Process Intensification for the Green Solvent Ethyl Lactate Production Based on Simulated Moving Bed and Pervaporation Membrane Reactors

Home , Ethyl lactate

Process Intensification for the Green Solvent Ethyl Lactate Production based on Simulated Moving Bed and Pervaporation Membrane Reactors

A Dissertation Presented to the Faculdade de Engenharia da Universidade do Porto for the degree of PhD in Chemical and Biological Engineering

by Carla Sofia Marques Pereira

Supervised by Professor Alírio Egídio Rodrigues and Dr. Viviana Manuela Tenedório Matos da Silva

Laboratory of Separation and Reaction Engineering, Associate Laboratory LSRE/LCM Department of Chemical Engineering, Faculty of Engineering, University of Porto

October 2009

FEUP-LSRE/LCM - Universidade do Porto

Acknowledgements

First of all, I want to thank my supervisors, Professor Alírio Rodrigues and Dr Viviana Silva. Professor Alírio, thank you for all the friendship, constant support and for always challenging me to reach higher goals within my work. Dr. Viviana, I want to thank you for all the encouragement, motivation, constant support, for all the long discussions and great ideas that make me go further and further within my work, and, also, for being a truly and special friend.

I am very grateful to Professor Simão Pinho, for the friendship and all the support in the framework of the project “POCI/EQU/61580/2004” and to Professor Madalena Dias for the support whenever needed.

To all my LSRE colleagues, especially Israel Pedruzzi, Pedro Sá Gomes, Michael Zabka, João Santos, Miguel Granato, João Pedro Lopes, Alexandre Ferreira and Nuno Lourenço for the friendship, collaboration, and support whenever I needed.

To my primary school teacher, Professor João Aveiro, for always believing in me and keeping me motivated along the years.

To Fundação para a Ciência e Tecnologia, for the financial support (Research Fellowship: SFRH / BD / 23724 / 2005).

To Sofia Rodrigues, Marta Abrantes, Fátima Mota, Miguel Teixeira and Nuno Garrido for all the great moments spent after work!!!

Last, but not least, I would deeply like to thank my family and friends, for all giving love, support and trust, especially to my grandmother, Dulcelina, that will always stay in my heart.

To my parents and sisters

“Green chemistry represents the pillars that hold up our sustainable future. It is imperative to teach the value of green chemistry to tomorrow’s chemists.”

Daryle Busch, President of the American Chemical Society (June 26, 2000)

Resumo

O principal objectivo deste trabalho foi o desenvolvimento de um novo processo eficiente para a produção do solvente verde, lactato de etilo, através da reacção de esterificação entre etanol e ácido láctico, utilizando tecnologias híbridas de reacção/separação baseadas em reactores de leito móvel simulado e processos de membranas por pervaporação. De forma a atingir a meta proposta, foram abordados os seguintes temas:

Aquisição de dados fundamentais: A resina de permuta iónica, Amberlyst 15-wet, foi avaliada tanto como catalisador para a reacção de esterificação, como adsorvente selectivo para a água. Os dados cinéticos e de equilíbrio da reacção foram medidos, na gama de temperaturas 50ºC-90ºC, e usados para a determinação da constante de equilíbrio e da lei cinética da reacção como função da temperatura, baseadas em actividades descritas pelo modelo UNIQUAC. Os dados de adsorção foram também medidos, a 20ºC e a 50ºC, e ajustados a uma isotérmica de Langmuir multicomponente, tendo-se assumido uma capacidade volumétrica da monocamada igual para todas as espécies, reduzindo o número de parâmetros de ajuste de 8 para 5, para cada temperatura. Membranas comerciais hidrófilas da Pervatech foram avaliadas para a desidratação do etanol, ácido láctico e lactato de etilo, por pervaporação. As permeâncias de todas as espécies foram determinadas em função da composição e temperatura na gama 48ºC-72ºC.

Intensificação de processo: Modelos matemáticos, considerando resistências internas e externas à transferência de massa e velocidade variável devido à mudança das propriedades da mistura multicomponente, foram desenvolvidos para reactores cromatográficos — reactores de leito fixo e de leito móvel simulado, SMBR — e validados pelos dados experimentais. O modelo matemático do reactor de membranas por pervaporação, PVMR, considera, adicionalmente, a permeação através da membrana, os efeitos de polarização por concentração e temperatura e operação não isotérmica. A avaliação teórica do comportamento da unidade SMBR foi realizada para analisar o efeito da configuração, da composição da alimentação e tempo de comutação nas regiões de separação/reacção e/ou no desempenho do processo nos pontos operacionais óptimos. O desempenho do PVMR foi avaliado para operação isotérmica e não isotérmica, e foram determinadas condições apropriadas para a maximização quer da conversão do ácido láctico quer da pureza do lactato de etilo. Finalmente, uma nova tecnologia foi desenvolvida e submetida a registo de patente, o reactor de membranas de leito móvel simulado, PermSMBR, o qual integra membranas selectivas dentro das colunas do SMBR.

Abstract

The main objective of this work was the development of a new efficient process to produce the green solvent ethyl lactate from the esterification reaction between ethanol and lactic acid by using hybrid reaction/separation technologies based on simulated moving bed reactors and pervaporation membrane processes. To accomplish this target, the following topics were addressed:

Basic data acquisition: The acidic ion exchange resin Amberlyst 15-wet was evaluated as both catalyst for esterification and selective adsorbent for water. Equilibrium and kinetic data were measured in the temperature range 50-90ºC, and used to obtain the equilibrium constant and kinetic law as function of temperature, which are based on liquid activities described by the UNIQUAC model. Adsorption data was also obtained and fitted to a multi-component Langmuir isotherm assuming a constant monolayer capacity in terms of volume for all species, reducing the adjustable parameters from 8 to 5, for each temperature. Pervatech hydrophilic commercial membranes were evaluated for the dehydration of ethanol, lactic acid and ethyl lactate, by pervaporation. The permeances of all species were determined as function of composition and temperature in the range 48-72ºC.

Process intensification: Mathematical models, considering external and internal mass-transfer resistances and velocity variations due to the change of multi-component mixture properties, were developed for chromatographic reactors — fixed bed and simulated moving bed reactor, SMBR — and validated by experimental data. The pervaporation membrane reactor, PVMR, model also takes into account, the permeation through the membrane, concentration and temperature polarization effects, and non-isothermal operation. The theoretical assessment of the SMBR unit behaviour was performed to analyse the effect of SMBR configuration, feed composition and switching time on the reactive/separation regions and/or on the process performance at the optimal operating points. The performance of the PVMR was evaluated for isothermal and non- isothermal operation, and suitable conditions for maximization of both lactic acid conversions and ethyl lactate purity were examined. Finally, a new technology was developed and submitted to patent registration, the simulated moving bed membrane reactor, PermSMBR, which integrates perm-selective membranes inside the SMBR columns.

Zusammenfassung

Ziel dieser Arbeit war es einen völlig neuartigen und effizienten Prozess für die Produktion von “Grünem Lösungsmittel” und Ethyl-Lakton, mittels der Verästerung aus Ethanol und Milchsäure über die Hybrid-Technologie, zu entwickeln. Die Reaktion/Trennung basiert auf simulierten Fliessbettreaktoren und Membranprozessen mittels Pervaporization. Um diese Zielsetzung zu erreichen, wurden folgende Themen diskutiert:

Folgende wichtige Daten wurden gesammelt: Für den Ionen-Austausch wurde das Harz, Amberlyst 15-wet, genutzt. Es wurde sowohl als Katalysator für die Reaktivverästherung, wie auch als selektiver Adsorbent für Wasser ausgewertet. Die Kinetischen- und chemischen Gleichgewichtsdaten wurden zwischen 50ºC und 90ºC gemessen. Als Vorlage für die Bestimmung der chemischen Gleichgewichtskonstante wie auch der Kinetik als Funktion der Temperatur, wurde das UNIQUAC Model genutzt. Die Adsorptionsdaten wurden zwischen 20ºC und 50ºC gemessen, und entsprechend einer Langmuir-Isotherme für Multikomponenten angeglichen, wobei eine gleich grosse Volumenkapazität der Monoschicht für alle Spezies angenommen wurde. Die Anzahl der zu justierenden Parameter wurde hierbei für jede Temperatur von 8 auf 5 reduziert. Kommerzielle hydrophile Membranen der Firma Pervatech wurden für die Dehydratisierung von Ethanol, Milchsäure und Ethyllakton mittels Permeation ausgewertet. Die Permeation aller Spezies wurde als Funktion der Zusammensetzung und Temperatur, zwischen 48ºC und 72ºC ermittelt.

Intensivierung des Prozesses: Es wurden mathematische Modelle für chromatographische Reaktoren entwickelt (Modelierte Festbett- und Fliessbettreaktoren SMBR) und experimentell ausgewärtet, unter Berücksichtigung des internen und externen Massenaustausches, sowie der variablen Geschwindigkeiten. Es herrschen unterschiedlichen Vermischungseigenschaften der Multikomponenten. Das mathematische Modell der Pervaporationsmembrane (PVMR) berücksichtigt gleichfalls die folgenden Effekte: Polarization aufgrund von Temperatur- und Konzentrationsgradienten sowie nicht-isotherme Reaktionsführung. Das theoretische Verhalten der SMBR Einheit wurde unter Berücksichting der Konfiguration, der Mischungszusammensetzung beim Eintritt und der Komutationszeit in den Trennungs- und Reaktionszonen und/oder für die Prozessleistung an den operationallen Optima ausgewertet. Der Wirkungsgrad der PVMR wurde für den isothermen und nicht-isothermen Betrieb ermittelt, und es wurden entsprechende Bedingungen für den maximale Umsatz der Milchsäure und der maximalen Reinheit für Ethyl-Lakton bestimmt. Der lezte Schritt war die Entwicklung einer neuen Technolgie, ein Membranreaktor mit simuliertem Fliessbett (PermSMBR), welche als Patent angemeldet wurde. Dieser Reaktor integriert selektive Membranen innerhalb der SMBR Säule.

Table of contents Pag.

1. Introduction...... 1 1.1 Relevance and Motivation...... 1 1.2 Objectives and Outline...... 3

2. State of the art on Green Solvent Ethyl Lactate...... 5 2.1 Green Chemistry ...... 5 2.2 Ethyl lactate applications ...... 7 2.2.1 Solvent Market Analysis ...... 8 2.3 Synthesis of ethyl lactate...... 9 2.3.1 Renewable Resources...... 11 2.3.1.1 Ethanol Platform...... 12 2.3.1.2 Lactic acid Platform...... 13 2.3.2 Patented Processes Overview...... 13 2.3.3 Reactive Separations ...... 16 2.3.3.1 Reactive Distillation (RD) ...... 17 2.3.3.2 Simulated Moving Bed Reactor (SMBR) ...... 19 2.3.3.3 Pervaporation Membrane Reactor (PVMR)...... 21 2.4 References ...... 26

3. Batch Reactor: Thermodynamic Equilibrium and Reaction Kinetics...... 35 3.1 Introduction ...... 36 3.2 Experimental Section...... 40 3.2.1 Chemicals and Catalyst ...... 40 3.2.2 Experimental set-up...... 41 3.2.3 Analytical method ...... 42 3.3 Thermodynamic Equilibrium Results ...... 42 3.3.1 Thermodynamic equilibrium constant...... 42 3.3.1.1 Activity coefficients estimation ...... 44 3.3.2 Equilibrium constant and reaction enthalpy for the synthesis of Ethyl Lactate...... 45 3.3.3 Application of this methodology to other works...... 47 3.4 Kinetic Studies ...... 49 3.4.1 Preliminary Studies ...... 50 3.4.1.1 Evaluation of external mass transfer limitations (effect of stirring speed)...... 50 3.4.1.2 Evaluation of internal mass transfer limitations (effect of particle size)...... 50 3.4.1.3 Evaluation of catalyst deactivation (effect of catalyst reusability)...... 51 3.4.2 Kinetic Model...... 52 3.4.2.1 Parameter estimation from experimental data...... 54 3.4.3 Modelling and discussion of results ...... 55 3.4.3.1 Effect of catalyst loading ...... 56 3.4.3.2 Effect of initial molar ratio of reactants ...... 57 3.4.3.3 Effect of reaction temperature...... 58 ii TABLE OF CONTENTS

3.4.3.4 Effect of Lactic acid and Ethyl Lactate oligomers...... 58 3.4.3.5 Effect of polar species ...... 62 3.5 Conclusions...... 63 3.6 Notation...... 64 3.7 References Cited ...... 66

4. Fixed Bed Adsorptive Reactor...... 71 4.1 Introduction...... 72 4.2 Experimental Section...... 73 4.2.1 Chemicals and Catalyst / Adsorbent ...... 73 4.2.2 Experimental Apparatus...... 74 4.2.2.1 Bed Porosity and Peclet Number ...... 75 4.3 Modelling of Fixed Bed ...... 76 4.3.1 Multi-component viscosity ...... 81 4.4 Results and Discussion ...... 84 4.4.1 Adsorption Isotherm ...... 84 4.4.1.1 Binary Adsorption experiments...... 85 4.4.2 Kinetic experiments ...... 90 4.4.2.1 Fixed Bed Reactor ...... 90 4.5 Conclusions...... 94 4.6 Notation...... 94 4.7 References...... 97

5. Simulated Moving Bed Reactor...... 101 5.1 Introduction...... 102 5.2 Modelling Strategies ...... 104 5.2.1 SMBR mathematical model...... 104 5.2.2 SMBR performance parameters...... 108 5.2.3 Numerical Solution...... 108 5.3 Experimental Section...... 109 5.3.1 Chemicals and Catalyst / Adsorbent ...... 109 5.3.2 The SMBR LICOSEP 12-26 Unit...... 109 5.4 Results and Discussion ...... 111 5.4.1 Experimental Results ...... 111 5.4.2 Simulated results...... 115 5.4.2.1 Comparison of SMBR and TMBR models...... 115 5.4.2.2 Reactive/separation regions ...... 116 5.4.2.3 Separation Region vs Reactive/Separation Region...... 117 5.4.2.4 Effect of the Feed Composition...... 118 5.4.2.5 Effect of the SMBR columns arrangement ...... 120 5.4.2.6 Effect of Switching Time...... 121 5.5 Conclusions...... 123 5.6 Notation...... 124 5.7 References Cited ...... 127

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION iii

6. Pervaporation Membrane Reactor...... 131 6.1 Introduction ...... 132 6.2 Experimental Section...... 135 6.2.1 Materials...... 135 6.2.2 Pervaporation Membrane Reactor Unit...... 135 6.3 Pervaporation Studies...... 136 6.3.1 Pervaporation Transport...... 137 6.3.2 Preliminary Studies ...... 138 6.3.2.1 Evaluation of the membrane quality ...... 138 6.3.2.2 Evaluation of mass transfer limitations in the boundary layer ...... 139 6.3.3 Detailed Studies ...... 140 6.3.3.1 Water/Ethanol System ...... 140 6.3.3.2 Water/Ethyl lactate System...... 140 6.3.3.3 Water/Lactic acid System ...... 141 6.3.3.4 Membrane performance evaluation...... 141 6.3.4 Parameters estimation ...... 143 6.3.4.1 Permeance temperature dependence ...... 143 6.3.4.2 Permeance temperature and water content dependence ...... 144

6.3.4.3 Estimation of the boundary layer mass transfer coefficient (kbl)...... 148 6.4 Modelling ...... 149 6.4.1 Batch Pervaporation Model...... 149 6.4.2 Pervaporation Membrane Reactor model...... 152 6.5 Results and Discussion...... 155 6.5.1 Batch Pervaporation ...... 155 6.5.2 Pervaporation Membrane Reactor...... 157 6.6 Conclusions ...... 161 6.7 Notation...... 162 6.8 References ...... 165

7. PermSMBR – A New Hybrid Technology ...... 171 7.1 Introduction ...... 172 7.2 Technical description of the PermSMBR technology...... 174 7.3 PermSMBR mathematical model...... 178 7.4 PermSMBR geometrical specifications ...... 181 7.5 Simulated Results...... 182 7.5.1 Reactive/Separation Region: PermSMBR vs SMBR ...... 183 7.5.2 PermSMBR 3 zones ...... 184 7.5.3 Comparison between PermSMBR, SMBR and RD technologies ...... 187 7.6 Conclusions ...... 187 7.7 Notation...... 188 7.8 References ...... 191

8. Conclusions and Suggestions for Future Work...... 193

APPENDIX A. Safety Data ...... A1 iv TABLE OF CONTENTS

APPENDIX B. Thermodynamic Properties ...... B1

APPENDIX C. Calibration ...... C1

APPENDIX D. Binary adsorption experiments at 293.15 K ...... D1

1. Introduction

1.1 Relevance and Motivation

Petroleum (“black gold”) is at the heart of today’s economics and politics problems. The traditional petroleum reserves are in decline; moreover, the environmental regulation is every day more severe, being, therefore, a great challenge to the design and implementation of green products and processes.

Green solvents, which are produced from the processing of agricultural crops, were developed as a more environmentally friendly alternative to petrochemical solvents. Lactate esters solvents are 100% biodegradable, easy to recycle, non-corrosive, non-carcinogenic and non-ozone depleting. Lactate esters have found industrial applications in specialty coatings, inks, cleaners and straight cleaning use.

Ethyl lactate is a green solvent derived from nature-based feedstocks and it is so benign that the U.S. Food and Drug Administration approved its use in food products. Ethyl lactate could replace a range of environment-damaging halogenated and toxic solvents, including ozone- depleting chlorofluorocarbons, carcinogenic methylene chloride, and toxic ethylene glycol ethers and chloroform. In Figure 1.1 the ethyl lactate life-cycle is shown.

Ethyl lactate is produced from the esterification of lactic acid with ethanol through a reversible reaction, having water as a by-product. Traditionally, ethyl lactate is synthesized in a reactor followed by separation units in order to recover it, to remove the by-product (water) and to recycle the unconverted reactants to the reactor; however, this represents high costs. The objective of this work is to study equipments and techniques that are more compact, 2 CHAPTER 1. Introduction

energy efficient, and environment-friendly sustainable processes for the ethyl lactate production.

From Nature To Nature

Figure 1.1- Ethyl Lactate life-cycle.

Process intensification, regarding the integration of reaction and separation processes into a single device, provides the most feasible engineering solution to the sustainable synthesis of ethyl lactate, since at least one of the products is being removed from the reaction medium to lead to depletion of the limiting reactant. In this perspective, continuous chromatographic reactor and membrane reactor will be considered for ethyl lactate production, namely the Simulated Moving Bed Reactor (SMBR) and the Pervaporation Membrane Reactor (PVMR), respectively. The SMBR is a competitive technology for systems involving equilibrium controlled reactions catalysed by ion exchange resins, which are also selective adsorbent for water (the by-product formed in the ethyl lactate synthesis), given that the products are formed and simultaneously separated and removed from the reaction medium. The PVMR technology is a clean and economic alternative to conventional processes, since equilibrium could be shifted by continuously removing water through a selective membrane, allowing costs reduction and higher product purity. Combining the advantages of both technologies, finally, a new hybrid technology will be developed, the Simulated Moving Bed Membrane Reactor (SMBMembR or PermSMBR) that combines a reactor with two different separation techniques into a single device: continuous counter-current chromatography (SMB) with a selective permeable membrane (Pervaporation or Permeation). PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 3

1.2 Objectives and Outline

The main goal of this work is the development of a process for simultaneous reaction and separation in a single device for ethyl lactate production with a high purity, yield and complete reactants conversion. Three technologies will be studied: the Simulated Moving Bed Reactor (SMBR) using a catalyst that is also a selective adsorbent for water/ethyl lactate separation, the Pervaporation Membrane Reactor (PVMR) using a hydrophilic water permselective membrane for continuous removal of the by-product water and the Simulated Moving Bed Membrane Reactor (PermSMBR), where the SMBR is integrated with the PVMR by using selective permeable membranes inside the columns of the SMBR.

The thesis comprises 8 chapters dealing with different aspects of ethyl lactate production, in addition to the present one.

In Chapter 2, the state of the art of production process aspects as patented processes for esters production; the advantages of heterogeneous catalyst, such as ion-exchange resins; the methods used to displace equilibrium towards ester formation are addressed. An overview in some reactive separations as reactive distillation, chromatographic reactors and membrane reactors applied to the production of oxygenates is reviewed in order to improve the overall efficiency of the process of ethyl lactate synthesis.

Chapter 3 addresses the kinetic studies for ethyl lactate production by heterogeneous catalysis; the influence of catalyst loading, temperature and initial molar ratio of reactants are analysed. A methodology based in the UNIQUAC model for determination of the thermodynamic equilibrium constant is developed.

Experimental and simulated results for the ethyl lactate production in a fixed bed adsorptive reactor are shown in Chapter 4. Dynamic adsorption experiments of binary non-reactive mixtures were performed in order to obtain multicomponent adsorption equilibrium isotherms of Langmuir type. The reaction kinetics and adsorption data were used in the mathematical model of the adsorptive reactor, which also included axial dispersion, velocity variations and external and internal mass-transfer resistances.

In Chapter 5, the simulated moving bed reactor technology for the ethyl lactate production is evaluated by experiments as well as by simulations. In order to describe the dynamic behaviour of this unit, a mathematical model considering external and internal mass-transfer resistances and variable velocities is developed. The influence of operational parameters, as 4 CHAPTER 1. Introduction

feed composition, SMBR configuration and switching time, on the SMBR performance is presented.

Pervaporation processes using hydrophilic silica membranes are evaluated in Chapter 6 for the ethyl lactate system. The effects of feed composition and operating temperature on the membrane performance are analyzed. Mathematical models, considering concentration and temperature polarization and non-isothermal effects, are developed and applied to analyze the performance of batch pervaporation and continuous pervaporation membrane reactor, in both isothermal and non-isothermal conditions.

In Chapter 7, a new technology, the simulated moving bed membrane reactor, is presented and applied for the ethyl lactate synthesis. The potential of this new equipment is demonstrated by comparing its performance to other reactive separation processes.

Finally, the general conclusions drawn from this work and the suggestions for future work will be presented in Chapter 8.

2. State of the art on Green Solvent Ethyl Lactate

In this chapter, a review on the green chemistry principles is made and a literature survey on applications of ethyl lactate and production processes (renewable resources, patents, reactive separations) is presented.

2.1 Green Chemistry

Green chemistry is the best use of chemistry for pollution prevention. More specifically, green chemistry is the design of chemical products and processes that reduce or eliminate the use and generation of hazardous substances. It is a highly effective approach to pollution prevention because it applies innovative scientific solutions to real-world environmental situations.

Currently, a great challenge is the design and implementation of completely green products and processes. There is not a systematic and reliable method for ensuring that the chemistry being implemented is green, since the number of chemicals synthesis pathways is enormous. Indeed, it is more correct to verify if a proposed manufacturing process is “greener” than other alternatives. Anastas and Warner have developed the “Twelve Principles of Green Chemistry” to aid one in assessing how green is a product or a process (Anastas and Warner, 1998), which are:

1. It is better to prevent waste than to treat or clean up waste after it is formed.

2. Synthetic methods should be designed to maximize the incorporation of all materials used in the process into the final product. 6 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

3. Wherever practicable, synthetic methodologies should be designed to use and generate substances that possess little or no toxicity to human health and the environment.

4. Chemical products should be designed to preserve efficacy of function while reducing toxicity.

5. The use of auxiliary substances (e.g. solvents, separation agents, etc.) should be made unnecessary whenever possible and, innocuous when used.

6. Energy requirements should be recognized for their environmental and economic impacts and should be minimized. Synthetic methods should be conducted at ambient temperature and pressure.

7. A raw material or feedstock should be renewable rather than depleting whenever technically and economically practical.

8. Unnecessary derivatization (blocking group, protection/deprotection, and temporary modification of physical/chemical processes) should be avoided whenever possible.

9. Catalytic reagents (as selective as possible) are superior to stoichiometric reagents.

10. Chemical products should be designed so that at the end of their function they do not persist in the environment and break down into innocuous degradation products.

11. Analytical methodologies need to be further developed to allow for real-time in- process monitoring and control prior to the formation of hazardous substances.

12. Substances and the form of a substance used in a chemical process should be chosen so as to minimize the potential for chemical accidents, including releases, explosions, and fires.

Based on these twelve principles, this thesis focuses:

1) The synthesis of the green solvent Ethyl lactate produced from renewable raw material that is a more environmentally friendly alternative to petrochemical solvent: 7th principle.

2) Ethyl lactate is 100% biodegradable, easy to recycle, non-corrosive, non- carcinogenic and non-ozone depleting: 3rd, 4th and 10th principles.

3) The use of solid acid catalysts to improve the reaction kinetics without using increase the stoichiometric of reactants, and it is more advantageous then

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 7

homogenous catalysts, since these are more corrosive and require a further step of neutralization: 1st and 9th principles.

4) The process intensification by using hybrid technologies where reaction and separation of at least one product take place in a single unity (SMBR, PVMR and PermSMBR) will reduce/eliminate the use of solvents and requires less energy consumption: 5th and 6th principles.

2.2 Ethyl lactate applications

The reaction between an alcohol and a carboxylic acid to form an ester and water is of considerable industrial interest (Dhanuka et al., 1977). Organic esters are a very important class of chemicals having applications in a variety of areas in the chemical industries such as perfumes, flavours, pharmaceuticals, plasticizers, solvents and intermediates (Weissermel and Arpe, 1997).

Ethyl lactate is an important organic ester, which is biodegradable and can be used as food additive, in perfumery, as flavour chemicals and solvent, which can dissolve acetic acid cellulose and many resins (Tanaka et al., 2002). It is a particularly attractive solvent for the coatings industry as a result of its high solvency power, high boiling point, low vapour pressure and low surface tension. Ethyl lactate is a desirable coating for wood, polystyrene and metals and also acts as a very effective paint stripper and graffiti remover. It has replaced solvents including N-methyl Pyrrolidone (NMP) (Reisch, 2008), toluene, acetone and xylene, which has resulted in the workplace being made a great deal safer. In Table 2.1 solvating properties of ethyl lactate and NMP are presented. Table 2.1 Solvating properties of ethyl lactate and N-methyl Pyrrolidone.

Ethyl Lactate N-methyl pyrrolidone

Kauri Butanol(KB) Value >1000 350

Hildebrand 21.3 23.1

Disperse 7.8 8.8

Polar 3.7 6.0

Hydrogen 6.1 3.5 Miscible in Water and Miscible in Water and Solubility Hydrocarbons Hydrocarbons

8 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

Other applications of ethyl lactate include being an excellent cleaner for the polyurethane industry and for metal surfaces, efficiently removing greases, oils, adhesives and solid fuels. Beyond all these applications, ethyl lactate can also be used in the pharmaceutical industry as a dissolving/dispersing excipient for various biologically active compounds without destroying the pharmacological activity of the active ingredient. It proves to be a very effective agent for solubilising biologically active compounds that are difficult to solubilise in usual excipients (Muse and Colvin, 2005).

Ethyl lactate can also be applied as a more environment friend alternative route to produce 1,2-propanediol, which is normally produced by the hydration of propylene oxide derived from petrochemical resource (Huang et al., 2008). In Table 2.2 the major benefits of the ethyl lactate are presented.

Table 2.2 Ethyl lactate major benefits.

Ethyl Lactate Benefits

100% Biodegradable Renewable - made from corn and other carbohydrates

FDA approved as a flavour additive EPA approved SNAP solvent

Non carcinogenic Non corrosive

Great penetration characteristics Stable in solvent formulations until exposed to water

Rinses easily with water High solvency power for resins, polymers and dyes

High boiling point Easy and inexpensive to recycle

Low VOC Not a Ozone Depleting Chemical

Low Vapor Pressure Not a Hazardous Air Pollutant

2.2.1 Solvent Market Analysis

Almost all manufacturing and processing industries depend on the use of solvents (see Figure 2.1). The world solvent market is estimated at 30 million pounds per year at prices from $0.90 to $1.70 per pound. The ethyl lactate green solvent has the potential to displace 80 % of these solvents (Energetics Incorporated, 2003).

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 9

Agricultural Dry Cleaning 1% Oil Seed Extract chemicals 2% 2% Others 8% Rubber/Polymer Manufacture 4% Paints 46% Metal/Industrial Cleaning 4%

House/Car 6%

Personal Care 6%

Printing Inks 6%

Pharmaceuticals Adhesives 6% 9%

Figure 2.1 Solvent demand (AAE Chemie, 2009).

Selling prices for ethyl lactate have ranged from $1.50 to $2.00 per pound, but processing advances could drive the price as low as $1.00 to $0.85 per pound (Argonne, 2006), enabling ethyl lactate to compete directly with the petroleum-derived toxic solvents currently used. Moreover, the crude prices have risen sharply, making of ethyl lactate green solvent more commercially attractive. Among this, due to an environmental consciousness, some consumers are willing to pay more for products that are less detrimental to the environment.

2.3 Synthesis of ethyl lactate

The conventional way to produce ethyl lactate is the esterification of lactic acid with ethanol catalyzed by an acid catalyst, according to the reaction:

+ Ethanol (Eth) + Lactic Acid (La) ←⎯→⎯H Ethyl Lactate (EL) +Water (W )

The use of these reactants (ethanol and lactic acid) has the advantage of both being produced from renewable resources (by glucose or sugar fermentation processes).

Esterifications are self-catalyzed reactions, since the H+ cation released from the partial dissociation of the carboxylic acid used as reactant catalyses the reaction. However, the use of catalyst is favourable for the reaction rate as the kinetics of the self-catalyzed reaction is extremely slow, since its rate depends on the autoprotolysis of the carboxylic acid. For

10 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

example, the lactic acid acidity constant is pKa=3.86 @ 25 ºC, and therefore an aqueous solution with 85 % of lactic acid (about 10.8 M) has a pH=1.4. Typically, the catalytic production of lactates is performed with homogeneous catalysts using acids, such as sulphuric acid, phosphoric acid and anhydrous hydrogen chloride. However, the use of heterogeneous catalyst (as for example, zeolites, ion-echange resins like Amberlyst 15-wet, Nafion NR50, among others) has clear advantages:

- easy to separate from the reaction medium;

- long life time;

- higher purity of products (side reactions can be eliminated or are less significant);

- elimination of the corrosive environment caused by the discharge of acid containing waste.

As previously mentioned, the esterification is a reversible reaction and, in order to obtain acceptable ester yields, the equilibrium must be displaced towards the ester production, which might be accomplished by different methods, such as:

1. to use a large excess of one of the reactants, in general the alcohol; however, this results in a relatively inefficient use of reactor space and in very diluted products, which will require an efficient separation afterwards;

2. to eliminate the water by azeotropic distillation between a solvent and water – the solvent and water must be partially miscible and the boiling points of the different components in the reaction medium must be compatible with that azeotrope;

3. to use reactive separations (as reactive distillation, simulated moving bed reactor, pervaporation reactor, etc.) in order to remove the products from the reaction medium.

In reactions limited by chemical equilibrium where more than one product is formed conversion can be enhanced in multifunctional reactor where the products are separated as they are formed. Novel reactor configurations and choice of operating conditions can be used to maximise the conversion of reactants and improve selectivity of desired product, thereby reducing the costs associated with the separation step. Recently, reactive distillation,

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 11

chromatographic reactors and membrane reactors have been intensively applied to esterifications processes, as it will be discussed latter within this chapter.

2.3.1 Renewable Resources

In recent years, an increasing demand on using biorenewable materials instead of petroleum based feedstocks for producing chemicals, driven by environmental concerns and by the concept of sustainability, has been noticed. Biobased products are one of the main pillars of a sustainable economy. Nature produces 170 billion tons of biomass per year by photosynthesis, 75 % of which belong to the class of carbohydrates; however, just 3-4 % of these compounds are used by humans for food and non-food purposes (Röper, 2002). Carbohydrates are very abundant renewable resources and they are currently considered as an important feedstock for the Green Chemistry of the future (Lichtenthaler, 1998; Lichtenthaler, 2002; Lichtenthaler and Peters, 2004). Industrial plants, named as biorefineries, have been created where biomass is converted economically and ecologically, in chemicals, materials, fuels and energy (see Figure 2.2). The biorefineries could be the basis of the new bioindustry and its concept is similar to the petroleum refinery; the difference is that the biorefinery is based on conversion of biomass feedstocks instead of crude oil.

Figure 2.2 Schematic diagram of a biorefinery for precursor-contained biomass. (Kamm and Kamm, 2004a; Kamm and Kamm, 2004b)

12 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

2.3.1.1 Ethanol Platform

Ethanol is an important raw material in the chemical industry and can also be used as transportation fuel. It can be produced from a variety of biomass crops, including sugar crops (e.g., sugarcane and sugar beet), starch crops (e.g., corn and cassava), or cellulosic feedstocks (e.g., wood, grasses and agricultural residues).The production of ethanol from starch crops involves as main steps: liquefaction and saccharification (conversion to sugar), milling, pressing, fermentation and distillation. The production from cellulosic feedstocks is similar, however it is significantly more difficult and costly to convert cellulose and hemicellulose into their component sugars (glucose and xylose, respectively) than is the case for starches (Sagar and Kartha, 2007). Currently, more than 37 billion litters of ethanol are produced worldwide per year from starch and sugar crops (Rass-Hansen et al., 2007; Tilman et al., 2006). In 2008, cellulosic ethanol industry developed some new commercial-scale plants. In the United States, plants with 12 million liters capacity per year were operational, and an additional 80 million liters per year of capacity (26 new plants) was under construction. In Canada, capacity of 6 million liters per year was operational. In Europe, several plants were operational in Germany, Spain, and Sweden, and capacity of 10 million liters per year was under construction (REN21, 2009). Ethanol derived from cellulosic crops is appealing since it broadens the scope of potential feedstocks beyond starch and sugar-based food crops. Moreover, cellulosic ethanol can be more effective and promising as an alternative renewable biofuel than corn ethanol because its use reduces even more the net greenhouse gas (GHG) emissions when compared with the petroleum fuel (Wang et al., 2008) (see Figure 2.3).

Figure 2.3 Percent change in greenhouse gas emissions (adapted from Wang et al, 2008).

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 13

2.3.1.2 Lactic acid Platform

Lactic acid (2-hydroxypropionic acid) is an important platform chemical for the biorenewable economy. It is an α-hydroxy acid containing a hydroxyl group adjacent to the carboxylic acid functional group; a review on the lactic acid chemistry can be found in literature (Holten, 1971). Lactic acid can be produced through chemical synthesis or through the fermentation of different carbohydrates, such as, glucose (from starch), maltose (produced by specific enzymatic starch conversion), sucrose (from syrups, juices, and molasses), or lactose (from whey) (Corma Canos et al., 2007). Nowadays, it is commercially produced by fermentation of glucose. One of the most important steps in the lactic acid production is the recovery from fermentation broths. The separation and purification stages represent about 50 % of the total production cost. However, current advances in membrane-based separation and purification technologies, particularly in microfiltration, ultrafiltration and electrodialysis, have originated new processes which should reduce the lactic acid cost production (Wasewar et al., 2004). The lactic acid production is around 350,000 tons per year and it is defended by some observers that the worldwide growth per year is of 12-15 % (Wasewar et al., 2004). A lot of products are derived from lactic acid; some of them are new chemical products and others, which represent biobased routes to chemicals, currently produced from petroleum. The most important ones are shown in Figure 2.4.

Figure 2.4 Some potential derivatives of lactic acid (Corma Canos et al., 2007).

2.3.2 Patented Processes Overview

There are a great number of patents related to esters production. A summary of those patented processes is presented in Table 2.3.

Table 2.3 Patented processes for esters production. Commercial Solvents Lactic acid is dehydrated and mixed with ethanol and concentrated sulphuric acid (catalyst). This mixture is refluxed for one hour and the Corporation esterification takes place to a determined extent. Afterwards, distillation is started to separate the components of the reaction mixture. (Bannister, 1936) Temperature range of the process 135ºC-145ºC. USA-Secretary of Process to produce esters from the reaction between the basic nitrogen salt of the carboxylic acid with an alcohol. It is a batch process and Agriculture distillation is applied to separate the final mixture components. The process achieved from 61 to 92% ammonia removal and from 49 to 67% (Filachione and Fisher, 1951) conversion to butyl lactate. Temperature range 89ºC-195ºC. Process, where the esterification extent is enhanced, using pervaporation membranes in order to selective remove water from the reaction zone. The American Oil Company It is a continuous method, which integrates reaction and separation in the same unit and uses as catalyst acid ion exchange resins. A (Jennings and Binning, 1960) temperature of 100ºC to 200ºC is maintained in the feed zone and the pressure is kept in order to maintain the water in liquid phase. BASF Aktiengesellschaft Process to produce optically pure alkyl D- or L-lactates by reaction of calcium lactate with an alcohol in the presence of a strong acid; the water (Bott et al., 1986) present in the reaction mixture or formed during the esterification is separated off by azeotropic distillation with the aid of an entraining agent. Battelle Memorial Institute Batch process for the preparation of esters of lactic acid directly from ammonium lactate and an alcohol. In this method the use of CO2 as a (Walkup et al., 1991; Walkup catalyst is required and the preferred range for the reaction mixture temperature is from 100ºC to 200ºC. A yield of lactate of about 75% is et al., 1993) reported. Recovery of high purity lactate ester from fermentation broth containing ammonium lactate or other basic salt of lactic acid; acidifying in the E. I. Du Pont de Nemours presence of an alcohol using continues addition of sulphuric acid or other strong acid and crystallizing to precipitate out some or all of the basic and Company salt of the strong acid; simultaneously or sequentially removing water while also esterifying the lactic acid with the alcohol to form impure (Cockrem and Johnson, 1993) lactate ester; removing the crystals formed; distilling the lactate ester to remove impurities. Musashino Chemical Method for producing a lactic ester by microorganic fermentation of lactic acid with a simple apparatus. A pressure in the range of 100 to Laboratory Ltd. 760 mmHg and a temperature of about 130ºC are recommended for this process. (Akira et al., 1994) BASF Aktiengesellschaft Process for the synthesis of lactates by fermentation of sugars mixtures, conversion of the lactic acid obtained during the fermentation to its (Sterzel et al., 1995) salts, followed by esterification. In this process lactic acid is esterified with ethanol in the presence of a catalyst such as p-toluenesulfonic acid. The catalyst, water and DAICEL CHEM IND LTD unreacted ethanol are removed to give a solution (A). The solution A is neutralized with (B) a solution of an alkali metal salt in an alcohol and (Yukio, 1996) distilled to give the ethyl lactate. Rectification process to produce ethyl lactate from lactic acid and ethanol. This patented process includes technological steps, such as, (Feng et al., 1996) determination of lactic acid, ethanol, sulphuric acid and benzene amounts, catalytic reaction, rectification for dewatering, neutralization and reduced distillation. Yield rate up to over 90% is reported. Process for the synthesis of high purity ethyl lactate and other lactate esters from carbohydrate feedstock. This process consists in a reactor Argonne National coupled with a pervaporation membrane unit for water removal and followed by separation of the reaction mixture in two consecutive Laboratory distillation columns. Alternatively, the reactor is followed by a plurality of pervaporation steps. It is reported a conversion greater than 99 %, (Rathin and Shih-Perng, 1998) for an initial ethanol/lactic acid molar ratio of 2:1, a reaction mixture temperature of 95ºC, a permeate-side vacuum pressure less than 0.5 mbar and as catalyst an ion-exchange resin, Amberlyst XN-1010, at 10% of lactic acid weight. Mitsubishi Gas Chemical Process to produce lactates from acetaldehyde and formate; The method described is characterized by the fact that there is no formation of Company, Inc. ammonium salts as by-products as in the case of the conventional techniques. (Abe et al., 1998) PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 15

A.E. Staley Manufacturing Process for the simultaneously production of an organic acid and of an ester of the organic acid. A mixture of ammonium salt of an organic acid Co. with alcohol is rapidly heating in order to produce a liquid stream containing acid, ester, and unreacted ammonium salt. (Cockrem, 2001) Process where a solution of carboxylic acid in a solvent and an alcohol are fed to a simulated moving bed reactor (SMBR), that contains a Eastman Chemical Company solid(s) (adsorber/catalyst) to produce two streams, one comprising a solution of the ester of the carboxylic acid and the alcohol and other (Arumugam et al., 2003) comprising the solvent. The process is particularly valuable for the preparation of an alkanol solution of an alkyl 2-keto-L-gulonate ester (AKLG). The simulated moving bed reactor is maintained at a temperature of 30 to 60ºC and a pressure of 3.5 to 20 bars. Cargill, Incorporated Techniques for processing lactic acid/lactate salt mixtures obtained from fermentation broths. These techniques generally concern the provision (Eyal et al., 2003; Eyal et al., of separated lactic acid and lactate streams from the mixtures. In this patent preferred methods of separation and processing of each of the 2006) streams are provided. Process to produce an ester that comprises the following steps: (a) feeding to a first vessel a mixture of organic acid, alcohol, and water, where A.E. Staley Manufacturing the organic acid and alcohol react to form monomeric ester and water (temperature of 150 to 220ºC) (b) feeding the mixture obtained in to a Co. second vessel (temperature range 30 to 100ºC), where are produced a vapour stream, that comprises alcohol, ester and water, and a liquid one, (Cockrem, 2003) that can be recycled to the first vessel. Process to produce esters in a chromatographic reactor in which the heterogeneous solid phase acts both as catalyst and as a means exhibiting La Chemical SpA preferential adsorption towards one of the reaction products (typically water). This process is particularly improved compared with the (Ruggieri et al., 2003) conventional technology since for regenerating the catalyst, it is used a desorbent mixed with a second compound, normally the anhydride of the acid used in the esterification reaction, which, by chemical reaction, completes the removal of the water adsorbed. Disclose a technique that uses ammonium lactate as raw material to make ethyl lactate by rectifying. The adopted equipment includes rectifying (Xueming and Jing, 2003) tower, condenser on the top of tower and oil-water separator. It uses metal halide as catalyst, an initial molar ratio of lactic acid and ethanol being 1:1 and benzene being 30%-50% of lactic acid ammonium weight. Process that relates to a continuous method to produce ethyl lactate from the esterification between lactic acid and ethanol in the presence of a catalyst (H SO 98%) ; this method consists in continuously extracting a mixture comprising ethyl lactate, ethanol, water and different heavy Arkema (FR) 2 4 products from the reaction medium at partial lactic acid conversion rate and, then, fed the mixture to a reduced-pressure flash separation, (Tretjak et al., 2006; Tretjak producing an overhead stream containing a mixture of ethyl lactate, ethanol and water, that is subjected to a fractional distillation column. A and Teissier, 2004) purity higher than 94.6 % of ethyl lactate is reported for an initial ethanol/lactic acid molar ratio equal to 2.5; esterification carried out at 80ºC; flash separation at 85ºC and 50 mbar, and fractional distillation at a column bottom temperature of 155°C and top temperature of 77.2ºC. Continuous method for preparing ethyl lactate which consists in reacting lactic acid with ethanol (ethanol/lactic acid molar ratio higher than Arkema (FR) 2.5) in the presence of a catalyst (H SO 98%) at a reflux of the reaction medium of about 100ºC under pressure ranging between 1.5 to 3 bars. (Martino-Gauchi and Teissier, 2 4 This method is characterized by the continuous extraction of a near-azeotropic water/ethanol gas mixture from the esterification reaction 2004; Martino-Gauchi and medium, followed by dehydration of this mixture using molecular sieves and recuperation from the dehydration mixture an ethanol gas stream Teissier, 2007) capable of being recycled to the esterification reaction medium and a flow consisting of water and ethanol which is fed to a distillation column. Lactic acid esterification by continuous countercurrent reactive distillation with alcohols, especially ethanol (to produce ethyl lactate). In this Board of Trustees of invention recycle of dimmers and trimmers and other oligomers of lactic acid are provided in order to improve yields. For absolute ethanol fed Michigan State near to the bottom of the column at 82ºC and lactic acid solution (85 wt % in water) fed near to the top of column at 25ºC (molar ratio of (Miller et al., 2006) ethanol to lactic acid of 3.3) it is reported a lactic acid conversion of 83% and a ethyl lactate yield of 82%. Method for preparing a lactic acid ester composition based on a lactic acid composition involving two steps: (a) transforming of the Roquette Freres composition into a lactic acid oligomeric composition; (b) mixing and reacting the oligomeric composition with an alcohol, in the presence of a (Fuertes et al., 2008) transesterification catalyst, to esterify all or part of the lactic acid contained in the oligomeric composition. This invention also discloses the use of ethyl lactate as solvent for preparing gelified compositions.

The esterification reaction is usually catalyzed by a strong acid, being most common sulphuric acid. However, the use of solid acid catalysts, as ion exchange resins, is also mentioned. In order to overcome equilibrium limitations excess of one reactant is commonly applied, normally the alcohol. Another technique is the use of a solvent, as benzene, substantially immiscible with water in order to extract the ester. Reaction and separation are, in almost all patented processes, separated steps, being distillation the most used separation technology.

2.3.3 Reactive Separations

In the last years, chemicals, petrochemicals and pharmaceuticals industries have been gone through a permanently increasing interest in the development of hybrid processes combining reaction and separation mechanisms into a single, integrated operation known as ‘reactive separation’. The combination of the two stages into a single unit brings important advantages, such as energy and capital cost reductions, increased yield and removal of some thermodynamic restrictions, e. g. azeotropes. A variety of separation principles and concepts can be incorporated into a reactor, see Figure 2.5.

Figure 2.5 Separation functions integrated into a reactor.

Important examples of reactive separations are reactive distillation, reactive absorption, reactive extraction or reactive membrane separation. Until now, such processes have had industrial application, mainly in areas like the homogeneously catalysed synthesis of acetates and the heterogeneously catalysed production of fuel additives. The potential is much wider; however, optimal functioning depends on careful process design, with appropriately selected PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 17

column internals, feed locations and catalyst placement. Greater understanding of the general and particular features of the process behaviour is equally essential.

2.3.3.1 Reactive Distillation (RD)

Reactive distillation (RD) is an unit operation that combines chemical reaction and distillation within a single vessel, thereby reducing equipment and recycle costs. A typical RD column is shown in Figure 2.6. Other advantages offered by reactive distillation include high selectivity, reduced energy uses, and reduction or elimination of solvents (Malone and Doherty, 2000). It is an effective method that has considerable potential for carrying out equilibrium-limited reactions, such as esterification and ester hydrolysis reactions; conversion can be increased far beyond chemical equilibrium conversion due to the continuous removal of reaction products from the reactive zone.

A Reactive B Section

Figure 2.6 Typical Reactive Distillation Column ( Tb,C < T b,B < T b,A < T b,D ).

Reactive distillation has received much attention in the last years (Sundmacher and Kienle, 2002; Taylor and Krishna, 2000; Tsai et al., 2008). It has been used for the esterification of fatty acids (Dimian et al., 2008; Steinigeweg and Gmehling, 2003), as well as being devised as a new method to clean industrial water from acetic acid (Bianchi et al., 2003). The RD technology was, also, applied on the ethyl lactate synthesis, first by Asthana and collaborators (Asthana et al., 2005), where it is reported higher lactic acid conversions (>95 %) and good ethyl lactate yields (>85 %), and, more recently, by Gao and co-workers (Gao et al., 2007).

18 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

However, applications of this technology in industry are still limited to a few reactive systems, mainly etherification (e.g. MTBE), esterification (e.g. methyl acetate), and alkylation (e.g. ethylbenzene or cumene) (Tuchlenski et al., 2001).

The production of methyl acetate is a classic example of successful RD (Agreda and Partin, 1984; Agreda et al., 1990). Conventional processes use one or more liquid-phase reactors with large excess of one reactant in order to achieve high conversions of the other. A typical flow sheet of a conventional process for the methyl acetate production is shown in Figure 2.7 in which the reaction section is followed by eight distillation columns, one liquid-liquid extractor and a decanter. This process requires a large capital investment, high energy costs and a large inventory of solvents. In the reactive distillation process for methyl acetate, the entire process is carried out in a single unit (see Figure 2.7), which represents one-fifth of the capital investment of the conventional process and consumes only one-fifth of the energy (Krishna, 2002).

Figure 2.7 Task-integrated methyl acetate column is much simpler than conventional plant (Stankiewicz and Moulijn, 2000).

In spite of all advantages of the RD technology, there are still some constraints and difficulties in its implementation, mainly due to volatility limitations. In order to maintain high concentrations of reactants and low concentrations of products in the reaction zone, the reactants and products must have suitable volatility. Also, it is necessary that both products have different boiling points to ensure the separation. The major disadvantage of RD

PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 19

technology, for exothermic reactions, is that chemical reaction has to show significant conversion at distillation temperature (Pöpken et al., 2000).

2.3.3.2 Simulated Moving Bed Reactor (SMBR)

Simulated Moving Bed (SMB) systems are used in industry for separations that are either impossible or difficult using traditional techniques. This technology uses differences in the adsorptivity of the different components involved rather than differences in their volatility, being an interesting alternative to distillation when the species involved exhibit small volatility differences, are non-volatile or are sensitive to temperature, as in the case of many fine chemical and pharmaceutical applications.

The combination of SMB and chemical reaction has been, in the last years, a subject of considerable attention in the scientific research, being this integrated reaction-separation technology called Simulated Moving Bed Reactor (SMBR). A schematic diagram of a SMBR unit is presented in Figure 2.8 where a reaction of type A+B↔C+D is considered, for the case of D being more adsorbed than C. The SMBR consists of a set of columns connected in series that are packed with a solid, which acts as both adsorbent and catalyst. Typically, there are two inlets (feed and desorbent) and two outlets (extract and raffinate). The component A is used as reactant and desorbent, therefore it is introduced in the system in the feed and desorbent streams. The other reactant B is used as feed. The products D and C are collected in the extract and the raffinate, respectively, since D is more adsorbed than C. At regular time intervals, called switching time period, all streams are switched for one bed distance in direction of the fluid flow. A cycle is completed when the number of switches is equal to a multiple of the columns number. In this way, the countercurrent motion of the solid is simulated with a velocity equal to the length of a column divided by the switching time. According to the position of the inlet and outlet stream the unit can be divided in four sections. In section I, positioned between the desorbent and extract nodes, the adsorbent is regenerated by desorption of the more strongly adsorbed product (D) from the solid. In section II (between the extract and feed node) and section III (between the feed and raffinate node) the reaction is taking place and products (C and D) are formed. The more strongly adsorbed product D is adsorbed and transported with the solid phase to the extract port. The less strongly adsorbed product C is desorbed and transported with the liquid in direction of the raffinate port. In section IV, positioned between the raffinate and desorbent node, the desorbent is regenerated before being recycled to section I.

20 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

Desorbent (A) Raffinate (A+C)

12 11 10 9 8 7 Direction of fluid flow A + B Ù C + D and port switching

1 2 3 4 5 6

Extract Feed (A+D) (A+B)

Figure 2.8 Scheme of a Simulated Moving Bed Reactor (SMBR).

The simultaneous reaction and separation in a SMBR has shown that considerable improvements in some processes performance can be achieved, as for example, on the synthesis of acetals (Pereira et al., 2008; Rodrigues and Silva, 2005; Silva and Rodrigues, 2005), fructose (Azevedo and Rodrigues, 2001; Da Silva et al., 2005; Zhang et al., 2004), lactosucrose (Kawase et al., 2001; Pilgrim et al., 2006), methylacetate (Lode et al., 2003) and MTBE (Zhang et al., 2001). In the last years, cation exchange resins are being widely used as catalyst of esterifications and acetalizations. Moreover, those resins adsorb selectively water, a by-product of those kind of reactions. Therefore, combining these two properties of acidic resins, and knowing that esterifications are reversible reactions, chromatographic reactors appear as promising technologies; in particular the SMBR, since the products are continuously separated and removed from the reaction medium, leading to complete conversion. This has motivated several studies on esterification reactions by means of the SMBR technology; examples are the esterification of acetic acid with methanol (Lode et al., 2003; Yu et al., 2003), ethanol (Mazzotti et al., 1996) and β-phenethyl alcohol (Kawase et al., 1996), and the esterification of acrylic acid with methanol to form methyl acrylate (Ströhlein et al., 2006).

Although the SMBR technology allows 100 % of conversion with 100 % of recovery of the desired product, a further step is necessary to separate the product from the raffinate mixture, and to recover the reactant A, used as desorbent, from both extract and raffinate streams, in order to recycle it to the SMBR unit.

PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 21

2.3.3.3 Pervaporation Membrane Reactor (PVMR)

Pervaporation is one of the membrane processes that can be employed for the separation of liquid mixtures that are difficult or not possible to separate by conventional methods, such as distillation. One example is the application of pervaporation to break the azeotrope of the mixture ethanol-water, where it is much more economical to use pervaporation or vapour permeation than other conventional methods (see Table 2.4).

Table 2.4 Dehydration costs of ethanol from 99.4 to 99.9 vol % by different methods (Drioli and Romano, 2001).

Molecular Vapor Entrainer Pervaporation Sieve Utilities Permeation Distillation ($/ton) Adsorption ($/ton) ($/ton) ($/ton)

vapor - 12.8 120.0 80.0

electricity 40.0 17.6 8.0 5.2

cooling water 4.0 4.0 15.0 10.0

Entrainer - - 9.6 -

Replacement of membranes and 19.0 30.6 - 50.0 molecular sieves

total costs 63.0 65.0 152.6 145.2

The pervaporation process has significant separation potential for various types of solutions, being specially suited for organic-water and organic-organic separations (Feng and Huang, 1996; Fleming and Slater, 1992; Huang and Rhim, 1991; Neel, 1991; Neel, 1995). It is used to separate a liquid mixture by partly vaporizing it through a nonporous permselective membrane, as shown in Figure 2.9. The feed liquid mixture is allowed to flow along one side of the membrane, and a fraction of it, the “permeate”, is recovered in the vapour state on the other side of the membrane, by means of vacuum or sweep gas. The mass transport through the membrane is induced by maintaining a low vapour pressure on the permeate side, eliminating thereby the effect of osmotic pressure. The permeate stream, enriched in the most

22 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

permeating component, might be then condensed in order to recover it. The remaining feed that does not permeate through the membrane, called the “retentate”, is depleted in the permeating component (Neel, 1995).

FeedLiquid Retentate

Pervaporation membrane

Vapor Permeate

Figure 2.9 Schematic representation of the pervaporation process.

There are a number of reviews on pervaporation processes (Dutta et al., 1996; Song et al., 2004; Van Hoof et al., 2004) and also on pervaporation membrane reactors (PVMRs) (Lim et al., 2002; Lipnizki et al., 1999; Waldburger and Widmer, 1996), since its application to equilibrium-limited reactions improves conversion by selectively removing one reaction product e.g. (Benedict et al., 2006; Castanheiro et al., 2006; David et al., 1991; Domingues et al., 1999; Lauterbach and Kreis, 2006; Peters et al., 2005a; Peters et al., 2005b; Sanz and Gmehling, 2006; Tanaka et al., 2002). PVMRs are, therefore, a type of membrane reactors that combines chemical reaction and separation by pervaporation, which is usually implemented by two different semi-batch processes: (i) the pervaporation unit (PV) is coupled to the reactor, i.e., the PV unit is an external process unit (see Figure 2.10a); and (ii) the reactor and the membrane are integrated in the same unit (see Figure 2.10b).

Although there is a recent interest in PVMRs, its discovery goes back to 1960, according to the first patent for a continuous process that integrates reaction and pervaporation in the same unit applied to an esterification reaction catalyzed by an acid ion exchange resin and using water selective membrane in order to remove it from the reaction zone and therefore enhancing the conversion (Jennings and Binning, 1960). Also, in 1986, another process was patented for the acetic acid esterification reaction with ethanol (Pearce, 1986), which reports complete conversion of the acetic acid by using a pervaporation membrane reactor consisting of two half-cells with a flat membrane disk (commercial PVA or Nafion) placed in the middle, as shown in Figure 2.11.

PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 23

(a) (b)

Membrane

Figure 2.10 Layout of a Semi-batch Pervaporation Membrane Reactor (SBPVMR): (a) external pervaporation unit; (b) membrane and reactor in the same unit.

Figure 2.11 Experimental set-up for the pervaporation membrane reactor (Pearce, 1986).

Even though PVMRs are finding broad uses, esterifications appear to be a key application (Marcano and Tsotsis, 2002). The esterification reactions are a typical example of equilibrium-limited reaction that produces by-product water. Considering a catalytic esterification reaction scheme of the type:

H + A + BCD←⎯⎯⎯⎯→ + where C is the desired ester product and D is the by-product (water). Due to the thermodynamic equilibrium limitation of the esterification reaction, a conventional reactor

24 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

will operate at low conversion; however, if a membrane is integrated in the reactor, as shown in Figure 2.12, wherein the water is removed through the permselective membrane from the reaction zone to the other side of the membrane, the reaction will proceed in the forward direction and therefore high conversion is expected to be attained in a reasonably short time.

Permeate (D)

Membrane Feed Retentate (A+B) (C)

Figure 2.12 Schematic representation of a membrane reactor for by-product withdrawal in a reversible reaction.

Zhu and collaborators studied the esterification reaction of acetic acid with ethanol both by experiment and simulation in a continuous-flow PVMR using a polymeric/ceramic composite membrane (Zhu et al., 1996). The same reaction was studied in a continuous tube membrane reactor (Waldburger and Widmer, 1996). This reaction was also studied in a PVMR, housing the reactor and membrane in the same unit, applying a zeolite T-membrane since it is stable under acidic conditions (Tanaka et al., 2001). Nafion tubular membranes, which also act as catalyst, were applied for the esterification of acetic acid with methanol and n-butanol, where the equilibrium conversions of 73 % and 70 % were increased to 77 % and 95 %, respectively (Bagnell et al., 1993). The esterification of acetic acid with butanol was more significantly improved than with methanol, due to the higher membrane selectivity towards water in butanol/water system. This particular esterification was also studied using Zr(SO4)·4H2O as catalyst and using cross-linked polyvinyl alcohol (PVA) membranes (Liu et al., 2001; Liu and Chen, 2002). Experiments and simulations were conducted to investigate the effects of several operating parameters, such as reaction temperature, initial molar ratio of acetic acid to n-butanol, ratio of the membrane area to the reacting mixture volume and catalyst concentration.

Regarding the operating modes of PVMR, semi-batch esterification process coupled by pervaporation is the most used (Xuehui and Lefu, 2001), being applied for the synthesis of ethyl tert-butyl ether (ETBE) from tert-butyl alcohol (TBA) and ethanol (Kiatkittipong et al., 2002); and applied for the esterification of acetic acid with isopropanol leading to conversions higher than 90 % (Sanz and Gmehling, 2006). The esterification of lactic acid and succinic

PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 25

acid with ethanol were studied in semi-batch well-mixed reactors coupled by pervaporation (Benedict et al., 2006), using two different pervaporation membranes, a GFT-1005 membrane (Deutsche Carbone AG) and a T-1b (TexSep 1) membrane (Texaco Research), both with organic acid compatibility and very high water permeation selectivity. The conversion of lactic acid obtained in the PVMR was 71 %, being the usual equilibrium value of 55 %. Based on the synthesis of methyl acetate from methanol and acetic acid catalysed by Amberlyst-15 and using a polyvinyl alcohol (PVA) membrane, three different configurations of PVMRs were compared (Assabumrungrat et al., 2003): (i) semi-batch (SBPVMR), (ii) plug-flow (PFPVMR) and (iii) continuous stirred tank (CSPVMR): Simulations, carried out using the experimental determined kinetic and permeation parameters, conclude that the PFPVMR is the most favourable mode, although there are some conditions (at low values of Damkohler number (0.5 and 1)) where CSPVMR is superior to PFPVMR. A new concept of a hybrid PVMR system, which integrates the pervaporation step through a membrane with adsorption in the permeate side, proved to enhance in 5 % the conversion reported in the PVMR in the absence of the adsorbent (Park and Tsotsis, 2004). Based on the concept of catalytic membranes (Bagnell et al., 1993), the performance of a composite catalytic membrane was examined for the esterification of acetic acid and butanol aiming to develop a continuous composite catalytic pervaporation membrane reactor (Peters et al., 2005a). It was proved by simulations that the outlet conversion for the catalytic pervaporation-assisted esterification reaction exceeds the conversion of a conventional pervaporation membrane reactor, with the same loading of catalyst dispersed in the liquid bulk. At the same conditions, a conversion of 85 % for the catalytic pervaporation-assisted esterification reactor is reported against 79 % for a conventional pervaporation membrane reactor.

The scientific literature on PVMRs is abundant, since it is an area where significant activity is under way and many advances are expected in the future, due to the many advantages of PVMRs: the simultaneous removal of a product from the reactor enhances the conversion; undesired side reactions can be suppressed; high conversion is possible at almost stoichiometric feed flow rates and the heat of reaction can be used for separation. Therefore, lower energy consumption and higher product yields make of the pervaporation membrane reactor an interesting alternative to conventional processes. However, no large-scale industrial applications have been reported yet. The main reason for this should be related to factors as small fluxes of the desired species and mechanical and thermal stability of the

26 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

membranes. Further developments in the field of materials engineering will certainly change this picture.

2.4 References

U.S. Department of Energy. "Environmental Benefits." Retrieved June, 2009, http://www1.eere.energy.gov/biomass/environmental.html.

Argonne National Laboratory. "Ethyl Lactate Solvents: Low-Costs and Environmental Friendly." Retrieved January 4, 2006, http://www.techtransfer.anl.gov/techtour/ethyllactate.html.

AAE Chemie. "Solvent Information." Retrieved June, 2009, http://www.aaechemie.com/html/aae_solvent_information.html.

Energetics Incorporated, "Industrial Bioproducts: Today and Tomorrow. ", U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, Office of the Biomass Program, Washington, D.C., 2003.

Abe T., Toshiyuki, T. Uchiyama, H. Higuchi, Y. Shima and K. Ikemoto, "Process for preparing lactate", U.S. Patent No. 5824818 (1998).

Agreda V. H. and L. R. Partin, "Reactive distillation process for the production of methyl acetate", U.S. Patent No. 4435595 (1984).

Agreda V. H., L. R. Partin and W. H. Heise, "High-purity methyl acetate via reactive distillation", Chem. Eng. Prog. 86(2): 40-46, 1990.

Akira k., A. Tomohiro, Y. Masaaki and M. Shigendon, "Method for production of lactic acid and lactic esters", CA Patent No. 2 116 710 (1994).

Anastas P. T. and J. C. Warner, "Green Chemistry: Theory and Practice", Oxford Science Publications, New York: 1-135, 1998.

Arumugam B. K., L. W. Blair, B. W. Boyd, N. A. Collins, D. A. Larkin, S. T. Perri and C. W. Sink, "Preparation of esters of carboxylic acids ", U.S. Patent No. 6518454 (2003).

Assabumrungrat S., J. Phongpatthanapanich, P. Praserthdam, T. Tagawa and S. Goto, "Theoretical study on the synthesis of methyl acetate from methanol and acetic acid in pervaporation membrane reactors: Effect of continuous-flow modes", Chem. Eng. J. 95(1): 57-65, 2003.

Asthana N., A. Kolah, D. T. Vu, C. T. Lira and D. J. Miller, "A continuous reactive separation process for ethyl lactate formation", Org. Process Res. Dev. 9(5): 599-607, 2005.

Azevedo D. C. S. and A. E. Rodrigues, "Design methodology and operation of a simulated moving bed reactor for the inversion of sucrose and glucose-fructose separation", Chem. Eng. J. 82(1-3): 95-107, 2001.

PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 27

Bagnell L., K. Cavell, A. M. Hodges, A. W. H. Mau and A. J. Seen, "The use of catalytically active pervaporation membranes in esterification reactions to simultaneously increase product yield, membrane permselectivity and flux", J. Membrane Sci. 85(3): 291-299, 1993.

Bannister W. J., "Esterification Process", U.S. Patent No. 2 029 694 (1936).

Benedict D. J., S. J. Parulekar and S. P. Tsai, "Pervaporation-assisted esterification of lactic and succinic acids with downstream ester recovery", J. Membrane Sci. 281(1-2): 435-445, 2006.

Bianchi C. L., V. Ragaini, C. Pirola and G. Carvoli, "A new method to clean industrial water from acetic acid via esterification", Appl. Catal., B: Environ. 40(2): 93-99, 2003.

Bott K., W. Boehm, G. Fritz, C. Martin and H. Siegel, "Process for the preparation of lactic- acid esters", EP Patent No. 0174624 (1986).

Castanheiro J. E., A. M. Ramos, I. M. Fonseca and J. Vital, "Esterification of acetic acid by isoamylic alcohol over catalytic membranes of poly(vinyl alcohol) containing sulfonic acid groups", Appl. Catal., A: gen 311(1-2): 17-23, 2006.

Cockrem M. C. M., "Process for production of organic acids and esters thereof ", U.S. Patent No. 6291708 (2001).

Cockrem M. C. M., "Process for production of esters ", U.S. Patent No. 6664413 (2003).

Cockrem M. C. M. and P. D. Johnson, "Recovery of lactate esters and lactic acid from fermentation broth", US Patent No. 5210296 (1993).

Corma Canos A., S. Iborra and A. Velty, "Chemical routes for the transformation of biomass into chemicals", Chem. Rev. 107(6): 2411-2502, 2007.

Da Silva E. A. B., D. P. Souza, A. A. U. De Souza, S. M. A. G. U. Souza and A. E. Rodrigues, "Analysis of the behavior of the simulated moving bed reactor in the sucrose inversion process", Sep. Sci. Technol. 40(12): 2373-2389, 2005.

David M. O., T. Q. Nguyen and J. Neel, "Pervaporation-esterification coupling. Part II. Modelling of the influence of different operating parameters", Chem. Eng. Res. Des. 69(4): 341-346, 1991.

Dhanuka V. R., V. C. Malshe and S. B. Chandalia, "Kinetics of the liquid phase esterification of carboxylic acids with alcohols in the presence of acid catalysts: Re-interpretation of published data", Chem. Eng. Sci. 32(5): 551-556, 1977.

Dimian A. C., C. S. Bildea, F. Omota and A. A. Kiss, "Innovative process for fatty acid esters by dual reactive distillation", Comput. Chem. Eng., 2008.

Domingues L., F. Recasens and M. A. Larrayoz, "Studies of a pervaporation reactor: Kinetics and equilibrium shift in benzyl alcohol acetylation", Chem. Eng. Sci. 54(10): 1461-1465, 1999.

Drioli E. and M. Romano, "Progress and new perspectives on integrated membrane operations for sustainable industrial growth", Ind. Eng. Chem. Res. 40(5): 1277-1300, 2001.

28 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

Dutta B. K., W. Ji and S. K. Sikdar, "Pervaporation: Principles and applications", Sep. Purif. Methods 25(2): 131-224, 1996.

Eyal A. M., J. N. Starr, R. Fisher, B. Hazan, R. Canari, D. R. Witzke, P. R. Gruber and J. J. Kolstad, "Lactic acid processing; methods; arrangements; and, products ", U.S. Patent No. 6534679 (2003).

Eyal A. M., J. N. Starr, R. Fisher, B. Hazan, R. Canari, D. R. Witzke, P. R. Gruber, J. J. Kolstad and X. Meng, "Lactic acid processing; methods; arrangements; and, products ", U.S. Patent No. 7144977 (2006).

Feng X., D. Yu and X. Liu, "Rectifying process for preparing ethyl lactate ", CN Patent No. 1114305 (1996).

Feng X. and R. Y. M. Huang, "Studies of a membrane reactor: Esterification facilitated by pervaporation", Chem. Eng. Sci. 51(20): 4673-4679, 1996.

Filachione E. M. and C. H. Fisher, "Production of esters ", U.S. Patent No. 2 565 487 (1951).

Fleming H. L. and C. S. Slater, in Membrane Handbook, W. S. W. Ho, K. K. Sirkar, Ed.; Van Nostrand Reinhold: New York, Chapter 7-10, (1992).

Fuertes P., R. Tamion, G. Fleche and S. Comini, "Method for preparing a lactic acid ester composition and use thereof as solvent ", U.S. Patent No. 7342050 (2008).

Gao J., X. M. Zhao, L. Y. Zhou and Z. H. Huang, "Investigation of ethyl lactate reactive distillation process", Chem. Eng. Res. Des. 85(4 A): 525-529, 2007.

Holten C. H., "Lactic acid", Verlag Chemie: Copenhagen, Denmark, (1971).

Huang L., Y. Zhu, H. Zheng, M. Du and Y. Li, "Vapor-phase hydrogenolysis of biomass- derived lactate to 1,2-propanediol over supported metal catalysts", Appl. Catal., A: gen 349(1-2): 204-211, 2008.

Huang R. Y. M. and J. W. Rhim, "Separation Characteristics of Pervaporation Membrane Separation Processes", in Pervaporation Membrane Separation Processes; R. Y. M. Huang, Ed.; Elsevier: RhimAmsterdam, The Netherlands, 111-180, (1991).

Jennings J. F. and R. C. Binning, "Organic Chemical Reactions Involving Liberation Of Water", U.S. Patent No. 2956070 (1960).

Kamm B. and M. Kamm, "Biorefinery - systems", Chem. Biochem. Eng. Q. 18(1): 1-6, 2004a.

Kamm B. and M. Kamm, "Principles of biorefineries", Appl. Microbiol. Biotechnol. 64(2): 137-145, 2004b.

Kawase M., A. Pilgrim, T. Araki and K. Hashimoto, "Lactosucrose production using a simulated moving bed reactor", Chem. Eng. Sci. 56(2): 453-458, 2001.

PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 29

Kawase M., T. B. Suzuki, K. Inoue, K. Yoshimoto and K. Hashimoto, "Increased esterification conversion by application of the simulated moving-bed reactor", Chem. Eng. Sci. 51(11): 2971-2976, 1996.

Kiatkittipong W., S. Assabumrungrat, P. Praserthdam and S. Goto, "A pervaporation membrane reactor for liquid phase synthesis of ethyl tert-butyl ether from tert-butyl alcohol and ethanol", J. Chem. Eng. Jpn. 35(6): 547-556, 2002.

Krishna R., "Reactive separations: More ways to skin a cat", Chem. Eng. Sci. 57(9): 1491- 1504, 2002.

Lauterbach S. and P. Kreis, "Experimental and theoretical investigation of a pervaporation membrane reactor for a heterogeneously catalysed esterification", Desalination 199(1-3): 418-420, 2006.

Lichtenthaler F. W., "Towards improving the utility of ketoses as organic raw materials", Carbohydr. Res. 313(2): 69-89, 1998.

Lichtenthaler F. W., "Unsaturated O- and N-heterocycles from carbohydrate feedstocks", Acc. Chem. Res. 35(9): 728-737, 2002.

Lichtenthaler F. W. and S. Peters, "Carbohydrates as green raw materials for the chemical industry", Comptes Rendus Chimie 7(2): 65-90, 2004.

Lim S. Y., B. Park, F. Hung, M. Sahimi and T. T. Tsotsis, "Design issues of pervaporation membrane reactors for esterification", Chem. Eng. Sci. 57(22-23): 4933-4946, 2002.

Lipnizki F., R. W. Field and P. K. Ten, "Pervaporation-based hybrid process: A review of process design, applications and economics", J. Membrane Sci. 153(2): 183-210, 1999.

Liu Q., Z. Zhang and H. Chen, "Study on the coupling of esterification with pervaporation", J. Membrane Sci. 182(1-2): 173-181, 2001.

Liu Q. L. and H. F. Chen, "Modelling of esterification of acetic acid with n-butanol in the presence of Zr(SO4)2·4H2O coupled pervaporation", J. Membr. Sci. 196: 171-178, 2002.

Lode F., G. Francesconi, M. Mazzotti and M. Morbidelli, "Synthesis of methylacetate in a simulated moving-bed reactor: Experiments and modeling", AlChE J. 49(6): 1516-1524, 2003.

Malone M. F. and M. F. Doherty, "Reactive distillation", Ind. Eng. Chem. Res. 39(11): 3953- 3957, 2000.

Marcano J. G. S. and T. T. Tsotsis, "Catalytic Membranes and Membrane Reactors", Wiley VCH, (2002).

Martino-Gauchi G. and R. Teissier, "Continues method for preparing ethyl lactate", WO Patent No. 2004052826 (2004).

Martino-Gauchi G. and R. Teissier, "Continues method for preparing ethyl lactate", U.S. Patent No. 7297809 A2 (2007).

30 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

Mazzotti M., A. Kruglov, B. Neri, D. Gelosa and M. Morbidelli, "A continuous chromatographic reactor: SMBR", Chem. Eng. Sci. 51(10): 1827-1836, 1996.

Miller D. J., N. Asthana, A. Kolah and C. T. Lira, "Process for production of organic acid esters ", US Patent No. 2006/0014977 (2006).

Muse J. and H. A. Colvin, "Use of ethyl lactate as an excipient for pharmaceutical compositions", US Patent No. 2005/0287179 A1 (2005).

Neel J., "Pervaporation", in Pervaporation Membrane Separation Processes; R. Y. M. Huang, Ed.; Elsevier: Amsterdam, The Netherlands, 1-109, (1991).

Neel J., "Pervaporation", in Membrane Separations Technology, Principles and Applications; R. D. Noble, S. A. Stern, Ed.; Elsevier: Amsterdam, The Netherlands, 143-211, (1995).

Park B. G. and T. T. Tsotsis, "Models and experiments with pervaporation membrane reactors integrated with an adsorbent system", Chem. Eng. Proc.: Proc. Intens. 43(9): 1171-1180, 2004.

Pearce G. K., "Esterification process", EP Patent No. 0210055 A1 (1986).

Pereira C. S. M., P. S. Gomes, G. K. Gandi, V. M. T. M. Silva and A. E. Rodrigues, "Multifunctional Reactor for the Synthesis of Dimethylacetal", Ind. Eng. Chem. Res. 47(10): 3515-3524, 2008.

Peters T. A., N. E. Benes and J. T. F. Keurentjes, "Zeolite-coated ceramic pervaporation membranes; pervaporation- esterification coupling and reactor evaluation", Ind. Eng. Chem. Res. 44(25): 9490-9496, 2005a.

Peters T. A., J. Van Der Tuin, C. Houssin, M. A. G. Vorstman, N. E. Benes, Z. A. E. P. Vroon, A. Holmen and J. T. F. Keurentjes, "Preparation of zeolite-coated pervaporation membranes for the integration of reaction and separation", Catal. Today 104(2-4): 288-295, 2005b.

Pilgrim A., M. Kawase, F. Matsuda and K. Miura, "Modeling of the simulated moving-bed reactor for the enzyme-catalyzed production of lactosucrose", Chem. Eng. Sci. 61(2): 353- 362, 2006.

Pöpken T., L. Götze and J. Gmehling, "Reaction kinetics and chemical equilibrium of homogeneously and heterogeneously catalyzed acetic acid esterification with methanol and methyl acetate hydrolysis", Ind. Eng. Chem. Res. 39(7): 2601-2611, 2000.

Rass-Hansen J., H. Falsig, B. JÃ¸rgensen and C. H. Christensen, "Bioethanol: Fuel or feedstock?", J. Chem. Technol. Biotechnol. 82(4): 329-333, 2007.

Rathin D. and T. Shih-Perng, "Esterification of fermentation-derived acids via pervaporation", US Patent No. 5723639 (1998).

Reisch M., "Solvent users look to replace NMP", Chemical and Engineering News 86(29): 32, 2008.

REN21 (2009). Renewables Global Status Report: 2009 Update.

PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 31

Rodrigues A. E. and V. M. T. M. Silva, "Industrial process for acetals production in a simulated moving bed reactor", Patent PT 103123, 2004; WO 2005/113476A1 (2005).

Röper H., "Renewable raw materials in Europe - Industrial utilisation of starch and sugar", Starch/Staerke 54(3-4): 89, 2002.

Ruggieri R., G. Ranghino, G. Carvoli, A. Tricella, D. Gelosa and M. Morbidelli, "Process for esterification in a chromatographic reactor", U.S. Patent No. 6586609 (2003).

Sagar A. D. and S. Kartha (2007). Bioenergy and sustainable development? Annual Review of Environment and Resources. 32: 131-167.

Sanz M. T. and J. Gmehling, "Esterification of acetic acid with isopropanol coupled with pervaporation. Part II. Study of a pervaporation reactor", Chem. Eng. J. 123(1-2): 9-14, 2006.

Silva V. M. T. M. and A. E. Rodrigues, "Novel process for diethylacetal synthesis", AlChE J. 51(10): 2752-2768, 2005.

Song K. H., K. R. Lee and J. M. Rim, "Pervaporation of esters with hydrophobic membrane", Korean J. Chem. Eng. 21(3): 693-698, 2004.

Stankiewicz A. I. and J. A. Moulijn, "Process intensification: Transforming chemical engineering", Chem. Eng. Prog. 96(1): 22-33, 2000.

Steinigeweg S. and J. Gmehling, "Esterification of a fatty acid by reactive distillation", Ind. Eng. Chem. Res. 42(15): 3612-3619, 2003.

Sterzel H., H. Vogel, H. Exner, D. Kratz and M. Brudermuller, "Preparation of lactates ", U.S. Patent No. 5453365 (1995).

Ströhlein G., Y. Assunção, N. Dube, A. Bardow, M. Mazzotti and M. Morbidelli, "Esterification of acrylic acid with methanol by reactive chromatography: Experiments and simulations", Chem. Eng. Sci. 61(16): 5296-5306, 2006.

Sundmacher K. and A. Kienle, "Reactive Distillation: Status and Future Directions", Willey- VCH, Weinheim (2002).

Tanaka K., R. Yoshikawa, C. Ying, H. Kita and K. I. Okamoto, "Application of zeolite membranes to esterification reactions", Catal. Today 67(1-3): 121-125, 2001.

Taylor R. and R. Krishna, "Modelling reactive distillation", Chem. Eng. Sci. 55(22): 5183- 5229, 2000.

Tilman D., J. Hill and C. Lehman, "Carbon-negative biofuels from low-input high-diversity grassland biomass", Science 314(5805): 1598-1600, 2006.

Tretjak S., E. Burtin and R. Teissier, "Continuous ethyl lactate preparation method", U.S. Patent No. 2006/0041165 A1 (2006).

32 CHAPTER 2. State of the art on Green Solvent Ethyl Lactate

Tretjak S. and R. Teissier, "Continuous ethyl lactate preparation method", WO Patent No. 2004052825 (2004).

Tsai R. C., J. K. Cheng, H. L. Huang, C. C. Yu, Y. S. Shen and Y. T. Chen, "Design and control of the side reactor configuration for production of ethyl acetate", Ind. Eng. Chem. Res. 47(23): 9472-9484, 2008.

Tuchlenski A., A. Beckmann, D. Reusch, R. Düssel, U. Weidlich and R. Janowsky, "Reactive distillation-industrial applications, process design and scale-up", Chem. Eng. Sci. 56(2): 387- 394, 2001.

Van Hoof V., L. Van den Abeele, A. Buekenhoudt, C. Dotremont and R. Leysen, "Economic comparison between azeotropic distillation and different hybrid systems combining distillation with pervaporation for the dehydration of isopropanol", Sep. Purif. Technol. 37(1): 33-49, 2004.

Waldburger R. M. and F. Widmer, "Membrane reactors in chemical production processes and the application to the pervaporation-assisted esterification", Chem. Eng. Technol. 19(2): 117- 126, 1996.

Walkup P. C., C. A. Rohrmann, R. T. Hallen and D. E. Eakin, "Production of esters of lactic acid, esters of acrylic acid, lactic acid, and acrylic acid ", U.S. Patent No. 5071754 (1991).

Walkup P. C., C. A. Rohrmann, R. T. Hallen and D. E. Eakin, "Production of esters of lactic acid, esters of acrylic acid, lactic acid, and acrylic acid ", U.S. Patent No. 5252473 (1993).

Wang M., M. Wu, H. Huo and J. Liu, "Life-cycle energy use and greenhouse gas emission implications of Brazilian sugarcane ethanol simulated with the GREET model", International Sugar Journal 110(1317): 527-545, 2008.

Wasewar K. L., A. A. Yawalkar, J. A. Moulijn and V. G. Pangarkar, "Fermentation of glucose to lactic acid coupled with reactive extraction: A review", Ind. Eng. Chem. Res. 43(19): 5969-5982, 2004.

Weissermel K. and H. Arpe, "Industrial Organic Chemistry", 3rd ed., VCH Publishers Inc., New York (1997).

Xuehui L. and W. Lefu, "Kinetic model for an esterification process coupled by pervaporation", J. Membrane Sci. 186(1): 19-24, 2001.

Xueming Z. and G. Jing, "Process for preparing ethyl lactate by reation and distillation using ammonium lactate as raw material", CN Patent No. 1438213 (2003).

Yu W., K. Hidajat and A. K. Ray, "Modeling, Simulation, and Experimental Study of a Simulated Moving Bed Reactor for the Synthesis of Methyl Acetate Ester", Ind. Eng. Chem. Res. 42(26): 6743-6754, 2003.

Yukio Y., "Production of ethyl lactate", JP Patent No. 8208565 (1996).

Zhang Y., K. Hidajat and A. K. Ray, "Optimal design and operation of SMB bioreactor: Production of high fructose syrup by isomerization of glucose", Biochem. Eng. J. 21(2): 111- 121, 2004.

PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 33

Zhang Z., K. Hidajat and A. K. Ray, "Application of simulated countercurrent moving-bed chromatographic reactor for MTBE synthesis", Ind. Eng. Chem. Res. 40(23): 5305-5316, 2001.

Zhu Y., R. G. Minet and T. T. Tsotsis, "A continuous pervaporation membrane reactor for the study of esterification reactions using a composite polymeric/ceramic membrane", Chem. Eng. Sci. 51(17): 4103-4113, 1996.

3. Batch Reactor: Thermodynamic Equilibrium and Reaction Kinetics

Abstract. The heterogeneous catalysis of lactic acid (88 wt. %) esterification with ethanol in presence of Amberlyst 15-wet, was studied for catalyst loading of 1.2 wt. % to 3.9 wt. %, initial molar ratio of reactants of 1.1 to 2.8 and temperature from 50ºC to 90ºC.

In this work a methodology based in the UNIQUAC model was developed to determine the thermodynamic equilibrium constant since in literature there is inconsistency concerning the temperature dependence of the thermodynamic equilibrium constant. A simplified Langmuir- Hinshelwood kinetic model was used to describe the experimental data. The proposed rate

2 law is r = kc (aEth aLa − aEL aW / K ) (1+ K Eth aEth + KW aW ) ; the kinetic parameters are the

7 −1 −1 pre-exponential factor, kc,0 = 2.70×10 mol.g .min and the activation energy,

Ea = 49.98kJ / mol . The equilibrium reaction constant is K =19.35exp (− 515.13 /T (K)) with reaction enthalpy 4.28 kJ/mol. The model reasonably predicts the kinetic experimental data and it will be very useful to apply for the design and optimization of industrial hybrid reactive separation processes.

Adapted from: Pereira C. S. M., S. P. Pinho, V. M. T. M. Silva and A. E. Rodrigues, "Thermodynamic Equilibrium and Reaction Kinetics for the Esterification of Lactic Acid with Ethanol Catalyzed by acid ion exchange resin", Ind Eng Chem Res. 47: 1453-1463, 2008.

36 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

3.1 Introduction

The ethyl lactate synthesis comprises a liquid-phase reversible reaction between ethanol and lactic acid, wherein water is a sub-product:

+ Ethanol (Eth) + Lactic Acid (La) ←⎯→⎯H Ethyl Lactate (EL) +Water (W )

The lactic acid contains a hydroxyl group adjacent to the carboxylic acid and because of its bifunctional nature undergoes intermolecular esterification in aqueous solutions above 20 wt. % to form linear dimer and higher oligomer acids (Montgomery, 1952; Vu et al., 2005):

2La1 ⇔ La2 +W (lactic acid dimer formation)

La1 + La2 ⇔ La3 +W (lactic acid trimer formation)

…

La1 + Lan−1 ⇔ Lan +W (lactic acid oligomer formation) with n ≥ 2

where:

OH OH O OH O OH O O n

Lactic acid (La1) Lactic acid oligomers (Lan+1)

The use of lactic acid as reactant is then complicated, since the extent of the self-esterification increases with the increase of the acid concentration. A method to purify the lactic acid is through the esterification with lower alcohols, such as methanol, ethanol or butanol; and then the produced ester is separated and hydrolyzed back into pure lactic acid and the alcohol is recovered and reused (Troupe and DiMilla, 1957).

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 37

For the process of ethyl lactate production it will be desirable to use a high lactic acid concentration, with lower water content, in order to increase the yield of the reaction and the concentration of the ethyl lactate. However, this will imply the presence of lactic acid oligomers during the esterification, which will be converted into the corresponding esters:

La1 + Eth ⇔ EL1 +W (ethyl lactate formation)

La2 + Eth ⇔ EL2 +W (ethyl lactate dimer formation)

La3 + Eth ⇔ EL3 +W (ethyl lactate trimer formation)

…

Lan + Eth ⇔ ELn +W (ethyl lactate oligomer formation) where:

OH O OH O OC2H5 OC2H5 O n O

Ethyl lactate (EL1) Ethyl lactate oligomers (ELn+1)

For an aqueous solution with 88 wt. % of lactic acid, the molar percentage of the monomer

(La1), dimer (La2) and trimer (La3) are of about 43.5 mol %, 9.2 mol % and 1.8 mol %, respectively; and about 45 mol % of water. While a 20 wt. % aqueous solution of lactic acid is constituted only by monomer and water, being the monomer molar percentage of about 5.6 mol % (Asthana et al., 2006). However, the percentage of lactic acid and ethyl lactate oligomers is less then 5% at equilibrium (Asthana et al., 2006; Tanaka et al., 2002) and the use of an aqueous solution with a high lactic acid concentration is desirable to produce ethyl lactate at industrial scale by means of a continuous hybrid process as reactive pervaporation, reactive chromatographic processes, among others. The kinetics of ethyl lactate production has been studied since 1957 (Troupe and DiMilla, 1957), but lately has deserved more attention since it is a green solvent and an alternative to the traditionally petroleum derived solvents. Troupe and Dimilla (1957) studied the esterification reaction between lactic acid and ethanol using sulphuric acid as catalyst.

38 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

However, this kind of homogeneous catalysts may be the origin of a lot of problems, because of their miscibility with the reaction medium, which causes separation problems; in addition, strong acid catalysts lead to corrosion of the equipment. The replacement of homogeneous catalysts by heterogeneous catalysts is gaining importance due to their ecofriendly nature. Besides being non-corrosive and easy to separate from the reaction mixture, the heterogeneous catalyst can be used repeatedly over a prolonged period without any difficulty in handling and storage. Many solid-acid catalysts have been used, such as acid treated clays (Yadav and Krishnan, 1998), heteropolyacids (Dupont et al., 1995; Lacaze-Dufaure and Mouloungui, 2000; Schwegler et al., 1991; Yadav and Krishnan, 1998), iodine (Ramalinga et al., 2002), MCM-41 (Koster et al., 2001), zeolite-T membrane (Tanaka et al., 2002), smopex- 101 (Lilja et al., 2002; Mäki-Arvela et al., 1999), HY zeolite (Chen et al., 1989; Kirumakki et al., 2003; Ma et al., 1996), zeolite beta (Kirumakki et al., 2003) and ZSM-5 (Kirumakki et al., 2003; Ma et al., 1996; Wu and Chen, 2004; Yadav and Krishnan, 1998). However, ion- exchange resins are the most commonly used solid catalysts and they have been proved to be effective in liquid phase esterification (Benedict et al., 2003; Lee et al., 2002; Lee et al., 2000; Liu and Tan, 2001; Yadav and Mujeebur Rahuman, 2002; Zhang et al., 2004). Since the heterogeneous catalysis is clearly advantageous, some studies have been already performed for the esterification of lactic acid with ethanol. Zhang and co-workers studied the kinetic of esterification of lactic acid (20 wt. %) with ethanol catalyzed by five different cation-exchange resins (Zhang et al., 2004). They proposed a simplified mechanism based on Langmuir-Hinshelwood model to describe the kinetic behaviour. Delgado et al. (2007b) also investigated the esterification of lactic acid with ethanol and the hydrolysis of the ethyl lactate in the presence of a commercial cation-exchange resin; an aqueous lactic acid solution of 20 wt. % and a mechanism based on the Langmuir-Hinshelwood model to describe the kinetics was used. This esterification with/without a solid catalyst (Amberlyst XN-1010) was also investigated by Benedict and their collaborators. A kinetic model based in concentrations to describe the behaviour of the reaction between an 88 wt. % of lactic acid solution with ethanol was used (Benedict et al., 2003). The presence of oligomers was not mentioned in their work. In Tanaka and co-workers studies about this esterification, the oligomers presence was considered and the reactions were described by simple nth-order reversible rate expressions based on the species concentration (Tanaka et al., 2002). Three different solutions of lactic acid (20 wt. %, 50 wt. % and 88 wt. %) were used in the esterification reaction with ethanol in Asthana’s and collaborators work (2006). The oligomers presence

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 39 was also taken into account and a similar model based on the species concentration was used to describe the reaction kinetics. In spite of the number of kinetic studies available in the literature, the thermodynamic equilibrium of the reaction in the liquid phase has not been clearly studied, and some of the authors do not report the values of the equilibrium constant (Zhang et al., 2004). In some works (Delgado et al., 2007b; Troupe and DiMilla, 1957) different equilibrium constants based in concentration (Kx) at the same temperature are reported. The authors of those studies have concluded that the equilibrium constant Kx vary significantly with the initial molar ratio of the reactants, being less sensitive to the temperature. However, the thermodynamic equilibrium constant defined as function of the species liquid activities, which is only temperature dependent, is not presented in their works. In order to overcome the lack of thermodynamic data, Delgado and co-authors have studied the vapor-liquid reactive equilibria for the ethyl lactate synthesis, and they have proposed the following expression to describe the reaction equilibrium constant (Delgado et al., 2007a): 2431.2 ln (K) = 7.893− (3.1) T (K) Nevertheless, they have found some difficulties in measuring the vapor phase composition, that affect significantly the experimental values of the experimental equilibrium constant, leading to high deviations between experimental values and those predicted by Equation 3.1, as shown in Figure 3.1.

2.50 experimental [26] ln K = 7.893 - 2431.2 / T (K) 2.00

1.50 ln (K) ln 1.00

0.50

0.00 0.00265 0.0027 0.00275 0.0028 0.00285 1 / T (K) Figure 3.1 Representation of experimental values of ln K as function of 1/T. (Data collected from Delgado et al. (2007b)).

40 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

This work was undertaken to obtain the reaction equilibrium and kinetic data for the synthesis of the ethyl lactate in the liquid phase, avoiding the vaporization of all species by working at 6 bar (helium pressurization). The pressure influence in the value of equilibrium constant is negligible for this system for the temperature and pressure operating range, as it can be estimated by the correction factor K P (Smith and Van Ness, 1987). Therefore the thermodynamic equilibrium constant was estimated by the UNIQUAC method and the esterification reaction catalyzed by the ion-exchange resin Amberlyst 15-wet was described by a simple activity based kinetic model, which will be applied in the modelling of some reactive separation processes, such as membrane reactors. In this work a high lactic acid concentration was used with the objective of maximizing the ethyl lactate productivity for an industrial process. The presence of oligomers was neglected, since at equilibrium the total amount of lactic acid and ethyl lactate oligomers represents less than 5 % according to Tanaka et al. (2002) and Asthana et al. (2006) studies. However, in the final section of this paper the presence of oligomers will be addressed.

3.2 Experimental Section

3.2.1 Chemicals and Catalyst

The chemicals used were ethanol (>99.9% in water), lactic acid (>85% in water) and ethyl lactate (>98% in water) from Sigma-Aldrich (U.K.). A commercial strong-acid ion-exchange resin named Amberlyst 15-wet (Rohm & Haas) was used as catalyst and adsorbent. This resin is a bead-form macroreticular polymer of styrene and divinylbenzene, with particle diameter varying between 0.3 to 1.2 mm, an ion exchange capacity of 4.7 meq H+/g of dry resin and inner surface area of 53 m2/g. According to Ihm et al. (Ihm et al., 1988) only 4 % of the active sites are located at the macropores (surface of the microspheres) and the others 96 % are inside gel polymer microspheres. Since the water adsorbed on the catalyst surface decreases the reaction kinetics, because it is one of the reaction products, it was necessary to guarantee anhydrous resin. For that, the resin was washed several times with deionised water and dried at 90ºC until the mass remains constant. This method is of value for the reuse experiments, since the water is abundantly present due to the lactic acid solution.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 41

3.2.2 Experimental set-up

The experiments were carried out in a glass-jacketed 1 dm3 autoclave (Büchi, Switzerland), operating in a batch mode, mechanically stirred at 600 rpm, equipped with pressure and temperature sensors and with a blow-off valve (Figure 3.2). The temperature was controlled by thermostated ethylene glycol/water solution (Lauda, Germany) that flows through the jacket of the reactor and feed vessel. To maintain the reacting mixture in liquid phase over the whole temperature range, the pressure was set at 0.6MPa with helium. The lactic acid solution is charged into the reactor and heated to the desired reaction temperature. The dry catalyst is placed in a basket at the top of the stirrer shaft. Ethanol is heated up to the desired temperature into the feed vessel and then charged to the reactor opening the on/off valve. The agitation is immediately turned on and the basket of catalyst falls down in the reactant solution. This time is considered to be the starting time of the esterification reaction. One of the outlets of the reactor was connected directly to a liquid sampling valve (Valco, USA), which injects 0.2 µl of pressurised liquid to a gas chromatograph.

vent BV

FV He PM NV vent PT vacuun M TT V1 NV BR

GC He

Figure 3.2 Experimental set-up for kinetic studies. BR-batch reactor; FV-feed vessel; M motor; TT-temperature sensor; PT-pressure sensor; PM-manometer; BV- blow-off valve; V1-sampling valve; V2-injection valve; NV-needle valve; GC-gas chromatograph; TB-thermostatic bath.

42 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

3.2.3 Analytical method

All the samples were analysed in a gas chromatograph (Chrompack 9100, Netherlands) using a fused silica capillary column (Chrompack CP-Wax 57 CB, 25 m x 0.53 mm ID, df = 2.0 μm) to separate the compounds and a thermal conductivity detector (TCD 903 A) to quantify them. The column temperature was programmed with a 1.5 min initial hold at 110ºC, followed by a 50ºC/min ramp up to 190ºC and held for 8.5 min. The injector and detector temperature were maintained at 280ºC and 300ºC, respectively. Helium N50 was used as the carrier gas with flowrate 10.50 ml/min. In order to analyze the lactic acid and ethyl lactate oligomers at equilibrium a HPLC system from Gilson (France) using an ICSep ION-300 column held at 20ºC was used. A 0.0085 N H2SO4 solution was used as mobile phase (0.4 ml / min) and species were quantified by a refractive index detector.

3.3 Thermodynamic Equilibrium Results

The experiments to measure the equilibrium constant were done in a temperature range of 323-363 K. At each temperature different experiments were performed using different initial molar ratios ( REth / La = 1.0 to REth / La = 2.8) and different mass of catalyst (2.3 wt. % to 6.0 wt. %) (see Table 3.1). All the experiments lasted long enough to ensure that the equilibrium was reached.

3.3.1 Thermodynamic equilibrium constant

In its most general form the chemical equilibrium constant ( K ) for a reaction is given by:

⎛ ΔGD ⎞ K = exp⎜− ⎟ = aν i (3.2) ⎜ ⎟ ∏ i ⎝ RT ⎠ i where ΔGD is the reaction standard free Gibbs energy, R the ideal gas constant, T the absolute temperature, ai is the activity of species i, and ν i its stoichiometric coefficient in the reaction.

For the esterification reaction, occurring in the liquid phase at low pressure:

aEL aW xEL xW γ ELγ W K = = = K x Kγ (3.3) aEth aLa xEth xLa γ Ethγ La

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 43

being x and γ the mole fraction and the activity coefficient of each species, respectively.

The thermodynamic chemical equilibrium constant is only temperature dependent. From the experimental point of view, however, changing isothermically the initial mass of each reactant will give, most probably, different values for the equilibrium constant. Naturally, that is a consequence of experimental errors as well as deficiencies in the thermodynamic models used to calculate the activity coefficients. For example, as can be observed in Table 3.1 three different runs were carried out at 323.15 K. Estimating the activities coefficients by the UNIFAC method (Fredenslund et al., 1977), using the relative molecular volume, surface area and the interaction parameters presented in literature (Reid et al., 1987); the resulting equilibrium constants are quite different; 4.126, 3.813, and 4.637, respectively, from top to bottom.

Table 3.1 Conditions of the experiments performed to measure the thermodynamic equilibrium constant.

Catalyst Initial number of moles Equilibrium number of moles loading Lactic Lactic Ethyl T (K) (wt. %) Ethanol acid Water Ethanol acid Lactate Water 323.15 5.67 0.8154 0.4051 0.2968 0.5459 0.1356 0.2695 0.5662 323.15 5.95 0.7329 0.4051 0.2966 0.4786 0.1508 0.2543 0.5509 323.15 5.63 0.4033 0.4051 0.2957 0.1983 0.2000 0.2051 0.5008 333.15 5.67 0.8154 0.4051 0.2968 0.5403 0.1300 0.2751 0.5719 333.15 5.95 0.7329 0.4051 0.2966 0.4744 0.1465 0.2586 0.5551 333.15 5.63 0.4033 0.4051 0.2957 0.1963 0.1981 0.2070 0.5027 343.15 5.67 0.8154 0.4051 0.2968 0.5351 0.1249 0.2802 0.5770 343.15 5.95 0.7329 0.4051 0.2966 0.4730 0.1452 0.2599 0.5565 343.15 5.63 0.4033 0.4051 0.2957 0.1969 0.1986 0.2065 0.5022 343.41 2.82 5.6988 2.0197 1.4837 4.1506 0.4716 1.5481 3.0319 343.41 2.82 5.6988 2.0197 1.4837 4.1361 0.4570 1.5627 3.0464 353.15 2.37 3.9033 3.4444 2.5154 1.9555 1.4966 1.9477 4.4632 362.87 2.82 5.6988 2.0197 1.4837 4.1173 0.4382 1.5815 3.0652 362.87 2.82 5.6988 2.0197 1.4837 4.1029 0.4239 1.5958 3.0796

One possibility to get a single equilibrium constant value for each temperature would be to fit all the equilibrium constants using an equation of the type:

44 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

ln K = a + b T (3.4)

This, however, presents a big deficiency. In fact, using a unique value for the equilibrium constant at a given temperature, the equilibrium composition must be recalculated for each specific initial condition, which immediately will introduce changes in the magnitude of the activity coefficients.

Assuming a given value for the equilibrium constant at 323.15 K, solving Equation 3.3 in order to the equilibrium composition involves an iterative procedure. First, all the activity coefficients must be assumed equal to one, obtaining the equilibrium composition of an ideal solution. After the activity coefficients can be determined using a model such as UNIQUAC or UNIFAC, and a new equilibrium composition can now be calculated. The procedure is repeated until convergence. It must be stressed that this final equilibrium composition will certainly not be the same as the one presented in Table 3.1, which was used to calculate the equilibrium constant and to regress the coefficients in Equation 3.4. So this inconsistency must be avoided.

To overcome this problem, in this work it is suggested to obtain the coefficients in Equation 3.4 that allow the calculation of the equilibrium composition as closer as possible to that observed experimentally. Therefore, the coefficients were estimated minimizing the following objective function (Fob):

2 ⎛ X exp − X calc ⎞ Fob = ⎜ k k ⎟ (3.5) ∑⎜ exp ⎟ k ⎝ X k ⎠

exp calc where X k and X k are the experimental and the calculated equilibrium conversion for experiment k, respectively.

3.3.1.1 Activity coefficients estimation

In this work, instead of using the UNIFAC method it was preferred to apply the UNIQUAC model. Indeed, there are some available experimental vapor-liquid equilibrium data involving mixtures of species involved in the reaction under study, which makes preferable to use a correlation model instead of a pure predictive method.

Initially the parameters between ethanol and ethyl lactate were estimated based on the data published by Peña-Tejedor et al. ( 2005) and Vu et al. (2006). Following, the parameters

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 45

between water and ethyl lactate were obtained based on the data by Vu et al. ( 2006). Finally the parameters between lactic acid and all other species were estimated using data from the quaternary system measured by Delgado et al. (2007a). The parameters were estimated minimizing the following objective function (Fob):

2 ⎛ γ exp − γ calc ⎞ Fob = ∑∑⎜ i i ⎟ (3.6) ⎜ γ exp ⎟ ji⎝ i ⎠ j where i is the species and j the experimental data point. It should be mentioned that the parameters between water and ethanol were found in the DECHEMA books (Gmehling et al., 1981). The interaction parameters are given in Table 3.2.

Table 3.2 UNIQUAC interaction parameters (K). Ethyl Ethanol Water Lactic acid lactate Ethanol 0.0000 -152.319 -17.554 + 0.2797 T (K)* -35.008 Ethyl 264.990 0.0000 207.789 -20.986 Lactate Water -21.987 + 0.2276 T (K)* -13.093 0.0000 -99.183 Lactic Acid 33.741 219.89 213.19 0.0000 *DECHEMA (Gmehling et al., 1981) The average relative deviation found for the activity coefficients were 12.9 %, but special difficulties were found when describing the behaviour of diluted solutions, which was never the case when the chemical equilibrium experimental studies were carried out in this work.

3.3.2 Equilibrium constant and reaction enthalpy for the synthesis of Ethyl Lactate

Using the data found experimentally for the chemical equilibrium compositions in 14 different runs, at 4 different temperatures in the range between 323.15 and 362.87 K, it was found the following relation for the equilibrium constant:

515.13 ln(K )=− 2.9625 (3.7) TK()

Using this result and the iterative procedure described before, the deviations found between the experimental and calculated equilibrium compositions for all experiments are given in Table 3.3. It can be seen that the higher deviation found between the experimental and calculated equilibrium composition is of about 5.6 %.

46 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

Table 3.3 Experimental and calculated equilibrium compositions for all the experiments performed and the correspondent deviation percent. Equilibrium composition (mole fraction) average T (K) Experimental Calculated deviation Ethyl Lactic Ethyl Lactic (%) Ethanol Water Ethanol Water lactate acid lactate acid 323.15 0.3598 0.1776 0.3732 0.0894 0.3561 0.1813 0.3769 0.0857 2.06 323.15 0.3336 0.1773 0.384 0.1051 0.3264 0.1845 0.3913 0.0978 3.77 323.15 0.1812 0.1854 0.4527 0.1808 0.1931 0.1734 0.4407 0.1928 5.58 333.15 0.3561 0.1813 0.3769 0.0857 0.3528 0.1846 0.3802 0.0824 1.87 333.15 0.3307 0.1802 0.3869 0.1021 0.3228 0.1881 0.3948 0.0943 4.11 333.15 0.1794 0.1871 0.4544 0.179 0.1881 0.1784 0.4457 0.1878 4.08 343.15 0.3527 0.1847 0.3803 0.0823 0.3498 0.1876 0.3832 0.0794 1.67 343.15 0.3297 0.1812 0.3879 0.1012 0.3194 0.1915 0.3982 0.0909 5.41 343.15 0.1799 0.1866 0.4539 0.1796 0.1835 0.1831 0.4504 0.183 1.64 343.41 0.451 0.1682 0.3295 0.0512 0.451 0.1683 0.3295 0.0512 0.01 343.41 0.4495 0.1698 0.3311 0.0497 0.451 0.1683 0.3295 0.0512 1.18 353.40 0.1983 0.1975 0.4525 0.1517 0.202 0.1937 0.4488 0.1555 1.78 362.87 0.4474 0.1719 0.3331 0.0476 0.4474 0.1718 0.3331 0.0477 0.07 362.87 0.4459 0.1734 0.3347 0.0461 0.4474 0.1718 0.3331 0.0477 1.30

In Table 3.4 the activity coefficients for the equilibrium composition and the thermodynamic equilibrium constant are presented. In terms of equilibrium conversion the average relative deviation found was 2.60%, being possible to find an average reaction enthalpy of 4.28 kJ·mol-1 in that temperature interval. Some authors (Zhang et al., 2004) have also indicated the reaction is slightly endothermic, but uncertainty is high. In fact using the values found for the standard state enthalpy of formation in the DIPPR 801 database (DIPPR, 1998) (see Table 3.5) it is possible to calculate -20.97 ± 186.7 kJ·mol-1 for the reaction enthalpy at 298.15 K. This considerable high error (± 186.7 kJ·mol-1) is mainly due to the high uncertainty on the ethyl lactate standard state enthalpy of formation. Nevertheless the value found is in good agreement to values given in the literature (Benedict et al., 2003; Zhang et al., 2004).

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 47

Table 3.4 Activity coefficients for the equilibrium composition and the thermodynamic equilibrium constant. Activity coefficients T (K) Lactic Ethyl K Ethanol Water acid lactate 323.15 1.0563 1.2391 1.3557 1.6928 3.9291 323.15 1.0563 1.2632 1.3941 1.6629 3.9291 323.15 1.0263 1.3238 1.7011 1.5288 3.9291 333.15 1.0652 1.2397 1.3386 1.6847 4.1217 333.15 1.0660 1.2643 1.3749 1.6547 4.1217 333.15 1.0398 1.3285 1.6644 1.5200 4.1217 343.15 1.0737 1.2396 1.3222 1.6771 4.3116 343.15 1.0755 1.2649 1.3564 1.6467 4.3116 343.15 1.0531 1.3325 1.6300 1.5117 4.3116 343.41 1.0607 1.1543 1.2338 1.7832 4.3165 343.41 1.0607 1.1543 1.2338 1.7832 4.3165 353.40 1.0749 1.3350 1.5295 1.5267 4.5035 362.87 1.0707 1.1511 1.2125 1.7715 4.6781 362.87 1.0707 1.1511 1.2125 1.7715 4.6781

Table 3.5 Standard state enthalpy of formation of the different species (Gmehling et al., 1981). Ethanol Lactic acid Ethyl lactate Water 0 -1 ΔH f (kJ·mol ) -234.950 -682.960 -695.084 -241.814 Error < 1 % < 10 % < 25 % < 0.2 %

3.3.3 Application of this methodology to other works

The equilibrium compositions and equilibrium conversions of other studies (Benedict et al., 2003; Delgado et al., 2007b; Troupe and DiMilla, 1957) were predicted by the UNIQUAC model and the equilibrium constant proposed in this work. It should be noticed that those equilibrium compositions were not used in the estimation of the UNIQUAC parameters nor in the estimation of chemical equilibrium constant presented by Equation 3.7. The comparison

48 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

between the experimental and calculated equilibrium compositions is shown in Figure 3.3.a for ethanol and water; and Figure 3.3.b for lactic acid and ethyl lactate.

1.0 Molar Fraction

0.8

0.6 calc x

0.4

Ethanol 0.2 Water

0.0 0.0 0.2 0.4 0.6 0.8 1.0 xexp Figure 3.3.a Experimental and calculated equilibrium molar fraction of ethanol and water species.

0.20 Molar Fraction

0.15

calc 0.10 x

Ethyl Lactate 0.05 Lactic Acid

0.00 0.00 0.05 0.10 0.15 0.20

xexp

Figure 3.3.b Experimental and calculated equilibrium molar fraction of ethyl lactate and lactic acid species.

The equilibrium compositions are very well predicted for ethanol and water; while for the ethyl lactate and lactic acid deviations are higher, but even so satisfactory. The larger deviations can be due to the fact that the interaction parameters used in UNIQUAC model for those particular species are not based in a large and consistent set of experimental vapour

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 49

liquid equilibrium (VLE) data as is the case for water and ethanol species. Furthermore the composition range is now much restricted (maximum mole fraction around 0.20), for which the model presents more difficulties to calculate accurately the activity coefficients and the analytical method for these species is not as precise as it is for water and ethanol compounds.

In terms of the deviation of experimental and calculated equilibrium conversion (Figure 3.4), it can be seen that when 85 wt. % lactic acid feed is used, the model prediction is quite good. In the remaining cases, for 44 wt. % lactic acid and mainly for 20 wt. % lactic acid solution the deviations are more significant. This is maybe due to the fact that diluted solutions in lactic acid and, obviously, in ethyl lactate are used, and like mentioned before the UNIQUAC model does not describe with the desired accuracy the behaviour of that kind of solutions. If possible it would be preferable to use activity coefficients at infinite dilution.

1.0 Equilibrium conversion

0.8

0.6 calc X 0.4

La 20 wt.% (Delgado et al., 2007a) La 44 wt.% (Troupe and Dimilla, 1957) 0.2 La 85 wt.% (Troupe and Dimilla, 2003) La 85 wt.% (Benedict et al., 2003) La 85 wt.% (This work) 0.0 0.00.20.40.60.81.0 Xexp

Figure 3.4 Experimental and calculated equilibrium conversion. (Data collected from Troupe and Dimilla (1957), Benedict et al. (2003), Delgado et al. (2007) and this work).

3.4 Kinetic Studies

The experimental results of the reaction kinetics of the esterification of lactic acid and ethanol catalyzed by the Amberlyst 15-wet resin are presented in this section. The effect of various conditions, such as, catalyst loading, initial molar ratio between ethanol and lactic acid and reaction temperature on lactic acid conversion as function of time is studied. This study was

50 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

performed varying the condition under evaluation and keeping constant the remaining conditions, in absence of mass transfer limitations and catalyst deactivation as shown by the preliminary studies performed.

3.4.1 Preliminary Studies

3.4.1.1 Evaluation of external mass transfer limitations (effect of stirring speed)

To quantify the influence of external mass transfer resistance preliminary experiments at different stirring speed were run using a molar ratio of ethanol to lactic acid of 1.82 at 324.18 K and 2.4 wt. % of Amberlyst 15-wet with particle diameter 0.5 < dp < 0.6 mm. Figure 3.5 shows the conversion of lactic acid as a function of time at different stirring speed. With a stirring speed of 600 rpm, there is no limitation due to external resistance, so all further experiments were done at 600 rpm.

0.6

0.5

0.4

0.3 600 rpm

Conversion 0.2 800 rpm

0.1

0 0 200 400 600 800 Time (min) Figure 3.5 Experimental data obtained at different stirrer speeds for a molar ratio of ethanol to lactic acid of 1.82 at 324.18 K using 2.4 wt. % of Amberlyst 15-wet as catalyst with particle diameter 0.5 < dp < 0.6 mm.

3.4.1.2 Evaluation of internal mass transfer limitations (effect of particle size)

The Amberlyst 15wet was separated by particle size and three classes with different diameters were obtained: 0.425mm

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 51

has an average diameter of 0.685 mm, on the conversion of lactic acid; it can be seen that there are no significant internal diffusion limitations for the experiments performed. Therefore, the unsieved resin was used for the following kinetic experiments performed in this work. This is in agreement with several works performed with this kind of resin and type of reaction (Delgado et al., 2007b; Liu and Tan, 2001; Pöpken et al., 2000; Zhang et al., 2004).

0.6

0.5

0.4

0.3 unsieved resin 0.6

Conversion 0.2 0.5

0 0 200 400 600 800 Time (min) Figure 3.6 Effect of catalyst particle size on the conversion of lactic acid history for a molar ratio of ethanol to lactic acid of 1.82 at 324.18 K using 2.4 wt. % of catalyst and stirrer speed of 600 rpm.

3.4.1.3 Evaluation of catalyst deactivation (effect of catalyst reusability)

The Amberlyst 15-wet catalyst reusability was studied at 344.05 K. The resin was reused up to three times. First, a reaction was carried out using fresh catalyst. Then the catalyst used was separated from the reaction mixture by filtration, washed several times with deionised water and dried at 90ºC until the mass remained constant and charged again to the reactor. A new esterification reaction was performed using the same conditions that the first one and so on. The lactic acid conversion as a function of time was analyzed and the three conversion histories are presented in Figure 3.7. As it can be seen no significant changes were observed. The resin activity was kept the same in the three runs. However, each experiment was performed using fresh Amberlyst 15-wet catalyst. Some authors (Dixit and Yadav, 1996) studied the reusability of the Amberlyst 15-wet in the alkylation reaction of o-xylene with styrene and they observed a drastic reduction in the conversion of styrene due to the direct deposition of the by-products on the active sites and the loss of accessibility of the active sites

52 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

due to pore blockage. So, it can be concluded that the catalytic activity of the resins depends on its interaction with the reaction medium.

0.7

0.6

0.5

0.4 fresh catalyst 0.3 2nd use

Conversion 3rd use 0.2

0.1

0 0 200 400 600 800 1000 1200 Time (min) Figure 3.7 Effect of the Amberlyst 15-wet reusability on the conversion of lactic acid history at 344.05 K for a molar ratio of ethanol to lactic acid of 1.82 using 2.4 wt. % of Amberlyst 15-wet catalyst with an average particle diameter of 0.685 mm and stirrer speed of 600 rpm.

3.4.2 Kinetic Model

The esterification reactions have been described using different models, such as, pseudo- homogeneous, Eley-Rideal and Langmuir-Hinshelwood model (L-H model) (Lee et al., 2002; Lilja et al., 2005; Pöpken et al., 2000; Sanz et al., 2002; Teo and Saha, 2004). However, the L-H model has been considered the most appropriate model for the reaction between lactic acid and ethanol (Delgado et al., 2007b; Zhang et al., 2004). Therefore, the model developed is based on an L-H mechanism and it considers the following steps:

Ö Ethanol and Lactic acid adsorption:

Eth + S ←⎯→K s⎯, Eth Eth.S

La + S ←⎯→Ks⎯, La La.S

Ö Surface reaction between the adsorbed species of ethanol and lactic acid:

K1 Eth.S + La.S ⇔ EL.S + W.S

Ö desorption of ethyl lactate and water:

* EL.S ←⎯→Ks⎯, EL EL + S

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 53

* W .S ←⎯→K⎯s , W W + S

Surface reaction is assumed to be the controlling step, with the other steps remaining in equilibrium. Multi-component Langmuir adsorption isotherms, written in terms of activities, are assumed to describe the adsorption behaviour of the compounds of the reaction mixture in the surface of the resin. Taking into account the above considerations the following rate expression, written in terms of activities of components, due to the non-ideality of the reaction mixture, is obtained:

a a a a − EL W Eth La K r = k c 2 (3.8) ⎛ W ⎞ ⎜1 + ∑ K s,i ai ⎟ ⎝ i= Eth ⎠

where kc is the kinetic constant, K s, i is the adsorption constant for species i and K is the equilibrium reaction constant.

In order to reduce the number of optimization parameters it was taken into account only those components that had the strongest adsorption. It was considered that the most polar molecules, water and ethanol, have the strongest adsorption strength on the Amberlyst 15-wet surface, being therefore neglected the adsorption of lactic acid and ethyl lactate. This consideration is corroborated by several works of adsorption in similar resins (Mazzotti et al., 1997; Sanz et al., 2002; Silva and Rodrigues, 2002; Zhang et al., 2004). Thus, the simplified rate expression used to describe the experimental data is:

a a a a − EL W Eth La K r = kc 2 (3.9) (1+ K s,Eth aEth + K s,W aW )

In this kinetic model (Equation 3.9) there are three parameters to be estimated, at each temperature, the kinetic constant (kc) and the two adsorption parameters (Ks,Eth and Ks,W), instead of five if the rate Equation 3.8 was used to describe the experimental data.

The temperature dependence of kinetic constant was fitted with the Arrhenius equation:

⎛ Ea ⎞ kC = k 0, C exp ⎜− ⎟ (3.10) ⎝ RT ⎠

where Ea is the reaction activation energy, k0,c is the pre-exponential constant, R is the gas

54 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

constant and T is the temperature.

The temperature dependence of the adsorption equilibrium constants were fitted with:

⎛ ΔH s ⎞ K s = K 0, s exp⎜− ⎟ (3.11) ⎝ RT ⎠

where K 0,s is the constant for Equation 3.11 and ΔH S is the adsorption enthalpy.

3.4.2.1 Parameter estimation from experimental data

The mass balance in the batch reactor for a component i, in liquid phase, at constant temperature is given by:

d n i = v w r (3.12) d t i cat

where ni is the number of moles of component i, t is the time, wcat is the mass of catalyst and r is the reaction rate expressed in moles i / (mass of catalyst .min).

The number of moles of component i (ni ) as a function of the conversion X of the limiting reactant (l ) is:

⎛ X ⎞ n = n ⎜ R +ν ⎟ (3.13) i l,0 ⎜ i / l i ⎟ ⎝ ν l ⎠

where nl,0 and ν l are, respectively, the initial moles number and the stoichiometric coefficient of the limiting reactant and Ril/ is given by:

ni,0 Ri / l = (3.14) nl,0

where ni,0 is the initial number of moles of component i .

Introducing Equation 3.13 into Equation 3.12 we get:

d X ν w r = l cat (3.15) d t nl, 0 with the initial condition: t = 0; X = 0 .

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 55

The differential Equation 3.15 combined with the suggested rate expression, Equation 3.9, was solved numerically with the DASOLV integrator implemented in gPROMS-general PROcess Modelling System version: 3.0.3, which is a commercial package from Process Systems Enterprise. For all simulations a tolerance equal to 10-5 was fixed. The estimation of the unknown parameters ( Ea , k0,c , K 0,Eth , K 0,W , ΔH Eth and ΔHW ) was carried out using the “Parameter estimation in gProms” that attempts to determine the values for the parameters, in order to maximize the probability that the mathematical model will predict the values obtained from experiments. Assuming independent, normally distributed measurements errors,ε ik , with zero means and standard deviations, σ ik , this maximum likelihood goal can be captured through the following objective function:

' N 1 ⎪⎧ NE NM i ⎡ X − X ⎤⎪⎫ Φ = ln(2π ) + ln(σ 2 ) + ik ik (3.16) min ⎨∑∑⎢ ik 2 ⎥⎬ 2 2 θ ⎩⎪ i==11k ⎣ σ ik ⎦⎭⎪

Where, N is the total number of measurements taken during the experiments ( N = 290), θ is the set of model parameters to be estimated (θ = 6 ), NE is the number of experiments performed ( NE = 15 ), NM i is the number of measurements of the conversion in the ith

2 experiment, σ ik is the variance of the kth measurement of the conversion in experiment i

2 −3 ' (σ = 1.83×10 ) , X ik is the kth measured value of conversion in experiment i and, finally, X ik is the kth (model-) predicted value of conversion in experiment i. The quality of the model fit was tested through the mean relative deviation (MRD) between the calculated conversion values (Xcalc) and the experimental ones (Xexp) (see Equation 3.17).

1 ⎛ X − X ⎞ MRD = ⎜ calc exp ⎟*100% (3.17) N ⎜∑ X ⎟ ⎝ N exp ⎠

3.4.3 Modelling and discussion of results

The simplified L-H model (Equation 3.9) was used to describe the kinetic behaviour of the esterification of lactic acid with ethanol using Amberlyst 15-wet as catalyst. In order to obtain the unknown parameters at least two different experiments (different initial molar ratio) were performed for each temperature, which varied from 323.15 K to 363.15 K. The values of the

56 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

optimized parameters along with the MRD value are presented in Table 3.6. A value of MRD between experimental and calculated conversion of 6.8 % was obtained.

As mentioned before in section 3.4.2 the Arrhenius equation was used to fit the kinetic (kc) constant (Figure 3.8) and the Equation 3.11 was used to fit the adsorption constants (Ks,Eth and

Ks,W) which are given by KW =15.19exp(12.01/T(K))and K Eth =1.22exp(359.63/T(K)). From the temperature dependence of the kinetic constant the apparent activation energy was calculated being the correspondent value 49.98 kJ/mol.

Table 3.6 Parameters of the kinetic model expressed in terms of activities (Equation 3.9) and the mean relative deviation, MRD.

k0,c Ea ΔHW ΔH Eth -1. -1 -1 K0, W K0, Eth MRD (%) (mol.g min ) (kJ.mol ) (J.mol-1) (J.mol-1) 2.70×107 49.98 15.19 - 99.85 1.22 - 29.95×102 6.84

0.5

0 c k -0.5 ln ln

-1

-1.5

-2 2.7 2.8 2.9 3 3.1 3.2 1000/T (K-1)

Figure 3.8 Representation of experimental values of ln kc as function of 1/ T and linear fitting.

3.4.3.1 Effect of catalyst loading

The catalyst loading was varied from 1.2 wt. % to 3.9 wt. % (weight of catalyst/total weight of reaction mixture) keeping the rest of the experimental conditions similar. The effect of the catalyst loading on the esterification reaction between lactic acid and water is shown in Figure 3.9. As may be observed the reaction rate increases with the percentage of Amberlyst 15-wet. This was expected since the increase of catalyst implies an increase in the number of active sites available for the reaction.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 57

0.7

0.6

0.5

0.4

0.3 1.2 wt.% 2.4 wt.% Conversion 0.2 3.9 wt.% 0.1 L-H model 0 0 50 100 150 200 250 300 Time (min) Figure 3.9 Effect of catalyst loading on the conversion of lactic acid history for a molar ratio of ethanol to lactic acid of 1.82 at 353.49 K and using an average particle diameter of 0.685 mm and stirrer speed of 600 rpm.

3.4.3.2 Effect of initial molar ratio of reactants

To study the effect of the initial molar ratio of reactants (REth/La) on the conversion of lactic acid, this condition was varied from 1.1 to 2.8, as presented in Figure 3.10. It can be seen that the equilibrium conversion increases with the increase of the initial molar ratio of ethanol to lactic acid and that the equilibrium is achieved faster for larger initial molar ratio values.

0.9 0.8 0.7 0.6 0.5

0.4 REth/La = 1.1 Conversion 0.3 REth/La = 1.8 0.2 REth/La = 2.8 L-H model 0.1 0 0 500 1000 1500 2000 2500 Time (min) Figure 3.10 Effect of initial molar ratio of ethanol to lactic acid on the conversion of lactic acid history at 353.40 K and using 2.4 wt. % of Amberlyst 15- wet catalyst with an average particle diameter of 0.685 mm and stirrer speed of 600 rpm.

58 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

It may be observed from Figure 3.10 that the experiments are initially well predicted by the model (until about 120 minutes); and due to the methodology developed in this work, the equilibrium is also well described. However, there is a transient state (between 120 minutes till the equilibrium), where the model fails to describe the experiments, being more significant for higher values of initial molar ratio between ethanol and lactic acid.

3.4.3.3 Effect of reaction temperature

The effect of reaction temperature is shown in Figure 3.11. The reaction rate increases with the reaction temperature. The same effect is noticed on the equilibrium conversion. Once again, it can be noticed that in the transient state, the model predicts higher conversion of lactic acid values than those obtained experimentally.

0.8

0.7

0.6

0.5 T = 324.14 K T = 333.11 K 0.4 T = 344.05 K T = 353.40 K

Conversion 0.3 T = 363.50 K 0.2 L-H model

0.1

0 0 500 1000 1500 2000 2500 3000 Time (min)

Figure 3.11 Effect of the reaction temperature on the conversion of lactic acid history for a molar ratio of ethanol to lactic acid of 1.82 and using 2.4 wt. % of Amberlyst 15-wet catalyst.

3.4.3.4 Effect of Lactic acid and Ethyl Lactate oligomers

The proposed model shows a kinetic behaviour with two limiting situations:

a) first slope at the beginning of the experiment, corresponding to the initial reaction rate;

b) final plateau, corresponding to the reaction equilibrium.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 59

However, experimentally, it seems that there are three limiting situations, where the first and third are the same that the ones described by the proposed model, but the transient state is represented by a line with very small slope. This could be due to the presence of the lactic acid oligomers, and consequently ethyl lactate oligomers that where formed. To confirm this assumption an evaluative modelling study was made considering the oligomers presence using the kinetic model and the correspondent’s parameters reported in literature (Tanaka et al., 2002).

In Figure 3.12a the molar fractions of the ethyl lactate, lactic acid, ethanol and water as function of time are presented, being the correspondent’s oligomers presented in Figure 3.12b. Analysing the simulated kinetic curve for the ethyl lactate monomer, here denominated just as ethyl lactate, one can conclude a similar behaviour with the experimental results shown in this work, where there are three limiting steps:

a) initial conversion of lactic acid oligomers in ethyl lactate oligomers, but the kinetic rate is much higher for the monomers than for the dimers and trimers. Therefore the experiments are initially well predicted;

b) transient state, where the ethyl lactate monomer concentration increases slowly, since several reversible reactions are occurring and the equilibrium is shifted towards lactic acid monomer formation and consequently ethyl lactate monomer. In this transient step, firstly, the dimers concentrations of lactic acid and ethyl lactate decrease due to their hydrolysis, and then the same happens to the trimers that are converted into dimers and finally again into monomers, till the equilibrium is reached;

c) reaction equilibrium, when all the reversible reactions are in equilibrium, the total amount of lactic acid and ethyl lactate oligomers are less then 0.4%. Therefore, the experiments are well predicted at the equilibrium since oligomers presence is negligible.

60 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

0.7 Et hyl l act at e Et hanol 0.6 Lacti c aci d Wat er

0.5

0.4

0.3 Molar fraction Molar 0.2

0.1

0.0 0 1020304050480490500 Time (min) Figure 3.12a Molar fraction histories of the compounds: ethyl lactate monomer, ethanol,

lactic acid monomer and water. T = 363.15K and REth / La = 3 . The kinetic model and the parameters used were taken from Tanaka et al. (2002).

0.06 Ethyl lactate dimer Ethyl lactate trimer 0.05 Lactic acid dimer Lactic acid trimer 0.04

0.03

Molar fraction Molar 0.02

0.01

0.00 0 10 20 30 40 50 480 490 500 Time (min)

Figure 3.12b Molar fraction histories of the oligomers: ethyl lactate dimer, ethyl lactate trimer, lactic acid dimer and lactic acid trimer.

T = 363.15K and REth / La = 3 . The kinetic model and the parameters used were taken from Tanaka et al. (2002).

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 61

The transient behaviour, where the hydrolysis of dimers and trimers of lactic acid and ethyl lactate are dominant to produce the ethyl lactate monomer, is even more noticeable for higher values of initial molar ratio of ethanol/lactic acid, as it can be shown in Figure 3.13a. This is due to the fact that the excess of ethanol benefits more the ethyl lactate dimer formation as shown in Figure 3.13b, that will be further hydrolysed and the lactic acid dimer formed will be converted into lactic acid monomer that will be finally converted into ethyl lactate monomer. Nevertheless, in the equilibrium composition the excess of ethanol leads to smaller amounts of oligomers (2.4 molar % in the case of REth / la = 1and 0.4 molar % in case of REth / la = 3).

0.25

0.20

0.15 Ethyl lactate, REth/La= 1

Lactic acid, REth/La= 1 0.10 Ethyl lactate, REth/La= 3

Molar fraction Lactic acid, REth/La= 3

0.05

0.00 0 10 20 30 40 50 480 490 500 Time (min) Figure 3.13a Molar fraction histories of ethyl lactate and lactic acid for different initial molar ratios. T = 353.15K . The kinetic model and the parameters used were taken from Tanaka et al. (2002).

62 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

0.018

0.016 Ethyl lactate dimer, REth/La= 1 0.014 Ethyl lactate dimer, REth/La= 3 0.012

0.010

0.008

Molar fraction 0.006

0.004

0.002

0.000 0 1020304050480490500 Time (min) Figure 3.13b Molar fraction histories of ethyl lactate dimer for different initial molar ratios. T = 353.15K . The kinetic model and the parameters used were taken from Tanaka et al. (2002).

3.4.3.5 Effect of polar species

The activity of the resin varies with the polarity of the reaction medium, since it influences the number of available sulfonic groups and their acidity (Fite et al., 1998). The polarity of the medium, mainly due to the water and alcohol concentrations, can affect the reaction rate in two ways:

i) the more adsorbed species (water and alcohol) inhibit the others to adsorb onto the active sites.

ii) water can ionize, solvate and dissociate the acidic protons of the sulfonic groups, depending on their concentration; when the sulfonic sites are completely dissociated the reaction occurs in the liquid phase as in the case of homogeneous catalysis (Gomez et al., 2004).

In this case, the reaction medium has a high water concentration; besides the water initially present in the reaction due to the lactic acid solution, more water is being formed during the course of the esterification of lactic acid with ethanol. Although the kinetic model proposed in this work has taken into account the inhibitory effect caused by adsorption, it doesn’t consider the remaining effects that could be due to water or ethanol presence. Therefore, this could

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 63

also be a reason for the deviations between the experimental results and the predicted by the kinetic model in the transient stage. Françoisse and Thyrion studied the ETBE synthesis catalyzed by Amberlyst 15 (Françoisse and Thyrion, 1991). In that study the influence of ethanol in the reaction rate was taken into account. A kinetic model to describe the behaviour of the reaction for low and high ethanol concentrations was developed. However, this case is different from the one presented in this work, since the most adsorbed component is ethanol and there is no water as reaction product. They use a non-polar solvent (n-pentane) and the other reactant (isobutene) is also non-polar. In the case under study, besides ethanol there is also lactic acid concentrated solution that has high water content. Therefore, at the beginning of the reaction there are three polar species competing to the acid sites; it was observed that the initial kinetic rate increases with the ethanol concentration till a plateau, which was not the behaviour observed in the ETBE kinetics. Most of the works about the esterification reaction of ethanol and lactic acid considers that water and ethanol are the most adsorbed species that supports our kinetic model assumptions (Delgado et al., 2007b; Zhang et al., 2004). Moreover, in a later work for the esterification of lactic acid with butanol catalyzed by Amberlyst 15, the authors didn’t observe different mechanisms for high and low alcohol concentration, and similarly with our model, they neglect the oligomers presence and only considered the water adsorption (Dassy et al., 1994).

The proposed model describes quite well the experimental data up to 80 % of the equilibrium conversion as well as the equilibrium stage, which are the most important to apply to a hybrid reactive separation technology.

3.5 Conclusions

The equilibrium composition for the liquid phase reaction of ethyl lactate synthesis catalyzed by the acid ion exchange resin Amberlyst 15-wet was measured in the temperature range of 50-90ºC, at 6 bars. The thermodynamic equilibrium constant estimated by the UNIQUAC method was ln K = 2.9625 − 515.13 T (K ) and the average reaction enthalpy was of 4.28 kJ·mol-1 in that temperature interval. This relation was also successfully applied to describe the equilibrium compositions of other published studies; better prediction was found for systems where high concentrations of lactic acid was used.

64 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

Because of the strong non-ideality of the liquid reaction mixture, the reaction rate model was formulated in terms of activities. The rate-controlling step for the esterification reaction between lactic acid and ethanol, heterogeneously catalyzed by the Amberlyst 15-wet, was the surface reaction, since external and internal mass resistances were insignificant for the temperature range of 50-90ºC. The catalyst reusability was also studied and it was not verified catalyst deactivation till 3 usages of the same resin sample.

A three-parameter model based on a Langmuir-Hinshelwood rate expression was proposed to describe the experimental kinetic results:

2 r = kc ()aEth aLa − aEL aW / K (1+ K Eth aEth + KW aW ) ; and the model parameters are

711−− kTKmolgc =×2.70 10 exp( − 6011.55/ ( )) ( . .min ) , KW =15.19exp(12.01/T(K)) and K Eth =1.22exp()359.63/T(K) . The agreement between experimental and simulated results was good for the following operating conditions: catalyst loading from 1.2 wt. % to 3.9 wt. %, initial molar ratio of reactants from 1.1 to 2.8 and temperature from 50ºC to 90ºC.

3.6 Notation

a liquid phase activity

d p pellet diameter (m) df film thickness (µm)

-1 Ea apparent reaction activation energy (J mol )

K equilibrium reaction constant

K x equilibrium constant based on molar fractions

K γ equilibrium constant based on activity coefficients

K s equilibrium adsorption constant

-1 -1 kc kinetic constant (mol g min )

-1 -1 k0,c pre-exponential factor (mol g min )

K 0,s constant for Equation 3.11

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 65

MRD mean residual deviation n number of moles (mol)

N total number of measurements taken during the experiments

NE number of experiments performed

NM i number of measurements of the conversion in experiment i

-1 ΔH s enthalpy of adsorption (J mol )

0 -1 ΔH f standard enthalpy of formation (J mol )

-1 ΔGº standard Gibs energy of reaction (J mol )

R gas constant (J mol-1 K-1) r reaction rate (mol g-1 min-1)

REth / La initial molar ratio of ethanol to lactic acid t time coordinate (min)

T temperature (K)

X conversion of the limiting reactant x molar fraction

wcat mass of dry catalyst (g)

X ik kth (model-) predicted value of conversion in experiment i

' X ik kth measured value of conversion in experiment i

Greek letters

ν stoichiometric coefficient

γ activity coefficient

θ set of model parameters to be estimated

2 σ ik variance of the kth measurement of conversion in experiment i

2 σ average of variance

66 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

Subscripts

0 initial value

Eth ethanol

La lactic acid

EL ethyl lactate

W water

exp experimental calc calculated

i relative to component i

l relative to limiting reactant

3.7 References Cited

Asthana N. S., A. K. Kolah, D. T. Vu, C. T. Lira and D. J. Miller, "A kinetic model for the esterification of lactic acid and its oligomers", Ind. Eng. Chem. Res. 45(15): 5251-5257, 2006.

Benedict D. J., S. J. Parulekar and S.-P. Tsai, "Esterification of Lactic Acid and Ethanol with/without Pervaporation", Ind. Eng. Chem. Res. 42(11): 2282-2291, 2003.

Chen Z. T., J. S. Zhong, Q. F. Li, Y. H. Li and Z. H. Ou, "Esterification on Zeolites", Science in China Series B-Chemistry 32(7): 769-775, 1989.

Dassy S., H. Wiame and F. C. Thyrion, "Kinetics or the liquid phase synthesis and hydrolysis of butyl lactate catalysed by cation exchange resin", J. Chem. Technol. Biotechnol. 59(2): 149-156, 1994.

Delgado P., M. T. Sanz and S. Beltran, "Isobaric vapor-liquid equilibria for the quaternary reactive system: Ethanol + water + ethyl lactate + lactic acid at 101.33 kPa", Fluid Phase Equilib. 255(1): 17-23, 2007a.

Delgado P., M. T. Sanz and S. Beltran, "Kinetic study for esterification of lactic acid with ethanol and hydrolysis of ethyl lactate using an ion-exchange resin catalyst", Chem. Eng. J. 126(2-3): 111-118, 2007b.

DIPPR, "Thermophysical Properties Database", (1998).

Dixit A. B. and G. D. Yadav, "Deactivation of ion-exchange resin catalysts. Part I: Alkylation of o-xylene with styrene", React. Funct. Polym. 31(3): 237-250, 1996.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 67

Dupont P., J. C. Védrine, E. Paumard, G. Hecquet and F. Lefebvre, "Heteropolyacids supported on activated carbon as catalysts for the esterification of acrylic acid by butanol", Appl. Catal., A: gen 129(2): 217-227, 1995.

Fite C., J. Tejero, M. Iborra, F. Cunill, J. F. Izquierdo and D. Parra, "The effect of the reaction medium on the kinetics of the liquid-phase addition of methanol to isobutene", Appl. Catal., A: gen 169(1): 165-177, 1998.

Françoisse O. and F. C. Thyrion, "Kinetics and mechanism of ethyl tert-butyl ether liquid- phase synthesis", Chem. Eng. Process. 30(3): 141-149, 1991.

Fredenslund A., J. Gmehling and P. Rasmussen, "Vapor-liquid equilibria using UNIFAC", Elsevier, (1977).

Gmehling J., U. Onken and W. Arlt, "Vapor-liquid equilibrium data collection, aqueous- organic systems (supplement 1)", DECHEMA, Frankfurt am Main, Germany (1981).

Gomez M. F., L. A. Arrúa and M. C. Abello, "Synthesis of 1,1-diethoxyethane using a continuous flow reactor: Catalyst deactivation and effect of feed purity and of solvent addition", J. Chem. Technol. Biotechnol. 79(4): 391-396, 2004.

Ihm S. K., M. J. Chung and K. Y. Park, "Activity difference between the internal and external sulfonic groups of macroreticular ion-exchange resin catalysts in isobutylene hydration", Ind. Eng. Chem. Res. 27(1): 41-45, 1988.

Kirumakki S. R., N. Nagaraju, K. V. R. Chary and S. Narayanan, "Kinetics of esterification of aromatic carboxylic acids over zeolites Hβ and HZSM5 using dimethyl carbonate", Appl. Catal., A: gen 248(1-2): 161-167, 2003.

Koster R., B. Van der Linden, E. Poels and A. Blieky, "The mechanism of the gas-phase esterification of acetic acid and ethanol over MCM-41", J. Catal. 204(2): 333-338, 2001.

Lacaze-Dufaure C. and Z. Mouloungui, "Catalyzed or uncatalyzed esterification reaction of oleic acid with 2-ethyl hexanol", Appl. Catal., A: gen 204(2): 223-227, 2000.

Lee M. J., J. Y. Chiu and H. M. Lin, "Kinetics of catalytic esterification of propionic acid and n-butanol over Amberlyst 35", Ind. Eng. Chem. Res. 41(12): 2882-2887, 2002.

Lee M. J., H. T. Wu and H. M. Lin, "Kinetics of catalytic esterification of acetic acid and amyl alcohol over dowex", Ind. Eng. Chem. Res. 39(11): 4094-4099, 2000.

Lilja J., D. Y. Murzin, T. Salmi, J. Aumo, P. Mäki-Arvela and M. Sundell, "Esterification of different acids over heterogeneous and homogeneous catalysts and correlation with the taft equation", J. Mol. Catal. A: Chem. 182-183: 555-563, 2002.

Lilja J., J. Wärna, T. Salmi, L. J. Pettersson, J. Ahlkvist, H. Grénman, M. Rönnholm and D. Y. Murzin, "Esterification of propanoic acid with ethanol, 1-propanol and butanol over a heterogeneous fiber catalyst", Chem. Eng. J. 115(1-2): 1-12, 2005.

Liu W. T. and C. S. Tan, "Liquid-phase esterification of propionic acid with n-butanol", Ind. Eng. Chem. Res. 40(15): 3281-3286, 2001.

68 CHAPTER 3. Thermodynamic Equilibrium and Reaction Kinetics

Ma Y., Q. L. Wang, H. Yan, X. Ji and Q. Qiu, "Zeolite-catalyzed esterification I. Synthesis of acetates, benzoates and phthalates", Appl. Catal., A: gen 139(1-2): 51-57, 1996.

Mäki-Arvela P., T. Salmi, M. Sundell, K. Ekman, R. Peltonen and J. Lehtonen, "Comparison of polyvinylbenzene and polyolefin supported sulphonic acid catalysts in the esterification of acetic acid", Appl. Catal., A: gen 184(1): 25-32, 1999.

Mazzotti M., B. Neri, D. Gelosa, A. Kruglov and M. Morbidelli, "Kinetics of Liquid-Phase Esterification Catalyzed by Acidic Resins", Ind. Eng. Chem. Res. 36(1): 3-10, 1997.

Montgomery R., "Acidic constituents of lactic acid-water systems", J. Am. Chem. Soc. 74(6): 1466-1468, 1952.

Peña-Tejedor S., R. Murga, M. T. Sanz and S. Beltrán, "Vapor-liquid equilibria and excess volumes of the binary systems ethanol + ethyl lactate, isopropanol + isopropyl lactate and n- butanol + n-butyl lactate at 101.325 kPa", Fluid Phase Equilib. 230(1-2): 197-203, 2005.

Ramalinga K., P. Vijayalakshmi and T. N. B. Kaimal, "A mild and efficient method for esterification and transesterification catalyzed by iodine", Tetrahedron Lett. 43(5): 879-882, 2002.

Reid R. C., J. M. Prausnitz and B. E. Poling, "The Properties of Liquids and Gases", McGraw-Hill, (1987).

Sanz M. T., R. Murga, S. Beltran, J. L. Cabezas and J. Coca, "Autocatalyzed and ion- exchange-resin-catalyzed esterification kinetics of lactic acid with methanol", Ind. Eng. Chem. Res. 41(3): 512-517, 2002.

Schwegler M. A., H. van Bekkum and N. A. de Munck, "Heteropolyacids as catalysts for the production of phthalate diesters", Applied catalysis 74(2): 191-204, 1991.

Silva V. M. T. M. and A. E. Rodrigues, "Dynamics of a fixed-bed adsorptive reactor for synthesis of diethylacetal", AIChE J. 48(3): 625-634, 2002.

Smith J. and H. Van Ness, "Introduction to Chemical Engineering Thermodynamics", McGraw-Hill, (1987).

Teo H. T. R. and B. Saha, "Heterogeneous catalysed esterification of acetic acid with isoamyl alcohol: Kinetic studies", J. Catal. 228(1): 174-182, 2004.

Troupe R. A. and E. DiMilla, "Kinetics of the Ethyl Alcohol-Lactic Acid Reaction", Ind. Eng. Chem. 49(5): 847-855, 1957.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 69

Vu D. T., A. K. Kolah, N. S. Asthana, L. Peereboom, C. T. Lira and D. J. Miller, "Oligomer distribution in concentrated lactic acid solutions", Fluid Phase Equilib. 236(1-2): 125-135, 2005.

Vu D. T., C. T. Lira, N. S. Asthana, A. K. Kolah and D. J. Miller, "Vapor - Liquid equilibria in the systems ethyl lactate + ethanol and ethyl lactate + water", J. Chem. Eng. Data 51(4): 1220-1225, 2006.

Wu K. C. and Y. W. Chen, "An efficient two-phase reaction of ethyl acetate production in modified ZSM-5 zeolites", Appl. Catal., A: gen 257(1): 33-42, 2004.

Yadav G. D. and M. S. Krishnan, "An ecofriendly catalytic route for the preparation of perfumery grade methyl anthranilate from anthranilic acid and methanol", Org. Process Res. Dev. 2(2): 86-95, 1998.

Yadav G. D. and M. S. M. Mujeebur Rahuman, "Cation-exchange resin-catalysed acylations and esterifications in fine chemical and perfumery industries", Org. Process Res. Dev. 6(5): 706-713, 2002.

Zhang Y., L. Ma and J. Yang, "Kinetics of esterification of lactic acid with ethanol catalyzed by cation-exchange resins", React. Funct. Polym. 61(1): 101-114, 2004.

4. Fixed Bed Adsorptive Reactor

Abstract. Multi-component adsorption equilibrium data were measured through binary adsorption experiments performed in a fixed bed column packed with Amberlyst 15-wet for the ethyl lactate system, at 20ºC and 50ºC. A novel approach based on multi-component Langmuir isotherm was used assuming a constant monolayer capacity in terms of volume for all species, reducing the adjustable parameters from 8 to 5, for each temperature. Reactive adsorption experiments were performed and used to validate a mathematical model developed for both fixed bed and simulated moving bed reactors, which involves velocity variations due to the change of multi-component mixture properties.

Adapted from: Pereira C. S. M., V. M. T. M. Silva and A. E. Rodrigues, "Fixed Bed Adsorptive Reactor for Ethyl Lactate Synthesis: Experiments, Modelling, and Simulation", Sep. Sci. Technol. 44(12): 2721 - 2749, 2009.

72 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

4.1 Introduction

The conventional way to produce ethyl lactate is by the esterification of lactic acid with ethanol in the presence of an acid catalyst. The conversion of this kind of reactions is limited by the chemical equilibrium; therefore, the technology for the industrial preparation of esters involves two consecutive steps. The first is the reaction itself, which stops when the equilibrium is reached. The second is the separation of the products from the equilibrium mixture containing products and unconverted reactants. The disadvantage of this technology is in its economics, because of the high energy costs and investment in several reaction and separation units. Multifunctional reactors, where reaction and separation take place into a single unit, allow, in addition to obvious savings in equipment costs, significant improvements in process performance for reactions limited by chemical equilibrium. By removing one of the products from the reaction zone, the equilibrium limitation can be overcome and the conversion can be driven to completion.

In the esterification reaction between lactic acid and ethanol the catalyst used is usually concentrated sulphuric acid (Zhang et al., 2004). However, its application has several drawbacks (as separation problems and corrosion of equipment) and, thus, the use of heterogeneous catalysts is preferable. Among this type of catalysts, strongly acid resins, like Amberlyst 15-wet (A15), are of great interest in the case of reversible reactions, as esterifications and acetalizations, since they can act as catalyst and also as selective adsorbent (Funk et al., 1995; Gandi et al., 2006; Kawase et al., 1996; Mazzotti et al., 1996; Ruggieri et al., 2003; Silva and Rodrigues, 2002). Accordingly, the synthesis of ethyl lactate in a chromatographic reactor using the A15 resin is very attractive. The simulated moving bed reactor (SMBR) is one of the most interesting chromatographic reactors and has been applied to several reversible reactions catalyzed by acidic resins (Borges da Silva et al., 2006; Kawase et al., 1996; Lode et al., 2003; Mazzotti et al., 1996; Minceva et al., 2008; Pereira et al., 2008; Silva and Rodrigues, 2005; Yu et al., 2003) improving the reaction conversion, given that the products are formed and simultaneously separated and removed from the reaction medium in two different streams (extract and raffinate). In order to provide a better understanding of the performance of the SMBR and to validate kinetic and adsorption data it is appropriate to evaluate first the dynamic behaviour of the fixed bed adsorptive reactor (Gandi et al., 2006; Kawase et al., 1999; Lode et al., 2001; Mazzotti et al., 1997; Silva and Rodrigues, 2002). The present work addresses the detailed experimental study of the ethyl lactate synthesis in a fixed bed adsorptive reactor in order to validate the mathematical model

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 73

to be later used to define operating conditions of the SMBR. Binary adsorption experiments were carried out on a fixed bed packed with A15 resin, in absence of reaction, at 293.15 and 323.15 K to determine the multi-component adsorption parameters.

4.2 Experimental Section

4.2.1 Chemicals and Catalyst / Adsorbent

The chemicals used in the experiments were ethanol (>99.9% in water), lactic acid (>85% in water) and ethyl lactate (>98% in water) from Sigma-Aldrich (U.K.)

The column was packed with Amberlyst 15-wet (A15), which is a highly cross-linked polystyrene-divinylbenzene ion exchange resin functionalized with sulfonic groups (SO3H), that acts as catalyst and adsorbent in this system. The properties of the A15 resin are presented in Table 4.1. A15 is a macroreticular-type resin, but according to Ihm et al. (Ihm et al., 1988) only 4 % of the active sites are located at the macropores (surface of the microspheres) and the others 96 % are inside gel polymer microspheres.

Table 4.1 Physical and chemical properties of resin A15.

Properties A15

Manufacturer Rohm and Haas

Concentration of acid sites (eq H+/kg) 4.7

Surface area (m2/g) 53

Average pore diameter (nm) 30

Particle diameter (mm) 0.3-1.2

This kind of resins swell selectively in the presence of a liquid phase constituted of a multi-component mixture. The swelling is due to the sorption of the different components of the mixture, depending on their relative affinities to the resin. The polymeric resins that contain sulfonic acid functional groups, like the A15, exhibit a strong selectivity for polar species. Some authors (Mazzotti et al., 1997) studied the sorption equilibrium of acetic acid,

74 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

ethanol, water and ethyl acetate on A15. They verified that the components can be listed by the following decreasing order of affinity to the resin: water, ethanol, acetic acid and ethyl acetate and that the swelling ratio is, respectively, 1.52, 1.48, 1.30 and 1.22. This conclusion agrees with the polarity of the components. So, it is expected for the system hereby presented that the order of decreasing affinity of the components to the resin will be: water, ethanol, lactic acid and ethyl lactate. The swelling effect might change the length and the bulk porosity of the fixed bed reactor; however, in the system under study, just a insignificant variation in the bed length was noticed; and, therefore, a constant bed length was considered.

4.2.2 Experimental Apparatus

The experiments were performed in a laboratory-scale jacketed glass column that was maintained at constant temperature through a thermostatic bath (293.15 or 323.15 K), at atmospheric pressure (see Figure 4.1). The experimental breakthrough curves were obtained by analysing with a gas chromatograph, small samples withdrawn at different times from the column exit.

Feed Stream

Thermostatic Resin

Bath

Refrigerating water

Sampling line Three-way valve

Reservoir

Figure 4.1 Experimental set-up (configuration: top-down flow direction).

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 75

4.2.2.1 Bed Porosity and Peclet Number

The bed porosity and the Peclet number were determined by pulse experiments of tracer using a blue dextran solution (5 kg/m3), which is a polymer whose molecule is large enough (M.W. = 2,000,000) to diffuse only in the bulk fluid phase between resin particles. Samples of the blue dextran solution (0.2 cm3) were injected under different flow rates and the column response was monitored using a UV-VIS detector at 300 nm. The bed porosity was calculated from the stoichiometric time of the obtained experimental curves. An average Peclet number was obtained for the range of flow rates to be used in the fixed bed experiments by calculating the second moment of the experimental curves. Figure 4.2 shows the different tracer experiments performed and the respective values of bed porosity and Peclet number are presented in Table 4.2. The characteristics of the fixed bed column are summarized in Table 4.3.

1.00 run 1 0.80 run 2 run 3

0 0.60

C/C 0.40 0.20 0.00 048121620 Time (min) Figure 4.2 Tracer experiments using a blue dextran solution. Points are experimental values and lines are simulated curves.

Table 4.2 Results obtained from tracer experiments.

Q (mL/min) τ (min) ε σ 2 Pe

run 1 1.985 11.23 0.347 3.528 71.48

run 2 4.981 4.69 0.364 0.556 79.24

run 3 8.026 2.92 0.365 0.229 74.54

76 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

Table 4.3 Characteristics of the fixed bed column.

Solid weight (A15) 25 g

Length of the bed (L) 12 cm

Internal diameter (Di) 2.6 cm

Average radius of resin beads (rp) 372.5 μm

3 Bulk density (ρ b) 390 kg/m

Bed porosity (ε) 0.36

(Lode et al., 2001) Resin particle porosity (εp) 0.36

Peclet number 75

All the samples removed from the column exit were analysed in a gas chromatograph; the analytical method is described in Chapter 3.

4.3 Modelling of Fixed Bed

Several mathematical models have been developed to explain the kinetic behaviour of the fixed adsorptive reactor and to predict the breakthrough curves. In order to interpret correctly the behaviour of a fixed bed adsorptive reactor, it is necessary to characterize on one hand the partitioning equilibrium on the solid sorbent and on the other hand the reaction kinetics on the solid catalyst. A rigorous modelling of the sorption in the swelling polymer should include an appropriate model to predict the polymer-phase activities. Mazzoti and co-workers used the extended Flory–Huggins model to determine the chemical activities of the species in the polymer phase for the esterification of acetic acid to ethyl acetate in the presence of A15 ion exchange resin catalyst (Mazzotti et al., 1996). They assumed, in their kinetic model, that the activities of the species in the bulk liquid phase (estimated by UNIFAC method) were in equilibrium with the ones in the resin polymer phase (estimated by the extended Flory– Huggins model). However, in latter papers (Lode et al., 2001; Pöpken et al., 2000), it is mentioned that the extended Flory–Huggins model is not well suited for high cross-linked resins bearing highly polar groups on almost all monomer and should be regarded as an empirical tool to correlate the equilibrium data and to predict the behaviour of

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 77

multi-component mixtures. Yu and collaborators state that using the extended Flory–Huggins model based in adsorption equilibrium data is not viable for most adsorption systems, given that non-reactive binary mixtures are scarce in the literature and the model involves complexity and inconvenience in computation (Yu et al., 2004). Thus, an alternative approach based on the multi-component Langmuir model was considered in this work, since it is able to represent satisfactorily the experimental adsorption data and is simpler than the Flory-Huggins model. Nevertheless, it has to be mentioned that one of the assumptions of the Langmuir model considers that the monolayer adsorption has energetically equal binding sites, which do not describe the actual physical adsorption phenomena. Therefore, some authors (Pöpken et al., 2000) proposed a multi-component Langmuir adsorption isotherm which assumes a constant monolayer capacity in terms of mass for all species, after having measured mono-component adsorption data expressed in terms of volumes, masses and moles of different species per gram of A15. In this work we will consider a constant volumetric monolayer capacity for all species, which will reduce the adjustable adsorption parameters from 8 (one molar monolayer capacity and one equilibrium constant for each component) to 5 (one volumetric monolayer capacity for all components and one equilibrium constant for each component).

A detailed model was used to describe the dynamic behaviour of the fixed bed adsorptive reactor taken into account the following assumptions:

- Isothermal operation;

- The flow pattern is described by axial dispersed plug flow model;

- External and internal mass transfer for absorbable species is combined in a global resistance;

- Constant column length and packing porosity;

- Extended Langmuir Isotherm model for multi-component adsorption;

- Velocity variations due to changes in bulk composition.

The model equations are constituted by the following system of four second order partial

th differential equations in the bulk concentration of the i component ( Ci ), four ordinary differential equations in the average concentration of the ith component into the particle pores

78 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

(Cpi, ) and four algebraic equations in the adsorbed concentration in equilibrium with

Cp,i ( q i ):

Bulk fluid mass balance to component i:

∂∂Cx∂()uC (1−∂ε ) 3 ⎛⎞ i i i (4.1) ++KCCLi,,() i −= pi D ax⎜⎟ C T ∂∂tzε rp ∂∂ zz⎝⎠

where KL,i is the global mass transfer coefficient of the component i, ε is the bed porosity, t is the time variable, z is the axial coordinate, Dax and u are the axial dispersion coefficient and the interstitial velocity respectively, rp is the particle radius and xi is the component molar fraction in liquid phase.

The axial dispersion coefficient Dax was calculated from the Peclet number:

uL Pe= (4.2) Dax

The interstitial fluid velocity variation is calculated using the total mass balance:

du (1−ε ) 3 NC =−∑ KVLi,, moli() C i − Cp,i (4.3) dzε rp i=1 where Vmol,i is the molar volume of component i .

Pellet mass balance to component i:

3 ∂ C p, i ∂ q i ν i ρb K L,i ()Ci −C p, i = ε p + (1− ε p ) − r()C p, i (4.4) rp ∂ t ∂ t 1− ε

Where ν i is the stoichiometric coefficient of component i , ρb is the bulk density, ε p is the particle porosity, q i is the average adsorbed phase concentration of species i in equilibrium with Cp,i , and r is the kinetic rate of the chemical reaction relative to the average particle concentrations in the fluid phase given by (see Chapter 3):

aaEth La− aa EL W K eq rk= c 2 (4.5) ⎛⎞NC ⎜⎟1+ ∑ Kasi, i ⎝⎠i=1

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 79

where kc is the kinetic constant, K s, i is the adsorption constant for species i and Keq is the equilibrium reaction constant, a is the species activity (calculated by UNIQUAC model) and the subscripts Eth, La, EL and W refer to ethanol, lactic acid, ethyl lactate and water, respectively. Kinetic and thermodynamic parameters, essential for the Equation 4.5, were taken from Chapter 3 and are given in Table 4.4. Table 4.4 Kinetic and thermodynamic parameters. Temperature k c K K K (K) (mol/(g.min)) s,w s,Eth eq

293.15 0.035 15.82 4.16 3.34

323.15 0.225 15.77 3.71 3.93

Langmuir Adsorption equilibrium isotherm to component i:

QKCiipi, qi = NC (4.6) 1+ ∑ KCj p, j j=1

where QQViVmoli= / , , QV is the volumetric monolayer capacity, Vmol, i is the molar volume of species i and Ki is the equilibrium constant for component i.

Initial and Danckwerts boundary conditions:

t = 0 Ci = C p, i = Ci, 0 (4.7)

∂ xi zuCDCuC=−=0 iaxT iF, (4.8a) ∂ z z=0

uu= 0 (4.8b)

∂ C z = L i = 0 (4.9) ∂ z z=L where subscripts F and 0 refer to the feed and initial states, respectively.

The proposed model considers a global mass transfer coefficient ( K L ) defined, for each component, as:

1 1 1 = + (4.10) K L ke ε p ki

80 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

wherein ke and ki are, respectively, the external and internal mass transfer coefficients.

Santacesaria and co-workers (Santacesaria et al., 1982) showed that the internal mass transfer coefficient varies in time, and the calculation of the rigorous values requires the solution of the complete model equations inside particles. As an approximation, the mean value estimated by the Equation 4.11 (Glueckauf, 1955) was used:

5/Dm τ ki = (4.11) rp

The external mass transfer coefficient was estimated by the Wilson and Geankoplis correlation (Ruthven, 1984):

1.09 Sh = ()Re Sc 0.33 0.0015 < Re < 55 (4.12) p ε p p

where Shp and Re p are, respectively, the Sherwood and Reynolds numbers, relative to particle:

ke d p Shp = (4.13) Dm

ρ dup Re = (4.14) p η and Sc is the Schmidt number:

η Sc = (4.15) ρ Dm

The infinite dilution diffusivities were estimated by the Scheibel correlation (Scheibel, 1954) which modified the Wilke-Chang equation in order to eliminate the association factor:

2/3 8.2× 10−8T ⎡ ⎛⎞3V ⎤ Dcms02(/)=+⎢ 1⎜⎟mol, B ⎥ (4.16) AB, η VV1/3 ⎢ ⎜⎟⎥ B mol,, A⎣ ⎝⎠ mol A ⎦

0 where D AB, is the diffusion coefficient for a dilute solute A into a solvent B, T is the temperature, Vmol, i is the molar volume of the component i, ηB is the viscosity of solvent B. Table 4.5 shows the diffusion coefficients for all binary systems:

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 81

Table 4.5 Diffusion coefficients for the binary mixtures estimated using the Scheibel correlation (equation 4.16).

2 Di,jº (cm /s) Ethanol Lactic acid Ethyl lactate Water

Ethanol 0 1.6796E-06 2.583E-05 2.3655E-05

Lactic acid 2.5056E-05 0 2.1087E-05 2.0236E-05

Ethyl lactate 1.8034E-05 9.8526E-07 0 1.5444E-05

Water 8.2195E-05 4.6733E-06 7.4567E-05 0

For binary systems, Vignes equation (Vignes, 1966) was used to predict DAB, in concentrated solutions from the infinite dilute coefficients as a simple function of composition:

00xx21 DD2,1== 1,2()() D 1,2 D 2,1 (4.17)

For concentrated multi-component systems there are several mixing rules (Umesi and Danner, 1981) to predict the molecular diffusivity coefficient of a solute in a mixture. In this work it was used the Perkins and Geankoplis method (Perkins and Geankoplis, 1969):

n 0.8 0 0.8 DxDAm,,ηη m= ∑ i Ai i (4.18) i=1 iA#

where ηi is the viscosity of pure component i and ηm is the viscosity of the mixture. The viscosities of the mixtures play an important role in the estimation of the diffusion coefficient. The methods used to predict binary and multi-component mixtures viscosities are detailed in the next sub-section.

In this model, the mixture viscosity and the components diffusivities in the liquid mixture were calculated at each time and at every axial position; therefore, the mass transfer parameters also depend on the liquid composition (estimated at each time and axial position from the equations presented).

4.3.1 Multi-component viscosity

One of the most used and recommended liquid mixture viscosity correlation is the one by Grunberg–Nissan (Grunberg and Nissan, 1949), which for binary mixtures is:

82 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

ln(ηm )=+xx112 ln(η ) ln(η 2121,2 ) + xxG (4.19)

where xi is the mole fraction of component i and G1,2 is an empirical interaction parameter adjusted by experimental data.

For aqueous lactic acid solutions, the interaction parameters were found to be

GCLA, W (20º )= 5.240 and GCLA, W (50º )= 4.369 using viscosities determined experimentally presented in literature (Troupe et al., 1951). As can be seen in Figure 4.3, the Equation 4.19 describes well the viscosity in a large range, from diluted solutions till concentrated lactic acid (85% mass fraction).

35 Exp. 20 ºC (Troupe et al, 1951) Exp. 50 ºC (Troupe et al,1951) 30 Theoretical (Eq. 4.19)

Viscosity (cP) 10

0 0.0 0.2 0.4 0.6 Mole fraction of lactic acid Figure 4.3 Viscosities of aqueous lactic acid solutions.

For the system water ethanol, experimental data at 20 ºC (Gonzalez et al., 2007) and 50 ºC

(Motin et al., 2005) are not well fitted by Equation 4.19 (GCEth, W (20º )= 2.568 and

GCEth, W (50º )= 1.799 ), as it can be seen in Figure 4.4. It was stated by several authors, that the Grunberg–Nissan correlation is not suitable for several aqueous system (Li and Carr, 1997; Macias-Salinas et al., 2003). Normally, the correlation used to describe the viscosity of those systems involves at least 3 adjustable parameters (Gonzalez et al., 2007; Motin et al., 2005). We propose a new correlation with just one adjustable parameter similar to that of Grunberg–Nissan, but the excess viscosity function is averaged by the volume fractions instead of molar ones:

ln(ηηηφφm )=+x112 ln( )xG ln( 2121,2 ) + (4.20)

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 83

where xi is the mole fraction of component i , φi is the volume fraction of component i and

G1,2 is an empirical interaction parameter adjusted by experimental data. Using the new interaction parameters in Equation 4.20 (GCEth, W (20º )= 3.833 andGCEth, W (50º )= 2.698 ), the ethanol/water system viscosities are very well predicted, as it can be seen in Figure 4.5.

3.5 Exp. 20 ºC (Gonzalez et al, 2007) 3.0 Exp. 50 ºC (Motin et al, 2005) Theoretical (Eq. 4.19) 2.5

2.0

1.5

Viscosity (cP) Viscosity 1.0

0.5

0.0 0.0 0.2 0.4 0.6 0.8 1.0 Mole fraction of ethanol Figure 4.4 Viscosities of ethanol/water mixtures. Solid line calculated by Equation 4.19.

3.5 Exp. 20 ºC (Gonzalez et al, 2007) 3.0 Exp. 50 ºC (Motin et al, 2005) Theoretical (Eq. 4.20) 2.5

2.0

1.5

Viscosity (cP) 1.0

0.5

0.0 0.0 0.2 0.4 0.6 0.8 1.0 Mole fraction of ethanol Figure 4.5 Viscosities of ethanol/water mixtures. Solid line calculated by Equation 4.20.

Since there are no experimental data available for the systems ethanol/ethyl lactate and ethanol/lactic acid, these systems are considered as ideal, i.e., GEth, EL = 0 andGEth, LA = 0 .

84 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

The liquid viscosity of the multi-component system ethanol, lactic acid, ethyl lactate and water is calculated by the following model:

ln(ηm )= xxxxxxGG Eth ln(η Eth )+ LA ln(η LA )+ EL ln(η EL )+ W ln(η W ) + LA W LA,, W+φ Ethφ W Eth W (4.21)

The viscosities of pure components at 20 and 50ºC are presented in Table 4.6 (see appendix B for details in calculation).

Table 4.6 Viscosities (cP) of ethanol, lactic acid, ethyl lactate and water at 20ºC and 50ºC.

Temperature Ethanol Lactic acid Ethyl lactate Water 20 ºC 1.1617 53.6564 2.5986 1.0254 50 ºC 0.6885 13.9919 1.1233 0.5530

Numerical Solution

The model equations were solved numerically using the gPROMS-general PROcess Modelling System version: 3.0.3 (www.psentreprise.com). gPROMS is a software package for modelling and simulation of processes with both discrete and continuous as well as lumped and distributed characteristics. The mathematical model involves a system of partial and algebraic equations (PDAEs). Third order orthogonal collocation over twenty one finite elements was used in the discretization of axial domain. The system of ordinary differential and algebraic equation (ODAEs) was integrated over time using the DASOLV integrator implementation in gPROMS. For all simulations was fixed a tolerance equal to 10-7.

4.4 Results and Discussion

4.4.1 Adsorption Isotherm

As the resin A15 acts simultaneously as adsorbent and as catalyst, in order to have information on the adsorptive equilibrium alone it is necessary to perform experiments with non reactive binary mixtures. So, the breakthrough curves of ethanol, lactic acid, ethyl lactate and water were measured in the absence of reaction. The resin was saturated with a certain component A and then the feed concentration of component B was changed stepwise. The

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 85

experimental data were used to calculate the number of moles adsorbed/desorbed of each component for all the experiments. The adsorption parameters were optimized by minimizing the difference between experimental and theoretical values, according to Equation 4.26:

∞ ads nexp =Q []CF −Cout (t) dt (4.22) ∫0

∞ des nexp =Q []Cout (t) − CF dt (4.23) ∫0

ads ntheo =([ε + (1− ε )ε p ](CF − C0 )+ (1− ε )(1− ε p )[q(CF ) − q(C0 )])V (4.24)

des ntheo = ([ε + (1− ε )ε p ](C0 − CF )+ (1− ε )(1− ε p )[q(C0 ) − q(CF )])V (4.25)

NE 22 fob=−+−⎡⎤ nads n ads n des n des ∑ ⎢⎥()()exptheo exp theo (4.26) k =1 ⎣⎦

4.4.1.1 Binary Adsorption experiments

As mentioned above, for the determination of the adsorption parameters over the A15 resin it was necessary to perform experiments with non reactive pairs of species. The possible binary mixtures to run the breakthrough experiments in absence of reaction are ethanol / water, ethyl lactate / ethanol and lactic acid / water. In all the experiments, the correct liquid flow direction (bottom-up or top-down) was considered in order to obtain reproducibility in the experimental results, since the hydrodynamic regime has an important effect on them and the difference in densities of the species can lead to axial backmixing driven by natural convection. The densities of ethanol, lactic acid, ethyl lactate and water at 293.15 K are 0.79 g/cm3, 1.21 g/cm3, 1.04 g/cm3 and 1 g/cm3, respectively. The concentration fronts moving within the column are hydrodynamically stable if the component above the front is less dense than the component below the front (Silva and Rodrigues, 2002). The multi-component adsorption equilibrium was measured at two different temperatures, 293.15 K and 323.15 K. In Table 4.7 the optimized adsorption parameters and the molar volumes at the two temperatures are presented.

86 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

Table 4.7 Adsorption parameters over A15 resin.

Component QV (ml/lwet solid) Q (mol/lwet solid) K (l/mol) Vmol (ml/mol) 20 ºC / 50 ºC 20 ºC / 50 ºC 20 ºC / 50 ºC 20 ºC / 50 ºC Ethanol 6.70 / 6.30 5.443 / 3.068 58.17 / 60.87 Lactic acid 5.22 / 4.94 4.524 / 4.085 74.64 / 77.56 390.0/383.5 Ethyl lactate 3.42 / 3.23 1.117 / 1.815 113.99 / 118.44 Water 21.57 / 20.58 15.353 / 7.055 18.08 / 18.63

In order to compare the selectivity of the resin to the components, two different experiments were performed. One, where ethanol is fed to the column initially saturated with water, and other, where ethanol is fed to the column initially saturated with ethyl lactate (see Figure 4.6). In the first case (Figure 4.6a) the concentration front of ethanol has a dispersive character and in the second one (Figure 4.6b) the ethanol concentration front is self-sharpening. This is due to the fact that water is the preferentially adsorbed component and ethyl lactate is the weakest adsorbed component.

(a) (b)

18 18 16 16 14 Ethanol 14 Ethanol 12 12 Theoretical Theoretical 10 10 8 8 6 6 4 4 2 2

Outlet concentration concentration (molL)Outlet 0 0 Outlet concentration (mol/L) 0 10203040 0 10203040 Time (min) Time (min) Figure 4.6 Breakthrough experiments: outlet concentration of ethanol as a function of time; Q = 5 mL/min; T = 293.15 K; top-down direction flow; ( a) ethanol displacing water; (b) ethanol displacing ethyl lactate.

The breakthrough curves of the four components at 323.15 K are presented next in this section, while the remaining breakthrough curves at 293.15 K used for the determination of the adsorption parameters are shown in appendix D.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 87

(a) 60 Water 50 Ethanol Theoretical 40

Outlet concentration (mol/L)10

0 0 50 100 150 200 250 300 350 Time (min)

(b) 60 Water 50 Ethanol Theoretical

Outlet concentration (mol/L)10

0 0 20 40 60 80 100 120 140 160 Time (min)

Figure 4.7 Breakthrough experiments: outlet concentration of ethanol and water as a function of time; Q = 5 mL/min; T = 323.15 K; (a) water displacing ethanol; Bottom up flow direction; (b) ethanol displacing water; Top-down flow direction.

The experimental and theoretical concentration history for the binary mixture ethanol / water is shown in Figure 4.7. Comparing the experimental amount of water adsorbed in the experiment of Figure 4.7a with the experimental amount of water desorbed in the experiment of Figure 4.7b, the error obtained is 0.3 %. For ethanol, the error between the amount adsorbed, Figure 4.7b, and the amount eluted, Figure 4.7a, is 0.31 %. The deviations found between the experimental amount adsorbed of water and the one predicted by the model is of 4 %, while in the case of ethanol the difference between experimental amount adsorbed and predicted is of 0.6 %.

88 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

(a) 18 Ethyl lactate 16 Ethanol 14 Theoretical

4 Outlet concentration (mol/L) 2

0 0 40 80 120 160 200 Time (min)

(b) 18

16 Ethyl lactate 14 Ethanol Theoretical 12

4 Outlet concentration (mol/L) 2

0 0 20406080100 Time (min)

Figure 4.8 Breakthrough experiments: outlet concentration of ethanol and ethyl lactate as a function of time; Q = 5 mL/min; T = 323.15 K; (a) ethyl lactate displacing ethanol; Bottom up flow direction (b) ethanol displacing ethyl lactate; Top-down flow direction.

The experimental and simulated results for the non reactive binary mixture ethanol / ethyl lactate are shown in Figure 4.8. The difference obtained between the experimental amount of ethyl lactate eluted in the experiment of Figure 4.8b, and the experimental amount of ethyl lactate adsorbed in the experiment of Figure 4.8a, was 0.21 %. In the case of ethanol, the error obtained between the experimental amount adsorbed, Figure 4.8b, and the experimental amount eluted, Figure 4.8a, was 4.22 %.

For the non reactive pair lactic acid / water (Figure 4.9) the deviation found between the experimental amount of lactic acid eluted (Figure 4.9b) and adsorbed (Figure 4.9a) was 0.45 % and between the experimental amount of water adsorbed (Figure 4.9b) and eluted (Figure 4.9a) was -1.07 %.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 89

(a) 60 Water 50 Lactic acid Theoretical 40

Outlet concentration (mo/L) 10

0 0 40 80 120 160 Time (min)

(b) 60

40 Water Lactic acid 30 Theoretical

Outlet concentration (mol/L) 10

0 0 102030405060 Time (min)

Figure 4.9 Breakthrough experiments: outlet concentration of water and lactic acid as a function of time; Q = 5 mL/min; T = 323.15 K; (a) lactic acid displacing water; Bottom up flow direction (b) water displacing lactic acid; Top-down flow direction.

All the binary adsorption experiments performed are very well described by the model, except for the case of the experiment of Figure 4.9b, where pure water displaces an 86 wt % lactic acid solution. This could be explained by the water/lactic acid high selectivity in A15 (about 7.2); however, this behaviour was not observed in the binary experiments ethanol/water where the selectivity is very similar (about 7.5). Therefore, that behaviour might be due to viscous fingering phenomenon, since the highest viscous fluid (85% lactic acid solution: 9.40 cP at 323.15 K), is being displaced by the lowest viscous fluid (water: 0.55 cP at 323.15 K), in the top-down direction, inducing an unstable interface between them.

90 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

The difference in their viscosities is of about 9 cP and in this case fingering effects are significant (Catchpoole et al., 2006). Viscous fingering is not accounted in the global mass transfer coefficient (Equation 4.10). In order to verify this assumption, the effect of viscous fingering can be described increasing the axial dispersion (Mallmann et al., 1998). Indeed, the theoretical curve using an axial dispersion coefficient increased by a factor of 7 (thin line in Figure 4.9b) fits the experimental data very well.

From the data reported it can be concluded that the most adsorbed component in A15 is water followed by ethanol and lactic acid, being the less adsorbed the ethyl lactate, as expected due to the polarity of the species.

4.4.2 Kinetic experiments

4.4.2.1 Fixed Bed Reactor

A reaction experiment was performed in the chromatographic reactor packed with A15 initially saturated with ethanol, at 20ºC. As shown in Figure 4.10, the lactic acid conversion is very low due to both mass transfer and reaction kinetics limitations. Therefore, the fixed bed reactor was operated at 50ºC in order to reduce mass transfer resistance and to increase reaction kinetics.

16 Ethanol Water 14 Ethyl lactate Lactic acid 12 Theoretical

4 Outlet concentration (mol/L) 2

0 0 102030405060 Time (min) Figure 4.10 Concentration histories at the outlet of the fixed bed adsorptive reactor; column initially saturated with ethanol and then fed with a mixture of ethanol and lactic acid solution:QmL= 4.6 / min , TK= 293.15 ,

CmolLLa, F = 6.76 / , CmolLEth, F = 6.76 / andCmolLWF, = 5.45 / .

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 91

A mixture of ethanol and lactic acid was fed to the reactor saturated with ethanol and the concentration history of ethanol, lactic acid, ethyl lactate and water at the end of the column is shown in Figure 4.11a.

(a) 18 Ethanol 16 Water Ethyl lactate 14 Lactic acid 12 Theoretical

4 Outlet concentration concentration Outlet (mol/L) 2

0 0 20 40 60 80 100 120 140 Time (min)

(b) 18 Ethanol 16 Water Ethyl lactate 14 Lactic acid 12 Theoretical

4 Outlet concentration (mol/L) 2

0 0 10203040506070 Time (min)

Figure 4.11 Concentration histories at the outlet of the fixed bed adsorptive reactor for production steps. a) Experiment 1: Column initially saturated with ethanol and then fed with a mixture of ethanol and lactic acid solution.

QmL=1.0 / min , TK= 323.15 , CmolLLa, F = 5.88 / , CmolLEth, F = 6.60 /

andCmolLWF, = 5.87 / . b) Experiment 2: Column initially saturated with ethanol and then fed with lactic acid solution (86 wt % in water).QmL=1.3 / min , TK= 323.15 .

92 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

As the lactic acid solution enters the column it is adsorbed, reacts with ethanol to produce ethyl lactate and water in the stoichiometric amount. Ethyl lactate is first eluted, since it has less affinity with the resin than water. This process continues until the equilibrium is reached; the resin is completely saturated with water and lactic acid. At this moment the selective separation of ethyl lactate and water is no longer possible. The local compositions remain constant and the steady state is achieved being the outlet stream constituted by a reactive mixture at the equilibrium composition. Analysing the outlet concentration curves of ethyl lactate and water, shown in Figure 4.11a, a difference between them is noticed although they are formed at the same stoichiometric amount. This is related to the difference in the adsorbed amount of these species in the resin and, mainly, due to the fact that in the feed there is already some water content due to the lactic acid solution.

A second experiment was performed by only feeding the lactic acid solution (86 wt % in water) to the fixed bed adsorptive reactor (see Figure 4.11b), in order to validate the model under very different conditions. Similarly to the previous experiment, as the lactic acid and water enter the column, they are adsorbed and start displacing the initially adsorbed ethanol. Simultaneously, the lactic acid reacts with ethanol in the resin phase until complete depletion of adsorbed ethanol and the formed ethyl lactate is adsorbed, forming a dispersive front with ethanol once this is more adsorbed than ethyl lactate. The adsorption of lactic acid solution continues till the resin saturation, displacing the ethyl lactate. Finally, lactic acid and water exit the column, being water the last eluted component due to high affinity of A15 towards this specie.

After the steady state is achieved and to perform a new reaction experiment it is necessary to first regenerate the column. This step requires the use of a solvent to displace the adsorbed species. The solvent used was ethanol, since lactic acid solution could not be used, because it contains water and then the column wouldn’t be completely regenerated. In Figure 4.12, the regeneration steps for each experiment are presented. As expected, ethyl lactate and lactic acid are rapidly desorbed; however, a large amount of ethanol is required in order to completely desorb all water.

Al the experimental results obtained for both production and regeneration steps, at 293.15 and 323.15 K, are well described by the proposed mathematical model.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 93

(a) 18

14 Ethanol 12 Water 10 Ethyl lactate Lactic acid 8 Theoretical

4 Outlet concentration (mol/L) 2

0 0 102030405060 Time (min)

(b)

14 Ethanol 12 Water Ethyl lactate 10 Lactic acid 8 Theoretical

4 Outlet concentration (mol/L) 2

0 0 10203040 Time (min) Figure 4.12 Concentration histories at the outlet of the fixed bed adsorptive reactor for regeneration steps with ethanol (QmL= 4.3 / min , TK= 323.15 ). a) Experiment 1. b) Experiment 2.

94 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

4.5 Conclusions

A detailed study of the ethyl lactate synthesis in a fixed bed adsorptive reactor was carried out. Adsorption experiments in absence of reaction at 20ºC and 50ºC were performed. For each temperature, the adsorption parameters were obtained: a single volumetric monolayer capacity for all components and one equilibrium constant for each component. A mathematical model for the fixed bed adsorptive reactor was developed which includes: axial dispersion, external and internal mass transfer resistances, multi-component Langmuir adsorption isotherm, the reaction rate measured in a previous work and velocity variations due to mixture molar volume variations. This model proved to be efficient in the prediction of the reaction and regeneration steps performed in the fixed bed reactor and it will be very useful for the study of the SMBR for the synthesis of ethyl lactate.

4.6 Notation

a liquid phase activity

3 C liquid phase concentration (mol/m )

3 CT total liquid phase concentration (mol/m )

3 C p average liquid phase concentration inside the particle (mol/m ) df film thickness (µm)

d p particle diameter (m)

0 2 D AB, diffusion coefficient for a dilute solute A into a solvent B (m /s)

2 DAB, mutual diffusion coefficient for binary concentrated solutions (m /s)

2 Dax axial dispersion (m /s)

2 Dm molecular diffusivity (m /s)

G1,2 interaction parameter in Equation 4.19

K Langmuir equilibrium parameter (m3/mol)

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 95

ki internal mass transfer coefficient (m/s)

ke external mass transfer coefficient (m/s)

kc kinetic constant (mol/kgres.s)

K eq equilibrium constant

K S adsorption constant in Equation 4.5.

K L global mass transfer coefficient (m/s)

L bed length (m) n number of moles (mol)

Pe Peclet number

3 q average solid phase concentration in equilibrium with Cp (mol/m res)

3 Q molar adsorption capacity, defined as QQViVmoli= / , (mol/m res)

3 3 QV volumetric monolayer capacity (m /m res) r reaction rate relative to fluid concentration inside the particle (mol/kgres.s)

rp particle radius (m)

Rep Reynolds number relative to particle

Sh p Sherwood number relative to particle

Sc Schmidt number

T temperature (K) t time (s) t* switching time (min) u interstitial velocity (m/s)

3 Vmol molar volume in the liquid phase (m /mol)

V volume of the bulk (m3) x liquid phase molar fraction

96 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

z axial coordinate (m)

Greek letters

ε bulk porosity

ε p porosity of pellet

η fluid viscosity (kg/m.s)

ηm mixture viscosity (kg/m.s)

φ volume fraction

ν stoichiometric coefficient

3 ρ liquid density (kg/m )

3 ρ b bulk density (kgres/m )

τ tortuosity

Subscripts

0 relative to initial conditions

exp experimental

F relative to the feed i relative to component i (i= Eth, La, EL, W) out at the end of the fixed bed column

p relative to particle theor theoretical

Eth relative to ethanol

La relative to lactic acid

EL relative to ethyl lactate

W relative to water

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 97

Superscripts ads adsorbed des desorbed

4.7 References

Borges da Silva E. A., A. A. Ulson de Souza, S. G. U. de Souza and A. E. Rodrigues, "Analysis of the high-fructose syrup production using reactive SMB technology", Chem Eng J. 118(3): 167-181, 2006.

Catchpoole H. J., R. Andrew Shalliker, G. R. Dennis and G. Guiochon, "Visualising the onset of viscous fingering in chromatography columns", J Chrom A 1117(2): 137-145, 2006.

Funk G. A., J. R. Lansbarkis and A. K. Chandhok, "Process for concurrent esterification and separation using a simulated moving bed", Patent 5,405,992 (1995).

Gandi G. K., V. M. T. M. Silva and A. E. Rodrigues, "Synthesis of 1,1-dimethoxyethane in a fixed bed adsorptive reactor", Ind Eng Chem Res. 45(6): 2032-2039, 2006.

Glueckauf E., "Theory of chromatography", Trans. Faraday Soc 51: 1540-1551 1955.

Gonzalez B., N. Calvar, E. Gomez and A. Dominguez, "Density, dynamic viscosity, and derived properties of binary mixtures of methanol or ethanol with water, ethyl acetate, and methyl acetate at T = (293.15, 298.15, and 303.15) K", J. Chem. Thermodyn. 39(12): 1578- 1588, 2007.

Grunberg L. and A. H. Nissan, "Mixture law for viscosity", Nature 164(4175): 799-800, 1949.

Kawase M., Y. Inoue, T. Araki and K. Hashimoto, "The simulated moving-bed reactor for production of bisphenol A", Catal Today 48(1-4): 199-209, 1999.

Kawase M., T. B. Suzuki, K. Inoue, K. Yoshimoto and K. Hashimoto, "Increased esterification conversion by application of the simulated moving-bed reactor", Chem Eng Sci. 51(11): 2971-2976, 1996.

Li J. and P. W. Carr, "Accuracy of Empirical Correlations for Estimating Diffusion Coefficients in Aqueous Organic Mixtures", Anal. Chem. 69(13): 2530-2536, 1997.

Lode F., G. Francesconi, M. Mazzotti and M. Morbidelli, "Synthesis of methylacetate in a simulated moving-bed reactor: Experiments and modeling", AIChE J. 49(6): 1516-1524, 2003.

98 CHAPTER 4. Fixed Bed Adsorptive Reactor: Experiments, Modelling and Simulation

Lode F., M. Houmard, C. Migliorini, M. Mazzotti and M. Morbidelli, "Continuous reactive chromatography", Chem Eng Sci. 56(2): 269-291, 2001.

Macias-Salinas R., F. Garcia-Sanchez and G. Eliosa-Jimenez, "An equation-of-state-based viscosity model for non-ideal liquid mixtures", Fluid Phase Equilib. 210(2): 319-334, 2003.

Mallmann T., B. D. Burris, Z. Ma and N. H. L. Wang, "Standing wave design of nonlinear SMB systems for fructose purification", AlChE J. 44(12): 2628-2646, 1998.

Mazzotti M., A. Kruglov, B. Neri, D. Gelosa and M. Morbidelli, "A continuous chromatographic reactor: SMBR", Chem Eng Sci. 51(10): 1827-1836, 1996.

Mazzotti M., B. Neri, D. Gelosa and M. Morbidelli, "Dynamics of a Chromatographic Reactor: Esterification Catalyzed by Acidic Resins", Ind Eng Chem Res. 36(8): 3163-3172, 1997.

Minceva M., P. S. Gomes, V. Meshko and A. E. Rodrigues, "Simulated moving bed reactor for isomerization and separation of p-xylene", Chem Eng J. 140(1-3): 305-323, 2008.

Motin M. A., M. H. Kabir and M. E. Huque, "Viscosities and excess viscosities of methanol, ethanol and n-propanol in pure water and in water + surf excel solutions at different temperatures", Phys. Chem. Liq. 43(2): 123-137, 2005.

Pereira C. S. M., P. S. Gomes, G. K. Gandi, V. M. T. M. Silva and A. E. Rodrigues, "Multifunctional Reactor for the Synthesis of Dimethylacetal", Ind Eng Chem Res. 47(10): 3515-3524, 2008.

Perkins L. R. and C. J. Geankoplis, "Molecular diffusion in a ternary liquid system with the diffusing component dilute", Chem. Eng. Sci. 24(7): 1035-1042, 1969.

Ruggieri R., G. Ranghino, G. Carvoli, A. Tricella, D. Gelosa and M. Morbidelli, "Process for esterification in a chromatographic reactor", Patent 6,586,609 (2003).

Ruthven D. M., "Principles of Adsorption and Adsorption Processes", Wiley & Sons, New York (1984).

Santacesaria E., M. Morbidelli, A. Servida, G. Storti and S. Carra, "Separation of xylenes on Y zeolites. 2. Breakthrough curves and their interpretation", Ind Eng Chem Process Des Dev 21(3): 446-451, 1982.

Scheibel E. G., "Correspondence. Liquid Diffusivities. Viscosity of Gases", Ind Eng Chem 46(9): 2007-2008, 1954.

Silva V. M. T. M. and A. E. Rodrigues, "Dynamics of a fixed-bed adsorptive reactor for synthesis of diethylacetal", AIChE J. 48(3): 625-634, 2002.

Silva V. M. T. M. and A. E. Rodrigues, "Novel process for diethylacetal synthesis", AIChE J. 51(10): 2752-2768, 2005.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 99

Troupe R. A., W. L. Aspy and P. R. Schrodt, "Viscosity and Density of Aqueous Lactic Acid Solutions", Ind Eng Chem 43(5): 1143-1146, 1951.

Umesi N. O. and R. P. Danner, "Predicting diffusion coefficients in nonpolar solvents", Ind Eng Chem Proc DD 20(4): 662-665, 1981.

Vignes A., "Diffusion in binary solutions: Variation of diffusion coefficient with composition", Ind. Eng. Chem. Fundam. 5(2): 189-199, 1966.

Yu W., K. Hidajat and A. K. Ray, "Modeling, Simulation, and Experimental Study of a Simulated Moving Bed Reactor for the Synthesis of Methyl Acetate Ester", Ind Eng Chem Res. 42(26): 6743-6754, 2003.

Yu W., K. Hidajat and A. K. Ray, "Determination of adsorption and kinetic parameters for methyl acetate esterification and hydrolysis reaction catalyzed by Amberlyst 15", Appl Catal A: Gen. 260(2): 191-205, 2004.

Zhang Y., L. Ma and J. Yang, "Kinetics of esterification of lactic acid with ethanol catalyzed by cation-exchange resins", React Funct Polym 61(1): 101-114, 2004.

5. Simulated Moving Bed Reactor

Abstract. A novel approach for the synthesis of ethyl lactate using a simulated moving bed reactor was evaluated by experiments as well as by simulations. A mathematical model considering external and internal mass-transfer resistances and variable velocity due to change of liquid composition was developed to describe the dynamic behaviour of the SMBR and it was validated by the experiments performed; it was observed that the experimental results were well predicted by the model. The effect of operating parameters, as the feed composition, SMBR configuration and switching time on the SMBR performance parameters at the optimal operating points and/or reactive/separation regions was studied. It was shown that the SMBR is a very attractive technology for the production of ethyl lactate, since under appropriate conditions the lactic acid conversion can be driven to completion and productivity as high as 32 kgEL/(Lads.day) and purity of 95 % can be obtained.

Adapted from: Pereira C. S. M., M. Zabka, V. M. T. M. Silva and A. E. Rodrigues, "A novel process for the ethyl lactate synthesis in a simulated moving bed reactor (SMBR)", Chem. Eng. Sci. 64(14): 3301-3310, 2009. 102 CHAPTER 5. Simulated Moving Bed Reactor

5.1 Introduction

Esterification and transesterification (Ferreira and Loureiro, 2004; Ma et al., 2008; Pereira et al., 2008b; Schmid et al., 2008; Yadav and Devi, 2004), acetalization (Chopade and Sharma, 1997; Gandi et al., 2005; Silva and Rodrigues, 2001; Yadav and Pujari, 1999), and etherification (Cruz et al., 2005; Yadav and Lande, 2005) reactions, that are limited by the thermodynamic equilibrium should be performed using hybrid technologies where reaction and separation take place in a single unit, like chromatographic reactors or membrane reactors, given that the reaction yield will be increased by removing one or more products. Moreover, the use of this kind of technologies allows cost reduction and higher product purity. In order to improve the process the use of solid acid catalysts is preferable. Since they are less corrosive and, in opposition to homogeneous ones, they not require a further step of separation and neutralization. Among the solid acid catalysts, ion exchange resins, like Amberlyst 15-wet (A15), offer to the ethyl lactate production the advantage of having a double role, they act as catalyst and as selective adsorber between the two reaction products, water and ethyl lactate. Therefore, the use of the Amberlyst 15-wet resin in a chromatographic reactor to produce ethyl lactate seems to be interest. The principle of the chromatography batch reactor has been developed by Dinwiddie and Morgan in the beginning of the 1960s (Dinwiddie and Morgan, 1961). The reactant is injected as a sharp pulse into a fixed-bed column; during its propagation along the column it reacts and leads to different products. The different affinity of the components with the solid phase leads to different propagation velocities and, because of that, it is possible to separate the products from the reaction mixture. For equilibrium limited reactions where complete separation of the products is achieved the conversion can overcome the thermodynamic equilibrium value. However, the batch chromatographic reactor is not the best technology to overcome that limitation, since it has the usual drawbacks of a batch operation, where high quantity of eluent is needed to perform the separation, resulting in highly diluted products. These led to the development of continuous chromatographic reactors in order to decrease the solvent consumption and to enhance the productivity, where the continuous mode of operation can be achieved by moving the adsorbent or simulating its motion. However, the flow of solid particles in the columns leads to some difficulties such as particle attrition, fluid velocity limited by fluidization phenomena, considerable pressure drops, lack of efficiency, that are overcome by the simulated moving bed (SMB) technology (Broughton and Gerhold, 1961), in which the counter-current movement of the solid is simulated by using a series of packed beds where PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 103

the inlet/outlet ports are synchronously shifted in direction of the liquid flow. The extension of the SMB technology to integrate reaction and separation steps into a single apparatus led to the Simulated Moving Bed Reactor (SMBR). This chromatographic reactor gained more and more attention in the last years because of its potential to improve process efficiency, as in the case of esterifications (Dunnebier et al., 2000; Kawase et al., 1996; Lode et al., 2003; Lode et al., 2001; Mazzotti et al., 1996; Meissner and Carta, 2002; Migliorini et al., 1999; Yu et al., 2003), etherifications (Zhang et al., 2001), sugars (Azevedo and Rodrigues, 2001; Da Silva et al., 2005; Da Silva et al., 2006; Dunnebier et al., 2000; Pilgrim et al., 2006; Zhang et al., 2004), bisphenol A (Kawase et al., 1999), acetals (Dubois, 2008a; Pereira et al., 2008a; Rodrigues and Silva, 2005; Silva and Rodrigues, 2005) and polyacetals (Dubois, 2008b) and will be applied in this work on the synthesis of the green solvent ethyl lactate. A schematic diagram of a SMBR unit is presented in Figure 5.1 where a reaction of type A+B↔C+D is considered. Similarly to the SMB, the SMBR consists of a set of columns connected in series packed with a solid, which could be a mixture of catalyst and selective adsorbent or a solid that acts both as catalyst and as adsorbent. Typically, there are two inlets, feed and desorbent, and two outlets, extract and raffinate. In this case, the component A is used as reactant and desorbent, therefore it is introduced in the system in the feed and desorbent streams. The other reactant B is used as feed. At regular time intervals, called the switching time, all streams are switched for one column distance in direction of the fluid flow. In this way, the countercurrent motion of the solid is simulated and its velocity is equal to the length of a column divided by the switching time. A cycle is completed when the number of switches is equal to a multiple of the total number of columns. According to the position of the inlet and outlet stream the unit can be divided in four sections. In section 1, positioned between the desorbent and extract nodes, the adsorbent is regenerated by desorption of the more strongly adsorbed product (D) from the solid. In section 2 (between the extract and feed node) and section 3 (between the feed and raffinate node) the reaction is taking place and the products C and D are formed. The more strongly adsorbed product, D, is adsorbed in sections 2 and 3 and transported with the solid phase to the extract port. The less strongly adsorbed product, C, is desorbed in sections 2 and 3 and transported with the liquid in direction of the raffinate port. In section 4, positioned between the raffinate and desorbent node, before being recycled to section 1, the desorbent is regenerated by adsorption of the less adsorbed product (C).

104 CHAPTER 5. Simulated Moving Bed Reactor

Desorbent (A) Raffinate (A+C)

12 11 10 9 8 7 Direction of fluid flow A + B Ù C + D and port switching

1 2 3 4 5 6

Extract Feed (A+D) (A+B)

Figure 5.1 Schematic diagram of a SMBR with three columns per section.

In this work an alternative technology, the SMBR, is proposed for the ethyl lactate production. The ethyl lactate synthesis is experimentally studied using a SMBR pilot unit and theoretically using the SMBR model proposed. This model results are verified by the experiments performed and the effect of various operating parameters, as feed composition, SMBR configuration and switching time, into the SMBR performance at the optimal operating points and/or reactive/separation regions is evaluated. The SMBR model is compared with the equivalent True Moving Bed Reactor (TMBR) model. All this study will allow to future compare the SMBR and the membrane reactor processes in economic and efficiency terms.

5.2 Modelling Strategies

5.2.1 SMBR mathematical model

The SMBR modelling strategy allows the visualization of the axial movement of concentration profiles and the variations in extract and raffinate concentrations within a period. The dynamic calculation of concentration profiles until the unit achieves the cyclic steady state, accounting for the inlet and outlet discontinuous shift, was made considering axial dispersion flow for the bulk fluid phase, linear driving force (LDF) approximation for the inter and intra-particle mass transfer rates, multi-component adsorption equilibrium and PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 105

velocity variations due to adsorption/desorption rates leading to more accurate results. The porosity of the packed bed and its length (packed bed length) were assumed to be constant.

The SMBR model equations are:

Bulk fluid mass balance to component i in column k:

∂∂CCuik ik k (1−ε ) 3 ∂ ⎛⎞ ∂ xik ++KCCLik,,,() ik −=pik, D axk⎜⎟ C T k (5.1) ∂∂tzε rp ∂ z⎝⎠ ∂ z

where Cik and C pik, are the bulk and average particle concentrations in the fluid phase of species i in column k respectively, K L,ik is the global mass transfer coefficient of the component i, ε is the bulk porosity, t is the time variable, z is the axial coordinate, Dax,k , and uk are the axial dispersion coefficient and the interstitial velocity in column k, respectively, and rp is the particle radius.

The global mass transfer coefficient ( K L ) is defined as:

1 1 1 = + (5.2) K L ke ε p ki

wherein ke and ki are, respectively, the external and internal mass transfer coefficients. The values of the external, internal and global mass transfer coefficients are shown in Table 5.1. The methods used to obtain these values and the determination of axial dispersion coefficient, Dax , are presented in detail in Chapter 4.

Table 5.1 Mass transfer coefficients estimated at 50 ºC at the inlet of the SMBR for a simulation where the feed is lactic acid solution (86 % in water) at a flowrate of 7 mL/min.

Component ke (cm/min) ki (cm/min) KL (cm/min) Ethanol 0.387 0.138 0.045 Lactic Acid 0.351 0.119 0.039

Ethyl Lactate 0.296 0.092 0.031

Water 0.623 0.280 0.090 106 CHAPTER 5. Simulated Moving Bed Reactor

Interstitial fluid velocity variation is calculated using the total mass balance:

du (1−ε ) 3 n k =−∑ KVLik,, moli() C ik − Cpik, (5.3) dzε rp i=1 where Vmol,,i is the molar volume of component i that is 60.87 mL/mol, 77.56 mL/mol, 118.44 mL/mol and 18.63 mL/mol at 50 ºC for ethanol, lactic acid, ethyl lactate and water, respectively.

Pellet mass balance to component i, in column k:

∂C pik, ∂qik 3 εεppLikikip+−()1() =KCC, () −pik,, + υρ rC pik (5.4) ∂∂ttrp

where qik is the average adsorbed phase concentration of species i in column k in equilibrium with C pik, , ε p the particle porosity, υi  the stoichiometric coefficient of component i, ρ p the particle density and r is the chemical reaction rate relative to the average particle concentrations in the fluid phase and is given by (Chapter 3):

a a a a − EL W Eth La K r = k c 2 (5.5) ⎛ W ⎞ ⎜1 + ∑ K s,i ai ⎟ ⎝ i=Eth ⎠

Multi-component adsorption equilibrium isotherm:

QKCpik, q = ads, i i (5.6) ik ⎛⎞n ⎜⎟1+ ∑ KCl pik, ⎝⎠l=1

where,Qads,i and Ki represent the total molar capacity per unit volume of resin and the equilibrium constant for component i, respectively, and n the total number of components. The adsorption parameters were determined and presented in Chapter 4.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 107

Initial and Danckwerts boundary conditions:

t = 0 : CCik==pik, C ik ,0 and qqik= ik,0 (5.7)

∂C ik z = 0 : uCkik−= D axk,, uC kikF (5.8a) ∂z z=0

uukk= ,0 (5.8b)

∂C ik z = Lc : = 0 (5.8c) ∂z z=Lc where F and 0 refer to the feed and initial states, respectively.

Mass balances at the nodes of the inlet and outlet lines of the SMBR:

uu1 D D Desorbent node: CCCij(4,)== zLc=− ij (1,0) == z i (5.9a) uu44

Extract(j=2) and Raffinate (j=4) nodes: Ci( j−1,z=Lc) = Ci( j,z=0) (5.9b)

u3 uF F Feed node: Ci(2,z=Lc) = Ci(3,z=0) − Ci (5.9c) u2 u2 where,

u1 = u4 + uDs Desorbent (Ds) node ; (5.10a)

u2 = u1 − uX Extract (X) node ; (5.10b)

u3 = u2 + uF Feed (F) node ; (5.10c)

u4 = u3 − uR Raffinate (R) node ; (5.10d)

The ratio between the fluid interstitial velocity, uj, and the simulated solid velocity, Us, L (defined by the column length and switching time relation,Us = ) could be defined for t * each section giving a new parameter:

u γ = j (5.10e) j Us 108 CHAPTER 5. Simulated Moving Bed Reactor

5.2.2 SMBR performance parameters

The SMBR process performance is calculated over a complete cycle according to the next equations:

tNt+ c * CdtR ∫ EL Raffinate Purity: PUR(%)= 100 t (5.11a) tNt+ c * CCCdtRRR++ ∫ ()La EL W t

tNt+ c * CdtX ∫ W Extract Purity: PUX (%)= 100 t (5.11b) tNt+ c * CCCdtXXX++ ∫ ()La EL W t

⎛ t++Ncct**t N t ⎞ ⎜ Q C X dt + Q C R dt ⎟ X ∫∫La R La ⎜ t t ⎟ Lactic acid Conversion: X (%) = 100⎜1− F ⎟ (5.12) ⎜ QF Cla N ct * ⎟ ⎜ ⎟ ⎝ ⎠

t+Nct* Q C R dt R ∫ EL ⎛⎞kg t Raffinate Productivity⎜⎟EL : PR = (5.13) ⎝⎠dayLresin ()1− ε Vunit N ct *

QDCEth,D + QF (CEth,F − XCLa,F ) Desorbent Consumption (LEth/kgEL): DC = N ct * (5.14) t+Nct* Q C R dt R ∫ EL t where Nc represents the total number of columns (the complete cycle). The productivity is defined considering the ethyl lactate (desired product) produced and withdrew from the raffinate stream. The ethanol consumed in the reaction is not taken into account to calculate the desorbent consumption, being only considered the amount of ethanol used as desorbent.

5.2.3 Numerical Solution

The above model equations were solved numerically by using the gPROMS-general PROcess Modelling System version: 3.0.3. The mathematical model involves a system of partial and algebraic equations (PDAEs). The axial domain was discretized using third order orthogonal collocation in finite elements method (OCFEM). Ten finite elements per column with two PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 109

collocation points in each element were used. The system of ordinary differential and algebraic equations (ODAEs) was integrated over time using the DASOLV integrator implementation in gPROMS. For all simulations was fixed a tolerance equal to 10-5. It is assumed that an SMB simulation has reached the cyclic steady state when the columns profiles in two consecutive cycles have less than 1.0 % of relative deviation and the global mass balance is verified with less than 1.0 % of relative error.

5.3 Experimental Section

5.3.1 Chemicals and Catalyst / Adsorbent

The chemicals used in the experiments were ethanol (>99.5 % in water), lactic acid (>85 % in water) and ethyl lactate (>98 % in water) from Sigma-Aldrich (U.K.).

The columns were packed with Amberlyst 15-wet (Rohm & Haas, France), which is a highly cross-linked polystyrene-divinylbenzene ion exchange resin functionalized with sulfonic groups (SO3H) that acts as catalyst and adsorbent in this system.

5.3.2 The SMBR LICOSEP 12-26 Unit

All SMBR experiments were performed in a pilot unit LICOSEP 12-26 by Novasep (Vandoeuvre-dès-Nancy, France), where 12 columns Superformance SP 230 x 26 (length x internal diameter, mm), by Götec Labortechnik (Mühltal, Germany), packed with the acid resin Amberlyst 15-wet were connected. The characteristics of the SMBR columns are presented in Table 5.2 and a side view of the pilot unit is shown in Figure 5.2. The bed porosity and the Peclet number were determined as in previous Chapter (Chapter 4). These columns can withstand up to 60ºC of temperature and 60 bar of pressure. The operating temperature was 50ºC; this temperature was ensured in the jacketed columns using a thermostatic bath (Lauda, Germany). Between every two columns exists a four-port valve (Top-Industrie, France) actuated by the control system. When required, according to the operating conditions set into the system, the valves allow either pumping of the feed or desorbent into the system or withdrawal of extract or raffinate from the system. HPLC pumps are used to pump each of the inlet streams (feed or desorbent) and outlet streams (extract or raffinate). The maximum flow rate in the desorbent and extract pumps is 30 mL/min, while in 110 CHAPTER 5. Simulated Moving Bed Reactor

the feed and raffinate pumps is 10 mL/min. The recycling pump, which is a positive displacement three-headed membrane pump (Dosapro Milton Roy, France), can deliver flow rates as low as 20 mL/min up to 120 mL/min and it can hold up to 100 bar pressure. A six- port valve, located between the twelfth and the first columns, was used to collect the samples for internal concentration measurements (Novasep). All the samples were analysed in a gas chromatograph using the analytical method described in Chapter 3.

Figure 5.2 Licosep 12-26 unit with the 12 packed columns.

Table 5.2 Characteristics of the SMBR columns.

Solid weight (A15) 47.6 g

Length of the bed (L) 23 cm

Internal diameter (Di) 2.6 cm

Average radius of resin beads (rp) 342.5 μm

3 Bulk density (ρ b) 390 kg/m

Bed porosity (ε) 0.4

Resin particle porosity (εp) 0.36 (Lode et al., 2001)

Peclet number 300 PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 111

5.4 Results and Discussion

5.4.1 Experimental Results

Three different experiments of ethyl lactate synthesis were performed at 50ºC in the SMBR unit. In all the experiments the feed composition was a lactic acid solution (85 % in water).

The flow rates of feed and raffinate streams were QF = 1.80 mL/min and QR = 8.80 mL/min, while for desorbent and recycle the flow rates were QD = 27.40 mL/min and

QRe c = 24.70 mL/min. Two different switching times (2.9 and 3.0 min) and two different configurations (3-3-3-3 and 3-3-4-2) where tested, as shown in Table 5.3. The SMBR performance parameters obtained experimentally are presented in Table 5.3, as well as the corresponding values obtained by simulation. In Figure 5.3, the cyclic steady-state (CSS) concentration profiles obtained experimentally at the middle of the switching time after 13 cycles are shown and compared with the profiles simulated by the SMBR model.

Table 5.3 Operating conditions for SMBR experiments and performance parameters obtained experimentally (bold) and by simulation (inside brackets).

Run 1 Run 2 Run 3

t* (min) 2.9 2.9 3

Configuration 3-3-3-3 3-3-4-2 3-3-4-2

PUR (%) 72.75 (79.24) 73.54 (79.80) 75.17 (81.32)

PUX (%) 95.54 (99.89) 95.22 (99.83) 97.76 (99.85)

X (%) 99.10 (97.46) 99.91 (97.87) 99.96 (98.00)

-1 -1 PR (kgEL.Lresin .day ) 3.20 (3.66) 3.31 (3.68) 3.36 (3.68)

DC (LEth/kgEL) 13.41 (11.74) 12.99 (11.70) 12.77 (11.68)

For the first experiment the lactic acid conversion is about 99 %, being for the second and third experiments around 100 %. However the raffinate purities are low, this is due to the fact that the SMBR experiments realized were performed under conditions of incomplete adsorbent regeneration in section 1 (see Figure 5.3) caused by equipment limitations (maximum allowable desorbent flow rate of 30 mL/min). 112 CHAPTER 5. Simulated Moving Bed Reactor

Figure 5.3 Experimental concentration profiles in SMBR unit at the middle of switching time at cyclic steady state (13th cycle) and simulated curves. (a) Run 1; (b) Run 2; (c) Run3.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 113

The feed is a lactic acid solution (85 % in water), so water is already introduced into the system and as the reaction proceeds more water is being formed. Since water is the most adsorbed component and there is a significant amount of this component in the SMBR, the desorbent flow rate used was not high enough to completely regenerate (removing the water adsorbed by the resin) the first section; because of that the adsorbed water was carried by the resin entering in section 4 and transported to section 3, hydrolysing the ethyl lactate producing ethanol and lactic acid (the opposite reaction). As consequence the ethyl lactate productivity and its concentration in the raffinate stream is low. The use of a higher desorbent flow rate will permit the complete regeneration of section 1 leading to higher purities and productivity. From Figure 5.3, it can be perceived that the reaction takes place in sections 2 and 3; since the esterification reaction between lactic acid and ethanol is very slow, the change of configuration from 3-3-3-3 to 3-3-4-2 increases the lactic acid conversion, due to the increase of the number of columns available in the reactive sections 2 and 3. Increasing the switching time (Figure 5.3c), increases the lactic acid conversion once the contact time between liquid and solid streams is higher. Moreover, the ethyl lactate purity increases since the value of γ1 is higher and therefore the resin is better regenerated. The CSS concentration profiles obtained experimentally at 25 %, 50 % and 75 % of the switching time, at the 13th cycle are compared with the simulated ones, in Figure 5.4. It is noticed that all the species concentration fronts propagate along the column without changing its shape presenting, therefore, a constant behaviour along the time. For the third experiment, the experimental composition of extract and raffinate streams collected for a whole cycle at the 3rd, 5th, 7th, 9th and 11th cycles, and the extract and raffinate concentration histories calculated from the SMBR model are shown in Figure 5.5.

As it can be observed, the mathematical model used predicts with reasonable accuracy the experimental concentration profiles (Figure 5.3 and Figure 5.4), the raffinate and extract histories (Figure 5.5) and the performance parameters (Table 5.3).

114 CHAPTER 5. Simulated Moving Bed Reactor

Figure 5.4 Experimental concentration profiles in SMBR unit at the 25 %, 50% and 75 % of the switching time at cyclic steady-state (13th cycle) and simulated curves for Run 3.

a) b)

Figure 5.5 Experimental and theoretical average concentration of all species (ethanol, lactic acid, ethyl lactate and water) for the conditions of the third experience in the extract (a) and raffinate (b) streams.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 115

5.4.2 Simulated results

5.4.2.1 Comparison of SMBR and TMBR models

The equivalent TMBR model was also studied since it is a simpler model that implies less computational time than that required by the SMBR model. However, comparing the concentration profiles at the cyclic steady state calculated from the SMBR model with those calculated from the equivalent TMBR model, as shown in Figure 5.6, some difference is noticed and, because of that, the SMBR model was used for all the simulations presented in this work. The operating conditions used in the simulations are presented in Table 5.4 and the feed flow rate and raffinate flow rate used were 5 mL/min and 17 mL/min, respectively.

Figure 5.6 Comparison of the concentration profiles at the middle of the switching time at cyclic steady state determined with the SMBR model (solid lines) and steady state concentration profiles calculated with the TMBR model (dashed lines), for the operating conditions presented in Table 5.

116 CHAPTER 5. Simulated Moving Bed Reactor

Table 5.4 Operating conditions.

Operating conditions Configuration 3-3-4-2 (analyzed in section 5.4.2.5)

Feed Concentration (mol/L) CEth,F =0.0; CEL,F = 0.0 (analyzed in section 5.4.2.4) CLA,F =10.75; CW,F = 9.48 Desorbent flow rate (mL/min) 58.0

Recycle flow rate (mL/min) 27.0 Switching time (min) 2.7 (analyzed in section 5.4.2.6)

5.4.2.2 Reactive/separation regions

The proper design of a SMBR unit involves the correct choice of the operating conditions as flow rates in each section of the unit, switching time, columns arrangement (SMBR configuration) and feed concentration, among others. In the next sub-sections, the reactive/separation regions and the optimal operating points for different operating conditions are presented in order to understand the behaviour of the SMBR process for ethyl lactate synthesis. The operating conditions used in the simulations indicating which parameter is being changed in each sub-section are presented in Table 5.4. The reactive/separation regions setting a criteria of 95 % for extract and raffinate purity and, also, for lactic acid conversion were determined from the average concentrations over a cycle obtained by the SMBR model at cyclic steady state. The cyclic steady state SMBR model was successively solved for several values of γ2 and γ3, keeping the values of γ1 (4.698) and γ4 (1.492). The reactive/separation region is located within the region between the diagonal γ2=γ3, the horizontal line γ3 =1.492 and γ3 axis. The γ3 value must be higher than γ2, since the diagonal

γ2=γ3 corresponds to zero feed flow rate. The algorithm used in the construction of the reactive/separation region begins by setting a feed flow rate of 0.01 ml/min and the value of

γ2 equal to 1.492. Then, the feed flow rate was kept constant and the γ2 values were gradually increased. The value of γ3 was calculated from the mass balance in the feed node for each value of γ2. For each set of γ2 and γ3, the conversion and the purities of extract and raffinate were estimated and the values that satisfy the criteria of 95 % were selected to build the reaction/separation region. After this set of simulations, the feed flow rate was increased and PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 117

the same procedure was repeated. The simulations procedure ends when the maximum value of feed flow rate that gives required product purities and conversion is achieved. Above that feed flow rate value the requirements can not be fulfilled for any pair of values of γ2 and γ3.

5.4.2.3 Separation Region vs Reactive/Separation Region

One can use a SMB unit to perform the separation of ethyl lactate and water, in absence of reaction, considering the purity criteria of 95 %, to process a feed with composition of

CW = 12.17 mol/L and CEL = 6.47 mol/L that corresponds to complete conversion of the lactic acid solution (CAL = 10.75 mol/L and CW = 9.48 mol/L) into ethyl lactate and water, taking into consideration the expression , where x is the molar fraction of Cxii= ∑( xV nmk, ) i k species i and Vm,n is the molar volume of species n.

In Figure 5.7, the comparison between the separation region of the SMB and the reactive/separation region of the SMBR indicates that the process is limited by the reaction. As mentioned by Fricke and Schmidt-Traub, the reactive separation region is reduced by decreasing either the equilibrium constant or the reaction kinetics (Fricke and Schmidt-Traub, 2003). The lactic acid esterification reaction is slow and, because of that, it is necessary to keep the reactant as long as possible in sections 2 and 3 (between the extract and raffinate ports). If the lactic acid is not fully consumed it will preferentially contaminate the raffinate stream, since the resin selectivity between lactic acid and ethyl lactate is smaller than the one between water and lactic acid. Thus, in the SMBR it is necessary to ensure the separation of the products (ethyl lactate and water), but is crucial to guarantee the complete conversion of the limiting reactant (lactic acid). One way of improving reaction kinetics is by increasing the system temperature that will also benefit the equilibrium constant since the ethyl lactate synthesis is an endothermic reaction. However, this leads to higher energy requirement and it will affect the products separation: (i) if water adsorption is more affected than ethyl lactate adsorption, less desorbent consumption is required, but maximum productivity decreases; (ii) in the opposite case, productivity is enhanced while the consumption of desorbent increases. The trade-off between those parameters should be evaluated through an economical assessment. Another way to get the full conversion is increasing the reaction rate by using a mixture of two stationary phases that increases the catalytic properties without losing 118 CHAPTER 5. Simulated Moving Bed Reactor

efficiency on the separation of the products (Silva and Rodrigues, 2008; Ströhlein et al., 2004).

4.0

3.5

3.0 3

γ 2.5

2.0

1.5 Feed: lactic acid solution (SMBR) Feed: ethyl lactate and water (SMB) 1.0 1.0 1.5 2.0 2.5 3.0 3.5 4.0 γ2 Figure 5.7 Reactive/separation region for the case where lactic acid solution is fed to the SMBR (CLA,F = 10.75 mol/L; CW,F = 9.48 mol/L) and separation region for ethyl lactate and water fed to the SMB (CEL,F = 6.47 mol/L; CW,F = 12.17 mol/L). (γ1 = 4.698; γ4 =1.492; 95 % purity).

5.4.2.4 Effect of the Feed Composition

In order to study the influence of the feed composition, this was varied by adding pure ethanol to the lactic acid solution remaining constant the other operation conditions (see Table 5.4). As it can be seen in Figure 5.8, the size of the reactive/separation region decreases with the increase of the lactic acid concentration in the feed. This could be justified by a lack of ethanol in the sections 2 and 3, where reaction occurs, limiting the lactic acid conversion to the thermodynamic equilibrium and so higher lactic acid concentrations will lead to higher quantities of unreacted lactic acid and therefore the raffinate stream will be contaminated. This could be confirmed by analyzing the influence of the lactic acid feed molar fraction in the ethyl lactate productivity and the desorbent consumption for the optimal operating points (see Figure 5.9). It can be seen that by increasing the lactic acid feed molar fraction until 72 %, where the lactic acid is the limiting reactant, the productivity increases with consequent decrease in desorbent consumption; however, further increase in the lactic acid feed molar fraction has a negative impact in the SMBR performance parameters since, as it was mentioned, ethanol is the limiting reactant in sections 2 and 3. Thus, for the values of γ1 and PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 119

γ4 and SMBR configuration defined in Table 5.4, the feed molar composition of 72 % of lactic acid leads to the optimal values of productivity (17.33 kgEL/(Lads.day)) and desorbent consumption (5.36 LEth/kgEL). So, it can be concluded that the lactic acid concentration in the SMBR feed should be high enough to increase the productivity without leading to a lack of ethanol in the reaction zone (section 2 and 3).

3.5

3.0 3

γ 2.5

100 % La solution 2.0 88 % La solution 38 % La solution 1.5 1.5 2.0 2.5 3.0 3.5 γ2

Figure 5.8 Reactive/separation regions for different feed concentrations of lactic acid solution. (γ1 = 4.698; γ4 =1.492; 95 % purity).

18.0 9.0 PR 17.0 DC 8.2

16.0 ) .day)) 7.4 EL ads /Kg

/(L 15.0 Eth EL 6.6 14.0 DC (L DC PR (Kg PR 5.8 13.0

12.0 5.0 35 45 55 65 75 85 95 Lactic acid solution (85 % in water) feed composition (%)

Figure 5.9 SMBR performance for the optimal operating points as a function of the lactic acid solution molar fraction in feed.

120 CHAPTER 5. Simulated Moving Bed Reactor

5.4.2.5 Effect of the SMBR columns arrangement

The influence of the columns arrangement on the SMBR system performance was studied by simulating different SMBR configurations (3-3-3-3, 3-3-4-2 and 3-3-5-1), keeping the remaining operating conditions as mentioned in Table 5.4. The performance parameters corresponding to the optimal operating point (vertex of the reactive/separation region) for each case are presented in Table 5.5, and the reactive/separation regions are shown in Figure 5.10. It can be perceived that, the introduction of more columns into the section 3 by taking of the section 4 and keeping the total number of columns constant, does not improve significantly the process performance. This is due to the fact that after about 3 mL/min of lactic acid solution feed flow rate the ethanol is the limiting reactant and even when the reaction zone is increased is not possible to completely convert the lactic acid, which will cause contamination of the raffinate stream. Although the increase of the reaction zone in these conditions does not improve much the ethyl lactate production on the SMBR unit it was noticed that only one column in section 4 is enough to allow the complete regeneration of the desorbent (see Figure 5.11).

The kinetics of the ethyl lactate synthesis is very slow and to allow full conversion of the lactic acid is necessary not only to increase the reaction zone, but also to feed to the SMBR a lactic acid concentration that does not lead to a lack of ethanol in the reaction zone.

3.5 conf 3-3-3-3 conf 3-3-4-2 3.0 conf 3-3-5-1 3

γ 2.5

2.0

1.5 1.5 2.0 2.5 3.0 3.5 γ2

Figure 5.10 Reactive/separation regions for different SMBR configurations. (γ1 = 4.698; γ4 =1.492; 95 % purity).

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 121

Table 5.5 Performance parameters of the SMBR unit at cyclic steady state for the optimal operation points for the different SMBR configurations.

Columns arrangement 3-3-3-3 3-3-4-2 3-3-5-1

Raffinate productivity (kgEL/LA15/day) 14.97 15.33 15.36

Desorbent consumption (LEth/kgEL) 5.83 5.68 5.67

Figure 5.11 Concentration profiles at the middle of the switching time at cyclic steady state for the optimal operating point of the SMBR configuration 3-3-5-1.

5.4.2.6 Effect of Switching Time

The switching time effect on the SMBR performance for the optimal operating point was studied by two different ways: one varying the switching time and keeping constant the remaining operating parameters of Table 5.4 (Figure 5.12); the other varying the switching time and the desorbent and recycle flow rate in order to keep constant the γ1 (4.698) and

γ4 (1.492) values (Figure 5.13). For the first case (Figure 5.12), as the switching time decreases until 2.1 minutes, the solid flow rate increases leading to a better SMBR performance. However, decreasing the switching time below 2.1 min, the restriction of γ1 is being violated and, therefore, is no longer possible to achieve the minimum purity of 95 % for the raffinate stream, since the solid is not being completely regenerated in section 1, and, because of that, the water appears in the raffinate stream, leading to a low purity. The SMBR 122 CHAPTER 5. Simulated Moving Bed Reactor

operation using a switching time of 2.1 min instead of 2.8 min enhances the productivity in

22.0 % (maximum of 18.06 kgEL/(Lads.day)) and decreases the desorbent consumption in

19.6 % (minimum of 4.75 LEth/kgEL), and will reduce downstream costs since the products are less diluted.

19.0 6.0 PR 18.0 DC

5.6 ) .day)) EL

ads 17.0 /Kg /(L

5.2 Eth EL 16.0 DC (L

PR (Kg PR 4.8 15.0

14.0 4.4 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 Switching time (min)

Figure 5.12 SMBR performance for the optimal operating points as a function of the switching time.

For the second case (Figure 5.13), where the γ1 and γ4 values are keeping the same, similarly to the previous case, the productivity increases by decreasing the switching time; contrarily, the desorbent consumption increases once the desorbent flow rate is increasing in order to keep the value of γ1 constant. Changing the switching time from 2.8 to 2.1, increases the productivity and desorbent consumption to 19.34 kgEL/(Lads.day) and 5.79 LEth/kgEL, corresponding to a variation of +30.7 % and +2.0 %, respectively. The reduction of 2.8 min to 1.0 min, in switching time, leads to a significant increase in productivity in 114.4 %

(31.7 kgEL/(Lads.day)), while the desorbent consumption is increased in 33.3 %

(7.6 LEth/kgEL). The choice of the optimal switching time value requires an economical assessment to the whole process, considering the downstream separation units, in order to verify if the enhancement of productivity reimburses the increase of the cost associated to the desorbent separation and recycle to the SMBR unit.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 123

32.0 7.6

PR 28.0 7.2 DC ) EL .day)) 24.0 6.8 ads /Kg /(L Eth EL 20.0 6.4 DC (L

PR (Kg PR 16.0 6.0

12.0 5.6 1.01.41.82.22.6 Switching time (min)

Figure 5.13 SMBR performance for the optimal operating points as a function of the switching time keeping γ1 = 4.698 and γ4 =1.492.

5.5 Conclusions

The ethyl lactate was produced in the Simulated Moving Bed Reactor pilot unit LICOSEP, using Amberlyst 15-wet resin as catalyst and selective adsorbent. The experiments conducted were not performed under optimal operating conditions since they were limited by restrictions of the experimental set-up; therefore, the desorbent flow rate used was not enough to achieve complete regeneration of the resin in section 1. The performance parameters and the experimental profiles were predicted with good accuracy using the SMBR model. The theoretical assessment of the SMBR unit behaviour was performed ensuring complete regeneration of the resin (in section 1) and desorbent (in section 4), by using the mathematical model to analyse the effect of SMBR configuration, feed composition and switching time into the reactive/separation regions or/and into the process performance at the optimal operating points. Since the kinetics of the ethyl lactate synthesis is very slow, even at 50ºC, its production in the SMBR is determined by the reaction, and therefore, the complete conversion of lactic acid is a decisive parameter in order to optimize the SMBR performance. To get that target, several aspects should be regarded: to play with feed composition to avoid a lack of ethanol in the reactive sections 2 and 3; to increase the number of columns in those sections to increase the reaction zone; to choose the best switching time, that affects significantly the SMBR process. The other key aspect that should be considered is the water 124 CHAPTER 5. Simulated Moving Bed Reactor

desorption in section 1 to ensure the complete regeneration of the resin; Amberlyst 15-wet is very selective to water and therefore higher desorbent flow rates are required. The effects of the various operating parameters, as the feed composition, the SMBR configuration and the switching time, on the SMBR process showed that there is a complex interaction among all of them in terms of their impacts on the performance of the process, as the extract and raffinate purities, lactic acid conversion, ethyl lactate productivity and desorbent consumption. Moreover, some set of parameters might act in conflicting ways, like the case of the switching time for fixed values of γ1 and γ4, that increase significantly the productivity but also increases the desorbent consumption. For a fixed switching time of 2.7 min the best feed molar composition was 72 % of lactic acid leading to a productivity of 17.33 kgEL/(Lads.day) and desorbent consumption of 5.36 LEth/kgEL. The higher value of productivity,

31.7 kgEL/(Lads.day), was obtained for a switching time of 1 min, 100% of lactic acid solution in feed, γ1 = 4.698 and γ4 =1.492.

5.6 Notation

a liquid-phase activity

C liquid phase concentration (mol/L)

CT total liquid phase concentration (mol/L)

2 Dax axial dispersion coefficient (m /min)

DC desorbent consumption (L/mol)

-1 -1 kc kinetic constant (mol kg min )

ke external mass transfer coefficient

Keq equilibrium reaction constant

ki internal mass transfer coefficient

Ki Langmuir equilibrium constant of component i (L/mol)

KL global mass transfer coefficient PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 125

L column length (m) n total number of components

Nc total number of columns

PR raffinate productivity (kgEL/(L resin.day))

PUR raffinate purity (%)

PUX extract purity (%) q solid phase concentration in equilibrium with the fluid concentration inside the particle (mol/L)

Qads, i adsorption capacity of component i (mol/Lwet solid)

Q j volumetric flowrate in section j (L/min) r rate of reaction (mol kg-1 min-1)

rp particle radius (m) t time variable (min) t * switching time (min)

Us solid velocity (m/min) u interstitial velocity (m/min)

Vmol, i molar volume of species i (L/mol)

Vunit volume of adsorbent in SMBR (L) x mole fraction

X lactic acid conversion z axial coordinate (m)

Greek letters

γ interstitial velocities ratio

ε bulk porosity 126 CHAPTER 5. Simulated Moving Bed Reactor

ε p  particle porosity

υi  stoichiometric coefficient of component i

ρ p particle density

Subscripts i relative to component i (i= Eth, La, EL, W)

j relative to section in SMBR (j = 1, 2, 3, 4) k relative to column in SMBR

Eth relative to ethanol

La relative to lactic acid

EL relative to ethyl lactate

W relative to water

0 relative to initial conditions

p relative to particle

F relative to the feed

R relative to raffinate

Re c relative to recycle

X relative to extract

Superscripts

F relative to the feed

R relative to raffinate

X relative to extract

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 127

5.7 References Cited

Broughton D. B. and C. G. Gerhold, "Continuous sorption process employing fixed bed of sorbent and moving inlets and outlets.", Patent 2 985 589 (1961).

Chopade S. P. and M. M. Sharma, "Reaction of ethanol and formaldehyde: use of versatile cation-exchange resins as catalyst in batch reactors and reactive distillation columns", React. Funct. Polym. 32(1): 53-65, 1997.

Cruz V. J., J. F. Izquierdo, F. Cunill, J. Tejero, M. Iborra and C. Fité, "Acid ion-exchange resins catalysts for the liquid-phase dimerization/etherification of isoamylenes in methanol or ethanol presence", React. Funct. Polym. 65(1-2): 149-160, 2005.

Da Silva E. A. B., A. A. U. De Souza, S. G. U. De Souza and A. E. Rodrigues, "Simulated moving bed technology in the reactive process of glucose isomerization", Adsorption 11(1 SUPPL.): 847-851, 2005.

Da Silva E. A. B., A. A. U. De Souza, S. G. U. De Souza and A. E. Rodrigues, "Analysis of the high-fructose syrup production using reactive SMB technology", Chem. Eng. J. 118(3): 167-181, 2006.

Dinwiddie J. A. and W. A. Morgan, " Fixed bed type reactor", Patent 2 976 132 (1961).

Dubois J.-L., "Procede de synthese d' acetals par transacetalisation dans un reacteur a lit mobile simule", Patent FR 2909669 (A1) (2008a).

Dubois J.-L., "Process for the synthesizing polyacetals in a simulated mobile-bed reactor", Patent WO 2008/053108 (2008b).

Dunnebier G., J. Fricke and K. U. Klatt, "Optimal Design and Operation of Simulated Moving Bed Chromatographic Reactors", Ind. Eng. Chem. Res. 39(7): 2290-2304, 2000.

Ferreira M. V. and J. M. Loureiro, "Number of actives sites in TAME synthesis: Mechanism and kinetic modeling", Ind. Eng. Chem. Res. 43(17): 5156-5165, 2004.

Fricke J. and H. Schmidt-Traub, "A new method supporting the design of simulated moving bed chromatographic reactors", Chem. Eng. Process. 42(3): 237-248, 2003.

Gandi G. K., V. M. T. M. Silva and A. E. Rodrigues, "Process Development for Dimethylacetal Synthesis: Thermodynamics and Reaction Kinetics", Ind. Eng. Chem. Res. 44(19): 7287-7297, 2005.

Kawase M., Y. Inoue, T. Araki and K. Hashimoto, "The simulated moving-bed reactor for production of bisphenol A", Catal. Today 48(1-4): 199-209, 1999. 128 CHAPTER 5. Simulated Moving Bed Reactor

Kawase M., T. B. Suzuki, K. Inoue, K. Yoshimoto and K. Hashimoto, "Increased esterification conversion by application of the simulated moving-bed reactor", Chem. Eng. Sci. 51(11): 2971-2976, 1996.

Lode F., G. Francesconi, M. Mazzotti and M. Morbidelli, "Synthesis of methylacetate in a simulated moving-bed reactor: Experiments and modeling", AlChE J. 49(6): 1516-1524, 2003.

Lode F., M. Houmard, C. Migliorini, M. Mazzotti and M. Morbidelli, "Continuous reactive chromatography", Chem. Eng. Sci. 56(2): 269-291, 2001.

Ma L., J. Hong, M. Gan, E. Yue and D. Pan, "Kinetics of esterification and transesterification for biodiesel production in two-step process", J. Chem. Ind. Eng. China 59(3): 708-712, 2008.

Mazzotti M., A. Kruglov, B. Neri, D. Gelosa and M. Morbidelli, "A continuous chromatographic reactor: SMBR", Chem. Eng. Sci. 51(10): 1827-1836, 1996.

Meissner J. P. and G. Carta, "Continuous regioselective enzymatic esterification in a simulated moving bed reactor", Ind. Eng. Chem. Res. 41(19): 4722-4732, 2002.

Migliorini C., M. Fillinger, M. Mazzotti and M. Morbidelli, "Analysis of simulated moving- bed reactors", Chem. Eng. Sci. 54(13-14): 2475-2480, 1999.

Novasep, "Licosep 12-26 Instructions manual, France",

Pereira C. S. M., P. S. Gomes, G. K. Gandi, V. M. T. M. Silva and A. E. Rodrigues, "Multifunctional Reactor for the Synthesis of Dimethylacetal", Ind. Eng. Chem. Res. 47(10): 3515-3524, 2008a.

Pereira C. S. M., S. P. Pinho, V. M. T. M. Silva and A. E. Rodrigues, "Thermodynamic equilibrium and reaction kinetics for the esterification of lactic acid with ethanol catalyzed by acid ion-exchange resin", Ind. Eng. Chem. Res. 47(5): 1453-1463, 2008b.

Pilgrim A., M. Kawase, F. Matsuda and K. Miura, "Modeling of the simulated moving-bed reactor for the enzyme-catalyzed production of lactosucrose", Chem. Eng. Sci. 61(2): 353- 362, 2006.

Rodrigues A. E. and V. M. T. M. Silva, "Industrial process for acetals production in a simulated moving bed reactor", Patent PT 103123, 2004; WO 2005/113476A1 (2005).

Schmid B., M. Döker and J. Gmehling, "Esterification of ethylene glycol with acetic acid catalyzed by Amberlyst 36", Ind. Eng. Chem. Res. 47(3): 698-703, 2008.

Silva V. M. T. M. and A. E. Rodrigues, "Synthesis of diethylacetal: thermodynamic and kinetic studies", Chem. Eng. Sci. 56(4): 1255-1263, 2001.

Silva V. M. T. M. and A. E. Rodrigues, "Novel process for diethylacetal synthesis", AlChE J. 51(10): 2752-2768, 2005.

Silva V. M. T. M. and A. E. Rodrigues, "Influence of Different Mixtures of Catalyst and Adsorbent on the Smbr Performance: Reactive-Separation and Regeneration Regions ", 2008. PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 129

Ströhlein G., F. Lode, M. Mazzotti and M. Morbidelli, "Design of stationary phase properties for optimal performance of reactive simulated-moving-bed chromatography", Chem. Eng. Sci. 59(22-23): 4951-4956, 2004.

Yadav G. D. and K. M. Devi, "Immobilized lipase-catalysed esterification and transesterification reactions in non-aqueous media for the synthesis of tetrahydrofurfuryl butyrate: comparison and kinetic modeling", Chem. Eng. Sci. 59(2): 373-383, 2004.

Yadav G. D. and S. V. Lande, "Rate intensive and selective etherification of vanillin with benzyl chloride under solid-liquid phase transfer catalysis by aqueous omega phase.", J. Mol. Catal. A: Chem. 244: 271, 2005.

Yadav G. D. and A. A. Pujari, "Kinetics of acetalization of perfumery aldehydes with alkanols over solid acid catalysts.", Can. J. Chem. Eng. 77, 1999.

Zhang Y., K. Hidajat and A. K. Ray, "Optimal design and operation of SMB bioreactor: Production of high fructose syrup by isomerization of glucose", Biochem. Eng. J. 21(2): 111- 121, 2004.

Zhang Z., K. Hidajat and A. K. Ray, "Application of Simulated Countercurrent Moving-Bed Chromatographic Reactor for MTBE Synthesis", Ind. Eng. Chem. Res. 40(23): 5305-5316, 2001.

6. Pervaporation Membrane Reactor

Abstract. Integrated membrane reactors are one of the most promising and sustainable technologies to carry out equilibrium limited reactions. However, that integration leads to lower process flexibility when trying to find suitable operating conditions for both reaction and pervaporation process.

Pervaporation process using commercial microporous silica water selective membranes was evaluated to contribute either for the separation of SMBR extract stream (water/ethanol mixtures) or for the ethyl lactate process intensification by continuous pervaporation membrane reactor (PVMR). Preliminary studies were performed in order to assess the existence of membrane defects and mass transfer limitations, by studying the influence of feed pressure and flowrate, respectively. After, in the absence of mass transfer limitations, membrane performance was evaluated experimentally, at different composition and temperature measuring the flux and selectivity of each species in binary mixtures (water/ethanol, water/ethyl lactate and water/lactic acid). Thus, species permeances were obtained for each experiment and correlated in order to account for the effect of temperature and feed composition. Permeances of ethanol and ethyl lactate depend solely on the temperature, following an Arrhenius equation; for water, its permeance follows a modified Arrhenius equation taking into account also the dependence on the feed water content. Mathematical models, considering concentration and temperature polarization, and non- isothermal effects as well, were developed and applied to analyze the performance of batch pervaporation and continuous pervaporation membrane reactor, in both isothermal and non- isothermal conditions. The PVMR with 5 membranes in series, operating at 70ºC, leads to 98 % of lactic acid conversion and 96 % of ethyl lactate purity.

132 CHAPTER 6. Pervaporation Membrane Reactor

6.1 Introduction

In the past few years, the interest in the application of reactive separations to many chemical processes has substantially increased. This is particularly true for equilibrium limited reactions, where the removal of at least one of the reaction products shifts the equilibrium towards the product formation. Among all the separation techniques, pervaporation is becoming a promising technology, potentially useful in applications such as dehydration of organic mixtures (Chapman et al., 2008; Slater et al., 2006; Van Hoof et al., 2006; Yoshikawa et al., 2002), separation of organic mixtures (Lin et al., 2009; Smitha et al., 2004) and removal of organic compounds from aqueous solutions (Khayet et al., 2008; Ohshima et al., 2005). Aiming to produce esters, where water is a by-product, membranes suitable for dehydration of organic mixtures can be applied for esters process intensification; being polymeric (Hasanoglu et al., 2007; Krishna Rao et al., 2007; Namboodiri et al., 2006) and ceramic (Asaeda et al., 2005; Casado et al., 2005; Li et al., 2009) membranes the most used. Based on this, several works have been made regarding the use of pervaporation membrane reactors for chemicals production (Feng and Huang, 1996; Lim et al., 2002; Park and Tsotsis, 2004; Peters et al., 2005; Zhu et al., 1996). In the literature, two main types of pervaporation membrane reactors can be found; reactor and membrane housed in separate units (Figueiredo et al., 2008; Korkmaz et al., 2009; Sanz and Gmehling, 2006) or membrane and reactor incorporated into a single unit (de la Iglesia et al., 2007). Although the number of works dealing with PVMR in the last years is significant, most of them use simplified mathematical models to optimize and predict the behaviour of PVMR units. Usually, concentration and temperature polarization effects on the pervaporation process are neglected, which can be very significant under certain PVMR working conditions (Gómez et al., 2007).

In pervaporation processes, the transport of the components from the feed liquid mixture to the vapor phase involves the following steps: (i) mass transfer from the feed bulk to the feed membrane interface; (ii) partition of penetrants between feed and membrane; (iii) selective transport (diffusion) through the membrane; and (iv) desorption into the vapor phase on the permeate side. The partitioning and desorption steps are normally neglected and the remaining steps are considered as the main contribution to the overall resistance to mass transfer. The mass transfer resistance in membranes depends on its properties and also on the chemical and physical properties of the feed components. In the past, the membrane materials and dimensions were not optimized and, therefore, high membrane mass transfer resistance was commonly observed. However, new developments in membrane materials and ultra-thin

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 133

composite membranes have led to much smaller membrane mass transfer resistances, which are now mainly due to the diffusive transport at the boundary layer, like noticed in the removal of volatile organic compounds from wastewaters (Wijmans et al., 1996) and in the dehydration of cyclohexane (Ortiz et al., 2006). Diffusive transport depends on the hydrodynamic conditions, physical properties of the fluid and of the solute, and the system geometry. Several correlations based on Sherwood number for determination of the mass transfer coefficient for transport in the boundary layer, kbl , have been proposed over the years on different membrane modules configurations (Bandini et al., 1997; Crowder and Cussler, 1998; Dotremont et al., 1994; Gekas and Hallstrom, 1987; Lipnizki and Field, 2001; Lipski and Côté, 1990; Oliveira et al., 2001; Urtiaga et al., 1999).

Pervaporation is often considered as an isothermal process, but this process always induces a temperature drop in the feed, since it involves vaporization. The heat transfer in pervaporation involves: (i) heat transfer from the feed bulk to the feed membrane interface; (ii) heat transfer through the membrane; and (iii) consumption of heat (vaporization followed by expansion) at the permeate side of the membrane. In systems operating under laminar conditions, where the membrane is thin and/or has high permeability, the temperature drop in the feed membrane interface is considerable (Favre, 2003) and, therefore, the boundary layer heat transfer resistance should be taken into account. The heat transfer coefficient in the boundary layer is often determined by Nusselt correlations (Karlsson and Trägardh, 1996).

The ethyl lactate synthesis on pervaporation and vapour-permeation membrane reactors was already studied by some authors. The two configurations adopted were: (i) batch reactor, where the esterification reaction takes place, followed by a membrane for water removal, and reflux of the retentate to the reactor (Benedict et al., 2003; Benedict et al., 2006; Rathin and Shih-Perng, 1998; Wasewar et al., 2009) and (ii) membrane inside a batch reactor (Jafar et al., 2002; Tanaka et al., 2002). In fact none of these studies considers a continuous integrated membrane reactor for ethyl lactate production. Regarding the type of membranes tested, polymeric membranes were used for pervaporation (Benedict et al., 2003; Benedict et al., 2006; Rathin and Shih-Perng, 1998) and zeolites for vapour-permeation (Jafar et al., 2002; Tanaka et al., 2002). In all of these studies, the pervaporation process is superficially addressed; important membrane parameters, like selectivity and species permeances, at different temperatures and feed concentrations were not determined. The most detailed pervaporation study for the ethyl lactate system was performed by Delgado and co-authors for binary and quaternary mixtures using a commercial polymeric membrane Pervap 2201 (from

134 CHAPTER 6. Pervaporation Membrane Reactor

Sulzer Chemtech) (2008; 2009, respectively). These studies, focused just on the separation, still did not consider concentration and temperature polarization, and were not extended to the reactive system.

Although most of studies involves the use of polymeric membranes that provide good selectivity and flux, these membranes do not normally support reaction conditions (in terms of concentrations, temperature, and pH, among others) and, therefore, are not appropriated for applications under those conditions. Inorganic membranes, mainly those of silica and zeolites, present better stability under acidic and high temperature conditions, and are the best alternative to perform dehydration of a reaction medium. Since the commercial microporous silica membrane (from Pervatech) revealed better selectivity and water flux than the commercial membrane Pervap SMS (from Sulzer Chemtech) in the dehydration of aqueous mixtures containing acetone and isopropanol (Casado et al., 2005), Pervatech membrane will be considered in this work. Moreover, for the dehydration of ethanol, the Pervatech membrane proved to have high flux and selectivity (Sommer and Melin, 2005).

The synthesis of ethyl lactate in a continuous integrated pervaporation membrane reactor, using the commercial tubular membrane with higher flux and high selectivity (Pervatech), is here, for the first time, implemented and evaluated. Furthermore, in order to better characterize the pervaporation process and aiming to describe the PVMR unit, the effect of feed pressure, flowrate, temperature and composition on the pervaporation performance is studied by batch experiments (BP), testing the different binary mixtures involved in the synthesis of ethyl lactate (ethanol/water, ethyl lactate/water and lactic acid/water). Additionally, a new mathematical model accounting for the mass transport phenomena under non-isothermal conditions is developed and applied to better understand and describe the ethyl lactate production by means of the PVMR.

6.2 Experimental Section

6.2.1 Materials

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 135

adsorbent. Commercial hydrophilic membrane supplied by Pervatech BV (The Netherlands) was used. It is resistant for any solvent at any concentration, but sensitive to acidic and alkaline media (preferred pH range from 2 up to 9) and so, exposure to inorganic acids or caustic should be avoided. This membrane has a modified silica (methyl silica) selective layer coated onto gamma alumina. The separation layer is applied inside of an asymmetric ceramic tube that has an outer diameter of 10 mm, an inner diameter of 7 mm and a length of 50 cm. It has an effective membrane area per tube of about 110 cm2. Two equal membrane modules were used to perform the experiments.

6.2.2 Pervaporation Membrane Reactor Unit

The experiments were carried out in a pervaporation membrane reactor unit, which can operate in batch (pervaporation studies) or in continuous mode (reaction pervaporation experiments). This unit is equipped with temperature (TI) (type K thermocouple, accuracy of about +/- 2.2ºC) and pressure sensors (PI) in order to monitor and register these two parameters. The absolute pressure is measured through 2 analogue dials (accuracy of about +/- 0.5 bar), filled with glycerine (Nuova Firma), while the pressure in the permeate side is measured by means of 1 digital dial ceraphant-T PTC31 (Endress+Hausser) (accuracy of about +/- 1 mbar). A schematic representation of the pervaporation membrane reactor unit is shown in Figure 6.1. The temperature was controlled by a thermostated bath (Lauda, Germany) with ethylene glycol/water solution that flows through the jackets of feed vessels 1 and 2; pressure was set at 2 bars by applying an overpressure of helium to the system in order to prevent vaporization of feed mixture over the whole temperature range.

Pervaporation experiments

The feed was charged into the feed vessel 1 (1 L capacity) and heated to the desired temperature. In order to heat the whole system (tubing and membranes) to that same temperature, the feed mixture is after re-circulated over the membrane modules using a positive displacement diaphragm pump (Hydra Cell G-03, Wanner International), in the absence of vacuum on the permeate side. When the steady state is reached, the pervaporation experiment starts by applying vacuum to the permeate side by means of a vacuum pump (Boc Edwards, U.K.). During the whole run, all vapour permeated was condensed on two parallel glass cold traps filled with liquid nitrogen. Finally, the collected permeate was defrosted, weighted and analyzed. The duration of the experiment is conditioned by the trade-off

136 CHAPTER 6. Pervaporation Membrane Reactor

between ensuring the nearly constant feed composition and enough amount of permeate. To verify the assumption of constant feed composition samples were collected before and after each experiment.

Figure 6.1 Set-up of the pervaporation membrane reactor unit.

6.3 Pervaporation Studies

The experimental results of the pervaporation studies for the different binary mixtures (ethanol/water, ethyl lactate/water and lactic acid/water) involved in the esterification reaction between lactic acid and ethanol are presented in this section. The effect of the absolute feed pressure, feed flowrate, operation temperature and water feed molar fraction onto the membrane performance is evaluated. The membrane permeabilities were determined in absence of mass transfer limitations in the boundary layer like shown by the preliminary studies performed.

6.3.1 Pervaporation Transport

The pervaporation performance of the membrane was evaluated in terms of pervaporation flux and separation factor (membrane selectivity). The process separation factor (α ) is defined as:

y x α = ij (6.1) xijy

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 137

where xi is the liquid mole fraction of component i on the feed side and yi is the mole fraction of component i on the permeate side.

The partial flux of one component through the membrane is given by:

Jipermitot= wJ, (6.2)

2 where Jtot it the total permeation flux expressed in kg/(m .s) and wperm, i is the mass fraction of component i on the permeate.

The solution-diffusion model (Wijmans and Baker, 1995) was successfully applied in the description of solvent dehydration using microporous silica membranes (Ma et al., 2009; Sommer and Melin, 2005). This model provides the following transport equation for the permeation molar flux of a component through the membrane:

0 JxpyPimembiiiiiperm=−Q(, γ ) (6.3)

2 where Qmemb, i is the permeance of component i through the membrane (mol/(m .s.Pa)), which is equal to Pi / A (Pi is the permeability coefficient of component i (mol/(m.s.Pa)), and A is the thickness of the selective layer of the membrane), γ i is the activity coefficient (calculated

0 by the UNIQUAC model using the parameters determined in chapter 3), pi is the saturation pressure of component i, Pperm is the total pressure on the permeate side and yi is the molar fraction of component i in the vapor phase. The saturation pressure of pure components was estimated by the Antoine equation and it is presented in Appendix B.

Some works consider the temperature influence on the total flux to measure an apparent activation energy (Delgado et al., 2009; Khayet et al., 2008; Slater et al., 2006). However, in this work, the activation energy of permeation is determined by an Arrhenius-type equation for the permeance temperature dependence (Feng and Huang, 1997):

⎛⎞−Eperm, i QQmemb,,0 i= memb exp⎜⎟ (6.4) ⎝⎠RT

138 CHAPTER 6. Pervaporation Membrane Reactor

where Q is the pre-exponential factor, E is the activation energy of permeation, memb,0 perm, i which is a combination of activation energy of diffusion and the heat of adsorption on the membrane ( EEHperm,, i=+Δ D i s ), T is the absolute temperature and R is the ideal gas constant.

6.3.2 Preliminary Studies

6.3.2.1 Evaluation of the membrane quality

The driving force for the transport through the membrane is based on the species chemical potential difference over the membrane and, theoretically, the absolute feed pressure affects the chemical potentials via the Poynting factor, which might affect either the flux or the selectivity. However, this influence is often negligible since the Poynting factor is one at low pressures and, therefore, a way to detect membrane imperfections is by performing pervaporation experiments at different feed pressures. If the permeate composition and the total flux remains constant over the studied pressure range the membrane quality can be, in principle, guaranteed (Pera-Titus et al., 2008).

The influence of the feed pressure onto the pervaporation membrane performance, is shown in Figure 6.2, where the total permeation flux and permeate composition are represented as a function of feed pressure, keeping constant the remaining conditions (temperature, flowrate, water feed concentration, permeate pressure). The variations observed either on the permeate compositions or on the total flux are within the experimental error, and, therefore, it is possible to conclude that pervaporation is not affected by the feed pressure. This is corroborated by other work (De Bruijn et al., 2007) where the effects of feed pressure on a silica membrane performance were evaluated. Within the pressure range studied, the Poynting factor is almost one for all species, and therefore these results indicate that the silica membranes used are free of significant defects.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 139

0.90 1.00

0.88 )) 2 0.98 0.86 (kg/(h.m 0.84 tot

J 0.96

0.82 Total flux Permeate composition 0.80 0.94 Permeate Water molefraction 0.00.51.01.52.02.5 Absolute feed pressure (bar)

Figure 6.2 Pressure dependence of total permeation flux and of permeates composition for the system ethanol/water.

(TK= 321.15 , xWF, =±0.183 0.001, Pperm = 20 mbar ).

6.3.2.2 Evaluation of mass transfer limitations in the boundary layer

The effect of concentration polarization on the pervaporation process was studied for water/ethanol mixtures, at constant feed water concentration and operating temperature, varying the feed flowrate. Analyzing Figure 6.3, it is possible to conclude that for feed flowrates higher than 19 L/h, the total flux remains constant, indicating absence of mass transfer resistance from the bulk liquid phase to the feed-membrane interface. Therefore, the remaining pervaporation experiments were performed at a feed flowrate of 20 L/h.

1.00

0.95

)) 0.90 2

0.85

(kg/(h.m 0.80 tot J 0.75

0.70 10 13 16 19 22 25 Feed flowrate (L/h)

Figure 6.3 Total permeation flux as a function of feed flowrate

(TK= 321.15 , xWF, = 0.182± 0.006 , Pmbarperm = 25 ).

140 CHAPTER 6. Pervaporation Membrane Reactor

6.3.3 Detailed Studies

6.3.3.1 Water/Ethanol System

The influence of feed composition at different operating temperatures on the total permeation flux and on the permeate composition is shown in Figure 6.4 for the water/ethanol pair. (a) (b)

6.45 1.00 T=321.15 K 5.45 T=336.15 K 0.99 )) 2 4.45 T=344.15 K 0.98 3.45 (kg/(h.m T=344.15 K

tot 0.97 J 2.45 T=321.15 K 0.96 T=336.15 K 1.45 Permeate water molefacrtion 0.45 0.95 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.00.10.20.30.40.50.6 Feed water mole fraction Feed water mole fraction

Figure 6.4 Pervaporation performance for water/ethanol mixtures. (a) Influence of feed water mole fraction on total permeation flux at different operating temperatures; (b) Influence of feed water mole fraction on permeate composition at different operating temperatures.

As can be seen, the total permeation flux increases with water concentration (linearly) and with the temperature in the feed (Figure 6.4a). From Figure 6.4b, it is also observed an increase in water permeate mole fraction with the feed water concentration, but little influence is noticed with the temperature.

6.3.3.2 Water/Ethyl lactate System

Like before, in the water/ethyl lactate system, it is observed an increase on the total flux with the increase in the feed water mole fraction, except when the system was kept at 321.15 K, where this effect of feed water composition on the total flux is not so significant (see Figure 6.5a). The temperature effect on the total flux is once again relevant, increasing with the operation temperature, for the same feed composition, it is observed an increase around 167 % in the total flux. This is justified by the fact that the permeation driving force increases with temperature. Figure 6.5b, shows that the water mole fraction in the permeate is almost independent (varies from 0.996 to 0.999) from the feed water concentration and temperature, considering the studied range.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 141

(a) (b) 1.000 8.15 T=321.15 K 7.15 T=334.15 K 0.998 T=343.15 K )) 2 6.15 0.996 5.15

(kg/(h.m 0.994 T=321.15 K tot

J 4.15 T=334.15 K 0.992 3.15 T=343.15 K Permeate water molefraction 2.15 0.990 0.35 0.45 0.55 0.65 0.75 0.35 0.45 0.55 0.65 0.75 Feed water mole fraction Feed water mole fraction

Figure 6.5 Pervaporation performance for water/ethyl lactate mixtures. (a) Influence of feed water mole fraction on total permeation flux at different operating temperatures; (b) Influence of feed water mole fraction on permeate composition at different operating temperatures.

6.3.3.3 Water/Lactic acid System

Two pervaporation experiments were performed for the water/lactic acid mixtures with a feed water molar fraction of 0.77, at two different temperatures (321.15 K and 343.15 K). Experimentally, the permeate stream was only composed by water and, therefore, no more experiments were performed for this mixture.

6.3.3.4 Membrane performance evaluation

For both ethanol/water and ethyl lactate/water systems, the flux increases with feed water composition and temperature. It was shown that the flux depends strongly on the temperature like already mentioned for dehydration in similar membranes (Casado et al., 2005; ten Elshof et al., 2003). In the case of water/lactic acid mixtures, no lactic acid was detected in the permeate side and no more measurements were made regarding this separation.

In the Figure 6.6 the process separation factor for the pairs water/ethanol and water/ethyl lactate is plotted as a function of the temperature. It can be seen that the membrane presents good selectivity towards water in both pairs, but the selectivity to water is higher in mixtures with ethyl lactate most probably due to the higher molecular size of ethyl lactate compared with that of ethanol. One indicative measure of the molecular size is given by the gyration radius and their values presented in Table 6.1 support that hypothesis. However, according to that, it was expected the permeation of lactic acid through the membrane. Naturally, this is

142 CHAPTER 6. Pervaporation Membrane Reactor

not the only factor affecting selectivity; the lower lactic acid vapor pressure compared to all other species (Table 6.1) significantly reduces the permeation driving force for this component. Moreover, the affinity between species and membrane should be considered as a secondary factor (the molecular diameter is the most important). This affinity is directly related with the dipole moment of each component (Table 6.1): small dipole moment might complicate the penetration of molecules into the hydrophilic silica layer. Comparing lactic acid with ethyl lactate, the molecular sizes are similar, but the first has the lowest dipole moment, which in addition to the lowest vapor pressure, leads to infinite membrane selectivity towards water in lactic acid mixtures.

Table 6.1 Physical properties of the different species (water, ethanol, ethyl lactate and lactic acid) (Delgado et al., 2008).

D Molecule Radius of gyration ( A ) Dipole moment (Debyes) Water 0.615 1.8497 Ethanol 2.259 1.6908 Ethyl lactate 3.622 2.4000 Lactic acid 3.298 1.1392

1200 water/ethanol xW,F = 0.435 ± 0.006

1000 water/ethyl lactate xW,F = 0.494 ± 0.008

800

600

Separation factor Separation 400

200

0 320 325 330 335 340 345 350 Temperature (K)

Figure 6.6 Influence of the temperature on the separation factor for the binary mixtures water/ethanol and water/ethyl lactate.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 143

6.3.4 Parameter Estimation

The pervaporation process design requires the knowledge of the permeance of each component as function of temperature (Equation 6.4). However, there is evidence that the feed water content affects permeance (Delgado et al., 2009) and therefore, both factors are needed to be accounted for.

6.3.4.1 Permeance temperature dependence

Experimental permeance for each species can be calculated using the experimental information found so far in Equation 6.3, simultaneously with the activity coefficients calculated by the UNIQUAC model. Those values are presented in Figure 6.7.

(a) (b)

0.0 -6.0

-0.5 -6.5

-1.0 -7.0

-7.5 memb, Eth

memb, w -1.5 ln Q ln ln Q ln -2.0 -8.0

-2.5 -8.5 -9.0 -3.0 2.85 2.90 2.95 3.00 3.05 3.10 3.15 2.85 2.90 2.95 3.00 3.05 3.10 3.15 -1 1000/T (K-1) 1000/T (K )

-3.0

-4.0

-5.0

memb, EL -6.0 ln Q ln -7.0

-8.0

-9.0 2.90 2.95 3.00 3.05 3.10 3.15 1000/T (K-1) Figure 6.7 Temperature dependence of species permeances (mol/(min.dm2.bar)) and linear fittings: (a) water; (b) ethanol; (c) ethyl lactate.

From the slope and intercept of the linear regression, the activation energy of permeation and the pre-exponential factor are determined for each species, like presented in Table 6.2.

144 CHAPTER 6. Pervaporation Membrane Reactor

Although, the permeation flux increases with temperature due to the increasing driving forces, the membrane permeability decreases with the temperature like the negative values of the activation energies of permeation indicates (Table 6.2). As mentioned previously, the activation energy is a sum of the diffusion activation energy and the heat of adsorption on the membrane, indicating that the permeation of water, ethanol and ethyl lactate are governed by the adsorption. Similar results were found for water/ethanol permeation through silica microporous membranes (Ma et al., 2009).

Table 6.2 Pervaporation parameters of Equation 6.4 and the mean relative deviation (MRD).

2 Component Qmemb,0 (mol/(min.dm .bar)) Eperm (kJ/mol) MRD (%) Ethanol 1.41×10-7 -22.60 14.51 Ethyl lactate 1.12×10-4 -10.42 24.13 Water 1.78×10-3 -13.93 13.94

The error introduced considering composition invariant permeances can be calculated from the mean relative deviation (MRD), defined by Equation 6.5.

1 ⎛⎞JJ− MRD =×⎜⎟icalc,,exp i 100% (6.5) nJ⎜⎟∑ exp⎝⎠nexp i ,exp

In order to decrease the deviations between theoretical and experimental fluxes, parameter estimation considering also composition dependence will be addressed in following sub- section.

6.3.4.2 Permeance temperature and water content dependence

The strategy developed is based on the linearization of equation 6.4 at each composition.

Considering water/ethanol and water/ethyl lactate mixtures, the permeances of water as a function of temperature, at a constant feed water mole fraction, are shown in Figure 6.8 and Figure 6.9, respectively. Except for an ethyl lactate aqueous solution with 68 % of water content, where the activation energy is positive, in all other situations the activation energy of permeation is negative has previously determined. This might indicate that, for water/ethyl

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 145

lactate mixtures, the water permeance at high feed water concentrations is controlled by diffusion instead of adsorption.

(a) (b)

-0.70 -1.10

-0.90 -1.20

-1.10 -1.30

-1.30 -1.40 memb, W memb, W ln Q ln -1.50 Q ln -1.50

-1.70 -1.60

-1.90 -1.70 2.85 2.90 2.95 3.00 3.05 3.10 3.15 2.85 2.90 2.95 3.00 3.05 3.10 3.15 1000/T (K-1) 1000/T (K-1) (c) (d)

-1.10 -1.20

-1.20 -1.30

memb, W -1.30 memb, W -1.40 ln Q ln Q ln

-1.40 -1.50

-1.50 -1.60 2.90 2.95 3.00 3.05 3.10 3.15 2.85 2.90 2.95 3.00 3.05 3.10 3.15 1000/T (K-1) 1000/T (K-1) Figure 6.8 Temperature dependence of water permeances (mol/(min.dm2.bar)) and

linear fittings (water/ethanol mixtures): (a) xWF, =±0.105 0.001;

(b) xWF, =±0.183 0.003; (c) xWF, =±0.298 0.006 (d) xWF, =±0.492 0.005 .

146 CHAPTER 6. Pervaporation Membrane Reactor

(a) (b)

-1.00 -1.00

-1.02 -1.10 -1.04

-1.20 -1.06 memb, W memb, W -1.30 -1.08 ln Q ln Q ln -1.10 -1.40 -1.12

-1.50 -1.14 2.90 2.95 3.00 3.05 3.10 3.15 2.90 2.95 3.00 3.05 3.10 3.15 -1 -1 1000/T (K ) 1000/T (K )

-1.12

-1.14 memb, W ln Q ln -1.16

-1.18 2.95 3.00 3.05 3.10 3.15 -1 1000/T (K ) Figure 6.9 Temperature dependence of water permeances (mol/(min.dm2.bar)) and linear fittings (water/ethyl lactate mixtures): (a) x =±0.435 0.006 ; (b) WF,

xWF, =±0.564 0.015 ; (c) xWF, = 0.680± 0.003 .

The pre-exponential factors as well as the activation energies of permeation calculated from Figure 6.8 and Figure 6.9 are presented in Table 6.3. Independently of binary mixture considered, the representation of those parameters as function of the water feed mole fraction is shown in Figure 6.10.

Table 6.3 Pre-exponential factor and activation energy for different feed water contents.

Parameters Water/ethanol mixtures Water/ethyl lactate mixtures

Water molar (%) 10.5 18.3 29.8 49.2 43.2 56.4 68.0

Eperm,W (kJ/mol) -30.24 -17.56 -17.52 -7.84 -13.56 -4.31 3.46

Q memb0,W 4.62×10-6 4.55×10-4 4.47×10-4 1.47×10-2 2.00×10-3 7.12×10-2 1.13 (mol/min.dm2.bar)

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 147

(a) (b)

1.2 10 5 1.0 .bar))

2 0 0.8 -5 -10 0.6 (kJ/mol) -15 perm E (mol/(min.dm 0.4 -20 -25 0.2 memb,0 -30 Q 0.0 -35 0.0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 Feed water mole fraction Feed water mole fraction Figure 6.10 Influence of feed water mole fraction on: (a) pre-exponential factor; (b) activation energy of permeation.

Analysing Figure 6.10, it is possible to conclude that the pre-exponential factor and water permeance dependencies on feed water mole fraction are well described by exponential and linear functions, respectively, leading to the following expressions:

−62 Qmemb,0 =×1.967 10 exp(18.64 x W ) (mol/(min.dm . bar )) (6.6)

Experm, w=−50.377 W 32.326 (KJ/mol) (6.7)

These expressions were combined in a final empirical correlation that describes the permeance of water as a function of the temperature and water feed content:

−6 ⎛⎞50377xW − 32326 2 Qperm, W=×1.967 10 exp(18.64 x W )exp⎜⎟ − (mol/(min.dm . bar )) (6.8) ⎝⎠RT The error between experimental and theoretical permeances calculated by Equation 6.8 is of 10.35 % (MRD), which is about 26 % smaller than the one obtained when just the temperature influence was considered. Although derived from binary mixtures, expression 6.8 can be applied to estimate the water permeance in quaternary mixtures, since it depends only on water content no matters the remaining constituents. This is in agreement with experimental pervaporation results using binary and quaternary mixtures involved in the ethyl lactate synthesis, since for the same temperature and water feed content, the value of water permeance is almost the same for binary or quaternary mixtures (Delgado et al., 2009). Therefore, it can be stated that the water transport is barely affected by the presence of the other components supporting the validity of Equation 1.8. For ethanol and ethyl lactate, no

148 CHAPTER 6. Pervaporation Membrane Reactor

relation was found between their permeances in the membrane and the feed contents. Thus, the parameters of Table 6.2 were used to determine the permeances of these species at different operating temperatures and water feed contents.

6.3.4.3 Estimation of the boundary layer mass transfer coefficient (kbl)

In pervaporation processes, in addition to the permeation resistance in the membrane there is also the resistance in the boundary layer (concentration polarization). This can be conveniently represented by the resistance in series model, where the overall resistance to transport is the sum of boundary layer and membrane resistances. The resistance in series equation for pervaporation was derived by Wijmans and collaborators (1996):

0 11γ iimolipV , =+b (6.9) kQov,, i memb i av F

2 in which kov, i is a global membrane mass transfer coefficient (mol/(s.m .bar )) , that combines the resistance due to the diffusive transport in the boundary layer with the membrane

3 b resistance, Vmol,i is the molar volume of component i (/mmol ), Kavbl= F where vF is the feed liquid velocity (/)ms. Equation 6.9 will be used to determine the mass transfer resistance in the boundary layer.

The pervaporation data experimentally obtained for water/ethanol mixtures using different feed flowrates was used to plot the inverse of the global membrane mass transfer coefficient

b (1/ kov ) of water and ethanol as a function of 1/vF , as shown in Figures 6.11 and 6.12, respectively. The value of parameter b was chosen in order to keep the data points on a straight line and also minimizing the difference between the inverse of the intercept of the line and the experimental permeance measured in absence of mass transfer limitations on the boundary layer (being the best fit for b=0.97), while the parameter a was obtained from the slope of this line. By this analysis it was possible to obtain the following expressions to determine the boundary layer mass transfer coefficients for both species (ethanol and water):

0.97 −5 ⎡ ⎤ Kmsbl, W(/)7.2710=×⎣ vms F (/)⎦ (6.10)

0.97 −7 ⎡ ⎤ Kmsbl, Eth(/)3.5610=×⎣ vms F (/)⎦ (6.11)

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 149

2.0 1000

1.9 900 1.8 .bar.s)/mol) .bar.s)/mol) 2 1.7 2 800 ((m 1.6 ((m

ov,W 700 1.5 ov,Eth 1/k 1/k 1.4 600 5678910 5678910 b -b 1/vF (m/s) b -b 1/vF (m/s) b b Figure 6.11 1/kov, W versus 1/vF , b=0.97. Figure 6.12 1/kov, Eth versus 1/vF , b=0.97.

6.4 Modelling

6.4.1 Batch Pervaporation Model

The mathematical model developed to describe the behaviour of the batch pervaporation membrane (BPM) considers:

- Plug flow for the bulk fluid phase;

- Total feed volume inside the tank and the retentate velocity variations (inside membrane) due to permeation of components;

- Concentration polarization, where the resistance due to the diffusive transport in the boundary layer is combined with the membrane resistance in a global membrane resistance;

- Non-isothermal operation due to heat consumption for species vaporization;

- Temperature polarization.

Following these assumptions the BPM model equations are:

Feed tank mass balance to component i

dVC( fi, ) =−QC QC (6.12) dt ret ret,, i f f i

150 CHAPTER 6. Pervaporation Membrane Reactor

where t is the time variable, V is the volume of the feed thank, Q f is the flowrate fed to the membrane modules and Qret is the flowrate at the end of the membrane modules, C f and Cret are the liquid phase concentration fed to the membrane and at the end of the membrane modules, respectively.

Feed volume variation

dV =−QQ (6.13) dt ret f

Retentate mass balance to component i

∂C ∂()vCret, i ret, i ++=AJ 0 (6.14) ∂∂tzmi

where z is the axial coordinate at the membrane modules, v is the superficial velocity, Am is the membrane area per unit membrane modules volume and Ji is the permeate molar flux of species i, through the membrane, defined as:

0 JxpyPioviiiiiperm=−k(, γ ) (6.15)

where kov, i is the global membrane mass transfer coefficient, that combines the resistance due to the diffusive transport in the boundary layer with the membrane resistance (Wijmans et al., 1996):

11γ pV0 =+iimoli, (6.16) kQov,, i memb i K bl

in which Vmol,i is the molar volume of component i and Kbl is the boundary layer mass transfer

2 coefficient. For laminar flow and Graetz number ( dvDLint/( m ) ) much greater than one, the mass transfer coefficient for transport in the boundary layer, kbl , is determined by the Lévêque correlation (Lévêque, 1928):

0.33 0.33 0.33 ⎛⎞dint Sh= 1.62Re Sc ⎜⎟ (Re< 2300) (6.17) ⎝⎠L

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 151

where Sh= kbl dint / D m , Redv= ρ int /η and Sc= ηρ/( Dm ) are the Sherwood, Reynolds and

Schmidt numbers, respectively, Dm is the solute diffusivity in the boundary layer, dint is the inside diameter of the membrane, L is the membrane length, ρ is the density and η is the viscosity. The accuracy of the Lévêque correlation to estimate the mass transfer coefficient in the laminar regime has been validated experimentally by several works (Crowder and Cussler, 1998; Wickramasinghe et al., 1992). The prediction of the solute diffusivity was made using the Perkins and Geankoplis method (Perkins and Geankoplis, 1969). Further details concerning its calculation can be found in Chapter 4.

The molar fraction of component i on the vapor phase (permeate side), yi , is defined as:

Ji yi = n (6.18) ∑ Ji i=1

Fluid velocity variation in the membrane feed side calculated from the total mass balance dv n =−AJVmimoli∑ , (6.19) dz i=1 where n is the total number of components.

Retentate heat balance

nn l ∂∂TTl (6.20) ∑∑Cpi,, Cret,, i+ vC pi C ret i+−= A m h F() T T m 0 ii==11∂∂tz

l where C p,i is the liquid heat capacity of component i, T is the absolute temperature in the feed side of the membrane, Tm is the membrane temperature and hF is the heat transfer coefficient in the liquid boundary layer.

Membrane heat balance

n ddint int +δ V 22hTF ()−= Tmii 22∑ Δ HJ (6.21) (rrint+−δδ / 2) int ( rr int +− / 2) int i=1

152 CHAPTER 6. Pervaporation Membrane Reactor

V where rint is the internal radius of the membrane, δ is the membrane thickness and ΔHi is the heat of vaporization of species i. The heat transport coefficient was estimated by the Sieder- Tate correlation, valid for laminar pipe flow (Welty et al., 2008):

0.33 0.14 ⎛⎞dint ⎛⎞ηb Nu =1.86⎜⎟ Re Pr ⎜⎟ (6.22) ⎝⎠L ⎝⎠ηw

l where Nu= h d / λ , Pr=ηC p / λ are the Nusselt and Prandtl numbers, respectively, η and F int b

ηw are the viscosity of the liquid in the feed and in the membrane wall, respectively and λ is the thermal conductivity.

Initial boundary conditions:

tCCC===0: f ,,,0iretii (6.23a)

VV= 0 (6.23b)

TT= F (6.23c)

z ==0: TTF (6.24a)

vv= F (6.24b)

CCret,, i= f i (6.24c) where subscripts 0 and F refer to initial state and membrane feed conditions, respectively.

6.4.2 Pervaporation Membrane Reactor Model

A mathematical model was also developed to describe the behaviour of the tubular pervaporation membrane reactor that similarly to the BPM model considers:

- Concentration polarization, where the resistance due to the diffusive transport in the boundary layer is combined with the membrane resistance in a global membrane resistance;

- Temperature polarization;

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 153

Additionally, it also takes in account:

- Axial dispersion flow for the bulk fluid phase;

- External and internal mass transfer for adsorbable species combined in a global particle resistance;

- Non-isothermal operation due to heat effects during species vaporization and reaction (slightly endothermic);

- Velocity variations due to permeation of components and adsorption/desorption rates;

- Constant column length and packing porosity.

Following this assumption the PVMR model equations are:

Bulk fluid mass balance to component i:

∂∂CxA∂ ()uC (1−∂ε ) 3 ⎛⎞ i i im (6.25) ++KL,,ii()CC −= piaxT D⎜⎟ C − J i ∂∂tzε rp ∂∂ zz⎝⎠ε

where K L,i is the global mass transfer coefficient of the component i, ε is the bed porosity,

Dax , and u are the axial dispersion coefficient and the interstitial velocity, respectively, rp is the particle radius, C p is the average particle concentration and all the other variables were already defined. The determination of the global mass transfer coefficient ( K L ) is presented in detail in Chapter 4 and the axial dispersion coefficient ( Dax ) is estimated from the empirical correlation (Butt, 1980) valid for liquids in packed beds:

ε Pe =+0.2 0.011Re0.48 (6.26) in which PeduD= / and Redu= ρ /η are the Peclet and Reynolds numbers, int ax int respectively.

The permeate flux of membrane is defined by Equation 6.15 (BPM model).

Interstitial fluid velocity variation calculated from the total mass balance du (1− ε ) 3 nnA m =−∑ KL,,imoliVCC() i −pi, − ∑ J i (6.27) dzεε rp ii==11

154 CHAPTER 6. Pervaporation Membrane Reactor

where Vmol,i is the molar volume of component i and n is the total number of components.

Pellet mass balance to component i

3 ∂C p,i ∂ qi ν i ρb K L,i ()Ci −C p,i = ε p + (1− ε p ) − r(C p,i ) (6.28) rp ∂t ∂t 1− ε where ν is the stoichiometric coefficient of component i , ρ is the bulk density, ε the i b p particle porosity, q i is the average adsorbed phase concentration of species i in equilibrium with Cp,i , and r is the kinetic rate of the chemical reaction relative to the average particle concentrations in the fluid phase. The reaction rate and adsorption isotherms are those determined in Chapters 3 and 4, respectively.

Retentate heat balance

nn∂∂TTA ρ l l mb (6.29) ∑∑CCpi,,iiFmr++−+Δ= uCC pi hTT() H r 0 ii==11∂∂tzεε

where ΔHr is the reaction enthalpy, determined in Chapter 3. It must be considered that the heat of adsorption was neglected since while some species are being adsorbed, others are being desorbed and there is compensation on the heat released or absorbed, respectively.

The equation of the membrane heat balance is the one represented in the BPM model (Equation 6.21).

Initial and Danckwerts boundary conditions

t = 0 : CCii==pi, C,0 , qqii= ,0 and TT= F (6.30)

∂C i z = 0 : uCiax−= D uC iF, (6.31a) ∂z z=0

uu= 0 (6.31b)

TT= F (6.31c)

∂Ci z = Lc : = 0 (6.31d) ∂z zLc= where F and 0 refer to the feed and initial states, respectively.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 155

The viscosities, heat capacities, thermal conductivities and heats of vaporization of pure components were calculated for each temperature by the expressions presented in Appendix B. The mixture heat capacity was estimated assuming linear mole fraction averages, while the mixture density and thermal conductivity were estimated assuming linear mass fraction averages. The methods used to predict binary and multi-component mixture viscosities were presented in detail in Chapter 4. Details on numerical solution are the same presented in Chapter 5.

6.5 Results and Discussion

6.5.1 Batch Pervaporation

Within the mathematical model proposed for the batch pervaporation, the mass transfer coefficient in the boundary layer is, for all the species, estimated through the Lévêque correlation. However, its accuracy should be validated experimentally. In Section 6.3.4.3, the boundary layer mass transfer coefficients, for water and ethanol, were determined from the experimental data (Equations 6.10 and 6.11) and will be applied in the BPM model. The water permeation fluxes were calculated either by using the experimental mass transfer coefficient in the boundary layer (Equations 6.10 and 6.11) or using the Lévêque correlation (Equation 6.17). As can be observed in Figure 6.13, there is a good agreement between the water permeation flux estimated using the experimental mass transfer coefficients with the one using the Lévêque correlation, proving its accuracy in the prediction of the boundary layer mass transfer coefficients. Besides, small deviations are observed between the experimental fluxes and the ones determined by the BPM model. In order to validate the BPM non-isothermal model, the dehydration of ethanol and ethyl lactate aqueous solutions was studied. The evolution of water mole fraction in the retentate stream is in agreement with the experimental data: for both ethanol and ethyl lactate cases the retentate composition after 15 and 10 minutes, respectively, is similar to the theoretical values, as shown in Figure 6.14. In terms of temperature drop, experimentally it was not observed any temperature variation for ethanol dehydration, while it has dropped 2ºC for the ethyl lactate experiment. According to the simulation it would be expected a decrease of 2 and 4ºC for ethanol and ethyl lactate dehydrations, respectively. This deviation might be due to experimental error of the thermocouple used (accuracy of 2.2ºC).

156 CHAPTER 6. Pervaporation Membrane Reactor

2.5 Estimated mass transfer coefficient (Lévêque correlation)

Experimental mass transfer coefficient 2.0 ) 2 1.5 (mol/(h.dm

W 1.0 J

0.5

0.0 0.0 0.5 1.0 1.5 2.0 2.5 J (mol/(h.dm2) W,exp Figure 6.13 Experimental water permeation flux as a function of water permeation flux determined by the BPM model with the Kbl estimated (Lévêque correlation) and determined from experimental data.

337 345 0.189 Feed (Experimental) 0.559 Feed (Experimental) Retentate (Experimental) Retentate (Experimental) 344 0.187 Retentate (BPM model) 0.555 Retentate (BPM model) 343 0.185 Temperature (BPM model) 336 Temperature (BPM model) 0.551 0.183 342 0.547 0.181 341 335

Temperature (K) 0.543 0.179 340 Temperature (K) Temperature Water molarfraction

0.177 Water molarfraction 0.539 339

0.175 334 0.535 338 02468101214 0246810 Time (min) Time (min) Figure 6.14 Evolution of water composition on retentate and temperature predicted by the non-isothermal BPM model: a) dehydration of 81 % of ethanol in aqueous solution at 336 K; b) dehydration of 44 % of ethyl lactate in aqueous solution at 344 K.

6.5.2 Pervaporation Membrane Reactor

The performance of the pervaporation membrane reactor unit, packed with A15-wet resin in the lumen side of the tubular membranes (where the active layer of the silica membranes is placed) was evaluated by simulation. It was considered that the resin was initially saturated with ethanol and then it was fed with a mixture of ethanol and 85 wt. % lactic acid aqueous

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 157

solution. The permeation fluxes were estimated using the expressions previously obtained in section 6.3.3 and a summary of the parameters used in the simulations is presented in Table 6.4.

Table 6.4 Parameters used in the simulations.

Parameters

Feed temperature 323.15 K

Feed composition xxEth,,W,F F==0.3546; LA F 0.3424; x =0.3030 Feed flowrate 1.0 mL/min Permeate pressure 10 mbar Bed porosity 0.424 Bulk density 374.0 g/L Length of the bed 100 cm Internal diameter 0.7 cm

The PVMR unit was evaluated considering isothermal and non-isothermal operation and its performance was compared with the one of a fixed bed reactor (FBR) in the same operational conditions. The concentration histories at the end of the PVMR and FBR, considering isothermal operation, are shown in Figure 6.15. It can be observed that the enhancement introduced by the water removal through the membranes is significant for the ethyl lactate production, where a 69 % lactic acid conversion was achieved in the PVMR unit, at the steady-state, while in the FBR the conversion obtained was just 29 %.

158 CHAPTER 6. Pervaporation Membrane Reactor

18 18 Ethanol Ethanol 16 16 Lactic acid Lactic acid 14 14 Ethyl lactate Ethyl lactate 12 12 Water Water 10 10 8 8 C (mol/L) 6 C (mol/L) C 6 4 4 2 2 0 0 0 20406080100 0 20406080100 Time (min) Time (min)

Figure 6.15 Concentration histories at the reactor outlet, considering isothermal operation: a) PVMR; b) FBR.

18 18 16 Ethanol 16 Ethanol Lactic acid Lactic acid 14 14 Ethyl lactate Ethyl lactate 12 12 Water Water 10 10 8 8 C (mol/L) C C (mol/L) C 6 6 4 4 2 2 0 0 0 20406080100 0 20406080100 Time (min) Time (min) Figure 6.16 Concentration histories at the reactor outlet, considering non-isothermal operation: a) PVMR; b) FBR.

Considering, alternatively, a non-isothermal operation (see Figure 6.16b), the lactic acid conversion at steady state of the FBR is about 29 %, much better than 11 % found for the PVMR (Figure 6.16a). This low performance of the PVMR is due to the temperature drop noticed in the bulk, around 36ºC, while only 2ºC was noticed for the FBR (Figure 6.17). This slightly decrease in the FBR is due to the heat needed to the reaction only (endothermic reaction), while for the PVMR there is a large heat consumption to vaporize the species. Since the flowrate is very small, the heat capacitance provided by the liquid stream is low and therefore its temperature is drastically decreased, leading to lower reaction kinetic rates, higher mass transfer resistances and lower water permeation fluxes through the membranes.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 159

330

320

310 FBR

T (K) PVMR 300

290

280 0 20 40 60 80 100 120 140 Time (min) Figure 6.17 Temperature histories at the outlet of the FBR and of the PVMR.

Clearly, to operate the PVMR for efficient ethyl lactate production, it is necessary to provide the heat required for vaporization and reaction, operating in isothermal conditions. As stated before, the silica membranes used in this work have the selective layer coated inside the membrane tube. Therefore, the most suitable way to supply heat to the retentate liquid is introducing a heated sweep gas in the permeate side instead of using vacuum. However, when the membrane selective layer is coated on the external side of the membrane tube (shell side), the heat can be supplied through an appropriate heated solution re-circulated through jacketed modules; in this situation, the heat is rapidly transferred to the retentate liquid stream.

Assuming that the PVMR unit could operate isothermally replacing vacuum by a heated sweep gas, it is possible to determine by simulation the number of membranes connected in series needed to maximize the lactic acid conversion of the same feed processed before. For an operating temperature of 50ºC (same conditions from Table 6.4), it is possible to achieve 93 % of lactic acid conversion and 84 % of ethyl lactate purity when using a 250 cm long membrane reactor (5 tubular membranes, effective area of about 550 cm2), as it can be seen in Figure 6.18.

160 CHAPTER 6. Pervaporation Membrane Reactor

18 16 Ethanol Lactic acid 14 Ethyl lactate 12 Water 10 8 C (mol/L) 6 4 2 0 0 40 80 120 160 200 240 Time (min)

Figure 6.18 Concentration histories at the outlet of the PVMR for a bed length of 250 cm at 323.15 K.

Trying to further enhance the separation performance, as well as the reaction rate, the PVMR temperature was increased to 70ºC. In this case, it was obtained a lactic acid conversion of 98 % and an ethyl lactate purity of 96 %, as shown in Figure 6.19. From the analysis of the internal concentration profiles on the retentate side, shown in Figure 6.20, it is concluded that using only two membranes it is possible to get about 90 % of lactic acid conversion, but only 82 % of ethyl lactate purity; being necessary 3 more membranes to maximize both lactic acid conversion and ethyl lactate purity.

18 16 Ethanol Lactic acid 14 Ethyl lactate 12 Water 10 8 C (mol/L) 6 4 2 0 0 40 80 120 160 200 240 Time (min)

Figure 6.19 Concentration histories at the outlet of the PVMR at 343.15 K (L=250 cm).

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 161

10 Ethanol Lactic acid 8 Ethyl lactate Water 6

C (mol/L) C 4

0 0 0.2 0.4 0.6 0.8 1 Z

Figure 6.20 Concentration profiles at steady state at 343.15 K.

These results demonstrate that the PVMR has a great potential for the ethyl lactate production, where high conversions and purities can be obtained, when this reactor is operated under isothermal conditions.

6.6 Conclusions

Commercial hydrophilic silica membrane from Pervatech BV (The Netherlands) was evaluated for the dehydration of ethanol, ethyl lactate and lactic acid, in the temperature range varied from 48ºC to 72ºC. First pervaporation studies indicate that the membranes have no major imperfections since the total flux and water selectivity is barely affect by absolute feed pressure. The influence of hydrodynamic conditions on the membrane polarization was analyzed, and for velocity values higher than 0.14 m/s polarization effects are eliminated. The effect of feed temperature and composition on the pervaporation performance was evaluated by batch experiments. It was concluded that the microporous silica membranes have high flux and high selectivity for water, while ethanol and ethyl lactate permeation is reduced and lactic acid does not permeate at all. In summary, the permeances for all species through the microporous silica membranes, as a function of temperature and feed water content, are described by the following equations:

3 −7 ⎛⎞22.60× 10 2 Qmemb, Eth =×1.41 10 exp⎜⎟mol/( dm bar min) ⎝⎠RT

162 CHAPTER 6. Pervaporation Membrane Reactor

Q = 0 mol/( dm2 bar min) memb, LA

3 −4 ⎛⎞10.42× 10 2 Qmemb, EL =×1.12 10 exp⎜⎟mol/( dm bar min) ⎝⎠RT

−6 ⎛⎞50377xW − 32326 2 Qxmemb, W=×1.967 10 exp(18.64 W )exp⎜⎟ − mol/( dm bar min) ⎝⎠RT

A mathematical model was developed for the batch pervaporation membrane, considering (i) plug flow for the bulk fluid phase; (ii) total feed volume inside the tank and retentate velocity inside the membrane variations due to permeation of components; (iii) concentration polarization, where the resistance due to the diffusive transport in the boundary layer is combined with the membrane resistance in a global membrane resistance; (iv) non-isothermal operation due to heat consumption for species vaporization; and (v) temperature polarization. The batch pervaporation membrane model was validated experimentally and, therefore, it was extended to the integrated pervaporation membrane reactor packed with the catalyst Amberlyst-15wet. This model was used to evaluate the performance of the PVMR in both isothermal and non-isothermal conditions. It was concluded that non-isothermal operation worsens the performance of the PVMR, being even worse than that obtained in the fixed bed reactor. In isothermal conditions, the PVMR is a very attractive solution leading to a lactic acid conversion of 90 % and ethyl lactate purity of 82 %, for the PVMR set-up designed in this work (100 cm of length, 2 membrane modules) operating at 70ºC. For near lactic acid depletion (98 % conversion), it is produced ethyl lactate with 96 % purity if the membrane modules has 250 cm (5 membranes in series), as indicated by the model predictions.

6.7 Notation

2 3 Am membrane area per unit membrane modules volume (m /m ) C liquid phase concentration (mol/m3) 3 C f liquid phase concentration in the thank (mol/m )

3 C p average particle concentration (mol/m particle) l C p liquid heat capacity (J/(mol.K))

3 Cret liquid phase concentration in the retentate (membrane feed side) (mol/m )

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 163

2 Dax axial dispersion coefficient (m /s) dint internal diameter of the membrane (m)

2 Dm solute diffusivity in the boundary layer (m /s)

ED activation energy of diffusion (J/mol)

Eperm activation energy of permeation (J/mol) hF heat transfer coefficient in the liquid boundary layer (W/K)

ΔH s heat of adsorption (J/mol)

ΔH V heat of vaporization (J/mol)

ΔH r enthalpy of reaction (J/mol) J permeate flux (mol/(m2s)) 2 Jtot total permeate flux (mol/(m s))

Kbl boundary layer mass transfer coefficient (m/s)

K L global mass transfer coefficient

2 kov global membrane mass transfer coefficient (mol/(m sPa)) L column length (m) n total number of components Nu Nusselt P permeability coefficient (mol/(m.s.Pa)) p0 saturation pressure (s)

Pperm total pressure on the permeate side (s) Pr Prandtl q solid phase concentration in equilibrium with the fluid concentration inside the particle (mol/m3) 2 Qmemb permeance (mol/(m sPa))

2 Qmemb,0 pre-exponential factor (mol/(m sPa))

3 Q f flowrate fed to the membrane modules (m /s)

3 Qret flowrate at the exit of the membrane modules (m /s) r rate of reaction (mol kg-1 s-1)

R gas constant (J/(mol.K))

164 CHAPTER 6. Pervaporation Membrane Reactor

rint internal radius of the membrane (m)

rp particle radius (m) Re Reynolds Sc Schmidt Sh Sherwood t time variable (s) T temperature in the feed side of the membrane (K)

Tm membrane temperature (K) u interstitial velocity (m/s) V volume of the feed thank (m3) v superficial velocity (m/s) 3 Vmol molar volume (m /mol) w mass fraction

xi liquid molar fraction of component i in the feed side

yi molar fraction in the vapor phase of component i z axial coordinate at the membrane modules (m)

Greek letters

γ i activity coefficient ε bulk porosity

ε p  particle porosity

υi  stoichiometric coefficient of component i

3 ρb bulk density (kg/m bed) ρ fluid density (kg/m3) μ viscosity (Pa.s) δ membrane thickness (m) λ thermal conductivity (W/(m.K)) A thickness of the selective layer of the membrane (m) α process separation factor

Subscripts

F relative to the feed

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 165

i relative to component i (i= Eth, La, EL, W)

0 relative to initial conditions

Perm relative to permeate

Eth relative to ethanol

La relative to lactic acid

EL relative to ethyl lactate

W relative to water

6.8 References

Asaeda M., M. Ishida and T. Waki, "Pervaporation of aqueous organic acid solutions by porous ceramic membranes", J. Chem. Eng. Jpn. 38(5): 336-343, 2005.

Bandini S., A. Saavedra and G. C. Sarti, "Vacuum Membrane Distillation: Experiments and Modeling", AlChE J. 43(2): 398-408, 1997.

Benedict D. J., S. J. Parulekar and S. P. Tsai, "Esterification of lactic acid and ethanol with/without pervaporation", Ind. Eng. Chem. Res. 42(11): 2282-2291, 2003.

Benedict D. J., S. J. Parulekar and S. P. Tsai, "Pervaporation-assisted esterification of lactic and succinic acids with downstream ester recovery", J. Membr. Sci. 281(1-2): 435-445, 2006.

Butt J. B., "Reaction kinetics and Reactor Design", Prentice-Hall: Englewood Cliffs, NJ (1980).

Casado C., A. Urtiaga, D. Gorri and I. Ortiz, "Pervaporative dehydration of organic mixtures using a commercial silica membrane: Determination of kinetic parameters", Sep. Purif. Technol. 42(1): 39-45, 2005.

Chapman P. D., T. Oliveira, A. G. Livingston and K. Li, "Membranes for the dehydration of solvents by pervaporation", J. Membr. Sci. 318(1-2): 5-37, 2008.

Crowder R. O. and E. L. Cussler, "Mass transfer resistances in hollow fiber pervaporation", Journal of Membrance Science 145(2): 173-184, 1998.

De Bruijn F., J. Gross, Å. OlujiÄ‡, P. Jansens and F. Kapteijn, "On the driving force of methanol pervaporation through a microporous methylated silica membrane", Ind. Eng. Chem. Res. 46(12): 4091-4099, 2007. de la Iglesia Ó., R. Mallada, M. Menéndez and J. Coronas, "Continuous zeolite membrane reactor for esterification of ethanol and acetic acid", Chem. Eng. J. 131(1-3): 35-39, 2007.

166 CHAPTER 6. Pervaporation Membrane Reactor

Delgado P., M. T. Sanz and S. Beltrán, "Pervaporation study for different binary mixtures in the esterification system of lactic acid with ethanol", Sep. Purif. Technol. 64(1): 78-87, 2008.

Delgado P., M. T. Sanz and S. Beltrán, "Pervaporation of the quaternary mixture present during the esterification of lactic acid with ethanol", J. Membr. Sci. 332(1-2): 113-120, 2009.

Dotremont C., S. Van den Ende, H. Vandommele and C. Vandecasteele, "Concentration polarization and other boundary layer effects in the pervaporation of chlorinated hydrocarbons", Desalination 95(1): 91-113, 1994.

Favre E., "Temperature polarization in pervaporation", Desalination 154(2): 129-138, 2003.

Feng X. and R. Y. M. Huang, "Studies of a membrane reactor: Esterification facilitated by pervaporation", Chem. Eng. Sci. 51(20): 4673-4679, 1996.

Feng X. and R. Y. M. Huang, "Liquid Separation by Membrane Pervaporation: A Review", Industrial & Engineering Chemistry Research 36(4): 1048-1066, 1997.

Figueiredo K. C. d. S., V. M. M. Salim and C. P. Borges, "Synthesis and characterization of a catalytic membrane for pervaporation-assisted esterification reactors", Catal. Today 133- 135(1-4): 809-814, 2008.

Gekas V. and B. Hallstrom, "Mass transfer in the membrane concentration polarization layer under turbulent cross flow. I. Critical literature review and adaptation of existing sherwood correlations to membrane operations", J. Membr. Sci. 30(2): 153-170, 1987.

Gómez P., R. Aldaco, R. Ibáñez and I. Ortiz, "Modeling of pervaporation processes controlled by concentration polarization", Computers & Chemical Engineering 31(10): 1326- 1335, 2007.

Hasanoglu A., Y. Salt, S. Keleser, S. Özkan and S. Dinçer, "Pervaporation separation of organics from multicomponent aqueous mixtures", Chemical Engineering and Processing: Process Intensification 46(4): 300-306, 2007.

Jafar J. J., P. M. Budd and R. Hughes, "Enhancement of esterification reaction yield using zeolite A vapour permeation membrane", J. Membr. Sci. 199(1): 117-123, 2002.

Karlsson H. O. E. and G. Trägardh, "Heat transfer in pervaporation", J. Membr. Sci. 119(2): 295-306, 1996.

Khayet M., C. Cojocaru and G. Zakrzewska-Trznadel, "Studies on pervaporation separation of acetone, acetonitrile and ethanol from aqueous solutions", Sep. Purif. Technol. 63(2): 303- 310, 2008.

Korkmaz S., Y. Salt, A. Hasanoglu, S. Ozkan, I. Salt and S. Dincer, "Pervaporation membrane reactor study for the esterification of acetic acid and isobutanol using polydimethylsiloxane membrane", Applied Catalysis A: General 366(1): 102-107, 2009.

Krishna Rao K. S. V., M. C. S. Subha, M. Sairam, N. N. Mallikarjuna and T. M. Aminabhavi, "Blend membranes of chitosan and poly(vinyl alcohol) in pervaporation dehydration of isopropanol and tetrahydrofuran", J. Appl. Polym. Sci. 103(3): 1918-1926, 2007.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 167

Lévêque M. A., "Les lois de transmission de chaleur par convection", Annal. Mines 13: 201, 1928.

Li X., H. Kita, H. Zhu, Z. Zhang and K. Tanaka, "Synthesis of long-term acid-stable zeolite membranes and their potential application to esterification reactions", J. Membr. Sci. 339(1- 2): 224-232, 2009.

Lim S. Y., B. Park, F. Hung, M. Sahimi and T. T. Tsotsis, "Design issues of pervaporation membrane reactors for esterification", Chem. Eng. Sci. 57(22-23): 4933-4946, 2002.

Lin L., Y. Zhang and Y. Kong, "Recent advances in sulfur removal from gasoline by pervaporation", Fuel, 2009.

Lipnizki F. and R. W. Field, "Mass transfer performance for hollow fibre modules with shell- side axial feed flow: using an engineering approach to develop a framework", J. Membr. Sci. 193(2): 195-208, 2001.

Lipski C. and P. Côté, "The Use of Pervaporation for the Removal of Organic Contaminants from Water", Environ Prog 9: 254-261, 1990.

Ma Y., J. Wang and T. Tsuru, "Pervaporation of water/ethanol mixtures through microporous silica membranes", Sep. Purif. Technol. 66(3): 479-485, 2009.

Namboodiri V. V., R. Ponangi and L. M. Vane, "A novel hydrophilic polymer membrane for the dehydration of organic solvents", Eur. Polym. J. 42(12): 3390-3393, 2006.

Ohshima T., T. Miyata and T. Uragami, "Selective removal of dilute benzene from water by various cross-linked poly(dimethylsiloxane) membranes containing tert-butylcalix[4]arene", Macromol. Chem. Phys. 206(24): 2521-2529, 2005.

Oliveira T. A. C., U. Cocchini, J. T. Scarpello and A. G. Livingston, "Pervaporation mass transfer with liquid flow in the transition regime", J. Membr. Sci. 183(1): 119-133, 2001.

Ortiz I., A. Urtiaga, R. Ibáñez, P. Gómez and D. Gorri, "Laboratory- and pilot plant-scale study on the dehydration of cyclohexane by pervaporation", J. Chem. Technol. Biotechnol. 81(1): 48-57, 2006.

Park B. G. and T. T. Tsotsis, "Models and experiments with pervaporation membrane reactors integrated with an adsorbent system", Chemical Engineering and Processing: Process Intensification 43(9): 1171-1180, 2004.

Pera-Titus M., J. Llorens and F. Cunill, "On a rapid method to characterize intercrystalline defects in zeolite membranes using pervaporation data", Chem. Eng. Sci. 63(9): 2367-2377, 2008.

Perkins L. R. and C. J. Geankoplis, "Molecular diffusion in a ternary liquid system with the diffusing component dilute", Chem. Eng. Sci. 24(7): 1035-1042, 1969.

168 CHAPTER 6. Pervaporation Membrane Reactor

Rathin D. and T. Shih-Perng, "Esterification of fermentation-derived acids via pervaporation", US Patent No. 5723639 (1998).

Sanz M. T. and J. Gmehling, "Esterification of acetic acid with isopropanol coupled with pervaporation. Part II. Study of a pervaporation reactor", Chem. Eng. J. 123(1-2): 9-14, 2006.

Slater C. S., T. Schurmann, J. MacMillian and A. Zimarowski, "Separation of diacteone alcohol-water mixtures by membrane pervaporation", Sep. Sci. Technol. 41(12): 2733-2753, 2006.

Smitha B., D. Suhanya, S. Sridhar and M. Ramakrishna, "Separation of organic-organic mixtures by pervaporation - A review", J. Membr. Sci. 241(1): 1-21, 2004.

Sommer S. and T. Melin, "Performance evaluation of microporous inorganic membranes in the dehydration of industrial solvents", Chem. Eng. Process. 44(10): 1138-1156, 2005.

Tanaka K., R. Yoshikawa, C. Ying, H. Kita and K. I. Okamoto, "Application of zeolite T membrane to vapor-permeation-aided esterification of lactic acid with ethanol", Chem. Eng. Sci. 57(9): 1577-1584, 2002. ten Elshof J. E., C. R. Abadal, J. Sekulic, S. R. Chowdhury and D. H. A. Blank, "Transport mechanisms of water and organic solvents through microporous silica in the pervaporation of binary liquids", Microporous Mesoporous Mater. 65(2-3): 197-208, 2003.

Urtiaga A. M., E. D. Gorri, J. K. Beasley and I. Ortiz, "Mass transfer analysis of the pervaporative separation of chloroform from aqueous solutions in hollow fiber devices", J. Membr. Sci. 156(2): 275-291, 1999.

Van Hoof V., C. Dotremont and A. Buekenhoudt, "Performance of Mitsui NaA type zeolite membranes for the dehydration of organic solvents in comparison with commercial polymeric pervaporation membranes", Sep. Purif. Technol. 48(3): 304-309, 2006.

Wasewar K., S. Patidar and V. K. Agarwal, "Esterification of lactic acid with ethanol in a pervaporation reactor: modeling and performance study", Desalination 243(1-3): 305-313, 2009.

Welty J., C. E. Wicks, G. L. Rorrer and R. E. Wilson, "Fundamentals of Momentum, Heat and Mass Transfer", John Wiley & Sons, (2008).

Wickramasinghe S. R., M. J. Semmens and E. L. Cussler, "Mass transfer in various hollow fiber geometries", J. Membr. Sci. 69(3): 235-250, 1992.

Wijmans J. G., A. L. Athayde, R. Daniels, J. H. Ly, H. D. Kamaruddin and I. Pinnau, "The role of boundary layers in the removal of volatile organic compounds from water by pervaporation", J. Membr. Sci. 109(1): 135-146, 1996.

Wijmans J. G. and R. W. Baker, "The solution-diffusion model: A review", J. Membr. Sci. 107(1-2): 1-21, 1995.

Yoshikawa M., K. Masaki and M. Ishikawa, "Pervaporation separation of aqueous organic mixtures through agarose membranes", J. Membr. Sci. 205(1-2): 293-300, 2002.

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 169

7. PermSMBR – A New Hybrid Technology

Abstract. In this Chapter, a new technology, the Simulated Moving Bed Membrane Reactor (PermSMBR), was presented and applied for the esterification of lactic acid with ethanol. Its conception was a result of ethyl lactate process re-intensification by integrating a permeable membrane reactor with a simulated moving bed reactor, since the products separation by selective adsorption is enhanced by continuous water removal through the membrane, shifting the equilibrium till high lactic acid conversion with high ethyl lactate purity. It was demonstrated that the PermSMBR technology applied for the ethyl lactate synthesis can reduce the desorbent consumption in 62 % and increase the ethyl lactate productivity in 33 % when compared with the SMBR and in 98 % when compared with reactive distillation for the same purity and conversion requirements.

Adapted from: V. M. T. M. Silva, Pereira C. S. M. and A. E. Rodrigues, "Reactor de membranas adsorptivo de leito móvel simulado, novo processo híbrido de separação e respectivas utilizações”, (PT 104496, patent pending, 2009) 172 CHAPTER 7. PermSMBR – A New Hybrid Technology

7.1 Introduction

The paradigm of chemical engineering process is changing. Traditional processes (where the reactor is followed by separation units in order to recover the desirable product, to remove the by-product and to recycle the unconverted reactants to the reactor) are being replaced by integrated processes where reaction and separation occur in the same device. These integrated processes are of considerable interest, mainly for equilibrium-limited reactions where the continuous removal of at least one reaction product shifts the equilibrium in order to increase conversion and reduce by-product formation. The term multifunctional reactor is often used to embrace reactive separations technology, which main advantages are higher yields, reduction of energy requirements, decrease of solvents consumption and lower capital investments. Typical examples of equilibrium-limited reactions include:

Esterification: R´-COOH + HO-R = R´COOR + H2O

Acetalization: R´-CHO + 2 HO-R = R´CH-(OR)2 + H2O

Ketalization: R´R´´CO + 2 HO-R = R´R´´C(OR)2 + H2O In the state of the art, the custom multifunctional reactors used for that type of reactions are: reactive distillations, reactive extractions, membrane reactors and chromatographic reactors. Regarding to reactive distillation (RD) the best example is the methyl acetate synthesis developed and patented by the company Eastman Kodak Company (Agreda and Partin, 1984). The entire process is carried out in a single column and represents one-fifth of the capital investment and consumes one-fifth of the energy of the traditional process (reaction followed by separation by distillation) (Krishna, 2002). However, there are some disadvantages in the use of RD for systems that exhibit azeotropes formation and/or the boiling points of the products are similar. Membrane reactors are widely used and typical examples are pervaporation and permeation reactors, where the catalyst is in fluidized (Alonso et al., 2001; Lee et al., 2006) or fixed bed (Lafarga and Varma, 2000; Zhu et al., 1996). Several times, processes where the reactor and membrane are housed in separate units in series or parallel are also regarded as membrane reactors (Datta and Tsai, 1998; Tsotsis et al., 2007). Chromatographic reactors include fixed bed (FBR) (Pereira et al., 2009a; Silva and Rodrigues, 2002), pressure swing adsorption (PSAR) and simulated moving bed reactors (SMBR) (Kawase et al., 1996; Pereira et al., 2009b; Silva and Rodrigues, 2005). However, from all the mentioned chromatographic reactors, the most common for process intensification for the production of oxygenated products is the SMBR. The SMBR is well- PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 173

known equipment (Broughton and Gerhold, 1961) that consists of a set of interconnected columns packed with an acid solid or a mixture of solids (catalysts and adsorbents). The description of an SMBR unit is presented in Chapter 5. The SMBR has several advantages, as the ones already mentioned for the reactive separations; however, it also has some disadvantages, as for example, the difficulty in removing the more adsorbed species, which implies high desorbent consumptions. Besides, in several applications, the feed is a mixture of reactants with one of the products, which will influence the performance of the unit resulting in low reactants conversion and products purities, low unit productivity and high eluent/desorbent consumption. This is even worse for the case where the product in the feed is the more retained one, since then the desorbent consumption will increase significantly. In order to overcome these issues, it was developed a novel technology, the Simulated Moving Bed Membrane Reactor (SMBMembR or PermSMBR), which comprises a reactor with two different separation techniques (chromatography – Simulated Moving Bed (SMB) with a selective permeable membrane – Pervaporation or Permeation) into a single device. It is a clean and economic alternative to conventional processes and even competitive when compared with the intensified processes. If reactive separations embrace the concept of Process Intensification, the reactive hybrid separations (as PermSMBR) embody the Process “ReIntensification” (Table 7.1).

Table 7.1 Chemical engineering process evolution.

Traditional Process Process Intensification Process ReIntensification

Reactor + Separator Reactive Separations Reactive Hybrid Separations Distillation • Reactive Distillation • SMB Membrane Reactor Adsorption • Membrane Reactor Crystallization • SMB Reactor Membranes Extraction

The PermSMBR technology is suitable mainly for oxygenated compounds as esters, acetals and ethers, used as biofuels, solvents, flavours, among others. In this Chapter, the PermSMBR will be exploited for the ethyl lactate synthesis. This new reactor leads to near depletion of lactic acid and high ethyl lactate purity with higher concentration, and, consequently, lower down streaming costs associated to the separation units. 174 CHAPTER 7. PermSMBR – A New Hybrid Technology

7.2 Technical description of the PermSMBR technology

The PermSMBR consists of a set of columns with membranes connected in series and packed with a solid, which could be a mixture of catalyst and selective adsorbent or a solid that acts both as catalyst and as adsorbent. Typically, there are two inlets, feed and desorbent, and three outlets, extract, raffinate and permeate. In Figure 7.1, there is schematic representation of a PermSMBR unit, where a reaction of type A+B ↔ C+D is considered.

Desorbent (A) Raffinate (A+C) Raffinate Feed

Liquid direction and ports switch A+B ⇔ C+D

Desorbent Extract Feed (A+B) Extract (A+D) Figure 7.1 Schematic diagram of a PermSMBR unit with 4 sections and three columns per section.

In this case, the component A is used as reactant and desorbent, therefore it is introduced in the system in the feed and desorbent streams. The other reactant B is used as feed. The products formed (C and D) are removed from the PermSMBR in three different streams: raffinate that is rich in the less strongly adsorbed product, C, extract, rich in the more strongly adsorbed product, D, and total permeate that combines all the permeate streams and that comprises mainly product D, for which the membranes are selective. All the inlet/outlet streams, with exception of the permeate streams, are introduced/removed from the system through ports positioned between the columns, and at regular time intervals, called the switching time, this streams are switched for one column distance in direction of the fluid flow. In this way, the countercurrent motion of the solid is simulated and its velocity is equal to the length of a column divided by the switching time. A cycle is completed when the number of switches is equal to a multiple of the total number of columns. The PermSMBR is PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 175

equipped with a rotary valve or a number of valves arranged in manner such that any feed stream may be introduced to any column and any outlet stream may be withdrawn from any column; it is also equipped with vacuum (or sweep gas flow) in order to withdraw the permeate stream from each column. The vacuum (or sweep gas flow) can be turned on in all the columns or just in some. Similarly to the SMBR, the position of the inlet/outlet streams defines different sections existing in the PermSMBR unit, each one accomplishing a certain function and containing a variable number of columns. The typical PermSMBR contains 4 sections, as represented in Figure 7.1 and Figure 7.2, where the section 1 is comprised between the desorbent and extract nodes, the section 2 is comprised between the extract and feed node; the section 3 is comprised between the feed and raffinate node, and the section 4 is comprised between the raffinate and desorbent node. In section 1, the adsorbent is regenerated by desorption of the more strongly adsorbed product (D) from the solid using the desorbent (A); In sections 2 and 3, reactive sections, the products C and D are separated as they are being formed. These products are continuously removed from the unit by adsorption and also through the selective membranes and, therefore, the reaction will proceed beyond the thermodynamic equilibrium, being possible to obtain 100 % of reactants conversion. In section 4, before being recycled to section 1, the desorbent is regenerated by adsorption of the less adsorbed product (C).

P P P P

Section 1 Section 2 Section 3 Section 4

D X F R

Figure 7.2 Schematic diagram of a PermSMBR unit with 4 sections: 2 inlet ports for feed (F) and desorbent (D) streams; 2 outlet ports for extract (X) and raffinate (R) streams; and outlet permeate streams (P).

The PermSMBR can have different configurations depending on the number of streams fed/removed from the unit. The total number of streams, with exception of the permeate streams, corresponds to the total number of sections. For example, the PermSMBR unit can be simplified to a unit of three sections: eliminating the extract stream, when the membrane is selective to the more adsorbed product (Figure 7.3); or eliminating the raffinate stream, when the membrane is selective to the less retained product (Figure 7.4).

176 CHAPTER 7. PermSMBR – A New Hybrid Technology

Figure 7.3 Schematic diagram of a PermSMBR unit with 3 sections: 2 inlet ports for feed (F) and desorbent (D) streams and 1 outlet port for raffinate (R) stream.

Figure 7.4 Schematic diagram of a PermSMBR unit with 3 sections: 2 inlet ports for feed (F) and desorbent (D) streams and 1 outlet port for Extract (X) stream.

If necessary, the PermSMBR unit can also be more complex, having five or more sections. For example, the schematic diagram of the process represented in Figure 7.5 is similar to the one described for the Figure 7.2, but it has 5 sections since an additional feed stream (F2) is introduced to the system. In this case, the regeneration of the solid and the desorbent is accomplished in sections 1 and 5, respectively; and the complete conversion of reactants and separation of the formed products occurs in section 2, 3 and 4. The feed F2 can comprise the same reactants as the ones in feed F1, but in different proportions, or other reactants in order to obtain the desired product. In the case where 3 products are formed, it will be necessary to have other extra stream designed as R2 in Figure 7.6. Additionally, if it is observed a decrease in the PermSMBR unit performance (decrease of products purity, reduction of reactants conversion and/or loss of productivity, among others) due to problems related to deactivation/poisoning of the catalyst/adsorbent and/or of the membranes, the PermSMBR unit can be operated in order to correct those problems, making a bypass to each one of the columns during a cycle to perform the necessary treatments (treatments with acids, solvents, replacement of the solid and/or of the membrane, thermal treatments, …). For example, if the desorbent is more expensive than the more adsorbed product, the treatment should by applied to the last column of section I (Figure 7.7), since this column is still saturated with the most PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 177

adsorbed product, avoiding the unnecessary consumption of desorbent if other column of the same section was selected. This procedure can also be performed in other sections in the most appropriate column. Other ways of correction along time can be performed to the operational variables of the PermSMBR unit, similarly to the procedure described in literature for the SMB unit (Sá Gomes et al., 2007).

P P P P P

Section 1 Section 2 Section 3 Section 4 Section 5

D X F1 F2 R Figure 7.5 Schematic diagram of a PermSMBR unit with 5 sections: 3 inlet ports for 2 feed streams (F1 and F2) and a desorbent stream (D), and 2 outlet ports for extract (X) and raffinate (R) streams.

Figure 7.6 Schematic diagram of a PermSMBR unit with 6 sections: 3 inlet ports for 2 feed streams (F1 and F2) and a desorbent stream (D), and 3 outlet ports for extract (X) and raffinate (R1 and R2) streams.

Figure 7.7 Schematic representation of the bypass to perform the regeneration/activation of catalyst/adsorbent and/or of the membrane in section 1 of the PermSMBR unit.

178 CHAPTER 7. PermSMBR – A New Hybrid Technology

7.3 PermSMBR mathematical model

Since the PermSMBR results from the integration of pervaporation membrane reactor and simulated moving bed reactor, its model will combine the specificities of each model previously described on Chapters 6 and 5, respectively; namely:

- axial dispersion flow for the bulk fluid phase;

- linear driving force (LDF) approximation for the inter and intra-particle mass transfer rates;

- multi-component adsorption equilibrium;

- velocity variations due to adsorption/desorption rates and species permeation;

- constant porosity and length of the packed bed;

- membrane concentration polarization;

- isothermal operation.

In Figure 7.8 a schematic representation of the fluxes inside one membrane of the PermSMBR is shown, where F is the molar flux in the feed side (retentate) and J is the permeate molar flux through the membrane.

membrane

Permeate b Feed Flow Permeate

F|z+dz Flow z+dz J|z J|z z

F|z

Figure 7.8 Schematic representation of the fluxes inside one membrane of the PermSMBR.

PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 179

Following this assumption the PermSMBR model equations are:

Bulk fluid mass balance to component i in column k

2 ∂∂CCuik() ik k(1−ε ) 3 ∂ CA ik m pik, (7.1) ++KL,,ik()CC ik −=− D ax k2 J ik ∂∂tzεε rp ∂ z

The permeate flux ( Ji ) is defined as:

0 JapyPiiiiiperm=−k(ov, ) (7.2)

where kov, i is a mass transfer coefficient based on a partial vapour pressure driving force (see

0 Chapter 6), ai is the activity of component i in bulk, pi is the saturation pressure of component i, Pperm is the total pressure on the permeate side and yi is the molar fraction of component i in the vapour phase (permeate side) defined as:

Ji yi = n (7.3) ∑ Ji i=1

The determination of the global mass transfer coefficient ( K L ) is presented in detail in

Chapter 4 and the calculation of the axial dispersion coefficient ( Dax ) is presented in previous Chapter (Chapter 6).

Interstitial fluid velocity variation calculated from the total mass balance du(1−ε ) 3 nn A km =−∑ KL,,ikVCC mol i() ik −pik, − ∑ J iK (7.4) dzεε rp ii==11 180 CHAPTER 7. PermSMBR – A New Hybrid Technology

where Vmol,i is the molar volume of component i (60.87 mL/mol, 77.56 mL/mol, 118.44 mL/mol and 18.63 mL/mol at 50 ºC for ethanol, lactic acid, ethyl lactate and water, respectively) and n is the total number of components.

Pellet mass balance to component i, in column k

∂C pik, ∂qik 3 εεppLikikip+−()1() =KCC, () −pik,, + υρ rC pik (7.5) ∂∂ttrp

where qik is the average adsorbed phase concentration of species i in column k in equilibrium with C p,ik , ε p the particle porosity, υi  the stoichiometric coefficient of component i, ρ p the particle density and r is the chemical reaction rate relative to the average particle concentrations in the fluid phase. The reaction rate and adsorption isotherms are those determined in Chapters 3 and 4, respectively.

Initial and Danckwerts boundary conditions

t = 0 : CCik==pik, C ik ,0 and qqik= ik ,0 (7.6)

∂C ik z = 0 : uCkik−= D axk,, uC kikF (7.7a) ∂z z=0

uukk= ,0 (7.7b)

∂C ik z = Lc : = 0 (7.7c) ∂z z=Lc where F and 0 refer to the feed and initial states, respectively.

Mass balances at the nodes of the inlet and outlet lines of the PermSMBR:

uu1 D D Desorbent node: CCCij(4,)== zLc=− ij (1,0) == z i (7.8a) uu44

Extract (j=2) and Raffinate (j=4) nodes: Ci( j−1,z=Lc) = Ci( j,z=0) (7.8b)

u3 uF F Feed node: Ci(2,z=Lc) = Ci(3,z=0) − Ci (7.8c) u2 u2 where, PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 181

u1 = u4 + uDs Desorbent (Ds) node ; (7.9a)

u2 = u1 − uX Extract (X) node ; (7.9b)

u3 = u2 + uF Feed (F) node ; (7.9c)

u4 = u3 − uR Raffinate (R) node ; (7.9d)

u j γ = (7.9e) j Us

The PermSMBR process performance and details on numerical solution are the same presented in Chapter 5.

7.4 PermSMBR geometrical specifications

The PermSMBR unit considered consists in 12 columns where each column has 13 commercial hydrophilic tubular membranes (Pervatech BV) in order to dehydrate the reaction medium. Regarding to the position of the membrane separation layer, it is possible to have different configurations. In the case of the silica membranes from Pervatech, the selective layer is inside the tube and consequently, it was considered that these membranes were packed with the resin Amberlyst 15-wet in the lumen side (inside the membrane tube). The PermSMBR parameters were chosen in order to compare its performance to the one of SMBR, since this technology will be the most competitive regarding to the PermSMBR. The same mass of catalyst and effective area was considered. The porosity of the SMBR unit was determined experimentally and presented in Chapter 5, while the PermSMBR porosity was estimated in agreement with the experimental results in fixed bed columns with ratio between tube diameter and particle diameter of 10 (Theuerkauf et al., 2006). The characteristics of the columns are presented in Table 7.2. 182 CHAPTER 7. PermSMBR – A New Hybrid Technology

Table 7.2 Characteristics of the columns for both SMBR and PermSMBR.

SMBR PermSMBR Solid weight (A15) 47.6 g 47.6 g Length of the bed (L) 23 cm 25.45 cm

Internal diameter (Di) 2.6 cm 0.7 cm Bed porosity (ε ) 0.4 0.424

3 3 Bulk density (ρb) 390 kg/m 374 kg/m

7.5 Simulated Results

The ethyl lactate synthesis using the SMBR was previously studied (Chapter 5) and the best performance obtained for that unit was under the following operation conditions: a feed of lactic acid solution (85 wt. % in water), a desorbent of ethanol (99.5 wt. % in water), a configuration of 3-3-4-2, a working temperature of 50ºC, a switching time of 2.1 min, andγγ14==3.654 and 1.161. In this section the PermSMBR will be evaluated and compared with the SMBR in order to demonstrate that the ethyl lactate production is enhanced by the integration of perm selective membranes in the SMBR unit and, therefore, the same operation conditions will be used. In order to keep the sameγ1 andγ 4 , the switching time (t *) was changed to 2.323, since the PermSMBR column length is different of the one of the SMBR unit (see Table 7.2). A summary of the operating conditions used is presented in Table 7.3.

Table 7.3 Operating conditions.

Operating conditions

Temperature 50 ºC Feed lactic acid (85 wt. % in water) Desorbent ethanol (99.5 wt. % in water) Configuration 3-3-4-2 Desorbent flow rate 58 mL/min Recycle flow rate 27 mL/min Switching time 2.323 min PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 183

Aiming to evaluate the geometrical parameters equivalence between the two units, the SMBR was compared to the PermSMBR in absence of permeation (considering in the model zero flux through the membranes) by using the parameters of Table 7.2 and operating conditions of Table 7.3. The feed flow rate and extract flow rate used were 7 mL/min and 39 mL/min, respectively. The performance parameters for both processes, presented in Table 7.4, are very similar, as expected.

Table 7.4 Comparison between SMBR and PermSMBR*

SMBR PermSMBR* PUX 99.92 99.94 PUR 96.19 97.85 X 99.51 99.37

-1 -1 PR (kgEL.Lresin .day ) 14.64 14.63

DC (LEth/kgEL) 5.98 5.98 * Results obtained from the PermSMBR model considering zero flux through the membranes (equal to a SMBR).

7.5.1 Reactive/Separation Region: PermSMBR vs SMBR

The reactive/separation region is a feasible region (in the γ2 - γ3 plane), which determines the operating conditions in sections 2 and 3, for given conditions on sections 1 and 4, to obtain products with specific purity requirement. The reactive/separation regions were calculated for both PermSMBR and SMBR processes, setting the conditions for complete regeneration on sections 1 and 4,γγ14(3.654) and (1.161) , and imposing a 95 % criteria for extract and raffinate purities and for lactic acid conversion. As it can be seen in Figure 7.9, the size of the reactive/separation region using the PermSMBR technology is higher than the one for the SMBR. This is justified by the fact that in the PermSMBR unit the water is removed not only by the selective adsorption onto the resin but also by the selective removal through the membrane; and therefore, the quantity of water removed from the system will be higher allowing higher feed flowrates of lactic acid solution without the contamination of the raffinate stream. It can be perceived that the PermSMBR allows working with higher feed flowrates (12.1 mL/min) than the SMBR (8.8 mL/min) to obtain the same purity and conversion requirements, and consequently higher productivity and lower desorbent consumption are achieved using this new technology. This is corroborated by the 184 CHAPTER 7. PermSMBR – A New Hybrid Technology

performance parameters obtained for both units working under the optimal operating conditions, presented in Table 7.5. As it can be seen, the ethyl lactate synthesis on the PermSMBR benefits its productivity in about 34 % and, additionally, decreases the desorbent consumption in 28 % which will reduce downstream costs since the products are less diluted.

3.0 PermSMBR SMBR 2.6

2.2 3 γ 1.8

1.4

1.0 1.0 1.4 1.8 2.2 2.6 3.0 γ2

Figure 7.9 Reactive/separation region for PermSMBR and SMBR processes (PermSMBR switching time of 2.323 min and SMBR switching time of 2.1 min; remaining conditions of Table 7.3)

Table 7.5 Performance parameters of the SMBR and PermSMBR for the EL synthesis.

SMBR PermSMBR Improvement

-1 -1 PR (kgEL.Lresin .day ) 18.06 24.19 33.94 %

DC (LEth/kgEL) 4.75 3.41 28.21 %

7.5.2 PermSMBR 3 zones

As stated before, the PermSMBR technology can be operated in different configurations, depending on the products to be separated and on the membrane performance. The ethyl lactate synthesis involves the formation of water, a by-product for which the membrane here considered is selective; therefore, it might be convenient to simplify the PermSMBR unit from 4 to 3 sections, eliminating the extract stream, which change the previous configuration

3-3-4-2 to 6-4-2. Setting the feed and desorbent flowrates at QmLF = 8 / min and

QmLD = 25 / min , respectively, and using the remaining conditions of Table 7.3, the PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 185

PermSMBR performance can be significantly enhanced by adjusting the permeate pressure, as shown in Table 7.6. Lactic acid conversion and ethyl lactate purity are the parameters most significantly improved. The analysis of internal concentration profiles at the cyclic steady- state, shown in Figure 7.10, allows better understanding the PermSMBR unit behaviour.

Table 7.6 Performance parameters for different permeate pressures.

Pperm 10 mbar 6 mbar 0 mbar

PUR (%) 92.85 96.15 99.63

X (%) 97.56 99.18 99.86

-1 -1 PR (kgEL.Lresin .day ) 16.27 16.51 16.59

DC (LEth/kgEL) 2.00 1.96 1.95

Water, the most adsorbed component is formed from the esterification reaction but is also fed to the system since an 85 % lactic acid aqueous solution is used. In order to have a high ethyl lactate purity and high lactic acid conversion, water should be removed from section 2, and desorbed from the resin in section 1. Since water is removed by pervaporation, the desorbent flowrate can be reduced below the value used in the SMBR unit, without compromising the resin regeneration. The reduction of the permeate pressure from 10 mbar (Figure 7.10a), to 6 mbar (Figure 7.10b) and 0 mbar (Figure 7.10c), increases the water permeation flux (Equation 7.2), enhancing significantly the resin regeneration on section 1, which avoids water to pass from section 1 to section 3 and increases the ethyl lactate purity. Since the reaction is equilibrium limited, this decrease on water content near the raffinate port prevents the ethyl lactate hydrolysis, increasing the lactic acid conversion, as consequence.

A SMBR unit to have the same raffinate purity, lactic acid conversion and ethyl lactate productivity as the PermSMBR ( Pperm = 6 mbar ), would have to operate in the following conditions: configuration of 3-3-4-2, t*2.1= min, QmLF = 8 / min , QmLX = 38 / min ,

QmLD = 58 / min and QmLRec = 27 / min . However, this would imply a desorbent consumption of 5.20 LEth/kgEL, which is 62 % higher than the one needed for the PermSMBR.

186 CHAPTER 7. PermSMBR – A New Hybrid Technology

18 (a) 16 Ethanol Lactic acid 14 Ethyl lactate

12 Water

6 Concentration (mol/L) Concentration 4

0 P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 D036912R F

18 (b) 16 Ethanol Lactic acid 14 Ethyl lactate

12 Water

6 Concentration (mol/L) Concentration 4

0 P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 D036912 F R

18 (c) 16 Ethanol Lactic acid 14 Ethyl lactate

12 Water

6 Concentration (mol/L) Concentration 4

0 P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 D036912R F Figure 7.10 Influence of permeate pressure on the concentration profiles at the middle of the switching time at cyclic steady state for the PermSMBR with 3

sections and configuration 6-4-2: a) Pmbarperm =10 ; b) Pperm = 6 mbar and

c) Pperm = 0 mbar . PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 187

7.5.3 Comparison between PermSMBR, SMBR and RD technologies The ethyl lactate synthesis using Amberlyst 15-wet as catalyst was successfully implemented by means of RD processes; and the best unit performance was obtained using 88 wt. % lactic acid solution, at 128ºC (bottom temperature), achieving 95 % of lactic acid conversion and 95 % of ethyl lactate purity (Asthana et al., 2005). From the previous studies on the SMBR and PermSMBR processes for the ethyl lactate synthesis at 50ºC, the best ethyl lactate productivity obtained for a criteria of 95 % of ethyl lactate purity and 95 % of lactic acid conversion (see Table 7.7), is 48 and 98 % higher than that of the RD process, respectively. However, in the SMBR and PermSMBR processes the excess of ethanol used to desorb water is higher than in the case of RD, leading to higher ethanol consumption.

Table 7.7 Performance parameters for RD, SMBR and PermSMBR technologies.

RD+ SMBR PermSMBR

-1 -1 PR (kgEL.Lresin .day ) 12.19 18.06 24.19

DC (LEth/kgEL) 2.17 4.75 3.41

+ (Asthana et al., 2005)

Another factor to take in consideration is that the RD operates at 128ºC, while the SMBR and PermSMBR systems work at 50ºC; thus, the energy consumption will be higher on the RD process. Therefore, comparison of technologies must be done in terms of economical assessment.

7.6 Conclusions

In this Chapter a new technology was proposed, the PermSMBR, that consists in a SMBR integrated with a permeable membrane reactor by using selective permeable membranes inside the columns of the SMBR. The PermSMBR proved to be more effective than the SMBR when applied to the ethyl lactate synthesis; higher ethyl lactate productivities and lower desorbent consumptions were obtained for the same purity and conversion criterions 188 CHAPTER 7. PermSMBR – A New Hybrid Technology

with this new reactor, since additionally to the driven force for product separation by selective adsorption there is the selective membrane removal of water.

The 3 sections PermSMBR operated in configuration 6-4-2 was studied, and it was concluded that it is not necessary the completely regeneration of the resin by water desorption in section 1 by introducing the sufficient amount of desorbent (ethanol), since water is also removed by the permeable membranes, preventing the ethyl lactate contamination. It was stated a decrease of 62 % in the desorbent consumption with this new technology using a 6 mbar vacuum pressure when comparing with the SMBR to attain the same ethyl lactate purity, productivity and lactic acid conversion.

When compared with the RD process, the PermSMBR requires higher ethanol consumption, but allows a significantly ethyl lactate productivity increase of 98 %, operating at quite lower temperatures.

Concluding, the PermSMBR is a very interesting technology, even compared with other intensified processes, that allows complete reactants conversion, high productivity, high purity, significant reduction of solvent consumption and consequently lowers downstreaming costs associated to the separation units.

7.7 Notation

ai liquid-phase activity of component i in bulk side

2 3 Am membrane area per unit reactor volume (m membrane/m bulk)

C liquid phase concentration (mol/L)

2 Dax axial dispersion coefficient (m /min)

DC desorbent consumption (L/mol)

2 Ji permeate flux of species i (mol/(m min))

KL global mass transfer coefficient

2 kov global membrane mass transfer coefficient (mol/(m sPa))

L column length (m) PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 189

n total number of components

0 pi saturation pressure of component i (bar)

Pperm total pressure on the permeate side (bar)

PR raffinate productivity (kgEL/(L resin.day))

PUR raffinate purity (%)

PUX extract purity (%) q solid phase concentration in equilibrium with the fluid concentration inside the particle (mol/L)

Q volumetric flowrate (L/min) r rate of reaction (mol kg-1 min-1)

rp particle radius (m) t time variable (min) t * switching time (min)

Us solid velocity (m/min) u interstitial velocity (m/min)

Vmol, i molar volume of species i (L/mol)

X lactic acid conversion

yi molar fraction in the vapor phase of component i z axial coordinate (m)

Greek letters

γ interstitial velocities ratio

ε bulk porosity 190 CHAPTER 7. PermSMBR – A New Hybrid Technology

ε p  particle porosity

υi  stoichiometric coefficient of component i

ρ p particle density

Subscripts i relative to component i (i= Eth, La, EL, W) j relative to section in SMBR (j = 1, 2, 3, 4) k relative to column in SMBR

0 relative to initial conditions

Eth relative to ethanol

La relative to lactic acid

EL relative to ethyl lactate

W relative to water

F relative to the feed

p relative to particle

R relative to raffinate

Re c relative to recycle

X relative to extract

Superscripts

F relative to the feed

R relative to raffinate

X relative to extract PROCESSES INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 191

7.8 References

Agreda V. H. and L. R. Partin, "Reactive distillation process for the production of methyl acetate", U. S. Patent 4435595 (1984).

Alonso M., M. J. Lorences, M. P. Pina and G. S. Patience, "Butane partial oxidation in an externally fluidized bed-membrane reactor", Catal. Today 67(1-3): 151-157, 2001.

Asthana N., A. Kolah, D. T. Vu, C. T. Lira and D. J. Miller, "A continuous reactive separation process for ethyl lactate formation", Org. Process Res. Dev. 9(5): 599-607, 2005.

Broughton D. B. and C. G. Gerhold, "Continuous Sorption Process Employing Fixed Bed of Sorbent and Moving Inlets and Outlets", US Patent No. 2 985 589 (1961).

Datta R. and S.-P. Tsai, "Esterification of Fermentation-Derived Acids via Pervaporation", WO Patent No. 9823579 (1998).

Kawase M., T. B. Suzuki, K. Inoue, K. Yoshimoto and K. Hashimoto, "Increased esterification conversion by application of the simulated moving-bed reactor", Chem. Eng. Sci. 51(11): 2971-2976, 1996.

Krishna R., "Reactive separations: more ways to skin a cat", Chem. Eng. Sci. 57(9): 1491- 1504, 2002.

Lafarga D. and A. Varma, "Ethylene epoxidation in a catalytic packed-bed membrane reactor: Effects of reactor configuration and 1,2-dichloroethane addition", Chem. Eng. Sci. 55(4): 749- 758, 2000.

Lee W. N., I. J. Kang and C. H. Lee, "Factors affecting filtration characteristics in membrane- coupled moving bed biofilm reactor", Water Res. 40(9): 1827-1835, 2006.

Pereira C. S. M., V. M. T. M. Silva and A. r. E. Rodrigues, "Fixed Bed Adsorptive Reactor for Ethyl Lactate Synthesis: Experiments, Modelling, and Simulation", Sep. Sci. Technol. 44(12): 2721 - 2749, 2009a.

Pereira C. S. M., M. Zabka, V. M. T. M. Silva and A. E. Rodrigues, "A novel process for the ethyl lactate synthesis in a simulated moving bed reactor (SMBR)", Chem. Eng. Sci. 64(14): 3301-3310, 2009b.

Sá Gomes P., M. Minceva and A. E. Rodrigues, "Operation strategies for simulated moving bed in the presence of adsorbent ageing", Sep. Sci. Technol. 42(16): 3555-3591, 2007.

Silva V. M. T. M. and A. E. Rodrigues, "Dynamics of a fixed-bed adsorptive reactor for synthesis of diethylacetal", AIChE J. 48(3): 625-634, 2002.

Silva V. M. T. M. and A. E. Rodrigues, "Novel process for diethylacetal synthesis", AlChE J. 51(10): 2752-2768, 2005.

Theuerkauf J., P. Witt and D. Schwesig, "Analysis of particle porosity distribution in fixed beds using the discrete element method", Powder Technol. 165(2): 92-99, 2006. 192 CHAPTER 7. PermSMBR – A New Hybrid Technology

Tsotsis T. T., M. Sahimi, B. Fayyaz-Najafi, A. Harale, B.-G. Park and P. K. T. Liu, "Hybrid adsorptive membrane reactor", US Patent No. 2007053811 (A1) (2007).

8. Conclusions and Suggestions for Future Work

This thesis focused on the development of a new efficient process to produce ethyl lactate based in the esterification reaction between ethanol and lactic acid by using hybrid technologies of reaction/separation based on the Simulated Moving Bed Reactor (SMBR) and Pervaporation Membrane processes. Therefore, several topics were addressed from the fundamentals (thermodynamic equilibrium and kinetics of reaction, multi-component adsorption equilibria and pervaporation data), to the process development (modelling and simulation) and intensification (simulated moving bed reactor, pervaporation membrane reactor and an integration of both). The main results and conclusions are:

(i) Batch Reactor: Thermodynamic Equilibrium and Reaction Kinetics

To circumvent the lack of a suitable estimative for the thermodynamic equilibrium constant as function of temperature, the equation ln K = 2.9625 − 515.13 T (K ), based on the UNIQUAC method, and valid in the temperature range of 50-90ºC, was proposed.

The kinetic law was described by a Langmuir-Hinshelwood rate expression based on activities, considering the surface reaction as the rate-controlling step. The following kinetic law was proposed for the temperature range of 50-90 ºC, for the Amberlyst 15-wet catalyst:

2 rkaa=−c( Eth La aaK EL W/(1) ++ Ka Eth Eth Ka W W ); and the model parameters are kmolg( /( .min))=× 2.70 107 exp − 6011.55/ TK ( ) , K =15.19exp 12.01/T(K) and c ( ) W ( )

K Eth =1.22exp(359.63/T(K)).

(ii) Fixed Bed Adsorptive Reactor

The multi-component adsorption equilibria were determined from dynamic adsorption experiments in absence of reaction, at 20 ºC and 50 ºC, using Amberlyst 15-wet as selective

194 CHAPTER 8 Conclusions and Suggestions for Future Work

adsorbent. The multi-component Langmuir adsorption isotherm was modified in order to reduce the adjustable adsorption parameters from 8 (one molar monolayer capacity and one equilibrium constant for each component) to 5 (one volumetric monolayer capacity for all components and one equilibrium constant for each component): using the following parameters: qQVKCiVmoliii=+()/1, ( ∑ KC jj)

Q (ml/l ) K (l/mol) V (ml/mol) Component V wet solid mol 20 ºC / 50 ºC 20 ºC / 50 ºC 20 ºC / 50 ºC Ethanol 5.443 / 3.068 58.17 / 60.87 Lactic acid 4.524 / 4.085 74.64 / 77.56 390.0/383.5 Ethyl lactate 1.117 / 1.815 113.99 / 118.44 Water 15.353 / 7.055 18.08 / 18.63

(iii) Simulated Moving Bed Reactor

The ethyl lactate was produced in the Simulated Moving Bed Reactor pilot unit LICOSEP, using Amberlyst 15-wet resin as catalyst and selective adsorbent. A mathematical model considering external and internal mass-transfer resistances and variable velocity due to change of liquid composition was developed to describe the dynamic behaviour of the SMBR and it was validated by the experiments performed. The theoretical assessment of the SMBR unit behaviour was performed ensuring complete regeneration of the resin (in section 1) and desorbent (in section 4), by using the mathematical model to analyse the effect of SMBR configuration, feed composition and switching time into the reactive/separation regions or/and into the process performance at the optimal operating points. It was shown that the SMBR is a very attractive technology for the production of ethyl lactate, since under appropriate conditions the lactic acid conversion can be driven to completion and productivity as high as 32 kgEL/(Lads.day) and purity of 95 % can be obtained.

(iv) Pervaporation Membrane Reactor

Pervaporation process using commercial Hydrophilic silica membranes from Pervatech was evaluated for the ethyl lactate system aiming to contribute either for the separation of SMBR extract stream (water/ethanol mixtures) or for the ethyl lactate process intensification by continuous pervaporation membrane reactor (PVMR). First pervaporation studies indicate that the membranes have no major imperfections since the total flux and water selectivity is

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 195

barely affected by absolute feed pressure. The influence of hydrodynamic conditions on the membrane polarization was analyzed, and for velocity values higher than 0.14 m/s polarization effects are eliminated. It was concluded that the permeances for all species as function of temperature and feed water content, are described by the following equations for the temperature range 48ºC - 72ºC:

3 −7 ⎛⎞22.60× 10 2 Qmemb, Eth =×1.41 10 exp⎜⎟mol/( dm bar min) ⎝⎠RT Q = 0 mol/( dm2 bar min) memb, LA

3 −4 ⎛⎞10.42× 10 2 Qmemb, EL =×1.12 10 exp⎜⎟mol/( dm bar min) ⎝⎠RT

−6 ⎛⎞50377xW − 32326 2 Qxmemb, W=×2 10 exp(18.642 W )exp⎜⎟ − mol/( dm bar min) ⎝⎠RT

A mathematical model was developed for the batch pervaporation membrane (BPM), considering (i) plug flow for the bulk fluid phase; (ii) total feed volume inside the tank and retentate velocity inside the membrane variations due to permeation of components; (iii) concentration polarization, where the resistance due to the diffusive transport in the boundary layer is combined with the membrane resistance in a global membrane resistance; (iv) non- isothermal operation due to heat consumption for species vaporization; and (v) temperature polarization. The BPM model was validated experimentally and, therefore, it was extended to the integrated pervaporation membrane reactor packed with the catalyst Amberlyst-15wet. In isothermal conditions, the PVMR is a very attractive solution being possible to achieve near lactic acid depletion (98 % conversion) and ethyl lactate with 96 % purity, at 70ºC.

(v) PermSMBR – A New Hybrid Technology

A new technology, the Simulated Moving Bed Membrane Reactor (PermSMBR), was developed and applied for the esterification of lactic acid with ethanol. Its conception was a result of process re-intensification by integrating a permeable membrane reactor with a simulated moving bed reactor. The PermSMBR proved to be more effective than the SMBR when applied to the ethyl lactate synthesis; higher ethyl lactate productivities and lower desorbent consumptions were obtained for the same purity and conversion criterions with this new reactor, since additionally to the driven force for product separation by selective adsorption there is the selective membrane removal of water. It was demonstrated that the

196 CHAPTER 8 Conclusions and Suggestions for Future Work

PermSMBR technology can reduce the desorbent consumption in 62 % and increase the ethyl lactate productivity in 33 % when compared with the SMBR and in 98 % when compared with RD for the same purity and conversion requirements. Concluding, the PermSMBR is a promising technology, even if compared with other intensified processes that allows complete reactants conversion, high productivity, high purity, significant reduction of solvent consumption and consequently lowers downstream costs associated to the separation units.

This thesis is focused on the ethyl lactate synthesis with a special contribution on its process intensification. The SMBR technology exhibits higher productivity than that of the PVMR, but has the disadvantage of requiring further separation units for recovery of ethanol from both extract and raffinate streams. Integrating the benefits of SMBR and PVMR, a new technology was invented, the PermSMBR. However, certain topics were not exploited and others are unfinished and, therefore, can be deeper investigated. For that reason it is suggested to explore, in a near future, the following research lines:

(i) Screening of new catalysts

Smopex 101 fibre has proved to be an efficient catalyst for acetalization and esterification reactions; since it has negligible adsorption capacity is not suitable for chromatographic reactor, namely the SMBR and PermSMBR, although would be valuable for the PVMR. Another interesting catalyst is the Amberlyst 36-wet (A36), a strongly acidic ion exchange resin, which proved to be more active than Amberlyst 15-wet (A15) for esterification reactions; moreover, as adsorbent has selectivity to water since the swelling from dry to water is 54 %, while A15 only swells 37 %. Therefore, A36 will enhance the performance of SMBR and PermSMBR by improving both reaction kinetics and products separation.

(ii) Deactivation studies

It is known that catalyst and/or adsorbent ageing affects significantly the performance of industrial processes, and might compromise the process feasibility. Ageing studies should consider not only pro-analytical reactants but also industrial reactants (lactic acid from fermentation broth and bio-ethanol). Its influence on the ion-exchange resin deactivation, in terms of reaction kinetics and multi-component adsorption equilibria, should be addressed, in

PROCESS INTENSIFICATION FOR THE GREEN SOLVENT ETHYL LACTATE PRODUCTION 197

order to understand its implications on the processes performance. Analysis of non-isothermal effects onto the SMBR and PermSMBR unit performance should also be performed.

(iii) Process Integration and Economical Evaluation

The economical assessment of the ethyl lactate synthesis should be carried out considering its integration with the subsequent separation units. Comparison between technologies (SMBR, PVMR, PermSMBR, RD, ...) should be based on financial project evaluation using the Net Present Value analysis, net annual profit or annual production cost. Alternatively, the evaluation of the most sustainable technology should be based on the Life Cycle analysis.

Appendix A: Safety Data

1.1 Ethyl lactate

NFPA 2 • Health: 3

30 • Flammability: 2

• Reactivity: 0

1.1.1 General

Synonyms: lactic acid ethyl ester

Molecular formula: C5H10O3 CAS No: 97-64-3

1.1.2 Physical Data

Appearance: colourless liquid Melting point: -26 ºC Boiling point: 154 ºC Vapour density: 4.07 (air = 1) Vapour pressure (mmHg): 5 @ 30ºC (86ºF) Density (g cm-3): 1.03 Flash point: 46 ºC (closed cup) Explosion limits: 1.5 – 11.4% Autoignition temperature: 400 ºC Water solubility: appreciable

1.1.3 Stability

Stable. Combustible. Incompatible with strong oxidizing agents. APPENDIX A. Safety Data A 2

1.1.4 Toxicology

Skin, eye and respiratory irritant.

Toxicity data:

Oral-mouse lethal dose 50% kill 2500 mg kg-1

Intraperitoneal – rat lowest published lethal dose 1000 mg kg-1

Subcutaneous – mouse lethal dose 50% kill 2500 mg kg-1

Risk phrases: • R36 Irritating to eyes. • R37 Irritating to respiratory system. • R38 Irritating to skin.

1.1.5 Personal Protection

Minimize contact. Safety phrases: • S26 In case of contact with eyes, rinse immediately with plenty of water and seek medical advice.

1.1.6 Hazard Summary

• Ethyl lactate can affect you when breathed in and may be absorbed through the skin.

• Prolonged contact can irritate the skin and eyes.

• Breathing ethyl lactate may cause dizziness, lightheadedness, and passing out.

1.1.7 How to determine if you are being expose

• Exposure to hazardous substances should be routinely evaluated. This may include collecting personal and area air samples. Under OSHA 1910.20, you have a legal right to obtain copies of sampling results from your employer. If you think you are experiencing any work-related health problems, see a doctor trained to recognize occupational diseases. Take this Fact Sheet with you. APPENDIX A. Safety Data A 3

1.1.8 Workplace exposure limits

No occupational exposure limits have been established for ethyl lactate. This does not mean that this substance is not harmful. Safe work practices should always be followed.

• It should be recognized that ethyl lactate may be absorbed through the skin, thereby increasing your exposure.

1.1.9 Ways of reducing exposure

Where possible, enclose operations and use local exhaust ventilation at the site of chemical release. If local exhaust ventilation or enclosure is not used, respirators should be worn.

Wear protective work clothing.

Wash thoroughly immediately after exposure to ethyl lactate and at end of the workshift.

Post hazard and warning information in the work area. In addition, as part of an ongoing education and training effort, communicate all information on the health and safety hazards of ethyl lactate to potentially exposed workers.

1.1.10 Health hazard information

Acute Health Effects

The following acute (short-term) health effects may occur immediately or shortly after exposure to ethyl lactate:

Prolonged contact can irritate the skin and eyes.

Breathing ethyl lactate may cause dizziness, lightheadedness and passing out.

Chronic Health Effects

The following chronic (long-term) health effects can occur at some time after exposure to ethyl lactate and can last for months or years:

Cancer Hazard

According to the information presently available to the New Jersey Department of Health and Senior Services, ethyl lactate has not been tested for its ability to cause cancer in animals.

APPENDIX A. Safety Data A 4

Reproductive Hazard

According to the information presently available to the New Jersey Department of Health and Senior Services, ethyl lactate has not been tested for its ability to affect reproduction.

Other Long-Term Effects

Ethyl lactate has not been tested for other chronic (long-term) health effects.

Medical Testing

There is no special test for this chemical. However, if illness occurs or overexposure is suspected, medical attention is recommended.

Any evaluation should include a careful history of past and present symptoms with an exam. Medical tests that look for damage already done are not a substitute for controlling exposure.

Request copies of your medical testing. You have a legal right to this information under OSHA 1910.1020.

1.1.11 Workplace controls and practices

Unless a less toxic chemical can be substituted for a hazardous substance, ENGINEERING CONTROLS are the most effective way of reducing exposure. The best protection is to enclose operations and/or provide local exhaust ventilation at the site of chemical release. Isolating operations can also reduce exposure. Using respirators or protective equipment is less effective than the controls mentioned above, but is sometimes necessary.

In evaluating the controls present in your workplace, consider: (1) how hazardous the substance is, (2) how much of the substance is released into the workplace and (3) whether harmful skin or eye contact could occur. Special controls should be in place for highly toxic chemicals or when significant skin, eye, or breathing exposures are possible.

In addition, the following controls are recommended:

• Where possible, automatically pump liquid Ethyl Lactate from drums or other storage containers to process containers.

Good WORK PRACTICES can help to reduce hazardous exposures. The following work practices are recommended:

• Workers whose clothing has been contaminated by Ethyl Lactate should change into clean clothing promptly. APPENDIX A. Safety Data A 5

• Contaminated work clothes should be laundered by individuals who have been informed of the hazards of exposure to Ethyl Lactate.

• Eye wash fountains should be provided in the immediate work area for emergency use.

• If there is the possibility of skin exposure, emergency shower facilities should be provided.

• On skin contact with Ethyl Lactate, immediately wash or shower to remove the chemical. At the end of the workshift, wash any areas of the body that may have contacted Ethyl Lactate, whether or not known skin contact has occurred.

• Do not eat, smoke, or drink where Ethyl Lactate is handled, processed, or stored, since the chemical can be swallowed. Wash hands carefully before eating, drinking, smoking or using the toilet.

Personal Protective Equipment

WORKPLACE CONTROLS ARE BETTER THAN PERSONAL PROTECTIVE EQUIPMENT. However, for some jobs (such as outside work, confined space entry, jobs done only once in a while, or jobs done while workplace controls are being installed), personal protective equipment may be appropriate.

The following recommendations are only guidelines and may not apply to every situation.

Clothing

• Avoid skin contact with Ethyl Lactate. Wear solvent-resistant gloves and clothing. Safety equipment suppliers/manufacturers can provide recommendations on the most protective glove/clothing material for your operation.

• All protective clothing (suits, gloves, footwear, headgear) should be clean, available each day, and put on before work.

Eye Protection

• Wear indirect-vent, impact and splash resistante goggles when working with liquids.

• Wear a face shield along with goggles when working with corrosive, highly irritant or toxic substances.

APPENDIX A. Safety Data A 6

Respiratory Protection

IMPROPER USE OF RESPIRATORS IS DANGEROUS. Such equipment should only be used if the employer has a written program that takes into account workplace conditions, requirements for worker training, respirator fit testing and medical exams, as described in OSHA 1910.134.

• Engineering controls must be effective to ensure that exposure to Ethyl Lactate does not occur.

• Where de potential for overexposure exists, use a MSHA/NIOSH a approved supplied-air respirator with a full facepiece operated in a pressure-demand or other positive-pressure mode. For increased protection use in combination with an auxiliary self-contained breathing apparatus operated in a pressure-demand or other positive-pressure mode.

1.1.12 Handling Storage

• Prior to working with Ethyl Lactate you should be trained on its proper handling and storage.

• Ethyl Lactate is not compatible with OXIDIZING AGENTS( such as PERCHLORATES, PEROXIDES, PERMANGANATES, CHLORATES, NITRATES, CHLORINE, BROMINE and FLUORINE).

• Store in tightly closed containers in a cool, dark, well-ventilated area.

• Sources of ignition, such as smoking and open flames, are prohibited where Ethyl Lactate is handled, used, or stored.

1.1.13 Fire Hazards

• Ethyl Lactate is a COMBUSTIBLE LIQUID.

• Use dry chemical, CO2, water spray or alcohol resistant foam extinguishers.

• CONTAINERS MAY EXPLODE IN FIRE.

• Use spray to keep fire-exposed containers cool.

• If employees are expected to fight fires, they must be trained and equipped as stated in OSHA 1910.156. APPENDIX A. Safety Data A 7

1.1.14 Spills and emergencies

If Ethyl Lactate is spilled or leaked, take the following steps:

• Evacuate person’s not wearing protective equipment from area of spill or leak until clean- up is complete.

• Remove all ignition sources.

• Ventilate area of spill or leak.

• Absorb liquids in vermiculite, dry sand, earth, or a similar material and deposit in sealed containers.

• It may be necessary to contain and dispose of Ethyl Lactate as a HAZARDOUS WASTE. Contact your Department of Environmental Protection (DEP) or your regional office of the federal Environmental Protection Agency (EPA) for specific recommendations.

• If employees are required to clean-up the spills, they must be properly trained and equipped. OSHA 1910.120(q) may be applicable.

1.1.15 First aid

Eye Contact

Immediately flush with large amounts of water. Continue without stopping for at least 15 minutes, occasionally lifting upper and lower lids.

Skin Contact

Remove contaminated clothing. Wash contaminated skin with soap and water.

Breathing

• Remove the person from exposure.

• Begin rescue breathing if breathing has stopped and CPR if heart action has stopped.

• Transfer promptly to a medical facility.

APPENDIX A. Safety Data A 8

1.2 Lactic Acid

NFPA 1 • Health: 2

20 • Flammability: 1 • Reactivity: 0

1.2.1 General

Synonyms: 2-hydroxypropanoic acid, ethylideneactic acid, 1-hydroxyethanecarboxylic acid

Molecular formula: CH3CHOHCOOH CAS No: 50-21-5 EC No: 200-018-0

1.2.2 Physical Data

Appearance: colourless to yellow liquid Melting point: 18 ºC Boiling point: 122 ºC @12mmHg Specific gravity: 1.05

1.2.3 Stability

Stable. Combustible. Incompatible with strong oxidizing agents.

1.2.4 Toxicology

Eye or skin contact may cause severe irritation or burns.

Toxicity data:

Oral-rat lethal dose 50% kill 3730 mg kg-1

Subcutaneous – mouse lethal dose 50% kill 4500 mg kg-1

Risk phrases:

• R34 Causes burns. APPENDIX A. Safety Data A 9

1.2.5 Personal Protection

Safety glasses.

Safety phrases:

• S26 In case of contact with eyes, rinse immediately with plenty of water and seek medical advice.

• S27 Take off immediately all contaminated clothing.

• S36 Wear suitable protective clothing.

• S37 Wear suitable gloves.

• S39 Wear eye / face protection.

1.2.6 Health Effects

• Inhalation: causes burns on contact with mucous membranes.

• Eye: Causes burns on contact with eyes.

• Skin: Causes burns on contact with skin.

• Ingestion: Harmful if swallowed. Can cause abdominal pain, diarrhea, and burns of digestive tract.

• Routes of Entry: Inhalation, Ingestion or skin contact.

1.2.7 First Aid

• Inhalation: Remove to fresh air; give artificial respiration if breathing has stopped. Get medical attention.

• Eye: Immediately flush eyes with large amounts of water for at least 15 minutes.

• Skin: Immediately flush thoroughly with large amounts of water. Remove contaminated clothing and wash before reuse.

• Ingestion: Do not induce vomiting; get medical immediate attention.

1.2.8 Fire Hazard

• Extinguishing Media: water fog, foam, carbon dioxide, dry chemical APPENDIX A. Safety Data A 10

• Special Procedures: Wear self-contained breathing apparatus.

• Unusual Hazard: Emits acrid fumes when heated.

1.3 Ethanol

NFPA

3 • Health: 0 00 • Flammability: 3

• Reactivity: 0

1.3.1 General

Synonyms: ethyl alcohol, grain alcohol, fermentation alcohol, alcohol, methylcarbinol, absolute alcohol, absolute ethanol, anhydrous alcohol, alcohol dehydrated, algrain, anhydrol.

Molecular formula: C2H5OH

CAS No: 64-17-5

1.3.2 Physical data

Appearance: colourless liquid

Melting point: -130 C

Boiling point: 78 C

Specific gravity: 0.789

Vapour pressure: 1.59

Flash point: 56 F

Explosion limits: 3.3% - 24.5%

Autoignition temperature: 683 F

Water solubility: miscible in all proportions APPENDIX A. Safety Data A 11

1.3.3 Stability

Stable. Substances to be avoided include strong oxidising agents, peroxides, acids, acid chlorides, acid anhydrides, alkali metals, ammonia, moisture. Forms explosive mixtures with air.

1.3.4 Toxicology

Causes skin and eye irritation. Ingestion can cause nausea, vomiting and inebriation; chronic use can cause serious liver damage. Note that “absolute” alcohol, which is close to 100% ethanol, may nevertheless contain traces of 2-propanol, together with methanol or benzene. The latter two are very toxic, while “denatured” alcohol has substances added to it which make it unpleasant and possibly hazardous to consume.

Risk phrases:

• R11 Highly flammable.

• R20 Harmful by inhalation.

• R21 Harmful in contact with skin.

• R22 Harmful if swallowed.

• R36 Irritating to eyes.

• R37 Irritating to respiratory system.

• R38 Irritating to skin.

• R40 Possible risk of irreversible effects.

1.3.5 Personal protection

Safety glasses. Suitable ventilation.

Safety phrases:

• S7 Keep container tightly closed.

• S16 Keep away from sources of ignition.

• S24 Avoid contact with skin.

• S25 Avoid contact with eyes. APPENDIX A. Safety Data A 12

• S36 Wear suitable protective clothing.

• S37 Wear suitable gloves.

• S39 Wear eye / face protection.

• S45 In case of accident or if you feel unwell, seek medical advice immediately (show the label whenever possible.)

Appendix B: Thermodynamic Properties

1.1 Available Literature Data

The data presented in this section are from Yaws (Yaws, 1999), excepted when mentioned.

In Table B.1 some physical and thermodynamic properties of lactic acid, ethanol, ethyl lactate and water are presented.

Table B.1 Basic properties of lactic acid, ethanol, ethyl lactate and water. Properties Lactic Acid Ethanol Ethyl Lactate Water Molecular weigh - M (g/mol) 90.079 46.069 118.133 18.015 3 Density- ρ (g/cm ) 1.209 0.789 1.031 1.027

Melting temperature - Tf (K) 289.95-291.15 159.15 248.25 273.15

Normal boiling temperature - Tb (K) 395.15 351.45 426.15-427.15 373.15

Critical temperature - Tc (K) 616.00 516.25 588.00 647.13

Critical pressure - Pc (bar) 59.65 63.84 38.60 221.20 3 Critical volume - Vc (cm /mol) 216.9 166.9 354.0 57.1 Acentric factor - ω 1.035 0.637 0.793 0.344

1.1.1 Density

The modified form of the Rackett equation was selected for correlation of saturated liquid density as a function of temperature.

T −(1− )n Tc ρ L = AB (B.1)

3 with ρ L (g / cm ) and T (K).

APPENDIX B. Thermodynamic Properties B2

Table B.2 Constants used for density calculation.

Constants Lactic Acid Ethanol Water Ethyl Lactate A 0.39816 0.26570 0.34710 0.33372 B 0.26350 0.26395 0.27400 0.21190 n 0.28570 0.23670 0.28571 0.45530

Tmin (K) 291.15 159.05 273.16 247.15

Tmax (K) Tc Tc Tc Tc

1.1.2 Viscosity

The correlation for liquid viscosity as a function of temperature is given by Equation B.2.

B log η = A + + CT + DT 2 (B.2) 10 L T with η L (cP) and T (K).

Table B.3 Constants used for viscosity calculation.

Constants Ethanol Water Ethyl Lactate A -6.4406E+00 -10.2158E+00 -20.0105E+00 B 1.1176E+03 1.7925E+03 3.2123E+03 C 1.3721E-02 1.7730E-02 4.1891E-02 D -1.5465E-05 -1.2631E-05 -3.2733E-05

Tmin (K) 240 273 247

Tmax (K) Tc 643 Tc

1.1.3 Vapour Pressure

The Antoine-type equation with extended term was selected for correlation of vapour pressure as a function of temperature:

B log P = A + + C log T + DT + ET 2 (B.3) 10 vp T 10

with Pvp (mmHg) and T (K).

APPENDIX B. Thermodynamic Properties B3

Table B.4 Constants used for vapour pressure calculation. Constants Lactic Acid Ethanol Water Ethyl Lactate A -27.0836E+00 23.8442E+00 29.8605E+00 32.0863E+00

B -3.9661E+03 -2.8642E+03 -3.1522E+03 -2.9164E+03

C 2.0233E+01 -5.0474E+00 -7.3037E+00 -9.5666E+00

D -4.2176E-02 3.7448E-11 2.4247E-09 6.5114E-03

E 2.0310E-05 2.7361E-07 1.8090E-06 4.5645E-13

Tmin (K) 291.15 159.05 273.16 247.15

Tmax (K) Tc Tc Tc Tc

1.1.4 Liquid Heat Capacity

The correlation for heat capacity of liquid is a series expansion in temperature, given by Equation B.4.

234 CABTCTDTp =+ + + + ET (B.4) with Cp (J/(Kmol.K)) and T (K), with exception for the ethyl lactate specie where Cp is given in (J/(mol.K)).

Table B.5 Constants used for heat capacity calculation.

Constants Ethanol* Water* Ethyl Lactate Lactic acid* A 1.0264E+05 2.7637E+05 -46.239E+00 6.1082E+04 B -1.3963E+02 -2.0901E+03 2.1823E+00 5.0343E+02 C -3.0341E-02 8.1250E+00 -5.9832E-03 ------D 2.0386E-03 -1.4116E-02 6.8683E-06

E ------9.3701E-06 ------

Tmin (K) 159.05 273.16 248.00 289.90

Tmax (K) 390.00 533.15 529.00 675.00 Error < 3% < 1% ------< 10%

* parameters taken from (DIPPR, 1998).

APPENDIX B. Thermodynamic Properties B4

1.1.5 Heat of Vaporization

The correlation selected for the calculation of the heat of vaporization as a function of temperature is given by Equation B.5 (DIPPR, 1998).

2 V B++CTrr DT ΔH = A ⎣⎦⎡⎤1−Tr ( ) (B.5)

V with TTTrc= / and ΔH in (J/(Kmol).

Table B.6 Constants used for heat vaporization calculation.

Constants Ethanol Water Ethyl Lactate Lactic acid A 5.5789E+07 5.2053E+07 8.0260E+07 1.0436E+08 B 3.1245E-01 3.1990E-01 4.0930E-01 3.8548E-01 C ------2.1200E-01 ------D ------2.5795E-01 ------

Tmin (K) 159.05 273.16 247.15 289.90

Tmax (K) 514.00 647.13 588.00 675.00 Error < 1% < 1% < 10% < 25%

1.1.6 Thermal Conductivity

The thermal conductivity was calculated by Equation B.6 (DIPPR, 1998).

234 λ =+ABTCTDT + + + ET (B.6) with λ (W/(m.K)) and T (K).

Table B.7 Constants used for thermal conductivity calculation.

Constants Ethanol Water Ethyl Lactate Lactic acid A 2.4680E-01 -4.3200E-01 2.8358E-01 3.4850E-01 B -2.6400E-04 5.7255E-03 -3.5110E-04 -3.7085E-04 C ------8.0780E-06 ------D ------1.8610E-09 ------Tmin (K) 159.05 273.16 247.15 289.90 Tmax (K) 353.15 633.15 427.65 490.00 Error < 5% < 1% < 25% < 10%

APPENDIX B. Thermodynamic Properties B5

1.2 Properties Estimation

1.2.1 Estimation of Liquid Viscosity

The lactic acid viscosity was estimated by the following expression (DIPPR, 1998):

⎛⎞4097.9 ηLA =−exp⎜⎟ 14.403 + − 0.4407 × ln(T ) (B.7) ⎝⎠T

with ηL (.)Pa s and T (K). Table B.8 Viscosity of lactic acid. T (K) 293.15 323.15 η (cP) 53.67 13.99

For the estimation of the viscosity of the remaining species (ethanol, ethyl lactate and water) it was used the Equation B.2 presented in section 1.1.2.

1.2.2 Estimation of Vapour Pressure

For correlation of vapour pressure as a function of temperature the Antoine equation with extended term was chosen (Equation B.3).

2.50

Lactic acid 2.00 Ethanol Water 1.50 Ethyl lactate (atm) vp

P 1.00

0.50

0.00 290 320 350 380 T (K) Figure B. 1 Variation of the vapour pressure of the compounds with the temperature.

APPENDIX B. Thermodynamic Properties B6

1.2.3 Estimation of Liquid Heat Capacity

The liquid heat capacity was estimated by the correlation given by Equation B.4 (see Figure B.2).

350 Ethanol Water 300 Ethyl lactate Lactic acid

250

200

150

Cp (J / (mol.K)) 100

0 290 310 330 350 370 T (K)

Figure B.2 Variation of the liquid heat capacity of the different species with the temperature.

1.2.4 Estimation of Heat of Vaporization

The heat of vaporization was estimated by the correlation given by Equation B.5 (see Figure B.3).

9.0E+04

7.5E+04 Ethanol Water Ethyl lactate 6.0E+04 (J / mol) (J Lactic acid V H Α 4.5E+04

3.0E+04 290 310 330 350 370 T (K)

Figure B.3 Variation of the heat of vaporization of the different species with the temperature.

APPENDIX B. Thermodynamic Properties B7

1.2.5 Estimation of Molar Volumes

The molar volumes for all components were estimated with Gunn-Yamada method (Reid et al., 1987):

f (T ) V ()T = V R (B.8) f ()T R where

f (T ) = H 1 (1−ωH 2 ) (B.9)

2 3 4 H1 = 0.33593 − 0.33953Tr +1.51941Tr − 2.02512Tr +1.11422Tr (B.10)

2 H 2 = 0.29607 − 0.09045Tr − 0.04842Tr (B.11)

T T R Tr = or (B.12) Tc Tc

VR is the molar volume at the reference temperature TR (cm3/mol), ω is the acentric factor and Tc is the critical temperature (K).

Gunn-Yamada method could be used just in the case when the liquid molar volume is known at some temperature (reference temperature). The reference molar volumes are presented at Table B.9.

Table B.9 Molar volumes at the reference temperature used for all calculations. VR (cm3/mol) Reference Temperature TR (K) Lactic Acid Ethanol Ethyl Lactate Water 293.15 74.640 58.174 113.986 18.082

1.3 References

DIPPR, "Thermophysical Properties Database", (1998). Reid R. C., P. J.M. and P. B.E., "The Properties of Gases and Liquids", McGraw-Hill, (1987). Yaws C. L., "Chemical Properties Handbook", McGraw-Hill, (1999).

Appendix C: Calibration

1.1 Calibration

1.1.1 Pure Components The calibration was realized by injecting different volumes of pure components at 20 ºC and registering the respective peak area values. The molar volume, VM, of each species was used to convert volume, V, in number of moles, n. In the case of lactic acid the volume was converted to number of moles using the solution density (ρ), since lactic acid is not pure (lactic acid 85 w/w % pure).

The response factor (fi) was defined as:

n i = f i A i (C.1)

where ni is the number of moles of component i (μmol) and Ai is the area of component i (u.a.)

Table C.1 Ethanol 3 VM (cm /mol) V (μL) n (μmol) A (u.a.) 10 0.1 1.7093 0.3394 nEth = 5.879AEth 8 2 0.1 1.7093 0.3410 R = 0.983 0.2 3.4185 0.6408 6 (µmol) 0.2 3.4185 0.6385 4 Eth 0.3 5.1278 0.9178 n 58.17 2 0.3 5.1278 0.9208 0.4 6.8371 1.1555 0 0.4 6.8371 1.1796 0 0.4 0.8 1.2 1.6 AEth (u.a.) 0.5 8.5464 1.4011 0.5 8.5464 1.3564 Figure C.1

APPENDIX C. Calibration C2

Table C.2 Lactic acid 3 6 ρ (g/cm ) V (μL) n (μmol) A (u.a.) nLA = 6.994ALA 5 R2 = 0.990 0.2 2.2375 0.3144 0.2 2.2375 0.3212 4 0.3 3.3562 0.4732 (µmol)

LA 3 0.3 3.3562 0.4574 n 1.209 2 0.4 4.4750 0.6262 0.4 4.4750 0.6612 1 0.5 5.5937 0.7800 0.3 0.5 0.7 0.9 0.5 5.5937 0.8294 ALA (u.a.)

Figure C.2

Table C.3 Ethyl lactate 3 VM (cm /mol) V (μL) n (μmol) A (u.a.)

0.1 0.8695 0.3074 5 0.1 0.8695 0.3020 nEL = 3.345AEL 4 R2 = 0.986 l) 0.2 1.7390 0.5713 o 0.2 1.7390 0.5721 m3 (µ L 0.3 2.6085 0.8174 E 113.99 n 2 0.3 2.6085 0.8202 0.4 3.4780 1.0289 1 0.4 3.4780 1.0371 0 0.5 4.3475 1.2496 0 0.4 0.8 1.2 1.6 0.5 4.3475 1.2407 AEL (u.a.) Figure C.3

Table C.4 Water 3 VM (cm /mol) V (μL) n (μmol) A (u.a.) 30 0.1 5.5253 0.5409 25 nW= 11.826AW 0.1 5.5253 0.5434 R2 = 0.988 20 0.2 11.0506 1.0200 15 0.2 11.0506 1.0287 (µmol) W 10 0.3 16.5759 1.4712 n 18.08 0.3 16.5759 1.4729 5 0.4 22.1012 1.8591 0 0.4 22.1012 1.8350 0 0.5 1 1.5 2 2.5 0.5 27.6266 2.2369 AW (u.a.) 0.5 27.6266 2.2602 Figure C.4

APPENDIX C. Calibration C3

1.1.2 Binary mixtures

Several standard binary mixtures with known concentration were prepared, analysed and the molar fraction of each component was determined, accordingly to Equation C.2.

fiiA xest, i = (C.2) ∑ fnnA n

In order to introduce a correction factor, the estimated molar fraction (obtained from Equation C.2) and the real molar fraction of each component were adjusted for each pair by a linear fitting (see Figure C.5).

xCalc,, i= abx+ est i (C.3)

1 0.4 Lactic acid/Water Ethyl lactate/Ethanol 0.8 0.3 0.6 real,EL real,LA 0.2

x 0.4 x

0.1 0.2 xreal,EL = 1.1304xest,EL -0.0177 xreal,LA = 1.089xest,LA + 0.0118 R2 = 0.9998 R2 = 0.9984 0 0 0 0.1 0.2 0.3 0.4 0 0.2 0.4 0.6 0.8 1 xest,EL xest,LA

0.8 1 Ethanol/Water Ethyl lactate/Water 0.8 0.6 0.6 0.4 real,EL real,Eth real,Eth 0.4 x x 0.2 x = 0.9706x - 0.0044 0.2 xreal,Eth = 1.0492xest,Eth -0.018 real,EL est,EL 2 R² = 0.9993 R = 0.9995 0 0 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 1 xest,EL xest,Eth Figure C.5 Real molar fraction as function of estimated molar fraction of a specie in different binary mixtures.

APPENDIX C. Calibration C4

1.1.3 Multicomponent mixtures Several standard multicomponent mixtures with known concentrations were prepared and analysed. The real molar fraction and the estimated molar fraction (from Equation C.2) of each component were adjusted by a linear fitting (Figure C.6).

1 0.6 xreal,Eth = 0.9986xest,Eth -0.016 xreal,EL = 0.8803xest,EL + 0.0031 0.8 R² = 0.9986 R² = 0.9993 0.4 0.6 real,Eth real,EL x 0.4 x 0.2 0.2

0 0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 x xest,EL est,Eth

0.6 xreal,W = 0.9969xest,W -0.0144 R² = 0.999 0.4 real,W

x 0.2

0 0 0.2 0.4 0.6 xest,W Figure C.6 Real molar fraction as function of estimated molar fraction of a specie in multicomponent mixtures.

The lactic acid molar fraction was calculated from mass balance: xLA=1(−++xxx Eht EL W ).

1.2 Validation of Calibration

The molar fraction of a component in a binary mixture is calculated using Equation C.2 and the equation obtained from the linear fitting of the respective pair (presented in Figure C.5). For the case of multicomponent mixtures the procedure for the calculation of a specie molar fraction is the same that the one used for binary mixtures; first it is used Equation C.2 and then the equation obtained from the linear fitting of the specie in a multicomponent mixture (presented in Figure C.6). For both cases, the concentration is then evaluated using the liquid molar volumes (Equation C.4):

APPENDIX C. Calibration C5

x i (C.4) Ci = ∑ xVnMn, n In order to verify the analysis accuracy for the binary and quaternary mixtures several samples with known concentration were prepared and analysed, as shown in Tables C.4-C.8.

Table C.4 Analysis of Ethanol/Ethyl lactate mixtures. Ethyl lactate molar fraction Sample Real Calculated Error (%) 1 0.1919 0.1935 0.82 2 0.5078 0.5075 -0.07 3 0.8968 0.8991 0.26

Table C.5 Analysis of Lactic acid/Water mixtures. Lactic acid molar fraction Sample Real Calculated Error (%) 1 0.0818 0.0809 -1.12 2 0.3182 0.3213 0.98 3 0.4098 0.4146 1.17

Table C.6 Analysis of Ethanol/Water mixtures. Ethanol molar fraction Sample Real Calculated Error (%) 1 0.1560 0.1574 0.87 2 0.6025 0.5912 -1.88

Table C.7 Analysis of Ethyl lactate/Water mixtures. Ethyl lactate molar fraction Sample Real Calculated Error (%) 1 0.2789 0.2781 -0.27 2 0.6565 0.6622 0.87

APPENDIX C. Calibration C6

Table C.8 Analysis of a quaternary mixture. Molar fraction Component Real Calculated Error (%) Ethanol 0.3969 0.3981 0.29 Lactic acid 0.1865 0.1900 1.88 Ethyl lactate 0.2275 0.2233 -1.85 Water 0.1891 0.1886 -0.24

Appendix D: Binary adsorption experiments at 293.15 K

The breakthrough curves for the binary adsorption experiments performed at 293.15 K are presented below.

(a)

60 Water 50 Ethanol Theoretical 40

Outlet concentration (mol/L) 10

0 0 40 80 120 160 200 Time (min) (b)

60 water 50 ethanol Theorical 40

10 Outlet concentration (mol/L) concentration Outlet 0 0 102030405060 Time (min) Figure C.1 Breakthrough experiments: outlet concentration of ethanol and water as a function of time; Q = 5 mL/min; T = 293.15 K; (a) water displacing ethanol; Bottom up flow direction; (b) ethanol displacing water; Top-down flow direction.

APPENDIX D. Breakthrough curves at 293.15 K D 2

(a) 18

16 Ethyl lactate Ethanol 14 Theoretical 12

4 Outlet concentration concentration Outlet (mol/L) 2

0 0 40 80 120 160 200 Time (min) (b)

18 16

14 ethyl lactate 12 ethanol Theorical 10 8 6 4

Outlet concentration concentration Outlet (mol/L) 2 0 0 10203040 Time (min)

Figure C.2 Breakthrough experiments: outlet concentration of ethanol and ethyl lactate as a function of time; Q = 5 mL/min; T = 293.15 K; (a) ethyl lactate displacing ethanol; Bottom up flow direction (b) ethanol displacing ethyl lactate; Top-down flow direction.

APPENDIX D. Breakthrough curves at 293.15 K D 3

Water 50 Lactic acid Theoretical 40

Outlet concentration concentration Outlet (mol/L) 10

0 0 20406080100 Time (min)

Figure C.3 Breakthrough experiments: outlet concentration of water and lactic acid as a function of time; Q = 5 mL/min; T = 293.15 K; Lactic acid displacing water; Bottom up flow direction.