CARBON DIOXIDE REMOVAL IN GAS TREATING PROCESSES

Transcription

1 HÅVARD LIDAL NO NEI-NO--562 CARBON DIOXIDE REMOVAL IN GAS TREATING PROCESSES TH UNIVERSITETET I TRONDHEIM NORCES TEKNISKE HØGSKOLE DOKTOR INGENIØR AVHANDLING 1992:26 INSTITUTT FOR KJEMITEKNIKK TRONDHEIM DISTRIBUTION OF THIS DOCUMENT IS'UNUMITED'f smamrnm

2 CARBON DIOXIDE REMOVAL IN GAS TREATING PROCESSES by Håvard Lidal A Thesis Submitted for the Degree of Dr. Ing. The University of Trondheim The Norwegian Institute of Technology Department of Chemical Engineering Trondheim, June 1992 MASTER NJH.IRVKH 199! DISTRIBUTION OF THIS DOCUMENT IS UNLIMITED

3 ADDENDUM The cooperation of the industrial participants in this SPUNG project, Norsk Hydro a.s and Kværner Engineering A/S, is greatly appreciated. I

4 ACKNOWLEDGEMENTS I am most obliged to my supervisor Olav Erga for all his professional and personal support. His encouragement, inspiring personality, and wholehearted interest in the field of gas treating have given me the backing I needed during this work. I wish to express my sincere appreciation to Dag Eimer of Norsk Hydro a.s. I learned a lot from discussions we had, and I enjoyed working with him on various projects. Thanks are due to Olav Juliussen of SINTEF for his technical assistance with the laboratory equipment. I also wish to acknowledge the contributions of A.R. Fossen-Helle, J. Bjørvik, W.E. Olsen, M. Schneider, and M. Tørnqvist for performing parts of the experiments. Thanks also to all those representatives of the gas industry, professors and staff members from our university and universities and research establishments around the world, and other people I had the opportunity to meet and have inspiring discussions with. In particular, I would like to thank Orville C Sandall from UCSB who accepted to serve on my dissertation committee, and travel all the way from California to do this. Above all, I would like to give special thanks ;,o God and my family, especially my late father, my mother, and my brother. I gratefully acknowledge the financial support of the Royal Norwegian Council for Scientific and Industrial Research (NTNF), given as a part of the SPUNG Programme (State R&D Programme for Utilization of Natural Gas). The support of the Foundation for Scientific and Industrial Research at the Norwegian Institute of Technology (SINTEF), as well as grants received from NTHs Fond, M.H. Lungreens Enkes Fond, and Lise og Arnfinn Hejes Fond, are greatly appreciated. in

5 ABSTRACT A semiempirical thermodynamic model which represents the equilibrium partial pressure of C0 2 over aqueous solutions of tertiary and sterically hindered amines, is presented. The model has been used on the tertiary amine methyldiethanolamine (MDEA), and on the sterically hindered amine 2-amino-2-methyl-1-propanol (AMP). Measurements of ph as a function of C0 2 concentration play an important role in the modelling procedure. The model is based on the ph data, together with solubility measurements performed in this work and also solubility data collected from the literature. Solubility and ph measurements were made over a temperature range of 25 to 70 C, and for amine concentrations of 3M AMP and 4 and 4.28M MDEA. The model relates the equilibrium partial pressure of CO2 as a function of the amine concentration, the C0 2 loading, and the temperature. For MDEA solutions, the model covers the temperature interval of 25 to 140 C, and can be used for C0 2 loadings between and 1 mol C0 2 /mol MDEA, and at C0 2 partial pressures between and 50 atm. The model is tested against experimental data from several literature references with amine concentrations ranging from 1.69 to 4.28M, and it is found to predict the experimental data very well. While the presented model covers both absorption and desorption conditions for MDEA solutions, the application range is restricted to absorption conditions for AMP solutions. The technique of utilizing measured ph data in the modelling of vapor-liquid equilibrium, distinguishes the present model from equilibrium models found in the literature. Establishing accurate relations for ph as a function of the C0 2 loading and the IV

6 temperature, constitute the backbone on which the model is based. The solubility of C0 2 has been measured over a temperature range of 30 to 70 C in mixed nonaqueous solutions of glycols and alkanolamines. The following systems have been studied: Triethyleneglycol (TEG) with either monoethanolamine (MEA) or diethanolamine (DEA), and diethyleneglycol (DEG) with MEA. Measurements were made with amine contents of 5, 10, and 13.6mol%. The solubility in these mixed solvents is compared with other mixed solvents and also with aqueous amine solutions. The effect of temperature and amine concentration on solubility is also discussed. To be able to estimate the CO2 partial pressure at temperatures above 70 C, a vapor-liquid equilibrium model is developed for the TEG/MEA-system. The model, which is in many aspects similar to the model developed for the aqueous system, shows satisfactory agreement with the available experimental data. The rates of C0 2 absorption into mixed solvents have been measured using a string-of-discs experimental set-up. These experiments were undertaken on five solvents with and without the addition of 5mol% MEA. The following solvents were investigated: N-methyl-pyrrolidone, ethanol, diethyleneglycol monomethylether, TEG, and water. Due to problems with temperature rise, only approximate data have been obtained. To improve our laboratory facilities, two new experimental setups have been designed and built. These are an apparatus for solubility measurements at temperatures above 70 C, and a onesphere apparatus for determinations of reaction kinetics. Both sets of equipment are described in this thesis. v

7 TABLE OF CONTENTS Acknowledgements Abstract List of tables List of figures iii iv ix xi Chapter One Introduction Acid Gas Removal Technologies Alkanolaraine Solutions Scope of the Work 7 Chapter Two Literature Review VLE Data in Gas Treating Processes VLE Measurements in Aqueous Alkanolamine Solutions VLE Modelling in Aqueous Alkanolamine Solutions VLE Measurements in Mixed Nonaqueous Solvents VLE Measurements in Pure Physical Solvents VLE Modelling Techniques in Physical Solvents Chemistry of C0 2 - Amine Systems Introduction Reactions between C0 2 and Amines in Aqueous Solutions Reaction Kinetics between C0 2 and Aqueous MDEA Reaction Kinetics between C0 2 and Aqueous AMP Reaction Kinetics in Nonaqueous Solutions Experimental Equipment for Kinetic Determinations 27 Chapter Three Experimental Vapor-Liquid Equilibrium Measurements Equilibrium Equipment for Temperatures up to 70 C A New Equipment for Temperatures up to 120 C ph Measurements Kinetic Measurements String-of-discs Column One-sphere Apparatus Chemicals and Gases Liquid Analysis C0 2 Concentration Amine Concentration Gas Analysis 36 vi

8 Chapter Four Experimental Results Vapor-Liquid Equilibrium Measurements C0 2 Solubility in Aqueous MDEA Solutions C0 2 Solubility in Aqueous AMP Solutions C0 2 Solubility in Nonaqueous Amine Solutions ph Measurements Aqueous MDEA Solutions Aqueous AMP Solutions Kinetic Measurements 49 Chapter Five A Model for Equilibrium Solubility of C0 2 in Aqueous Solutions of the Tertiary Amine MDEA Introduction C0 2 Equilibrium Model for Aqueous 4M MDEA Approximations The Basic Model A Correlation for ph A Correlation for logk A Preliminary Final Model Comparison with Experimental Equilibrium Data Extended Equilibrium Model, Valid for Aqueous Solutions with 1-4.5M MDEA at Temperatures between 25 and 1 40 C Introducing VLE Data from the Literature A New Correlation for the Parameter K The Final Model Comparison with Experimental Equilibrium Data Accuracy of the Model Conclusions 71 Chapter Six A Model for Equilibrium Solubility of C0 2 in an Aqueous Solution of the Sterically Hindered Amine AMP Introduction The Equilibrium Model for C Approximations The Basic Model A Correlation for ph A Correlation for logk The Final Model Comparison with Experimental Equilibrium Data Limitations Conclusions 80 vii

9 Chapter Seven Vapor-Liquid Equilibria of Mixed Nonaqueous Solvents Equlibrium Solubility Model for C0 2 in TEG/MEA Solutions Background Modelling Procedure Comparison with Experimental Equilibrium Data Comparison with Aqueous Amine Solutions Comparison with other Mixed Solvents Comparison with Pure Physical Solvents 92 Chapter Eight Conclusions and Recommendations Conclusions Recommendations 94 Nomenclature 96 References 98 Appendix A Appendix B Tabulated Data of C0 2 Solubility in Aqueous Systems 108 Tabulated Data of C0 2 Solubility in Nonaqueous Systems 110 Appendix C Tabulated ph Data for Aqueous Systems 116 Appendix D Tabulated Results of Kinetic Measurements Appendix E HP-42S Program for Calculation of Equilibrium Partial Pressure of C0 2 over Aqueous MDEA viii

10 LIST OF TABLES Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7 Table 8 Table 9 Table 10 Table 11 Table 12 Solubility of C0 2 in aqueous solutions of 4.00M MDEA at 30, 45, and 60 C 108 Solubility of C0 2 in aqueous solutions of 4.28M MDEA at 25, 40, and 70 C 109 Solubility of C0 2 in aqueous solutions of 3.00M AMP at 40 and 50 C 109 Solubility of C0 2 in solutions of TEG and 5mol% MEA at 30, 50, and 70 C 110 Solubility of C0 2 in solutions of TEG and 10mol% MEA at 30, 50, and 70 C 111 Solubility of C0 2 in solutions of TEG and 5mol% DEA at 30, 50, and 70 C 112 Solubility of C0 2 in solutions of TEG and 10mol% DEA at 30 and 50 C 113 Solubility of C0 2 in solutions of TEG and 13.6mol% DEA at 30, 50, and 70 C 114 Solubility of C0 2 in solutions of DEG and 5mol% MEA at 40 C 115 Solubility of C0 2 in solutions of DEG and 10mol% MEA at 40 C 115 ph values as a function of C0 2 loading in aqueous solutions of 4.00M MDEA at 30, 40, 50, and 60 C. 116 ph values as a function of C0 2 loading in aqueous solutions of 3.00M AMP at 20, 30, 40, and 50 C Table 13 Rate of absorption of C0 2 in water at 20 C 118 Table 14 Table 15 Table 16 Rate of absorption of C0 2 in a solution of water and 5mol% MEA at 20 C 118 Rate of absorption of C0 2 in n-methylpyrrolidone at 20 C 119 Rate of absorption of C0 2 in a solution of n-methyl-pyrrolidone and 5mol% MEA at 20 C 119 IX

11 Table 17 Rate of absorption of C0 2 in ethanol at 20 C 120 Table 18 Rate of absorption of C0 2 in a solution of ethanol and 5mol% MEA at 20 C 120 Table 19 Rate of absorption of C0 2 in triethyleneglycol at 20 C C 121 Table 20 Rate of absorption of C0 2 in a solution of triethyleneglycol and 5mol% MEA 121 Table 21 Rate of absorption of C0 2 in diethyleneglycol monomethylether at 20 C 122 Table 22 Rate of absorption of C0 2 in a solution of diethyleneglycol monomethylether and 5mol% MEA at 20 C 122 x

12 LIST OF FIGURES Figure 2.1 Figure 3.1 Figure 3.2 Figure 3.3 Figure 3.4 Figure 4.1 Figure 4.2 Figure 4.3 Figure 4.4 Figure 4.5 Figure 4.6 Figure 4.7 Figure 4.8 Figure 4.9 Figure 4.10 Figure 4.11 Molecular structure of amines used in acid gas removal processes 19 Gas-liquid equilibrium equipment 30 New gas-liquid equilibrium equipment, capable of measuring solubilities at temperatures encountered in desorption units String-of-discs absorber 32 Schematic of operation of string-of-discs and one-sphere apparatus 34 Solubility of C0 2 in aqueous 4.00M MDEA solutions at 30, 45, and 60 C 38 Solubility of C0 2 in aqueous 4.28M MDEA solutions at 25, 40, and 70 C, compared with literature data 39 Solubility of C0 2 in aqueous 3.00M AMP solutions, compared with literature data Solubility of C0 2 in TEG solutions containing 5mol% MEA at 30, 50, and 70 C Solubility of C0 2 in TEG solutions containing 10mol% MEA at 30, 50, and 70 C Solubility of C0 2 in TEG solutions containing 5mol% DEA at 30, 50, and 70 C Solubility of C0 2 in TEG solutions containing 10mol% DEA at 30 and 50 C 44 Solubility of C0 2 in TEG solutions containing 13.6mol% DEA at 30, 50, and 70 C. 45 Solubility of C0 2 in DEG solutions containing 5mol% MEA and 1 Omol% MEA at 40 C 46 Experimental ph data for aqueous 4.00M MDEA solutions at 30, 40, 50, and 60 C 47 Experimental ph data for aqueous 3.00M AMP solutions at 20, 30, 40, and 50 C 48 XI

13 Figure 4.12 Figure 4.13 Figure 5.1 Figure 5.2 Figure 5.3 Figure 5.4 Figure 5.5 Figure 5.6 Figure 5.7 Figure 5.8 Figure 5.9 Rate of absorption of C0 2 in physical solvents as a function of wetting rate at 20 C. 50 Rate of absorption of C0 2 in physical solvents containing 5mol% MEA as a function of wetting rate at 20 C 51 pkp' as a function of the temperature for aqueous 4.00M MDEA solution 57 logk as a function of the temperature for aqueous 4.00M MDEA solution 58 Comparison of the present model with experimental data from the literature on the system of 4.28M MDEA aqueous solution at 25, 40, 70, 100, and 120 C 63 Comparison of the present model with experimental data from the literature on the system of 4.28M MDEA aqueous solution at 140 C 64 Comparison of the present model with present experimental data and data taken from the literature on the system of 4.28M MDEA aqueous solution at 40 C 65 Comparison of the present model with experimental data from the literature on the system of 2.00M MDEA aqueous solution at 25, 40, 70, 100, and 120 C 66 Comparison of the present model with experimental data from the literature on the system of 2.00M MDEA aqueous solution at 40 C 67 Comparison of the present model with experimental data from the literature on the system of 3.04M MDEA aqueous solution at 40 and 100 C Comparison of the present model with experimental data from the literature on the system of 1.69M MDEA aqueous solution at 100 C 69 xii

14 Figure 5.10 Figure 6.1 Figure 6.2 Figure 7.1 Figure 7.2 Figure 7.3 Figure 7.4 Figure 7.5 Figure 7.6 Figure 7.7 Figure 7.8 Comparison of the present model with present experimental data for aqueous solutions of 4.00M MDEA at 30 C 70 pkp* as a function of 1/T for aqueous 3.00M AMP solution 77 logk as a function of C0 2 loading for aqueous 3. 0OM AMP solution 78 Equilibrium partial pressure of CO2 presented as a function of 1000/T for eight different C0 2 loadings in aqueous solutions Of 4.28M MDEA 83 Equilibrium partial pressure of CO2 presented as a function of 1000/T for five different C0 2 loadings in a solution of TEG and 10mol% MEA 84 Comparison of the present model with present experimental data for a solution of TEG and 10mol% MEA at 30, 50, and 70 C, and predicted equilibrium curves for 100 and 150 C. 86 Comparison of equilibrium curves at 40 C for three different solvents, all containing 5mol% MEA 88 Comparison of equilibrium curves for the TEG/DEA system at different amine concentrations at 30 C 89 Comparison of equilibrium curves at 50 C for TEG solutions containing 10mol% MEA and 10mol% DEA 90 Present C0 2 solubility data in a mixed TEG/MEA solution compared with the solubility in NMP/MEA solutions at 50 C 91 C0 2 solubility data for 5mol% and 10mol% MEA in TEG, compared with the solubility in pure TEG 92 xiii

15 Chapter One Introduction 1.1 ACID GAS REMOVAL TECHNOLOGIES Acid gases such as carbon dioxide (C0 2 ), hydrogen sulfide (H 2 S), and sulfur dioxide (S0 2 ) are removed from a variety of gas streams, including natural gas, flue gas, synthesis gas, and refinery gases. Acid gas treating generally refers to removal of C0 2 and H 2 S, while the removal of S0 2 is often denoted r"lue gas desulfurization, although the technology used is often very similar. Removal of organic sulfur compounds such as carbonyl sulfide (COS), carbon disulfide (CS 2 ), mercaptans (RSH), thiophene, and other impurities present at low concentration levels (HCN, NH 3, S0 3 ), are often required as well. Kohl and Riesenfeld (1985) divides all gas purification processes into three categories: absorption into liquid, adsorption on a solid, and chemical conversion to another compound. In addition both cryogenic and membrane technology can be applied favorably in certain cases. Absorption into a liquid is the most used method (Astarita et al. (1983)), and is the method studied in this thesis. The liquid solution can consist of either a physical solvent, a chemical solvent (or a blend of chemical solvents) in water, or a mixed solvent containing both a chemical active and a nonaqueous physical solvent. The best method for a certain application is decided by parameters such as feed gas composition, pressure and quantity of gas treated, as well as the cleanup target. Since the process 1

16 to be chosen, will be the one that shows the best economics and the most reliable operation, it is important to have at hand design data for the processes. In the case of absorption processes, models describing the gas-liquid equilibria are important tools in the process design. The acid gas content in feed gases to treating units can range from less than 1% to well above 50%. The specification of acid gas in treated gas varies markedly from application to application. For example, according to Astarita et al. (1983), the pipeline specification for natural gas is maximum 4 ppm H 2 S and 1% C0 2. For natural gas to LNG plants the C0 2 content is usually limited to 50 ppm, and in ammonia manufacturing, the C0 2 impurity of the feed gas must be reduced to 10 ppm. As one can see, the range over which the feed gas compositions and the desired treated gas specifications varies, is quite large. The capability to remove acid gases at these different levels, depends highly on the process chosen. For example, the pure physical solvents are well suited for bulk C0 2 removal when the inlet partial pressure of C0 2 is relatively high, above approximately 7 atm according to Astarita et al. (1983), while at the same time the C0 2 specification in the treated gas is quite loose. If deep acid gas removal is required, the addition of an alkanolamine may help. Aqueous alkanolamine solutions are often used when the partial pressure of C0 2 in feed gas is relatively low and C0 2 removal down to ppm levels is required. The use of alkanolamines is discussed in some detail in the next section. Other useful chemical solvents for certain applications are aqueous solutions of salts of amino acids as well as promoted hot carbonate solutions. Membrane technology will have its potential for treating of high 2

17 pressure gases with high levels of acid gas (Funk and Li (1989)). For small acid gas removal units, savings in capital and operating costs might be expected. However, to minimize hydrocarbon losses, membranes with high C0 2 /CH 4 selectivity must be developed or complex recirculation schemes must be used. Work has also been done on a laboratory scale to use facilitated gel membranes containing amines to separate hydrocarbons and acid gases (Pellegrino et al. (1989) and Chakma (1992)). Pressure swing adsorption processes can be competitive with absorption in small plants (Astarita et al. (1983)). Adsorption are suited for trace removal of acid gases. 1.2 ALKANOLAMINE SOLUTIONS Alkanolamines are the most used chemical active agent in acid gas removal processes (Astarita et al. (1983)). Since Bottoms (1930) introduced the amines to "separate acidic gases", and recommended the tertiary triethanolamine (TEA) because of its higher boiling point, several new and more suited amines have become commercially available. Among the most important ones are monoethanolamine (MEA), diethanolamine (DEA), diisopropanolamine (DIPA), B, B'-hydroxy-aminoethylether (DGA, also known as diglycolamine), and methyldiethanolamine (MDEA). The tertiary amine MDEA has come to extensive use quite recently for a number of gas treating applications. MDEA is the major constituent in solvent processes offered by Dow Chemical Co. (Gas/Spec solvents), Union Carbide (Ucarsol solvents), Texaco Chemical Co. (Textreat solvents), and BASF (Activated MDEA). These are proprietary formulated solvents containing inhibitors, activators, and other additives. MDEA solutions exhibit large 3

18 acid gas capacity as well as easy regenerability. The use of corrosion inhibitors have made it possible to increase the amine concentration in the solutions markedly, and thereby reduce the solvent circulation rate, giving lower operating and capital costs. For example, aqueous MEA solutions can now be used in concentrations up to 5M, compared to typically 3M previously (Astarita et al. (1983)). According to Niswander et al. (1992), the last generation of MDEA based solutions have diminished their corrosiveness with a factor of 10, compared to the first generation of MDEA solvents. In recent years a new class of amines has been introduced: the sterically hindered amines. Sartori and Savage (1983) define a sterically hindered amine to be a primary amine where the amino group is attached to a tertiary carbon atom, or a secondary amine in which the amino group is attached to a secondary or a tertiary carbon atom. Examples are 2-amino-2-methyl-1-propanol (AMP), 2- (tert-butylamino) ethanol (TBE), and 2-piperidine ethanol (PE). Due to the bulky substituent attached to the amino group, a strong bonding of CO2 to the nitrogen atom is prevented, and the result is a low tendency to form carbamates. As we shall see in Chapter 6, an improved thermodynamic capacity exceeding 0.5 mol C0 2 /n\ol amine can be expected, at favorable absorption rates. Besides, the sterically hindered amines are well suited as promoters for the hot carbonate process (Say et al. (1984)). They are also suited for selective removal of H 2 S when CO2 is present, as an alternative to tertiary amines. By using blends of amines, one can make use of each amine' s attractive properties. For example, the large capacity and easy stripping of an MDEA solution can be combined with an MEA solution's ability to produce high purity sweet gas (Chakravarty 4

19 (1985)). This opens for an interesting possibility of tailormaking blended amine solutions to meet specific acid gas removal requirements. Evidently, there is a great need for more fundamental research into these "new" chemical solvents. Rochelle (1991) suggests that further studies should be undertaken to obtain solubility and kinetic data for both sterically hindered amines and MDEA based solutions mixed with primary and secondary amines. Alkanolamines are also used in nonaqueous solutions. Savings due to easier regeneration can be obtained. An example of this is the Sulfinol process using a mixture of DIPA, sulfolane (tetrahydrothiophene dioxide), and water. This process has shown capability of removing carbonyl sulfide (COS) and mercaptans (RSH) together with H 2 S and C0 2 (Kohl and Riesenfeld (1985)). Information given in Gas Process Handbook (1990) indicates that the Sulfinol process can deliver treated gas specified to 50 ppm C0 2, and thus be used prior to liquefaction in an LNG plant. Another example using a combined chemical and physical solvent is the Amisol process where methanol is mixed with MEA or DEA. Quite recently, Institut Francais du Petrole (IFP) has introduced a 2-stage process called IFPEXOL. Based on methanol as the major constituent, this process is capable of removing acid gases and water, giving hydrate protection and controlling the dew point (Minkkinen and Levier (1992)). The use of di- and triethyleneglycol together with alkanolamines, as studied in some detail in this thesis, was first described by Hutchinson (1939). Kohl and Riesenfeld (1985) discuss the advantages and the problems arising when such mixtures are used. Among the problems, the most important one was that the glycolamine system requires a high reboiler temperature, causing a 5

20 corrosive environment in the stripper and the heat exchanger. This eventually led to a decrease in the use of amine-glycol solutions for gas treating purposes. In recent years more resistant metals have been developed, and additives such as corrosion inhibitors have become available. This can lead to a renaissance for such processes, when it is important to reduce the number of process units to save either space or weight. Such solvents is capable of removing water and CO2 in one step. Savings should be obtained in cases where removal of both these components is necessary. One advantage using the glycol-amine process is that the steam consumption can be lowered compared to aqueous systems. In addition, these solutions will have the capability to reduce the C0 2 content in the gas down to extremely low levels, because the CO2 is more readily stripped from the solution. Vaporization losses and degradation problems may occur because of the high temperatures. According to McCartney (1948), this can be reduced by introducing a glycol wash after the glycol-amine absorber. To minimize degradation, amines other than MEA could be used. Secondary amines such as DEA look promising, while tertiary and sterically hindered amines will face problems in nonaqueous solvents due to their resistance towards the formation of carbamates. Versteeg (1986) has shown that tertiary amines (MDEA) do not show significant effect on C0 2 solubility in nonaqueous solutions. Hydrocarbons are in general more soluble in nonaqueous physical solvents than in aqueous solvents. 6

21 1.3 SCOPE OF THE WORK This thesis deals with the problems related to acid gas treating in general, and specifically to CO2 removal using alkanolamines. Most emphasis has been put on developing simple and reliable modelling procedures for vapor-liquid equilibria of aqueous amine solutions. The modelling technique presented here has been tested with the tertiary amine MDEA and the sterically hindered amine AMP. The model predicts equilibrium partial pressures of C0 2 in good agreement with experimental values. Experimental equilibrium and ph data are presented, and the model is based on these data and solubility data from the literature. The objective was to develop a model which could be used at both absorption and desorption conditions. In the case of MDEA we have succeeded in covering the temperature interval from 25 to 140 C. C0 2 loadings between and 1 mol C0 2 /mol amine, and C0 2 partial pressures between and 50 atm, are correlated. The model is tested against present experimental data and data published previously by several investigators, and found to be accurate for the following amine molarities: 1.69, 2.00, 3.04, 4.00, and 4.28M. In the case of AMP the application range of the model is restricted to absorption conditions. This work also includes equilibrium measurements for nonaqueous alkanolamine solutions at absorption temperatures. The following systems are investigated: TEG/MEA, TEG/DEA, and DEG/MEA. The measurements are undertaken to compare the C0 2 solubility in these mixed nonaqueous solvents with the solubility in aqueous amine solutions and other solvents containing amines, reported in the literature. The influence of temperature and amine concentration on the C0 2 solubility is also investigated. 7

22 The TEG/MEA-system with 0.79M MEA (10mol%) is modelled to enable estimation of C0 2 partial pressures at elevated temperatures, well outside the range where the measurements were undertaken. Some screening measurements to determine absorption rates of C0 2 into five different solvents including water, are also reported. The same solvents, with addition of MEA, are also investigated with respect to kinetics. A string-of-discs column was used for these experiments. The work described in this thesis is in many ways the first comprehensive treatment of acid gas removal processes done in our laboratory. Previous studies have mostly been related to S0 2 absorption. As a result of preliminary experiments undertaken in the start-up phase of this study, we realized that both a new experimental set-up for high temperature solubility measurements, as well as an improved apparatus for kinetic determinations, were desirable. These two new experimental set-ups were built in 1991 and are now in use. Having these apparatuses at our hands, we can conduct experimental research in most areas related to gas treating technology. The new experimental facilities are described in some detail in Chapter 3. 8

23 Chapter Two Literature Review 2.1 VAPOR-I.IOUID EQUILIBRIUM DATA IN GAS TREATING PROCESSES VXE MEASUREMENTS IN AQUEOUS ALKANOLAMINE SOLUTIONS A number of investigators have presented vapor-liquid equilibrium data on aqueous C02-alkanolamine systems. Some examples are: - For MEA systems, contributions are made by Mason and Dodge (1936), Leibush and Shneerson (1950), Muhlbauer and Monaghan (1957), Jones et al. (1959), and Lee et al. (1975, 1976). - DEA systems are investigated by Bottoms (1931), Mason and Dodge (1936), Reed and Wood (1941), Murzin and Leites (1971), Lee et al. (1972, 1974), Lawson and Garst (1976), Kennard and Meisen (1984), and Lai et al. (1985). - For TEA systems measurements are presented by Bottoms (1931), Mason and Dodge (1936), Byudkovskaya and Leibush (1949), and Jou et al. (1985). - VLE data for aqueous DIPA are given by Isaacs et al. (1977). More recently, equilibrium data for the tertiary amine MDEA and also for sterically hindered amines have become available. Equilibrium data for the MDEA system is given by Jou et al. (1982, 1986), Bhairi (1984), Chakma and Meisen (1987), Austgen (1989), and Lidal and Erga (1991). Sharma (1965) observed that 9

24 sterical hindrance has a pronounced effect on the stability of the carbamates, see section 2.2 and 6.1. The sterically hindered amines were later introduced to acid gas treating by Exxon (Chem. Eng. News (1981)). A few investigators have reported equilibrium data in some hindered amines. Measurements on one of the best known hindered amines, AMP, have been undertaken by Sartori and Savage (1983), Komorowicz and Erga (1987), Roberts and Mather (1988a), Teng and Mather (1989, 1990), Erga and Lidal (1990), and Tontiwachwuthikul et al. (1991). In a research report from the Gas Processors Association, equilibrium solubility of CO2 in aqueous solutions of MEA, DGA, DEA, and MDEA are given (Maddox et al. (1987)). Equilibrium data for DGA are also presented by Martin et al. (1978) and Dingman et al. (1983) VLE MODELLING FOR C0 2 IN AQUEOUS ALKANOLAMINE SOLUTIONS Mason and Dodge (1936) made the first attempt to correlate the equilibrium solubility data for C0 2 in alkanolamines. Since the reactions between amines and CO2 had not been properly investigated at that time, they were limited to use curv*. itting methods. A method for predicting C02/amine equilibria in aqueous solutions based on the use of apparent equilibrium constants, without activity coefficients, was described by Danckwerts and McNeil (1967). They used the same approach as Van Krevelen (1949) had developed for aqueous solutions of ammonia and C0 2. The apparent equilibrium constants, and their dependence on the ionic strength of the solutions, are used to describe the chemical equilibria. A similar approach was used by Kent and Eisenberg (1976) for MEA 10

25 and DEA solutions/ the main difference being that the apparent equilibrium constants were regarded as constant, irrespective of the Ionic strength. In the Kent-Eisenberg method, the approach made by Danckwerts and McNeil (1967) was modified by forcing the apparent equilibrium constants to fit published equilibrium data as a function of the temperature. An early attempt to include activity coefficients into a predictive model for C0 2 /amine equilibria was made by Klyamer and Kolesnikova (1972), and was further developed to describe the C02/H 2 S/amine equilibria by Klyamer et al. (1973). They used a method proposed for the H 2 S/amine system by Atwood et al. (1957), where the activity coefficients of all ionic species are assumed to be equal. According to Deshmukh and Mather (1981), the generalized model given by Klyamer et al. is algebraically equivalent to the Kent-Eisenberg model if the activity coefficients are set equal to unity. These earlier models exhibit a useful description of the chemical equilibria for many compositions of the aqueous amine solutions. However, they often fail at compositions outside the range where the apparent equilibrium constants are fitted, or the activity coefficients are determined. To be able to broaden the range where such models could be applied, one has to use equilibrium constants expressed as functions of C0 2 concentration, amine molarity, and temperature. Realizing that the Kent-Eisenberg model has certain limitations, improvements of the method have been achieved by several investigators over the years, such as Chakma and Meisen (1990) for the C0 2 /DEA/water system. The most important improvement is that the apparent equilibrium constant of the main DEA-C0 2 reaction is recorrelated using a more comprehensive set of 11

26 experimental data. In the new correlation the apparent equilibrium constant is expressed as a function not only of the temperature» but also of the CO2 concentration and the amine molarity- Jou et al. (1982) used a similar procedure to correlate their VI.E data for the MDEA system. A "new generation" of equilibrium models has been developed in recent y&ars. A. thermodynamic framework was established by Edwards et al. (1975, 1978) to calculate gas-liquid equilibria in aqueous solutions containing one or more volatile weak electrolytes, such as C0 2. The framework was so constructed that the equilibrium compositions of multisolute systems could be predicted using only binary interaction parameters. Beutier and Renon (1978) used a similar approach, in which two ternary interaction parameters were fitted to the actual ternary experimental data. In this way, a better agreement between calculated and expsrimental data for a two-solute system, was obtained. Deshmukh and Mather (1981) proposed a mathematical model based on the extended Debye-Hiickel theory of electrolyte solutions, using the Guggenheim (1935) equation, which represents the activity coefficients by the use of two terms. The first of these terms is the standard Debye-Huckel term representing the electrostatic forces. The second term includes binary interaction parameters accounting for short-range Van der Waals forces. Because many of these interaction parameters were unavailable, Deshmukh and Mather adjusted some of them to ternary VLE data (CO2/MEA/H2C) and H 2 S/MEA/H 2 0). In this, they made use of the assumption that the interaction parameters for the species which were present in very small concentrations could be neglected. The fugacity coefficients were calculated using the Peng-Robinson (1976) equation of state. The model exhibits a good fit to the 12

27 experimental data for the MEA system except for high C0 2 loadings. This especially applies to the quaternary system; C02/H 2 S/MEA/H20, where the assumptions made seem to lead to an underprediction of the equilibrium partial pressures. Chakravarty (1985) in his work used a similar approach. By extensive use of literature data an equilibrium model was developed, applicable to four single amine systems (MEA, DEA, DIPA, and MDEA) as well as blends of amines (MEA/MDEA and DEA/MDEA). Based on a generalized excess Gibbs energy model that treats both long-range electrostatic interactions between ions, and shortrange interactions between all liquid phase species, Austgen (1989) has developed a thermodynamically consistent model describing the vapor-liquid equilibria in acid gas-amine-water systems. The vapor phase fugacity coefficients were calculated by the use of the Redlich-Kwong-Soave equation of state (Soave (1972)). The Electrolyte-NRTL equation (Chen and Evans (1986)) was used to represent the liquid phase activity coefficients. The Electrolyte-NRTL equation requires binary interaction parameters to be estimated from experimental data. In addition the carbamate stability constant was treated as an adjustable parameter within the VLE model. The model was extended to describe C0 2 solubilities in blends of amines (MEA/MDEA and DEA/MDEA). An attempt to correlate the CC^/AMP/water system was made by Chakraborty et al. (1986) based on equilibrium constants at vanishingly small ionic strength. As would be expected, the model could not describe the equilibrium curve very well at high C0 2 loadings. Tontiwachwuthikul et al. (1991) proposed a modified Kent-Eisenberg model for the same system, and obtained a better agreement between calculated and experimental data. VLE models for aqueous DGA solutions have been developed by 13

28 Dingman et al. (1983) and Hu and Chakma (1990). While Hu and Chakma based their method on similar principles as Kent and Eisenberg (1976), Dingman et al. took a more fundamental approach, by using the framework introduced by Edwards et al. (1975), and thereby included activity coefficients in the description of the vapor-liquid equilibria. In this review of the literature, no reports were found regarding the use of measured ph data in the modelling of VLE in alkanolamine systems VLE MEASUREMENTS IN MIXED NONAQUEOUS SOLVENTS Parts of this thesis concern the absorption of C0 2 into a mixed solution, containing an alkanolamine and a glycol solvent, with virtually no water present. Since Hutchinson (1939) proposed the glycol-amine process for simultaneous acid gas removal and dehydration, few investigations on the C0 2 solubility in these systems have been reported. Most of the literature published deals with the technical specifications of the process, or with the limitations and problems related to the process. Examples are Chapin (1947), Kohl and Blohm (1950), Polderman et al. (1955), and Holder (1966). Literature data are scarce for all systems combining amines and nonaqueous physical solvents. Murrieta-Guevara and Tre jo Rodriguez (1984) presented solubility data for C0 2, H 2 S, and methane in nonaqueous mixtures of alkanolamines (MEA, DEA) and physical solvents (n-methyl-pyrrolidone (NMP), propylene carbonate (PC)). Murrieta-Guevara et al. (1992) introduced some additional data for the solubility of C0 2 in NMP solutions containing either MEA or DEA. Solubilities of C0 2, H 2 S, and 14

29 ethane in PC, NMP, and sulfolane (tetrahydrothiophene dioxide) in mixtures with alkanolamines, were measured by Rivas and Prausnitz (1979). Dimov et al. (1976) measured low pressure VLE data for MEA solutions of ethyleneglycol, NMP and tetrahydrofurfuryl alcohol at different water levels (also without water). Leites et al. (1972) compared the C0 2 solubility between several nonaqueous solvents containing MEA. Takeshita and Kitamoto (1988) measured the C0 2 solubility in complete water free solutions of methanol/ octane and triethylamine with different primary and secondary amines. Woertz (1972) investigated a number of mixtures containing an amine, a physical solvent, and a small amount of water (3 or 10vol%). In the literature one can find VLE data for aqueous systems containing both an amine and a physical solvent, an example being the data of Roberts and Mather (1988b), where the solubility of acid gases in a mixed solvent of 16.5wt% AMP, 32.2wt% sulfolane, and 51.3wt% water was reported. Oyevaar et al. (1989) measured the C0 2 solubility in aqueous ethyleneglycol solutions containing DEA VLE MEASUREMENTS IN PORE PHYSICAL SOLVENTS TEG-CO2 equilibria without amine present were measured by Takahashi et al. (1984) and Jou et al. (1987). Takahashi et al. also presented solubility data for the DEG-C0 2 system. C0 2 solubilities in other useful physical solvents are reported by a series of investigators. Some examples are: Isaacs et al. (1977), Laddha et al. (1981), Sweeney (1984, 1988), Roberts and Mather (1988c), Murrieta-Guevara et al. (1988), Jou et al. (1990a, 1990b), and Yogish (1991). Fogg and Gerrard (1990) have collected published C0 2 solubility data for more than

30 different solvents VLE MODELLING TECHNIQUES IN PHYSICAL SOLVENTS The solubility of acid gas in a pure physical solvent can be described by Henry's law (Eqn. (5.7)). However, at higher concentrations and partial pressures most systems show a considerable deviation from the linearity assumed in the simple form of Henry's law (Fogg and Gerrard (1990) and Carroll (1991)). Thus, in order to successfully correlate experimental results up to high concentrations, one needs a method based on an equation of state valid for the solvent and dilute solutions of the solute in the solvent. Such an approach was used by Jou et al. (1987, 1990a) to correlate the solubility of C0 2 and H 2 S and the lower alkanes in solutions of TEG and sulfolane. They used the Peng- Robinson (1976) equation of state, and obtained interaction parameters for these systems. These interaction parameters were further used to determine the three parameters in the equation developed by Krichevsky and Iliinskaya (1945). The Krichevsky- Iliinskaya equation has been shown to be applicable also for mixtures of components with strong intermolecular interactions. For such systems simple equations of state are insufficient for description of the phase behaviour (Jou et al. (1987)). For the mixed solvents described in section 2.1.3, correlations for the C0 2 solubility are given by Rivas and Prausnitz (1979) and Roberts and Mather (1988b). Rivas and Prausnitz determined equilibrium constants to describe the chemical equilibria for the absorbed gas and the chemical solvent. Roberts and Mather used the solubility model of Deshmukh and Mather (1981) to predict the equilibrium partial pressures of C0 2 in a mixture of a chemical active agent (AMP), a physical solvent (sulfolane), and water. 16

31 2.2 CHEMISTRY OF COo - AMINE SYSTEMS INTRODUCTION In this work, emphasis has been put into the developing of simple and reliable methods for representing the equilibria of C0 2 -amine systems. This cannot be done without having an understanding of the chemistry encountered in these systems. Several comprehensive investigations have been undertaken to study the kinetics of the reactions in alkanolamine processes. As a result, rate-based process models for acid gas removal are developed, see for example Glasscock (1990) and Carey (1990). Carey gives an overview of rate-based models available. Tomcej (1987) developed a nonequilibrium stage model to simulate acid gas absorption into alkanolamine solutions. This model has become commercially available as a simulation program under the name of AMSIM. Other models for commercial use are the TSWEET program. According to informations given at the last GPA convention in 1992, TSWEET will soon offer the capability of simulating systems using blends of amines (Bullin et al. (1992)). A simulation program described by Sardar and Weiland (1985) is also commercially available. Several of the larger companies such as DOW Chemical Co. are known to have in-house amine process simulators (Katti and Langfitt (1985)). Based on a mass transfer model described in the literature (Versteeg et al. (1989, 1990)), researchers at Twente University have developed a simulation program called SIMULTER for the calculation of the absorption rates of Co 2 and H 2 S into aqueous solutions of tertiary amines. Versteeg (1986) in his work studied the reaction between C0 2 and different alkanolamines both in aqueous and nonaqueous solutions. Khalil (1984), Yu (1985) and Al-Ghawas (1988) studied the 17

32 kinetics of absorption of H 2 S and C0 2 in aqueous MDEA solutions. Al-Ghawas (1988) and Glasscock and Rochelle (1989) recapitulate the different mass transfer models presented in the literature. Such models are the film theory model, still surface models, surface renewal models, penetration models, and combinations of these models. This theory will not be taken any further in this thesis. Complete understanding of the mechanism of the reaction of C0 2 with alkanolamines is still ahead of us. For well investigated systems, however, kinetic expressions which are in good agreement with experimental data, are established. In the following sections, the basic C0 2 - amine chemistry, and the kinetics proposed in the literature for MDEA and AMP systems, are presented. The last section in this chapter presents different experimental techniques for determinations of reaction kinetics REACTIONS BETWEEN C0 2 AND AMINES IN AQUEOUS SOLUTIONS Compared with the instantaneous proton transfer reaction when H 2 S reacts with an alkanolamine, the reaction between C0 2 and alkanolamines is more complex, and the reaction rate depend highly on the structure of the alkanolamine molecule. Primary and secondary amines, have the capability to react with C0 2, forming carbamate ions. These are amines with one or two carboncontaining groups attached to the nitrogen atom. Tertiary amines, like TEA and MDEA, with three carbon-containing groups attached to the nitrogen atom, cannot form carbamates, and bicarbonate formation becomes the only main reaction. 18

33 H I H - N - C 2 H 4 OH Monoethanolamine (MEA) I C 2 H 4 OH H - N - C 2 H 4 OH Diethanolamine (DEA) C 2 H 4 OH I C 2 H 4 OH - N - C 2 H 4 OH C 2 H 4 OH CH 3 - N - C 2 H 4 OH TriethanolaminetTEA) Methyldiethanolamine(MDEA) CH 3 HO - CH 2 - C - NH2 I CH amino methyl - I - propanol (AMP) Figure 2.1 Molecular structure of amines used in acid gas removal processes 19

34 Figure 2.1 shows the molecular structure of MEA, DEA, and TEA, as well as the two amines especially studied in this investigation, MDEA and AMP. AMP is denoted a sterically hindered amine (Sartori and Savage (1983)), since the amino group is attached to a tertiary carbon atom. For definition of a sterically hindered amine, see Chapter 1. An important reaction in aqueous solutions containing C0 2 is the "OH"-reaction": C0 2 + OH" = HC0 3 " (2.1) At ph values above 8, the most important reaction mechanism of this reaction is the direct one, where Eqn. (2.1) is the actual kinetic step (Astarita et al. (1983)). At lower ph values/ a competing mechanism occurs. In this C0 2 is first hydrated: C0 2 + H 2 0 = H 2 C0 3 (2.2) Reactxon (2.2) is then followed by the dissociation of the carbonic acid: H 2 C0 3 = HC0 3 " + H + (2.3) For reactions involving amines at sufficiently high ph-vajues Astarita et al. (1983) suggest that in general three main reactions should be considered. Taking a primary amine (RNH 2 ) as an example: Carbamate Formation,CF: C RNH 2 = RNH RNHCOCT (2.4) Bicarbonate Formation,BF: C0 2 + RNH 2 + H 2 0 = RNH HC0 3 " (2.5) Carbamate Reversion,CR: RNHCOO" + H 2 0 = RNH 2 + HC0 3 " (2.6) 20

35 We now introduce the C0 2 loading, y, expressed as mol C0 2 /mol amine. For primary and secondary amines, CF will take place at y<0.5, CR at y>0.5, and BF at all values of y. For tertiary amines, CF does not take place, and BF is the only reaction. For hindered amines, CF may be very small or negligible. Al-Ghawas (1988) in his work also includes the direct formation of carbonic acid by the reaction of C0 2 and H 2 0 (Eqns. ( )), and also an alkylcarbonate formation reaction. However, according to Astarita et al. (1983) and Yu et al. (1985), both of these reactions will proceed to a negligible extent at the phvalues and temperatures usually encountered in gas treating processes. According to Danckwerts (1979) and Astarita et al. (1983), with later support also by other investigators, the CF mechanism is believed to proceed by the steps: C0 2 + RNH 2 = RN + H 2 COO" (2.7) RN + H 2 COO" + RNH 2 = RNH RNHCOO" (2.8) This zwitterion mechanism was first proposed by Caplow (1968) for amines without alcoholic groups. The rate-determining step in this mechanism is believed to be the zwitterion formation (Egn. (2.7)). This is verified for the MEA system, where a rate equation as follows has been verified: r = k CF C c02 C RNH2 (2.9) For some of the other amines, such as DEA, there are data supporting Eqn. (2.9), while other data suggest the reaction to be second-order with respect to the amine (Hikita et al. (1977)). 21

36 Versteeg and van Swaaij (1988a) explain how the same reaction, for different amines, can assume different reaction orders. Among the first to investigate the reactions between C0 2 and amines were Danish researchers. They studied the carbamate formation from a number of amines, such as dimethylamine (Faurholt (1925)) and glycine (Jensen et al. (1954)). The reactions between C0 2 and alkanolamines were also studied. The rate of reaction of C0 2 with both MEA and DEA (Ballund Jensen et al. (1954)), as well as TEA (Jørgensen and Faurholt (1954)), was measured. Their work is also commented on in the next section. A possible mechanism of the CF reaction is described in the next section REACTION KINETICS BETWEEN C0 2 AND AQUEOUS MDEA MDEA is today the most used tertiary amine for acid gas removal. MDEA has outdone for example TEA, which was the first amine to be used for gas sweetening purposes (Bottoms (1930)). When the correct additives are used, MDEA offers several advantages over other amines also for bulk C0 2 removal (Bullin et al. (1990, 1992)). This is discussed elsewhere in this thesis (Chapter 5). A number of investigations have been conducted on the kinetics of the MDEA-C0 2 system in recent years. Examples are Barth et al. ( ), Haimour and Sandall (1984), Yu et al. (1985), Versteeg (1986), Haimour et al. (1987), Tomcej and Otto (1989), Crooks and Donnellan (1990), and Al-Ghawas and Sandall (1991). There are some controversies in the literature about the reaction rate of C0 2 with MDEA. Glasscock (1990) suggests that the discrepancies found in the literature is due to the fact that the 22