CONTAMINANT SELECTIVITY, MODEL DEVELOPMENT, AND PILOT STUDY OF INNOVATIVE ION EXCHANGE FOR SMALL PUBLIC WATER SYSTEMS

Transcription

1 CONTAMINANT SELECTIVITY, MODEL DEVELOPMENT, AND PILOT STUDY OF INNOVATIVE ION EXCHANGE FOR SMALL PUBLIC WATER SYSTEMS By YUE HU A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2017

2 2017 Yue Hu

3 To my family and friends

4 ACKNOWLEDGMENTS I would like to express my special thanks to my advisor Dr. Treavor Boyer who guided and encouraged me through my PhD life. My appreciation also extends to my committee members: Dr. Ben Koopman for his kindly and patiently guided me for my model development part of research as well as coached me for courses, Dr. Paul Chadik for this insightful comments and suggestions for my research and practice presentations as well as taught me courses, and Dr. Bin Gao for providing me comments on my research. I would also like to thanks the operators at Cedar Key to support my pilot studies. Special thanks to my family. My lovely grandparents always support me and understand me. My parents encouraged me to pursue an advanced degree abroad. My younger cousin who grew with me together at my grandparents home, always cares about me. I would also like to thank my lab mates who supported my research, encouraged me overcoming challenges, and make me feel our research group is my second home outside my home country. Finally, I would also thank my friends in our department for always willing to answer my questions, no hesitate to let me use the instruments, work on projects together and establish outreach student chapter. 4

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS... 4 LIST OF TABLES... 8 LIST OF ABBREVIATIONS ABSTRACT CHAPTER 1 INTRODUCTION Treatment Challenges of Small Public Water Systems A Robust Process Innovative Ion Exchange Bicarbonate-form Ion Exchange Combined Ion Exchange Organization of Dissertation SELECTIVITY OF BICARBONATE-FORM ANION EXCHANGE FOR DRINKING WATER CONTAMINANTS: INFLUENCE OF RESIN The Application of Bicarbonate-form Resin Materials and Methods Anion Exchange Resins Chemical Contaminants Batch Equilibrium Experiments Analytical Methods Data Analysis Binary Exchange Plots and Separation Factors Adsorption Isotherm Models Results Binary Exchange Plots and Separation Factors Adsorption Isotherms Contaminant Removal in Ideal System Discussion Selectivity of Bicarbonate- and Chloride-form Resins Influence of Resin Properties on Contaminant Removal Differences in Selectivity Sequences Implications of Separation Factor on Resin Selection Concluding Remarks

6 3 INTEGRATED BICARBONATE-FORM ION EXCHANGE TREATMENT AND REGENERATION FOR DOC REMOVAL: MODEL DEVELOPMENT AND PILOT PLAN STUDY The Process for DOC Removal Process Description and Model Formulation Process Description Model Formulation Materials and Methods Materials Experimental Methods Batch equilibrium tests Batch kinetic tests Pilot plant tests Analytical Methods Results and Discussion Raw Water Characteristics Laboratory Batch Experiments Pilot Plant Results Process Model Calculation and Validation Concluding Remarks SIMUTANEOUS MULTIPLE CONTAMINANTS REMOVAL IN A COMBINED ION EXCHANGE PROCESS FOR SMALL WATER SYSTEMS The Concept of Combined Ion Exchange Operation Materials and Methods Materials Experiments Batch equilibrium and kinetic tests Pilot-scale, completely mixed flow reactor study Analytical Methods Results and Discussion The Alternative Counterion on the Equilibrium and Kinetic Characteristics DOC and Calcium Removal at Varying Effective Resin Dose DOC and Calcium Removal of Multiple Treatment and Regeneration Cycles 108 Multiple Contaminants and Salt Water Intrusion Study Spiking strontium and nitrate to Cedar Key raw water Impact of seawater intrusion on DOC and hardness removal DOC and calcium removal using conventional AER and CER Conclusions Remarks CONCLUSIONS APPENDIX A SUPPLEMENTARY INFORMATION FOR CHAPTER

7 B SUPPLEMENTARY INFORMATION FOR CHAPTER C SUPPLEMENTARY INFORMATION FOR CHAPTER LIST OF REFERENCES BIOGRAPHICAL SKETCH

8 LIST OF TABLES Table page 2-1 Properties of each resin Separation factor of selected bicarbonate-form resin Langmuir parameter of monovalent contaminants using bicarbonate-form anion exchange resin Freundlich parameter of monovalent contaminants using bicarbonate-form anion exchange resin Langmuir parameter of divalent contaminants using bicarbonate-form anion exchange resin Freundlich parameter of divalent contaminants using bicarbonate-form anion exchange resin Thermochemical radius and hydration energy of anions investigated in ion exchange batch experiments Lab experiment plan Pilot plant test plan Summary of equilibrium and kinetic parameters Summary of pilot plant data and model prediction Experimental plan for batch tests and pilot tests Summary of raw water and treated water parameters for pilot tests A-1 ph change before and after equilibrium A-2 Langmuir parameter of divalent contaminants using bicarbonate-form anion exchange resin (corrected the molar concentration during calculation) A-3 Freundlich parameter of divalent contaminants using bicarbonate-form anion exchange resin (corrected the molar concentration during calculation) B-1 Summary of mass coefficient from literatures and this research B-2 The comparison of pilot plant data and model prediction

9 Figure LIST OF FIGURES page 1-1 The goal of innovative ion exchange is for small public water systems The organization of dissertation Binary equilibrium plot for selected contaminants and selected resins Experimental nitrate adsorption to bicarbonate-form resins and isotherm model fits to experimental data Experimental bromide adsorption to bicarbonate-form resins and isotherm model fits to experimental data Experimental perchlorate adsorption to bicarbonate-form resins and isotherm model fits to experimental data Experimental chromate adsorption to bicarbonate-form resins and isotherm model fits to experimental data Experimental sulfate adsorption to bicarbonate-form resins and isotherm model fits to experimental data The system diagram of CMFR type ion exchange process The concept of integrated treatment and regeneration ion exchange process Isotherm carry out by virgin resin with Cedar Key raw water (data symbols show mean value ± one standard deviation for triplicate samples) DOC concentration in liquid phase as varying the salt concentration in regeneration solution (NaHCO3, NaCl) (data symbols show mean value ± one standard deviation for triplicate samples) Adsorption kinetic carry out by virgin resins with Cedar Key raw water (data symbols show mean value ± one standard deviation for samples in three parallel test) Adsorption kinetic carry out by DOC-loaded resins with regeneration solution (data symbols show mean value ± one standard deviation for samples in three parallel test) The DOC removal percentage of pilot plant data and model prediction (data symbol shows mean of duplicate samples) Dynamic model predictions for runs 4 and Isotherm fitting and equilibrium and kinetic parameters

10 4-2 Contaminant removal percentage and concentrations in effluent under varying the conditions of effective resin dose level Contaminant removal percentage and concentration in effluent during alternative-ion-form and original-ion-form operations Contaminant removal percentage and concentration in effluent for scenario of Cedar Key raw water spiking with SrCl2 and NaNO Contaminant removal percentage and concentration in effluent for scenario of Cedar Key raw water blending with Atlantic Ocean water DOC and calcium of removal percentage and concentration in effluent during CIX operation using conventional AER and CER The application of innovative ion exchange to small communities, and the pathways of water, contaminants and chemicals A-1 Experimental nitrate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose) A-2 Experimental bromide adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose) A-3 Experimental perchlorate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose) A-4 Experimental chromate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose) A-5 Experimental sulfate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose) A-6 Experimental chromate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (corrected the molar concentration during calculation) A-7 Experimental sulfate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (corrected the molar concentration during calculation) A-8 Experimental chromate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose, corrected the molar concentration during calculation) A-9 Experimental sulfate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose, corrected the molar concentration during calculation)

11 B-1 Yue Hu, Pilot plant set up in Cedar Key drinking water treatment plant, March 2016, Cedar Key, Florida, USA. Photo courtesy of Yue Hu B-2 DOC concentration in resin phase as varying the salt concentration in regeneration solution (NaHCO3, NaCl) C-1 Isotherm fitting for equilibrium batch tests for potassium-form and sodiumform CER C-2 Model prediction for calcium removal during multiple loading jar-tests performed in previous literatures and continuous pilot tests for Cedar Key raw water treatment when dosed 6 ml/l of Plus resin

12 LIST OF ABBREVIATIONS AER Br - CER CMFR Ca 2+ CaCO3 CaSO4 Cl - CrO4 2- ClO4 - CIX CER DIC DOM DOC EPA EC ERD HRT IX MCL NaHCO3 NaCl Na + Anion exchange resin Bromide ion Cation exchange resin Completely mixed flow reactor Calcium ion Calcium carbonate Calcium sulfate Chloride ion Chromate ion Perchlorate ion Combined ion exchange Cation exchange resin Dissolved inorganic carbon Dissolved organic matter Dissolved organic carbon US Environmental Protection Agency Enhanced coagulation Effective resin dose Hydraulic residence time Ion exchange Maximum concentration level Sodium bicarbonate Sodium chloride Sodium ion 12

13 NF NO3 - RO SO4 2- UV nanofiltration Nitrate ion Reverse osmosis Sulfate ion Ultraviolet absorbance 13

14 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CONTAMINANT SELECTIVITY, MODEL DEVELOPMENT, AND PILOT STUDY OF INNOVATIVE ION EXCHANGE FOR SMALL PUBLIC WATER SYSTEMS By Yue Hu August 2017 Chair: Treavor Boyer Major: Environmental Engineering and Sciences Small public water systems (PWS), serving less than 10,000 residents, encounter numerous challenges to comply with regulations set by the Safe Drinking Water Act. Also, some unregulated constituents can become target contaminants considering the subsequent processes operation. Therefore, small PWS require an effective process that is affordable and suitable for a wide range contaminant removal. The innovative ion exchange (IX) presented in this research aims to help small PWS to remove problematic chemical contaminants, including disinfection byproduct precursors, nitrate, perchlorate, chromate, hardness, and strontium. Specifically, the innovative IX includes bicarbonate-form IX and combined ion exchange (CIX). Bicarbonate-form IX uses bicarbonate as a counterion and bicarbonate salts for regeneration to remove negatively charged contaminants. The benefit of bicarbonate-form IX is the disposal of bicarbonate salt has less harmful environmental impacts compared to a traditional sodium chloride regeneration solution. In the multiple contaminants removal, the CIX uses the anion exchange resin (AER) and cation exchange resin (CER) in the same vessel and regenerates exhausted AER and CER simultaneously. The CIX process can 14

15 save the chemical usage and footprint compared with a traditionally separate AER and CER operation. Realizing the benefits, the overall goal of this research aims to evaluate the treatment performance of innovative IX for a wide range of contaminant removal for small PWS. The work in this research includes a selectivity study of bicarbonate-form AER for a variety of negatively charged contaminants, namely nitrate, bromide, perchlorate, chromate, sulfate and dissolved organic carbon (DOC). Additionally, model development and pilot studies were performed to evaluate the bicarbonate-form IX for DOC removal. Lastly, the pilot study for CIX was performed to expand the treatment scope to remove both negatively and positively charged contaminants. The target contaminants included DOC, calcium, strontium, and nitrate. Results in this research identified the key resin properties that influence the treatment performance for common water contaminants. The spacing of functional groups and the type of resin matrix were the important properties for inorganic contaminants removal, and the pore structure play an important role in DOC removal. With respect to DOC removal in a practical IX configuration, the model predictions and pilot studies all showed a comparable treatment performance when comparing with traditional chloride-form AER and sodium chloride for regeneration. Moreover, the DOC treatment was improved by increasing the salt concentration in the regeneration solution. For the multiple contaminants removal, the CIX process showed the substantial removal of DOC, calcium and strontium with varying regeneration conditions and source water parameters. 15

16 CHAPTER 1 INTRODUCTION U.S. Environmental Protection Agency (EPA) defines small public water systems (PWS) as systems that serve less than 10,000 residents. Small PWS face unique and complex challenges to be in compliance with drinking water standards and operate sustainably. Unlike many large utilities, most small PWS have limited technical, managerial, and financial resources. For example, usually, only a few staff are responsible for water plant operation and customer service. Due to limited financial resources, many small PWS cannot afford to install automated data monitoring and recording systems, and policy makers are reluctant to increase water rates on a small base of customers. As the need for environmental protection increases, small water systems are inclined to operate in more sustainable ways, meaning cost-effective performance, low energy demand, less material usage and minimal environmental impact. This research focuses on a new type of ion exchange (IX) to remove a wide range of chemical contaminants to help small PWS provide safe and reliable drinking water. Treatment Challenges of Small Public Water Systems In 1976, the Safe Drinking Water Act was passed, and established maximum contaminant levels (MCLs) for 22 contaminants and all PWS were forced to comply. These regulations are continuously being reviewed and updated, and by 2010, 91 contaminants were regulated. A recent publication summarized the health-based violations for small PWS, and nitrate exceedances and Disinfection Byproducts (DBPs) Stage 1 Rule violations are among the most common (Oxenford and Barrett, 2016). U.S. EPA regulated nitrate in 1992 with MCL of 10 mg/l as N. Nitrate naturally occurs in 16

17 groundwater, and excess nitrate is often detected in active agricultural areas (Bernard et al., 1998, Bernard and Kerie, 2006). In 2006, U.S. EPA released the DBP Stage 2 Rule, and all small PWS were required to start compliance monitoring in DBPs are formed when dissolved organic carbon (DOC) reacts with chlorine reagents during the disinfection process. In addition to federal regulations, small PWS are also required to meet state standards. Hexavalent chromium was regulated in California in 2014 with an MCL of 0.01 mg/l. Hexavalent chromium can naturally exist in groundwater and can also enter the aquifer through leaching of metallic wastewater. A recent study detected hexavalent chromium exceedances in over 406 wells in California (GAMA, 2016). In addition to meeting existing regulations, small PWS must also plan for future regulated contaminants. Perchlorate and strontium were both listed in Contaminant Candidate List 3, and strontium was listed in the Regulatory Contaminant Determination 3, which indicates strontium may be regulated in the future. Some unregulated constituents can still be listed as target contaminants because they negatively affect subsequent treatment processes or cause taste issues in the finished water. For example, a common problem is high hardness in raw water, which could lead to membrane fouling (Shirazi et al., 2010). Hardness is formed from the dissolution of divalent cations, including calcium and magnesium. Moreover, in certain extreme events, the source water quality changes drastically and other problems arise and become problems for water treatment, i.e., saltwater intrusion (Seatta et al., 2015). Small PWS must select effective, affordable and sustainable treatment processes to meet established compliance and adapt to the change of treatment to provide high quality drinking water for the community. 17

18 A Robust Process Reliable, affordable and sustainable processes are required to aid small PWS in meeting existing compliance standards, adapting to new regulations, and maximizing the long-term benefits to the public and environment. The best available technologies that can address common violated chemical contaminants and have the capability to adapt to changing regulations include the ion exchange (IX), nanofiltration (NF), and reverse osmosis (RO). In IX treatment, anion exchange resin (AER) can be used to remove negatively charged contaminants such as DOC, nitrate, perchlorate, and hexavalent chromium; cation exchange resin (CER) can be used to remove positively charged contaminants such as hardness and strontium. The removal performance of IX processes for a wide range of contaminants has been widely studied in recent years (Boyer et al., 2008b, Darracq et al., 2014, Edebali and Pehlivan, 2010, Li et al., 2016, Mergen et al., 2008). Regarding membrane processes, NF and RO can be used to remove dissolved inorganic ions and DOC. Most NF installations are focused on hardness removal, but the use of NF has since expanded to remove additional dissolved ions. RO is mainly used for seawater and brackish water desalination. The effective removal performance of NF and RO for a variety of contaminants has been well studied and documented (Garcia et al., 2006, Mahvi et al., 2012, Malaeb and Ayoub, 2011, Uyak et al., 2008, Yoon et al., 2009). However, IX, NF and RO processes all generate a concentrated salt brine, and waste disposal is a major issue (Jensen and Darby, 2016, McTigue and Cornwell, 2009). The benefits to the community of using the above listed technologies are minimized without adequate and affordable brine waste management. Brine waste disposal methods include deep-well injection, pond evaporation, land application, 18

19 crystallization and sewer discharge. Major concerns associated with those disposal methods include high cost, stringent permitting and harmful effects on aquatic species and plant growth (Jensen and Darby, 2016, McTigue and Cornwell, 2009). Brine reuse technologies have been widely evaluated (Bae, 2002, Choe et al., 2015, Ibáñez et al., 2013, Lehman et al., 2008); however, the additional cost may not be feasible for small PWS. The unique mechanism of the IX reaction and operation approach enable conventional IX processes to be modified. The innovative IX processes presented in this research aim to fulfill the ultimate goal of this research: to aid small PWS in meeting compliance requirements while maximizing the benefits to the community and environment (Figure 1-1). Innovative Ion Exchange The development of innovative IX aims to modify current IX to achieve comparable, effective treatment performance as well consider the waste generation. The major problem of current IX is the disposal of highly concentrated sodium chloride brine (NaCl) to the environment. The NaCl is used during the regeneration step to recover the resin capacity. Alternatively, the bicarbonate-form IX, one of the innovative IX processes, uses bicarbonate salt for regeneration, has multiple benefits related to subsequent treatment processes and waste disposal. Similarly, the new combined ion exchange (CIX) operation can reduce the quantity of brine because resins are regenerated in the same vessel. Specifically, bicarbonate-form IX can be used to remove a wide range of negatively charged contaminants, and the CIX process can further expand the treatment scope for multiple contaminants removal. 19

20 Bicarbonate-form Ion Exchange The mechanism of IX treatment is through the stoichiometric exchange of a counterion in the resin with a contaminant ion in raw water. Using the bicarbonate salts for resin regeneration requires switching the counterion to the bicarbonate ion for AER. One advantage of using bicarbonate salt is that the brine can be disposed to the sewer system, which is low cost and bicarbonate can be used as an alkalinity source for nitrification during wastewater treatment. Additionally, treated water with an elevated bicarbonate concentration, instead of chloride in conventional IX, would increase the buffering capacity of the finished water and be beneficial to corrosion control (Nguyen et al., 2011a, Nguyen et al., 2011b). Previous studies have focused on the removal of DOC and inorganic contaminants, such as nitrate and sulfate, using bicarbonate salts regeneration in laboratory batch tests (Rokicki and Boyer, 2011, Walker and Boyer, 2011). Expanding the application of bicarbonate-form IX, Chapter 2 evaluated AERs with a wide range of properties to test selectivity for a variety of negatively charged constituents, some of which have regulated MCLs or were listed in Contaminant Candidate List 3. Furthermore, Chapter 3 focused on bicarbonate-form IX for disinfection byproducts precursor removal. The model was developed according to the recently developed completely mixed flow reactor type of configuration, and pilot-scale of bicarbonate-form IX was performed for operational understanding. Combined Ion Exchange In CIX, AER and CER operation takes place in the same vessel. CIX can be used to remove both negatively and positively charged contaminants simultaneously and uses the same batch salt solution for combined AER and CER regeneration, where 20

21 the anions in the salt solution is used to recover AER capacity and the cations is used to regenerate cation resin. Typically, the separate anion and cation processes are implemented, or individual anion or cation processes are coupled with other process units. That means either the anion or cation counterion in the salt solution is used, but the remaining ion is wasted without any function. CIX can be used to achieve multiple contaminant removal and has many benefits, particularly for small PWS. CIX has been evaluated for the removal of DOC and hardness in laboratory batch studies (Apell and Boyer, 2010, Comstock and Boyer, 2014). Recently published work focused on the large-scale study of CIX process for DOC and hardness removal (Arias-Paic et al., 2016, Locke and Smith, 2016). Regarding regeneration, the simultaneous regeneration of exhausted nitrate AER and calcium AER was evaluated with sodium bicarbonate (NaHCO3), potassium chloride and NaCl (Maul et al., 2014). To further expand the treatment of CIX process, Chapter 4 focused on pilot-scale evaluation of CIX operation considering multiple contaminants, alternative regeneration chemicals, variable water quality and alternative resins. Organization of Dissertation The investigation of innovative IX processes aims to provide an effective solution for small PWS to remove a wide range of contaminants along with considering the waste generation. Specifically, the goal of Chapter 2 is to generate new data on selectivity of AERs using bicarbonate as the mobile counterion for anion exchange removal of important drinking water contaminants. The selectivity analysis is conducted for six bicarbonate-form AER paired with six common anionic water contaminants. The selectivity sequence, the influence of resin properties and treatment performance are investigated to provide the resin selection recommendation according to the treatment 21

22 purpose. Chapter 2 was published in the Separation and Purification Technology in The goal Chapter 3 is to evaluate the DOC removal of using bicarbonate-form resin for treatment and NaHCO3 for regeneration in CMFR process. The process model is developed to couple the treatment and regeneration to predict the treatment performance. The pilot studies are performed for model validation as well as provide a practical evaluation. Chapter 3 was published in the Water Research in The goal of Chapter 4 is to evaluate the performance of the CIX process for small PWS to achieve multiple contaminants removal. The work of CIX process considers the mixture of multiple contaminants treatment, using alternative salts for regeneration and variety of resins for CIX operation. The research of Chapter 4 plan to submit to Water Research in May Lastly, the key results, application, and challenges of innovative IX are summarized and discussed in the Conclusion chapter. The schematic organization of this dissertation is shown in Figure

23 Figure 1-1. The goal of innovative ion exchange is for small public water systems 23

24 Figure 1-2. The organization of dissertation 24

25 CHAPTER 2 SELECTIVITY OF BICARBONATE-FORM ANION EXCHANGE FOR DRINKING WATER CONTAMINANTS: INFLUENCE OF RESIN * The Application of Bicarbonate-form Resin Ion exchange is a water treatment process in which the mobile counterion on the resin phase is stoichiometrically exchanged with contaminant ions in the aqueous phase. In drinking water treatment, anion exchange is used to remove regulated and problematic contaminants, such as nitrate, perchlorate, and natural organic matter (NOM). Other advantages of ion exchange include low capital cost and high efficiency at producing high quality water (Clifford Dennis, 1993). However, concentrated sodium chloride solution is required for regeneration of strong-base anion exchange resin, and thus the management of waste brine is a major challenge to implementing ion exchange, especially for areas that lack disposal options. The direct discharge of brine to land or surface water can lead to harmful impacts on plant growth and wildlife (L.Bernstein, 1975, M.Canedo-Arguelles, 2013). The discharge of brine to wastewater treatment plants requires a large dilution and can be adverse to biological treatment (T.Panswad, 1999). In order to improve the management of waste brine, an alternative salt for resin regeneration has been proposed to broaden the acceptable disposal options. Previous work has demonstrated the potential of using sodium bicarbonate for resin regeneration and using bicarbonate-form resin for contaminant removal. The similar performance of using sodium bicarbonate and sodium chloride in terms of * Reproduced with permission from Yue Hu, Jerrine Foster, Treavor H. Boyer. Selectivity of bicarbonateform anion exchange for drinking water contaminants: Influence of resin properties. Separation and Purification Technology 163, , Copyright 2016 Elsevier Ltd. 25

26 regeneration and contaminant removal was observed (Rokicki and Boyer, 2011). Moreover, the use of sodium bicarbonate for resin regeneration and bicarbonate-form resin for contaminant removal is beneficial both for waste disposal and treated water. Bicarbonate contributes alkalinity to natural water, and it is also a benefit to biological wastewater treatment. For example, lack of alkalinity will prevent nitrification (Spellman, 2008). In addition, increasing alkalinity of treated water is beneficial to drinking water distribution systems by reducing certain types of corrosion (Cong et al., 2009, Sarver and Edwards, 2012). Only a few studies have reported using sodium bicarbonate for resin regeneration and using bicarbonate-form resin for contaminant removal. Bicarbonateform magnetic ion exchange (MIEX) resin was previously compared with chloride-form MIEX resin with regard to its affinity, regeneration, and stoichiometry for sulfate, nitrate, and NOM (measured by dissolved organic carbon (DOC) and UV absorbance at 254 nm (UVA254)) (Rokicki and Boyer, 2011). The similar performance was demonstrated for both bicarbonate-form and chloride-form resins with respect to selectivity and regeneration. Moreover, the long-term DOC and bromide removal and regeneration were also investigated by using bicarbonate-form MIEX resin (Walker and Boyer, 2011). The chloride-form MIEX resin was used as a baseline for comparison. Both bicarbonate-form MIEX resin and chloride-form MIEX resin demonstrated similar trends for long-term regeneration. The maximum contaminant removal efficiency was achieved with virgin resin, and contaminant removal decreased with increasing number of regeneration cycles. The contaminant removal sequence was the same for both bicarbonate-form and chloride-form MIEX resin. In another research, the regeneration 26

27 efficiency of several different strong-base anion exchange resins was studied using different bicarbonate salts, namely sodium bicarbonate and potassium bicarbonate, and compared with chloride-form resin (Maul et al., 2014). The regeneration efficiency of bicarbonate was higher than chloride for polyacrylic type resin, whereas it was lower for polystyrene type resin. Furthermore, the use of resin in bicarbonate form, hydroxide form, and chloride form, to exchange with chloride, nitrate, and sulfate was investigated with several selected adsorption isotherm models and thermodynamic models (Dron and Dodi, 2011a, b), and showed the Dubinin-Astakhov had the best data fitting among adsorption models and nonideal thermodynamic model had the best fitting than other models. The previous studies have demonstrated the efficiency of using sodium bicarbonate for resin regeneration and the potential of a few different bicarbonate-form resins, mostly MIEX resin, in terms of contaminant removal. Therefore, there remains a major gap in the literature pertaining to the treatment efficiency of bicarbonate-form resin across a range of resin types and contaminants. In this work, six strong-base anion exchange resins were selected to represent a wide range of resin properties. The interactions between these resins and six regulated and/or problematic drinking water contaminants were studied. The contaminants included nitrate, bromide, perchlorate, sulfate, chromate, and Suwannee River natural organic matter (NOM). Nitrate, perchlorate, and chromate (i.e., hexavalent chromium) were selected because of adverse effects on human health. Bromide and Suwannee River NOM were selected because of their contribution to disinfection byproduct formation (Hsu and Singer, 2010). 27

28 Lastly, sulfate was included in this study because it is a known competitor for ion exchange sites (Song et al., 2012). The goal of this research was to generate new data on ion exchange selectivity using bicarbonate as the mobile counterion for anion exchange removal of important drinking water contaminants. The specific objectives were to: 1) evaluate the effect of resin properties on selectivity of each contaminant; 2) determine the resin selectivity sequence for each bicarbonate-form resin based on separation factors and isotherm parameters; and 3) quantify contaminant removal as a function of resin type and resin dose. All data were generated using batch equilibrium experiments in which bicarbonate-form resin was mixed with synthetic water containing a single contaminant. The purpose of using synthetic contaminant solutions was to enable an isolated investigation of each contaminant s interactions with the selected resins and varying resin properties. Materials and Methods Anion Exchange Resins The resin properties of the six selected resins are listed in Table 2-1. All resins were obtained from the manufacturer in the chloride form. Resin density was calculated in triplicate by measuring 1 ml of wet settled resin, allowing the resin to dry, and measuring the dry mass. Density was used for determination of the dry resin dose. All resins were converted from the chloride form to the bicarbonate form by mixing the resins in highly concentrated sodium bicarbonate solution. The concentration of sodium bicarbonate (CAS# ) was 25 times greater than the equivalent amount of anion exchange resin capacity. ph was measured to confirm that the bicarbonate ion was the dominant ion in solution. A jar test apparatus was used for conditioning the resin by 28

29 mixing at 100 rpm for 24 h. After mixing, the resin was carefully rinsed with deionized (DI) water until the conductivity was ~1 µs/cm. The conditioned resins were stored in a desiccator to dry until used. Chemical Contaminants The chemical contaminants used in this work were sodium nitrate (CAS# ), sodium bromide (CAS# ), sodium chromate (CAS# ), sodium sulfate (CAS# ), sodium perchlorate (CAS# ) and Suwannee River NOM (SRNOM, 1R101N, International Humic Substances Society). The synthetic contaminant solutions were prepared by dissolving 5 meq/l of each inorganic contaminant ion (0.24 meq/l Suwannee River NOM) in DI water. The concentration of 0.24 meq/l NOM was chosen to make the concentration of dissolved organic carbon (DOC) equal to 30 mg/l as C. Batch Equilibrium Experiments Batch equilibrium tests were conducted with various resin doses and each synthetic contaminant solution in 125 ml amber glass bottles. Resin doses were selected for all inorganic contaminants to theoretically achieve 5, 25, 50, 100, 150, and 300% removal of the equivalent initial contaminant concentration. Resin doses were selected for NOM to theoretically achieve 5, 25, 50, 100, and 150% removal of the equivalent initial contaminant concentration. The calculation of percent resin dose was resin dose (meq/l) divided by initial contaminant concentration (meq/l), expressed as percent. For each resin dose, control samples included contaminant solution without resin (measured in duplicate), and all other samples were measured in triplicate. All samples were put on a shaker table for 24 h at room temperature (approx. 22 ºC). The 29

30 ph of each sample was measured before and after the batch equilibrium test. After equilibrium, all samples were filtered through 0.45 µm filter paper. Analytical Methods Ion chromatography (IC) (Dionex 3000 equipped with AS22 guard column and analytical column) was used for analysis of nitrate, sulfate, bromide, perchlorate, and chloride. The IC method followed U.S. EPA Method (EPA, 1997). The eluent was prepared as 4.5 mm Na2CO3/1.4 mm NaHCO3 and filtered through 0.45 µm filter paper before use. A chloride selective electrode (Thermo Orion) was used for analysis of chloride concentration in NOM batch test. Total organic carbon (TOC) analyzer (Shimadzu TOC-VCH) was used for analyzing dissolved inorganic carbon (DIC) and DOC. Bicarbonate concentration was calculated based on ph and DIC (Benjamin, 2010). Inductively coupled plasma (ICP) (Thermo ICAP 6200) was used to analyze hexavalent chromium following U.S. EPA Method 6010C (EPA, 2007). Chemical standards were purchased and analyzed to assess the accuracy of the results, i.e., seven anion standard (Dionex) for IC, inorganic carbon and organic carbon standards (Ricca) for TOC analyzer, and hexavalent chromium standard (Ricca) for ICP. All samples were measured in duplicate, with relative difference less than 10%. In each run, one calibration point was used for the periodic check after 10 to 15 samples. The entire calibration curve was re-analyzed at the end of each run. External calibration standards were analyzed with relative difference less than 10% at concentration > 1 mg/l and less than 20% at concentration < 1 mg/l. 30

31 Data Analysis Binary Exchange Plots and Separation Factors The binary exchange plot is a qualitative method to evaluate the affinity of an ion exchange resin for a contaminant. The plot shows the contaminant concentration fraction in the aqueous phase on the x-axis and the contaminant concentration fraction in the resin phase on the y-axis. The concentration fraction was calculated as the ion concentration normalized to the total concentration in the aqueous phase or resin phase. The separation factor, αi/j, for each pair of contaminant and mobile counterion, is a quantitative method to indicate the selectivity of an ion exchange resin for a contaminant. It is calculated using the contaminant ion concentration distribution between the aqueous phase and resin phase, divided by the mobile counterion concentration distribution between the two phases (Equation 2-1) (Edzwald and Tobiason, 2010). The separation factor varies with contaminant fraction in the aqueous and resin phase. Therefore, the representative separation factor was calculated at the contaminant fraction in the aqueous phase at xi = 0.5 (Edzwald and Tobiason, 2010). Linear interpolation was used to approximate the separation factor value at xi = 0.5. The linear interpolation was applied for nitrate, bromide, and sulfate. The data nearest to the point at xi = 0.5 was used to calculate the separation factor for perchlorate and chromate. For those two contaminants, the separation factor changed sharply with contaminant fraction in both phases in nonlinear trend; therefore, the selection of closest data point to xi = 0.5 was more appropriate compared to linear interpolation method. The separation factor of the data that most closed to the point to xi = 0.5 was applied to NOM calculation. The separation factor was calculated as a binary system (i.e., bicarbonate/contaminant) for inorganic contaminants and NOM; however, chloride 31

32 was founded in NOM solution, which can compete with NOM uptake. Additional details on this are discussed in the section of Influence of resin properties on contaminant removal. y i / x i α i/hco3 = y HCO3 /x HCO3 (2-1) Contaminant ion concentration in aqueous phase, ce (meq/l), was measured by IC, and contaminant ion concentration on resin phase, qe (meq/g), was calculated from Equation 2-2, where v is volume of synthetic contaminant solution (L), m is dry mass of resin (g), and c0 is initial contaminant solution concentration in aqueous phase (meq/l). q e = v m (c 0 c e ) (2-2) Adsorption Isotherm Models Langmuir and Freundlich isotherm models were selected and calculated using nonlinear regression in MATLAB (Matlab 2014b). In the resulting isotherm graphs, the x-axis, ce (mmol/l), equilibrium contaminant ion concentration in the aqueous phase, is plotted with the y-axis, qe (mmol/g), equilibrium contaminant ion concentration in the resin phase. Using the Langmuir isotherm modeling approach, monolayer coverage and constant adsorption energy are assumed (Foo and Hameed, 2010). The equilibrium constant, kl (L/mmol), and maximum capacity, q0 (mmol/g), are obtained by fitting data to the Langmuir model (Equation 2-3). The Freundlich isotherm model is empirically derived and includes the assumption that the resin has heterogeneous surface and unbounded resin capacity (Foo and Hameed, 2010). The Freundlich parameter kf (mmol/g) and n (dimensionless) are obtained by fitting data to Equation

33 q e = q 0k L c e 1+k L c e (2-3) 1/n q e = k F c e (2-4) G = RT lnk L (2-5) Results Binary Exchange Plots and Separation Factors Binary exchange plots and separation factor calculations were used to evaluate the selectivity of each resin for the selected contaminants. The x-axis is the equivalent concentration fraction of contaminant ion i in the aqueous phase, and the y-axis is the equivalent concentration fraction of contaminant ion i in the resin phase (Figure 2-1. a f). Qualitatively, data points above the 1:1 line indicate favorable selectivity with data points further above the 1:1 line showing higher selectivity. The magnitude of the separation factor, α i/j, quantitatively demonstrates the selectivity of the resin for a contaminant. Values of the separation factor for each combination of resin and contaminant are listed in Table 2-2. A separation factor > 1 indicates favorable selectivity, while a separation factor < 1 indicate unfavorable selectivity. The higher the value of the separation factor, the higher the selectivity of the resin for a given contaminant. The separation factor value of NOM was very low (Table 2-2) due to the presence of chloride, and is therefore excluded from the following explanation of the results. Bicarbonate-form A520E resin had the highest selectivity for nitrate, bromide, and perchlorate as illustrated by the A520E data points being furthest above the 1:1 line (Figure 2-1). The quantitative affinity of bicarbonate-form A520E resin for nitrate, bromide and perchlorate was calculated with corresponding separation factors of 86.3, 33

34 7.3, and 2.2, respectively (Table 2-2), which were also the highest values for each resin. Bicarbonate-form Marathon11 resin showed the highest selectivity for chromate (Figure 2-1d) with a separation factor of 68.1 (Table 2-2). Lastly, the bicarbonate-form IRA958 resin had the highest selectivity for sulfate (Figure 2-1e) with a separation factor of 5 (Table 2-2), and also for Suwannee River NOM (Figure 2-1f) with a separation factor of 1.3 (Table 2-2). The separation factor for each combination of bicarbonate-form resin and contaminant was used to generate resin-specific selectivity sequences as follows: A520E: Perchlorate > chromate > nitrate > sulfate > bromide Marathon11: Chromate > perchlorate > sulfate > bromide > nitrate Dowex22: Perchlorate > chromate > sulfate > nitrate bromide A300: Chromate > perchlorate > sulfate > bromide nitrate IRA958: Chromate > sulfate > perchlorate > bromide nitrate A850: Chromate > perchlorate > sulfate > bromide nitrate Adsorption Isotherms Langmuir and Freundlich isotherms were used to investigate adsorption behavior. In Figures 2 6, the x-axis is the contaminant concentration in the aqueous phase, and the y-axis is the contaminant concentration in resin phase. For each combination of resin and contaminant, the Langmuir (q0 and kl) and Freundlich (kf and n) model parameters were calculated and listed in Tables The change in Gibbs free energy (ΔG 0 ) calculated from the Langmuir model indicates the spontaneous nature of the sorption process, and it can be used to determine the contaminant selectivity sequence for each resin (Dron and Dodi, 2011a). The ΔG 0 < 0 indicates spontaneous and favorable sorption, whereas the ΔG 0 > 0 indicates nonspontaneous and unfavorable sorption. The dimensionless parameter 1/n obtained from the Freundlich model can also be used to evaluate sorption behavior, where 0 < 1/n < 1 34

35 indicates favorable sorption and chemisorption process, whereas 1/n > 1 indicates unfavorable sorption and cooperative adsorption (Foo and Hameed, 2010, Rauf et al., 2008). The statistical metrics, R-squared (R 2 ), average relative error (ARE), and the sum of squares of errors (SSE), were calculated to evaluate the fit of the models to the data. SSE can only be used to evaluate the data within the same pair of resin and contaminant. R 2 and ARE can be used to compare among different models and pair of resin and contaminant. Data points corresponding to the lowest resin dose showed high variability (Appendix Figures A-1 A-5), and these points were excluded as discussed in the following paragraphs. For the monovalent inorganic contaminants (Figures and Tables ), both the Langmuir and Freundlich models demonstrated a good fit to the data for most bicarbonate-form resins. All bicarbonate-form resins showed nonlinear and favorable sorption behavior. The ΔG 0 < 0 (calculated from the Langmuir model) indicated the favorable exchange of bicarbonate on A520E resin to nitrate and bromide, and bicarbonate on all resins except IRA958 to perchlorate. The value of 1/n also indicated favorable sorption to A520E resin for all monovalent contaminants. The resin capacity was determined from the q0 parameter of the Langmuir model. Overall, the calculated resin capacity was larger than the value provided by the manufacture. The calculated value of the resin capacity for perchlorate was in closer agreement with the manufacturer value than nitrate or bromide. The relative error of the calculated q0 and manufacturer value ranged from % for perchlorate. The most obvious difference between the calculated resin capacity with the manufacturer value was for A850 resin with nitrate and bromide. 35

36 For the divalent inorganic contaminants (Figures and Tables ), both the Langmuir and Freundlich models showed a good fit to the data for the majority of bicarbonate-form resins. All divalent ions and resins had ΔG 0 < 0, which indicated favorable exchange. Similarly, the value of 1/n indicated favorable sorption for all divalent contaminants. The q0 parameter value was in relatively close agreement with the manufacturer value for all bicarbonate-form resins with divalent contaminants. The relative error of the calculated q0 and manufacturer value ranged from % and 3 20% for chromate and sulfate, respectively. The ΔG 0 obtained from Langmuir model provided an estimate of the contaminant selectivity sequence for each resin, which were as follows: A520E: Perchlorate > chromate > bromide nitrate > sulfate Marathon11: Chromate > sulfate > perchlorate > nitrate bromide Dowex22: Chromate > perchlorate > sulfate > bromide nitrate A300: Chromate > Perchlorate > sulfate > bromide nitrate IRA958: Chromate > sulfate > perchlorate bromide nitrate A850: Chromate > perchlorate sulfate > bromide nitrate Contaminant Removal in Ideal System Contaminant removal in an ideal system (i.e., no competing contaminants) was investigated as a function of bicarbonate-form resin dose and type (Figure 2-7). For nitrate removal (Figure 2-7), A520E resin achieved the highest removal. The remaining resins had lower and similar removal results. At the 100% resin dose, A520E resin reached over 80% nitrate removal, whereas the other resins achieved 54 59% removal. Results for bromide removal were similar to nitrate. At the 100% resin dose, A520E resin showed > 70% bromide removal, while the other resins had approximately 60% bromide removal. All resins showed very high chromate removal. At the 100% resin dose, Marathon11 and A850 resins reached approximately 90% chromate removal, and 36

37 the other resins reached approximately 80% chromate removal. All resins showed high sulfate removal (~80%), except Dowex22 resin, which achieved 60%. Among the contaminants, resin performance was the most variable for perchlorate removal. A520E resin had the highest perchlorate removal. At the 100% resin dose, A520E resin reached > 90% removal, followed by Dowex22 and Marathon11 resins with > 85% removal; A300, A850, and IRA958 resins had 80%, 70%, and 60% removal, respectively. Chloride was present in the NOM synthetic water (due to the source of NOM isolate), which may have decreased NOM removal due to a competing anion present. The NOM removal was calculated without considering the influence of chloride. At the 100% resin dose, IRA958 resin achieved > 80% NOM removal, whereas the other resins achieved < 60% removal. Discussion Selectivity of Bicarbonate- and Chloride-form Resins The separation factor of chloride-form resin for perchlorate, chromate, nitrate, sulfate, and bromide is 150, 100, 3.2, 9.1, and 2.3, respectively (Edzwald and Tobiason, 2010). The resin used in this data was a strong-base resin with trimethylamine (R N + (CH3)3) functional groups and polystyrene composition, while other characteristics were not specified (Edzwald and Tobiason, 2010). The most selective bicarbonate-form resins for each contaminant are comparing to above published data, which was listed in Table 2-2. The separation factor in this work for perchlorate, chromate, nitrate, sulfate, and bromide is 86.3, 68.1, 7.3, 5, and 2.2, respectively. Although it is difficult to directly compare the separation factors between the chloride-form resin and bicarbonate-form resin because of different resin properties and experimental conditions, this result still shows the matchup magnitude of the separation factor value, which further indicates the 37

38 potential for a similar performance when using bicarbonate-form resin to remove negatively charged contaminants. Influence of Resin Properties on Contaminant Removal For inorganic contaminants, the functional group and polymer composition of each resin play the most important role on selectivity of each pair of resin to contaminant. A520E resin is the most selective resin for nitrate, bromide, and perchlorate based on the separation factor and isotherm parameter values. The functional group for A520E resin is triethylamine (R N + (CH2CH3)3), which has larger spacing between functional groups, and as a result is selective for monovalent ions over divalent ions (Clifford, 1988, Dennis, 1982). The wide spacing of functional groups on A520E resin can be easily accessed by nitrate, bromide, and perchlorate ions and favorably interact with charged sites. In addition, the hydrophobic character of the resin and contaminant ion is also an important consideration for selectivity. Resin hydrophobicity is influenced by the type of polymer composition and the chain length of the functional group (Baohua Gu, 2004, Zhong Xiong, 2007). Polystyrene resin has repeating benzene structure and is more hydrophobic than polyacrylic resin with repeating aliphatic carbonyl structure. The longer chain length of triethyl is more hydrophobic than trimethyl functional group. As such, resin with polystyrene composition is more selective for hydrophobic contaminant ion (Arup K. Sengupta, 1986, Zhong Xiong, 2007). In this work, the nitrate, bromide and perchlorate all shows the preferable selective towards to polystyrene, longer chain length functional group type resin, which corresponds to other research results (Gu et al., 2007, Hsu and Singer, 2010, Phetrak et al., 2014, Song et al., 2012, Xiong et al., 2007). Moreover, this preference also reflects the hydrophobic characteristics of ions. Respect to the 38

39 contaminant ions, the extent of hydrophobicity is only compared within inorganic ions, including nitrate, bromide, perchlorate, chromate, and sulfate. The anion structure, thermochemical radii, charge density and hydration energy, can be used to evaluate the extent of hydrophobicity. A general rule (Arup K. Sengupta, 1986), having the same charge, an anion with tetrahedral structure is less hydrated and more hydrophobic, followed by anion with trigonal structure, and finally, anion with spherical structure is the most hydrophilic. Therefore perchlorate is more hydrophobic than nitrate and bromide. Moreover, the thermochemical radii, charge density, and hydration energy can also reflect the extent of hydrophobicity (Gu et al., 2007). For same charge ion, the larger the thermochemical radii, the lower the charge density and hydration energy, which was characterized as more hydrophobic ion. Perchlorate ion had the largest radii among monovalent ions (Table 2-8), for the same charge, resulted in lowest charge density which corresponded to the lowest hydration energy of perchlorate. Nitrate had the larger thermochemical radii, lower charge density and hydration energy than bromide ion. Therefore, theoretically, he selectivity sequence for same resin should be: perchlorate > nitrate > bromide, which agrees the same result in this work. Other research also showed same selectivity sequence, perchlorate > nitrate > chloride, where chloride ion had very similar properties with bromide (Table 2-8) (Gu et al., 2007). Overall, resin with wide functional group spacing and hydrophobic polymer composition (e.g., A520E resin) ranks the highest selectivity with nitrate, bromide, and perchlorate. For conventional functional groups, resin with hydrophobic polymer composition (e.g., Marathon11, A300, and Dowex22 resins) is more selective than hydrophilic polymer composition (e.g., 39

40 IRA958 and A850 resin) for nitrate, bromide, and perchlorate. And perchlorate is more preferred selective than nitrate and bromide for the same resin. Marathon11 resin has the highest selectivity for chromate. Marathon11 resin has small spacing functional group trimethylamine (R N + (CH3)3) and a hydrophobic polystyrene composition. While, IRA958 resin demonstrates the highest selectivity to sulfate. IRA958 resin has trimethylamine functional group (R N + (CH3)3), which is also smaller spacing functional group, but polyacrylic, hydrophilic composition. Chromate (CrO4 2- ) is divalent ion at the conditions in this work (ph > 8, pka value for chromate species can be found in SI) and requires two charge sites to interact. The smaller spacing of functional groups allows for easier electrostatic interactions with the divalent ion than wider spacing functional groups (Clifford, 1988, Dennis, 1982). Similarly, sulfate is also divalent ion and selective for resin with smaller spacing functional groups. However, having same polyatomic tetrahedral structure and similar thermochemical radii, chromate has lower hydration energy comparing to sulfate ion (Table 2-8), which is characterized as relative hydrophobic than sulfate ion. Therefore, chromate is selective to a hydrophobic polystyrene resin (i.e. Marathon11), while sulfate is selective to the hydrophilic polyacrylic resin (i.e. IRA958). Same results also demonstrated from separation factor results from a Type-I, polystyrene resin (Edzwald and Tobiason, 2010), which showed higher selectivity of chromate than sulfate. Moreover, the relative hydrophilic characters of sulfate ion also discussed in other works (Arup K. Sengupta, 1986, Guter, 1982). The above explanation is discussed among contaminant ion with the same valence; it is complex to compare selectivity preference between monovalent and 40

41 divalent ion across the characteristic of charge and hydrophobic characteristics. Generally, the higher valence of ion is preferred with ion exchanger, which is explained with Donnan potential (Helfferich, 1962). The Donnan potential is described as the ability to attract ion into the ion exchanger in order to balance its tendency to the liquid phase. The Donnan potential is proportional to the ion valence. From selectivity sequence listed in the section of Selectivity of bicarboante- and chloride-form resins, however, chromate is generally preferred by ion exchanger than perchlorate, followed by sulfate and finally nitrate and bromide. Therefore, the synthesized investigation with both valence and hydrophobicity characteristics should be applied to evaluate selectivity preference among monovalent and divalent contaminant ions. For Suwannee River NOM, only IRA958 resin shows selective removal with separation factor > 1. The separation factor for the other resins with NOM was very low (Table 2-2). MIEX resin is widely studied for NOM removal and had a separation factor of 5.7 with Suwannee River fulvic acid (Singer, 2008). The reason for the low value was believed owing to the existence of competitive ions. After analyzing chloride concentration with chloride electrode, the decreasing chloride concentration with increasing resin dose was observed. This observation indicated that chloride was competing with NOM for exchange sites on the resin. The source of chloride is possible from adjusting ph with hydrochloric acid. In addition, it is also possible chloride and other ions were left over after NOM isolation process. Although chloride was likely competing with NOM, IRA958 resin still shows selective to NOM because its resin properties satisfied NOM uptake preference. Different than inorganic contaminants, pore structure and polymer matrix are important resin properties for selective removal of 41

42 NOM (Singer, 2008). Macroporous type resin has larger pore size than gel type resin. The larger pore size can enable easier access by large NOM molecules. In addition, resin with polyacrylic composition has higher water content and open structure than polystyrene resin, which reinforces selectivity for NOM. As such properties, IRA958 has the highest selectivity to NOM. Several other research also demonstrated the same result that NOM had a preference to macroporous, polyacrylic type resin (Hsu and Singer, 2010, Levitskaia et al., 2007, Phetrak et al., 2014). Differences in Selectivity Sequences Langmuir isotherm was originally derived for adsorption on activated carbon, however, it was also widely used on ion exchange system (Darracq et al., 2014, Ding et al., 2012, Liu et al., 2015, Song et al., 2012). A good data fitting was showed by using isotherm model for those studies. In this work, by comparing the selectivity sequence based on separation factor and Langmuir isotherm (i.e., ΔG 0 ), some differences were observed. Few papers had compared the ion exchange equilibrium method, including separation factor, selectivity coefficient and mass-action law, and the Langmuir isotherm model for ion exchange systems (Borba et al., 2011, Dron and Dodi, 2011a, b, Hekmatzadeh et al., 2012, Lin and Juang, 2005). A good data fitting was observed for all works using those two methods, however, several researchers had the recommendation on using ion exchange equilibrium calculation. The mass-action law, separation factor and selectivity coefficient are common methods for the study of ion exchange equilibrium. Different with separation factor equation, the selectivity coefficient has the ionic valences as exponents. Therefore, the value of separation factor was same as the selectivity coefficient when applying monovalent ion exchange system (Edzwald and Tobiason, 2010).The mass-action law equation was calculated 42

43 when activity correction was applied in selectivity coefficient equation (Helfferich, 1962). The equation of selectivity coefficient and mass-action law can be found in Appendix Equations A.1-A.5. Dron and Dodi (Dron and Dodi, 2011a, b) determined contaminant selectivity sequences by using both the selectivity coefficient and ΔG 0 calculated from Langmuir model. A good agreement was reported and the contaminant selectivity sequence was as follows: OH - < HCO3 - < Cl - < NO3 - < SO4 2-. Hekmatzadeh et al. (Hekmatzadeh et al., 2012) modeled nitrate exchange to chloride-form resin with mass-action law and Langmuir isotherm. The mass-action law equation in that paper was calculated without activity correction for monovalent system, in other word, the author (Hekmatzadeh et al., 2012) compared the method of separation factor with Langmuir isotherm. From comparing total resin capacity of these two methods, the author (Hekmatzadeh et al., 2012) concluded that the mass-action law was more effective and reliable to model ion exchange system than Langmuir isotherm. In a separate work, Borba et al. (Borba et al., 2011) commented on mass-action law and multicomponent Langmuir isotherm when modelling multicomponent ion exchange system. The author (Borba et al., 2011) also recommended using mass-action law because Langmuir isotherm was lacking ion exchange mechanism meaning. However, the authors also commented that Langmuir isotherm can be used for preliminary estimation because of the calculation simplicity, which can facilitate mass-action law method for the further ion exchange investigation. In a cation exchange study, another author (Lin and Juang, 2005) also recommend using the mass-action law for ion exchange system. 43

44 Finally, the calculation method can also contribute to slight differences in the selectivity sequence when using separation factor and adsorption isotherm. For separation factor calculation, one or two data points were used to approximate separation factor in this work. In contrast, the calculation of kl of Langmuir and Freundlich isotherm used all data points for nonlinearization. In summary, several previous authors (Borba et al., 2011, Hekmatzadeh et al., 2012, Lin and Juang, 2005) recommended using ion exchange equilibrium method, including separation factor and mass-action law to estimate resin selectivity for contaminants. Langmuir isotherm is robust on utilizing all data points and is calculation favorable, which can be used for preliminary estimation. However, care should be account for this method which is lacking of ion exchange meaning. Implications of Separation Factor on Resin Selection The resin selectivity for each contaminant is determined by the separation factor in order to narrow down the selection of resins for a target contaminant. The resin selectivity sequence can be used to consider competing influence of contaminants and the breakthrough time for ion exchange process. For example, sulfate should be taken into account if bicarbonate-from A300 resin is chosen for perchlorate removal (contaminant selectivity sequence in the section of Binary exchange plots and separation factors). Moreover, resin selection should also include the resin regeneration, especially for hazardous contaminants with high affinity to resins, such as perchlorate and hexavalent chromium. Logically, the higher affinity to resin, the more efficiency on contaminant removal, the less efficiency on resin regeneration. If brine disposal option is available for perchlorate removal, the low selective resin can be selected and resin can be used for several times; while, if brine disposal is not available, 44

45 the high selective resin can be used for once without brine regeneration (Edzwald and Tobiason, 2010). Similar logic for resin selection can also be applied on hexavalent chromium. In summary, the contaminant sequence for each bicarbonate-form resin in this work provides the preliminary resin screen for target contaminant removal. Care also needs to take into account the resin regeneration and bicarbonate brine disposal. Concluding Remarks Bicarbonate-form, strong-base, anion exchange resin demonstrates the potential to replace conventional chloride-form, strong-base, anion exchange resin as indicated by comparable separation factors for different contaminants. The most selective bicarbonate-form resin for each contaminant is recommended, and the important resin properties for contaminant selectivity were identified. For monovalent anions with relative hydrophobic character, triethylamine, polystyrene resin was most selective because of wider functional group spacing of triethylamine over trimethylamine and more hydrophobic character of polystyrene over polyacrylic. For divalent anions with relative hydrophobic character, trimethylamine, polystyrene resin was most selective because of closer functional group spacing of trimethylamine than triethylamine and more relative hydrophobic character of polystyrene than polyacrylic. For divalent anions with relative hydrophilic character, trimethylamine, polyacrylic resin was most selective because of closer functional group spacing and more hydrophilic polymer composition. The contaminant selectivity sequence of each resin was determined by using separation factor and Langmuir model parameter kl. The order was slightly different between the separation factor method and Langmuir isotherm model because of these methods use the data. 45

46 Table 2-1. Properties of each resin Resin Manufacture Resin type Functional group a. Polystyrene b. Polyacrlyic c. Macroporous d. gel Capacity (meq/ml) Polymer composition Pore structure Water content A520E Purolite Type I, strong base R N + (CH 2CH 3) PS a MP c 50-56% IRA958 Dow Type I, strong base R N+(CH 3) PA b MP 66-72% Marathon11 Dow Type I, strong base R N + (CH 3) PS G d 50-60% A850 Purolite Type I, strong base R N + (CH 3) PA G 70-82% Dowx22 Dow Type II, strong base R N + (CH 3) 2(CH 2OH) 1.2 PS MP 48~56 % A300 Purolite Type II, strong base R N + (CH 3) 2(C 2H 4OH) 1.4 PS G 45-51% 46

47 Table 2-2. Separation factor of selected bicarbonate-form resin Resin α NO3 /HCO 3 α Br /HCO 3 - α ClO4 /HCO 3 α CrO4 2 /HCO 3 α SO4 2 /HCO 3 α NOM/HCO3 A520E Marathon Dowex A IRA A

48 Table 2-3. Langmuir parameter of monovalent contaminants using bicarbonate-form anion exchange resin Resin Nitrate q 0 (mmol/g) Langmuir q 0 (mmol/g) manufacture kl (L/mmol) G 0 R 2 ARE (%) SSE (mmol 2 /g 2 ) A520E Marathon Dowex A IRA E04 A Resin Bromide q 0 (mmol/g) Langmuir q 0 (mmol/g) manufacture kl (L/mmol) G 0 R 2 ARE (%) SSE (mmol 2 /g 2 ) A520E Marathon Dowex A IRA E-03 A Resin Perchlorate q 0 (mmol/g) Langmuir q 0 (mmol/g) manufacture kl (L/mmol) G 0 R 2 ARE (%) SSE (mmol 2 /g 2 ) A520E Marathon Dowex A IRA A

49 Table 2-4. Freundlich parameter of monovalent contaminants using bicarbonate-form anion exchange resin. Resin Nitrate k F 1/n R 2 ARE (%) SSE (mmol 2 /g 2 ) (mmol 1-1/n L 1/n g -1 ) A520E Marathon Dowex A IRA A Resin Bromide k F 1/n R 2 ARE (%) SSE (mmol 2 /g 2 ) (mmol 1-1/n L 1/n g -1 ) A520E Marathon Dowex A E-04 IRA A Resin Perchlorate k F (mmol 1-1/n L 1/n g -1 ) 1/n R 2 ARE (%) SSE (mmol 2 /g 2 ) A520E Marathon Dowex A IRA A

50 Table 2-5. Langmuir parameter of divalent contaminants using bicarbonate-form anion exchange resin Resin Chromate q 0 (mmol/g) Langmuir q 0 (mmol/g) manufacture kl (L/mmol) G 0 R 2 ARE (%) SSE (mmol 2 /g 2 ) A520E Marathon Dowex A IRA A Resin Sulfate q 0 (mmol/g) Langmuir q 0 (mmol/g) manufacture kl (L/mmol) G 0 R 2 ARE (%) SSE (mmol 2 /g 2 ) A520E Marathon Dowex A IRA A

51 Table 2-6. Freundlich parameter of divalent contaminants using bicarbonate-form anion exchange resin Resin Chromate k F 1/n R 2 ARE (%) SSE (mmol 2 /g 2 ) (mmol 1-1/n L 1/n g -1 ) A520E Marathon Dowex A IRA A Resin Sulfate k F (mmol 1-1/n L 1/n g -1 ) 1/n R 2 ARE (%) SSE (mmol 2 /g 2 ) A520E Marathon Dowex A IRA A

52 Table 2-7. Thermochemical radius and hydration energy of anions investigated in ion exchange batch experiments Anion NO3 - Thermochemical radius (pm) 179 a 196 b Hydration energy, ΔG 0 (kj/mol) -314 b -320 c Br b -314 b -360 c ClO4-240 a,b -259 b -260 c CrO a 240 b b c SO a 230 b a Jenkins and Thakur, 1979 b Bianchi et al., 1997 c Smith, b c 52

53 YNOM A520E 0.4 Dowex A Marathon IRA A XNO XBr YBr- a Figure 2-1. Binary equilibrium plot for selected contaminants and selected resins. a) nitrate-bicarbonate binary equilibrium plot for six selected resins, b) bromidebicarbonate binary equilibrium plot for six selected resins, c) perchlorate- YNO3- c d b 1.0 YClO XClO4- e YCrO XCrO42- f YSO XSO XNOM 53

54 bicarbonate binary equilibrium plot for six selected resins, d) chromatebicarbonate binary equilibrium plot for six selected resins, e) sulfate binary equilibrium plot for six selected resins, f) NOM-bicarbonate binary equilibrium plot for six selected resins. 54

55 Figure 2-2. Experimental nitrate adsorption to bicarbonate-form resins and isotherm model fits to experimental data. 55

56 Figure 2-3. Experimental bromide adsorption to bicarbonate-form resins and isotherm model fits to experimental data. 56

57 Figure 2-4. Experimental perchlorate adsorption to bicarbonate-form resins and isotherm model fits to experimental data. 57

58 Figure 2-5. Experimental chromate adsorption to bicarbonate-form resins and isotherm model fits to experimental data. 58

59 Figure 2-6. Experimental sulfate adsorption to bicarbonate-form resins and isotherm model fits to experimental data. 59

60 % Removal % Removal % Removal % Removal % Removal % Removal a c e A520E Dowex22 A300 Marathon11 IRA958 A Dose % Dose % Dose % b d f Dose % Dose % A520 Dowex22 A300 Marathon11 Amberlite IRA958 A Dose % Figure 2-7. Contaminant removal as function of resin dose of selected resins. a) nitrate removal as function of resin dose of selected resins, b) bromide removal as function of resin dose of selected resins, c) perchlorate removal as function of resin dose of 60

61 selected resins, d) chromate removal as function of resin dose of selected resins. e. Sulfate removal as function of resin dose of selected resins, f) NOM removal as function of resin dose of selected resins. 61

62 CHAPTER 3 INTEGRATED BICARBONATE-FORM ION EXCHANGE TREATMENT AND REGENERATION FOR DOC REMOVAL: MODEL DEVELOPMENT AND PILOT PLAN STUDY The Process for DOC Removal Dissolved organic carbon (DOC) is a common constituent in drinking water supplies and is known as a disinfection by-product precursor, which contributes to the formation of carcinogenic compounds (Hrudey, 2009). Among the available technologies, ion exchange (IX) is an effective water treatment process to remove DOC (Bolto et al., 2004, Graf et al., 2014, Metcalfe et al., 2015). The mechanism of IX reaction is through the stoichiometric exchange of counterions on the resin with DOC in solution with chloride as the traditional counterion on anion exchange resin. After the resin is exhausted with contaminants, sodium chloride (NaCl) is used to recover the resin capacity and covert the resin back to the original chloride-form. However, the prevailing implementation of the traditional IX process is limited by the disposal of this highly concentrated NaCl solution (Jensen and Darby, 2016). There is also a concern about chloride ion release to finished water during treatment. This research, therefore, proposes an alternative option that uses bicarbonate ion as the counterion and sodium bicarbonate (NaHCO3) for resin regeneration. Using alternative-ion-form of resin and alternative type of salt for regeneration has advantages that are to minimize downstream environment impact and to benefit treated water. The highly concentrated NaHCO3 solution can be disposed to the sewer and used as an alkalinity source for the biological process during wastewater treatment. Produced by bicarbonate-form resin IX Reproduced with permission from Hu, Y, T.H. Boyer, Integrated bicarbonate-form ion exchange treatment and regeneration for DOC removal: Model development and pilot plant study. Water Research, 115, /j.watres Copyright 2017 Elsevier Ltd. 62

63 process, the finished water has lower chloride concentration and higher level of alkalinity, which may be beneficial to distribution system and premise plumbing corrosion. The elevated chloride ion in treated water produced by traditional IX may increase corrosion potential especially on lead corrosion (Edwards and Triantafyllidou, 2007, Ishii and Boyer, 2011, Nguyen et al., 2011b, Willison and Boyer, 2012). Several previous researchers have evaluated the treatment performance of bicarbonate-form resin and regeneration performance of NaHCO3. Hu et al. (Hu et al., 2016) conducted the selectivity analysis for six bicarbonate-form resins paired with six common drinking water contaminants and showed similar contaminant selectivity sequence between bicarbonate-form and chloride-form resins. In another research (Rokicki and Boyer, 2011), the selectivity analysis was also evaluated, which was performed under the conditions of mixed inorganic contaminants with absence and presence of DOC in raw water. Moreover, the same contaminant selectivity sequence was determined and a comparable removal performance was demonstrated for bicarbonate-form and chloride-form resin, even for multiple treatment and regeneration cycles (Rokicki and Boyer, 2011). In addition, 21 treatment and regeneration cycles of DOC removal tests were performed and showed 87% to 54% of DOC removal from the beginning of experiment to the end (Walker and Boyer, 2011) Less than 10% difference was observed comparing bicarbonate-form and chloride-form resin (Walker and Boyer, 2011). With respect to using alternative salts for regeneration, previous research (Maul et al., 2014) used NaHCO3 to regenerate nitrate-exhausted resin and showed a comparable regeneration efficiency with NaCl. Moreover, the higher regeneration performance was demonstrated for polyacrylic than polystyrene resin when using 63

64 bicarbonate salt. For the regeneration of DOC exhausted resin, the same researchers (Rokicki and Boyer, 2011, Walker and Boyer, 2011) showed less than 10% difference of using bicarbonate salt regeneration compared with using chloride salt for multiple (up to 21) treatment and regeneration cycles. In summary, previous research showed the viable option of using bicarbonate-form resin for treatment and bicarbonate salt for regeneration. Nevertheless, all previous research was performed in laboratory batch or jar tests, and there is no published information regarding the performance of bicarbonate-form resin for treatment and sodium bicarbonate for regeneration in realistic processes such as in a completely mixed flow reactor (CMFR). The CMFR IX process is often operated with magnetic ion exchange (MIEX) resin and sometimes referred to as the MIEX process. Other applications of operating resin in a suspended manner similar to the CMFR IX have also been developed such as suspended ion exchange (SIX) (Metcalfe et al., 2015). The CMFR IX process is different than the traditional IX column configuration because it uses resin in a well-mixing mode with resin recycle and partial resin regeneration. Due to the completely mixed operation, the CMFR IX process can be placed at the beginning of a treatment train and overcome the challenge of the sensitive response to the level of turbidity from source water, which was identified as a problem for column operation (Boyer and Singer, 2006). The benefits of operating the CMFR IX process as the first stage of treatment include decreasing coagulant chemical requirements and reducing membrane fouling (Boyer, 2015, Fabris et al., 2007, Fearing et al., 2004, Huang et al., 2012, Humbert et al., 2007, Singer and Bilyk, 2002, Singer et al., 2007, Warton et al., 2007). 64

65 Process models have been developed to understand the characteristics and operation of the CMFR IX process. The model developed for DOC removal in previous research (Boyer et al., 2010, Boyer et al., 2008a) consisted of macro-scale and microscale submodels: ideal CMFR model to describe macro-scale process behavior at treatment contactor with resin tracking described using Monte-Carlo method. The linear equilibrium isotherm and intra-particle diffusion model were assumed for micro-scale model. In another study targeting copper removal (Dahlke et al., 2006), the same macro-scale model was also applied. Differently, a linear-driven force model and the composite equilibrium isotherm was used for micro-scale model calculation. The predicted results in this research (Dahlke et al., 2006) showed less than 10% difference with the experimental data. However, the regeneration efficiency in above mentioned publications was assumed as 100% or not mentioned. Other studies have accounted for the regeneration efficiency coupled with treatment, and those that did were developed for IX column process (Gomes et al., 2001, Semmens et al., 1981, Zhang et al., 2015). With respect to pilot and full scale study of the CMFR IX process, Boyer and Singer investigated the influence of the operating parameters, i.e., resin concentration and regeneration ratio, on the removal of DOC and additional inorganic contaminants (Boyer and Singer 2006). Another pilot study has also conducted under the conditions of varying the operating parameters at multiple locations (Singer et al., 2007). Both research demonstrated the substantial DOC removal as well as inorganic ions removal to a certain degree. A large-scale plant study (Warton et al., 2007) compared DOC removal between the combined process of CMFR IX followed by coagulation and enhanced coagulation (EC) process only. The results (Warton et al., 2007) showed 65

66 more DOC removal by the combined process than the EC process only. Although previous researchers have evaluated the DOC treatment in a large scale study, there remains a gap in knowledge on the impact of varying salt concentration in the regeneration solution on overall DOC removal. The goal of this research was to evaluate the DOC removal of using integrated bicarbonate-form resin for treatment and NaHCO3 for regeneration (integrated bicarbonate-form IX) in CMFR process. The specific objectives were to: 1) develop a CMFR IX process model that coupled treatment and regeneration to predict the effluent DOC concentration, and 2) conduct a pilot plant study of using integrated bicarbonateform IX to investigate the DOC removal under the conditions of varying salt concentration in the regeneration solution, and to compare with the integrated chlorideform resin for treatment and NaCl for regeneration (integrated chloride-form IX). The required equilibrium and kinetic parameters for the process model were obtained from laboratory experiments. The DOC removal data from the pilot plant study were used to evaluate the influence of regeneration conditions on treatment performance, as well as to validate the process model. Process Description and Model Formulation Process Description The typical full-scale implementation of the CMFR IX process and pilot setup in this research is shown in Figure 3-1 a and b. The raw water enters a well-mixing IX treatment contactor where contaminant removal takes place via stoichiometric exchange on the resin. Treated water and resin carried over then enter the settler where the water exits the process, and the majority of settled resin is recycled back to the IX treatment contactor. At the same time, a side stream of settled resin is continuously 66

67 collected for periodic regeneration as well as the same mass of fresh regenerated resin is continuously added to the IX treatment contactor (Figure 3-1a). In this approach, a relative constant resin capacity is maintained in the IX treatment contactor, which enables a steady-state operation. In order to minimize the complexity of the pilot plant version, the resin collected for regeneration was removed from the IX treatment contactor (Figure 3-1b). The rate of this side stream was determined according to the rate of freshly regenerated resin entering the contactor; therefore, the location of resin collection for regeneration is not expected to impact the process performance. The resin used in this research was MIEX anion exchange resin manufactured by IXOM. Compared with traditional resin, which has a particle size of cm, the MIEX resin has a smaller particle size of cm (Graf et al., 2014). The smaller particle size of MIEX resin enables a faster rate of DOC uptake, and the magnetic iron oxide incorporated into the resin is intended to aid settling of the resin in the CMFR configuration. Model Formulation The concept of the integrated treatment and regeneration framework is described in Figure 3-2. The DOC concentration in the fresh brine and DOC-loaded resin are initial conditions for the regeneration model. After simulating the regeneration model for predetermined time, the result of DOC-loaded resin was updated as fresh regenerated resin. Then the DOC concentration in raw water and fresh regenerated resin was used as initial condition for the treatment model. In this approach, the developed process model can be used to couple the contaminant treatment and resin regeneration. In addition, the DOC concentration in the fresh regenerated resin is influenced by the salt concentration in regeneration solution, thus, the regeneration performance is ultimately 67

68 linked to the DOC concentration in treated water. The developed model was solved dynamically with MATLAB ODE 15s solver. The algorithm of this solver is based on backward differentiation formulas ( The mass balance analysis was conducted for liquid and resin phase during treatment and regeneration. The process model was developed according to the mass balance analysis, and the assumptions included neglecting make-up resin addition and resin loss due to attrition; modeling the settler as point tank, which only considers mass distribution; linear equilibrium relationship between contaminant concentration in the liquid and solid phase; spherical resin particles assuming an average particle size and age distribution. In the IX treatment contactor, the mass balance analysis of macro-scale DOC in the liquid phase and resin phase is described using the ideal CMFR model (Equation 3-1 and 3-2). There are removable and non-removable portions of DOC during IX treatment. In the following text, the DOC in liquid phase all refer to the removable portion for model calculation. c in (mg/l) is the DOC concentration in raw water and c out (mg/l) is the DOC concentration in treated water. q out (mg/ml resin) is the DOC concentration in resin phase leaving the IX treatment contactor and q f (mg/ml resin) is the DOC concentration in fresh regenerated resin phase pumped from resin storage tank. τ H (min) is the hydraulic residence time and is calculated by dividing the volume of IX treatment contactor by the raw water flow rate. τ s is the average solids (i.e., resin) residence time (SRT) and is calculated from dividing τ H (min) by the resin regeneration ratio (f R ) (Equation 3-3), as described in the previous research (Boyer et al., 2010). In 68

69 the micro-scale model, the linear-driving force approximation is used to describe the mass transfer rate in liquid and resin phase(dahlke et al., 2006, Zhang et al., 2015). k L1 (cm/min) is the lumped mass transfer coefficient, c (mg/l) is the DOC concentration in the liquid phase at equilibrium. The determination of c is according to Equation 3-4 where K 1 (L liquid/ml resin) is the adsorption equilibrium parameter and was chosen to describe the local equilibrium in liquid phase with resin phase. α IX (ml resin/l) is the resin concentration in IX treatment contactor. Rp (cm) is the average resin radius, and cm was selected in this research. The DOC-loaded resin is regenerated in a batch reactor, accordingly, the ideal batch reactor model is selected. The rate of mass transfer is modeled as the same method as the treatment model, but DOC is exchanged from resin phase to liquid phase (Equation 3-5). c reg (mg/l) is the DOC concentration in liquid phase after regeneration. k L2 (cm/min) is the lumped mass transfer coefficient during regeneration and α reg (ml resin/l) is the resin concentration in batch regeneration reactor. The same mass transfer coefficient in the treatment model was assumed for regeneration, and c reg (mg/l), which is the DOC concentration in brine at equilibrium, is determined from laboratory experiments. dc out dt = 1 τ H (c in c out ) 3α IX Rp k L1(c out c ) (3-1) dq out dt = 1 τ s (q f q out ) + 3 R p k L1 (c out c ) (3-2) 1 τ s = f R τ H (3-3) q out = K 1 c (3-4) 69

70 dc reg dt = 3α reg k Rp L2(c reg c reg ) (3-5) Materials and Methods Materials The raw water was from the groundwater supply to the Cedar Key Water and Sewer District drinking water treatment plant located in Cedar Key, Florida, USA. The raw groundwater was transported to the laboratory, stored at 4 C, and was filtered through 0.45 µm filter within 24 h of collection. The 8.6% (w/v) NaHCO3 and 6% (w/v) NaCl regeneration solution was prepared in the laboratory by dissolving the appropriate salt in deionized (DI) water. The 6% (w/v) NaCl solution was selected according to the same equivalent strength as the 8.6% (w/v) NaHCO3 solution. The fresh virgin resin was washed with DI water and conditioned into bicarbonate-form or chloride-form. The purpose of conditioning the chloride-form resin, which was the original form, was to follow the same procedure as the bicarbonate-form resin. The conditioning step was followed with previous literature (Hu et al., 2016, Rokicki and Boyer, 2011, Walker and Boyer, 2011), which entails mixing highly concentrated salt solution (50 times equivalent capacity of the resin) with the virgin resin. The DOC-loaded resin was collected directly from the spent resin storage tank at the Cedar Key water treatment plant and was transported to the laboratory and rinsed with DI water. Experimental Methods Laboratory batch tests were performed to determine the isotherm and kinetic parameters and DOC concentration in the liquid phase after regeneration. The detailed experimental plan is summarized in Table

71 Batch equilibrium tests The batch equilibrium tests were performed to obtain isotherm coefficient during treatment and determine the DOC concentration in the liquid phase after regeneration (Table 3-1, equilibrium test 1 and 2). In equilibrium test 1, The tests were performed with bicarbonate-form and chloride-form resins, respectively. In equilibrium test 1, Cedar Key raw water was mixed with varying dry resin doses, namely 5, 10, 20, 40, 80, 150 mg, in a 125 ml amber glass bottle. In equilibrium test 2, the varying amount of wet DOC-loaded resin were mixed with 8.6 % (w/v) NaHCO3 and 6 % (w/v) NaCl, respectively. The loaded wet resin was dosed at 1, 5, 10, 15, 20, 25 ml. The tests were conducted using a shaker table at 250 rpm for 24 h at ambient temperature. Two control samples were prepared for each test, and three samples were prepared at each resin dose. After equilibrium, all samples were filtered with 0.45 μm filters and stored at 4 C. Batch kinetic tests The Batch kinetic tests were conducted to obtain mass transfer coefficient during treatment and regeneration (Table 3-1, kinetic tests 1 and 2). In kinetic test 1, 1 ml/l wet virgin resin was dosed in 500 ml beaker and mixed with raw water on a jar tester. The kinetic experiment was performed in three parallel tests and samples were collected via syringe at 2.5, 5, 10, 20, 40, 60, 90, 180 min. In kinetic test 2, 10 ml/l DOC-loaded resin was dosed in 500 ml beaker, and was mixed with concentrated regeneration solution of 8.6% (w/v) NaHCO3 and 6% (w/v) NaCl, respectively. According to the operation at the Cedar Key drinking water treatment plant, the regeneration period is operated for 20 min, therefore, the time interval for regeneration was chosen at 2.5, 5, 10, 20, 40 min. All samples were filtered through 0.45 μm filter and stored at 4 C. 71

72 Pilot plant tests The detailed pilot plant tests are summarized in Table 3-2. The CMFR IX pilot system was assembled by the authors (Appendix Figure B-1). Before the operation, the fresh regenerated resin was stored in the resin storage tank that was continuously mixed. At the start-up phase, the virgin resin was dosed in the IX treatment contactor, and the fresh regenerated resin was continuously pumped from the storage tank into the IX treatment contactor, whereas, the same amount of the mixed resin in the IX treatment contactor was collected in a small funnel settler. The same operational approach was applied for both integrated bicarbonate-form IX and integrated chloridefrom IX. The operational parameters were selected to maintain a stable hydraulic operation and achieve the treatment goal. For all pilot plant tests, an 8 min hydraulic residence time was selected, and the typical hydraulic residence time in the CMFR IX process is min. The selection of 14% for regeneration ratio was intended to provide a SRT equal to approximately 1 h according to the Equation The 1 h SRT was selected due to the logistics of sample collection. The resin concentration of 15 ml/l and 20 ml/l in the IX treatment contactor was tested in preliminary experiments, and the 20 ml/l was selected to enable a greater than 60% DOC removal. Accordingly, the effective resin dose in this research was 3.8 ml/l as defined in previous research (Boyer and Singer, 2006). Each pilot experiment was performed for 6 h to achieve steady-state operation. Previous research showed that steady-state operation could be reached after three SRT (Boyer et al., 2008a). The fresh regenerated resin was prepared in the laboratory by mixing 150 ml DOC-loaded resin (from Cedar Key water treatment plant) with 750 ml regeneration solution, i.e., 8.6% (w/v) NaHCO3 or 6% (w/v) 72

73 NaCl, in jar tester for 20 min. The hourly sample was collected after two hours of operation. The resin concentration in IX treatment contactor was measured every hour by syringe to check that it was operating in a constant condition. Analytical Methods DOC was analyzed using Total Organic Carbon (TOC) analyzer (Shimadzu TOC- VCPH) following Standard Method 5310 B (Eaton et al., 2005). Chloride, sulfate, and nitrate were analyzed by ion chromatography (Dionex ICS-5000) following U.S. EPA method Ultraviolet absorbance at 254 nm (UV254) was measured on a Hitachi U-2900 spectrophotometer following Standard Method 5310 B (Eaton et al., 2005). ph was measured with portable ph (Hanna phep 4 probe), conductivity was measured with a conductivity meter (Oakton ECTestr 11 probe), the dissolved iron was measured with HACH iron test kit (DR 850 colorimeter), and the alkalinity was measured following Standard Method 2320 (Eaton et al., 2005). For accuracy check, the external standards for TOC (Ricca) and inorganic anions (Dionex seven anion standard) were analyzed. In addition, the calibration standard solutions were measured again at the end of each run, and one selected calibration solution was analyzed as an unknown sample for every samples as accuracy check. All samples were analyzed in duplicate with difference less than 10%. The relative difference of calibration standard and the external standard was less than 10% at concentration > 1 mg/l and less than 20% at concentration < 1 mg/l. Results and Discussion Raw Water Characteristics The Cedar Key raw water contained relatively high concentration of DOC, iron, and alkalinity. The total DOC concentration in raw water was measured at the level of 73

74 7 9 mg/l for laboratory experiments, which were conducted between October to December 2015, while DOC was 6 7 mg/l for pilot plant study, which was conducted between March to April The non-removable fraction of total DOC is discussed in the section of Pilot plant results. The alkalinity was approximately 250 mg/l as CaCO3, and the range of dissolved iron was mg/l. In order to minimize the effect of iron fouling on resin, the raw water was treated with permanganate addition at the well and settling for 24 h in a 1000 gal (3785 L) equalization tank before pumping to the pilot system. The ph was measured in the range of 7 8 and conductivity was µs/cm. The UV254 in raw water was approximately /cm, which corresponded to specific UV absorbance (SUVA) of 2.5 L/mg m. Compared with the reported SUVA from a variety of water sources (e.g., L/mg m) (Boyer and Singer 2006, Warton et al. 2007), the raw water in this research was characterized as low to moderate range (i.e., < 3 L/mg m) (Boyer and Singer 2005) meaning the DOC was composed of a mixture of humic material. The selected inorganic species were also analyzed, and the concentration of chloride, sulfate, and nitrate was 11 mg/l, 3.5 mg/l, and 1 mg/l as N, respectively. This low concentration of inorganic anions indicated a minimal competition with DOC for available sites on resin phase. Laboratory Batch Experiments The results of batch studies are summarized in Table 3-3. Overall, the results showed a comparable equilibrium and kinetic parameters between bicarbonate-form and chloride-form resin for treatment, as well as a comparable kinetic parameter between using bicarbonate salt and chloride salt for regeneration. 74

75 Figure 3-3 shows the results of the equilibrium isotherm during treatment for bicarbonate-form and chloride-form resin, respectively. The x-axis is the DOC concentration in the liquid phase, and the y-axis is the DOC concentration in the resin phase. The linear equilibrium isotherm was selected to fit the experimental data as described in previous research (Boyer et al. 2008). The linear isotherm coefficient was obtained by estimating the slop of isotherm. The non-removable DOC concentration (C N ) was estimated by extending the isotherm line to the x-axis, and the non-removable fraction of DOC (f N ) was calculated through dividing C N by DOC concentration in the raw water. The reason that there is part of DOC cannot be removed by IX treatment is due to size exclusion and charge characteristics of the DOC (Boyer et al., 2008a, Qi et al., 2012, Wolska, 2015). The DOC concentration in the liquid phase after regeneration was directly obtained from experimental data, which was generated by mixing DOC-loaded resin with regeneration solution. The purpose of estimating DOC in the liquid phase after regeneration was in order to calculate the DOC concentration remaining in the resin, and used as input as q f in section of Model formulation. The DOC remaining in the resin phase is plotted with the salt concentration in regeneration solution in Figure B-2. The DOC released from resin is plotted in Figure 3-4, and shows it increased as the salt concentration increased in the regeneration solution. In addition, the difference of the DOC released between bicarbonate and chloride salts regeneration was very similar at the lowest salt concentration, while the difference was greater at the higher salt concentration. For example, the DOC concentration was 100 mg/l and 120 mg/l in 1.4% (w/v) NaHCO3 and 1% (w/v) NaCl, respectively, whereas the DOC concentration 75

76 was 560 mg/l and 730 mg/l in 8.6% (w/v) NaHCO3 and 6% (w/v) NaCl, respectively. Theoretically, the same extent of DOC displacement should be observed because both regeneration solutions were prepared based on the same equivalent (molar) concentration. The salt concentration in the highest regeneration solution (Figure 3-4) was equaled to 10 the equivalent capacity of dosed resin, which is the same value used in previous research that reported similar DOC displacement between NaCl and NaHCO3 over multiple regeneration cycles(rokicki and Boyer, 2011, Walker and Boyer, 2011). However, the opposite phenomenon observed in this research raises a question if additional parameters were playing a role during regeneration. It is speculated that the ionic strength of the liquid phase influenced the Donnan potential and resulted in this performance difference. The Donnan potential is an electric potential that attracts ions and balances their tendency to diffuse from resin phase to liquid phase (Helfferich, 1962), and is influenced by the ionic strength of solution and the ionic valance. The Donnan potential is higher in dilute ionic concentration in the liquid phase, and the force acts on ions is for a higher valence of ions. Although at the same equivalent ratio, i.e., 10, the salt concentration in previous researches are 20.8 mmol/l (Rokicki and Boyer, 2011) and 52 mmol/l (Walker and Boyer, 2011), which is more dilute than 1026 mmol/l in this study, hence, a stronger effect of Donnan potential can possibly attract bicarbonate ion from regeneration solution to resin phase. The mass transfer coefficient during treatment was estimated by fitting first-order equation with experimental data (Figure 3-5). The data was generated by mixing raw water with the virgin bicarbonate-form and chloride-form resin, respectively, and the samples were collected at certain time interval. The DOC concentration in the liquid 76

77 phase decreased from time zero to the end of operation due to exchange with counterions (i.e., HCO3 - or Cl - ). Compared with the estimated parameter in the previous research for three sources of raw water (Boyer et al., 2008a) (Boyer et al. 2008), the kinetic coefficient in this research is in the same magnitude with estimation of raw water from Durham (Appendix Table B-1). Moreover, the raw water characteristics play an important role on kinetic parameter estimation for DOC treatment, as demonstrated the change of value correlated to the source of raw water (Appendix Table B-1). With respect to the estimation of the mass transfer coefficient during regeneration, the same equation was used to fit experimental data (Figure 3-6). The lab data was generated through mixing the DOC-loaded resin with the bicarbonate and chloride salt solution, respectively, and the samples were collected at certain time interval. The estimation mass transfer coefficient showed the comparable results between treatment and regeneration, as shown in Table 3-3. There is limited published information on the estimation of kinetics parameter during regeneration. One study assumed the same value of the kinetic parameter during treatment and regeneration in column process and demonstrated a comparable modeling prediction with experimental data (Zhang et al., 2015). Pilot Plant Results The results of DOC and UV254 in treated water are summarized in Table 3-4 for pilot studies. The ph and conductivity in the raw water and treated water were relatively unchanged, therefore, the data was not shown. The tests were designed to evaluate the extent of regeneration on DOC removal, therefore, the salt concentration in regeneration solution is the only varying parameter. Run 1 was the control test which used virgin chloride-form resin in the IX treatment contactor and resin storage tank. 77

78 Using chloride salt to prepare fresh regenerated resin in runs 2 and 4 was intended to provide a baseline test for runs 3 and 5. Greater than 61% DOC removal and 78% reduction in UV254 was achieved, which was comparable with previous pilot and full scale studies, e.g., 35 73% DOC removal and 55 88% UV254 reduction (Boyer and Singer 2006, Singer et al. 2007, Warton et al. 2007). In terms of DOC treatment, run 1 demonstrated the highest DOC removal (74%), followed by runs 4 (71%) and 5 (64%), and runs 2 (64%) and 3 (61%). For run 1, the virgin chloride-form resin was dosed in IX treatment contactor at start-up, and the same fresh resin was continuously pumped from the resin storage tank to the IX treatment contactor (Figure 3-1b). In this operating mode, the highest resin capacity was maintained in the IX treatment contactor, and as a result, the highest DOC removal was achieved. For runs 4 and 5 the resin in the storage tank was prepared through regenerating DOC-loaded resin with 6% (w/v) NaCl and 8.6% (w/v) NaHCO3, respectively. Therefore, the prepared fresh regenerated resin in the storage tank was less fresh than virgin fresh resin, which decreased the resin capacity in the IX treatment contactor. The same reason led to the least DOC removal performance for runs 2 and 3, which used 1% (w/v) NaCl and 1.4% (w/v) NaHCO3 to prepare the fresh regenerated resin. Across comparing the bicarbonate-form with chloride-form IX, a similar DOC removal performance was observed. In terms of UV254, the removal performance was relatively stable and was not significantly influenced by the varying concentration in regeneration solution. This indicated that the maximum UV254 was already achieved with the least effective IX operation condition (i.e., runs 2 and 3), and that the IX process is more preferable for removal of UV-absorbing substances than DOC. 78

79 Process Model Calculation and Validation The process model was developed to predict the DOC concentration in treated water leaving the CMFR IX process under the conditions of varying salt concentration in regeneration solution. The model inputs included the raw water characteristics, equilibrium and kinetic parameters, and operating conditions. As stated in the section of Model formulation, the developed model was calculated using removable part of DOC. The non-removable portion of DOC was added to the removable DOC and plotted in Figures 3-7 and 3-8. The non-removable DOC was determined according to the equilibrium test that using virgin chloride-form resin, i.e., 1.7 mg/l. The calculated nonremovable DOC for bicarbonate-form resin was 2.3 mg/l, which was higher than the DOC concentration in effluent water (Table 3-4). This result is most likely due to the variability of raw water, for example, the Cedar Key water plant alternated between two groundwater wells over the course of this research. In previous research also showed that the non-removable part of DOC was changed with the source of raw water (Boyer et al., 2008a). Considering the possible variations, the estimation from the chloride-form equilibrium test was therefore selected, and the results of using the bicarbonate-form equilibrium test for model predictions were summarized in Appendix B, Table B-2. The process model validation was conducted through comparing the model results with five pilot plant runs in terms of percentage removal (Figure 3-7). The model predicted, and pilot results are summarized in Table 3-4. The effluent DOC concentration was calculated by averaging the DOC concentration of the last three samples collected from 3 to 6 SRT and was compared with the average DOC concentration collected at the same time interval for pilot studies. The process model results showed a less than 10% difference of DOC removal percentage and also less 79

80 than 10% of DOC absolute value of concentration in treated water. This agreement demonstrated that the model responded to the varying regeneration conditions. There was only one exception that run 1 had 18% difference in terms of DOC concentration in treated water. It is possible that the developed model underestimated the actual process treatment efficiency when using virgin resin. The reason is speculated that there is a limitation of predicting removable DOC in treated water that is as low or close to zero. Pilot plant data in run 1 and linear equilibrium isotherm data (Figure 3-1) showed that the DOC in treated water is very close to the estimated non-removable DOC concentration, which indicated that the removable DOC in treated water is very low. In addition, the process model was developed in a dynamic condition to investigate the DOC in treated water as a function of time. Figure 3-8 shows the model prediction in the liquid phase for runs 4 and 5, and was also compared with the pilot plant data. The model was calculated for six SRT. At time zero, the DOC concentration in the liquid phase was equal to the raw water. The decrease in DOC concentration to the lowest point at the start-up phase was because the resin was relatively fresh, and as a result, there were more sites available to exchange with negatively charged DOC. As time progresses, the resin gradually becomes saturated with DOC. Simultaneously, the continuous addition of resin from the storage tank and the removal of used resin from the treatment contactor enable a constant resin capacity to achieve a steady-state DOC concentration. Concluding Remarks Overall, this research aims to evaluate the performance of bicarbonate-form resin for treatment, and bicarbonate salt for regeneration in the CMFR IX process. The process model was developed to predict DOC concentration, and was validated with 80

81 pilot tests. In addition, the pilot tests performed a comprehensive evaluation of this alternative-ion-form resin for treatment and alterative salt for regeneration. The pilot plant study showed over 60% DOC removal and 80% UV254 reduction for integrated bicarbonate-form IX, which is comparable to integrated chloride-form IX operation. The DOC removal performance was influenced by the salt concentration used in regeneration solution. The treatment performance was improved by using higher salt concentration because less DOC remained on the resin, which resulted in an increased resin capacity in the IX treatment contactor. The process model was able to couple treatment and regeneration with less than 10% prediction difference in terms of DOC percentage removal. The results from process model responded to the change of salt concentration used for regeneration. The mass transfer coefficient was similar for IX process during treatment and regeneration. The linear isotherm showed the bicarbonate-form and chloride-form resin had similar affinity for the DOC from the same source of groundwater in this work. From the regeneration batch studies, it showed that bicarbonate salt has lower regeneration performance than chloride salt, and the difference becomes more apparent at the higher salt concentration. 81

82 Table 3-1. Lab experiment plan Equilibrium test 1. Treatment IX (Cedar key raw water with virgin resins) 2. Regeneration IX (Regeneration solution with DOC-loaded resins) Kinetic test 1. Treatment IX (Cedar key raw water with virgin resins) 2. Regeneration IX (Regeneration solution with DOC-loaded resins) a. bicarbonate-form resin b. chloride-form resin a. 8.6% (w/v) NaHCO3 b. 6% (w/v) NaCl a. bicarbonate-form resin Resin dose = 1 (ml/l) b. chloride-form resin Resin dose = 1 (ml/l) a. 8.6% (w/v) NaHCO3 Resin dose = 10 (ml/l) b. 6% (w/v) NaCl Resin dose = 10 (ml/l) Test condition 5, 10, 20, 40, 80,150 mg of resins plus 125 ml water 1, 5, 10, 15, 20, 25 ml of resins plus 100 ml regeneration solution Sampled after 2.5, 5, 10, 20, 40, 60, 90,180 min Sampled after 2.5, 5, 10, 20, 40 min 82

83 Table 3-2. Pilot plant test plan Run (#) Salt Salt concentration in regeneration solution (w/v %) 1 None NaCl 1 3 NaHCO NaCl 6 5 NaHCO

84 Table 3-3. Summary of equilibrium and kinetic parameters Equilibrium test 1. IX reaction during treatment (Cedar key raw water with virgin resins) Kinetic test 1. IX reaction during treatment (Cedar key raw water with virgin resins) 2. IX reaction during regeneration (Regenerants with DOC-loaded resins) Equilibrium coefficient (L/g) a. bicarbonate-form resin 49.0 b. chloride-form resin 50.9 Mass transfer coefficient (cm/min) a. bicarbonate-form resin b. chloride-form resin a. 8.6% (w/v) NaHCO3 solution b. 6% (w/v) NaCl solution

85 Table 3-4. Summary of pilot plant data and model prediction Run # Operation condition Raw. DOC (mg/l) Pilot. effl. DOC (mg/l) Model.effl. DOC (mg/l) Percent difference between effl. DOC Raw.UV254 (1/cm) Effl. UV254 (1/cm) 1 Virgin fresh resin % % 2 1% (w/v) NaCl reg % % 3 1.4% (w/v) NaHCO3 reg % % 4 6% (w/v) NaCl reg % % 5 8.6% (w/v) NaHCO3 reg % % UV254 percentage of removal (%) 85

86 Figure 3-1. The system diagram of CMFR type ion exchange process. a) the system diagram of CMFR type ion exchange process, b) the system diagram of pilot plant set up. 86

87 Figure 3-2. The concept of integrated treatment and regeneration ion exchange process. 87

88 qe (mg/g) Bicarbonate-form resin Chloride-form resin Linear (Bicarbonateform resin) Linear (Chloride-form resin) y = x R² = y = x R² = Ce (mg/l) Figure 3-3. Isotherm carry out by virgin resin with Cedar Key raw water (data symbols show mean value ± one standard deviation for triplicate samples). 88

89 DOC concentration in liquid phase (mg/l) NaHCO3 regenerant NaCl regenerant 1.4 (1) 2.9 (2) 4.3 (3) 5.7 (4) 7.2 (5) 8.6 (6) Salt concentration in regeneration solution % (w/v) Figure 3-4. DOC concentration in liquid phase as varying the salt concentration in regeneration solution (NaHCO3, NaCl) (data symbols show mean value ± one standard deviation for triplicate samples). 89

90 DOC Concentration in liquid phase (mg/l as C) Bicarbonate-form resin data Bicarbonate-form resin model fitting Chloride-form resin data Chloride-form resin model fitting time (min) Figure 3-5. Adsorption kinetic carry out by virgin resins with Cedar Key raw water (data symbols show mean value ± one standard deviation for samples in three parallel test). 90

91 DOC concentration in liquid phase (mg/l) NaHCO3 8.6% (w/v) regeneration data NaHCO3 8.6% (w/v) regeneration model fitting NaCl 6% (w/v) regeneration data NaCl 6% (w/v) regeneration model fitting Time (min) Figure 3-6. Adsorption kinetic carry out by DOC-loaded resins with regeneration solution (data symbols show mean value ± one standard deviation for samples in three parallel test). 91

92 DOC removal percentage (%) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Pilot plant data Model prediction Run (#) Figure 3-7. The DOC removal percentage of pilot plant data and model prediction (data symbol shows mean of duplicate samples). 92

93 Figure 3-8. Dynamic model predictions for runs 4 and 5. a) dynamic model prediction and fit to pilot test run 4, b) dynamic model prediction and fit to pilot test run 5. 93

94 CHAPTER 4 SIMUTANEOUS MULTIPLE CONTAMINANTS REMOVAL IN A COMBINED ION EXCHANGE PROCESS FOR SMALL WATER SYSTEMS The Concept of Combined Ion Exchange Operation Multiple contaminants removal can pose a treatment challenge for small public water systems (PWS) that lack of technical, managerial, and financial capacity. A recent report summarized the most common compliance violations by small PWS in the U.S. included the nitrate maximum contaminant level (MCL), and Stage 1 Disinfectants/Disinfection Byproducts Rule (Oxenford and Barrett, 2016). Nitrate is regulated in the U.S. and elsewhere because of acute toxicity concerns (Knobeloch et al., 2000). Disinfection byproducts are formed when dissolved organic carbon (DOC) react with disinfection reagents during disinfection process (Hua and Reckhow, 2007) and are of human health concern due to carcinogenicity and other adverse health impact (Hrudey, 2009). Moreover, there are unregulated contaminants that are being considered for regulation. For example, strontium is currently listed in Regulatory Determination 3 and may be regulated in the future. In addition, the presence of problematic constituents can also require treatment by small PWS. For example, calcium and magnesium ions (typically referred to as hardness) are present in many water supplies, and high concentration of hardness can cause operational problems at the water plant, e.g., membrane fouling (Shirazi et al., 2010), and scaling in residential water heaters and industrial equipment. Due to the possible presence of variety of contaminants and problematic ions in source of raw water, small PWS requires an effective process to handle those contaminants or constituents to produce high quality water for the community. 94

95 The ion exchange (IX) process is a suitable technology for small PWS to remove a wide range of negatively charged and positively charged contaminants, such as nitrate, DOC, calcium, and strontium (Boyer and Singer, 2006, Foster et al., 2017, Hu et al., 2016). Typically, anion exchange resin (AER) and cation exchange resin (CER) are operated in separate vessels in series for the multiple contaminants removal. After exhausted with contaminant ions, the multiple salt solutions, typically sodium chloride (NaCl), are used for AER and CER regeneration independently. Such independent operation of AER and CER requires larger footprint and generates multiple waste streams. The disposal of waste brine, i.e., NaCl, is the major drawback of the IX implementation for small PWS, especially which located away from the coast (Jensen and Darby, 2016, M.Canedo-Arguelles, 2013, McTigue and Cornwell, 2009). Alternatively, the combined ion exchange (CIX) process presented in this research is intended to upgrade the conventional IX to benefit the small PWS on mulitple contaminants removal. The CIX process uses the AER and CER in the same veseel and regenerates the exhausted combined resins of AER and CER simutanously. The signle process unit of CIX operation can be used to handle variety contaminants in raw water. Moreover, compared with the independent regeneration of AER and CER in conventional IX, the combined resins regeneration effciently utilizes both anionic and cationic ions, which can reduce the chemcial usage and waste production. In addition, this research proposes the CIX in alternative-ion-form, which operates bicarbonate-form AER and potassium-form CER in the same vessel and uses potassium bicarbonate (KHCO3) for combined resins regeneration, to further minimze the disposal issue of waste brine. KHCO3 can be disposed to the sewer system where the high concentration 95

96 of bicarboante can be used as an alkalinity source for nitrification process, and potassium ion is an important factor for enhaced phosphrous removal in wastewater treatment plant (Choi et al., 2011, Spellman, 2008). Moroever, releasing bicarbonate ion in finished water can benefit corossion control in distribution system (Edwards and Triantafyllidou, 2007). Previous research on the CIX process has mostly focused on DOC and hardness removal (Apell and Boyer, 2010, Arias-Paic et al., 2016, Comstock and Boyer, 2014, Locke and Smith, 2016). The research conducted by (Apell and Boyer, 2010) had showed 70% DOC and over 55% hardness removal from real groundwater using the approach of single-loading jar tests. Another research (Comstock and Boyer, 2014) utilized CIX process for DOC and hardness removal from real groundwater and nanofiltration concentrate. The multiple-loading jar tests showed 76% DOC and 97% hardness removal for groundwater and 83% DOC and 86% hardness removal for nanofiltration concentrate. Research by (Arias-Paic et al., 2016) performed CIX operation in the laboratory batch studies and pilot-scale tests in the fluidized configuration with majority resin recycle and partial resin regeneration. The pilot tests showed over 75% DOC and 50% hardness removal. In addition, the enhanced removal of DOC was observed during CIX process comparing with AER treatment only (Apell and Boyer, 2010, Arias-Paic et al., 2016). The possible reason for this phenomenon was explained that because there was possible removing of the complexes formed with calcium and dissolved organic matter (DOM) lead to the additional DOC treatment. Finally, a full-scale CIX was documented which is located at the City of Bunnell, Florida, USA. The CIX was operated in the fluidized configuration with majority resin recycle and 96

97 partial resin regeneration (Locke and Smith, 2016). The full-scale CIX process achieved 78% DOC removal and 46% hardness removal, which addressed previous problems of elevated DBP formation. As illustrated here, there are only a few previous studies (Apell and Boyer, 2010, Arias-Paic et al., 2016, Comstock and Boyer, 2014, Locke and Smith, 2016) on the CIX process in bench- and large-scale, and all of the studies have focused on DOC and hardness removal during chloride-form AER and sodium-form CER operation and NaCl regeneration. Moreover, the previous studies only focused on the magnetic AER, i.e., MIEX, and magnetic and non-magnetic CER during CIX operation. It is also interesting to evaluate the CIX operation using the conventional resins, typically operated in the column systems. As such, there remains a gap in knowledge on the treatment performance of CIX process considering multiple contaminants, alternative regeneration chemicals, and variable water quality and alternative resins. Accordingly, the goal of this research was to evaluate the performance of the CIX process on multiple contaminants removal for small PWS. The CIX process was tested using a pilot-scale completely mixed flow reactor (CMFR) configuration with majority resin recycle and partial resin regeneration. A wide range of contaminants or problematic ions were evaluated including DOC, calcium, strontium, nitrate, sodium, and chloride. The raw water tested in this research was a real groundwater with high DOC and high calcium. The same groundwater was spiked with strontium and nitrate salts to evaluate problematic multiple anionic and cationic contaminants. In addition, the real groundwater was blended with real seawater to evaluate the effect of variable water quality. The specific objectives were to 1) study the influence of the alternative counterion on the equilibrium and kinetic characteristics of CER, 2) evaluate the CIX 97

98 process for DOC and calcium removal under the conditions of varying the effective resin dose (ERD), 3) test the removal of DOC and calcium using CIX in alternative-ion-form of resins for treatment and KHCO3 for regeneration in multiple treatment and regeneration cycles, and compare with the CIX in original-ion-form of resin for treatment with NaCl for regeneration, 4) expand the CIX process to remove problematic multiple contaminants in groundwater including DOC, calcium, strontium, nitrate, and seawater intrusion and 5) evluate the treatment performance using the conventional AER and CER in CIX process for the DOC and calciurm removal. Materials and Methods Materials The raw water used for the batch experiments was prepared in the laboratory by dissolving the appropriate mass of calcium chloride (CaCl2, CAS#: ) in deionized (DI) water. The raw water for pilot tests was from the groundwater supply to the Cedar Key Water and Sewer District located in Cedar Key, Florida, USA. The raw water used for the multiple contaminants test was prepared through spiking strontium chloride (SrCl2, CAS#: ) and sodium nitrate (NaNO3, CAS#: ) to the Cedar Key raw water. The raw water used for saltwater intrusion study was prepared by blending Atlantic Ocean water, collected at the St. Augustine, FL, USA, with the Cedar Key raw water. The salt solution used for preparing the bicarbonate-form AER was prepared through dissolving sodium bicarbonate (NaHCO3, CAS#: ) in DI water to make 25 excess salt equivalents than the resin equivalents. The same procedure was also applied to prepare the potassium-form CER, except the salt solution was prepared by dissolving the potassium chloride (KCl, CAS#: ) in DI water. The regeneration solution for pilot tests was prepared by dissolving the appropriate 98

99 mass of salt, either KHCO3 (CAS #: ) or NaCl (CAS#: ) in DI water to make the salt concentration of 12% (w/v). The chemical reagents mentioned above were all ACS grade and purchased from manufactures. The AER and CER used for tests of objective 1-4 are the MIEX and Plus, respectively, which are classified as the novel resin groups designed for well-mixed mode of operation. The MIEX and Plus are manufactured from IXOM. The MIEX has the resin properties of quaternary ammonium functional group, macroporous pore structure, and polyacrylic matrix. The resin capacity of 0.52 meq/ml. The special characteristic of AER in novel resin group is the incorporation of magnetic iron oxide. Also, the MIEX has a smaller particle size (150 µm- 200 µm) compared with other conventional resins (550 µm- 810 µm) (Graf et al., 2014). The Plus CER has no magnetic characteristics, and has properties of gel pore structure and polystyrene matrix. The resin capacity is 1.9 meq/ml. For objective 5, the AER and CER used are the conventional resin group, which are designed to operating in the column configuration. The AER in conventional resin group is the A30 manufactured from Thermax and the CER is the C150 from Purolite. The AER in conventional resin group has the same resin properties with the AER in novel resin group, except the AER in conventional resin group is not incorporated with iron oxide and has relatively larger particle size. The resin capacity of the A30 is 0.7 meq/ml. The CER in conventional resin group has the same polystyrene matrix property with the CER in novel resin group, however has different pore structure that is the macroporous. The C150 CER has the resin capacity of 1.8 meq/ml. Regarding the alternative-ion-form resin preparation, the bicarbonate-form AER was prepared through mixing the individual AER 99

100 in chloride-form, the original-form from manufacture, with the NaHCO3 solution in a jar tester for 24 h. The salt solution has 20 excess of equivalents than resin equivalents. The same procedure was applied to prepare the potassium-form CER through mixing the original-form of resin with KCl solution. The exception was the bicarbonate-form AER and potassium-form CER used for salt water intrusion scenario which was prepared by mixing the combined resins with 20% (w/v) KHCO3 in jar tester for 24 h. After mixing step, the resins in alternative-ion-form were carefully rinsed with DI water before testing. Experiments According to each objective, the detailed batch and pilot test plan and operation conditions are summarized in Table 4-1. Batch equilibrium and kinetic tests The batch equilibrium and kinetics tests for CER in novel resin group followed the same procedure described in the previous study (Hu and Boyer, 2017). The batch equilibrium tests were conducted by mixing the synthetic CaCl2 solution (containing 5 meq/l of calcium) with six varying resin doses on a shaker table for 24 h (Table 4-1, test 1). Triplicate samples were prepared for the control sample and the samples at each resin dose. The kinetic tests were performed by mixing 1 ml/l CER with the synthetic CaCl2 solution (containing 5 meq/l of calcium) in a jar tester; and samples were collected at certain intervals (Table 4-1, test 2). Three parallel kinetic tests were conducted, and the two samples were collected at each time interval. The samples collected for batch tests were all filtered through 0.45 μm filter and stored at 4 C. 100

101 Pilot-scale, completely mixed flow reactor study The pilot system is the CMFR configuration which was described in detail in the previous study (Hu and Boyer, 2017). The combined resins were dosed in the IX treatment contactor, and the fresh combined resins were dosed in the resin storage tank. The combined resins and treated water were separated in a settler, where the combined resins were recycled back to the IX treatment contactor, but treated water was overflowed. Simultaneously, a side stream of fresh combined resins were pumped from the resin storage tank into the IX treatment contactor; and the same amount of exhausted combined resins were collected from IX treatment contactor into the resin collector for regeneration. One course of operation was performed for 6 h. For the multiple cycles of treatment and regeneration tests (Table 4-1, tests 6 and 7), the regeneration step occurred at the end of treatment operation. The exhausted combined resins accumulated in the resin collector were mixed with the regeneration solution in a jar tester for 20 min. The volume of regeneration solution relative to the total volume of combined resins was 5 to 1. After regeneration, the combined resins were carefully rinsed with DI water until the conductivity less than 5 µs/cm. The fresh regenerated combined resins were placed in the resin storage tank for the following cycle of treatment. The treatment and regeneration cycle was performed for three times. The 2 L jar-tester container was used as the IX treatment contactor. The volume ratio of AER over CER was 1 to 1 for the entire pilot studies. The previous study (Comstock and Boyer, 2014) suggested the volume ratio of 1 to 4, which used different type of resins for batch study. After the preliminary tests, this research selected the volume ratio of 1 to 1 because it showed an effective contaminants removal and stable 101

102 hydraulic operation. The theoretical regeneration ratio of 20% and hydraulic residence time of 12 min was selected for all pilot tests to make the mean solids residence time (SRT) equal to 1 h. The SRT was calculated from the formula of hydraulic residence time divided by regeneration ratio (Boyer et al., 2008a). One hour was selected as the sampling interval, and duplicated samples were collected at effluent. The raw water was stored in a 10 gallon bucket and the sample of raw water was collected once for each test before testing. ph and conductivity were measured immediately after the sample collection. The dissolved ions were measured using the required instruments described in the following section. All collected samples were filtered through 0.45 μm filter and stored at 4 C. Analytical Methods The DOC was measured with Total Organic Carbon Analyzer (Shimadzu TOC- Vcph) following the Standard Method 5310 B (Eaton et al., 2005). Anions, including chloride, nitrate, and sulfate, and cations, including calcium and sodium, were measured by ion chromatography (Dionex ICS-5000) following EPA method and Dionex Method (EPA, 1997, Thomas et al., 2002). Strontium and the cations in test 8 (Table 4-1) were analyzed by inductively coupled plasma (ICP-OES) following EPA Method 6010C. ph and conductivity were measured with Accument AB15 ph meter coupled with Thermoscientific Orion Dual Star panel and Oakton ECTestr 11 probe, respectively. Moreover, Ricca TOC, Dionex six cation and Dionex seven anion of standard solutions were used as external standards for DOC, anion, and cation analysis, respectively. The external standards and calibrations were analyzed at the end of each run, and one selected calibration solution was analyzed every for 15 measurements to monitor the accuracy. Duplicate measurements were tested for 102

103 majority of samples with less than 10% difference, and the relative difference of calibration point and the external standard was analyzed with less than 10% at the concentration > 1 mg/l and less than 20% at the concentration < 1 mg/l for majority of measurements. Results and Discussion The Alternative Counterion on the Equilibrium and Kinetic Characteristics The batch tests were performed to investigate the equilibrium and kinetic properties of the potassium-form CER in the novel resin group and compare them with sodium-form to understand the influence of counterion. A similar study for AER was documented in the previous research (Hu and Boyer, 2017). The batch equilibrium tests were performed by mixing CaCl2 synthetic water with six varying amount of CER in potassium-form and sodium-form, respectively (Table 4-1, test 1). In Figure 4-1a and b, the equilibrium isotherm was plotted of calcium in the liquid phase along with the calcium in the resin phase for potassium-form and sodium-form CER, respectively. The data at the lowest resin dose was eliminated due to the variation from dosing a low mass of resin, but was plotted in the Appendix Figure C-1. The equilibrium data was also used to fit the Langmuir and the Freundlich isotherm, and the related parameters of each model were tabulated in Figure 4-1d. The detailed calculation procedure was followed in the previous study (Hu et al., 2016). The parameter of k L from the Langmuir model and 1/n from the Fruendlich model can be used to indicate the selectivity preference of the CER to calcium. The higher number of k L of potassium-form CER reflects a preferable selectivity to calcium than sodium-form CER, which was also observed from the previous study (Foster et al., 2017). The similar value of 1/n in Freundlich model indicates a comparable selectivity of potassium-form CER compared 103

104 with sodium-form CER. The kinetic data is plotted in Figure 4-1c, where the x-axis is time, and the y-axis is the calcium concentration in the liquid phase. The first-order kinetic equation was used to fit experimental data to obtain the lumped mass transfer coefficient described in the same previous publication (Hu and Boyer, 2017). The comparable coefficient of potassium and sodium form of CER showed that the counterion has a similar impact on the calcium uptake. In summary, potassium-form CER has a higher or comparable selectivity to calcium compared with sodium-form CER, and potassium-form CER has a comparable calcium removal rate as sodium-form CER. DOC and Calcium Removal at Varying Effective Resin Dose The raw water quality for the pilot tests is summarized in Table 4-2, tests 3-5. The Cedar Key raw water had DOC in the range of 7-8 mg/l and calcium in the range of mg/l (~250 mg/l as CaCO3). The major source of hardness in this raw water is from calcium; therefore, only calcium data was recorded. The effluent concentrations from the pilot tests were also summarized in Table 4-2. The pilot tests were operated in a dynamic condition, and the samples collected at the number of SRT from 3 to 6 were averaged to represent the mean effluent concentration (Boyer et al., 2008a, Hu and Boyer, 2017). The pilot tests were performed under the varying parameter conditions to understand the influence of ERD on treatment performance (Table 4-1, tests 3-5). Previous research has shown that the ERD and SRT were the important process parameters to describe IX performance in CMFR type of IX configuration (Boyer et al., 2008a, Boyer and Singer, 2006). Therefore, the varying levels of ERD were selected for testing, and the AER in chloride-form and CER in sodium-form was used. The resin 104

105 dose concentration varied as 10 ml/l, 20 ml/l, and 30 ml/l, respectively. Then the ERD can be calculated as 2 ml/l, 4 ml/l, and 6 ml/l, respectively, according to the equation of ERD equals the resin dose concentration times the regeneration ratio (Boyer et al., 2008a, Boyer and Singer, 2006). The ERD level was in the range of 1 ml/l to 5 ml/l as in previous studies (Apell and Boyer, 2010, Comstock and Boyer, 2014), except the test under the 6 ml/l condition. Figure 4-2a shows the results of DOC and calcium percentage removal during the CIX operation. The DOC removal for 2 ml/l, 4 ml/l, and 6 ml/l ERD condition was 55%, 63%, and 63%, respectively, and the calcium removal was 32%, 68%, and 70%, receptively. The Figure 4-2 b shows the dynamic results of effluent DOC and calcium conditions along with the operation. The x-axis is the number of SRT and the y- axis is the concentration of contaminant ions. The performance of DOC removal increased from the lowest ERD to moderate ERD condition, but no further improvement was observed in the 6 ml/l condition. A similar trend of calcium removal at varying ERD conditions was also shown. Comparing the calcium removal with DOC, a relatively varied effluent was shown, the explanation of this phenomenon is discussed in the following section. The phenomenon that improved DOC removal was not obvious under the highest ERD level can be explained by the charge of the dissolved organic matter (DOM) and the size exclusion (Boyer et al., 2008b, Wolska, 2015). The natural DOM cannot be removed by AER through stoichiometric IX reaction. The large size of DOM is difficult to diffuse into the resin to have the IX reaction. Regarding the calcium removal, the treatment efficiency was lower than expect at each ERD. The phenomenon can be 105

106 explained by the operational conditions. The equivalent analysis was performed for calcium removal, since the IX reaction is in the stoichiometric mechanism. The equivalents of resin indicate the theoretical exchange sites for contaminants removal, which was calculated by the resin dose concentration multiplies by the reins capacity. The equivalents of CER dosed in 2 ml/l, 4 ml/l and 6 ml/l conditions corresponded to 3.8 meq/l, 7.6 meq/l and 11.4 meq/l, respectively, comparing with 5 meq/l calcium in raw water. Therefore, the treatment efficiency, theoretically, should achieve 76% calcium removal in 2 ml/l and almost 100% removal in 4 ml/ and 6 ml/l conditions, respectively. The results showed in Figure 4-2 indicates that the counterions of CER is not completely exchanged with calcium ions. However, the calcium removal from previous studies agreed with the equivalents analysis (Arias-Paic et al., 2016, Comstock and Boyer, 2014). The previous research (Comstock and Boyer, 2014) performed multiple-loading jar tests with regeneration during CIX operation and demonstrated an almost 100% calcium removal when dosing CER with excess equivalents than calcium in raw water. Another study (Arias-Paic et al., 2016) conducted multiple-loading jar tests without regeneration which showed nearly 50% calcium removal when dosed with CER having about half of equivalents than calcium in raw water. Therefore, it is possible that the operational conditions in different approaches affect the calcium treatment. It is possible that the CER was not well-mixed in treatment contactor used in current research because the CER used in this research has a larger particle size and higher density. At such mixing condition, it can bring variations during the continuous withdrawal and replacement resin in the IX treatment contactor. The relatively heavier CER may stay at the bottom in the IX treatment reactor and get exhausted, but the 106

107 relatively less heavy CER was continuously taken out with partial use of exchange sites on resin. Regarding multiple-loading jar test, the CER is operated in batch system without withdrawal and replacing of resin; rather, the raw water is decanted and replaced after certain contact time. Under such operational condition, the dosed CER is completely used for calcium removal. To further illustrate this possible reason, the model developed from previous research (Hu and Boyer, 2017) for CMFR type of IX process were used to predict the calcium removal in current research and for previous multiple-loading jar tests (Arias-Paic et al., 2016, Comstock and Boyer, 2014). The results were plotted in Appendix Figure C-2. The predicted results from the model agreed with results in multiple-loading jar tests in previous research when compared to the continuous pilot system in this research, which reflects that the operation of multipleloading jar tests match up with the assumptions of the developed model that assumed a well-mixed condition in the treatment contactor, and constant operation of continuous withdrawal and replacement of resins. Therefore, the possible variations during continuous pilot systems operation in this research may lower the calcium removal. Future research is required to understand the influence of operational conditions, physical properties of resins, and system configuration on contaminant removal. In summary, the DOC and calcium removal responded to the ERD with increased removal when compared to the lowest ERD with the moderate ERD condition, but additional removal was not achieved at the highest ERD condition. The 4 ml/l ERD was selected as the optimal condition based on DOC and calcium removal and resin usage. 107

108 DOC and Calcium Removal of Multiple Treatment and Regeneration Cycles Multiple cycles of treatment and regeneration were tested using CIX in alterativeion-form and comparing it with the CIX in original-ion-form for DOC and calcium removal. The DOC and calcium removal percentage during each operation condition were plotted in Figure 4-3 a and b. The x-axis is the number of cycles, and the y-axis is the percent of contaminant removed. Comparing the DOC treatment under the two conditions (Figure 4-3 a and b), the bicarbonate-form AER in alternative-ion-form CIX operation (Figure 4-3 a) showed 71% DOC removal (cycle 1), followed by 67% (cycle 2) and then 66% (cycle 3). This was comparable to the chloride-form AER in the originalion-form CIX system (Figure 4-3 b) having about 65% DOC removal for each cycle. This comparable DOC treatment was also observed in the previous studies that used the bicarbonate-form AER only for DOC removal in batch-and pilot-scale studies (Hu and Boyer, 2017, Rokicki and Boyer, 2011, Walker and Boyer, 2011). The calcium removal of potassium-form CER in the alternative-ion-form CIX operation (Figure 4-3 a), decreased from 69% (cycle 1) to 52% (cycle 2) but increased to 82% removal (cycle 3). The treatment performance of the sodium-form CER in original-ion-form CIX operation was decreased from 84% (cycle 1), followed by 59% (cycle 2), and to 54% (cycle 3). From the result of contaminant percentage removal, the calcium is relatively fluctuated compared to the DOC removal. To understand the dynamic characteristic of the CIX/CMFR process, the concentration at each collected interval are plotted in Fig.3 c and d and the average effluent concentrations of DOC and calcium are summarized in Table 4-2 tests 6 and 7. In Figure 4-4 c and d, the x-axis is the number of SRT (1 SRT equals to 1 hr. section of pilot-scale completely mixed flow reactor study), and the y-axis is the concentration of contaminant ions in the liquid phase. Cross comparison of DOC 108

109 removal in Figure 4-3 c and d, the concentration in raw water was relatively constant and was in the range of 7 to 8 mg/l. The DOC concentration in the effluent ranged from 2 to 3 mg/l, and showed a relatively stable trend during the three cycles of operation. The calcium in raw water ranged from 100 to 120 mg/l (~250 mg/l as CaCO3). The effluent calcium data showed wide variation during operation in the range of 20 mg/l to 60 mg/l (~50 mg/l as CaCO3 to 150 mg/l as CaCO3), with a few data points approaching the raw water concentration. Due to the variation of calcium removal (discussed in the section of DOC and calcium removal at the varying effective resin dose level), it is difficult to directly compare the calcium removal of potassium-form CER and sodium-form CER during CIX operation. Theoretically, a comparable treatment should be expected according to the discussion in the section of The Alternative Counterion on the Equilibrium and Kinetic Characteristics. Moreover, care should be taken into account the calcium variation for CIX system design. The design of a storage tank after the CIX process can be an option to minimize the variation of calcium treatment (Locke and Smith, 2016). Regarding regeneration, solutions of KHCO3 and NaCl at 12% (w/v) were prepared in the laboratory to regenerate combined AER and CER simultaneously. The vertical line in Figure 4-3 c and d indicates the occurrence of regeneration. The same mass salt concentration, i.e., 12% (w/v), was used to prepare KHCO3 and NaCl. Regarding the molar concentration, the anionic counterion in KHCO3 solution is 1200 mmol/l, which is 24 equivalents of the theoretical equivalents of AER during regeneration. The anionic counterion in NaCl solution is 4100 mmol/l, which is 41 equivalents of the theoretical equivalents of AER during regeneration. Previous studies 109

110 have shown 10 excess equivalents in salt concentration was sufficient to recover the resin exchange sites (Rokicki and Boyer, 2011, Walker and Boyer, 2011). For the DOC treatment, the comparable DOC concentration in the effluent reflected a minimum influence of using KHCO3 compared with NaCl to regenerate. This was also observed in the previous study of DOC treatment with AER solely using NaHCO3 regeneration compared with NaCl regeneration (Hu and Boyer, 2017). With respect to CER in combined resin regeneration, KHCO3 contains 6 equivalents more than the theoretical equivalents of CER, and NaCl has 8 excess equivalents. Comparing KHCO3 with NaCl regeneration, although there was variation during the multiple cycles of operation, over 50% calcium removal indicates that sufficient exchange sites can also be recovered to maintain an effective calcium removal. During combined AER and CER regeneration, mineral precipitation was observed, which may be formed from calcium reacting with sulfate ions or carbonate ions. The previous study discussed that the calcium carbonate precipitation may be beneficial to the resin recovery efficiency (Apell and Boyer, 2010), which may also be applied to the combined AER and CER regeneration in this research. Moreover, the calcium carbonate or calcium sulfate precipitation can reduce the calcium in used brine, which may enable reusing brine solution several more times. Furthermore, the batch regeneration step happened along with the treatment in CMFR IX system uses less amount of regeneration solution and less contaminants released in regeneration solution comparing to regenerate entire used resin in a column system, which can also benefit to brine reuse. Furthermore, previous research has demonstrated an enhanced DOC removal in CIX operation (Apell and Boyer, 2010, Arias-Paic et al., 2016). The enhanced DOC 110

111 removal showed the CIX operation in a batch test using both magnetic AER and CER to treat raw water containing DOC and calcium (Apell and Boyer, 2010). The possible explanation of enhanced removal discussed in the research was that CER can remove the positively charged complexes of DOM-Ca (Apell and Boyer, 2010). A separate research experiment used the same magnetic CER only to treat water containing DOM and calcium observed DOC reduction (Indarawis and Boyer, 2012). The mechanism of this phenomenon was explained by the negatively charged magnetic nanoparticles on the CER were able to adsorb the DOM-Ca through electrostatic interactions (Indarawis and Boyer, 2013). The enhanced DOC removal was also observed in a recently published study which used magnetic AER and non-magnetic CER (Plus, the same CER in this research) to perform laboratory batch CIX operation for DOC and calcium removal (Arias-Paic et al., 2016). The two mechanisms proposed were that CER can remove additional DOC through removing positive charged calcium; that can act as a bridge to dissolved organic matter through complexing with carboxylic acid groups or other weekly acidic functional groups and can bond to the weakly acidic functional group (Arias-Paic et al., 2016). However, the enhanced DOC removal was not observed in current research, using the same AER and CER with the recently published work (Arias-Paic et al., 2016), comparing with AER treatment only in authors previous study (Hu and Boyer, 2017). No additional DOC treatment was shown in another study which used the same magnetic AER and conventional CER (Amberlite 200C) during CIX operation in multiple-loading jar tests (Comstock and Boyer, 2014). The 67%-70% of DOC removal in CIX operation was comparable with ~70% DOC removal in AER treatment only for the same source of water treatment (Comstock and Boyer, 2014). It is 111

112 still interesting to know if the enhanced DOC removal can be observed in a different source of raw water treatment and during CIX operation with resins having different properties. Multiple Contaminants and Salt Water Intrusion Study The CIX operation has been comprehensively investigated for DOC and hardness removal, the application of the CIX process to other scenarios is also of great interest in this research to help the small PWS that have problematic source water. A pilot study was performed for two scenarios including spiking SrCl2 and NaNO3 to Cedar Key raw water, and blending Atlantic Ocean water with Cedar Key raw water to mimic seawater intrusion. The raw water properties are summarized in Table 4-1, tests 8 and 9. The 4 ml/l dose level was used for test 8, and 8 ml/l was used in test 9. Spiking strontium and nitrate to Cedar Key raw water Strontium (22 mg/l) and nitrate (30 mg/l as N) were spiked in to the high DOC and hardness of Cedar Key raw water to make a complex raw water that contained multiple anionic and cationic contaminants. The results in Figure 4-5 showed removals of 62% DOC, 68% calcium, 67% strontium, and 30% nitrate. The effluent concentration of contaminant ions was plotted in Figure 4-5 b and the average concentration was summarized in Table 4-2, test 8. The results showed a substantial removal of strontium, although the competing effect can result from the presence of high calcium, 20 times of equivalent concentration than strontium. The reason for this phenomenon can be explained by the strontium has a higher selectivity to the CER used in current research, having the polystyrene and gel resin properties, which was demonstrated from the previous study (Foster et al., 2017). The CER is preferable selective to the ions having a larger ionic radius and smaller hydrated radius, which are the properties of strontium 112

113 compared with calcium. Regarding the lower nitrate removal, the reason is that nitrate is not selective to the AER used in this research, having a narrow spacing of functional groups and polyacrylic resin matrix. The previous study showed that nitrate has high selectivity to the resin with wider space of the triethylamine (R-N + (CH2CH3)3) functional group and polystyrene composition (Hu et al., 2016). To improve nitrate removal, adding a nitrate selective resin to operate two AER and one CER can be an option for complex multiple contaminants removal. Impact of seawater intrusion on DOC and hardness removal In coastal areas, small PWS may face changes in raw water quality due to extreme events such as prolonged drought and floods. In 2012, Cedar Key, Florida experienced a seawater intrusion event that increased the sodium levels and resulted in a Do not drink the water advisory for two months (Seatta et al., 2015). Sodium in raw water increased from a background level of ~10 mg/l to beyond 160 mg/l, which is the MCL in Florida. The chloride in the raw water increased from a background level of 10 mg/l to 500 mg/l, and conductivity increased from ~500 µs/cm to ~2000 µs/cm. The original process train at the water plant: permanganate addition, anion exchange, lime softening, sand filtration, and chlorination, was not capable of reducing the concentration of sodium and chloride to drinking water levels. To help small PWS handle sudden changes in water quality, CIX using the alternative-ion-form was evaluated for removal of DOC and calcium (hardness) as well as sodium and chloride. The pilot test was performed using 8 ml/l ERD (Table 4-1, test 9) of bicarbonate-form AER and potassium-form CER for treatment, and KHCO3 for regeneration. The results for DOC, chloride, sulfate, calcium, and sodium removal are shown in Figure 4-7 a. The CIX process achieved the following removals: 72% DOC, 69% calcium, 54% sulfate, 113

114 27% sodium, and 23% chloride. The effluent concentration of contaminant ions was plotted in Figure 4-7 b and the average concentration was summarized in Table 4-2, test 9. The 8 ml/l ERD corresponded to 4.16 meq/l equivalents of AER. The removal of chloride from 500 mg/l to 390 mg/l equals to a 3 meq/l reduction. Additionally, there was a 0.06 meq/l DOC reduction (assuming a charge density of 10 meq/g C (Indarawis and Boyer, 2013)) and 0.7 meq/l sulfate reduction. Therefore, the AER exchange sites approached saturation with the contaminant ions. If desired, the chloride removal could be improved by increasing the AER resin dose. The design of operating the large quantity of resin regarding the selection of mixer and size and shape of contactor reactor needs to be taken into account. It is interesting to observe the substantial DOC removal under the high concentrations of the sulfate and chloride conditions. The effective DOC was also observed from a previous study that used the same type of AER to remove DOC (4.2 mg/l) in the presence of sulfate (46 mg/l) and other inorganic ions (Rokicki and Boyer 2011). The reason for this occurrence was explained by both DOC and sulfate were preferable selective to AER and had similar selectivity. The 8 ml/l ERD corresponded to 15.2 meq/l of CER equivalents. The reduction of calcium and sodium used 7 meq/l of resin equivalents, which means that approximately half of the equivalents were used for treatment. The reason for the low sodium removal was because the potassium-form of gel, polystyrene CER is not selective to sodium (Foster et al., 2017). The strong acid of CER only has two types of resin properties, both of them showed low selectivity for sodium. Another impact is the operational conditions may influence the cationic contaminants removal in this pilot set up (section of DOC and Calcium Removal at the Varying the Effective Resin Dose). The 114

115 options to increase the sodium removal are to increase the CER dose since the effluent sodium concentration was close to meeting the MCL (188 mg/l vs. 160 mg/l), or select an alternative process configuration, e.g. fluidized configuration. Hence, the CIX process in alternative-ion-form is a possible short-term option to handle changes in raw water quality including elevated sodium and chloride levels. DOC and calcium removal using conventional AER and CER The resin used for CIX operation is typically magnetic AER, and magnetic or nonmagnetic CER for treatment evaluation (Apell and Boyer, 2010, Arias-Paic et al., 2016, Comstock and Boyer, 2014). It is also beneficial to test the performance using both conventional AER and CER during CIX operation in a CMFR configuration. The A30 AER in chloride-form and C150 CER in sodium-form with NaCl for regeneration were used for Cedar Key raw water treatment targeting DOC and calcium removal (Table 4-1, test 10). The tests were performed following the procedure of test 4 and 5 in Table 4-2 that is in multiple treatment and regeneration cycles under the condition of dosing 4 ml/l ERD. The DOC and calcium removal percentages are shown in Figure 4-7 a, and the dynamic concentration is shown in Figure 4-7 b, where the vertical lines indicate the occurrence of regeneration with 12 % (w/v) NaCl. The results showed removals of greater than 40% DOC and 60% calcium (Figure 4-7 a). The effluent DOC concentration ranged between 3 mg/l to over 4 mg/l, and the calcium concentration ranged between 20 mg/l (~50 mg/l as CaCO3) to over 40 mg/l (~100 mg/l as CaCO3) over multiple cycles of treatment and regeneration (Figure 4-7 b). Compared with the results of using a novel resin group, namely MIEX and Plus (Figure 4-3), the CIX using conventional resin group showed relatively lower DOC removal and comparable calcium removal. The calcium in the effluent over the time of operation was relatively more 115

116 stable than the novel resin group (Figure 4-3 and Figure 4-6). The possible reason is that the A30 and C150 have less difference of particle size compared with MIEX and Plus. The A30 has the same resin properties as MIEX, except the resin radius, iron incorporation, and resin capacity (0.7 meq/ml of A30 compared to 0.52 meq/ml of MIEX). The lower removal of DOC treatment observed in the conventional resin group is possibly due to the resin breakdown during operation since the conventional AER is designed for column operation and not the mixing and pumping in the CMFR process. Fine resin particles were observed in the settler and the effluent. The C150 CER has polystyrene resin properties but a macroporous pore structure compared to Plus with polystyrene and gel properties. The C150 has 1.8 meq/ml and Plus has 1.9 meq/ml of resin capacity. The calcium removal was comparable, and the treatment was relatively consistent during operating C150 in multiple cycles. The possible reason is macroporous structure is relative swelling in water than gel type of resin; therefore, the resins can be easier to operate during mixing, withdrawal and recycle. Moreover, the resin strength is another factor during mixing and pumping in the CMFR configuration, which can affect treamtent performance. Overall, the conventional resin group demonstrated that it is feasible to use conventional AER and CER in the CIX process using a CMFR type of IX configuration. Conclusions Remarks The CIX process is a novel operation for multiple contaminants removal. The DOC and calcium removal was evaluated under the condition of varying the parameter of ERD and regeneration salts. The treatment performance was also evaluated for a complex mixture of multiple contaminants and seawater intrusion, which showed effective treatment performance. Lastly, the CIX process can be operated using 116

117 conventional AER and CER for removal of DOC and calcium, which allows CIX process to be expanded to numerous types of resins. The key results in this research showed CIX can be applied for a variety of multiple contaminants removal to achieve over 60% DOC, calcium and strontium removal and additional contaminants removal to some degree. The multiple treatment and regeneration cycles showed over 60% DOC and over 50% calcium removal during alternative-ion-form resin for treatment and KHCO3 for regeneration, which is comparable to the original-ion-form resin for treatment and NaCl for regeneration. The DOC and calcium removal was improved along with increasing the ERD level. However, limited improvement was observed in the excess resin dose condition. When varying effective resin levels of operation, the calcium treatment was lower than expect according to equivalent analysis. The reason was that the CER cannot achieve a well-mixed condition and the resin capacity cannot be constantly maintained with continuous resin withdrawal and replace. Lastly, the change in the counterion for the cation resin did not change the equilibrium and kinetic properties of the resin. 117

118 Table 4-1. Experimental plan for batch tests and pilot tests Objective Test Operation cycle Raw water Type of resin Regen. ratio Contactor conc. ERD HRT (#) (#) (#) AER CER (%) (ml/l) (ml/l) (min) Synthetic Ca 2+ raw -- a. Potassium -- Varying -- 60*24 water b. Sodium resin dose a Synthetic Ca 2+ raw -- a. Potassium b water b. Sodium Cedar key raw Chloride Sodium water Cedar key raw Chloride Sodium water Cedar key raw Chloride Sodium water Cedar key raw water Chloride Sodium Cedar key raw water Bicarbonate Potassium Cedar key raw water spiking with Chloride Sodium SrCl2 and NaNO Cedar key raw water blending with seawater Cedar key raw water Bicarbonate Potassium Chloride Sodium a. The CER was dosed to have CER equivalents of 5%, 25%, 50%, 100%, 150% and 300% of Ca 2+ in the synthetic raw water b. The samples were collected at the time intervals of 5, 10, 20, 40, 60, min 118

119 Table 4-2. Summary of raw water and treated water parameters for pilot tests Test (#) ph Conductivity (µs/cm) Infl.DOC (mg/l) Eff.DOC (mg/l) Infl.Ca 2+ (mg/l) a b Eff.Ca 2+ (mg/l) a. Strontium and nitrate was spiked in Cedar Key raw water. The strontium concentration in raw water was 23 mg/l and nitrate was 31 mg/l. The strontium concentration in effluent was 7.5 mg/l and nitrate was 22 mg/l. b. Atlantic Ocean water was blended with Cedar Key raw water. The chloride concentration in raw water was 503 mg/l, sulfate was 148 mg/l, and sodium is 258 mg/l. The chloride concentration in effluent was 390 mg/l, sulfate is 69 mg/l and sodium was 188 mg/l. 119

120 Calcium in liquid phase (mg/l) c Potassium-form data Potassium-form fitting Sodium-form data Sodium-form fitting Time (min) d Parameter Potassiumform Sodiumform k L (Langmuir) (L/mmol) q 0 (Langmuir) (mmol/g) k F (Freundlich) (mmol 1-1/n L 1/n g -1 ) 1/n (Freundlich) Mass transfer coefficient (cm/min) Figure 4-1. Isotherm fitting and equilibrium and kinetic parameters. a) isotherm data for potassium-form CER with synthetic Ca 2+ raw water (data symbols show mean value + one standard deviation of triplicate samples), b) isotherm data for sodium-form CER with synthetic Ca 2+ raw water (data symbols show mean value + one standard deviation of triplicate samples), c) kinetic data for potassium-form and sodium-form CER, respectively with synthetic Ca 2+ raw water (data symbols show mean value + one standard deviation of triplicate samples), d) equilibrium parameters of Langmuir, Freundlich isotherms and mass transfer coefficient of first-order kinetic model. 120

121 Calcium concentration (mg/l) DOC concentration (mg/l) Containminant removal percentage a 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% b DOC Calcium 2 ml/l 4 ml/l 6 ml/l Effective resin dose level ml/l 4 ml/l 6 ml/l Raw water (calcium) Effluent (calcium) Raw water (DOC) Effluent (DOC) Number of solids residence time Figure 4-2. Contaminant removal percentage and concentrations in effluent under varying the conditions of effective resin dose level. a) contaminant removal in percentage under the varying the level of effective resin dose (averaged mean value for the samples collected at the last three solids residence time + one standard deviation), b) contaminant concentration in raw water and effluent for six solids residence time at each effective resin dose condition (data collected at each time interval). 121

122 Calcium concentration (mg/l) DOC concentration (mg/l) Contaminant removal percentage Contaminant removal percentage a 100% 90% 80% DOC removal (bicarbonate-form) Calcium removal (potassium-form) b 100% 90% 80% DOC removal (chloride-form) Calcium removal (sodium-form) 70% 70% 60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% cycle 1 cycle 2 cycle 3 Number of operation cycle 0% cycle 1 cycle 2 cycle 3 Number of operation cycle c Number of solids residence time Raw water (Calcium) Effluent (Calcium) Raw water (DOC) Effluent (DOC) 122

123 Calcium concentration (mg/l) DOC concentration (mg/l) d Raw water (Calcium) Effluent (Calcium) Raw water (DOC) Effluent (DOC) Number of solids residence time Figure 4-3. Contaminant removal percentage and concentration in effluent during alternative-ion-form and original-ion-form operations. a) contaminant removal in percentage during operation of CIX in alternative-ion-form for treatment and KHCO3 for regeneration (averaged mean value for the samples collected at the last three solids residence time for three cycles of operation + one standard deviation) b) contaminant removal in percentage during operation of CIX in original-ion-form and NaCl for regeneration (averaged mean value for the samples collected at the last three solids residence time for three cycles of operation + one standard deviation), c) contaminant concentration in raw water and effluent during operation of CIX in alternative-ion-form for treatment and KHCO3 for regeneration (data collected at each time interval for multiple cycles of treatment and regeneration cycles), d) contaminant concentration in raw water and effluent during operation of CIX in original-ion-form and NaCl for regeneration (data collected at each time interval for multiple cycles of treatment and regeneration cycles). 123

124 Inorganic contaminant concnetration (mg/l) DOC concentration (mg/l) Contaminant removal percentage a 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% DOC Calcium Strontinum Nitrate b Calcium (raw water) Nitrate (raw water) Strontium (raw water) Calcium (effluent) Nitrate (effluent) Strontium (effluent) DOC (raw water) DOC (effluent) Number of solids residence time Figure 4-4. Contaminant removal percentage and concentration in effluent for scenario of Cedar Key raw water spiking with SrCl2 and NaNO3. a) contaminant removal in percentage of CIX process using chloride-form AER and sodiumform CER during CIX operation (averaged mean value for the samples collected at the last three solids residence time + one standard deviation), b) contaminant effluent concentration of CIX process using chloride-form AER and sodium-form CER during CIX operation (data collected at each time interval)

125 inorganic anion concentration (mg/l) DOC concentration (mg/l) Contaminant removal percentage a 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% DOC Calcium Sodium Chloride Sulfate b Chloride (raw water) Sulfate (raw water) Calcium (raw water) Sodium (raw water) Chloride (effluent) Number of solids residence time Figure 4-5. Contaminant removal percentage and concentration in effluent for scenario of Cedar Key raw water blending with Atlantic Ocean water. a) contaminant removal in percentage of CIX process using bicarbonate-form AER and potassium-form CER during CIX operation (averaged mean value for the samples collected at the last three solids residence time + one standard deviation), b) contaminant effluent concentration of CIX process using bicarbonate-form AER and potassium-form CER during CIX operation (data collected at each time interval) Sulfate (effluent) Calcium (effluent) Sodium (effluent) DOC (raw water) DOC (effluent) 125

126 Calcium concentration (mg/l) DOC concentration (mg/l) Contaminant removal percentage a 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% b 120 DOC Calcium cycle 1 cycle 2 cycle 3 Number of operation cycle Raw water (calcium) Effluent (calcium) Raw water (DOC) Effluent (DOC) Number of solids residence time Figure 4-6. DOC and calcium of removal percentage and concentration in effluent during CIX operation using conventional AER and CER. a) DOC and calcium removal in percentage of CIX operation using the chloride-form conventional AER and sodium-form conventional CER (averaged mean value for the samples collected at the last three solids residence time + one standard deviation), b) DOC and calcium effluent concentration of CIX operation using the chloride-form conventional AER and sodium-form conventional CER (data collected at each time interval). 126

127 CHAPTER 5 CONCLUSIONS Small PWS encounter numerous challenges to remove a wide range of contaminants as well as minimize the harmful impact generated during treatment. The innovative IX presented in this research adopts a holistic methodology to upgrade current IX process to become suitable and sustainable for small PWS treatment. By using alternative salts for regeneration, the bicarbonate-form IX can reduce the brine disposal burden placed on the waste stream while also removing a wider range of negatively charged contaminants. Additionally, the traditional operation IX can be changed to operate AER and CER in the same vessel for multiple contaminants removal and regenerate combined AER and CER simultaneously to minimize waste brine production. Specifically, the results from selectivity analysis for the wide range of contaminants showed that the bicarbonate-form AER is a viable option to address the problems of the common violated chemical contaminants or problematic ions. The key resin properties for the target contaminant treatment were summarized and discussed. The contaminant selectivity sequences were also established to guide small PWS for the resin selection and process design for target contaminant removal. The results that focused on bicarbonate-form IX in the CMFR IX configuration for DOC removal showed the comparable removal performance to the traditional IX process, which further strengthened the viability of using bicarbonate-form IX in a practical process. The process model, incorporating the regeneration to the treatment, was developed to predict the DOC concentration in treated water. The developed model can be utilized as a technical tool for small PWS regarding the treatment process 127

128 design. Incorporating the regeneration efficiency to the treatment can aid small PWS in understanding the impact of varying the regeneration conditions on the contaminant removal. Moreover, the pilot tests were performed to validate the developed model as well as provide a practical understating regarding operating the bicarbonate-form IX. The pilot tests of CIX operation showed the substantial removal of a varying mixture of multiple contaminants, including real groundwater having high DOC and hardness, real groundwater spiked with strontium and nitrate, and groundwater affected by saltwater intrusion. The CIX designed for a group of contaminants removal can aid the small PWS to remove multiple contaminants as well as prepare for the viability of source of water. The implementation of the innovative IX also requires addressing certain considerations. The cost of sodium bicarbonate is approximately five times that of sodium chloride. Considering, however, the cost of sodium bicarbonate may be offset by the low cost of the bicarbonate salt disposal. Regarding the cost of CIX operation, a recently published work of a full-scale CIX plant mentioned that the fluidized IX configuration with majority resin recycle and partial resin regeneration had relative lower capital and operation cost. Therefore, the fluidized IX configuration was selected for the CIX pilot test before full-scale implementation. However, the detailed cost was not specific. Future work is required to quantify the cost estimation for innovative IX implementation during treatment and waste disposal. The conceptual implementation of innovative IX for small PWS and the material pathways are depicted in Figure

129 Figure 5-1. The application of innovative ion exchange to small communities, and the pathways of water, contaminants and chemicals. 129

130 APPENDIX A SUPPLEMENTARY INFORMATION FOR CHAPTER 2 Figure A-1. Experimental nitrate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose). 130

131 Figure A-2. Experimental bromide adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose). 131

132 Figure A-3. Experimental perchlorate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose). 132

133 Figure A-4. Experimental chromate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose). 133

134 Figure A-5. Experimental sulfate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose). 134

135 Chromic acid deprotonation reaction: H 2 CrO 4 H + + HCrO 4 Pka 1 = 0.9 (A-1) HCrO 4 H CrO 4 Pka 2 = 6.5 (A-2) Ion exchange reaction: z B A s Z A + z A B R Z A z B A R Z A + z A B s Z B (A-3) Selectivity coefficient equation: K AB = ( q Z A B Z C A B ) ( ) C A q B (A-4) Mass-action law equation: K AB = ( q Z B Aγ RA ) ( C Z A Bγ SB ) C A γ SA q B γ RB (A-5) A: contaminant ion B: mobile counter ion z j : The valence of species j C j : The equivalent concentration of species j in the liquid phase q j : The equivalent concentration of species j in the resin phase γ Sj : The acivity coefficient of species j in the liquid phase γ Rj : The acivity coefficient of species j in the liquid phase 135

136 Table A-1. ph change before and after equilibrium Contaminant Synthetic nitrate, bromide, sulfate and perchlorate solution ph before equilibrium ph after equilibrium Net change of hydroxide ion ~ 6 >8 Increasing meq/l Synthetic chromate solution ~8 ~9 Increasing meq/l Prepared NOM solution ~8 ~8 No 136

137 Table A-2. Langmuir parameter of divalent contaminants using bicarbonate-form anion exchange resin (corrected the molar concentration during calculation) Resin Chromate q 0 (mmol/g) Langmuir q 0 (mmol/g) manufacture kl (L/mmol) G 0 R 2 ARE (%) SSE (mmol 2 /g 2 ) A520E Marathon Dowex A IRA A Resin Sulfate q 0 (mmol/g) Langmuir SSE (mmol 2 /g 2 ) q 0 (mmol/g) manufacture kl (L/mmol) G 0 R 2 ARE (%) A520E Marathon Dowex A IRA A

138 Table A-3. Freundlich parameter of divalent contaminants using bicarbonate-form anion exchange resin (corrected the molar concentration during calculation) Resin Chromate k F 1/n R 2 ARE (%) SSE (mmol 2 /g 2 ) (mmol 1-1/n L 1/n g -1 ) A520E Marathon Dowex A IRA A Resin Sulfate k F (mmol 1-1/n L 1/n g -1 ) 1/n R 2 ARE (%) SSE (mmol 2 /g 2 ) A520E Marathon Dowex A IRA A

139 Figure A-6. Experimental chromate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (corrected the molar concentration during calculation). 139

140 Figure A-7. Experimental sulfate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (corrected the molar concentration during calculation). 140

141 Figure A-8. Experimental chromate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose, corrected the molar concentration during calculation). 141

142 Figure A-9. Experimental sulfate adsorption to bicarbonate-form resins and isotherm model fits to experimental data (including data at lowest resin dose, corrected the molar concentration during calculation). 142

143 Corrected information in page 36 A520E: Perchlorate > sulfate > chromate > bromide nitrate Marathon11: Chromate > sulfate > perchlorate > nitrate bromide Dowex22: Chromate > sulfate perchlorate > bromide nitrate A300: Chromate > sulfate > Perchlorate > bromide nitrate IRA958: Chromate > sulfate > perchlorate bromide nitrate A850: Chromate > sulfate > perchlorate > bromide nitrate Corrected unit of k F to mmol 1-1/n L 1/n g

144 Nomenclature APPENDIX B SUPPLEMENTARY INFORMATION FOR CHAPTER 3 c in : Removable DOC concentration in raw water (mg/l) c out : Removable DOC concentration in treated water (mg/l) c reg : Removable DOC concentration in liquid phase after regeneration (mg/l) c : Removable DOC concentration in liquid phase at equilibrium during treatment (mg/l) c reg : Removable DOC concentration in liquid phase at equilibrium after regeneration (mg/l) q out : DOC concentration in resin phase exiting IX treatment contactor (mg/ml resin) q f : DOC concentration in fresh resin phase which pass through regeneration (mg/ml resin) k L1 : Lumped mass transfer coefficient of during treatment (cm/min) k L2 : Lumped mass transfer coefficient during regeneration (cm/min) K 1 : Linear equilibrium coefficient of IX reaction during treatment (L liquid/ml resin) Rp: Resin radius (cm) α IX : Resin concentration in IX treatment contactor (ml resin/l) α reg : Resin concentration in batch regeneration reactor (ml resin/l) τ H : Hydraulic residence time (min) τ s : Solids residence time (min) f R : Regeneration ratio (%) t: Time (min) 144

145 Table B-1. Summary of mass coefficient from literatures and this research Water Mass transfer coefficient (cm/min) Sweetwater Lake a (Boyer et al., 2008a) North Bay Aqueduct a (Boyer et al., 2008a) Durham a (Boyer et al., 2008a) Cedar Key a. The estimation of mass transfer coefficient from diffusivity is based on a literature (Dahlke et al., 2006). 145

146 Table B-2. The comparison of pilot plant data and model prediction Run # Operation condition Raw. DOC (mg/l) Pilot. effl. DOC (mg/l) Non-removable DOC (mg/l) Model.effl. DOC (mg/l) 1 Virgin fresh resin % 2 1% (w/v) NaCl reg % 3 1.4% (w/v) NaHCO3 reg % 4 6% (w/v) NaCl reg % 5 8.6% (w/v) NaHCO3 reg % DOC percent difference (%) 146

147 Figure B-1. Yue Hu, Pilot plant set up in Cedar Key drinking water treatment plant, March 2016, Cedar Key, Florida, USA. Photo courtesy of Yue Hu. 147