dissertation of candidate for the Ph.D. degree *. (signed) (chair of the committee) (date) *We also certify that written approval has been obtained for any proprietary material contained therein.

3 Table of Contents List of Tables List of Figures Acknowledgement List of Abbreviations Abstract ix x xxvii xxix xxxii Chapter 1. Introduction 1.1 Overview on the TiO 2 for photocatalysis Electronic band structures in semiconductors for photocatalysis Water splitting using semiconductor photocatalysis Photooxidation at the liquid-solid interface on TiO 2 catalysts Limitation of TiO 2 as an efficient photocatalyst and the modification of TiO Non-metal doped TiO 2 for photocatalysis Synthesis of non-metal doped TiO 2 nanomaterials Property investigation methods for non-metal doped TiO 2 nanomaterials Properties of non-metal doped TiO 2 nanomaterials Optical property of non-metal doped TiO 2 nanomaterials Electronic property of non-metal doped TiO 2 nanomaterials Photoelectrochemical property of non-metal doped TiO 2 nanomaterials Photocatalytic property of non-metal doped TiO 2 nanomaterials 25 i

4 1.7 Specific aims for this dissertation References 28 Chapter 2. Nitrogen Doped TiO 2 Nanoparticles for Visible-Light Photocatalysis Abstract Introduction Experimental Results and discussion Conclusion References 43 Chapter 3. Formation of Oxynitride as the Photocatalytic Enhancing site in Nitrogen Doped Titania Nanocrystals: Comparison to a Commercial Nanopowder Abstract Introduction Experimental Results UV-vis reflectance spectroscopy FTIR spectroscopy Raman spectroscopy Transmission electron microscopy X-ray powder diffractometry X-ray photoelectron spectroscopy 60 ii

5 3.3.7 Photocatalytic activity Discussion Conclusions References 69 Chapter 4. Investigation of the Relation between the Structure, Chemical Composition, Optical and Photocatalytic Property of Nitrogen-doped Titania Nanomaterials Using Bottom-up method Abstract Introduction Experimental Results Discussion The relationship between the nitrogen concentration and the absorption property and catalytic property of the doped TiO The doping reaction chemistry of titanium isopropoxide with amine Conclusions References 89 Chapter 5. Photocatalytic degradation of azo dyes by nitrogen-doped TiO 2 nanocatalysts Abstract 94 iii

6 5.1 Introduction Experimental Results and discussion Characterization AO7 decolorization by UV/doped-TiO AO7, MX-5B and RB5 degradation by solar light/doped-tio Conclusions References 109 Chapter 6. Visible-light Sensitive C-, N- and S-Doped Titanium Dioxide Oxidized from Titanium Carbide, Nitride and Sulfide Abstract Introduction Experimental Results and discussion Structural transformation Chemical composition transformation Lattice vibrational transformation UV-visible absorption transformation Valence band transformation Conclusions References 127 iv

7 Chapter 7. Visible-light Sensitive Carbon-, Nitrogen- and Sulfur-Doped TiO 2 Nanocrystals Directly Derived from Micrometer-sized Inorganic Precursors Abstract Introduction Experimental Results and discussion Conclusions References 139 Chapter 8. The Electronic Structure of Visible-light Responsive N-, C- and S- Doped Titania Obtained by X-ray Spectroscopy Abstract Introduction Experimental Results and discussion Structural, chemical and optical properties X-Ray Absorption of TiO 2 and TiO 2-x N x X-Ray Emission of TiO 2-x N x Contribution of the N, C and S dopants to the electronic band structures of TiO 2-x N x, TiO 2-y C y and TiO 2-z S z Correlation between XPS, XAS and XES results to build complete electronic band structure of pure TiO v

8 8.4 Conclusions References 165 CHAPTER 9. Towards Understanding the Connections between Molecules and Nanocrystals via the Uniquely Stable CdSe Molecular Clusters Abstract Introduction Experimental Results and discussion Conclusions References 188 Chapter 10. The Crystallization Process in 2 nm CdSe Quantum Dots and Related Surface Induced Optical Property Changes Abstract Introduction Experimental Results Discussion The electronic structure change of the small CdSe nanoparticles in the crystallization process The surface induced absorption evolution during the crystallization 213 vi

9 The surface induced PL and PLE evolutions during the crystallization Surface induced bleach in the femtosecond TDA measurements Conclusions References 227 Chapter 11. Coherency Strain Effects on the Optical Response of Core/Shell Hetero-Nanostructures: A Spectroscopic Investigation Abstract Introduction Experimental Results Discussion Strain/stress effects Quantum confinement Electronic confinement Conclusions References 248 CHAPTER 12. Femtosecond Carrier Relaxation and Time-Resolved Temperature Profiles of Semiconductor Quantum Dot Ensembles Abstract 251 vii

10 12.1 Introduction Experimental Results Discussions Conclusions References 266 Appendix 271 Bibliography 291 viii

11 List of Tables Table 5.1 Characteristics of Acid Orange 7, Reactive Black 5, and Procion Red MX-5B. Table 6.1 Raman vibrational mode frequencies of the TiO 2-x C x, TiO 2-x N x and TiO 2-x S x compared to pure rutile TiO 2. ix

12 List of Figures Figure 1.1 Schematic photoexcitation in a solid followed by deexcitation events. Figure 1.2 Energies for various semiconductors in aqueous electrolytes at ph = 1. Figure 1.3 Space charge layer formation and band bending between n-type semiconductor and solution. Figure 1.4 Schematic of Schottky barrier in a semiconductor-metal system. Figure 1.5 Potential energy diagram for the H 2 /H 2 O and O 2 /H 2 O redox couples relative to the band-edge positions for TiO 2. Figure 1.6 Water splitting in a semiconductor photoelectrochemical cell. Figure 1.7 Water splitting of on composite semiconductor system. Figure 1.8 Water splitting in a semiconductor system with sacrificial donor. Figure 1.9 Solar spectrum at sea level with the sun at zenith with the absorption region of TiO 2. Figure 1.10 Metal-modified semiconductor photocatalyst particle. Figure 1.11 Photoexcitation in composite semiconductorsemiconductor photocatalyst. Figure 1.12 Excitation steps using dye molecule sensitizer. x

13 Figure 1.13 Distribution of the dopants used in the TiO 2 nanomaterials in the periodic table. Figure 1.14 (A) Reflectance measurements showing the red shift in optical response due to the nitrogen doping of TiO 2 nanoparticles. (B) Diffuse reflectance spectra of S- doped and pure TiO 2 powders (PT-101: rutile). Figure 1.15 (A) Total DOSs of doped TiO 2 and (B) the projected DOSs into the doped anion sites, calculated by FLAPW, for the dopants F, N, C, S, and P located at a substitutional site for an O atom in the anatase TiO 2 crystal (the eight TiO 2 units per cell). The results for N doping at an interstitial site (N i -doped) and that at both substitutional and interstitial sites (N i+s -doped) are also shown. The energy is measured from the top of the valence bands of TiO 2, and the DOSs for doped TiO 2 are shifted so that the peaks of the O 2s states (at the farthest site from the dopant) are aligned with each other. Arb. unit, arbitrary units. Figure 1.16 (A) Total DOS of: (A) undoped, (B) S-doped TiO 2, (C) partial DOS of S atoms in S-doped TiO 2. (B) Total DOSs of (a) pure and (b) F-doped TiO 2 calculated by FLAPW. The dopant F is located at the substitutional site for an O atom in the rutile TiO 2 crystal (the two TiO 2 unit cells). The energy on the horizontal axis is measured from the top of the VBs. E g indicates the (effective) bandgap energy of the semiconductors. The impurity states are labeled (I) and (II). Figure 1.17 (A) IPCE vs λ for (Δ) N-doped TiO 2 and ( ) nondoped TiO 2 in 0.1 M HClO 4 at 0.5 M vs Ag/AgCl. The inset shows an expanded plot of IPCE vs λ in the xi

14 visible-light region. (B) Photocurrent spectra of the (a) pure TiO 2 and (b) S-doped TiO 2 synthesized by ion implantation and subsequent annealing. Figure 1.18 Photocatalytic properties of TiO 2-x N x samples ( ) compared with TiO 2 samples ( ). (A) Decomposition rates [measuring the change in absorption of the reference light (Δabs)] of methylene blue as a function of the cutoff wavelength of the optical high-path filters under fluorescent light. (B) Decomposition rates of methylene blue in the aqueous solution under visible light as a function of the ratio of the decomposed area in the XPS spectra with the peak at 396 ev to the total area of N 1s. The total N concentrations for the powder samples were evaluated to be 1.0 atomic %, a; 1.1 atomic %, b; 1.4 atomic %, c; 1.1 atomic %, d; and 1.0 atomic %, e. Figure 2.1 XRD spectrum of a TiO 2-x N x nanoparticle sample with an average diameter of 10 nm. Figure 2.2 TEM images of TiO 2-x N x nanocolloid particles (A) at low resolution (bar size 100 nm), (B) at high resolution image of the region in the indicated area (bar size 2 nm) inset on the left side showing a polycrystalline anatase phase. Figure 2.3 XPS spectrum of a TiO 2-x N x nanoparticle sample with an average diameter of 10 nm. Figure 2.4 Reflectance measurements showing the red-shift in optical response due to the nitrogen doping of TiO 2 nanoparticles. xii

15 Figure 2.5 Comparison of the photocatalytic decomposition of methylene blue in presence of doped and undoped titania nanoparticles, as monitored by the changes in absorbance at 650 nm after (a) 390 nm laser excitation and (b) 540 nm excitation. The inset in 3a shows the photodegradation of methylene blue in water at neutral ph. Figure 3.1 Visual comparison (A) and UV-visible reflectance spectra (B) of (a) TiO 2 nanoparticles; (b) Degussa P25 TiO 2 powder; (c) Degussa P25 TiO 2 powder nitrided with triethyl-amine; and (d) nitrided TiO 2-x N x nanocrystal powder. Figure 3.2 IR transmission spectra of P25 TiO 2 powder (red); P25 TiO 2 powder nitrided after triethyl-amine treatment (blue); TiO 2 nanocolloid particles (black) and nitrided TiO 2-x N x nanocrystals (green). Figure 3.3 Raman Spectra of P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocrystals (black) and nitrided TiO 2-x N x nanocrystals (green). Figure 3.4 TEM images of Nitrogen-doped TiO 2 nanocolloid particles (a) at low resolution (bar size 100 nm), (b) at high resolution image of the region in the indicated area (bar size 2 nm) inset on the left side and (c) the diffraction pattern showing a polycrystalline anatase phase. Figure 3.5 Comparison of the powder x-ray diffraction patterns of P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocrystals (black) and nitrided TiO 2-x N x nanocrystals (green). xiii

16 Figure 3.6 Comparison of the X-ray photoelectron spectra of P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocrystals (black) and nitrided TiO 2-x N x nanocrystals (green). The inset shows the N 1S peak for these samples in the 400 ev region. Figure 3.7 Photocatalytic decomposition of methylene blue on TiO 2-x N x nanocolloid (A) and P25 TiO 2 (B), as monitored by the decrease in optical density at 650 nm following 780, 540, and 390 nm laser excitation after excitation of a 5 nl volume of a 2 mm aqueous methylene blue solution. C) Relative photoreaction rate on the decomposition of methylene blue under 540 nm and 390 nm light with commercial P25 TiO 2 (white column), commercial P25 TiO 2 nitrided with triethylamine (black column) and N-doped TiO 2 (red column) nanocolloid particles as catalysts. The bars shown in the graph depict the relative reactivities obtained from the decolorization of methylene blue at 650 nm per unit time where the photocatalytic decomposition rate under 540 nm with commercial P25 TiO 2 is set to 1. All rates were corrected for the decomposition of methylene blue without any catalyst under identical conditions. Figure 3.8 The Ti 2p peak around 460 ev (A) and O1s peak (B) of P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocolloid particles (black) and nitrided TiO 2-x N x nanocrystals (green). The inset in (A) shows the Ti 2p 3/2 peak of these samples around 460 ev. Figure 3.9 X-ray photoelectron spectrum of the oxygen 1s peak displays a second signal shifted to higher binding energy for the nitrided nanocolloid due to the formation of O- Ti-N bond. xiv

17 Figure 4.1 XRD spectra of the resultant samples: dried TiO 2-x N x (black curve), TiO 2-x N x after heat treatment for 3 hours at 200 o C (red curve), 300 o C (green curve), 350 o C (blue curve), 400 o C (cyan curve) and 500 o C (magenta curve). Figure 4.2 TEM images of the nitrogen-doped titania nanocatalyst before (A) and after calcination (B) Figure 4.3 Global and N 1S XPS spectra of the resultant samples: dried TiO 2-x N x (black curve), TiO 2-x N x after heat treatment for 3 hours at 200 o C (red curve), 300 o C (green curve), 350 o C (blue curve), 400 o C (cyan curve) and 500 o C (magenta curve). Figure 4.4 UV-visible reflectance spectra of the resultant samples: dried TiO 2-x N x (black curve), TiO 2-x N x after heat treatment for 3 hours at 200 o C (red curve), 300 o C (green curve), 350 o C (blue curve), 400 o C (cyan curve) and 500 o C (magenta curve). Figure 4.5 FT-IR transmission spectra of the resultant samples: dried TiO 2-x N x (black curve), TiO 2-x N x after heat treatment for 3 hours at 200 o C (red curve), 300 o C (green curve), 350 o C (blue curve), 400 o C (cyan curve) and 500 o C (magenta curve). Figure 4.6 Photocatalytic decomposition (decolorization) of methylene blue on the resultant samples: dried TiO 2-x N x (black curve), TiO 2-x N x after heat treatment for 3 hours at 200 o C (red curve), 300 o C (green curve), 400 o C (cyan curve) and 500 o C (magenta curve). Figure 4.7 The relationship between the UV-visible absorption property and the nitrogen concentration in the resultant samples. xv

18 Figure 4.8 The relationship between the photocatalytic activity and the nitrogen concentration in the resultant samples. Figure 5.1 (A) XRD patterns of Degussa P25 TiO 2 powder (a), nitrogen-doped TiO 2 nanocatalyst before calcination (b) and after calcination at 400 C (c), and (B) UV-visible absorption spectra of Degussa P25 (a); nitrogen-doped TiO 2 (b) and nitrogen-doped TiO 2 after 400 o C (c). Figure 5.2 TEM images of the nitrogen-doped titania nanocatalyst before (A) and after calcination (B), the bar scale in the TEM is 50 nm. Figure 5.3 Decolorization of AO7 with three different nanoparticles. Condition: 0.06 mm AO7 mixed with 10 mg L -1 TiO 2 particles, with UV light illumination 800 W m -2. Figure 5.4 Decolorization of azo dyes by doped TiO 2 nanoparticles. Condition, 0.03 mm dyes mixed with 10 mg L -1 N-doped TiO 2 nanoparticles without calcination, illuminated by solar light with average light intensity 120 W m -2. Figure 5.5 Decolorization of azo dyes by doped TiO 2 nanoparticles. Condition, 0.03 mm azo dyes mixed with 10 mg L -1 N-doped TiO 2 nanoparticles with calcination, illuminated by solar light with average light intensity 120 W m -2. Figure 5.6 TOC disappearance (A) and SO 4 2- generation (B) of azo dyes. Condition, 0.03 mm azo dyes mixed with 10 mg L -1 N-doped TiO 2 nanoparticles with calcination, illuminated by solar light with average light intensity 120 W m -2. xvi

19 Figure 5.7 Energy Scheme of TiO 2 (bulk) and the three dyes as calculated by a semiempirical AM1 method. CBE: conduction band edge, VBE: valence band edge, HOMO: the highest occupied molecular orbital, LUMO: the lowest unoccupied molecular orbital. Figure 6.1 XRD pattern evolution after TiC (A), TiN (B) and TiS 2 (C) powder heated at different temperatures from 350 o C to 1000 o C for 6 hr. (#) anatase TiO 2 (o) rutile TiO 2, (*) TiC in (A), TiN in (B), and TiS 2 in (C). Figure 6.2 XPS pattern changes after TiC, TiN and TiS 2 after heated at different temperatures from 500 o C to 1000 o C before Ar sputtering. Figure 6.3 After Ar + sputtering for 10 min, core-level XPS spectra of the C-, N- and S-doped TiO 2 compared to that of pure TiO 2. (A) C 1s binding energy region for pure TiO 2 and TiO 2-y C y, (B) N 1s binding energy region for pure TiO 2 and TiO 2-x N x, (C) S 2p binding energy region for pure TiO 2 and TiO 2-z S z ; (D) Ti 2p binding energy region for pure TiO 2 and TiO 2-y C y, TiO 2-x N x and TiO 2-z S z ; (i) prue TiO 2, (ii) C-doped TiO 2, (iii) N- doped TiO 2, (iv) S-doped TiO 2 derived from TiC, TiN and TiS 2 powder after heated at 1000 o C for 6hr. Figure 6.4 FT-IR spectrum evolution from TiC, TiN and TiS 2 powder to the resultant C-, N- and S-doped TiO 2 after heating at different temperatures from 350 o C to 1000 o C. Peaks a, b, c, d can be assigned to the vibrations of Ti-O bonds in anatase and rutile phases of TiO 2, and Ti-S bonds in layered TiS 2. See text for details. Figure 6.5 Raman spectra for the resultant C-, N- and S-doped TiO 2 powders from TiC, TiN and TiS 2 after heating at 1000 o C. xvii

20 Figure 6.6 Absorption spectra of TiC, TiN and TiS 2 powders and the resultant TiO 2 particles after heating at different temperatures higher than 500 o C. Figure 6.7 Valence band (VB) XPS of (A) TiO 2 (i), TiC (ii), TiN (iii) and TiS 2 (iv); (B) (i) pure rutile TiO 2 and the (ii) C-, (iii) N- and (iv) S-doped TiO 2 derived from TiC, TiN and TiS 2 respectively. Figure 7.1 (A) XRD patterns of the (a, *) TiC, (c, +) TiN and (e, ^) TiS 2 and the resultant (b) C-, (d) N- and (f) S- doped TiO 2 samples after heating at 350ºC, 650ºC and 450ºC for 96 hours, respectively; (o) anatase, and rutile (#) TiO 2 structures. (B) The corresponding reflectance spectra. Figure 7.2 TEM and HRTEM images of C-doped TiO 2 nanoparticles. TiC powder (A) and the resultant C-doped TiO 2 nanoparticles after heating before (B) and after size selective centrifugation (C) and (D). HRTEM images of three C-doped TiO 2 nanocrystals (E), (F), and (G). The insets in (A) and (B) show the selected area electron diffraction patterns of cubic TiC and the C-doped TiO 2 nanoparticles with anatase phase. Figure 7.3 (A) C 1s, (B) N 1s, (C) S 2p, (D) Ti 2p and (E) VB XPS spectra of TiO 2-x C x, TiO 2- yn y, TiO 2-z S z and pure TiO 2. VB XPS displayed additional electronic states above the valence bandedge caused by C 2p, N 2p and S 3p orbitals as enlarged in (F). Figure 7.4 (A) Photocatalytic decomposition curves of the methylene blue under Xe lamp illumination with TiO 2-x C x, TiO 2-y N y and TiO 2-z S z as photocatalysts compared to Degussa P25. (B) The correlation between the amount of photodecomposition of xviii

21 methylene blue after 500 min Xe lamp illumination and the integral of the absorption curve of these doped TiO 2 from 800 nm to 400 nm. Figure 8.1 XRD patterns for the titanium compounds: (a) commercial TiN; (b) TiN after heating; (c) commercial TiC; (d) TiC after heating; (e) commercial TiS 2 ; (f) TiS 2 after heating. The resultant samples from TiN, TiC and TiS 2 display the typical diffraction pattern of rutile TiO 2. Figure 8.2 Before Ar sputtering, global XPS of the N-, C- and S-doped TiO 2 powder: TiO 2-x N x, TiO 2-y C y and TiO 2-z S z. The carbon 1s signals at ev in the spectra were from the carbon tape. Figure 8.3 After Ar + sputtering for 10 min, partial XPS spectra of the N-, C- and S- doped TiO 2 compared to that of pure rutile TiO 2. (A) N 1s binding energy region for pure TiO 2 and TiO 2-x N x, (B) C 1s binding energy region for pure TiO 2 and TiO 2-y C y, (C) S 2p binding energy region for pure TiO 2 and TiO 2-z S z ; (D) Ti 2p binding energy region for pure TiO 2 and TiO 2-x N x, TiO 2-y C y and TiO 2-z S z ; (i) pure TiO 2, (ii) TiO 2-x N x, (iii) TiO 2-y C y, (iv) TiO 2-z S z. Figure 8.4 Reflectance spectra for the titanium compounds: (a) commercial TiN; (b) TiO 2-x N x ; (c) commercial TiC; (d) TiO 2-y C y ; (e) commercial TiS 2 ; (f) TiO 2-z S z. The N-, C- and S-doped TiO 2 display the typical bandgap around 3.1 ev and additional lower energy tail absorption. Figure 8.5 XAS spectra of Ti 2p (a) and O 1s (b) in pure TiO 2 and TiO 2-x N x samples. xix

22 Figure 8.6 XES spectra of Ti L (a) and O K α (b) in the TiO 2 and the TiO 2-x N x sample. Figure 8.7 (a) Ti 2p XAS and Ti L XES spectra, (b) O 1s XAS and O K α XES specra and (c) XPS valence band spectra of (i) pure rutile TiO 2, (ii) TiO 2-x N x, (iii) TiO 2-y C y and TiO 2-z S z. Figure 8.8 Assignment of the O K α x-ray emission spectrum (a) from the transitions between the O 1s core-level and the valence band structure (fitting curve) from XPS measurement (b); Assignment of the Ti L x-ray emission spectrum of (c) from the transitions between the Ti 2p core-level and the valence band structure (fitting curve) from XPS measurement (d). Figure 8.9 Comparison of the sum of the PVBs of O and Ti to the total valence band (VB) from XPS measurement with its fitting. Figure 8.10 (a) The correlation of the O 1s core-level XPS and O 1s x-ray absorption spectra used to construct the partial conduction band (PCB) structure having O 2p characters; (b) The correlation of Ti 2p core-level XPS and Ti 2p x-ray absorption spectra used to construct the partial conduction band (PCB) structure having Ti 3d characters. Figure 8.11 The constructed conduction band (CB) and the partial conduction bands (PCBs) from O and Ti in TiO 2. Figure 8.12 (A) Comparison of the inverse-photoemission spectrum (a) [63] and theoretical calculation of the band structure (b) of TiO 2 to the constructed conduction band structure (c) in this contribution. (B) Comparison of the constructed band structures xx

23 of the TiO 2 in this contribution to the theoretical single particle calculations of the band structures of TiO 2. (a), (c), (e): constructed band structures; (b), (d), (f): theoretical calculations. Figure 9.1 MALDI-MS spectrum of Cd m Se n clusters (m, n = 3-9) with 2-(4- hydroxyphenlyazo) benzoic acid (HABA) as the matrix. The inset shows the whole mass spectrum from 500 to 10,000. For brevity Cd m Se n cluster peaks are labled as (m,n). Figure 9.2 (A) Steady state UV-visible absorption spectra of Cd 7 Se 7 clusters, dotted lines: absorption spectra, red solid lines: fitted absorption spectra, green gaussian curves: fitted absorption peaks, inset: the transition energy against the number of state, the upward arrows shows the excitation energy in the TDA measurement; (B) TDA spectra of CdSe cluster after different delay times; (C) Dynamics of the TDA spectra detected at different transition energies, the inset shows the early rise of the TDA signals; (D) Lifetimes of the TDA dynamics at different detection energies, τ r : rise times of the TDA, τ d1, τ d2 : the first and second components of the decays of the TDA signals. Figure 9.3 (A) PL and PLE spectra of the Cd 7 Se 7 clusters, the upwards arrows indicate the position where lifetimes were measured as in B; (B) Time-resolved decay of the PL of Cd 7 Se 7 clusters. Figure 9.4 (A) Comparison of the UV-visible absorption spectra (solid line) and the calculation (dashed line) by the TDLDA method. (B) Comparison of the calculated and experimental transition energies in the absorption spectra shown in A. xxi

24 Figure 9.5 Proposed structure of Cd 3 Se 4, Cd 4 Se 4, Cd 4 Se 6 and Cd 7 Se 7 with the corresponding carbon frameworks of norbornane, bicyclo-[2,2,2]-octane, adamantane and diamantane, zincblende CdSe and diamond. S-Figure 1 The fitting of the PL spectrum of Cd 7 Se 7 cluster S-Figure 2 The fitting of the PLE spectrum for Cd 7 Se 7 cluster S-Figure 3 Comparison of the experimental absorption spectrum with the calculated absorption spectrum using LDA method. S-Figure 4 Comparison of the experimental absorption spectrum with the calculated absorption spectrum using LDA method. S-Figure 5 The Comparison of energies of the corresponding peaks in the absorption, PL, PLE spectra and the calculated absorption spectra for Cd 7 Se 7 cluster. Figure 10.1 UV-visible (solid lines) and photoluminescence (PL) (dash lines) spectra of CdSe NPs observed after 10 min (black), 15 min (red), 30 min (green), 45 min (blue), 60 min (cyan) and 90 min (magenta) reaction time. Figure 10.2 Evolution of the Stokes shift of the CdSe NPs during synthesis. Figure 10.3 Temporal evolution of the PLE spectra of the CdSe NPs during the crystal formation. The emission wavelengths of the PLE spectra for the sample of 10 min (black), 15 min (red), 30 min (green), 45 min (blue), 60 min (cyan), and 90 min xxii

25 (magenta) reaction time are 474 nm, 490 nm, 500 nm, 510 nm, 510 nm, and 525 nm, respectively. (For clarity the spectra were normalized and offset.) Figure 10.4 HRTEM images of CdSe NPs after 10 min (Left) and 90 min (Right) particle formation. Compared to the amorphous phase of the 10 min sample, the 90 min sample was well crystallized. Figure 10.5 XRD patterns of the CdSe NPs formed for 10 min and 90 min. Due to the small size of these samples, the XRD signal were weak for both samples. The XRD results shows that these CdSe NPs had similar sizes around 2.1 ± 0.1 nm, and the CdSe NPs formed after 90 min s reaction had better crystallinity from their stronger diffraction pattern. Figure 10.6 a). The changes of the TDA spectra extracted at 2 ps delay time for the different CdSe nanoparticles over the crystallization process. b). The evolution of different components (τ r : rise lifetime, τ 1 : first component of bleach decay, τ 2 : second component of bleach decay) of lifetimes of decay dynamics of the bleaches for different samples. Figure 10.7 The illustration of the origin of absorption evolution over the crystallization of small CdSe nanoparticles, where the blue curve illustrates the steady-state absorption spectra. 0 refers to the ground state, while 1S 1S and 3 / 2 e 2S / 21S e 3 refer to the lower two states in the optical transitions in the small CdSe nanoparticles, the distribution of the surface states are indicated as the discrete lines. The boldness of the lines and the green Gaussian shape areas illustrate the density of the electronic states. xxiii

26 Figure 10.8 The illustration of the origin of PL evolution during the crystallization of small CdSe nanoparticles, where the blue curve illustrates the steady-state emission spectra. T refers to the dark states or the triplet states. The boldness of the magenta down arrows illustrates the probability of the optical transitions. Other signs have the same meanings as in Figure Figure 10.9 The illustration of the origin of emission evolution in the formation of small CdSe nanoparticles. The signs have the same meanings as in Figure Figure The illustration of the origin of bleach in the TDA of small CdSe nanoparticles, where the TDA is red-shift to the onset of the UV-visible absorption spectra, and the blue curves illustrate the steady-state and transient absorption spectra after the pump light. Other signs have the same meanings as in Figure 4. Similar to the state-filling of the intrinsic bandedge states in high quality CdSe nanoparticles, the bleach here is due to the block of the optical transitions between the surface states above the bandedge of the valence band and the surface states below the conduction band, caused by the state-filling of the defect states, and occurs at the lower energy tail of the steadystate absorption spectra. Figure 11.1 Absorption spectra of CdSe/CdS core/shell nanoparticles with CdS capping of different thickness after the indicated capping times. For clarity, the spectra are offset by 0.01 absorbance units, each. The red-shift of the absorbance maximum is caused by the increasing shell thickness. xxiv

27 Figure 11.2 a) Steady state PL spectra of CdSe/CdS core/shell nanoparticles with different shell thickness (ML: monolayer) after the indicated capping times. Excitation wavelength is 400 nm. b) PL quantum yield of CdSe/CdS core/shell nanoparticles as the synthesis time for the shell progresses. Quantum yields of nanocrystal solutions were calculated by comparing the integrated emission to that of Rhodamine 6G in ethanol, with an excitation wavelength of 490nm. Figure 11.3 a) PL lifetimes of CdSe (0 min) and CdSe/CdS nanoparticles (5 min to 360 min capping time) recorded at 570 nm. b) PL lifetime components of CdSe/CdS nanoparticles versus capping reaction time. The experimental error is < 5%. Figure 11.4 Comparison of shifts in absorption maximum (a) shifts in emission maximum (b) and Stokes shifts (c) of CdSe/CdS core/shell nanoparticles obtained at different capping reaction times and leading to different capping layer thickness. The experimental errors for the absorption and emission experiments are 0.2 nm, and 0.4 nm for the derived Stokes shift. Figure 11.5 Two-dimensional sketch of the coherency strain in a 1-monolayer capped core/shell nanoparticle (left) and dislocation formation in a multiplayer core-shell nanoparticle (right). Figure 11.6 Illustration of the quantum confinement of core/shell nanoparticle systems. Figure 11.7 Qualitative description of the frontier orbital energy levels of a nanoparticle confined in a quantum shell. xxv

28 Figure 12.1 UV-visible absorption (blue) and PL (red dashed, arbitrary scale) spectra of CdSe QDs indicating a high-quality quantum dot sample with no detectable low-energy surface states. The numbered markers refer to the energies associated with the electronhole pair states shown. Figure 12.2 (A) Contour plot of the experimental femtosecond TDA spectra ( ev, nm) measured with 100 fs time resolution in a 2.9 ns time window. (B) The relaxation dynamics recorded at different electron-hole pair energies: (from bottom to top) (+) 2s 1/2 (h)1s(e) (2.76 ev); (X)1p 3/2 (h)1p(e) (2.62 ev); (Δ)1s 1/2 (h)1s(e) (2.54 ev); ( ) 2s 3/2 (h)1s(e) (2.39 ev); (Ο) 1s 3/2 (h)1s(e) (2.30 ev); and ( ) 1s 3/2 (h)1s(e) (2.28 ev). The inset expands the first three kinetic traces at higher time resolution. Figure 12.3 (A) Contour plot of the time-resolved n rel matrix in the energy range of ev, and a time window of 3 nanoseconds. (B) n rel (open circles) fitted with Fermi-Dirac function from equation (4) (solid lines) at different delay times during the first 500 ps. For clarity, each curve is offset by Figure 12.4 (A) Fermi-Dirac distribution functions at different delay times after the femtosecond laser pulse excitation. (B) Carrier temperature ( ) as a function of delay time obtained from the distributions in 4A. The curves come from three different models for net non-exponential relaxation, as described in the text. xxvi

29 Acknowledgements I would like to express my deepest regards and sincere gratitude to Professor Clemens Burda for his guidance, enthusiasm and support. I really appreciate his constant encouragement throughout my Ph. D studies, his faith in the abilities of the students and his ready availability on all matters. He is not only a helpful advisor, but also a dear friend. I am grateful to all my labmates, for the past and present helpful discussions and providing a friendly environment to work and study. Professors Robert C. Dunbar, Michael Zargoski, Daniel A. Scherson and Pirouz Pirouz are gratefully acknowledged for taking time to serve on my Ph. D. committee. Professor Robert C. Dunbar is very appreciated for his encouragement, help and discussion on the theoretical calculations on the small CdSe clusters. Professor Fred L. Urbach is also appreciated for his help on the use of his UV-visible reflectance spectrometer. Professors Fred L. Urbach, Alfred B. Anderson, Frank Ernst and Zhong- Wu Guo are appreciated for their wonderful courses. Thanks to Dr. James Gole at the School of Physics, Georgia Institute of Technology for collaboration. Thanks to Dr. Jin Lin and Yang Liu at University of Wisconsin-Milwaukee for help on carrying out photocatalytic experiments with dyes under sun light. Thanks to Dr. Jing-Hua Guo at Lawrence Berkeley National Lab for the XAS and XES measurements. Thanks to Dr. Jennings and staff of the CSAM center for helping with XPS, XRD and TEM measurements, and Dr. Huolei Peng and Nakul Maiti for help on Raman measurements. xxvii

30 I need to thank my parents who always supported me in every step of my life. I also would not have made it through my Ph. D. study without the love and support of my wife, Xiaoping. She was always there with me, helping me and encouraging me, so I can resume my confidence in spite of frustrations and continue to work through it. The financial support from the College of Art & Science at Case, National Science Foundation (NSF), National Institutes of Health (NIH), Petroleum Research Fund - American Chemical Society (ACS-PRF) and the Ohio Board of Regents is gratefully appreciated. xxviii

31 List of Abbreviations ALS the Advanced Light Source AO7 acid orange 7 CB CBE CCD DF DFT DOS E f E g FLAPW FTIR FWHM GGA HABA HOMO HRTEM IAA IPCE LDA LUMO MALDI-TOFMS conduction band conduction band edge charge-coupled device distribution function density functional theory the density of state Fermi level bandgap the framework of the local density approximation Fourier-transformed infrared the full width at half-maximum the generalized gradient approximation 2-(4-hydroxyphenlyazo) benzoic acid the highest occupied molecular orbital high-resolution transmission electron microscopy indole-3-acetic acid the incident photo-to-current efficiency local density approximation the lowest unoccupied molecular orbital matrix-assisted laser desorption time-of-flight mass spectroscopy xxix

32 MX-5B NHE NP PCB PDOS PL PLE PVB QD RBS procion red MX-5B normal hydrogen electrode nanoparticle partial conduction band the partial density of state photoluminescence photoluminescence excitation partial valence band quantum dot Rutherford backscattering RB5 reactive black 5 SEM SIMS TDA TDLDA TDPA TEM TEY TOC TOF TOP TOPO TPP scanning electron microscopy secondary ion mass spectrometry time-resolved differential absorption the time-dependent local density approximation n-tetradecylphosphonic acid transmission electron microscopy total electron yield total organic carbon time-of-flight trioctylphosphine trioctylphosphonic oxide triphenylphosphine xxx

33 XAS XES XPS XRD UV UV-vis VB VBE x-ray absorption spectroscopy x-ray emission spectroscopy X-ray photoelectron spectroscopy X-ray diffraction ultraviolet ultraviolet-visible valence band valence band edge xxxi

34 Synthesis and Investigation of Novel Nanomaterials for Better Photocatalysis Abstract by XIAOBO CHEN Since the discovery of the photocatalytic splitting of water on TiO 2 electrode by Fujishima and Honda in 1972, enormous effort has been spent on the study of TiO 2 under light illumination, due to its various potential applications, such as photovoltaics and photocatalysis. The optical properties, in particular the absorption, of TiO 2 are essential to its photon-driven applications. Typically, TiO 2 absorbs in the UV regime, which is only a small fraction of the sun s energy (< 10%). The performance of TiO 2 can be enhanced by shifting the onset of its absorption from the UV to the visible region. Metals have been employed to tune the electronic structure of TiO 2 -based material. The photocatalytic reactivity of metal-doped TiO 2 depends on many factors, and metal doping can result in thermal instability and increased carrier trapping. The desired visible-light absorption of TiO 2 can be also achieved by using main group dopants. In this dissertation, different non-metal elements, C, N and S, are incorporated to the lattice of TiO 2 to induce the absorption in the visible-light regime. Both bottom-up and top-down methods are used to synthesize these doped TiO 2 nanoparticles. The optical, physical, electronic and photocatalytic properties of these doped TiO 2 nanoparticles are explored with different techniques. The relationship between the xxxii

35 optical, electronic and photocatalytic properties are elucidated. The photocatalytic performance of the doped TiO 2 nanoparticles is applied not only to the model photodegradation of methylene blue, but also on other industrial dyes under natural sunlight illumination. The non-metal-doped TiO 2 nanoparticles demonstrated improved photocatalytic performance over the non-doped TiO 2 nanoparticles, i.e. in the visiblelight regime. On the other hand, as the size of nanoparticles decreases, the surface-to-volume ratio increases dramatically (~ 1/r), so does the surface area (1/r 2 ). The high surface area brought by the small size of nanoparticles becomes more important for the optical and electronic properties of nanomaterials, compared to the bulk materials. Besides the wellknow quantum-confinement effect, the surface effect should be taken into account for small nanoparticles. Thus in this dissertation, the synthesis and properties of II-VI (CdSe, CdSe/CdS) semiconductor nanoparticles are investigated to elucidate the surface effect on the properties of nanoparticles, which helps to understand the photocatalytic property of TiO 2 nanomaterials, since the main catalytic reactions occur on the surface. The gradual crystallization of small nanoparticles, as well as its effect on the optical properties is elucidated. The interface strain/stress in the CdSe/CdS core/shell system is explored on their optical properties, as well as the hot carrier relaxation dynamics in CdSe nanoparticles. xxxiii

36 CHAPTER 1 INTRODUCTION * 1.1 Overview on the TiO 2 for photocatalysis The scientific and engineering interest in the application of semiconductor photocatalysis has grown dramatically over the last decades. 1-6 Several simple oxide and sulfide semiconductors have band-gap energies sufficient for promoting or catalyzing a wide range of chemical reactions of environmental interest. The primary criterion for good semiconductor photocatalysts for organic compound degradation is that the redox potential of the H 2 O/ OH (OH- = OH + e-; Eº= -2.8 V) couple lies within the bandgap domain of the material and that the material is stable over prolonged periods of time. Among the semiconductors including TiO 2 (E g = 3.2 ev), WO 3 (E g = 2.8 ev), SrTiO 3 (E g = 3.2 ev), a-fe 2 O 3 (E g = 3.1 ev), ZnO (E g = 3.2 ev), and ZnS (E g = 3.6 ev). 1,2 TiO 2 has proven to be the most suitable for widespread environmental applications, since it is biologically and chemically inert; it is stable with respect to photocorrosion and chemical corrosion; and it is inexpensive. Semiconductor photocatalysis with a primary focus on TiO 2 as a durable photocatalyst has been applied to a variety of problems of environmental interest in addition to water and air purification, for the destruction of microganisms, 7,8 for the inactivation of cancer ce11s, 9,10 for the photosplitting of water to produce hydrogen gas, and for the clean up of oil spills In the parlance of solid-state physics, semiconductors (and insulators) are defined as solids in which at 0 K (and without excitations) the uppermost band of occupied * This chapter is partially published in J. Nanoscience and Nanotechnology,

37 electron energy states is completely full. It is well-known from solid-state physics that electrical conduction in solids occurs only via electrons in partially-filled bands, so conduction in pure semiconductors occurs only when electrons have been excited-- thermally, optically, etc.--into higher unfilled bands. At room temperature, a proportion (generally very small, but not negligible) of electrons in a semiconductor have been thermally excited from the "valence band," the band filled at 0 K, to the "conduction band," the next higher band. The ease with which electrons can be excited from the valence band to the conduction band depends on the energy gap between the bands, and it is the size of this energy bandgap that serves as an arbitrary dividing line between semiconductors and insulators. Here, semiconductors are defined as with bandgap less than 4 ev at room temperature. Figure 1.1 Schematic photoexcitation in a solid followed by deexcitation events. [4] Semiconductor electronic structures are characterized by a filled valence band VB, and an empty conduction band, CB, and can act as sensitizers for light-reduced redox processes. When a photon with an energy of hv matches or exceeds the bandgap energy, E g of the semiconductor, an electron, e - cb, is promoted from the valence band, into the 2

38 conduction band, leaving a hole, h + vb behind. Excited state conduction-band electrons and valence-band holes can react with electron donors and electron acceptors adsorbed on the semiconductor surface or within the surrounding electrical double layer of the charged particles or recombine and dissipate the input energy as heat, get trapped in metastable surface states. The above process in the photocatalysis is illustrated in Figure 1.1. Once excitation occurs across the band gap there should be a sufficient lifetime (in the nanosecond regime) for the created electron-hole pair to undergo charge transfer to adsorbed species on the semiconductor surface from solution or gas phase contact. If the semiconductor remains intact and the charge transfer to the adsorbed species is continuous and exothermic the process is termed heterogeneous photocatalysis. 4 The initial process for heterogeneous photocatalysis of organic and inorganic compounds by semiconductors is the generation of electron-hole pairs in the semiconductor particles. Upon excitation, the fate of the separated electron and hole can follow several pathways. Recombination of the separated electron and hole can occur on the surface (pathway A) in the volume of the semiconductor particle (pathway B) or with the release of heat. The photoinduced electron/hole can migrate to the semiconductor surface. At the surface the semiconductor can donate electrons to reduce an electron acceptor (usually oxygen in an aerated solution) (pathway C); in turn, a hole can migrate to the surface where an electron from a donor species can combine with the surface hole oxidizing the donor species (pathway D). The electron transfer process is more efficient if the species are spreadsorbed on the surface. 20 The probability and rate of the charge transfer processes for electrons and holes depends upon the respective positions of the 3

39 band edges for the conduction and valence bands and the redox potential levels of the adsorbate species Electronic band structures in semiconductors for photocatalysis The ability of a semiconductor to undergo photoinduced electron transfer to adsorbed species on its surface is governed by the band energy positions of the semiconductor and the redox potentials of the adsorbate. The relevant potential level of the acceptor species is thermodynamically required to be below (more positive than) the conduction band potential of the semiconductor. The potential level of the donor needs to be above (more negative than) the valence band position of the semiconductor in order to donate an electron to the vacant hole. Figure 1.2 Energies for various semiconductors in aqueous electrolytes at ph = 1. [4] The band edge positions of several semiconductors are presented in Figure 1.2. The internal energy scale is given on the left for comparison to the vacuum level and on 4

40 the right for comparison to normal hydrogen electrode (NHE). The positions are derived from the flat band potentials in a contact solution of aqueous electrolyte at ph = 1. The ph of the electrolyte solution influences the band edge positions of the various semiconductors compared to the redox potentials for the adsorbate. Contact between a semiconductor and another phase (i.e. liquid, gas, or metal) generally involves a redistribution of electric charges and the formation of a double layer. The transfer of mobile charge carriers between the semiconductor and the contact phase, or the trapping of charge carriers at surface states at the interface, produces a space charge layer. For semiconductor-gas phase interactions, an n-type semiconductor such as TiO 2 can have surface states available for electron trapping. The surface region will become negatively charged. To preserve electrical neutrality a positive space charge layer develops just within the semiconductor causing a shift in electrostatic potential and a bending of bands upward toward the surface. Figure 1.3 illustrates the space charge layers produced from the mobility of charge across a semiconductor-solution interface for an n-type semiconductor. 4 If isolated, the semiconductor has a flat band potential diagram in the absence of a space charge layer, and contains a uniform distribution of charge (Figure 1.3a). When a semiconductor and a liquid solution are in contact (solid-liquid interface), a charge transfer occurs until an electrostatic equilibrium is achieved, i.e. their Fermi levels (E f ) reach the same energy and a depletion layer of major carriers is produced within the semiconductor, with a subsequent bending of the bands at the interface. If the Fermi level (E f ) of the semiconductor is higher than E redox of the solution, electrons are transferred to the solution and the semiconductor and the solution acquire positive and 5

41 Figure 1.3 Space charge layer formation and band bending between n-type semiconductor and solution. negative charges, respectively. 21 The excess of charge in the semiconductor is not confined to the surface, but is distributed in a region called the space charge region expanding from the surface to the bulk of the semiconductor. The electric field resulting from the formation of the space charge region is responsible for the band bending. The bands bend up when the semiconductor is positively charged, and bend down when the semiconductor is negatively charged. The existence of a positive charge on the interface increases the majority carrier concentration of electrons near the surface within the region of the space charge layer (Figure 1.3b). The space charge layer formed is called an accumulation layer. The bands of the semiconductor will bend down as they move toward the surface as a result of the decrease of electron potential energy as they move toward the positively charged outerlayer. 6

42 When negative charges accumulate at the interface the majority electron carrier concentration is less than in the interior of the semiconductor (Figure 1.3c). The space charge layer formed is a depletion layer and the bands bend upward toward the surface. When the depletion of the majority charge carriers extends far into the semiconductor, the Fermi level near the interface can decrease below the intrinsic level, which is half way between the bottom of the conduction band and the top of the valence band. The surface region of the semiconductor appears to be p-type while the bulk still exhibits n-type behavior. This space charge layer is called an inversion layer (Figure 1.3d). Figure 1.4 Schematic of Schottky barrier in a semiconductor-metal system. The energy diagram of a semiconductor-metal system with the formation of a Schottky barrier is shown in Figure 1.4. Electrically neutral and isolated from each other, the metal and the n-type semiconductor have different Fermi level position. When the metal with a higher work function (φ m ) is in contact with a semiconductor with a lower work function (φ s ), electron migration from the semiconductor to the metal occurs until the two Fermi levels are aligned and forms a space charge layer. The surface of the metal acquires an excess negative charge while the semiconductor exhibits an excess positive 7

43 charge as a result of electron migration away from the barrier region. The bands of the semiconductor bend upward toward the surface, and the layer is said to be depleted. The barrier formed at the metal-semiconductor interface is called the Schottky barrier. The height of the barrier, φ b, is given by φ b = φ m - Ex where E x, is the electron affinity, measured from the conduction band edge to the vacuum level of the semiconductor. The Schottky barrier produced at the metal-semiconductor interface can serve as an efficient electron trap preventing electron-hole recombination in photocatalysis Water splitting using semiconductor photocatalysis H 2 O cannot be photodecomposed on clean TiO 2 surfaces, even though TiO 2 can be effectively photoexcited under band-gap irradiation. Figure 1.5 illustrates the bandedge positions of TiO 2 relative to the electrochemical potentials of the H 2 /H 2 O and O 2 /H 2 O couple. 22 According to this electron energy diagram, water photolysis is energetically favorable. However, due to the presence of a large overpotential for the evolution of H 2 and O 2 on the TiO 2 surface, TiO 2 alone becomes inactive. Sustained photodecomposition of water has been achieved under conditions where photogenerated electrons and holes are separated for maximum photoreaction yield. (1) Water splitting can be achieved in a closed circuit photoelectrochemical cell employing a TiO 2 anode and a metal (Pt in most cases) cathode as shown in Figure 1.6. The photoexcitation of TiO 2 injects electrons from its valence band into its conduction band. The electrons flow through the external circuit to the Pt cathode where water 8

44 molecules are reduced to hydrogen gas and the holes remain in the TiO 2 anode where water molecules are oxidized to oxygen. Hydrogen evolves from Pt and oxygen from TiO 2. A small externally applied electrical potential (> 0.25 V) may be necessary but this is much smaller than that required in an electrochemical cell for H 2 O electrolysis (> 1.23 V). 4 Figure 1.5 Potential energy diagram for the H 2 /H 2 O and O 2 /H 2 O redox couples relative to the band-edge positions for TiO 2. Figure 1.6 Water splitting in a semiconductor photoelectrochemical cell. (2) Water splitting can also be realized in a composite semiconductor system as shown in Figure 1.7 where TiO 2 powder are deposited with metal particles (such as Pt)for H 2 evolution and metal oxide particles (such as RuO 2 ) for O 2 evolution. 23 This system behaves as a short-circuited micro photoelectrochemical cell in which Pt is the cathode 9

45 and RuO 2 is the anode. The presence of Pt and RuO 2 significantly reduces the overpotential for H 2 and O 2 production, respectively. Band-gap excitation in the TiO 2 substrate injects negatively charged electrons into the Pt particles and positively charged holes into the RuO 2 particles. Trapped electrons in Pt reduce water to hydrogen and trapped holes in RuO 2 oxidize water to oxygen. Figure 1.7 Water splitting on a composite semiconductor system. Figure 1.8 Water splitting in a semiconductor system with sacrificial donors. (3) Water splitting can also be achieved in a semiconductor system with sacrificial donors as shown in Figure 1.8. The sacrificial species are used to remove one of the photodecomposition products so that the reaction equilibrium is shifted toward further 10

46 decomposition. The sacrificial species may be oxidized by the hole-reaction product (presumably O 2 ) or reduced by the electron-reaction product (presumably H 2 ). For example, when alcohols such as CH 3 OH are added to a TiO 2 aqueous suspension, sustained H 2 production is observed upon irradiation and the alcohol molecules are oxidized to CO Photooxidation at the liquid-solid interface on TiO 2 catalysts By using TiO 2 as a photocatalyst, the oxidation of organics may be used for decontamination treatment in polluted waters. For instance, chlorobenzene undergoes mineralization over TiO 2. 25,26 Continuous band-gap irradiation of an aqueous semiconductor dispersion excites an electron from the valence band to the conduction band, creating an electron-hole pair. The electrons possess the reducing power of the conduction band energy and the holes have the oxidizing power of the valence band energy. From the band-edge positions of the valence band and conduction band, the redox capability of a photoexcited semiconductor particle in the aqueous solution can be estimated. The bulk photoelectrons and photoholes can recombine to produce thermal energy, or rapidly migrate to the surface and react with adsorbed species at the surface. In a steady state photocatalytic reaction, the rate of oxidation by the holes has to be balanced by the rate of reduction by the electrons. Either of these reactions can be rate determining. Although the above physical events are generally accepted as the initial step for the photooxidation process, the subsequent chemical events at the liquid-solid interface remain an ambiguous and controversial issue. The trapped holes have been proposed to directly oxidize adsorbate molecules, 27 or to react with surface hydroxyl groups to 11

47 produce hydroxyl radicals which are strong oxidizing agents. 28,29 The chemical identification of hydroxylated oxidation intermediates and the ESR detection of hydroxyl radicals appear to support the hydroxyl radical mechanism. A recent diffuse reflectance flash photolysis experiment in nonaqueous solution presented evidence in favor of the direct hole oxidation route. 27 The trapped electrons are believed to react with preadsorbed molecular oxygen to produce O 2- and O 2-2 anions. They may directly oxidize organic species, protonate to generate hydroperoxide radicals and hydroxyl radicals, or further react with more trapped electrons to eventually form water. The oxygen plays a specific role during photooxidation in addition to scavenging the trapped electron. 1.5 Limitation of TiO 2 as an efficient photocatalyst and the modification of TiO 2 Figure 1.9 Solar spectrum at sea level with the sun at zenith with the absorption region of TiO 2. TiO 2 is a quite stable photocatalyst, but since its band gap is large (E, = 3.2 ev) it is only active in the ultraviolet region which is < 10% of the overall solar intensity as 12

48 shown in Figure The limitations of a particular semiconductor as a photocatalyst for a particular use can be surmounted by modifying the surface of the semiconductor. (1) Metal semiconductor modification In photocatalysis the addition of noble metals to a semiconductor can change the photocatalytic process by changing the semiconductor surface properties. The metal can enhance the yield of a particular product or the rate of the photocatalytic reaction. 30 The addition of a metal to a semiconductor surface can also change the reaction products. Figure 1.10 Metal-modified semiconductor photocatalyst particle. Figure 1.10 is an illustration of the electron capture properties at the Schottky barrier of the metal in contact with a semiconductor surface. The picture schematically illustrates the small area of the semiconductor surface that the metal actually covers. Transmission electron microscopy measurements have found that when the added metal is Pt, the Pt particles form clusters on the surface. 31 After excitation the electron migrates to the metal where it becomes trapped and electron-hole recombination is suppressed. The migration of electrons to the metal particles was confirmed by studies showing the reduction in the photoconductance of the semiconductor for the Pt deposited TiO 2 13

49 compared to TiO 2 alone. 32 The hole is then free to diffuse to the semiconductor surface where oxidation of organic species can occur. The addition of Pt to the TiO 2 surface is beneficial for photocatalytic reactions evolving gas, especially hydrogen. The metal is important also because of its own catalytic activity and its modification on the semiconductor by changing the distribution of electrons, i.e. the decrease in electron density within the semiconductor. This leads to an increase in the hydroxyl group and in turn affects the photocatalytic process on the semiconductor surface. The addition of silver to the TiO 2 surface increases the production of H 2 from alcohol. The increase in H 2 production is attributed to the trapping of electrons at the metal sites. 33 (2) Composite semiconductors Figure 1.11 Photoexcitation in composite semiconductor photocatalyst. Coupled semiconductor photocatalysts provide an interesting way to increase the efficiency of a photocatalytic process by increasing the charge separation, and extending the energy range of photoexcitation for the system. Figure 1.11 illustrates geometrically and energetically the photoexcitation process for the composite (coupled) semiconductorsemiconductor photocatalyst CdS-TiO The energy of the excitation light is too small 14

50 to directly excite the TiO 2 portion of the photocatalyst, but it is large enough to excite an electron from the valence band across the band gap of CdS (E g = 2.5 ev) to the conduction band. The hole produced in the CdS valence band from the excitation process remains in the CdS particle while the electron transfers to the conduction band of the TiO 2 particle. The electron transfer from CdS to TiO 2 increases the charge separation and efficiency of the photocatalytic process. The separated electron and hole are then free to undergo electron transfer with adsorbates on the surface. The quantum yield for the reduction of methylviologen drastically increased and approached an optimum value of 1 when the concentration of TiO 2 was increased in a CdS-TiO 2 system. 34,35 The coupling of semiconductors with the appropriate energy levels can produce a more efficient photocatalyst via better charge separation. (3) Surface sensitization Figure 1.12 Excitation steps using dye molecule sensitizer. Surface sensitization of a wide band-gap semiconductor photocatalyst (TiO 2 ) via chemisorbed or physisorbed dyes can increase the efficiency of the excitation process. The photosensitization process can also expand the wavelength range of excitation for the photocatalyst through excitation of the sensitizer followed by charge transfer to the 15

51 semiconductor. Some common dyes which are used as sensitizers include erythrosin B, 36 thionine, 37 and analogs of Ru(bpy) ,39 Figure 1.12 illustrates the excitation and charge injection steps involved for the regenerative dye sensitizer surface process. Excitation of an electron in the dye molecule occurs to either the singlet or triplet excited state of the molecule. If the oxidative energy level of the excited state of the dye molecule with respect to the conduction band energy level of the semiconductor is favorable (i.e. more negative), then the dye molecule can transfer the electron to the conduction band of the semiconductor. The surface acts as a quencher accepting an electron from the excited dye molecule. The electron in turn can be transferred to reduce an organic acceptor molecule adsorbed on the surface. (4) Transition metal doping The influence of dissolved transition metal impurity ions on the photocatalytic properties of TiO 2 has become another interesting area of semiconductor modification. The benefit of transition metal doping species is the improved trapping of electrons to inhibit electron-hole recombination during illumination. Different metals have been employed to tune the electronic structure of TiO 2 -based material either by the ion implantation method or by a wet chemical method. 1,40-72 The photocatalytic reactivity of metal-doped TiO 2 depends on many factors, including the dopant concentration, the energy level pattern of the dopants within the TiO 2 lattice, their d-electronic configuration, and the distribuion of the dopants. Enhanced photocatalytic activity has been reported for Fe(III), Mo(V), Ru(III), Re(V), V(IV), and Rh(III) metals doped TiO 2 at 0.5% atomic ratio. It has been reported that the metal ions physically implanted have 16

52 produce significant changes in the TiO 2 material. However, metal doping can result in thermal instability and increased carrier trapping. 1,43, Non-metal doped TiO 2 for photocatalysis The visible-light response for nitrogen-doped TiO 2 was first discovered by Sato, where an NO x impurity was attributed for the sensitization of the visible light. 73 Recent theoretical and experimental studies have shown that the desired bandgap narrowing of TiO 2 can be achieved by using main group dopants to enhance photoactivity in the visible spectral range The elements used as the dopants in TiO 2 cover almost the whole periodic table and is summarized in Figure Figure 1.13 Distribution of the dopants used in the TiO 2 nanomaterials in the periodic table Synthesis of non-metal doped TiO 2 nanomaterials 17

53 The preparation methods of non-metal doped TiO 2 nanomaterials can be divided into three types: wet chemistry, high temperature treatment and ion implantation of TiO 2 nanomaterials. The wet chemistry usually involves the hydrolysis of titanium precursor, in the mixture of water and other reagents, followed with heating. We reported the preparation of nitrogen-doped TiO 2 nanomaterials by hydrolysis of titanium tetraisoporopoxide in water/amine mixture, or the postreatment on the TiO 2 sol with amines Similarly, nitrogen-doped TiO 2 was obtained by hydrolysis of titanium tretrachloride with a nitrogen source, including ammonia, ammonium carbonate, or ammonium bicarbonate by Kisch and co-worker, 82 and directly from Ti-bipyridine complex. 83 Sulfur- and Fluorinedoped TiO 2 nanomaterials were synthesized by mixing titanium isopropoxide with ethanol containing thiourea, 90,92 and H 2 O-NH 4 F mixture. 76,100 Chlorine- and Brominedoped nanomaterials were synthesized by adding titanium tretrachloride in ethanol containing HBr. 98 High temperature synthesis of doped TiO 2 nanomaterials includes the heat treatment of TiO 2 nanomaterials under different atmospheres, and the direct synthesis of doped TiO 2 materials from other titanium compounds. For example, nitrogen-, fluorinedoped TiO 2 were obtained by heating TiO 2 under NH 3 flux at o C, 74,80,84,87,90 and hydrogen fluoride, 99 and Carbon- and Sulfur-doped TiO 2 were made by heating titanium carbide, 95,96 and sulfide powder, 92,93 respectively. Direct burning titanium metal sheet in a natural gas flame can also result in Carbon-doped TiO 2 nanomaterials under proper conditions

54 Nitrogen-, Sulfur- and Fluorine-doped TiO 2 were also prepared using sputtering or ion-implanting techniques with nitrogen 84 or N + 2 gas flux, 81 F + and S + ion flux, 94,97 respectively. The chemical states between the dopants in these doped TiO 2, and the doping sources, are usually different. 74,81,94,97 Different doping methods can induce a different valence state of the dopant. For example, the incorporated S from thiourea showed S 4+ state, 90,92 while direct heating of TiS 2 or sputtering with S + induced S 2- anion Unlike the chemical doping methods, the mechanisms behind the physical doping method are somewhat elusive to chemists Property investigation methods for non-metal doped TiO 2 nanomaterials X-ray photoelectron spectroscopy (XPS) and X-ray diffraction (XRD) are two very important techniques employed in the characterization of non-metal doped TiO 2 nanomaterials. XPS is essential in the determination of the chemical composition, the chemical states of each element, the valence band structure, and the doping sites in the materials, e.g. the difference between the substitutional and the interstitial sites. XRD is essential in the determination of the crystal structure and the crystallinity, and in the estimation of the crystal grain size according to the Scherrer equation Kλ D = (1) β cosθ Where D is the size of the crystal grain, K is a dimensionless constant, 2θ is the diffraction angle, λ is the wavelength of the X-ray radiation, and β is the full width at half-maximum (FWHM) of the diffraction peak. 86 Besides, scanning electron microscopy (SEM) and transmission electron microscopy can be used to explore the morphology and crystallinity of the nanomaterials. 77,87 Rutherford backscattering (RBS) and secondary ion 19

55 mass spectrometry (SIMS) can be used to get information about the chemical composition, and elemental distribution, if the nanomaterials are prepared into films. 97 UV-visible absorption, or reflectance spectroscopy is commonly used to explore the optical properties of the materials. 74,77 And theoretical calculations are also helpful in understanding the electronic structure changes after doping, which includes spinpolarized density functional theory (DFT) within the generalized gradient approximation (GGA), 88 ab initio band calculation using the super-cell approach, 93 and the full-potential linearized augmented plane wave (FLAPW) in the framework of the local density approximation (LDA) 74 or based on the DFT within GGA. 94 Experimentally, the electronic structure can also be plotted in the action spectrum as current versus wavelength, as measured via the response in the simple photon-to-current setup. 86,87, Properties of non-metal doped TiO 2 nanomaterials Optical property of non-metal doped TiO 2 nanomaterials Figure 1.14 (A) Reflectance measurements showing the red shift in optical response due to the nitrogen doping of TiO 2 nanoparticles.[77] (B) Diffuse reflectance spectra of S- doped and pure TiO 2 powders (PT-101: rutile).[90] 20

56 Usually, after TiO 2 is doped with other non-metal elements, its color changes from white into yellow or even light gray, and the onset of absorption spectra is redshifted to a longer wavelength Figure 1.14 shows two examples of TiO 2 doped with nitrogen 77 and sulfur, 90 respectively. The nitrogen-doped TiO 2 nanomaterials had its absorption extended up to 600 nm, and the band gap absorption onset shifted from 380 to 600 nm. 54 The sulfur-doped TiO 2 also displays strong absorption in the region from 400 nm to 600 nm. The redshift in the absorption spectrum of doped TiO 2 reflects the narrowing of band gap in the electronic structure, and this enhances its photosensitivity, i.e. photocatalytic, photochemical, and photoelectrochemical activities, in the visible region, as expected Electronic properties of non-metal doped TiO 2 nanomaterials As stated above, the optical property of the material is eventually determined by its underlying electronic structure. The electronic band structures shown in Figure 1.15, i.e. the density of states (DOSs) of anatase TiO 2 with different substitutional dopants, including C, N, F, P or S, were calculated by Asahi and co-workers, using the fullpotential linearized augmented plane wave (FLAPW) in the framework of the local density approximation (LDA). 74 They found that substitutional doping of N was the most effective because its p states contribute to the band-gap narrowing by mixing with O 2p states. The substitutional site N doping was found more efficient than the molecularly existing species, e.g. NO and N 2 dopants, which give rise to the bonding states below the O 2p valence bands and antibonding states deep in the band gap (N i and N i+s ), and are well screened and hardly interact with the band states of TiO 2. However, a different 21

57 suggestion as to the origin of the band structure changes by nitrogen doping was recently proposed by Valentin and co-workers, where for nitrogen doping in both anatase and rutile polymorphs, N 2p localized states were found just above the top of the O 2p valence band. 88 Interestingly, they found that in anatase, these dopant states caused a Figure 1.15 (A) Total DOSs of doped TiO 2 and (B) the projected DOSs into the doped anion sites, calculated by FLAPW, for the dopants F, N, C, S, and P located at a substitutional site for an O atom in the anatase TiO 2 crystal (the eight TiO 2 units per cell). The results for N doping at an interstitial site (N i -doped) and that at both substitutional and interstitial sites (N i+s -doped) are also shown. The energy is measured from the top of the valence bands of TiO 2, and the DOSs for doped TiO 2 are shifted so that the peaks of the O 2s states (at the farthest site from the dopant) are aligned with each other. Arb. unit, arbitrary units. [74] redshift of the absorption band edge towards visible region, while in rutile, an overall blueshift was found by the N-induced contraction of the O 2p band. 88 Recently, experimental evidence was found to support the statement that nitrogen-doped TiO 2 has nitrogen-induced midgap level, formed slightly above the oxygen 2p valence band, and that this level is being responsible for the visible-light response

58 Also from Asahi s results, C and P dopants introduce deep states in the gap, and S dopant induces a similar band-gap narrowing as nitrogen. The latter was also shown from the calculations of Umebayashi and co-workers, where the mixing of the sulfur 3p states with the valence band was found to contribute to the increased width of the valence band, leading to the narrowing of the band gap. 93,94 In the case where S exists as S 4+, replacing Ti 4+, sulfur 3s induce states just above the O 2p valence states, and S 3p states Figure 1.16 (A) Total DOS of: (A) undoped, (B) S-doped TiO 2, (C) partial DOS of S atoms in S-doped TiO 2.[92] (B) Total DOSs of (a) pure and (b) F-doped TiO 2 calculated by FLAPW. The dopant F is located at the substitutional site for an O atom in the rutile TiO 2 crystal (the two TiO 2 unit cells). The energy on the horizontal axis is measured from the top of the VBs. E g indicates the (effective) bandgap energy of the semiconductors. The impurity states are labeled (I) and (II).[97] contribute to the conduction band of TiO 2 as shown in Figure 1.16A. 92 From the calculations by Yamaki and co-workers as shown in Figure 1.16B, 97 when F replace the O in the TiO 2 lattice, the F 2p states are localized below the O 2p valence states without any mixing with the valence or conduction band, and they were expected not to contribute to the optical absorption spectrum. Additional states appeared just below the conduction edge, which originated from the occurrence of the electron occupied level 23

59 composed of the t 2g state of the Ti 3d orbital. The electronic change induced by F dopant was considered to be similar to the O vacancy, thus reducing the effective bandgap and visible-light photoresponse Photoelectrochemical properties of non-metal doped TiO 2 nanomaterials The narrowing band gap of doped TiO 2 can be measured by a photo-to-carrier conversion set-up. 86,87,94 Light from a xenon lamp was irradiated onto the sample after passing through a monochromator. The photocurrents from the film electrodes made with doped TiO 2 were measured as a function of wavelength. 86,87,94 The incident photo-tocurrent efficiency, IPCE λ, can be calculated from IPCE λ hc I ph, λ = (2) e Pλλ where I ph,λ is the photocurrent, P λ is the power intensity of the light at wavelength λ, and h, c, and e are Planck s constant, speed of light and elementary charge, respectively. 86 Figure 1.17 (A) IPCE vs λ for (Δ) N-doped TiO 2 and ( ) nondoped TiO 2 in 0.1 M HClO 4 at 0.5 M vs Ag/AgCl. The inset shows an expanded plot of IPCE vs λ in the visible-light region.[87] (B) Photocurrent spectra of the (a) pure TiO 2 and (b) S-doped TiO 2 synthesized by ion implantation and subsequent annealing. [94] 24

60 IPCE λ is called an action spectrum. Figure 1.17A shows such a action spectrum, where nitrogen-doped TiO 2 displayed a higher response in the visible region than pure TiO Figure 1.17B shows the photocurrent spectra for the pure and S-doped crystals. The photocurrent spectrum edge shifted to the low-energy region below 2.9 ev for the S- doped crystal, compared to 3.0 ev for pure TiO 2, due to the transition of electrons across the narrowed bandgap between the VB and CB Photocatalytic property of non-metal doped TiO 2 nanomaterials The visible-light response of the non-metal doped TiO 2 nanomaterials have been demonstrated with much higher photocatalytic properties than pure TiO 2 nanomaterials, Figure 1.18 Photocatalytic properties of TiO 2-x N x samples ( ) compared with TiO 2 samples ( ). (A) Decomposition rates [measuring the change in absorption of the reference light (Δabs)] of methylene blue as a function of the cutoff wavelength of the optical high-path filters under fluorescent light. (B) Decomposition rates of methylene blue in the aqueous solution under visible light as a function of the ratio of the decomposed area in the XPS spectra with the peak at 396 ev to the total area of N 1s. The total N concentrations for the powder samples were evaluated to be 1.0 atomic %, a; 1.1 atomic %, b; 1.4 atomic %, c; 1.1 atomic %, d; and 1.0 atomic %, e. [74] 25

61 especially in the visible-light region. 74,77,78,80,81,83,84,89 Figure 1.18 shows the results on the decomposition of methylene blue using nitrogen-doped TiO 2 conducted by Asahi and coworkers. 74 It was found that nitrogen-doped TiO 2 had much higher photocatalytic activity than pure TiO 2 in the visible-light region, while a lower activity in the UV-light region. And the photocatalytic activity of the nitrogen-doped TiO 2 showed nitrogen concentration dependent performance in the visible region. 74 The nitrogen concentration dependent photocatalytic activity of the nitrogen-doped TiO 2 was also found in the study of Irie and co-workers, 96 where they also suggested that the band structure of the nitrogen-doped TiO 2 with a lower nitrogen concentration (< 2%) was different from that with a higher concentration. The photocatalytic oxidation of organic compounds in nitrogen-doped TiO 2 under visible illumination was suggested by Nakamura and coworkers, mainly via reactions with surface intermediates of water oxidation or oxygen reduction, not by direct reactions with holes trapped at the nitrogen-induced midgap level. 87 The photocatalytic activity of sulfur-doped TiO 2 has also been studied, where the sulfur-doped TiO 2 displayed a higher photocatalytic activity in the visible region, but a lower photocatalytic activity in the UV region. 90,91 The carbon-doped TiO 2 made from TiC precursor demonstrated a noticeable photocatalytic activity in the visible region from the study of Choi, Irie and co-workers. 95,96 The carbon-doped TiO 2 made by pyrolyzing Ti metal in a natural gas flame displayed a much higher photoactivity compared to pure TiO In the photodecomposition study of acetone, Fluorine-doped TiO 2 showed a higher photocatalytic activity than that of Degussa P25 under proper preparation conditions. 76 And the chlorine and bromine co-doped TiO 2 displayed a much higher 26

62 photocatalytic activity than chlorine- or bromine-doped TiO 2 and P25 in the study carried out by Luo and co-workers Specific aims of this dissertation From the above discussion, in order to shift the absorption property of TiO 2 from UV into the visible regime, different non-metal elements can be introduced into the lattice of TiO 2. Thus, one goal of this dissertation is the synthesis and property investigations (including electronic, optical and catalytic) of doped TiO 2 nanoparticles with different elements (C, N, S). The interest in the nanoscale-dimension materials stems from the fact that new properties are acquired at this length scale, and equally important, that these properties change with their size or shape. The change in the properties at this length scale is not a result of scaling factors. It results from different causes in different materials. In semiconductors, it results from the further confinement of the electronic motion to a length scale comparable or smaller than the length scale characterizing the electronic motion in bulk semiconducting material (called the electron Bohr radius). For semiconductors, when the size of materials is decreased to less than 10 nanometer, besides the composition of the material, there are two more important aspects, which must be taken into account. First, the intrinsic properties of the materials are influenced by the quantum size confinement or the quantum size effect, especially, when the size is comparable to or less than the Bohr exciton radius, the natural length scale of the electron-hole pair of the bulk material. This confinement induces discrete electronic states in the valence and conduction bands in the quantum dots compared to the 27

63 continuum states of the corresponding bulk materials and an increase of the band edge transition energy as the particle size is reduced. Second, the surface atoms play a more important role as the size of the material decreases. The surface atoms usually have fewer adjacent coordinate atoms and unsaturated sites or more dangling bonds. Thus these atoms are chemically more active compared to the bulk atoms, which renders them to be more accessible for chemical modification. Thereby, the additional surface electronic states of the nanoparticle modify the electronic structure of the particle. Particularly, they induce extra electronic states in the band gap, which act as electron or hole trap centers. As stated above, the surface has a large effect on the properties of nanomaterials. Thus the other goal is the synthesis and property investigations of II-VI (CdSe, CdSe/CdS) semiconductor nanoparticles to elucidate the surface effect on the property of nanoparticles, which helps to understand the photocatalytic properties of TiO 2 nanomaterials, since the main catalytic reactions occur on the surface. 1.8 Reference 1 Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W. Chem. Rev. 1995, 95, Grätzel, M. Nature, 2001, 414, Millis, A.; Hunte, S. L. J. Photochem. Photobiol. A, 1997, 108, 1. 4 Linsebigler, A. L.; Lu, G.; Yates Jr., J. T. Chem. Rev. 1995, 95, Fujishima, A.; Rao, T. N.; Tryk, D. A J. Photochem. Photobio. C 2000, 1, 1. 6 Tryk, D. A.; Fujishima, A.; Honda, K. Electrochimica Acta 2000, 45,

64 7 Ireland, J. C.; Klostermann, P.; Rice, E. W.; Clark, R. M. Appl. Environ. Microbiol. 1993, 59, Sjogren, J. C.; Sierka, R. A. Appl. Enuiron. Microbiol. 1994, 60, Cai, R. X.; Kubota, Y.; Shuin, T.; Sakai, H.; Hashimoto, K; Fujishima, A. Cancer Res. 1992, 52, Cai, R.; Hashimoto, K.; Kubota, Y.; Fujishima, A. Chem. Lett. 1992, Karakitsou, K. E.; Verykios, X. E. J. Phys. Chem. 1993, 97, Gratzel, M. ACC. Chem. Res. 1981, 14, Kalyanasundaram, K; Borgarello, E.; Duonghong, D.; Gratzel, M. Angew. Chem., Int. Ed. Engl. 1981, 20, Duonghong, D.; Borgarello, E.; Gratzel, M. J. Am. Chem. Soc. 1981, 103, Borgarello, E.; Kiwi, J.; Pelizzetti, E.; Visca, M.; Gratzel, M. Nature 1981, 289, Wold, A. Chem. Mater. 1993, 5, Gerischer, H.: Heller, A. J. Electrochem. Soc. 1992, 139, Jackson, N. B.; Wang, C. M.; Luo, 2.; Schwitzgebel, J.; Ekerdt, J. G.; Brock, J. R.; Heller, A. J. Electrochem. Soc. 1991, 138, Nair, M.; Luo, Z. H.; Heller, A. Ind. Eng. Chem. Res. 1993, 32, Matthews, R. W. J. Catal. 1988, 113, Schiavello, M. Heterogeneous photocatalysis, New York: Wiley, Jaeger, C. D.; Bard, A. J. J. Phys. Chem. 1979, 83, Duonghong, D.; Borgarello, E.; Gratzel, M. J. Am. Chem. Soc. 1981, 103, Kawai, T.; Sakata, T. J. Chem. Soc., Chem. Commun. 1980, Ollis, D. F.; Hsiao, C. Y.; Budiman, L.; Lee, C.-L. J. Catal. 1984, 88,

65 26 Matthews, R. W. J. Catal. 1986, 97, Draper, R. B.; Fox, M. A. Langmuir 1990, 6, Turchi, C. S.; Ollis, D. F. J. Catal. 1989, 119, Turchi, C. S.; Ollis, D. F. J. Catal. 1990, 122, Sato, S.; White, J. M. Chem. Phys. Lett. 1980, 72, Pichat, P.; Mozzanega, M.-N.; Disdier, J.; Herrmann, J.-M. Nouu. J. Chim. 1982, 11, Disdier, J.; Herrmann, J.-M.; Pichat, P. J. Chem. Soc. Faraday Tran , 79, Sclafani, A.; Mozzanega, M.-N.; Pichat, P. J. Photochem. Photobiol.A: Chem. 1991, 59, Gopidas, K. R.; Bohorquez, M.; Kamat, P. V. J. Phys. Chem. 1990, 94, Spanhel, L.; Weller, H.; Henglein, A. J. Am. Chem. SOC. 1987, 109, Kamat, P. V.; Fox, M. A. Chem. Phys. Lett. 1983, 102, Patrick, B.; Kamat, P. V. J. Phys. Chem. 1992, 96, Vlachopoulos, N.; Liska, P. Augustynski, J.; Gratzel, M. J. Am. Chem. Soc. 1988, 110, Desilvestro, J.; Gratzel, M.; Kavan, L.; Moser, J.; Augustynski, J. J. Am. Chem. Soc. 1986, 107, Wang, C.; Bahnemann D.W.; Dohrmann, J. K. Chem. Comm. 2000, 16, Wang, Y.; Hao, Y.; Cheng, H.; Ma, H.; Xu, B.; Li, W.; Cai, S. J. Mater. Sci. 1999, 34, Coloma, F.; Marquez, F.; Rochester, C. H.; Anderson, J. A. Phys. Chem. Chem. Phys. 2000, 2,

66 43 Yamashita, H.; Honda, M.; Harada, M.; Ichihashi, Y.; Anpo, M.; Hirao, T.; Itoh, N.; Iwamoto, N. J. Phys. Chem. B 1998, 102, Wang, Y.; Cheng, H.; Hao, Y.; Ma, J.; Li, W.; Cai, S. Thin Solid Films 1999, 349, Moser, J.; Grätzel, M.; Gallay, R. Helu. Chim. Acta 1987, 70, Grätzel, M.; Howe, R. F. J. Phys. Chem. 1990, 94, Soria, J.; Conesa, J. C.; Augugliaro, V.; Palmisano, L.; Schiavello, M.; Sclafani, A. J. Phys. Chem. 1991, 95, Sclafani, A.; Palmisano, L.; Schiavello, M. Res. Chem. Intermed. 1992, 18, Palmisano, L.; Augugliaro, V.; Sclafani, A.; Schiavello, M. J. Phys. Chem. 1988, 92, Choi, W.; Termin, A.; Hoffmann, M. R. J. Phvs. Chem. 1994, 98, Choi, W. Y.; Termin, A.; Hoffmann, M. R. Angew. Chem. 1994, 106, Choi, W.; Termin, A.; Hoffmann, M. R. Angew. Chem., Int. Ed. Engl. 1994, 33, Martin, S. T.; Herrmann, H.; Hoffmann, M. R. Trans. Faraday Soc. 1994, 90, Borgarello, E.; Kiwi, J.; Grätzel, M.; Pelizzetti, E.; Visca, M. J. Am. Chem. Soc. 1982, 104, Herrmann, J. M.; Disdier, J.; Pichat, P. Chem. Phys. Lett. 1984, 108, Mu, W.; Herrmann, J. M.; Pichat, P. Catal. Lett. 1989, 3, Luo, Z.; Gao, Q. H. J. Photochem. Photobiol. A: Chem. 1992, 63, Karakitsou, K. E.; Verykios, X. E. J. Phys. Chem. 1993, 97, Peng, S.; Li, Y.; Jiang, F.; Lu, G.; Li, S. Chem. Phys. Lett. 2004, 398,

67 60 Gracia, F.; Holgado, J. P.; Caballero, A.; Gonzalez-Elipe, A. R. J. Phys. Chem. B ACS ASAP. 61 Yang, Q. -J.; Xu, Z. -L.; Xie, C.; Xue, B. -Y.; Du, Y. -G.; Zhang, J. -H.; Gaodeng Xuexiao Huaxue Xuebao 2004, 25, Liu, Y.; Liu, C.-Y.; Rong, Q.-H.; Zhang, Z.; Wang, C.-Y. Res. Chem. Interm. 2004, 30, Bessekhouad, Y.; Robert, D.; Weber, J. -V.; Chaoui, N. J. Photochem. Photobio. A 2004, 167, Sakatani, Y.; Ando, H.; Okusako, K.; Koike, H.; Nunoshige, J.; Takata, T.; Kondo, J. N.; Hara, M.; Domen, K. J. Mater. Res. 2004, 19, Li, F. B.; Li, X. Z.; Hou, M. F. Appl. Catal. B 2004, 48, Cao, Y.; Yang, W.; Zhang, W.; Liu, G.; Yue, P. New J. Chem. 2004, 28, Wang, C.; Xi, H.; Wang, R. Chem. Lett. 2004, 33, Li, W.; Wang, Y.; Lin, H.; Shah, S. I.; Huang, C. P.; Doren, D. J.; Rykov, S. A.; Chen, J. G.; Barteau, M. A. Appl. Phys. Lett. 2003, 83, Li, J.; Luo, S.; Yao W.; Zhang, Z. Mater. Lett. 2003, 57, Frindell, K. L.; Bartl, M. H.; Robinson, M. R.; Bazan, G. C.; Popitsch; A.; Stucky, G. D. J. Solid State Chem. 2003, 172, Wang, W. Y.; Zhang, D. F.; Chen, X. L. J. Mater. Sci. 2003, 38, Xu, X. H.; Wang, M.; Hou, Y.; Yao, W. F.; Wang, D.; Wang, H. J. Mater. Sci. Lett. 2002, 21, Sato, S. Chem. Phys. Lett. 1986, 123, Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293,

68 75 Khan, S. U. M.; Al-Shahry, M.; Ingler, Jr. W.B. Science 2002, 297, Yu, J. C.; Yu, J. G.; Ho, W. K.; Jiang, Z. T.; Zhang, L. Z. Chem. Mater. 2002, 14, Burda, C.; Lou, Y.; Chen, X.; Samia, A. C. S.; Stout, J.; Gole, J. L. Nano Lett. 2003, 3, Gole, J. L.; Stout, J. D.; Burda, C.; Lou, Y.; Chen, X. J. Phys. Chem. B 2004, 108, Chen, X.; Burda, C. J. Phys. Chem. B 2004, 108, Diwald, O.; Thompson, T. L.; Goralski, E. G.; Walck, S. D.; Yates, Jr. J. T. J. Phys. Chem. B 2004, 108, Diwald, O.; Thompson, T. L.; Zubkov, T.; Goralski, E. G.; Walck, S. D.; Yates, Jr. J. T. J. Phys. Chem. B 2004, 108, Sakthivel, S.; Kisch, H. ChemPhysChem 2003, 4, Sano, T.; Negishi, N.; Koike, K.; Takeuchi, K.; Matsuzawa, S. J. Mater. Chem. 2004, 14, Irie, H.; Watanabe, Y.; Hashimoto, K. J. Phys. Chem. B 2003, 107, Lindgren, T.; Lu, J.; Hoel, A.; Granqvist, C. G.; Torres, G. R.; Lindquist, S. -E. Solar Energy Mater. Solar Cells 2004, 84, Lindegren, T.; Mwabora, J. M.; Avendaňo, J. J.; Hoel, A.; Granqvist, C. G.; Lindquist, S. -E. J. Phys. Chem. B 2003, 107, Nakamura, R.; Tanaka, T.; Nakato, Y. J. Phys. Chem. B 2004, 108, Valentin, C. D.; Pacchioni, G.; Selloni, A. Phys. Rev. B 2004, 70, Wei, H.; Wu, Y.; Lun, N.; Zhao, F. J. Mater. Sci. 2004, 39,

69 90 Ohno, T.; Mitsui, T.; Matsumura, M. Chem. Lett. 2003, 32, Ohno, T.; Akiyoshi, M.; Umebayashi, T.; Asai, K.; Mitsui, T.; Matsumra, M. Appl. Catal. A. 2004, 265, Umebayashi, T.; Yamaki, T.; Tanaka, S.; Asai, K. Chem. Lett. 2003, 32, Umebayashi, T.; Yamaki, T.; Itoh, H.; Asai, K. Appl. Phys. Lett. 2002, 81, Umebayashi, T.; Yamaki, T.; Yamaki, T.; Yamamoto, S.; Miyashita, A.; Sumita; T. Asai, K. J. Appl. Phys. 2003, 93, Choi, Y.; Umebayashi, T.; Yoshikawa, M. J. Mater. Sci. 2004, 39, Irie, H.; Watanabe, Y.; Hashimoto, K. Chem. Lett. 2003, 32, Yamaki, T.; Umebayashi, T.; Sumita, T.; Yamamoto, S.; Maekawa, M.; Kawasuso, A.; Itoh, H. Nucl. Instr. Meth. Phys. Res. 2003, 206, Luo, H.; Takata, T.; Lee, Y.; Zhao, J.; Domen; K.; Yan, Y. Chem. Mater. 2004, 16, Subbarao, S. N.; Yun, Y. H.; Kershaw, R.; Dwinghta; K.; Wold, A. Inorg. Chem. 1979, 18, Hattori, A.; Yamamoto, M.; Tada, H.; Ito, S. Chem. Lett. 1998, 27, Yamaki, T.; Sumita, T.; Yamamoto, S. J. Mater. Sci. Lett. 2002, 21, Yang, S.; Gao, L. J. Am. Ceram. Soc. 2004, 87, Zhao, X.-F.; Zhang, Z.-H.; Meng, X.-F.; Liu, L.; Jia, D.-Z. Fenzi Cuihua 2003, 17, Yu, J. -G.; Yu, J. C.; Cheng, B.; Hark, S. K.; Lu, K. J. Solid State Chem. 2003, 174,

70 CHAPTER 2 Photocatalysis Nitrogen Doped TiO 2 Nanoparticles for Visible-Light * Abstract Nitrogen doped TiO 2 : TiO 2-x N x nanoparticles were prepared by employing the direct amination of 6-10 nm sized titania. Doping at the nanometer scale led to an enhanced nitrogen concentration of up to 8%, compared to 2% in thin films and micrometer scale TiO 2 powders. The synthesized TiO 2-x N x nanocrystals are catalytically active and absorb well into the visible region up to 600 nm, thus exemplifying the use of nanostructure-based synthesis as a means of producing novel photocatalytic materials. * This work is in collaboration with Prof. James Gole at the School of Physics, Georgia Institute of Technology, and this chapter has been partially published in Nano Lett. 2003, 3(8),

71 2.1 Introduction The efficient utilization of solar energy is one of the major goals of modern science and engineering that will have a great impact on technological applications. 1-5 Of the materials being developed for photocatalytic applications, titanium dioxide (TiO 2 ) remains the most promising because of its high efficiency, low cost, chemical inertness, and photostability However, the widespread technological use of TiO 2 is impaired by its wide bandgap (3.2 ev), which requires ultraviolet irradiation for photocatalytic activation. Since UV light accounts for only a small fraction (8 %) of the sun s energy compared to visible light (45%), any shift in the optical response of TiO 2 from the UV to the visible spectral range will have a profound positive effect on the photocatalytic efficiency of the material. 11 Here we report, a simple nitrogen-doping method for nanometer-sized visible-light TiO 2 photocatalysts. The prepared photocatalysts show an enhancement in the photodegradation efficiency of methylene blue under visible light (wavelength 390 nm) irradiation compared to commercially available TiO 2 catalyst. An initial approach to shift the optical response of TiO 2 from the UV to the visible spectral range has been the doping of TiO 2 with transition metal elements However, metal doping has several drawbacks. The doped materials have been shown to suffer from thermal instability and the metal centers act as electron traps, which reduce the photocatalytic efficiency. Furthermore, the preparation of transition metal doped TiO 2 requires more expensive ion-implantation facilities. 19,20 Recently, it was shown that the desired bandgap narrowing of TiO 2 can be better achieved by using anionic dopant species rather than metals ions. 11,21,22 Substitutional doping of nitrogen was found to be most effective because its p states contribute to the bandgap narrowing by mixing with the O 2p states. Recently, Asahi et al. showed that TiO 2 films can be doped with nitrogen 36

72 by sputtering methods, and exhibit thereafter enhanced photoactivity in the visible spectral range. 11 Considerable efforts have been undertaken to dope TiO 2 thin films and powders with nitrogen by annealing TiO 2 at elevated temperature under NH 3 flow for several hours. Nevertheless, the doping process on these micron-sized TiO 2 systems resulted in only small amounts ( 2%) of nitrogen incorporation. 11 We have developed an alternative nanoscale synthesis route, which leads to an increased nitrogen dopant concentration (up to 8%) in titania. TiO 2-x N x was synthesized at room temperature, by employing the direct amination of TiO 2 nanoparticles. The synthesized TiO 2-x N x nanoparticles are photocatalytically active with an absorbance that extends into the visible region up to 600 nm. 2.2 Experimental Preparation of nitrogen doped TiO 2 nanoparticles Small TiO 2 nanocrystals were prepared by the controlled hydrolysis of titanium (IV) isopropoxide in water, under controlled ph. 23 By adjusting the ph of the solution, TiO 2 nanocrystals in the 3 to 10 nm size range can be synthesized as transparent colloidal solutions, which are stable for extended periods. 23 To introduce nitrogen dopant into the titania nanoparticles, triethylamine is added to the colloidal nanoparticle solution. The addition of amine to the nanoparticle solution results in the formation of yellow nanocrystals (mean diameter of 10 nm). Synthesis of the TiO 2 nanocrystals entails the drop wise addition (1 ml / min) of a 5 ml aliquot of Ti[OCH(CH 3 ) 2 ] 4 (Aldrich, 97%) dissolved in isopropyl alcohol (5:95) to a 900 ml doubly distilled water (2 o C) adjusted to ph 2 by HNO 3 addition. After continuous stirring of the reaction mixture for 12 hours, a colloidal solution of TiO 2 nanocrystals is formed. Using dynamic light scattering (Coutler Plus 4), it has been 37

73 demonstrated that the size of the nanoparticles depends critically on the amount of HNO 3 added to the reaction mixture. By controlling the amount of added nitric acid, TiO 2 nanoparticles of sizes ranging from 3 to 10 nm can be synthesized, which are stable for extended periods under refrigeration. Treatment of the initial nanoparticle solution and gel with an excess of triethylamine results in the formation of a yellowish solution. The solution was treated directly with an alky ammonium compound to facilitate nitrogen incorporation. Upon vacuum drying (5 x 10-2 Torr) for several hours, the treated nanoparticle solution forms deep yellow crystallites. Characterization The X-ray diffraction (XRD) patterns were obtained using a Philips PW 3710 X-ray powder diffractometer. The reflectance spectra were obtained on a Cary 50 UV-visible spectrometer equipped with a reflectance unit. The low resolution TEM images were taken on a JEOL 1200EX transmission electron microscope operated at 80 kv. High-resolution transmission electron microscopy (HRTEM) images were obtained on a Tecnai F30 HRTEM machine operated under 300 kv. Samples for TEM were prepared by depositing a drop of the nanocrystal solution in water onto a copper grid supporting a thin film of amorphous carbon, and drying in air. The chemical compositions were determined with X-ray photoelectron spectroscopy (XPS), taken on a Perkin-Elmer PHI 5600 XPS System with the samples on a carbon tape sticking to the aluminum support. The XPS binding energies were calibrated with respect to the C 1s peak from the carbon tape at ev. The photocatalytic activity of the TiO 2-x N x particles was evaluated at excitation wavelengths of 390 and 540 nm using a Clark MXR 2001 fs laser system. The laser beam (800 fs, 1 khz, 120 fs laser pulse train) was sent either through a BBO crystal to generate second harmonic 390 nm (10 mw) light pulses or to an optical parametric amplifier to generate stable 540 nm (4 mw) pulses. The laser 38

74 light intensity was adjusted with a neutral density filter wheel. The pulse train was guided into a quartz cuvette filled with a 2 ml aqueous solution of methylene blue (optical density) 0.8) and 10 mg of the new catalyst, to excite a pump volume of about 5 nl (0.5 mm is the diameter of the excitation beam at the reaction cell). The decomposition of the solute was followed by measuring the decolorization of the methylene blue in solution with a Varian Cary Bio50 UV-visible spectrometer. The reported quantum yield for decomposition was based on the analysis of the photon flux, taking into account the volume of the excitation region and including the dilution factor for evaluating the optical density changes depicted. The process requires low intensity when compared with high-intensity sunlight. 2.3 Results and Discussion Intensity / a.u Theta / o Figure 2.1 XRD spectrum of a TiO 2-x N x nanoparticle sample with an average diameter of 10 nm. 39

75 The XRD spectrum shown in Figure 2.1 and the HRTEM micrograph shown in Figure 2.2 demonstrate that the treated nanostructures are of the anatase crystalline phase. Both the XRD and HRTEM show that the doped TiO 2 nanomaterials were highly crystallized, indicated by the high-intensity of the XRD diffraction peaks and the wellresolved lattice fringes in the HRTEM image. B 2 nm 100 nm A Figure 2.2 TEM images of TiO 2-x N x nanocolloid particles (A) at low resolution (bar size 100 nm), (B) a high resolution image of the region in the indicated area (bar size 2 nm) inset on the left side showing the projection of the polycrystalline anatase phase along (001) direction. O 1s Intensity / a.u. O KLL Ti 2s Ti 2p C 1s Atomic % Element O(1s) 64 N(1s) 8 Ti(2p) 28 N 1s Ti 3s Ti 3p Binding Energy / ev Figure 2.3 XPS spectrum of a TiO 2-x N x nanoparticle sample with an average diameter of 10 nm. 40

76 The XPS spectrum of the TiO 2-x N x nanoparticle sample (Figure 2.3) shows that up to 8 % atomic ratio of nitrogen has been incorporated into the lattice of TiO 2 host. The change in color of the nanocrystals upon nitrogen incorporation demonstrates a profound effect on their optical response in the visible wavelength range. In contrast to the nanoparticle reactivity, which we have described above, no significant reaction was observed when TiO 2 micropowders were treated with triethylamine. Furthermore, the treatment of Degussa P25 nanopowder (particle size 30 nm) resulted in a much slower doping reaction. 100 a % Reflectance b Wavelength / nm TiO 2 nanoparticles TiO 2-x N x nanoparticles Figure 2.4 Reflectance measurements showing the red-shift in optical response due to the nitrogen doping of TiO 2 nanoparticles. Shown in Figure 2.4 are the UV-visible reflectance spectra of pure TiO 2, and nitrogen-doped TiO 2 nanoparticles. The reflectance measurements on the doped TiO 2 nanoparticles show that the bandgap absorption onset of the nanocrystals shifted from 380 to 600 nm. The figure compares the optical reflectance spectra of commercial 41

77 Degussa P25 TiO 2, onsetting at 380 nm (spectrum a) and the reflectance spectrum for TiO 2-x N x nanocrystals with size 10 nm, rising at 600 nm (spectrum b) OD (Methylene Blue) W avelength / n m OD (Methylene 650 nm) (b) λ ex : 540 nm OD 0.6 OD no nanocatalyst TiO 2 nanoparticles TiO 2-x N x nanoparticles (a) λ ex : 390 nm Reaction time / min no nanocatalyst TiO 2 nanoparticles TiO 2-x N x nanoparticles Reaction time / min Figure 2.5 Comparison of the photocatalytic decomposition of methylene blue in presence of doped and undoped titania nanoparticles, as monitored by the changes in absorbance at 650 nm after (a) 390 nm laser excitation and (b) 540 nm excitation. The inset in 3a shows the photodegradation of methylene blue in water at neutral ph. The photocatalytic activity of the nanoparticles was evaluated by measuring the decomposition of methylene blue at 650 nm, upon photoexcitation with 390 and 540 nm visible light. Figure 2.5 shows the photodegradation of methylene blue in water at neutral ph. The nitrogen-doped nanoparticles showed enhanced photocatalytic activity. On the other hand, the undoped TiO 2 nanoparticles (Degussa P25), did not show much activity under visible-light radiation compared to the reference experiment without nanoparticles. The observed gradual decrease in absorption of methylene blue through time is attributed to the direct decomposition of the dye upon laser irradiation, and is not due to sensitization via the Degussa P25 TiO 2 nanocrystals. 24 The difference in the photocatalytic activity of the nanoparticles observed after 390 and 540 nm excitation can be correlated to their reflectance spectra presented in 42

78 Figure 2.4. At 390 nm a larger difference in optical response is observed, which explains the significant difference in the photocatalytic activity at 390 nm irradiation (Fig. 2.5a). On the other hand, at wavelengths > 500 nm the differences in the optical responses are smaller, hence the photocatalytic activity under 540 nm irradiation is less pronounced. Nevertheless, the nitrogen doped nanocrystals still exhibited a higher photocatalytic activity. 2.4 Conclusions In conclusion this study demonstrates the effectiveness of using nanometer scaled materials in developing efficient visible-light activated photocatalysts by doping. Furthermore, it is shown that modification of the TiO 2 photocatalyst can be realized by using a simple, room temperature synthesis process, which results in enhanced nitrogen incorporation that is less efficient at the micrometer scale. This study could provide a pathway for the production of environmentally benign photocatalysts, which exceed the efficiency of current catalysts, particularly for visible light activation. 2.5 References 1. Fujishima, A.; Honda, K. Nature 1972, 238, Schiavello, M. Photoelectrochemistry, Photocatalysis, and Photoreactors: Fundamentals and Developments, NATO ASI Series; D. Reidel Publishing Company: Dordrecht, Ollis, D.S.; Al-Ekabi, H. Photocatalytic Purification and Treatment of Water and Air; Elsevier: Amsterdam,

79 4. Khan, S.U.M.; Akikusa, J. J. Phys. Chem. B 1999, 103, Licht, S.; Wang, B.; Mukerji, S.; Soga, T.; Umeno, M.; Tributsch, H. J. Phys. Chem. B. 2000, 104, Wilcoxon, J.P. Photocatalysis Using Semiconductor Nanoclusters, Advanced Catalytic Materials, MRS Proc. Boston, MA, Serpone, N.; Pelizzetti, E. Photocatalysis: fundamentals and applications; Wiley: New York, Kozhukharov, V.; Vitanov, P.; Stefchev, P.; Kabasanova, E.; Kabasanov, K;. Machkova, M.; Blaskov, V.; Simeonov, D.; Tzaneva, G. J. Environ. Protec. Eco. 2001, 2, Schiavello, M.; Dordrecht H., Eds. Photoelectrochemistry, photocatalysis, and photoreactors: fundamentals and developments; Kluwer Academic: Boston, Linsebigler, A.L.; Lu, G.; Yates, J.T. Chem. Rev. 1995, 95, (a). Asahi, R.; T. Morikawa, T. Ohwaki, K. Aoki, Y. Taga, Science 2001, 293, 269. (b). Horishi, I.; Wanatabe, Y.; Hashimoto, K. J. Phys. Chem. B Shah, S.I.; Li, W.; Huang, C.P.; Jung, O.; Ni, C. Proc. Natl. Acad Sci. U.S.A. 2002, 99, Xu, A.; Zhu, J.; Gao, Y.; Liu, H. Chem. Res. Chin. Univ. 2001, 17, Wang, C.; Bahnemann, D.W.; Dohrmann, J.K. Chem. Comm. 2000, 16,

80 15. Wang, Y.; Hao, Y.; Cheng, H.; Ma, H.; Xu, B.; Li, W.; Cai, S. J. Mater. Sci. 1999, 34, Coloma, F.; Marquez, F.; Rochester, C.H.; Anderson, J.A. Phys.Chem. Chem. Phys. 2000, 2, Altynnikov, A.A.; Zenkovets, G.A.; Anufrienko, V.F. React. Kin. Cat. Lett. 1999, 67, Umebayashi, T.; Yamaki, T.; Itoh, H.; Asai, K. J. Phys. Chem. Sol. 2002, 63, Wang, Y.; Cheng, H.; Hao, Y.; Ma, J.; Li, W.; Cai, S. Thin Solid Films 1999, 349, Yamashita, H.; Honda, M.; Harada, M.; Ichihashi, Y.; Anpo, M.; Hirao, T.; Itoh, N.; Iwamoto, N. J. Phys. Chem. B 1998, 102, Yu, J.C.; Yu, J.G.; Ho, W.K.; Jiang, Z.T.; Zhang, L.Z. Chem. Mat. 2002, 14, Khan, S.U.M.; Al-Shahry, M.; Ingler, Jr. W.B. Science 2002, 297, Weng, Y.; Wang, Y.; Asbury, J.B.; Ghosh, H.N.; Lian, T. J. Phys. Chem. B 2000, 104, We have observed clear evidence for the photodegradation of methylene blue through the so-called sensitization process in the visible range at λ> 500 nm (K. Vinodgopal, et al. Environ. Sci. Technol. 1996, 30, 1660). Note that the data depicted in Figure 3 obtained for TiO 2 have a significant component due to the direct photodegredation of methylene blue. 45

81 CHAPTER 3 Formation of Oxynitride as the Photocatalytic Enhancing site in Nitrogen Doped Titania Nanocrystals: Comparison to a Commercial Nanopowder * Abstract: Nitrogen-doped titania (TiO 2 ) nanocolloids with an average size of 10 nm are successfully prepared and characterized using a variety of spectroscopic methods. Using a combination of UV-visible reflectance spectroscopy, Raman scattering, FT-IR transmission and x-ray photoelectron spectroscopy, the degree of nitrogen incorporation and the nature of chemical bonding on the nanocrystal surface and in the bulk was evaluated. Furthermore, x-ray powder diffractometry and transmission electron microscopy are utilized to determine the crystal phase, grain size and nanocrystal shape. The enhanced photocatalytic activity of the nitrogen-doped nanocrystals is evaluated through a study of the decomposition of methylene blue under visible light excitation in comparison to a selected series of titania-based nanomaterials. Spectroscopic measurements provide conclusive evidence for O-Ti-N bond formation during the doping process. This substitutional doping is held accountable for the significant increase in photocatalytic activity in the nitrogen-doped titania nanocrystals. * This chapter has been partially published in J. Phys. Chem. B 2004, 108, 15446; Adv. Funct. Mater. 2005, 15,

82 3.1 Introduction Nanomaterials have attracted much attention over the past decade due to their unique properties, which include quantum confinement and a heightened reactivity associated with changes in their molecular electronic structure and/or an increase in surface-to-volume ratio. 1-6 Particularly, nanostructured titanium dioxides (TiO 2 ) have gained great interest due to their potential as photovoltaics photochromic sensors and photocatalysts The research on semiconductor-based photocatalysis was initiated and promoted in 1972 with the discovery of the photocatalytic splitting of water on TiO 2 electrodes using ultraviolet light Among the different semiconductors, TiO 2 has however proven to be the most promising photocatalyst due to its low cost, non-toxicity, high stability, and high efficiency for mediating difficult-to-remove pollutants Specifically, TiO 2 nanocrystals could serve as an excellent material for the photocatalytic decomposition of organic pollutants in water. It is anticipated that in the presence of TiO 2 nanocrystals, sewage water could be gradually purified under sunlight. Currently however, the highly efficient usage of TiO 2 in photocatalysis applications is prevented by its wide bandgap (3.2 ev), which responds to only a small fraction of the sun s energy spectrum. Therefore, one of the endeavors to improve the performance of TiO 2 is to increase its optical activity by shifting the onset of its response from the UV to the visible region Underlying the optical response of any material is its electronic structure, which is related to its chemical composition, the nature of the bonds between the atoms or ions, and, for nanometer sized materials, its physical dimension (confinement of carriers). By changing the chemical composition of TiO 2 through main-group element doping, in particular with nitrogen, it 48

83 has been shown that the resultant TiO 2-x N x is capable of light absorption at much longer wavelength For substitutional doping of TiO 2, either the metal (titanium) or the non-metal (oxygen) component can be replaced. Over the past years, different metals have been employed to tune the electronic structure of TiO 2 -based materials either by the ion implantation method or by wet chemical methods. 8,14 The photocatalytic reactivity of metal-doped TiO 2 depends on several factors, which includes the dopant concentration, the energy levels of the dopants within the TiO 2 lattice, their d-electronic configuration, and the spatial distribution of the dopants. Enhanced photocatalytic activity has been reported for metal-doped TiO 2 at 0.5% doping. 8 In addition, it has been reported that metal ions physically implanted as zero-valent metals produce significant changes in the TiO 2 host material. 14,18-20 However, studies also reveal that metal doping can result in thermal instability and increased carrier trapping. 21,22 Alternatively, recent theoretical and experimental studies have shown that the desired bandgap narrowing of TiO 2 can be achieved by using main-group dopants to enhance the photoactivity of TiO 2 in the visible spectral range. 15,16,23,24 When employing dopants to change the optical response of TiO 2 there are at least two factors to consider. First, it is desirable to maintain the integrity of the crystal structure of the photocatalytic host material; a second requirement is to produce favorable changes in its electronic structure. In the case of the crystal structure of a material, it is directly related to the ratio of the cation and anion sizes in the crystal lattice. While it appears easier to substitute the Ti 4+ cation in TiO 2 with other transition metals, it is more difficult to replace the O 2- anion with other anions due to differences in charge states and ionic radii. However, the inherent lattice strain in nanometer-sized materials provides an 49

84 opportunity to dope TiO 2 to a larger extent. This scenario was first predicted by theory and demonstrated in experiments on thin films, 15 where the nitrogen-doped TiO 2 film exhibited improved catalytic activity under visible-light excitation. Very recently, the formation of TiO 2-x N x nanocrystals has also been reported. 16,17 However, the actual formation of O-Ti-N domains in the lattice of the host semiconductor has not yet been demonstrated and therefore structural evidence for the formation of substitutional O-Ti-N sites within the nitrogen-doped TiO 2 nanocrystals still remains elusive to date. Here, we report on the preparation and spectroscopic evidence of substitutional nitrogen incorporation in TiO 2 nanostructured materials. Furthermore, the synthesized nitrogen-doped nanocrystals are compared to a series of titania-based nanomaterials, which includes the commercially available TiO 2 Degussa P25 nanopowder. A number of techniques have been used to explore the structure and properties of the investigated nanocrystals. Furthermore, using the photodecomposition of methylene blue under visible light irradiation as a model, considerable enhanced photocatalytic activity is found for the TiO 2-x N x nanocolloid in comparison to the nitrided Degussa P25 nanopowder sample. 3.2 Experimental Preparation of Nitrogen-doped TiO 2 Nanomaterials. The synthesis of TiO 2 nanocrystals is accomplished with the drop wise addition (1 ml / min) of a 5 ml aliquot of Ti[OCH(CH 3 ) 2 ] 4 (Aldrich, 97%) dissolved in isopropyl alcohol (5:95), to 900 ml of doubly distilled water adjusted to ph = 2.0 using HNO 3. After continuous stirring of the reaction mixture for four hours, a clear colloidal solution of TiO 2 nanocrystals is formed. 25 Treatment of the nanoparticle solution with an excess of triethylamine (Alfa Aesar 99+%) results in the formation of a yellow solution. Upon centrifugation and 50

85 vacuum drying (5 x 10-2 Torr), the treated nanoparticles are isolated as yellow crystallites. The choice of triethylamine represents the best compromise for the direct solution phase nitridation of the TiO 2 nanocolloid. The process could also be successfully carried out using hydrazine and to some extent with an ammonium hydroxide solution. On the other hand, the nitrided Degussa sample was obtained by reacting approximately 0.5 gram of Degussa with 10 ml of triethylamine for > 4 hours at 80 C. Upon vacuum drying (5 x 10-2 Torr) for several hours, this treatment resulted in the isolation of yellow-brown crystallites. Characterization. The UV-visible reflectance spectra of the nanocrystal samples were measured using a Cary 5000 UV-visible-NIR spectrometer. Samples for Fouriertransformed infrared (FTIR) measurement were prepared as KBr pellets and analyzed using a Thermo Nicolet Nexus 870 FTIR spectrometer. On the other hand, the Raman spectra were collected from a dye laser-pumped Raman system using an excitation wavelength of 458 nm 26. The incident laser employed in the measurements was operated at 20 Hz with 10 ns pulses and the laser power was attenuated to 5 mw using neutral density filters. Furthermore, the entrance slit on the half-meter spectrometer (Chromex, 500i; 2400/300 grating) was set to 100 μm for all spectra. In addition, a notch filter was used to filter out the excitation light. During measurement, the beam was focused on the sample and data were collected in a 135 back-scattered configuration. Each scan was accumulated for 120 seconds and the spectra were calibrated using Ar and Xe lamps. X-ray diffraction patterns were obtained for the different nanocrystal samples using a Philips PW 3710 X-ray powder diffractometer. For XPS measurements a Perkin- Elmer PHI 5600 XPS System was used. Samples for XPS measurement were coated on carbon tapes attached to the sample holder. The transmission electron microscopy (TEM) 51

86 images were taken using a Philips CM20 transmission electron microscope operated at 200 kv. Samples for TEM analysis were prepared by depositing a drop of aqueous nanocrystal solution onto a copper grid supporting a thin film of amorphous carbon. The grid was dried in vacuum for 24 hours prior to TEM measurement. Evaluation of the photocatalytic activity of the nanocrystals was conducted by measuring the changes in absorption of a methylene blue solution at 650 nm during the dye s photocatalytic decomposition, upon irradiation with 390, 540 and 780 nm visible light. The tunable single-wavelength light excitation was produced by an amplified Tisapphire-based femtosecond laser system (Clark MXR CPA 2001), which is pumped by an Er-doped fiber laser and using an OPA (TOPAS, light conversion). During irradiation of the sample, the laser beam intensity was adjusted with a neutral density filter wheel and the pulse train was guided into a quartz cuvette filled with a 2 ml aqueous solution of methylene blue (optical density 0.8) and 10 mg of the nanocrystal sample. The absorption spectra of methylene blue were then obtained on a Varian Cary Bio50 UVvisible spectrometer under aerobic conditions and were always corrected for methylene blue decomposition in the absence of any titania nanocrystals. 3.3 Results UV-vis Reflectance Spectroscopy Recently, the first successful nitrogen doping of up to 8% into titania nanocrystals was reported, 16 forming TiO 2-x N x nanostructures. The precursor Ti[OCH(CH 3 ) 2 ] 4 in isopropyl alcohol (volume ratio 5 : 95) was added drop-wise (1 ml / min) into distilled water under ph = 2.0 to form a colloidal solution after continuous stirring for > 4 hours, Figure 3.1A-a. 25 When, an excess of triethylamine is added, the solution becomes turbid 52

87 and the color of the solution appears pale yellow. After centrifugation and vacuum drying for 12 hours, yellow crystallites were obtained (Figure 3.1A-d). The change in color of the nanocrystals upon nitrogen treatment indicates a shift of absorption into the visible wavelength range and is distinctly different from that which accompanies the amine treatment of Degussa P25 (Figure 3.1A-b). The color change upon nitrogen doping of Degussa P25 is to light brown (Figure 3.1A-c). A Reflectance / %R B a b c d a b c d a TiO 2 Nanoparticle b Degussa P25 c Nitrated Degussa d TiO 2.x N x Nanoparticle Wavelength / nm Figure 3.1 Visual comparison (A) and UV-visible reflectance spectra (B) of (a) TiO 2 nanoparticles; (b) Degussa P25 TiO 2 powder; (c) Degussa P25 TiO 2 powder nitrided with triethyl-amine; and (d) nitrided TiO 2-x N x nanocrystal powder. The chemistry involved in the formation of our TiO 2-x N x nanocrystals can be envisioned as a two-step reaction process: hydrolysis and condensation. Titanium alkoxides (Ti-OR) are widely used precursors for the hydrolysis (reaction 1) and condensation (reaction 2): 53

88 Ti OR + H O( H NR ') hydrolysis Ti OH ( NHR ') + ROH (3.1) 2 2 Ti OH ( NHR ') + ( NHR ') HO Ti condensation Ti O( N) Ti + H O( H NR ') (3.2a) 2 2 Ti OH ( NHR ') + RO Ti condensation Ti O( N) Ti + ROH ( R ' OR). (3.2b) The hydrolysis reaction (reaction 1) occurs upon the addition of water, when a hydroxyl group undergoes nucleophilic substitution on the metal center resulting in the exchange of the alkyl group (OR). This is followed by condensation, which involves the formation of Ti-O-Ti bonds and, in the presence of amines, the O-Ti-N bond is formed through substitution of oxygen with nitrogen in the initial TiO 2 crystal. Figure 1B compares the UV-visible reflectance spectra of the presented TiO 2 samples obtained with and without amine treatment. The reflectance spectrum of the prepared nanoscale TiO 2 is blue-shifted compared to the larger Degussa P25 powder, whose reflectance spectrum rises at 380nm. The reflectance spectrum for nitrided Degussa TiO 2 appears quite complex as it shows a slow increase in optical absorption to wavelengths exceeding 800 nm. This optical response in the long wavelength range is, however, not reflected in the activity of the nitrided sample and is, therefore, believed to be due to absorption into defect-related states. In contrast, the nitrogen-doped TiO 2 nanoparticles show a sharp onset of absorption, which has shifted from 340 to 500 nm (d) FTIR Spectroscopy Figure 3.2 compares the IR transmittance spectra for these TiO 2 -based samples. FTIR spectra provide information on surface functional groups. Interestingly, the different samples exhibit quite similar FTIR spectra. The signals in the range

89 cm -1 are characteristic of the formation of an O-Ti-O lattice. After vacuum drying we find no Transmission / a.u. TiO 2-x N x Nanocolloid TiO 2 Nanocolloid N-treated Degussa P25 Degussa P25 OH OH O-Ti-O Wavenumber / cm -1 Figure 3.2 IR transmission spectra of P25 TiO 2 powder (red); P25 TiO 2 powder nitrided after triethyl-amine treatment (blue); TiO 2 nanocolloid particles (black) and nitrided TiO 2-x N x nanocrystals (green). evidence for adsorbed organic alkoxides, such as -OC 3 H 7. Nor do we find evidence for the formation of amine surface complexes with the Ti(IV) ion. 27 The peak at 1630 cm -1 results from O-H bending of adsorbed water molecules. 28 The larger O-H peaks observed in the Degussa samples are attributed to a higher water content in these samples, compared to the synthesized nanosized titania-based samples. The broad adsorption band observed at 2900 to 3600 cm -1 corresponds to the O-H stretch region. 29,30 The peaks of the nanosized titania-based samples are shifted to lower energy and narrowed compared to those of the Degussa samples. However, the transmission band for the Degussa P25 sample shifts to higher energy from 1376 cm -1 to 1436 cm -1 after amine treatment, while the transmission band of the nanosized titania-based sample shifts to lower energy, from 55

90 1316 cm -1 to 1225 cm -1 after amine treatment. The typical infrared peaks of NO 2 occur at 1618 cm -1 (anti-symmetric stretch, very strong), 1318 cm -1 (symmetric stretch, weak), and 750 cm -1 (bend, strong). Since these peaks are not detected, significant formation of NO 2 species on the surface can be ruled out. 31 Triethylamine displays sharp infrared peaks around 2950 cm -1, and 1000 cm cm These peaks are not observed for either the synthesized nitrited TiO 2 nanoparticle nor the Degussa P25. Thus, on the surface there are no detectable triethylamine residues Raman Spectroscopy Figure 3.3 shows the Raman spectra taken for the titania-based samples. The Ti- O-Ti network peaks over the cm -1 range are characteristic of anatase and rutile structures Typically there are six Raman active fundamental modes observed at 144 cm -1 (E g ), 197 cm -1 (E g ), 397 cm -1 (B 1g ), 518 cm -1 (A 1g + B 1g ) and 640 cm -1 (E g ) for A TiO 2-x N x Nanoparticle TiO 2 Nanoparticle B Nitrited P25 Raman Intensity / a.u. TiO 2-x N x Nanoparticle TiO 2 Nanoparticle Nitrated P25 Raman Intensity / a.u. Degussa P25 Degussa P Raman Shift / cm Raman Shift / cm -1 Figure 3.3 Raman Spectra of P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocrystals (black) and nitrided TiO 2-x N x nanocrystals (green). 56

91 anatase TiO 2. For rutile TiO 2, there are four Raman active modes at 144 cm -1 (B 1g ), 448 cm -1 (E g ), 613 cm -1 (A 1g ) and 827 cm -1 (B 2g ), respectively The Raman shift peaks below 200 cm -1 are not observed due to the use of a filter to block the excitation light. The P25-based and our nanocolloid-tio 2 particles displayed Raman vibrations around 396 cm -1, 513 cm -1, and 633 cm -1, which can be attributed to the vibrational modes of the anatase phase. For P25 TiO 2, there is one more peak observed at 445 cm 1, which can be attributed to the vibrational mode of the rutile phase in the Degussa P25 samples. The Raman peaks shift slightly in the prepared TiO 2 nanocrystals compared to the Raman peaks for P25 powder. For example, compared with the vibrational mode at 396 cm -1 for P25, this mode shifts to 401 cm -1 for the TiO 2 nanocrystals and to 394 cm -1 for the nitrogen-doped TiO 2-x N x nanocrystals. The different Raman shifts in the untreated samples can be attributed to the different sizes of the TiO 2 samples. 39 There is a notable shift on nitrogen doping for the latter. When compared to the P25-based samples, the Raman peaks are broader for the 5-10 nm TiO 2 nanocolloid particles. The full widths at half maximum (FWHM) are 23 ± 0.5 cm -1, 22 ± 1.0 cm -1, 29 ± 0.5 cm -1, and 28 ± 1.0 cm - 1, for P25, nitrided P25, and pure and nitrogen-doped TiO 2 nanocolloid particles, respectively. This broadening likely results from the finite lifetime of the vibrational modes in the smaller size particles. 38,39 However, the formation of any O-Ti-N bond does not lead to any observable new Raman band. 40,41 This can be attributed to the small change in the Raman vibrational modes due to the partial replacing of O with N, which can be too small to be detected. The lacking of Raman modes above 800 cm -1 shift also rules out the formation of traceable amounts of surface NO, NO 2 species Transmission Electron Microscopy 57

92 In Figure 3.4, the transmission electron microscopy (TEM) images show that the synthesized nitrogen-doped TiO 2 nanoparticles are partially aggregated and polycrystalline in nature. The individual grain size can be confirmed to be about ~10 nm. Furthermore, the diffraction pattern (Inset c) suggests that the nitrogen-doped TiO 2 nanoparticles have an anatase phase crystal structure, consistent with the X-ray powder diffraction pattern in Figure 3.5. B C 2 nm A 100 nm Figure 3.4 TEM images of Nitrogen-doped TiO 2 nanocolloid particles (a) at low resolution (bar size 100 nm), (b) at high resolution image of the region in the indicated area (bar size 2 nm) inset on the left side and (c) the diffraction pattern showing a polycrystalline anatase phase X-ray Powder Diffractometry Figure 3.5 depicts the X-ray powder diffraction (XRD) patterns for both the initial and nitrided titania nanocolloid and P25 samples, respectively. We examine the significant differences between the XRD profiles for these systems. From the intensity distribution of the particular reflections and the integral intensity of the X-ray diffraction 58

93 patterns, crystallite size distributions and average crystallite sizes can be calculated. 43 With diminishing crystallite size, the measured XRD pattern exhibit broadened and often overlapping reflections. The broadening of the reflections is inversely proportional to the crystallite size. The size of coherently diffracting domains can be obtained from Scherrer s equation D = 0.9λ βcosθ (3.3) where D is the crystal size, λ is the wavelength of X-ray radiation ( nm for Cu K α TiO 2-x N x Nanocolloid TiO 2 Nanocolloid N-treated P25 Degussa P25 Intensity / a.u θ / O Figure 3.5 Comparison of the powder x-ray diffraction patterns of P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocrystals (black) and nitrided TiO 2-x N x nanocrystals (green). radiation), β is the full width at half maximum, and θ is the diffraction angle. 43,44 The P25 and the amine-treated P25 were found to consist of a mixture of anatase and rutile phases in the ratio 3:1, while the prepared TiO 2 nanocolloid particles and the nitrogen-doped TiO 2-x N x nanocolloids displayed only the anatase crystal phase, which is considered to be 59

94 the more photoreactive phase. 40 Since the diffraction patterns of the P25 and the synthesized nanosized TiO 2 particles did not show any change after amine treatment, it suggests that the doping process does not significantly changes the crystalline phase and has little effect on the grain size. Compared to the 30 nm average diameter of a commercial P25 sample, the average grain size of the N-doped TiO 2 is close to 10 nm as estimated from the Debye-Scherrer equation. The diffraction peaks of the nano-sized TiO 2 are broad as some peaks coalesce due to the small size of the particles. 17, X-ray Photoelectron Spectroscopy X-ray photoelectron spectroscopy (XPS) profiles for the titania and nitrided titania samples are shown in Figure 3.6. The XPS technique monitors the electron binding energy of sites within a few nanometers of the particle surfaces. Recently, Gyorgy et al. 47 have carried out an excellent depth profiling study on combined titanium oxide-titanium nitride systems, which is the subject of the current study. From our measurements, we find that between 4 and 8 percent nitrogen is incorporated into the lattice of the synthesized TiO 2 nanoparticles. In comparison, less than 1 % nitrogen can be detected in the amine-treated P25 sample. It appears that the synthesized nanometersized TiO 2 particles, with an average diameter of ~10 nm, are more receptive for nitrogen up take. 60

95 TiO 2-x N x Nanocolloid O 1S N1s peak TiO 2 Nanocolloid N-treated Degussa Degussa P25 Ti 2P Intensity / a.u. Ti LMM O KLL Ti 2S C 1S N 1S Ti 3P Ti3S Energy / ev Figure 3.6 Comparison of the X-ray photoelectron spectra of P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocrystals (black) and nitrided TiO 2-x N x nanocrystals (green). The inset shows the N 1S peak for these samples in the 400 ev region Photocatalytic Activity The photocatalytic activity of the nanoparticles is evaluated by measuring the changes in absorption of the methylene blue solution at 650 nm during photocatalytic decomposition, upon irradiation with 540 nm (visible) and 390 nm (UV) light (Figure 7). The results presented in Figure 7A demonstrate that the synthesized nitrogen-doped TiO 2 nanocrystals have, under visible light excitation, a much higher photocatalytic activity compared to commercially available P25 TiO 2 nanoparticles, Figure 7B. The significant efficiency for 390 nm excitation is consistent with the change in the reflectance spectrum of P25 TiO 2 treated for an extended period with triethylamine. At 540 and 780 nm, P25 61

96 TiO 2 is virtually ineffective. The undoped nanocolloid was ineffective at all wavelengths (data not shown). The photodecomposition activity with the nitrogen-doped TiO 2 70 O.D. / (a.u.) nm 540 nm A) TiO 2-x N x Nanoparticle nm Time / min 790 nm 540 nm 390 nm B) Degussa P Relative Reactivity C White: Degussa P25 Black: N-treated P25 Striped: TiO 2-x N x Nanocolloid 540 nm 390 nm Figure 3.7 Photocatalytic decomposition of methylene blue on TiO 2-x N x nanocolloid (A) and P25 TiO 2 (B), as monitored by the decrease in optical density at 650 nm following 780, 540, and 390 nm laser excitation after excitation of a 5 nl volume of a 2 mm aqueous methylene blue solution. C) Relative photoreaction rate on the decomposition of methylene blue under 540 nm and 390 nm light with commercial P25 TiO 2 (white column), commercial P25 TiO 2 nitrided with triethylamine (black column) and N-doped TiO 2 (red column) nanocolloid particles as catalysts. The bars shown in the graph depict the relative reactivities obtained from the decolorization of methylene blue at 650 nm per unit time where the photocatalytic decomposition rate under 540 nm with commercial P25 TiO 2 is set to 1. All rates were corrected for the decomposition of methylene blue without any catalyst under identical conditions. nanocolloid as a catalyst is found to be 7 and 4.3 times higher than that of P25 TiO 2 under 540 and 390 nm light excitation, respectively. The nitrided P25 TiO 2 clearly does not show a promising catalytic activity under visible light irradiation as does the synthesized nitrogen-doped TiO 2 nanocolloid. This difference in photocatalytic activity 62

97 under 540 nm and 390 nm light irradiation can be easily related to the distinctly different optical responses of the two types of nanoparticles as presented in Figure 1B. 3.4 Discussion Like other common organic pollutants, methylene blue can be decomposed under light irradiation and eventually mineralized on TiO 2 nanoparticles into CO 2, NH + - 4, NO 3 and SO ,10,12,14,48,49 As electrons are promoted from the valence band to the conduction band, leaving a hole in the valence band, they can migrate to the surface of the TiO 2 nanoparticles, and combine with absorbed oxygen molecules to form O 2 O2 - centers. O 2 -, neutralized by protons in water, can form HO 2 radicals, which can react with each other to form H 2 O 2 and O 2 or OH. As the holes in the valence band migrate to the surface, they can react with the OH - groups to form OH -. These transient species as well as holes break down the methylene blue to smaller fragments, which eventually completely decompose into simple inorganic minerals. Among the many factors affecting the efficiency of the photocatalyst, the number of excited electrons and holes created by irradiation plays an important role. The more efficiently these carriers are created, the better the photocatalyst. The number of excited carriers is proportional to the absorption of the photocatalyst. Thus, photocatalytic activity should increase with the shift of the absorption edge from the UV to the visible region. However, there seem to be other factors co-controlling the photocatalytic efficiency. These include chemical structure, crystallinity, size, and the adsorption properties of the pollutants on the catalyst 14, 48,49 substrate. 63

98 In combination, the XPS, TEM and UV-vis reflectance results suggest that nitrogen has been incorporated into the lattice of the TiO 2 nanoparticles. The incorporation of nitrogen into the TiO 2 lattice clearly leads to a pronounced red-shift in the optical response of the synthesized 10 nm nanocrystals, which is clearly distinct from the effect that nitrogen-doping has on the larger 30 nm particles of the P25 powder. In addition IR and Raman measurements do not show any evidence for surface -NO 2 or alcoxyl termination. For further analysis of the chemical structure of the investigated titania samples we have examined three areas of the XPS spectrum, the Ti 2p region near 460 ev, the O 1s region near 530 ev, and the N 1s region near 400 ev (inset in Figure 3.6). In these three regions there are distinct differences between the synthesized nanocrystals and the commercially available P25 samples. The binding energy peak for the nitrogen-doped TiO 2 nanocrystals is broad, which extends from to ev. It is centered at ev that is greater than the typical binding energy of ev in TiN, and where O-Ti-N electrons are to be expected. 50 This shift in binding energy can be understood by the fact that the N 1s electron binding energy is higher when the formal charge of N is more positive (e.g. 408 ev in NaNO 3 ) compared to a zero or negative formal charge (398.8 ev in NH 3 ). 32 The local electron density around N is lower for positive formal charges while higher in the latter case. When nitrogen substitutes the oxygen in the initial O-Ti-O structure, the electron density around N is lowered compared to the one in a TiN crystal due to the oxygen vicinity. Thus, the N 1s binding energy in O-Ti-N environment is higher than that in an N-Ti-N environment. When scanning the Ti 2p and O 1s XPS regions, major differences for the original binding energies and their nitrogen-treated counterparts are noted. First, for the Ti 2p 3/2 64

99 region, depicted in Figure 3.8 (A), one observes a peak close to ev for the P25 powder, located at notably higher binding energy than for the remaining samples. In comparison, the corresponding peak of the synthesized nanocrystals appeard at ev, a notably lower binding energy. This demonstrates that the shift in the XPS spectra with Ti 2p Peak TiO 2-x N x Nanocolloid TiO 2 Nanocolloid N-treated Degussa Degussa P25 A TiO 2-x N x nanoparticle TiO 2 nanoparticle Nitrited Degussa Degussa B O 1s Peak Intensity / a.u Intensity / a.u Energy / ev Energy / ev Figure 3.8 The Ti 2p peak around 460 ev (A) and O1s peak (B) of P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocolloid particles (black) and nitrided TiO 2-x N x nanocrystals (green). The inset in (A) shows the Ti 2p 3/2 peak of these samples around 460 ev. nitrogen incorporation is most significant in the prepared nanocrystal sample, where a shift towards lower binding energy upon nitrogen-treatment signifies the successful incorporation of nitrogen into the TiO 2 lattice. 50 The treated P25 sample shows a peak at ev and the nitrided nanocrystals at ev. This is significantly lower (0.9 ev) in binding energy, indicating the highest nitrogen incorporation in the synthesized titania nanocrystals. The typical binding energy of Ti 2p 3/2 peak in TiO 2 crystals is ev, which is higher than that in TiN crystals, typically at ev. 32, Also, the binding energy of the Ti 2p 3/2 peak shifts to lower energies when the valence state of Ti 4+ is lowered to Ti 3+ and Ti This is also true for TiO, TiS, TiN, TiP compounds, where 65

100 the valence states of Ti are Ti 2+, Ti Thus, the observed Ti 2p 3/2 binding energy after nitrogen treatment can be attributed to the formation of O-Ti-N by partially substituting the O atom in the TiO 2 lattice with nitrogen. This shifts the binding energies to lower values. The observed changes in the XPS spectra are consistent with the observed changes in the infrared spectra in that the IR transitions red-shift upon N-doping. The observed XPS peaks for the Ti 2p region and their change with nitrogen incorporation are strikingly consistent with the recent XPS depth profiling characterization of Gyorgy et.al. 47 These authors have also observed a shift in the Ti 2p binding energy to lower energies as a TiO 2 surface is nitrided. Further, the observations in the present study are consistent with the earlier results of Saha and Tomkins 50 who have used XPS to characterize the oxidation of a titanium nitride surface; in essence, the reverse of the current experiment and that of Gyorgy et.al. 47 The O 1s XPS spectra in Figure 3.8 (B) also show significant changes upon nitrogen incorporation, the most significant being an additional signal at higher binding energy than the main oxygen feature for the prepared nanocrystals (Figure 3.8 and 3.9). In Figure 3.8B, the P25 sample shows an oxygen 1s peak at ev vs. the nanocrystals at ev. The effect of nitrogen-treatment is also very different for the P25 sample as we observe a shift to approximately a ev binding energy for both the P25 and nanocolloid samples. These peak shifts are again possibly due to geometrical constraints in the different nanostructures. Of greater importance, is the additional peak, which appears at 532 ev (Figure 3.9) clearly noted in the spectrum of the synthesized nitrogen-treated nanocrystals. It is a feature first observed by Saha and Tomkins 50 for a native oxide sample and most 66

101 recently characterized by Gyorgy et al. 47 in their depth profiling study on titanium nitride surfaces. Gyorgy et.al. have assigned this feature to the formation of oxidized Ti-N, which leads to the O-Ti-N structure. Our study suggests that the appearance of this peak is also a consistent feature for the N-substitution in TiO 2 and signifies the formation of an O-Ti-N structure. 47 TiO 2-x N x nanoparticle O 1S Peak Intensity / a.u Energy / ev Figure 3.9 X-ray photoelectron spectrum of the oxygen 1s peak displays a second signal shifted to higher binding energy for the nitrided nanocolloid due to the formation of O- Ti-N bond. In analogy to the classic paper of Saha and Tomkins, we suggest that the Ti 2p 3/ ev XPS peak is consistent with the formation of a crystalline TiO 2 sample. The nitrogen-incorporation shifts the XPS spectrum to a lower binding energy (peak ~ ev). This effect is much less pronounced for the P25 TiO 2 nanocrystals. However, because of surface strain and lattice distortion present in the synthesized TiO 2 nanocrystal lattice, 16,55,56 one would expect that the incorporation of the amine and its subsequent reaction can proceed more readily in the prepared nanocrystals. The smaller nanocrystals 67

102 take up the nitrogen easier because of larger lattice strain and the larger P25 cannot do that so readily because the surface is already relaxed and any doping poses a steric problem, which is also suggested as significant perturbation in the binding energies. It is, therefore, not surprising that we observed a much higher binding energy of the XPS Ti 2p and O 1s peaks in the P25 sample. The structure of the P25 sample suggests that the nitrogen perturbs the TiO 2 lattice and changes the Ti-O bond shifting the infrared transitions to lower energy. In the TiO 2 nanocolloid, the nitrogen should have a much smaller effect on the Ti-O bond as these structures are already considerably more strained. 55,56 The nitrogen, which is incorporated to form the O-Ti-N should in fact somewhat strengthen the Ti-O bond, thus shifting the O-Ti-O infrared feature to higher frequency. Thus, the observed changes in the XPS, FTIR, and Raman spectra are providing consistent structural information for O-Ti-N formation, the substitutional doping of nitrogen for an oxygen, that leads to the enhanced photocatalytic activity observed in smaller nitrogen-doped nanocrystals with an average size of ~ 10 nm. 3.5 Conclusions A highly efficient visible light activated nanosized-photocatalyst was prepared by doping TiO 2 nanoparticles with nitrogen. The doping was characterized by several techniques, such as UV-visible reflectance, IR transmission, Raman Scattering, XRD, XPS, and TEM. The formation O-Ti-N is identified as the chemical structure formed during the substitutional doping process and is attributed to be the one responsible for the significant increase in photocatalytic activity of the synthesized nitrogen-doped nanocrystals. The properties of this readily nitrided nanocrystal were compared with an undoped and nitrided commercial P25 TiO 2 nanomaterial. The photocatalytic 68

103 decomposition of methylene blue in water was found to be faster by a factor of ~ 4 in the synthesized samples. Thus, the promise of the nitrogen-doped TiO 2 nanoparticles for photocatalysis under visible light has been demonstrated using a known and often used standard model reaction. 3.6 References 1. Alivisatos, A. P. J. Phys. Chem. 1996, 100, Nirmal, M.; Brus, L. Acc. Chem. Res. 1999, 32, Murray, C. B.; Kagan, C. R.; Bawendi, M. G. Annu. Rev. Mater. Sci. 2000, 30, Burda, C.; El-Sayed, M. A. Pure Appl. Chem. 2000, 72, Gole, J. L.; White, M.G. J. Catal. 2001, 204, Gole, J. L. Shinall, B.D. et al., Chem. Phys. Chem., in press. 7. Gerfin, T.; Graetzel, M.; Walder, L.; Molecular and Supermolecular Surface Modification of Nanocrystalline TiO 2 Films: Charge-Separating and Charge- Injecting Devices; Gerfin, T.; Graetzel, M.; Walder, L. Ed.; John Weley and Sons: Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W. Chem. Rev. 1995, 95, Grätzel, M. Nature, 2001, 414, Millis, A.; Hunte, S. L. J. Photochem. Photobiol. A: Chem., 1997, 108, 1. 69

104 11. Fujishima, A.; Honda, K. Nature 1972, 37, Fujishima, A.; Rao, T. N.; Tryk, D. A J. Photochem. Photobio. C: Photochem. Rev. 2000, 1, Tryk, D. A.; Fujishima, A.; Honda, K. Electrochimica Acta 2000, 45, Kaneko, M.; Okura, I. Photocatalysis Science and Technology Springer 2002, New York. 15. Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293, Burda, C.; Lou, Y.; Chen, X.; Samia, A. C. S.; Stout, J.; Gole, J. L. Nano Lett. 2003, 3, Gole, J. L.; Stout, J.; Burda, C.; Lou, Y.; Chen, X, J. Phys. Chem. B, 2004, 108, Wang, C.; Bahnemann, D.W.; Dohrmann, J. K. Chem. Comm. 2000, 16, Wang, Y.; Hao, Y.; Cheng, H.; Ma, H.; Xu, B.; Li, W.; Cai, S. J. Mater. Sci. 1999, 34, Coloma, F.; Marquez, F.; Rochester, C. H.; Anderson, J. A. Phys. Chem. Chem. Phys. 2000, 2, Yamashita, H.; Honda, M.; Harada, M.; Ichihashi, Y.; Anpo, M.; Hirao, T.; Itoh, N.; Iwamoto, N. J. Phys. Chem. B 1998, 102, Wang, Y.; Cheng, H.; Hao, Y.; Ma, J.; Li, W.; Cai, S. Thin Solid Films 1999, 349,

105 23. Yu, J. C.; Yu, J. G.; Ho, W. K.; Jiang, Z. T.; Zhang, L. Z. Chem. Mater. 2002, 14, Khan, S. U. M.;. Al-Shahry, M; Ingler, Jr. W.B. Science 2002, 297, In an alternate synthesis using acetic acid, 250 ml of doubly ionized water and 80 ml of acetic acid were combined in a one liter flask as the mixture was cooled to 0 C in an ice bath under stirring. 10 ml of 2-propanol followed by 3.7 ml of Ti[OCH(CH 3 ) 2 ] 4 are added to a dropping funnel fixed to the flask as this solution is added slowly and dropwise, under a dry nitrogen atmosphere, again with vigorous stirring. Continued stirring of the initial mixture for 24 hours produces a clear colloid solution with particle sizes ranging from 5 to 20 nm. 26. Copeland, T.; Shea, M. P.; Milliken, M. C.; Smith, R. C.; Protasiewicz, J. D.; Simpson, M. C. Analytica Chimica Acta 2003, 496, This is not the case for metal doped systems, in particular for copper doped TiO 2-x N x we find evidence for amine surface complexation, work in progress. 28. Deng, C., James, P.F., Wright, P.V., J. Mater. Chem. 1998, 8, Park, H. K., Kim, D.K, Hee, C., J. Am. Ceram. Soc. 1997, 80, Zheng, M., Gu, M., Jin, Y., Jin, G., Mater. Sci. Engin. 2000, B77, Shimanouchi, T., J. Phys. Chem. Ref. Data, 1972, 6, NIST/EPA Gas-Phase Infrared Database. 71

106 33. Takahashi, T.; Nakabayashi, H.; Tanabe, J.; Yamada, N.; Mizuno, W. J. Vac. Sci. Technol. A 2003, 21, Ohsaka, T.; Izumi, F.; Fujiki, Y. J. Raman Spectrosc. 1978, 7, Berger, H.; Tang, H.; Lévy, F. J. Cryst. Growth 1993, 130, Tang, H.; Prasad, K.; Sanjinès, R.; Schmid, P. E.; Lévy, F. J. Appl. Phys. 1994, 75, Robert, T. D.; Laude, L. D.; Geskin, V. M.; Lazzaroni, R.; Gouttebaron, R. Thin Solid Films 2003, 440, Bersani, D.; Antonioli, G.; Lottici, P. P.; Lopez, T. J. Non-Crystalline Solids 1998, , Ivandaa, M.; Musić, S.; Gotić, M.; Turković, A.; Tonejc, A.M.; Gamulin, O. J. Mole. Struct. 1999, , Bernard, M.; Deneuville, A.; Thomas, O.; et al. Thin Solid Films 2000, 380, Cheng, Y. H.; Tay, B. K.; Lau, S. P.; et al. J. Appl. Phys. 2002, 92, Bao, X.; Wild, U.; Muhler, M.; Pettinger, B.; et al. Surf. Sci. 1999, 425, Pielaszek, J. X-ray Diffraction from Nanostructured Materials, in Nanostructured Materials: selected synthesis methods, properties and applications, Editors: Philippe Knarth, Joop Schoonman, Kluwer Academic Publishers, Boston, Klug, H. P.; Alexander, L.E. X-ray Diffraction Procedures for Polycrystalline and Amorphous Materials, 2 nd. Ed., Wiley: New York,

107 45. Lin, J.; Yu, J. C.; Lo, D.; Lam, S. K. J. Catalysis, 1999, 183, Prokes, S. M.; Carlos, W. E.; Gole, J. L.; She, C.; Lian, T. Material Research Society Proceedings, 2003, Gyorgy, E., Perez del Pino, A., Serra, P., Morenza, J. L., Surface and Coatings Technology, 2003, 173, Houas, A.; Lachheb, H.; Ksibi, M.; Elaloui, E.; Guilard, C.; Herrmann, J. Appl. Cata. B: Environmental , Oh, Y.; Bao, Y.; Jenks, W. S. J. Photochem. Photobio A: Chem. 2003, 161, Saha, N.C.; Tomkins, H. C. J. Appl. Phys., 1992, 72, Hashimoto, S.; Murata, A.; Sakurada, T.; Tanaka, A. J. Surf. Anal. 2002, 9, Yoshitake, M.; Thananan, A.; Aizawaki, T.; Yoshihara, K. Surf. Interface Anal. 2002, 34, Wang, Y. L Surf. Coat. Tech. 2002, 150, Gall, D.; Haasch, R. T.; Finnegan, N.; et al. Surf. Sci. Spec. 2000, 7, Chen, X.; Lou, Y.; Samia, A. C.; Burda, C. Nano Lett. 2003, 3, Zhang, X.; Wang, M. X.; Qu, L.; Peng, X. Appl. Phys. Lett. 2002, 81,

108 CHAPTER 4 Investigation of the Relation between the Structure, Chemical Composition, Optical and Photocatalytic Property of Nitrogen-doped Titania Nanomaterials Using Bottom-up method Abstract Nitrogen-doped titania nanoparticles were prepared in a mixture of alcohol/amine. The UV-visible absorption of the nitrogen-doped titania nanoparticles had a strong linear relation with the doped nitrogen concentration. The photocatalytic activity of the nitrogen-doped titania nanoparticles is monotonically related to the nitrogen concentration, and also affected by the crystal size, surface area, band structure, and surface property of the nanoparticles. The doped nitrogen is chemically bonded to the titanium atoms and interstitial hydrogen atom in the aminolysis process, and the binding energy of the doped nitrogen is related to its complicated local bonding environment. The reaction mechanism of titanium isopropoxide with amine is discussed, which provides a clue to the current doping chemistry on titanium dioxide. 75

109 4.1 Introduction Since the discovery of the photocatalytic splitting of water on TiO 2 electrode by Fujishima and Honda in 1972, 1 enormous effort has been spent on the study of TiO 2 under light illumination, due to its various potential applications, such as photovoltaics and photocatalysis. 1-9 The properties and performance of TiO 2 and TiO 2 -based devices are affected by the size of the basic TiO 2 particles. As the size of TiO 2 particles decreases, the surface-to-volume ratio increases dramatically (~ 1/r), so does the surface area (1/r 2 ). The high surface area brought by the small size of TiO 2 nanoparticle is beneficial to photocatalysis because most photochemical reactions occur on the surface. 1-7,10 Accompanied with the high surface area brought by the small size of TiO 2, is the increase of the bandgap energy, caused by the quantum confinement effect due to the spatial confinement of the motion of the electrons and holes in the particle. 9,11 The optical properties, i.e. the absorption, of TiO 2 is essential in its photon-driven applications. Typically, TiO 2 absorbs in the UV regime (3.0 ev for rutile phase and 3.2 ev for anatase phase), which is only a small fraction of the sun s energy (< 10%). The performance of TiO 2 can be enhanced by shifting the onset of its absorption from the UV to the visible region Metals have been employed to tune the electronic structure of TiO 2 -based material The photocatalytic reactivity of metal-doped TiO 2 depends on many factors, and metal doping can result in thermal instability and increased carrier trapping. The desired visible-light absorption of TiO 2 can be also achieved by using main group dopants In the study of nitrogen-doped TiO 2 from Asahi, 16 substitutional nitrogen, related to binding energy of 396 ev as atomic β-n form, was found responsible for the enhanced 76

110 activity both from their calculation and experiments. A linear increase of the visible-light response to the XPS-intensity of the nitrogen at 396 ev was reported by Irie and coworkers. 23 Diwald studied the nitrogen-doped TiO 2 with photoactivities on rutile single crystals and found different conclusions. 20 In the doped TiO 2 sample prepared by sputtering TiO 2 N 2+ /Ar + gas mixture, followed by annealing, the N peak was found at 396 ev and attributed to chemically bound N - species within the crystalline TiO 2 lattice. 20 When the doped TiO 2 sample was prepared by treating TiO 2 with NH 3 flow at high temperature, although a nitrogen peak at 396 ev was detected, the enhanced photocatalytic activity was attributed to nitrogen with binding energy of ev, which is chemically bound to hydrogen and interstitially existed in the TiO 2 lattice. 20 In the recent report on nitrogen-doped titania by Sakthivel, a nitrogen peak around 404 ev was observed, while the signal at 396 ev was completely absent. 21 We found that the absorption of the nitrogen-doped titania nanoparticles could be extended up to 600 nm, and its photocatalytic property could be enhanced. 19 Using x-ray photoelectron spectroscopy, we found the nitrogen atom around ev replaced part of the oxygen atoms in the initial titania lattice, and the newly formed O-Ti-N local bond environment was responsible for the enhancement of the absorption and photocatalytic activity. 19 Here, we used a simple method to dope titania nanomaterials with nitrogen and investigated their chemical compositions, structures, optical properties and photocatalytic activities. We found that in the nitrogen-doped titania nanoparticles, the nitrogen concentration has a strong correlation with the optical response, and the photocatalytic activity. By analyzing the reaction mechanism involved in the doping process, we found a reasonable explanation for the physical and chemical status of nitrogen, which can 77

111 further explain the observations in other studies. This discussion could help the understanding and manipulating of doped titanium dioxide for visible-light applications. 4.2 Experimental TiO 2 nanoparticle was prepared by adding of a 9.0 ml Ti[OCH(CH 3 ) 2 ] 4 (Aldrich, 97%) to a 150 ml isopropanol containing 5.0 ml NH 3 H 2 O very slowly. After continuous stirring of the reaction mixture for 2 hours, the resultant colloidal solution was boiled for 4 hours at around 85 o C. After removing the solvent by vacuum and drying for several hours, the samples were then calcined at ºC for 3 hours. The UV-visible reflectance spectra of the nanocrystal samples were measured on a Cary 50 UV-visible spectrometer with a fiber optical reflectance unit. X-ray diffraction (XRD) patterns were obtained for the different nanocrystal samples using a Philips PW 3710 X-ray powder diffractometer. The transmission electron microscopy (TEM) images were obtained using a JEOL 1200EX transmission electron microscope operated at 80 kv. For x-ray photoelectron spectroscopy (XPS) measurements, a Perkin-Elmer PHI 5600 XPS System was used. The energy resolution of the spectrometer was ev for the XPS measurements. Samples for XPS measurement were coated on carbon tape attached to the sample holder. The pressure in the vacuum chamber during the measurements was below mbar. All the measurements were made at room temperature. FT-IR spectra were obtained on a ThermaNicolet 870 FT-IR instrument. The samples were mixed with KBr powder (5:95 in volume ratio) and pressed into thin pellets. Evaluation of the photocatalytic activity of the nanocrystals was conducted by measuring the changes in absorption of a methylene blue solution at 650 nm during the 78

112 dye s photocatalytic decomposition, upon irradiation from a 150 W Xe lamp (# 6253, Oriel) in a parallel photochemical reactor described previously. The light was focused onto a cuvette filled with a 2 ml aqueous solution of methylene blue (optical density 1.0) and 10 mg of the nanocrystal sample. The absorption spectra of methylene blue were then obtained on a Varian Cary Bio50 UV-visible spectrometer under aerobic conditions and were always corrected for methylene blue decomposition in the absence of any titania nanocrystals Results Intensity / a.u. TiON 500 O C TiON 400 O C TiON 350 O C TiON 300 O C TiON 200 O C TiON Theta / O Figure 4.1 XRD spectra of the resultant samples: dried TiO 2-x N x (black curve), TiO 2-x N x after heat treatment for 3 hours at 200 o C (red curve), 300 o C (green curve), 350 o C (blue curve), 400 o C (cyan curve) and 500 o C (magenta curve). From the XRD spectra of the resultant samples shown in Figure 4.1, their structures can be determined. It can be seen that the samples baked at low temperature (< 350 o C) had broad peaks, suggesting an amorphous structure; the samples baked at high temperatures ( 350 o C) had sharp and well-defined peaks, featuring the anatase TiO 2 structure. The crystallization from amorphous phase occurs from 300 o C to 350 o C. Using 79

113 the width of the (101) peak, the average grain size estimated from the Scherrer s equation, is 7.68 nm, 9.44 nm and 14.8 nm in diameter for the sample heated at 350 o C, 400 o C and 500 o C, respectively. A B Figure 4.2 TEM images of the nitrogen-doped titania nanocatalyst before (A) and after calcination (B) Figure 4.2 shows TEM images for the doped TiO 2 nanocatalyst before (A) and after (B) the calcination. Before calcination, the doped TiO 2 nanomaterial had a gel-type amorphous structure (Figure 4.2A). After calcination, the sample crystallized and the crystallite size is centered around 10 nm (Figure 4.2B). From the global and N 1S XPS spectra of powder shown in Figure 4.3, the chemical composition determined for the resultant samples is 2.81% atomic ratio of nitrogen to total atoms for the dried sample, and is 2.75%, 2.66%, 0.90%, 0.37% and 0.38% atom ratio for the sample calcined at 200 o C, 300 o C, 350 o C, 400 o C and 500 o C, 80

114 Intensity / a.u. A 500 o C 400 o C 350 o C 300 o C 200 o C TiO 2-x N x Intensity / a.u. 500 o C 400 o C 350 o C 300 o C 200 o C B TiO 2-x N x Binding Energy / ev Binding Energy / ev Figure 4.3 Global and N 1S XPS spectra of the resultant samples: dried TiO 2-x N x (black curve), TiO 2-x N x after heat treatment for 3 hours at 200 o C (red curve), 300 o C (green curve), 350 o C (blue curve), 400 o C (cyan curve) and 500 o C (magenta curve). respectively. It is obvious that the nitrogen concentration in these samples decreased monotonously as the temperature of heat treatment increased. The binding energy of N1s electron occurred at ev. With the increasing of calcining temperature, it slightly decreased. This is attributed to local environment change of the nitrogen atom in the doped TiO 2 sample during the baking process. As analyzed later, at lower baking temperatures, the nitrogen atom is bonded to titanium and hydrogen (one or two) in the complicated amorphous gel structure; at higher baking temperature, the bonded hydrogen atoms are gradually expelled due to the condensation and crystallization process. The samples baked at low temperatures (< 350 o C) had a gray color, suggesting long-range absorption, and the samples baked high temperatures ( 350 o C) displayed a yellow to light-yellow color, due to the absorption from 400nm to 500nm; as seen from the UV-visible reflectance spectra of these samples shown in Figure

115 TiO 2-x N x Absorbance / a.u. 200 O C 300 O C 350 O C 400 O C 500 O C Wavelength / nm Figure 4.4 UV-visible reflectance spectra of the resultant samples: dried TiO 2-x N x (black curve), TiO 2-x N x after heat treatment for 3 hours at 200 o C (red curve), 300 o C (green curve), 350 o C (blue curve), 400 o C (cyan curve) and 500 o C (magenta curve). FTIR spectra provide information on surface functional groups. Interestingly, the different samples exhibit quite similar FTIR spectra as shown in Figure 4.5. The signals in the range cm -1 are characteristic of the formation of an O-Ti-O lattice. The peak at 1630 cm -1 results from O-H bending of adsorbed water molecules. The larger O- H peaks observed in the samples calcined at lower temperatures, suggests that these samples had more surface OH groups. The peak at 1539 cm -1 and 1453 cm -1 can be attributed to the newly formed N-H bond on the surface. The broad adsorption band observed at 2900 to 3600 cm -1 corresponds to the O-H stretch region. The typical infrared peaks of NO 2 occur at 1618 cm -1 (anti-symmetric stretch, very strong), 1318 cm -1 (symmetric stretch, weak), and 750 cm -1 (bend, strong). The typical vibrational mode for Ti-N (str) occurs around 1037 cm -1, while the Ti-O locates at 705 cm -1. The development 82

116 Transmittace / a.u. 500 O C 400 O C 350 O C 300 O C 200 O C TiO 2-x N x OH OH Wavenumber / cm -1 Figure 4.5 FT-IR transmission spectra of the resultant samples: dried TiO 2-x N x (black curve), TiO 2-x N x after heat treatment for 3 hours at 200 o C (red curve), 300 o C (green curve), 350 o C (blue curve), 400 o C (cyan curve) and 500 o C (magenta curve). of FT-IR spectra over 900 cm -1 to 1700 cm -1 displayed the structural evolution of the samples, which is mainly involved with the removal of the NH group during the formation of the TiO 2 lattice in the calcination (the disappearance of the peaks at 1539 cm -1 and 1453 cm -1 ), and the development of the O-Ti-N bond network during the condensation process (the increase in the absorption of the peak at 1350 cm -1 in the region from below 1400 cm -1 ) in the bulk. To evaluate and compare the photocatalytic performance of the N-doped nanomaterials, we have used a photoreaction based on the decomposition of methylene blue. The photocatalytic activity of the nanoparticles is further evaluated by measuring the change in absorption of a methylene blue solution at 650 nm during photocatalytic decomposition. From the results shown in Figure 4.6, the dried powder displayed the highest photocatalytic activity on the decomposition of methylene blue, which decreased as the temperature of the heat treatment increased. 83

117 Absorbance / a.u o C 400 o C 300 o C 200 o C TiO 2-x N x Time / min Figure 4.6 Photocatalytic decomposition (decolorization) of methylene blue on the resultant samples: dried TiO 2-x N x (black curve), TiO 2-x N x after heat treatment for 3 hours at 200 o C (red curve), 300 o C (green curve), 400 o C (cyan curve) and 500 o C (magenta curve). 4.4 Discussion The relationship between the nitrogen concentration and the absorption property and catalytic property of the doped TiO 2 To explore the relationship between the UV-visible absorption properties and the nitrogen concentration in the nitrogen-doped samples, we integrated the absorption over the wavelength range from 800 nm to 300 nm, since the cuvettes used in the photocatalytic decomposition property displayed high absorption, and the Xe lamp used had a similar flux over this range. From the plotted results in Figure 4.7, there existed a strong correlation between the UV-visible absorption properties and the nitrogen concentration in the nitrogen-doped samples. Unlike the strong correlation between the UV-visible absorption property and the nitrogen concentration, the relationship between the photocatalytic activity and the 84

118 Absorption / a.u Nitrogen % TiON 200C 300C 350C 400C 500C Sample Figure 4.7 The relationship between the UV-visible absorption properties and the nitrogen concentration in the resultant samples Δ (O.D.) Nitrogen % TiON 200C 300C 400C 500C Sample 0.0 Figure 4.8 The relationship between the photocatalytic activity and the nitrogen concentration in the resultant samples. nitrogen concentration shown in Figure 4.8 did not display an apparent linear relationship, although both showed the same trend of a decrease as the baking temperature increased. This can be attributed to the many other factors co-controlling the photocatalytic reactivity of the nitrogen-doped titania nanoparticles, besides the absorption factor. First, the samples baked at lower temperatures had amorphous structures and smaller sizes, which induced larger surface areas and larger adsorbing ability, and different relaxation and trapping dynamics of the photoexcited charge carriers, thus the efficiency of charge 85

119 transfer into the methylene adsorbed on the surface. Second, the samples baked at lower temperatures had more OH groups on the surface, which facilitated the adsorption of the methylene blue in the first step of photodecomposition. Third, the surface NH groups can facilitate the adsorption of the methylene blue due to the local basic environments on the surface induced by NH groups. Fourth, the band structures of the different samples may differ from each other, which induced different reducing/oxidizing ability of the excited charge carriers during the photocatalytic process The doping reaction chemistry of titanium isopropoxide with amine The traditional reaction in the hydrolysis of titanium isopropoxide with water can be illustrated as follows: Pr i i i i i H 2 O + Ti( O Pr ) 4 H 2O Ti( O Pr ) 4 ( HO)( O Pr ) 3Ti O H Ti( O Pr ) 3OH + Pr i OH The overall reaction can be written as i i i Ti( O Pr ) 4 + nh 2O Ti( O Pr ) 4 ( OH ) + n Pr OH ( n < 4) After hydrolysis, the precipitates could have formed due to condensationpolymerization of the alkoxyloxide by the liberation of water as i i i { Ti( O Pr ) ( OH ) } {(Pr O) ( OH ) Ti O Ti( OH ) ( O Pr ) } + ph O( n 4) 2 p 4 n n 4 n n 1 n 1 4 n 2 < p The hydrolysis in alcohol is very similar to that in water. The difference is that the hydrolysis rate is much slower in alcohol than in water, due to the fact that water is a stronger acid with stronger nucleophilic attacking ability than alcohol. In the presence of amine, similar reaction of aminolysis will occur as follows: n n 86

120 Pr i i i i i i H3 N + Ti( O Pr ) 4 H3N Ti( O Pr ) 4 ( H 2N)( O Pr ) 3Ti N H 2 Ti( O Pr ) 3 NH 2 + Pr OH and the overall reaction can be written as i i i Ti( O Pr ) 4 + nh 3 N Ti( O Pr ) 4 n ( NH 2 ) n + n Pr OH ( n < 4) After aminolysis, the following reaction could occur in the precipitates due to condensation-polymerization of the alkoxylxide by the liberation of amine as or i i i { Ti( O Pr ) ( NH ) } (Pr O) ( NH ) Ti N Ti( NH ) ( O Pr ) + ph N ( n 4) 2 p 4 n 2 n 4 n 2 n 1 2 n 1 4 n 3 < i { Ti( O Pr ) ( NH ) } H i ( NH ) ( Pr ) 2 n 1Ti O n i i 4 n 2 n (Pr O) 4 n ( NH 2 ) n 1Ti N Ti( NH 2 ) n 1( O Pr ) 4 n + 2 ph N ( n < 4) 3 p 3 The former aminolysis pathway induces anatase or rutile TiO 2 -like structure, while the latter tends to form cubic TiN structure. The latter pathway is unlikely due to the fact that in the XRD measurement only the anatase structure is observed; This mechanism is consistent with the nitrogen 1s binding energy is at ev in the XPS measurement, which is close to the typical energy for nitrogen in a compound containing the NH group, contrary to that in bulk TiN materials at ev. Thus, the nitrogen in the doped titania materials form a complex structure with titanium and hydrogen in the TiO 2 lattice structure. This mechanism is supported by the evolution of the FT-IR spectra of the samples and the status of nitrogen in these samples is consistent with the results of the recent study by Yates and co-workers. 20 p p 87

121 In the present reaction in the alcohol/amine mixture, there exists a competition between the alcohol hydrolysis and the amine aminolysis with the titanium isopropoxide precursor. The hydrolysis dominates over the aminolysis for the same reason that the alcohol acts as a stronger neucleophilic-attacking reagent than amine. Although the above reaction mechanism is different from that of the nitridation of TiO 2 crystals or films with NH 3 gas, both mechanisms can induce the same possible local nitrogen bond environment in the intermediate complex compounds. The nitridation of TiO 2 using NH 3 gas at high temperatures was studied as early as 1991 by Shin and coworkers. 39 In the nitridation process of TiO 2, titania was reduced to a fcc-phase nonstoichiometric TiO x (0.9 < x < 1.25) and followed by the gradual replacing of O with N atom. 39 The reaction can be summarized as TiO 2 TiO TiN x O y. 39 In the first step, titania was reduced to nonstoichiometric TiO, by the atomic hydrogen produced during the ammonia dissociation. 39 And the process of coreduction/nitridation can be viewed as a kinetic competition between the rates of oxygen removal and nitrogen diffusion. 39 The dissociation of ammonia on TiO 2 surface was also found by other researchers The removal of oxygen is initiated by the attack of hydrogen on the oxgen to eventually form water. Thus, based on the above studies, the substitutional nitrogen, related to a binding energy of 396 ev in the atomic β-n form 16,20,22 can be explained as the formed local Ti-N bond, while the nitrogen observed with a binding energy of ev 20 can be attributed to the NH x (x < 3) species existing in the TiO 2 lattice. These species can form local O-Ti- N-H x or O-Ti-N-Ti complex environments in the intermediate step of the nitridation of TiO 2. 88

122 We found previously that the absorption of the nitrogen-doped titania nanoparticles could be extended up to 600 nm, and the photocatalytic property could be enhanced. 19 Using x-ray photoelectron spectroscopy, we found that nitrogen atoms around ev replaced part of the oxygen atoms in the initial titania lattice, and the newly formed O-Ti-N local bond environment was responsible for the enhancement of the absorption and photocatalytic activity. 19 Here we further found that the substitutional nitrogen atom is bonded to not only the titanium atom, but also to the hydrogen atom, and is responsible for the enhanced photocatalytic activity of the doped TiO 2 nanoparticles. 4.5 Conclusions Nitrogen-doped titania nanoparticles were prepared in a mixture of alcohol/amine. The nitrogen concentration of the samples decreased when calcined at increased temperatures. The UV-visible absorption of the nitrogen-doped titania nanoparticles had a strong linear relation with the doped nitrogen concentration. The photocatalytic activity of the nitrogen-doped titania nanoparticles is monotonically related to the nitrogen concentration, and also affected by the crystal size, surface area, band structure, and surface property of the nanoparticles. The doped nitrogen is chemically bonded to the titanium atoms and interstitial hydrogen atoms in the aminolysis process. The chemistry of doping TiO 2 with nitrogen is discussed for the reaction of titanium isopropoxide with amine, which could provide a clue to manipulating the doping of titanium dioxide. 4.6 References 1. Fujishima, A.; Honda, K. Nature 1972, 37,

123 2. Tryk, D. A.; Fujishima, A.; Honda, K. Electrochimica Acta 2000, 45, Linsebigler, A. L.; Lu, G.; Yates Jr., J. T. Chem. Rev. 1995, 95, Fujishima, A.; Rao, T. N.; Tryk, D. A. J. Photochem. Photobio. C 2000, 1, Millis, A.; Hunte, S. L. J. Photochem. Photobiol. A 1997, 108, Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W. Chem. Rev. 1995, 95, Grätzel, M. Nature, 2001, 414, 338. Hagfeldt, A.; Grätzel, M. Chem. Rev. 1995, 95, Kaneko, M.; Okura, I. Photocatalysis Science and Technology Springer 2002, New York. 9. Beydoun, D.; Amal, R.; Low, R.; McEvoy, S. J. Nanoparticle Res. 1999, 1, Diebold, U. Surface Sci. Reports 2003, 48, Alivisatos, A. P. J. Phys. Chem. 1996, 100, a) Moser, J.; Grätzel, M.; Gallay, R. Helu. Chim. Acta 1987, 70, b) Grätzel, M.; Howe, R. F. J. Phys. Chem. 1990, 94, a) Choi, W.; Termin, A.; Hoffmann, M. R. J. Phvs. Chem. 1994, 98, b) Choi, W.; Termin, A.; Hoffmann, M. R. Angew. Chem. 1994, 106, c) Choi, W.; Termin, A.; Hoffmann, M. R. Angew. Chem., Int. Ed. Engl. 1994, 33, Mu, W.; Herrmann, J. M.; Pichat, P. Catal. Lett. 1989, 3, Luo, Z.; Gao, Q. H. J. Photochem. Photobiol. A 1992, 63, Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293, Khan, S. U. M.; Al-Shahry, M.; Ingler, Jr. W. B. Science 2002, 297,

124 18. Yu, J. C.; Yu, J. G.; Ho, W. K.; Jiang, Z. T.; Zhang, L. Z. Chem. Mater. 2002, 14, a) Burda, C.; Lou, Y.; Chen, X.; Samia, A. C. S.; Stout; J.; Gole, J. L. Nano Lett. 2003, 3, b) Gole, J. L.; Stout, J. D.; Burda, C.; Lou, Y.; Chen, X. J. Phys. Chem. B 2004, 108, c) Chen, X.; Burda, C. J. Phys. Chem. B 2004, 108, a) Diwald, O.; Thompson, T. L.; Goralski, E. G.; Walck, S. D.; Yates, Jr. J. T. J. Phys. Chem. B 2004, 108, b) Diwald, O.; Thompson, T. L.; Zubkov, T.; Goralski, E. G.; Walck, S. D.; Yates, Jr. J. T. J. Phys. Chem. B 2004, 108, Sakthivel, S.; Kisch, H. ChemPhysChem 2003, 4, Sano, T.; Negishi, N.; Koike, K.; Takeuchi, K.; Matsuzawa, S. J. Mater. Chem. 2004, 14, Irie, H.; Watanabe, Y.; Hashimoto, K. J. Phys. Chem. B 2003, 107, a) Lindgren, T.; Lu, J.; Hoel, A.; Granqvist, C. G.; Torres, G. R.; Lindquist, S-E. Solar Energy Mater. Solar Cells 2004, 84, 145. b) Lindegren, T.; Mwabora, J. M.; Avendaňo, J. J.; Hoel, A.; Granqvist, C. G.; Lindquist, S. -E. J. Phys. Chem. B 2003, 107, Nakamura, R.; Tanaka, T.; Nakato, Y. J. Phys. Chem. B 2004, 108, Valentin, C. D.; Pacchioni, G.; Selloni, A. Phys. Rev. B 2004, 70, Wei, H.; Wu, Y.; Lun, N.; Zhao, F. J. Mater. Sci. 2004, 39, a) Ohno, T.; Mitsui, T.; Matsumura, M. Chem. Lett. 2003, 32, 364. b) Ohno, T.; Akiyoshi, M.; Umebayashi, T.; Asai, K.; Mitsui, T.; Matsumra, M. Appl. Catal. A. 2004, 265,

125 29. a) Umebayashi, T.; Yamaki, T.; Tanaka, S.; Asai, K. Chem. Lett. 2003, 32, 330. b) Umebayashi, T.; Yamaki, T.; Itoh, H.; Asai, K. Appl. Phys. Lett. 2002, 81, 454. c) Umebayashi, T.; Yamaki, T.; Yamaki, T.; Yamamoto, S.; Miyashita, A.; Sumita, T.; Asai, K. J. Appl. Phys. 2003, 93, Choi, Y.; Umebayashi, T.; Yoshikawa, M. J. Mater. Sci. 2004, 39, Irie, H.; Watanabe, Y.; Hashimoto, K. Chem. Lett. 2003, 32, a) Yamaki, T.; Umebayashi, T.; Sumita, T.; Yamamoto, S.; Maekawa, M.; Kawasuso, A.; Itoh, H. Nucl. Instr. Meth. Phys. Res. 2003, 206, 254. b). Yamaki, T.; Sumita, T.; Yamamoto, S. J. Mater. Sci. Lett. 2002, 21, Luo, H.; Takata, T.; Lee, Y.; Zhao, J.; Domen, K.; Yan, Y. Chem. Mater. 2004, 16, Subbarao, S. N.; Yun, Y. H.; Kershaw, R.; Dwinghta, K.; Wold, A. Inorg. Chem., 1979, 18, Hattori, A.; Yamamoto, M.; Tada, H.; Ito, S. Chem. Lett. 1998, 27, Yang, S.; Gao, L. J. Am. Ceram. Soc. 2004, 87, Zhao, X.-F.; Zhang, Z.-H.; Meng, X.-F.; Liu, L.; Jia, D.-Z. Fenzi Cuihua 2003, 17, Yu, J.-G.; Yu, J. C.; Cheng, B.; Hark, S. K.; Iu, K. J. Solid State Chem. 2003, 174, Shin, C.; Bugli, G.; Djega-Mariadassou, G. J. Solid State Chem. 1991, 95, Wilson, J. N.; Idriss, H. Langmuir 2004, 20, Teramura, K.; Tanaka, T.; Yamazoe, S.; Arakaki, K.; Funabiki, T. Appl. Catal. B 2004, 53,

126 42. Siu, W. K.; Bartynski, R. A.; Hulbert, S. L. J. Chem. Phys. 2000, 113,

127 CHAPTER 5 Photocatalytic degradation of azo dyes by nitrogendoped TiO 2 nanocatalysts * Abstract: This study examined the photocatalytic degradation of three azo dyes, acid orange 7 (AO7), procion red MX-5B (MX-5B) and reactive black 5 (RB5) using a new type of nitrogen-doped TiO 2 nanocrystals. These newly developed doped titania nanocatalysts demonstrated high reactivity under visible light (λ > 390 nm), allowing more efficient usage of solar light. The doped titania were characterized by X-ray diffraction (XRD) and transmission electron microscopy (TEM). Experiments were conducted to compare the photocatalytic activities of nitrogen-doped TiO 2 nanocatalysts and commercially available Degussa P25 powder using both UV illumination and solar light. It is shown that nitrogen-doped TiO 2 after calcination had the highest photocatalytic activity among all three catalysts tested, with 95 % of AO7 decolorized in one hour under UV illumination. The doped TiO 2 also exhibited substantial photocatalytic activity under direct sunlight irradiation, with 70 % of the dye color removed in one hour and complete decolorization within three hours. Degussa P25 did not cause detectable dye decolorization under identical experimental conditions using solar light. The decrease of total organic carbon (TOC) and evolution of inorganic sulfate (SO 2-4 ) ions in dye solutions were measured to monitor the dye mineralization process. * This work has been in collaboration with Prof. Jin Li and Ms. Yang Liu at the Department of Civil Engineering and Mechanics, University of Wisconsin-Milwaukee. 94

128 5.1 Introduction Azo compounds are an important class of synthetic dyes commonly used as coloring agents in the textile, paint, ink, plastics and cosmetics industries. They are characterized by the presence of one or more azo group (-N=N-) bound to aromatic rings. The release of azo dyes into the environment is of great concern due to the coloration of natural waters, the toxicity, the mutagenicity, and carcinogenicity. 1-4 Doped TiO 2 -based nanoparticles are materials that recently demonstrated enhanced photocatalytic efficiency, when used with electron deficient model compounds such as methylene blue in water. 5 The application of these materials for relevant industrial dyes is an important step in order to establish semiconductor-based photocatalysis for environmental use. Since the discovery of its photocatalytic activity in the early 1970s, the technology of semiconductor-based photocatalysis has progressed rapidly. Along with several other metal oxide semiconductors (ZnO, Fe 2 O 3, and WO 3 ), titanium dioxide (TiO 2 ) has received considerable attention for its capability to completely mineralize recalcitrant contaminants, which can not be effectively removed by conventional methods. TiO 2 is non-toxic, efficiently photocatalytic, chemically stable and relatively inexpensive. 6-9 The primary process of semiconductor based photocatalysis can be described as follows: S + hν e-cb + h+vb where S represents a semiconductor material. Upon absorption of light energy equal to or larger than the band gap energy, a valence band electron of the semiconductor can be excited to the conduction band, leaving a positive hole in the valence band. The positive hole is a strong oxidant, which can either oxidize a compound directly or react with 95

129 electron donors like water or hydroxide ions to form hydroxyl radicals ( OH), which are also potent oxidants. Although TiO 2 has been used to degrade a large variety of organic and inorganic contaminants, 6 wide technological usage of TiO 2 is impaired by its wide band-gap (3.2 ev), which requires ultraviolet light irradiation for photocatalytic activation. Since UV light accounts for only a fraction < 10 % of the sun s energy compared to visible light (45 %), any shift in the optical response of TiO 2 from the UV to the visible spectral range will have a profound positive effect on the photocatalytic efficiency of the material Several attempts have already been made to lower the band-gap energy of crystalline TiO 2, including reduction with hydrogen and transition metal doping Both methods, however, produced only a small change in the TiO 2 band-gap. Moreover, the metal-doped materials suffer from thermal instability. 19 Current research supports the conclusion that the desired band-gap narrowing of TiO 2 can be better achieved by using anionic dopants rather than cationic metals. 5,20,21 In this paper, we report the solar photocatalytic degradation of three azo dyes, acid orange 7 (AO7), procion red MX-5B (MX-5B) and reactive black 5 (RB5), using a recently developed class of TiO 2 nanoparticles doped with nitrogen. 5 Comparisons were made between the photocatalytic activities of nitrogen-doped TiO 2 nanocatalysts, undoped TiO 2 nanoparticles, and commercially available Degussa P25 powder using UV illumination and solar light. 96

130 5.2 Experimental TiO 2 Nanoparticle Preparation and Characterization. The nitrogen-doped titania nanocatalysts were prepared by a sol-gel method. 5 Specifically, the nitrogen-doped TiO 2 nanocrystals were synthesized by adding very slowly 9 ml of Ti[OCH(CH 3 ) 2 ] 4 (Aldrich, 97%) to 150 ml ethanol and 5 ml aqueous ammonia solution. After continuous stirring of the reaction mixture at 85 C for 4 hours, a colloidal solution of TiO 2 nanocrystals is formed. After centrifugation and removing the solvent under vacuum for several hours, gray powders were obtained. Calcination at 400 ºC for 3 hours resulted in a light-yellow powder. Pure TiO 2 nanocrystals were prepared with the same procedure without the addition of ammonia during the hydrolysis process of titanium precursor. The nanocatalysts have been characterized by X-ray diffraction (XRD) and transmission electron microscopy (TEM). The X-ray diffraction patterns were obtained using a Philips PW 3710 X-ray powder diffractometer. The TEM images were obtained using a JEOL 1200EX transmission electron microscope operated at 80 kv. The specimen for TEM analysis was prepared by depositing a drop of the nanocrystal solution in water onto a copper grid supporting a thin film of amorphous carbon. The grid was dried in the air. The UV-visible reflectance spectra of the nanocrystal samples were measured on a Cary 50 UV-visible spectrometer with a reflectance unit. Azo Dyes. Acid orange 7, procion red MX-5B and reactive black were obtained from Sigma-Aldrich and were used as-received without further purification. Detailed information regarding these dyes is given in Table 5.1. All aqueous dye solutions were prepared with water from a Millipore Waters Milli-Q purification unit. 97

131 Table 5.1 Characteristics of Acid Orange 7, Reactive Black 5, and Procion Red MX-5B Dye Chemical formula M w (g/mol) Absorbance (λ) Acid Orange nm Reactive Black nm Procion Red MX-5B nm Light Sources and Photo Reactors. Both solar light and UV illumination were used as an energy source in this study. The UV light source was a 150 W high-pressure mercury lamp (UXL-151H) with maximum output between 250 and 450 nm. Experiments were carried out in Petri dishes with 10 cm diameter. Each Petri dish contained a 15 ml of dye and TiO 2 suspension. The Petri dishes were covered with plastic film to prevent evaporation of the dye solution. Neither forced aeration nor stirring of the dye solution was conducted in all the experiments. Experiments using solar light were carried out from 9 am to 5 pm during the summer season in Milwaukee, WI. For each set of experiments, eight dishes were prepared and placed under direct sunlight. One of the Petri dishes was wrapped with aluminum foil at the end of each 98

132 experimental period, and the solution was transferred into a 15 ml graduated tube covered with aluminum foil for further analysis. Sample Analysis. A broad range LUX/FC luxmeter (Sper Scientific) was used to measure the light intensity for solar experiments. Decolorization of azo dyes was determined by examining the concentration of dyes using their maximum absorbance in a UV-Vis spectrophotometer (Milton Roy, SPECTRONIC GENESYS). The concentration of sulfate (SO 2-4 ) ion was analyzed by ion chromatography (Dionex IC25). Total organic carbon (TOC) of the dye solutions was measured using a TOC monitor (Shimadzu, TOC- 5000). 5.3 Results and discussion Characterization Fig. 5.1A shows the X-ray powder diffraction (XRD) patterns of the used samples. From the intensity distribution of the X-ray diffraction signals and their integral intensity, the average nanocrystallite sizes was calculated according to the Debye-Scherer equation. 22 With reducing nanocrystal size a measured XRD pattern exhibits broader peaks. The Degussa P25 powder consists of a mixture of anatase and rutile phases in the ratio 3:1; while the nitrogen-doped TiO 2-x N x nanoparticles, before and after calcinations, display only the anatase crystal phase, which is generally considered more photoreactive. Comparing to the 30 nm average diameter of a commercial Degussa P25 sample, the average grain size of the N-doped TiO 2 is close to 10 nm after calcination, as estimated from the Debye-Scherer equation. The diffraction peaks of the nano-sized TiO 2 are broad and some peaks coalesce due to the small size of these nanoparticles. The X-ray patterns 99

133 clearly show, that before calcination, the doped TiO 2 nanomaterial is completely amorphous (curve b), while after calcination it is well crystallized (curve c). Intensity / a.u Theta / ο A c b a Absorbance / a.u. (b) (c) (a) Wavelength / nm B Figure 5.1 (A) XRD patterns of Degussa P25 TiO 2 powder (a), nitrogen-doped TiO 2 nanocatalyst before calcination (b) and after calcination at 400 C (c), and (B) UV-visible absorption spectra of Degussa P25 (a); nitrogen-doped TiO 2 (b) and nitrogen-doped TiO 2 after 400 o C (c). Figure 5.1B shows the UV-visible absorption spectra of these samples. Degussa P25 displays clear UV absorption, while nitrogen-doped TiO 2 shows huge long-tailed absorption through the visible to the near-ir before heating. Part of the huge absorption of the doped TiO 2 may be due to absorbed NH 3 on the TiO 2 nanoparticles. After heating at 400 o C, the doped TiO 2 maintained absorption in the region from 400 nm to 800 nm, though with much less intensity. This altered absorption in the visible-light region can be attributed to the change of the electronic structure of the TiO 2 due to the nitrogen dopant. 100

134 A B 50 nm 50 nm Figure 5.2. TEM images of the nitrogen-doped titania nanocatalyst before (A) and after calcination (B), the bar scale in the TEM is 50 nm. Figure 5.2 shows TEM images for the doped TiO 2 nanocatalyst before (A) and after (B) the calcination. Before calcination, the doped TiO 2 nanomaterial had gel-type and amorphous structure. After calcination, the sample crystallized and the crystallite size is centered around 10 nm. This is consistent with the results from XRD measurements. The photocatalytic activity of pure TiO 2 nanocrystals was studied extensively as a reference material and were compared with that of the doped and undoped TiO 2 nanocrystals for their photocatalytic decomposition of methylene blue using a Xe lamp and wavelength-tunable laser excitation. 5,23 We confirmed experimentally that the doped TiO 2 had a much higher photocatalytic performance than pure TiO 2 nanocrystals upon visible light excitation. In the following, we will further present the results of tests on the photocatalytic activity of doped TiO 2 under natural sun light and compare it to commercial Degussa P25 TiO 2 nanocrystals. Degussa P25 is commercially available and 101

135 is used as a standard in many publications while the performance of pure TiO 2 nanoparticles made by different groups displays varied photocatalytic activity. Thus the comparison of the doped TiO 2 with Degussa P25 provides a common standard when comparing with the results from other groups. [AO7] / mg L NC without calcination NC with calcination Degussa P25 A -ln (C/Co) B Time / hours Time / hours Figure 5.3 Decolorization of AO7 with three different nanoparticles. Condition: 0.06 mm AO7 mixed with 10 mg L -1 TiO 2 particles, with UV light illumination 800 W m AO7 decolorization by UV/doped-TiO 2 Doped TiO 2 nanomaterials and Degussa P25 were used to degrade acid orange 7 (AO7) using the UV light source. The reactions were carried out with solutions containing 0.06 mm AO7 and 10 mg L -1 catalysts. Figure 5.3A shows the relationship between dye concentration and the irradiation time for each catalyst during the photocatalytic degradation of AO7. N-doped TiO 2 nanoparticles after calcinations exhibited the best decolorization efficiency with 95 % of AO7 decolorized in 1 hour, while N-doped TiO 2 without calcination showed 75 % of conversion and Degussa P25 showed less than 40% conversion in the same test period. The kinetics of AO7 decolorization is presented in Figure 5.3B by plotting the logarithm of the normalized dye 102

136 concentration against irradiation time. Fairly good linear relationships were observed, indicating that all reactions followed the first-order kinetics. The decoloration kinetics of the dye can be rationalized by a modified Langmuir-Hinshelwood (L-H) mechanism, where dc dt kr K ec = 1+ k C e (5.1) with C being the dye concentration, k r the apparent reaction rate constant, and K e the apparent equilibrium constant for the adsorption of the dye on the nanocatalyst surface. The integrated form of Eq. (1) is: 1 C0 1 t = ln + ( C0 C) (5.2) K k C k e r r When the concentration of the dye is sufficiently low (and it is here, as well as in contaminated waste waters, << 10-3 M), equation (2) can be expressed as: C0 ' ln = kr K et = k t (5.3) C The overall rate constants for AO7 decolorization in reciprocal hours are given in the following order: k calcinated NCs (3.22 hr -1 ) > k uncalcinated NCs (1.20 hr -1 ) > k P25 (0.73 hr -1 ), indicating faster dye decolorization and higher catalytic activity for N-doped TiO 2 by a factor of ~ 4. The better performance of N-doped nanoparticles could be explained by the fact that the nitrogen doped nanoparticles were also excited with longer-wavelength light. 5 The enhanced photocatalytic activity of the calcinated sample compared to the uncalcinated one can be explained by the fact that after calcination the sample were better crystallized. Before calcination the sample is amorphous and after absorbing photons the 103

137 created electrons and holes can be deactivated in the catalyst itself before they can migrate to the surface and undergo photocatalytic reactions AO7, MX-5B and RB5 degradation by solar light/doped-tio 2 These experiments were carried out using the three different titania nanocatalysts under direct sunlight irradiation, with an initial dye concentration 0.03 mm and catalyst loading of 10 mg L -1. After 8 hours of illumination, no detectable decolorization was observed for all three dyes with P25-assisted photocatalysis. In contrast, nitrogen-doped TiO 2 nanostructures exhibited substantial photocatalytic activity within the time frame of the tests. In the following section, the decolorization of AO7, MX-5B and RB5 using N- doped TiO 2 nanocatalysts with and without calcination is presented. The disappearance of total organic carbon and evolution of final mineralization products of the three investigated azo dyes, are also described. [Dye] / mm A AO7 MX-5B RB Time / hours -ln(c/co) B Time / hours Figure 5.4 Decolorization of azo dyes by doped TiO 2 nanoparticles. Condition, 0.03 mm dyes mixed with 10 mg L -1 N-doped TiO 2 nanoparticles without calcination, illuminated by solar light with average light intensity 120 W m

138 Three dye solutions were prepared with AO7, MX-5B and RB5. Experiments were carried out under solar light radiation with an average light intensity of 120 W/m 2. The disappearance of the dye color using N-doped nanocatalysts without calcination is plotted as a function of irradiation time. As illustrated in Figure 5.4A, 53 % of the dye color was removed within 1 hour and 95 % decolorization was achieved in 3 hours. The decolorization rate was lower than that obtained with UV illumination under similar experimental conditions, as can be expected due to the absorption profile of the catalysts. It should also be noted that direct sunlight is a low intensity light source, with an average 120 W/m 2, which is several orders of magnitude lower than that used in the UV/TiO 2 experiments. The kinetics of dye decolorization is presented in Figure 5.4B. Linear regressions show that all reactions followed the apparent first-order kinetics. The overall reaction rate constants were 0.77 hr -1 for AO7, 0.76 hr -1 for MX-5B and 0.77 hr -1 for RB5. [Dye] / mm AO7 RB5 MX-5B A -ln(c/co) B Time / hours Time / hours Figure 5.5 Decolorization of azo dyes by doped TiO 2 nanoparticles. Condition, 0.03 mm azo dyes mixed with 10 mg L -1 N-doped TiO 2 nanoparticles with calcination, illuminated by solar light with average light intensity 120 W m -2. The decolorization of dye solutions using N-doped NCs with calcination is presented in Figure 5.5. The decolorization in the dye solutions reached 70 % within 1 105

139 hour, and 100 % within 3 hours, with a catalyst loading of 10 mg L -1. Three straight lines indicate that all reactions followed the first-order kinetics. The slopes giving the apparent rate constants show that k AO7 (1.44 hr -1 ) > k MX-5B (1.36 hr -1 ) > k RB5 (1.24 hr -1 ). The three dyes chosen in these experiments have similar molecular structure with the presence of electron withdrawing azo groups. However, they are different in terms of molecular weight, number of aromatic rings, number of azo bonds and the presence of halogen groups. It is interesting to note that molecular structure had no significant effect on the photodegradation kinetics of selected dyes when nanocatalysts without calcinations were used as the catalysts, suggesting that the controlling reaction mechanism was related to the properties of the catalyst. TOC disappearance and SO 2-4 evolution. The kinetics of total organic carbon (TOC) disappearance was examined for the NCs after calcination. As shown in Figure 5.6A, all three dyes followed the same pattern with about 8 % of the TOC degraded during the first hour and more markedly during the second hour of the reaction with over 60 % TOC reduction. About 80 % of carbon disappeared after 4 hours of sunlight irradiation. All reactions followed first-order kinetics verified by linear regression of f(t) = ln(c/c o ). The reaction rate constants for the three dyes are quite similar with k RB5 (0.34 hr -1 ) > k AO7 (0.32 hr -1 ) > k MX-5B (0.28 hr -1 ). These reaction rates are much lower than those for the decolorization, suggesting that the breakage of azo bond is the first step of the photocatalytic dye degradation. The relatively slow TOC reduction was most likely caused by the transformation of parent compounds into smaller organic intermediates, such as acetic acids, phenols and aldehydes, which still contribute to the TOC of the solution. The early breakdown products then underwent further oxidation leading to the 106

140 production of CO 2. The presence of a low but constant level of TOC in solution after extended irradiation, suggests the accumulation of photocatalytic reaction end products that can not be completely mineralized to water and CO 2 during that time. [TOC] /mg L AO7 MX-5B RB5 A [SO 4 2- ] / mg L B Time / hours Time / hours Figure 5.6 TOC disappearance (A) and SO 2-4 generation (B) of azo dyes. Condition, 0.03 mm azo dyes mixed with 10 mg L -1 N-doped TiO 2 nanoparticles with calcination, illuminated by solar light with average light intensity 120 W m -2. During the course of dye degradation, the inorganic anion SO 4 2- formed progressively. As shown in Figure 5.6B, the amount of SO 2-4 increased more than 30% per hour, which is faster than the CO 2 (TOC) disappearance rate but slower than the decolorization rate. RB5 shows a 7.43 mg L -1 initial sulfate ion concentration, which can be explained by the presence of some impurity in the RB5 sample. Dyes are often degraded rather rapidly under solar illumination in the presence of TiO 2 due to the electron injection mechanism. From the results above, it seems that this mechanism is not operative here due to the electronic nature of the dyes: The dyes, which are electron acceptors compared to the doped TiO 2 nanoparticles, have too low-lying energy levels to be used for photoinduced charge injection. Figure 5.7 shows the energy scheme of TiO 2 (bulk) and the three dyes as calculated by the semi-empirical PM3 107

141 method. It is apparent that the energetic match between the nanoparticles and those of the dyes is poor, which prevents the efficient charge injection from the dyes to the nanoparticle photocatalyst. The HOMO and LUMO of these dyes are located in the bandgap of TiO 2, thus the charge injection mechanism from the dyes to the nanoparticle photocatalyst is probably not efficient. 0 Energy / ev CBE LUMO LUMO LUMO HOMO HOMO VBE HOMO TiO 2 AO7 RB5 PR MX-5B Figure 5.7 Energy Scheme of TiO 2 (bulk) and the three dyes as calculated by the semiempirical AM1 method. CBE: conduction band edge, VBE: valence band edge, HOMO: the highest occupied molecular orbital, LUMO: the lowest unoccupied molecular orbital. 5.4 Conclusions The effect of various nanometer sized photocatalysts on the photodegradation of three azo dyes, acid orange 7 (AO7), procion red MX-5B (MX-5B) and reactive black 5 (RB5) was examined. The nitrogen doped titania nanoparticles showed significant improvement in reactivity compared to Degussa P25 particles and to pure TiO 2 nanoparticles, particularly in the visible range. The concentration dependence of the different constituents was found and the reaction kinetics followed an apparent first-order reaction and was rationalized by a Langmuir-Hinshelwood type mechanism. The 108

142 superior photoactivity of doped nanocatalysts could be attributed to their smaller grain sizes and improved visible light adsorption. Calcinated nanoparticles had better performance than uncalcinated ones due to better crystallization as evidenced by TEM studies. 5.5 References 1. Oh, S. W.; Kang, M. N.; Cho, C. W.; Lee, M. W. Dyes Pigments 1997, 33, Stylidi, M.; Kondarides, D. I.; Verykios, X. E., Appl. Catal. B: Environ. 2003, 40, Li, J.; Bishop, P. L., Wat. Sci. Technol., 2002, 46, Li, J.; Bishop, P. L. Wat. Sci. Technol. 2004, 49, Burda, C.; Lou, Y.; Chen, X.; Samia, A. C. S; Stout, J.; Gole, J. L., Nano Letters. 2003, 3, Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W. Chem. Rev. 1995, 95, Andrew, M.; Stephen, L. H., J. Photochem. Photobiol. A: Chem. 1997, 108, Silva, C. A.; Madeira, L. M.; Boaventura, R. A.; Costa, C. A. Chemosphere. 2004, 55, Son, H. S., Lee, S. J., Cho, I. H., and Zoh, K. D., Chemosphere 2004, 57, Heller, A. Accounts Chem. Res. 1995, 28, Linsebigler, A. L.; Lu, G.; Yates, J. T., Chem. Rev. 1995, 95, Sauer, M. L.; Ollis, D. F. J. Catal. 1996, 163,

143 13. Khader, M. M. K., F. M.N.; El-Anadouli, B.E.; Ateya, B.G., J. Phys. Chem. 1993, 97, Kilic, C.; Zunger, A., Appl. Phys. Letters 2002, 81: Palmer, R. A.; Doan, T. M.; Lloyd, P.G.; Jarvis, B.L.; Ahmed, N.U. Plasma Chem. and Plasma Process. 2002, 3, Wang, C.; Bahnemann, D. W.; Dohrmann, J. K. Chem. Comm. 2000, 16, Shah, S. I.; Li, W.; Huang, C. P.; Jung, O.; Ni, C. Proceed. National Academy of Sciences 2002, 99, Liu, G.; Zhang, X.; Xu, Y.; Niu, X.; Zheng L.; Ding, X. Chemosphere 2004, 55, Di Paola, A.; Ikeda, S.; Marci, G.; Ohtani, B. Intern. J. Photoenergy. 2001, 3, Khan, S. U. M.; Al-Shahry, M.; Ingler, W. B. Jr.. Science 2002, 297, Yu, J. C.; Yu, J. G.; Ho, W. K.; Jiang, Z. T.; Zhang, L. Z. Chem. Mater. 2002, 14, Pielaszek, J X-ray Diffraction from Nanostructured Materials, in Nanostructured Materials: selected synthesis methods, properties and applications, Editors: Philippe Knarth, Joop Schoonman, Kluwer Academic Publishers, Boston, Gole, J. L.; Stout, J. D.; Burda, C.; Lou, Y.; Chen, X. J. Phys. Chem. B 2004, 108,

144 CHAPTER 6 Visible-light Sensitive C-, N- and S-Doped Titanium Dioxide Oxidized from Titanium Carbide, Nitride and Sulfide Abstract Visible-light sensitive titanium dioxide becomes very important in the expansion of its applications in the visible-light region. In this paper, C-, N- and S-doped titanium dioxide materials were prepared from titanium carbide, nitride and sulfide compounds by oxidation at elevated temperatures. Their optical, structural and chemical properties in the transformation were investigated with x-ray diffraction, x-ray photoelectron emission, FT-infrared, Raman and UV-visible absorption spectroscopy. The transformation temperatures from anatase into rutile structure of the C-, N- and S-doped titanium dioxide were different, caused by C, N and S dopants in the doped TiO 2. The Raman vibrational modes of the doped TiO 2 were also modified by the C, N and S dopants to different extents. The structural changes can be tracked by FT-IR spectra in addition to x-ray diffraction pattern changes. These C-, N- and S-doped TiO 2 had a common bandedge absorption at 3.0 ev, and long-wavelength absorption in the visible-light region. The long-wavelength absorption of these doped titanium dioxide can be clearly attributed to the modification of the electronic states above the valence band of TiO 2 brought by the C, N and S dopants as revealed from the valence band XPS. This study promotes our current understanding and research effort in the doping chemistry and physics of TiO 2 for visible-light applications. 111

145 6.1 Introduction Titanium dioxide (TiO 2 ) has been a fascinating material since late 1960s, due to its applications in various areas such as pigments, photoconductors, dielectric materials and photocatalysts. 1-7 The wide application of TiO 2 as a photocatalyst is limited by its UV or near-uv absorption properties. 8-9 Thus much effort has been devoted recently to improve the optical response of TiO 2 in the visible region. Doping TiO 2 with different elements such as nitrogen, carbon, sulfur has displayed promising results in visible-light photocatalysis, where visible-light absorption (> 400 nm) has been achieved Titanium nitride (TiN) and titanium carbide (TiC) are good metallic conductors, with a partially filled band and a chemical bond simultaneously of metallic, covalent, and ionic character. 15 Titanium sulfide (TiS 2 ) is regarded as a semiconductor or semimetal, 16,17 with a narrow bandgap of about 0.9 ev. 18 In the optical spectrum, these materials show absorption from IR via visible into UV regime. These compounds can be oxidized into TiO 2 at high temperatures in air or oxygen Thus, it is reasonable to postulate that by oxidizing these materials under proper conditions, one can obtain C-, N- and S-doped TiO 2, which potentially has visible-light absorption properties. Here, we investigated the synthesis and properties of C-, N- and S-doped TiO 2 materials with visible-light absorption by oxidizing titanium carbide, nitride and sulfide at elevated temperatures. Different techniques were employed to elucidate the chemical and structural changes of this oxidation and to unveil the physical origin of the visiblelight absorption properties of the doped TiO 2 caused by the C, N and S dopants. This study would promote the current understanding and research effort in the doping chemistry and physics of TiO 2 for visible-light applications. 112

146 6.2 Experimental The titanium nitride (TiN), carbide (TiC) and sulfide (TiS 2 ) were obtained from Strem Chemicals and used as obtained. The C-, N- and S-doped titanium dioxide (TiO 2 ) samples were prepared by heating the above reagents at o C for 6 hours. About 0.5 g TiX powder was loaded in a ceramic sample boat (4.0/l 0.5/h 0.5/w in inch), and then placed in the middle of a quartz tube in a Lindberg tube furnace. The two ends of this tube were left open to the atmosphere. The temperature was slowly ramped up and cooled down at a rate of 2ºC/min during the heating and cooling processes. After cooling to room temperature, the samples were taken out for further investigations. The UV-visible reflectance spectra were measured on a Cary 50 UV-visible spectrometer with a reflectance unit. X-ray diffraction (XRD) patterns were obtained on a Philips PW 3710 X-ray powder diffractometer. For x-ray photoelectron spectroscopy (XPS) measurements, a Perkin-Elmer PHI 5600 XPS System was used. Samples for XPS measurement were coated on carbon tape attached to the sample holder. FT-IR spectra were obtained on a ThermaNicolet 870 FT-IR instrument. The samples were mixed with KBr powder (5:95 in volume ratio) and pressed into thin pellets. Raman spectra were recorded in a Raman microscope system. A small amount of the sample was put over an aluminum tray in the microscope stage. A nm krypton laser beam (50 mw) is introduced in the microscope s optical axis using a fiber optic, and focused on the samples to generate the Raman scattering. 180 back-scattered Raman light was collected from the sample by using a long-focal-length X20 microscope objective lens. The sample can be viewed via a long focal length objective and a charge-coupled device (CCD) camera for optical imaging. The Raman microscope was calibrated by using standardized 113

147 neon and tungsten lamps before the measurement. Data analysis was performed using GRAMS/32 software (Galactic Industries, Inc.). 6.3 Results and Discussion Structural Transformation TiN and TiC have a NaCl-like cubic structure, 22 while TiS 2 has a CdI 2 -like layered structure XRD is used to determine the structure of materials. 26 Figure 6.1 shows the XRD pattern changes after TiC (A), TiN (B) and TiS 2 (C) powders oxidized at different temperatures from 350 o C to 1000 o C for 6 hr. The XRD pattern of TiC only A Intensity / a.u. o 1000 o o o C o o o o oo o o 900 o C 800 o C 700 o C # 600 o C 500 o C 400 o C # # * 350 o C TiC * * * B Intensity / a.u. o o 1000 o C 900 o C 800 o C 700 o C 600 o C # 500 o C * 400 o C 350 o C TiN Theta / O 2 Theta / O C Intensity / a.u. o o 1000 o C o 900 o C o o o oo o o # 800 o C 700 o C 600 o C 500 o C 400 o C ## # ## # # # 350 o C * TiS 2 * * * * * * * o o o o o o o oo * * * 2 Theta / o Figure 6.1 XRD pattern evolution after TiC (A), TiN (B) and TiS 2 (C) powder heated at different temperatures from 350 o C to 1000 o C for 6 hr. (#) anatase TiO 2 (o) rutile TiO 2, (*) TiC in (A), TiN in (B), and TiS 2 in (C). 114

148 showed (111), (200), (220), (311) and (222) peaks due to its rock-salt structure, while after heating, the XRD pattern gradually changed into anatase or rutile phase TiO 2. This transformation already started after heating at 350 o C. After heating at above 500 o C, the structure of TiC transformed into anatase- and rutile-phase TiO 2 or the mixture. The anatase TiO 2 diffraction pattern completely disappeared, and only rutile TiO 2 patterns were detected after heating at temperature above 700 o C. The widths of the diffraction peaks are inversely proportional to the size of the crystal grains in the sample as expressed by the Scherrer s equation. 26,27 The width of the diffraction peaks of these TiO 2 powders became narrower as the oxidizing temperature increased. This suggests that at higher temperatures, TiC powders transformed into TiO 2 structures with bigger grain sizes. The structural transformations of TiN and TiS 2 into the TiO 2 structure are very similar to that of TiC, with some minor differences. In the case of TiN, first, TiN needed higher temperature (700 o C 6 hrs) than TiC (600 o C 6 hrs) to transform into the TiO 2 structure at similar reaction environments, based on the result that TiN residue was detected in the XRD patterns at 600 o C for 6 hrs, while TiC was not detected. Second, the crystal-grain sizes of the TiO 2 structures after heating are bigger, derived from the narrower diffraction peaks. Third, rutile TiO 2 structure rather than anatase TiO 2 occurred in the transformation of TiN from the higher rutile/anatase ratios. In the case of TiS 2, below 800 o C, only the anatase TiO 2 structure patterns was detected, while at 800 o C, it mostly transformed into the rutile phase TiO 2, with some anatase TiO 2 structure, and above 800 o C, only the rutile TiO 2 structure were formed. The transformation from 115

149 anatase into rutile phase of the TiO 2 structure from TiS 2 occurred at a higher temperature (800 o C-900 o C) than that from TiC (600 o C-700 o C) and TiN (500 o C-600 o C). Thus, the same type of structure, anatase TiO 2, derived from different sources, showed different thermal stabilities. The anatase TiO 2 structure from TiS 2 showed the highest stability, than that from TiC, and that from TiN. The thermal stability difference of these TiO 2 structures can be attributed to the different influence on the bonding structure in these samples brought by the C, N and S dopants Chemical Composition Transformation The chemical transformation often occurred at the interface between the surface Intensity / a.u. A O KLL Ti 2S O 1S Ti 2P C 1S TiC 1000 O C TiC 900 O C TiC 800 O C TiC 700 O C TiC 600 O C TiC 500 O C TiC Ti 3S Ti 3P Intensity / a.u. B Ti LMM O KLL O 1S Ti 2P Ti 2S N 1S C 1S 1000 o C 900 o C 800 o C 700 o C 600 o C 500 o C TiN Ti 3P Ti3S Binding Energy / ev Binding Energy / ev Intensity / a.u. C Ti LMM O KLL O 1s Ti 2p C 1s 1000 o C 900 o C 800 o C 700 o C 600 o C 500 o C TiS 2 S 2p Ti 3p Ti 3s Binding Energy / ev Figure 6.2 XPS pattern changes after TiC, TiN and TiS 2 was heated at different temperatures from 500 o C to 1000 o C before Ar sputtering. 116

150 of the material and the active atmosphere. Thus, the surface composition change can be a front-sign of the bulk chemical composition change. XPS is used to determine the chemical composition of the materials, i.e. the surface composition Figure 6.2 shows the XPS patterns for TiC powder and the resultant TiO 2 powders after heating at different temperatures. The XPS binding energies from the samples were calibrated with respect to the C 1s peak from the carbon tape at ev. For TiC powder, the carbon 1s binding energy occurred at ev, 31,32 and the titanium 2p binding energy occurred at ev (for 2p 3/2 ). 31,32 After heating above 500 o C, the carbon 1s signal at ev completely disappeared, while the titanium 2p signals moved to ev (for Ti 2p 3/2 ) and (for Ti 2p 1/2 ), which are the typical binding values of Ti 4+ in TiO 2. 28,31,32 And the only peak centered at ev for oxygen 1s suggested the chemical composition of TiO 2. In the case of TiN and TiS 2, their chemical composition changes after heating were similar to that of TiC, from the XPS measurements. For TiN, after heating above 500 o C, the nitrogen 1s peak at ev completely disappeared, and the two peaks at ev for Ti 2p 3/2 and ev for Ti 2p 1/2 as Ti 3+ in the TiN moved to ev and ev, respectively as Ti 4+ in the resultant TiO ,33 For TiS 2, after heating, the sulfur 2s and 2p peak at ev and ev completely disappeared. 31 Thus, the surfaces of these materials are completely transformed into pure TiO 2. With the depth profiling method, the bulk composition can also be obtained. Depth profiling was performed on the samples after heating them at 1000 o C for 6 hr. Small amounts of the C, N and S signals were detected in the XPS spectra shown in Figure 6.3 after the samples were sputtered with Ar + ions for 10 min. The sputtering rate was at 57Å/min. In Fig. 6.3A, the signal at a binding energy of ev can be 117

151 Intensity / a.u. A C 1s XPS (ii) Carbon tape Ti-C Intensity / a.u. B N1s XPS (ii) (i) Ti-N (i) Binding Energy / ev Binding Energy / ev Intensity / a.u. C (iv) (i) S2p XPS Binding Energy / ev Ti-S Intensity / a.u. D Ti2p XPS Binding Energy / ev (vi) (iii) (ii) (i) Figure 6.3 After Ar + sputtering for 10 min, core-level XPS spectra of the C-, N- and S- doped TiO 2 compared to that of pure TiO 2. (A) C 1s binding energy region for pure TiO 2 and TiO 2-y C y, (B) N 1s binding energy region for pure TiO 2 and TiO 2-x N x, (C) S 2p binding energy region for pure TiO 2 and TiO 2-z S z ; (D) Ti 2p binding energy region for pure TiO 2 and TiO 2-y C y, TiO 2-x N x and TiO 2-z S z ; (i) prue TiO 2, (ii) C-doped TiO 2, (iii) N- doped TiO 2, (iv) S-doped TiO 2 derived from TiC, TiN and TiS 2 powder after heated at 1000 o C for 6hr. attributed to C1s electrons from the carbon tape, and the signal at ev can be attributed to the Ti-C bond in the the C-doped TiO 2 sample 14,34. In Figure 6.3B, the signals at ev and ev in the N- doped TiO 2 sample can be attributed to N1s electrons having Ti-N bonds within different bonding environments 8,9,33, In Figure 118

152 6.3C of the S-doped TiO 2, the signal at and can be attributed to the S2p electrons in the S-Ti bond and the formed SO 2 species trapped in the lattice during the oxidation 40,41. The peaks in the Ti2p 3/2 XPS spectra in Fig. 6.3D at binding energies of ,42, , and ev 42 verified the O-Ti, C-Ti, N-Ti, and S-Ti bonds in the C-, N- and S-doped TiO 2 samples, respectively. Overall, the C, N and S dopant in these doped TiO 2 samples are about 0.5%, 0.2% and 4.2% atomic percent. Combined with the results from XRD, the composition of these samples can be summarized as: the samples derived at a temperature higher than 400 o C from TiC, 500 o C from TiN and 350 o C from TiS 2 were C-, N- and S-doped TiO 2 ; the other samples were a mixture of doped anatase and/or rutile TiO 2 with the TiC, TiN or TiS Lattice Vibrational Transformation As the chemical composition and structure changes, the vibrational motion in the materials also changes. Thus, the vibrational motions in materials provide information about their chemical composition and structure. Figure 6.4 shows the evolution of the FT- IR spectra for TiC, TiN and TiS 2 and the resultant doped TiO 2 powders after heating at different temperatures. TiC, TiN and TiS 2 displayed high absorbance from 800 cm -1 to 4000 cm -1. After heating, the resultant TiO 2 powders displayed high transmittance in the region from 700 cm -1 to 2000 cm -1. The signals in the range cm -1 are characteristic of the formation of a Ti-O-Ti lattice. The evolution of the FT-IR spectra from TiN and TiS 2 to the resultant TiO 2 after heating was similar to that of TiC. The fine vibrational motion of the Ti-O-Ti lattice were more pronounced in the region of 500 cm

153 Transmittance % a c b Wavenumber / cm -1 A 1000 o C 900 o C 800 o C 700 o C 600 o C 500 o C 400 o C 350 o C TiC Transmittance % c b a Wavenumber / cm -1 B 1000 o C 900 o C 800 o C 700 o C 600 o C 500 o C 400 o C 350 o C TiN Transmittance % d c Wavenumber / cm -1 Figure 6.4 FT-IR spectrum evolution from TiC, TiN and TiS 2 powder to the resultant C-, N- and S-doped TiO 2 after heating at different temperatures from 350 o C to 1000 o C. Peaks a, b, c, d can be assigned to the vibrations of Ti-O bond in anatase and rutile phases of TiO 2, and Ti-S bond in layered TiS 2. See text for details. b C 1000 o C 900 o C 800 o C 700 o C 600 o C 500 o C 400 o C 350 o C TiS 2 to 1000 cm -1 in the FT-IR spectra of the TiO 2 made from TiS 2. The IR absorption peak around 830 cm -1 (peak a) can be attributed to the Ti-O vibrations in anatase phase TiO 2. The IR absorption peak around 530 cm -1 (peak b) can be attributed to the Ti-O vibrations in rutile phase TiO 2. The IR absorption peak around 750 cm -1 (peak c) can be attributed to common Ti-O vibrations in both anatase and rutile phase TiO 2. The IR absorption peak around 1130 cm -1 (peak d) can be attributed to the Ti-S vibrations in layered TiS 2. Thus, the structural changes between different phases are reflected in the IR absorption peaks, 120

154 since the different arrangements of atoms produce different bond structures and bond strengths. Conversely, from the different IR peaks, different structures can be recognized. TiO 2-x S x Intensity / a.u. TiO 2-X N X TiO 2-x C x Wavenumber / cm -1 Figure 6.5 Raman spectra for the resultant C-, N- and S-doped TiO 2 powders from TiC, TiN and TiS 2 after heating at 1000 o C. Table 6.1 Raman vibrational mode frequencies of the TiO 2-x C x, TiO 2-x N x and TiO 2-x S x compared to pure rutile TiO 2 Second-order process E g A 1g position/cm -1 width/cm -1 position/cm -1 width/cm -1 position/cm -1 width/cm -1 TiO TiO 2-x C x TiO 2-x N x TiO 2-x S x Complementary to IR spectra, Raman is a powerful technique to explore the vibrational modes and to characterize materials. For TiC, there are three Raman 121

155 vibrational modes around 270 cm -1, 450 cm -1, and 860 cm For TiN, there are four Raman peaks around 235cm -1 (transverse acoustic-ta), 320 cm -1 (longitudinal acoustic- LA), 440 cm -1 (second-order acoustic-2a), and 570 cm -1 (transverse optical-to). 47 For TiS 2, there are two Raman peaks around 150cm -1, 230 cm -1, 330 cm -1, and 380 cm ,49 For rutile TiO 2, there are Raman active modes at 448 cm -1 (E g ) and 613 cm -1 (A 1g ). 46,47 The results of the Raman measurements are shown in Figure 6.5 and summarized in Table 6.1 for the resultant TiO 2 obtained after heating at 1000 o C for 6 hrs. For all the three samples, there were three peaks around 245 cm -1, 430 cm -1, and 606 cm -1, which is attributed to the TiO 2 rutile structure. The peak around 245 cm -1 was attributed to a second-order process. This vibrational mode could be attributed to the second-order process. The small difference in the positions of these peaks reflects the difference in the vibrational motions, possibly due to the structural difference because of the possible lattice distortion brought by the C, N and S dopants. The widths of these peaks for the TiO 2 increased in the order TiS 2 > TiN > TiC. This suggested that the size of these TiO 2 decreased in the order of TiC < TiN < TiS 2, since the width of the peak results from the finite lifetime of the vibrational modes in the small size particles and is inversely proportional to the size of the particle, 50,51 and may be also possibly related to the dopants influence on the lattice distortion UV-Visible Absorption Transformation The above structural changes of these materials directly lead to the optical and electronic property changes. TiN and TiC are good metallic conductors with partially filled bands. TiS 2 is regarded as a semiconductor or semimetal, with a narrow bandgap of 122

156 Absorbance / a.u o C 900 o C 800 o C 700 o C 600 o C 500 o C 450 o C 350 o C TiC A Absorbance / a.u o C 900 o C 800 o C 700 o C 600 o C 500 o C 400 o C 350 o C TiN B Wavelength / nm Wavelength / nm Absorbance / a.u o C 900 o C 800 o C 700 o C 600 o C 500 o C 400 o C 350 o C TiS 2 C Wavelength / nm Figure 6.6 Absorption spectra of TiC, TiN and TiS 2 powders and the resultant TiO 2 particles after heating at different temperatures higher than 500 o C. about 0.9 ev. Optical techniques, i.e. absorption spectroscopy help to elucidate the optical properties, the bandgap and the electronic structure of materials. TiN, TiC and TiS 2 have gray, black and black-gray color, respectively. In the optical spectrum, they show absorption from IR via visible into UV regime. After heating, these samples changed into yellow or slight yellow color (from TiS 2 ). Figure 6.6 shows the optical reflectance spectra of the commercial TiC, TiN and TiS 2 powder and the resultant C, N- and S-doped TiO 2 particles after oxidizing at different temperatures from o C for 6 hrs. The samples after heating TiC at 350 o C and 400 o C, and those 123

157 heated at 350 o C-500 o C from TiN, were mixtures of C- or N-doped TiO 2 and TiC or TiN, consistent with the XRD results. For the samples from TiC, after heating they displayed a bandgap around 3.0 ev due to the chemical and structural changes. Besides the bandgap absorption, these C-doped TiO 2 materials show a lower energy absorption from near-ir to the visible. The low-energy absorption region of the C-doped TiO 2 can be divided into two parts. The first part is from 800 to 580 nm, where only weak absorptions were observed. The second part expands from 580 to 400 nm, where substantial absorption was observed for all the C-doped TiO 2 samples after heating. The relative absorption intensity compared to the bandgap absorption at 400 nm increased as the heating temperature increased from 350 o C to 1000 o C. Also, the bandgap increased slightly as the heating temperature increased. The absorption changes of the commercial TiN and TiS 2 powders and the resultant N- and S-doped TiO 2 particles after heating at different temperatures are analogous to those of TiC. The difference is that the N-doped TiO 2 particles showed a higher absorption than the C-doped TiO 2 made from TiC, while the S-doped TiO 2 particles showed a lower absorption. Besides, the S-doped TiO 2 samples by heating TiS 2 at 500 o C to 700 o C displayed a bandgap around 3.3 ev, while the samples heated above 800 o C displayed a bandgap around 3.0 ev. This was due to the different phases (anatase and rutile, respectively) of TiO 2 formed as shown by XRD results. The optical properties of a material reflects the properties of its electronic structure. Thus, these optical changes are attributed to the electronic structural changes due to the chemical composition and structural changes from titanium carbide, nitride and sulfide into C-, N- and S-doped TiO

158 6.3.5 Valence Band Transformation TiC, TiN and TiS 2 have gray to black color with absorption from IR via visible into UV regime as shown above. This is determined from their electronic structures. This can be directly seen from the valence band XPS spectra shown in Figure 6.7A. Both TiC and TiN have a valence band at or near 0 ev in binding energy compared to the Intensity / a.u. A (iv) (iii) (ii) (i) Intensity / a.u. B (iv) (iii) (ii) (i) Binding Energy / ev Binding Energy / ev Figure 6.7 Valence band (VB) XPS of (A) TiO 2 (i), TiC (ii), TiN (iii) and TiS 2 (iv); (B) (i) pure rutile TiO 2 and the (ii) C-, (iii) N- and (iv) S-doped TiO 2 derived from TiC, TiN and TiS 2 respectively. valence band edge at 3.0 ev of pure rutile TiO 2, while TiS 2 displayed a band edge around 0.5 ev. These are consistent with the previous study that TiN and TiC are good metallic conductors with a partially filled band, 15 and TiS 2 is a semiconductor or semimetal 16,17 with a narrow bandgap of about 0.9 ev. 18 As shown in Figure 6.7B, compared to pure rutile TiO 2 with the valence band edge at 3.0 ev, TiO 2-x C x, TiO 2-y N y and TiO 2-z S z display additional electronic states above the valence band edge. These additional states can be attributed to C, N and S dopants in these doped materials, as compared with the electronic states of pure TiO 2 and TiC, TiN and TiS 2 as shown. The valence band XPS spectra exactly explain the absorption spectra of these doped TiO 2, with the absorption at 125

159 3.0 ev corresponding to the optical transitions involved with the bandedge, and the longtail absorption in the visible region to the optical transitions with the intra band electronic states brought by the C, N and S dopants. 6.4 Conclusions In this study, the chemical, structural and optical property changes of TiO 2-x C x, TiO 2-y N y and TiO 2-z S z powders derived from TiC, TiN and TiS 2 after oxidizing from 350 o C to 1000 o C in atmosphere were investigated with XRD, XPS, FT-IR, Raman and UV-visible absorption spectroscopy. The TiO 2-x C x, TiO 2-y N y and TiO 2-z S z powders showed different structural stabilities on transformation from anatase into rutile, depending on the source. The structural evolution can be elucidated from the XRD patterns and the FT-IR patterns. The C, N and S dopants displayed the modification of the Raman vibrational modes of pure TiO 2. The TiO 2-x C x, TiO 2-y N y and TiO 2-z S z displayed bandgap absorption around 3.0 ev of the typical rutile TiO 2 and visible-light absorption from 400 nm to 800 nm caused by the C, N and S dopants. Valence band XPS clearly revealed the additional electronic states in the bandgap caused by these dopants, compared to pure TiO 2. Thus it clearly revealed the physical origin for the visible-light absorption property of these doped TiO 2. This study would promote the current understanding and research effort in the doping chemistry and physics of TiO 2 for the visible-light applications, and stimulate our further effort to tune the properties of TiO 2 for our needs. 126

160 6.5 References 1. Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W. Chem. Rev. 1995, 95, Grätzel, M. Nature, 2001, 414, Millis, A.; Hunte, S. L. J. Photochem. Photobiol. A, 1997, 108, Linsebigler, A. L.; Lu, G.; Yates Jr., J. T. Chem. Rev. 1995, 95, Fujishima, A.; Rao, T. N.; Tryk, D. A J. Photochem. Photobio. C 2000, 1, Tryk, D. A.; Fujishima, A.; Honda, K. Electrochimica Acta 2000, 45, Choi, H. C.; Ahn, H.-J.; Jung, Y. M.; Lee, M. K.; Shin, H. J.; Kim, S. B.; Sung, Y-E. Appl. Spectros. 2004, 58, Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293, Burda, C.; Lou, Y.; Chen, X.; Samia, A. C. S.; Stout, J.; Gole, J. L. Nano Lett. 2003, 3, Miyauchi, M.; Ikezawa, A.; Tobimatsu, H.; Irie, H.; Hashimoto, K. Phys. Chem. Chem. Phys. 2004, 6, Sakthivel, S.; Kisch, H. Chem. Phys. Chem. 2003, 4, Ohno, T.; Mitsui, T.; Matsumura, M. Chem. Lett. 2003, 32, Umebayashi, T.; Yamaki, T.; Tanaka, S.; Asai, K. Chem. Lett. 2003, 32, Irie, H.; Watanabe, Y.; Hashimoto, K. Chem. Lett. 2003, 32, Fischer, D. W.; Baun, W. L. J. Appl. Phys. 1968, 39,

161 16. Faba, M. G.; Gonbeau, D.; Pfister-Guillouzo, G. J. Electron Spectros. Related Phenom. 1995, 73, Wu, Z. Y.; Lemoigno, F.; Gressier, P.; Ouvrard, G.; Moreau, P.; Rouxel, J. Phys. Rev. B 1996, 54, R Fischer, D. W. Phys. Rev. B 1973, 8, (a) Shimada, S. J. Mater. Sci. 1996, 31, 673. (b) Stewart, R. W.; Cutler, I. B J. Am. Ceram. Soc. 1967, 50, 4, (a) Esaka, F.; Furuya, K.; Shimada, H.; Imamura, M.; Matsubayashi, N.; Sato, H.; Nishijima, A.; Kawana, A.; Ichimura, H.; Kikuchi, T. J. Voc. Sci. Technol. A 1997, 15, (b) Soriano, L.; Abbate, M.; Pen, H.; Czyżyk, M. T.; Fuggle, J. C. J. Electron Spectros. Related Phenom. 1993, 62, 197. (c) Soriano, L.; Abbate, M.; Fuggle, J. C.; Prieto, P.; Jiménez, C.; Sanz, J. M.; Galán, L.; Hofmann, S. J. Voc. Sci. Technol. A 1993, 11, (a) Basu, S. K.; Taniguchi, M. Therm. Acta. 1986, 109, 253. (b) Sriram, M. A.; Kumta, P. N. J. Am. Ceram. Soc. 1994, 77, Didziulis, S. V.; Lince, J. R.; Stewart, T. B.; Eklund, E. A. Inorg. Chem. 1994, 33, Ohno, Y.; Hirama, K.; Nakai, S.; Sugiura, C.; Okada, S. Phys. Rev. B 1983, 27, Fischer, D. W. Phys. Rev. B 1973, 8, Šimůnek, A, Šipr, O, Bocharov, S.; Heumann, D.; Dräger, G. Phys. Rev. B 1997, 56, Klug, H. P.; Alexander, L. E. X-ray Diffraction Procedures for Polycrystalline and Amorphous Materials, 2 nd. Ed., Wiley: New York,

162 27. Pielaszek, J. in Nanostructured Materials: selected synthesis methods, properties and applications (Eds: P. Knarth, J. Schoonman), Kluwer Academic Publishers, Boston, Hashimoto, S.; Murata, A.; Sakurada, T.; Tanaka, A. J. Surf. Anal. 2002, 9, Yoshitake, M.; Thananan, A.; Aizawaki, T.; Yoshihara, K. Surf. Interf. Anal. 2002, 34, Wang, Y. L. Surf. Coat. Tech. 2002, 150, NIST XPS Database, Zhang, L.; Koka, R. V. Mater. Chem. Phys. 1998, 57, Saha, N. C.; Tompkins, H. G. J. Appl. Phys. 1992, 72, Choi, Y., Umebayashi, T., Yoshikawa, M. J. Mater. Sci. 2004, 39, Diwald, O., Thompson, T. L., Goralski, E. G., Walck, S. D., Yates, Jr. J. T. J. Phys. Chem. B 2004, 108, Diwald, O., Thompson, T. L., Zubkov, T., Goralski, E. G., Walck S. D., Yates, Jr. J. T. J. Phys. Chem. B 2004, 108, Chen, X., Burda, C. J. Phys. Chem. B 2004, 108, Gole, J. L., Stout, J. D., Burda, C., Lou, Y., Chen, X. J. Phys. Chem. B 2004, 108, Irie, H., Watanabe, Y., Hashimoto, K. J. Phys. Chem. B 2003, 107, Umebayashi, T., Yamaki, T., Yamamoto, S., Miyashita, A., Tanaka, S., Sumita, T., Asai, K. J. Appl. Phys. 2003, 93, Umebayashi, T., Yamaki, T., Itoh, H., Asai, K. Appl. Phys. Lett. 2003, 81, Martinez, H., Auriel, C., Gonbeau, D., Loudet, M., Pfister, G. G. Appl. Surf. Sci. 1996, 93,

163 43. Choi, Y., Umebayashi, T., Yamamoto, S., Tanaka, S. J. Mater. Sci. Lett. 2003, 22, Park, H. K.; Kim, D. K.; Hee, C. J. Am. Ceram. Soc. 1997, 80, Zheng, M.; Gu, M.; Jin, Y.; Jin, G. Mater. Sci. Engin. 2000, B77, Begun, G. M.; Bamberger, C. E. Appl. Spectrosc. 1989, 43, Cheng, Y. H.; Tay, B. K.; Lau, S. P.; Kupfer, H.; Richter, F. J. Appl Phys. 2002, 92, Carmalt, C. J.; Dinnage, C. W.; Parkin, I. P. J. Mater. Chem. 2000, 10, Unger, W. K.; Reyes, J. M.; Singh, O.; Curzon, A. E.; Irwin, J. C.; Frindt, R. F. Solid State Comm. 1978, 28, Bersani, D.; Antonioli, G.; Lottici, P. P.; Lopez, T. J. Non-Crystalline Solids 1998, , Ivandaa, M.; Musić, S.; Gotić, M.; Turković, A.; Tonejc, A.M.; Gamulin, O. J. Mole. Struct. 1999, ,

164 CHAPTER 7 Visible-light Sensitive Carbon-, Nitrogen- and Sulfur- Doped TiO 2 Nanocrystals Directly Derived from Micrometer-sized Inorganic Precursors Abstract: C-, N- and S-doped TiO 2 nanocrystals were further developed directly from micrometer-sized inorganic precursors using the versatile method described in the previous chapter and under optimal conditions. Their electronic origins for the visiblelight absorption were elucidated for the first time with X-ray photoelectron spectroscopy, which is consistent with the previous theoretical and experimental studies and explains the photocatalytic activity in the visible-light regime of doped TiO

165 7.1 Introduction Since the photoelectrochemical splitting of water on TiO 2 electrodes was discovered, semiconductor-based photocatalysis has attracted extensive attention, especially in the application of photodegradation of organics in polluted water and air under ultraviolet light (wavelength < 400 nm) illumination. 1 The visible-light applications of TiO 2 have recently been pursued by doping this material with metal 1 and nonmetal elements, e.g. nitrogen, 2,3 carbon 4 and sulfur. 5 Various methods including solgel, hydrolysis, 3 and pyrolysis, have been developed to synthesize these titania-based visible-light sensitive nanostructures. Most of this work was done on bulk TiO 2 powders or films. 3-5 Although much progress has occurred, TiO 2 nanocrystals doped with carbon and sulfur dopants still remain a rather rare specialty, so does the direct experimental evidence on the electronic origin for the visible-light absorption properties of doped TiO 2. Room temperature sol-gel derived TiO 2 is usually amorphous 3 or incompletely crystallized. 6 Following heat or hydrothermal treatment leads to crystallization at higher temperatures. 7 For the bulk it is known that titanium carbide, 8 nitride 9 and sulfide 10 can be oxidized in air or oxygen atmosphere into titanium dioxide. Here, a simple but versatile method to prepare high-quality C-, N- and S-doped TiO 2 nanoparticles from bulk inorganic precursors under ambient conditions is presented. These doped TiO 2 nanocrystals showed visible-light absorption and high photocatalytic activity. The electronic origin for the visible-light response of the doped TiO 2 was revealed with valence band (VB) XPS spectroscopy, directly confirming the theoretical prediction of the modification of the electronic band structure by C, N and S dopants

166 7.2 Experimental Preparation of C-, N- and S-doped TiO 2 nanocrystals: The experiments were conducted as follows. About 0.5 g TiX powder was loaded in a ceramic sample boat (4.0/l 0.5/h 0.5/w in inch), and then placed in the middle of a quartz tube in a Lindberg tube furnace. The two ends of this tube were left open to the atmosphere. These powders were oxidized at 350ºC-650ºC for 96 hrs. The temperature was slowly ramped up and cooled down at a rate of 2ºC/min during the heating and cooling process. After cooling to room temperature, they were taken out for further investigations. Characterizaton of C-, N- and S-doped TiO 2 nanocrystals: The X-ray diffraction patterns were obtained using a Philips PW 3710 X-ray powder diffractometer. The reflectance spectra were obtained on a Cary 50 UV-visible spectrometer equipped with a reflectance unit. The low resolution TEM images were taken on a JEOL 1200EX transmission electron microscope operated at 80 kv. High-resolution transmission electron microscopy (HRTEM) images were obtained on a Tecnai F30 HRTEM machine operated under 300 kv. Samples for TEM were prepared by depositing a drop of a nanocrystal solution in water onto a copper grid supporting a thin film of amorphous carbon, and drying in air. The chemical compositions were determined with X-ray photoelectron spectroscopy (XPS), taken on a Perkin-Elmer PHI 5600 XPS System with the samples on a carbon tape sticking to the aluminum support after the samples were sputtered with Ar + flux for 5 min. The Ar + sputtering rate was 57Å/min. The XPS binding energies were calibrated with respect to the C 1s peak from the carbon tape at ev. The photocatalytic activities of these C-, N- and S-doped TiO 2 photocatalysts with comparison to commercial pure TiO 2 Degussa P25 were evaluated by measuring the 133

167 decomposition of methylene blue under illumination from a 150-W xenon arc lamp system. Evaluation of the photocatalytic activity of the nanocrystals was conducted by measuring the changes in absorption of a methylene blue solution at 650 nm during the dye s photocatalytic decomposition. The light was focused onto a cuvette filled with a 2 ml aqueous solution of methylene blue (optical density = 1.0) and 10.0 mg of the nanocrystal sample. The absorption spectra of methylene blue were then measured on a Varian Cary Bio50 UV-visible spectrometer under aerobic conditions and were always corrected for methylene blue decomposition in the absence of any titania nanocrystals. 7.3 Results and Discussion Figure 7.1A shows the X-ray diffraction (XRD) patterns of commercial TiC, TiN and TiS 2 powders and the resultant C-, N-, S-doped TiO 2 after heating at 350, 650, and 450 o C for 96hr, respectively. After heating, the micrometer-sized cubic structure TiC and layered TiS 2 converted into the C- and S-doped TiO 2 nanoparticles with anatase phase XRD Intensity / a.u. (f ) (e) (d) (c) (b) o o o o o ^ ^ ^ ^ ^ ^ # # # # # # # + + o o o o o * * (a) A θ / O Absorbance / a.u. B (f) (b) (d) (a) (e) (c) Wavelength / nm Figure 7.1 (A) XRD patterns of the (a, *) TiC, (c, +) TiN and (e, ^) TiS 2 and the resultant (b) C-, (d) N- and (f) S- doped TiO 2 samples after heating at 350ºC, 650ºC and 450ºC for 96 hours, respectively; (o) anatase, and rutile (#) TiO 2 structures. (B) The corresponding reflectance spectra. 134

168 and an average size of 25 nm in diameter from TEM images and estimated from the Scherrer s equation. 11 After heating, the initially black or gray TiC, TiN and TiS 2 powders changed into yellow. As shown in the reflectance spectra (Figure 7.1B), TiC, TiN and TiS 2 absorbed light from UV into infrared regime; the resultant doped TiO 2 samples showed band-edge absorptions around 3.2 ev (390 nm) for C- and S-doped TiO 2 (anatase phase) and 3.0 ev (420 nm) for N-doped TiO 2 (rutile phase) and additional longwavelength absorptions up to 800 nm, which is attributed to the optical transitions caused by the additional bandgap states induced by the C-, S-, and N-dopants in the TiO 2 host lattice. E nm F nm 200 nm A B 50 nm G C D nm 50 nm 50 nm 5 nm Figure 7.2 TEM and HRTEM images of C-doped TiO 2 nanoparticles. TiC powder (A) and the resultant C-doped TiO 2 nanoparticles after heating before (B) and after size selective centrifugation (C) and (D). HRTEM images of three C-doped TiO 2 nanocrystals (E), (F), and (G). The insets in (A) and (B), respectively, show the selected area electron diffraction patterns of cubic TiC and the C-doped TiO 2 nanoparticles with anatase phase. Figure 7.2 shows the TEM images of the commercial TiC powder and the C- doped TiO 2 nanoparticles. The size of the TiC powder was above 500 nm (Figure 7.2A), 135

169 while after heating, well-dispersed nanometer sized C-doped TiO 2 particles were obtained as a mix of different sizes (Figure 7.2B). The size distribution could be narrowed down by a size selection centrifugation as shown in Figure 7.2C and Figure 7.2D. The selected area electron diffraction pattern shown in the insets of Figure 7.2A and Figure 7.2B confirmed the structural change from cubic TiC into anatase TiO 2 structure. High-resolution transmission electron microscopy (HRTEM) images (Figure 7.2E - Figure 7.2G) showed that the doped TiO 2 nanocrystals had nearly spherical shapes and were highly crystallized in single (Figure 7.2E and 2F) or multiple (Figure 7.2G) domains with well-resolved lattice structures for TiO 2. The observed spacing between the lattice planes of the TiO 2 nanocrystals was obtained as nm for (002) and nm for (100) planes. The presented XPS spectra show that the overall C, N and S dopant concentration is 0.5%, 0.2% and 4.2% in atomic percent in the samples of TiO 2-x C x, TiO 2-y N y and TiO 2- zs z, respectively. The XPS binding energies were calibrated with respect to the C 1s peak from the carbon tape at ev (Figure 7.3A). The signals at ev (Figure 7.3A), ev and ev (Figure 7.3B), and ev (Figure 7.3C) verified the existence of Ti-C, 12 Ti-N 2,3,13 and S-Ti 5,14 in TiO 2-x C x, TiO 2-y N y, and TiO 2-z S z, respectively. The signal at ev in TiO 2-z S z could also be trapped SO 2 in the lattice. 5,14 The shoulders in the Ti 2p3/2 XPS spectra in Fig. 3D at binding energies around 455 ev also confirmed the existence of C-Ti, 8 N-Ti, 9 and S-Ti 10 bonds in TiO 2-x C x, TiO 2-y N y and TiO 2-z S z, respectively. 136

170 Intensity / a.u. A) C1s XPS in the C-doped TiO2 TiO 2-x C x Ti-C Intensity / a.u. B) N1s XPS in N-doped TiO2 TiO 2-y N y Ti-N TiO 2 TiO Binding Energy / ev Binding Energy / ev Intensity / a.u. C) S2p XPS in S-doped TiO2 TiO 2-z S z Ti-S Intensity / a.u. D) TiO 2-z S z TiO 2-y N y TiO 2-x C x TiO 2 Ti2p XPS Ti-O Ti-S Ti-C Ti-N TiO Binding Energy / ev Binding Energy / ev Intensity / a.u. E) VB XPS TiO 2-z S z TiO 2-y N y TiO 2-x C x Intensity / a.u. S 3p N 2p C 2p F) VB XPS TiO 2-z S z TiO 2-y N y TiO 2-x C x TiO 2 TiO Binding Energy / ev Binding Energy / ev Figure 7.3 (A) C 1s, (B) N 1s, (C) S 2p, (D) Ti 2p and (E) VB XPS spectra of TiO 2-x C x, TiO 2- yn y, TiO 2-z S z and pure TiO 2. VB XPS displayed additional electronic states above the valence bandedge caused by C 2p, N 2p and S 3p orbitals as enlarged in (F). Direct evidence on the modification of the TiO 2 valence band caused by the C-, N-, and S- dopants was found from the valence band (VB) XPS, which can probe directly the total density of the state distribution in the valence band. 15 Additional electronic states below 2.0 ev were observed above the valence bandedge of pure TiO 2 (Figure 7.3E and Figure 7.3F). These states can be attributed to the C 2p, N 2p and S 3p orbitals in TiO 2-137

171 xc x, TiO 2-y N y. This finding directly supported the theoretical predictions 2,14 and previously mentioned indirect experimental evidence, 11,16 that the C-, N- and S-dopants provide additional electronic states above the valence band edge of pure TiO 2. These additional states are responsible for the visible-light absorption of these doped TiO 2. Absorbance / a.u A Degussa P25 TiO 2-x C x TiO 2-y N y TiO 2-z S z Δ Abs. (after 500 min) B Integ. (Abs.) Time / min 0.0 C-TiO2 N-TiO2 S-TiO2 Sample Name 0 Figure 7.4 (A) Photocatalytic decomposition curves of the methylene blue under Xe lamp illumination with TiO 2-x C x, TiO 2-y N y and TiO 2-z S z as photocatalysts compared to Degussa P25. (B) The correlation between the amount of photodecomposition of methylene blue after 500 min Xe lamp illumination and the integral of the absorption curve of these doped TiO 2 from 800 nm to 400 nm. Among the four presented nanomaterials, TiO 2-y N y displayed the highest photocatalytic activity, followed by TiO 2-z S z, TiO 2-x C x and pure TiO 2 Degussa P25 as shown in Figure 7.4A. This can be explained by the fact that TiO 2-y N y has the highest visible-light absorption, followed by TiO 2-z S z and TiO 2-x C x. The photocatalytic activity of these doped TiO 2 samples had a strong correlation with the integral of the absorption of these doped TiO 2 in the visible regime as displayed in Figure 7.4B. This suggests that the visible-light photocatalytic performance is mainly correlated to the visible-light absorption of these materials, consistent with previous reports. 2-5,11-14 This further 138

172 suggests that if their visible-light absorption can be increased, the performance could be further enhanced. At this point, the doping chemistry for TiO 2 with nonmetals is at the developing stage, and any improvement in optical response offers great potential for improved photoinduced applications. 7.4 Conclusions C-, N- and S-doped TiO 2 nanocrystals can be developed directly from micrometer-sized inorganic precursors using the top-down method under optimal conditions. Their electronic origins for the visible-light absorption were elucidated for the first time with X-ray photoelectron spectroscopy, which is consistent with the previous theoretical and experimental studies and explains the photocatalytic activity in the visible-light regime of the doped TiO References 1. a) Fujishima, A.; Honda, K. Nature 1972, 238, 37. b) Linsebigler, A. L.; Lu, G.; Yates, Jr. J. T. Chem. Rev. 1995, 95, 735. c) Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W. Chem. Rev. 1995, 95, 69. d) Millis, A.; Hunte, S. L. J. Photochem. Photobio. A 108, 1997, Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293, a) Burda, C.; Lou, Y.; Chen, X.; Samia, A. C. S.; Stout, J.; Gole, J. L. Nano Lett. 2003, 3, b) Chen, X.; Burda, C. J. Phys. Chem. B 2004, 108, c) Gole, J. L.; Stout, J. D.; Burda, C.; Lou, Y.; Chen, X. J. Phys. Chem. B 2004, 108, d) Chen, X.; Lou, Y.; Samia, A. C. S.; Burda, C.; Gole, J. L. Adv. Funct. Mater. 2005, 139

173 15, 41. e) Prokes, S. M.; Gole, J. L.; Chen, X.; Burda, C.; Carlos, W. E. Adv. Funct. Mater. 2005, 15, Khan, S. U. M.; Al-Shahry, M.; Jr. Ingler, W. B. Science 2002, 297, Umebayashi, T.; Yamaki, T.; Yamamoto, S.; Miyashita, A.; Tanaka, S.; Sumita, T.; Asai, K. J. Appl. Phys. 2003, 93, a) Park, H. K.; Moon, Y. T.; Kim, D. K.; Kim, C. H. J. Am. Ceram. Soc. 1996, 79, b) Hague, D. C.; Mayo, M. J. J. Am. Ceram. Soc. 1994, 77, c) Slunecko, J.; Kosec, M.; Holc, J.; Drazic, G. J. Am. Ceram. Soc. 1998, 81, Bacsca, R. R.; Grätzel, M. J. Am. Ceram. Soc. 1996, 79, a) Li, Y.-L.; Ishigaki, T. Chem. Mater. 2001, 13, b) Choi, Y.; Umebayashi, T.; Yamamoto, S.; Tanaka, S. J. Mater. Sci. Lett. 2003, 22, Saha, N. C.; Tompkins, H. G. J. Appl. Phys. 1992, 72, Martinez, H.; Auriel, C.; Gonbeau, D.; Loudet, M.; Pfister, G. G. Appl. Surf. Sci. 1996, 93, Lindegren, T.; Mwabora, J. M.; Avendaňo, J. J.; Hoel, A.; Granqvist, C. -G.; Lindquist, S.-E. J. Phys. Chem. B 2003, 107, a) Irie, H.; Watanabe, Y.; Hashimoto, K. Chem. Lett. 2003, 32, 772. b) Choi, Y.; Umebayashi, T.; Yoshikawa, M. J. Mater. Sci. 2004, 39, a) Diwald, O.; Thompson, T. L.; Goralski, E. G.; Walck, S. D.; Yates, Jr. J. T. J. Phys. Chem. B 2004, 108, b) Diwald, O.; Thompson, T. L.; Zubkov, T.; Goralski, E. G.; Walck, S. D.; Yates, Jr. J. T. J. Phys. Chem. B 2004, 108, 52. c) Irie, H.; Watanabe, Y.; Hashimoto, K. J. Phys. Chem. B 2003, 107, Umebayashi, T.; Yamaki, T.; Itoh, H.; Asai, K. Appl. Phys. Lett. 2003, 81,

174 15. a) Finkelstein, L. D.; Kurmaev, E. Z.; Korotin, M. A.; Moewes, A.; Schneider, B.; Butorin, S. M.; Guo, J. H.; Nordgren, J.; Hartmann, D.; Neumann, M.; Ederer, D. L. Phys. Rev. B 1999, 60, b) Woicik, J. C.; Nelson, E. J.; Kronik, L.; Jain, M.; Chelikowsky, J. M.; Heskett, D.; Berman, L. E.; Herman, G. S. Phys. Rev. Lett. 2002, 89, / Nakamura, R.; Tanaka, T.; Nakato, Y. J. Phys. Chem. B 2004, 108,

175 CHAPTER 8 The Electronic Structure of Visible-light Responsive N-, C- and S-Doped Titania Obtained by X-ray Spectroscopy * Abstract Visible-light sensitive TiO 2 -based materials have become important due to the numerous industrial applications associated with solar energy. We report here the study of visible-light responsive N-, C- and S-doped TiO 2 materials using x-ray absorption, emission, and photoemission spectroscopy. We found that the N, C and S dopants had negligible influence on the O 2p and Ti 3d orbitals in the doped TiO 2 compared to pure TiO 2, and had a small effect on the crystal structure of the original TiO 2. For the first time, direct experimental evidence was obtained for the modification of the electronic band structure of TiO 2 caused by the N, C and S dopants in these doped materials. The complete electronic band structure of pure TiO 2 was constructed by combining the results from x-ray absorption & emission with x-ray photoelectron spectroscopy. The comparison of the constructed band structure of pure TiO 2 supported this assignment and thus explained the origin of the visible-light absorption properties of these doped TiO 2. * This work has been carried out in collaboration with Dr. Jing-Hua Guo at the Advanced Light Source, Lawrence Berkeley National Laboratory. 142

176 8.1 Introduction Titanium dioxide (TiO 2 ) has been the subject of extensive studies since the late 1960s, due to its applications in various areas such as pigments, photoconductors, dielectric materials, and photocatalysts. 1-9 The electronic structure of this material is well understood, both theoretically and experimentally, 7,10-20 and is characterized by a band gap located in between the oxygen 2p valence band and the titanium 3d conduction band. 10 X-ray emission spectroscopy (XES) and x-ray absorption spectroscopy (XAS) are powerful element-oriented probes for the partially occupied and unoccupied electron densities of states in materials, respectively. 7,10-29 In XAS, a core-level electron absorbs an x-ray photon and is excited to an unoccupied state above the Fermi level, E F. 7,10,12,20,22 The transition is controlled by the dipole selection rule, thus only transitions with Δl = ± 1 are allowed. 22 In addition, due to the higher probability of intra-atomic transitions, XAS is site selective with on-site transitions dominating the spectrum. 17,22 In XES, a valence band electron relaxes via the emission of an x-ray photon, filling a core hole created by the x-ray absorption process. 12,15,18,24 The XES spectrum reflects the element specific valence band density of states (DOS), resolved into its orbital angular momentum components, i.e the partial density of states (PDOS). 30,31 In contrast, X-ray photoelectron spectroscopy (XPS) measures the total DOS of the valence band, as well as the binding energy of the core-level electrons. 20,21,25,32-36 The photocatalytic ability of TiO 2 in many reactions has been demonstrated and widely studied. 1-6 However, the wide application of TiO 2 as a photocatalyst is limited by its UV or near-uv absorption properties. 8-9 Doping TiO 2 with different elements such as 143

177 nitrogen, carbon, and sulfur has shown promising results for visible-light photocatalysis, where visible-light absorption has been achieved. 8,9,37-41 It is known that titanium nitride (TiN), carbide (TiC), and sulfide (TiS 2 ) can be oxidized to become TiO 2 by hightemperature sintering in air or oxygen environments TiN and TiC are metallic conductors with a partially filled band and a chemical bond of simultaneously metallic, covalent, and ionic characters, 12,21-26 while TiS 2 is regarded as a semiconductor or semimetal, 33,42 with a bandgap of about 0.9 ev. 28 And thus the oxidation of these materials provides an attractive technique to prepare the doped TiO 2. We report here a comprehensive spectroscopic study of the electronic structures of N-, C- and S-doped TiO 2 prepared from the oxidation of TiN, TiC, and TiS 2 at high temperatures. The valence band, conduction band, and core level states have been measured using a combination of XAS, XES and XPS. Direct experimental evidence was observed for the first time on the modification of the electronic band structures of TiO 2 brought about by the N, C and S dopants in these doped materials. A comprehensive picture of the band structure of pure TiO 2 is derived from these measurements, which are in good agreement with other experimental and theoretical results, 8 leading to additional understanding of the origin of the long-tail visible-light absorption of these doped TiO Experimental The titanium nitride (TiN), carbide (TiC) and sulfide (TiS 2 ) were purchased from Strem Chemicals. The resultant doped titanium dioxide TiO 2-x E x (E = N, C, S) samples were prepared by oxidation of the above reagents at 1000 o C for 6 hours in a quartz tube under the atmosphere inside a Lindberg tube furnace with a digital temperature control 144

178 unit. These samples were then allowed to cool, and doped titania were obtained. All measurements were made at room temperature. X-ray diffraction (XRD) patterns were obtained using a Philips PW 3710 X-ray powder diffractometer. The UV-visible reflectance spectra of the nanocrystal samples were measured on a Cary 50 UV-visible spectrometer with a reflectance unit. The X-ray spectroscopic experiments were measured at the undulator beamline 7.0 of the Advanced Light Source (ALS), Lawrence Berkeley National Laboratory with a spherical grating monochromator. 43 Ti L 2,3 XES and O K α XES spectra of these samples were recorded by using the Nordgren-type grating spectrometer. 44 The spectrometer was mounted perpendicular to the incoming photon beam in the polarization plane and the resolution was 0.3 ev and 0.4 ev, respectively, for Ti L 2,3 XES and O K α XES spectra. The monochromator was set to the excitation energies of 475 ev and 565 ev respectively, for Ti L 2,3 and O K α spectra with a resolution of 0.5 ev. The samples were mounted to have a beam incidence angle of 30 to the sample surface. For energy calibration of the Ti L 2,3 XES and O K α XES, the spectra of the reference samples Ti, TiO 2, and Zn were measured. The base pressure of the chamber was 2 x 10-9 mbar. The absorption spectra at the Ti 2p and O 1s edges were measured by means of total electron yield (TEY) and with a monochromator resolution set to 0.2 ev. The absorption intensity was normalized by the current from a clean gold mesh placed in the incoming beam to compensate fluctuations of the incoming photon intensity. The x-ray fluorescence and absorption spectra were brought to a common energy scale using an elastic peak in the fluorescence spectra recorded at the excitation energy set at the absorption edge. 145

179 For the x-ray photoelectron spectroscopy (XPS) measurements, a Perkin-Elmer PHI 5600 XPS System was used. The energy resolution of the spectrometer is ev for the XPS measurements. Samples for XPS measurement were coated on carbon tape attached to the sample holder. The pressure in the vacuum chamber during the measurements was below 3 x 10-8 mbar. Global and partial XPS spectra were taken before and after Ar + sputtering for 10 min at a sputter rate of 57Å/min. 8.3 Results and Discussion Structural, Chemical and Optical Properties TiN and TiC have NaCl-like cubic structure, while TiS 2 has a CdI 2 -like layered structure. TiN and TiC crystallize in the rock-salt structure, with N or C atoms occupying interstitial positions in a close packed arrangement of Ti atoms. 26 This structure gives rise to strong metal to metal and metal to non-metal interactions. 26 TiS 2 has a layered structure consisting of sulfur-titanium-sulfur slabs The titanium ions are in a regular octahedral coordination to the sulfur ions and each slab is formed by a two-dimensional hexagonal titanium sublattice sandwiched by two closely adjacent sulfur hexagonal sublattices. 29 The sulfur-titanium bonds are quite strong within a single slab, while there is only a weak coupling between individual slabs. 29 The change in crystal structure of the starting materials TiN, TiC and TiS 2 into TiO 2 upon heat treatment is shown in Figure 8.1. After treatment at 1000 o C in air, these materials transformed into rutile structures. 146

180 (f) Intensity / a.u. (e) (d) (c) (b) Theta / o Figure 8.1 XRD patterns for the titanium compounds: (a) commercial TiN; (b) TiN after heating; (c) commercial TiC; (d) TiC after heating; (e) commercial TiS 2 ; (f) TiS 2 after heating. The resultant samples from TiN, TiC and TiS 2 display the typical diffraction pattern of rutile TiO 2. (a) O 1s Ti 2p Ti 3s Intensity / a.u. O KLL TiO 2-z S z TiO 2-y C y C 1s Ti 3p Ti 3s TiO 2-x N x Binding Energy / ev Figure 8.2 Before Ar sputtering, global XPS of the N-, C- and S-doped TiO 2 powder: TiO 2-x N x, TiO 2-y C y and TiO 2-z S z. The carbon 1s signals at ev in the spectra were from the carbon tape. Chemical transformation often occurred at the interface between the surface of the material and the active atmosphere. Thus, the surface composition change can be a frontsign of the bulk chemical composition change. Figure 8.2 shows the global XPS spectra 147

181 of the N-, C- and S-doped TiO 2 : TiO 2-x N x, TiO 2-y C y and TiO 2-z S z before Ar + sputtering. The XPS binding energies from the samples were calibrated with respect to the C 1s peak from the carbon tape at ev. For all doped TiO 2 samples, typical binding energies for Ti 2p and O1s were detected at ev (for Ti 2p 3/2 ) and (for Ti 2p 1/2 ), ev, respectively. No N, C or S signals were detected on the surface. Thus, the surfaces of these materials were completely transformed into pure TiO 2. Even before heating, some oxygen signal was detected for TiN, TiC and TiS 2 precursors, since the surface of these compounds can be partially oxidized in air With the depth profiling method, the bulk composition can also be obtained. Depth profilings were preceded for 10 min with the sputtering rate of 57Å/min. Small amount of the C, N and S signal were detected in the partial XPS spectra shown in Figure 8.3. In Figure 8.3A, the signals at ev and ev in the TiO 2-x N x can be attributed to N1s electrons in Ti-N bonds within different bonding environments 8,9, In Figure 8.3B, the signal at a binding energy ev can be attributed to C1s electrons from the carbon tape, and the signal at ev can be attributed to the Ti-C bond in the TiO 2-y C 41,51 y. In Figure 8.3C of the TiO 2-z S z, the signal at and can be attributed to the S2p electrons in the S-Ti bond and the formed SO 2 species trapped in the lattice during the oxidation 52,53. The peaks in the Ti2p 3/2 XPS spectra in Figure 8.3D at binding energies of , , and ev 55 verified existence of the O- Ti, N-Ti, C-Ti, and S- Ti bonds in the TiO 2-x N x, TiO 2-y C y and TiO 2-z S z samples, respectively. Overall, the N, C and S dopant concentration in these doped TiO 2 samples is about 0.2%, 0.5% and 4.2% in atomic ratios. 148

182 Intensity / a.u. A N1s XPS (ii) (i) Ti-N Intensity / a.u. B C 1s XPS (ii) Carbon tape Ti-C (i) Binding Energy / ev Binding Energy / ev Intensity / a.u. C (iv) (i) S 2p XPS Binding Energy / ev Ti-S Intensity / a.u. D Ti2p XPS Binding Energy / ev (iv) (iii) (ii) (i) Figure 8.3 After Ar + sputtering for 10 min, partial XPS spectra of the N-, C- and S-doped TiO 2 compared to that of pure rutile TiO 2. (A) N 1s binding energy region for pure TiO 2 and TiO 2-x N x, (B) C 1s binding energy region for pure TiO 2 and TiO 2-y C y, (C) S 2p binding energy region for pure TiO 2 and TiO 2-z S z ; (D) Ti 2p binding energy region for pure TiO 2 and TiO 2-x N x, TiO 2-y C y and TiO 2-z S z ; (i) pure TiO 2, (ii) TiO 2-x N x, (iii) TiO 2-y C y, (iv) TiO 2-z S z. The chemical oxidation and structural transformation of the TiN, TiC and TiS 2 directly affect the optical properties of these materials. Figure 8.4 shows the optical reflectance of TiN, TiC, TiS 2 and the TiO 2-x N x, TiO 2-y C y and TiO 2-z S z. TiN, TiC and TiS 2 have gray, black and black-gray color. After heating, these samples changed into yellow or slight yellow color (for TiS 2 ). As noted earlier, TiN and TiC are good metallic conductors with a partially filled band at E 12,24,25,34,35 F, while TiS 2 is regarded as a 149

183 Wavelength / nm Absorbance / a.u. (f) (d) (b) (e) (c) (a) Optical Energy / ev Figure 8.4 Absorption spectra for the titanium compounds: (a) commercial TiN; (b) TiO 2-x N x ; (c) commercial TiC; (d) TiO 2-y C y ; (e) commercial TiS 2 ; (f) TiO 2-z S z. The N-, C- and S-doped TiO 2 display a typical bandgap of around 3.1 ev and additional lower energy tail absorption. semiconductor or semimetal, 33,42 with a bandgap of about 0.9 ev. 28 In the optical spectrum, the original starting materials all show absorption from IR into UV regime. After heating they oxidized into N-, C- and S-doped TiO 2, and displayed a bandgap around 3.0 ev due to the resulting rutile TiO 2 structures and low energy absorption from the near-ir to the visible range which can be attributed to the modification of the original TiO 2 electronic states caused by the N, C and S dopants in TiO 2-x N x, TiO 2-y C y and TiO 2- zs z as discussed below X-Ray Absorption of TiO 2 and TiO 2-x N x The Ti 3d, 4s, and 4p atomic orbitals and the nitrogen, carbon, oxygen 2s and 2p or sulfur 3s and 3p atomic orbitals are involved in the bonding for TiN, TiC, TiS 2 and 150

184 TiO 2, 12,24,28 and thus for TiO 2-x N x, TiO 2-y C y and TiO 2-z S z too. The valence and conduction bands of the 3d metals consist primarily of mixed 3d, 4s and 4p states with the s states mainly responsible for electrical conductivity and cohesion and the d states responsible for the majority of the fundamental transition-metal characteristics. 12 To obtain information about the distribution of s and d states, one can study x-ray emission or absorption spectra arising from transitions to an inner level of p symmetry, 12,24,28 such as the transition 3d4s 2p 3/2 (the L 3 band or L α line), and the transition 3d4s 2p 1/2 (L 2 band or L β line). 12,28 The intensity distribution of a valence emission band and the variation in the absorption coefficient μ are directly related to the density of states and the transition probability given by the relation I(E), μ(e) P(E) N(E) (1) where P(E) is the transition probability and N(E) the density of states. 12,28 The energy differences between the signals of the emission and absorption spectra are equal to the corresponding distances on the density-of-state curve. 12,28 For the 3d transition metals, the L 2,3 absorption spectra give information regarding the unfilled band (conduction band), while the L 2,3 emission band contains information about the filled portion of the 3d4s band (valence band). 12,28 For the nonmetal O, the K absorption or emission band of the oxygen in these TiO 2-x N x, TiO 2-y C y and TiO 2-z S z is assigned the contribution of the TiO 2-x lattice. Figure 8.5a shows x-ray absorption spectra from TiO 2 and TiO 2-x N x samples, obtained by the total electron yield (TEY) method. The spectrum of TiO 2-x N x is very similar to that of pure rutile TiO 2. This suggested that the influence of the N dopant on the TiO 2 structure and crystal field in the TiO 2-x N x is very small. In 3d transition metal 151

185 Absorbance / a.u. t 2g L 3 L 2 a) e g t 2g e g C Ti 2p XAS TiO 2-x N x TiO 2 D E Absorbance / a.u. t 2g e g C D E b) F O 1s XAS TiO 2-x N x TiO 2 G Phonton Energy / ev Photon Energy / ev Figure 8.5 XAS spectra of Ti 2p (a) and O 1s (b) in pure TiO 2 and TiO 2-x N x samples. oxides, the 2p x-ray absorption spectrum is completely dominated by the strong 2p 3d dipole transitions. 14 To a first approximation, the spectrum is related to the density of Ti 3d unoccupied states, distorted by the influence of the Ti 2p core-hole potential. 23 The features of the Ti L 2,3 -edge spectrum reflect the local coordination number and symmetry of a titanium ion in the material and can be explained by the ligand-multiplet theory, due to the large Coulombic and exchange interactions of the 2p-3d and 3d -3d orbitals, and thus is directly related to the local symmetry and electron configuration of both the ground state and the final state. 7,11,15,16 The crystal field also splits the final state multiplet, and the spectrum is dependent upon the crystal field effect of the final state. 45,46 For both TiO 2 and TiO 2-x N x, the 2p spin-orbit coupling splits the initial state into 2p 3/2 and 2p 1/2, resulting in two L-edge features, denoted L 3, and L 2, respectively. 7,11,14,16, Both the L 3 and L 2 features further split into L 3 -t 2g, L 3 -e g, L 2 -t 2g, and L 2 -e g features because of the low symmetry of the ligand field O h. 7,11,14,15,16, The L 2 -e g feature splits further into a double peak band centered at 461 ev due to the slight distortion of the TiO 6 8- or TiO 6-x N x 6-152

186 octahedron that results from the configurational deformation. 7 The energy splitting for the centers of t 2g -e g of L 3 is 3.0 ev in TiO 2-x N x and TiO 2. The first two small peaks are related to the splitting of a single allowed j-j coupling transition by the crystal field. 11,57 There are three final states in increasing energy: 2p 3/2 3d 3/2, 2p 3/2 3d 5/2 and 2p 1/2 3d 3/2. The 2p 3/2 3d 3/2 state gives rise to the two small peaks in the spectrum through the crystal field splitting. 57 These two small peaks are also assigned to the subbands of the L 3 -t 2g band by the crystal field splitting. The third and fourth peaks are related to t 2g and e g symmetry for the L 3 edge. 7,11,14,16 In the photon energy range ev, the peak is attributed to e g. 7,11,14,16 The e g peak at lower energy side originates from the long Ti-O bonds due to a hybridization effect weaker than the short Ti-O bonds. 15 The same consideration is used for the fifth and sixth peaks for the L 2 edge. 11 Details of the L 2 edge are blurred, because the intrinsic broadening is considerably larger due to an extra Auger decay channel. 11 Above 469 ev, there are weak and broad satellite peaks for both TiO 2-x N x and TiO 2. O 2p and Ti 2p states give the main contribution to the peak C in the satellite, without the titanium 3d and 4s states. 14 Peak D is from O 2p and titanium 4p states with comparable contribution from Ti 3d states. 14 Peak E is attributed to the titanium 4s and 4p states with the O 2p contribution. 14 Interestingly, the energy splitting for the centers of t 2g -e g of L 2 is larger in TiO 2-x N x (3.2 ev) than in TiO 2 (3.0 ev). This can be attributed to a possible crystal field change due to the chemical environment changes after some of the surrounding O atoms are replaced by N atoms. The O 1s absorption spectra of the TiO 2 and TiO 2-x N x are almost identical as presented in Figure 8.5b. The O 1s x-ray absorption spectrum corresponds to transition from the oxygen 1s core level into empty or partially filled O 2p states. 7 The final state of 153

187 oxygen 1s x-ray absorption contains a core hole in the oxygen 1s level. At the absorption edge, the excited electron is then located in the Ti 3d band, and from dipole selection rules must have oxygen p character. 10 Thus, the oxygen 1s x-ray-absorption spectrum gives a direct picture of the oxygen p-projected density of states. 10,22 The intensity and the spread of the absorption signal illustrate the importance of covalent contribution to the bonding in the material. 22,47 The low-energy range of the spectra ( ev) is dominated by two strong broad bands with an energy splitting close to the energy difference of peaks, i. e. to the splitting of t 2g and e g states in the Ti 2p XAS. 10,14,16 This region is attributed to the oxygen 2p contribution to states with predominantly titanium character (the titanium 3d band), or the O 2p states which are hybridized with the empty Ti 3d bands. 10,11,14,16 Crystal field effects split the spectrum here. 11 The prominent doublets can be assigned respectively to transitions into the t 2g and e g bands of the titanium, and are sensitive to the short-range order, i.e the nearest neighbor order, and are more localized. 7,10,14,16,22,23 For both TiO 2-x N x and pure TiO 2, the t 2g peak is centered at ev and the e g peak at ev, thus with the energy splitting of 2.7 ev. The crystal field splitting (10 Dq) is very sensitive to the coordination number, the distribution of the ligand and to the strength of the hybridization. 22,23 This suggested that there is no apparent difference in the crystal filed in TiO 2-x N x and pure TiO 2. The highenergy part of the spectrum is formed by the delocalized states derived from the antibonding O 2p and Ti 4sp band with principally oxygen 2p character. 11,16 The features above 536 ev are due to the covalent mixing of O 2p and Ti 4sp orbitals and are sensitive to the long-range order. 7,10,14,16,22,23 The origin for the bands C-E is the same as analyzed above in the Ti XAS spectra. Peaks F and G are attributed to O 2p with contributions 154

188 from the Ti 4p. 14 The states in the 10 ev to 20 ev region above the threshold of the absorption are dominated by oxygen p character and they are the antibonding combinations of direct oxygen-oxygen interactions. 10 The identical features of the O K absorption spectra of TiO 2-x N x and TiO 2 suggest that the N dopant did not affect the O 2p oribtals X-Ray Emission of TiO 2-x N x X-ray emission bands result from removal of an inner-level electron followed by the transition of a valence-band electron into the inner hole. 12,15-17,24 Such transitions give rise to the K band (valence band 1s), the L 2,3 bands (valence band 2p), and the M 2,3 bands (valence band 3p). 12,24 The K band results from probing the part of the valence band having p symmetry while the L band results form the distribution of s and d symmetry. 12,24,28 The band structures of TiN, TiC, TiS 2, TiO 2 and thus of TiO 2-x N x, TiO 2- yc y and TiO 2-z S z, are composed of a strong admixture of p, d, and s symmetry, so that the information present in the O K and Ti L bands can be combined to explore the contribution of TiO 2-x to the valence band structure in the TiO 2-x N x, TiO 2-y C y and TiO 2- zs z. Figure 8.6 shows the XES spectra of Ti L and O K α in the TiO 2 and the TiO 2-x N x samples. The valence band of TiO 2 and TiO 2-x N x is mainly composed of the contribution from the major oxygen or nitrogen 2p states with a minor contribution from the titanium 3d states. 10,12 The titanium L emission spectrum of TiO 2-x N x and pure TiO 2 were very similar in that there was a major peak at ev due to transition from the oxygen 2p band to the Ti 2p 3/2 band. For TiO 2-x N x, there was a small shoulder at the higher energy 155

189 side (454.6 ev), which can be attributed to the overlap with the transition from the nitrogen 2p band to the titanium 2p 1/2 band. 12,15 Intensity / a.u. (a) Ti L XES TiO 2-x N x TiO 2 Intensity / a.u. (b) O K α XES TiO 2-x N x TiO Emission Energy / ev Emission Energy / ev Figure 8.6 XES spectra of Ti L (a) and O K α (b) in the TiO 2 and the TiO 2-x N x sample. The O K α x-ray emission spectra of both TiO 2 and the TiO 2-x N x are almost identical, with one major peak at ev and a smaller shoulder at ev. The O K emission can be attributed to transition from the filled oxygen 2p states to the oxygen 1s hole state. 12 The same splitting energies between the main peak and the shoulder for TiO 2 and TiO 2-x N x suggest that the change on the crystal field caused by the N dopant was negligible, since the surrounding environment is composed of the Ti atoms at the same lattice in TiO 2-x N x as in pure TiO 2. The very similar Ti absorption and emission spectra with almost identical features for TiO 2-x N x and pure TiO 2 suggest that the N dopants had a small influence on the crystal field or the crystal structures, no influence on the O 2p orbitals, and a small influence on the Ti 2p orbitals in TiO 2-x N x as compared to pure TiO 2. Furthermore, this 156

190 suggests that the N atoms were substitutionally replaced the original O atoms in the rutile TiO 2 lattice Contribution of the N, C and S dopants to the electronic band structures of TiO 2-x N x, TiO 2-y C y and TiO 2-z S z Intensity / a.u. a) EMS ABS Ti 2p (iv) (iii) (ii) (i) Photon Energy / ev Intensity / a.u. b) EMS ABS O 1s (iv) (iii) (ii) (i) Photon Energy / ev Intensity / a.u. c) Binding Energy / ev VB XPS (iv) (iii) (ii) (i) Figure 8.7 (a) Ti 2p XAS and Ti L XES spectra, (b) O 1s XAS and O K α XES spectra and (c) XPS valence band spectra of (i) pure rutile TiO 2, (ii) TiO 2-x N x, (iii) TiO 2-y C y and TiO 2-z S z. In order to elucidate the effect of the N, C and S dopants in the TiO 2-x N x, TiO 2-y C y and TiO 2-z S z samples on the structural and electronic modification of the TiO 2, the X-ray absorption, emission and XPS spectra of pure rutile TiO 2 and the N-, C- and S-doped TiO 2 are compared in Figure 8.7. It is very interesting that we find the Ti 2p, O 1s XAS and the Ti L, O K α XES specra for all the N-, C- and S-doped TiO 2 to be very similar or identical to those of pure TiO 2 as shown in Figure 8.7a and 8.7b. This again suggests that these N, C, and S dopants have substitutionally replaced the oxygen atoms in the TiO 2 lattice and that this replacement has not brought any apparent structural or crystal field distortion in the doped titania compared to pure titania. In contrast to XAS and XES, which probes the partial density of the conduction/valence band, the XPS valence band 157

191 (VB), probes the total density of states distribution in the valence band. 18,58-62 In the valence XPS spectra shown in Figure 8.7c, there are additional widely dispersed electronic states above the valence band edge in the N-, C- and S-doped TiO 2 compared to pure TiO 2. These states are caused by the N, C and S dopants in the doped TiO 2. This assignment is clearly seen from a comparison of the Ti 2p and O 1s X-ray absorption and emission spectra of the doped and undoped TiO 2 with the valence XPS spectra. This is more clearly seen from the complete electronic band structure of pure TiO 2 as show below by combining the above results from XPS, XAS and XES spectroscopy Correlation between XPS, XAS and XES Results to build a complete electronic band structure of pure TiO 2 Since the valence band contains the contribution from the Ti 3d states, when the Ti 2p core electron is excited enough above the continuum state and the excited electron spreads out of the excited core site, one obtains a Ti 3d 2p emission spectrum. 15 The Ti L emission corresponds to the transition from the Ti 2p core hole state to the valence hole state, which corresponds to the final state of Ti 2p and valence-band photoemission, respectively. 15 Therefore, it is useful to relate the x-ray emission spectra with Ti 2p core and valence photoemission spectra, which can help elucidate the electronic structure of the material. 15 Figure 8.8 shows the assignment of the O K α x-ray emission spectrum (a) from the transitions between the O 1s core-level and the valence band structure (fitting curve) from XPS measurement (b). The peaks A and A are assigned to the transitions from the valence density of state levels to the O 1s core level. Although the two bands in the valence band almost show the density of states as determined from the XPS, the 158

192 transitions to the O 1s core level show one main peak and one shoulder peak. This suggests that the transition probability to the same O 1s core level from the different states in the valence band is different. Since the states in the valence band are strongly hybridized, it is plausible that the transition probability varies. Similarly, the peaks in the Ti L x-ray emission spectrum can be assigned to transitions (a) A' O Kα XES in TiO 2 (b) O 1s XPS Valence XPS in TiO 2 Intensity / a.u. A Intensity / a.u. A A' Photon Energy / ev Binding Energy / ev (c) A Ti L XES in TiO 2 (d) Ti 2p XPS Valence XPS in TiO 2 Intensity / a.u. A' B' B Intensity / a.u. A A' B B' Photon Energy / ev Binding Energy / ev Figure 8.8 Assignment of the O K α x-ray emission spectrum (a) from the transitions between the O 1s core-level and the valence band structure (fitting curve) from XPS measurement (b); Assignment of the Ti L x-ray emission spectrum of (c) from the transitions between the Ti 2p core-level and the valence band structure (fitting curve) from XPS measurement (d). 159

193 from the valence band levels to the Ti 2p core level (c and d), and the transition probability to the same Ti 2p core level is different from the different states in the valence band containing different O 2p and Ti 3d character. Given the Δl = ± 1 selection rule, the features in the O K α and Ti L emission spectra reflect primarily transitions from valence band states having a O 2p character, and states having Ti 3d or 4s character, respectively. Having successfully determined the structure of the O 1s x-ray emission spectrum in TiO 2 by consideration of transitions from the measured core-level and valence-band states, we should also be able to retrieve/construct the structure of the valence band if the O K α x-ray emission spectrum and the O 1s XPS spectrum are known. From the XPS valence band photoemission, the total density of states is obtained, while from the x-ray emission, only the states involving a high transition probability and having the same character, i.e. O 2p, can be obtained. It is reasonable to assume that the transition probabilities from the different valence states with the same character, i.e. O 2p, to the O 1s core level are the same. Thus, the partial valence band structure having O 2p character can be constructed. Intensity / a.u. Expt. VB XPS Fit. VB XPS Total Constr. VB from Ti 2p and O 1s XES Constr. Partial VB from O 1s XES Constr. Partial VB from Ti 2p XES Binding Energy / ev Figure 8.9 Comparison of the sum of the PVBs of O and Ti to the total valence band (VB) from XPS measurement with its fitting. 160

194 In the reconstruction of the conduction band, the basic idea is to relate the x-ray absorption features with the core-level XPS spectrum. In XPS, electrons are excited to the vacuum state (zero binding energy), while in x-ray absorption spectra, the electrons are assumed to be excited to the conduction band. Thus, the energy difference of the corresponding XPS spectra relative to the x-ray absorption spectra (the core-level binding energy minus the x-ray absorption peak energy) gives the energy distribution of the states in the conduction band. If the core-level binding energy is larger than the x-ray absorption peak energy for the same O 1s, or Ti 2p electrons, the corresponding binding energy is a positive value. Otherwise, it has negative values. The other parameters of the peak, such as the width and relative height, are obtained from fitting the corresponding x- ray absorption spectrum. The constructed partial valence band (VB) structure having O 2p character from O 1s core-level XPS and O K α x-ray emission spectra can be constructed as well as the partial valence band having Ti 3d character. Figure 8.9 shows the comparison of the sum of the partial VBs of O and Ti to the total valence band from XPS measurement with its fit. The total density of states as the sum of the partial VBs of the O and Ti matches quite well with the VB from the XPS measurement. The comparison of the reconstructed partial valence bands to the valence band from XPS measurement (with its fit) displays the validity of this method, where the partial VBs from O and Ti are contained in the total VB in TiO 2. Thus, from the components corelevel x-ray emission and XPS results, we can successfully retrieve the partial and total valence band structures, which match with the experimental measurement from XPS. Using the same idea of the correlation of XPS and XAS as shown in Figure 8.10, we can construct the O and Ti partial conduction bands from the O 1s or Ti 2p core-level 161

195 XPS and O 1s or Ti 2p x-ray absorption spectra as shown in Figure The former contains mainly the component having or coupling with states of O 2p character, and the latter contains mainly the Ti 3d component in the conduction band. Intensity / a.u. O 1s XPS O 1s XAS in TiO 2 A B C D E F G H (a) Intensity / a.u. Ti 2p XPS A A' A'' A"' A"" B B' C C' D E Ti 2p XAS in TiO 2 (b) Binding Energy / ev Photon Energy / ev Binding Energy / ev Photon Energy / ev Figure 8.10 (a) The correlation of the O 1s core-level XPS and O 1s x-ray absorption spectra used to construct the partial conduction band (PCB) structure having O 2p characters; (b) The correlation of Ti 2p core-level XPS and Ti 2p x-ray absorption spectra used to construct the partial conduction band (PCB) structure having Ti 3d characters. Constructed Partial Conduction Band from O 1s Constructed Partial Conduction Band from Ti 2p Constructed Conduction Band DOS / a.u Binding Energy / ev Figure The constructed conduction band (CB) and the partial conduction bands (PCBs) from O and Ti in TiO

196 The sum of the O and Ti partial conduction bands should also reflect the total conduction band structure of TiO 2. Figure 8.12 shows the constructed conduction band in pure rutile TiO 2. The comparison between the constructed conduction band displays the consistence with the inverse-photoemission spectrum, 63 the calculation of the band structure, 64 as shown in Figure 8.12A, and also the results from previous bremstrahlung isochromat study on TiO Figure 8.12 (A) Comparison of the inverse-photoemission spectrum (a) [63] and theoretical calculation of the band structure (b) of TiO 2 [64] to the constructed conduction band structure (c) in this contribution. (B) Comparison of the constructed band structures of the TiO 2 in this contribution to the theoretical single particle calculations of the band structures of TiO 2. (a), (c), (e): constructed band structures; (b), (d), (f): theoretical calculations.[10] We can also directly compare the constructed band structure to the theoretical calculations in the literature, which are well established and quite accurate. 10,18 Figure 163

197 8.12B compares the constructed band structures of the pure TiO 2 to a single particle calculation of the band structures in TiO In the comparison, the constructed bands are shifted by the same value to match the positions of the t 2g and e g peaks in the O partial band structures. The good match shown in Figure 8.12 displays again the validity of themethod to build the band structure from XPS and x-ray absorption/emission spectra. It is clear that there are no electronic states available in the bandgap region (> 2.5 ev) of pure rutile TiO 2 above the valence band edge. This again confirms that the additional electronic states in the XPS spectra above the valence band edge of pure TiO 2 are due to the contributions from the N, C and S dopants in the TiO 2-x N x, TiO 2-y C y and TiO 2-z S z samples, which explain exactly the electronic origin of the visible-light absorption property of these doped TiO 2 materials. In summary, the N-, C- and S-doped TiO 2 displayed long tail absorption from the visible into the near-infrared region. The additional states in the valence band can be attributed to the N, C and S dopants in these TiO 2-x N x, TiO 2-y C y and TiO 2-z S z samples. These additional states are consistent with the results that additional states appeare at the edge of the valence band from the calculation by Asahi, where the full-potential linearized augmented plane wave formalism in the framework of the local density approximation (LDA) for C, N, F, P, or S doped TiO 2 crystal was used. 8 This likely explains the origin of the long-tail absorption, and is consistent with indirect experimental evidence from open circuit potential measurement, 65 and photocurrent measurement

198 8.4 Conclusions Visible-light active N-, C- and S-doped titanium dioxide with rutile phase was obtained by treating titanium nitride, carbide and sulfide at high temperatures, and were fully explored with XPS, x-ray absorption and x-ray emission spectroscopy. We found that the N, C and S dopants had a small effect on the crystal structure of the original TiO 2, and had no or a small influence on the O 2p and Ti 3d orbitals in the doped TiO 2 compared to pure TiO 2. Direct experimental evidence was observed for the first time on the modification of the electronic band structure of TiO 2 caused by the N, C and S dopants in these doped materials. Using the x-ray absorption/emission spectra with the core-level XPS, the conduction/valence band structures of pure TiO 2 were then in turn constructed. The validity of this method was tested through a comparison of the constructed valence band with the measured valence band (from XPS), and via comparison of the partial band structure from oxygen and titanium with the theoretical calculations of band structure. The constructed band structure of pure TiO 2 further confirmed the electronic origin of the long-tail absorption of TiO 2-x N x, TiO 2-y C y and TiO 2-z S z. This study helps current research efforts in the doping chemistry and physics of TiO 2 in the visible-light applications. 8.5 References 1. Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W. Chem. Rev. 1995, 95, Grätzel, M. Nature, 2001, 414,

199 3. Millis, A.; Hunte, S. L. J. Photochem. Photobiol. A: Chem., 1997, 108, Linsebigler, A. L.; Lu, G.; Yates Jr., J. T. Chem. Rev. 1995, 95, Fujishima, A.; Rao, T. N.; Tryk, D. A. J. Photochem. Photobio. C: Photochem. Rev. 2000, 1, Tryk, D. A.; Fujishima, A.; Honda, K. Electrochimica Acta 2000, 45, Choi, H. C.; Ahn, H.-J.; Jung, Y. M.; Lee, M. K. et al. Appl. Spectrom. 2004, 58, Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293, Burda, C.; Lou, Y.; Chen, X.; Samia, A. C. S.; Stout, J.; Gole, J. L. Nano Lett. 2003, 3, de Groot, F. M. F.; Faber, J.; Michiels, J. J. M; Czyżyk, M. T.; Abbate, M.; Fuggle, J. C. Phys. Rev. B, 1993, 48, Hwu, Y.; Yao, Y. D.; Cheng, N. F.; Tung, C. Y.; Lin, H. M. NanoStruct. Mater. 1997, 9, Fischer, D. W.; Baun, W. L. J. Appl. Phys. 1968, 39, Henningsson, A.; Rensmo, H.; Sandell, A.; Siegbahn, H.; Södergren, S.; Lindström, H.; Hagfeldt, A. J. Chem. Phys. 2003, 118, Finkelstein, L. D.; Zabolotzky, E. I.; Korotin, M. A.; Shamin, S. N.; Butorin, S. M.; Kurmaev, E. Z.; Nordgren, J. X-ray Spectrosc. 2002, 31,

200 15. Harada, Y.; Kinugasa, T.; Eguchi, R.; Matsubara, M.; Kotani, A.; Watanabe, M.; Yagishita, A.; Shin, S. Phys. Rev. B 2000, 61, Harada, Y.; Watanabe, M.; Eguchi, R.; Ishiwata, Y.; Matsubara, M.; Kotani, A.; Yagishita, A.; Shin, S. J. Electron Spectrosc. Related Phen. 2001, , Ruus, R.; Kikas, A.; Saar, A.; Ausmees, A.; Nõmmiste, E.; Aarik, J.; Aidla, A.; Uustare, T.; Martinson, I. Solid State Comm. 1997, 104, Augustsson, A.; Henningsson, A.; Butorin, S. M.; Siegbahn, H.; Nordgren, J.; Guo, J. H. J. Chem. Phys. 2003, 119, Finkelstein, L. D.; Kurmaev, E. Z.; Korotin, M. A.; Moewes, A.; Schneider, B.; Butorin, S. M.; Guo, J. H.; Nordgren, J.; Hartmann, D.; Neumann, M. D.; Ederer, L. Phys. Rev. B 1999, 60, Okada, K.; Kotani, A. J. Electron Spectros. Related Phenom. 1993, 62, de Groot, F. M. F. J. Electron Spectros. Related Phenom. 1993, 62, Esaka, F.; Furuya, K.; Shimada, H.; Imamura, M.; Matsubayashi, N.; Sato, H.; Nishijima, A.; Kawana, A.; Ichimura, H.; Kikuchi, T. J. Voc. Sci. Technol. A 1997, 15, Soriano, L.; Abbate, M.; Fuggle, J. C.; Prieto, P.; Jiménez, C.; Sanz, J. M.; Galán, L.; Hofmann, S. J. Voc. Sci. Technol. A 1993, 11, Soriano, L.; Abbate, M; Pen, H.; Czyżyk, M. T.; Fuggle, J. C. J. Electron Spectrosc. Relat. Phenom. 1993, 62, Fisher, D. W. J. Appl. Phys. 1970, 41,

201 25. Ihara, H.; Kumashiro, Y.; Itoh, A. Phys. Rev. B 1975, 12, Didziulis, S. V.; Lince, J. R.; Stewart, T. B. E. A. Eklund, Inorg. Chem. 1994, 33, Ohno, Y.; Hirama, K.; Nakai, S.; Sugiura, C.; Okada, S. Phys. Rev. B 1983, 27, Fischer, D. W. Phys. Rev. B 1973, 8, Šimůnek, A.; Šipr, O.; Bocharov, S.; Heumann, D.; Dräger, G. Phys. Rev. B 1997, 56, Georgson, M.; Bray, G.; Claeson, Y.; Nordgren, J.; Ribbing, C. G.; Wassdahl, N. J. Vac. Sci. Technol. A 1991, Nordgren, J.; Bray, G.; Claesson, Y.; Georgson, M.; Ribbing, C. G.; Wassdahl, N. Surface Coat. Technol. 1990, 43-44, Dartigeas, K.; Benoist, L.; Gonbeau, D.; Pfister-Guillouzo, G.; Ouvrard, G.; Levasseur, A. J. Electron Spectrosc. Relat. Phenom. 1997, 83, Faba, M. G.; Gonbeau, D.; Pfister-Guillouzo, G. J. Electron Spectrosc. Relat. Phenom. 1995, 73, Didziulis, S. V.; Lince, J. R.; Stewart, T. B. Inorg. Chem. 1994, 33, Porte, L.; Roux, L.; Hanus, J. Phys. Rev. B 1983, 28, Vasile, M. J.; Emerson, A. B.; Baiocchi, F. A. J. Vac. Sci. Technol. A 1990, 8, Miyauchi, M.; Ikezawa, A.; Tobimatsu, H.; Irie, H.; Hashimoto, K. Phys. Chem. Chem. Phys. 2004, 6, Sakthivel, S.; Kisch, H. Chem. Phys. Chem. 2003, 4, Ohno, T.; Mitsui, T.; Matsumura, M. Chem. Lett. 2003, 32, Umebayashi, T.; Yamaki, T.; Tanaka, S.; Asai, K. Chem. Lett. 2003, 32,

202 41. Irie, H.; Watanabe, Y.; Hashimoto, K. Chem. Lett. 2003, 32, Wu, Z. Y.; Lemoigno, F.; Gressier, P.; Ouvrard, G.; Moreau, P.; Rouxel, J. Phys. Rev. B 1996, 54, R Warwick, T.; Heimann, P.; Mossessian, D.; McKinney, W.; Padmore, H. Rev. Sci. Instrum. 1995, 66, Nordgren, J.; Bray, G.; Cramm, S.; Nyholm, R.; Rubensson, J. -E.; Wassdahl, N. Rev. Sci. Instrum. 1989, 60, Saha, N. C.; Tompkins, H. G. J. Appl. Phys. 1992, 72, Diwald, O.; Thompson, T. L.; Goralski, E. G.; Walck, S. D.; Yates, Jr. J. T. J. Phys. Chem. B 2004, 108, Diwald, O.; Thompson, T. L.; Zubkov, T.; Goralski, E. G.; Walck S. D.; Yates, Jr. J. T. J. Phys. Chem. B 2004, 108, Chen, X.; Burda, C. J. Phys. Chem. B 2004, 108, Gole, J. L.; Stout, J. D.; Burda, C.; Lou, Y.; Chen, X. J. Phys. Chem. B 2004, 108, Irie, H.; Watanabe, Y.; Hashimoto, K. J. Phys. Chem. B 2003, 107, Choi, Y.; Umebayashi, T.; Yoshikawa, M. J. Mater. Sci. 2004, 39, Umebayashi, T.; Yamaki, T.; Yamamoto, S.; Miyashita, A.; Tanaka, S.; Sumita, T.; Asai, K. J. Appl. Phys. 2003, 93, Umebayashi, T.; Yamaki, T.; Itoh, H.; Asai, K. Appl. Phys. Lett. 2003, 81, Choi, Y.; Umebayashi, T.; Yamamoto, S.; Tanaka, S. J. Mater. Sci. Lett. 2003, 22, Martinez, H.; Auriel, C.; Gonbeau, D.; Loudet, M.; Pfister, G. G. Appl. Surf. Sci. 169

203 1996, 93, de Groot, F. M. F.; Fuggle, J. C.; Thole, B. T.; Sawatzky, G. A. Phys. Rev. B 1990, 42, de Groot, F. M. F.; Fuggle, J. C.; Thole, B. T.; Sawatzky, G. A. Phys. Rev. B 1990, 41, Pedio, M.; Fuggle, J. C.; Somers, J.; Umbach, E.; Haase, J.; Lindner, Th.; Hofer, U.; Grioni, M.; de Groot, F. M. F.; Hillert, B.; Becker, L.; Robinson, A. Phys. Rev. B 1989, 40, NIST XPS srdata.nist.gov/xps. 60. Stefan, H. Photoelectron Spectroscopy: Principles and Applications, Springer- Verlag, New York, Cardona, M.; Ley, L. Photoemission in Solids I General Principles, Springer-Verlag, New York, Ley, L.; Cardona, M. Photoemission in Solids II: Case Studies, Springer-Verlag, New York, See, A. K.; Thayer, M. R.; Bartynski, A. Phys. Rev. B 1993, 47, Glassford, K. M.; Chelikowsky, J. R. Phys. Rev. B 1992, 46, Lindegren, T.; Mwabora, J. M.; Avendaňo, J. J.; Hoel, A.; Granqvist, C. G.; Lindquist, S-E. J. Phys. Chem. B 2003, 107, Nakamura, R.; Tanaka, T.; Nakato, Y. J. Phys. Chem. B 2004, 108,

204 CHAPTER 9 Towards Understanding the Connections between Molecules and Nanocrystals via the Uniquely Stable CdSe Molecular Clusters Abstract The study of small clusters helps to elucidate the connection between molecules and nanoparticles. In this paper, the synthesis and investigation of Cd 7 Se 7 clusters are presented. Cd 7 Se 7 clusters are synthesized using the same method to prepare larger CdSe nanocrystals, but at a much lower reaction temperature to arrest their growth. The optical properties of this molecular cluster show characteristics typically attributed to surfacelike states in large nanocrystals from both the steady-state and time-resolved absorption and emission measurements. They are compared and appear to be consistent with calculations using the time-dependent local density approximation method. The possible structure of the Cd 7 Se 7 is proposed with analogy to the carbon frame of diamantane. This structure is energetically stable from previous theoretical calculations and bulk CdSe nanocrystals can be constructed in the same way as diamond is constructed from adamantane. Thus, this study of diamantine-like cluster Cd 7 Se 7 could provide a viable pathway to understand the structure and property connections between the basic molecular building units and larger nanocrystals or bulk crystals. 171

205 9.1 Introduction A remarkable characteristic of semiconductor nanoparticles is the evolution of their physical and chemical properties with size. As atoms arrange to form a crystal a great change of fundamental properties can be observed. The size dependence of the electronic structure and the optical properties of larger semiconductor quantum dots, revealing quantum confinement effects, have been studied extensively. 1-5 Surface and surface-induced properties become increasingly important for smaller systems on the nanometer scale. 2,6 The influence of surface atoms is usually discovered in the photoluminescence (PL) properties of nanoparticles. 6 Deep trap emission occurs in cases of imperfect passivation of the surface, which creates intra-gap electronic states. 1-6 This emission is broad, substantially red shifted from the absorption onset, and is characterized by long lifetimes. The direct observation of the absorption of the surface states was thought unlikely until the recent studies on ZnS nanoparticles. 7 Commonly, low surface contribution causes a lack of direct transitions between surface states in the absorption spectra. 7 Surface and surface-induced properties become increasingly important for smaller systems on the nanometer scale. This trend is expected to be finite and for sufficiently some small cluster sizes, nanocrystalline properties are taken over by molecular properties. This transition between molecule and nanocrystal is anticipated in an intermediate species with very low numbers of participating atoms, namely clusters. Thus, the study of clusters is essential to understand the connection between the molecules and larger nanocrystals or bulk materials. Cadmium chalcogenide clusters have been synthesized and well studied In clusters with less than 100 atoms the surface 172

206 influence on the emission should be enhanced. Moreover, the emission could exhibit characteristics that resemble those of the deep trap emission of nanocrystals. 8 Recent theoretical studies of Cd n Se n clusters show that a significant tailing of sub-threshold absorption associated with unsaturated atoms is expected, 9 and differentiation of absorption and emission transitions from surface states or intrinsic states becomes unnecessary. In this section, we synthesize one of these clusters, Cd 7 Se 7, in the same way we synthesize larger CdSe nanocrystals, but at a much lower temperature. The reason is to arrest the intermediate species, which are the building blocks for larger nanocrystals, before growth into larger nanocrystals. Thus the study of arrested clusters could allow us to make both structure and property connections between the molecules and larger nanocrystals or bulk materials. We find the Cd 7 Se 7 cluster to be unique and stable among the small CdSe clusters and its structure is proposed to have a diamantane framework of Cd and Se atoms. It is the largest cluster in which every atom is surface-exposed in the sense that every atom has less than 4 neighbors within the cluster. Smaller clusters can be derived by removing atoms from the cluster tip, and larger ones can be derived by simply adding atoms onto the exposed hexagons. CdSe nanocrystals or bulk crystals can be constructed from these clusters in the same way as diamond is constructed from adamantane. The steady-state optical properties of the Cd 7 Se 7 cluster were found to be in good agreement with previous theoretical calculations. This cluster shows optical characteristics typically attributed to surface-like states in large nanocrystals from both the steady-state and time-resolved absorption and emission measurements. Thus the 173

207 optical and structural connections have been bridged between the molecular end and the bulk crystal/nanocrystal end via this cluster. 9.2 Experimental The CdSe cluster was synthesized by adding g cadmium stearate (90%, Strem Chemicals) and g trioctylphosphonic oxide (TOPO) (90%, Strem Chemicals) into a three-necked flask and heating up to 200 C for 2 hours under N 2 flow on a Schlenk line. Then hot selenium stock solution ( g Se (99.5%, Alfa Aesar) dissolved in g trioctylphosphine (TOP) (90 %, Aldrich)) was quickly injected. After the injection, the temperature of the mixed solution was dropped to 190 C. This solution was quickly heated up to and reacted at 200 C for 3 hours. The solution was taken out and quenched in cold toluene. The fraction of CdSe nanoparticles in toluene was placed in 1-cm quartz cuvettes. The steady-state UV-visible absorption and PL spectra were collected at room temperature on a Varian Cary 50 and a Varian Eclipse Fluorescence spectrophotometer, respectively. The MALDI-MS spectrum of the CdSe clusters was obtained using 2-(4- hydroxyphenlyazo) benzoic acid (HABA) as the matrix and an indole-3-acetic acid (IAA) matrix in an Applied Biosystems Voyager-DE STR matrix-assisted laser desorption timeof-flight mass spectrometer (MALDI-TOFMS). Femtosecond time-resolved TDA measurements were recorded on a previously described femtosecond laser pump-probe system. 16 The laser system consists of a femtosecond erbium fiber laser oscillator, which is frequency doubled to 780 nm and amplified (Clark MXR CPA 2001). This femtosecond laser produces fundamental pulses 174

208 at 780 nm, 120 fs FWHM duration, and 800 µj output energy per pulse at a repetition rate of 1 khz. A small portion of the fundamental output pulse train is used to generate white light in a 2 mm sapphire crystal while the remaining laser light is used to frequencydouble the fundamental to achieve a wavelength of 390 nm, respectively. The probewavelength range is extended beyond the white light spectrum by using an OPA. The pump-probe experiments are all carried out at ambient temperature and the excitation beam is modulated by a chopper with a 100 Hz frequency. The probe light is used with reflective optics in order to avoid white light dispersion. Measurements are conducted with the excitation beam focused to a spot diameter of about 500 µm and the probe beam to 100 µm. The quantum dot (QD) solution was placed in a 2 mm path length quartz cuvette and continuously stirred by a cell stirrer to avoid permanent bleaching of the pump-probe volume element in the solution. Time-resolved differential absorption (TDA) spectra are obtained by comparing transmitted probe light, with and without the application of the pump pulse. Labview-assisted data acquisition results finally in 2- dimensional matrices (wavelength versus delay time), which are then analyzed by the single-value decomposition method. This results in a global analysis of the spectrum-time matrix as opposed to a kinetic analysis at single wavelengths. Time-resolved emission was measured with a time-correlated single photon counting system. The excitation source was a Spectra-Physics (SP) Tsunami Ti-sapphire laser pumped with a SP Millennia diode-pumped Nd-YVO 4 laser. The repetition rate was set at 4 MHz with a 2.2 ps pulse width. The output was frequency doubled by a SP GWU HG flexible harmonic generator and used as the excitation source. The emission was detected by a cooled Hamamatsu (Hamamatsu, Japan) R3809U-50 microchannel plate 175

209 photomultiplier, with a SPC PC module (Becker & Hickl GmbH Intelligent Measurement and Control Systems, Berlin, Germany) as the photon counting electronics. The excitation energy used was 3.19 ev, and the emission was monitored at 2.63 ev and 2.28 ev. 9.3 Results and Discussion Cd m Se n = (m,n) (7,7) 1386 Intensity / a.u. (3,4) 669 (4,4) 774 (4, 6) 925 (5,6) 1055 (6,5) 1080 (8,8) (9,9) Mass / m/z Figure 9.1 MALDI-MS spectrum of Cd m Se n clusters (m, n = 3-9) with 2-(4- hydroxyphenlyazo) benzoic acid (HABA) as the matrix. The inset shows the whole mass spectrum from 500 to 10,000. For brevity Cd m Se n cluster peaks are labled as (m,n). Mass spectrometry experiments were conducted on these samples. This method has been recently been shown to be effective in evaluating the size of small nanoparticles. 12 Shown in Figure 9.1 is a MALDI-MS spectrum of the CdSe clusters using 2-(4-hydroxyphenlyazo) benzoic acid (HABA) as the matrix. Different Cd m Se n clusters (m, n = 4-9) can be assigned assuming the CdSe clusters were singly charged, consistent with the MALDI-MS spectrum obtained using an indole-3-acetic acid (IAA) 176

210 matrix. From the MS spectrum, the dominant presence of the Cd 7 Se 7 species was found with some mass and size distribution, which is commonly observed in the synthesis of CdSe nanoparticles using similar methods. 2-5 Thus it is reasonable to assume that the optical properties shown below are mainly determined by Cd 7 Se 7. Furthermore, as shown below, the absence of a sizable wavelength dependence of PLE suggests that Cd 7 Se 7 has formed practically as the only species and as a uniquely stable cluster molecule during the synthesis. This assignment is also strengthened by the good match between the experimental measurement and the theoretical prediction of the absorption spectrum as shown later. The smaller fragments may, after further investigation, be explained by cluster fragmentation, while the larger ones can be attributed to the aggregation of fragmented species from Cd 7 Se 7 under the MALDI-MS experimental conditions. Figure 9.2A shows the steady-state UV-visible absorption spectra of the CdSe clusters after different reaction times. Up to ten transitions can be resolved by fitting with Lorentzian functions, which don t change for three samples extracted at different reaction times, suggesting that under the applied conditions with time the concentration increases but not the particle size. The inset shows the spectral positions of the transition energies calculated and experimentally obtained. The shape and positions of the experimentally observed absorption spectra are in good agreement with the absorptions of Cd 7 Se 7 cluster calculated with the time-dependent local density approximation (TDLDA) method. 9 For larger CdSe nanocrystals, the absorption is attributed to electronic transitions from valence band to conduction band states and often a low number of surface states warrant the absence of surface-related transitions in the absorption spectrum. On the other hand, in small CdSe clusters, the valence band and the conduction band states are 177

211 reduced to a few discrete electronic levels, and the distinction between intrinsic and surface states becomes unnecessary. In Cd 7 Se 7, all the atoms are exposed to the surface, and the observed spectral features clearly look like being surface-related. For example, the low energy tail of the absorption spectrum can be assigned to surface-related states as Absorbance / a.u # State Energy / ev A O.D. / a.u ps 0.59 ps 1.17 ps 2.93 ps 89.3 ps 317 ps 1685 ps B O.D. / a.u. C ev 2.68 ev 2.59 ev 2.52 ev 2.39 ev 2.26 ev 2.07 ev 1.91 ev Energy / ev Δ O.D (a.u.) Delay Time (ps) Delay Time / ps Lifetime / ps ~ Energy / ev D τ d2 τ d1 ~ ~ τ r Energy / ev ~ Figure 9.2 (A) Steady state UV-visible absorption spectra of Cd 7 Se 7 clusters, dotted lines: absorption spectra, red solid lines: fitted absorption spectra, green gaussian curves: fitted absorption peaks, inset: transition energy against the number of states, the upward arrow shows the excitation energy in the TDA measurement; (B) TDA spectra of CdSe cluster after different delay times; (C) Dynamics of the TDA spectra detected at different transition energies, the inset shows the early rise of the TDA signals; (D) Lifetime of the TDA dynamics at different detection energies, τ r : rise times of the TDA, τ d1, τ d2 : the first and second components of the decay of the TDA signals. 178

212 theoretically predicted. 9 Figure 9.2B shows the femtosecond time-resolved transient differential absorption (TDA) spectra at different delay times after a laser pump pulse. Figure 9.2C and Figure 9.2D show the decay dynamics and the trend of the lifetimes detected at different energies. In larger CdSe nanocrystal systems, the TDA spectra are featured with a bleach band at the band edge absorption energy due to state-filling effects after laser excitation On the other hand, transient absorption is usually attributed to the trapped charges at surface sites. In this paper, femtosecond time-resolved TDA was used to distinguish between electronic transitions from/to intrinsic states and surface states. While the first ones cause temporary bleaching (due to state filling) of the ground state absorption, the latter ones cause transient excited state absorption (due to excited surface-state population). Here, we have used the femtosecond TDA technique to distinguish the nature of the involved states. For the cluster sample studied, transient absorption spectra were observed in the range from 1.8 ev to 2.8 ev. In analogy to larger nanocrystals, the transient absorption is akin to surface-state transient absorption after initial relaxation. This situation is quite similar to those observed in fullerenes. 19 The envelope shape and position of the early transient absorption spectra resemble those of the steady state UV-visible absorption spectrum but are red-shifted, suggesting both transitions are surface-related. The rise times of the transient absorption increased with the transition energies (transient absorption blue shift) in a time range from 200 to 700 fs, which suggests that upon 3.2 ev (388 nm) laser excitation, the carriers relax on this timescale. The decay of the transient absorption in the visible spectral range had two time components of ps and ps, dependent on the energy of the transition. The decay of the transient absorption had the same trend as the rise, i.e. that the 179

213 population and depopulation/relaxation process is faster for the higher states than that for the lower states. This blue-shift in transient absorption measurements is commonly observed for molecules, in contrast to nanoparticles where bleach is observed that redshifts with delay time. In summary, the rise of the transient absorption occurs within 700 fs, while the decay has two components of 50 ps and 500 ps, with a concurrent blue-shift corresponding to a relaxation rate of 0.45eV/μs. PL Intensity / a.u. A PLE Intensity / a.u. Intensity / a.u. B Energy / ev Delay Time / ns Figure 9.3 (A) PL and PLE spectra of the Cd 7 Se 7 clusters, the upwards arrows indicate the position where lifetimes were measured as in B; (B) Time-resolved decay of the PL of Cd 7 Se 7 clusters. Figure 9.3A shows the PL and photoluminescence excitation (PLE) spectra of the CdSe cluster. In the PL measurement the sample was excited at 4.05 ev and in the PLE measurement the emission at 2.48 ev was monitored. The PL spectrum, which shows a broad emission band around 2.48 ev, exhibits emissions from higher excited states that can be resolved. The PL and PLE spectra can be fitted using a sum of Gaussian bands, and up to 9 peaks can be resolved in each spectrum (See supporting infomation). In the overlap region between 2.5 ev and 3.8 ev, these different states can be found correspondingly in the PL, PLE, and absorption spectra. The transition states in the PL 180

214 and PLE spectra overlapped also very well with those states resolved in the absorption spectrum. Usually, the PL process is affected not only by the direct absorption but also by successive relaxation processes involving structural changes. Here, the PL seems to be exactly centered at the absorption onset. And the large overlap area between the PL and the absorption suggests that there is little geometrical distortion in the excited state compared to the ground state. At the same time, in the absorption spectrum the oscillator strength increases gradually with energy, while the emission intensity decreases. Again, this mirror image spectral behavior is very much akin to molecules rather than to nanocrystals. Usually, the PL is dominated by radiative relaxation from the lowest excited state, since the lowest excited state often has a significantly longer lifetime. The possibility that the well-resolved peaks in the PL would be due to vibrational sublevels in the lowest electronically excited state is excluded, since the vibrational energy of Cd between Se is typically around 205 cm -1, which is about 1/8 of the energy separation between these peaks in the PL spectrum. These peaks can be attributed to radiative relaxation from different excited states with similar lifetimes. This assignment is further verified by lifetime measurement at different PL energies as shown below. The investigated CdSe clusters have a very low PL quantum yield of ~ 0.6% at room temperature referred to anthracene (27% in ethanol) excited at 3.3 ev (380 nm). Since the PL is determined by the competing radiative and non-radiative relaxation rates of the excited states, the low PL yield suggests that for these clusters the non-radiative relaxation pathways are dominant. This phenomenon can be attributed to the presence of surface states (in the language for crystals) or the occurrence of fast vibrational energy dissipation (molecular description) in the investigated clusters. The PL and PLE spectra 181

215 at different excitation and detection energies were obtained. There was no wavelength dependence observed, providing evidence for purity of the sample, and suggesting Cd 7 Se 7 as a uniquely stable cluster and possibly the only molecule synthesized under the current synthetic conditions. Figure 9.3B shows the PL lifetimes of the Cd 7 Se 7 cluster. By using a picosecond single photon counting system 20 to monitor the decay of the emission at 2.63 ev near the peak and at 2.28 ev, which corresponds to two resolved peaks in the PL spectrum, the lifetime and nature of these states can be obtained. Three components with lifetimes of 10 ± 3 ns, 37 ± 7 ns, and 140 ± 20 ns can be fitted at both energies. The lifetimes are very similar at all different energies, suggesting that the nature of transition at these energies is similar as well. A lifetime in tens of nanoseconds suggests that the emission is from different electronically excited states instead of from different vibrational levels within the lowest electronically excited state, since relaxation from higher vibrational levels to the lowest vibrational level in the same electronic states occurs in picoseconds, while the lifetimes here are in nanoseconds. 21 A dependency of the PL lifetime on the growth time was not observed within the experimental precision. Again, this speaks for the formation of pure and discrete molecular cluster Cd 7 Se 7. The low PL yield of < 1% and short lifetime in the nanosecond regime of the cluster molecules could well be due to effective singlet-triplet state splitting, 22 meaning that the observed PL is more of a pure fluorescence than the mixed PL observed in larger nanoparticles due to a large exchange interaction. The almost equally spaced resonances with ΔE 200 mev in the absorption, PL and PLE spectra are not due to the vibrational 182

216 energies in the Cd 7 Se 7 cluster, but due to the different excited electronic states with similar excited state lifetimes. Exp. Trans. Energy / ev B Cal. Trans. Energy / ev Figure 9.4 (A) Comparison of the UV-visible absorption spectra (solid line) and the calculation (dashed line) performed by the TDLDA method. (B) Comparison of the calculated and experimental transition energies in the absorption spectra shown in A. We compared the experimental absorption spectrum with the calculated spectra using local density approximation (LDA) (See supporting information) and TDLDA methods, and found that the experimental results matched better with results of the TDLDA method than those of the LDA method. Figure 9.4A shows a comparison of the UV-visible absorption spectrum of the Cd 7 Se 7 cluster to theoretical calculations using the TDLDA method, where the Cd 7 Se 7 was assumed free of ligands. Qualitatively the curves agree well in terms of the threshold and the peak energies since in most cases the peaks line up. In two cases the calculated peak occurs at an experimentally observed minimum (at 3.1 and 3.7 ev). Figure 9.4B shows a comparison of the experimentally measured and calculated energies of corresponding electronic transitions of Cd 7 Se 7 clusters. The 2.59 ev, 3.11 ev and 3.76 ev transition energies are shown by a circle. These two sets of energy states were simply numbered from low energy to high energy. Clearly, the 183

217 experimental and calculated values of the corresponding electronic states match well. The circled transitions can be interpreted as forbidden transitions (from the calculations). Possibly, the transition rules are not strictly observed in the experimental measurements. In the calculations, the Cd 7 Se 7 cluster was treated as an isolated system, while in the measurement the solvent and ligands can affect the effective symmetries and transition moments. We found that transition energies are not very much dependent on the inclusion of surface ligands like TOPO. This is consistent with previous studies. 23,24 The interaction of ligands with the cadmium atoms on the surface is expected to be rather weak, and thus these clusters are considered to be models for ligand-free clusters. 1,2,23 The theoretical calculation by Puzder et al 24 suggests that the presence of surfactant ligands has little effect on the optical properties and the relaxed structures of clusters. For a cluster smaller than the attenuation length of the excitation light, the PLE data are proportional to the absorption coefficient. This means that all the peaks that are present in the absorption should be more or less present in the PLE spectrum. This is exactly the case here, and this fact was used for a comparison between absorption and PLE and between PLE data and theoretical data (See supporting information). This also demonstrates the good correlation between the corresponding optical transitions, obtained by theory and experiment. Figure 9.5 shows the proposed structures for the stable cluster Cd 7 Se 7 with other fractional clusters. The structural similarity of Cd 7 Se 7 to that of diamantane can explain why it has such a unique stability. The zincblende CdSe nanocrystals have the same structural framework as that of diamond. CdSe nanocrystals or bulk crystals can be built up with Cd 7 Se 7 units in the same way as diamond can be built using diamantine units. The proposed structure for Cd 7 Se 7 shown above is maintained after geometrical 184

218 optimization. The optimized structure has a little structural distortion from the ideal diamantane structure due to the difference in the bond covalent strength. This structure is consistent with the recent theoretical calculation of Deglmann and co-workers. 23 From their calculation for Cd n Se n (n =1-7) clusters, the stable structures had common features of six-membered rings, and exhibited high symmetry. The calculated atomization energy per CdSe unit showed the Cd 7 Se 7 to be the most stable one among these clusters with the highest atomization energy. Figure 9.5 Proposed structure of Cd 3 Se 4, Cd 4 Se 4, Cd 4 Se 6 and Cd 7 Se 7 with the corresponding carbon frameworks of norbornane, bicyclo-[2,2,2]-octane, adamantane and diamantane, zincblende CdSe and diamond. 185

219 From the mass spectrum in Figure 9.1, we find that besides the dominant Cd 7 Se 7, there exist other CdSe clusters in the time of flight (TOF) experiment. For example Cd 5 Se 6, Cd 6 Se 5, Cd 4 Se 6, Cd 4 Se 4, Cd 3 Se 4, are found towards the molecular end. The structural similarity of Cd 3 Se 4 and Cd 4 Se 4 to the carbon frameworks of norbornane and bicyclo-[2,2,2] Octane, explain the existence of these clusters. The Cd 4 Se 6 and Cd 6 Se 4 structures are particularly stable and indeed can be envisioned as fractures of the adamantane structure. It seems reasonable to believe that Cd 4 Se 6 or Cd 6 Se 4 is less stable than Cd 7 Se 7, considering its charge balance and the integrated mass peaks. In addition, Cd 8 Se 8, and Cd 9 Se 9 are found towards the nanocrystal end, where any of the exposed hexagon surfaces can be capped with another Cd or Se atom. All these clusters can be easily built around this diamantane-type Cd 7 Se 7 cluster by adding or removing Cd or Se atoms. Furthermore, Cd 5 Se 5 was not found in the mass spectrum. And indeed, following the model of an adamantane-derived structure strictly, it is not possible to build an alternating Cd 5 Se 5 structure. On the other hand, Cd 4 Se 6 has the structure of adamantane. Cluster fragments other than Cd 7 Se 7 may have been produced during the laser induced ionization of the TOF measurement by fragmentation and aggregation. MALDI-MS is a soft ionization technique, where minor fragmentation can be expected, leading to smaller clusters. On the other hand, aggregation cannot be completely excluded, particularly within the laser-induced surface plasma, which could form larger clusters. From mass spectrometry we can confirm that particles with masses around Cd 7 Se 7 are of highest abundance and unique stability. With further investigation of optical properties, we conclude that Cd 7 Se 7 is by far the dominant species as discussed above. In previous calculations, Cd 6 Se 6 was predicted to have a magic number behavior. 9 This is 186

220 not observed here, and there are several possible reasons. First, the bare cluster system calculated was different from the synthetic environment. Second, in the calculations, only several model clusters were investigated, while under the present experimental conditions, there may exist structures with higher stability. Third, the synthetic condition here may preferably form Cd 7 Se 7 clusters, but not Cd 6 Se 6. This does not necessarily mean that Cd 7 Se 7 is more stable than Cd 6 Se 6. It is remarkable that the experimental optical properties of the Cd 7 Se 7 cluster show an excellent agreement with the theoretical calculations for the Cd 7 Se 7 cluster. In the study by Deglmann et al., 23 the excitation spectra for two larger clusters Cd 30 Se 30, and Cd 44 Se 44 showed the onset of the absorption around 1.75 ev and 1.0 ev, respectively. Their spectra also showed almost linearly increasing energies for transition states for these two clusters, similar to the spectrum observed here. Springborg et al. 25 also reported some theoretical calculations on some CdSe and CdS clusters, and found that with increasing cluster size the gap between the highest occupied and lowest unoccupied molecular orbitals (HOMO/LUMO gap) is decreasing. Since there was no calculation on Cd 7 Se 7, direct comparison of the observed spectrum here with their results is not straightforward. 9.4 Conclusions In summary, the building blocks for bigger nanocrystals, CdSe clusters were synthesized in the same way to synthesize larger CdSe nanocrystals, but at a much lower temperature, before growth into larger nanocrystals. The Cd 7 Se 7 cluster was identified as one with a unique stability. The absorption spectrum for Cd 7 Se 7 was consistent with theoretical calculations using the TDLDA method. The optical properties show 187

221 characteristics of surface-related/molecular transitions, demonstrating that optical transitions in these small nanoclusters are similar in nature to the surface state transitions in larger nanocrystals. A possible structure for the Cd 7 Se 7 is proposed in analogy to the carbon frame of diamantane. The proposed Cd 7 Se 7 is stable, consistent with previous theoretical calculations. Bulk CdSe crystals/nanocrystals can be constructed based on Cd 7 Se 7 in the same way as diamond can be constructed from diamantane. Thus, this study of diamantane-like Cd 7 Se 7 cluster could provide a viable pathway to understand the structure and property connections between nanocrystals or bulk crystals and the basic molecular building units. 9.5 References 1. Alivisatos, A. P. J. Phys. Chem. 1996, 100, Murray, C. B.; Norris, D. J.; Bawendi, M. G. J. Am. Chem. Soc. 1993, 115, ; Murray, C. B.; Kagan, C. R.; Bawendi, M. G. Annu. Rev. Mater. Sci. 2000, 30, Nirmal, M.; Brus, L. Acc. Chem. Res. 1999, 32, ; Bawendi, M. G.; Carroll, P. J.; Wilson, W. L.; Brus, L. E. J. Chem. Phys. 1992, 96, Peng, Z., Peng X. J. Am. Chem. Soc. 2001, 123, ; Peng, X. Adv. Mater. 2003, 15, ; Peng, Z. A.; Peng, X. J. Am. Chem. Soc. 2002, 124, Burda, C.; Link, S.; Mohamed, M.; El-Sayed, M. J. Phys. Chem. B 2001, 105, ; Chen, X.; Lou, Y.; Samia, A. C.; Burda, C. Nano Lett. 2003, 3,

222 6. Bawendi, M. G.; Carroll, P. J.; Wilson, W. L.; Brus, L. E. J. Chem. Phys., 1992, 96, Nirmal, M.; Norris, D. J.; Kuno, M.; Bawendi, M. G.; Efros, Al. L. M. Rosen, Phys. Rev. Lett. 1995, 2001, Chen, W.; Wang, Z.; Lin, Z.; Lin, L. J. Appl. Phys. 1997, 82, Chen, W. in Handbook of Nanostructured Materials and Nanotechnology, Ed. by Nalwa, H. S. Vol. 4: Optical Properties, Chapter 5, p , Academic Press, New York, Behrens, S.; Bettenhausen, M.; Eichhöfer, A.; Fenske, D. Angew. Chem. Int. Ed. Engl. 1997, 36, ; Soloviev, V. N.; Eichhöfer, A.; Fenske, D.; Banin, U. J. Am. Chem. Soc. 2001, 123, Soloviev, V. N.; Eichhöfer, A.; Fenske, D.; Banin, U. J. Am. Chem. Soc. 2000, 122, Troparevsky, M. C.; Kronik, L.; Chelikowsky, J. R. Phys. Rev. B, 2001, 65, / / Herron, N.; Calabrese, J. C.; Farneth, W. E.; Wang, Y. Science 1993, 259, Dean, P. A. W.; Vittal, J. J.; Payne, N. C. Inorg. Chem. 1987, 26, Anjali, K. S.; Vittal, J. J. Inorg. Chem Comm. 2000, 3, Hursthouse, M. B.; Malik, M. A.; Motevalli, M.; O Brien, P. J. Mater. Chem. 1992, 2, Abrahams, I.; Malik, M. A.; Motevalli, M.; O Brien, P. J. Organomet. Chem. 1994, 465, 73-77; Khitrov, G. A.; Strouse, G. F. J. Am. Chem. Soc. 2003, 125, Lee, G. S. H.; Fisher, K. J.; Craig, D. C.; Scudder, M. L.; Dance, I. G. J. Am. Chem. Soc. 1990, 112,

223 14. Vossmeier, T.; Reck, G.; Schulz, B.; Katsikas, L.; Weller, H. J. Am. Chem. Soc. 1995, 117, Brennan, J. G.; Siegrist, T.; Carroll, P. J.; Stuczynski, S. M.; Brus, L. E.; Steigerwald, M. L. J. Am. Chem. Soc. 1989, 111, Cheng, Y.; Emge, T. J.; Brennan, J. G. Inorg. Chem. 1994, 33, Lou, Y.; Samia, A. C. S.; Cowen, J.; Banger, K.; Chen, X.; Lee, H.; Burda, C. Phys. Chem. Chem. Phys. 2003, 5, Klimov, V. I. J. Phys. Chem. B 2000, 104, Klimov, V. I.; McBranch, D. W.; Leatherdale, C. A.; Bawendi, M. G. Phys. Rev. B 1999, 60, Burda, C.; Link, S.; Green, T. C.; El-Sayed, M. A J. Phys. Chem. B 1999, 103, Burda, C.; Link, S.; Mohamed, M.; El-Sayed, M. J. Phys. Chem. B 2001, 105, Burda, C.; Samia, A. C. S.; Hathcock, D. J.; Huang, H.; Yang, S. J. Am. Chem. Soc. 2002, 124, Cody, J.; Dennisson, J.; Gilmore, J.; VanDerveer, D. G.; et al. Inorg. Chem 2003, 42, Lakowicz, J. R. Principles of Fluorescence Spectroscopy, Kluwer Academic/Plenum Publishers, New York, Franceschetti, A.; Wang, L. W.; Fu, H.; Zunger, A. Physical Review 1998, 58, R13367-R Deglmann, P.; Ahlrichs, R.; Tsereteli, K J. Chem. Phys. 2002, 116, Puzder, A.; Williamson, A. J.; Gygi, F.; Galli, G. Phys. Rev. Lett. 2004, 92, / /4. 190

224 25. Sarkar, P.; Springborg, M. Phys. Rev. B 2003, 68, / /7; Joswig, J.-O.; Seifert, G.; Niehaus, T. A.; Springborg, M. J. Phys. Chem. B 2003, 107,

225 Supporting information for chapter 9 S-Figure 1. The fitting of the PL spectrum of Cd 7 Se 7 cluster S-Figure 2. The fitting of the PLE spectrum for Cd 7 Se 7 cluster S-Figure 3. Comparison of the experimental absorption spectrum with the calculated absorption spectrum using LDA method. S-Figure 4. Comparison of the experimental absorption spectrum with the calculated absorption spectrum using LDA method. S-Figure 5. The Comparison of energies of the corresponding peaks in the absorption, PL, PLE spectra and the calculated absorption spectra for Cd 7 Se 7 cluster. 192

226 Wavelength / nm PL Intensity / a.u Energy / ev S-Figure 1. The fitting of the PL spectrum of Cd 7 Se 7 cluster 193

227 PLE Intensity / a.u Energy / ev S-Figure 2. The fitting of the PLE spectrum for Cd 7 Se 7 cluster 194

228 S-Figure 3. Comparison of the experimental absorption spectrum with the calculated absorption spectrum using the LDA method. 195

229 S-Figure 4. Comparison of the experimental absorption spectrum with the calculated absorption spectrum using the LDA method. 196

230 Energy / ev Cal PLE Emission Abs # Transition S-Figure 5. The Comparison of energies of the corresponding peaks in the absorption, PL, PLE spectra and the calculated absorption spectra for the Cd 7 Se 7 cluster. 197

231 CHAPTER 10 The Crystallization Process in 2 nm CdSe Quantum Dots and Related Surface Induced Optical Property Changes * Abstract Investigation of the growth of CdSe nanocrystals (~ 160 atoms) to the uniquely stable size of 2 nm allows the monitoring of the crystallization process in semiconductor quantum dots. By using a combination of optical techniques, high-resolution transmission electron microscopy (HRTEM), and powder X-ray diffractometry (XRD), new phenomena were explored during the CdSe nanocrystal growth process, which involved significant morphological reconstruction and crystallization of the initially formed amorphous nanoparticles. During crystallization, the absorption onset of the CdSe quantum dots blue shifted towards higher energies at 3 ev (414 nm), while the photoluminescence red shifted to lower energies. Furthermore, an apparent increase of the Stokes shift was observed during the formation of small CdSe nanoparticles. On the other hand, the photoluminescence excitation spectrum showed constant features over the entire reaction time. The transient differential absorption displayed bleach with peak position on the red-side of the first peak in the corresponding absorption spectrum, and this bleach peak blue-shifted as the nanoparticles crystallized over time. Additionally, * Part of this chapter is published in J. Am. Chem. Soc. 2005, 127,

232 results from HRTEM and XRD studies show that the CdSe nanoparticles were amorphous at the early reaction stages and became increasingly crystallized after longer reaction times, while the particle size remained the same during the crystallization process. These observations demonstrate the important role of the surface on the optical properties of small CdSe quantum dots and facilitated spectroscopic monitoring of the crystallization process in the quantum dots. Accompanying the structural reconstruction of surface atoms during crystallization of the small nanoparticles, the underlying electronic structures of the surface states underwent dramatic changes. The absorption, transient differential absorption, and the photoluminescence spectra were attributed to optical transitions involved with the surface (defect) states in the bandgap of the small CdSe nanoparticles, while the photoluminescence excitation peaks were attributed to intrinsic transitions from the valence band to the conduction band with energies that were independent of the surface configurations. The blue-shifts in the absorption, transient differential absorption, and photoluminescence during the crystallization process are caused by the reconstruction of surface atoms towards configurations with lower free energies and shallower defects near the electronic (valence and conduction) bandedge. In the steadystate absorption, all optical transitions involved with the surface states were allowed, while in the femtoseconds transient absorption experiments, the optical transitions between the surface states above the valence bandedge and those below the conduction bandedge were blocked, inducing the lower-energy bleach. 199

233 10.1 Introduction Quantum size confinement occurs as the size of materials is reduced to the nanometer regime. This phenomenon induces discrete electronic states and an increase or blue shift of the band edge transition energy in nanoparticles (NPs) compared to the corresponding bulk materials. 1-4 Another important consequence is the manifestation of highly surface-related properties, from structural transformation to the emergence of unique optical properties, 1-3,5,6 due to the high surface-to-volume ratio in NPs. In a NP the surface atoms usually have fewer adjacent coordination atoms and more dangling bonds, and can be treated as defects compared to the bulk atoms. 1-4 These defects induce additional electronic states in the band gap and can mix with the intrinsic states to a substantial extent, and may also influence the spacing of the energy levels and optical properties of NPs. 1-4 At high densities of surface defects, a decrease in the observed transition energy and a red shifted emission band can be observed due to defect band formation, 3,7 which can be investigated by monitoring the photoluminescence (PL) of the NPs. 1-6 Theoretical studies suggest that there should be absorption peaks at lower energies in the bandgap with considerable contribution from dangling surface orbitals, or mixed (intrinsic/surface) states. 8,9 However, such low energy features are usually due to the low defect content of NPs 7 and the small transition dipole moments between intrinsic and surface states, and were not observed in absorption experiments 8 until recent studies on ZnS NPs. 7 The crystallization process during particle growth involves structural reconstruction and surface relaxation, which have been proposed theoretically 8,11 and experimentally, 10 such morphological changes are due to the surface strain in 200

234 nanocrystals 6,10 and the minimization of the total free energy. 8 The surface can induce reversible structural transformations in different environments due to the different interfacial free energies assumed by the surface of the NPs. 13 Surface reconstruction and structural relaxation can also reduce the permanent dipole moment in CdSe nanocrystals, 11,12 thus changing their optical properties. Conversely, the change of optical properties of nanocrystals can be used to investigate the crystallization process. The growth and crystallization processes are highly important issues in the preparation of NPs. Highly crystalline CdSe NPs can be obtained at elevated temperatures, and several methods have been developed to improve the crystallinity of NPs. 2,5 Recently, the nucleation and growth of CdSe NPs have been investigated in-situ. 4 Usually, the CdSe NPs size grow over time; this process involves simultaneous growth and crystallization, and the optical properties of the NPs reflect mostly the size change of the NPs caused by the growth process. 2,4 In this report, the crystallization process is observed separately from the growth process and investigated during the controlled synthesis of CdSe NPs towards a particle size of 2.0 nm, which is considered as an especially stable size for CdSe NPs with an associated band edge absorption centered at 414 nm that has been observed earlier. 2,4 Specially stable sizes have been known for many years as certain electronic and geometric combinations of atoms can yield stable structures with unique atomic arrangements 2,4,5 that can prevent the growth of the NPs. This in turn facilitates isolated observation of the crystallization process, during NP growth if the reaction is carefully controlled. During this process, new phenomena, including blue-shift of the absorption and the increasing Stokes shift over time, have been observed as the reaction proceeds. 201

235 10.2 Experimental For a typical synthesis of the CdSe NPs, g cadmium stearate (tech. 90%, Strem Chemicals) and g TOPO (tech. 90%, Strem Chemicals) were loaded in a three-neck flask and heated up to 240 C for 2 hours under Ar flow in a Schlenk line. Then hot selenium stock solution ( g Se (99.5%, Alfa Aesar) dissolved in g TPP (99 %, Strem Chemicals)) was quickly injected. After the injection, the temperature of the mixed solution dropped to 200 C. This solution was quickly heated to 220 C and reacted at this temperature for 90 minutes. As the reaction progressed, 0.1 ml fractions of the CdSe sample were taken out and quenched in 2 ml cold toluene. Afterwards, the fractions of CdSe NPs in toluene were placed in 1-cm quartz cuvettes and their steady-state UV-visible absorption and PL spectra were collected at room temperature on a Varian Cary 50 and a Varian Eclipse Fluorescence spectrophotometer, respectively. The high-resolution transmission electron microscopy (HRTEM) was performed on a Tecnai F30 machine operated at 300 kv. Specimens for TEM analysis were prepared by depositing a drop of NP solution onto a copper grid supporting a film of amorphous carbon. The grid was dried in air prior to HRTEM measurements. The X-ray diffraction patterns were obtained using a Philips PW 3710 X- ray powder diffractometer. The time-resolved spectra were measured using the femtosecond differential absorption (TDA) method in a femtosecond laser pump-probe system. This system consisted of an amplified erbium-doped fiber laser, which was frequency doubled to

236 nm and amplified in a regenerative amplifier (Clark MXR CPA 2001). It produced 1 khz pulses with 120 fs FWHM duration and 800 µj output energy per pulse. A small part (4%) of the fundamental generated a white light continuum as probe in a 2 mm sapphire plate. The remaining laser light (96%) was frequency doubled as pump with a 0.2 mm BBO crystal. The pump beam was modulated by a chopper with a 100 Hz frequency. Reflective optics was used in order to avoid white light dispersion for the probe light. The pump beam was focused to a spot diameter of about 500 µm and the probe beam to 100 µm. The pump-probe measurements were all carried out at ambient temperature. The nanoparticle solution was placed in a quartz cuvette of a 2 mm path length and continuously stirred by a cell stirrer to avoid permanent bleaching of the pump-probe volume element in the solution during the measurement. 2-dimensional matrixes (wavelength versus delay time) were obtained with a labview-assisted data acquisition method, and were analyzed by the single-value decomposition method. This resulted in a global analysis of the spectrum-time matrix as opposed to a kinetic analysis at single wavelengths Results The temporal evolution of the UV-visible spectra of the CdSe NPs is shown in Figure 1. The reaction time was counted after the injection of the Se stock solution into the Cd TOPO mixture. Within the first 10 min of the reaction, no absorption was detected in the visible range. However, after 10 minutes of reaction time, a small absorption peak starting at 2.56 ev (486 nm) was observed. This absorption band then shifted to 2.86 ev (434 nm) after 30 min and was observed to move to 3.00 ev (414 nm) as the reaction was 203

237 prolonged to 90 min. Overall, a blue-shift of the absorption peak was observed as the reaction proceeded over time. Wavelength / nm PL Intensity / a.u. t Absorbance / a.u Energy /ev Figure 10.1 UV-visible (solid lines) and photoluminescence (PL) (dashed lines) spectra of CdSe NPs observed after 10 min (black), 15 min (red), 30 min (green), 45 min (blue), 60 min (cyan) and 90 min (magenta) reaction time. On the other hand, the emission spectra (Figure 10.1) recorded under 3.27 ev (380 nm) light excitation, featured a broad band around 2.49 ev (500 nm) and a long tail up to 1.78 ev (700 nm). Unlike the blue-shift of the absorption band, the emission spectra red-shifted as the NPs formed. The emission peak was measured at 2.55 ev (488 nm) after 10 min and then shifted to 2.44 ev (508 nm) and 2.37 ev (523 nm) after 30 min and 90 min reaction times, respectively. Furthermore, the emission spectra initially had an asymmetric shape, and as the CdSe NPs formed, the long wavelength tail of the emission became less pronounced. Shown in Figure 10.2 is the temporal evolution of the Stokes shift, defined as the energy difference between the first peak of the absorption spectrum and the emission peak, of the CdSe NPs investigated in this study. From Figure 10.2 one can see that the 204

238 Stokes shift increased from ev to ev and ev as the CdSe NPs formed in 10 min, 30 min, and 90 min, reaction times, respectively. Energy /ev Photoluminescence Stokes shift Absorption Growth Time / min Wavelength /nm Figure 10.2 Evolution of the Stokes shift of the CdSe NPs during synthesis. PLE Intensity/ a.u. Wavelength / nm S 90min 3/2 1S e 2S 3/2 1S e 60min 45min 30min 15min 10min Energy / ev Figure 10.3 Temporal evolution of the PLE spectra of the CdSe NPs during crystal formation. The emission wavelengths of the PLE spectra for a sample of 10 min (black), 15 min (red), 30 min (green), 45 min (blue), 60 min (cyan), and 90 min (magenta) reaction time are 474 nm, 490 nm, 500 nm, 510 nm, 510 nm, and 525 nm, respectively. (For clarity the spectra were normalized and offset.) 205

239 By performing photoluminescence excitation (PLE) measurements, one can gain insights on the origin of the observed emission. Furthermore, it is also useful for revealing the purity and the size of the samples. 7,14,15 As seen in Figure 10.3, all the samples have two peaks in their PLE spectra, one around 2.94 ev (423 nm) and the other around 3.28 ev (379 nm). Unlike the absorption and emission spectra, the energies of these two excitation peaks in the PLE spectra were independent of the formation time of CdSe NPs. By taking a series of PL spectra at different excitation energies and a series of PLE spectra at different emission energies one observes that the PL spectra as well as the PLE are consistent in shape and lack wavelength dependence, suggesting that the samples are of good homogeneity. From these measurements, evidence for a narrow size distribution and a uniquely stable size was derived. 2 nm Figure 10.4 HRTEM images of CdSe NPs after 10 min (Left) and 90 min (Right) particle formation. Compared to the amorphous phase of the 10 min sample, the 90 min sample was well crystallized. The HRTEM studies performed for the 10 min and 90 min samples are shown in Figure 10.4, which displayed that both these samples had similar sizes around 2.0 nm, 206

240 consistent with the results in the literature. 2 The darker areas in left image are the amorphous CdSe nanoparticles, and their sizes are around 2.0 nm. For the 10 min sample, no lattice fringes were observed in the HRTEM images of the randomly oriented NPs (Figure 4). Thus, the 10 min sample was suggested to be amorphous, since from the statistical point of view, there should be a chance to resolve the lattice fringe at least for some NPs if the samples were well crystallized and randomly oriented on the carbon grid. In contrast, the 90 min sample was well crystallized and the lattice structure could be resolved (Figure 10.4). 90 min 10 min Intensity / a.u Theta / o Figure 10.5 XRD patterns of the CdSe NPs formed after 10 min and 90 min. Due to the small size of these samples, the XRD signal was weak for both samples. The XRD results shows that these CdSe NPs had similar sizes around 2.1 ± 0.1 nm, and the CdSe NPs formed after 90 min s reaction had better crystallinity as evidenced from their stronger diffraction pattern. This observation is consistent with the XRD results, where the 90 min sample exhibited a stronger diffraction pattern than the 10 min sample (Figure 10.5) after peak fittings. The similar FWHMs of the corresponding diffracting peaks suggests that the size of these CdSe NPs was similar, which was estimated to be 2.1 ± 0.1 nm in diameter, 2 207

241 containing about 160 atoms in total. 2,8,11 The similar sizes of these two samples suggested that the crystallization process is by far dominant over the growth process in the formation of CdSe NPs, at least in between 10 min. and 90 min. reaction times. From the HRTEM, well-crystallized CdSe nanocrystals form over time from the amorphous structures with similar sizes via a gradual crystallization process. -Δ(OD) norm Wavelength / nm a) 10min 15min 30min 45min 60min 90min t Energy / ev Lifetime / ps τ 2 ~ 12 ~ 9 6 τ 1 3 ~ ~ τ r Reaction Time / min b) Figure 10.6 a). The changes of the TDA spectra extracted at 2 ps delay time for the different CdSe nanoparticles during the crystallization process. b). The evolution of different components (τ r : rise lifetime, τ 1 : first component of bleach decay, τ 2 : second component of bleach decay) of lifetimes of decay dynamics of the bleaches for different samples. The change of TDA spectra for these CdSe nanoparticles during the crystallization process is shown in Figure 10.6a. TDA displayed bleach with peak position on the red-side of the first peak in the corresponding absorption spectrum, and this bleach peak blue-shifted as the CdSe nanoparticles crystallized over time, similar to the blue-shift of the absorption of the CdSe nanoparticles. The bleach peak moved from around 2.45 ev (506 nm) (10 min), to 2.57 ev (483 nm) and 2.67 ev (464 nm), after 30 min and 60 min, respectively. For the 90 min sample, the bleach shifted to below 2.78 ev 208

242 (450 nm) by estimation, which is on the edge of the probe light in our pump-probe experiment. The lifetimes of the decay dynamics of the bleaches over time are shown in Figure 10.6b. As seen in this figure, the rise time of the bleach peak occurred at less than 1 ps, the first component of the bleach decay occurred in the range of several ps, while the second component of the bleach decay occurred on the hundred ps time scale. As the CdSe nanoparticles crystallized with time, the lifetime of the bleach formation increased, while the lifetime of the bleach decreased Discussion The slow crystallization was possible by using triphenylphosphine (TPP) instead of TOP as the capping material. The large steric demand of the TPP capping material can cause inefficient capping on the surface of CdSe NPs (coordinated to the Se atom) and thus can result in more surface defects. At the same time, the formation temperature was lowered to control the transformation rate of the CdSe NPs, thus allowing for a longer formation time for small CdSe NPs in order to extend the observation time window. In general, higher quality CdSe NPs can be obtained at higher formation temperatures and with longer formation times. 2,4 At lower temperatures, however, the CdSe NPs formed usually have many surface states at the initial stage, and the growth and crystallization rate is much slower, thus providing an opportunity to investigate the phenomena related to the growth and crystallization process of NPs. As observed in the HRTEM images and verified in the PLE measurements, the crystallization process of crystallinity is dominant over the growth process in size in the presented reaction. Thus, the growth process can be 209

243 safely considered to be negligible, and the observed optical property changes of NPs can be attributed to the isolated crystallization process of the particularly stable 2.0 nm CdSe NPs. As the reaction proceeds, a blue shift in absorption and a red shift in emission were observed, with similar PLE features. These phenomena can be elucidated in light of the gradual removal of surface defects during the crystallization process. Usually as NPs grow, the intrinsic absorption, due to transitions from the valence band to the conduction band levels, shifts towards red. 1-4 The absorption properties observed here can therefore be attributed to transitions involving surface defects states (in 2 nm NPs, 70 % of the atoms are on the surface). Theory predicts weak and broad surface-state absorption within the bandgap with a low oscillator strength 8,9 occurring as a long-wavelength tail in the optical absorption spectrum. These transitions are likely to be coupled with intrinsic states. 8 In this study, the CdSe NPs formed at shorter times show deep trap states giving rise to the long-wavelength absorption. 7,9 During the crystallization process of the CdSe NPs, surface atoms can be rearranged until a minimization in surface energy is reached. 8,13 The blue-shift of the absorption, as the NPs develop, reflects the correspondingly gradual removal of the initial trap and surface states during the cystallization process. According to the model involving the narrowing of the NP size distribution during the preparation of CdSe NPs, 4 the initially observed broad absorption peak could also be explained by an inhomogeneous size distribution, with subsequent concentration of NPs around 2.0 nm. However, this possibility is excluded since the blueshifted absorption would mean that the NPs significantly shrink with time, which is considered unlikely under these growth conditions. 210

244 In the assignment of the PL contributions, the band gap PL and the trap state PL should be considered. The band gap PL is narrow (width determined by the size distribution) and is only slightly Stokes-shifted from the absorption onset. 1-4 Trap-state PL is broad and is substantially red shifted from the absorption onset (typically > 0.2 ev). 1-4 The initial broad widths and large red shifts of the PLs are consistent with the characteristics of deep-trap emissions in the NPs. This PL is thus assigned to transitions involving the trap states. The long tail emission suggests broad distribution of surface and defect states and relatively low density of states at each energy. 14,16 The change from broad and asymmetric into narrow and symmetric shape of the emission spectra of these CdSe NPs confirms the conclusions from the UV-visible measurement that during the formation of these small CdSe NPs, the initial deep trap surface states are gradually removed. Moreover, the distribution of surface defect states becomes narrower as the surface gradually reconstructs and becomes more ordered. During this reconstruction the NPs gradually crystallize when the reaction proceeds over time 1,17. Previous works on larger NPs have reported that the Stokes shift becomes smaller as the NPs grow. 1-4,14 In contrast, the Stokes shift of a molecule or small cluster is very small, since it originates from the energy difference of the ground state vibrational levels. The increase in the Stokes shift is consistent with the gradual crystallization of NPs in that the NPs possess mesostable structures with many defects (amorphous) in the beginning and they reconstruct and crystallize to minimize the total energy, thereby removing defects. The main contribution of the increased Stokes shift derives from the shift of absorption to blue during growth, with smaller contributions from the shift of PL to red. This can be understood since the observation of the emission from surface-related states requires a much lower surface defect density, while the observation of surface- 211

245 related absorption requires a high surface defect density. With longer formation times (e.g. 90 min sample), the absorption quickly resembles the intrinsic transitions, while the emission is still quite affected by the surface defects. The Stokes shift of the CdSe sample after the 90 min reaction, is consistent with the extrapolated values from previous work. 14,16 In the PLE spectra, the energies of the two excitation peaks were independent of the formation time of the CdSe NPs. This suggests that these are the intrinsic state transitions, and that these CdSe NPs have the same size during formation. They can be assigned to the lowest two excited states: 1S 3/2 1S e and 2S 3/2 1S e, respectively, which match quite well with the extrapolated values from the PLE and transient differential absorption (TDA) measurements. 14 This assignment is also consistent with theoretical predictions that the excitonic energy is not sensitive to surface relaxation and reconstructions. 8 In addition, the transition intensity at 2.94 ev (423 nm) increases faster than tht of the 2S 3/2 1S e transition at 3.28 ev (379 nm). This suggests that the lower excited state 1S 3/2 1S e is more strongly coupled to the surface states than the higher excited state 2S 3/2 1S e. The gradual change in relative intensities is due to changes in coupling with the surface states. At the same time, the transition energies are not affected, supporting further that the measurements presented here describe the gradual crystallization of the CdSe NPs at a given size of 2.0 nm. In the following, we will discuss how the surface affects the optical property of the small CdSe nanoparticles during the crystallization process. 212

246 The electronic structure change of the small CdSe nanoparticles in the crystallization process. Accompanied with the structural ordering of the CdSe nanoparticles, i.e. the surface reconstruction/relaxation during crystallization, the electronic structure of the CdSe nanoparticles change accordingly. First, the initially formed deeper surface states or defect states of the amorphous phase are removed and shallower surface states form near the bandedge during the crystallization process. Second, in the amorphous phase of CdSe nanoparticles, the surface states are widely spread and have low density at each energy interval; when CdSe nanoparticles become more crystallized, these surface states gather to a narrower region with a higher density at each energy interval, and move closer to the bandedge. The change of the electronic structure of CdSe nanoparticles during crystallization induces the observed optical changes in both steady-state and timeresolved spectra as discussed in the following The surface induced absorption evolution during crystallization Absorption spectroscopy is the commonly used technique to explore the quantum effects in semiconductor nanoparticles, especially the development of discrete features in the spectra and the enlargement of the energy gap. 1-12,20-32 The absorption intensity is determined by the numbers (or density) of the occupied levels in the ground state, the numbers (or density) of the unoccupied levels in the excited state, and the transition probability. The absorption constant is given by 2 πe K = N( E) f ij (10.1) nm 213

247 where N(E) is the density of states function and f ij is the oscillator strength determining the transition probability between i and j states. 7 The inhomogeneous distribution of the samples (including shape and the surface defect distributions) can change N(E), which may change the overlap of the transitions and result in the broadening of the absorption band. It has been suggested that the disappearance of the direct transition involved with the surface states in the absorption spectra is caused the low surface content, 7 since most nanoparticles form with a high surface quality with different stabilizing agents by capping or coating with organic or inorganic materials and transitions. The disappearance of the optical transitions in the absorption can be understood from equation (1). First, in bulk or nanomaterials with sizes larger than 5 nm, 99% or at least 70% of the atoms are in the bulk but NOT on the surface. Second, unlike the ordered bulk atoms, the surface atoms have more irregular environments, or much broader site distributions. Third, the above two reasons induce a much lower electronic density at each energy interval for the surface Figure 10.7 The illustration of the origin of absorption evolution over the crystallization of small CdSe nanoparticles, where the blue curve illustrates the steady-state absorption spectra. 0 refers to the ground state, while 1S 1S and 3 / 2 e 2S / 21S e 3 refer to the lower two states in the optical transitions in the small CdSe nanoparticles, the distribution of the 214

248 surface states are indicated as the discrete lines. The boldness of the lines and the green Gaussian shape areas illustrate the density of the electronic states. atoms compared to the high electronic density in the condensed electronic band for the bulk atoms. Thus, the optical transition probability involved with surface states is much less compared to that of the bulk states in materials with a size larger than 5 nm without a capping ligand on the surface. If the surface atoms are capped with a ligand, the number of surface states and thus the optical transition probability from surface states dramatically diminishes. However, in 2.0 nm CdSe nanoparticles, more than 60% of the atoms are on the surface, and the large ligand allows an incomplete capping on the surface. This allows the optical transitions from the surface states to become apparent. The optical transitions in the absorption, which include transitions from the bulk and the surface, during the crystallization process, are illustrated in Figure 10.7 for the small CdSe nanoparticles. The unique absorption properties observed here are attributed to transitions to the manifold of surface defects states. Under the conditions described here, the first peak of the absorption blue-shifts as the nanoparticles crystallize with time, due to the movement of the initial deep surface states during the crystallization process. The CdSe nanoparticles initially formed have an amorphous structure, which means that the surface atoms are highly disordered. This induces a broad and low-density with a broad defect distribution including deep and shallow surface defects; such defects give rise to the long wavelength absorption with a weak oscillation strength. During crystallization, surface atoms rearrange to optimize their structure until a minimization in total energy occurs, where the deeper defects move to shallower sites and the distribution of defects becomes 215

249 narrower and denser. This process eventually induces high density of surface defects near the bandedge, which causes the observed blue-shift in the onset of the absorption spectra The surface induced PL and PLE evolutions during the crystallization PLE is a very useful technique for exploring the optical and electronic properties of semiconductor nanoparticles, by monitoring a narrow spectral region of the full luminescence while scanning the excitation energy. 7,15,22,26 The photo-excitation rate is given by R = Cq( ν ) Kdν (10.2) where C is the velocity of radiation within the material, q(ν) is the photon density of radiation, and K is the absorption constant. 7 As q(ν) increases very rapidly with increasing wavelength, the significant part of the integral in the above equation is concentrated around the absorption edge. A particular transition may not be observed by PLE if q(ν) is zero, even if absorption transitions between the two states occur. The transition between two states may be detected by absorption as long as the transition is allowed. With PLE, only radiative transitions can be observed. This means that in a PLE spectrum, optical transitions involved with bulk states predominate over those involved with surface states, since non-radiative transitions dominate over radiative transitions in the latter case as seen in the low PL quantum yield below. Thus PLE measurements reveal what transitions are most likely from the bulk. In the PLE spectra shown in Figure 10.3, these two excitation peaks display constant energies of 2.94 ev (423 nm) and 3.28 ev (379 nm), independent of the different crystallization stages of the CdSe nanoparticles. They can be assigned to the 216

250 lowest two excited intrinsic states: 1S 3/2 1S e and 2S 3/2 1S e, respectively, which match quite well with the extrapolated values from the PLE and transient differential absorption (TDA) measurements. 14,26 The energy of the lowest excited state 1S 3/2 1S e also matched with the energy of the first absorption peak for the sample crystallized for a longer time (90 min). The relative intensity between these two peaks is affected by the crystallization time or by the extent of crystallinity of the CdSe nanoparticles. The lower excitation peak at 2.94 ev (423 nm) increases faster than the higher one, 2S 3/2 1S e at 3.28 ev (379 nm). This suggests that the lower excited state 1S 3/2 1S e was is more strongly coupled to the surface states than the higher excited state 1S 3/2 1S e, due to the electronic structure changes of surface atoms during the crystallization process. During crystallization, the surface states are moving much closer and thus inducing stronger coupling with the lower intrinsic state 1S 3/2 1S e than the higher instrinsic state 2S 3/2 1S e, thus inducing a stronger overlap between the surface states and the lower intrinsic state. Since the PLE spectra are due to the bulk (or intrinsic) states, while the absorption spectra can be from both the bulk states or the surface states, this proves to be very useful in differentiating the origin of optical transitions in the absorption spectra. For nanoparticles with few surface defects, the PLE spectra for both the band-edge and the deep-trap emissions are similar to the absorption spectra, 7,18 since both are characterized by the optical transitions of intrinsic (or bulk) states. However, if absorption is dominated by transitions to the surface states, the PLE spectra could be different from those due to intrinsic absorption, which is exactly the case here by comparing the PLE and absorption spectra. This further verifies that the absorption spectra of CdSe nanoparticles are dominated by transitions of surface defect states in the small CdSe nanoparticles due to the high surface state content. 217

251 The quantum yield of these CdSe nanoparticles excited at 380 nm is between 4 ~ 6 %, and increases slightly as CdSe nanoparticles crystallize. The reference used in the quantum yield measurement is Anthracene (27% in ethanol). The optical density of CdSe nanoparticle solutions is adjusted to between 0.3 and 0.5. The PL quantum yield is a result of a competition between the radiative and nonradiative relaxation rates. The low PL yields suggests that the dominant relaxation pathways are nonradiative, consistent with the high surface defect content within the small CdSe nanoparticles. Figure Illustration of the origin of PL evolution during crystallization of small CdSe nanoparticles, where the blue curve illustrates the steady-state emission spectrum. T refers to the dark states or the triplet states. The boldness of the magenta down arrows illustrates the probability of the optical transitions. Other signs have the same meanings as in Figure The involvement and influence of the surface states are usually found in the PL of CdSe nanoparticles. 1-4,9-12 In the assignment of the PL, band gap PL and deep trap PL should be distinguished. In the relaxation process of the optically excited nanoparticles, the higher excited states can directly relax to the ground states in a radiative manner, or nonradiatively relax into the lower excited state and radiatively relax, or continue nonradiative relaxation into dark states or surface defect states and then radiatively relax 218

252 to the ground state as illustrated in Figure And the final emission is the competition between the nonradiative relaxation on the defect with the radiative relaxation via either dark state 14 or surface defect states. Also, there is a competition between these two radiative relaxation pathways via dark states and surface states. The band gap PL is narrow (width determined by the size distribution), and is only slightly Stokes-shifted from the absorption onset. The trap state PL is broad and is substantially red shifted from the absorption onset (typically > 0.2 ev). 14,15,22,26,29 The broad widths and large red shifts of the PLs are consistent with the characteristics of deep-trapped emissions in the nanoparticles. The PL is thus assigned to transitions involving the surface. This centered emission can be attributed to transitions from near band edge states. The long tail emission suggests broad distribution of surface defect states. 2,26,14,29 The change of the main PL channels during the crystallization process is also demonstrated in Figure 10.8 and Figure In the amorphous phase, the surface defect states are widely distributed and have a low density, and because of the weak coupling with the dark state, the resultant PL shows long and asymmetric tail. When the nanoparticles crystallize over time, the distribution of surface defect states is narrowed with an increased density and moves near the bandedge, which can overlap with the dark state in energy, resulting in the round and symmetric PL spectra. As the CdSe nanoparticle crystallizes over time, the coupling between the surface defect states and the dark state also become stronger. The induced splitting explains the red shift of the PL peak over the crystallization process as illustrated in Figure When the surface defect states are broadly distributed over energy and the local density is low, the coupling between the states and the dark state is weak, and the main radiative relaxation pathway is via the dark state, featured with a sharp peak at higher energies and 219

253 a long PL tail at low energies (see PL for the 10 min sample). As the distribution of surface defect states is narrowed and moves up over energy and the local density is enhanced, the coupling between these states and the dark state becomes stronger. This results in a lowering of the energy of dark states by the coupling, and the main radiative relaxation pathway via the dark state features with a round peak at lower energies (see PL for the 90 min sample). Therefore, the PL peak moves to lower energies during the crystallization process of CdSe nanoparticles. Figure 10.9 Illustration of the origin of emission evolution during the formation of small CdSe nanoparticles. The signs have the same meanings as in Figure Surface Induced Bleach in the Femtosecond TDA Measurements The TDA spectra (pump on minus pump off) contains both bleached transitions and induced absorptions arising from the nanoparticles optically excited by the pump, and displays the absorption change induced by the 120 fs laser pump pulse. The bleaching in the TDA spectra is caused by the state-filling of surface states in the photoexcited nanoparticles. 5,14,22 The blue-shift of the bleach peaks over time is due to the state-filling of the blue-shifted surface states during the crystallization process. 220

254 Compared to the ground state absorption peak, the bleach peak of the surface states occurs at a lower energy. This could be explained by state-filling of surface defect states of lower energies and lower densities, due to the much higher radiation intensity of laser pulse during the laser experiment compared to that of conventional UV lamp light. Figure Illustration of the origin of bleach in the TDA of small CdSe nanoparticles, where the TDA is red-shift to the onset of the UV-visible absorption spectra, and the blue curves illustrate the steady-state and transient absorption spectra after the pump light. Other signs have the same meanings as in Figure 4. Similar to the state-filling of the intrinsic bandedge states in high quality CdSe nanoparticles, the bleach here is due to the block of the optical transitions between the surface states above the bandedge of the valence band and the surface states below the conduction band, caused by the state-filling of the defect states, and occurs at the lower energy tail of the steady-state absorption spectra. The state filling of the surface states is similar to the state-filling of the bandedge (intrinsic) states in high quality CdSe nanoparticles as reported in many publications. 5,14,17,22,30,31 The difference is that the bleach due to the filling of defect states occurs at a lower energy compared to the steady-state absorption energy, while the bleach due to the filling of bandedge (intrinsic) states occurs at the exact energy of the steady- 221

255 state absorption. The origin of this difference is that in near defect free nanoparticles, the number of surface states at lower energies is much smaller than that of bandedge states; while in small CdSe nanoparticles with high surface defect content, the number of surface states below the peak of the surface states is comparable to that of the peak. Besides the parameters in equation (10.1) determining the real optical transition probability or intensity, the different intensity of the illumination light in the UV-visible absorption and the pump-probe measurements also affect the results. The much higher radiation intensity in the laser pump light compared to the low radiation intensity in the conventional UV-absorption lamp or the white probe laser light, compensates for the difference in the number of lower surface states by that of the peak. As a result, enough optical transitions are induced below the energy of the peak of surface states, and result in a red-shift of the bleach compared to the UV-visible absorption peak which corresponds to the peak of surface states. The detailed possible optical transitions in the UV-visible absorption measurement are illustrated in Figure 10.10a. The optical transitions involved with the surface defect states in CdSe nanoparticles include optical transitions from the bandedge of the valence band to surface states below the conduction band, transitions from the surface states above the bandedge of the valence band to the conduction band, and transitions from the surface states above the bandedge of the valence band to the surface states below the conduction band. As demonstrated in Figure 10.10b in the femtosecond transient absorption measurement, the optical transitions between the surface states above the bandedge of the valence band and the surface states below the conduction band are blocked, due to the filling of these surface states by the fast relaxation of the excited holes and electrons from 222

256 higher states. The absorption peak in the steady-state absorption spectrum can be attributed to transitions between the bandedge states and the surface defect states, since their transition probabilities are much higher than those between the surface states, which are blocked in the femtoseconds transient absorption experiment. Thus, the bleach of optical transitions from the surface states to surface states occurs at the low energy tail of the steady-state absorption spectra, and the peak of the bleach is red-shifted to the absorption peak. The assignment of the bleach due to state-filling of surface states is supported by the decay dynamics of the bleach previously shown in Figure 10.6b. The decay dynamics for these samples could only be fitted to a two-component exponential function; contrary to the three-component exponential functions used for the bandgap bleach dynamics. This displays the different origin of the bleach signals observed. The bleaches are attributed to the state-filling of surface-induced defect or bandgap states, contrary to the normal state-filling of the bulk or bandgap states. The relaxation channels for defect states here are less than those for the bandgap states, since bandgap states can first relax to the bandgap states. The relaxation pathways of the bleach from surface states include radiative and nonradiative relaxation, where the former is the origin of the PL, namely surface state PL. 7,14,29,33 The lifetimes of the surface state decay are consistent with the results obtained by femtosecond fluorescence upconversion studies. 33 The fast component of the decays of the surface state bleach was attributed to the ultrafast relaxation of the surface selenium dangling bond electrons to the valence band where they combine radiatively with the photogenerated holes. 28 The slow component of the decays can be attributed to the relaxation to the triplet states or the dark states, which couple with the surface states and cause phosphorescence. 14,33 223

257 During the crystallization process of the CdSe nanoparticles, the time to fill surface defect states from higher excited states increases, and the time to dump these states decreases, since as a CdSe nanoparticle crystallizes, shallower surface defect states are formed, and thus the bleach decays faster due to nonradiative and radiative relaxation as luminescence Conclusions Although in most studies the growth and crystallization of NPs occur simutaneously, 2,4 here the crystallization process is isolated and studied in the controlled crystallization of CdSe NPs at an especially stable size of 2.0 nm under suitable conditions. The combination of HRTEM, XRD, UV-vis absorption, PL and PLE results suggest that the CdSe NP samples over different formation times had similar sizes and the initially formed 2.0 nm CdSe NPs underwent gradual crystallization at this particularly stable size, for the following reasons. First, the similar FWHMs of the XRD and the similar features of the PLE spectra over time suggest that these NPs have similar sizes, since the PLE are mostly due to transitions between the intrinsic states in the bands. 3,7,14 Second, the HRTEM and XRD reveal that the initially formed CdSe NPs have amorphous structures, while the NPs formed at later stages were better crystallized. Third, the changes in absorption and emission of these CdSe NPs over time further displays this process. A size around 2.0 nm with an absorption at 3.00 ev (414 nm) has been suggested to be a very stable structure for CdSe NP. 2,4,5 The initially formed amorphous CdSe nanoparticles gradually crystallize in a way that the deeper surface defects are first removed followed by the shallower defects. This involves rearrangement and 224

258 reconstruction of atoms towards thermodynamic stability. 4 Thus, the gradual crystallization of CdSe NPs involves surface reconstruction, consistent with other reports. 8,13 The Stokes shift of the CdSe NPs increases from ev to ev as the particle forms due to the energy difference between the surface states and intrinsic states (PL peak). The surface-induced optical property changes during the crystallization process of small CdSe nanoparticles were investigated on both their steady-state and time-resolved spectra. 1). During the crystallization process, the absorption of CdSe nanoparticles blue shifts to higher energies as they grow. The absorption spectra are attributed to optical transitions involved with the surface defect states in the bandgap. The blue-shift of the absorption spectra can be explained by the gradual removal of the surface defect states from the deeper states to shallower ones in the bandgap due to gradual reconstruction of the surface during crystallization. 2). As the CdSe nanoparticles crystallizes, the PL with a broad band and a low energy tail, shifts to lower energies corresponds to transitions between the dark states and the surface defect states. 3). During the crystallization process, the energies of the PLE peaks, attributed to transitions from the intrinsic excited states: 1S 3/2 1S e and 2S 3/2 1S e,were independent of the crystallization time and the crystallinility of the CdSe nanoparticles; while the relative intensity of the PLE peaks was affected, with the lower excitation peak increasing faster than the higher one, due to the stronger coupling between the lower excited state and the surface states. 225

259 4). The Stokes shift of the CdSe nanoparticles increases from ev to ev as the particle crystallized. This is due to the coupling between the surface states and the dark states. 5). In the TDA spectra, the bleach was red-shifted to the first peak in the absorption spectrum, due to state-filling of the surface states of lower energies; the blueshift of the bleach during the crystallization process was due to state-filling of the shifted surface states as the CdSe nanoparticles become more crystallized. 6.) The relaxation dynamics of the bleach of surface state transitions in the TDA spectra, can be fitted to a two-exponential functions, contrary to the three exponential functions required for the bandedge state transitions. This is because the number of the relaxation channels of surface states was less than that of the intrinsic bandedge transitions. 7). The optical transitions in the steady state absorption include transitions from the bandedge of the valence band to the surface states below the conduction band, transitions from the surface states above the bandedge of the valence band to the conduction band, and transitions from the surface states above the bandedge of the valence band to the surface states below the conduction band. While in the TDA spectra, the transitions from surface states above the bandedge of the valence band to the surface states below the conduction band are blocked. In summary, optical techniques were combined with HRTEM and XRD to investigate the crystallization of small CdSe nanocrystals towards a stable size of 2.0 nm. This involves gradual removal of surface defects and reconstruction of CdSe NPs. Thus, the crystallization process is also a surface related phenomenon. The unique stability of this size of NPs allows the separation of the crystallization process from the growth 226

260 process. While the optical properties due to the intrinsic states (PLE) are not much affected by the crystallization process, other optical properties, which are related to defect changes, are strongly influenced in this crystallization process References 1. (a) Alivisatos, A. P. J. Phys. Chem. 1996, 100, (b) Alivisatos, A. P. Endeavour 1997, 21, (a) Murray, C. B.; Norris, D. J.; Bawendi, M. G. J. Am. Chem. Soc. 1993, 115, (b) Murray, C. B.; Kagan, C. R.; Bawendi, M. G. Annu. Rev. Mater. Sci. 2000, 30, (a) Nirmal, M.; Brus, L.; Acc. Chem. Res. 1999, 32, (b) Bawendi, M. G.; Carroll, P. J.; Wilson, W. L.; Brus, L. E. J. Chem. Phys. 1992, 96, (a) Peng, Z.; Peng, X. J. Am. Chem. Soc. 2001, 123, (b) Peng, X. Adv. Mater. 2003, 15, ; Peng, Z. A. (c) Peng, X. J. Am. Chem. Soc. 2002, 124, (c) Qu, L.; Yu, W. W.; Peng, X. Nano Lett. 2004, 4, (d) Qu, L.; Peng, X. J. Am. Chem. Soc. 2002, 124, (a) Burda, C.; Link, S.; Mohamed, M.; El-Sayed, M. J. Phys. Chem. B 2001, 105, (b) Landes, C.; Braun, M.; Burda, C.; El-Sayed, M. A. Nano Lett. 2001, 1, Chen, X.; Lou, Y.; Samia, A. C.; Burda, C. Nano Lett. 2003, 3, Chen, X.; Lou, Y.; Samia, A. C. S.; Burda, C. J. Am. Chem. Soc. 2005, 127,

261 7. (a) Chen, W.; Wang, Z.; Lin, Z.; Lin, L. J. Appl. Phys. 1997, 82, (b) Chen, W. Handbook of Nanostructured Materials and Nanotechnology, Ed. by Nalwa, H. S. Vol. 4: Optical Properties, Chapter 5, p , Academic Press, New York, Leung, K.; Whaley, K. B. J. Chem. Phys. 1999, 110, (a) Troparevsky, M. C.; Kronik, L.; Chelikowsky, J. R. Phys. Rev. B, 2001, 65, / /4. (b) Troparevsky, M. C.; Kronik, L.; Chelikowsky, J. R. J. Chem. Phys. 2003, 119, McGinley, C.; Riedler, M.; Möller, T.; Borchert, H.; Haubold, S.; Haase, M.; Weller, H. Phys. Rev. B 2002, 65, / / (a) Rabani, E. J. Chem. Phys. 2001, 115, (b) Rabani, E.; Hetényi, B.; Berne, B. J.; Brus, L. E. J. Chem. Phys. 1999, 110, (a) Blanton, S. A.; Leheny, R. L.; Hines, M. A.; Guyot-Sionnest, P. Phys. Rev. Lett. 1997, 79, (b) Empedocles, S. A.; Bawendi, M. G. Science 1997, 278, Zhang, H.; Gilbert, B.; Huang, F.; Banfield, J. F. Nature 2003, 424, (a) Norris, D. J.; Bawendi, M. G. J. Chem. Phys. 1995, 103, (b) Norris, D. J.; Bawendi, M. G. Phys. Rev. B 1996, 53, (c) Norris, D. J.; Sacra, A.; Murray, C. B.; Bawendi, M. G. Phys. Rev. Lett. 1994, 72, (d) Nirmal, M.; Norris, D. J.; Kuno, M.; Bawendi, M. G.; Efros, Al. L.; Rosen, M. Phys. 228

262 Rev. Lett. 1995, 2001, (e) Nirmal, M.; Murray, C. B.; Bawendi, M. G. Phys. Rev. B 1994, 50, (a) Soloviev, V. N.; Eichhöfer, A.; Fenske, D.; Banin, U. J. Am. Chem. Soc. 2001, 123, (b) Soloviev, V. N.; Eichhöfer, A.; Fenske, D.; Banin, U. J. Am. Chem. Soc. 2000, 122, Efros, Al. L.; Rosen, M.; Kuno, M.; Nirmal, M.; Norris, D. J.; Bawendi, M. Phys. Rev. B 1996, 54, Zhang, J.; Wang, X.; Xiao, M.; Qu, L.; Peng, X. Appl. Phys. Lett. 2002, 81, a) Hoheisel, W.; Colvin, V. L.; Johnson, C. S.; Alivisatos, A. P. J. Chem. Phys. 1994, 101, b) Greenham, N. C.; Peng, X.; Alivisatos, A. P. Phys. Rev. B 1996, 54, Behrens, S.; Bettenhausen, M.; Deveson, A. C.; Eichhöfer, A.; Fenske, D.; Lohde, A.; Woggon, U.; Angew. Chem., Int. Ed. Engl. 1996, 35, Behrens, S.; Bettenhausen, M.; Eichhöfer, A.; Fenske, D. Angew. Chem., Int. Ed. Engl. 1997, 24, Behrens, S.; Fenske, D. Ber. Bunsen-Ges. Phys. Chem. 1997, 101, Soloviev, V. N.; Eichhöfer, A.; Fenske, D.; Banin, U. Phys. Stat. Sol. (B) 2001, 224, 1, Shim, M.; Guyot-Sionnest, P. Nature 2000, 407, 981. b) Shim, M.; Wang, C.; Guyot-Sionnest, P. J. Phys. Chem. B 2001, 105,

263 24. Xu, C.; Wang, W. N.; Zhang, W. H.; Zhuang, J.; Liu, L.; Kong, Y. Q.; Zhao, L.; Long, Y. C.; Fan, K. N.; Qian, S. X.; Li, Y. F. J. Phys. Chem. A 2000, 104, Eichkorn, K.; Ahlrichs, R. Chem. Phys. Lett. 1998, 288, Landes, C.; Braun, M.; Burda, C.; El-Sayed, M. A. Nano Lett. 2001, 1, Burda, C.; Link, S.; Mohamed, M. B.; El-Sayed, M. J. Chem. Phys. 2002, 116, Burda, C.; Link, S.; Green, T. C.; El-Sayed, M. A. J. Phys. Chem. B, 1999, 103, Mohamed, M. B.; Burda, C.; El-Sayed, M. A. Nano Lett. 2001, 1, Klimov, V. I. J. Phys. Chem. B 2000, 104, 6112; Klimov, V. I.; McBranch, D. W.; Leatherdale, C. A.; Bawendi, M. G. Phys. Rev. B 1999, 60, Klimov, V. I.; Schwarz, Ch. J.; McBranch, D. W.; Leatherlade, C. A.; Bawendi, M. G. Phys. Rev. B 1999, 60, R Klimov, V. I. Handbook of Nanostructured Materials and Nanotechnology, Ed. by H. S. Nalwa, Vol. 4: Optical Properties, Chapter 7, 451, Academic Press, New York, Underwood, D. F.; Kippeny, T.; Rosenthal, S. J. J. Phys. Chem. 2001, 105,

264 CHAPTER 11 Coherency Strain Effects on the Optical Response of Core/Shell Hetero-Nanostructures: A Spectroscopic Investigation * Abstract Core/shell-structured CdSe/CdS nanoparticles were prepared by a one-pot procedure. The interface effect was studied by steady-state and time-resolved photoluminescence spectroscopy. Coherency strain across the CdSe/CdS interface is addressed to rationalize the measured optical response. The optical properties of the different core/shell systems agree with Matthews-Blakeslee theory, which predicts a critical thickness for a defect-free shell of less than two monolayers capping material for the CdSe/CdS system. * Part of this chapter is published published in Nano Lett. 2003, 3, 799; Int. J. Nanotech. 2004, 1,

265 11.1 Introduction As the size of a semiconductor decreases and becomes similar to or smaller than that of the exciton Bohr radius of the bulk material, quantum confinement occurs because of the space-confined motion of electrons and holes in nanomaterials. 1-3 This induces discrete electronic states in the valence and conduction bands and an increase of interband transition energies in the semiconductor nanoparticles compared to their bulk counterparts. Furthermore, as the size of the material is reduced, surface atoms become more important. For bulk materials, internal atoms dominate the properties of the materials while for nanomaterials the surface and interface atoms may be dominant. 1 Surface atoms usually have unsaturated or dangling bonds, while heterostructure interfaces contain strain-induced defects. This induces extra electronic states in the band gap, which act as electron and hole trapping centers and potentially quench photoluminescence (PL) from the material. Spectroscopy is useful for investigating the optical response due to the interface effects of core/shell nanoparticles. Both steady-state and time-resolved PL spectroscopy have been used to explore the electronic structure of II-VI semiconductor nanoparticles alone and with surface adsorbates in solution. 4-6 Studies on the optical dynamics of CdSe nanoparticles reveal that excitation above the band gap energy leads to a bleach of the lowest absorption transition within a few hundred femtoseconds. 6-8 The bleach occurs through band-filling and its decay occurs through several routes, e.g. through surface and bulk trapping Electron-hole recombination can occur nonradiatively by phonon-assisted relaxation and Auger mechanisms. 11,13 The observed recovery is multiexponential and occurs on several time scales within a few picoseconds to microseconds. 7,14 232

266 Since the surface of a nanoparticle is made up of unsaturated atoms, it is reactive for overgrowth with other reagents (e.g. other semiconductor capping materials). Previous work 15 has shown that epitaxial growth can occur in these systems. Organic surfactants and inorganic materials have been used as passivating materials. 2,14,15 The capped nanoparticles can display higher a PL quantum yield, typically associated with longer PL lifetimes. Compared to organic capping materials, inorganic epitaxial growth can eliminate both the anionic and cationic surface dangling bonds and also generate new nanoparticle core/shell systems with novel properties, which depend on both core and shell materials. 14,15 The coherency strain, which enables the shell material at the interface to adapt the lattice parameters of the core, can play an important role in such systems. In this study, CdSe/CdS core/shell nanostructures are synthesized and investigated using steady-state and time-resolved spectroscopy. The results are then discussed in the light of a coherency strain model, based on the Matthews-Blakeslee theory for heterostructure semiconductor interfaces. CdSe/CdS core/shell nanoparticles were synthesized using similar methods employed in the preparation of CdSe/ZnS, 14 and CdSe/CdS 15 core/shell nanoparticles with some modifications. Usually, these core/shell nanoparticles are synthesized by a two-step synthetic procedure. First, nearly monodispersed core CdSe nanoparticles capped with TOPO are synthesized using reported methods. 2,14-15 Then, these CdSe nanoparticles in TOPO are coated by growth of ZnS or CdS after the addition of Me 2 Zn or Me 2 Cd solutions together with the sulfur precursor TOPS. A crystalline overlayer of the capping semiconductor may be formed on semiconductor nanoparticles by heterogeneous nucleation (precipitation on the core surface). The reaction temperature in 233

267 the second step is usually lower than that in the first step to facilitate the formation of ZnS or CdS layers on the surface of the core CdSe nanoparticle and to prevent Ostwald ripening. In our study, the Se and Cd precursors were mixed at C with an excess amount of Cd precursor. The excess Cd precursor rendered the CdSe nanoparticles to have a Cd 2+ -rich surface, leading to good capping with TOPO and to optically good quality CdSe nanoparticles. 2,15 It also facilitated the sulfur precursor to adsorb and react on the surface to form core/shell structures. A complementary amount of the sulfur stock solution was then added slowly, in order to suppress homogenous nucleation of CdS, forming a CdS layers on the surface of CdSe nanoparticles. To prepare specific thicknesses of the shell layer, the method established by Peng et al. 15 was used. Since TEM does not give reliable data below 0.3 nm, the size was rather determined by the amount of the added capping material. 15 Here, additional amounts of CdS precursor were added to form 1, 2, or 3 monolayers on the surface of the CdSe core nanoparticles. Aliquots are taken at different reaction times and quenched in cold toluene. In the following, the optical response of the synthesized core/shell systems is summarized. The obtained results were consistent with the step-wise addition of CdS layers. Moreover, the optical changes are consistent with previously reported data

268 11.2 Experimental Chemicals Technical grade (90%) trioctylphosphonic oxide (TOPO), sulfur powder (sublimed 99+ %), and Cadmium oxide (99%) were purchased from Strem Chemicals. Selenium powder (99.5%) and n-tetradecylphosphonic acid (TDPA) (98%) were purchased from Alfa Aesar. Trioctylphosphine (TOP) (tech. 90 %) was purchased from Aldrich. All the chemicals were used as purchased. Preparation of CdSe/CdS core/shell Nanoparticles For a typical synthesis, g cadmium oxide, g TDPA and g TOPO were loaded in a threenecked flask and heated up to 240 C for 2 hours under N 2 flow in a Schlenk line. Hot selenium stock solution ( g Se dissolved in g TOP) was then quickly injected. After injection, the temperature of the mixed solution dropped to 200 C, so the solution was quickly heated to ~ C and allowed to react for 8 minutes. The resulting solution was cooled down to 180 C within the next 7 minutes. A hot sulfur stock solution (0.0084g S dissolved in g TOP) was slowly injected and reacted at 170 C for 6 hours. Fractions of the samples were taken out during the reaction and quenched in toluene. Characterization. Absorption spectra were obtained using a Cary 50 Bio UV- Visible spectrophotometer while steady-state PL experiments were conducted on a Cary Eclipse PL Spectrophotometer. The PL lifetimes were measured on a nanosecond laser system (Spectra Physics GRC 230 with a MOPO 730) equipped with a Chromex Monochromator attached to a Hamamatsu R928 photomultiplier, which is read out with a LeCroy LT 342 oscilloscope. 235

269 11.3 Results Figure 11.1 shows the UV/visible absorption spectra of the CdSe/CdS core/shell nanoparticles at various stages of the overgrowth process, which are consistent with previous reports The observed band-edge absorption of CdSe occurrs at nm, indicating a size of about 4.0 nm in diameter from optical and TEM calibrations. 2,16 After the S solution was injected, CdS formed on the surface of the CdSe nanoparticles to form the core/shell structured nanoparticles. The band-edge absorption of these core/shell nanoparticles showed within the first 30 min a red-shift from to nm as the reaction proceeded. Following this, the absorption red-shift slowed down and converged at almost nm. This is attributed to the growth process of CdS layers on the surface of the CdSe nanoparticles. At early times, there is a higher concentration of available Cd 0.3 capping time Absorbance min 180min 90min 30min 10min 5min 0min Wavelength / nm Figure 11.1 Absorption spectra of CdSe/CdS core/shell nanoparticles with CdS capping of different thickness after the indicated capping times. For clarity, the spectra are offset by 0.01 absorbance units, each. The red-shift of the absorbance maximum is caused by the increasing shell thickness. 236

270 and S in the solution and CdS layers are formed rapidly. As the available Cd and S are depleted, the growth of the CdS layers slows down. Compared to the absorption spectrum of the CdSe core nanoparticles, the spectra of the CdSe/CdS core/shell system red-shifted by approximately 5 nm during the course of the capping process. This phenomenon can be explained with the growing nanoparticle size, and mixing of the core and shell LUMOs in a molecular orbital model. 15 Alternatively, the red-shift of the core/shell system can also be explained by an electronic confinement effect, which occurs in the core/shell and host/guest-type materials. 17 CdSe/CdS core/shell nanostructures can be treated as CdSe nanoparticles confined in a CdS matrix, analogous to a guest molecule confined in a CdS host nanocavity. When a CdSe nanoparticle is confined in a CdS nanocavity, the corresponding Fermi level is raised while the band gap decreases. Coherency strain will also lead to a red-shift, as discussed later. Because CdS has a larger bulk band-gap energy of 2.5 ev ( 490 nm) than that of CdSe (1.7 ev 725 nm), the absorption of CdS nanoparticles can be concealed by the absorption of CdSe nanoparticles. Using information from UV/visible absorption spectra solely cannot verify if there are separate CdS nanoparticles formed in the solution instead of on the surface of the CdSe core. However, PL spectroscopy can help to clarify this. If the sample contains isolated CdS nanoparticles, the PL spectra of CdS should be obtained after it is excited above the band-gap energy even when energy transfer to CdSe takes place. However, there was no emission observed below 500 nm when the samples were excited at 300 nm, providing further evidence that clean samples of core-shell nanoparticles were obtained. 237

271 Normalized PL a Energy (ev) min 5 min 10 min 30 min 90 min 180 min 360 min Quantum Yield / % ~1ML ~2ML b ~3ML Wavelength / nm Capping Time / min Figure 11.2 a) Steady state PL spectra of CdSe/CdS core/shell nanoparticles with different shell thicknesses (ML: monolayer) after the indicated capping times. Excitation wavelength is 400 nm. b) PL quantum yield of CdSe/CdS core/shell nanoparticles as the synthesis time for the shell progresses. The quantum yield of nanocrystal solutions was calculated by comparing the integrated emission to that of Rhodamine 6G in ethanol, with an excitation wavelength of 490nm. The PL spectra of CdSe/CdS nanoparticles are presented in Figure 11.2a. All the samples had an optical density OD = 0.25 at the band-edge absorption maximum around nm and were excited with 400 nm light. The CdSe nanoparticle and CdSe/CdS core-shell nanoparticles have good optical qualities and no emission from deep trap states at wavelengths longer than 640 nm was detected. The full width at half maxima (FWHM) of the PL of these samples was about 30 nm. Interestingly, the similar FWHMs of the emission spectra of all the samples suggests that the CdSe/CdS core/shell nanoparticles have a similar size distribution as that of the CdSe core, implying that Ostwald ripening of CdSe cores was effectively suppressed during growth of the CdS layers. The observed Stokes shift (the difference between PL maximum and lowest-energy transition of the 238

272 absorption spectrum) is around 16.0 nm with small fluctuations. These fluctuations are the result of subtle changes of the emitting state. 2,11,18-20 The PL yield changed drastically during the synthesis time as shown in Figure 11.2b. The PL yield first increased from 12% to 23% (relative to Rhodamine 6G) then decreased to about ~ 11%. In addition, the Stokes shift between the peaks of emission and band-edge absorption changed as the capping reaction proceeded. For about one monolayer of CdS, the nanoparticles showed the highest PL yield, suggesting that the optimal shell thickness (without introduction of additional interface defects) is < 2 monolayers. At this stage, the CdS acts as a good inorganic capping material and helps in removing surface defects and the corresponding low-lying electronic states of the CdSe core crystals, thus increasing the PL yield. At the interface, the 3.9% lattice mismatch causes strain (coherency strain) in the shell material, that enables the CdS to adapt the parameters of the CdSe core. As the number of CdS layers continues to increase (t > 10 min.), dislocations are formed at the interface to offset some of the coherency strain required to maintain partial lattice matching elsewhere. This brings about defect-related states and lower PL yields The interface can thus drastically affect the optical properties of a core/shell nanoparticle. This coherency strain effect can be verified by the measured changes of the PL quantum yield and the PL lifetimes of the CdSe/CdS nanoparticles. Due to the diversity of the nature of the defects, it would be a difficult task to derive a formula directly relating the coherency strain and the measured photophysical parameters, although correlations were previously indicated However, the PL lifetimes of the CdSe/CdS core/shell structured nanoparticles can be conveniently measured on a 239

273 nanosecond time scale and are summarized in Figures 11.3, where all the samples were excited at 460 nm. counts (normalized) min [core] 5 min [~1 ML] 30 min 90 min [~2 ML] 180 min 360 min [~3 ML] a Photoluminescence Lifetime (ns) ~ 1 ML ~ 2 ML τ 2 τ 1 ~ 3 ML b Time / ns Capping time / min Figure 11.3 a) PL lifetimes of CdSe (0 min) and CdSe/CdS nanoparticles (5 min to 360 min capping time) recorded at 570 nm. b) PL lifetime components of CdSe/CdS nanoparticles versus capping reaction time. The experimental error is < 5%. For the CdSe nanoparticles, the PL decayed with a lifetime of 52.2 ns. For the CdSe/CdS core/shell nanoparticles, the PL decays were fitted best biexponentially, with a fast component of about 25.0 ns and a dominant slow component of about ns. For example, for the core-shell nanoparticles of 5 min. and 10 min. capping time, the first component is a fast decay with a lifetime of ~25 ns, while the second component is a slower decay with a lifetime of ~75 ns. Both lifetime components reach a maximum for the sample with 90 min. capping reaction time and level off for samples with longer capping reaction times. On the other hand, the quantum yield reaches a maximum at very early capping times (5 min capping time, ~ 1 monolayer). Thereafter it drops sharply to much lower values. From a comparison of Figures 11.2 and 11.3, it becomes evident, that the highest quantum yields do not correlate with the longest PL lifetimes. The increase in 240

274 PL lifetimes occurred while the PL quantum yield decreased to a minimum, suggesting that the long PL lifetimes are not strictly from band edge luminescence, but originate from shallow trap states and the localization of electrons and holes 23 in interface dislocations (near band gap PL). While the band edge depopulation is accelerated by additional defect states, the radiative charge carrier recombination time (the observed PL lifetime) from these localized states is long-lived. ~1ML ~2 ML ~3 ML λ Abs / nm a λ PL / nm 568 b 564 Stokes shift / nm Capping time / min c Figure 11.4 Comparison of shifts in the absorption maximum (a) shifts in emission maximum (b) and Stokes shifts (c) of CdSe/CdS core/shell nanoparticles obtained at different capping reaction corresponding to different capping layer thicknesses. The experimental errors for the absorption and emission experiments are, respectively, 0.2 nm, and 0.4 nm for the derived Stokes shift. The values of the absorption and PL maximum, as well as the Stokes shift, are presented in Figure 11.4 as a function of increasing capping time. The Stokes shift in 241

275 figure 4c shows similar changes as the PL yield (Figure 11.2) and PL lifetime (Figure 11.3). An increased Stokes shift is evidence for defect-related carrier trapping. Such an increased Stokes shift is observed at 5 min. reaction time, after addition of ~ 1 monolayer shell material, and decreases (probably up to the point when the first strain-induced defects appear) during the following 30 min. From then on, the Stokes shift increases for larger thicknesses of shell material. A plausible explanation for the changes in optical properties with shell thickness is that for CdSe capped with less than one complete layer CdS, the CdS adopts the lattice parameter of the CdSe core (coherency strain) as shown in Figure The thin CdS capping layer removes the original surface defect states of the CdSe core and therefore enhances the PL yield (fig. 4, t = 5 min). It also could lead to a compression of the core, as indicated in Figure With the continuous growth of CdS on the surface of the CdSe core, the CdS capping layers readjust to the parameters of bulk CdS and induce lattice dislocations (semi-coherency) at the interface. This allows relief of some of the accumulated strain (Figure 11.5) Discussion Strain/Stress Effects The red-shift of the CdSe/CdS core/shell nanoparticles can also be explained by the electronic structure distortion of the CdSe core by the developed electric field at the CdS interface. As discussed in the results part, the stress induced by the lattice mismatch of CdSe and CdS results in an asymmetric internal electric field across the interface, which affects the electronic states, thus the absorption and emission properties of the CdSe nanoparticle. 242

276 b CdSe CdS b: Burgers Vector T: Dislocation Figure 11.5 Two-dimensional sketch of the coherency strain in a 1-monolayer capped core/shell nanoparticle (left) and dislocation formation in a multiplayer core-shell nanoparticle (right). There is to our knowledge not yet a developed theory for defect generation during the core-shell formation of nanoparticles. On the other hand, the Matthews-Blakeslee theory 25 describes the formation of strain-induced misfit dislocations on planar semiconductor interfaces, which could serve as a first-order approximation towards an upper-value estimation of the critical layer thickness h c for a dislocation-free and coherent interface 25,26 : h c 2 b 1 ν cos = 8πε m cosλ 1+ ν β αh ln b c (11.1) where b is the magnitude of the Burgers vector b (a vector indicating the direction and magnitude of dislocation); β is the angle between the misfit dislocation line and the Burgers vector b; v is the Poisson ratio (the ratio of transverse contraction to longitudinal extension in the direction of a stretching force in a simple tension experiment) of the bulk 243

277 crystal, assuming that v of the shell material is similar to that of the core; ε m is the misfit of the two crystals across the interface, defined as ε m = (d b -d a )/d a, d a and d b are the interplane spacing of the core and shell material, respectively; λ is the angle between Burgers vector b and the direction orthogonal to the dislocation line and the interface. Applying this model leads to a good agreement with the experimental data. A critical thickness of < 1 nm is obtained for a 3.9% lattice mismatch at the interface, 27 which equates to < 2 epitaxial layers of CdS shell on a CdSe core. This also explains the rapid degradation of the optical quality as soon as one passivation layer is completed. Beyond this limit, strain is partially relaxed through generation of misfit dislocations. These dislocations relieve some, but not all, strain, which induces as a result an inhomogeneous strain field. 26 With increasing thickness of the shell material, all the coherency stress will be relieved. This will lead to the observed the PL quantum yield changes of the first dramatic increase and then slow decrease until to a constant. This leads to trapped charge carriers, low quantum yields, long PL lifetimes (trapped carrier recombination), and enlarged Stokes shifts. It should to be pointed out, however, that the provided analytical form for critical thickness describes a thin-film/semi-infinite substrate configuration, for which the substrate remains unstrained. As pointed out above, the coherency strain also leads to changes in the electronic structure. 26 For direct band gap materials, such as CdSe, the deformation potential of the conduction band is larger then that of the valence band (because it is inversely proportional to the density of states) and the changes in band gap are negative. Thus, under a shell-induced strain, the band gap of the core material should decrease and a red shift of absorption and emission should occur. This is consistent with the data presented 244

278 in Figure 11.1 and 11.2 and leads to the following mechanistic picture about the photophysics of core/shell hetero-nanostructures. When core/shell nanoparticles with a shell thickness > h c are excited, the photoinduced charge carriers migrate to the interface of the CdSe and CdS and are trapped in the misfit dislocations. Thus, the nature and the number of trapping sites directly affects the optical quality of the nanoparticles and the thickness of the shell layer is therefore critical for the optical response. It should be noted at this point that previous HRTEM work 15 suggests that such core-shell systems grow (at least to a large part) epitaxially. Time-resolved spectroscopy, on the other hand, is a very sensitive tool to specifically detect the dynamics of carriers due to trapping processes. Thus, by means of time-resolved laser spectroscopy, carrier trapping is easily observed in cases of heteronanostructured materials. A reasoning for the apparent difference is that electron microscopy is selective for individual particles or small assemblies. Laser spectroscopy on nanoparticle solutions is, on the other hand, an ensemble technique and measures the averaged response of >10 6 nanoparticles Quantum confinement Compared to absorption and emission spectra of the CdSe core nanoparticles, the spectra of the CdSe/CdS core/shell system red-shifted by approximately 5 nm during the course of the capping reaction. This phenomenon can be explained with a particle-in-abox model 28 and the high degree of mixing of the core and shell LUMOs in a molecular orbital model. There are several more aspects to consider. When the size of the material 245

279 E Large potential well Nanosize core Size effect Potential well effect Figure 11.6 Illustration of the quantum confinement of core/shell nanoparticle systems. is similar to or less than that of the exciton Bohr radius, the continuous electronic states become discrete and the energy difference between the electronic states increases. This effect is well known as the size confinement or the quantum confinement effect and has been well studied theoretically and experimentally. The observed red-shift of the absorption and emission of CdSe/CdS core/shell system could be explained by using the same physical principles. When a CdS shell is formed on the surface of the CdSe core, the total size of the CdSe/CdS is larger than that of the core. This allows a further delocalization of the electronic wavefunction and creates a red-shift of the band edge transition (size effect). On the other hand, because the bandgap of the CdS is larger than that of the CdSe, the potential well of the CdSe/CdS nanoparticle becomes deeper than that of the pure CdSe nanoparticle. This increased potential well can lead to an increase of the transition energies of the CdSe core and blue-shifts the absorption and emission 246

280 spectra of the CdSe/CdS systems (potential well effect). As illustrated in Fig. 11.6, the observed red-shift of the absorption and emission spectra could be explained as the overall result of the size effect overlapping with the potential-well effect. CBE E f + χβ - χβ Δα CBE -χβ = χβ Δβ E f -χβ = χβ Δβ VBE VBE Nanoparticle Nanoparticle confined in a quantum shell Figure 11.7 Qualitative description of the frontier orbital energy levels of a nanoparticle confined in a quantum shell Electronic Confinement The observed red-shift is also caused by another confinement effect, which becomes more pronounced for confined entities, the electronic confinement. The electronic confinement is illustrated in Scheme Fig CdSe/CdS core/shell nanostructures can be treated as CdSe nanoparticles confined in a CdS shell. This structure is analogous to a conjugated molecule confined in a nanocavity, a common theme in host-guest chemistry. 29 When a nanoparticle is confined in a nanocavity, the corresponding fermi level, the valence band edge (VBE) and the conduction band edge (CBE) are lifted up, while the energy difference between the VBE and CBE decreases. This effect can induce the observed red-shift of the absorption and emission spectra. 247

281 11.5 Conclusions In conclusion, CdSe/CdS core/shell nanoparticles were synthesized with a one-pot synthesis and the thickness of the shell layer was thereby altered in situ. The core/shellstructured nanoparticles showed red-shifted absorption and emission spectra as a CdS shell formed on the CdSe core. The red-shift of the absorption and emission spectra of CdSe/CdS core/shell nanoparticle was explained by the phenomena of quantum confinement and coherency strain effects. Optical spectroscopy supports the model derived from the Matthews-Blakeslee theory, that passivation and epitaxial overgrowth of nanostructures is successful and leads to a defect-free shell, that increases the optical quality of the system, up to a critical shell thickness h c, which is < 2 monolayers in the case of CdSe/CdS. Beyond that thickness, further capping leads to dislocation formation accompanied by a decrease in PL quantum yield, carrier trapping (which can lead to longer emission lifetimes), and an increase in PL Stokes shift. This model rationalizes and agrees with the spectroscopic observations made References 1. Alivisatos, A. P. J. Phys. Chem. 1996, 100, Murray, C. B.; Norris, D. J.; Bawendi, M. G. J. Am. Chem. Soc. 1993, 115, Nirmal, M.; Brus, L. Acc. Chem. Res. 1999, 32, Greenham, N. C.; Peng, X.; Alivisatos, A. P. Phys. Rev. B 1996, 54, Logunov, S.; Green. T.; Marguet, S.; El-Sayed, M. A. J. Phys. Chem. A 1998, 102,

282 6. Landes, C.; Burda, C.; Braun, M. and El-Sayed, M. A. J. Phys. Chem. B 2001, 105, Burda, C.; Link, S.; Mohamed, M.; and El-Sayed, M. J. Phys. Chem. B 2001, 105, Klimov, V. I.; McBranch, D. W.; Leatherdale, C. A.; Bawendi, M. G. Phys. Rev. B. 1999, 60, Zhang J. Z. Acc. Chem. Res. 1997, 30, Weller, H.; Eychmüller, A. In Advances in Photochemistry; John Wiley and Sons: New York, 1995; Vol. 20, pp Nirmal, M.; Norris, D. J.; Kuno, M.; Bawendi, M. G.; Efros, A. L.; Rosen, M. Phys. Rev. Lett. 1995, 75, Efros, A. L.; Rosen, M.; Kuno, M.; Nirmal, M.; Norris, D. J.; Bawendi, M. G. Phys. Rev. B 1996, 54, Klimov, V. I.; McBranch, D. W. Phys. Rev. Lett. 1998, 80, Hines, M. A.; Guyot-Sionnest, P. J. Phys. Chem. B 1996, 100, Peng, X.; Schlamp, M. C.; Kadavanich, A. V. et al. J. Am. Chem. Soc. 1997, 119, Kuno, M. K. MIT, PhD thesis, Zhang, L. Z.; Cheng, P. Wuji Huaxue Xuebao 2003, 19,

283 18. Strassburg, M.; Dworzak, M.; Heitz, R. et. al Mater. Sci. Engi. 2002, Türck, V.; Rodt, S.; Stier, O.; et. al J. Lumi. 2000, 87-89, Türck, V.; Rodt, S.; Heitz, R.; et. al Physica E 2002, 13, Zhang, J.; Wang, X.; Xiao, M.; Qu, L.; Peng, X. Appl. Phys. Lett. 2002, 81, Rabani, E. J. Chem. Phys. 2001, 115, Nirmal, M.; Murray, C. B.; Bawendi, M. G. Phys. Rev. B 1994, 50, Melo, O. D.; Vargas-Hernández, C.; Hernández-Calderón, I. Appl. Phys. Lett. 2003, 82, Matthews, J. W.; Blakeslee, A. E. J. Cryst. Growth, 1974, 27, Romanov, A. E.; Pompe, W.; Mathis, S., et. al J. Appl. Phys. 1999, 85, Pearsall, T. P. Mat. Res. Soc. Symp. Proc. 1996, 405, Peng, X., Schlamp, M.C., Kadavanich, A.V. et al. J. Am. Chem. Soc., 1997, 119, Zhang, L. Z.; Cheng, P. Wuji Huaxue Xuebao, 2003, 19,

284 CHAPTER 12 Femtosecond Carrier Relaxation and Time-Resolved Temperature Profiles of Semiconductor Quantum Dot Ensembles Abstract: Hot carrier energy distributions in a CdSe quantum dot (QD) ensemble are recorded and analyzed to give their temperature relaxation with femtosecond time resolution. The relative number of excited electrons and holes over an energy window from ev is obtained by comparing the chirp-free femtosecond transient differential absorption spectra with the ground-state oscillator strength. The resulting temperature relaxation is analyzed with various models. A multi-exponential analysis reveals three well-separated time regimes of 450 fs, 35 ps, and 2 ns, which are consistent with the time scales of carrier-carrier, carrier-phonon, and phonon-phonon interactions, respectively. Empirical agreement with a percolation model suggests that large regions of correlated nanoparticles relax faster than small regions, so that the cooling process can be modified by the correlation range of hot carriers in the nanoparticle system. Experimental access to fs-time resolved hot carrier distribution functions and carrier temperatures facilitates the design and application of QD-based devices. 251

285 12.1 Introduction Heat dissipation is now a major concern in designing electronic devices. 1-2 Most of the desired characteristics of electronic components deteriorate with increasing temperature. 3-4 A better understanding and control of heat dissipation in devices becomes essential as the size of electronic building blocks is reduced towards nanometer dimensions, particularly in QD-based systems like sensor, 5 diode, 6 laser, 7-8 transistor, 9 and communication devices The development of optical gain in chemically synthesized semiconductor QDs has been recently studied as the first step towards QD lasers. 13 The dynamics of hot charge carrier relaxation have been studied for bulk, quantum-well, and QD semiconductor systems Unlike the parabolic and steplike increase in density of electronic states in bulk and quantum wells, respectively, the electronic spectrum of a QD consists of well-separated atom-like states, often modified with delta functions, with an energy spacing that increases as the dot size is reduced Reports on carrier relaxation in QDs have shown that phonon bottleneck effects exist in QDs on substrates grown in the Stranski-Krastanow mode, where the cooling of electrons or holes is slowed down to the hundred picosecond time scale, 18 while a phonon bottleneck could be circumvented in colloidal QDs by electron-hole Auger processes. 13 When a semiconductor QD is excited with an energy larger than the bandgap, excited electrons and holes are produced with excess energies Δ 1 Ee = ( hν E g )(1 + me / mh ) (12.1) Δ E = ( hν E ) ΔE (12.2) h g e 252

286 m e m h Here and are the effective masses of the electron and hole, respectively, ΔE e ( ΔE ) is the energy difference between the conduction (valence) band edge and the h initial energy of the photogenerated electron (hole). 14 These excess energies add to the kinetic energies of the respective charge carriers. Initially the electrons and holes are far from thermal equilibrium, so their energies are far from the Fermi-Dirac distribution determined by the temperature of the system. If photon absorption produces electrons and holes with initial excess kinetic energies above the conduction and band edges, then both initial carrier temperatures are above the lattice temperature; these carriers are called hot carriers (i.e. hot electrons and hot holes). 14 After sufficient thermalization time, the electron and hole pair can be treated as a single quasiparticle, with a single carrier temperature T c. 15 The initial energy distribution of the photoexcited carriers is determined by the spectral width of the laser and the dispersions of the conduction and the valence bands. Immediately following photoexcitation by an ultrashort laser pulse, the photopopulated electronic states still retain their coherence. The dephasing time (τ c ), is usually very fast and after loss of coherency, the carrier distribution function (DF) is nonthermal. And via carrier-carrier scattering the initial non-thermal distribution converts into a thermalized Fermi-Dirac type distribution with characteristic temperatures T e and T h for electrons and holes, different from the lattice temperature T L. Electrons and holes then thermalize with each other usually in the regime of less than 100 fs by electron-hole collisions to a single temperature T c higher than T L. 14,15 In the following hundreds of femtoseconds to picoseconds, the cooling of the carriers to the lattice temperature occurs by carrier/phonon-coupling, mainly through the interaction of the high-energy tail of the distribution with optical phonons. 14,15 If the carriers have sufficient lifetime, the 253

287 photoexcited carriers will cool until they have the same temperature as the lattice. The carriers are then able to either relax in more localized surface states or directly recombine to return the semiconductor quantum dot back to thermal equilibrium conditions. Although femtosecond transient differential absorption spectroscopy (TDA) provides an excellent means to probe the excited-state electron-hole relaxation in semiconductor quantum dots, deducing their energy distribution and resulting carrier temperature with comparable time resolution is an experimental challenge. Here, we present a method to observe the hot carrier energy distribution in a semiconductor QD assemble with fs time resolution. Chirp-free fs pump-probe spectroscopy is used to investigate the electron-hole cooling processes in CdSe QDs. The time-resolved energy distribution of hot carriers in CdSe QDs after electronic excitation is obtained from the experimental TDA data matrix (energy (ε i ) as a function of time (t j )), by normalizing with the ground-state oscillator strength (spectral vector α(ε i )). This method provides a technique to characterize the fundamental cooling and heat dissipation processes in QDbased systems on ultrafast time scales, which is also useful in the practical design of QDbased electronics Experimental Sample Preparation Samples of CdSe QDs were synthesized similar to the method described previously. 22 Technical grade (90%) trioctylphosphonic oxide (TOPO), and Cadmium oxide (99%) were purchased from Strem Chemicals. Selenium powder (99.5%) and n-tetradecylphosphonic acid (TDPA) (98%) were purchased from Alfa 254

288 Aesar. Trioctylphosphine (TOP) (tech. 90 %) was purchased from Aldrich. All chemicals were used as purchased. For a typical synthesis, g cadmium oxide, g TDPA and g TOPO were loaded in a three-necked flask and heated up to 240 C for 2 hours under N 2 flow in a Schlenk line. Hot selenium stock solution ( g Se dissolved in g TOP) was then quickly injected. After the injection, the temperature of the mixed solution dropped to 200 C, so the solution was quickly heated to ~ C and allowed to react for 8 minutes. The resulting solution was cooled down to room temperature. Fractions of the samples were taken out during the reaction and quenched in toluene. UV-visible absorption and PL spectroscopy Absorption spectra were obtained at room temperature using a Cary 50 Bio UV-Visible spectrophotometer while steadystate PL experiments were conducted on a Cary Eclipse PL Spectrophotometer. Femtosecond TDA Spectroscopy The femtosecond laser pump-probe system consists of an amplified erbium-doped fiber laser, which is frequency doubled to 780 nm and amplified in a regenerative amplifier (Clark MXR CPA 2001). The system produces pulses with 120 fs FWHM duration and 800 µj output energy per pulse at a repetition rate of 1 khz. The excitation beam is frequency doubled by a 0.2 mm BBO crystal, and then modulated by a 100 Hz frequency chopper. The continuum probe of white light comes from a small part (4%) of the fundamental beam that is split off and passed through a 2 mm sapphire. Reflective optics is utilized for the probe beam to avoid white light dispersion. The excitation beam is focused to a spot diameter of about 500 µm and the probe beam to 100 µm. The pump-probe measurements were all carried out with the sample bath at ambient temperature. The QD solution was placed in a quartz cuvette with 255

289 a 2 mm path length, which was continuously stirred by a cell stirrer to avoid permanent bleaching of the pump-probe volume element in the solution during measurement. Data acquisition, involving Labview software, gave 2-dimensional matrices of wavelength versus delay time. Unlike kinetic analysis at single wavelengths, the matrices were analyzed by the single-value decomposition method, 17 resulting in a global analysis of the spectrum vs. time matrix Results Shown in Fig is the UV-visible absorption spectrum and the photoluminescence (PL) of a CdSe QD sample prepared in our laboratory. The numbers identify energies associated with particular electron-hole pair states. 23,24 The observed Optical Density (a.u. ) E 2s E 1p E 1s s 3/2 (h)1s(e) 2 2s 3/2 (h)1s(e) 3 1s 1/2 (h)1s(e) 4 1p 3/2 (h)1p(e) 5 2s 1/2 (h)1s(e) 6 1p 1/2 (h)1p(e) 7 2s 3/2 (h)2s(e) Energy (ev) Figure 12.1 UV-visible absorption (blue) and PL (red dashed, arbitrary scale) spectra of CdSe QDs indicating a high-quality quantum dot sample with no detectable low-energy surface states. The numbered markers refer to energies associated with the electron-hole pair states shown. 256

290 bandgap absorption of the QDs occurs at 2.28 ev, indicating an average particle diameter of 4.0 nm, as confirmed by transmission electron microscopy (TEM) measurements. 21,23,24 The FWHM of the PL is consistent with a narrow size distribution of QDs, as confirmed by the TEM studies. The high optical quality of the prepared CdSe QDs is evidenced by the lack of emission from trap states at energies lower than 1.94 ev. Delay Time (ps) Δ (OD) - Δ (OD) Δ (OD) Delay Time / ps Energy ( ev) Delay Time (ps) A B Figure 12.2 (A) Contour plot of the experimental femtosecond TDA spectra ( ev, nm) measured with 100 fs time resolution in a 2.9 ns time window. (B) The relaxation dynamics recorded at different electron-hole pair energies: (from bottom to top) (+) 2s 1/2 (h)1s(e) (2.76 ev); (X)1p 3/2 (h)1p(e) (2.62 ev); (Δ)1s 1/2 (h)1s(e) (2.54 ev); ( ) 2s 3/2 (h)1s(e) (2.39 ev); (Ο) 1s 3/2 (h)1s(e) (2.30 ev); and ( ) 1s 3/2 (h)1s(e) (2.28 ev). The inset expands the first three kinetic traces at a higher time resolution. The experimental TDA matrix, measured at room temperature after a pump intensity of 0.8 μj per laser pulse, is presented as a contour plot in Fig. 12.2A. It is important to note that this pump energy corresponds to an average number of electronhole pairs per QD of N eh 0.5. Thus, most QDs had at most one carrier pair, so that the 257

291 quantum dot ensemble is equilibrated primarily via coupling to the solvent bath through the Auger process where the hot electrons cool down by transferring the extra energy to the coupled hole, which is excited to the solvent continuum state. 13 The carrier relaxation dynamics were recorded at different transition energies and are presented in Fig. 12.2B. In the time frame of the measurement, the transients exhibited three exponential components: a first component with a lifetime < 1 ps, followed by a second component in the time scale of ~ 40 ps, and a third component in the nanosecond time regime. The highest energy states clearly depopulate faster than the lowest ones, but not much difference in lifetime is observed for the 1p 3/2 (h)1p(e), 1s 1/2 (h)1s(e) and 2s 3/2 (h)1s(e) states Discussion Femtosecond TDA spectroscopy provides an excellent means to investigate the electron-hole cooling processes in semiconductor QDs by measuring the carrier energy distribution spectrum and the dynamics of carrier cooling At low excitation densities (N eh < 1), the transient differential absorption (TDA) spectrum can be expressed by TDA (hv) Δα (hv) = -α o (hv)(f e + f h ) (12.3) where f e and f h are the electron and hole energy Fermi-Dirac distribution functions, respectively, at the wave vector appropriate for the photon energy hv (direct transitions), and α o is the absorption coefficient without excitations. 13,15 Thus, the measurement of the transient absorption spectrum allows the determination of the sum of excited carrier distributions. Most experiments have been analyzed using this low carrier-density 258

292 description, where many-body effects can be neglected. 13,15 It has been reported that for CdSe QDs, the spectral dynamics of the bleach in the TDA is strongly dominated by the electrons. 13,17 Thus relaxation of the excited carriers can be simplified to relaxation of the hot electron. This simplifies Eq. (12.3) to TDA (hv) Δα (hv) -α o (hv)f e (12.4) The thermal population of electronic states has a negative-exponential dependence on the energy, i.e. n(ε i ) exp(-ε i /kt) and therefore the relative population is n rel = n(ε i )/n(ε 0 ) = exp(-δε i /kt), where Δε = ε i - ε 0 with energy ε 0 set to the lowest excited state energy ε gap. The TDA spectra were converted to a relative density of states by subtracting the bandedge depopulation time trace to give Δα(t j ), normalized with the ground-state absorption spectrum (α o (ε i )), and shifting the energy baseline to coincide with the lowest excited state energy to yield the relative population matrix n rel = n(ε i, t j )/n(ε gap, t j ). The results are shown in Figure The n rel (ε i,t j ) matrix provides the time- and energy-resolved carrier distribution maps of the QD ensemble. The spectra at different delay times t j display the relative carrier population, from which one can directly extract the electronic temperature of the QD ensemble with a ~ 100 fs time resolution. The energy- and time-resolved carrier distribution from 2.28 ev (band edge) to 2.76 ev is plotted as a contour plot in Figure 12.3A,where the carrier temperatures at different delay times can be extracted. The timeresolved carrier distributions and their fits to the Fermi-Dirac distribution function at different delay times are shown in Figure 12.3B. As can be seen from Figure 12.3B, upon laser excitation, the quantum dots initially exhibit an inverted carrier population (<

293 Delay Time (ps) Energy (ev) 2.76 nrel. n rel ps 0.21ps 0.32ps 0.43ps 0.64ps 1.07ps 2.56ps 10.83ps 75.33ps ps Carrier Energy / ev A B Figure 12.3 (A) Contour plot of the time-resolved n rel matrix in the energy range of ev, and a time window of 3 nanoseconds. (B) n rel (open circles) fitted by the Fermi-Dirac function from equation (4) (solid lines) at different delay times during the first 500 ps. For clarity, each curve is offset by fs), where the high-energy excited electronic states (> 2.55 ev) are more highly occupied than the lower ones (< 2.55 ev), but they quickly thermalize (< 640 fs) to excited states that can be described by an effective carrier temperature, T c. Indeed, an effective carrier temperature is appropriate to describe the electron energy distribution for most of the QD ensemble over the entire time range. After the thermalization within hot electrons/holes (> 640 fs), the cooling dynamics are expected to couple with phonons and other radiative and non-radiative band edge depopulation pathways to eventually cool down, as demonstrated with time-resolved laser spectroscopic studies We observe that the higher excited states 1p 3/2 (h)1p(e) and 2s 1/2 (h)1s(e) are more highly occupied during the initial 200 fs after the pump pulse, and that the population inversion in the conduction 260

294 band can be directly observed at this early stage. The depopulation of the higher energy levels occurred within the following 700 fs. The carrier temperatures were determined by fitting the obtained relative population distribution n rel spectra with the Fermi-Dirac function multiplied by the density of state distribution function using Eq. (12.5). as shown below: ( E Ef )/ kt i ( E Ef )/ kt ε e i 2 ( E εi ) 2 2ω 1 1 N( E) G( ε ) = ( Ae ) e i (12.5) ε i Here G(ε i ) is the Gaussian function at energy level ε i describing the homogeneous and inhomogeneous broadening due to phonon coupling and the size distribution of the QD ensemble. For simplicity the fitting was limited to the energy region < 2.6 ev. The sum of Gaussian functions describes the electron density in terms of the discrete energy levels of CdSe QDs. The temperature can be safely introduced after 640 fs in the whole spectrum region from 2.28 to 2.76 ev, where the relative population displays a Fermi- Dirac shape. For simplification, the fitting was limited to the energy region < 2.6 ev, where only thermalization to the lowest excited state was considered. And the temperature can be introduced approximately after 110 fs. The carrier distribution function is initially non-thermal at high energies, but the low-energy electrons and holes equilibrate to a Fermi-Dirac distribution within the ~100 fs time resolution of the measurement. During the first picosecond the energy in the carrier system is redistributed, presumably as a result of electron/hole scattering, 13,25,26 Thereafter, the carrier temperature T c is obtained by treating the electrons and holes as mutually thermalized.. The relative population distributions n rel, Figure 12.3B, could be 261

295 fitted best with Gaussian transition bands of h ω = 26 mev band width in Eq. (12.5). We note that 26 mev matches well with the longitudinal optical phonon energy of bulk CdSe, which indicates that carrier-phonon cooling occurs through phonon emission. The cooling process is characterized by narrowing of the energy distribution and red shift of the distribution peak as time progresses. f(e) ps t 0.21ps 0.32ps 0.43ps Energy (ev) A Carrier Temperature (K) B Stretched Exponential Multi Exponential Percolation Time (ps) Figure 12.4 (A) Fermi-Dirac distribution functions at different delay times after femtosecond laser pulse excitation. (B) Carrier temperature ( ) as a function of delay time obtained from the distributions in Fig12.4A. The curves come from three different models for net non-exponential relaxation, as described in the text. Figure 12.4A shows the changes in the calculated Fermi-Dirac distribution at different delay times obtained by fitting the energy distribution of the excited carriers with the Fermi-Dirac distribution function for CdSe QDs. These results clarify that during the first hundreds of femtoseconds, states higher than the 1s 3/2 (e)1s(h) have a high probability of being populated, but already after 1 ps delay the population is clearly restricted to the states near the conduction band edge. The depopulation dynamics for higher states can be attributed to the combined effect of the red-shifting and the 262

296 narrowing of carrier energy distribution during cooling. This partly explains the multiexponential depopulation dynamics of the higher states, which is observed even in highquality QD systems. Figure 12.4B shows the carrier temperature versus delay time obtained from the distribution functions in Figure 12.4A Three different functions for net non-exponential relaxation are tested against the data in Fig. 4B. The dashed curve in Figure 12.4B is a fit to the data using the stretched exponential relaxation formula, T () t = e (/ t τ ) Δ T 0 e β + T. The relatively poor fit shown in Fig. 4B indicates that the data do not follow this standard relaxation formula. The dot-dashed curve in Figure 12.4B is a fit to the data using the sum of three exponential relaxation functions yielding three distinct relaxation times, τ i. Under these experimental conditions, the fastest carrier cooling process has a lifetime of τ 1 = 450 fs, which is a typical time scale for electron-hole scattering. 13,15-17 The next cooling component yields τ 2 = 35.7 ps, which is a typical time scale for electron-phonon interactions, consistent with other recent experiments The latter occurs mainly through interaction of the high-energy tail of the carrier distribution with optical phonons. 15 At delay times greater than 100 ps (about an order of magnitude greater than for bulk semiconductors), the drastic slowing down of the carriers gives τ 3 ~ 2 ns, characteristic of a phonon bottleneck. 13 This phonon-phonon component is not observable in the data shown in Figure 12.4B, but it is indicated in the two nanoseconds time regime of the TDA matrix as shown in Figure 12.2B. If the carriers have sufficient lifetime, the photoexcited carriers will cool until they have the same temperature as the lattice. The carriers are then able to either relax to more localized surface or bulk defect states, or directly recombine to return the QDs back to ground state equilibrium. The 263

297 assignment of these three time regimes to carrier/carrier, carrier/phonon, and phonon/phonon interactions is based on previous reports and is still a subject of investigations The fitted cooling rate components involving phonon interactions are 530K/ps and 200K/ns, respectively. These cooling rates appear fast enough for a reliable cooling of QD-based electronic devices, where operations per second may be envisioned. 30 The solid curve in Figure 12.4B is a fit to the data using a percolation model function. 31 The model is based on the assumption that the sample contains a distribution of domain sizes p x, and that the domains relax with a relaxation rate that varies Cx / exponentially with inverse size w w e. Here the size x is proportional to the x = number of excited electrons in the domain, w is the relaxation rate for asymptotically large domains, and C is a correlation coefficient. Integrating over all possible domain sizes, the effective temperature of the hot electrons as a function of time can be written as wt Te() t = ΔT0 xpx e x dx/ 15. Γ( 19 / 6 ) + T (12.6) 0 We have tried fitting the measured relaxation using several standard size distributions, including a Gaussian and various distributions from percolation theory We find best agreement with the data using a percolation distribution for finite domains in 3-dimensions above the percolation threshold 19 / p x e x x 2/ 3, which is typical of systems with quenched disorder The solid curve in Figure 12.4B is obtained using ΔT 0 = 9300 K, T = 680 K, 1/w = ps, and C =

298 The percolation model describes the behavior of a set of objects (here nanoparticles) with quenched local randomness in their interactions that yields a random distribution of region sizes The randomness can be attributed to interactions between randomly located nanoparticles. Here, the inverse size dependence is due to thermal fluctuations within each region, in contrast to the behavior of inter-particle correlations found for magnetic nanoparticles at much lower temperatures. 36 Agreement with the model suggests that after the initial laser pulse, the hot carriers become correlated into regions of nanoparticles having a percolation distribution of sizes, and that the cooling process requires cooperative activation of the entire region. A key feature of the model is the inverse size dependence in the relaxation rate. Specifically, since we find C>0 from the data here, the model indicates that larger regions relax faster than smaller ones. Similar behavior has been found in previous work, 13,14 and is a general property of slow relaxation in highly correlated systems. This inverse size dependence can be attributed to the fact that large regions have more internal degrees of freedom than small ones, so that large regions can find pathways around an energy barrier, rather than having to activate directly over a barrier as is required for small regions. Agreement with the model indicates that it may be possible to accelerate the cooling process by increasing the size of the correlated regions, thereby increasing the number of energy transfer pathways to cool the carriers. To summarize the results from the fitting functions shown in Fig. 12.4B, the stretched exponential clearly does not give good agreement with the deduced relaxation. The multiple exponential function gives the best fits to the data, and reasonable values for the relaxation times are obtained that are in general agreement with other measurements. 265

299 However, the three exponentials contain a total of seven adjustable parameters, which results in some uncertainty in the parameters. By comparison, the stretched percolation model function has only four adjustable parameters, and may provide a means for obtaining the fundamental distribution of time scales in the QD ensemble Conclusions In conclusion, femtosecond time-resolved TDA spectra provide a tool to measure the experimental hot carrier distribution matrix and the cooling processes in semiconductor nanomaterials. The presented analysis yielding the population matrix n rel (ε i,t j ) provides a more direct and detailed understanding of heat dissipation in electronically excited nanosystems. Hot carrier distributions and temperatures were directly obtained on the fastest possible time scale for thermalized electron/hole-pairs. Three fitting functions: stretched exponential, multiple exponential, and percolation model, are applied to analyze the cooling dynamics of the hot carriers in an ensemble of QD nanoparticles. The measured hot carrier relaxation as characterized by the percolation model suggests that larger regions of nanoparticles relax faster than smaller regions, so that heat dissipation can be tuned by changing the correlation between nanoparticles. This could prove significant for designing high-performance devices in many fields of applied nanoscience and nanotechnology References 1. Starner, T.; Maguire, Y. Mobile Networks and Applications 1999, 4,

300 2. Höhberger, E. M.; Krämer, T.; Wegscheider, W., Blick, R. H. Appl. Phys. Lett. 2003, 82, Kotchetkov, D.; Zou, J.; Balandin, A. A.; Florescu, D. I.; Pollak, F. H. Appl. Phys. Lett. 2001, 79, Chen, G.; Shakouri, A. ASME J. Heat Trans. 2001, 124, Ivanisevic, A.; Yeh, J.; Mawst, L.; Kuech, T. F.; Ellis, A. B. Nature 2001, 409, Zrenner, A.; Beham, E.; Stufler, S.; Findeis, F.; Bichler, M.; Abstreiter, G. Nature 2002, 418, Coe, S.; Woo, W.; Bawendi, M.; Bulović, V. Nature 2002, 420, Fafard, S.; Hinzer, K.; Raymond, S.; Dion, M.; McCaffrey, J.; Feng, Y.; Charbonneau, S. Science 1996, 274, Klein, D. L.; Roth, R.; Lim, A. K. L.; Alivisatos, A. P.; McEuen, P. L. Nature 1997, 389, Herz, L. M.; Phillips, R. T. Nature Mater. 2002, 1, Amlani, I.; Orlov, A. O.; Toth, G.; Bernstein, G. H.; Lent, C. S.; Snider, G. L. Science 1999, 284, Bachtold, A.; Hadley, P.; Nakanishi, T.; Dekker, C. Science 2001, 294, Klimov, V. I. J. Phys. Chem. B 2000, 104, 6112; Klimov, V. I.; Mikhailovsky, A. A.; Xu, S.; Malko, A.; Hollingsworth, J. A.; Leatherdale, C. A.; Eisler, H.-J.; 267

301 Bawendi, M. G. Science 2000, 290, 314; Klimov, V. I.; Schwarz, Ch. J.; McBranch, D. W.; Leatherdale, C. A.; Bawendi, M. G. Phys. Rev. B 1999, 60, R2177. Guyot- Sionnest, P.; Shim, M.; Matranga, C.; Hines, M. Phys. Rev. B 1999, 60, R Nozik, A. J. Annu. Rev. Phys. Chem. 2001, 52, Shah, J. Ultrafast Spectroscopy of Semiconductors and Semiconductor Nanostructures (Springer, New York, 1999); Prabhu, S. S. et al. Phys. Rev. B 1995, 51, Kash, K.; Shah, J.; Block, D.; Gossard, A. C.; Wiegmann, W. Physica B 1985, 134, Burda, C.; Link, S.; Mohamed, M. B.; El-Sayed, M. J. Chem. Phys. 2002, 116, Lou, Y.; Chen, X.; Samia, A. C.; Burda, C. J. Phys. Chem. B 2003, 107, Adler, F.; Geiger, M.; Bauknecht, A.; Haase, D.; Ernst, P.; Dörnen, A.; Scholz, F.; Schweizer, H. J. Appl. Phys. 1998, 83, Prabhu, S. S.; Vengurlekar, A. S.; Roy, S. K.; Shah, J. Phys. Rev. B 1995, 51, Alivisatos, A. P. J. Phys. Chem. 1996, 100, Murray, C. B.; Norris, D. J.; Bawendi, M. G. J. Am. Chem. Soc. 1993, 115, Chen, X.; Lou, Y.; Samia, A. C.; Burda, C. Nano Lett. 2003, 3, Norris, D. J.; Bawendi, M. G. Phys. Rev. B 1996, 53,

302 24. Norris, D. J.; Efros, A. L.; Rosen, M.; Bawendi, M. G. Phys. Rev. B 1996, 53, Immediately following laser excitation, the photopopulated electronic states still retain their coherency. The dephasing time (τ c ) is usually very short (< s, usually several femtoseconds) for quantum well and bulk materials. After loss of coherency the carrier DF is still non-thermal and and can be directly monitored and temporally resolved. 26. The thermalization process within the first picosecond does not change the total carrier temperature via carrier-carrier scattering (but it will change the temperature of the lower energy carriers, whose temperature we obtained here, compared to the higher excited ones). After 1 ps the electrons and holes thermalize to the same temperature, and the cooling process can be monitored through hot electron cooling via state filling and transient bleaching dynamics. 27. Chamberlin, R. V.; Mozurkewich, G.; Orbach, R. Phys. Rev. Lett. 1984, 52, 867. Chamberlin, R. V. Phys. Rev. B 1984, 30, Grassberger, P.; Procassia, I. J. Chem. Phys. 1982, 77, Phillips, J. C. Phys. Rev. B 1995, 52, R Heath, J. R.; Kuekes, P. J.; Snider, G. S.; Williams, R. S. Science 1998, 280, Chamberlin, R. V.; Haines, D. N. Phys. Rev. Lett. 1990, 65, Stauffer, D. Introduction to Percolation Theory (Taylor and Francis, Philadelphia, 1985). 269

303 33. Isichenko, M. B. Rev. Mod. Phys. 1992, 64, Chamberlin, R. V.; Böhmer, R.; Sanchez, E.; Angell, C. A. Phys. Rev. B 1992, 46, Chamberlin, R. V.; Scheinfein, M. R. Ultramicroscopy 1992, 47, Chamberlin, R. V.; Hemberger, J.; Loidl, A.; Humfeld, K. D.; Farrell, D.; Yamanuro, S.; Ijiri, Y.; Majetich, S. A. Phys. Rev. B 2002, 66, /1. 270

304 Appendix Experimental Part 1. Sample Preparation 1.1. Doped TiO 2 Nanoparticles Sol-gel method Doping Afterwards Doped TiO 2 nanoparticles were first prepared by doping TiO 2 gel afterwards using the sol-gel method. For a typical synthesis, small TiO 2 nanocrystals were prepared by the controlled hydrolysis of titanium (IV) isopropoxide in water, under controlled ph. By adjusting the ph of the solution, TiO 2 nanocrystals in the size range of 3 to 10 nm can be synthesized as transparent colloidal solutions, which are stable for extended periods. To introduce nitrogen dopant into the titania nanoparticles, triethylamine is added to the colloidal nanoparticle solution. The addition of amine to the nanoparticle solution results in the formation of yellow nanocrystals (mean diameter of 10 nm). Synthesis of the TiO 2 nanocrystals entails the drop wise addition (1 ml / min) of a 5 ml aliquot of Ti[OCH(CH 3 ) 2 ] 4 (Aldrich, 97%) dissolved in isopropyl alcohol (5:95) to a 900 ml doubly distilled water (2 o C) adjusted to ph 2 by HNO 3 addition. After continuous stirring of the reaction mixture for 12 hours, a colloidal solution of TiO 2 nanocrystals is formed. Using dynamic light scattering (Coutler Plus 4), it has been demonstrated that the size of the nanoparticles depends critically on the amount of HNO 3 added to the reaction mixture. By controlling the amount of added nitric acid, TiO 2 nanoparticles of sizes 271

305 ranging from 3 to 10 nm can be synthesized, which are stable for extended periods under refrigeration. Treatment of the initial nanoparticle solution and gel with an excess of triethylamine results to the formation of a yellowish solution. The solution was treated directly with an alky ammonium compound to facilitate nitrogen incorporation. Upon vacuum drying (5 x 10-2 Torr) for several hours, the treated nanoparticle solution forms deep yellow crystallites. The samples prepared with this method were fully investigated in Chapter 2 and Chapter Doping from the Beginning In another series, Doped TiO 2 nanoparticles were prepared simultaneously with the preparation of TiO 2 gel. TiO 2 nanoparticle was also prepared by adding of a 9.0 ml Ti[OCH(CH 3 ) 2 ] 4 (Aldrich, 97%) to a 150 ml isopropanol containing 5.0 ml NH 3 H 2 O very slowly. After continuous stirring of the reaction mixture for 2 hours, the resultant colloidal solution is boiled for 4 hours at around 85 o C. After removing the solvent by vacuum and drying for several hours, the samples were then calcined at ºC for 3 hours. The samples prepared with this method were fully investigated in Chapter 4 and Chapter Solid-State Method In this method, N, C and S-doped TiO 2 were prepared using solid-state oxidation reaction of TiN, TiC and TiS2 micormeter-sized powder with air at proper temperatures. For example, the experiments for N, C and S-doped TiO 2 nanoparticles were conducted as follows. About 0.5 g TiX (X=N, C, S 2 ) powder was loaded in a ceramic sample boat 272

306 (4.0/l 0.5/h 0.5/w in inch), and then placed in the middle of a quartz tube in a Lindberg tube furnace. The two ends of this tube were left open to the atmosphere. These powders were oxidized at 350ºC-650ºC for 96 hrs. The temperature was slowly ramped up and cooled down at a rate of 2ºC/min during the heating and cooling process. After cooling to room temperature, they were taken out for further investigations. The samples prepared with this method were fully investigated in Chapter 6, Chapter 7 and Chapter CdSe Nanoparticles Growth of CdSe Nanoparticles Technical grade (90%) trioctylphosphonic oxide (TOPO), and Cadmium oxide (99%) were purchased from Strem Chemicals. Selenium powder (99.5%) and n- Tetradecylphosphonic acid (TDPA) (98%) were purchased from Alfa Aesar. Trioctylphosphine (TOP) (tech. 90 %) was purchased from Aldrich. All chemicals were used as purchased. For a typical synthesis, g cadmium oxide, g TDPA and g TOPO were loaded in a three-necked flask and heated up to 240 C for 2 hours under N 2 flow in a Schlenk line. Hot selenium stock solution ( g Se dissolved in g TOP) was then quickly injected. After the injection, the temperature of the mixed solution dropped to 200 C, so the solution was quickly heated to ~ C and allowed to react for 8 minutes. The resulting solution was cooled down to room temperature. Fractions of samples were taken out during the reaction and quenched in toluene. The samples prepared with this method were fully investigated in Chapter Growth of CdSe/CdS Core/shell Nanoparticles 273

307 The synthetic scheme for the growth of CdSe/CdS core/shell nanoparticles is similar to that of the growth of CdSe nanoparticles, except that the CdSe nanoparticles were first growth, and then the elemental sources for the growth of CdS nanoparticles were added during the growth of CdSe nanoparticles. For a typical synthesis, g cadmium oxide, g TDPA and g TOPO were loaded in a three-necked flask and heated up to 240 C for 2 hours under N 2 flow in a Schlenk line. Hot selenium stock solution ( g Se dissolved in g TOP) was then quickly injected. After the injection, the temperature of the mixed solution dropped to 200 C, so the solution was quickly heated to ~ C and allowed to react for 8 minutes. The resulting solution was cooled down to 180 C within the next 7 minutes. A hot sulfur stock solution (0.0084g S dissolved in g TOP) was slowly injected and reacted at 170 C for 6 hours. Fractions of samples were taken out during the reaction and quenched in toluene. The samples prepared with this method were fully investigated in Chapter Growth of small CdSe Nanoparticles The growth process of small CdSe nanoparticles was similar to that of the growth of CdSe nanoparticles, except that small CdSe nanoparticles were grown under a much lower temperature in order to slow down the reaction rate and allow the observation of the crystallization process during the growth. For a typical synthesis of the CdSe NPs, g cadmium stearate (tech. 90%, Strem Chemicals) and g TOPO (tech. 90%, Strem Chemicals) were loaded in a three-neck flask and heated up to 240 C for 2 hours under Ar flow in a Schlenk line. Then hot selenium stock solution ( g Se (99.5%, Alfa Aesar) dissolved in g TPP (99 %, Strem Chemicals)) was quickly 274

308 injected. After the injection, the temperature of the mixed solution dropped to 200 C. This solution was quickly heated to 220 C and reacted at this temperature for 90 minutes. As the reaction progressed, 0.1 ml fractions of the CdSe sample were taken out and quenched in 2 ml cold toluene. The samples prepared with this method were fully investigated in Chapter Growth of small CdSe Nanoclusters The growth process of small CdSe nanoclusters was similar to that of the growth of small CdSe nanoparticles, except that small CdSe nanoclusters were grown under a much lower temperature in order to slow down the reaction rate and the reaction were stopped in the middle of growth to arrest these small CdSe nanoclusters. The CdSe cluster was synthesized by adding g cadmium stearate (90%, Strem Chemicals) and g TOPO (90%, Strem Chemicals) into a three-necked flask and heating up to 200 C for 2 hours under N 2 flow on a Schlenk line. Then hot selenium stock solution ( g Se (99.5%, Alfa Aesar) dissolved in g TOP (90 %, Aldrich)) was quickly injected. After the injection, the temperature of the mixed solution was dropped to 190 C. This solution was quickly heated up to and reacted at 200 C for 3 hours. The solution was taken out and quenched in cold toluene. The samples prepared with this method were fully investigated in Chapter 9. 2 Characterization 2.1 X-ray diffractometry (XRD) 275

309 The X-ray diffraction patterns were obtained using a Philips PW 3710 X-ray powder diffractometer. From the intensity distribution of the particular reflections and the integral intensity of the X-ray diffraction patterns, crystallite size distributions and average crystallite sizes can be calculated. With diminishing crystallite size, the measured XRD pattern exhibit broadened and often overlapping reflections. The broadening of the reflections is inversely proportional to the crystallite size. The size of coherently diffracting domains can be obtained from Scherrer s equation D = 0.9λ βcosθ where D is the crystal size, λ is the wavelength of X-ray radiation ( nm for Cu K α radiation), β is the full width at half maximum, and θ is the diffraction angle. Figure 1 demonstrated the XRD patterns for the N-doped TiO 2 nanoparticles prepared at different conditions, which will be discussed in detail in Chapter 4. Intensity / a.u. TiON 500 O C TiON 400 O C TiON 350 O C TiON 300 O C TiON 200 O C TiON Theta / O Figure 1 XRD spectra of the resultant samples: dried TiO 2-x N x (black curve), TiO 2-x N x after heat treatment for 3 hours at 200 o C (red curve), 300 o C (green curve), 350 o C (blue curve), 400 o C (cyan curve) and 500 o C (magenta curve) UV-visible Absorption, Reflectance and Photoluminescence (PL) 276

310 The optical property of the samples can be obtained by measuring the UV-visible absorption, reflectance and photoluminescence spectra. The UV-visible absorption and reflectance spectra were obtained on a Cary 50 UV-visible spectrometer equipped with a reflectance unit. The PL spectra were collected at room temperature on a Varian Eclipse Fluorescence spectrophotometer. Figure 2 demonstrates the UV-visible absorption (blue) and PL (red dashed, arbitrary scale) spectra of CdSe QDs discussed in Chapter 12, and Figure 3 shows the reflectance spectra of N-doped TiO 2 nanoparticles. Optical Density (a.u. ) E 2s E 1p E 1s s 3/2 (h)1s(e) 2 2s 3/2 (h)1s(e) 3 1s 1/2 (h)1s(e) 4 1p 3/2 (h)1p(e) 5 2s 1/2 (h)1s(e) 6 1p 1/2 (h)1p(e) 7 2s 3/2 (h)2s(e) Energy (ev) Figure 2 UV-visible absorption (blue) and PL (red dashed, arbitrary scale) spectra of CdSe QDs indicating a high-quality quantum dot sample with no detectable low-energy surface states. The numbered markers refer to the energies associated with the electronhole pair states shown. 277

311 100 a % Reflectance b TiO 2 nanoparticles TiO 2-x N x nanoparticles Wavelength / nm Figure 3 Reflectance measurements showing the red-shift in optical response due to the nitrogen doping of TiO 2 nanoparticles Transmission Electron Microscopy (TEM) and High-Resolution Transmission Electron Microscopy (HRTEM) The morphology of the materials can be revealed with Transmission Electron Microscopy (TEM). The low resolution TEM images were taken on a JEOL 1200EX transmission electron microscope operated at 80 kv. High-resolution transmission electron microscopy (HRTEM) images were obtained on a Tecnai F30 HRTEM machine operated under 300 kv. Samples for TEM were prepared by depositing a drop of a nanocrystal solution in water onto a copper grid supporting a thin film of amorphous carbon, and drying in air. Figure 4 shows the low resolution TEM images and HRTEM of the C-doped TiO 2 nanocrystals, which will be discussed in Chapter 7. E nm F nm 200 nm A B 50 nm G C D nm 50 nm nm 5 nm

312 Figure 4 TEM and HRTEM images of C-doped TiO 2 nanoparticles. TiC powder (A) and the resultant C-doped TiO 2 nanoparticles after heating before (B) and after size selective centrifugation (C) and (D). HRTEM images of three C-doped TiO 2 nanocrystals (E), (F), and (G). The insets in (A) and (B) show the selected area electron diffraction patterns of cubic TiC and the C-doped TiO 2 nanoparticles with anatase phase X-ray photoelectron spectroscopy (XPS) The chemical compositions of materials can be determined with X-ray photoelectron spectroscopy (XPS). In XPS, core-level electrons are pumped to the vacuum level by the incident light, and the binding energies of the outgoing electrons are measured. Different core-level electrons for different elements have different and fingerprint binding energies. All the XPS spectra were determined with, taken on a Perkin-Elmer PHI 5600 XPS System with the samples on a carbon tape sticking to the aluminum support. The XPS binding energies were calibrated with respect to the C 1s peak from the carbon tape at ev. Figure 5 demonstrates the XPS spectra for the P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocrystals (black) and nitrided TiO 2-x N x nanocrystals (green). For details, please see Chapter

313 TiO 2-x N x Nanocolloid O 1S TiO 2 Nanocolloid N-treated Degussa Degussa P25 Ti 2P Intensity / a.u. Ti LMM O KLL Ti 2S C 1S N 1S Ti 3P Ti3S Energy / ev Figure 5 Comparison of the X-ray photoelectron spectra of P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocrystals (black) and nitrided TiO 2-x N x nanocrystals (green). The inset shows the N 1S peak for these samples in the 400 ev region Fourier Transform Infrared Spectroscopy (FTIR) Fourier Transform Infrared Spectroscopy (FTIR) is a very powerful technique to explore the surface property of a material. This technique measures the absorption of various infrared light wavelengths by the material of interest. Absorption bands in the range of wavenumbers are typically due to functional groups (e.g. -OH, C=O, N-H, CH3, etc.). The region between wavenumbers is referred to as the fingerprint region. Absorption bands in this region are generally due to intra-molecular phenomena, and are highly specific for each material. The specificity of these bands allows computerized data searches to be performed against reference libraries to identify a material. Samples for FTIR measurement were prepared as KBr pellets and analyzed using a Thermo Nicolet Nexus 870 FTIR spectrometer. Figure 6 demonstrates the FT-IR 280

314 transmission spectra of P25 TiO 2 powder; nitrided P25 TiO 2 powder; TiO 2 nanocolloid particles and TiO 2-x N x nanocrystals. For details, please see Chapter 3. Transmission / a.u. TiO 2-x N x Nanocolloid TiO 2 Nanocolloid N-treated Degussa P25 Degussa P25 OH OH O-Ti-O Wavenumber / cm -1 Figure 6 FT-IR transmission spectra of P25 TiO 2 powder (red); P25 TiO 2 powder nitrided after triethyl-amine treatment (blue); TiO 2 nanocolloid particles (black) and nitrided TiO 2-x N x nanocrystals (green) Raman Spectroscopy Infrared (IR) and Raman spectroscopy both measure the vibrational energies of molecules but these method rely only different selection rules. Recall that for a vibrational motion to be IR active, the dipole moment of the molecule must change. Therefore, the symmetric stretch in carbon dioxide is not IR active because there is not change in the dipole moment. The asymmetric stretch is IR active due to a change in dipole moment. For a transition to be Raman active, there must be a change in polarizability of the molecule. 281

315 Raman Intensity / a.u. TiO 2-x N x Nanoparticle TiO 2 Nanoparticle Nitrited P25 Degussa P Raman Shift / cm -1 Figure 7 Raman Spectra of P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocrystals (black) and nitrided TiO 2-x N x nanocrystals (green). Raman spectra were recorded on a Raman microscope system. Small amount of sample was put over an aluminum tray on the microscope stage. A nm krypton laser beam (50 mw) is introduced on the microscope s optical axis using a fiber optic, and focused on the samples to generate the Raman scattering. 180 back-scattered Raman light was collected from the sample by using a long-focal-length X20 microscope objective lens. The sample can be viewed via a long focal length objective and a chargecoupled device (CCD) camera for optical imaging. Raman microscope was calibrated by using standardized neon and tungsten lamps before the measurement. Data analysis was performed using GRAMS/32 software (Galactic Industries, Inc.). Figure 7 demonstrates the Raman Spectra of P25 TiO 2 powder (red), nitrided P25 TiO 2 powder (blue), TiO 2 nanocrystals (black) and nitrided TiO 2-x N x nanocrystals (green). For details, please see Chapter Photocatalytic Activity Measurements 282

316 Photocatalytic activity measurements of doped TiO 2 nanoparticles under single-wavelength incident light illumination The photocatalytic activity of the TiO 2-x N x particles was evaluated at excitation wavelengths of 390 and 540 nm using a Clark MXR 2001 fs laser system. The laser beam (800 fs, 1 khz, 120 fs laser pulse train) was sent either through a BBO crystal to generate second harmonic 390 nm (10 mw) light pulses or to an optical parametric amplifier to generate stable 540 nm (4 mw) pulses. The laser light intensity was adjusted with a neutral density filter wheel. The pulse train was guided into a quartz cuvette filled with a 2 ml aqueous solution of methylene blue (optical density) 0.8) and 10 mg of the new catalyst, to excite a pump volume of about 5 nl (0.5 mm is the diameter of the excitation beam at the reaction cell). The decomposition of the solute was followed by measuring the decolorization of the methylene blue in solution with a Varian Cary Bio50 UV-visible spectrometer. The reported quantum yield for decomposition was based on the analysis of the photon flux, taking into account the volume of the excitation region and including the dilution factor for evaluating the optical density changes depicted. The process requires low intensity when compared with bright sunlight Photocatalytic activity measurements of doped TiO 2 nanoparticles in a simple parallel photochemical reactor under a simulating solar-lamp For overall photocatalytic activity of the doped TiO 2 nanoparticles, we have designed a simple and useful parallel photochemical reactor intended to study the photodecomposition of dyes. The photochemical reactions are followed through timedependent changes in the ground state absorption spectra of the dyes. A distinctive 283

317 characteristic of this reactor is that four reaction cells can be run under identical conditions and turned on/off simultaneously. Four different reaction conditions, such as four different concentrations of dye or catalysts (or four different catalysts or dyes) can be monitored over precisely the same time frame and analyzed after the same amount of photolysis. Also, the power dependence of the photodecomposition can be easily monitored by removing one or two beamsplitters in this system. Such a reactor can be used for a multitude of parallel photochemical transformations beyond the single reaction described in the experimental example. Figure 8 Illustration of the parallel photochemical reactor set-up. The main elements of this system (Figure 8) consist of a Xe lamp (150 W), three 50/50 beamsplitters, two Ag 100% reflecting mirrors, four convex lenses with short focusing distance (f = 50 mm), one weak convex lens, two magnetic stirrers, four cuvettes, four mini stir bars, and one electric shutter. The Xe lamp (Model 6137) and the lamp power supply (Model 8500) are from Oriel Corp. All the optical components can be obtained from ThorLab or Melles Griot. The three 50/50 beamsplitters (Melles Griot, Part No. 03BTF007) and the two 100% reflecting mirrors (ThorLab, Part No. BB1-E02) are arranged at a 45 angle with regard to the light path, which ensures that four light paths with equal light flux reaching the cuvettes containing the dye solution with catalyst. The 284