Design of a Continuous-Flow, Immobilized-Cell Fermentor System for Production of Bioethanol

Lan Ma, B.S.

Dissertation

Chemical Engineering

Submitted to the Graduate Faculty of Texas Tech University in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

Approved

Dr. Ronald C. Hedden Chair of Committee

Dr. M. Nazmul Karim

Dr. Mark W. Vaughn

Dr. Carla M. R. Lacerda

Dr. Michael J. D. San Francisco

Dr. W. Andrew Jackson

Dr. Mark Sheridan Dean of the Graduate School

December 2014

Texas Tech University, Lan Ma, December 2014

Acknowledgements

First and foremost I want to thank my advisor Dr. Hedden for his support, advice and guidance during my doctoral study at Texas Tech University. I appreciate Dr. Hedden for introducing me to the areas of polymer and ethanol fuel. At the beginning of study, I was not even sure what polymer is, he expressed his confidence on me. I am very grateful to

Dr. Hedden for the time, ideas and encouragements he gave to me, especially when I encountered difficulties and lost confidence during the research. It has truly been a privilege for me to be his student.

I would like to thank Dr. Karim for his kindness in guiding my early study and allowing me to use his equipment. A special thank you goes to Dr. Lacerda who gave me a lot of valuable advices. I would also like to thank Dr. Vaughn, Dr. San Francisco and

Dr. Jackson for being my dissertation committees.

To my colleges, Dr. Jun Zhao, Dr. Huipeng Chen, Dr. Seunghyun Ryu, Ziniu Yu,

Rutvik Godbole, Michael Sees and undergraduate students who worked in our laboratory,

I am thankful for all your friendship, support and help.

All my friends, you know how I appreciate your support and accompany.

Finally, I could never complete my research without the love, support and

encouragement of my parents, Ruiqing Ma and Shuqin Luan. I love you so much.

Texas Tech University, Lan Ma, December 2014

Table of Contents

Acknowledgements ...... ii Table of Contents ...... iii Abstract ...... vi List of Tables ...... viii List of Figures ...... ix Chapter 1 Introduction ...... 1 1.1 Biofuel as a renewable fuel ...... 1 1.2 Cellulosic Feedstocks ...... 2 1.3 Development of ethanologenic microorganisms ...... 7 1.3.1 Engineering and performance of E. coli ...... 7 1.3.2 Engineering and performance of other microorganisms ...... 9 1.3.3 Comparison of the microorganisms fermentation performance ...... 11 1.4 Fermentation reactor design ...... 12 1.4.1 Continuous immobilized cell reactor design ...... 12 1.4.2 Immobilization scaffold materials and methods ...... 13 1.4.3 Immobilized cell reactor for EtOH production ...... 17 1.5 Bioethanol separation ...... 19 1.6 Objectives and organization of this study ...... 22 1.7 References ...... 26 Chapter 2 Batch and chemostat bio-ethanol fermentation process with strain Escherichia coli. LY01 ...... 36 2.1 Introduction ...... 36 2.2 Materials and methods ...... 40 2.2.1 Microorganism and growth medium ...... 40 2.2.2 Inoculum preparation ...... 40 2.2.3 Fermentation conditions ...... 41 2.2.3 Analytical methods ...... 41 2.3 Results ...... 42 2.3.1 Batch fermentations ...... 42 2.3.2 Chemostat fermentaions ...... 50 iii

Texas Tech University, Lan Ma, December 2014

2.4 Conclusion ...... 55 Chapter 3 Design of a continuous-flow, packed-bed fermentation process ...... 59 3.1 Introduction ...... 59 3.2 Materials and methods ...... 61 3.2.1 Segmented vertical column reactor and immobilized cell stirred tank reactor 61 3.2.2 Inoculum preparation ...... 63 3.2.3 Analytical methods ...... 63 3.3 Results and discussion ...... 64 3.3.1 Cell immobilization with synthetic porous polymer scaffold (SPPS) ...... 64 3.3.2 Cell immobilization with poly(ethylene terephthalate) fiber scaffold...... 79 3.4 Discussion ...... 96 3.5 Conclusion ...... 97 3.6 References ...... 99 Chapter 4 Kinetics study of ethanol fermentation with Escherichia coli. LY01 ...... 101 4.1 Batch fermentation kinetics ...... 101 4.1.1 General unstructured kinetic model ...... 102 4.1.2 Models with growth inhibitors ...... 104 4.1.3 Kinetics model of ethanol production by E. coli LY01 ...... 105 4.2. Continuous fermentation kinetics ...... 111 4.3. Mathematical modeling of cell immobilized fermentation with packed-bed column reactor at steady state ...... 117 4.4 Dimensionless parameters estimation ...... 126 4.5 Conclusion ...... 128 4.6 References ...... 131 Chapter 5 Combinatorial screening of selective materials for extractive fermentation . 133 5.1 Introduction ...... 133 5.2 Experimental section ...... 136 5.2.1 Network synthesis...... 136 5.2.2 Swelling experiments ...... 137 5.2.3 Selectivity ...... 137 5.3 Theory and calculations (collaborated with Rutvik Godbole) ...... 139

Texas Tech University, Lan Ma, December 2014

5.4 Results and discussion ...... 144 5.4.1 Equilibrium swelling ...... 144 5.4.2 Distribution coefficient and equilibrium selectivity ...... 147 5.4.3 Predicted distribution coefficient and selectivity from the model ...... 149 5.4.4 Diffusivities and Permeability Selectivity ...... 151 5.4.5 Optimum Material Selection ...... 153 5.5 Conclusion ...... 155 5.6 References ...... 156 Chapter 6 Effects of carbon black nanoparticles on two-way reversible shape memory in crosslinked polyethylene ...... 163 6.1. Introduction ...... 163 6.2. Experimental section ...... 165 6.2.1. Materials ...... 165 6.2.2. Sample preparation ...... 165 6.2.3. Swelling experiments ...... 166 6.2.4. Shape memory effect (SME) analysis ...... 166 6.2.5. Wide-angle X-ray diffraction (2D-WAXD) measurements ...... 168 6.3 Results ...... 168 6.3.1 Thermal, mechanical, and shape memory properties ...... 168 6.3.2. WAXD analysis ...... 178 6.4 Discussion ...... 181 6.5 Conclusions ...... 183 6.6 References ...... 184 Chapter 7 Summary and future work ...... 190 7.1 Summary ...... 190 7.2 Future work ...... 192

Texas Tech University, Lan Ma, December 2014

Abstract

Due to limited global supplies of fossil fuel resources, rising prices, and environmental problems, future energy needs must be met in part or entirely by fuels derived from renewable resources. Production of ethanol (EtOH) from cellulosic biomass is one of the more attractive alternatives for producing a portable, liquid fuel. While many processes and economic factors are involved in the conversion of lignocellulosic biomass to anhydrous ethanol, a significant cost reduction can be achieved by designing more energy-efficient fermentations and separation processes for alcohol recovery.

One emphasis of our work is design of an efficient continuous-flow reactor system for bio-ethanol production. While continuous-flow reactor systems potentially offer great advantages in terms of increased volumetric productivity and continuous operation, the biofuels industry continues to prefer batch fermentation processes due to instabilities encountered in continuous-flow systems. We have developed a continuous-flow fermentation process based upon E. coli strain LY01, which is immobilized on a low volume fraction (<0.01) of poly(ethylene terephtalate) fibers of 40-55 μm diameter.

Stirred-tank and vertical column reactor designs have been compared with simple glucose feedstocks. Stable fermentation with glucose conversion greater than 90 % can be obtained with the tank reactor using 70 g/L glucose in the feed, while the immobilization of cells allows dilution rate and volumetric productivity to exceed the values obtained in chemostat mode of operation. The column reactor provides exceedingly high cell densities and productivities when glucose concentration in the feed is low (< 20 g/L), but suffers from instabilities and poor conversion when the feedstock concentration approaches commercially viable levels. Thus, the stirred tank with an immobilized cell

Texas Tech University, Lan Ma, December 2014

section appears to be the more attractive design. Current experiments involve assessing the performance of the continuous-flow system with mixed sugars.

The second emphasis of this dissertation is to design of an alternative, energy-efficient process for recovery of ethanol from dilute (15 to 50 g/L) fermentation effluents, which is vital to the biofuels industry, especially for cellulose-derived systems that produce low concentrations of product. A variety of separation techniques, such as pervaporation, have been studied due to high-energy requirements in distillation. To accelerate materials optimization, an innovative high-throughput screening approach has been applied to polymeric materials for pervaporation. A combinatorial approach is taken to screen a matrix of copolymer gels having orthogonal gradients in crosslinker concentration and copolymer composition. Using a combination of swelling in pure solvents (water or alcohol) and solvent mixtures (water + alcohol), the selectivity and distribution coefficients of alcohols in the gels can be obtained. Theoretically computed selectivities are compared with the values obtained by high performance liquid chromatography technique. Furthermore, the best performance gel for the pervaporation was picked up based on the measured diffusivities of the gels. This work is not only designed to achieve materials that exhibit unparalleled separation efficiency, but also to provide the industry with methods to rapidly tune material performance to meet the evolving needs of biofuel recovery process. The materials can be used in either pervaporation or else in the gel stripping methods being developed by Godbole & Hedden.

vii

Texas Tech University, Lan Ma, December 2014

List of Tables

Table 1-1. Important traits for EtOH production...... 11 Table 2-1. Batch and chemostat fermentation results of E. coli KO11 and LY01...... 39 Table 2-2. Specific growth rate of E. coli LY01 in the anaerobic batch fermentations. .. 50 Table 3-1. Characteristics of porous polymer scaffold materials determined by USANS...... 67 Table 4-1. List of parameters used in the batch fermentation model...... 107 Table 4-2. Compilation of fitted kinetic parameters (continuous fermentation without fibers)...... 115 Table 4-3. List of parameters used in the column reactor model...... 121 Table 4-4. Constants used for calculating the dimensionless parameters...... 127 Table 4-5. Dimensionless parameters...... 127 Table 5-1. Measured diffusivities of EtOH and water for selected membranes (boldface). The remaining diffusivities were estimated by double interpolation ...... 152 Table 6-1. Properties of cPE and cPE/CB nanocomposites with various vCB...... 172

viii

Texas Tech University, Lan Ma, December 2014

List of Figures

Figure 1.1. Lignocellulose structure...... 4 Figure 2.1. Ethanol fermentation with E. coli KO11 (black dots) and LY01 (red dots). 43 Figure 2.2. Cell culture with different glucose and LB concentrations...... 44 Figure 2.3 Anaerobic batch fermentations with different initial glucose concentrations. 46 Figure 2.4. Cell mass concentration from the anaerobic batch fermentation using 2.5 L stirred tank bioreactor with different initial glucose concentrations ...... 49 Figure 2.5. Anaerobic chemostat fermentation using 2.5 L stirred tank bioreactor with initial glucose concentrations of 70 g/L. Dilution rate was 0.026 h-1...... 53 Figure 2.6. Anaerobic chemostat fermentation using 2.5 L stirred tank bioreactor with initial glucose concentrations of 70 g/L. Dilution rate was 0.04 h-1...... 54 Figure 3.1. Scheme of fermentation process with fiber immobilized segmented vertical column reactor ...... 62 Figure 3.2. Low-vacuum SEM image of partly desiccated SPPS surface...... 66 Figure 3.3. USANS data for poly(hydroxyethylmethacrylate) SPPS ...... 66 Figure 3.4. Particle size distribution of SPPS after grinding...... 68 Figure 3.5. Control experiment without gels inside column...... 70 Figure 3.6. Glucose and EtOH concentrations exiting the column reactor ...... 71 Figure 3.7. Packed-bed section of glass column reactor...... 71 Figure 3.8. ICR fermentations with different dilution rates at 1.2 h-1 and 3.6 h-1...... 73 Figure 3.9. ICR fermentations with different dilution rates at 1.2 h-1 and 0.1 h-1...... 75 Figure 3.10. Column reactor fermentations conducted by increasing substrate concentration in the feed...... 76 Figure 3.11. Comparison of glucose conversion during the fermentation with SPPS particles of different diameters...... 78 Figure3.12. Column reactor fermentation conducted with 85 % internal porosity...... 79 Figure 3.13. (a): PET fibers (as received) used in packed bed section. (b): SEM of the E. coli LY01 cells immobilized onto a PET fiber...... 80 Figure 3.14. Flask fermentations with and without fibers inside. (a): Glucose and ethanol concentrations; (b): Optical cell density...... 82

Texas Tech University, Lan Ma, December 2014

Figure 3.15. ICR fermentation with PET fibers as immobilization scaffold. Glucose and LB concentrations in the feed tank were 20 g/L and 25 g/L, respectively. Dilution rate was 0.84 h-1. (a): Residual glucose and ethanol concentrations; (b) Immobilized cell dry mass density...... 84 Figure 3.16. ICR fermentation with PET fibers as immobilization scaffold. Glucose and LB concentrations in the feed tank were 30 g/L and 40 g/L, respectively. Dilution rate was 0.84 h-1. (a): Residual glucose and ethanol concentrations; (b) Immobilized cell dry mass density...... 85 Figure 3.17. ICR fermentation with cells immobilized on PET fibers. Initial glucose and LB were 50 g/L and 75 g/L, respectively...... 87 Figure 3.18. Fermentation with an ICR whose outlet is connected with a tank reactor. . 90 Figure 3.19. Reactor was running with fibers packed between two porous plates. Agitation was provided by the cell-impeller...... 92 Figure 3.20. ICR fermetnation with 2.5 g immobilized fiber inside...... 93 Figure 3.21. ICR fermentation with different amount of fibers...... 95 Figure 4.1. Experimental and simulation results of the sugar and ethanol concentrations at different initial glucose concentrations ...... 108 Figure 4.2. The effect of initial glucose concentrations on (a) the saturated growth

constant (Ks (g/L)) and (b) the saturated ethanol production constant (Ksp (g/L))...... 109 Figure 4.3. Experimental results and simulation of the sugar and ethanol concentrations at different initial glucose concentrations; ...... 110 Figure 4.4. Fitting of the anaerobic chemostat fermentation with glucose feed concentration of 70 g/L. Dilution rate was 0.026 h-1...... 113 Figure 4.5 . Fitted and experimental results of the chemostat fermentation. Dilution rate was 0.04 h-1...... 116 Figure 4.6. Predicted residual glucose and ethanol concentration profiles for ICR with D= 0.04 h-1 and an glucose feed concentration of 70 g/L...... 117 Figure 4.7. Fitting curve of the cell mass concentration versus column axial distance. 122 Figure 4.8. Predicted glucose and ethanol concentrations of the column reactor with different glucose feed concentrations: (a) 10 g/L and (b) 20 g/L...... 123

Texas Tech University, Lan Ma, December 2014

D Figure 4.9. Predicted ethanol concentrations with different dispersion numbers ( ): uL 0.03, 0.005 and 0.001. Glucose feed concentration was 10 g/L for all the calculations. 124 Figure 4.10. The residual glucose concentration inside the column reactor by assuming different percentages (10 %, 50 %, 70 % and 100 %) of the immobilized cells are the viable cells...... 125 Figure 5.1. Equilibrium swelling of copolymer networks ...... 146

Figure 5.2. Experimentally determined distribution coefficient of EtOH ka (a), and

equilibrium selectivity aa* (b) for copolymer networks...... 148 Figure 5.3. Predicted distribution coefficient (a) and selectivity (b) for ethanol...... 150

Figure 5.4. Permeability selectivity for EtOH (αa) ...... 153

Figure 5.5. Flux of EtOH though the membranes with aa > 1.0. Membrane thickness is 1 mm...... 155 Figure 6.1. Illustration of the 2W-SME in the cPE/CB nanocomposite with vCB of 5.0 vol. %...... 170 CB Figure 6.2. Dependence of Rf, Rr, Ra, R'r, and Xc on v ...... 171 Figure 6.3. Shear storage modulus (G', MPa) vs. frequency (ω, Hz) at 130 °C for cPE and cPE/CB nanocomposites with various vCB...... 172 Figure 6.4. (a) Influence of applied stress on the 2W-SME for the cPE/CB nanocomposite with vCB of 1.0 vol. %. (b) Relationship between applied stress and

observed elongation during cooling (ε2-ε1)...... 177 Figure 6.5. 2D-WAXD patterns of the cPE/CB nanocomposites with vCB of 1.0 vol %...... 180 Figure 6.6. Diffracted X-ray intensity vs. azimuthal angle (χ) for cPE/CB nanocomposites with vCB of 1.0 vol % before stretching...... 181

Texas Tech University, Lan Ma, December 2014

Chapter 1 Introduction

1.1 Biofuel as a renewable fuel

The increasing demand on global supplies of fossil fuel resources causes

environmental problems and, with prices rising, future energy needs must be met in part

or entirely by fuels derived from renewable resources [1]. With the advances in metabolic

engineering and synthetic biology, a new sustainable production of biofuels can be

constructed [2]. Methanol, ethanol (EtOH), propanol, and butanol can be used for motor fuels. However, only methanol and ethanol fuels are economically suitable for internal combustion engines and offers efficiency advantages over gasoline [3, 4]. Compared with

gasoline, ethanol has a higher octane number, flame speed, and heat of vaporization as

well as broader flammability limits, which allows for a higher compression ratio, shorter burn time, and a cleaner burning engine [5]. Furthermore, complete combustion of EtOH

produces only water and CO2 as by-products [3]. The highest blends level of denatured

EtOH (E85) reaches to 85 % at fueling stations in the U.S, mainly in the Midwest. With

2,224 fuel stations already selling E85 in the U.S. and new stations appearing daily [4], the continued growth of EtOH as an automotive fuel appears likely.

Commercial-scale EtOH production relies mostly upon catalytic reforming of ethylene.

However, 93 % of EtOH in the world in 1995 was bioethanol produced by fermentation.

Bioethanol is an alternative fuel derived almost entirely from renewable sources of feedstocks. EtOH fermentation normally involves three steps: (1) formation of fermentation sugars; (2) microorganisms consume sugars to produce EtOH; (3) EtOH separation and purification. Most microorganisms can only use 6-carbon sugars like

Texas Tech University, Lan Ma, December 2014

glucose. It is easier to convert the biomass materials with high level glucose or precursors to EtOH. However, most sugar materials are human food, which are expensive to be used for EtOH production. Another potential fermentation material is starch, which is made up of long chains of glucose molecules. Commonly, for ethanol production, starchy materials include cereal grains, potato, sweet potato, and cassava.

Production of bio-EtOH from renewable, non-food feedstocks will eliminate the competition with food sources. However, the cost of cellulosic ethanol is higher than the ethanol produced from sugar or starch feedstocks [6]. There are three costs can be reduced for industrial-scale EtOH fermentation process: raw material cost, capital equipment and operating costs. Among these costs, the feedstock cost comprises a very large fraction of the overall ethanol production. In order to improve both existing and future biofuels economics, we need better process intensity and multiple technologies [7].

Fermentable sugars from enzymatic hydrolysis of cellulose must be less than 4-5 cents/lb if we want to get an economically viable fermentation compared with sugars derived from corn [8].

1.2 Cellulosic Feedstocks

The varied raw materials used for EtOH fermentation can be classified in to three main types: sugars, starches, and cellulosic feedstocks [9]. Microorganism can typically consume 6-carbon sugars, therefore, glucose is the easiest one to convert to ethanol. The other biomass feedstocks involving rich sugars are: sugar beet, sweet sorghum, various fruits, and sugar cane. Brazil is the largest single producer of sugar cane, therefore, EtOH production from Brazil is mainly from sugar cane [10]. However, these sugar feedstocks are still in the human food chain, which is expensive to be used for ethanol production.

Texas Tech University, Lan Ma, December 2014

Starch is a biopolymer, made up of long chains of glucose molecules. Starchy materials are required to break down into simple glucose molecules to produce EtOH by fermentation. Examples of starchy materials commonly include corn, wheat, potatoes, cassava root, corn stover [11, 12]. Starchy materials are used widely in North America and Europe [10]. The bio-ethanol production of United States is especially from corn, and production is concentrated in Midwestern state, which has abundant corn supplies [13].

However, like sugar materials, starchy materials are also in human food chain and kind of expensive to be used for bio-EtOH production. Therefore, cellulosic materials, the most abundant reproducible resources and outside the human food chain, are a very good choice.

Cellulosic materials, also called lignocellulosic biomass, are comprised of cellulose, hemicellulose and lignin. The structure of lignocellulose is shown in Fig. 1.1 The examples of cellulosic materials are wood, paper, agricultural residues, waste sulfite liquor from pulp and other fibrous plant material [1]. Cellulose consists of long chains of glucose molecules similar to starch molecules, but has different structural configuration and are more difficult to break into fermentation sugars. In contrast to cellulose, hemicellulose is comprised of long chains of glucose involving both 6-carbon sugars

(glucose, mannose, and galactose) and 5-carbon sugars (xylose) [14]. The sugar composition is different depending on the type of plant. Lignin, which does not contain any sugars, provides structural support for the plant and make enclosed cellulose and hemicellulose molecules difficult to reach. Also, the raw lignocellulosic biomass is a highly crystalline complex that cannot be directly hydrolyzed for conversion to fuels.

Fermentation of starches and cellulose materials are more complex than fermentation of

Texas Tech University, Lan Ma, December 2014

sugars because starches and cellulose need to be first hydrolyzed into fermentable sugars, then converted by microorganisms to EtOH.

Figure 1.1. Lignocellulose structure. Adapted from the figure draw by Mohamadmahdi Kowsari [15]. Lignocellulosic ethanol production normally involves four steps: (1) pretreatment; (2) hydrolysis; (3) fermentation; (4) ethanol separation/distillation [16]. The purpose of pretreatment is to break the crystal structure of lignocellulose and expose the cellulose and hemicellulose molecules for hydrolysis [17]. There are two pretreatment methods: physical or chemical. Physical pretreatment involves steam explosion, liquid hot water

(LHW), milling, radiation, or freezing, which all need high energy consumption [18, 19].

Chemical pretreatments involve ammonia fiber expansion (AFEX), acid, alkaline, wet oxidation, and organosolv pretreatment to break and dissolve crystal structure [20].

The most basic and common methods to hydrolyze the cellulosic biomass are acid and enzymatic hydrolysis. Acid hydrolysis, involving dilute acid and concentrated acid hydrolysis process, is one of the oldest and most applied technologies [21]. Dilute acid process normally needs high temperature and pressure. The advantage of dilute acid

Texas Tech University, Lan Ma, December 2014

hydrolysis is its quick reaction rate, which can be used in the continuous fermentation.

The disadvantage is sugar decomposition, low sugar yield and some other by-products like furfural, which are poisonous for later fermentation [22]. Concentrated acid hydrolysis has high sugar recovery efficiency; however, the reaction process is slow.

Therefore, a cost effective acid recovery system is difficult to develop [23]. After acid hydrolysis, which is normally used as a prehydrolysis (or pretreatment), hemicelluolose degrades to monosacchrides and cellulose receives more access to cellulase for enzymatic hydrolysis. Cellulose is hydrolyzed by three classes of enzymes: endoglucanase, exoglucanase, and β-glucosidase [19, 24]. Endoglucanase (EG, EC 3.1.2.4) attacks the endogenous part of cellulose chain into insoluble cellooligosaccharides [25].

Exoglucanase or cellobiohydrolase (CBH, 1, 4- β –D-glucan cellobiohydrolase, or EC

3.2.1.91) attacks the ends of polymer, releasing cellobiose that finally degrades into glucose by β-glucosidase [26]. The advantages of enzyme hydrolysis are: (1) high efficiency and controllable byproduct production; (2) mild process conditions; and (3) relatively low process energy requirements [9]. However, the cost of enzymes is still high like the other processes and researchers are continuously working to find a low cost process. Another ethanol production process is thermochemical process, which is a hybrid thermochemical and biological system. Biomass materials are thermochemically gasified to a mixture of gas, and then a specific process condition is required for microorganism to convert the synthesis gas to EtOH [27].

The EtOH production cost can be decreased by reducing either raw materials cost or cellulose enzyme cost. The efficiency of enzyme hydrolysis can be improved from several areas: (1) increase the activity of cellulose enzyme; (2) raise the amount of

Texas Tech University, Lan Ma, December 2014

enzyme production; (3) develop recombinant microorganism, which have broad substrate

utilization and can be applied in simultaneous saccharification and fermentation process

(SSF) [21] [28]. SSF process, which offers important potential advantages in reducing

costs, is extensively studied recently [21]. In this process, sugars produced from cellulose

hydrolysis are converted to EtOH simultaneously, which improves the inhibition problem

[29]. SSF also involves the other advantages such as lower enzyme requirements, shorter

process time, lower requirements for sterile conditions, and less reactor volume because it

only requires a single reactor [30]. In order to have SSF system, a genetic engineered

microorganism that can convert xylose and or/ pentose to EtOH are preferred. The first

generation large-scale bio-EtOH productions from S. cerevisiae has been proved in

industry. Therefore, genetic techniques are gaining interest on expressing heterologous

cellulases in S. cerevisiae or E. coli to create new recombinant cellulolytic microorganisms. Wood et al. reported the expression of endoglucanase genes from

Erwinia chrysanthemi P86021 in E. coli KO11 [31]. Den Hann et al. reported co- expression of endoglucanase and β-glucosidase from T. reesei and Saccharomycopsis fibuligera in S. cerevisiae [32]. It showed that after pretreatment, the amorphous cellulose was hydrolyzed to glucose efficiently by this recombinant yeast. Similar work was reported by Fujta et al. that two cellulolytic enzymes –endoglucanase and β-glucosidase separately from T. reesei and A. aculeatus were expressed in S. cerevisiae [33].

Further investigations of lignocellulosic - production still face several challenges such as maintaining the stability of genetically engineered microorganisms in the commercial scale fermentation, developing more efficient pretreatment technologies, and integrating the optimal components into economic EtOH production system [30]. In order to have an

Texas Tech University, Lan Ma, December 2014

industrial-scale bio-EtOH production, all the strategies mentioned above should be integrated with an approach toward technical and economic feasibility [34, 35].

1.3 Development of ethanologenic microorganisms

5-carbon sugars comprise a high percentage in cellulosic materials; however, most of microorganisms can only consume 6-carbon sugars. In order to get EtOH production more efficiency and economics, it is important to find a single microorganism that is suitable for single-stage fermentation of hexose and pentose mixture. The ethanologenic microorganisms that currently have been engineered and show the most promise in industry are E. coli, Klebsiella oxytoca and Zymomonas mobilis [36].

1.3.1 Engineering and performance of E. coli

Historically, Saccharomyces is the main microorganism used for commercial EtOH production. Saccharomyces and Z. mobilis are strains with homo-ethanol pathway because they comprised of pyruvate decarboxylase (pdc) and alcohol dehydrogenas (adh), produced two EtOH molecules per glucose [37]. However, both organisms cannot utilize pentose sugars, which are the major components of hemicellulose. E. coli can utilize both hexose and pentose sugars. However, it lacks native ability to produce EtOH as the major product [38]. The ability to ferment a wide range of sugars, no requirements for complex growth factors and ease of genetic manipulation and prior industrial use for protein production made E. coli an obvious choice for metabolic engineering of biocatalyst for lignocellulosic EtOH production [37].

Therefore, researchers expected to express pdc in E. coli to produce only EtOH. The first recombinant expression of Z. mobilis homoethanol pathway in E. coli was described

Texas Tech University, Lan Ma, December 2014

by Ingram et al. in 1987. A pet (production of EtOH) operon was introduced into E. coli

in plasmids expressing both adhІІ and pdc genes, which allows EtOH as the main

fermentation product [39]. However, this initial engineered strain is not stable and

plasmids carrying the adh and pdc genes were easily lost in the absence of antibiotics.

Therefore, Prof. Ingram's research group tried to integrate the PET operon into

chromosome at the pyruvate formate lyase (pfl) locus with an antibiotic resistance marker

in order to eliminate an enzyme competing for pyruvate. Also, succinate pathway was

terminated by eliminating fumarate reductase (frd) and further increased EtOH yield. The

final strain KO11 could produce EtOH at a yield of 95 % in rich medium [40]. KO11 has

been evaluated for fermentation of hemicellulose hydrolysates of agricultural residues

such as steep liquor and crude yeast autolysate, and achieved over 40 g EtOH L-1 within

48 h [41]. There are studies that have raised concerns regarding the phenotypic stability

of KO11 for high EtOH production. Lawford and Rousseau reported that the strain lost

all of its ability to produce EtOH in less than 12 generations when grown on glucose and

mannose in serial batch culture, even when 40 µg ml-1 chloramphenicol (Cm) was

included in the medium [42]. Also, strain KO11 lost its ethanologenicity rapidly when

grown in a continuous culture with 25 g/L glucose or xylose and it produced a lot of organic acids [43]. In contrast with Lawford and Rousseau’s results, Dumsday et al. reported that relative stability of E .coli KO11 during chemostat culture on glucose [44].

It is reported that KO11, which was grown from continuous glucose-feed culture has high

EtOH yields, but not from xylose or glucose/xylose mixtures. When xylose alone was used as feeding source, EtOH yields started to decrease after 5 days and cells lost Cm hyper-resistance (600 mg/l) after 30 days.

Texas Tech University, Lan Ma, December 2014

Strain KO11 has similar EtOH production rate with yeast, however, the EtOH

tolerance is lower than commercial yeast. In complex medium, KO11 grew poorly in the

presence of 35 g L-1 EtOH and only 10 % survival after 0.5 min exposure to 100 g L-1 EtOH [37, 45]. A 3-month metabolic evolution with a combination of liquid and

solid medium was used to isolate high EtOH tolerance strain. Strain LY01 was

obtained after evolution, which can grow in the presence of 50 g L-1 EtOH and exhibited over 50 % survival from 0.5 min exposure to 100 g L-1 EtOH. Both KO11 and LY01 can

produce EtOH from multiple substrates; however, they only attained high EtOH yields in rich media. It is reported that KO11 produced 45 g L-1 EtOH from 100 g L-1 glucose in

72 hours in rich medium compared with 30 g L-1 in 96 hours in minimal medium [46].

Also, the cell mass of LY01 and produced EtOH were tenfold lower attained in minimal

medium than in rich medium [47]. Another limitation is antibiotic dependence. The

stability to retain ethanologenicity over time without antibiotic is important to reduce cost.

The continuous fermentations of E. coli KO11 have been studied and reported [43, 44,

48]. However, no report about continuous fermentation of LY01 was published.

1.3.2 Engineering and performance of other microorganisms

Z. mobilis has a homoethanol fermentation pathway and high EtOH tolerance up to 120

g L-1[49]. However, it only ferments glucose, fructose and sucrose. Researchers have

been working to transform Z. mobilis with a plasmid encoding the genes involving ability

to ferment xylose and arabinose. Recently, the newest strain AX101 which can ferment

both arabinose and xylose was developed by Zhang and collaborators [50]. Z. mobilis

fermentation has byproducts such as sorbitol, acetoin, glycerol, acetic acid and

Texas Tech University, Lan Ma, December 2014

extracellular levan polymer. The major disadvantage of AX101 cultures is their low

tolerance to acetic acid, which is commonly found in hydrolysates .

In addition to E. coli, Ingram and colleagues have also transformed pet operon to K.

oxytoca strains, which are especially well suited for cellulosic EtOH production [40].

The best mutant isolated (P2) has the ability to ferment either glucose (100 g L-1) or cellobiose (100g L-1) to EtOH with yields of 44-45 g/L within 48 h. It has been

successfully tested in multiple feedstocks including sugarcane bagasse [51], corn fiber

[52] and mixed office paper [53]. Recently, variant engineering strains have been constructed that express endoglucanase based on strain P2. Doran et al.[51] compared strain K. oxytoca P2 with E. coli KO11 in the fermentation of sugar beet pulp with simultaneous enzymatic hydrolysis of cellulose. Actually, strain KO11 produced 40 % more EtOH than strain P2.

Yeast, particularly Saccharomyces spp., is the workhorse for glucose-based feedstocks to EtOH in industry, traditionally was unable to ferment xylose or utilize xylose for growth. The first successful engineered Saccharomyces yeast was achieved at Purdue

University in 1993, which can effectively convert both glucose and xylose to EtOH by developing two plasmids pLNH32 and pLNH33 that involve cloned xylose-metabolizing genes [54]. In 1995, a more perfect yeast 1400 (LNH-ST) was engineered, which requires no selection pressure to maintain its ability for xylose-fermenting at any stage of growth.

At that time, another two stable glucose-xylose-cofermenting Saccharomyces yeast strains were also developed based on the plasmid LNH-ST: 295A (LNH-ST) and 424A

(LNH-ST). Compared with E. coli, yeast is able to survive in a more severe environment

Texas Tech University, Lan Ma, December 2014

without requirements of rich medium. Therefore, most biofuel companies prefer to use

engineered yeast instead of E. coli for the EtOH production.

1.3.3 Comparison of the microorganisms fermentation performance

Among the ethanologenic strains, E. coli, Z. mobilis, and S. cerevisiae have been widely studied and applied for cellulosic EtOH production. Each strain has its unique advantages and disadvantages. Economically-attractive strain should meet the industrial requirements, which is summarized in Table 1.1 [55, 56].

Table 1-1. Important traits for EtOH production. Adapted from (Ming W. Lau, Christa Gunawan, Venkatesh Balan and Bruce E. Dale. Biotechnology for biofuels 2010, 3, 11- 21.)

Trait Requirements Ethanol yield >90 % of theoretical Ethanol tolerance >40g L-1 Ethanol productivity >1 g L-1h-1 Robust grower and simple growth requirements Inexpensive medium formulation Able to grow in undiluted hydrolysates Resistance to inhibitors

In order to compare the fermentation performances of these strains systematically,

Bruce E. Dale et al. established a common platform to obtain the fermentation parameters

using S. cerevisiae 424A (LNH-ST), Z. mobilis AX101 and E. coli KO11 from corn

steep liquor (CSL) and ammonia fiber explosion (AFEX) pretreated corn stover [57].

All three ethanologens are able to produce EtOH with a rate greater than 0.42 g/g

consumed sugar, 40g L-1 EtOH concentration and 0.7 g L-1 h-1 (0-48 h) productivity from

a CSL-supplemented co-fermentation. S. cerevisiae 424A (LNH-ST) shows slower

fermentation rate in xylose-only CSL fermentation and requires higher nutrients than the

Texas Tech University, Lan Ma, December 2014

other two microorganisms[57]. E. coli KO11 and Z. mobilis AX101 are relatively more effective to ferment EtOH from consumed sugars.

The development of future EtOH-producing microorganisms should consider what parameters are most important for industrial fermentations. The critical traits are pentose utilization, genetic stability, high EtOH productivity and tolerance, and hardiness, which means the fermentation can perform even in imperfect large industrial process.[56]

Genetic stability is more significant for continuous fermentation because of large number of generations during the process.

1.4 Fermentation reactor design

1.4.1 Continuous immobilized cell reactor design

Mode of bioreactor operation can be classified as batch, fed batch, and continuous.

Microorganisms growing inside bioreactor are suspended or immobilized. Immobilized cell bioreactors (ICR) include moving bed biofilm reactor, packed bed, fibrous bed or membrane reactor [58]. Although anaerobic fermentation based upon genetically engineered microorganisms are promising, fermentation economics are adversely affected by the need for large-scale batch or semi-batch fermentation reactors. Traditional

EtOH fermentation is using batch reactor with suspended cells, which is expensive to produce and operate due to high down-time. In addition, EtOH concentration is required to keep below a threshold (often in the range of 50 to 100 g/L) preventing cell death or product inhibition. By contrast, continuous fermentation with immobilized cells is recognized as a good approach to improve the economics of EtOH fermentation. The operation of continuous reactor is easy to control with steady-state fermentation

Texas Tech University, Lan Ma, December 2014

conditions [59]. The other advantages involve higher conversion rates, faster

fermentation rates, reduced product losses and improved produce consistency [60].

Especially for continuous ICR fermentation, microorganisms are attached to a solid support or scaffold that permits exchange of substrate and products with the surrounding liquid. ICR can provide high cell densities, normally 5 to 10 times higher than free-cell suspensions [61], which in combination with high dilution rates, leads to high volumetric efficiency [62]. In a column-type ICR, EtOH is flushed away continuously, which keeps its concentration low near the reactor inlet, reducing product inhibition [63, 64].

Compared with batch reactor, there is no cost on separating cells from the product mixture in ICR mode [64, 65]. Especially in the lignocellulosic fermentation, ICR could prevent reactor washout when inhibitor or toxins exists [66]. The inhibitor problem can be mediated by continuous flow, as long as dilution rate is kept below a certain level [67].

Based on the advantages mentioned above, the size of ICR system is about 5 or 10 times smaller than batch or semi-batch reactor for the similar production capacity. Furthermore, the sugar conversion of ICR is more efficient, which reduce raw material cost that is the largest annual operating cost for EtOH producer.

1.4.2 Immobilization scaffold materials and methods

Cell immobilization is a technique to fix cells in a distinct matrix to separate cells from the bulk phase containing substrates and products, but permits exchange of substrates, products, and inhibitors, etc. It is recognized that microbial cell density is significant to attain high volumetric productivity. Immobilized cell system allows continuous mode without cell washout and decouples microbial growth from cellular synthesis of favored compounds. There is evidence to indicate that bound-cell system is more tolerant to the

Texas Tech University, Lan Ma, December 2014

perturbed reaction environment and similarly less sensitive to toxic substances present in the liquid medium [68]. Generally, cell immobilization methods are classified into four major categories: (1) attachment to a surface; (2) entrapment within a porous matrix; (3) containment behind a barrier; and (4) self aggregation (Fig. 1.2) [69].

The adhesion between cells and solid support are normally due to natural adsorption or chemically bonding. The first example of cell immobilization was reported by Hattori and Furusaka about E. coli binding on to an ion exchange resin [70]. Subsequently, the adsorption of cells was found on different substrate such as woodchips, kieselguhr, glass ceramic and plastic materials, etc. The reasons for those adsorption phenomena are attributed to electrostatic interactions (van der Walls force), ionic (Lewis acids/base) and hydrophobic interactions [71]. The industrial application of surface attachment of microorganisms has been examined in wastewater treatment and mammalian cell culture.

Mammalian cell was adhered on roller bottles or microcarriers, which are small porous beads with 100-200 μm diameters [72]. The cell laden microcarriers were then placed in a packed bed reactor or continuous stirred tank reactor [73].

Cells can also bind to a surface through covalent bonding or cross linking. The formation of covalent bonding needs activated inorganic support and cells in the presence of a binding agent. A widely known reaction is introducing the reactive organic group on inorganic silica surface such as treating the silica surface with glutaraldehyde and isocyanate or chelation to metal oxides [74]. However, the toxicity of chemicals used in the cross-linking limits its application. There are no barriers between cells and bulk solutions in the cell attachment system. Therefore, cells are possibly detached and relocated, cannot be used where cell-free effluent is required. Also, it is difficult to

Texas Tech University, Lan Ma, December 2014

control the thickness of biofilm, which is decided by the medium flow rate and type of strain.

The second category is cell entrapment within porous matrix. Two methods are available. One is that cells diffuse into a preformed matrix and are trapped inside after cells begin to grow. There are many types of porous matrix such as sponge, sintered glass, cordierite, bricks, ceramics, silicon carbide, polyurethane foam, chitosan and stainless steel fibers [68, 75, 76]. The preformed matrix normally can resist more compression and disintegration in the stirred vessels. Also, cell adsorption is not harmful. However, cell volumetric packing density is not high compared with other immobilization method [69].

In the second method, porous matrix is synthesized in situ around cells. A most

extensively method is gel entrapment. Entrapment by ionic network formation, such as

alginate beads, is formed by dripping cell suspension-sodium alginate solution mixture

into a stirred calcium chloride solution [77]. Another material is κ-carrageenan, which has similar procedure with alginate, using either calcium or potassium. Particle hardness is increased by adding potassium [68]. Several other materials such as agar and agarose are also used for cell immobilization by precipitation [78]. Gel entrapment by polymerization is widely carried out using polyacrylamide. The hardness and the pore size of polyacrylamide gel are influenced by the ratio of cells to acrylamide and the content of acrylamide. The particle size will determine activity, stability, and the pressure drop inside reactor.

Gel entrapment is accompanied with some problems like bed compaction, rupture of the gel by escaping gas (causing cell washout), and improper CO2 ventilation (leading to

floating of the gel beads) [79]. Transport of nutrients is resistant due to mass transfer,

Texas Tech University, Lan Ma, December 2014

which creates a stress factor for secondary metabolite synthesis. In a study of S.

cerevisiae immobilized in hollow-fiber polysulfone and polypropylene membranes,

Robertson et al. found that EtOH production was inhibited and fermentation efficiency was reduced by holdup of CO2 in the reactor [62]. Also, Taherzadeh et al. found that calcium alginate beads were unstable in fermentation of hydrolyzed cellulosic material by

S. cerevisiae, unless a source of Ca2+ was added continuously during fermentation [67].

Some modified methods were used for cell immobilization. Vorlop et al. reported that

drying of immobilized cell sphere could increase the beads' hardness [80]. Also, stability of beads can be improved by glutaraldehyde treatment or coating gel particle with catalyst-free polymer. Recoating the calcium alginate beads with a double layer in a two-

step preparation procedure was reported by Yamagiwa et al. [81]. Zhengqiang Li et al.

designed a "fish-in-net" approach for encapsulation of Z. mobilis cells in mesoporous

silica based materials[82]. Promising results were obtained with some porous materials

such as Loofa sponges [83]. The use of porous soft materials may be an optimal scaffold

for ICR. The presence of (interconnected) porosity, gel-encapsulated systems, and CO2

bubbles can escape easily through pores. Loofa sponges are readily available and

inexpensive, but their primary disadvantage is that the pore size and connectivity cannot

be controlled directly.

Containment of cells behind a barrier can be performed with semi-permeable

membrane or microcapsule. It is normally used when cell-free product or high molecular

weight product are required. A simple barrier can be a liquid/liquid interface between two

immiscible fluids. Cell suspension was emulsified with a surfactant into a hydrocarbon

solvent, and then the surfactant-enclosed droplets were resuspended in the aqueous

Texas Tech University, Lan Ma, December 2014

substrate phase. Cells did not leak from the aqueous phase to organic phase [69]. Semi-

permeable membrane is similar to microcapsules. Cells stay in one side of the barrier while supplied nutrients and products are removed on the other side of the barrier.

However, this type of membrane is expensive, and normally have mass transfer limitation and possible membrane fouling caused by cell growth.

Based on the definition of cell immobilization, cell aggregation or flocculation can also be considered as immobilized. Large size of cell aggregation is possible used in the packed bed reactor, fluidized beds and CSTR due to its simplicity and low cost. However, aggregation will be influenced by a lot of factors such as nutrient condition, pH,

fermentation temperature, and storage conditions [84]. It is difficult to control and predict

the impact of these parameters on cell aggregation. Artificial flocculation agent or cross- linker can enhance aggregation process for cells that do not naturally flocculate.

1.4.3 Immobilized cell reactor for EtOH production

Reactor configurations for immobilized cells normally involve three categories: (1) particles or large surface that is fixed in space; (2) large moving surfaces and (3) particles

suspended and mixed in solution [69]. Therefore, immobilized cell reactor is separated

into several types: packed bed reactors, mechanically agitated and airlift reactors, bubble

column reactors, hollow fiber membrane reactors, and flat bed reactors. The most

common immobilized cell reactor for EtOH production is a three-phase packed bed

reactor contains gas (CO2), liquid (medium) and solid (immobilization material). The

advantages of packed bed reactor include simplicity of operation and high mass transfer

rates. However, it is difficult to control cell growth, and pressure increases due to

compression of the particles. Also, there is a lot of inadvertent formation of gas bubbles.

Texas Tech University, Lan Ma, December 2014

Promising results regarding ICR fermentations have been obtained on the laboratory or

pilot plant scale as early as the 1980s [68], some examples of which follow. Arcuir et al.

(1982) used glass-fiber pads to attach Z. mobilis for continuous EtOH production, which

lasted for 28 days and got the maximum EtOH volumetric productivity of 152 g/(L·h)

[85]. Nojima et al. (1983) developed an ICR pilot plant for continuous EtOH production,

which was based upon a photo-crosslinkable resin scaffold. Their system operated for

8,000 h continuously, producing 250 L of EtOH per day. Nagashima et al. (1984) also

operated a pilot plant based upon alginate-entrapped yeast cells, which was capable of

producing 600 L per day of EtOH in concentrations of up to 9.0 % by volume [86]. Amin

and Doelle (1990) used polyurethane foam to encapsulate Z. mobilis in a vertical rotating

ICR [87], resulting in EtOH productivity of 63 g/(L·h) via glucose fermentation for a

period of 72 days. Gil et al. (1991) developed an ICR system based upon yeasts adsorbed

onto an aluminum silicate scaffold, which allowed production of EtOH from sucrose in

concentrations of up to 98 g/L for a period of 2 years [88]. Ghorbani et al.(2011)

designed an ICR system with sodium-alginate immobilized yeast, which got maximum

EtOH production of 19.15 g/L from cane molasses [89]. A similar operation was done by

Mongkolkajit et al. (2011) using alumina-doped alginate gel. Alumina particles not only

improved the gel porous structure, but also provided good mechanical strength and high

immobilization yield, which achieved 19.3 g/L immobilized cell concentration after 30

days of continuous fermentation and EtOH productivity of 12.6 g/(L·h) at the dilution

rate of 0.28 h-1 [90]. Yan et al. (2012) presented a study to examine EtOH production from concentrated food waste hydrolysates using yeast immobilized on corn stalks, which allowed maximum EtOH volumetric productivity of 43.45 g/(L·h) [91]. Chen et al.

Texas Tech University, Lan Ma, December 2014

developed a fibrous packed-bed bioreactor with high specific surface area and good

adsorption efficiency of yeast resulting in a maximum EtOH concentration of 108.14 g/L

and a productivity of 14.71 g/(L·h) at a dilution rate of 0.136 h-1 with 250 g/L glucose

[92]. A graft copolymer of carboxymethylcellulose-g-poly (N-vinyl-2-pyrrolidone) was

developed by Gokgoz et al. (2013) for Saccharomyces bayanus immobilization.

Maximum EtOH concentration and productivity were 98.78 g/L and 8.23 g/(L·h) and this

material can be used seven times without losing its activity [93].

Additional studies are focused on improving reactor performance due to some

combination of higher cell density or increased cell lifetime, with higher EtOH yields,

higher volumetric production rates and long reactor running time. There is ample

evidence suggesting that continuous flow ICR systems could become the preferred

fermentation technology for future cellulosic EtOH production from biomass. However,

the recently emerged commercial cellulosic EtOH industry has so far relied upon free-cell fermentors instead of ICR systems. Arguments against ICR systems usually amount to mass transfer limitations, gas holdup issues, or instability of the immobilization matrix.

1.5 Bioethanol separation

In order to make an economically competitive fermentation process, one has to

overcome yet another major drawback of high product recovery cost. The cost mainly

includes the equipments used for separation and overall energy consumption during

recovery. Almost 80 % of the equipment cost is spent on the separation equipments and

more than 25 % of the total manufacturing cost is utilized for the energy consumption of

plant. The traditional distillation method used for ethanol separation after fermentation is

only economically favorable when ethanol concentration is above a certain threshold

Texas Tech University, Lan Ma, December 2014

(~70 g/L for EtOH) [94]. The highest achievable ethanol concentration is 140 g/L or so.

However, biomass-derived ethanol is usually below 5 wt. % due to the presence of toxic

inhibitory compounds in the feed [95].

For this reason production of EtOH using low-cost recovery system has been

considered as a necessary change in order to improve economics of the process. One

approach is separating EtOH during the continuous fermentation in order to reduce the

inhibition effect. There are several proposed EtOH recovery techniques: vaccum

distillation, solvent extraction, stripping by CO2, and membrane pervaporation [96].

Pervaporation is very competitive among all these techniques because it has many

advantages such as increased energy efficiency, cost effectiveness, and protection of

environment [96, 97]. Pervaporation is a process in which a liquid mixture containing

two or more components is placed in one side of nonporous polymeric or inorganic

membrane while the other side is applied with a vaccum or gas purge [98]. Based on the

properties of membrane (hydrophobic or hydrophilic), organic compound or water will

preferentially permeate through membrane and evaporate into vapor phase. The

performance of membrane can be characterized by flux and selectivity. Although many

membranes have been studied for EtOH recovery, there are only a few membranes

meeting the industrial requirement for the purpose of recovering EtOH from diluted fermentation broth.

The most representative membrane is PDMS, whose hydrophobic property makes

PDMS a good material for alcohol removal from low alcohol concentration solution. The

fluxes and permeability selectivity of unmodified PDMS membrane range from 1 to

Texas Tech University, Lan Ma, December 2014

1000 g m-2 h-1 and less than 10 for removal of EtOH by the pervaporation process [99]. A

flux of 150 g·m-2·h-1 and permeability selectivity of 10.3 for EtOH to water would be cost

competitive for a commercial-scale EtOH plant [100]. The energy required to recover

EtOH by pervaporation is only comparable with the energy required by distillation when

permeability selectivity is around 20 or higher than 20 [97]. In order to improve PDMS

performance, a lot of attempts have been made on PDMS modification such as using

different supports (PS, PTFE and CA etc.), and synthesizing PDMS block or graft

copolymers with other polymers. Miyata et al. has reported a bulk copolymerization of

styrene and divinylbeneze in the PDMS network to form an interpenetrating polymer

network (IPN) membrane (PDMS/PSt IPN) [101]. Zhe et al. designed a novel silicone

membrane containing both PDMS and ladder-like phenylsilsesquioxane segments, which

has a high EtOH separation factor and increased EtOH flux [102]. A high EtOH- permselective membrane was synthesized by Kashiwagi et al. by surface treatment of silane compounds on thin PDMS membrane, which can get a separation factor of 18 and a flux of 150 g·m-2·h-1 [103]. There are also some other reported polymers used for EtOH

recovery membrane design such as poly (1-trimethylsilyl-1-propyne) (PTMSP) [104], polydimethylsiloxane-imide (PSI) copolymers, a hydrophobic polyoctylmethyl siloxane

(POMS) [105] and poly (ether block amide) [106].

Inorganic membrane such as hydrophobic zeolites has shown higher separation performance than PDMS [97]. Furthermore, zeolite membrane has other advantages such as low swelling, high chemical resistance, and thermal stability. To combine the advantages of inorganic membrane and polymeric membrane, various inorganic-polymer

or polymer-polymer composite membranes have been studied, such as

Texas Tech University, Lan Ma, December 2014

polystyrenesulfonate/alumina [107], polyelectrolytes multi-layer [108], KA zeolite- incorporated cross-linked PVA multiplayer mixed matrix membranes [109].

One pervasive problem in membrane design is that materials exhibiting the highest selectivity for the desired permeant usually exhibit very low solubility, and therefore low transport rates are observed. The problem is further complicated by the use of diverse array of biomass-derived feedstocks to produce bio-alcohols. The development of new pervaporation membranes is based on three approaches: (1) the modification of polymers or membranes; (2) the incorporation of adsorber agents such as zeolites into polymer material; and (3) the development of complete new polymers [96]. However, there are only few combinations of miscible polymers and one pervaporation membrane is unlikely to perform optimally in all alcohol recovery processes. Therefore, design of highly selective, highly permeable polymeric membranes that have tunable chemical composition and separation characteristics could greatly benefit the growth of the young biofuels industry in the U.S. and abroad.

1.6 Objectives and organization of this study

As stated previously, multiple studies have been focused on the continuous bio-EtOH production and separation. However, there is no successful implementation of ICR for

EtOH production on a pilot or commercial scale with cellulosic biomass. Also, previous studies simply used inexpensive, commercially available gels without considering the materials design aspects of the problem. There are no systematically studies to optimize scaffold materials for cell immobilization such as the effects of particle size, pore size, and pore connectivity.

The objectives of research focus on the following items.

Texas Tech University, Lan Ma, December 2014

1. Design a continuous-flow, cell-immobilized fermentation process based on E. coli

LY01 for production of EtOH.

Different types of fermentation: batch, chemostat and cell-immobilized reactor should be compared via E .coli LY01 with glucose or mixed sugars as substrate. By studying the batch fermentation, the effect of cell specific growth rate on the chemostat fermentation can be quantified. Construct and design cell-immobilized reactors, and compare the performance of small-scale column reactor with other types of continuous-flow rectors.

Once the type of cell-immobilized reactor is determined, the EtOH productivity and glucose conversion can be optimized.

2. Evaluate scaffold materials that perform optimally in ICR fermentations with E. coli

LY01.

Cell-immobilized reactor is evaluated with different methods based on different scaffold materials. Material properties are characterized. Although numerous studies of

ICR fermentations have been published since the 1980s, only a few have examined porous immobilization materials. The influences of pore size and volume fraction on the performance of ICR have barely been investigated for any immobilization matrix. Porous material, as one of our studied scaffold materials, will be systematically investigated

3. Evaluate the maximum sugar concentration in the feed that can achieve ≥ 90 % sugar conversion

The ethanologenicity of E. coli LY01 is influenced by multiple factors such as LB concentration, flow rate, and cell-immobilization method. Therefore, the maximum sugar concentration (glucose or xylose) that can achieve 90 % conversion also depends on several factors. Once higher sugar concentration can be consumed with high conversion,

Texas Tech University, Lan Ma, December 2014

sugar mixtures and hydrolyzed cellulosic biomass can be used as feeding source instead

of single sugar in our ICR system.

4. Kinetics study of EtOH production by E. coli LY01.

The reaction model of the batch and continuous fermentations should be developed.

The model is used to predict the kinetic parameters and immobilized viable cell

concentration. The reactor actually contains three phases: liquid (medium), solid (cell and

scaffold material) and gas (CO2). Substrate and product are continuously flowing from

reactor bottom to the top. CO2 is releasing with the medium. A good reactor model should match with experimental results, and predict the results from different factors. Also, the model is able to show the sugar and EtOH concentrations along with the column positions.

5. Design materials and methods to screen materials for extractive fermentation. The concentration of cellulosic EtOH is influenced by inhibiting compounds and toxins; recovery of the alcohol by traditional multi-stage distillation is not feasible due to high energy consumption. We pursue an energy-friendly new separation process for recovery of dilute alcohols from ICR systems. The new process is either pervaporation or gel stripping, which all require highly selective materials. A series of copolymer gels can be developed using combinatorial chemistry principles to identify a material that swells to a high degree and is reasonably selective for EtOH. Both experiment and model were developed in this study. Optimized material is screened out by combining the calculation and easy measurement.

The rest of the dissertation is organized as follows. Chapter 2 presents the basic batch and chemostat fermentation results of E. coli LY01. Design and operation of cell immobilized continuous EtOH fermentation process are described in Chapter 3. Chapter

Texas Tech University, Lan Ma, December 2014

4 presents the kinetics study of bioreactor fermentation. In Chapter 5, the combinatorial

screening technique for EtOH separation with a copolymer network is described. A fundamental four component Flory-Rehner equation was developed with another student,

Mr. Rutvik Godbole.

Chapter 6 deals with unrelated work in the area of polymer-carbon nanocomposites, which was completed in the course of my dissertation research. Finally,

Chapter 7 summarizes the contributions of the research and suggests future work.

Texas Tech University, Lan Ma, December 2014

1.7 References

[1] Yamashita, Y.; Kurosumi, A.; Sasaki, C.; Nakamura, Y. Ethanol production from paper sludge by immobilized Zymomonas mobilis. Biochemical Engineering Journal 2008, 42 (3), 314-319.

[2] Wen, M.; Bond-Watts, B. B.; Chang, M. C. Production of advanced biofuels in engineered E. coli. Current opinion in chemical biology 2013, 17 (3), 472-479.

[3] Wyman, C. E. Ethanol from lignocellulosic biomass - technology, economics, and opportunities. Bioresource Technol 1994, 50 (1), 3-16.

[4] Ogbonna, J. C.; Mashima, H.; Tanaka, H. Scale up of fuel ethanol production from sugar beet juice using loofa sponge immobilized bioreactor. Bioresource Technology 2001, 76 (1), 1-8.

[5] Demirbas, A. Producing and using bioethanol as an automotive fuel. Energy Sources Part B 2007, 2 (4), 391-401.

[6] Tao, L.; Aden, A. The economics of current and future biofuels. In Vitro Cellular & Developmental Biology-Plant 2009, 45 (3), 199-217.

[7] Ingledew, W. M.; Jones, A. M.; Bhatty, R. S.; Rossnagel, B. G. Fuel alcohol production from hull-less barley. Cereal Chemistry 1995, 72 (2), 147-150.

[8] Ladisch, M. R.; Svarczkopf, J. A. Ethanol-production and the cost of fermentable sugars from biomass. Bioresource Technol 1991, 36 (1), 83-95.

[9] Badger, P. C. Ethanol from cellulose: a general review. Janick, J., Whipkey, A., Ed.; ASHS Press: Alexandria, VA, 2002; Vol. 5; p 17-21.

[10] Viikari, L.; Vehmaanperä, J.; Koivula, A. Lignocellulosic ethanol: From science to industry. Biomass and Bioenergy 2012, 46, 13-24.

[11] Khaw, T. S.; Katakura, Y.; Ninomiya, K.; Moukamnerd, C.; Kondo, A.; Ueda, M.; Shioya, S. Enhancement of ethanol production by promoting surface contact between starch granules and arming yeast in direct ethanol fermentation. Journal of Bioscience and Bioengineering 2007, 103 (1), 95-97.

[12] Calinescu, I.; Chipurici, P.; Alexandrescu, E.; Trifan, A. Saccharomyces cerevisiae yeast immobilized on marrow stem sunflower and polyacrylamide hydrogels. Central European Journal of Chemistry 2014, 12 (8), 851-857.

[13] Asher, A. Opportunities in biofuels creating competitive biofuels markets. Biofuels Australasia 2006 conference, Sydney, Australia, Nove 20-22, 2006.

Texas Tech University, Lan Ma, December 2014

[14] Kovalenko, G. A.; Beklemishev, A. B.; Perminova, L. V.; Mamaev, A. L.; Rudina, N. A.; Moseenkov, S. I.; Kuznetsov, V. L. Immobilization of recombinant E. coli thermostable lipase by entrapment inside silica xerogel and nanocarbon-in-silica composites. Journal of Molecular Catalysis B-Enzymatic 2013, 98, 78-86.

[15] Kowsari, M. Speciale water pollution sources. http://biofuel.webgarden.com/sections/blog/pictures-for-lignocellulose.

[16] Ogier, J. C.; Ballerini, D.; Leygue, J. P.; Rigal, L.; Pourquie, J. Ethanol production from lignocellulosic biomass. Oil & Gas Science and Technology- Revue d’IFP Energies Nouvelles 1999, 54 (1), 67-94.

[17] Lee, J. Biological conversion of lignocellulosic biomass to ethanol. Journal of Biotechnology 1997, 56 (1), 1-24.

[18] Semencenko, V. V.; Mojovic, L. V.; Petrovic, S. D.; Ocic, O. J. Recent trends in bioethanol production. Hemijska Industrija 2011, 65 (2), 103-114.

[19] Margeot, A.; Hahn-Hagerdal, B.; Edlund, M.; Slade, R.; Monot, F. New improvements for lignocellulosic ethanol. Current Opinion in Biotechnology 2009, 20 (3), 372-380.

[20] Sarkar, N.; Ghosh, S. K.; Bannerjee, S.; Aikat, K. Bioethanol production from agricultural wastes: an overview. Renewable Energy 2012, 37 (1), 19-27.

[21] Brethauer, S.; Wyman, C. E. Review: Continuous hydrolysis and fermentation for cellulosic ethanol production. Bioresour Technol 2010, 101 (13), 4862-4874.

[22] Penner, M. H.; Hashimoto, A. G.; Esteghlalian, A.; Fenske, J. J. Acid-catalyzed hydrolysis of lignocellulosic materials. In Agricultural Materials as Renewable Resources: Nonfood and Industrial Applications; ACS Symposium Series; 1996; Vol. 647; p 12-31.

[23] Rinaldi, R.; Schueth, F. Acid hydrolysis of cellulose as the entry point into biorefinery schemes. ChemSusChem 2009, 2 (12), 1096-1107.

[24] Ryu, S. Production of bioethanol and biobutanol using genetically engineered Escherichia coli. Ph.D. Dissertation, Department of Chemical Engineering, Texas Tech Universiy, 2010.

[25] Blumer-Schuette, S. E.; Kataeva, I.; Westpheling, J.; Adams, M. W. W.; Kelly, R. M. Extremely thermophilic microorganisms for biomass conversion: status and prospects. Current Opinion in Biotechnology 2008, 19 (3), 210-217.

[26] Singh, A.; Hayashi, K. Microbial cellulases-protein architecture, molecular- properties, and biosynthesis. Advances in Applied Microbiology 1995, 40(1), 1-44.

Texas Tech University, Lan Ma, December 2014

[27] van der Heijden, H.; Ptasinski, K. J. Energy analysis of thermochemical ethanol production via biomass gasification and catalytic synthesis. Energy 2012, 46 (1), 200-210.

[28] Jonsson, L. J.; Alriksson, B.; Nilvebrant, N.-O. Bioconversion of lignocellulose: inhibitors and detoxification. Biotechnology for Biofuels 2013, 6(1), 6-16.

[29] Wyman, C. E. What is (and is not) vital to advancing cellulosic ethanol. Trends in Biotechnology 2007, 25 (4), 153-157.

[30] Sun, Y.; Cheng, J. Y. Hydrolysis of lignocellulosic materials for ethanol production: a review. Bioresource technology 2002, 83 (1), 1-11.

[31] Wood, B. E.; Beall, D. S.; Ingram, L. O. Production of recombinant bacterial endoglucanase as a co-product with ethanol during fermentation using derivatives of Escherichia coli KO11. Biotechnology and Bioengineering 1997, 55 (3), 547- 555.

[32] Den Haan, R.; Rose, S. H.; Lynd, L. R.; van Zyl, W. H. Hydrolysis and fermentation of amorphous cellulose by recombinant Saccharomyces cerevisiae. Metabolic Engineering 2007, 9 (1), 87-94.

[33] Fujita, Y.; Ito, J.; Ueda, M.; Fukuda, H.; Kondo, A. Synergistic saccharification, and direct fermentation to ethanol, of amorphous cellulose by use of an engineered yeast strain codisplaying three types of cellulolytic enzyme. Applied and Environmental Microbiology 2004, 70 (2), 1207-1212.

[34] Limayem, A.; Ricke, S. C. Lignocellulosic biomass for bioethanol production: Current perspectives, potential issues and future prospects. Progress in Energy and Combustion Science 2012, 38 (4), 449-467.

[35] Banerjee, S.; Mudliar, S.; Sen, R.; Giri, B.; Satpute, D.; Chakrabarti, T.; Pandey, R. A. Commercializing lignocellulosic bioethanol: technology bottlenecks and possible remedies. Biofuels Bioproducts & Biorefining-Biofpr 2010, 4 (1), 77-93.

[36] Dien, B. S.; Cotta, M. A.; Jeffries, T. W. Bacteria engineered for fuel ethanol production: current status. Applied Microbiology and Biotechnology 2003, 63 (3), 258-266.

[37] Jarboe, L. R.; Grabar, T. B.; Yomano, L. P.; Shanmugan, K. T.; Ingram, L. O. Development of ethanologenic bacteria. Advances in Biochemical Engineering / Biotechnology 2007, 108, 237-261.

[38] Geddes, C. C.; Nieves, I. U.; Ingram, L. O. Advances in ethanol production. Current opinion in biotechnology 2011, 22 (3), 312-319.

Texas Tech University, Lan Ma, December 2014

[39] Ingram, L. O.; Conway, T.; Clark, D. P.; Sewell, G. W.; Preston, J. F. Genetic- engineering of ethanol-production in Escherichia coli. Applied and Environmental Microbiology 1987, 53 (10), 2420-2425.

[40] Ohta, K.; Beall, D. S.; Mejia, J. P.; Shanmugam, K. T.; Ingram, L. O. Metabolic engineering of Klebsiella-Oxytoca M5A1 for ethanol-production from xylose and glucose. Applied and Environmental Microbiology 1991, 57 (10), 2810-2815.

[41] Asghari, A.; Bothast, R. J.; Doran, J. B.; Ingram, L. O. Ethanol production from hemicellulose hydrolysates of agricultural residues using genetically engineered Escherichia coli strain KO11. Jounal of Industrial Microbiology 1996, 16, 42-47.

[42] Lawford, H. G.; Rousseau, J. D. Loss of ethanologenicity in Escherichia coli B recombinants PLOI297 and KO11 during growth in the absence of antibiotics Biotechnology Letters 1995, 17 (7), 751-756.

[43] Lawford, H. G.; Rousseau, J. D. Factors contributing to the loss of ethanologenicity of Escherichia coli B recombinants pLOI297 and KO11. Applied Biochemistry and Biotechnology 1996, 57-8, 293-305.

[44] Dumsday, G. J.; Zhou, B.; Yaqin, W.; Stanley, G. A.; Pamment, N. B. Comparative stability of ethanol production by Escherichia coli KO11 in batch and chemostat culture. Journal of Industrial Microbiology & Biotechnology 1999, 23 (1), 701-708.

[45] Yomano, L. P.; York, S. W.; Ingram, L. O. Isolation and characterization of ethanol-tolerant mutants of Escherichia coli KO11 for fuel ethanol production. Journal of Industrial Microbiology & Biotechnology 1998, 20 (2), 132-138.

[46] York, S. W.; Ingram, L. O. Ethanol production by recombinant Escherichia coli KO11 using crude yeast autolysate as a nutrient supplement. Biotechnology Letters 1996, 18 (6), 683-688.

[47] Martinez, A.; York, S. W.; Yomano, L. P.; Pineda, V. L.; Davis, F. C.; Shelton, J. C.; Ingram, L. O. Biosynthetic burden and plasmid burden limit expression of chromosomally integrated heterologous genes (pdc, adhB) in Escherichia coli. Biotechnology Progress 1999, 15 (5), 891-897.

[48] Zhou, B.; Martin, G. J.; Pamment, N. B. Increased phenotypic stability and ethanol tolerance of recombinant Escherichia coli KO11 when immobilized in continuous fluidized bed culture. Biotechnology and Bioengineering 2008, 100 (4), 627-633.

[49] Rogers, P. L.; Lee, K. J.; Skotnicki, M. L.; Tribe, D. E. Ethanol production by Zymomonas mobilis. Advances Biochem Eng 1982, 23, 37-84.

Texas Tech University, Lan Ma, December 2014

[50] Zhang, M.; Eddy, C.; Deanda, K.; Finkestein, M.; Picataggio, S. Metabolic engineering of a pentose metabolism pathway in ethanologenic Zymomonas mobilis. Science 1995, 267 (5195), 240-243.

[51] Doran, J. B.; Aldrich, H. C.; Ingram, L. O. Saccharification and fermentation of sugar cane bagasse by Klebsiella oxytoca P2 containing chromosomally integrated genes encoding the Zymomonas mobilis ethanol pathway. Biotechnology and Bioengineering 1994, 44 (2), 240-247.

[52] Moniruzzaman, M.; Dien, B. S.; Ferrer, B.; Hespell, R. B.; Dale, B. E.; Ingram, L. O.; Bothast, R. J. Ethanol production from AFEX pretreated corn fiber by recombinant bacteria. Biotechnology Letters 1996, 18 (8), 985-990.

[53] Brooks, T. A.; Ingram, L. O. Conversion of mixed waste office paper to ethanol by genetically-engineered Kiebsiella oxytoca strain P2. Biotechnology Progress 1995, 11 (6), 619-625.

[54] Ho, N. W. Y.; Chen, Z.; Brainard, A. P.; Sedlak, M. Genetically engineered Saccharomyces yeasts for conversion of cellulosic biomass to environmentally friendly transportation fuel ethanol. In Green Chemical Syntheses and Processes; ACS Symposium Series; 2000; Vol. 767; p 143-159.

[55] Zaldivar, J.; Nielsen, J.; Olsson, L. Fuel ethanol production from lignocellulose: a challenge for metabolic engineering and process integration. Applied Microbiology and Biotechnology 2001, 56 (1-2), 17-34.

[56] Bothast, R. J.; Nichols, N. N.; Dien, B. S. Fermentations with new recombinant organisms. Biotechnology Progress 1999, 15 (5), 867-875.

[57] Lau, M. W.; Gunawan, C.; Balan, V.; Dale, B. E. Comparing the fermentation performance of Escherichia coli KO11, Saccharomyces cerevisiae 424A(LNH-ST) and Zymomonas mobilis AX101 for cellulosic ethanol production. Biotechnology for Biofuels 2010, 3, 11-21.

[58] Dombek, K. M.; Ingram, L. O. Effects of ethanol on the Escherichia coli plasma membrane. Journal of Bacteriology 1984, 157 (1), 233-239.

[59] Guidoboni, G. E. Continuous fermentation systems for alcohol production. Enzyme and Microbial Technology 1984, 6 (5), 194-200.

[60] Verbelen, P. J.; De Schutter, D. P.; Delvaux, F.; Verstrepen, K. J.; Delvaux, F. R. Immobilized yeast cell systems for continuous fermentation applications. Biotechnology Letters 2006, 28 (19), 1515-1525.

[61] Nigam, J. N. Continuous ethanol production from pineapple cannery waste using immobilized yeast cells. Journal of Biotechnology 2000, 80 (2), 189-193.

Texas Tech University, Lan Ma, December 2014

[62] Inloes, D. S.; Taylor, D. P.; Cohen, S. N.; Michaels, A. S.; Robertson, C. R. Ethanol-production by Saccharomyces cerevisiae immobilized in hollow-fiber membrane bioreactors. Applied and Environmental Microbiology 1983, 46 (1), 264-278.

[63] Jamuna, R.; Ramakrishna, S. V. High-concentration ethanol-production using immobilized yeast-cells. Biomass & Bioenergy 1992, 3 (2), 117-119.

[64] Najafpour, G.; Younesi, H.; Syahidah Ku Ismail, K. Ethanol fermentation in an immobilized cell reactor using Saccharomyces cerevisiae. Bioresour Technol 2004, 92 (3), 251-60.

[65] Najafpour, G. D. Immobilization of microbial cells for the production of organic acids. Journal of Sciences, Islamic Republic of Iran 1990, 1(3), 172-176.

[66] Palmqvist, E.; Hahn-Hagerdal, B. Fermentation of lignocellulosic hydrolysates. I: inhibition and detoxification. Bioresource Technology 2000, 74 (1), 17-24.

[67] Taherzadeh, M. J.; Millati, R.; Niklasson, C. Continuous cultivation of dilute-acid hydrolysates to ethanol by immobilized Saccharomyces cerevisiae. Applied Biochemistry and Biotechnology 2001, 95 (1), 45-57.

[68] Ramakrishna, S. V.; Prakasham, R. S. Microbial fermentations with immobilized cells. Current Science 1999, 77 (1), 87-100.

[69] Karel, S. F.; Libicki, S. B.; Robertson, C. R. The immobilization of whole cells- engineering principles. Chemical Engineering Science 1985, 40 (8), 1321-1354.

[70] Hattori, T.; Furusaka, C. Chemical activites of E. coli adsorbed on a resin. The Journal of Biochemistry 1960, 48 (6), 831-836.

[71] Oliveira, R. Understanding adhesion: A means for preventing fouling. Experimental Thermal and Fluid Science 1997, 14 (4), 316-322.

[72] Levine, D. W., Wang, D.I.C.; Thilly, W.G. Optimization of growth surface parameters in microcarrier cell culture. Biotechnology and bioengineering 1979, 21(5), 821-845.

[73] Microcarrier Cell Culture, Principles and Methods, 1st ed.; Published by Pharmacia Fine Chemicals, Sweden; 1981.

[74] Navarro, J. M.; Durand, G. Modification of yeast metabolism by immobilization onto porous glass. Europe Journal of Applied Microbiology and Biotechnology 1977, 4, 243-254.

Texas Tech University, Lan Ma, December 2014

[75] Scott, J. A.; O'Reilly, A. M. Use of a flexible sponge matrix to immobilize yeast for beer fermentation. Journal of the American Society of Brewing Chemists 1995, 53(2), 67-71.

[76] Shen, H. Y.; Moonjai, N.; Verstrepen, K. J.; Delvaux, F. R. Impact of attachment immobilization on yeast physiology and fermentation performance. Journal of the American Society of Brewing Chemists 2003, 61 (2), 79-87.

[77] Barron, N.; Marchant, R.; McHale, L.; McHale, A. P. Ethanol production from cellulose at 45 degrees C using a hatch-fed system containing alginate- immobilized Kluyveromyces marxianus IMB3. World Journal of Microbiology & Biotechnology 1996, 12 (1), 103-104.

[78] Williams, P. D.; Mavituna, F. Immobilized plant cells. In Plant Biotechnology: Comprehensive Biotechnology; Fowler, M.W., Warren,G.S., Moo-Young, M., Ed.; Pergamon Press: Oxford; 1992; p 63-78.

[79] Riley, M. R.; Muzzio, F. J.; Buettner, H. M.; Reyes, S. C. A simple correlation for predicting effective diffusivities in immobilized cell systems. Biotechnology and bioengineering 1996, 49 (2), 223-227.

[80] Vorlop, K. D.; Klein, J. Formation of spherical chitosan biocatalysts by iontropic gelation. Biotechnology Letters 1981, 3(1), 9-14.

[81] Yamagiwa, K.; Shimizu, Y.; Kozawa, T.; Onodera, M.; Ohkawa, A. Ethanol production by encapsulated and immobilized yeast. Biotechnology Techniques 1994, 8(4), 271-274.

[82] Niu, X.; Wang, Z.; Li, Y.; Zhao, Z.; Liu, J.; Jiang, L.; Xu, H.; Li, Z. "Fish-in-Net", a novel method for cell immobilization of Zymomonas mobilis. PLOS ONE 2013, 8 (11), 1-11.

[83] Roble, N. D.; Ogbonna J. C.; Tanaka, H. A novel circulating loop bioreactor with cells immobilized in loofa (Luffa cylindrica) sponge for the bioconversion of raw cassava starch to ethanol. Applied Microbiology and Biotechnology 2003, 60 (6), 671-678.

[84] Verstrepen, K. J.; Derdelinckx, G.; Verachtert, H.; Delvaux, F. R. Yeast flocculation: what brewers should know. Applied Microbiology and Biotechnology 2003, 61 (3), 197-205.

Texas Tech University, Lan Ma, December 2014

[85] Arcuri, E. J. Continuous ethanol production and cell growth in an immobilized cell bioreactor employing Zymomonas mobilis. Biotechnology and Bioengineering 1982, 24 (3), 595-604.

[86] Nagashima, M.; Azuma, M.; Noguchi, S.; Inuzuka, K.; Samejima, H. Large-scale preparation of calcium alginate immobilized yeast cells and its application to industrial ethanol production. Methods in Enzymology 1987, 136, 394-405.

[87] Amin, G.; Doelle, H. W. Production of high ethanol concentrations from glucose using Zymomonas mobilis entrapped in a vertical rotating immobilized cell reactor. Enzyme and Microbial Technology 1990, 12 (6), 443-446.

[88] Gil, G. H.; Jones, W. J.; Tornabene, T. G. Continuous ethanol production in a 2- stage, immobilized suspended cell bioreactor. Enzyme and Microbial Technology 1991, 13 (5), 390-399.

[89] Ghorbani, F.; Younesi, H.; Sari, A. E.; Najafpour, G. Cane molasses fermentation for continuous ethanol production in an immobilized cells reactor by Saccharomyces cerevisiae. Renewable Energy 2011, 36 (2), 503-509.

[90] Mongkolkajit, J.; Pullsirisombat, J.; Limtong, S.; Phisalaphong, M. Alumina- doped alginate gel as a cell carrier for ethanol production in a packed-bed bioreactor. Biotechnology and Bioprocess Engineering 2011, 16 (3), 505-512.

[91] Yan, S. B.; Chen, X. S.; Wu, J. Y.; Wang, P. C. Ethanol production from concentrated food waste hydrolysates with yeast cells immobilized on corn stalk. Applied Microbiology and Biotechnology 2012, 94 (3), 829-838.

[92] Chen, Y.; Liu, Q. G.; Zhou, T.; Li, B. B.; Yao, S. W.; Wu, J. L.; Ying, H. J. Ethanol production by repeated batch and continuous fermentations by Saccharomyces cerevisiae immobilized in a fibrous bed bioreactor. Journal of Microbiology and Biotechnology 2013, 23 (4), 511-517.

[93] Gokgoz, M.; Yigitoglu, M. High productivity bioethanol fermentation by immobilized Saccharomyces bayanus onto carboxymethylcellulose-g-poly(N- vinyl-2-pyrrolidone) beads. Artificial Cells Nanomedicine and Biotechnology 2013, 41 (2), 137-143.

[94] Mariano, A. P.; Filho, R. M. Improvements in biobutanol fermentation and their impacts on distillation energy consumption and wastewater generation. Bioenergy Research 2012, 5, 504-514.

[95] Gaykawad, S. S.; Zha, Y.; Punt, P. J.; Van Groenestijn, J. W.; van der Wielen, L. A. M.; Straathof, A. J. J. Pervaporation of ethanol from lignocellulosic fermentation broth. Bioresource Technology 2013, 129, 469-476.

Texas Tech University, Lan Ma, December 2014

[96] Peng, P.; Shi, B. L.; Lan, Y. Q. A review of membrane materials for ethanol recovery by pervaporation. Separation Science and Technology 2011, 46 (2), 234- 246.

[97] Vane, L. M. A review of pervaporation for product recovery from biomass fermentation processes. Journal of Chemical Technology and Biotechnology 2005, 80 (6), 603-629.

[98] Huang, H.-J.; Ramaswamy, S.; Tschirner, U. W.; Ramarao, B. V. A review of separation technologies in current and future biorefineries. Separation and Purification Technology 2008, 62 (1), 1-21.

[99] Beaumelle, D.; Marin, M.; Gibert, H. Pervaporation with organophilic membranes: state of the art. Food and Bioproducts Processing: Transactions of the Institution of Chemical Engineers Part C 1993, 71 (2), 77-89.

[100] O'Brien, D. J.; Roth, L. H.; McAloon, A. J. Ethanol production by continuous fermentation-pervaporation: a preliminary economic analysis. Journal of Membrane Science 2000, 166 (1), 105-111.

[101] Miyata, T.; Higuchi, J.; Okuno, H.; Uragami, T. Preparation of polydimethylsiloxane/polystyrene interpenetrating polymer network membranes and permeation of aqueous ethanol solutions through the membranes by pervaporation. Journal of Applied Polymer Science 1996, 61 (8), 1315-1324.

[102] Guo, Z.; Hu, C. Y. Pervaporation of organic liquid/water mixtures through a novel silicone copolymer membrane. Chinese Science Bulletin 1998, 43 (6), 487- 490.

[103] Kashiwagi, T.; Okabe, K.; Okita, K. Separation of ethanol from ethanol water mixtures by plasma polymerized membranes from silicone compounds. Journal of Membrane Science 1988, 36, 353-362.

[104] Hickey, P. J.; Juricic, F. P.; Slater, C. S. The effect of process parameters on the pervaporation of alcohols through organophilic membranes. Separation Science and Technology 1992, 27 (7), 843-861.

[105] Garcia, M.; Teresa Sanz, M.; Beltran, S. Separation by pervaporation of ethanol from aqueous solutions and effect of other components present in fermentation broths. Journal of Chemical Technology and Biotechnology 2009, 84 (12), 1873- 1882.

[106] Liu, F. F.; Liu, L.; Feng, X. S. Separation of acetone-butanol-ethanol (ABE) from dilute aqueous solutions by pervaporation. Separation and Purification Technology 2005, 42 (3), 273-282.

Texas Tech University, Lan Ma, December 2014

[107] Chen, W. J.; Aranda, P.; Martin, C. R. Pervaporation separation of ethanol/water mixtures by polystyrenesulfonate/alumina composite membranes. Journal of Membrane Science 1995, 107 (3), 199-207.

[108] Krasemann, L.; Tieke, B. Highly efficient composite membranes for ethanol- water pervaporation. Chemical Engineering & Technology 2000, 23 (3), 211-213.

[109] Huang, Z.; Guan, H.-M.; Tan, W. I.; Qiao, X.-Y.; Kulprathipanja, S. Pervaporation study of aqueous ethanol solution through zeolite-incorporated multilayer poly (vinyl alcohol) membranes: effect of zeolites. Journal of Membrane Science 2006, 276 (1-2), 260-271.

Texas Tech University, Lan Ma, December 2014

Chapter 2

Batch and Chemostat Bio-ethanol Fermentation Process with Strain Escherichia coli LY01

2.1 Introduction

Escherichia coli is potentially an important microorganism because of its well-known

metabolic characteristics, rapid growth, and potential applicability to mixed sugar

feedstocks that are derived from cellulosic biomass. The design and construction of the

ethanologenic strain KO11 is a big breakthrough in metabolic engineering of E. coli for

improving ethanol fermentation. Ingram and coworkers developed KO11 from E. coli W

by integrating the PET operon under the control of pfl promoter along with an antibiotic

resistance marker (chloramphenicol, Cm) [1].

E. coli KO11 has been tested with batch fermentation using glucose and diverse

biomass residues[2, 3]. Leite and coworkers carried out a fermentation with sweet whey

(58 g/L sugars) and got an ethanol yield of 100 %, but the yield went down to 38 % when

yeast extract and a trace metal mixture were removed [4]. Fermentation of sugar cane

bagasse hydrolysate (68 g/L sugars) with tryptone and yeast extract was carried out by

Takahashi and coworkers [5], which reached an ethanol yield of 92 % and volumetric

productivity of 0.66 g L-1 h-1 ethanol. Like above examples, other batch fermentation

results of E. coli KO11 are shown in Table 2.1. The disadvantage is the high cost of LB

(10g tryptone, 5 g yeast extract, 10 g NaCl for 1 L of LB) medium for industrial fuel production. Fermentation of E. coli KO11 without rich medium can only get the ethanol

yield lower than 70 % of theoretical maximum and the volumetric productivity decreases to half [6, 7].

Texas Tech University, Lan Ma, December 2014

E. coli KO11 was reported to show instability with regard to ethanol production when it was transferred several times in batch culture or cultivated in continuous cultures. In

the absence of antibiotic, KO11 exhibited dramatic loss of ethanologenicity, resulting a

reduction in ethanol yield. When fermentation was running with glucose, galactose,

mannose, xylose and LB medium without antibiotic, the ethanol yield decreased to

around 50 % of the theoretical maximum in only 12 generations except with xylose as

substrate [8]. When fermentation was running with antibiotic, the ethanol yield

dramatically decreases a lot only for mannose. Therefore, the stability of E. coli KO11 to

maintain a high efficiency of sugar to ethanol conversion is dependent on the substrate

species and antibiotic based on Lawford’s research. Using glucose-LB or xylose-LB as the only substrate in continuous fermentation at dilution rate of 0.14 and 0.07/h, strain

KO11 exhibited a rapid loss of ethanologenicity with and without antibiotic. Increasing the concentration of chloramphenicol from 40 to 300 mg/L can only delay the rate of loss of ethanologenicity, but it did not prevent it [9].

Dumsday et al. compared stability of ethanol production by E. coli KO11 in both batch and chemostat culture. The organism was stable on fermentation with glucose, mannose, xylose and galactose (20 g/L) for at least three serial transfers, even without antibiotic.

Chemostat culture on glucose was remarkably stable at dilution rate of 0.06/h. However, the organism lost their hyperethanologenicity on chemostat culture with mannose, xylose and a xylose/glucose mixture after 10 days [10]. The loss of hyperethanologenicity was accompanied with the production of high concentrations of acetic acid and biomass yield.

The reason of ethanol reduction or loss of ethanologenicity of E. coli KO11 in continuous

fermentation has been suggested due to the genetic instability [8] and loss of

Texas Tech University, Lan Ma, December 2014

heterologous ethanologenic genes [9, 11]. In 2008, Zhou et al. examined the stability of

KO11 using continuous fluidized bed culture. A vertical tubular fermentor was operated in their research [12]. The fermentation was run on two types with the normal chemostat culture and continuous fluidized bed culture. The result showed the phenotypic stability and ethanol tolerance during the cell immobilization fermentation was improved. Though immobilizing the cell cannot prevent the decline, it delayed the onset and slowed the rate.

Strain LY01 is was obtained after cultivating KO11 in a series LB medium with 50 g/L glucose and exogenous 30-50 g/L ethanol. The selected strain E. coli LY01 shows the resistance to growth inhibition by 45 g/L ethanol and is able to produce 60 g/L ethanol from 140 g/L xylose in 72 h, which is the highest reported ethanol concentration with ethanologenic E. coli LY01 [13]. When strains were exposed to high ethanol concentration (100 g/L), E. coli LY01 exhibited over 50 % survival (CFU) during 0.5 min exposure; however, KO11 only had less than 10 % survival. Gonzalez et al. studied

the difference of KO11 and LY01 through gene array-based identification of changes

[14]. Ethanol tolerance in microorganisms is related with cell membrane lipid composition including the change of unsaturated fatty acids amount, phospholipid composition, and the ratio of phospholipids to protein [15].

Texas Tech University, Lan Ma, December 2014

Table 2-1. Batch and chemostat fermentation results of E. coli KO11 and LY01.

Fermentation Strain Sugar (g/L) Ethanol yield Fermentation Dilution Ref Type (% theor.) Time Rate (h-1)

Batch KO11 glucose(100) 100 % 75h [2] Batch KO11 glucose(140) 73.8 % 96 h [13] Batch KO11 xylose(80) 100 % 75h [2] Batch KO11 xylose(90) 89.3 % 48h [13] Batch KO11 Xylose (100) 85.1 % 168h [16] Batch KO11 xylose(140) 83.3 % 120h [13] Batch KO11 corn stover (90) 90 % 60 h [17] Batch KO11 corn hulls and 86 % 72h [18] fibers (90) Batch KO11 Pinus sp. 85 %-91 % 100h [19] Hydrolysate(72) Batch LY01 glucose(140) 85.4 % 96h [13] Batch LY01 xylose(90) 91.9 % 48h [13] Batch LY01 xylose(140) 88.5 % 96h [13] chemostat KO11 glucose (20) 82 % 20 days 0.06 [10] chemostat KO11 glucose(25) 24 % 48h 0.07 [9] chemostat KO11 mannose (20) 59 % 27days 0.06 [10] chemostat KO11 xylose (20) 55 % 25 days 0.06 [10]

During the thermochemical treatment of biomass, some toxic molecules are generated

such as furan derivates, organic acids, aldehydes, alcohols, and aromatic compounds that

inhibit cell growth and ethanol production [20, 21]. Zaldivar et al. systematically studied

the effects of organic acids, selected aldehydes and alcohol components of hemicellulose

hydrolysate on the growth and fermentation of E. coli LY01 [22]. Organic acids, aldehydes and alcohols inhibit cell growth and ethanol production through damaging the cellular membrane [22, 23]. Hemicellulose hydrolysates involve a combination of toxic compounds rather than a single toxic compound [24]. E. coli LY01 has more resistance to

furfural and 5-hydroxymethylfurfural than KO11 [25]. Furthermore, ethanologenic E.

coli strains have a greater resistance to these toxic effects compared to other strains, such

as S. cereviciae and Z. mobilis [21, 22, 25].

There are only a few reports about continuous fermentation of E. coli KO11 and no

published reports about continuous fermentation of E. coli LY01. An economically-

Texas Tech University, Lan Ma, December 2014

attractive cellulosic technology certainly requires an ethanol yield, concentration and rate higher than 90 %, 40g/L, and 1.0 g L-1 h-1, respectively [16, 26]. In this chapter, the effects of glucose and nutrient on the batch fermentation of E. coli LY01 were studied.

Results of batch fermentations between LY01 and KO11 were compared. We also carried out an extensive investigation of chemostat fermentations of strain LY01 at both high and low dilution rates.

2.2 Materials and methods

2.2.1 Microorganism and growth medium

Escherichia coli KO11 and LY01 were obtained from Professor L.O. Ingram

(University of Florida). Both of them are engineered ethanol producing strains with resistance to chloramphenicol (Cm). Lysogeny Broth (LB) was used as growth medium with a make-up of 10 g/L tryptone, 5 g/L yeast extract, and 10 g/L Nacl. All the media contained 25 g/L LB and 40 μg/ml Cm (MP Biomedicals, LLC) and were prepared using ultrapure water. LB and sugar solution were sterilized separately by autoclaving at 121°C for 15 min. Chloramphenicol was added to the autoclaved medium after cooling. LB agar plates for initial cell growth were prepared with 15 g/L agar and the standard concentrations of LB and antibiotic. For long term storage, a medium of similar composition without agar was prepared for stock culture and cultivated at 35 °C with 200 rpm in a shaker. After 9 hours, the culture was diluted in sterilized 20 % glycerol (v/v), mixed well, and then stored freeze-dried in ampoules at -80 °C.

2.2.2 Inoculum preparation

To prepare the inoculum, cells were firstly plated onto LB agar plates and incubated at 35 °C overnight. After overnight culture, a single colony from the agar plate was

Texas Tech University, Lan Ma, December 2014

transferred into a flask containing 200 mL of liquid LB medium with 10 g/L glucose.

Two of these cultures were prepared. The liquid cultures were incubated for 16 h in a

shaker at 35 °C and 200 rpm. The pre-cultivated cells were collected by centrifugation at

5000 g, 4°C for 10 min and rinsed with sterilized ultrapure water to get rid of byproducts,

and then resuspended in 100 ml of fresh LB medium containing no additional glucose.

Cell plates were stored at 4 °C and were used for up to 45 days.

2.2.3 Fermentation conditions

After nitrogen sparging, anaerobic fermentation was initiated by adding the inoculum to the fermentation medium in the tank reactor. The volume of innoculum added was enough to provide approximately 1.0 OD at 600 nm (330 mg dry cell mass/L). The

temperature was maintained at 35 °C, the pH at 6.0, and the stir rate at 200 rpm. Nitrogen

was added to the tank at a rate of 0.1 mL/min to ensure anaerobic operation. Batch

fermentations were conducted at 35 °C and the stir rate of 200 rpm in a 2.5 L bioreactor

(New Brunswick Scientific BioFlo 310) containing 1.4 L of sugar medium. pH was

controlled at 6.0 by automatically adding 4.0 M NaOH. Nitrogen was added to the tank at

a rate of 0.1 mL/min to ensure anaerobic operation. For continuous fermentations, the

reactor was initially operated in batch mode. After 7 h, the feed and outlet pump were

started at a constant dilution rate. Samples were taken periodically and the OD and pH

were recorded. After filtration, the residual glucose and ethanol concentrations were

measured and recorded.

2.2.3 Analytical methods

Texas Tech University, Lan Ma, December 2014

Ethanol and residual glucose were measured by a high-performance liquid

chromatography with a Rezex ROA (Phenomenex )-organic acid column, equipped with a refractive index (RI) detector (Smartline RI detector 2300, Knauer). The column was operated at 85 °C with 0.005 N H2SO4 (0.6 ml/min) as the mobile phase.

Cell density was measured by the optical density of an appropriately diluted sample to

insure the optical density was in the range of 0.1 to 0.4 at 600 nm using an Evolution 260

BIO UV-visible spectrophotometer and converted to dry cell mass.

2.3 Results

2.3.1 Batch fermentations

It was reported that mutant LY01 is more resistant to growth inhibition by ethanol than

the parent KO11. Batch anaerobic fermentations with the same initial glucose

concentration of 55 g/L and optical density of 1.0 were operated using a 2.5 L stirred tank

reactor. Fig. 2.1 showed that glucose was completely consumed by both KO11 and LY01.

Glucose was consumed in about 23 hours by LY01, compared with 35 hours by KO11.

Also, ethanol yield from LY01 is about 96 % of theoretical maximum, which is higher

than the ethanol yield from KO11 (85 % of the theoretical maximum).

Texas Tech University, Lan Ma, December 2014

Figure 2.1. Ethanol fermentation with E. coli KO11 (black dots) and LY01 (red dots). Ethanol yield was reported to decrease when the fermentation of strain KO11 was not

conducted in a rich medium. Therefore, a series of flask fermentations were run in order

to examine the effects of substrate and LB concentrations on the cell growth of strain

LY01 (Fig. 2.2). Cell cultures were started with the same initial cell density of OD=0.1.

When LB concentrations were kept same at 50 g/L, cell growth rate was slightly decreased when initial glucose concentration changed from 10 g/L to 50 g/L. When glucose concentrations were kept same at 50 g/L, cells grew faster in the medium with high LB concentration (50 g/L) compared with medium with 10 g/L LB. LB

concentration not only influences cell growth rate, but also influences cell density. Cell

density of the steady state with low LB medium only reached half of the cell density with

high LB medium. Cell culture with both high glucose and LB medium has the lowest

Texas Tech University, Lan Ma, December 2014

growth rate due to substrate inhibition. In order to get high ethanol yield and glucose conversion, rich medium is required for E. coli LY01.

Figure 2.2. Cell culture with different glucose and LB concentrations. The average of duplicate experimental data is plotted from two trials. The error bars are standard deviation.

The effects of initial glucose concentrations on cell growth, substrate utilization, ethanol production, and duration of fermentation were studied. Fig. 2.3 showed a series of anaerobic batch fermentations with E. coli LY01 in a pH controlled 2.5 L bioreactor.

Fermentations were running with different initial glucose concentrations ranging from 10 to 73 g/L with same LB concentration 25 g/L and 40 μg ml-1 Cm. Glucose was completely consumed in all the fermentations. The curves of residual glucose concentration versus time in Fig. 2.3a are paralle, which means the glucose reaction rates in these four fermentations are almost same. The calculated glucose consumption rate is

Texas Tech University, Lan Ma, December 2014

1.83±0.27 g·L-1·h-1. It can be hypothesized that there is no substrate inhibition within the observed glucose concentration range due to the nearly identical rates of glucose

consumption and ethanol production. Fig. 2.3b presents the optical cell density as a

function of time during batch fermentation. Exponential growth phase started

immediately at the beginning of incubation in all experiments without a significant period

of adaptation. Cell specific growth rate slightly decreased and the duration of the

exponential growth phase increased with the initial glucose concentration increased.

Specific growth rates were calculated. The average ethanol yield was about 0.49±0.02 g/g

of sugar consumed (Fig. 2.3c), which is close to 100 % of the theoretical maximum.

Specific growth rate was also found decreasing slightly for E. coli KO11 with the initial

glucose concentration increased [27].

(a )

Texas Tech University, Lan Ma, December 2014

(b)

Figure 2.3 Anaerobic batch fermentations with different initial glucose concentrations. (a) Residual glucose; (b) Cell growth; (c) Ethanol.

Texas Tech University, Lan Ma, December 2014

Specific cell growth rate is a way of measuring how fast the cells are dividing in a culture. It is a very important parameter in the fermentation, especially for continuous fermentation. Specific cell growth rate is influenced by culture conditions such as substrate, nutrient, antibiotic and aerobic or anaerobic environment. It is normally calculated from exponential phase. In order to calculate cell growth rate more accurately, we converted cell optical density to cell dry mass (OD=1 is about 330 mg dry cell mass

/L) [28] and fitted data from both exponential and stationary phase (Fig. 2.4).

(a)

Texas Tech University, Lan Ma, December 2014

(b)

(c)

Texas Tech University, Lan Ma, December 2014

(d)

Figure 2.4. Cell mass concentration from the anaerobic batch fermentation using 2.5 L stirred tank bioreactor with different initial glucose concentrations : 10 g/L (a), 36 g/L (b), 54 g/L (c) and 73 g/L (c). Initial LB concentration was 25 g/L in all cases. The experimental data are shown as dots, and the solid lines are best fits to eqn. 2.2 below.

The Logistic equation, which characterizes growth in terms of carrying capacity, was used to fit the cell growth curve.

X µg =k(1 − ) (2.1) X ∞

The analytical solution obtained by integration of equation 2.1 yields eqn. 2.2.

Xekt X = 0 (2.2) X 1−−0 (1ekt ) X ∞

In eqn. 2.2, X 0 and X ∞ are the initial and maximum cell masses, respectively.

Equation 2.2 is equation 6.56 in the book Bioprocess Engineering, Second Edition 49

Texas Tech University, Lan Ma, December 2014

(Michael Shuler & Fikret Kargi). The parameter k is calculated by fitting experimental

data using eqn. 2.2. The specific growth rate µg is calculated from equation 2.1 by

1 setting XX= (midpoint of the exponential phase). Calculated specific growth rates 2 ∞

are shown in the Table 2.2 below. In these batch fermentations, the cell specific growth

rate was found to decrease as initial glucose concentration increased. In all cases, the μg

of LY01 (under the conditions employed) was lower than the maximum specific growth

rate for E. coli, which can be as high as 0.8 to 1.4 h-1, which may due to the product

inhibition.

Table 2-2. Specific growth rate of E. coli LY01 in the anaerobic batch fermentations.

Glucose Concentration (g/L) k Specific growth rate (h-1) (h-1) 10 0.768 0.041 0.384 0.021 36 0.515 0.013 0.257 0.007 54 0.420 0.021 0.210 0.010 73 0.461 0.016 0.231 0.008

2.3.2 Chemostat fermentaions

The good performance in batch fermentation of E. coli KO11 showed promise for its

commercial use. However, continuous fermentation would be preferable to batch

operations considering decreasing operating and capital cost. The stability of ethanol

production by E. coli KO11 was studied with low concentrations of sugar fermentation in

continuous culture [9, 10, 12]. The disagreements from previous studies were about

whether strain KO11 will lose its ethanologenicity during long-term continuous ethanol

fermentation with different sugars. The highest glucose concentration used in the chemostat fermentation, which achieved above 90 % conversion was 20 g/L [10]. An

Texas Tech University, Lan Ma, December 2014

important parameter in continuous fermentation is the strain’s ethanol tolerance. Strain

LY01 is more resistant to ethanol compared with KO11. However, the continuous growth of E. coli LY01 on sugar has not been previously investigated. In our study, we investigated the stability of ethanol production by LY01 at high glucose concentration with different dilution rates.

One of the most important features of chemostats is that micro-organisms can be grown in a physiological steady state and all culture parameters remain constant (culture volume, reaction rate, nutrient and product concentrations, pH, cell density, etc.) in

steady state. At steady state, cell specific growth rate µg should be equal to dilution rate

(D), which is defined as the rate of flow of medium over the volume of culture in the

bioreactor (eqn. 2.3).

Medium flow rate F D= = Culture volume V (2.3)

The specific growth rate of E. coli LY01 in the anaerobic batch fermentation with 73 g/L glucose is 0.231 h-1 (Table 2.2). Therefore, the highest dilution rate we can use in our

continuous fermentation should be 0.231 h-1 without washing cells at the same glucose

concentration. However, this dilution rate is so high to consume all the glucose at steady

state. The chemostat was operated as batch fermentation for the first 7 h to reach the

exponential cell growth phase. The glucose reaction rate of batch fermentation at this

phase is 1.83 g L-1 h-1. In order to consume all the glucose at steady state, the calculated

dilution rate for chemostat should be equal to 0.026 h-1 according to the batch

fermentation result (eqn. 2.4).

DC()s0 −= C ss r (2.4)

Texas Tech University, Lan Ma, December 2014

In equation 2.4, Cs0 is inlet glucose concentration, Cs is outlet glucose concentration,

and rs is glucose reaction rate, which was got from batch fermentation.

(a)

Texas Tech University, Lan Ma, December 2014

(b)

Figure 2.5. Anaerobic chemostat fermentation using 2.5 L stirred tank bioreactor with initial glucose concentrations of 70 g/L. Dilution rate was 0.026 h-1. Initial LB and chloramphenicol concentration were 25 g/L and 40 µg/ml. The first 7 h was operated as anaerobic batch fermentation. (a) Ethanol and residual glucose; (b) cell optical density. Fig. 2.5. shows that the steady state was reached after 2 days and all the glucose was completely consumed in about 4 days. The experiment was run continuously for 11 days without washing out of cells based on OD measurements. The average ethanol yield was

0.47 g g-1 of glucose consumed (92 % of the theoretical maximum). Lactic and acetic

acid concentrations were about 1.45 g/L and 1.11 g/L at steady state during the chemostat

fermentation. The productivity of chemostat fermentation is 0.873±0.03 g L-1 h-1, which is comparable with batch of 0.916±0.047 g L-1 h-1.

Dilution rate was increased to 0.04 h-1 with same feed concentration of 70 g/L glucose

(Fig. 2.6). Reaction reached steady state after 2 days. Glucose conversion decreased from

100 % (D=0.026 h-1) to about 78 % with productivity of 1.069 ±0.07 g L-1 h-1. It took less

Texas Tech University, Lan Ma, December 2014

time to reach the steady state and obtained higher productivity; however, the glucose

conversion was decreased.

(a)

(b)

Figure 2.6. Anaerobic chemostat fermentation using 2.5 L stirred tank bioreactor with initial glucose concentrations of 70 g/L. Dilution rate was 0.04 h-1. Initial LB and 54

Texas Tech University, Lan Ma, December 2014

chloramphenicol concentrations were 25 g/L and 40 µg/ml. The first 7 h was operated as anaerobic batch fermentation. (a) Ethanol and residual glucose; (b) cell optical density.

2.4 Conclusion

Our investigation appears to be the first report of continuous fermentation by E. coli

LY01. Specific cell growth rates were calculated from a set of batch fermentations, which was slightly decreased with initial glucose concentration increased. The growth of E. coli

LY01 was also dependent on LB concentration. The arguments of the relative stability of

E. coli KO11 during chemostat culture on glucose and sugar mixtures were reported [10].

As the ethanol-tolerant mutants of E. coli KO11, our results proved that the chemostat fermentation of E. coli LY01 with glucose is stable. At steady state of chemostat fermentation, the dilution rate is equal to the specific cell growth rate. However, this steady state only means the cell growth rate and death rate is equal, do not matches with high glucose conversion or productivity. During the chemostat fermentation with

D=0.026 h-1, we got comparable volumetric productivity with batch close to 1 g L-1 h-1.

When dilution rate slightly increased to 0.04 h-1, cells were not washed out; however, the glucose conversion dropped.

The EtOH productivity from a cell-immobilized reactor (ICR) should be higher than batch and chemostat fermentations. With the same dilution rate of 0.04 h-1, glucose conversion should be increased compared with the results from chemostat fermentation.

With different cell-immobilized method, ICR is able to immobilize 5 to 10 times higher cell density than cell suspended reactor. Therefore, with the same reactor scale, ICR should have higher EtOH productivity. In the Chapter 3, two kinds of materials (porous gel and fiber) were used for cell immobilization. And both column reactor and perfusion mode reactor were designed and compared.

Texas Tech University, Lan Ma, December 2014

2.5 References

[1] Jarboe, L. R.; Grabar, T. B.; Yomano, L. P.; Shanmugan, K. T.; Ingram, L. O. Development of ethanologenic bacteria. Advances in Biochemical Engineering /Biotechnology 2007, 108, 237-261.

[2] Ohta, K.; Beall, D. S.; Mejia, J. P.; Shanmugam, K. T.; Ingram, L. O. Genetic improvement of Escherichia coli for ethanol production: chromosomal integration of Zymomonas mobilis genes encoding pyruvate decarboxylase and alcohol dehydrogenase-II. Applied and Environmental Microbiology 1991, 57 (4), 893- 900.

[3] Dien, B. S.; Hespell, R. B.; Ingram, L. O.; Bothast, R. J. Conversion of corn milling fibrous co-products into ethanol by recombinant Escherichia coli strains K011 and SL40. World Journal of Microbiology & Biotechnology 1997, 13 (6), 619-625.

[4] Leite, A. R.; Guimaraes, W. V.; de Araujo, E. F.; Silva, D. O. Fermentation of sweet whey by recombinant Escherichia coli KO11. Brazilian Journal of Microbiology 2000, 31 (3), 212-215.

[5] Takahashi, C. M.; Lima, K. G. D.; Takahashi, D. F.; Alterthum, F. Fermentation of sugar cane bagasse hemicellulosic hydrolysate and sugar mixtures to ethanol by recombinant Escherichia coli KO11. World Journal of Microbiology & Biotechnology 2000, 16 (8-9), 829-834.

[6] Martinez, A.; York, S. W.; Yomano, L. P.; Pineda, V. L.; Davis, F. C.; Shelton, J. C.; Ingram, L. O. Biosynthetic burden and plasmid burden limit expression of chromosomally integrated heterologous genes (pdc, adhB) in Escherichia coli. Biotechnology Progress 1999, 15 (5), 891-897.

[7] Huerta-Beristain, G.; Utrilla, J.; Hernandez-Chavez, G.; Bolivar, F.; Gosset, G.; Martinez, A. Specific ethanol production rate in ethanologenic Escherichia coli strain KO11 is limited by pyruvate decarboxylase. Journal of Molecular Microbiology and Biotechnology 2008, 15 (1), 55-64.

[8] Lawford, H. G.; Rousseau, J. D. Loss of ethanologenicity in Escherichia coli B recombinants PLOI297 and KO11 during growth in the absence of antibiotics. Biotechnology Letters 1995, 17 (7), 751-756.

[9] Lawford, H. G.; Rousseau, J. D. Factors contributing to the loss of ethanologenicity of Escherichia coli B recombinants pLOI297 and KO11. Applied Biochemistry and Biotechnology 1996, 57 (8), 293-305.

[10] Dumsday, G. J.; Zhou, B.; Yaqin, W.; Stanley, G. A.; Pamment, N. B. Comparative stability of ethanol production by Escherichia coli KO11 in batch

Texas Tech University, Lan Ma, December 2014

and chemostat culture. Journal of Industrial Microbiology & Biotechnology 1999, 23 (1), 701-708.

[11] Martin, G. J.; Knepper. A.; Zhou, B.; Pamment, N. B. Performance and stability of ethanologenic Escherichia coli strain FBR5 during continuous culture on xylose and glucose. Journal of Industrial Microbiology & Biotechnology 2006, 33 (10), 834-844.

[12] Zhou, B.; Martin, G. J.; Pamment, N. B. Increased phenotypic stability and ethanol tolerance of recombinant Escherichia coli KO11 when immobilized in continuous fluidized bed culture. Biotechnology and Bioengineering 2008, 100 (4), 627-633.

[13] Yomano, L. P.; York, S. W.; Ingram, L. O. Isolation and characterization of ethanol-tolerant mutants of Escherichia coli KO11 for fuel ethanol production. Journal of Industrial Microbiology & Biotechnology 1998, 20 (2), 132-138.

[14] Gonzalez, R.; Tao, H.; Purvis, J. E.; York, S. W.; Shanmugam, K. T.; Ingram, L. O. Gene array-based identification of changes that contribute to ethanol tolerance in ethanologenic Escherichia coli: Comparison of KO11 (Parent) to LY01 (resistant mutant). Biotechnology Progress 2003, 19 (2), 612-623. . [15] Dombek, K. M.; Ingram, L. O. Effects of ethanol on the Escherichia coli plasma- membrane. Journal of Bacteriology 1984, 157 (1), 233-239.

[16] Lau, M. W.; Gunawan, C.; Balan, V.; Dale, B. E. Comparing the fermentation performance of Escherichia coli KO11, Saccharomyces cerevisiae 424A(LNH-ST) and Zymomonas mobilis AX101 for cellulosic ethanol production. Biotechnology for Biofuels 2010, 3, 11-21.

[17] Ingram, L. O.; Gomez, P. F.; Lai, X.; Moniruzzaman, M.; Wood, B. E.; Yomano, L. P.; York, S. W. Metabolic engineering of bacteria for ethanol production. Biotechnology and Bioengineering 1998, 58 (2-3), 204-214.

[18] Asghari, A.; Bothast, R. J.; Doran, J. B.; Ingram L. O. Ethanol production from hemicellulose hydrolysates of agricultural residues using genetically engineered Escherichia coli strain KO11. Jounal of Industrial Microbiology 1996, 16, 42-47.

[19] Barbosa, M. D. S.; Beck, M. J.; Fein, J. E.; Potts, D.; Ingram, L. O. Efficient fermentation of pinus sp. acid hydrolysates by an ethanologenic strain of Escherichia coli. Applied and Environmental Microbiology 1992, 58 (4), 1382- 1384. [20] Zaldivar, J.; Nielsen, J.; Olsson, L. Fuel ethanol production from lignocellulose: a challenge for metabolic engineering and process integration. Applied Microbiology and Biotechnology 2001, 56 (1-2), 17-34.

Texas Tech University, Lan Ma, December 2014

[21] Klinke, H. B.; Thomsen, A. B.; Ahring, B. K. Inhibition of ethanol-producing yeast and bacteria by degradation products produced during pre-treatment of biomass. Applied Microbiology and Biotechnology 2004, 66 (1), 10-26.

[22] Zaldivar, J.; Ingram, L. O. Effect of organic acids on the growth and fermentation of ethanologenic Escherichia coli LY01. Biotechnology and Bioengineering 1999, 66 (4), 203-210.

[23] Ingram, L. O.; Buttke, T. M. Effects of alcohols on micro-organisms. Advances in Microbial Physiology 1984, 25, 253-300.

[24] Zaldivar, J.; Martinez, A.; Ingram, L. O. Effect of alcohol compounds found in hemicellulose hydrolysate on the growth and fermentation of ethanologenic Escherichia coli. Biotechnology and Bioengineering 2000, 68 (5), 524-530.

[25] Zaldivar, J.; Martinez, A.; Ingram, L. O. Effect of selected aldehydes on the growth and fermentation of ethanologenic Escherichia coli. Biotechnology and Bioengineering 1999, 65 (1), 24-33.

[26] Dien, B. S.; Cotta, M. A.; Jeffries, T. W. Bacteria engineered for fuel ethanol production: current status. Applied Microbiology and Biotechnology 2003, 63 (3), 258-266.

[27] Li, X. Kinetic study and modelling of ethanol production from glucose/xylose mixtures using the genetically-engineered strain Escherichia coli KO11. Master Dissertation, Department of Chemical and Biological Engineering, University of Ottawa, 2008. [28] Ryu, S. Production of bioethanol and biobutanol using genetically engineered Escherichia coli. Ph.D. Dissertation, Department of Chemical Engineering, Texas Tech Universiy, 2010.

Texas Tech University, Lan Ma, December 2014

Chapter 3

Design of a Continuous-Flow, Packed-Bed Fermentation Process

3.1 Introduction

Compared with traditional batch fermentation, immobilized cell fermentation (ICR) is

a better approach to improve ethanol economics with higher conversion rates, faster

fermentation rates, and reduced product losses [1-3]. Ethanol concentration is normally

kept below a threshold (often in the range of 50 to 100 g/L) to prevent cells from dying in

the free-cell suspended reactor, which is not a constraint in an ICR. ICR can provide

normally 5 to 10 times higher cell densities by attaching or entrapping cells to a solid

scaffold and at the same time allowing the exchange of substrate and product with the

surrounding liquid [4]. In a column-type ICR, especially in the lignocellulosic

fermentation, inhibitor or toxins could continuously flow out, which keeps their

concentration low near the reactor inlet, reducing product inhibition [5-8]. For similar

production capacity, the size of a typical ICR system is about 5 or 10 times smaller than a

comparable batch or semi-batch reactor. However, the stability of ICR systems is a big issue for industrial application.

The technique of cell immobilization in ICR is very important, which directly influences the performance of the ICR system. Generally, there are four major immobilization methods: attachment to a surface, entrapment within a porous matrix, containment behind a barrier, and self-aggregation [9]. The entrapment method turned out to be a very popular one since the first publication of alginate-entrapped yeast cells used for EtOH production by Nagashima et al. [10]. Hereafter, modified alginate gels with sodium, potassium and alumina were reported for yeast immobilization [8, 11]. In

Texas Tech University, Lan Ma, December 2014

this method, the porous matrix is synthesized in situ around cells. Some other types of

porous matrix include sponge, porous glass microspheres, bricks, ceramics and

polyurethane foam [8, 12, 13]. Glass-fiber pads and aluminum silicate scaffold have been

used to attach cells for EtOH production [14, 15] However, it is difficult to control pore

size and connectivity. Instead of cell entrapment, immobilization can also be achieved by

natural adsorption or chemical bond between cells and solid support due to electrostatic interactions (van der Walls force), ionic (Lewis acid/base) and hydrophobic interactions

[16]. The disadvantages of adsorption are possible detachment and of cells, therefore, this

method may be not useful when cell-free effluent is required.

The maximum productivity is often observed with the maximum dilution rate. ICR allows high dilution rate to exceed the specific growth rate of the microorganism. Also, maximum productivity does not correspond to complete substrate utilization [17]. One disadvantage of ICR system is that the cellular physiology and morphology cannot be controlled. The immobilization procedure can alter the metabolic activity or viability of cells, which may cause the chief product to decline and affect the selectivity of a reaction complex. There are some examples in the literature showing that microorganisms are distorted by the pressure of the surrounding organisms. In addition, the high cell density in immobilization systems greatly reduces the diffusivity of nutrients, which can cause a portion of the otherwise viable biomass to become inactive [18]. Therefore, the viability of overall immobilized biomass has been estimated to be less than 20 % in some systems

[19]. On the other hand, cell immobilization is also reported to improve the phenotypic stability and ethanol tolerance of E. coli KO11 during continuous culture [20].

Texas Tech University, Lan Ma, December 2014

Our work focuses on designing of an efficient continuous-flow reactor system for bio-

ethanol production. Poly(ethylene terephthalate) (PET) fibers with 40-55 μm diameters were applied in the ICR. Stirred-tank and vertical column reactor designs have been compared with simple glucose feedstocks. With 70 g/L glucose in the feed, the immobilization of cells allows dilution rate and volumetric productivity to exceed the values obtained in chemostat mode of operation. Another cell immobilization method with synthetic porous polymer scaffold (SPPS) was also investigated. We studied the effects of gel particle size, porous volume, dilution rate, nutrient and substrate on the reactor performance. The advantages of our ICR system are high cell density, high productivity, and lack of gas holdup issue. However, it is difficult to control cell growth and reactor stability in the packed-bed column reactor, resulting in poor substrate conversion when feed concentration exceeds 20 g/L glucose.

3.2 Materials and methods

3.2.1 Segmented vertical column reactor and immobilized cell stirred tank reactor

The ICR is a segmented vertical column reactor with a total volume of 110 ml. Its

overall length/diameter (L/D) ratio is 18. An illustration of the reactor system is presented

in Fig. 3.1. Coiled tubing was used as a heat exchanger to control the fermentation temperature at 35 °C. A flow breaker was installed between the pump and column reactor to prevent feed tank contamination. The reactor was packed with SPPS particles or PET.

About 45 % of reactor volume was occupied by SPPS particles. However, fibers only

took 10 % of reactor volume and the mass of fiber was controlled at 3 g in each segment.

The reactor is separated into five segments constrained between autoclavable plastic

disks with numerous small holes (sieve plates). The hole size is slightly smaller than the

Texas Tech University, Lan Ma, December 2014

SPPS particle size. These sieve plates were used to eliminate problems with bed

compaction and pressure buildup. Prior to startup of the bioreactor, the whole

fermentation system was sterilized by autoclaving in 121°C for 15 minutes. SPPS

particles or fibers were packed into the reactor and sterilized within the reactor before

inoculation with cells. No pH control was performed during the fermentation. The initial

pH of feed solution is between 6.5 to 7.0.

Figure 3.1. Scheme of fermentation process with fiber immobilized segmented vertical column reactor

Another fermentor design tested is immobilized cell stirred tank reactor. Fibers were

stored between two porous steel plates in the center of a 2.5 L bioreactor (New

Brunswick Scientific BioFlo 310). Operating liquid volume is 1.4 L. Fiber amount was

varied between 2.5 g to 7.0 g, which was only 1 %-4 % of the fermentation volume. The

temperature was maintained at 35 °C and the stir rate at 200 rpm. pH was controlled at

6.0 by automatically adding 4.0 M NaOH.

Texas Tech University, Lan Ma, December 2014

3.2.2 Inoculum preparation

The microorganism information and cell culture process have already been

described in Chapter 2. Medium and inoculum preparation methods were same as in

batch fermentation runs. After nitrogen sparging, pre-cultivated cells were pumped into the reactor from the inlet of immobilized cell tank or column reactor. The volume of inoculum added was enough to provide approximately 1.0 OD at 600 nm. Column reactor was placed horizontally during the first 7 hours and run as a batch reactor without stirring, in a fashion similar to the tank reactor, which provided give cells enough time to enter the void space within the SPPS or to attach to fiber surface.

3.2.3 Analytical methods

The free cell density was measured at the outlet via optical density at 600 nm with

Evolution 260 BIO UV-visible spectrophotometer. After fermentation, fibers in each segment were taken out and dried in air for 3 days. The dry mass of immobilized cells was determined from the total dry mass of fiber segment minus the fiber mass. Ethanol

and glucose concentrations were determined by high-performance liquid chromatography

with a Phenomenex Rezex ROA column, equipped with a RI detector (Smartline RI

detector 2300, Knauer).

Scanning electron microscopy (SEM) images of porous gels were obtained under low

vacuum (50 Pa) using Hitachi TM-1000 Tabletop Microscope. A freshly cleaved surface was generated by slicing the swollen gel with a razor blade. The wet gel was then placed under vacuum in the SEM sample chamber and imaged immediately. Images were recorded for the gels' surfaces in a partially hydrated state, as the samples began to lose water immediately upon being placed under vacuum.

Texas Tech University, Lan Ma, December 2014

SEM images of PET fibers (selected fibers after fermentation with immobilized cells)

were taken with Hitachi S-4300 scanning electron microscope. Samples were immersed

overnight at 4 ºC in 2.5 % glutaraldehyde (Grade I, 50 % in H2O, Sigma) in 0.1 M

phosphate buffer (pH 6.5) for fixation, then rinsed thoroughly with phosphate buffer, and

dehydrated with 25 %, 50 %, 70 %, 80 %, 95 % and 100 % ethanol by submerging

samples in each concentration for 10 min. After dehydration, the dry samples were put in

a 50:50 mixture of acetone and HMDS [1,1,1,3,3,3-hexamethyldisilazane] (Sigma) for 30

min, followed by another 30 min in pure HMDS and then air-dried on filter paper [21].

The dried samples were sputter-coated with a thin layer of gold.

3.3 Results and discussion

Our study focuses on designing and optimizing ICR system by adjusting

immobilization materials, nutrient contents, dilution rate and packing volume. Three scaffold materials were studied in our research for cell immobilization. One is the SPPS material, which was synthesized by our own group [22]; the other two are glass fiber (7

μm in diameter) and PET fiber (Fairfield POLY-FIL, 40 to 55 μm in diameter). The

performance of different scaffold materials was examined in both the stirred tank and

column reactors.

3.3.1 Cell immobilization with synthetic porous polymer scaffold (SPPS)

3.3.1.1 Synthetic porous gel characterization

The porous, sponge-like, crosslinked networks of poly (hydroxyethylmethacrylate),

poly(HEMA) were developed earlier in Dr. Hedden's group [22]. The gel was

synthesized by a free-radical polymerization of a mixture of 2-hydroxyethylmethacrylate

(HEMA), crosslinker, and a certain volume fraction of porogen, a water-soluble poly

Texas Tech University, Lan Ma, December 2014

(ethylene glycol). In the water-swollen state, poly (ethylene glycol) was dissolved in the water and a randomly structured, interconnected pore network of volume fraction 0.3 to

0.8 was contained in the gel. The mean pore size was measured between 0.5 to 30 µm, which is enough to allow rod-shaped, ethanolognic bacteria (of length 1 to 5 µm and diameter≈1 µm) to enter the gel. Compared with traditional hydrogel material such as calcium alginate, poly (HEMA) gel matrix absorbs approximately 35 mass % water

(excluding water in the pores), so the water content in the polymer phase is lower and it has higher mechanical strength. There is not any macroscopic degradation of SPPS observed during autoclaving and fermentation. Fig. 3.2 shows a low-vaccum (~50 Pa)

SEM image of the surface of a partly hydrated SPPS particle. The shrinkage of pores during drying frustrates quantitative analysis of pore size. To obtain a more quantitative picture of the average pore size and the volume fraction of water-filled pores in the SPPS, we therefore characterized their structure (while swollen in D2O) using non-invasive ultra small-angle neutron scattering (USANS) methods (Fig. 3.3). Table 3.1 summarizes characteristics of several SPPS determined by fitting to a scattering model, including the mass % of porogen prior to extraction (wp), volume fraction of pores in the swollen state

(φ1s), and correlation length of porosity ξpores (µm). The average chord length for the pores, a measure of the mean pore diameter for a randomly structured porous material, is equal to (ξpores/(1-φ1s)). The size and volume fraction of the pores can be varied by changing the mass fraction of the porogen. SPPS having wp higher than about 70 % have dramatically increased pore size, at the expense of mechanical toughness [22].

Texas Tech University, Lan Ma, December 2014

Figure 3.2. Low-vacuum SEM image of partly desiccated SPPS surface. The porogen has already been removed. Pores shrank as the material dries under vacuum, so the pores are assumed to be somewhat larger in the fully hydrated state. This SPPS contained about 75 mass % porogen prior to extraction. Adapted from (Lannuzzi, M. A.; Reber III, R.; Lentz, D. M.; Zhao, J.; Ma, L.; Hedden, R. C. Polymer 2010, 51 (9), 2049-2056.)

Figure 3.3. USANS data for poly(hydroxyethylmethacrylate) SPPS (extracted, porogen removed). Adapted from (Lannuzzi, M. A.; Reber III, R.; Lentz, D. M.; Zhao, J.; Ma, L.; Hedden, R. C. Polymer 2010, 51 (9), 2049-2056.)

Texas Tech University, Lan Ma, December 2014

Table 3-1. Characteristics of porous polymer scaffold materials determined by USANS. Adapted from (Lannuzzi, M. A.; Reber III, R.; Lentz, D. M.; Zhao, J.; Ma, L.; Hedden, R. C. Polymer 2010, 51 (9), 2049-2056.)

(mass %) φ1s ξpores (µm) 40 0.370 6.15 50 0.388 1.60 60 0.408 1.20 70 0.508 1.13 80 - 28.4

Another critical parameter of interest is the diameter of the SPPS particles to be used in the packed-bed ICR (generally in the range of 200 µm to 1 cm). To obtain different diameters of SPPS particles, SPPS were synthesized in the Teflon jar and soaked in water for 2 days to remove the porogen, then ground up in a commercial food processor.

The particles were subsequently fractionated through a series of sieves to obtain particles with different mash sizes. This fabrication method is straightforward, and it could easily be scaled up for industrial production. Fig. 3.4 shows a typical size distribution of particles after grinding. The fermentation effects from gel particles with different pore volume fractions and particle sizes were studied in this paper.

An important distinction between these SPPS and the traditional gels (e.g. calcium alginate) is the mode of loading with cells. In traditional ICR systems, cells are usually encapsulated in the gel prior to the fermentation. SPPS particles were loaded into the reactor with medium and sterilized before inoculation with cells, which were able to enter the particles through surface-accessible pores. After cells reached the exponential growth phase, continuous flow was initiated. Because cells were free to enter and exit the SPPS, immobilization was partial, and some continuous loss of cells was possible. However, the loss of some small amount of cells was not enough to cause washout and it was not necessary to employ cell-recycle schemes. The advantages of facile gas ventilation and 67

Texas Tech University, Lan Ma, December 2014

substrate transport through the pores outweigh any lost productivity due to consumption of substrate to produce new cells. The SPPS packing also makes sterilization straightforward on any scale, as the entire reactor (including packing) can be autoclaved prior to introduction of cells.

Figure 3.4. Particle size distribution of SPPS after grinding. Particles can be separated into fractions by standard sieving method. Shared bar represents particles of 1.0 to 1.7 mm diameter, which was selected for bioreactor runs.

3.3.1.2 Vertical column reactor with SPPS

Batch fermentation was conducted to examine the effects of SPPS on biocompatibility. One flask fermentation containing 25 mass % SPPS, 10 g/L glucose and

25 g/L LB, was conducted at 35° C under anaerobic conditions. A control flask fermentation without SPPS was run under identical conditions. Both optical cell density and conversion of glucose vs. time profiles were unaffected by the presence of SPPS.

Glucose was completely consumed in 5 h with volumetric productivity of 0.94 g/L/h

(final EtOH concentration 4.75 g/L; 95 % of theoretical maximum yield).

Texas Tech University, Lan Ma, December 2014

Column reactor was firstly operated as a control experiment without any

immobilization scaffold. Reactor was fed with 30 g/L glucose and 25 g/L LB at dilution

rate of 0.2 h-1, which is close to the specific cell growth rate (0.257/h) obtained from

batch fermentation with the same initial glucose concentration. The reactor was run as a

batch for the first 7 hours, and then started to run as a continuous fermentation. Cells were washed out from OD=4.25 to OD=1.5 in about 100 hours (Fig. 3.5). Also, remaining measured glucose at the exit of column increased with time. Without cell immobilization in the column, ethanol concentration was very low (4 g/L) and cells were washed out.

(a)

Texas Tech University, Lan Ma, December 2014

(b)

Figure 3.5. Control experiment without gels inside column. The initial glucose concentration was 50 g/L with dilution rate of 0.2 h-1. (a) Gucose and ethanol concentrations; (b) Optical cell density. The gel particle diameter was chosen from 1.0-1.7mm size with 75 % internal porosity.

The external void volume fraction between the SPPS particles was approximately 0.55

(determined by draining free liquid). Fermentation was started to operate at dilution rate

of 1.39 h-1 over a 4 day period (Fig. 3.6). The glucose concentration in the feed was 10

g/L and steady-state operation was achieved after 20 h. The glucose conversion was more

than 95 %. The effluent ethanol concentration reached to 4.7±0.5 g/L over the experiment,

and the measured volumetric productivity was 6.533±0.70 g/L/h, which is 7 times higher

-1 -1 than batch productivity (0.94 g L h ). During the fermentation, the evolution of CO2

bubbles was noted after about 20 hours from the SPPS-packed segments (Fig. 3.7). Since

the packed gel was separated by plastic sieve-plates in the reactor, the fluidization of 1.0

Texas Tech University, Lan Ma, December 2014

mm mesh size particle was not an issue. The problem with alginate immobilization was usually that the gels floated to the top due to CO2 holdup. With the segmented column

design, the particles in our column only tended to float to the top of each segment due to

CO2 generation

Figure 3.6. Glucose and EtOH concentrations exiting the column reactor (dilution rate 1.41 h-1).

Figure 3.7. Packed-bed section of glass column reactor. The coiled tubing is a heat exchanger.

Texas Tech University, Lan Ma, December 2014

Column reactor fermentation was repeated with different dilution rates. Continuous

fermentation was run with dilution rate of 1.2 h-1 at the beginning, and then increased to

3.6 h-1 when all the glucose was consumed (Fig. 3.8a). Glucose conversion was more

than 95 % even after dilution rate was increased. The effluent ethanol concentration for

D=3.6 h-1 reached 4.1± 0.1 g/L and the volumetric productivity was 14.76 ± 0.36 g L-1 h-1,

which is about 15 times higher than batch with 10 g/L glucose. Besides immobilized cells,

there were still some free-suspended cells in the reactor. It is difficult to measure the

immobilized cell density; however, we can measure the free-suspended cell density at the

exit of the column reactor. Fig. 3.8b shows that the cell density at the exit of column

reactor increased and reached a steady state during the first stage (D=1.2 h-1). When dilution rate was increased to 3.6 h-1, cell density was also increased due to high nutrient

flux, and then reached steady state after 30 hours. Compared with the column reactor

control experiment (Fig. 3.5), the dilution rates of these cells immobilized fermentations

were much higher and cells were not washed out. Based on glucose conversion and cell

density results, it is clear that our system can immobilize cells during fermentation

without washing them away. However, cell dry mass is not measurable due to the mass

change of the swollen gel and dry gel.

Texas Tech University, Lan Ma, December 2014

(a)

(b) Figure 3.8. ICR fermentations with different dilution rates at 1.2 h-1 and 3.6 h-1. Except dilution rate, the other operating conditions were kept the same as Fig. 3.4. (a): Glucose and EtOH concentrations at column exit; (b) cell density at column exit.

Texas Tech University, Lan Ma, December 2014

The gel packed-bed column reactor works well with SPPS as immobilization scaffold on the fermentation with 10 g/L glucose concentration in the feed. We continuously worked on fermentations with high glucose concentration up to 50 g/L with different dilution rates at 0.1 h-1 and 1.2 h-1; however, the highest glucose conversion we could get was only about 50 % (Fig. 3.9). Lower dilution rate actually decreased glucose conversion. Normally, decreasing dilution rate will achieve longer retention time and higher cell mass. However, diffusion of substrate and nutrients may have caused a problem when cell mass was increased. Cellular physiology and morphology are difficult to control in the cell immobilized reactor. Therefore, the reasons of low conversion with higher glucose concentrations in the feed may be various

(a)

Texas Tech University, Lan Ma, December 2014

(b)

Figure 3.9. ICR fermentations with different dilution rates at (a): 1.2 h-1 and (b): 0.1 h-1. Glucose concentration in the feed was 50 g/L. Glucose and EtOH concentrations at column exit were measured.

A series of experiments were conducted to probe the reason for the poor conversion

observed with increasing glucose concentration in the feed. We firstly tried to start the fermentation with low glucose concentration and increased concentration step by step.

The purpose was to avoid environmental stress to the cells. Two feeding tanks were

prepared; one had 10 g/L glucose and the other one had 100 g/L glucose. Final glucose

concentration was set by controlling relative flow rates of these two tanks. Also, the feeding medium was pre-warmed before pumping to reactor. The packed gel was still chosen from 1.0-1.7mm size with 75 % porous volume. Dilution rate was 1.4 h-1.

Fermentation was repeated and both results are shown in Fig. 3.10. The conversion

reached almost 100 % with 10 g/L glucose, but immediately started to decrease when

Texas Tech University, Lan Ma, December 2014

initial glucose concentration increased to 15 g/L. The conversion was 60 % to 70 % with

15 g/L and decreased to only 45 % with 20 g/L glucose. Step by step feeding did not

increase glucose conversion at this dilution rate with the amount of time permitted in between concentration steps. The data did not provide any indication that conversion was improving after 90 h fermentation time, so this strategy was judged to be unsuccessful.

Figure 3.10. Column reactor fermentations conducted by increasing substrate concentration in the feed. Dilution rate was kept constant during the whole process at 1.4 h-1. We also examined the effects of SPPS particle size on the reactor performance

(glucose conversion) and operating problems (pressure buildup, gas holdup). Small

particle size normally facilitates the exchange of sugars, EtOH, and by-products between

particles and surroundings because the specific surface area increases as particle size

Texas Tech University, Lan Ma, December 2014

decreases. In addition, higher cell density will likely be found with small particles. A comparison experiment was performed with three different ranges of particle size (Fig.

3.11). Fermentation was run with particle size of 1.75-3.5 mm at feed concentration of 50 g/L glucose and 25 g/L LB. Dilution rate was 0.96 h-1 and 50 % of the column volume was occupied with SPPS. The highest glucose conversion was 45 %. At a similar dilution rate of 0.99 h-1, fermentation was taken with a small particle size of 1.0-1.75 mm. The

highest glucose conversion was 40 % before the run was terminated by a reactor leak.

The smaller particles are expected to increase the pressure over the column, so leaking is

more likely. In order to increase glucose conversion, a much smaller particle size with

diameter of 0.71 mm was used with 0.17 h-1 dilution rate. However, the result was even worse with much lower glucose consumed (20 % conversion). Therefore, the results showed the glucose conversion was not improved by decreasing the particle size. Even though the cell density was increased, the pressure buildup and mass transfer limitations made the reactor very unstable and therefore decreased the reaction overall rate.

We also studied the effects of gel porous volume on the reactor performance between

75 % and 85 % volume. SPPS having 85 % internal porosity are softer than those having

75 % internal porosity (Fig. 3.12). Glucose conversion and ethanol productivity were found to similar, regardless of internal porosity.

Besides the low substrate conversion observed with glucose feed concentrations

> 20 g/L, the second serious problem encountered with the main operational problem.

Higher packing volume and cell density made the reactor very unstable and subject to

leaking due to build-up of biomass. Cell mass was difficult to control; therefore, each fermentation was difficult to repeat, and runs could not be continued for long periods of

Texas Tech University, Lan Ma, December 2014

time without risk of pressure build-up. Our SPPS packed column reactor can obtain a 15 times higher volumetric productivity than batch reactor with low glucose feed concentration. It also decreases the gas holdup problem compared with calcium alginate gels packed reactor. However, column reactor is not industrially feasible if it only works great with low substrate feed concentration, and the operation stability is also a concern.

Figure 3.11. Comparison of glucose conversion during the fermentation with SPPS particles of different diameters.

Texas Tech University, Lan Ma, December 2014

Figure3.12. Column reactor fermentation conducted with 85 % internal porosity. Dilution rate was 0.2 h-1 at the first 70 hours, then increased to 1.0 h-1. Glucose concentration in the feed was 50 g/L. 3.3.2 Cell immobilization with poly(ethylene terephthalate) fiber scaffold

3.3.2.1 Polyethylene terephthalate fiber characterization

poly(ethylene terephthalate) fibers (Fairfield POLY-FIL, 40 to 55 μm in diameter) were used as received from a local vendor without any modification. The fibers occupy less volume than the SPPS, leaving more room for cells to grow and for fluid to occur. To test the dependence of pressure drop on fiber volume fraction, we emptied a HPLC column and filled it with different amount of fibers, then measured the pressure. The

HPLC pump pressure gauge showed 0 until the column was fully and tightly packed with about 18 % volume of fibers. Problems encountered early in the project with pressure buildup using SPPS packing were circumvented by switching to the fibers, which occupied a column reactor volume fraction of only 10 % compared to 45 % for the SPPS.

Texas Tech University, Lan Ma, December 2014

Work supporting the use of fibers for immobilization of ethanologens has been described

elsewhere and showed promising results [23]. Glass fibers of diameter 7 µm were also

tested, but performed poorly. The pressure in a fully packed glass fiber column was easily

buildup, which caused serious leaking problem. Fig. 3.13 show the microscope image of

PET fibers and SEM image of E. coli LY01 immobilized on the fibers.

(a)

(b)

Figure 3.13. (a): PET fibers (as received) used in packed bed section. (b): SEM of the E. coli LY01 cells immobilized onto a PET fiber.

Texas Tech University, Lan Ma, December 2014

3.3.2.2 Vertical column reactor with polyester fiber scaffold

From previous experiments with SPPS, we know fermentation with a high volume fraction of packing is very unstable since it created enough pressure buildup in the column to cause leaking on a regular basis, which is not observed for the column reactor with PET fibers. The immobilized cell dry mass can be measured after fermentation because fibers do not absorb significant water compared with hydrogel (SPPS). Also, a void volume fraction in a fully packed fiber column reactor is approximately 90 %, which is much higher than packed gel column (55 %). In addition, cells are immobilized by attaching on the fiber surface, not by entrapping inside the gels; therefore, mass transfer limitations are alleviated.

Flask fermentations were conducted firstly with some PET fibers inside to test fiber toxicity. Both achieved ethanol and cell concentrations indicated that fiber is not toxic to the cells (Fig. 3.14).

(a)

Texas Tech University, Lan Ma, December 2014

(b)

Figure 3.14. Flask fermentations with and without fibers inside. (a): Glucose and ethanol concentrations; (b): Optical cell density. The column was still separated into five sections and each section was filled with 3 g

fiber. Column and fibers were autoclaved together before pouring cell culture inside. The

initial cell density was controlled at OD=1 at 600 nm. Column was held horizontally for

7 hours as a batch fermentation before continuously pumping fresh medium through reactor. The fermentation process was the same as the process with SPPS. After fermentation, fibers from each segment were taken out and dried, and then the dry cell mass was measured by subtracting the dry fiber mass.

The highest glucose concentration that could be completely consumed in the SPPS- packed column reactor was 15 g/L. With same flow rate and LB concentration, the highest glucose concentration that the fiber-packed column reactor could completely convert to ethanol was is 20 g/L, only a slight improvement. In Fig. 3.15, substrate and nutrient concentrations in the feed tank were 20 g/L glucose and 25 g/L LB. Dilution rate

Texas Tech University, Lan Ma, December 2014

was 0.84 h-1. Results show that all the glucose was consumed in about 30 hours with glucose conversion more than 95 % (Fig. 3.15a). The effluent ethanol concentration

reached 9.0±0.5 g/L over the fermentation, and the volumetric productivity at steady state

was 7.56±0.42 g/L/h. Fig. 3.15b presents dry cell density in each column section, which

indicated very high cell mass was immobilized by fibers (OD=1 is about 330 mg dry cell mass /L). The average immobilized cell mass density is 57 g dry cells/ L, which is about

95 % of the total cell mass inside the reactor. Immobilized cell density was decreasing from column bottom to the top and one third of cells were gathered in the bottom section.

Cells accumulation may be due to two reasons: cells prefer to accumulate in the nutrient- rich environment, which is in the bottom of column; and cells tend to sink to the bottom due to high specific gravity. The color of the fibers at the bottom was observed to be a darker brown.

(a)

Texas Tech University, Lan Ma, December 2014

(b)

In the fiber immobilized column reactor, 30 g/L glucose cannot be completely consumed with 25 g/L LB; however, the glucose was completely consumed with 40 g/L

LB. The effluent ethanol concentration was 7.0±0.5 g/L, which was only 52 % of theoretical maximum. The volumetric productivity was lower than that observed with 20 g/L glucose fermentation; however, the average cell density was higher (Fig. 3.16b).

Lactic acid concentration from the last two samples was observed with 3.1 g/L. Most of glucose was probably used for cell growth or maintenance, and acid production, not for ethanol production. Besides immobilized cells, heavy precipitation of “floc” was observed in the bottom of column. It caused instability in the column’s performance, flow resistance and mass transfer limitations.

Texas Tech University, Lan Ma, December 2014

(a)

(b)

Figure 3.16. ICR fermentation with PET fibers as immobilization scaffold. Glucose and LB concentrations in the feed tank were 30 g/L and 40 g/L, respectively. Dilution rate was 0.84 h-1. (a): Residual glucose and ethanol concentrations; (b) Immobilized cell dry mass density.

Texas Tech University, Lan Ma, December 2014

The fiber-packed column was then examined with a rich medium involving 50 g/L glucose, 75 g/L LB at different dilution rates. Even with very high LB concentration, only less than 50 % of the glucose was consumed at dilution rate of 0.84 h-1 (Fig. 3.17a).

Dilution rate was then decreased to 0.147 h-1; however, the reactor was very unstable and oscillation was observed (Fig. 3.17b), which similar to that observed in the gel column reactor. In the packed-bed column reactor, decreasing dilution rate caused cell flocculation and decreased reactor performance. However, when the column reactor was started with dilution rate of 0.84 h-1 for 5 days and decreased to 0.147 h-1, glucose conversion immediately increased to 90 % and then started to drop (Fig. 3.17c). Fig. 3.17 again illustrates the stability problem of the continuous-flow immobilized column reactor.

(a)

Texas Tech University, Lan Ma, December 2014

(b)

(c)

Figure 3.17. ICR fermentation with cells immobilized on PET fibers. Initial glucose and LB were 50 g/L and 75 g/L, respectively. The fiber amount inside column was 15 g. Residual glucose and ethanol concentrations were measured during the fermentation with constant dilution rate at (a) 0.84 h-1; (b) 0.147 h-1 and (c) changing dilution rate from 0.84 h-1 to 0.147 h-1 during the fermentation.

Texas Tech University, Lan Ma, December 2014

A new column was designed with four small sample ports evenly spaced along the

column and one large sample port on the top (Fig. 3.18c). Fibers were packed in the same

way inside the column. Because column reactor cannot consume high concentration glucose, we therefore designed a new process by connecting the outlet of column with a tank reactor. The flushed free cells and medium were fed into the tank reactor, and fermentation was continuously run as a chemostat. Dilution rate in the column reactor was 0.84 h-1. Inlet flow rate of column reactor and outlet flow rate of tank reactor were

kept at 1.4 ml/min. Glucose and LB concentrations were 100 g/L and 25 g/L. Half of

glucose was consumed. About 25 g/L glucose was consumed by the column reactor and

another 25 g/L was consumed in the tank reactor. By comparing ethanol concentration

from different sample ports (Fig. 3.18b), about 60 % of the ethanol was actually produced

from bottom section of column. It may explain why column reactor did not work well

with high glucose concentration, because almost half of column did not contribute to the

fermentation. At steady state, the overall volumetric productivity and conversion were

1.14 g L-1 h-1 and 50 %. When fermentation reached steady state, the pump was stopped

and started to run as a batch until all the leftover glucose was completely consumed. The overall volumetric productivity and substrate conversion will be increased if the

fermentation is running with column and tank in series, which may make the whole

fermentation system more economically favorable.

Fiber-packed reactors have been developed elsewhere for biofuel production. Yang

designed a fed-batch fermentation system for n-butanol production with a fibrous bed

bioreactor for immobilized-cell fermentation and a flask with temperature and pH control

[23]. Cells were immobilized in a glass column (50 mm diameter, 300 mm length, 250 ml

Texas Tech University, Lan Ma, December 2014

working volume) packed with spiral wound cotton towel and stainless steel wire cloth. In

their system, cells were cultivated in the flask firstly, and then re-circulated between

column and flask until the cell density in the broth no longer decreased. In our system, the column reactor was used to both immobilize and cultivate cells. The suspended cells

continuously consumed the remaining glucose in the tank reactor. Based on Yang’s work,

we could also try to circulate the cells between tank and column reactor or add series of

tank reactor after column in order to increase the ethanol productivity in the future.

(a)

Texas Tech University, Lan Ma, December 2014

(b)

(c)

Figure 3.18. Fermentation with an ICR whose outlet is connected with a tank reactor. Glucose (a) and ethanol (b) were measured from samples taken from five ports of column reactor and the outlet of tank reactor. (c): A flow diagram of the reactor.

Texas Tech University, Lan Ma, December 2014

3.3.2.3 Stirred tank reactor with polyethylene terephthalate fibers scaffold

Results of chemostat fermentation of E. coli LY01 with dilution rate 0.4 h-1 are

shown in Fig. 2.6. At steady state, the conversion of 70 g/L glucose was only 78 % with

productivity of 1.069 ±0.07 g/L/h. With the same dilution rate, we conducted another run

with 2.5 g fibers inside the reactor to immobilize cells. Fibers were stabilized in the

middle of the reactor between two steel porous plates (Fig. 3.19). Agitation was provided by the cell-lift impeller, shown in Fig. 3.19, and was maintained at 200 rpm. The impeller could continuously recirculate medium and cells through the impeller shaft, discouraging accumulation of floc on the bottom of the tank. Above and below the plates were stirred

medium, which allows enough substrate and product diffusion between immobilized and

free cells. Fig. 3.20a showed the fermentation results of residual glucose and ethanol

concentrations; 2.5 g of fiber were used to immobilize cells. A large increase in glucose

conversion and productivity was achieved by using fibers inside the reactor (Fig. 3.20b).

The glucose was almost completely consumed in less than 2 days with ethanol

productivity of 1.354 ± 0.08 g∙L-1∙h-1. The fibers managed to immobilize 10.5 g of dry

cell mass. Cells were not found to lose their ethanologenicity during 10 days of

fermentation.

Texas Tech University, Lan Ma, December 2014

Figure 3.19. Reactor was running with fibers packed between two porous plates. Agitation was provided by the cell-impeller.

Texas Tech University, Lan Ma, December 2014

(a)

(b)

Figure 3.20. ICR fermetnation with 2.5 g immobilized fiber inside. Initial glucose, Luria broth and chloramphenicol concentration were 70 g/L, 25 g/L and 40 µg/ml. Dilution rate was 0.04 h-1. The first 7 h was operated as anaerobic batch fermentation. (a) Ethanol and residual glucose; (b) cell optical density.

Texas Tech University, Lan Ma, December 2014

The designed continuous-flow fiber-immobilized tank fermentation is very stable and

easy to operate. Compared with chemostat fermentation, both reaction rate and glucose

conversion were increased during the immobilized cell fermentation.

The reactor performance was further investigated by adjusting the fiber mass and dilution rate in order to get the highest ethanol productivity on 70 g/L glucose. These results are given in Fig. 3.21. At a higher dilution rate of 0.06 h-1, 70 % of available

glucose was consumed with a productivity of 1.59 ± 0.11 g∙L-1∙h-1. Glucose conversion

was increased from 70 % to 90 % by increasing the fiber mass to 3.5 g fiber. Achieved

ethanol productivity was 1.8 ± 0.05 g∙L-1∙h-1, which was twice of the productivity

observed in batch fermentation. Immobilized dry cell mass was increased from 11.3 g to

14.2 g.

Compared with fiber packed-bed column reactor, cell immobilized tank reactor has pH, nitrogen purge and stirring control, which provides a suitable anaerobic environment

for cell growth and adequate substrate supply to the immobilized cells. Stirred-tank

reactor with cell immobilized fibers is more efficient for long-term bio-ethanol production than the column reactor.

Texas Tech University, Lan Ma, December 2014

(a)

(b) Figure 3.21. ICR fermentation with different amount of fibers. Initial glucose, Luria broth and chloramphenicol concentrations were 70 g/L, 25 g/L and 40 µg/ml. Dilution rate was 0.06 h-1. The first 7 h was operated as anaerobic batch fermentation. (a) 2.5 g immobilized fiber; (b) 3.5 g immobilized fiber.

Texas Tech University, Lan Ma, December 2014

3.4 Discussion

Our study appears to be the first report of conversion of sugar to ethanol by LY01 in continuous culture, especially in continuous immobilized-cell culture. In immobilized-

cell tank reactor, greater than 90 % glucose conversion was achieved with a dilution rate of 0.06 h-1. Ethanol productivity was doubled when compared with batch and continuous fermentations. In the continuous fermentation, we observed no significant loss of

ethanologenicity on glucose within the timeframe of the trials. The minimum

requirements proposed by Dien et al. for an economically-attractive cellulosic technology

are 90 % ethanol yield, 40 g/L ethanol concentration in effluent, and a minimum

productivity of 1.0 g∙L-1∙h-1 [24]. If feed concentration were increased by 10 to 20 %, it is

possible that the requirements for commercial production might be met.

Glucose conversion and productivity were increased when PET fiber was used as an

immobilization scaffold. It was observed that fermentations in the column reactor

suffered from oscillations in glucose conversion and poor repeatability. The cause of

these oscillations, as well as the low glucose concentration required to prevent oscillation,

was not investigated in this study due to lack of repeatability. Evidence also exists that

having too much fiber in the reactor can lead to poor reactor performance. It is believed

that this is due to a combination of poor mixing and mass limitations. The column reactor

is able to produce a lot of biomass, but did not perform well when glucose feed was

above 20 g/L.

The low dilution rate involved in this study was chosen to increase the one-pass

utilization of glucose. As expected, increasing the dilution rate reduced the glucose

utilization but improved the productivity. It is therefore expected that an even higher

Texas Tech University, Lan Ma, December 2014

productivity can be achieved by raising the dilution rate from its current value. While not

investigated for technical reasons, it is believed that the current reactor scheme can be

improved by using several tank reactors in series, as is common in the chemical process

industry.

There are several directions for future work. Since LY01 is of particular interest for

liganocellulose-derived bioethanol production, the major direction of future work will be

into evaluating reactor performance with mixed substrates such as glucose-xylose blends.

Other directions include improving the fiber immobilization capabilities, investigating

reactor performance over longer time frames, and fine tuning reactor parameters to

achieve optimal performance.

3.5 Conclusion

Both synthetic porous polymer scaffolds （ SPPS） and PET fibers with

diameter of 40-55 μm were chosen as cell immobilized materials. PET fibers provided

more void space than SPPS in the column reactor, so the fibers were judged to be more

effective scaffolds for reducing pressure buildup. SPPS particle size and internal pore volume fraction were studied and we found no significant effects on fermentation results.

Compared with the packed-bed column reactor, the stirred, immobilized-cell tank reactor with PET fibers is more stable without pH control, diffusion and pressure buildup problems. Both glucose reaction rate and conversion were increased due to high cell density achieved in the fermentation. The immobilized tank reactor showed remarkable stability during 10 days fermentation with a productivity of 1.8 ± 0.05 g L-1 h-1, which is

twice that batch fermentation with 70 g/L glucose. The designed fiber packed-bed

column reactor achieved a productivity of 9.0±0.5 g L-1 h-1 at dilution rate of 0.84/h with

Texas Tech University, Lan Ma, December 2014

20 g/L glucose, representing an improvement of a factor of 10× compared to batch fermentation. However, it could not be run with industrially relevant sugar concentrations in the feed (70 g/L or higher). Therefore, fiber immobilized column reactor is only good for fermentation with low substrate concentrations. The immobilized-cell tank reactor is better equipped to provide a stable and efficient process for continuous bio-ethanol production with high sugar concentrations in the feed and high substrate conversion.

Texas Tech University, Lan Ma, December 2014

3.6 References

[1] Guidoboni, G. E. Continuous fermentation systems for alcohol production. Enzyme and Microbial Technology 1984, 6(5), 194-200.

[2] Verbelen, P. J.; De Schutter, D. P.; Delvaux, F.; Verstrepen, K. J.; Delvaux, F. R. Immobilized yeast cell systems for continuous fermentation applications. Biotechnology Letters 2006, 28 (19), 1515-1525.

[3] Inloes, D. S.; Taylor, D. P.; Cohen, S. N.; Michaels, A. S.; Robertson, C. R. Ethanol-production by Saccharomyces cerevisiae immobilized in hollow-fiber membrane bioreactors. Applied and Environmental Microbiology 1983, 46 (1), 264-278.

[4] Nigam, J. N. Continuous ethanol production from pineapple cannery waste using immobilized yeast cells. Journal of Biotechnology 2000, 80 (2), 189-193.

[5] Jamuna, R.; Ramakrishna, S. V. High-concentration ethanol-production using immobilized yeast-cells. Biomass & Bioenergy 1992, 3 (2), 117-119.

[6] Najafpour, G.; Younesi, H.; Syahidah Ku Ismail, K. Ethanol fermentation in an immobilized cell reactor using Saccharomyces cerevisiae. Bioresour Technol 2004, 92 (3), 251-260.

[7] Taherzadeh, M. J.; Millati, R.; Niklasson, C. Continuous cultivation of dilute-acid hydrolysates to ethanol by immobilized Saccharomyces cerevisiae. Applied Biochemistry and Biotechnology 2001, 95 (1), 45-57.

[8] Ramakrishna, S. V.; Prakasham, R. S. Microbial fermentations with immobilized cells. Current Science 1999, 77 (1), 87-100.

[9] Karel, S. F.; Libicki, S. B.; Robertson, C. R. The immobilization of whole cells- engineering principles. Chemical Engineering Science 1985, 40 (8), 1321-1354. . [10] Nagashima, M.; Azuma, M.; Noguchi, S.; Inuzuka, K.; Samejima, H. Large-scale preparation of calcium alginate immobilized yeast cells and its application to industrial ethanol production. Methods in Enzymology 1987, 136, 394-405.

[11] Ghorbani, F.; Younesi, H.; Sari, A. E.; Najafpour, G. Cane molasses fermentation for continuous ethanol production in an immobilized cells reactor by Saccharomyces cerevisiae. Renewable Energy 2011, 36 (2), 503-509.

[12] Scott, J. A.; O'Reilly, A. M. Use of a flexible sponge matrix to immobilize yeast for beer fermentation. Journal of the American Society of Brewing Chemists 1995, 53(2), 67-71.

Texas Tech University, Lan Ma, December 2014

[13] Shen, H. Y.; Moonjai, N.; Verstrepen, K. J.; Delvaux, F. R. Impact of attachment immobilization on yeast physiology and fermentation performance. Journal of the American Society of Brewing Chemists 2003, 61 (2), 79-87.

[14] Arcuri, E. J. Continuous ethanol production and cell growth in an immobilized cell bioreactor employing Zymomonas mobilis. Biotechnology and bioengineering 1982, 24 (3), 595-604.

[15] Gil, G. H.; Jones, W. J.; Tornabene, T. G. Continuous ethanol production in a 2- stage, immobilized suspended cell bioreactor. Enzyme and Microbial Technology 1991, 13 (5), 390-399.

[16] Oliveira, R. Understanding adhesion: A means for preventing fouling. Experimental Thermal and Fluid Science 1997, 14 (4), 316-322.

[17] Sitton, O. C.; Gaddy. J. L. Ethanol production in an immobilized-cell reactor. Biotechnology and Bioengineering 1980, 22 (8), 1735-1748.

[18] Piret, J. M.; Devens, D. A.; Cooney, C. L. Nutrient and metabolite gradients in mammalian-cell hollow fiber bioreactors. Canadian Journal of Chemical Engineering 1991, 69 (2), 421-428.

[19] Gikas, P.; Livingston, A. G. Use of ATP to characterize biomass viability in freely suspended and immobilized cell bioreactors. Biotechnology and Bioengineering 1993, 42 (11), 1337-1351.

[20] Zhou, B.; Martin, G. J. O.; Pamment, N. B. Increased phenotypic stability and ethanol tolerance of recombinant Escherichia coli KO11 when immobilized in continuous fluidized bed culture. Biotechnology and bioengineering 2008, 100 (4), 627-633.

[21] Chu, Y.-F.; Hsu, C.-H.; Soma, P. K.; Lo, Y. M. Immobilization of bioluminescent Escherichia coli cells using natural and artificial fibers treated with polyethyleneimine. Bioresource Technology 2009, 100 (13), 3167-3174.

[22] Lannuzzi, M. A.; Reber III, R.; Lentz, D. M.; Zhao, J.; Ma, L.; Hedden, R. C. USANS study of porosity and water content in sponge-like hydrogels. Polymer 2010, 51 (9), 2049-2056.

[23] Lu, C.; Zhao, J.; Yang, S.-T.; Wei, D. Fed-batch fermentation for n-butanol production from cassava bagasse hydrolysate in a fibrous bed bioreactor with continuous gas stripping. Bioresource Technology 2012, 104, 380-387.

[24] Dien, B. S.; Cotta, M. A.; Jeffries, T. W. Bacteria engineered for fuel ethanol production: current status. Applied Microbiology and Biotechnology 2003, 63 (3), 258-266.

100

Texas Tech University, Lan Ma, December 2014

Chapter 4

Kinetics Study of Ethanol Fermentation with Escherichia coli LY01

4.1 Batch fermentation kinetics

A mathematical kinetic model of the process is often required to help understand

and predict microbial behavior, optimize the fermentation process, and determine suitable

control strategies for a given fermentation. Fermentation is a complex process where the

cellular physiological activities and the bioprocess are influenced by the numerous

interactions between microbes and culture environment. In general, the system can be

described only by a set of first order ordinary differential equations when perfect mixture

is assumed inside reactor and the only independent variable is time [1]. However,

depending on the level of complexity of the models to be developed, a large number of

mechanisms should be considered in the development of a kinetic model, such as the

inhibition effect of substrates, products of cell growth and fermentation, and the response

of cellular biological function to the variations of environmental pH, temperature, shear stress, and other fermentation conditions.

The most general mathematical models, which are used to describe the behavior of different biological systems, are classified into unstructured, structured and segregated models. Unstructured models consider that all cells are homogeneous, describing the microbial growth without considering the internal structure of cells and the population

diversity [2, 3]. The Monod equation is a common unstructured model, which describes

cellular mass and is only dependent on total biomass concentration. When inhibition of

cell growth occurs, the Monod equation is formulated with inhibition of high substrate

101

Texas Tech University, Lan Ma, December 2014

and/or product concentration [4, 5]. In general, unstructured models are not good models

to describe the dynamic experiments where composition and biomass activity change [6].

Structured models consider an individual cell as a multicomponent system, describing microbial growth with more than one variable, and some basic mechanisms of cellular behavior are considered inside the model [7]. Even though the cell population may be heterogeneous in a growing culture, an unstructured model was widely used and proved to provide a very useful description of experimental reality of the cells [8]. However, the parameters appearing in the unstructured kinetic models, which seldom have a real physical meaning and vary with cellular species or for the same cells exposed to different experimental conditions, normally need to be interpreted more carefully [8].

4.1.1 General unstructured kinetic model

The simplest enzyme-catalyzed kinetics reactions are often described by Michaelis-

Menten kinetics, as expressed by the following equation:

vS[] υ = m (4.1) Km '+ [S]

-1 νm is the maximum ethanol production rate (h ). [S] is the substrate concentration (g/L).

-1 -1 Km ' is the Michaelis-Menten constant. υ is the rate of product formation (g L h ). νm

changes if more enzyme is added, but the addition of more substrate has no influence on

νm. In a fermentation system, cell growth rate is proportional to the biomass concentration;

there usually exists a saturation limit for growth rate on each substrate and cells

metabolize substrate to product even when they do not grow [6]. If enzyme

concentrations are replaced by the cell concentration, the kinetics of cell growth rate can

be expressed using an analogous expression of the Michaelis-Menten equation:

102

Texas Tech University, Lan Ma, December 2014

1 dX mmS =mg = (4.2) X dt Ks + S

-1 In eqn. 4.2, X is cell concentration (g/L). µg is specific cell growth rate (h ). mm is

-1 maximum specific growth rate (h ). Ks is saturation growth constant (g/L). Eqn. 4.2 is called the Monod kinetic equation. The Monod equation is derived based on a premise that a single enzyme system with Michaelis-Menten kinetics is responsible for uptake of substrate and the amount or activity of enzyme is low enough to limit cell growth rate.

Even though this premise is rarely correct, the Monod equation empirically fits a wide range of data satisfactorily.

Many other equations for the specific growth rate have been proposed [9].

Blackman equation: mmgm= , if S ≥2Ks

mm mg = S , if S<2Ks (4.3) 2Ks

−KS Tessier equation: mmgm=(1 − e ) (4.4)

m S n mm=m = + −−n 1 Moser equation: g n ms(1KS ) (4.5) KSs +

mmS Contois equation: mg = (4.6) KXsx + S

dX X Logistic equation: =kX (1 − ) (4.7) dt X ∞

In eqn. 4.3 to 4.7, n, k and K are constants. Ksx is a growth coefficient of the Contois

equation. X ∞ is carrying capacity (g/L).

The Moser equation is the most general form of these equations, and it reduces to the Monod equation when n=1. The Logistic model, proposed by Pierre Verhulst (1838),

103

Texas Tech University, Lan Ma, December 2014

has been widely used due to its “goodness to fit” [10]. It assumes the specific growth rate is limited by the maximum biomass concentration that can be achieved, but the equation neglects the substrate and product inhibition effects. Based on the various experimental environments, these equations sometimes give better fit to the experimental data than the

Monod equation [11].

4.1.2 Models with growth inhibitors

In some fermentation, microbial growth rate is inhibited and the metabolic activities of cells are reduced at high concentrations of substrate. The inhibition is due to multiple reasons such as the modification of the chemical potential of substrates, the alteration of cell membrane permeability or the variation of enzyme activity [12]. In the glucose fermentation to ethanol, the noncompetitive substrate inhibition is normally observed as the following equation [9]:

m S m = m g 2 (4.8) KsI++ SS/ K

KI is substrate inhibition term for cell growth (g/L). Similarly to product inhibition, cell membrane and constituents are influenced at high ethanol concentration. The equation is expressed as follows:

mmS P n mg =(1 − ) (4.9) KSsm+ P

P is product concentration (g/L). Pm is ethanol inhibition term for growth (g/L).

Cellular behavior is also influenced by the temperature, pH or mixture of substrates.

There is not a unique equation that counts for all the factors that will influence the cell

104

Texas Tech University, Lan Ma, December 2014

growth rate. The choice among the above equations is dependent on how to effectively represent the experimental results.

4.1.3 Kinetics model of ethanol production by E. coli LY01

The most popular model to describe the cell growth rate is the unstructured Monod equation [8]. Besides the Monod equation, the Logistic equation is also widely used for fitting the experimental cell mass concentration [13, 14]. An empirical kinetic model was developed to describe the ethanol production by E. coli KO11 in batch fermentation by

Olsson and Hahn-Hagerdal [15]. They used modified Monod kinetics equations to describe the ethanol production rate. A constant cell mass concentration was used in the model and it was dependent on the final cell mass concentration, which is similar to the

Logistic equation. Kinetic parameters were estimated by nonlinear, least-square fitting to the experimental substrate and ethanol concentration data.

An empirical kinetic model was developed to describe the batch fermentation of glucose by E. coli LY01 in our study. Cell growth was described by two equations:

Monod equation and the Logistic equation. Earlier batch fermentation results showed no substrate inhibition effects with the chosen glucose concentration. Therefore, only the product inhibition effect was considered in the model. The ethanol production rate, substrate consumption rate and cell growth rate were described by [15, 9]

dCppCs C n' =υmcC ( )(1− ) (4.10) dt Ksp+ C s P m '

dC11 dC dC s=−+() Pc (4.11) dt Yps// dt Yxs dt

dCm C C C c= max sc(1− p ) n (4.12) dt Kss+ C P m

105

Texas Tech University, Lan Ma, December 2014

Eqn.4.13 describes the kinetics of cell growth rate by the Monod equation, which is also presented with the Logistic equation in eqn.4.14.

dC C cc=k(1 − ) (4.13) dt Cc∞

The ethanol inhibition was described by the generalized nonlinear equation proposed by Levenspiel [16]. The rate of substrate consumption is related to the biomass and ethanol production rates. In equation 4.10 to 4.14, Ks and Pm are, respectively, the saturation constant and the ethanol inhibition term for cell growth. Parameters Pm and n in the modified Monod equation do not only include the inhibition effect of ethanol, but

also other metabolites such as lactate and acetate. Cc0 and Cc∞ represent the initial cell density and the stationary cell density (the carrying capacity). Ksp and Pm' are, respectively, the saturation constant and the ethanol inhibition term for ethanol

production. The factors Yps/ and Yxs/ correspond to the amount of ethanol and cells produced from a given amount of glucose, and they were calculated based on experimental data. n, n’, Pm, Pm' and m were obtained from literature [8, 15, 17]. The best fit values of parameters were evaluated using least-squares method by minimizing the sum of squared errors between the predicted and experimental data:

22 J=∑( Spe − S )( +∑ PP pe − ) (4.14)

Fig. 4.1 and 4.3 show the experimental results, as well as the modeled values of the glucose and ethanol concentrations from the batch fermentations. The model estimation gave an adequate fit to the experimental data for both ethanol and glucose concentrations.

The cell mass concentration predicted from the model was the viable cells, not the measured total cell density. In a growing culture, some cells were dormant or died due to

106

Texas Tech University, Lan Ma, December 2014

the errors in autosynthesis. Therefore, the growth of total cell population is sometimes

described by the sum of the viable cells growth and the cell death rate. In Fig. 4.1 and 4.2,

the cell growth rate was investigated with the Monod equation. Significant effects of the

initial substrate concentrations on the saturation constant have been found. Both the

saturation growth constant (Ks) and saturation ethanol production constant (Ksp) were

found to increase when the initial glucose concentration increased.

A good agreement between the model predictions and the experimental glucose

fermentation data were obtained for all the four different batch fermentations by the model with either the Monod equation or the Logistic equation. The ethanol concentrations also got a good fit except for the one with 73 g/L initial glucose, for which the predicted data were somewhat smaller than experimental data. When cell growth rate was described by the Logistic equation, the initial input of cell mass concentrations were around 0.33 g/L. Instead of fitting Ks and Ksp, the obtained parameters were Ksp , k and

Cc∞ ,which did not exhibit any trend with respect to the initial glucose concentration.

Table 4-1. List of parameters used in the batch fermentation model.

-1 -1 Parameter P (g/L) P ' (g/L) ν (h ) Y p/s n n' m m m mm (h ) Yxs/

value 68 49.2 1.5 1.67 0.51 0.35 0.146 0.52

107

Texas Tech University, Lan Ma, December 2014

(a)

(b)

Figure 4.1. Experimental and simulation results of the sugar and ethanol concentrations at different initial glucose concentrations ; lines ( -×-) correspond to model simulations while dots correspond to experimental data; (■) initial glucose concentration 10 g/L; (●) 36 g/L; (▲) 54 g/L; (▼)73 g/L. Cell growth rate was described by the Monod equation.

108

Texas Tech University, Lan Ma, December 2014

(a)

Figure 4.2. The effect of initial glucose concentrations on (a) the saturated growth constant (Ks (g/L)) and (b) the saturated ethanol production constant (Ksp (g/L)).

109

Texas Tech University, Lan Ma, December 2014

Figure 4.3. Experimental results and simulation of the sugar and ethanol concentrations at different initial glucose concentrations; lines ( -×-) correspond to model simulation while dots correspond to experimental data; (■) initial glucose concentration 10 g/L; (●) 36 g/L; (▲) 54 g/L; (▼)73 g/L. Cell growth rate was described by the Logistic equation.

110

Texas Tech University, Lan Ma, December 2014

4.2. Continuous fermentation kinetics

An empirical kinetic model was developed to describe the continuous fermentation

of glucose with E. coli LY01. First, the kinetics of the chemostat fermentation was described. The substrate and product inhibition terms were included in the model. Cell

growth was strongly affected by the substrate and product concentrations. Cell growth

was separately described by two equations: the Monod equation and the logistic equation.

Second, the kinetics for the cell immobilized continuous fermentation was modeled.

For the chemostat fermentation, the rates of cell growth, ethanol production and

substrate consumption were related to the dilution rate (D), cell concentration (Cc), ethanol concentration (Cp) and substrate concentration (Cs) as follows [20]:

Accumulation=In- Out+ Generation

dc Ethanol: p =−+DC r (4.15) dt pp

dC Substrate: s =DC −+ DC r (4.16) dt s0 ss

dC Cell: c =−+DC r (4.17) dt cg

In the preceding equations, rp, rs and rg represent for ethanol production rate, substrate

consumption rate and cell growth rate, respectively. Cs0 is initial substrate concentration.

If modified Monod equation was used [8, 15] to describe cell growth, rg was

expressed with both substrate and product inhibition term as follows:

mmaxCCsc Cp n rg = 2 (1− ) (4.18) C s Pm KCss++ Kss

In eqn. 4.18, Kss is substrate growth inhibition term (g/L). If cell growth was described by

the Logistic equation [9], then rg is expressed as follows:

111

Texas Tech University, Lan Ma, December 2014

Cc rkg =(1 − ) (4.19) Cc∞ and by integration of equation 4.19 yields equation 4.20,

Cekt C = c0 (4.20) c C 1−−c0 (1ekt ) Cc∞

The substrate and ethanol inhibition were described separately by an uncompetitive inhibition and the generalized nonlinear equation proposed by Levenspiel [16]. The rate of substrate consumption is related to the rates of production of biomass and ethanol, and includes the rate if substrate consumption due to metabolic activities.

Cs Cp n' rCp=υ mc( 2 )(1− ) (4.21) Cs Pm ' KCsp++ s Kssp

rrpg rs=−+−()mCcc (4.22) YYps// xs

In eqn. 4.21, Kssp is substrate production inhibition term (g/L).

The initial guesses of υm , Ksp and Kssp parameters were estimated from the experiment of batch fermentation. The other parameters were estimated by manual adjustment until a good visual fit with the chemostat experimental data was obtained. The best fit values of parameters were evaluated using the least-squares method by minimizing the sum of squared errors between the predicted and experimental data.

112

Texas Tech University, Lan Ma, December 2014

(a)

(b)

Figure 4.4. Fitting of the anaerobic chemostat fermentation with glucose feed concentration of 70 g/L. Dilution rate was 0.026 h-1. (a) Ethanol and residual glucose from experiment and calculation using Monod equation to describe cell growth; (b) Ethanol and residual glucose from experiment and calculation using the Logistic equation to describe cell growth. The experimental data are shown as dots, and the solid lines are predicted data.

113

Texas Tech University, Lan Ma, December 2014

Fig. 4.4. shows the fitted residual glucose and ethanol concentration profiles with the

experimental chemostat fermentation results. The kinetic parameters in the proposed

model were determined on the basis of experimental data (Table 4.2). The measured

optical cell density did not allow the analysis of the viable cell mass concentrations

accurately and frequently enough. However, only the viable cell mass was counted in the

model. The Monod equation (a) and the Logistic equation (b) were separately applied to

express the cell growth kinetics. In eqn. 4.20, the use of initial cell mass concentration

was known. The final cell mass concentration and the value Ksp, Kssp, m were obtained

from the model by using least square fits of experimental glucose and ethanol

concentrations. Fig. 4.4(b) shows a better fit with the Logistic equation than Fig. 4.4(a) with the Monod equation. When the Monod equation was used to describe cell growth, the predicted data did not agree with the experimental results very well for the first 50 hours, especially for glucose. At the beginning of fermentation, product inhibition for cell growth was not supposed to be significant, but the deviation between the data and the model predictions suggest that production inhibition may have already been a factor.

114

Texas Tech University, Lan Ma, December 2014

Table 4-2. Compilation of fitted kinetic parameters (continuous fermentation without fibers).

Parameter Monod equation The logistic equation

D=0.026h-1 D=0.04h-1

Ksp (g/L) 0.01 3.22 13.01

Kssp (g/L) 1.00 17.72 131.33

Ks (g/L) 0.456 ------

Kss (g/L) 65.0 ------

Cc∞ (g/L) --- 2.810 2.171

k --- 0.639 0.449

mc 0.003 0.013 0.009

From Chapter 3, we know glucose conversion decreased from 100 % (D=0.026 h-1) to about 78 % when dilution rate increased to 0.04 h-1. The same method was used to fit the experimental glucose and ethanol concentrations at D=0.04 h-1 (Fig. 4.5). The model with the Monod equation to describe the cell growth rate did not provide a good fit to the data.

However, the Logistic equation provided a better fit and yielded the parameters shown in

Table 4.2.

115

Texas Tech University, Lan Ma, December 2014

Figure 4.5 . Fitted and experimental results of the chemostat fermentation. Dilution rate was 0.04 h-1. Cell growth was described by the Logistic equation. The experimental data are shown as dots, and the solid lines are predicted data from eqns. 4.15-4.17 and eqns. 4.19-4.22. When there was 2.5 g immobilized fiber inside the reactor, glucose was completely consumed in less than 2 days with the same dilution rate. It is difficult to measure the viable cell mass concentration in the cell immobilized reactor. However, the viable cell mass may be predicted by the model above with some assumptions. The immobilized cells were assumed to have the same fermentation and growth ability as the free cells.

Also, the reaction was assumed to occur without diffusion limitation because of the very

small fiber volume fraction and rapid stirring. Therefore, all the kinetic parameters were

kept the same with the chemostat fermentation at D=0.04 h-1, except for the initial and final cell density. The final viable cell mass in the cell immobilized fermentation reached to 5.201 g/L, which is two times higher than 2.171 g/L (measured value) from the non-

116

Texas Tech University, Lan Ma, December 2014

immobilized cell fermentation. Based on the assumptions, the model is able to predict the viable cells from the immobilized cell reactor by fitting the glucose and ethanol concentrations. The measured immobilized cell mass is about 5 X higher than the suspended cell mass in the reactor as reported in Chapter 3, which includes some dead cells and some cells without producing any ethanol.

(b)

Figure 4.6. Predicted residual glucose and ethanol concentration profiles for ICR with D= 0.04 h-1 and an glucose feed concentration of 70 g/L. Cells were immobilized with 2.5 g PET fibers. The experimental data are shown as dots, and the solid lines are predicted data. 4.3. Mathematical modeling of cell immobilized fermentation with packed-bed column reactor at steady state

A mathematical model was developed to describe the continuous fermentation in a packed-bed column reactor. The inhibition by the substrate and product were present

117

Texas Tech University, Lan Ma, December 2014

when fermentation was running at high substrate concentration. The Monod kinetics

equation was modified to describe the dynamic process of ethanol fermentation by

engineered E. coli responding to changes in the environment conditions. Our column

reactor did not seem reasonable to be described by plug flow model due to the lack of

turbulent flow in the column. To account for the deviations from plug flow, the dispersion model was introduced in the calculation, which accounted for some back mixing in the axial direction due to different flow velocities and molecular and turbulent diffusion [18, 19]. The following assumptions were made to model the reactor: a. The flow rate of liquid is constant. b. Bed porosity is constant throughout the bed. d. No diffusion limitation exists between fibers (The volume of fiber is very small).

The basic mass balance equation for every species by referring to an elementary

section of reactor is:

Input - output -disappearance by reaction + generation= accumulation (4.23)

The above equation becomes for component glucose:

(in-out) bulk flow + (in-out)axial dispersion+ generation= accumulation.

The individual terms (in g glucose /time) are as follows:

entering by bulk flow=(g glucose/volume)(flow velocity)(cross-sectional area)

=Cs,l uS

leaving by bulk flow= Cs,l+∆l uS

dN dC entering by axial dispersion= ss= −()DS dt dl l

dN dC leaving by axial dispersion= ss= −()DS dt dl ll+D

118

Texas Tech University, Lan Ma, December 2014

generation= rVss= rS∆ l(1−ε )

In the above equations, S is the tube transverse cross sectional area (m2). D is the dispersion coefficient (m2/h). ∆l is the section length (m). u is the fluid velocity (m/h). V

3 is volume of finite column segment (m ). Ns is the dispersed substrate amount (g). ε is the

column volume void fraction.

The difference between this material balance and plug flow reactor is the inclusion

of two dispersion terms, because material was flowing through the differential section not

only by bulk flow but by dispersion [18][19] as well. At steady state, accumulation is

equal to zero. Entering all these term into eqn. 4.23, dividing by Sl∆ and taking limits as

∆l→0, eqn. 4.24 was obtained.

d2 C dC Dss− ur +−(1ε ) = 0 (4.24) dl2 dl s

. In dimensionless form where z=l/L and τ=L/u=V/υ , we can obtain []

D d2 C dC ss− +−τε(1 )r = 0 (4.25) Lu dz2 dz s

L is the reactor length (m). υ is volumetric flow rate (m3/h); The ethanol production rate

r p from glucose fermentation was described using the Monod kinetics including

substrate and product inhibition, which was expressed as same as eqn.4.21. The yield

factor Yps/ gives the amount of ethanol produced from a given amount of substrate. The substrate utilization rate is therefore related to the ethanol production rate by eqn. 4.26

rp rs=−− mCcc (4.26) YPS/

119

Texas Tech University, Lan Ma, December 2014

At steady state, substrate is used for ethanol production and cell maintenance. Average

cell dry mass is measured from each section of column after fermentation. An equation of

cell mass concentration is achieved with the reactor tube axial position by fitting the

experimental data. Initial estimates of the parametersν m , Ksp and Kssp were based on the

D fitted results of batch and chemostat fermentations. is called dimensionless uL

dispersion group, which is approaching zero when dispersion is negligible (as plug flow) and is approaching infinity when dispersion is large (as mixed flow). For a small extents

D of dispersion from plug flow, is smaller than 0.01, while it is greater than 0.01 for uL

D large deviations. Therefore, was assumed with different values in our reactor. uL

120

Texas Tech University, Lan Ma, December 2014

Table 4-3. List of parameters used in the column reactor model.

Parameters Quantity

S (m2) 4.91*10-4

3 -5 ν (m /h) 8.4*10

-1 υm (h ) 1.5

V ( m3) 1.1*10-4

ε 10 %

Immobilized dry cell mass was measured as described in Chapter 3, then converted to the cell concentration in each segment of the column. Cell concentration was fitted versus column axial direction(Z) using the non-linear least squares fitting routine in

WaveMetrics IgorPro6 (Fig. 4.7). The equations of cell mass concentration are described in eq. 4.27. All the immobilized cells were assumed to be viable.

When glucose feed concentration is 10 g/L, cell mass concentration is

XZ( )=+− 21.711 71.042*exp( 8.8675l ) (4.27a)

When glucose feed concentration is 20 g/L, cell mass concentration is

XZ( )=+− 30.941 75.275*exp( 13.074l ) (4.27b)

121

Texas Tech University, Lan Ma, December 2014

(a)

Figure 4.7. Fitting curve of the cell mass concentration versus column axial distance. The experimental data are shown in points. The fitted data are shown as a line. The influence of different initial glucose concentrations (10 g/L and 20 g/L) on the reactor performance was investigated. At steady state, cell mass concentration versus column axial direction was predicted with an equation by fitting the experimental data.

The dilution rate used in the model was 0.84/h, which was much higher than the specific growth rate of LY01 determined in batch cultures, as reported in Chapter 2. When

D glucose feed concentration is 10 g/L, and was assumed equal to 0.005, Fig. 4.8a uL shows that about 100 % conversion can be achieved only with fifth of the column. When initial glucose concentration goes up to 20 g/L, the glucose was completely consumed with 30 % of column (Fig. 4.8b).

122

Texas Tech University, Lan Ma, December 2014

(a)

(b)

Figure 4.8. Predicted glucose and ethanol concentrations of the column reactor with different glucose feed concentrations: (a) 10 g/L and (b) 20 g/L.

123

Texas Tech University, Lan Ma, December 2014

D With different assumptions of the dispersion number ( ), ethanol concentrations uL were calculated and compared with the same glucose feed concentration at 10 g/L (Fig.

D 4.9). Compared with the results from equal to 0.005, the ethanol production rate is uL

D higher when is equal to 0.001. Small extents of dispersion will increase the ethanol uL

D production rate. However, if was assumed with 0.03, which means large deviation uL

D existing inside the column ( >0.01), the ethanol concentration shows fluctuation uL during the beginning of the run.

D Figure 4.9. Predicted ethanol concentrations with different dispersion numbers ( ): uL 0.03, 0.005 and 0.001. Glucose feed concentration was 10 g/L for all the calculations.

124

Texas Tech University, Lan Ma, December 2014

The results from Fig. 4.8 were calculated with the assumption that 100 % of the immobilized cells were the viable cells. However, that is not right, because some of the immobilized cells were dead cells or some of the cells lost the ability to produce any ethanol during the run. Therefore, fermentations with different percentages of the viable cells were calculated (Fig. 4.10). Fig. 4.10 shows the glucose consumption rate decreases with the viable cell mass decreases. The glucose is not totaly consumed when there is only 10 % of the immobilized cells are the viable cells.

Figure 4.10. The residual glucose concentration inside the column reactor by assuming different percentages (10 %, 50 %, 70 % and 100 %) of the immobilized cells are the viable cells.

125

Texas Tech University, Lan Ma, December 2014

4.4 Dimensionless parameters estimation

For cell-immobilized column reactor with SPPS, several dimensionless parameters were estimated based on the following equations.

Reynolds number for fluid flow through a packed bed reactor:

ρVd Re = sp (4.28) µ(1−e )

3 In eqn. 4.28, ρ is density of the fluid (kg/m ), which is assumed as water. Vs is the superficial velocity (m/h). dp is the size of immobilized enzyme particle (m). µ is the dynamic viscosity of the fluid (kg/(m·h)) and ε is the void volume fraction.

Damköhler numbers (Da): maximum rate of reaction V ' Da = = m (4.29) maximum rate of external diffusionkSLb [ ]

2 ׳ Vm is the maximum reaction rate per unit of external surface area (kg/(m ·h)) .kL is is the liquid mass transfer coefficient (m/s). [Sb] is the substrate concentration in bulk solution

(kg/m3).

2/3 1/2 DAB U k L ≈× 0.6  ×  (4.30) ν 1/6 1/2 d p

2 DAB is mass diffusivity of the substrate in the liquid phase (m /h). ν is the kinematic viscosity (m2/h). U is the free-system liquid velocity (velocity of the fluid flowing past the particle) (m/h).

Schmidt number is described by eqn. 4.31:

u viscous diffusion rate Sc = = (4.31) D molecular(mass) diffusion rate

Sherwood number is described by eqn. 4.31:

126

Texas Tech University, Lan Ma, December 2014

= (4.32)

convective mass transfer coefficient According to Froessling𝑆𝑆ℎ equation,diffusive eqn. mass4.32 cantransfer be rewrittencoefficient with Reynolds and Schmidt numbers; for a sphere, it is expressed as:

/ / = 2 + 0.6 (4.33) 1 2 1 3 Table 4-4. Constants used𝑆𝑆ℎ for calculating𝑅𝑅𝑒𝑒 𝑆𝑆 the𝑐𝑐 dimensionless parameters. Parameter number Parameter number 3 2 -6 ρ (kg/m ) 1000 DAB (m /h) 2.41×10 µ (kg/(m·h)) 2.872 U (m/h) 0.813 -3 2 dp (m) 1.75×10 ν (m /h) 0.0029 3 Vs (m/h) 0.184 [Sb] (kg/m ) 20 2 ׳ ε 0.55 Vm (kg/(m ·h)) 0.612

Table 4-5. Dimensionless parameters.

Dimensionless parameter number Re 0.2 Da 10.5

Reynolds number and Damköhler number can be estimated for our column reactor system. The estimated Reynolds number shows the fluid inside the reactor is a laminar flow. Since the Damköhler number is bigger than 1, the reaction inside the column reactor is external diffusion rate limited. The Schmidt number cannot be calculated due to the difficulty to obtain the substrate mass transfer rate at the boundary layer of the particle. However, when fluid velocity is high and particle size is very small, the resistance to external mass transfer can be negligible. It means the outside surface concentration is equal to the bulk concentration.

127

Texas Tech University, Lan Ma, December 2014

4.5 Conclusion

Kinetic models for batch fermentation, chemostat fermentation and cell-immobilized column reactor fermentation were derived. With the derived model, the viable cell mass can be predicted from the cell-immobilized fermentation.The experimental results show that the ICR system is most applicable to fermentation of dilute sugar mixtures (30 g/L or less). Even when dilution rate decreased to a value close to the specific growth rate, the

reactor still could not get high conversion with initial 50 g/L glucose. The model is

actually based on the final cell mass, which is dependent on the initial glucose

concentration. In the calculation, all the cells were assumed viable along the column

reactor. However, the experimental results showed that most glucose is consumed by the

cells at the bottom of the column. The reactor model here was constructed with the

assumption that mass transfer limitations were not present inside the column reactor.

However, mass transfer limitations were a limitation for fermentation experiments within

column reactor, which was described in Chapter 3. In addition, the model did not

consider the gas (CO2) effect on the fermentation. Therefore, a more complex model is

needed in order to describe the real fermentation condition in the cell-immobilized

column reactor.

Nomenclature

υ the rate of product formation (g/L·h) showing in the Michaelis-Menten equation

[S ] substrate concentration (g/L) showing in the Michaelis-Menten equation

Km ' Michaelis-Menten constant

-1 vm maximum ethanol production rate (h )

128

Texas Tech University, Lan Ma, December 2014

-1 µg specific cell growth rate (h )

-1 mm maximum specific growth rate (h )

Ks saturation growth constant (g/L)

Kss substrate growth inhibition term (g/L)

Ksp saturation production constant (g/L)

Kssp substrate production inhibition term (g/L)

Pm ethanol inhibition term for growth (g/L)

Pm ' ethanol inhibition term for production (g/L)

Cp ethanol concentration (g/L)

Cc , X cell concentration (g/L)

Cc0 initial cell concentration (g/L)

Cc∞ maximum cell concentration (the carrying capacity) (g/L)

Cs substrate concentration (g/L)

Cs0 initial substrate concentration (g/L)

rs rate of substrate utilization (g/L·h)

rp rate of ethanol production (g/L·h)

rg rate of cell growth (g/L·h)

Yps/ yield coefficient, g ethanol/g glucose (g/g)

Yxs/ yield coefficient g cell/g glucose (g/g)

-1 mc maintenance constant (h ) n, n', k exponents

129

Texas Tech University, Lan Ma, December 2014

D dispersion coefficient (m2/h)

l reactor length at any point (m)

L total reactor length (m)

V volume of finite column segment (m3)

Ns dispersed substrate amount (g)

Z dimensionless length of finite column segment u fluid velocity (m/h)

ε column volume void fraction

τ residence time (h)

ρ density of the fluid (kg/m3)

Vs superficial velocity (m/h)

dp size of immobilized enzyme particle (m)

µ dynamic viscosity of the fluid (kg/(m·h))

2 DAB mass diffusivity of the substrate in the liquid phase (m /h).

ν kinematic viscosity (m2/h)

U free-system liquid velocity (m/h)

Re Reynolds number

Da Damköhler numbers

Sc Schmidt number

Sh Sherwood number

130

Texas Tech University, Lan Ma, December 2014

4.6 References

[1] Paz Astudillo, I. C.; Cardona Alzate, C. A. Importance of stability study of continuous systems for ethanol production. Journal of Biotechnology 2011, 151 (1), 43-55.

[2] Sinclair, C. G.; Kristiansen, B. Fermentation kinetics and modelling. Bu’Lock, J. D., Ed.; Open University Press: New York, Taylor and Francis; 1987.

[3] Vazquez, J. A.; Murado, M. A. Unstructured mathematical model for biomass, lactic acid and bacteriocin production by lactic acid bacteria in batch fermentation. Journal of Chemical Technology and Biotechnology 2008, 83 (1), 91-96.

[4] Dadamo, P. D.; Rozich, A. F.; Gaudy, A. F. Analysis of growth data with inhibitory carbon sources. Biotechnology and Bioengineering 1984, 26 (4), 397- 402.

[5] Pinheiro, I. Q.; De Souza, M. B.; Lopes, C. E. The dynamic behaviour of aerated continuous flow stirred tank bioreactor. Mathematical and Computer Modelling 2004, 39 (4-5), 541-566.

[6] Nielsen, J.; Villadsen, J.; Liden, G. Bioreaction Engineering Principles, 2nd ed.; 2002.

[7] Lei, F.; Rotboll, M.; Jorgensen, S. B. A biochemically structured model for Saccharomyces cerevisiae. Journal of Biotechnology 2001, 88 (3), 205-221.

[8] Phisalaphong, M.; Srirattana, N.; Tanthapanichakoon, W. Mathematical modeling to investigate temperature effect on kinetic parameters of ethanol fermentation. Biochemical Engineering Journal 2006, 28 (1), 36-43.

[9] Michael, L.; Shuler, F. K. Bioprocess Engineering-Basic Concepts, 2nd ed.; 2010.

[10] Speers, R. A.; Rogers, P.; Smith, B. Non-linear modelling of industrial brewing fermentations. Journal of the Institute of Brewing 2003, 109 (3), 229-235.

[11] Wang, D.; Xu, Y.; Hu, J.; Zhao, G. Fermentation kinetics of different sugars by apple wine yeast Saccharomyces cerevisiae. Journal of the Institute of Brewing 2004, 110 (4), 340-346.

[12] Jones, R. P.; Pamment, N.; Greenfield, P. F. Alcohol fermentation by yeasts-the effect of environmental and other variables. Process Biochemistry 1981, 16 (3), 42-49.

[13] Dondo, R. G.; Marques, D. Optimal control of a batch bioreactor: A study on the use of an imperfect model. Process Biochemistry 2001, 37 (4), 379-385.

131

Texas Tech University, Lan Ma, December 2014

[14] Cacik, F.; Dondo, R. G.; Marques, D. Optimal control of a batch bioreactor for the production of xanthan gum. Computers & Chemical Engineering 2001, 25 (2- 3), 409-418.

[15] Olsson, L.; Hahnhagerdal, B.; Zacchi, G. Kinetics of ethanol production by recombinant Escherichia coli KO11. Biotechnology and Bioengineering 1995, 45 (4), 356-365.

[16] Levenspiel, O. The Monod equation: A revisit and a generalization to product inhibition situations. Biotechnology and Bioengineering 1980, 22 (8), 1671-1687.

[17] Li, X. Kinetic study and modelling of ethanol production from glucose/xylose mixtures using the genetically-engineered strain Escherichia coli KO11. Master Dissertation, Department of Chemical and Biological Engineering, University of Ottawa, 2008.

[18] Melick, M. R.; Karim, M. N.; Linden, J. C.; Dale, B. E.; Mihaltz, P. Mathematical modeling of ethanol production by immobilized Zymomonas mobilis in a packed bed fermenter. Biotechnology and Bioengineering 1987, 29 (3), 370-382;

[19] Levenspiel, O. Chemical Reaction Engineering, 3rd ed.; 1999.

[20] Fogler, H. S. Elements of Chemical Reaction Engineering, 4th ed.; 2005.

132

Texas Tech University, Lan Ma, December 2014

Chapter 5

Combinatorial Screening of Selective Materials for Extractive Fermentation

(Adapted from the paper in preparation “A combinatorial approach for tailoring selectivity and permeability in polymeric pervaporation membranes”)

5.1 Introduction

A pressing need has arisen to develop energy-efficient purification processes for renewable fuels in the emerging biofuels industry. Cost-efficient recovery of ethanol

(EtOH) and n-butanol (BuOH) from dilute aqueous fermentations poses a challenge, due to the low concentration of alcohols obtained [1-4]. Traditional distillation method only works economically favorable when alcohol concentration is above a certain threshold

(~70 g/L for EtOH) [5]. While EtOH concentrations of up to 140 g/L are achievable in

fermentations of pure glucose under ideal conditions, realistic EtOH concentrations

produced from biomass-derived, cellulosic feedstocks is usually below 5 wt. % due to the

presence of toxic inhibitory compounds in the feed [1-4,6,7]. Bio-BuOH concentration is limited to less than 20 g/L under ideal conditions (growth on glucose in rich media) by the toxicity of the BuOH, and the realistically achievable 10-12 g/L BuOH with most feedstocks is far below the threshold for economic feasibility [5]. Thus, design of alternative, energy-efficient processes for recovery of BuOH and EtOH from dilute fermentations is vital to the biofuels industry.

A variety of separation techniques have been studied due to high-energy requirements in distillation. As a membrane based separation technique, pervaporation is used to recover biofuels from fermentation broths [8, 9]. It is a process involving selective permeation of one liquid component through a membrane, followed by evaporation. The

133

Texas Tech University, Lan Ma, December 2014

component with high affinity and/or quicker diffusivity in the membrane is preferentially

removed from the mixture [10]. Pervaporation membranes fall into two categories, the

homogeneous membrane and the composite membrane, the latter normally offering a

higher flux than the homogeneous membrane [10]. One pervasive problem in membrane

design is that materials exhibiting the highest selectivity for the desired permeate usually

exhibit very low solubility, and therefore low transport rates are observed [5, 8, 11].

Optimizing the rate of permeation of the desired product is a non-trivial problem in

materials design not only due to this trade-off in properties, but because impurities (salts, proteins, soluble organic compounds) in the liquid feed complicate analysis of transport processes. Thus, much recent literature has been devoted to the design of pervaporation membrane materials, especially those based upon polymers [12].

A good separation material could allow running a continuous fermentation, reducing

the inhibitory effect of high ethanol concentration, and improving the productivity of

EtOH fermentation. However, the problem is further complicated by use of a diverse

array of biomass-derived feedstocks to produce bio-alcohols. These feedstocks (sugar

cane, corn, hydrolyzed cellulosic material derived from wood, or switch-grass) are not

pure substances, which makes the separation more complicated during cellulosic

fermentation. Furthermore, a more subtle problem is that one pervaporation membrane

material is unlikely to perform optimally in all alcohol recovery processes. The current

benchmark membrane used in the pervaporation is poly(dimethyl siloxane) (PDMS),

poly(1-trimethylsilyl-1-propyne) (PTMSP) or hydrophobic porous zeolite [8,13,14]. Most

of the reported membrane design is based on modifications of PDMS or other silicone

materials [15-46], including their blends with other polymers (e.g., polycarbonate [38],

134

Texas Tech University, Lan Ma, December 2014

poly(tetrafluoroethylene) [8, 38, 47], and poly(vinylidene fluoride) [23, 47] and their

composites with fillers such as silicalite [46-54], titania nanospheres [24], and carbon

nanotubes [55]. A wide variety of other polymers have been studied, such as cellulose

acetate [56,57], block copolymers [58-62], poly(vinylidene fluoride) [63], and

polyphosphazenes [64-66], but none of the alternative materials has challenged the

performance of PDMS-based membranes. Therefore, design of highly selective, highly

permeable polymeric membranes with tunable chemical composition and separation

characteristics and development of methodology for rapid optimization of their properties

could greatly benefit the growth of the young biofuels industry in the U.S. and abroad.

This paper describes a combinatorial technique developed by us to screen a high-

throughput network for ethanol-water separation. The work is conducted based on the theory of multi-component polymer-solvent systems. Networks comprised of random

copolymer with both hydrophobic and hydrophilic functionality are synthesized. A four- component (two monomer units, two solvents) extension of Flory-Rehner theory is developed for the first time to permit quantitative description of selectivity of random copolymer networks in water/alcohol mixtures. The interaction parameters (χij) required

for the same purpose are determined from simplified equilibrium swelling experiments.

Theoretically computed selectivities are compared with the values obtained by high

performance liquid chromatography (HPLC) technique. This work is not only to achieve materials that exhibit unparalleled separation efficiency, but also to provide the industry with methods to rapidly tune material performance to meet the evolving needs of biofuel recovery processes. In addition, the materials we designed are readily manufacturable and

135

Texas Tech University, Lan Ma, December 2014

thus could realistically be manufactured on-site and on-demand at a biorefinery (i.e., without requiring sophisticated synthetic techniques or processing equipment).

5.2 Experimental section

5.2.1 Network synthesis

Networks were synthesized by thermally initiated free radical polymerization of a

mixture of hydrophilic monomer 2-hydroxyethyl acrylate (H), hydrophobic monomer

butyl acrylate (B), cross-linker pentaerythritol tetraacrylate (C), and initiator 2,2'-

azobisisobutyronitrile (AIBN). No significant solvent was present during crosslinking.

Both H and B contained monomethyl ether hydroquinone as inhibitor, which was

removed by inhibitor remover (Sigma-Aldrich) prior mixing together by pipette. 2-

hydroxyethyl acrylate, butyl acrylate, pentaerythritol tetraacrylate and AIBN (98 %) were

obtained from Sigma-Aldrich. An 11×11 matrix networks were prepared using

combinatorial technique. The composition of material was defined by a two-dimensional

parameter space. The cross-linker concentration (or alternatively, the mole fraction of

reactive groups belonging to cross-linker, XC) and mole fraction of the hydrophilic

monomer (XH) were two independent dimensional parameters in our matrix. A typical

network was prepared by mixing the H, B, C with AIBN in a sealed glass bottle at

ambient temperature. The range of XC was from 0.0005 to 0.0105 by the increment of

0.001. XH was from 0 to 0.9895 by the increment of 0.1. The concentration of AIBN was

0.5 mass % of the total mixture in all cases and dissolved in monomer as a stock solution

before adding. All the samples had the same volume (1500μl) and were cured at 47 oC

overnight. As polymerization proceeded at 47 oC, no bubbles were observed in the

136

Texas Tech University, Lan Ma, December 2014

network. All the samples were prepared and crosslinked under the same conditions,

minimizing day-to-day variations in the investigator's lab technique.

5.2.2 Swelling experiments

The sample was immersed in large volume of ethanol and swollen at room

temperature until an equilibrium mass was reached. The ethanol was exchanged daily to

extract uncrosslinked component. The swollen mass of the gel was measured. Drying was

accomplished by placing the gel in air for 2 days, followed by drying under vacuum until

the mass of the extract sample reached an equilibrium value (Mex). The soluble fraction

(Wsol) of each sample was then determined by comparing the original, unextracted mass

(Munex) to Mex. The same method was used to measure the equilibrium-swelling ratio of

gel in water.

M ex Wsol =1 − (5.1) M unex

The mass equilibrium swelling ratio (Qse and Qsw) of ethanol and water were calculated

M se M sw according to Qse = and Qsw = M ex M ex

Where Mse and Msw are the total network mass at ethanol and water equilibrium swelling.

5.2.3 Selectivity

The samples were soaked in the 100 g/L EtOH-water mixture solution for several days to get equilibrium state. The mass of mixture solution is 1.5 times of the dry mass of network. After equilibrium state obtained, ethanol concentration of the leftover liquid was measured by HPLC technique and the swollen gel mass was also measured. If mass fractions of water and alcohol inside the membrane are ωwm and ωam, while the mass

137

Texas Tech University, Lan Ma, December 2014

fractions of water and alcohol in the external solution are ωws and ωas, then distribution

coefficient of ethanol and water, ka and kw are given by eqns. 5.2a and 5.2b.

ωam wwm ka = k w = (5.2a, 5.2b) ωas wws

An "equilibrium selectivity factor" aa*, is defined as the selectivity of the network for alcohol when swollen to thermodynamic equilibrium in the water/alcohol mixture.

w /wasam ka /φφ asam a a * = ≈= (5.3) w /wwswm kw /φφ wswm

In eqn. 5.3, the Φij are component volume fractions, and the stated approximation is

justified if volumes are additive. The value of aa* includes no information about the rates

of transport of water and alcohol through the network, but it is significant as readily

accessible experimental quantity.

The permeabilities of alcohol and water (Pa≡alcohol, Pw≡water) through the membrane

are given by eqns. 5.4a and 5.4b, respectively.

= kDP aaa = kDP www (5.4a, 5.4b)

Da and Dw are the diffusivities of alcohol and water in the membrane material,

respectively. The usual definition of permeability selectivity for alcohol (aa) of the

membrane, which includes both solubility and diffusivity information, is given by eqn.5.5.

Pa kD aa a a == (5.5) Pw kD ww In general, the objective for biofuels separations is to identify membrane materials

that exhibit the highest possible Pa, subject to the constraint that aa must remain above a

targeted minimum value. For example, it might be specified arbitrarily that the

alcohol/water mixture exiting the membrane can contain no more than 5 mol % water. In

138

Texas Tech University, Lan Ma, December 2014

practice, there is often a trade-off between solubility and selectivity, presenting a

complex materials optimization problem.

5.3 Theory and calculations (collaborated with Rutvik Godbole)

Basic research in polymer-solvent thermodynamics has traditionally focused on binary (polymer, solvent) or less commonly, ternary (polymer and 2 solvents) mixtures

[67-71]. The classical Flory-Rehner model [17] is a two-component (one polymer, one

solvent), lattice-based model that describes the equilibrium swelling of a network in

terms of a balance between the free energy change of mixing (DGM) and the elastic free

energy cost of deforming the network chains (DGE). Adjustable parameters are the

concentration of elastically effective chains (νe) and a polymer-solvent thermodynamic interaction parameter (χij). Okeowo and Dorgan [67] developed a three-component

extension of the Flory-Rehner model to describe swelling of a homopolymer (i) network

in a mixture of solvents (j and k) using three thermodynamic interaction parameters (χij,

χik, and χjk). However, the three-component model is insufficient to describe for

description of the swelling of random copolymer network in mixed solvents. The main

reason is local clustering of hydrophilic polymer segments, which leads to uneven

distributions of different solvent molecules in the swollen network, and an incorrect

calculation of DGM. The situation is exacerbated if reactivity ratios of different monomers

in copolymer are unequal, which is generally the case. Therefore a four-component (two polymers, two solvent) extension of Flory-Rehner theory is developed to capture the consequences of specific solvent-polymer interactions and quantitatively predict selectivities.

139

Texas Tech University, Lan Ma, December 2014

Derivation of a four-component Flory-Rehner extension begins with the Gibbs free energy of mixing of two solvents (1,2) with a polymer having two types of repeat units

(3,4) inside the membrane (subscript "m"), which is given by eqn. 5.6 [73].

∆G M lnln ln nnnn ln +++++= nn φχφχφφφφ kT m m m 44332211 m 2112 m 3113 m (5.6)

4114 m ++ 3223 m + nnn 4224 m + n φχφχφχφχ 4334 m

In eqn. 5.6, nj is the number of "j" molecules in the system, φj are their volume fractions, and subscript "m" refers to the membrane phase. Six interaction parameters (χij) exist for each 4-component system, with components defined as follows: (1=EtOH, 2=water,

3="B" monomer, 4="H" monomer). The elastic free energy change upon swelling of the membrane is given by eqn. 5.7.

∆G ν E e (α α α 222 3 −−++= lν()α α α ) (5.7) kT 2 zyx zyx

In eqn. 5.7, αp is the deformation factor in p-direction and νe is the effective number of chains between cross-links. The free energy of mixing of the two solvents in the outside solution (subscript "s") is given by eqn. 5.8.

∆G M nn lnln ++= n φχφφ (5.8) kT s s 21122211 s At equilibrium, the chemical potential of an alcohol molecule in the external solution

0 (compared to the reference state of a pure solvent μi ) is equal to the same for an alcohol molecule in the membrane phase, leading to eqn. 5.9.

m s ∆m1 = ∆m1 (5.9) Simplifying left hand side of eqn. 5.9 leads to

140

Texas Tech University, Lan Ma, December 2014

m ∆m1 ∂ ∆GM ∆GE  V1 =  +  ln 1m ()1 1m m +−−+= (m 1222 )(1−φχφφφφ 1m ) RT ∂n1  RT RT m V2 ,, nTP j  V  −−++  1  + χφχφφφχφχφ ( m m 144133 )(1 1m )  ( mm m 1441332 ) (5.10) V2 

 3/1 ()+φφ  + ()()V  φφρ −+ 43 mm  e  431 mm 2  3 In eqn.5.10, Vi is the molar volume of component i, ρe is elastic chain density (mol/cm ).

The above simplification is based on the following assumptions:

1. The network is amorphous and homogeneous and it swells isotropically i.e.

= =

𝑥𝑥 𝑦𝑦 𝑧𝑧 2. V𝛼𝛼 3 and𝛼𝛼 V4 are𝛼𝛼 very large compared to V1 and V2, because network is a single

molecule whereas there is large number of solute molecules.

Simplifying the right hand side of eqn. 5.9 leads to

s ∆m1 ∂ ∆GM  V1 2 =   ln 1s ()1 1s 2s +−−+= ( )(φχφφφ 212 m ) (5.11) RT ∂n1  RT s V2 ,, nTP j From eqn. 5.9, 5.10, and 5.11

V1 ln 1m ()1 1m m +−−+ (m 1222 )(1 1m ) ( m m 144133 )(1−++− φχφχφφχφφφφ 1m ) V2

 V   3/1 ()+φφ  −  1  ++ φφρχφχφφ −+ 43 mm  ( mm m 2442332 )()()eV  431 mm  (5.12) V2   2 

V1 2 ln 1s ()1 1s 2s +−−+= ( )(φχφφφ 212 s ) V2 Similarly, at equilibrium, the chemical potential of a water molecule in the external solution is equated to that of a water molecule in the membrane phase, leading to eqn.

5.13.

m s ∆m2 = ∆m2 (5.13) Which on further simplification as shown above yields eqn. 5.14.

141

Texas Tech University, Lan Ma, December 2014

V V  2 +−−+  2  ++− −φχφχφφχφφφφ ln 2m ()1 2m 1m  m 121 ()(1 2m m m 244233 )(1 2m ) V1  V1 

V   3/1 ()+ φφ  −  2  ++ φφρχφχφφ −+ 43 mm  ( mm m 1441331 )()()eV  432 mm  (5.14)  V1   2 

V2 V2 2 ln 2s ()1 2s 1s +−−+= (12 )(1−φχφφφ 2s ) V1 V1 The cross-linker used in the experiment is an octuple functional group; therefore

elastic chain density is four times of cross-linker density. χ13 and χ14 in eqn. 5.12 and 5.14 are obtained from equilibrium swelling experiment in pure ethanol, based on three components (2 monomers and a solvent) Flory-Rehner equations. Four components

Flory-Rehner equation is derived in a same manner as shown above, leading to eqn. 5.15.

' '  3/1 +φφ  ' ' ' ' ' +−++−+  ' φφρφχφχφφφ ' −+ ()3 4mm  = ln 1m ()(1 1m 3m 413 m 14 )(1 1 ) ()em V ()31 4mm  0 (5.15)  2  As pure solvent is used, there will not be any free energy of mixing for 2 solvents and

' ' thus right hand side of eqn. 5.15 is 0. φ3m and φ4m mentioned in eqn. 5.15 are volume

fractions of 2 polymers swollen in pure solvent are thus are different from the φ3m and

φ4m mentioned in eqn. 5.12 and 5.14. However, their ratio in both the scenarios is same

and is equal to their ratio in dry state (subscript "d"), leading to eqn. 5.16. The stated

approximation is justified if volumes are additive.

= = = (5.16) ′ ∅3𝑚𝑚 ∅3𝑚𝑚 ∅3𝑑𝑑 ′ ∅4𝑚𝑚 ∅4𝑚𝑚 ∅4𝑑𝑑 𝑇𝑇 φ3d and φ4d in eqn. 5.16 are volume fractions of both the monomers in dry state and T is

a known constant as both φ3d and φ4d are known. Mathematical modifications in eqn.

5.16 yields eqn. 5.17.

′ ′ + = ′ + (5.17) ∅4𝑚𝑚 ∅3𝑚𝑚 ∙ 𝜒𝜒13 ∅4𝑚𝑚 ∙ 𝜒𝜒14 ∅4𝑚𝑚 �∅3𝑚𝑚 ∙ 𝜒𝜒13 ∅4𝑚𝑚 ∙ 𝜒𝜒14� 142

Texas Tech University, Lan Ma, December 2014

' ' Furthermore, ()3 +φφ 4mm in eqn. 5.15 is calculated from measurement of equilibrium swelling ratio in pure ethanol (Qse) as shown in eqn. 5.18.

( + ) = ( ) (5.18) ′ ′ 1 Thus, from eqn. 5.15, 5.16,∅3𝑚𝑚 5.17∅ 4𝑚𝑚and 5.18,𝑄𝑄𝑠𝑠𝑠𝑠 we can get

( + ) =

3𝑚𝑚 13 4𝑚𝑚 14 ( ) ( + 1) ∅ ∙ 𝜒𝜒 ∅ ∙ 𝜒𝜒 (5.19) 1� 2 𝜈𝜈𝑒𝑒 1 3 1 𝑄𝑄𝑠𝑠𝑠𝑠−1 1 ∅4𝑚𝑚 ∙ �𝑄𝑄𝑠𝑠𝑠𝑠 ∙ 𝑇𝑇 � ∙ �𝑉𝑉𝑜𝑜 ∙ 𝑉𝑉1 ∙ ��𝑄𝑄𝑠𝑠𝑠𝑠� − �2∙𝑄𝑄𝑠𝑠𝑠𝑠�� − 𝑙𝑙𝑙𝑙 � 𝑄𝑄𝑠𝑠𝑠𝑠 � − 𝑄𝑄𝑠𝑠𝑠𝑠� as ( + + ) = 1 ′ ′ ′ 1𝑚𝑚 3𝑚𝑚 4𝑚𝑚 Similarly,∅ χ∅23 and χ∅24 in eqn. 5.12 and 5.14 are obtained from equilibrium swelling experiment in pure water as,

( + ) =

∅3𝑚𝑚 ∙ 𝜒𝜒23 ∅4𝑚𝑚 ∙ 𝜒𝜒24 ( ) ( + 1) (5.20) 1� 2 𝜈𝜈𝑒𝑒 1 3 1 𝑄𝑄𝑠𝑠𝑠𝑠−1 1 ∅4𝑚𝑚 ∙ �𝑄𝑄𝑠𝑠𝑠𝑠 ∙ 𝑇𝑇 � ∙ �𝑉𝑉𝑜𝑜 ∙ 𝑉𝑉2 ∙ ��𝑄𝑄𝑠𝑠𝑠𝑠� − �2∙𝑄𝑄𝑠𝑠𝑠𝑠�� − 𝑙𝑙𝑙𝑙 � 𝑄𝑄𝑠𝑠𝑠𝑠 � − 𝑄𝑄𝑠𝑠𝑠𝑠� is equilibrium swelling ratio in pure water.

𝑠𝑠𝑠𝑠 𝑄𝑄 Mulder and Smolders in 1984 calculated interaction parameter between water and

o ethanol ( ) at 25 C using excess free energy of mixing [72]. The parameter was

12 12 found to 𝜒𝜒be concentration dependent and a fourth grade polynomial function𝜒𝜒 was used to express as a function of volume fraction of ethanol. Thus,

12 𝜒𝜒 2 3 4 o χ12 e e e 89.031.315.435.198.0 ⋅+⋅−⋅+⋅−= φφφφ e at 25 C (5.21)

φ1m = φφ 1se in external solution and φe = in membrane phase. +φφ 21 mm b Chuang et al. in 2000 used a functional form of c12 a −= to fit χ12 calculated 1 c ⋅− φe at 32oC and 37oC from excess Gibb's energy data using group contribution method of

UNIFAC [74]. Thus,

143

Texas Tech University, Lan Ma, December 2014

191.0 o χ12 786.0 −= at 32 C (5.22) 706.01 ⋅− φe

197.0 o χ12 777.0 −= at 37 C (5.23) 704.01 ⋅− φe

φ1m = φφ 1se in external solution and φe = in membrane phase. χ12 is calculated from +φφ 21 mm

different expressions above mentioned. The values, however, are found to be very similar

with less than 10 % errors.

Thus, all the interaction parameters (χij) in eqn. 5.12 and 5.14 are explicitly known.

Solving eqn. 5.12, 5.14 and 5.16 simultaneously while considering the fact that +

1𝑚𝑚 + + = 1 , all the four unknowns (Φ1m, Φ2m, Φ3m, and Φ4m ) can∅ be

2𝑚𝑚 3𝑚𝑚 4𝑚𝑚 obtained.∅ ∅ ∅

5.4 Results and discussion

5.4.1 Equilibrium swelling

The network described in the current work is a random copolymer involve both

hydrophilic and hydrophobic polymer, so that the cohesive energy density of the

copolymer covers a wide range. Also, the glass transition temperature of poly (butyl

acrylate) and poly (hydroxyehtyl acrylate) is around -49 oC and -20 oC; both are below

room temperature, which is useful for enhancing swelling kinetics and are amenable to scale-up and commercial production.

The soluble or extractable fraction of the networks (Wsol) is readily obtained from the

sample mass before and after swelling in EtOH, which is a thermodynamically good

solvent for all of the materials. For the model system, Wsol is less than 0.01 for the

majority of samples and less than 0.02 for the weakest crosslinked samples. The Wsol in

144

Texas Tech University, Lan Ma, December 2014

EtOH/water mixtures will be truly negligible due to the poor solvent quality of water for

the more selective network.

Fig. 5.1 presents the equilibrium swelling behavior of network in pure EtOH and

water. The biggest EtOH swelling ratio is in the copolymer composition range near XH ~

0.3, Xc ~ 0.001 and the swelling ratio is gradually decreasing from the center of this range to the edge of this matrix. The figure shows clearly that the swelling ratio of

copolymer is higher than the respective homopolymer (XH =0 and 1.0), which means that

the copolymer has a greater affinity for EtOH than either of the corresponding homo-

polymers. And the overall cohesive energy density of copolymer is much more similar to

that of the targeted EtOH compared with homo-polymers. The network has XH = 0.3 and

XC ≈ 0.001, absorbed more than 200 % of its initial (dry) mass in pure EtOH (Qse >3),

while absorbing less than 2 % of its mass when swollen in pure water (Qsw < 1.02). In addition, the swelling maximum for water is located at XH=1, far away in compositional

parameter space, indicating that the chosen monomer pair has good potential to produce a

selective membrane.

The effects of cross-linker concentration are also readily apparent from Fig. 5.1.

Weakly crosslinked systems (low XC) that barely reach the gel point have the greatest

tendency to swell to a high degree, as expected. However, care must be taken to ensure

that XC is high enough such that the gel point is reliably reached during polymerization.

Network must not dissolve to any appreciable extent in EtOH/water mixtures.

145

Texas Tech University, Lan Ma, December 2014

(a)

(b)

Figure 5.1. Equilibrium swelling of copolymer networks ("B" monomer = butyl acrylate, "H" monomer= 2-hydroxyethylacrylate) in pure EtOH (a) and water (b). xc is the mole fraction of acrylates belonging to crosslinker groups and xH (=1- 0.25xc - xB) is the mole fraction of hydrophilic monomer.

146

Texas Tech University, Lan Ma, December 2014

5.4.2 Distribution coefficient and equilibrium selectivity

Samples were swelled in 100g/L EtOH-water mixture until equilibrium station is

reached. The composition of external liquid of each sample in the matrix is computed

from HPLC result. By measuring the total mass of water and alcohol absorbed, the

relative concentration and amount of alcohol and water, inside and outside the network,

can be calculated by performing a simple mass balance on the system. Excess volume of

mixing in water and ethanol mixture was considered in the calculation. Fig. 5.2a shows

the distribution coefficient of EtOH. The range of EtOH distribution coefficient is from

0.1 to 0.7. The homopolymer poly (hydroxyethyl acrylate) has the highest EtOH distribution coefficient and the coefficient is decreasing with decreasing XH.

(a)

147

Texas Tech University, Lan Ma, December 2014

(b)

Figure 5.2. Experimentally determined distribution coefficient of EtOH ka (a), and equilibrium selectivity aa* (b) for copolymer networks.

* Fig. 5.2b shows the experimental equilibrium selectivity (aa ) for the copolymer

networks in 100 g/L EtOH solution. The most selective networks are actually the ones

having higher crosslink density and are rich in the hydrophobic monomer (XH ~ 0).

Except two highest point in Fig. 5.2b, the relative highest selectivity is 15.3 for the

sample with Xc=0.0075, XH=0. The selectivity is decreasing by increasing the composition of hydrophilic monomer. Lowering the crosslink density tends to absorb some water during the EtOH absorption.

As is often the case, selectivity does not correlate well with the equilibrium swelling data in pure solvents. The maxima for selectivity (Fig. 5.2b) and distribution coefficient

(Fig. 5.2a) do not coincide. The property that ultimately must be optimized is

148

Texas Tech University, Lan Ma, December 2014

* permeability, which will be discussed below. In addition, certain criteria for aa must be

met, which depend on the economics of the dehydration scheme to be employed by the

biorefinery and should not be specified up front.

5.4.3 Predicted distribution coefficient and selectivity from the model

In our method, selectivity can be predicted by just measuring the equilibrium swelling

in pure solvent. In the four-component Flory-Rehner equation, 6 interaction parameters

exist for each network in the sample matrix; χ 12 (1=EtOH, 2=water) and χ 34 (3="B" monomer, 4="H" monomer) are calculated from known cohesive energy densities. The remaining 4 parameters ( χ 13, χ 14, χ 23, and χ 24) were separately combined with

volume fraction as a term X13 (ϕ3m·χ13+ ϕ4m·χ14) and X 23 (ϕ3m·χ23+ ϕ4m·χ24), which can be

determined experimentally from swelling measurements in pure solvents as explained in

theory and calculation section. Swelling measurement in an 11x11 matrix provided a rich

library of information that will permit values of Xij to be determined with excellent

* * accuracy. Knowledge of the Xij permits calculation of aa as in Fig. 5.3. aa is also computed using HPLC data to check the correctness of predicted values. Similar with experimental data, the model also predicts the most selective network is the one with higher cross-linker density and highest hydrophobic monomer. The highest predicted selectivity is 12.4 for the sample with Xc=0.0105, XH=0, compared with 15.3 that is

observed experimentally. The success of the derived model means that the procedure of measuring selectivity by high-throughput HPLC analysis could eventually be eliminated and the membrane screening can be done by simulation based on our methodology and theory.

149

Texas Tech University, Lan Ma, December 2014

(a)

(b)

Figure 5.3. Predicted distribution coefficient (a) and selectivity (b) for ethanol.

150

Texas Tech University, Lan Ma, December 2014

5.4.4 Diffusivities and Permeability Selectivity

The final stage in the screening is the measurement of pure component diffusivities in

the membrane samples, which permits computation of permeability selectivities (αa) according to eqn. 5.5. The measurement of diffusivities is more time-consuming than the synthesis, swelling, and HPLC experiments. Thus, a reduced number of materials from the sample matrix were studied, and a double-interpolation procedure was employed to estimate the diffusivities in the remaining materials. An alternative approach that may prove preferable in some situations is to obtain component diffusivities by computational techniques. Computational methods can potentially account for the fact that the diffusivity of each component might be affected by the presence of the other. In this study, we assume the diffusivity of EtOH in a membrane is independent of the presence of some water to a first approximation, and vice versa.

Diffusion of water and EtOH in the membrane samples was assumed to obey Fick's second law and diffusion into both top and bottom surfaces was allowed to occur simultaneously. Diffusion into the sides was neglected. Crank has provided analytical solutions to Fick's second law for diffusion in flat polymer slabs [75, 16]. If Mt is the

mass of the slab at time ti, M0 is its initial (dry) mass, and M∞ is its equilibrium swollen

mass, then for (Mt-M0)/(M∞-M0)<<0.5, the sorption data follow eqn. 5.24.

MM− D t 0 ≈ 1/2 1/2 2(2 ) t (5.24) MM∞ − 0 π L

In eqn. 5.24, D is the diffusivity of the solvent in the membrane and 2L is the full

thickness of the polymer slab. Measured diffusivities for nine membrane samples are shown in Table 5.1.

151

Texas Tech University, Lan Ma, December 2014

Table 5-1. Measured diffusivities of EtOH and water for selected membranes (boldface). The remaining diffusivities were estimated by double interpolation

-6 2 Da, Ethanol Diffusivity×10 (cm /s) 0.0105 7.511 8.309 9.106 9.904 9.827 9.749 9.671 9.594 7.745 5.896 4.048 ) ) C 0.0095 7.652 8.285 8.919 9.553 9.541 9.530 9.518 9.507 7.832 6.157 4.483 0.0085 7.793 8.262 8.732 9.201 9.256 9.311 9.366 9.420 7.919 6.418 4.917 0.0075 7.934 8.239 8.544 8.849 8.970 9.092 9.213 9.334 8.007 6.679 5.352 0.0065 8.075 8.216 8.357 8.498 8.685 8.872 9.060 9.247 8.094 6.940 5.787 0.0055 8.216 8.193 8.169 8.146 8.400 8.653 8.907 9.161 8.181 7.201 6.221 0.0045 8.126 7.868 7.609 7.351 7.546 7.742 7.937 8.132 7.538 6.945 6.351 0.0035 8.036 7.543 7.049 6.556 6.693 6.830 6.967 7.104 6.896 6.688 6.480 0.0025 7.947 7.218 6.489 5.761 5.839 5.918 5.997 6.075 6.253 6.432 6.610 Mole Fractionof Acrylates

Belonging to Crosslinker (x Crosslinker to Belonging 0.0015 7.857 6.893 5.929 4.966 4.986 5.006 5.027 5.047 5.611 6.175 6.739 0.0005 7.767 6.568 5.369 4.171 4.133 4.094 4.056 4.018 4.969 5.919 6.869 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.9895 Mole Fraction of Acrylates Belonging to Hydrophilic Monomer (xH)

-5 2 Dw, Water Diffusivity×10 (cm /s) 0.0105 8.780 7.902 7.024 6.146 5.316 4.487 3.657 2.827 2.796 2.764 2.733 ) ) C 0.0095 8.645 7.846 7.047 6.249 5.392 4.536 3.679 2.823 2.772 2.721 2.669 0.0085 8.509 7.790 7.070 6.351 5.468 4.585 3.702 2.819 2.748 2.677 2.606 0.0075 8.374 7.734 7.093 6.453 5.544 4.634 3.724 2.815 2.724 2.633 2.542 0.0065 8.238 7.677 7.117 6.556 5.619 4.683 3.747 2.810 2.700 2.590 2.479 0.0055 8.103 7.621 7.140 6.658 5.695 4.732 3.769 2.806 2.676 2.546 2.416 0.0045 8.125 7.366 6.608 5.850 5.069 4.287 3.506 2.725 2.563 2.400 2.238 0.0035 8.146 7.111 6.076 5.041 4.442 3.842 3.243 2.644 2.449 2.255 2.060 0.0025 8.168 6.856 5.545 4.233 3.815 3.398 2.980 2.562 2.336 2.109 1.883 Mole Fractionof Acrylates

Belonging to Crosslinker (x Crosslinker to Belonging 0.0015 8.190 6.601 5.013 3.424 3.188 2.953 2.717 2.481 2.222 1.964 1.705 0.0005 8.212 6.347 4.481 2.616 2.562 2.508 2.454 2.400 2.109 1.818 1.527 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.9895 Mole Fraction of Acrylates Belonging to Hydrophilic Monomer (xH)

With diffusivities of water and EtOH known in each membrane, permeability selectivities can be computed by eqn. 5.5. Fig. 5.4 shows the calculated values of aa for all samples. Because the ratio of Da/Dw does not vary drastically among the samples, the highest values of aa are found for essentially the same samples that had the highest equilibrium aa*. In this particular system, the permeability selectivity was decided mostly by the solubility of the liquid components in the copolymers, which is not necessarily always the case.

152

Texas Tech University, Lan Ma, December 2014

Figure 5.4. Permeability selectivity for EtOH (αa) .

5.4.5 Optimum Material Selection

The permeability selectivity of the material alone does not provide a basis to select

an optimum material. The material selection algorithm is more complicated, as it is based

upon process economics that depend on specific design considerations and product purity

requirements. However, the data provided by the screening approach are sufficient for a

process engineer to select a material that performs appropriately for a given process. A simplified example is presented here to illustrate the utility of the library of information gained by the screening approach.

To demonstrate the material selection process, we first assume that the membrane

must meet some minimum value of aa, which is determined by other process considerations. For example, a high value of aa (e.g., 30) may be targeted for a process in

which the permeant is to be dehydrated directly by addition of a dessicant. On the other

153

Texas Tech University, Lan Ma, December 2014

hand, a lower value of aa (e.g., 1.0) may be acceptable if the permeant is to be enriched

by pervaporation prior to feeding to a multi-stage distillation column. For example,

increasing the concentration of n-butanol from a typical value of 12 g/L to a value of 36

g/L by pervaporation would result in a significant reduction in energy consumption in

multi-stage distillation (see Fig. 3 in Ref. 5). For illustrative purposes, we assume a value

of aa = 1.0 is acceptable for the system under study.

Assuming the targeted value of aa is met, the optimal material will then be the

composition that provides the highest flux of EtOH at steady state. Increasing selectivity

beyond the acceptable value may be counter-productive, as highly selective membranes

typically absorb less alcohol, and thus provide a lower flux of alcohol. Fig. 5.5 then

shows that many of the material compositions can be eliminated immediately on the basis

of insufficient selectivity. Among the remaining samples, the steady-state flux of EtOH

is estimated from Fick's first law. The diffusivity of EtOH is taken as its measured value

in the absence of water, to a first approximation. The liquid phase boundary condition is

taken to be the equilibrium swelling condition for the membrane material, and the

concentration of alcohol in the liquid phase is assumed to be constant. The concentration

at the face through which EtOH emerges is maintained effectively at zero concentration.

The flux through the membrane is then given by eqn. 5.25.

dC JD= − m (5.25) dd J is the permeation flux of EtOH through the membrane; D is the diffusion coefficient of the EtOH in the membrane. The thickness of membrane (δ ) is 1 mm in eqn. 5.25. Fig.

5.5 shows the estimated flux of EtOH (g ethanol m2 h-1) through the membrane samples

that have aa > 1.0. The optimal material based on this hypothetical scenario is the

154

Texas Tech University, Lan Ma, December 2014

membrane having XC=0.0095 and XH= 0.3. This membrane is not the most selective among the compositions studied, nor is it the most highly swelling in mixed solvents or pure EtOH. It is, however, the material composition that provides the highest possible

flux of EtOH with aa > 1.0.

Figure 5.5. Flux of EtOH though the membranes with aa > 1.0. Membrane thickness is 1 mm.

5.5 Conclusion

A novel combinatorial methodology developed by our group is able to screen the

selective membrane and predict selectivity by simple equilibrium swelling measurement.

A 11×11 matrix of random copolymer networks involving different cross-linker density,

different hydrophobic and hydrophilic monomer composition are synthesized. The

network having XH = 0.3 and XC ≈ 0.001 absorbs more than 200 % of its initial (dry)

155

Texas Tech University, Lan Ma, December 2014

mass in pure EtOH, while absorbing less than 2 % of its mass when swollen in pure water.

The equilibrium swelling ratio of copolymer in EtOH-water mixture outperforms either

of homopolymer. However, the equilibrium selectivity factor shows that the networks

with higher volume fraction of hydrophobic monomer preferably absorbs EtOH molecule

rather than water molecule, therefore, maximum equilibrium selectivity for EtOH is

observed for network having XH = 0.0. The selectivity predicted by four component

Flory-Rehner model developed by our group is close to the experimental selectivity from

HPLC measurements. The optimal material is the membrane having XH = 0.3 and XC =

2 -1 0.0095 with a flux of 1.015 g m h and permeability selectivity (aa) > 1.0.

5.6 References

[1] Balat, M.; Balat, H.; Oz, C. Progress in bioethanol processing. Progress in Energy and Combustion Science 2008, 34 (5), 551-573.

[2] Gray, K. A.; Zhao, L. S.; Emptage, M. Bioethanol. Current Opinion in Chemical Biology 2006, 10 (2), 141-146.

[3] Gray, K. A. Cellulosic ethanol state of the technology. International Sugar Journal 2007, 109 (1299), 145-151.

[4] Lin, Y.; Tanaka, S. Ethanol fermentation from biomass resources: current state and prospects. Applied Microbiology and Biotechnology 2006, 69 (6), 627-642.

[5] Mariano, A. P.; Maciel Filho, R. Improvements in biobutanol fermentation and their impacts on distillation energy consumption and wastewater generation. Bioenergy Research 2012, 5 (2), 504-514.

[6] Dien, B. S.; Cotta, M. A.; Jeffries, T. W. Bacteria engineered for fuel ethanol production: current status. Applied Microbiology and Biotechnology 2003, 63 (3), 258-266.

[7] Palmqvist, E.; Hahn-Hagerdal, B. Fermentation of lignocellulosic hydrolysates. I: inhibition and detoxification. Bioresource Technology 2000, 74, 17-24.

156

Texas Tech University, Lan Ma, December 2014

[8] Peng, P.; Shi, B.; Lan, Y. A Review of membrane materials for ethanol recovery by pervaporation. Separation Science and Technology 2011, 46 (2), 234-246.

[9] Vane, L. M. A review of pervaporation for product recovery from biomass fermentation processes. Journal of Chemical Technology & Biotechnology 2005, 80 (6), 603-629.

[10] Shao, P.; Huang, R. Y. M. Polymeric membrane pervaporation. Journal of Membrane Science 2007, 287 (2), 162-179.

[11] Shao, P.; Huang, R. Y. M. Polymeric membrane pervaporation. Journal of Membrane Science 2007, 287 (2), 162-179.

[12] Campillo-Ferndez, A. J.; Salmerchez, M.; Serra, R.; Meseguer Dues, J. M. a.; Monle Pradas, M.; Gez Ribelles, J. L. Water-induced (nano) organization in poly(ethyl acrylate-co-hydroxyethyl acrylate) networks. European Polymer Journal 2008, 44 (7), 1996-2004.

[13] Takegami, S.; Yamada, H.; Tsujii, S. Pervaporation of ethanol/water mixtures using novel hydrophobic membranes containing polydimethylsiloxane. Journal of Membrane Science 1992, 75 (1), 93-105.

[14] Jin, T.; Ma, Y.; Matsuda, W.; Masuda, Y.; Nakajima, M.; Ninomiya, K.; Hiraoka, T.; Daiko, Y.; Yazawa, T. Ethanol separation from ethanol aqueous solution by pervaporation using hydrophobic mesoporous silica membranes. Journal of the Ceramic Society of Japan 2011, 119 (1391), 549-556.

[15] Aroujalian, A.; Raisi, A. Pervaporation as a means of recovering ethanol from lignocellulosic bioconversions. Desalination 2009, 247 (1-3), 509-517.

[16] Chen, H.; Lentz, D. M.; Rhoades, A. M.; Pyles, R. A.; Haider, K. W.; Vanapalli, S. A.; Nunley, R. K.; Hedden, R. C. Surface infusion micropatterning of elastomeric substrates. Microfluidics and Nanofluidics 2012, 12 (1-4), 451-464.

[17] Flory, P. J.; Rehner, J. Statistical mechanics of cross-linked polymer networks II. swelling. The Journal of Chemical Physics 1943, 11 (11), 521-526.

[18] Han, X.; Wang, L.; Li, J.; Zhan, X.; Chen, J.; Yang, J. Separation of ethanol from ethanol/water mixtures by pervaporation with silicone rubber membranes: effect of silicone rubbers. Journal of Applied Polymer Science 2011, 119 (6), 3413-3421.

[19] Jeon, E. J.; Kim, A. S.; Lee, Y. T. Pervaporation of butanol/water mixtures using siloxane polymer/ceramic composite membranes. Desalination and Water Treatment 2012, 48 (1-3), 17-26.

[20] Lazarova, M.; Boesch, P.; Friedl, A. POMS membrane for selective separation of ethanol from dilute alcohol-aqueous solutions by pervaporation. Separation Science and Technology 2012, 47 (12), 1709-1714.

157

Texas Tech University, Lan Ma, December 2014

[21] Liu, W.; Li, B.; Cao, R.; Jiang, Z.; Yu, S.; Liu, G.; Wu, H. Enhanced pervaporation performance of poly (dimethyl siloxane) membrane by incorporating titania microspheres with high silver ion loading. Journal of Membrane Science 2011, 378 (1-2), 382-392.

[22] Lue, S. J.; Chien, C. F.; Mahesh, K. Pervaporative concentration of ethanol-water mixtures using heterogeneous polydimethylsiloxane (PDMS) mixed matrix membranes. Journal of Membrane Science 2011, 384 (1-2), 17-26.

[23] Okeowo, O.; Dorgan, J. R. Multicomponent swelling of polymer networks. Macromolecules 2006, 39 (23), 8193-8202.

[24] Yang, T. H.; Lue, S. J. UNIQUAC and UNIQUAC-HB models for the sorption behavior of ethanol/water mixtures in a cross-linked polydimethylsiloxane membrane. Journal of Membrane Science 2012, 415, 534-545.

[25] Anim-Mensah, A. R.; Mark, J. E.; Krantz, W. B. Use of solubility parameters for predicting the separation characteristics of poly(dimethylsiloxane) and siloxane- containing membranes. Science and Technology of Silicones and Silicone- Modified Materials 2007, 964, 203-219.

[26] Aroujalian, A.; Belkacemi, K.; Davids, S. J.; Turcotte, G.; Pouliot, Y. Effect of protein on flux and selectivity in pervaporation of ethanol from a dilute solution. Separation Science and Technology 2003, 38 (12-13), 3239-3247.

[27] Aroujalian, A.; Belkacemi, K.; Davids, S. J.; Turcotte, G.; Pouliot, Y. Effect of residual sugars in fermentation broth on pervaporation flux and selectivity for ethanol. Desalination 2006, 193 (1-3), 103-108.

[28] Bueso, L.; Osorio-Galindo, M.; Alcaina-Miranda, I.; Ribes-Greus, A. Swelling behavior of pervaporation membranes in ethanol-water mixtures. Journal of Applied Polymer Science 2000, 75 (11), 1424-1433.

[29] Chen, X.; Ping, Z. H.; Long, Y. C. Separation properties of alcohol-water mixture through silicalite-I-filled silicone rubber membranes by pervaporation. Journal of Applied Polymer Science 1998, 67 (4), 629-636.

[30] Fouad, E. A.; Feng, X. Pervaporative separation of n-butanol from dilute aqueous solutions using silicalite-filled poly(dimethyl siloxane) membranes. Journal of Membrane Science 2009, 339 (1-2), 120-125.

[31] Garcia, M.; Teresa Sanz, M.; Beltran, S. Separation by pervaporation of ethanol from aqueous solutions and effect of other components present in fermentation broths. Journal of Chemical Technology and Biotechnology 2009, 84 (12), 1873- 1882.

158

Texas Tech University, Lan Ma, December 2014

[32] Hofmann, D.; Fritz, L.; Paul, D. Molecular modelling of pervaporation separation of binary mixtures with polymeric membranes. Journal of Membrane Science 1998, 144 (1-2), 145-159.

[33] Jitesh, K.; Pangarkar, V. G.; Niranjan, K. Pervaporative stripping of acetone, butanol and ethanol to improve ABE fermentation. Bioseparation 2000, 9 (3), 145-154.

[34] Kang, Y. S.; Shin, E. M.; Jung, B.; Kim, J. J. Composite membranes of poly(1- trimethylsilyl-1-propyne) and poly(dimethyl siloxane) and their pervaporation properties for ethanol-water mixture. Journal of Applied Polymer Science 1994, 53 (3), 317-323.

[35] Kittur, A.; Choudhari, S.; Kariduraganavar, M. Preparation of zeolite- incorporated PDMS membranes for pervaporation separation of isopropanol- water mixtures. Composite Interfaces 2006, 13 (4-6), 507-521.

[36] Kittur, A. A.; Kariduraganavar, M. Y.; Kulkarni, S. S.; Aralaguppi, M. I. Preparation of zeolite-incorporated poly(dimethyl siloxane) membranes for the pervaporation separation of isopropyl alcohol/water mixtures. Journal of Applied Polymer Science 2005, 96 (4), 1377-1387.

[37] Li, S. Y.; Srivastava, R.; Parnas, R. S. Separation of 1-butanol by pervaporation using a novel tri-layer PDMS composite membrane. Journal of Membrane Science 2010, 363 (1-2), 287-294.

[38] Matsuyama, H.; Kariya, A.; Teramoto, M. Effect of Siloxane Chain Lengths of Monomers on Characteristics of Pervaporation Membranes Prepared by Plasma Polymerization. Journal of Applied Polymer Science 1994, 51 (4), 689-693.

[39] Ohshima, T.; Kogami, Y.; Miyata, T.; Uragami, T. Pervaporation characteristics of cross-linked poly(dimethylsiloxane) membranes for removal of various volatile organic compounds from water. Journal of Membrane Science 2005, 260 (1-2), 156-163.

[40] Olsson, J.; Tragardh, G.; Lipnizki, F. The influence of permeant and membrane properties on mass transfer in pervaporation of volatile organic compounds from dilute aqueous solutions. Separation Science and Technology 2002, 37 (6), 1199- 1223.

[41] Roizard, D.; Clement, R.; Lochon, P.; Kerres, J.; Eigenberger, G. Synthesis, characterization and transport properties of a new siloxane-phosphazene copolymer. Extraction of n-butanol from water by pervaporation. Journal of Membrane Science 1996, 113 (1), 151-160.

[42] Schmidt, S. L.; Myers, M. D.; Kelley, S. S.; McMillan, J. D.; Padukone, N. Evaluation of PTMSP membranes in achieving enhanced ethanol removal from

159

Texas Tech University, Lan Ma, December 2014

fermentations by pervaporation. Applied Biochemistry and Biotechnology 1997, 63 (5), 469-482.

[43] Shaban, H. I.; Ali, S. H.; Mathew, J. Alcohol recovery with pervaporation: effect of high 2-butanol concentration. Journal of Applied Polymer Science 2001, 82 (13), 3164-3171.

[44] Ulutan, S.; Nakagawa, T. Separability of ethanol and water mixtures through PTMSP-silica membranes in pervaporation. Journal of Membrane Science 1998, 143 (1-2), 275-284.

[45] Vane, L. M. A review of pervaporation for product recovery from biomass fermentation processes. Journal of Chemical Technology and Biotechnology 2005, 80 (6), 603-629.

[46] Yahaya, G. O. Separation of phenol from aqueous streams by a composite hollow fiber based pervaporation process using polydimethyl siloxane (PDMS)/ polyether-block-amide (PEBA) as two-layer membranes. Separation Science and Technology 2009, 44 (12), 2894-2914.

[47] Feng, X. S.; Huang, R. Y. M. Separation of isopropanol from water by pervaporation using silicone-based membranes. Journal of Membrane Science 1992, 74 (1-2), 171-181.

[48] Hickey, P. J.; Juricic, F. P.; Slater, C. S. The effect of process parameters on the pervaporation of alcohols through organophilic membranes. Separation Science and Technology 1992, 27 (7), 843-861.

[49] Qureshi, N.; Blaschek, H. P. Butanol recovery from model solution/fermentation broth by pervaporation: evaluation of membrane performance. Biomass & Bioenergy 1999, 17 (2), 175-184.

[50] Slater, C. S.; Hickey, P. J.; Juricic, F. P. Pervaporation of aqueous ethanol mixtures through poly(dimethyl siloxane) membranes. Separation Science and Technology 1990, 25 (9-10), 1063-1077.

[51] Uragami, T.; Morikawa, T. Permeation and separation characteristics of alcohol water mixtures through poly(dimethyl siloxane) membrane by pervaporation and evapomeation. Journal of Applied Polymer Science 1992, 44 (11), 2009-2018.

[52] Vane, L. M.; Namboodiri, V. V.; Bowen, T. C. Hydrophobic zeolite-silicone rubber mixed matrix membranes for ethanol-water separation: effect of zeolite and silicone component selection on pervaporation performance. Journal of Membrane Science 2008, 308 (1-2), 230-241.

160

Texas Tech University, Lan Ma, December 2014

[53] Zhan, X.; Li, J.; Huang, J.; Chen, C. Enhanced pervaporation performance of multi-layer PDMS/PVDF composite membrane for ethanol recovery from aqueous solution. Applied Biochemistry and Biotechnology 2010, 160 (2), 632- 642.

[54] Zhang, W. d.; Sun, W.; Yang, J.; Ren, Z. Q. The study on pervaporation behaviors of dilute organic solution through PDMS/PTFE composite membrane. Applied Biochemistry and Biotechnology 2010, 160 (1), 156-167.

[55] Huang, J. C.; Meagher, M. M. Pervaporative recovery of n-butanol from aqueous solutions and ABE fermentation broth using thin-film silicalite-filled silicone composite membranes. Journal of Membrane Science 2001, 192 (1-2), 231-242.

[56] Ikegami, T.; Morita, T.; Nakayama, S.; Negishi, H.; Kitamoto, D.; Sakaki, K.; Oumi, Y.; Sano, T.; Haraya, K.; Yanagishita, H. Processing of ethanol fermentation broths by Candida krusei to separate bioethanol by pervaporation using silicone rubber-coated silicalite membranes. Journal of Chemical Technology and Biotechnology 2009, 84 (8), 1172-1177.

[57] Ikegami, T.; Negishi, H.; Sakaki, K. Selective separation of n-butanol from aqueous solutions by pervaporation using silicone rubber-coated silicalite membranes. Journal of Chemical Technology and Biotechnology 2011, 86 (6), 845-851.

[58] Ikegami, T.; Negishi, H.; Yanase, H.; Sakaki, K.; Okamoto, M.; Koura, N.; Sano, T.; Haraya, K.; Yanagishita, H. Stabilized production of highly concentrated bioethanol from fermentation broths by Zymomonas mobilis by pervaporation using silicone rubber-coated silicalite membranes. Journal of Chemical Technology and Biotechnology 2007, 82 (8), 745-751.

[59] Kaewkannetra, P.; Chutinate, N.; Moonamart, S.; Kamsan, T.; Chiu, T. Separation of ethanol from ethanol-water mixture and fermented sweet sorghum juice using pervaporation membrane reactor. Desalination 2011, 271 (1-3), 88-91.

[60] Qureshi, N.; Meagher, M. M.; Hutkins, R. W. Recovery of butanol from model solutions and fermentation broth using a silicalite silicone membrane. Journal of Membrane Science 1999, 158 (1-2), 115-125.

[61] Qureshi, N.; Meagher, M. M.; Huang, J.; Hutkins, R. W. Acetone butanol ethanol (ABE) recovery by pervaporation using silicalite-silicone composite membrane from fed-batch reactor of Clostridium acetobutylicum. Journal of Membrane Science 2001, 187 (1-2), 93-102.

[62] Yen, H. W.; Chen, Z. H.; Yang, I. K. Use of the composite membrane of poly(ether-block-amide) and carbon nanotubes (CNTs) in a pervaporation system incorporated with fermentation for butanol production by Clostridium acetobutylicum. Bioresource Technology 2012, 109, 105-109.

161

Texas Tech University, Lan Ma, December 2014

[63] Li, L.; Xiao, Z. Y.; Tan, S. J.; Liang, P.; Zhang, Z. B. Composite PDMS membrane with high flux for the separation of organics from water by pervaporation. Journal of Membrane Science 2004, 243 (1-2), 177-187.

[64] Jha, A. K.; Chen, L.; Offeman, R. D.; Balsara, N. P. Effect of nanoscale morphology on selective ethanol transport through block copolymer membranes. Journal of Membrane Science 2011, 373 (1-2), 112-120.

[65] Jha, A. K.; Tsang, S. L.; Ozcam, A. E.; Offeman, R. D.; Balsara, N. P. Master curve captures the effect of domain morphology on ethanol pervaporation through block copolymer membranes. Journal of Membrane Science 2012, 401, 125-131.

[66] Liu, F. F.; Liu, L.; Feng, X. S. Separation of acetone-butanol-ethanol (ABE) from dilute aqueous solutions by pervaporation. Separation and Purification Technology 2005, 42 (3), 273-282.

[67] Okeowo, O.; Dorgan, J. R. Multicomponent swelling of polymer networks. Macromolecules 2006, 39 (23), 8193-8202.

[68] Flory, P. J.; Rehner, J. Statistical mechanics of cross‐linked polymer networks I. Rubberlike elasticity. The Journal of Chemical Physics 1943, 11, 512-521.

[69] Jonquieres, A.; Perrin, L.; Durand, A.; Arnold, S.; Lochon, P. Modelling of vapour sorption in polar materials: comparison of Flory-Huggins and related models with the ENSIC mechanistic approach. Journal of Membrane Science 1998, 147 (1), 59-71.

[70] Kargupta, K.; Datta, S.; Sanyal, S. K. Quantitative approach for the prediction of preferential sorption in the case of pervaporation of a physico-chemically similar binary mixture. Journal of Membrane Science 1997, 124 (2), 253-262.

[71] Mulder, M. H. V.; Franken, T.; Smolders, C. A. Preferential sorption versus preferential permeability in pervaporation. Journal of Membrane Science 1985, 22 (2), 155-173.

[72] Mulder, M. H. V.; Smolders, C. A. On the mechanism of separation of ethanol/water mixtures by pervaporation I. Calculations of concentration profiles. Journal of Membrane Science 1984, 17 (3), 289-307.

[73] Flory, P. J. Principles of Polymer Chemistry, 1st ed.; 1953.

[74] Chuang, W. Y.; Young, T. H.; Wang, D. M.; Luo, R. L.; Sun, Y. M. Swelling behavior of hydrophobic polymers in water/ethanol mixtures. Polymer 2000, 41 (23), 8339-8347.

[75] Crank, J. Diffusion in Polymers, 1st ed.; 1968.

162

Texas Tech University, Lan Ma, December 2014

Chapter 6

Effects of Carbon Black Nanoparticles on Two-Way Reversible Shape Memory in Crosslinked Polyethylene

(Adapted from the submitted paper “Effects of carbon black nanoparticles on two-way reversible shape memory in crosslinked polyethylene”)

6.1. Introduction

Shape memory polymers (SMPs) are smart materials that can recover from a temporary shape that is held after deformation to a permanent shape after applying an

external stimulus [1-12]. SMPs have great potential to be used in sensors, actuators, and

other devices. In contrast to the widely used shape memory alloys (SMAs) and shape

memory ceramics (SMCs), SMPs have advantages such as high extensibility (up to 500-

800 %), tunable elastic modulus, tunable response temperature, and low mass density.

Generally speaking, SMPs contain chemical and/or physical cross-linking that defines the

permanent shape, and they can undergo a reversible phase transition that triggers shape recovery.

In the past decade, many studies of SMPs focused on enhancing the mechanical and thermal properties or on achieving remote triggering [2, 4, 5, 13-22]. Other studies have

targeted SMPs with biocompatibility to be used in medical applications [23-25] and

SMPs with multiple SMEs to provide complex actuation [16, 20, 26-30]. Compared with

the well-documented one-way SMPs (1W-SMPs) that can only remember a high

temperature shape, relatively few two-way reversible SMPs (2W-SMPs) that can

remember both high and low temperature shapes have been reported [25, 31-36].

2W-SMPs could be very useful for achieving reversible actuation in devices such as

163

Texas Tech University, Lan Ma, December 2014

artificial muscles. Reversible actuation triggered by temperature, light, or electric fields

has already been demonstrated in several studies of oriented or "monodomain" liquid

crystalline elastomers, for example [37-41]. However, there is still a driving force for

design of new 2W-SMPs based on economically attractive and readily processable

polymeric materials such as polyolefins.

Recently, 2W-SMPs based on the chemically cross-linked, semicrystalline

poly(cyclooctene) (PCO) were reported, and their shape memory behavior was attributed

to stress-induced crystallization [31, 34]. The same strategy has been used to achieve

shape memory in some other semicrystalline polymers such as polyurethanes (PU),

poly(ε-caprolactone) (PCL), and poly(ethylene-co-vinyl acetate) (PEVA) [25, 32, 33, 35,

36, 42].

In our previous work, the preparation of 1W-SMPs from cross-linked polyethylene

(cPE) and carbon black (CB) nanoparticles was described, and their self-healing behavior was studied [22]. The cPE and cPE/CB nanocomposites showed both high strain fixity

ratio (Rf) and high strain recovery ratio (Rr). Crystallization-induced elongation was observed for all the SMPs prepared and the effect became less remarkable with increasing volume fraction of CB nanoparticles (vCB) [22].

This work concerns the 2W-SME in crosslinked polyethylene, in which the 1W-SME is well-known [43-45]. These new investigations focus on characterizing the 2W-SME in cPE and documenting the effects of added CB nanoparticles. Whereas studies of nanofillers in SMPs often aim at enhancing the modulus or toughness of the SMPs, the aim of this study is instead to examine the ability of CB nanoparticles to improve shape memory characteristics and alter the response temperature range. Using a series of

164

Texas Tech University, Lan Ma, December 2014

samples having systematic variations in CB loading, the effects of CB particle loading on

the actuation ratio (Ra) are reported, where Ra≡(ε2-ε1)*100 %, ε1 is the initial strain under

load above the crystal melting temperature, and ε2 is the strain following crystallization

under stress. In addition, the effects of CB nanoparticles on switch temperatures in

heating and cooling steps are documented, revealing a simple approach for lowering the

response temperature range of PE-based SMPs. The molecular and physical factors underlying the observed thermomechanical phenomena are elucidated through equilibrium swelling experiments, differential scanning calorimetry, and X-ray

diffraction.

6.2. Experimental section

6.2.1. Materials

Low density PE (type LD100BW) pellets with density of 0.923 g cm-3 and melt flow

index of 2.0 g (10 min)-1 were supplied by Sinopec Beijing Yanshan Company (China).

CB nanoparticles (Vulcan XC68) with diameter of 30-50 nm, iodine number of 68 mg g-1,

and dibutylphthalate number (DBP) of 123 mL (100 g)-1 were purchased from Cabot

Corporation (USA). 2,5-dimethyl-2,5-di-(tert-butylperoxy)-hexane (DHBP, 92 %) was

supplied by Acros Organics (Belgium).

6.2.2. Sample preparation

The preparation procedure is the same as in a previous work [22]. PE pellets and CB

nanoparticles were dried completely in a vacuum oven at 80 °C before use. PE pellets

were mixed with various volume fractions of CB nanoparticles in a Thermal Scientific

Haake MiniLab II mixer (Germany) at 140 °C and stirred at 60 rpm for 10 min. The

mixtures were extruded and cut into pieces before being soaked in DHBP in a

165

Texas Tech University, Lan Ma, December 2014

hermetically sealed glass flask at room temperature for 72 h. The soaked pieces were

subsequently molded by hot pressing at 200 °C and 20.0 MPa for 30 min to crosslink the

PE component. The free-radical mediated crosslinking process has been described

elsewhere [27, 46]. Finally, thin films with a thickness of 0.5-1.0 mm were obtained by

slow cooling in air.

6.2.3. Swelling experiments

Samples were immersed in large volume of 1,3-dichlorobenzene and stirred gently at

140 °C until an equilibrium mass was reached after about two days. During the swelling,

the solvent was replaced several times in order to remove the extractable fraction. The

extracted samples were dried completely in a vacuum oven at room temperature until

their masses were invariant with time. The equilibrium mass swelling ratio (Qs) was

calculated according to eqn. 6.1.

M Q = s (6.1) s M ex

In eqn. 6.1, Ms is the mass at equilibrium swelling and Mex is the dry mass of the extracted sample. The gel mass fraction of the network (fg) is defined by eqn.6.2.

M f = ex (6.2) g M unex

In eqn. 6.2, Munex is the dry mass of the sample prior to extraction.

6.2.4. Shape memory effect (SME) analysis

The cPE and cPE/CB samples having different volume fractions of CB nanoparticles

(vCB) were cut into rectangular strips with a width of ca. 3.0 mm and thickness of 0.5-1.0

mm. The shear storage moduli, G'(ω), were characterized in shear sandwich geometry on

a dynamic mechanical thermal analyzer (TA Instruments Q800 DMTA, USA).

166

Texas Tech University, Lan Ma, December 2014

Isothermal frequency scans from 0.1 Hz to 40 Hz were carried out with 0.5 % strain amplitude at 130 °C.

1W-SME and 2W-SME analyses were performed on the DMTA in tension mode. The

1W-SME was characterized using a four-step program. In the 1st step, the samples were

kept isothermally at 130.0 °C for 3 min, after which the nominal stress was increased

from 0 to 0.855 MPa at 0.171 MPa min-1. In the 2nd step, the samples were cooled down

to 45.0 °C at 3.0 K min-1 under a constant nominal stress of 0.855 MPa. In the 3rd step,

the stress was completely unloaded at 0.171MPa min-1 at 45.0 °C. In the 4th step, the

samples were heated to 130.0 °C at 3.0 K min-1 with no applied stress. For the 2W-SME

analysis, the difference from 1W-SME analysis is that the unloading step was skipped;

the third step (re-heating and strain recovery) was conducted at a constant nominal stress

of 0.855 MPa. Three consecutive cycles were run for both 1W-SME and 2W-SME

analyses and the average value and standard deviation of measured properties are

reported. The applied stress was also varied in order to judge its effect on the 2W-SME.

The relevant quantities to describe the 1W-SME are the strain fixity ratio (Rf) (Eqn. 6.3)

and the strain recovery ratio (Rr) (Eqn. 6.4) [11].

ε R 3 ×= %100 (6.3) f ε 2

− εε 43 Rr = × %100 (6.4) ε3

nd rd th In eqns. 6.3 and 6.4,ε2, ε3, and ε4 are the strains after the 2 , 3 , and 4 steps,

respectively. The relevant quantities to describe the 2W-SME are the actuation ratio (Ra)

’ (Eqn.6.5) and the strain recovery ratio (R r) (Eqn.6.6) [30].

R ()εε 12a ×−= %100 (6.5)

167

Texas Tech University, Lan Ma, December 2014

' − εε 32 R r = × %100 (6.6) − εε 12

st nd rd In eqns. 6.5 and 6.6, ε1, ε2, and ε3 are the measured strains after the 1 , 2 , and 3 steps

of the 2W-SME cycle, respectively.

6.2.5. Wide-angle X-ray diffraction (2D-WAXD) measurements

Samples for 2D -WAXD measurements were prepared from the nanocomposite

having vCB of 1.0 vol. %. 2D-WAXD measurements on the unmodified material were

performed on a Rigaku S/MAX 3000with a Cu (λ= 0.154 nm) rotating anode operated at

40 kV and 30 mA at room temperature. The 2D-WAXD data were collected by an image

plate detector (size: 150 mm × 150 mm, Fuji Co., Japan) with an exposure time of 20 min

per sample, with the X-ray beam normal to the original sample surface. The 2D-WAXD data were calibrated with a silicon powder standard (2θ = 28.44 at wavelength of 0.154

nm). For stretched samples, rectangular strips of the material with a width of ca. 3.0 mm,

a thickness of 0.5-1.0 mm, and a length of 5.0 mm were heated to 140 °C and kept for 2

min before they were stretched to a nominal stress of either σ=0.05 MPa or 0.46 MPa via

the Q800 DMTA with the same strain rate used in 2W-SME experiments. Each sample

was cooled down to room temperature under fixed stress and the material was removed

from the clamps. The 2D-WAXD measurements of 20 min duration were performed on a

Rigaku D/MAX Rapid II instrument using a graphite monochromator, 300 µm pinhole

collimator, and CuKa source (wavelength = 1.5418 Å) in transmission. The incident

X-ray beam was normal to the original sample surface.

6.3 Results

6.3.1 Thermal, mechanical, and shape memory properties

168

Texas Tech University, Lan Ma, December 2014

Fig. 6.1(a) illustrates the presence of the 2W-SME in the cPE/CB nanocomposite with vCB of 5.0 vol. %. The sample was initially subjected to a complete cycle of 1) increasing

temperature to 130 °C; 2) cooling under nominal stress of 0.855 MPa; 3) unloading at 45

°C; 4) heating to 130 °C without load. The cycle was then repeated, except the 2nd step

(cooling under load) was reversed by heating back to 130 °C to illustrate the existence of

the 2W-SME under tension (Fig. 6.1(b)). The sample remembers both its high-

temperature and its low-temperature shape at the nominal stress of 0.855 MPa. The

cooling-induced elongation (crystallization) and heating-induced contraction (crystal

melting) can be seen in each cycle. This cooling/heating cycle was repeated two

additional times to characterize the repeatability of the 2W-SME. The sample returns to

nearly the same elongation after each cooling step, with just a slight increase in the strain

ε2 in each cycle.

Complete loading/unloading cycles for the remaining samples are shown in the

Supporting Information. All samples exhibited a 2W-SME, including the cPE without

CB; the added CB was not necessary to observe a 2W-SME. The chosen amount of crosslinker (DHBP) was appropriate to impart a strong memory of the permanent shape, yet allowing the sample to remain readily crystallizable. The materials must not be crosslinked so densely as to significantly reduce their ability to crystallize during cooling, which provides the physical basis for the observed elongation. Another critical factor is the ability of the system to form highly oriented crystals, which is essential to achieve a high value of Ra. All samples exhibited the ability to elongate during cooling, but the

sample with vCB of 20.0 vol. % exhibited greatly decreased elongation, even when the

169

Texas Tech University, Lan Ma, December 2014

applied stress was increased to 3.5 MPa. Thus, an excess of CB nanoparticles can detrimentally impact the 2W-SME.

Figure 6.1. Illustration of the 2W-SME in the cPE/CB nanocomposite with vCB of 5.0 vol. %. (a) Complete loading and unloading cycle showing the 1W-SME. (b) Plot of the three consecutive 2W-SME cycles (cooling and heating under load). Note that ε3 is defined differently for 1W-SME and 2W-SME cycles.

170

Texas Tech University, Lan Ma, December 2014

The measured crystallinity from DSC (Xc) and the equilibrium mass swelling ratio (Qs)

in 1,3-dichlorobenzene at 140 °C for each sample are given in Fig. 6.2 and Table 6.1.

’ Measured values of Rf, Ra, Rr, and R r for all samples are also provided. Addition of CB

particles does decrease crystallinity, but only slightly; Xc decreases from 21.3 % (cPE) to

15.7 % (vCB = 20 vol. %) as CB is added. The swelling experiments are a qualitative indicator of the crosslink density of each cPE sample, including both chemical and

physical crosslinking effects. Judging by the trend in Qs, addition of CB particles does

increase the effective crosslink density (in the absence of crystals). Measurements of the

rubbery shear moduli of the samples above the crystal melting

temperature (140 °C) also support this assertion (Fig. 6.3).

CB Figure 6.2. Dependence of Rf, Rr, Ra, R'r, and Xc on v . Due to the logarithmic scale, the data for cPE are plotted at the position for vCB of 0.2 vol. % (dashed line).

171

Texas Tech University, Lan Ma, December 2014

Figure 6.3. Shear storage modulus (G', MPa) vs. frequency (ω, Hz) at 130 °C for cPE and cPE/CB nanocomposites with various vCB.

Table 6-1. Properties of cPE and cPE/CB nanocomposites with various vCB. vCB (vol %) 0 0.5 1.0 5.0 20.0

Rf (%) 93.9±0.3 92.7±0.2 93.1±0.3 92.2±0.3 86.8±1.7

Rr (%) 97.0±1.8 95.6±2.4 93.5±3.3 92.4±3.4 90.1±3.8

Ra (%) 40.9±0.5 66.6±0.4 62.6±1.8 41.6±0.3 0.9±0.2

’ R r (%) 99.3±0.3 93.5±2.6 94.7±1.9 91.6±0.7 31.1±1.6

Xc (%) 21.3 19.4 19.1 17.8 15.7 fg (%) 88.6 94.0 95.5 88.7 91.8

Qs 5.80 5.43 5.47 5.28 3.06

172

Texas Tech University, Lan Ma, December 2014

’ The effects of CB nanoparticles on Rf, Ra, Rr, and R r can be rationalized in terms of the

changes in crystallinity and effective crosslink density. The cPE sample has Rf=94 % and

Rr=97 %. Addition of CB nanoparticles causes only a slight decrease in the ratios Rf and

CB Rr for v up to 20 vol. %, a trend which mirrors the slight decrease in crystallinity. Any

decrease in crystallinity affects the ability of the elongated sample to hold its shape after

tension is removed, reducing Rf. The slight reduction in Rr indicates that the shape

’ recovery is less complete in the samples having added CB; creep is slightly enhanced. R r

CB CB ’ decreases gradually as v increases, except at v = 20 vol. %, where R r decreases dramatically to 31.1 %. This change is attributed to the disrupting effects of the concentrated CB nanoparticles, which encourage formation of less highly oriented crystals under tension. Xc measurements from DSC (Table 1) show that crystallinity is

’ only slightly decreased for this sample, so the dramatic drop in R r is inferred to be due to

a change in crystal orientation and/or morrphology.

Perhaps the most significant observation from Table 1 is the large increase in Ra,

when a small volume fraction of CB is present. The cPE sample exhibits Ra of 40 %, but

CB the samples having v of 0.5 % and 1.0 % each have Ra> 60 %, an increase of more than a factor of 1.5. A comparison of the heating/cooling cycles under load for each sample is useful to identify the reason for the improvement in Ra. Fig. 6.4(a) illustrates such a

comparison at a nominal stress of 0.855 MPa. It is evident that adding a small volume

CB fraction of CB (v =0.5 or 1.0 vol. %) has little effect on ε2, whereas ε1 decreases

substantially due to the increased effective crosslink density. The insensitivity of the

strain ε2 to the presence of CB indicates that the nanoparticles at low concentrations have

little effect on the ability of the sample to crystallize under tension, as supported by Xc

173

Texas Tech University, Lan Ma, December 2014

measurements. The elongation during cooling under load results from the formation of

oriented crystals, a process which, evidently, is essentially unaffected when vCB is ≤ 1.0

CB vol. %. In contrast, when v is increased to 5.0 vol. % or 20.0 vol. %, both ε2 and ε1

decrease substantially, producing a lower value of Ra. In fact, the 2W-SME is effectively eliminated when vCB is 20.0 vol. %. Thus, there exists an optimum loading of CB

CB nanoparticles that maximizes Ra, which is approximately v =0.5 vol. % for this system.

Another observation from Fig. 6.4(a) is that the CB particles affect the switch

h c h temperatures observed during heating and cooling (Tsw ,Tsw ). The Tsw (midpoint of the

c contraction during heating) is substantially higher than the Tsw (midpoint of the

elongation during cooling) in each case due to the slow kinetics of crystallization during

cooling. The CB particles do not act as nucleating agents, as their presence actually

c delays the onset of crystallization during cooling. The lowering of Tsw by CB leads to crystallization at a lower average temperature and formation of thermodynamically less

h c h stable crystals, which melt at a lower Tsw when re-heated. Both Tsw and Tsw decrease

with increasing νCB, providing a means to lower the response temperature range for shape-memory applications.

174

Texas Tech University, Lan Ma, December 2014

a Thick: cooling σ = 0.855 MPa Thin: hεating vCB (vol%) 0 100 0.5 1.0 5.0 80 20.0 (%) ε 60

Strain, 40

0 40 60 80 100 120 140 Tεmpεraturε, T (oC)

Exothermic up Thick: cooling 2.0 Thin: heating lPE )

-1 1.0

W g cPE 0.0 cPE/CB 0.5 vol%

cPE/CB 1.0 vol% -1.0 Heat flow, HF ( cPE/CB 5.0 vol%

-2.0 cPE/CB 20.0 vol%

40 60 80 100 120 140 Temperature, T (oC)

Fig.6.4. (a) 2W-SME cycle and (b) DSC traces showing crystallization during cooling and melting during subsequent heating of cPE and cPE/CB nanocomposites. In (b), non- crosslinked lPE is also shown as a reference. (This figure was done by Xiaoyan, Wang and Jun Zhao at University of Science and Technology)

175

Texas Tech University, Lan Ma, December 2014

The effects of CB addition on crystallinity and crystallization kinetics are further

evident from DSC analysis. Fig. 6.4(b) presents the DSC cooling and heating traces of

non-crosslinked lPE, cPE, and cPE/CB nanocomposites with various vCB. (Note that the

crystallization and melting temperatures obtained from DSC cannot be compared directly

c h to the Tsw and Tsw ranges in Fig. 6.4(a), which were determined for samples under

tension). Comparing lPE and cPE shows that the cross-linking causes a significant decrease in both the average crystallization temperature (Tc) and the average melting

temperature (Tm) in cooling/heating ramps at 5 °C/min. The Tc and Tm for lPE are about

95 °C and 108 °C, respectively, while they decrease to about 75 °C and 90 °C,

respectively, for cPE. The observed depression of both Tc and Tm indicates that the crystal

lamellae formed in the cross-linked samples are thermodynamically less stable due to the

presence of the non-crystallizable cross-link junctions. In addition, the appearance of a

double exotherm during cooling of cPE indicates that two populations of crystal lamellae

are formed at different average temperatures. This observation possibly indicates that

there is an inhomogeneous spatial distribution of crosslink junctions in the cPE, such that

the less densely crosslinked regions crystallize at higher average Tc during cooling.

From Fig. 6.4(b), the addition of CB nanoparticles causes a further decrease of

CB both Tc and Tm, indicating that crystal lamellae become progressively less stable as v

increases. The kinetics of crystallization are slowed by the disturbing effect of the CB

nanoparticles, meaning crystallization occurs at a lower average Tc during cooling. The

less stable crystal lamellae formed subsequently melt at a lower Tm when reheated. The

disturbing effect of the CB nanoparticles, which affects Tc and Tm, provides the basis for

lowering the switch temperatures in 2W-SME cycles.

176

Texas Tech University, Lan Ma, December 2014

Nominal

Figure 6.4. (a) Influence of applied stress on the 2W-SME for the cPE/CB nanocomposite with vCB of 1.0 vol. %. (b) Relationship between applied stress and observed elongation during cooling (ε2-ε1).

177

Texas Tech University, Lan Ma, December 2014

Another parameter which clearly affects the 2W-SME is the applied nominal stress (σ).

Fig. 6.5 illustrates the effect of σ on the 2W-SME cycle and the dependence of(ε2−ε1) on

σ for the cPE/CB nanocomposite with vCB of 1.0 vol. %. Fig. 6.5(a) shows that the 2W-

SME appears when σ is 0.46 MPa or 0.15 MPa (elongation during cooling and

contraction during heating). In contrast, the opposite deformation behavior (contraction

during cooling and elongation during heating) is seen whenσ=0.01 MPa. A transition

between these two different behaviors occurs nearσ=0.05 MPa. Thus, there is a critical

minimum stress (σc) that must be applied to observe a 2W-SME. Fig. 6.5(b) shows that

(ε2−ε1) increases approximately linearly with σ. By extrapolating the line to (ε2−ε1)=0, a

σc value of 0.055 MPa is obtained, which is consistent with the experimental

observations.

6.3.2. WAXD analysis

Wide-angle X-ray diffraction (WAXD) was applied to characterize crystal orientation

in the sample with vCB of 1.0 vol % in order to better understand the physical processes

responsible for the 2W-SME and the observed σc. WAXD patterns are shown in Fig.

6.6(a) for the unmodified film (before any deformations were applied) and in Fig. 6.6(b,c)

for the end of step 2 in the 2W-SME cycle at strain=ε2 (σ=0.05 MPa and 0.46 MPa, respectively). Intensity vs. azimuthal angle (χ, degrees) profiles are shown in Fig. 6.7.

Two strong diffraction peaks appearing at 2θ of 21.6 and 23.8 are assigned to the (110)

and (200) reflections of orthorhombic PE, respectively.[48] The incident X-ray beam

was oriented along the direction normal to the original film surface (ND).

178

Texas Tech University, Lan Ma, December 2014

Fig. 6.6(a) shows that before stretching, poles to (110) and (200) planes are randomly

oriented with respect to the ND. Therefore, there was a random orientation of chain axes

within the plane of the films as prepared. When the film was stretched, both (110) and

(200) reflections transitioned to two arcs enhanced at an azimuthal angle (χ) of

approximately 90 ° relative to the stretching direction (SD), as shown in Fig. 6.7 (b,c).

The observed orientation of the (110) and (200) poles indicates that the crystal c-axes

became preferentially oriented parallel to the SD during crystallization under stress.

Elongation of the samples during cooling under stress indicates the formation of highly

aligned crystal structures with chain axes aligned preferentially in the axis of elongation.

Crystallization of polyethylene under stress in the 2W-SME cycle most likely produces

row-nucleated or fibrillar morphology, based on the observed orientation of the c-axes.

179

Texas Tech University, Lan Ma, December 2014

Figure 6.5. 2D-WAXD patterns of the cPE/CB nanocomposites with vCB of 1.0 vol %. (a) unmodified film at σ=0 MPa. Film crystallized under stress at σ=0.05 MPa (b) and 0.46 MPa (c). (X-ray image was taken by Yongrui Liang at Beijing Institute of Petrochemical Technology)

(a)

180

Texas Tech University, Lan Ma, December 2014

(b)

(c)

Figure 6.6. Diffracted X-ray intensity vs. azimuthal angle (χ) for cPE/CB nanocomposites with vCB of 1.0 vol % before stretching (a) and after stretching with nominal stress σ=0.05 MPa (b) and σ=0.46 MPa (c).

6.4 Discussion

CB is included among many fillers that have been studied as modifiers for SMPs, such as graphene, carbon nanotubes, clays, carbon fibers, glass fibers, Kevlar fibers,

181

Texas Tech University, Lan Ma, December 2014

nano-SiC and nano-SiO2. The fillers are normally intended to improve the mechanical

properties, especially shape recovery stress. Improvements in mechanical response have

been observed with graphene-reinforced hyperbranched polyurethane, polyimide or

polyester/carbon nanofiber composites,[49-52], carbon fiber or glass fiber-reinforced

epoxy laminates,[53-55] carbon nanotube or carbon nanofiber-reinforced polyurethane

composites,[56-60] and carbon nanoclay-tethered shape memory polyurethane

nanocomposites.[61] Another common approach is to realize electroactive or photoactive

SMPs by addition of carbon nanotubes [62, 63] or carbon black.[64, 65]

High loadings of CB (>10 vol. %) were previously examined with polyurethane SMPs.

Both Li et al. [66] and Jana et al. [57] found that CB could improve the strain fixity in

polyurethane SMPs. However, CB addition decreased the shape recovery ratio and shape

recovery speed, and did not have much effect on the response temperature.[57, 66] Vaia

et al. [67] studied remote actuation in Morthane (polyester-polyurethane) SMPs, which

were made sensitive to infrared light by addition of 1 to 5 vol.% of carbon nanotubes.

Compared to CB-loaded polyurethane SMPs, substantially less carbon nanotubes were

required to achieve the same improvement in strain fixity. However, large volume

fractions of CB nanoparticles are not always required to improve properties of SMPs.

The results of this study as well as the results of Jana et al. [57]clearly illustrate that low loadings (1 to 5 vol.%) of CB nanoparticles can have a significant impact on certain properties. In addition, the processing of CB into SMPs is straightforward, and CB does not suffer from the potential health risks associated with CNTs.

182

Texas Tech University, Lan Ma, December 2014

6.5 Conclusions

The effects of CB nanoparticles on the 2W-SME observed in cPE have been

systematically investigated. CB nanoparticles are effective additives that can impact

characteristics of the two-way shape-memory cycle, especially Ra and switch

temperatures, at low volume fractions (0.5 to 1.0 vol. %). In the 2W-SME, the

elongation during cooling under load is attributed to the formation of highly oriented

crystals under stress, while the contraction during heating is due to melting of crystals, as

supported by 2D-WAXD and DSC measurements.

In the cPE/CB system studied, the value of Ra improved by more than 60 % after

addition of dilute CB nanoparticles. The CB increased the tensile modulus in the rubbery,

amorphous state without interfering with the material's ability to form highly oriented

crystals during cooling under tension. However, the 2W-SME was lost at high loadings

of CB (20 vol. %) due to the disrupting effects of the concentrated particles on the

formation of oriented crystals. A critical value of the nominal stress (σc ≈ 0.055 MPa) was required to observe the 2W-SME in the cPE samples studied. Below σc, cooling-

induced contraction and heating-induced elongation were observed instead due to the

inability of the material to form highly oriented crystals. CB nanoparticles have some

attractive features that may facilitate their use as additives in SMPs designed for

consumer applications, such as ease of processing, low cost, and low toxicity.

183

Texas Tech University, Lan Ma, December 2014

6.6 References

[1] Liu, C.; Qin, H.; Mather, P. T. Review of progress in shape-memory polymers. Journal of Materials Chemistry 2007, 17 (16), 1543-1558. [2] Ratna, D.; Karger-Kocsis, J. Recent advances in shape memory polymers and composites: a review. Journal of Materials Science 2008, 43 (1), 254-269. [3] Yoshida, M.; Lahann, J. Smart nanomaterials. ACS Nano 2008, 2 (6), 1101-1107. [4] Meng, Q.; Hu, J. A review of shape memory polymer composites and blends. Composites Part A: Applied Science and Manufacturing 2009, 40 (11), 1661- 1672;. [5] Liu, Y.; Lv, H.; Lan, X.; Leng, J.; Du, S. Review of electro-active shape-memory polymer composite. Composites Science and Technology 2009, 69 (13), 2064- 2068. [6] Liu, F.; Urban, M. W. Recent advances and challenges in designing stimuli- responsive polymers. Progress in Polymer Science 2010, 35 (1-2), 3-23. [7] Voit, W.; Ware, T.; Dasari, R. R.; Smith, P.; Danz, L.; Simon, D.; Barlow, S.; Marder, S. R.; Gall, K. High-strain shape-memory polymers. Advanced Functional Materials 2010, 20 (1), 162-171. [8] Lendlein, A.; Shastri, V. P. Stimuli-sensitive polymers. Advanced Materials 2010, 22 (31), 3344-3347. [9] Lendlein, A. Progress in actively moving polymers. Journal of Materials Chemistry 2010, 20 (17), 3332-3334 [10] Behl, M.; Razzaq, M. Y.; Lendlein, A. Multifunctional shape-memory polymers. Advanced Materials 2010, 22 (31), 3388-3410. [11] Xie, T. Recent advances in polymer shape memory. Polymer 2011, 52 (22), 4985- 5000. [12] Hu, J.; Zhu, Y.; Huang, H.; Lu, J. Recent advances in shape-memory polymers: structure, mechanism, functionality, modeling and applications. Progress in Polymer Science 2012, 37 (12), 1720-1763 . [13] Jiang, H. Y.; Kelch, S.; Lendlein, A. Polymers move in response to light. Advanced Materials 2006, 18 (11), 1471-1475. [14] Rezanejad, S.; Kokabi, M. Shape memory and mechanical properties of crosslinked polyethylene/clay nanocomposites. European Polymer Journal 2007, 43 (7), 2856-2865. [15] Iijima, M.; Kobayakawa, M.; Yamazaki, M.; Ohta, Y.; Kamiya, H. Anionic surfactant with hydrophobic and hydrophilic chains for nanoparticle dispersion

184

Texas Tech University, Lan Ma, December 2014

and shape memory polymer nanocomposites. Journal of the American Chemical Society 2009, 131 (45), 16342-16343. [16] Luo, X.; Mather, P. T. Triple-shape polymeric composites (TSPCs). Advanced Functional Materials 2010, 20 (16), 2649-2656. [17] Xu, B.; Fu, Y. Q.; Ahmad, M.; Luo, J. K.; Huang, W. M.; Kraft, A.; Reuben, R.; Pei, Y. T.; Chen, Z. G.; De Hosson, J. T. M. Thermo-mechanical properties of polystyrene-based shape memory nanocomposites. Journal of Materials Chemistry 2010, 20 (17), 3442-3448. [18] Golbang, A.; Kokabi, M. Temporary shape development in shape memory nanocomposites using magnetic force. European Polymer Journal 2011, 47 (8), 1709-1719. [19] Le, H. H.; Schoss, M.; Ilisch, S.; Gohs, U.; Heinrich, G.; Pham, T.; Radusch, H. J. CB filled EOC/EPDM blends as a shape-memory material: manufacturing, morphology and properties. Polymer 2011, 52 (25), 5858-5866. [20] Bae, C. Y.; Park, J. H.; Kim, E. Y.; Kang, Y. S.; Kim, B. K. Organic-inorganic nanocomposite bilayers with triple shape memory effect. Journal of Materials Chemistry 2011, 21 (30), 11288-11295. [21] Kumpfer, J. R.; Rowan, S. J., Thermo-, photo-, and chemo-responsive shape- memory properties from photo-cross-linked metallo-supramolecular polymers. Journal of the American Chemical Society 2011, 133 (32), 12866-12874. [22] Wang, X.; Zhao, J.; Chen, M.; Ma, L.; Zhao, X.; Dang, Z.-M.; Wang, Z. Improved self-healing of polyethylene/carbon black nanocomposites by their shape memory effect. Journal of Physical Chemistry B 2013, 117 (5), 1467-1474. [23] Neffe, A. T.; Hanh, B. D.; Steuer, S.; Lendlein, A. Polymer networks combining controlled drug release, biodegradation, and shape memory capability. Advanced Materials 2009, 21 (32-33), 3394-3398. [24] Serrano, M. C.; Carbajal, L.; Ameer, G. A. Novel biodegradable shape-memory elastomers with drug-releasing capabilities. Advanced Materials 2011, 23 (19), 2211-2215.

[25] Pandini, S.; Passera, S.; Ricco, T.; Borboni, A.; Bodini, I.; Vetturi, D.; Dassa, L.; Cambiaghi, D.; Paderni, K.; Degli Esposti, M.; Toselli, M.; Pilati, F.; Messori, M. Tailored one-way and two-way shape memory response of poly(epsilon- caprolactone)-based systems for biomedical applications. Advances in Science and Technology 2013, 77, 313-318. [26] Bellin, I.; Kelch, S.; Lendlein, A. Dual-shape properties of triple-shape polymer networks with crystallizable network segments and grafted side chains. Journal of Materials Chemistry 2007, 17 (28), 2885-2891.

185

Texas Tech University, Lan Ma, December 2014

[27] Kolesov, I. S.; Radusch, H. J. Multiple shape-memory behavior and thermal- mechanical properties of peroxide cross-linked blends of linear and short-chain branched polyethylenes. Express Polymer Letters 2008, 2 (7), 461-473. [28] Xie, T. Tunable polymer multi-shape memory effect. Nature 2010, 464 (7286), 267-270. [29] Ware, T.; Hearon, K.; Lonnecker, A.; Wooley, K. L.; Maitland, D. J.; Voit, W. Triple-shape memory polymers based on self-complementary hydrogen bonding. Macromolecules 2012, 45 (2), 1062-1069.

[30] Zhao, J.; Chen, M.; Wang, X.; Zhao, X.; Wang, Z.; Dang, Z.-M.; Ma, L.; Hu, G.- H.; Chen, F., Triple shape memory effects of cross-linked polyethylene/polypropylene blends with cocontinuous architecture. ACS Applied Materials & Interfaces 2013, 5 (12), 5550-5556. [31] Chung, T.; Rorno-Uribe, A.; Mather, P. T. Two-way reversible shape memory in a semicrystalline network. Macromolecules 2008, 41 (1), 184-192. [32] Zotzmann, J.; Behl, M.; Hofmann, D.; Lendlein, A. Reversible triple-shape effect of polymer networks containing polypentadecalactone- and poly(epsilon- caprolactone)-segments. Advanced Materials 2010, 22 (31), 3424-3429. [33] Hong, S. J.; Yu, W.-R.; Youk, J. H. Two-way shape memory behavior of shape memory polyurethanes with a bias load. Smart Materials & Structures 2010, 19 (3), 1-9. [34] Westbrook, K. K.; Parakh, V.; Chung, T.; Mather, P. T.; Wan, L. C.; Dunn, M. L.; Qi, H. J. Constitutive modeling of shape memory effects in semicrystalline polymers with stretch induced crystallization. Journal of Engineering Materials and Technology 2010, 132 (4), 041010-1-9. [35] Li, J.; Rodgers, W. R.; Xie, T. Semi-crystalline two-way shape memory elastomer. Polymer 2011, 52 (23), 5320-5325. [36] Raquez, J.-M.; Vanderstappen, S.; Meyer, F.; Verge, P.; Alexandre, M.; Thomassin, J.-M.; Jerome, C.; Dubois, P. Design of cross-linked semicrystalline poly(epsilon-caprolactone)-based networks with one-way and two-way shape- memory properties through Diels-Alder reactions. Chemistry-A European Journal 2011, 17 (36), 10135-10143. [37] Finkelmann, H.; Nishikawa, E.; Pereira, G. G.; Warner, M. A new opto- mechanical effect in solids. Physical Review Letters 2001, 87 (1), 15501-015501. [38] Ahir, S. V.; Tajbakhsh, A. R.; Terentjev, E. M. Self-assembled shape-memory fibers of triblock liquid-crystal polymers. Advanced Functional Materials 2006, 16 (4), 556-560.

186

Texas Tech University, Lan Ma, December 2014

[39] Qin, H.; Mather, P. T. Combined one-way and two-way shape memory in a glass- forming nematic network. Macromolecules 2009, 42 (1), 273-280. [40] Sanchez-Ferrer, A., Fischl, T.; Stubenrauch, M.; Wurmus, H.; Hoffmann, M.; Finkelmann, H. Photo-crosslinked side-chain liquid-crystalline elastomers for microsystems. Macromolecular Chemistry and Physics 2009, 210 (20), 1671- 1677. [41] Sanchez-Ferrer, A.; Fischl, T.; Stubenrauch, M.; Albrecht, A.; Wurmus, H.; Hoffmann, M.; Finkelmann, H. Liquid-crystalline elastomer microvalve for microfluidics. Advanced Materials 2011, 23 (39), 4526-4530. [42] Pandini, S.; Passera, S.; Messori, M.; Paderni, K.; Toselli, M.; Gianoncelli, A.; Bontempi, E.; Ricco, T. Two-way reversible shape memory behaviour of crosslinked poly(epsilon-caprolactone). Polymer 2012, 53 (9), 1915-1924. [43] Charlesby, A. Atomic Radiation and polymers, 1st ed.; 1960. [44] Morshedian, J.; Khonakdar, H. A.; Mehrabzadeh, M.; Eslami, H. Preparation and properties of heat-shrinkable cross-linked low-density polyethylene. Advances in Polymer Technology 2003, 22 (2), 112-119. [45] Li, F. K.; Chen, Y.; Zhu, W.; Zhang, X.; Xu, M. Shape memory effect of polyethylene nylon 6 graft copolymers. Polymer 1998, 39 (26), 6929-6934. [46] Kolesov, I. S.; Kratz, K.; Lendlein, A.; Radusch, H.-J. Kinetics and dynamics of thermally-induced shape-memory behavior of crosslinked short-chain branched polyethylenes. Polymer 2009, 50 (23), 5490-5498. [47] Brandrup, J.; Immergut, E. H.; Grulke, E. A. Polymer Handbook, 4th ed.; 1999.

[48] Silvestre, C.; Cimmino, S.; Raimo, M.; Duraccio, D.; del Amo Fernandez, B.; Lafuente, P.; Leal Sanz, V. Structure and morphology development in films of mLLDPE/LDPE blends during blowing. Macromolecular Materials and Engineering 2006, 291 (12), 1477-1485. [49] Tang, Z. H.; Kang, H. L.; Wei, Q. Y.; Guo, B. C.; Zhang, L. Q.; Jia, D. M. Incorporation of graphene into polyester/carbon nanofibers composites for better multi-stimuli responsive shape memory performances. Carbon 2013, 64, 487-498. [50] Mahapatra, S. S.; Ramasamy, M. S.; Yoo, H. J.; Cho, J. W. A reactive graphene sheet in situ functionalized hyperbranched polyurethane for high performance shape memory material. RSC Advances 2014, 4 (29), 15146-15153. [51] Oh, S. M.; Oh, K. M.; Dao, T. D.; Lee, H. I.; Jeong, H. M.; Kim, B. K. The modification of graphene with alcohols and its use in shape memory polyurethane composites. Polymer International 2013, 62 (1), 54-63.

187

Texas Tech University, Lan Ma, December 2014

[52] Yoonessi, M.; Shi, Y.; Scheiman, D. A.; Lebron-Colon, M.; Tigelaar, D. M.; Weiss, R. A.; Meador, M. A. Graphene polyimide nanocomposites; thermal, mechanical, and high-temperature shape memory effects. ACS Nano 2012, 6 (9), 7644-7655. [53] Kalita, H.; Karak, N. Bio-based hyperbranched thermosetting polyurethane/ triethanolamine functionalized multi-walled carbon nanotube nanocomposites as shape memory materials. Journal of Nanoscience and Nanotechnology 2014, 14 (7), 5435-5442. [54] Wei, Z. G.; Sandstrom, R.; Miyazaki, S. Shape memory materials and hybrid composites for smart systems - Part II Shape-memory hybrid composites. Journal of Materials Science 1998, 33 (15), 3763-3783. [55] Gall, K.; Mikulas, M.; Munshi, N. A.; Beavers, F.; Tupper, M. Carbon fiber reinforced shape memory polymer composites. Journal of Intelligent Material Systems and Structures 2000, 11 (11), 877-886. [56] Mondal, S.; Hu, J. L. Shape memory studies of functionalized MWNT-reinforced polyurethane copolymers. Iranian Polymer Journal 2006, 15 (2), 135-142. [57] Gunes, I. S.; Cao, F.; Jana, S. C. Evaluation of nanoparticulate fillers for development of shape memory polyurethane nanocomposites. Polymer 2008, 49 (9), 2223-2234. [58] Bai, Y. K.; Zhang, Y. M.; Wang, Q. H.; Wang, T. M. Shape memory properties of multi-walled carbon nanotube/polyurethane composites prepared by in situ polymerization. Journal of Materials Science 2013, 48 (5), 2207-2214. [59] Gu, S. Y.; Yan, B. B.; Liu, L. L.; Ren, J. Carbon nanotube-polyurethane shape memory nanocomposites with low trigger temperature. European Polymer Journal 2013, 49 (12), 3867-3877. [60] Fonseca, M. A.; Abreu, B.; Goncalves, F.; Ferreira, A. G. M.; Moreira, R. A. S.; Oliveira, M. S. A. Shape memory polyurethanes reinforced with carbon nanotubes. Composite Structures 2013, 99, 105-111. [61] Cao, F.; Jana, S. C. Nanoclay-tethered shape memory polyurethane nanocomposites. Polymer 2007, 48 (13), 3790-3800. [62] Ahir, S. V.; Terentjev, E. M. Photomechanical actuation in polymer-nanotube composites. Nature Materials 2005, 4 (6), 491-495. [63] Yang, L.; Setyowati, K.; Li, A.; Gong, S.; Chen, J. Reversible infrared actuation of carbon nanotube-liquid crystalline elastomer nanocomposites. Advanced Materials 2008, 20 (12), 2271-2275. [64] Wei, K.; Zhu, G. M.; Tang, Y. S.; Li, X. M. Electroactive shape-memory effects of hydro-epoxy/carbon black composites. Polymer Journal 2013, 45 (6), 671-675.

188

Texas Tech University, Lan Ma, December 2014

[65] Le, H. H.; Kolesov, I.; Ali, Z.; Uthardt, M.; Osazuwa, O.; Ilisch, S.; Radusch, H. J. Effect of filler dispersion degree on the Joule heating stimulated recovery behaviour of nanocomposites. Journal of Materials Science 2010, 45 (21), 5851- 5859. [66] Li, F. K.; Qi, L. Y.; Yang, J. P.; Xu, M.; Luo, X. L.; Ma, D. Z. Polyurethane/conducting carbon black composites: Structure, electric conductivity, strain recovery behavior; and their relationships. Jounral of Applied Polymer Science 2000, 75 (1), 68-77. [67] Koerner, H.; Price, G.; Pearce, N. A.; Alexander, M.; Vaia, R. A. Remotely actuated polymer nanocomposites - stress-recovery of carbon-nanotube-filled thermoplastic elastomers. Nature Materials 2004, 3 (2), 115-120.

189

Texas Tech University, Lan Ma, December 2014

Chapter 7

Summary and Future Work

7.1 Summary

With an increasing need for liquid fuels, petroleum displacement and rural economic development, future energy needs must be met in part or entirely by fuels derived from renewable resources. Ethanol fuel occupied about 5.4 % of the global gasoline type fuel use in 2008. However, bio-ethanol production is now mainly coming from corn, which is not sustainable and cost-effective. Cellulosic ethanol production will provide a sustainable, cost effective and environment preferable method. Technically, in order to get cellulosic ethanol, researchers are focusing on bacterial engineering to consume both hexose and pentose; cellulose degradation to get consumable fermentation sugars; and bioreactor design to increase productivity with the existence of inhibitors.

In this dissertation, a continuous immobilized cell fermentation process for ethanol production from glucose with Escherichia Coli LY01 was developed. Comparisons between batch and continuous (both chemostat and immobilized cell) fermentations were made. Our investigation appears to be the first report of continuous fermentation by E. coli LY01. Specific cell growth rates were calculated from a set of batch fermentations, which was slightly decreased with initial glucose concentration increased. The growth of

E. coli LY01 was also dependent on LB concentration. The arguments of the relative stability of E. coli KO11 during chemostat culture on glucose and sugar mixtures were reported. As the ethanol-tolerant mutants of E. coli KO11, our results proved that the chemostat fermentation of E. coli LY01 with glucose is stable. During the chemostat

190

Texas Tech University, Lan Ma, December 2014

fermentation, we got comparable volumetric productivity with batch close to 1 g L-1 h-1 at

dilution rate of 0.026 h-1 and 70 g/L feed glucose concentration.

Both packed-bed column reactor and stirred tank reactor were designed and studied

as a cell-immobilized reactor. Compared with packed-bed column reactor, cell- immobilized tank reactor is with a perfusion mode to allow the cells to circulate between the bottom and top of the packed-bed section. The glucose reaction rate and conversion were increased due to high cell density achieved in the fermentation. Synthetic porous polymer gel and PET fiber were chosen as cell immobilized materials. The designed cell immobilized tank reactor with PET fiber was able to achieve over double the volumetric productivity of batch fermentation and showed remarkable stability during 10 days fermentation. Packed-bed column reactor was able to improve the ethanol volume productivity by a factor of 10×; however it only works great with low concentration glucose. Kinetics of ethanol production from batch and chemostat fermentation was studied. The obtained kinetic parameters were used in the immobilized tank reactor in order to estimate the immobilized cell density. A preliminary model was built for packed- bed column reactor.

A novel combinatorial methodology developed by our group is able to screen the selective membrane and predict selectivity by simple equilibrium swelling measurement.

An 11×11 matrix of random copolymer networks involving different cross-linker density, different hydrophobic and hydrophilic monomer composition was synthesized. A fundamental study of the mixture solvent transport into the polymer network was present in this dissertation. Equilibrium selectivity and distribution coefficient for all the 121 gels were both calculated based on experimental data and predicted from the model.

191

Texas Tech University, Lan Ma, December 2014

Diffusivities were measured for several samples and the gel with the highest permeability selectivity was picked as a possible material used in the pervaporation or gel stripping separation process.

The last chapter is about shape memory polymers, which were some minor projects I did during Ph. D study.

7.2 Future work

The results of this study are a critical first step toward enabling continuous-flow production of bioethanol from more complex feedstock, such as those from cellulosic sources; below are some suggestions following aspects:

1. Continue work with xylose, mixture of glucose and xylose, and some cellulose

materials as feed substrate. A study that investigates the stability of bacterial

during the fermentation with these substrates is important.

2. Fiber amount affects the immobilized cell mass and substrate diffusion. Future

work should optimize fiber amount and dilution rate in the immobilized tank

reactor in order to reach the highest volumetric productivity without decreasing

glucose conversion.

3. In this work, packed-bed column reactor was able to achieve 10 times higher

productivity than batch fermentation; however, it only works for low glucose

concentration. The reason for the fluctuated fermentation results from the column

reactor need to be studied and prevented. Fermentation process and reactor

structure can be modified in order to consume higher substrate concentration. The

suggested methods are replacing E. coli with yeast which can survive in more

severe environment; cell circulation to increase the immobilized cell mass; or

192

Texas Tech University, Lan Ma, December 2014

work with series of tank reactor after column reactor. In addition, the viability of

immobilized cell needs to be investigated.

4. In chapter 4, kinetics ethanol production of both batch and chemostat

fermentations were studied. The model and fitted parameters that were used to

calculate the viable cell mass in the ICR were based on several assumptions.

These assumptions should be modified with the developed model based on the

fermentation conditions. In addition, a study that investigates the model of the

packed-bed column reactor is intriguing.

5. An efficient combinatorial screening method was provided in Chapter 5 for

ethanol separation material. Future study should focus on improving material

property with increased diffusivity and studying the pervaporation process

applied in the fermentation process on both theoretical and experimental aspects.

193