£5 0 ., r. . . . mi». , a: 5.43 . . ‘2 53." , . L a. f . . v. r Lawlw . g , . r . . . for 1» , $3.... a: :Z .933... it 5‘ t [4 t- : flak? .\. . .ruwaaf... x. . i. I ( t: iv... a": 1.325.. 2M..V$y )3 49.2... a; . n in. u :»..u a. . h? ~ .2. : a . J Jud“ » t x .3. .9 3.3.4.1.: is. ‘ it»; s . 5.5.2. . 5... L. .7 I. c stu- ‘\ ~.Vs«..vstp. .9. A i. ‘ &£.rcmuqn t 5.: x}. ‘ .L. , t, ~ ~ THESI‘Q 2 2 OCT)\ This is to certify that the dissertation entitled Aqueous-Phase Hydrogenation of Biomass Derived Lactic Acid to Propylene Glycol presented by Zhigang Zhang has been accepted towards fulfillment of the requirements for Ph.D. degree in Chemical Engineering [:2 ‘ ermine Major flofessor Date 8/37/00 MS U i: an Affirmative Action/Equal Opportunity Institution 042771 LIBRARY Mlchlgan State University PLACE IN REI'URN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE our DATE DUE DATE DUE APR (‘2 33 "23023 ,"l C "x fa“? U\.~h.. M' g. M 11 20 AUG 1 12009 woo WM.“ AQUEOUS-PHASE HYDROGENATION OF BIOMASS DERIVED LACTIC ACID TO PROPYLENE GLYCOL Zhigang Zhang A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Chemical Engineering College of Engineering July 19, 2000 0‘? :“:~ «1“ : Litjr ABSTRACT AQUEOUS-PHASE HYDROGENATION OF BIOMASS DERIVED LACTIC ACID TO PROPYLENE GLYCOL BY ZHIGAN G ZHAN G Aqueous phase hydrogenation of biomass-derived lactic acid (L-I- 2-hydroxy- propionic acid) to propylene glycol (PG) has been performed using a stirred batch reactor and a continuous trickle bed reactor. In the optimal reaction conditions and catalysts, over 90% PG selectivity with 95% lactic acid conversion can be achieved in both batch and trickle bed reactors. The major side reactions are the formation hydrocarbon (methane, ethane, and propane). When reaction temperature is lower than 170°C, PG is the only liquid product, which makes the product separation very simple. The best active metal is ruthenium and the best catalyst supports are selected activated carbons. In the stined batch reactor, optimal selectivity and high reaction rate are reached at a temperature of 150°C and high pressure of hydrogen (1500~2000psi). In the trickle bed reactor, the reaction temperature can be as low as 80°C with a pressure as low as 800psi without significant sacrifice of PG selectivity. The reaction temperature and pressure used in this process are very mild compared to carboxylic acid hydrogenation reported in literature. In the stirred batch reactor, the measured gas-liquid mass transfer coefficient and theoretical mass transfer analysis have shown that gas-liquid, liquid-solid, and intra- particle mass transfer are negligible at our reaction conditions. The intrinsic kinetics have been analyzed and the activation energy for lactic acid hydrogenation is 96kJ/mole. Lactic acid consumption rate is sensitive to reaction temperature and catalyst loading but insensitive to hydrogen pressure. The performance of laboratory prepared carbon- supported ruthenium catalysts are as good as commercial catalysts in the batch reactor. Granular carbon supported ruthenium catalysts were prepared and used in the trickle bed reactor to continuously hydrogenate lactic acid to PG. Experiments and calculation have verified that gas-liquid, liquid-solid and intra-particle mass transfers and surface chemical reaction together control the lactic acid hydrogenation reaction in the trickle bed reactor. Gas-liquid mass transfer is the major resistance. Lactic acid conversion increases with temperature at the same pressure and fixed hydrogen to lactic acid molar ratio. Like the reaction in the batch reactor, propylene glycol selectivity increases with hydrogen pressure. The overall activation energy in the trickle bed reactor is only 48kJ/mole, which indicates that mass transfer controls this hydrogenation reaction. Hydrogen solubility at high pressure and trickle bed dynamic liquid holdup, which are two key parameters in the trickle bed reactor modeling, were measured in the trickle bed reactor. A one-dimensional trickle bed reactor model was derived. This model consists of two differential and two algebraic equations and forms a typical two-point boundary value problem mathematically. With simplification, the model was solved in Mathematica. Liquid phase hydrogen and lactic acid concentration and catalyst surface hydrogen and lactic acid concentration profiles were plotted. The results give us more information about the role of gas-liquid and liquid-solid mass transfer. v! ‘ . Ltiitli ACKNOWLEDGEMENTS I am very grateful for the suggestions and express guidance of Dr. Dennis J Miller and Dr. James Jackson throughout the development of this work. Thanks to Dr. Kris Berglund, Dr. Ramani Narayan and Dr. Smith, Milton R III for their Service as members of my doctoral committee. Thanks to Bryan P Hogle for performing catalyst characterization. Thanks to Dushyant Shekhawat, Frank Jere, and Shubham Chopade for excellent suggestions and collaboration. Thanks to the Chemical Engineering Department at Michigan State University, Cargill Corporation and the consortium plant Biotechnology Research for their financial support of this work. iv Dedicated to peOple who help me in the past and the future! These people include: §§fi%fifi§flfifi,fififlfififi$fi%,mfififi%% i!fi%%¥$%fi,fifififiKTfifififififlmfifififi fit!$fifl§fiMfififik§fl%fiA$fi§fiWfififl% fi!fififlfififfifiklm,fiTWfi%%%fifififlifl klfioéfifiE9ffififiififi!$flfl§%Mfififik $fl%fifikfi%§fi£fi%filfififlfiTfififififiofl fiTfififiWfi%.fiEfl%%M,fiTWflfl¥fiogfifi E9ffififiifiil@flfiMfifiEkI%%*flfii,fi emaeameetmem,maeeaexeiag!ea fifififl.fifififi$%§l,Efimfifig$fiififil $A%EIE+%$W$E T¢Eflt€fiim%%fi fifi.%$¥fflfikl§ fi.%%§%klfil¢3 e.mea¢aeeama maeemaaammae £fi%%$T%lEfiifl EE!W%E%EIW§E $.fififiEWfiMfikg klfifififiiiTfikl fii.fi$$%fiklfl% filfififlfiififlfififi i. USTO USTC TABLE OF CONTENTS LIST OF TABLES ......................................................................................................... xiv LIST OF FIGURES ....................................................................................................... xvii ABBRIVIATIONS ......................................................................................................... xxi NOTATION ................................................................................................................. xxiii CHAPTER 1. BACKGROUND ....................................................................................... l 1.1. III'ERATURE REVIEW ............................................................................................... 2 1.1.1.1.actic acid ......................................................................................................... 4 1.1.1.1. History of discovery ................................................................................... 4 1.1.1.2. Production of lactic acid ............................................................................. 5 1.1.1.3. Synthetic technology .................................................................................. 7 1.1.1.4. The production of lactic acid from fermentation ........................................ 9 1.1.1.5. Lactic acid market .................................................................................... 10 1.1.2. Propylene glycol ............................................................................................. 12 1.2. LACTIC ACID REACTIONS ........................................................................................ 13 1.2.1. Lactic acid reduction ....................................................................................... 14 1.2.2. Polymerization ................................................................................................ 17 1.2.3. Dehydration .................................................................................................... 17 1.2.4. Condensation to 2,3-pentanedione ................................................................. 18 1.3. HYDROGENATING CARBOXYLIC ACIDS wrrH MOLECULAR HYDROGEN ................... 19 1.3.1. Heterogeneous catalysts .................................................................................. 23 vi [\J 5. n J J 1.3.2. Homogeneous hydrogenation catalysts ........................................................... 24 1.3.3. Hydrogenation of carboxylic acid salts ........................................................... 26 1.3.4. Hydrogen solubility ........................................................................................ 27 1.3.5. Hydrogenation mechanism ............................................................................. 27 1.4. TRICKLE BED REACTOR (TBR) AND MODELING ...................................................... 28 1.5. RATIONALE OF THIS RESEARCH .............................................................................. 31 CHAPTER 2. EQUIPMENT AND EXPERIMENTAL METHODS ............................. 33 2.1. REAGENTS .............................................................................................................. 33 2.2. CATALYSTS ............................................................................................................ 34 2.2.1. Commercial catalyst samples ......................................................................... 34 2.2.2. Catalyst preparation ........................................................................................ 35 2.2.2.1. Impregnation ............................................................................................ 35 2.2.2.2. Drying ....................................................................................................... 36 2.2.2.3. Reduction ................................................................................................. 36 2.2.3. Powder catalyst ............................................................................................... 37 2.2.4. Granular catalyst ............................................................................................. 38 2.3. BATCH REACTOR (AUTOCLAVE) ............................................................................. 38 2.3.1. Reactor system ................................................................................................ 39 2.3.2. Operating procedure for the batch rector ........................................................ 40 2.4. CONTINUOUS TRICKLE BED REACHON SYSTEM ....................................................... 42 2.4.1. Specifications of the trickle bed system ......................................................... 43 2.4.2. Operating procedure for the trickle bed reactor .............................................. 43 vii IiPRE 2.5.1. 2.5. PRODUCTS ANALYSIS ............................................................................................. 45 2.5.1. Mass spectrometer .......................................................................................... 45 2.5.1.1. Gas phase by-products identification ....................................................... 46 2.5.1.2. Quantitative analysis ................................................................................ 51 2.5.2. High performance liquid chromatography (I-IPLC) ........................................ 53 2.5.2.1. Calibration of HPLC ................................................................................ 55 2.5.2.2. Liquid compound peak assignment .......................................................... 57 2.5.3. Catalyst characterization ................................................................................. 59 CHAPTER 3. LACTIC ACID HYDROGENATION IN AUTOCLAVE ...................... 61 3.1. COMMERCIAL CATALYST TESTING .......................................................................... 61 3.1.1. Catalyst screening ........................................................................................... 61 3.1.2. Temperature and pressure optimization (Matrix -1) ...................................... 63 3.1.3. Catalyst loading effects (Matrices 2 and 3) .................................................... 66 3.1.4. Catalyst pie-reduction effect ........................................................................... 69 3.1.5. Lactic acid concentration effects .................................................................... 70 3.1.6. Gas product evolution ..................................................................................... 71 3.2. LACTIC ACID CONVERSION OVER LABORATORY PREPARED CATALYSTS .................. 74 3.3. LACTATE SALT HYDROGENATION ........................................................................... 77 3.3.1. Potassium lactate hydrogenation .................................................................... 78 3.3.2. Calcium ion .................................................................................................... 80 3.4. CARBON BALANCE ................................................................................................. 82 3.5. CONCLUSION .......................................................................................................... 84 viii CHINE idem CHAPTER 4. CONVERSION OF LACTIC ACID TO PROPYLENE GLYCOL IN A TRICKLE BED REACTOR ........................................................................................... 86 4.1. CONTROL PARAMETERS AND CATALYSTS IN TRICKLE BED REACTOR ...................... 86 4.1.1. Control parameters ......................................................................................... 87 4.1.1.1. Liquid superficial velocity ........................................................................ 87 4.1.1.2. Gas superficial velocity ............................................................................ 87 4.1.1.3. Flash vaporization .................................................................................... 88 4.1.1.4. Conditions of operation ............................................................................ 89 4.1.2. Catalysts .......................................................................................................... 89 4.1.2.1. Catalyst characterization .......................................................................... 89 4.1.2.2. Initial autoclave test of granular catalysts ................................................ 91 4.2. TRICKLE BED REACTION (RACEMIC LACTIC ACID) ................................................... 92 4.2.1. Results ............................................................................................................ 93 4.2.1.1. Charges 1 and 2 ........................................................................................ 93 4.2.1.2. Charge 3 ................................................................................................... 94 4.2.2. Temperature effect .......................................................................................... 95 4.2.3. Pressure effect (Charge 3, fully filled CG6M) ............................................... 97 4.2.4. Hydrogen/lactic acid molar ratio .................................................................... 99 4.2.5. The effect of changing liquid flow-rate ........................................................ 100 4.2.6. Long time and low concentration lactic acid hydrogenation ........................ 102 4.2.7. High lactic acid concentration feed .............................................................. 102 4.2.8. Addition of sulfur to lactic acid feed ............................................................ 104 4.3. CONVERSION OF UNREFINED LACTIC ACID To PG IN TRICKLB BED REACTOR ........ 106 ix 4.3.1. Reaction conditions ...................................................................................... 107 4.3.2. Results and discussion .................................................................................. 107 4.3.2.1. Hydrogenation at different Pressure and Temperature ........................... 107 4.3.2.2. Catalyst deactivation from impurities in Cargill lactic acid ................... 109 4.3.2.3. The effect of lactic acid sources ............................................................. 111 4.3.3. Summary ....................................................................................................... 113 4.4. CONCLUSION ........................................................................................................ 113 CHAPTER 5. MASS TRANSFER, KINETICS AND MODELING ........................... 115 5.1. H2 SOLUBILITY MEASUREMENT ............................................................................ 115 5.1.1. Apparatus ...................................................................................................... 116 5.1.2. Experimental steps ........................................................................................ 116 5.1.3. Calculation and results .................................................................................. 117 5.1.4. Solubility of hydrogen in 10% lactic acid .................................................... 118 5.2. CHARACTERIZATION OF MASS TRANSFER IN THE BATCH REACTOR (AUTOCLAVE). 1 19 5.2.1. Suspension of catalyst ................................................................................... 119 5.2.2. Maximum reaction rate and pseudo first order rate constant in batch reactor120 5.2.3. Gas-liquid mass transfer ............................................................................... 121 5.2.3.1. Principle and procedure .......................................................................... 122 5.2.3.2. Data analysis ........................................................................................... 122 5.2.3.3. Results and comparison with literature .................................................. 123 5.2.3.4. Comparing G-L mass transfer with reaction rate ................................... 125 5.2.3.5. Comparing with literature correlation .................................................... 126 5.2.3.6. Verification by investigating the stirring speed effect ............................ 127 5.2.4. Liquid-solid mass transfer ............................................................................ 128 5.2.5. Inna-particle mass transfer ........................................................................... 130 5.2.6. Summary of mass transfer in the batch reactor ............................................. 131 5.2.7. Batch reactor macro kinetics ........................................................................ 131 5.2.7.1. Initial reaction rate..... ............................................................................. 132 5.2.7.2. Activation energy ................................................................................... 134 5.2.8. Kinetics model .............................................................................................. 136 5.2.8.1. Model derivation .................................................................................... 136 5.2.8.2. Fitting the data of the reactions at 130°C ............................................... 138 5.3. CONTINUOUS REACTOR (TRICKLE BED) ................................................................ 140 5.3.1. Dynamic Liquid holdup ................................................................................ 140 5.3.2. Residence time distribution (species adsorption on catalyst) ....................... 142 5.3.3. Reaction rate and pseudo first order constant in the trickle bed ................... 145 5.3.4. Mass transfer coefficients in the trickle bed ................................................. 147 5.3.5. Inna-particle mass transfer in the trickle bed ............................................... 151 5.3.6. Trickle bed kinetics ...................................................................................... 152 5.3.6.1. Macro kinetics in trickle bed .................................................................. 152 5.3.6.2. The limiting reactant in liquid phase ...................................................... 154 5.4. TRICKLE BED MODELING ...................................................................................... 155 5.4.1. Model equations and boundary conditions ................................................... 155 5.4.2. Analysis of the model ................................................................................... 157 5.4.3. Modeling of trickle bed ................................................................................ 159 xi 5.4. 5.4 55.5 CHIP 6.1.1 6.] 5.4.4. Model parameters ......................................................................................... 160 5.4.5. Simulation results at 100°C .......................................................................... 161 5.5. SUMMARY ............................................................................................................ 163 CHAPTER 6. MECHANISTIC INSIGHT .................................................................... 165 6.1. CONTROL EXPERIMENTS FOR MECHANISM ELUCIDAIION ...................................... 165 6.1.1. PG hydrogenation (M47, M48) .................................................................... 166 6.1.2. Ethanol and methanol hydrogenation on RulC catalyst (M49) .................... 168 6.1.3. Propanoic acid hydrogenation ...................................................................... 170 6.1.4. Adding PG in lactic acid hydrogenation ....................................................... 171 6.2. GAS BY-PRODUCT INFORMATION AT DIFFERENT REACTION CONDITIONS ............... 173 6.2.1. Gas composition for low pressure hydrogenation (330psi) .......................... 17 5 6.2.2. Gas product distribution in trickle bed reactor ............................................. 175 6.2.3. Catalyst deactivation by unrefined Cargill lactic acid samples .................... 177 6.3. RA‘I'IONALIZATION OF'I'HEREACTION PATHS ........................................................ 177 6.3.1. Side reaction 1(methane formation) ............................................................. 181 6.3.2. Side reaction 2 (propanol formation) ............................................................ 182 6.3.3. Side reaction 3 (propane formation) ............................................................. 183 6.3.4. Side reaction 4 (ethane formation) ............................................................... 183 6.4. SUMMARY ............................................................................................................ 184 CHAPTER 7. SUMMARY AND RECOMMENDATIONS ........................................ 185 7.1. SUMMARY ............................................................................................................ 185 7.1.1. Hydrogenation of lactic acid in autoclave .................................................... 185 xii 71.2. H 7.1M 7.1.4.1 71.5.1 72. chc 7.7.1.5 722. 8 7.1.2. Hydrogenation reaction in trickle bed reactor .............................................. 187 7.1.3. Catalyst characterization and deactivation ................................................... 187 7.1.4. Kinetic parameter measurement and trickle bed modeling .......................... 188 7.1.5. Reaction pathway .......................................................................................... 188 7.2. RECOMMENDATIONS ............................................................................................ 189 7.2.1. Surface reaction pathway investigation ........................................................ 189 7.2.2. Selective deactivation of the catalyst and yield enhancement ...................... 190 7.2.3. Production of optically active propylene glycol ........................................... 190 APPENDIX A. PARAMETERS CALCULATION AND PHYSICAL DATA ........... 191 A. l . LACIIC ACID CONSUMPIION RATE AND HYDROGEN CONSUMPIION RATE ............ 191 A2. BULK DENSITY OF CATALYST .............................................................................. 191 A3. EXTERNAL POROSTTY .......................................................................................... 192 AA. INTERNAL POROSTTY AND CATALYST DENSITY ..................................................... 192 A5. DIFFUSION COEFFICIENTS .................................................................................... 192 A.6. BASIC PHYSICAL DATA USED IN KINETIC CALCULATION ....................................... 193 A61. Lactic acid .................................................................................................... 193 A.6.1.1. Density of aqueous solution of lactic acid ............................................. 193 A.6.1.2. Viscosity of aqueous solution of lactic acid .......................................... 194 A62. Saturation pressures of lactic acid and propylene glycol ............................. 195 A63. Equilibrium of lactic acid hydrogenation to PG .......................................... 195 REFERENCE ................................................................................................................. 197 xiii 1.1 Ti 11 1:. 1:; 1r .. .. e we LIST OF TABLES TABIE 1-1. PROPYLENE GLYCOL PRODUCTION ................................................................. 12 TABLE 1-2. IMPORTANT LACIIC ACID REACTIONS ............................................................. 14 TABLE l-3.LACI'1C HYDROGENATION RESULTS FROM PATENT U8573l479 ..................... 15 TABLE 14. TYPICAL RESULTS OF CARBOXYLIC ACID HYDROGENATION ........................... 21 TABLE 1-5. BROADBENT’S CARBOXYLIC ACID HYDROGENATION RESULTS ....................... 21 TABLE 1-6. PROPIONATE HYDROGENATION (5°) ................................................................. 27 TABLE 1-7. SOLUBILITY (MIMI-120) OF HYDROGEN IN WATER 0 .................................... 27 TABLE 2-1. REAGENTS USED IN THE HYDROGENTION 0F LACIIC ACID .............................. 34 TABLE 2-2. COMMERCIAL CATALYSTS ............................................................................. 35 TABLE 2-3. RESULTS OF INCIPIENT WETNESS TESTING ...................................................... 35 TABLE 2-4. DETAILS OF MSU RU/C POWDER CATALYSTS ............................................... 38 TABLE 2-5. EXPERIMENTAL PARAMETERS ........................................................................ 42 TABLE 2-6. HPLC OPERATION PARAMETERS ................................................................... 54 TABLE 2-7. REAL PEAK POSITION AND MENU VALUE ........................................................ 59 TABLE 2-8. PEAK ASSIGNMENT ........................................................................................ 59 TABLE 3-1. OPTIMIZATION MATRIX-l .............................................................................. 63 TABLE 3-2. EXPERIMENT NUMBERS FOR MATRICES 2 AND 3 ............................................ 65 TABLE 3-3. DETAILS OF MSU RUTHENIUM CATALYSTS AND TEST CONDIIIONS ................ 74 TABLE 34. CONVERSION AND SELECTIVITY ..................................................................... 75 TABLE 3-5. SUMMARY OF ALL EXPERIMENTS FOR LACTATE SALT HYDROGENATION ........ 78 TABLE 3-6. SUMMARY OF CARBON BALANCE ................................................................... 83 TABLE 4-1. GAs SUPERFICIAL VELOCIIY AT 1200PSI AND 100°C ..................................... 88 xiv TABLE Irsu 11311 1.1311 1.13: 1132. 118’. TEL 113: 1.13: Ia: TABLE 4-2. CATALYST SUPPORTS SPECIFICATION ............................................................. 90 TABLE 4-4. CATALYST PROPERTIES .................................................................................. 90 TABLE 4-5. THREE CATALYST CHARGES USED IN TRICKLE BED ........................................ 92 TABLE 4—6. TRICKLE BED REACTION SUMMARY (CHARGE 1 AND CHARGE 2) ................... 94 TABLE 4-7. RESULTS OF TRICKLE BED WITH 48GRAM CATALYST (CHARGE 3) .................. 95 TABLE 4-8. RESULTS OF 17.2% LACIIC ACID FEEDING ................................................... 103 TABLE 49. RESULTS OF ADDING SULFUR ....................................................................... 104 TABLE 4-10. CALCULATION SHOWS THE DEACIIVAIION IS FAST .................................... 105 TABLE 5-1. SOLUBILITY IN HPLC WATER (ML (STP)/G) ................................................ 117 TABLE 5-2. SOLUBIIITY OF HYDROGEN IN 10% LACIIC ACID (ML(STP)/G) .................... 119 TABLE 5-3. REACTION RATE FOR THREE CATALYSTS LOADING AT 150°C AND 2000PSI.. 121 TABLE 5-4. PSEUDOFIRSTORDER CONSTANT ................................................................. 121 TABLE 5-5. SUMMARY OF MASS TRANSFER COEFFICIENT ................................................ 124 TABLE 5-6. STIRRING SPEED EFFECTS ............................................................................. 128 TABLE 5-7. REGRESSION RESULTS .................................................................................. 135 TABLE 5-8. REGRESSION RESULTS FOR 130°C IN AUTOCLAVE ....................................... 138 TABLE 5-9. RESIDUAL SPECIES AFTER SWITCHING TO WATER ......................................... 145 TABLE 5-10. BALANCE OF DESORPTION AND ADSORPITON ............................................. 145 TABLE 511. PSEUDO FIRST ORDER REACTION CONSTANTS IN TRICKLE BED ................... 146 TABLE 5-12. COMPARISONS OF H; MASS TRANSFER AND OBSERVED REACTION RATE 150 TABLE 5-13. PSEUDOFIRSTORDER CONSTANT AT DIFFERENT TEMPERATURE ................. 153 TABLE 5-14. MODEL EQUATIONS AND BOUNDARY CONDITIONS FOR TRICKLE BED ......... 157 TABLE 5-15. MODEL PARAMETERS AT 100°C ................................................................ 160 XV TABLE 1.15:5 : TAM-I'- TABLE I 113T: TEE. 11315., 1.3.815..- TABLE. 1.13:2. TABLE 5-16. L-S MASS TRANSFER COEFFICIENTS FOR HYDROGEN AND LACHC ACID ...... 161 TABLE 5-17. SIMULATION RESULTS FOR K1 .................................................................... 161 TABLE 61. PRODUCT DISTRIBUTION FOR PG HYDROGENATION AT 170°C ..................... 167 TABLE 62. AVERAGE BOND ENERGY ............................................................................. 178 TABLE 6-3. AB INIIIo ENERGY (KCAI/MOLE) FOR DIFFERENT STRUCTURES ‘7’ .............. 179 TABLE 6-4. AB INTIIO ENERGY CHANGE DURING REACTION ........................................... 180 TABLE A-l. CATALYST DENSITY ANDPOROSITY ............................................................ 192 TABLE A-2. DmsrvrmcmZ/ssc) ............................................................................... 193 TABLE A-3. THERMODYNAMIC PROPERTIES FOR LA AND PG ........................................ 193 TABLE A-4. DENSIIIES OF AQUEOUS SOLUTIONS OF LACTIC ACID ................................... 194 TABLE A-5. VISCOSITIES AS A FUNCTION OF CONCENTRATION AND TEMPERATURE (CP) 194 xvi F 1131' F131; F13; F13; LIST OF FIGURES FIGURE l-l. ANNUAL LACTIC ACID PRODUCTION ............................................................... 7 FIGURE 1-2. CHEMICAL SYNTHESIS OF LACTIC ACID ........................................................... 8 FIGURE 1-3. PROPYLENE GLYCOL USE DISTRIBUTION 0 ..................................................... 12 FIGURE 14. EB EFFICIENCY vs. TEMPERATURE IN L (+) LACIIC ACID HYDROGENATION .. 16 FIGURE 2-1. HYDROGEN CHLORINE EVOLUTION PROFILE DURING REDUCTION .................. 37 FIGURE 2-2. BATCH REACTOR AND CONTROLLER ............................................................. 39 FIGURE 2-3. BATCH REACTION SYSTEM ............................................................................ 40 FIGURE 24. TRICKLE BED REACTOR SYSTEM ................................................................... 43 FIGURE 2-5. MASS SPECTROMETER SYSTEM ..................................................................... 45 FIGURE 2-6. FINAL GAS PHASE ANALYSIS FROM MASS SPECTROMETER (AUTOCLAVE) ...... 46 FIGURE 2-7. GAS PHASE ANALYSIS FROM MASS SPECIROMETER (TRICKLE BED REACTOR) 47 FIGURE 2-8. STANDARD MASS SPECTRUMS FOR MEIHANE ................................................ 47 FIGURE 2-9. STANDARD MASS SPECTRUMS FOR ETHANE .................................................. 48 FIGURE 2-10. STANDARD MASS SPECTRUMS FOR ETHENE ................................................. 48 FIGURE 2-11. STANDARD MASS SPECTRUMS FOR PROPANE ............................................... 49 FIGURE 2-12. STANDARD MASS SPECTRUMS FOR PROPENE ............................................... 49 FIGURE 2-13. BACKGROUND OF PURE HYDROGEN ............................................................ 50 FIGURE 2-14. CALIBRATION OF CH4, CO AND C02 .......................................................... 51 FIGURE 2-15. TYPICAL CHROMATOGRAPH FOR LIQUID ANALYSIS ..................................... 53 FIGURE 2-16. SCHEMAIIC OF HPLC SYSTEM ................................................................... 54 FIGURE 2-17. RESPONSE OF LACHC ACID ANALYSIS Is A CONSTANT ................................. 56 FIGURE 2-18. RESPONSE OF PROPYLENE GLYCOL ANALYSIS IS A CONSTANT ..................... 57 xvii FER 11611 Fit-LI FELT SA FIG I] s I My. (J y... FIGURE 2-19. TYPICAL HPLC CHROMATOGRAPH OF LIQUID PRODUCTS ........................... 58 FIGURE 31. FIVE-HOUR CONVERSION (TEMPERATURE AND PRESSURE IN PARENTHESEs). 62 FIGURE 32. LAch ACID CONVERSION vs. PRESSURE & TEMPERATURE AFTER 5 HOURS 64 FIGURE 3.3. LIQUID BY-PRODUCTS DISTRIBUTION AFTER 3 HOURS ................................... 65 FIGURE 3-4. CATALYST LOADING EFFECT AT Low TEMPERATURE AND PRESSURE ............ 67 FIGURE 3-5. CATALYST LOADING EFFECT AT HIGH TEMPERATURE AND PRESSURE ............ 67 FIGURE 3-6. SELECTIVITY CHANGES WITH CATALYST LOADING AT 150°C ........................ 68 FIGURE 3-7. SELECTIVITY CHANGE WITH CATALYST LOADING AT 130°C ......................... 69 HGURE 3-8. PRE-REDUCIION EFFECT ............................................................................... 70 FIGURE 3-9. REACTIONS AT DIFFERENT LACTIC ACID CONCENTRATION ............................ 71 FIGURE 3-10. GAS PRODUCT EVOLUTION FOR RulC (PMC) ............................................. 73 FIGURE 3-11. GAS PRODUCT EVOLUTION FOR RU/ALUMINA (DEGUSSA) .......................... 73 FIGURE 3-12. FIVE-HOUR CONVERSION AND SELECTIVITY FOR MSU CATALYSTS ............ 76 FIGURE 3-13. GAS BY-PRODUCI DISTRIBUTION AFTER FIVE-HOUR REACTION ................... 77 FIGURE 3-14. K*IONEFFBCTON LACHC ACID HYDROGENATION AT170°C ...................... 79 FIGURE 3-15. ADDING POTASSIUM SULFATE To LACHC ACID AT 150°C ........................... 79 FIGURE 3-16. CALCIUM TONEFFECT AT 150°C ................................................................. 81 FIGURE 3-17. CATALYST LOADING EFFECT ON THE GAS PRODUCTS AT1500PSI AND 150°C84 FIGURE 4-1. FLASH VAPORIZATION AND STEAM SATURATION PRESSURE .......................... 88 FIGURE 4-2. HYDROGEN DESORPTION PROFIIE DURING DISPERSION MEASUREMENT ........ 90 FIGURE 43. GRANULAR CATALYSTS PERFORMANCE IN BATCH REACTOR ......................... 91 FIGURE 44. LACIIC ACID CONVERSION vs. REACTION TEMPERATURE ............................. 96 FIGURE 4-5. PG SELECTIVITY vs. REACTION TEMPERATURE ............................................ 96 xviii FIGURE 4-6. CONVERSION PROFILE vs. TEMPERATURE AND PRESSURE ............................. 97 FIGURE 4—7. SELECTIVITY PROFILE VS. TEMPERATURE AND PRESSURE ............................. 98 FIGURE 4—8. EFFECIOF MOLAR RATIO ON CONVERSION ................................................... 99 FIGURE 4-9. EFFECIOF MOLAR RATIO ON SELECTIVITY .................................................. 100 FIGURE 4-10. CONVERSION AND SELECTIVITY vs. WEIGHT HOUR SPACE VELOCITY ........ 101 FIGURE 4—11. EFFECIOFWEIGHT HOUR SPACEVELOCITY .............................................. 101 FIGURE 4-12. CONVERSION AND SELECTIVITY OF EXTENDED TIME REACTION ................ 102 FIGURE 4-13. SELECTIVITY AND CONVERSIONS FOR TWO CONCENTRATIONS .................. 103 FIGURE 4—14. CONVERSION AND SELECTIVITY CHANGE WITH ADDITION OF SULFUR ....... 106 FIGURE 4-15. CONVERSION PROFILE COMPARISON OF CARGILL AND PURE LACIIC ACID 108 FIGURE 4-16. SELECTIVITY PROFILE OF CARGILL AND REGULAR LACTIC ACID ................ 108 FIGURE 4-17. CATALYST DEACIIVAIION (SWITCH FROM CARGILL To PURE LACHC ACIDl 10 FIGURE 4-18. CONVERSION, SELECTIVITY AND YIELD OF THREE LACHC ACIDS .............. 111 FIGURE 4-19. GAS PHASE COMPOSITION CHANGES WITHREACITONTIME ANDFEED ....... 112 FIGURE 5-1. APPARATUS FOR SOLUBILITY MEASUREMENT ............................................. 116 FIGURE 5-2. COMPARISON OF MEASURED SOLUBILITY AND LITERATURE DATA .............. 1 18 FIGURE 5-3. MINIMUM STIRRING SPEED FOR CATALYST SUSPENSION .............................. 120 FIGURE 5-4. HYDROGEN-WATER MASS TRANSFER COEFFICIENT IN THE AUTOCLAVE ...... 125 FIGURE 5-5. COMPARISON OF BERN’S CORRELATION AND MEASUREMENT ..................... 127 FIGURE 5-6. L-S MASS TRANSFER COEFFICIENT FROM SANo’s CORRELATION ................ 129 FIGURE 5-7. MASS TRANSFER COEFFICIENT FROM BOON-LONG’S EQUATION .................. 130 FIGURE 5-8. OBSERVABLE MODULUS CHANGES WITH CATALYST DIAMETER .................. 131 FIGURE 5-9. FIT CONCENTRATION CURVE TO 4TH ORDER POLYNOMIAL .......................... 132 xix PEI! FIGHT FEE FIGLRE IETT 116m: FIGURE 5-10. GET INITIALREACIION RATE FROM EXTRAPOLATING THE RATE CURVE 133 FIGURE 5-1 1. INITIAL REACTION RATES WITH CATALYST LOADING ................................. 133 FIGURE 5-12. COMPARISON OF EXPERIMENT AND PREDICIED RATES .............................. 135 FIGURE 5-13. COMPARISON OF MODEL PREDICTION (130°C) .......................................... 139 FIGURE 5-14. LIQUID HOLDUP AT DIFFERENT LIQUID FLOW RATES .................................. 141 FIGURE 5-15. LIQUID RESIDENCE'I'IMEIN'IIIETRICKIE BED ........................................... 141 FIGURE 5-16. OUTLET CONCENTRATION CHANGE WITH TIME AFTR SWITCH TO SOLUTION143 FIGURE 5-17. OUTLET CONCENTRATION CHANGE WITH TIME AFTER SWITCHING ............ 143 FIGURE 5-18. G-L MASS TRANSFER COEFFICIENT VS. FLOW RATE ................................... 148 FIGURE 5-19. L—S MASS TRANSFER COEFFICIENT vs. LIQUID FLOW RATE ........................ 150 FIGURE 5-20. OBSERVABLE MODULUS IN TRICKLE BED (100°C AND 1200PSI) ............... 151 FIGURE 5-21. ARRHENIUS PLOT FOR PSEUDO FIRST ORDER REACTION ............................ 153 FIGURE 5-22. Hz IN BULK LIQUID WITH CONSTANT LIQUID REACTANT CONCENTRATION. 159 FIGURE 5-23. SIMULATION RESULTS AT 100°C AND 400~1200PSI ................................. 162 FIGURE 5-24. COMPARISON OF BULK LIQUID AND SURFACE CONCENTRATION ................ 162 FIGURE 61. GAS PRODUCTS FROM PROPYLENE GLYCOL HYDROGENATION .................... 167 FIGURE 6-2. LIQUID PRODUCTS FROM PG HYDROGENATION AT 170°C .......................... 168 FIGURE 63. CONVERSION OF METHANOL AND ETHANOL HYDROGENATION AT 150°C 169 FIGURE 6-4. GAS PHASE ANALYSIS OF METHANOL AND ETHANOL HYDROGENATION ..... 169 FIGURE 6-5. COMPARISON OF LACTIC ACID AND PROPANOIC ACID HYDROGENATION ..... 171 FIGURE 6-6. CONVERSION PROFILES FOR ADDING DIFFERENT PG CONCENTRATION ........ 172 FIGURE 67. PG ADDITION EFFECT LACTIC ACID HYDROGENATION AT 150°C ................. 17 3 FIGURE 6-8. GAs PRODUCT COMPOSITION CHANGE WITH CATALYST LOADING ............... 174 XX Fm: 6 FIG FIGURE 6-9. GAS PRODUCT COMPOSITION CHANGE WITH CATALYST LOADING ............... 174 FIGURE 6-10. GAS PRODUCT COMPOSITION CHANGE WITH CATALYST LOADING ............. 174 FIGURE 6-1 1. GAS PRODUCT COMPOSITION CHANGE WITH CATALYST LOADING ............. 174 FIGURE 6-12. GAS PHASE SPECTRUM AFTER 12 HOURS AT 330PSI AND 150°C ................ 175 FIGURE 6-13. GAS PHASE ANALYSIS FOR CG6M CATALYST IN TRICKLE BED .................. 176 FIGURE 6-14. GAS PHASE ANALYSIS FOR CGSP CATALYST IN TRICKLE BED ................... 177 FIGURE 6-15. Hz ADSORPTION AND DESORPTION CHANGE AFTER DEACIIVAIION .......... 178 FIGURE 6-16. PG FORMATION PATH ............................................................................... 181 FIGURE 6-17. SCHEME OF PG HYDROLYSIS .................................................................... 182 FIGURE 6—18. PROPANOL FORMATION SCHEME ............................................................... 182 FIGURE 6-19. PROPANE FORMATION SCHEME ................................................................. 183 FIGURE 6-20. ETHANE FORMATION SCHEME ................................................................... 183 FIGURE A-l. SATURATION PRESSURES OF LACTIC ACID AND PROPYLENE GLYCOL .......... 195 FIGURE A-2. LA EQUILIBRIUM CONVERSION .................................................................. 196 xxi LA BET €06)? C6511 (i1 BET C2 C3 CG6M CGSM G-L HPLC WHSV H-W L-S Mass PMC Rpm RulC W% PD ABBREVIATIONS Lactic acid (water solution) Surafce are measured by BET method Hydrocarbon contained two carbon atoms Hydrocarbon contained three carbon atoms Active carbon support and the catalyst based on this support Active carbon support and the catalyst based on this support Gas-liquid Gas chromatography High performance liquid chromatography Weight hour space velocity Hougen-Watson model liquid-solid (catalyst surface) The rule in mass spectrum Michigan State University Propylene glycol (water solution) Pressure Chemical Corporation Round per minutes Ruthenium on carbon support Temperature (°C) Weight percentage Propane-1,1,2-triol 2—Hydroxy-propionaldehyde xxii NOTATION a Gas liquid interfacial area per unit volume of catalyst (cmz/cm3) 0,, Gas liquid interfacial area per unit volume of reactor (cmZ/cm3) AL Hydrogen concentration in bulk liquid (mole/cm3) Au Hydrogen concentration in liquid in inlet (at x=O) (mole/cm3) AS Hydrogen surface concentration (mole/cm3) A' Saturation concentration (hydrogen solubility) (mole/cm3) BL Lactic acid concentration in bulk liquid (mole/cm3) Bu Lactic acid concentration in liquid in inlet (at x=0) (mole/cm3) BS Liquid concentration on catalyst surface (mole/cm3) CM Lactic acid concentration (mole/ml) (autoclave) CfA Lactic acid concentration at start (mole/ml) (autoclave) D A Hydrogen diffusivity (cmzls) DB Lactic acid diffusivity (cm2/s) De Effective diffusivity (cmzls) FL Liquid feed flow rate (ml/min) K L Gas to liquid mass-transfer coefficient (m/s) k5 Liquid to solid mass-transfer coefficient (m/s) hd Dynamic liquid holdup {ml(liquid)/ml(catalyst)} LHSV Liquid hourly space velocity (hr") 2 Dimensionless distance xxiii m P: C L Dimensionless concentration of hydrogen C5 Dimensionless surface concentration of hydrogen C u Dimensionless inlet concentration of hydrogen aGL Dimensionless gas -liquid mass transfer coefficient at: Dimensionless liquid solid mass transfer coefficient aR Dimensionless reaction rate constant VR Catalyst volume in trickle bed (ml) 77C Catalytic effectiveness factor uL Liquid flow velocity (m/s) p Liquid viscosity (g/cm.sec) m, n Reaction order respecting to A and B p Catalyst density (0.84 g/mL) p3 Catalyst bulk density 0.44 ,0,_ Liquid density g/mL (1 .l for 10% lactic acid solution) L PeA = 1"“ Peclet number for H2 A Pen = u—LL— Peclet number for lactic acid 8 RC , rA Reaction rate (mole/gcat.min) robs Observed reaction rate (mole/gcat.min) xxiv Chapter 1. Background The current chemical production technology is deeply rooted in petroleum as the carbonaceous resource, for which the storage is limited and will be depleted in the near future. So, alternative carbon sources should be found to replace it. Of course, the best way is to find a renewable carbon resource. On our earth, it seems that plants (biomass) are the only renewable carbon resource. Biomass offers potential for producing a wide Variety of chemical species, either as unique new products or as competitive alternatives t0 traditionally petroleum-derived species. Biomass-based feedstocks (organic acids, BIC. . .) are attractive in making chemicals because they have the potential to be converted 10 a series of useful chemicals. Also, they can be produced via highly selective fermentation pathways and have high hydrogen to carbon ratio, and have low cOncentrations of undesirable impurities such as ash, trace metals, sulfur and phosphorus. It is well known that sulfur and phosphorus are deadly poisons for many catalysts and the purification processes (removing S and P) are very costly. Lactic acid will be used as a starting point for further utilizing biomass carbon resource, because it can be used as a precursor to synthesize many valuable chemicals and is readily available and very cheap in the future. Worldwide production of lactic acid (2-hydroxy-propionic acid) has dramatically increased since the early 1990s. The viability of making an inexpensive biodegradable polymer (poly-lactic acid) from lactic acid has sparked extensive interest and research in the area of producing and recovering relatively pure lactic acid from fermentation of corn starch. Recent advancements in fermentation technologies for lactic acid production will make it even more inexpensive. The primary focus of this work is on the formation of propylene glycol from lactic acid by aqueous phase hydrogenation over supported metal catalysts in mild reaction conditions. 1.1. Literature Review The possible carbon resources in the future are coal and renewable biomass. The coal resource is very large compared to petroleum, but the impurities (especially, sulfur and phosphorus) in coal make its utilization very difficult and costly, and also it could run out some day. The C02 in atmosphere could be used as a carbon resource, but from the present techniques and energy sources, it is not practical to get carbon directly from air. We have to use plant as a media, which can adsorb C02 directly from air and store it in the form of biomass. So, renewable biomass seems to be the final choice for mankind as a carbon resource. Also, some chemicals cannot be synthesized with present techniques or the procedures are too complicated and the cost is too high for chemical synthesis, and the only way is to use biomass fermentation (for example, most of the anti-bacteria drugs). With technical progress in fermentation and separation, it is also possible to produce chemicals as competitive alternatives to traditional petroleum-derived chemicals. The process of using biomass as a feedstock will produce less toxic waste and less pollution to the environment. This is another advantage compared to petroleum-based technology. Biomass consists of collectible plant derived materials that are abundant in nature, inexpensive, and potentially convertible to feedstocks in chemical production by fermentation processes. Currently, biomass used for fermentation mainly comes from agriculture, such as starch (corn, wheat, potato, and sago palm), because these are mass- produced, uniform in quality and oversupplied in some countries and the US. However, lignoeellulose (wood, agricultural residue, grass) should be the major source of biomass in the future since it is both available in large quantities and has no competing use as food. From fermentation by using bacteria, fungi or yeast under mostly anaerobic conditions, various organic acids can be yielded. However, the fermentation process always produces some by-products other than the main compound, and the process economics are decided mostly by product separations. The separations will take most of the production cost. Propylene glycol (PG) is a valuable and mass produced chemical and has many uses in food and chemical industries. The trickle bed reactor (TBR) is a concurrent downfiow packed column. It provides a good means of carrying out a reaction in which gaseous and liquid reactants are to be contacted with catalyst particles. TBR is a relatively new type of reactor, so some review will be given here. 1.1 c." (:11 via- 1.1.1. Lactic acid CH3CHOHC02H, is a colorless liquid organic acid. It is miscible with water or ethanol. Lactic acid is a fermentation product of lactose (milk sugar); it is present in sour milk, koumiss, leban, yogurt, and cottage cheese. The protein in milk is coagulated (curdled) by lactic acid. Lactic acid is produced in the muscles during intense activity. Calcium lactate, a soluble lactic acid salt, is used as a source of calcium in the diet. Lactic acid is produced commercially for use in pharmaceuticals and foods, in leather tanning and textile dyeing, and in making plastics, solvents, inks, and lacquers. Chemically, lactic acid occurs as two optical isomers, a dextro and a levo form; only the levo form takes part in animal metabolism. Currently, commercial synthetic lactic acid is racemic mixture and currently, all most all fermentation lactic acid is an optically active L (+) form. 1.1.1.1. History of discovery Lactic acid is the simplest hydroxyl acid having an asymmetric carbon atom and it therefore exists in a racemic form and in two optically active forms with opposite rotations of polarized light. Lactic acid occurs widely in nature as the racemic form as well as the optically active acid. The story of lactic acid is the history of modern biology and chemistry; the development of the industrial making of lactic acid is the history of the modern chemical industry (Benninga (1)). The following story is also from this reference. In 1780, famous Swedish chemist C_arl Wilhelm Scheele (2) found a new acid from sour milk, and named it after its origin Mjolksyra (acid of milk). In 1813, French chemist Henri Braconnot (1781-1855) found “another organic acid” (he called nanceic acid) from an' long Gan: R gad maxi the 52.3 weed BQRC rdéd mfirp mama; Chars- I.%SI rice water, beet juice, boiled beans and peas, and soured suspensions of baker’s yeast. He thought it was different from the acid discovered by EEOC—16. in the sour milk. Although German chemist Egg] proved the identity of lactic and nanceic acid later, many scientists believed that the lactic acid was no more than impure acetic acid in that time. Originally the lactic acid from fermentation and that found in muscle tissue were regarded as identical. In 1847, m re-examined meat extract and suspected that the two acids might not be same. Engelhardt found that the crystallization and solubility of the salts of the two lactic acids were different, and thus he concluded that these two acids were different. Heintz examined the two acids further and named the muscle lactic acid as sarco-lactic acid. The confusing properties of fermentation lactic acid and meat lactic acid did not get solved until the discovery of chemical’s optical activity. A lengthy paper on the properties of sareo-lactic acid in 1873 explained that the two lactic acids have identical chemical structure, nevertheless differ in optical activity. Finally in 1874, Dutch chemist Van’t Hoff proposed a geometrical model to explain the phenomenon of optical activity; in the same year French scientist I._ie;B_el (1847-1930) independently arrived at almost the same explanation of the optical rotation and connected the optical rotation to molecular asymmetry. 1.1.1.2. Production of lactic acid The principle source of lactic acid is fermentation, even today. Charles Elleg Am was the first lactic acid manufacturer in the US (I). The US $100,000 plant was built in 1882 in Littleton, Boston and the production began around 1883. Although the production capacity is unclear, the annual coal consumption was high, up to 22000 tons per year! In 1893, Merck began the production of lactic acid in Germany, and from then Germany was the major lactic acid maker in the world for a long time. Its lactic acid was exported to many industrialized countries such as England, France, Japan and US. For a fairly long time (until WWII), German export of lactic acid was over 2000 tons (pure) per year. The first preparation of synthetic lactic acid was performed by ME in 1850 by the reaction of alanine, 2-aminopropionic acid, with nitrous acid. A more suitable synthesis was found by Wislicenus (I) in 1863 by reacting acetaldehyde with hydrogen cyanide to from lactonitrile, which can be hydrolyzed to lactic acid in the presence of hydrochloric acid. Although the study of synthetic lactic acid started in the same time as the discovery of lactic acid, fermentation was the only method to produce lactic acid until 1960s. Monsanto’s 4500 t/a synthetic lactic acid plant was built in 1962 in Texas City and doubled in its capacity in 1969. Du Pont closed the old lactic acid plant at Gray’s Ferry in 1963. America Maize finished its lactic acid operation at 1964. Clinton, the largest manufacturer at that time, kept lactic acid production until 1982, but never made a sizable investment again. It looked like the synthetic route would replace fermentation for lactic acid production in no time. But the invention of producing enantiomerically pure L (+) lactic acid from fermentation in HV A laboratory (around 1967) saved the fermentation lactic acid production because the major uses of lactic acid are the food industry. L (+) isomer occurs normally in human metabolism, but the D (-) isomer is a foreign substance that metabolized differently. L (+) lactic acid will be converted to glycogen and D (-) lactic acid is oxidized or excreted in the urine, so L (+) lactic acid is natural for humans and D (-) is not. Only L (+) lactic acid is natural to humans and does not have any side effects, so it can be used in food safely. 140000«// 1200004 1000004 80000~.' 60000- 40000—. 20000- 7 World lactic acid production (t/a) .4 1 962 1 972 1 982 1997 1999 Year Figure 1-1. Annual lactic acid production Worldwide lactic acid production is shown Figure 1-1. The top parts in l962~ 1982 are synthetic production, but none of these data are available for 1997 and 1999, as most is from fermentation according to the sources. This figure is compiled with data from Bennina (I) and Chemical Market Report ‘3). 1.1.1.3. Synthetic technology The currently used chemical synthetic method was first found by German chemist Wislicenus in 1863, and the first patent was requested by a German company in 1930. Du Pont was granted another patent (1984415) in 1933 for preparation of lactonitrile. In 1949, Musashino Chemical Laboratory Ltd. of Tokyo, Japan realized Wislicenus’s idea in pilot scale plant with output of 330 tons per year. Because synthetic lactic acid is a racemic mixture of two enantiomers, it is not good for food industry. The largest synthetic plant (Monsanto) is based on a byproduct from acrylonitrile synthesis, so no other competitor appears except the only pilot plant in Japan. In base-catalyzed liquid phase, hydrogen cyanide reacts with acetaldehyde. The lactonitrile produced is distilled and hydrolyzed using a concentrated mineral acid like hydrochloric or sulfuric acid to produce lactic acid. The crude lactic acid produced is esterified to methyl lactate and recovered by distillation and hydrolysis by water under acid catalysts (4). CH3CHO + HCN < > CH3CHO HCN CH3CH0 HCN 4» H20 +l/2 HzSO.e %3mOHCWH + 1/2 (NI-L)ZSO4 CH3CHOH COOH + CH3OH < >CH3CHOH COOCH; + H20 CH3CHOHCOOCH3 +H20 % >CH3CHOH COOH + CH3OH Figure 1-2. Chemical synthesis of lactic acid A heavily researched and even piloted process is the oxidation of propylene with nitrogen peroxide. This process is based on the discovery of Lm and S_oaifg (5) in 1944. French’s Rh6ne-Poulenc and German’s BASF filed patents in 1966 and 1965 for this process and even the Russian scientists worked on its reaction mechanism (1). In this process, propylene gas mixed with oxygen is passed at a temperature near the freezing point through a solution of about 16% nitrogen peroxide in concentrated nitric acid (70%). Propylene gas is absorbed quantitatively and the oxygen immediately oxidizes low nitrogen oxides formed in reaction back to nitrogen peroxide at same time. Finally, this process was given up, the possible reason is the unstable by-products, which may decompose uncontrollably and even explode (6). In addition, the major market is the food industry and only pure L (+) lactic acid is preferable. Another not commercialized mastic athlom; 1.1.1.-l. l 0 mixture. ' and met: 1.1+) LII fermentet MUG ham ” 51711;) am Nut! 11 LTmedj a: synthetic routes include direct oxidation of ethanol or propylene glycol and hydrolysis of oz-chloropropionic acid (7). 1.1.1.4. The production of lactic acid from fermentation Originally, the lactic acid produced from natural fermentation was a racemic mixture. With the pressure from synthetic lactic acid and the knowledge of racemic lactic acid metabolism, HV A laboratory at Schiedam began to work on the production of pure L (+) lactic acid from fermentation. With their successful work, at the end of 1967, the fermented 90% L (+) lactic acid was on the market (1). Currently, most commercial production of lactic acid via carbohydrate fermentation uses lactobacillus delbreuckii bacteria (8’ to make pure L (+) lactic acid. Many carbohydrates, including whey, corn syrup, and cane, can be sources for the fermentation. Because lactic acid is an end product inhibitor, it must be converted to salt or extracted from the fermentation broth immediately as it is produced. Traditionally, calcium hydroxide has been used for this purpose. The recovered calcium lactate is then purified by evaporation and acidified with sulfuric acid to give lactic acid (9). However, since calcium lactate has a low solubility in water (8 wt %), only low concentration of lactate can be produced from fermentation to prevent formation of a large fraction of solid hydrated calcium lactate in solution and maintain product flow throughout the process. The cost of purifying dilute solution of calcium lactate and disposal of large amount of CaSO4 has always been a drawback in the production of lactic acid via fermentation. Recent developments in simultaneous or coupled extractive technology have led (10) to the use of nano—filtration membranes , anion-exchange resin (11) activated carbon column, and solvent extraction (’2) for recovering lactic acid from fermentation broth. fen: ‘IICI: Electrodialysis also is a possible way for recovering highly purified lactic acid from its salt form "2). These new innovations will lower the production cost of lactic acid. Another technique is to use ammonia as a pH control instead of calcium hydroxide (’3’. Because ammonium lactate is highly soluble in water, the concentration of lactate in broth can be up to 30% before activity of the fermentation organism is impaired. Ammonium lactate can subsequently be converted to lactate esters and ammonia in the presence of gaseous CO2 and alcohol at around 160°C and moderate pressure (1"). The ammonia is then recycled back to the fermentation broth. These advantages have made ammonia very attractive as a replacement for calcium hydroxide. Another very interesting process for producing and separating lactic acid from fermentation was invented by C_argiflus’. In this process, the lactate solution obtained from a fermentation broth is extracted by a water immiscible trialkyl amine in the presence of carbon dioxide. Lactic acid is recovered from the resulting organic phase, and regenerated extractant is recycled for reuse in the extraction. So, the consumption of acids and bases are avoided and the generation of waste salts and other by-production are substantially reduced, if not eliminated. Theoretically, CO2 is the only material for consumption. If such a advanced technique can be commercialized, the lactic acid production cost will be even lower. 1.1.1.5. Lactic acid market The primary user of L (+) lactic acid is the food industry. It is used as a food acidulant/flavoringlpH-buffering agent or as an inhibitor of bacterial spoilage in several processed foods like soups, candy, bread, etc. Another application is as an emulsifying agent in foods such as bakery goods. Another major use is the leather tanning. 10 Pharmaceutical and cosmetic applications include use in topical ointrnents, lotions, and biodegradable polymers for medical applications. The use of lactic acid and other 2- hydroxycarboxylic acids have been shown to alleviate or improve signs of skin, nail, and hair changes associated with intrinsic or extrinsic aging. The only known manufacturer of D (-) lactic acid is Rh6ne-Poulenc, which built a fermentation plant at Melle to produce exclusive D (-) lactic acid as a building block for stereo-isomeric herbicides (1' p448). Since the early 19905, worldwide production of lactic acid has dramatically increased and is currently estimated to be over 130000 tons/year ‘3’ (Figure 1-1), of which most is produced by fermentation. The possibility of making an inexpensive biodegradable polymer (poly lactic acid) from lactic acid has further stimulated the interest in investment and research in the area of producing pure lactic acid from fermentation of carbohydrates. For enlarging the lactic acid production ability, the major corporations including ADM, Cargrill, Purac, and A. E. Staley have ventured into the manufacture of lactic acid (’6) . The availability of a high volume of inexpensive lactic acid has also inspired the research of using lactic acid as an alternative feedstock for the production of many special and commercial chemicals such as acrylic acid “7), propionic acid, 2,3-pentanedione, pyruvic acid, and propylene glycol (18). The current market price of lactic acid ranges between $0.70llb to $0.85/1b depending on its purity ('9). In 1997, US consumes over 9000 tons of lactic acid annually, of which 85% is used in food-related applications to improve meat shelf-life and flavors, pH buffering agent, acidulant, and the production of emulsifying agents. 11 1.1.2. Propylene glycol miscellaneous Unsaturated tobacco polyester resins paints and coatings pet foods functional fluids liquid detergents food. drug and antifreeze and de- cosmetics uses icing fluids (20) Figure 1-3. Propylene glycol use distribution Propylene glycol is a nontoxic chemical and has many industrial applications. The use distribution is shown in Figure 1-3. UPR (unsaturated polyester resin), antifreezelcoolants and deicing fluids and food products are the major consumption. Tobacco, paints, synthetic marbles, propylene glycol ether (PGEs), high-performance industrial solvents for paints and coatings, cleaners, inks, and a variety of other pharmaceuticals consume the other part. Propylene glycol is a mass produced commercial chemical (450000 tons /year, 1998). The production capacity profiles are given in Table 1-1 (20). Table 1-1. Propylene glycol production CAPACITY 250 - 230 72 120 75 1312 Total (per year) Currently, all commercial production of PG is fi'om the hydration of propylene oxide. Di- and tri-propylene glycols, as well as small quantities of higher glycols, are also produced in the same plant. Propylene glycol (PG) capacities at some locations can be supplemented by shifting hydration equipment normally used for ethylene glycol (EG) to the production of PG. Among the five major US producers of PG, only Arco and Dow are back integrated into propylene oxide. Eastman Chemical acquired its South Charleston, W. Va., propylene glycol plant from Arco Chemical in 1991, and Huntsman purchased the Port Neches glycol plant from Texaco in 1994. The market demand for propylene glycol 460000 tons in 1997 and 476300 tons in 1998. The predicted demand in 2002 is 570000 tons. The production has grown at about 3~4% per year since 1988. The market price is high, up to 0.68$/lb in 1997 and the current price is 0.6$/lb USP grade and $0.47/lb industry grade ‘2". Lancaster Synthesis Ltd. (UK) produces optical active propylene glycol, but the market demand is unclear. 1.2. Lactic acid reactions Lactic acid (C3HGO3) and starch (Cd-IloOS-l-H2O) have same C, H, 0 ratio, therefore, starch can be converted to lactic acid at very high yield in fermentation process. In addition, the price of starch from corn is only $0.07~$0.10/lb. With the Progress of fermentation technology, it is possible to produce lactic acid at a price lower than 20 cent/1b. This potential further simulates the research about lactic acid derivatives. Lactic acid is a simple compound containing both hydroxyl and carboxylic acid groups, which permit it to participate in many interesting and valuable chemical feaCtions. Among the known reactions of lactic acid (Table 1-2), the dehydration to 13 1.2.1. 1 atrium from Iii: acrylic acid, the polymerization to poly (lactic acid) and the newly discovered condensation to 2,3-pentanedione (22’ are potentially the most profitable pathways. Table 1-2. Important lactic acid reactions Reaction Product Dehydration Acrylic acid Condensation 2,3-pentanedione Reduction Propylene glycol Polymerization Poly lactic acid 1.2.1. lactic acid reduction Broadbent et al (23) began the catalytic hydrogen reduction of lactic acid by using rhenium black as catalyst. Their work was about high activity rhenium powder prepared from rhenium heptoxide. They could hydrogenate many organic acids including lactic acid to corresponding alcohols. For lactic acid hydrogenation, they used 1 g ruthenium per mole lactic acid with no solvent. The total reaction time was 8 hours at 150°C and 3800 psi. The final products are 84% pr0pylene glycol and 16% of lactic acid and dilactide. They did not give any information about the purity of lactic acid used. Because Broadbent er al used in situ reduced high active rhenium black, they could use mild temperature to achieve high conversion. However, the pressure he used was still quite high from the viewpoint of process equipment requirements. In addition, lactic acid is only one of many acids used to test their catalyst, so no more details are given. Recently, m (2") patented a process for the preparation of optically active alcohols by reducing optically active carboxylic acids with hydrogen in the presence of ruthenium catalysts. In this patent, carboxylic acids with an alpha or beta branch are reduced with hydrogen at temperatures below 160°C and pressures below 3000 psi in the presence of ruthenium contained catalysts. For hydrogenating lactic acid, 4g of Ru black 14 and 89g of L- (+)-lactic acid were placed in 700g of water in a 1.3 L stainless steel autoclave. After flushing with nitrogen, the apparatus was closed and brought to a hydrogen pressure of lOObar. Over 2 hours the temperature was raised to 80°C and the hydrogen pressure to 200bar. The mixture was stirred at 80°C and 200bar until the uptake of hydrogen had ended. It was then cooled to room temperature, the catalyst was filtered off, and the water was distilled off. The residue obtained was distilled under nitrogen at 16 mbar to give 64g of L-(+)-propane-l,2—diol (b.p.= 74° C.; [a];0 +16.2°; ee>97%). Other examples of lactic acid hydrogenation are summarized in Table 1-3. Table 1-3. Lactic hydrogenation results from patent US5731479 Catalyst Amount Yield ee % Ru black 4 85 >97 10% by weight Ru-on-carbon 20 74 >97 RuO2 reduced at 150° C 2 88 >97 5% by weight Ru-on-Al2 03 20 68 >97 RuO2 reduced at 150°C 10 86 97 5% by weight Ru—on-carbon 10 35 >97 5% by weight Ru-on—carbon 20 64 >97 To maintain optical activity, Antons used very low temperature and high pressure. Although he did not give the reaction time, the reaction should very long due to the low temperature used. This patent shows the ee efficiency of lactic acid hydrogenation is only sensitive to the reaction temperature (Figure 1—4). The catalyst state (metal, supported metal or oxide) and catalyst support type (carbon or A1203) and catalyst loading do not affect the ee efficiency of formed L (+) propane 1,2-diol. The patent did not give the Pressure effect on the ee efficiency. 15 1 00% I I E 80% - I 2 I + 60% . —J 89 g Lactic acid hydrogenated at 200 bar with "5 a 4g ruthenium black. Summarized from patents ..\° 4° /° * US5731479, 0 ° 20% . 0% i i r . I 80 90 1 00 1 1 0 120 130 140 Temperature (C) Figure 1-4. ee efficiency vs. temperature in L (+) lactic acid hydrogenation For the above two published lactic acid hydrogenation studies, the first one used very high pressure, and the catalyst had to be prepared in situ. They did not mention by- products and corrosion in their paper, which are obvious problems for concentrated lactic acid at high temperature and pressure. The purpose of patent USS731479 is to produce an optically actively alcohol; to avoid racerrrization, a very low temperature was used, so the reaction time is possibly too long to be practical in industry. In summary, lactic acid reduction is not an easy task as highly active catalyst and elevated temperature and pressure are needed. Beside hydrogenation to propylene glycol, lactic acid also can be directly reduced 10 Propanoic acid by bacteria in fermentation process. However, this reduction path may be unimportant in industry, so only a few papers mention it ‘25). 16 1.2.2. Polymerization Because lactic acid contains both the hydroxyl and carboxylic acid groups, it can undergo self-esterification to form either dilactide or lactoyllactic acid (26). Similar to all esterifications, the latter reaction is acid-catalyzed. The production of dilactide is catalyzed by primarily weak base ‘27). Continued esterification of lactoyllactic acid gives poly lactic acid, but the produced polymer usually has low average molecular weight because of the equilibrium constraint imposed by water concentration. The commercialized route patented by Du-Pont for the production of high molecular weight biodegradable poly(lactic acid) from lactic acid involves two catalytic steps. After the formation of lactide, water is removed by distillation and then a simple catalytic ring- opening polymerization of the purified lactide produces polylactide. Currently, poly (lactic acid) is estimated to cost in the range of $1.00-$l.50/1b. Co-polymerization of lactic acid and diisocyanate has also been examined in the production of poly (ester- urethane) (28). Because the poly lactic acid is a biodegradable polymer, the market prospect is very good due to the ever tighter environmental regulation. The numerous patents (29) about polylactic acid show the demand for cheap biodegradable polymer. 1.2.3. Dehydration The dehydration of lactic acid to acrylic acid has been the focus of many studies in the past due to the demand for acrylate-based polymers. In 1956, a patent by 391m claims the invention of a process in which 68% acrylic acid yields are obtained at around 400°C using sulfate and phosphate catalysts. It has been known that phosphate salts are commonly used in catalytic dehydration, but no complete mechanistic explanation has been proposed. Monobasic sodium phosphate catalyst buffered with sodium bicarbonate l7 has also been used to obtain 58% acrylic acid yield from lactic acid (3°). Recently, ml; published an in-depth study of lactic acid reactions in supercritical water (31). A mechanism for the dehydration of lactic acid to acrylic acid without catalyst was proposed which involved the leaving of the (Jr-hydroxyl group and the carboxyl hydrogen instead of hydrogen from the methyl group, forming a lactone as an intermediate. They had also identified the decarbonylation of lactic acid to acetaldehyde as the major competing reactions since both go through similar transition states. Further studies on lactic acid conversion in supercritical water using phosphate catalysts resulted in a 58% acrylic acid yield (32’. Extensive conversion studies of lactic acid to acrylic acid over various sodium salts and supports have been performed by @1116; and Langford (33’. It was found that sodium metasilicate and bromate exhibit the highest selectivity toward the formation of acrylic acid. 1.2.4. Condensation to 2,3-pentanedione The condensation of lactic acid to 2,3-pentanedione over basic sodium salt catalysts was discovered by Gunter et a1 (3"), and extensively investigated by Mina”. Their research shows lactic acid obtained from fermentation can be converted to 2,3- pentanedione with acetaldehyde, acrylic acid, and propionic acid as lower value by- products in a fixed-bed, down-flow reactor. Formation of acrylic acid from lactic acid is relatively low at 23% yield over NaOH at 350°C because of the competing decarboxylation reaction to acetaldehyde at high temperature. However, 2,3- pentanedione can be produced in high yield over alkali metal catalysts at temperatures between 280-300°C with yield increasing in the order of Na < K < Cs. 2,3-Pentanedione yield as high as 60% theoretical with a 80% selectivity is obtained over a 2 mmol 18 CsOH/g silica catalyst at 280°C. Catalyst loading also increases the yield to 2,3- pentanedione proportionally up to a saturation limit of 2 mmol of metal per gram of support. Post reaction FI‘IR spectra of these alkali metal catalysts after exposure to lactic acid vapor indicate the formation of alkali lactate as the dominant species on the surface at 260-320°C. He concluded that the anions of initial sodium salts used do not participate in the condensation to 2,3-pentanedione, and the formation of 2,3-pentanedione involves presence of both lactic acid and alkali lactate. Conversion of the alkali salt to lactate is found to be the greatest when a low melting point salt with a volatile conjugate acid is used. The decarbonylation of lactic acid to acetaldehyde can be greatly reduced by using a silica support with low surface acidity. 1.3. Hydrogenating carboxylic acids with molecular hydrogen Because molecular hydrogen is not reactive chemically at low temperature and its solubility in aqueous phase is very low at low pressure (35’, almost all hydrogenation reactions need catalysts and high temperature and pressure. The carboxylic group is very stable to molecular hydrogen even with catalyst at strenuous conditions. The corresponding ester is much easier to hydrogenate to alcohol at relative mild reaction conditions. Therefore, some researchers hydrogenate the ester instead of free carboxylic acid to prepare corresponding alcohols. Before 19603, hydrogenation catalysts were largely confined to heterogeneous systems involving metals, metal oxides and some salts. Since the development of transition metal complexes, which replicate the catalytic properties of the metals, and are effective in homogeneous reaction systems, homogeneous catalysts also attract the attention of many researchers (F. J. Mcquillin (3“). 19 The earliest works of ester hydrogenation date back to the publications of Brown & Adkins ‘37) in 1934. In their work, copper-chromium oxide and nickel catalyst were used to hydrogenate optically active esters and ketones to optically active alcohols at temperature of ~250°C and pressure of 150-200atm. Most esters could maintain their optical activity during the hydrogenation, except for butyl lactate that lost it optical activity after 2 hours hydrogenation at 225°C and 200atm. The paper of m et at (38) in 1947 was about hydrogenation of esters to alcohols over Raney nickel. By using very large catalyst loading, they could hydrogenate esters of alpha amino acids to corresponding amino acohols at 50°C and 150~200atm. The extremely low temperature is advantageous in save the amino group and avoiding racemization. In early 19303, catalytic hydrogenation of carboxylic acids to the corresponding alcohols were accomplished with copper catalysts at temperatures above 300°C, with promoted copper catalysts at temperatures above 240°C, or with cobalt catalysts at temperatures above 220°C and at pressure above 200atm (39). The high temperature lead to very poor alcohol yields. Around 1952, Fwd (4°) and Carnahan at al ‘39) used ruthenium based catalysts to directly hydrogenate carboxylic acids at much low temperature, but they had to use pressure in excess of 500atm for best results. Their typical results are shown in Table 1-4. According to their investigation, the chief side reaction appeared to be further hydrogenation of the formed alcohol. They found byproduct ethanol during the hydrogenation of oxalic acid, and methane and ethane were detected when they hydrogenated hydroxyl acetic acid at 250°C. At some conditions, ester also formed and slowed the reaction. The problems with all of these studies are that the pressure is too high to be practical and/or the yield is very poor. 20 Table 1-4. Typical results of Carnahan et al 09’ carboxylic acid hydrogenation Substrate T (° C) t (hr) P (atm) Catalyst Yield % Acetic acid 147-170 10 700-950 Ru02 88 Oxalic acid 94-170 10.5 630-990 Ru02 47 Adipic acid 150-175 0.5 520-700 Ru02 48 Succinic acid 152-192 4-5 720-950 Ru02 59 Hydroxyacetic acid 145-149 0.16 700-710 Ru/C 83 In Broadbent’s paper (23), in situ preparation of high activity rhenium powder catalyst could reduce formic, acetic, propionic, butyric, capric, lauric, stearic, lactic, maleic, succinic and glutaric acids to corresponding alcohols at 137~286 °C and 150- 325atm. But the highly branched trimethylacetic (pivalic) acid could not be reduced to any recognizable product even at 264°C. Their carboxyl acid results are summarized in Table 1-5. Compared to Camahan’s works, Broadbent lowered the reaction pressure with good yield. The in situ catalyst preparation is a disadvantage. The strenuous reaction conditions used by Broadbent , FA!!! and Carnahan show that directly catalytic ' hydrogenation of carboxyl acids is very difficult even with highly active catalyst. Table 1-5. Broadbent’s carboxylic acid hydrogenation results Acid T (° C) P atm) t (hr) Yield % and products Formic 240 238 12 C02, methane Acetic acid 150 168 10 100 Ethanol Trifluoroacetic 270 300 18.5 100 Trifluoroethanol Propionic 165 252 1.5 92 Propyl alcohol Butyric 150 178 11 89 butyl alcohol Isobutyric 165 156 4 75 Isobutyl alcohol Caprylic 200 188 2 93 n-Hexyl alcohol Capric 137 173 3.5 100 n-decyl alcohol Laurie 160 186 10 100 n-Dodecyl alcohol Stearic 265 245 23.5 43 n-0ctadecyl alcohol Lactic 150 258 8 84 propylene glycol Maleic acid 196 286 12 91 succinic acid Succinic acid 210 245 4 94 (1,4-diol) Glutaric 250 179 50 1001 ,S-Pentanedi aol 21 who: ten: iii. ‘ «Lg: Special neighbor atom (or group) can activate the carboxylate group, so the carboxylic acid can be hydrogenated with catalyst in relatively mild conditions. In the patent of Antons (2") (1998), the carboxylic acids have to have a neighboring group at the a or B position. This group could be linear or branched C1 -C4 -alkyl, benzyl, and hydroxyl. The reduction is carried out in the presence of a solvent for optically active carboxylic acids and the product is optically active alcohol. Examples of suitable solvents are water, water-miscible organic solvents and mixtures of the two. Suitable reaction conditions are temperatures in the range 50 to 150°C. and pressure in the range 5 to 250 bar. The process according to this invention can be carried out continuously or batch wise. The surprising advantages of the process are that it provides access to optically active alcohols in a simple manner, at relatively low temperatures and pressures, at low cost and with a high selectivity (enantiomeric excess, ee, usually over 90%). Miroslav’s patent (1981) (41) shows that fluorine-containing alkyl, cycloalkyl, and benzene carboxylic acids could be hydrogenated to the corresponding primary alcohols with heterogeneous catalysts. The hydrogenation can be carried out in the liquid or vapor phase in the presence of a solid rhodium or iridium catalyst employed as the metal, metallic oxide, or mixture. In the liquid phase, the hydrogenation can be carried out in batch reactor at 50-150°C and 5-15 atmospheres. The main purpose of this invention is to hydrogenate trifluoroacetic acid in the liquid phase to 2,2,2-trifluoroethanol, which is an intermediate in the synthesis of the anesthetic, isoflurane, CF3CHC10CHF2. Some catalysts can hydrolyze or-hydroxyl group while saving the carboxyl group at the same time. A 1987 patent granted to ¥e_le_nfl' et al. describes a catalytic process for the conversion of a-hydroxyl carboxylic acid to aliphatic carboxylic acid and aldehyde 22 the silica nirogcn l WC an: acid from showed a Silppress t I-l-l- H61: 111 Ira] oxid 3.111? Cam Wide 54 11%)de hi‘mgenat Ofajd Will efime ca trim are 6 who] OT 2 (’2). The catalysts used have the formula MaM’bOX. The catalyst is prepared by impregnation of 30 % metal oxides onto 70 % silica (or silica-alumina support). Because of the complexity of the catalyst system, the exact structures of these oxide complexes on the silica support are not known. The feed (~26 wt % of lactic acid) is carried by nitrogen through a down flow, fixed bed reactor. Using MosCu4Sn0x on silica-alumina, mg was able to obtain a 64% propionic acid yield with a total conversion of 99% at 350°C and atmospheric pressure. Mag (22’ also investigated the formation of propionic acid from lactic acid over molybdenum and other mixed transition metal catalysts, and showed a 28% yield at 350°C and 5 second residence time. But, his purpose was to suppress this reaction to achieve high yield of 2,3-pentanedione. 1.3.1. Heterogeneous catalysts The most commonly used catalysts in organic acid hydrogenation are metals and metal oxides in solid state. Generally, organic acid hydrogenation is not easy, even with active catalysts, and vigorous conditions are still needed for successful reduction on a synthetic scale. At 150 °C and 2000 psig, Rh203 become a useful catalyst for carboxylic acid hydrogenation (’3’. Ruthenium is well known for its ability to promote the hydrogenation of aromatic rings without hydrolysis of hydroxyl groups, and the presence of acid will completely deactivate its ability to hydrogenate C=C bonds. It is also a very effective catalyst for organic acid hydrogenation. Ruthenium dioxide and ruthenium on carbon are effective catalysts for hydrogenation of mono and di-carboxylic acids to alcohol or glycol. High-pressure (5000-10000 psi) and elevated temperatures (130~225°C) have been used in prior research (43’. Yields of alcohol tend to be less than 23 ré Ix“ 0,513 far in, also on $113 obtair. 132 Mars: Will he perfect because of esterification of the alcohol. Near quantitative yields of alcohol can be obtained by mixing ruthenium and copper chromite catalysts to reduce the esters. Hydrocarbon by—products increase if the catalyst is reused or with increased temperature, but decrease with increased pressure. Rhodium or palladium with rhenium also shows synergistic effects (44). Catalysts made from Re207 and Pd (N 03)2 on carbon gave a 97% yield of 1,6-hexanediol from adipic acid. Copper chromite and barium- promoted copper chromite have been used for organic acid reductions but very high temperature (300 °C) are needed (43). Rhenium oxides are also useful in reduction of aromatic acids to alcohols without ring saturation. Strongly synergistic effects were found on substituting half of the Re207 with ruthenium on carbon, and excellent results can be obtained at part attributable to competition of substrate, hydrogen, and solvent for catalyst sites (43). 1.3.2. Homogeneous hydrogenation catalysts While heterogeneous catalysts still dominate the hydrogenation practice in industry, an increasing number of studies and patents indicate that homogeneous catalysts will become increasingly important. Calvin (45) reported the first homogeneous catalyst in 1938. _Igr_rc_hi (4°) found the activation of molecular hydrogen by rhodium (III) complexes in 1939. The finding of adsorption of hydrogen by cobalt cyanide solution at room temperature in an amount corresponding to almost one hydrogen atom per cobalt atom has led to the widely study of homogeneous hydrogenation catalysts. In 1961, 11mm) reported that under mild conditions, aqueous solutions of chlororuthenate(II) were effective catalysts for the hydrogenation of olefins. In 1966, Man (’8) found that aqueous hydrochloric acid solutions containing chlororuthenate(II) complexes (thought to 24 be RuCl42') hydrogenate maleic to succinic acids at 65 to 90 °C and up to 1 atrn hydrogen pressure. The initial blue solutions, produced by titanous reduction of ruthenium (III, IV) solutions, rapidly turned yellow in the presence of unsaturated organic acids and the rate of hydrogen adsorption was proportional to H2 pressure and Rull concentration. m (49) also formd the homogeneous catalysts can hydrogenate organic acids and anhydrides, but not esters. Mcquillin (3°) summarized the homogeneous hydrogenation development in “Homogeneous hydrogenation in organic chemistry” (1976) Currently, the available homogeneous catalyst hydrogenations are to a considerable extent complementary to heterogeneous counterparts. Compared to heterogeneous catalysts, most homogeneous catalytic systems still stay in the scientist’s lab. Homogeneous catalysts are inherently simple chemically and kinetically, much more amenable to detailed study, need even milder reaction conditions than heterogeneous catalysts, and give high selectivity in the hydrogenation of double bonds while saving hydroxyl groups (For example, hydrogenating maleic acid to succinic acid (48’). The mild reaction condition and specially designed homogeneous catalysts have the ability to maintain optical activity during the hydrogenation. There is, however, the practical disadvantage that recovery of hydrogenation products may commonly require chromatography, which makes its industrial utilization very difficult and costly. In addition, the catalytic complex may consequently be effectively lost during the products separation, even though it is likely that this disadvantage may be overcome by introduction of supported homogeneous catalyst. 25 1.3.2 1110 of? Sin; '3 1.3.3. Hydrogenation of carboxylic acid salts Hydrogenating carboxylic acid salts is difficult in aqueous phase because the formed base (hydroxyl) will change the solution’s pH quickly and resists further reaction. Adam et at ‘50) (1952) tried to directly hydrogenate Cd-Ni propionate. The key point for this reaction is removing water, and not all salt can be hydrogenated to alcohol. In their work, 1.44g of sodium propionate was treated with cation exchange resin to convert the salt to the free acid, then 1.7g cadmium chloride and 0.24g nickel nitrate was added. The formed precipitate was carefully dried in an air stream. The reactor vessel containing 1.5 g copper chromite catalyst was evacuated and maintained at a pressure of 50-micron Hg for 4-8 hours to remove final traces of water. It was observed that the presence of moisture at this point markedly reduced yield. The reactor vessel was filled to a pressure of 235 atrn with electrolytic hydrogen and heated to 240°C with shaking for nine hours. The product is 74% propanol in water (94% yield). More data are shown in Table 1-6. The other salt hydrogenation related works are fatty acid salts. Richardson and Taylor (5 ” showed at 133 atrn and 340°C for 3 hours, Cd-N i oleate can be hydrogenated to propyl propionate. That means that the C=C bond is broken. To continue the reduction of ester to the alcohol in one step process, copper chromite catalyst was added to the reaction mixture but the product consisted chiefly of a low boiling hydrocarbon, presumably propane or propylene. These are the only works to investigate carboxylate salt hydrogenation. 26 Table 1-6. Propionate hydrogenation (5°) H2 Pressure Temperature Reaction Products atrn (°C ) time (hour) Propane Propanol 133 340 0.5 12 60 133 340 3 68 15 150 320 1 17 57 150 280 9 2 47 200 270 9 59 30 200 240 9 2 74 235 240 9 <1 92 1.3.4. Hydrogen solubility Because the dissociation of gaseous hydrogen into atoms is endothermic (AH=104 Kcal/mole), its reactivity is very low. The solubility of hydrogen in aqueous phase is very low and not even measurable at higher temperature and atmospheric pressure (see Table 1-7). High pressure favors the dissolution of hydrogen, so high pressures will indubitably increase the reaction rate. Table 1.7. Solubility (ml/mlH20) of hydrogen in water (52’ T(°C) Solubility (1 atm) T (°C) Solubility P (atrn)20 °C Solubility 0 0.0214 60 0.0129 10 0.19 10 0.0193 80 0.0085 20 0.38 20 0.0178 100 0.0000 30 0.57 30 0.0163 40 0.76 40 0.0153 50 0.95 1.3.5. Hydrogenation mechanism Investigation of the reaction mechanism of carboxylic acid hydrogenation in condensed phase with heterogeneous catalysts is rarely found in the literature. In homogeneous catalytic hydrogenation of fumaric acid and maleic acid to succinic acid (hydrogenate C=C bond, not carboxyl), H_alm (48’ used isotopic tracers to investigate the hydrogenation mechanism. He found that the hydrogenation of fumaric or maleic acid 27 0‘. -m. 1.4 with D2 in H20 solution yielded un-deuterated succinic acid, while H2 (or D2) in D20 solutions yielded 2,3-dideutero-succinic acid They concluded that the hydrogen atoms, which added to the double bond, originate from the solvent rather than the hydrogen gas. In addition, at the reaction condition, no isotopic exchange between D2 and H20 was observed, i.e., there was no appearance of HD or H2 in the gas phase. This implies that the uptake of hydrogen by this catalyst system, at least beyond the disassociate stage, is not reversible. There are two possible paths for the carboxylic acid hydrogenation. One is to directly hydrogenate free acid, and another is to hydrogenate the corresponding ester. The second path probably is easier than the first one. flag (””3 work shows that without adding copper chromite, the hydrogenation will stop at ester stage. These possible pathways make the hydrogenation kinetic investigation more complicated. 1.4. Trickle bed reactor (TBR) and modeling Trickle-bed reactors are packed beds of catalyst over which liquid and gas reactants flow cocurrently downward. One of the first practical applications of a cocurrent trickle bed reactor was the synthesis of butynediol (53). Due to the wide range of operating conditions that they can accommodate, trickle-bed reactors are used extensively in industrial practice, both at high pressures (e.g., hydro-processing, etc.) and at the nomral pressures (e.g., bio-processing, etc.)(5"). Currently, some of the well-known applications in chemical processing are hydrodesulfurization of petroleum fractions, synthesis of 2-butyne-1, 4, -diol from acetylene and formaldehyde, and selective hydrogenation of acetylene (Henry and Gilbert (55’). 28 supei plant super react: Vino pack: pulsil liquic Trickle bed reactor is physically similar to the packed bed adsorption column. However, they are different in gas and liquid velocities and in the role of solid phase. In a packed bed, the packing is an inert that has the main purpose of improving the gas-liquid contact, while in the trickle bed, the packing is a porous active catalyst. The range of superficial liquid velocity encountered in trickle beds is from 0.01 to 0.3 cm/s in a pilot plant trickle-bed reactor and from 0.1 to 2 cm/s in commercial reactors. Similarly, the superficial gas velocity based on operating pressure can be from 2 to 45 cm/s in pilot reactors and from 15 to 3000 cm/s in commercial reactors. Sato et a1 (5°) observed the various flow regimes in cocurrent downflow in packed beds using glass spheres as packing. They classified these into three distinct flow pattern: trickling at low liquid rate, pulsing flow at higher gas and/ or liquid rates, and dispersed bubble flow at very high liquid rate and low gas rate. For laboratory scale trickle bed-reactors, catalyst partial wetting is a very uncertainly and yet important parameter. The accurate estimation of catalyst wetting efficiency is essential to determine trickle bed performance (5"). The reaction rate over externally incompletely wetted catalyst can be greater or smaller than the rate observed over completely wetted catalyst. This depends on whether the limiting reactant is present only in the liquid phase or in both gas and liquid phases. For instance, if the reaction is liquid limited and the limiting reactant is nonvolatile, such as occurs in many hydrogenation processes, then a decrease in the catalyst-liquid contacting efficiency reduces the surface available for mass transfer between the liquid and catalyst, causing a decrease in the observed reaction rate. However, if the reaction is gas limited, the gaseous reactant can easily access the catalyst pores from externally dry areas, and consequently a 29 higher reaction rate is observed with a decreased level of external catalyst wetting. Of course, the above analysis is based on the assumption that particles in trickle beds are always internally wetted. Kim et at (57) (1981) present criteria that allow one to estimate when internal pore dry-out can occur. Actually, internal wetting occurs in almost all trickle bed reactors. Compared to three-phase slurry reactors, trickle-bed reactors do not need special equipment for the separation of deactivated powder catalyst, so the initial investment of equipment and operation cost is low. In addition, its flow pattern is close to the plug flow (if the bed diameter is not too large), which is very convenient if high conversion is required. The low effectiveness factor of catalyst pellets in TRB reactor results from their large size, and the requirement for high mechanical strength of the pellets to avoid erosion by the liquid reaction mixture, are the two major disadvantages of trickle bed reactors. Modeling of trickle beds always is a frustrating task due to the lack of reliable mass transfer data, although many researchers put many efforts on this issue. Because mass transfer controls the trickle bed reaction in most cases, modeling the trickle bed actually equates to modeling the mass transfer. The complicated three-phase fluid dynamics in a trickle bed makes parameter measurement and estimation very difficult. Since catalyst wetting, liquid holdup, gas-liquid liquid-solid and pore diffusion all depend on the special system and catalyst used, it is very difficult to find generalized correlations to calculate the parameters. Goto and Smith ‘58) gave a systematical investigation of a trickle bed reactor. For the first time, they used a true trickle bed system to measure liquid hold up, gas-liquid 30 mass transfer and liquid-solid mass transfer coefficient. Based on their parameter measurements, a one-dimension trickle bed model was derived and used for trickle bed oxidation of formic acid (59). Recently the residence time distribution was measured and modeled by Stegeman et a1 (°°’. They concluded that the residence time of liquid phase could be well correlated to the Reynolds and modified Galileo numbers. Wammes investigated the hydrodynamics in trickle beds at elevated pressure (on. Their results show that the hydrodynamic states are the same at equal gas densities. In summary, compared to fixed bed reactors, the knowledge of the trickle bed still is in the early stage. Its mathematical modeling and scaling up still are still far from real use. However, its inherent advantage and wide use in industrial practices show that it is a good choice for our aqueous phase hydrogenation investigation. Specially, it is well fitted to our requirements for hydrogenation of organic acids. 1.5. Rationale of this research This research has both scientific and economic significance. The aqueous phase catalytic conversion of biomass-derived lactic acid to propylene glycol is part of the worldwide effort of utilizing renewable carbon resource. This project could provide carbon resource for mankind after the depletion of petroleum and other fossil carbons. Comparing to petroleum pathway of production of propylene glycol, the proposed techniques will produce less toxic waste and be less polluting to the environment. The even more important part of this research lies on its potential industrial value. Even though the market price of lactic acid is 0.7~0.8$/lb (88% food grade) (2”, the purification process occupies over 70% of the total production cost. With the use of new fermentation technology, the unrefined lactic acid could only cost about $0.15~0.20/1b. 31 The market price of 1,2 propylene glycol from petroleum synthesis is $0.6/lb (21). Apparently, if we can use unpurified lactic acid to produce 1,2 propylene glycol, there is a large margin between feedstock and product. It is apparent that this margin is large enough for commercialization. 32 Chapter 2. Equipment and experimental methods The hydrogenation reactions were conducted in a batch reactor (autoclave) and a trickle bed (continuous) reactor. The liquid phase products are analyzed by high performance liquid chromatography (HPLC) and gas phase by mass spectrometer. A Micromeretics Chernisorb 2700 was used to characterize the surface properties of catalysts. 2.1. Reagents The main reactants were lactic acid and hydrogen. For catalyst screening and reaction condition optimization, 85% racemic lactic acid (J. T. Baker) and L (+) lactic acid (Purac) at 88% (weight) water solution were used. Unrefined L (+) lactic acid samples provided by Cargill were used to test catalyst deactivation. Compressed hydrogen gas with 99.999% purity was a product of AGA. All reagents are listed in Table 2-1. 33 .nEu_un_au_rm__nmmmm summer. we C3151”- SIIppori 12.14 Table 2-1. Reagents used in the hydrogention of lactic acid Regent Puril‘ib Use Source L(+) lactic acid 88 Reaction Purac Lactic acid 85 Reaction J. T. Baker L(+) lactic acid 50 Reaction, catalyst deactivation Cargill Lactic acid 85 Calibration Aldrich Propylene glycol 99.9 Calibration, reaction Aldrich HPLC water N/A Solution for reaction J. T. Baker Hydrogen 99.999 Reaction AGA H2+C0+C02 N/A Mass spectrometer calibration AGA Ethanol 99.9 By-products identification Sigma Methanol 99.9 By-products identification J. T. Baker Acrylic acid 99 By-products identification Aldrich Propionic acid 99.9 By-products identification Sigma 1-propanol 99.9 By-products identification Aldrich 2-propanol 99.9 By-products identification Aldrich CG6M N/A Active carbon, catalyst support Yakima CGSP N/A Active carbon, catalyst support Yakima RuCl; hydrate ~49% Ru Preparing catalyst Aldrich Ru-nitrosyl nitrate 1.5% Ru Preparing catalyst Alfa 2.2. Catalysts The catalysts included commercial materials, our laboratory prepared powder catalyst, and granular carbon supported catalysts. All catalysts are metal, metal oxide, or supported metals. 2.2.1. Commercial catalyst samples Commercial catalysts were used in the initial stage of this project. These catalysts were either samples of commercial catalysts or recently developed new catalysts (Table 2-2). Table 2-2. Commercial catalysts Active metal S Manufacturer Pd/C on alumina Alumina Rull940C -new ' ' metal Ru on Titania Alumina Chemical ~99B-l3 & Chromium None United Inc A-7063 (Ni) Nickel None Activated Metals & Chemicals carbon Corporation carbon carbon corporation 1 l Corporation 1 2 3 4 5 6 7 8 9 p—e O 2.2.2. Catalyst preparation The preparation of supported metal catalysts consisted of impregnating, drying and reduction. For powder catalyst preparation, granular active support was ground into powder in a food blender and a 100~200-mesh fraction was collected as catalyst support. 2.2.2.1. Irnpregnation To control the metal loading in supported catalysts, the incipient wetness of support has to be measured before impregnation. First, the support was dried for 5 hours at 100°C and 30 in of Hg of vacuum. Then, HPLC water was added to about 5-gram dried support until the appearance of liquid phase. From the maximum water addition just before the appearance of liquid, incipient wetness could be calculated. The incipient wetness values for the three granular active carbon supports used are shown in Table 2-3. Table 2-3. Results of incipient wetness testing Sgrport Support weight(g) HLLC water (g) Incipient wetness CGSP 4.4 5.0 1.14 g water/ g support CG6M 6.3 9.7 1.54 g water/ g support Nuchar 3.5 6.3 1.80 g water/ g supmrt . 35 Precursor solution was prepared by dissolving ruthenium salt into HPLC water. The weight of salt was calculated from the metal loading requirement and the water requirement was computed from incipient wetness. Then the weighed support was added all at once to the precursor solution in a beaker and the mixture was stirred for 5 minutes to ensure the salt solution was well distributed in the catalyst support. 2.2.2.2. Drying For preparing powder catalysts, the mixture was placed on the rotating evaporator and the heat and vacuum were slowly increased to 80°C and 25 in of Hg. The total drying time was 2 hours. Then the mixture was cooled in room temperature for 5~10 hours. After that, the mixture was transferred to quartz tube reactor for reduction. For granular catalysts preparation, the mixture was dried on a metal sheet in ambient conditions for 24 hours and then placed in quartz reactor under 30 in Hg of vacuum at room temperature for additional 4 hours. After that, the temperature was slowly increased to 50°C in 2 hours and held for 5 hours to complete the drying. The mixture was left in reactor and cooled to room temperature for reduction 2.2.2.3. Reduction The same quartz tube used in granular catalyst drying was the reactor for catalyst reduction. First, the reactor was briefly purged with argon at room temperature. Hydrogen was then passed over the catalyst at 30 mllmin, and the temperature was ramped at 2°C/min to 400°C and held there for 16 hours. Finally, the catalyst was cooled under a helium (or argon) flow to room temperature and passivated by helium with 2% oxygen for 1 hour. 36 During the reduction, effluent gas was monitored by a mass spectrometer. For CGSP catalyst, the formed hydrogen chloride change with temperature is shown in Figure 2-1. The reduction began at about 200°C, and reached to a maximum at 280°C. The high background at 400°C came from the condensed yellow liquid (HC1+H20) in the outlet of the quartz reactor. HCI In effluent gas (Relative concentration) l 0 100 200 300 400 Temperature (C) Figure 2-1. Hydrogen chlorine evolution profile during reduction 2.2.3. Powder catalyst Powder Ru/carbon catalysts were prepared by Paul Fanson in our laboratory. The detailed parameters are given in Table 2-4. 37 Table 2-4. Details of MSU Ru/C powder catalysts Precursor Support: . . . Catalyst Name BET mzlg Drspersron Loading CGSP - G Cameron - Yakima INC. CGSP-200 Mesh 64° ° 9° 50% CG6M - F Cameron - Yakima macroporous-IOO +200 728 13% 5 % SAl35-C Aldrich Silica-alumina, _gr_ade 135. -100 mesh 440 14% 50% 56° ' D Aldrich ruthenium (III) C“. ‘°’°“ ' Yam” INC" 777 10% 4.4% chloride h drate ( ) nucro-porous 100 mesh cosp - A " y “‘1 Cameron — Yakima INC. 648 10% 5 4% CG5P(20*50 Mesh) ' AL100 - B Aldrich alpha aluminum oxide (-100 + 200 Mesh) 0'24 0% 5% ALg - E Alfa gamma alumina Sizez+100 mesh 45 13.5% 4.7% TiP25-Cl9 - J Degussa P25 Titania-200 49 N l A 5 0% Mesh ' CGSP - H Aldrich ruthenium(III) Cameron - Yakima INC. 648 3 % 5 0% chloride hydrate (EtOH) CG5P-200 Mesh ' CGSP-NOl-I Alfa ruthenium nitrosyl Cameron - Yakima INC. nitrate hydrate (egg crisp-200 Mesh 64° 38% 50% 2.2.4. Granular catalyst Granular catalysts were prepared for trickle bed reactor. Three active carbon supports were used in the first batch of catalyst preparation. After three catalysts were tested in autoclave, the catalyst prepared from WV-B (Nuchar) was discarded because its activity was too low. After trickle bed evaluation of CG5P and CG6M, CG5P was discarded because CG6M is apparently superior to CG5P. Therefore, only CG6M support was used in the second batch preparation. 2.3. Batch reactor (autoclave) Batch reactor was extensively used in this research for catalyst selection and optimization of reaction conditions. 38 2.3.1. Reactor system Figure 2-2. Batch reactor and controller The batch reactor used in this project was purchased from Parr Instrument Company. It is a 300-ml mini stirred tank reactor (Model 4561) shown in Figure 2-2. The internal diameter is 3-in and height is 4-in; the total volume (excluding the cooling loop and stirrer) is 300 ml. It is mounted in a bench top stand designed for conducting liquid- gas reactions. The maximum temperature is 350°C and maximum pressure is 3000 psi. The whole reactor is made of T316 stainless steel. A quartz liner also is used for further protection from corrosion at elevated temperature. This reactor is equipped with a gas inlet valve for continuously charging gas into the reactor, and a gas release valve for releasing pressure and gas sampling. A dip tube connected to a liquid sampling valve is used for withdrawing liquid sample from the reactor under pressure without interrupting the reaction. A safety rupture disk provides over pressure protection. A thermocouple is located in the vessel for temperature measurement and control. A stirring shaft with attached impellers (or gas entrainer) is used to suspend catalyst and entrain gas into the liquid phase. A stirrer driving system with a packless magnetic drive and a self-sealing 39 packing gland maintains a gas-tight seal around the rotating shaft. This reactor is controlled by a Model 4852 controller (also from Parr Instrument Company), which provides adjustable stirring speeds and automatic temperature control via electrical control of the heating mantle and the air-cooling loop inside the reactor. The whole reactor system consists of liquid feeding, gas feeding, reactor, controller and sampling tank (Figure 2-3). Frunhydrcgentank Coolhg air % To Mass Figure 2-3. Batch reaction system 2.3.2. Operating procedure for the batch rector Operation of batch reactor included reactor setup, steady state reaction and final liquid & gas volume measurements and analysis. First, weighed catalyst was loaded into quartz liner to avoid direct contact of lactic acid with the steel vessel. A stainless steel screen was put into the inlet of dip tube to act as a filter to keep the powder catalyst in the reactor during the liquid sampling. The reactor then was sealed. After reactor was installed on the stand, gas-lines and stirrer motor were hooked up. Then the reactor was purged by inert and hydrogen. For the safety of operation, the sealed reactor was first purged with inert gas (helium or argon reason), which was slowly filled into the reactor to 500psi. Then the reactor was slowly depressurized to atmosphere. In this stage, the oxygen concentration in the reactor should be less than 0.6 % (volume) if the inert gas was completely mixed with the air in sealed reactor. It should be safe to fill hydrogen into the reactor. However, for safety considerations, the reactor was filled with inert gas one more time and depressurized to atmosphere to make the theoretical oxygen concentration inside the reactor less than 0.02 %. After that, hydrogen was used to replace inert gas to purge the reactor twice to make the inert concentration in reaction less than 0.1 %. The next step was to reduce the catalyst. The role of this pre-reduction was to reduce the metal possibly oxidized during the catalyst transfer. The catalyst pre-reduction was done at stirring speed of 20~50rpm, temperature of 150°C, and hydrogen pressure of 500psi. The temperatrue stabilization needed about 10 minutes and additional 0.5-12 hours was used to complete catalyst reduction. After the pre-reduction, the reactor was depressurized again to atmospheric pressure. The reactant solution was filled into the feeding tank, and then the feeding tank was pressurized to 500psi. With the opening of liquid inlet valve, the solution was transferred to reactor by pressure. Then, the reactor was depressurized again. After about 10 minutes, the reactor was heated to desired temperature. When the temperature was stabilized, reactor was charged with hydrogen to desired pressure. Then the stirring speed was increased to 1000~1200 rpm, which is the time zero for this reaction. One (or half) hour later, the first sample was taken from the sampling loop. Before sampling, the sampling tank and sampling loop (a piece of tubing) was purged by 41 high-pressure hydrogen to clean the residual liquid there. Normally, the sampling valve had to opened three times to collect 1~2ml liquid sample. The liquid sample was then filtered by micro filter and mixed with reference solution for HPLC analysis. Gas samples were taken out the reactor from a needle valve on the top of reactor and continuously fed into the mass spectrometer at rate of 5 ml/min. After 5 (or more) hours reaction, all valves were closed and electrical power was turned off after the last liquid sample was taken. Then the reactor was left in air for cooling. When the reactor temperature was decreased below 40°C, the final gas phase volume was measured by water replacement (if necessary) while reactor was depressurized. After that, the rector was de-assembled and liquid volume was measured. Table 2-5. Experimental parameters Parameter Range Typical Reactant volume 100~150g 120g Temperature 100 to 170°C 150°C Pressure 500 to 2000-psi 1500psi Reactant concentration 5~30% 5%, 10% Catalyst (g)/100g solution 1~4g 1~2g Pre-reduction (temperature) 100~200°C 150°C Pre-reduction (time) 0.5~12 hours 1 hour Reaction time 2 to 10 hour 5 hours 2.4. Continuous trickle bed reaction system A trickle bed reactor system was constructed in our lab for related hydrogenation studies. A gas liquid separator was added to continuously analyze the effluent gas composition. 42 2.4.1. Specifications of the trickle bed system Mass-Flow controller Hydrogen A Inert (He) OIL Reactor Feed " Solution Vang ' 4* I“ Rotameters Back pressure Reoula To Mass QA spectrometer Vent Sampling #tank Figure 2-4. Trickle bed reactor system The reactor tube was 1.57cm ID and 61cm length 316 Stainless Steel tube. Its total volume was 118 ml. Gas flow rate (controlled by mass flow controller) could be 25~500 mI/min (STD) and liquid flow rate (depend on high-pressure liquid pump) could be 0.1 to 10 ml/min. The maximum temperature was 350°C, and the maximum pressure (limited by mass flow controller and pressure gauge) was 1280psi. The liquid distributor was 10cm height and packed with 2mm (diameter) glass beads. 2.4.2. Operating procedure for the trickle bed reactor The first thing was to prepare catalyst column. Stainless Steel screens were used to support the catalyst on the bottom and separate liquid distributor (glass beads) and catalyst at the top of reactor tube. A thermocouple was placed into the catalyst bed to measure and control the reactor temperature. Then the top and bottom of reactor were 43 connected to the tube to complete the preparation of catalyst column. Then, the column was hooked up to the trickle bed system, and all power cords and thermocouple lines were connected. Helium was charged into trickle bed system from bypass valve to 1000psi for leaking testing. Soap solution was used to test the all connecting-points. When pressure could hold for one-half hour (pressure drop less than 10psi), the leaking test was done and reactor was depressurized for catalyst reduction. During the catalyst reduction, reactor pressure was controlled at 800psi by the back pressure regulator. Then, the reactor was heated to 150°C and held for 2~8 hours to complete the in situ pre-reduction. This reduction was done for new catalyst and after 20 hours reaction. After reduction, HPLC water was pumped into trickle bed at lml/min and temperature and pressure were controlled at desired values. When the trickle bed was fully stabilized, liquid feed was switched from water to lactic acid solution and the gas- sampling valve was opened to monitor the gas composition by mass spectrometer. It typically took 90-120 minutes for steady state product compositions to be achieved at most of reaction conditions, so several different reaction conditions could be evaluated over the course of a day. After about 2 hours reaction, the system was fully stabilized, liquid sample was taken from the sampling tank, filtered with a 0.2 micro syringe filter, mixed with internal reference solution and analyzed by HPLC. After finishing all desired experiments, the system needed to be shut down. In shutdown operation, liquid feed was switched from lactic acid solution to pure water for l~2 hours to purge lactic acid out of the reactor, then all gas valves were closed and all power was turned off. 2.5. Products analysis The gas phase was mainly analyzed by a Quadrupole mass spectrometer. Gas chromatography (GC) was used only for verification. High performance liquid chromatography (HPLC), manufactured by Thermo Separation Products, was used for liquid phase analysis. 2.5.1. Mass spectrometer An Ametek Dycor M100M Quadrupole Mass mounted on a vacuum chamber that is capable of achieving a pressure down to 10'°torr, is equipped with an electron multiplier for analysis of species concentrations down to 100ppm, and is interfaced with a personal computer for data collection and manipulation. Product gases from both reactors as well as calibration gases pass by one end of a one-meter long quartz fine capillary tube which continuously draws sample gases to the vacuum chamber, and achieves the final pressure reduction to 104-10'5torr. A system sketch is given in Figure 2-5. The mass spectrum can be displayed on the screen of the controller or interfaced by another computer to collect data and save to files for further analysis. ‘ Mass Mass spectrometer controller computer Molecular pump Figure 2-5. Mass spectrometer system 45 2.5.1.1. Gas phase by-products identification Figure 2—6 and Figure 2-7 are typical gas phase mass spectrums in the final stage of reaction in batch reactor and trickle bed reactor. Compared to the standard mass spectrum (4») of methane (Figure 2-8), it can be seen that methane is one of the major gas products (characteristic peaks are MasslS & Massl6, and Massl6>MasslS). The big Mass 28 peak could be either ethane or ethene. From the relative height of Mass 28 and the small peaks (Mass 26, Mass 27 and Mass 30) (Figure 2—6 and Figure 2-9), we can conclude that the major C2 gas by-product is ethane, but we cannot exclude a trace amount of ethene forming. For C3 gas product, the major component should be propane because the big Mass 29 peak (Figure 2-6 and Figure 2-11) and the small peak around Mass 39~Mass 44. However, we cannot exclude the trace amount of propene because of the peaks at Mass 39~Mass 43 (Figure 2-12). 1.52 -7 - M18 Ml M16 M29 .90 1 M15 . l so1 l Il .30 M40 ‘ M27 M44 M30 Afl M26 M1M32 rrrrrrrr ioITttTIIIvaTvIvvvvvvvmjrvvvvvtiqvofrTT so! Figure 2-6. Final gas phase analysis from mass spectrometer (autoclave) 3a.! -7 1.15 lesi 130 C m 3; MC r ~90. methane ethane prelim -"°a ......... . IL. . .lllmJl/Um 10 20 30 40 50 Figure 2-7. Gas phase analysis from mass spectrometer (trickle bed reactor) ion , MAW m -l m .. Momma “1 ' wuss SPECTRUM m -r no....fi-.,,r,, 12 13 14. 1i 15. 17 ml: Figure 2-8. Standard mass spectrums for methane "’ http:/lwebbooknistgov/ 47 1 91W 4:- Ethane MASS SPECTRUM 100. d OFF ml: Figure 2-9. Standard mass spectrums for ethane 80.1 60 ."l 40.— 20.— Ethane MASS SPECTRUM 24. 28. Figure 2-10. Standard mass spectrums for ethene 48 cu— 32 Prepare Bd‘filbrm MASS SPECTRUM X - 23 Y - 77.5 l m. m. 411.. 2nd ‘ l to.rl..-;..I t....:l.s, 10. 211 30 411 50 ml: Figure 2-11. Standard mass spectrums for propane 100. 80.— Propene 60.- MASS SPECTRUM 40.— 20.— 0. I I I I I 1 I |T l I I l I I l I F l I I I I l T I I 0. 10. 20. z 30. 40. Figure 2-12. Standard mass spectrums for propene 49 In summary, the major by-products in gas phase consist of methane, ethane and propane and may contain traces amount of ethene and propene. The composition highly depends on the catalyst, reaction conditions, and the type of reactor. In Figure 2-6, the Cl , C2 and C3 are almost in the same level; for Figure 2-7 methane peak is higher than that of ethane and propane. In addition, the backgrounds of carrier gas (pure H2) are much less than the peak height of gas sample (Figure 2-13). For Mass 16 (Methane), the background is only 1% of that in most gas samples; for Mass 28, the background is less than 3% of that in gas samples. Therefore, the gas analysis is reliable. Figure 2-14 shows the calibration spectrum and the response of CH4, C0 and CO2 is very close. ll“ Background of pure hydrogen 'YfrvajvvvvvVIYvyvvvvrvvvvjfrrv'vrvvavvvvvlr‘r‘f Figure 2-13. Background of pure hydrogen 50 . CaibrationofCH4.CO and 002 co co2 Figure 2-14. Calibration of CHa, CO and CO2 2.5.1.2. Quantitative analysis Calibration of the mass spectrometer was performed by scanning a blend of AGA certified standard multicomponent gas mixture, which contains 2.05% CO, 2.03% C02, 2.01% CH4, and balance hydrogen. Five scans of pure hydrogen (containing no essential species of interest) were taken to obtain background levels, averaged, and then subtracted from the average of five scans of calibration gases or samples to obtain actual peak values. The mole fractions of key species in the calibration gas were then divided by the corrected mass spectrometer peak values to obtain actual responses (R). Then the gas phase sample was continuously passed for at least 5 minutes to stabilize the mass spectrometer to measure or continuously monitor the gas compositions. For the batch reactor, a flow rate of 5-ml per minute was used to minimize the disturbance in the reaction. 51 For 1 volume Vi . displlttmd bi meter. B. Tile {013 Ill fill the slab) [hill [he I 3 v C‘ Willi For autoclave reactor, final gas composition was analyzed and the total gas volume VT (produced and un-reacted hydrogen) in reactor was measured by water displacement. From these data, we could calculate the quantity of gas by-product moles by Ri = Ci/(H Ci - H0i) Ci=Ri*(H si -H0i) mi = Ci* VT/22.4 R: Response factor I: Methane, ethane and propane. HC Calibration peak height H0 Background peak height m; Gas moles for species i. Effluent gas flow rate vt in the trickle bed (ml/min) was measured by soap bubble meter. By-product flow rate fmi (mole/min) is calculated by fini = 1711' * Ci /22400 The total carbon in the gas phase equals mmethanz/ 3+ 2 methane/ 3 +mpr0pane moles lactic acrd equrvalent for batch finmerhane / 3 + 2 * finer/rang / 3 + finpmpane mole/minute for trickle bed reactor The analytical errors for gas analysis depend on the gas volume measurement and the stability of the mass spectrometer. For spectrometer, repeated experiments showed that the maximum error was less than 2%. The gas volume error in batch was less than 2%, which was introduced by the temperature uncertainty during the volume 52 lllfl’lilllfllll depends 0| ’9 . 3‘) 3:. (7': Ti hmmi 3:1de dc exchange wacmt Thin noble; lumen figs: measurement. The error of gas flow rate in trickle bed was less than 3%, which mainly depends on the stability of the trickle bed system. 2.5.2. High performance liquid chromatography (HPLC) The liquid phase was analyzed by a Spectra Tech P1000 HPLC manufactured by Thermo Separation Products. It consists of an HPLC (mobile phase) pump, UV detector, and RI detector (Figure 2-16). The separation column was an Aminex HPX 87H H” ion exchange type, which was a product of BIO-RAD Company. The column temperature was controlled to 50°C by a heating tape for better resolution and lower pressure dr0p. The liquid sample was filtered before injecting to HPLC to protect the column. The mobile phase was 5 mmole H2804 in HPLC grade water and the other operation parameters are shown in Table 2-6. Typical reaction progress chromatograph is shown in Figure 2-15. 200. 150 j . fl 1004 50 C_fi — ‘ l r — 5 IO 15 20 25 3O Figure 2-15. Typical chromatograph for liquid analysis 53 HPX Colum Mobile phase ’ I A Sample Figure 2-16. Schematic of HPLC system Table 2-6. HPLC operation parameters Column Aminex HPX 87H Mobile phase 5mmH2SO4 in HPLC water Column Temperature 50 °C Mobil phase flow rate ml/min 0.6 Lactic acid peak time (min) 20 Propylene glycol peak time (min) 26 Sucrose 12 Ethanol 26 54 2.5.2.1. Calibration of HPLC An internal reference was used in HPLC analysis. Sugar (sucrose) peak appeared in the beginning of chromatograph, but its peak could split into two peaks in some column and conditions. Ethanol peak is located after lactic acid and propylene glycol, but its position overlapped with some liquid by-products. So, each reference was used at certain conditions. If no liquid byproduct was ascertained, ethanol was used. Otherwise, sugar was used as internal reference. For calibration, Wsc gram standard solution (about 3% lactic acid CW and 3% propylene glycol Cm in water) was mixed with reference solution (W RC gram) to get the calibration chromatography. The peak area was obtained for lactic acid (Am ), propylene glycol (Am) and reference peak (ARC ). From the concentrations and corresponding peak areas, we could calculate the response factors for lactic acid (RM ) and propylene glycol (RM). RM = Cm. stc xi Wk *AIAC Cm stc xARC Rm = W, * Am The exact concentration of reference was not required, but same reference solution should be used for both calibration and actual analysis. The concentration of lactic acid and propylene glycol in real sample could be calculated by WRxAM andCPG=RPGWRXAPG. C =R “ “WSI'A, Ws'rA, The reliability of HPLC analysis was confirmed by analyzing lactic acid and propylene glycol solution with known concentration. When WR xAM IWS * AR vs. C u 55 was plotted for different lactic acid concentration, all data were located on a straight line, which shows the response factor RM was a constant over lactic acid concentration range. The average error was less than 2.6% (see Figure 2-17) for lactic acid analysis. In the same way, we could verify the reliability of propylene glycol analysis, which is shown in Figure 2-18. The average analysis error was 1.8% for propylene glycol. This error was smaller than that of lactic acid because lactic acid has two peaks at high concentration (dilactate). 4.51303 4.0E-O3 — 3.51303 — W. x A“ 3.0E-03 - w, *A, 2.51303 - 2.0E-03 . 1.5503 - 1.01303 - 5.01304 — 0.0E'I'm I I I I I 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Lactic acid (mole/L) Figure 2-17. Response of lactic acid analysis is a constant 56 6.0E—03 5.0E-03 ~ 4.0E—03 - RXPG W: *4n3.01303 - 2.0E-03 - 1.0E-03 r 0 .OE'I'W I I I I I I 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Propylene glycol (mole/L) Figure 2-18. Response of propylene glycol analysis is a constant 2.5.2.2. Liquid compound peak assignment Figure 2-19 is a typical calibration chromatograph with sugar as internal reference. The reference peak is far from lactic acid and propylene glycol, even though a small peak always following the main peak for unknown reason. Lactic acid and products peaks are well separated from each other. The peak assignments for reference, lactic acid and propylene glycol were done by injecting pure compounds. 57 sucrose lactic acrd Propylene glycol l 5 20 25 Time (minutes) Figure 2-19. Typical HPLC chromatograph of liquid products For peaks assignment of liquid by-products, the first step was to list some possible by-product compounds and get their peak positions (residence times, called “real”) by injecting pure compounds to our HPLC. Then their peak positions in the technical reference of HPLC column (called “menu”) were collected. The real/menu ratio, which is listed in Table 2-7, was obtained by comparing the real and menu positions for each compound. This table shows that the ratios are very close to a constant for LA, PG, EG, Acrylic and Propionic acid. Therefore, we can estimate the “real peak” position from “menu” with quite good approximation. Then real/menu ratio was used to convert unknown peak position in our chromatography to menu residence time, by which the compound can be initially assigned from the technical reference of HPIE column. Finally, pure compound was used to verify the assignment. The peak assignments for all possible liquid by-products are listed Table 2-8. 58 Table 2-7. Real peak position and menu value 59 Chemicals Real Menu Real/menu LA 20.5 16.7 1.23 1,2PG 25.5 21.5 1.19 2-Propanol 33.0 28.5 1.16 EG 24.8 20.7 1.20 Acgylic 30.0 25.2 1.19 Propionic acid 28.4 24.0 1.19 Table 2-8. Peak assignment Chemicals Real Menu Convert to menu Convert to real LA 20.5 16.7 20.2 1,2 PG 25.5 21.5 26.0 1,3 PG 22.2 26.8 l-Propanol 39.0 32.5 2-Propanol 33.0 28.5 27.3 EG 24.8 20.7 24.8 Ethanol 27.0 32.4 Methanol 23.8 28.8 Acrylic 30.0 25.2 25.0 Propionic 28.4 24.0 23.5 2.5.3. Catalyst characterization The BET surface and dispersion of catalyst were measured by Micromeretics Pulse Chemisorb 2700. Total surface area (BET) was measured by nitrogen adsorption at 77 K and analyzed with the BET method. Twenty to fifty milligrams of catalyst was loaded into a quartz sample tube which was sealed to the Chemisorb 2700 and heated to 423 K for 20 minutes to drive off any weakly adsorbed species, most of which was water. After calibration of nitrogen, a continuous flow of 5% nitrogen in helium was passed over the sample and the effluent gas composition was tested for nitrogen concentration. First a liquid nitrogen bath was used to cool the sample to 77K to get the adsorption profile, then liquid nitrogen bath was removed and the sample was heated to room temperature (by air) to get the desorption profile. The same procedure was followed for 10% nit given Cl surface measun lllCflSUll 10% nitrogen and 18.75% nitrogen in helium. The volume of nitrogen adsorbed at a given composition along with pressure was used in BET analysis to determine catalyst surface areas down to the micro pore level. The metal dispersion on catalyst support was measured by hydrogen adsorption using same instrument using the method of BET measurement. These analyses were finished by Bryan Hogle. Chapter 3. Lactic acid hydrogenation in autoclave This first part of the research includes catalyst screening, experimental condition optimization, testing MSU catalysts, hydrogenation of lactate and development of carbon balance. 3.1. Commercial catalyst testing The first thing for this project was to choose suitable catalysts. We know transition metal, supported metal and metal oxide have the potential to be good catalysts for carboxylic acid hydrogenation. Therefore, the first step was to obtain these kinds of catalyst samples from manufacturers. 3.1.1. Catalyst screening In the first batch of experiments, we emphasized lactic acid conversion and looked for catalyst with the highest activity. These samples are commercial catalysts or 61 llldUSll 11s to ranged 10300 loading are 113‘ Com'Ersi SUPPOIIS [lie Slippy “who. “ion-tn industrially developed new catalysts, which were listed in Chapter 2. Catalyst screening was conducted in the Parr autoclave with a stirring speed of 1200 rpm. Reaction pressure ranged from 1000 to 2000psi (7.2~l4.5MPa) and temperature was increased from 100°C to 200°C until a reasonable conversion was reached. For all experiments, the catalyst loading was lgram (as received basis) in 100 gram 5 % lactic acid in water. The results are visualized in Figure 3-1. Nickel/Alimina Pd/C (EC) Ru on alumina CuCrG-99B-l3 Ru/RB carbon A-7063 (Ni) Ru on Titania Ru on alumina G} Ru on alumina (9% Ru on alumina (J Ru/CP Ru on alumina Ru/C (EC) Ru/C(PMC) Ru/#l940C conversion Figure 3-1. Five-hour conversion (temperature and pressure in parentheses) The results clearly show that only supported ruthenium (Ru) can give reasonable conversion after 5 hours at such mild conditions. Ru on carbon and Ru on alumina supports showed almost the same performance. The problem with Ru on alumina is that the support degraded into very fine powder and formed an emulsion phase after 5 hours reaction. It is very difficult for the fine powder to precipitate. On the contrary, active carbon-supported ruthenium could survive under fast stirring and separated from the 62 product solution in a short time after reaction. So, if we use a powder catalyst in the stirred tank reactor for lactic acid hydrogenation, Ru/carbon is the best choice. From Figure 3-1, we also qualitatively know that the hydrogenation reaction rate increases with temperature and H2 pressure for the same catalyst. High temperature leads to high conversion at the same reaction time (Ru/Alumina DC). For carbon supported ruthenium catalysts, different carbon supports have different activity. For example, the activity of Ru on RB carbon support (CCC) is much lower than that of Ru on 1940C (PMC). 3.1.2. Temperature and pressure optimization (Matrix -1) On the base of catalyst screening, an optimization matrix was designed to identify optimal reaction conditions. The experiment conditions are given in Table 3—1. Table 3-1. Optimization Matrix-l Pressure Temperature 500 Psig 1000 Psig 1500 Psig 2000 Psig (3.4MPa) (6.9MPa) (10.3MPa) (13.8MPa) 130°C M14 M2 M12 M10 150°C M1 M13 M1 1 M9 170°C M8 M7 M6 M3 From catalyst screening, Ru/ 1940C from PMC, which was 5% ruthenium on active carbon with 63% H20) was the best catalyst. In this matrix, the catalyst loading was fixed at 1 gram (as received basis) catalyst in 100 gram 5% Purac lactic acid in water. For all runs, the catalyst pre-reduction was conducted at 170°C and 300psi (2Mpa) Hz for 30 minutes. The reaction time ranged from S to 9 hours. The stirring speed was fixed at 1200 rpm for all runs. The lactic acid conversions are visualized in Figure 3-2. It can be seen that the three highest S-hour conversions come from the runs at 150°C. In addition, 63 Figure 3-2 shows that the optimal reaction conditions for conversion should be located around 150°C. In the temperature range used, high pressure always favors the reaction rate. N. . “5‘ ‘3 90%+ 33% to 45% 78% to 90% 22% to 33% 67% to 78% “‘70 to 22% 56% to 67% 0% to 11% 45% to 56% Figure 3-2. Lactic acid conversion vs. pressure & temperature after 5 hours Experiment M8 (500psi and 170°C) was unusual. Although the 5-hour conversion was not as high as that in higher pressure at same reaction temperature, the initial reaction was very fast. Until 2 hours, the conversion was the highest in this matrix. The reasons are not clear yet, but it seems that something enhanced the initial catalyst activity. Methane, ethane and propane were found in the gas phase for all runs. At reaction temperature higher than 170°C, liquid by-products were detectable. The major liquid by- products are l-propanol, ethanol and a trace amount of 2-propanol, which are shown in Figure 3-3. i9 .3 Chromatograph of liquid products , .9 , E6 _ E .3 8 8 8 i ‘6 (u (U 9 _J E 8‘ 0'. i at ‘- l J PG N . l lAl _ l l A | 10 15 20 25 30 35 40 Figure 3-3. Liquid by-products distribution after 3 hours at 180°C and 2000psi Table 3-2. Experiment numbers for Matrices 2 and 3 T=150°C Matrix-2, Catalyst loading (50% moisture) Pressure 1 2 3 1 OOOpsi 99-9 99-7 99—8 1500psi 99-1(99-15) 99-2 99-3 2000psi 99-10 99-6 99-1 1 Matrix 3, Catalyst loading (50% moisture) T=130°C 1 2 3 1000psi 99-16 99-18 99-19 1500psi 99- 14 99-17 99-20 2000psi 99-12 99-13 99-22 3.1.3. Catalyst loading effects (Matrices 2 and 3) To investigate the catalyst loading effects, two matrices were designed. The catalyst used was PMC 5% Ru/C with 50% water (new PMC). The catalyst pre-reduction 65 3.1.3. Catalyst loading effects (Matrices 2 and 3) To investigate the catalyst loading effects, two matrices were designed. The catalyst used was PMC 5% RulC with 50% water (new PMC). The catalyst pre-reduction conditions were fixed at 500psi in hydrogen and 150°C for 12 hours. Reaction temperature was fixed at either 130°C or 150°C. Reaction pressure ranged from 1000 to 2000psi and catalyst loading was l~4gram in 100 gram 10% J. T. Baker lactic acid water solution. Table 3-2 is the conditions of these two matrices. The catalyst used in Matrix 2 and 3 was different from Matrix 1. The active metals are the same, but the active carbon support is different. So, we cannot compare these two Matrices with Matrix-l directly. The results of Matrix 2 (150°C) are shown in Figure 3-5. For Matrix 3 (130°C), the results are shown in Figure 3-4. These data show that this RuIC is also a good catalyst for hydrogenation of lactic acid. Over 90% selectivity and 97 % conversion can be achieved at optimal reaction conditions. Once again high pressure increases the reaction rate and enhances the selectivity at the same catalyst loading. As expected, lactic acid conversion linearly increased with loading at low temperature and low pressure because the conversion was low and reaction was controlled by the availability of catalyst surface (Figure 3-4). For high temperature and high pressure, it showed the similar trend in the first few hours because the lactic concentration was high and the reaction was still controlled by the availability of catalyst. But at high conversion, as the lactic acid concentration was getting lower and lower, lactic acid diffusion became more and more important, so the catalyst loading effect became less and less significant (Figure 3—5). 90% 80% 4 70% - 60% 4 50% ~ 40% 30% 20% 10% 0% r Conversion +t=1-hr +t=2—hr +t=3-hr ,, —I—t=4-hr —e—t=5-hr ‘ ’ C + T=1 30°C, P=1000psi 3 l l 5 2 2.5 Catalyst loading (g/100g) Figure 3-4. Catalyst loading effect at low temperature and pressure 1 00% 90% - 80% - 70% - 60% 50% 40% 30% 20% 10% l Conversion 0% T=150°C, P=2000psi 1 .5 2 2.5 3 Catalyst loading (g/100g) Figure 3-5. Catalyst loading effect at high temperature and pressure 67 The effect of catalyst loading on selectivity is much more complicated than conversion of lactic acid. It is an indication of how the catalyst loading affects the side reaction. For high temperature (150°C) hydrogenation, selectivity slightly decreased with catalyst loading at high pressure, which is shown in Figure 3-6. Although these decreases are within the analytical error, this trend was repeatedly shown at other reaction times. The decreasing trend in selectivity with catalyst loading is easily understood because the high loading favors side reactions. Here, we choose the 3-hr selectivity for comparison because the lactic acid and propylene glycol concentrations at this time are about half of the maximum, and so we can analyze the liquid sample with the highest reliability. 100% 95% _ +1000psi +1500psi +2000psi 1? F 0,? 90% - + ‘5 0° .— '— E 85/ .2 g 80% - c3 75% 70% I I l l 1 1.5 2 2.5 3 Catalyst loading (g/1 009) Figure 3-6. Selectivity changes with catalyst loading at 150°C The most unusual thing is the selectivity trend at low pressure (1000psi). It almost linearly increases with catalyst loading. This is also related to the low selectivity at low pressure, from which we can get a clue about the reaction pathway. At low temperature (130°C), the conversion was low and the loading effect is much more predictable. When the loading changes from 1 to 2 gram/ 100g, the selectivity was 68 maintained unchanged. But when the loading further was increased to 3 gram/ 100g solution, the catalyst favored extensively the side reaction and thus selectivity went down (Figure 3-7). 100% 95% 90% Na. 85% 'i 80% r 75% r 70% . i I 1 1 .5 2 2.5 3 catalyst loading (9/1009) +1000psi —l— 1500psi + 2000psi Selectivity at t=3-hr Figure 3-7. Selectivity change with catalyst loading at 130°C 3.1.4. Catalyst pre-reduction effect The pretreatment of catalysts will affect the activity substantially. Figure 3-8 shows that 30 minutes pre-reduction doubles the initial reaction rate. Over a 20% difference in conversion still exists, even after 5 hour. That means that part of the active sites cannot catalyze this reaction if the catalyst is not reduced before the reaction. 69 100% 80% e C g 60% ~ 0) > 5 40% ~ 0 + Pre-reduced catalyst 20% - —I— No pre-reduction 0% I i ' l f o 1 2 3 4 5 6 Reaction time (hour) Figure 3-8. Pre-reduction effect 3.1.5. Lactic acid concentration effects To investigate possible hydrogenation of high concentration lactic acid, 10% and 30% lactic acid were used in run M15 and M16. The catalyst loading was 1 gram of 5% Ru/C per 100g solution. Standard temperature and pressure (150°C and 2000psi) were used. The comparison with 5% lactic acid is summarized in Figure 3-9. At the same catalyst loading, high lactic acid concentration will lead to lower conversion at the same reaction time. The actual reaction rate (mole lactic acid/hr.gcat) for high concentration lactic acid was higher than that of low concentration lactic acid, but the conversion vs. time curve was low. Therefore, this reaction is not a first order reaction because conversion vs. time curve depends on reactant concentration. When using 30% lactic acid, with long enough reaction time (13.5 hrs), we still could get 85% conversion. From the trend of conversion curve (M16), we also could say that the catalyst still had activity after 13.5 hour because the curve still went up. That means that catalyst activity can be maintained for a long period of reaction. 70 100% A o ‘ A . 80% - . A . O A. {I g l. “a 60% r . . h l. “2’ ‘ O o 40% ~ 0 o 3 . 200/ _ o OM9(5% LA) ° ‘ o AM15(10% LA) 0 M16(30°/o LA) 0% I l l l l I O 2 4 6 8 10 12 14 Thne(mn Figure 3-9. Reactions at different lactic acid concentration 3.1.6. Gas product evolution The continuous gas phase monitoring in the autoclave (Figure 3-10 and Figure 3-11) shows that most gas products were formed in the first two hours of the hydrogenation reaction. One possible reason is that most by—product gas comes from the extremely active sites, which may exist in fresh catalyst and disappear after several hours of reaction. How these highly active sites deactivate is not yet clear. Metal leaching, poisoning from the products, and collapse of the support during the stirring all could contribute to the deactivation of extremely active sites. The gas product composition strongly depends on the catalyst support. For RulAlumina, the dominant gas by-product was methane; for PMC Ru/1940C (e) we got almost equal amounts of methane, ethane and propane at the same temperature and pressure (See Figure 3-10 and Figure 3-11). This indicates that support plays an ‘ We used two PMC catalyst samples, the first one is Rul1940C and second one is Ru/C-new 71 important role in this catalytic hydrogenation. Even for active carbon supports, different gas by-products may be produced for different active carbons. The gas product composition of PMC Ru/C-new was different from that of PMC Ru/1940C (see Table 3-6). 72 Relative concentration Relative Concentration ...... ......... : _: __ ..... ,/"”'/ CH4 If". ........... C2H6 4'” - - -C3H8 - " Ru/C PMC (old) 20001>si and 150°C 1 5 2 . . 3 Reaction time (hour) Figure 3-10. Gas product evolution for Ru/C (PMC) - - - CH4 wmczl-Ie , — - C3H8 I" v 5% Ru/ alumina Degussa .' 2000psi and 150°C a.. nan—’--~,’—.---.‘ ~--~-..---~ u" , r amWaW" "‘ ;~,: 3mg». .Wm.~«mr-mw I M I WW ’ 1 2 3 4 5 Reaction time (hour) Figure 3-11. Gas product evolution for RulAlumina (Degussa) 73 3.2. Lactic acid conversion over laboratory prepared catalysts This section focuses on hydrogenation over the catalysts prepared at MSU. These catalysts are supported ruthenium on active carbon, titania, or alumina. The precursors are ruthenium (Ill) chloride hydrate and Alfa ruthenium nitrosyl nitrate hydrate. Different reduction schemes, precursors, and supports led to different ruthenium loadings and dispersions. The specifications are shown in Chapter 2. The reaction conditions are given in Table 3-3. Table 3-3. Details of MSU ruthenium catalysts and test conditions No Exp. No Catalyst No P(psi) T(C) Dispersion Loadingfiulcatalyst) 1 M53 CGSP - G 2000 150 6 % 5.0% 2 M52 CGSP - H 2000 150 3 % 5.0% 3 M44 CG6M - F 2000 150 13% 5 % 4 M54 SAl35-C 2000 150 14% 5.0% 5 M57 SG6 - D 2000 150 10% 4.4% 6 M56 CGSP-NOl-I 2000 150 38% 5.0% 7 M55 CGSP - A 2000 150 10% 5.4% 8 M58 TiP25-Cl9 - J 2000 150 N/A 5.0% 9 M59 ALg - E 2000 150 13.5% 4.7% 10 M60 AL100 - B 2000 150 0% 5% The reactions were conducted with 100 gram 5% lactic acid in water with l-gram catalyst m. The catalyst was pre-reduced at 400 psi H2 and 150°C for 1 hour before adding solution. The final conversions and selectivity are given in Table 3-4. For run M60 (Table 3-3), only 5-hour data is shown in this table because the AL100 - B supported ruthenium has zero dispersion and the 5-hour conversion was less than 10%. These runs show those laboratories prepared catalysts are as good as commercial samples. Selectivity as high as 92% of the theoretical was observed with nearly complete lactic acid conversion. With twice the loading, the performance of these catalysts was ' The MSU catalysts contain no water, which are different from commercial one (over 50% P120), 74 close to the best commercial catalyst (Ru/C PMC). The interesting thing is the relationship between the performance and dispersion, which is shown in Figure 3-12. Low dispersion definitely is not good for this reaction (N01, 2, 10), but the highest dispersion does not mean the highest activity (N o 6). A moderate dispersion (around 10~15%) is required for both good conversion and selectivity. Table 3-4. Conversion and selectivity No Exp. No Support l-hr 2-hr 3-hr 4.11: 5-11: 6-hr 7-1.: 8-hr Dispersion 6 M56 CGSP-NOl-I 82:22 :3 :2, g g; a: 38% 7 M55 CGSP-A $611.53. 33 S 33 33 3; 3 33 10% Con.% 26 49 7o 78 84 90 100 4 M54 SA135-C Select. % 92 9O 93 96 94 14% ‘0 N 91 O 1 M53 CGSP-G Con.%. 18 29 41 47 65 6% Select. % 67 69 71 77 75 Con. % 7 26 36 44 56 o 2 M52 CGSP'H Select. % 66 61 66 66 69 3" Con.% 35 74 90 96 99 , 3 M44 CG‘SM'F Select.% 83 66 91 92 92 13" . Con.% 26 55 76 91 96 8 M58 “1,2509" Select. % 77 91 67 91 94 "IA Con. % 9 37 63 79 91 . 9 M59 Al‘g'E Select.% 90 69 83 61 61 13'5” Con.% 9 10 M60 AL100-B Select. % 76 0% ‘ the metal loading actually is equal to 2 gram commercial catalyst 75 100% 90% — 80% - 70% — 60% d 50% ~ 40% — 30% ~ 20% ~ 1 0% - 0%- _ 1.. 1 2 3 4 5 6 7 8 9 1 0 I Conversion 65% 56% 99% 84% 96% 93% 91 % 96% 91 % 9% El Selectivity 75% 89% 92% 94% 91 % 88% 82% 94% 91 % 78% ! Dispersion 6% 3% 13% 14% 10% 38% 10% N/A 14% 0% Con., selectivity & dispersion Figure 3-12. Five-hour conversion and selectivity for MSU catalysts The gas by—products depend on both the type of support and the dispersion of catalyst. No simple relationship between dispersion and gas phase composition can be found. Figure 3-13 shows gas by product distribution vs. dispersion for MSU laboratory prepared catalysts. Just like commercial Ru/C catalysts, no apparent liquid by-product was found since we only used low reaction temperature (150°C). The change of selectivity with time is an indication of how the side reactions occur over different catalysts. This issue will be further discussed in Chapter 6. 76 I LA->CH4 I LA->CZH6 I LA->C3H8 II Total 3% 6% 10% 10% 13.5% 14% 38% Ddispersion Figure 3-13. Gas by-product distribution after five-hour reaction 3.3. Lactate salt hydrogenation As we lmow the direct hydrogenation of lactate salt in aqueous phase was not reported in literature and is not favorable thermodynamically. Our experiments have verified this unfavorability. Experiments conducted at 150~l70°C and 2000psi over 5% Ru/C (PMC) for five hours showed that calcium and potassium lactate did not produce any detectable propylene glycol, and only ammonia lactate give a barely detectable peak in liquid HPLC chromatograph. The experiments in this section are summarized in Table 3-5. 77 Table 3-5. Summary of all experiments for lactate salt hydrogenation T P Conversion % at hour . . . No °C Psig 1 2 3 4 5 6 7 Reaction specrficatron M3 170 2000 17 36 52 66 78 5% lactic hydrogenation atl70°C M4 170 2000 17 33 53 69 79 5 % LA, adding KOH to K‘lLA =0.02 M5 170 2000 17 31 45 56 63 75 5 % LA. adding KOH to K‘lLA =0.1 M9 150 2000 ll 33 58 78 91 5% Lactic hydrogenation at 150°C M17 150 2000 19 37 46 53 67 KLA, , add H2804 to HILA=1 at 1 hr M18 150 2000 12 30 42 52 63 LA +K2804, K‘lLA ratio =0.6 M19 150 2000 12 33 48 62 75 LA +K2804, K‘lLA ratio =0.3 M20 150 2000 0 0 14 29 44 6O 74 CaLA, add H280. to HILA =1 at 2 hr M21 ~170 2000 No reaction KLA+700 psi C02 M22 ~17O 2000 No reaction CaLA+ 800 psi CO2 M26 150 2000 21] 411 581 76L 92] [ CaSO4 saturated lactic acid WKLA is potassium lactate. CaLA is calcium lactate. Their mole concentrations equal to 5% Lactic acid. LA is lactic acid. K‘lLA is potassium ion and lactic acid mole ratio. MA is free H’ and lactic acid mole ratio 3.3.1. Potassium lactate hydrogenation After the failure of direct hydrogenation of lactate, hydrogenating acidified salt (without purification) was the next step. First, a low concentration of potassium hydroxide (KOH) was added to the lactic acid to test the potassium ion effect. Figure 3-14 shows the reaction rate (conversion) decreases with the addition of potassium hydroxide. Addition of potassium hydroxide in an amount up to 2% of lactate present did not affect lactic acid hydrogenation up to 5-hour reaction. But higher KOH concentration [10 % mole of initial lactic acid present] slowed the hydrogenation noticeably after 2- hour reaction. The difference increases with time because the free acid concentration becomes lower and lower. These experiments show that K+ only slows the reaction and does not terminate it; with two more hours of reaction, almost the same conversion can be achieved. Adding KOH brings K+ into solution and decreases the free lactic acid concentration at the same time. To distinguish these two factors, potassium sulfate was added to lactic acid. 78 100% c\° 80% ~ P=2000 Psig g T=170 °C a A '8 60°/- 9 a 15 A o A E 40% ~ 3 fi UM4(K/LA=0.017) E 20% - XM3(N0 KOH) ' AMS (K/LA=0.1) 0% I I l l l r I 0 1 2 3 4 5 7 Time (hour) Figure 3-14. K+ ion effect on lactic acid hydrogenation at 170°C 100% El Lactic acid (M9) El 807 X Lactic acid +K2$O4 K/LA=0.6 (M18) ° A Lactic acid +K2$O4 K/LA=0.3 (M19) '3 ‘ °\° X .5 60% ~ E] A 8 g ‘ X g 40% - x 5 Q 20% - 150 °C g 2000psi 0% , . j . . 0 1 2 Time (hour) 4 5 Figure 3-15. Adding potassium sulfate to lactic acid at 150°C 79 Figure 3-15 shows that potassium salt retards the hydrogenation reaction. When the potassium to free lactic acid molar ratio equals 0.3, the 5-hour conversion decreased about 15% compared to no salt addition. For the ratio increased to 0.6, the conversion wereased up to 30% at S-hour. These runs clarify that potassium ions do affect the hydrogenation in the some way and that the effect increases with ion concentration. The mechanism is unclear yet. The second pK. of sulfuric acid is 1.9 and is lower than that of lactic acid (pKa=2.9), therefore, the proton transfer cannot occur. 3.3.2. Calcium ion Calcium lactate could not be hydrogenated to PG at our reaction conditions (the first two hours of M20, see Table 3—5 and Figure 3-16), even with 800psi CO2 over the reaction mixture “’ (M22, see Table 3-5). The purpose of addition of co; was to acidify the lactate salt, but the first pK,l W of H2C03 is 6.4, which is much larger than that of lactic acid (pKa=2.9). Therefore, it did not work. For hydrogenating acidified calcium lactate, the situation is a little different from that of potassium because the solubility of formed calcium sulfate is extremely low (0.2g/100water). That means that very limited concentration of Ca“ exist in the liquid phase and most of the calcium will precipitate in the form of calcium sulfate. In run M20, the initial solution was calcium lactate equivalent to 5% lactic acid in molar concentration. After two hour, no detectable PG was formed. Then an equal molar quantity of H2804 was added to the reactor from the feed tank. The hydrogenation immediately began after the addition of acid, but the reaction rate (the slope in conversion profile) was lower than pure lactic acid hydrogenation because the precipitated CaSOs still stayed in the reactor. The solution became a slurry at 80 the end of the experiment. The slurry might clog catalyst pores and increase resistance to lactic acid diffusion. In M26, lactic acid saturated with C3804 was hydrogenated at 150°C. The results show that the trace amount of calcium ion does not affect the lactic acid hydrogenation at all. Therefore, if we can simply filter out precipitated C3304 before hydrogenation, the calcium ions remaining in solution will not slow down the following hydrogenation reaction. 1 00% A Lactic acid I 30% - e CaLA+H2$O4 (M20) 1 , I - I- -CaSO4 saturated LA , fl .5 60% -* Linear (Shifted M20)- , ’ . e I . o 40% 7 O O 20% -« M20: Add H2804 at two hours 0% O C , , I ‘ ° 1 3 4 5 6 7 Time(hr) Figure 3-16. Calcium ion effect at 150°C In summary, direct hydrogenation of lactate salts is impossible at our hydrogenation conditions and with our catalyst. The potassium and calcium metal ion in lactic acid solution will retard hydrogenation reaction. The effect depends on ion concentration. Low concentration (<2%) potassium and calcium ion do not poison the catalyst and slow down the reaction. For calcium lactate, after acidifying and simple filtration, the solution can be directly hydrogenated without the requirement of further purification. ‘ The idea here is to use CO2 (H2CO3) as acid, the experiments showed it cannot free lactic acid ‘ pic“ is at 25°C and atmosphere, high pressure and temperature may slightly change its value. 81 3.4. Carbon balance To verify the reliability of analytical methods, we have calculated the carbon balance for the experiments of Matrices 2 and 3. We did this balance by combining gas phase analysis from the mass spectrometer with liquid phase HPLC analysis. This calculation needs the total gas phase volume, which includes the gas in the reactor after reaction and the gas removed during sampling. The total gas volume was measured by water displacement after reaction, and gas removed by sampling was estimated. In the sampling process, liquid and gas escaped the reactor together. For each sample, we took out about 1.2 ml liquid. The sample loop is about 2.5 ml, and we need to fill the sampling loop 3~4 times. That means that we lost about 50m] gas (at reaction pressure) for all sampling (5~7 samples). Because the gas phase by-products change with reaction time and we only analyzed the gas phase after the reaction for most of experiments, this method will introduce a maximum error of 5% if we assume the by-products concentrations linearly increase with time. This is the biggest source of error in carbon balance. The carbon balance is defined as: Carbon lost from liquid (HPLC) - Carbon recovered in gas (mass spectrometer) Inital carbon in liquid The data and calculation are summarized in Table 3-6. The overall carbon balance in these reactions closed to within 14%, an indication that the experimental results are extremely reliable. Table 3-6 also includes a very low pressure run (330psi), which produced lot of gas methane, ethane and propane after 12 hour. The major gas product was methane for the runs at 150°C, and the catalyst loading seems only to increase methane formation (Figure 3-17). 82 Table 3-6. Summary of carbon balance Rx Cat. Mass s ometer ks V B - roducts in 2‘? P“ g peliight pea P120 P": At Fax ancentrationgflf psi Hour 9 M15 M28 M29 psi psi psi CH4 C2H3 CaHa 9 1000 8 1 335-07 4.6E-08 125-07 69 931 307 3.6 0.3 1.1 7 1000 4.2 2 3.5E-07 6.0E-08 1.6E-07 69 931 307 3.8 0.4 1.4 8 1000 4.5 3 4.05-07 1.3E-07 1.3E-07 69 931 307 4.3 0.9 1.2 15 1500 6 1 4.3E-08 3.3E-08 3.3E-08 69 1431 307 0.5 0.2 0.3 5 o 1500 4.2 1.5 1.4E-07 6.0E-08 5.0E-08 69 1431 307 1.5 0.4 0.4 2 08 1500 9.8 2 215—07 605-08 5.0E-08 69 1431 307 2.3’ 0.4 0.4 3 '11 1500 5.3 3 4.3E-07 3.25-08 3.2E-08 69 1431 307 4.6 0.2 0.3 4 I" 1500 3.8 4 5.1E-07 4.0E-08 4.0E-08 69 1431 307 5.5 0.3 0.4 10 2000, 4.7 1 1.3E-07 1.2E-08 1.0E-08 69 1931 307 1.4 0.1 0.1 6 2000 4 2 1.2E-07 4.5E-08 4.0E-08 69 1931 307 1.3 0.3 0.4 11 2000 4 3 9.8E-08 6.0E-08 555-08 69 1931 307 1.0 0.4 0.5 23 330 12 2 1.8E-06 1.2E-06 3.1E-O7 69 261 307 18.9] 7.9 2.8 16 1000 8.3 1 8.0E-08 1.25—07 7.1E-08 39 961 307 0.9[ 0.8 0.6 18 1000 5.1 2 755-08 9.6E-08 4.5E-08 39 961 307 0.8l 0.6 0.4 19 1000 5.1 3 9.1E-08 1.1E-07 6.6E-08 39 961 307 1.0 0.7 0.6 14 9 1500 7.4 1 2.2E-08 1.8E-08 1.8E-08 39 1461 307 0.2 0.1 0.2 17 8 1500 6.3 2 1.2E-08 2.4E-08 1.4E-08 39 1461 307 0.1 0.1 0.1 20 E 1500 5 3 3.7E-08 1.2E-07 5.0E-08 39 1461 307 0.4 0.6 0.4 12 2000 4 1 1.4E-08 8.0E-09 8.0E-09 39 1961 307 0.1 0.0 0.1 13 2000 6.1 2 5.6E-08 5.05-08 8.0E-09 39 1961 307 0.6 0.3 0.1 22 2000 6.1 3 1.2E-07 1.1E-07 1.9E-08 39 1961 307 1.3 0.7 0.2 Cat. total (1 as Initial carbon come final Ii uid . . Recovered g P” g volumg 9 LA from as phage LA '9“ “mm Carbon psi 9 mole L(STD) Mole Mole %LA Con% 861% %LA Mole % 9 1000 1 0.56 12.5 0.12 0.0139 11.6 72 78 16.0 0.0192 96 7 1000 2 0.56 12.5 0.12 0.0165 13.8 85 82 15.3 0.0183 99 8 1000 3 0.56 12.5 0.12 0.0178 14.8 96 82 17.6 0.0211 97 15 1500 1 0.86 19.3 0.12 0.0050 4.2 71 92 6.0 0.0072 98 5 o 1500 1.5 0.86 19.3 0.12 0.0103 8.6 76 90 7.5 0.0090 101 2 as 1500 2 0.86 19.3 0.12 0.0127 10.6 98 92 8.4 0.0100 102 3 vi.- 1500 3 0.86 19.3 0.12 0.0169 14.0 96 80 18.8 0.0226 95 4 I" 1500 4 0.86 19.3 0.12 0.0203 16.9 99 79 21.0 0.0252 96 10 2000 1 1.16 26.0 0.12 0.0068 5.7 46 88 5.3 0.0064 101 6 2000 2 1.16 26.0 0.12 0.0114 9.5 95 86 13.5 0.0161 96 11 2000 3 1.16 26.0 0.12 0.0127 10.6 97 90 9.5 0.0115 101 23 330 2 0.16 3.5 0.12 0.0224 25.4 69 78 15.3 0.0183 103 16 1000 1 0.61 13.6 0.12 0.0088 7.3 54 90 5.5 0.0066 102 18 1000 2 0.61 13.6 0.12 0.0066 5.5 65 93 4.4 0.0052 101 19 1000 3 0.61 13.6 0.12 0.0083 7.0 79 86 11.3 0.0136 96 14 9 1500 1 0.92 20.7 0.12 0.0026 2.3 53 97 1.6 0.0019 101 17 8 1500 2 0.92 20.7 0.12 0.0024 2.0 81 95 4.0 0.0047 98 20 E 1500 3 0.92 20.7 0.12 0.0100 6.4 64 90 8.0 0.0096 101 12 2000 1 1.24 27.7 0.12 0.0017 1.4 47 91 4.3 0.0052 97 13 2000 2 1.24 27.7 0.12 0.0059 7.6 87 95 4.3 0.0052 101 22 2000 3 1.24 27.7 0.12 0.0134 11.2 93 87 12.2 0.0146 1.0 83 O o\° 5- .5 , '5’ 4~ ‘5 2 3* 9 CH4 8 +02H6 82" . ~~X--03H6 ‘5- o 0 272”“ mg; ___________ 0 1 2 3 4 5 Catalyst loading Figure 3-17. Catalyst loading effect on the gas products atlSOOpsi and 150°C 3.5. Conclusion Studies on batch hydrogenation of lactic acid in the autoclave indicate the following: 0 Only supported ruthenium catalyst is active enough to give reasonable conversion in mild conditions. 0 Lactic acid can be hydrogenated to propylene glycol at around 150°C and 1000~2000psi with carbon or alumina supported ruthenium catalyst. 0 With RulC catalysts, selectivity over 90% at a conversion close to 100% can be achieved at optimal reaction conditions. 0 Direct hydrogenation of lactate salt is impossible in aqueous phase for this catalyst system and at these reaction conditions. Acidified calcium lactate can be 84 hydrogenated to propylene glycol if the precipitated calcium sulfate is filtered before hydrogenation. 0 The carbon balance shows that our analytic methods are reliable 85 Chapter 4. Conversion of Lactic Acid to Propylene Glycol in a Trickle Bed Reactor After successfully converting lactic acid to propylene glycol (PG) in batch reactor, a laboratory scale trickle bed reactor was used to continuously hydrogenate lactic acid to PG. Racemic lactic acid, L+ lactic acid, and unrefined lactic acid were tested in this study. The granular catalysts used in this investigation were carbon-supported ruthenium prepared in our laboratories. 4.1. Control parameters and catalysts in trickle bed reactor The upper limits of trickle-bed reactor system are 1280psi for pressure and 300°C for temperature. Catalysts were prepared by ruthenium salt and activated carbon. The details of trickle bed reactor system and catalyst preparation were shown in Chapter 2. 86 4.1.1. Control parameters Before experiments in the trickle bed reactor, operating parameters, such as gas and liquid superficial velocity were calculated and compared with literature values to ensure our operation is in the range of typical operation of a trickle bed reactor. 4.1.1.1. Liquid superficial velocity Superficial velocity is defined as the velocity in the trickle bed reactor tube without catalyst present. Liquid superficial velocity is calculated from liquid flow rate. The liquid flow rate used is 0.5~4 mllnrin and the corresponding superficial velocity is 0.24~2.0cm/min. The equation used to calculate superficial velocity is d is tube diameter (1.57cm) FL Liquid flow rate mL/min 4F . u L = Zfi (cm/min) { According to Ramachandran (6‘), the commonly used liquid superficial velocity is 0.6cmlmin in pilot reactors and 6~12 cm/min in commercial reactors. Therefore, our trickle bed reactor was operated in the same liquid superficial velocity range as a pilot 13301012 4.1.1.2. Gas superficial velocity Gas superficial velocity is defined as the gas velocity in column without catalyst. The gas flow rate used is 30~500 ml/min (STP) and the corresponding gas superficial velocity is 0.2~4 cm/min at 100°C and 1200psi, which is shown in the Table 4-1. This number is much smaller than that used in commercial reactor (15~300 cm/sec, according to Rarnachandran), because of the high pressure used. 87 Table 4-1. Gas superficial velocity at 1200psi and 100°C Gas flow ml (STD) 30 100 200 300 400 500 Superficial velocity (cm/s) 0.004 0.013 0.027 0.040 0.054 0.067 Superficial velocity (cm/min) 0.2 0.6 1.6 2.4 3.2 4.0 4.1.1.3. Flash vaporization When dry hydrogen encounters the lactic acid water solution, part of the water will flash or vaporize. The maximum extent of flash vaporization depends on steam saturation pressure (temperature), liquid/gas flow rate ratio, and the pressure in the reactor. Figure 4-1 shows the water vapor pressure change with temperature and the extent of flash vaporization at different H2/LA molar ratios and temperature. The maximum amount of water vaporized is less than 0.5% of that fed at our operating conditions (<120°C), so no steam saturator is needed. 3.5% 250 3 0y - - - -H2/LA ratio=2 ' ° ------ H2/LA ratio=4 _ 200 A g 2.5% - — - - H2/LA ratio=6 E 3 Steam pressure CL 8 23% ‘ ‘4' 150 E '9' .' 3 81.5% . 10% lactic acid, ., ._ 100 .6 ‘3 density 1.1 g/mL ,' .5 g 1.0% - I. ' E! o / 3 — v ~- 50 4- ‘“ 0.5% - // 8 0.0% - , 7" f 0 50 100 150 200 Temperature (C) Figure 4-1. Flash vaporization and steam saturation pressure 88 4.1.1.4. Conditions of operation All reactions in the trickle bed were canied out at temperatures of 70~150°C and pressures of 200~1200psi. Lactic acid was fed to the reactor in aqueous solutions of 5~l7.2 w% lactic acid. The liquid flow rate ranged from O.5~4.0 mllmin, giving a weight hourly space velocity (kg lactic acid/kg catalyst/hr) of 0.3~2. The hydrogen to lactic acid feed molar ratio varied from 2:1 to 10:1. 4.1.2. Catalysts First, the prepared granular catalysts were characterized by physical and chemical adsorption, and then tested in the autoclave reactor. Only the ones with good performance in the batch reactor were used in the trickle bed reactor studies. 4.1.2.1. Catalyst characterization Three granular active carbon supports were used to prepare catalysts for trickle bed uses. The procedure described in Chapter 2 was used to prepare 5% ruthenium on activated carbon catalysts. The BET surface areas of both carbon support and prepared catalyst were measured (Table 4-2). Chemisorption was used to measure active metal dispersion and the bond strength between metal and hydrogen. Dispersion is defined as the percentage of active metal accessed by hydrogen and is an indicator of how much metal is available for catalyzing the reaction. Intermediate temperature (the position of the first peak in the hydrogen desorption curve) during the hydrogen desorption is a reflection of how strong the metal-hydrogen bond is (Figure 4-1). The hydrogen desorption peak above 300 °C most likely comes from strong adsorbed hydrogen on ruthenium, which may not involve the hydrogenation reaction and cannot come from carbon support because the hydrogen adsorbed on carbon cannot desorb below 700°C (75’. 89 Table 4—2. Catalyst supports specification Support Size BET (ml/g) Maker or name WV-B #1 14x35 Nuchar CGSP #2 20x50 648 Cameron - Yakima CG6M #3 12x40 728 Cameron - Yakima C o o PLPNP'P ‘mwafih 111111 Desorbed hydrogen 0.05 - CG6M Intermediate temperature T=178 0 0 1 100 200 Temperature (C) l l 300 400 Figure 4-2. Hydrogen desorption profile during dispersion measurement Table 4-3. Catalyst properties No Support BET (mzlg) Dispersion Intermediate temperature (°C) A WV-B 881 7.4% 245 B CGSP 670 8.7% 220 C CG6M 697 5.0% 178 D" CGGM N/A 5.0% N/ A *Catalyst D is prepared by the same support and the same procedure as catalyst C 4.1.2.2. Initial autoclave test of granular catalysts The three catalysts (A, B, C) described in Table 4-3 were first tested in the batch reactor at standard reaction conditions (2gram catalyst/100g 10% solution, 150°C and 2000psi). The conversion profiles for three catalysts are given in Figure 4-3, which shows that the activity of catalyst from WV-B is apparently lower than the other two. Therefore, only the catalysts prepared from CG5P and CG6M carbon supports were used in the trickle bed reactor. Compared to a powder catalyst (CG6M-F, 5% ruthenium on ground CG6M support, shown in Table 2-4), the conversions from granular catalysts are much lower (at same catalyst loading and reaction condition) than that of powder catalyst because of intra-particle mass transfer resistance. 100% 90% - o 70% ~ _ E 60% l 1&8 066M / / '5 50% - ,5 40% - / s 30% /- CE, 20% g / ° 10% ’ /l/./T 0% . . . . . 1 2 3 4 5 Reaction time Figure 4-3. Granular catalysts performance in batch reactor 91 hyd in ll per. dis; low 42 5110 Wit Yeti: The dispersion of CGSP is larger than that of CG6M (see Table 4-3), but the hydrogen-catalyst bond is stronger than CG6M as shown by the temperature of first peak in hydrogen desorption (T=178 and 220°C). Therefore, the two catalysts have similar performance in batch reactor. Although WV-B catalyst has the highest BET area and dispersion, the very strong hydrogen-catalyst bond (T =245°C) makes its reactivity very low as shown in Figure 4-3. 4.2. Trickle bed reaction (racemic lactic acid) Three catalysts charges were prepared for trickle bed reactions. The details are shown in Table 4-4. Table 4-4. Three catalyst charges used in trickle bed Charge 1 Charge 2 Charge 3 Catalyst B C D Support name CG5P CG6M CG6M Ruthenium loading 5% 5% 5% Weight (gram) 30 27 48 Volume (ml) 71 64 1 18 Temperature Temperature Temperature . Pressure Pressure Pressure USCd for testing H2/LA ratio H2/LA ratio Unrefined Adding sulfur sample The first two catalyst Charges (1 and 2) were tested at different temperatures, pressures, H2/LA ratios, and for rate enhancement by sulfur addition. Charge 3 fully filled with catalyst was used to investigate the temperature and pressure effects in trickle bed reactor. 92 4.2 its 4.2 at I wa 301 by ex; lac bal cal CO 4.2.1. Results The racemic lactic acid hydrogenation in the trickle bed reactor with three charges is shown in this section. 4.2.1.1. Charges 1 and 2 Before reaction, the catalyst was reduced at 150°C for 12 hours in pure hydrogen at 250psi. All feeding in this section was 10 % lactic acid (racemic) in water; flow rate was fixed at 1.0 mllmin unless specified. The hydrogen flow rate varied from 50 to 300ml/min to change the hydrogen to lactic acid molar ratio from 2:1 to 7.6:1. The hydrogen pressure was fixed at 1200psi and temperature ranged from 80~140 °C. The experimental results are summarized in Table 4-5. These data clearly demonstrate that lactic acid hydrogenation can be conducted in a continuous mode of operation. The reactivity of CG6M is higher than CG5P at all conditions, which is different from the batch reactor results. We cannot see any difference in autoclave reactions for these two catalysts (Section 4.1.2.2). The reason may be the particle size difference. Because CG6M is smaller than CG6P and the mass transfer is the major resistance in a trickle bed reactor, therefore the activity of CG6M is higher than that of CGSP. Experiment n9 and n10, both run at 120°C, show a selectivity of 86% with a conversion over 95%. The implication of these runs, of maintaining high selectivity at nearly complete lactic acid conversion, is that the downstream separation and purification of PG will be very simple. 93 nhl- - 4-— l I CGSP (30 gram) (Charge 1) A 5: 5a Catalyst Slippon; Picssure it. The [imperal 5mm 01 Table 4-5. Trickle bed reaction summary (Charge 1 and Charge 2) No Temperature (C) HZILA mole Ratio Conversion % Selectivity % N 8 n1 60 2.5 35 77 5 n7 100 2.5 67 67 X n6 120 2.5 62 66 5 n2 60 5.1 31 79 5’ n6 100 5.1 66 66 8 n9 120 5.1 95 67 § n4 60 7.6 36 61 0 n5 100 7.6 62 90 n10 120 7.6 96 65 13-n2 60 2 25 79 :: 14-n1 100 2 50 67 go 14-n2 120 2 67 65 5 13-n3 60 4 23 76 : 13-n7 100 4 54 64 E 14-n3 120 4 9o 62 :0 14-n6 130 4 93 77 :1 14-n5 140 4 97 74 91 13-n4 80 6 22 77 8 13-n6 100 6 64 62 14-n4 120 6 95 63 4.2.1.2. Charge 3 After the successful hydrogenation with Charges 1 and 2, a third charge of catalyst was prepared by filling the trickle bed tube to the top with catalyst D (CG6M support). Then a matrix of conditions was designed at three temperatures and six pressures at fixed H2/LA ratio=4:l. The conversion and selectivity are shown in Table 4—6. The general trends are that conversion and selectivity are sensitive to reaction temperature, pressure, H2/lactic acid molar ratio, catalyst type and catalyst loading. The effects of each parameter will be discussed in the following sections. 94 Table 4-6. Results of trickle bed with 48gram catalyst (Charge 3) Catalyst D (CGSM) (489mm), H2:LA=4:1 T°C Pressure (psi) 200 400 600 600 1000 1200 80 C°""°'s'°" 21.0% 33.5% 43.3% 50.7% 56.0% 63.2% S°'°°”“"" 59.4% 70.5% 76.0% 77.7% 76.6% 76.6% 100 C°""°'s‘°“ 52.0% 70.7% 61.5% 67.6% 91.1% 93.6% S°'°°""’"" 73.6% 76.5% 77.6% 79.6% 60.4% 60.7% 120 C°""°’S'°" 77.4% 66.9% 94.3% 97.7% 96.7% 99.6% S°'°°“""Y 34.1% 42.1% 46.9% 56.2% 63.2% 66.6% 4.2.2. Temperature effect Temperature effect was investigated in Charge 1(CGSP) and Charge 2 (CG6M). Lactic acid conversions increase almost linearly with temperature at 80~120 °C (Figure 4-4) for all hydrogen to lactic acid molar ratios. 95 100% 120 R is hydrogen / lactic c 80% _ acid molar ratio .9 g g 60% - 8 § 40% ‘ § +CGSP(R=2) - Cl- CG6M (6:2.5) 0 3 2‘“ +CGSP(R=4) - A- 066M (R=5) +CGSP(R=6) - o- 066M (R=7.6) 0% l I l l 60 , 100 1 Reaction Temperature (C) Figure 4-4. Lactic acid conversion vs. reaction temperature 100% —I—CG$P(R=2) - 1:1- CG6M (R=2.5) 95% - +CGSP(R=4) - A- CGGM (R=5) +CGSP(R=6) - o- CGGM (6:7.6) 90°/ 1 ,G . g o 0 ~ ~ - ‘ - - _ D ‘2: v . - -‘r - ‘ '5 (I) (D 0. R is hydrogen / lactic acid molar ratio 70% I l l l l 80 90 100 110 120 130 Reaction Temperature (C) Figure 4-5. PG selectivity vs. reaction temperature 96 I 140 p‘ The selectivity change with temperature is very similar to that in the autoclave reactor, with maximum selectivity located around 100~120°C (Figure 4-5) for the trickle bed reactor instead of 150°C in the batch reactor. High temperature and low temperature are not good for selectivity. The existence of an optimal temperature for selectivity is for high and low temperatures both favor side reactions. No clear relation can be found between hydrogen and lactic acid molar ratio and the PG selectivity. High selectivity for CG6M catalyst presumably comes from its support properties. 4.2.3. Pressure effect (Charge 3, fully filled CG6M) The pressure effect was investigated in Charge 3(48-g CG6M catalyst). Feed solution is 8.6% J. T. Baker lactic acid in water, hydrogen to lactic acid molar ratio was fixed at 4.7, and liquid flow rate was fixed at 2 mllmin. Three temperatures (80, 100 and 120°C) and six pressures (200, 400, 600, 800, 1000, 1200 psi) were used. 100% c 80% .9. 2 o . E 60% - / 8 E 8 40% ~ 5% +T=1 20 °C 3 o a -E-T=80 °C 20 /o -e—'r:1 00 °C 0% I l l l 200 400 600 800 1 000 1200 Pressure (Psig) Figure 4-6. Conversion profile vs. temperature and pressure in E com 36160 % i a D 1" 1») VI 2: t. r7 '5 '43 60% ,a 2 3 °/ (9 4° ° +T:120°c °' 0 ¢T=80°C 00/0 I l I i 200 400 600 800 1000 1200 Pressure (Psig) Figure 4-7. Selectivity profile vs. temperature and pressure The selectivity and conversion change with pressure and temperature are shown in Figure 4-6 and Figure 4-7. At high temperature (120°C), even at 200psi, the conversion reaches 80%, but the selectivity at high temperature was extremely low (35%). At low temperature (80°C), the conversion increased with pressure linearly. At high temperature (120°C), the increase became slower because it was already close to 100% conversion and no difference can be seen for the reaction at 1000psi and 1200psi from the point of view of conversion. For selectivity, 100°C is apparent better than 80°C and 120°C, which is consistent with the reaction in Charge 1 and 2. At high pressure (>600psi), the selectivity difference for 100°C and 80°C almost disappears. 4.2.4. Hydrogen/lactic acid molar ratio Hydrogen to lactic acid molar ratio only slightly affects the conversion and selectivity. Figure 4-8 shows conversion vs. hydrogen to lactic acid molar ratio for 98 Charge 1 and 2. At low temperature (80°C), the ratios did not change the lactic acid conversion at all because the lactic acid conversion was low and plenty of hydrogen was available. However, at high temperature (120°C), the rapid reaction (high conversion) made the reaction hydrogen limited at the catalyst surface and bulk liquid. Therefore, a high ratio always enhances hydrogen mass transfer and increases the lactic acid conversion. The selectivity vs. temperature and hydrogen/lactic acid ratio is shown in 3 Figure 4-9. High hydrogen to lactic acid ratio lowers selectivity at high temperature i because the extra hydrogen favors deep hydrogenation (side reactions). Like conversion, selectivity is not affected by hydrogen to lactic acid ratio at low temperature. 100% 90% - 60% - CV 70% - 60% J mleGSPU=80) wD-CGSM (T=80) 50% 1 +CGSP(T=120) ~o— CG6M (T=120) 40% ~ 30% _ D . W1?!" - ' “"5 20% F “' ' "7' 10% I l l l l 2 3 4 5 6 7 8 H2/Lactic acid ratio Conversion of lactic acid Figure 4-8. Effect of molar ratio on conversion 4.} Spa 1in line Selc ml! Cat; lnCn flat. thee 66% - “'7---~-~~::-::_er %%- ‘F“““M»::-MO 64% W\. g 62% 4 if} 60% - _ c.»- “a g 78% k meflmwfi‘ww‘i‘ww " _. 76% ~ 74% . -—-I--CGSP(T=80) WEI-CGSM (T=80) 72% ‘ +CGSP(T=120) ~o— CG6M 0:120) 70% I f r I I 2 3 4 5 6 7 6 H2/Lactic acid ratio (CGSP) Figure 4-9. Effect of molar ratio on selectivity 4.2.5. The effect of changing liquid flow-rate Figure 4:10 is the effect of liquid flow-rate (maintaining the hydrogen/lactic acid molar ratio=4zl) for Charge 1 at 100 °C and 1200psi. The X—axis is WHSV (weight hour space velocity), which is defined as lactic acid feed rate (g/hr) per gram catalyst. The liquid feed is 10 % lactic acid and the flow rate was 0.5~3 mllmin (WHSV=O.1~0.6). The linear decrease of lactic acid conversion with liquid flow rate is expected However, the selectivity also decreases from 90% to 80% when the liquid flow rate changes from 0.5 mllmin to 3 mllmin. The reasonable explanation is excess hydrogen available on the catalyst surface, which enhances the side reactions (deep hydrogenation). The PG output (g/min) vs. liquid flow rate is given in Figure 4-11. PG output increases quickly at the beginning (from 0.5 to 2 mllmin), and then the curve tends to be flat. When flow rate change from 2 to 3 mllmin, the PG output was unchanged because the conversion goes down quickly while selectivity changes little. 100 100% 90% ~ 0 . . 60% ~ . ‘ ' ' ' -O ----- 0 70% J . . 60% 1 - C- Selectlvrty 1: 50% _ +Conversion 5 40% i 30% 4 20% 1 10% 1 0% . . 1 0 0.2 0.4 0.6 0.8 WHSV(hr-1)[ (lactic acid feed(g/hr)/catalyst(g)] Figure 4-10. Conversion and selectivity vs. weight hour space velocity lOfl COIWGI’S IVIty Select 0.35 0.06 0.30 . __ g 0 25 0.05 E g ' -- 0.04 g v 0.20 . v 2:; 015 ~- 0.03 g ‘g ' . . ’ - e- LAinput(g) w E 5 0.10 . e 0'02 E 0.05 - o ’ * PG °"tp‘" (9) ._ 0.01 0.00 . : T 0 0 0.2 0.4 0.6 0.6 WHSV(hr—1)[ (lactic acid feed(g/hr)/catalyst(g)] Figure 4-11. Effect of weight hour space velocity 101 4.2.6. Long time and low concentration lactic acid hydrogenation This experiment used Charge 2 (27g CG6M) and 5.6% lactic acid feed at 100°C and 1200psi. Hydrogen/lactic acid molar ratio was maintained constant at 4:1; liquid flow rate was 1 mllmin for 27 hour and 0.3 mllmin for 67 hour. Total reaction time was 94 hours and the total lactic acid feed was about 3000ml. The selectivity almost was unchanged during this period, and low flow-rate led to high conversion (Figure 4-12). Catalyst deactivation is not a detectable in this extended time reaction. 100% -—1 A t a; 95% ~ ‘ ‘5 90% - A m g 60% « g 75% 4 Elselectivity “g 700/ - Aeonc. at 1ml/min 9 ° Aconc. at 0.3mein C 65% - 8 600/0 I I I l 0 20 40 60 80 100 Reaction time (hr) Figure 4-12. Conversion and selectivity of extended time reaction 4.2.7. High lactic acid concentration feed Using high lactic acid feed concentration has the potential to improve productivity and reduce the quantity of water that has to be handled in this process. Nevertheless, the corrosion problems of higher lactic acid concentrations are always technical challenges in industry. Therefore, 17.2 % lactic acid was used to show the concentration effect. Reactions were conducted at 1200psi and 100°C with 1 mllmin liquid feed rate with Charge 1(30gram CGSP). The hydrogen to lactic acid molar ratio was l.16~5.83:l. The 102 results are shown in Table 4—7. The comparison with 10 % lactic acid feed at the same temperature and pressure is shown Figure 4-13. When lactic acid feed concentration changed from 10% to 17.2%, the conversion decreased about 10 % at different hydrogen to lactic acid molar ratios. The PG selectivity for 17.2% feed is only slightly lower than that of 10% lactic acid feed at low H2/LA ratio. The selectivity and conversion at high concentration are relatively unaffected by hydrogen to lactic acid molar ratio (Figure 4-13). One possible explanation is that G-L mass transfer does not control the reaction because the increase of hydrogen (ratio) does not affect the reactions. Table 4-7. Results of 17.2% lactic acid feeding H2/LA molar ratio Conversion Yield Selectivity 1.16 45% 33% 73% 2.33 43% 33% 77% 3.5 43% 34% 79% 4.67 43% 34% 79% 5.63 43% 34% 79% . 100% 3. 90% - I— ; 80% ~ M fl {3 g 70% — I w 60% ~ g 50% - O—A A ° 5 40% - C ‘1’ v W 0 E 30% ~ -O-Conversion for C=17.2°/o g 20% - +Conversion forC =10°/o c —Cl— selectivrty for C=17.2°/o 8 10% 7 +selectivity for C=10% O 0/0 T I I I 1 2 3 4 5 6 H2/lactic acid molar ratio Figure 4-13. Selectivity and conversions for two lactic acid concentrations 103 4.2.8. selecti patent poisor formal may f2 hydro) longer fromt (27gm Hl'dm The St 4.2.8. Addition of sulfur to lactic acid feed The idea to add sulfur to the liquid feed stream is to find a way to enhance the selectivity. In related work on sugar hydrogenolysis over ruthenium catalysts (U .S. patents 5,600,028 and 4,430,253), the addition of small quantities of sulfur partially poisons the ruthenium catalyst, which lowers conversion and essentially eliminates the formation of methane and ethane from the sugar feedstock. Applying the same principal may facilitate an increase in selectivity to PG by eliminating lactic acid deep hydrogenation to hydrocarbon by-products. Slightly higher reaction temperatures or longer space velocities may compensate for the decrease in catalyst activity stemming from the addition of sulfur. The experiments were conducted at 100°C and 1200psi in Charge 2 (27gram CG6M). Sodium sulfide (N328) solution was mixed with 10% lactic acid feed Hydrogen to lactic acid molar ratio was fixed at 4:1 and liquid flow rate was 1 mllmin. The sulfur concentration in lactic acid was continuously increased (Table 4-8). Table 4—8. Results of adding Sulfur Conversion % Yield % Selectivity ppm of N828.9H20 Time (min) 68 59 88 0 120 63 55 87 100 120 60 51 85 100 130 57 49 87 300 120 56 49 88 300 60 56 48 86 300 70 53 45 84 600 120 52 45 85 600 60 47 41 88 600 60 33 27 84 600 100 Total sulfur (S) mole 0.0013 104 The addition of Na2S in liquid feed slowly decreased the lactic acid conversion, and the selectivity was almost unchanged. That means that the deactivation by sulfur is not selective. A simple calculation is given in Table 4-9 to show the deactivation process. In this calculation, we assume that only surface ruthenium calculated from dispersion is active and that lactic acid conversion is proportional to the amount of active surface ruthenium. The calculation indicates that for every two sulfur atoms passing over the catalyst, one surface ruthenium atom will be deactivated (Table 4-9). This shows that sulfur deactivates ruthenium active sites with a very high efficiency. Table 4-9. Calculation shows the deactivation is fast Catalyst (CGGM) Weight (g) 27 otal Flu (g) G 1.36 Total Ru Mole 0.013 Dispersion % 5 Ru on Surface Mole 0.00067 Total Na2$.9H20 addition Gram 0.304 otal S Mole 0.0013 Sulfur added lSurface Ru Mole/mole 2 Lactic acid conversion (before) % 68 Lactic acid conversion (after addition of sulfur) % 33 Fraction of sites deactivated % 49 105 Cali add at l. 4.3. resic' acid) 100% I; 60% «‘T ‘ *6 .9 1 8 60% 1 ‘- ._ _ _ .0 - - I I. g ‘\ c 40% ‘ ~ ‘ .3 —Selectivity § 20% _ - - 'Conversion 6 00/0 I l I 0 0.1 0.2 0.3 NaZS.9H20 (gram) passed through reactor 0.4 Figure 4-14. Conversion and selectivity change with addition of sulfur Figure 4-14 visualizes the catalyst activity change with the addition of sulfur. The catalyst was continuously deactivated by adding sulfur-containng lactic acid solution. No equilibrium state was found; thus with enough sulfur, all active sites will disappear. In addition, the catalyst was deactivated permanently. With a 12-hour reduction in hydrogen at 150°C and 300psi, no reactivity could be recovered. 4.3. Conversion of unrefined lactic acid to PG in trickle bed reactor The use of unrefined lactic acid to produce propylene glycol has the potential to further lower production costs. To investigate possible catalyst deactivation from the residual impurities from fermentation present in unrefined lactic acid, a raw lactic acid sample from Cargill was used as feed. The sample is unrefined lactic acid (L(+)-lactic acid), which contains 50% lactic acid by weight and some amount of unknown 106 impurities (6. By comparing its reactivity with refined DL-lactic acid (J. T. Baker, 85%) and Purac L(+)-lactic acid (FCC grade, 88%), the effect of impurities in the Cargill sample on catalyst performance was investigated. 4.3.1. Reaction conditions All experiments were conducted in the trickle bed reactor with Charge 3 described in Section 4.2.1. Reactions were carried out at temperatures from 80 - 120°C and pressures from 200 — 1200psi. Lactic acid was fed to the reactor in aqueous solutions of ~10% by weight. The liquid feed rate to the reactor was fixed at 2.0 mllmin, giving a weight hourly space velocity (kg lactic acid/kg catalyst/hr) of 1.3. The hydrogen to lactic acid feed molar ratio was fixed at 4: l, which is an optimal ratio from former experiments. 4.3.2. Results and discussion 4.3.2.1.Hydrogenation at different Pressure and Temperature Conversion of unrefined and reagent grade lactic acids is shown in Figure 4-15, and the selectivity profile comparison is given in Figure 4-16 as a function of temperature and pressure. ° Impurities here refer to components other than lactic acid and water; HPLC analysis showed no other peaks, so we are confident there are minimal quantities of other organic acids present. 107 5 80% l ' . 9 2 1: 60% - 8 :2 g 40% g g 20% JQIIKBO C) - * JT Baker *Cargillnoo C) - b JT Baker —I—Cargill(120 C) - I' JT Baker 0% I I I I I 200 400 600 800 1000 1200 Pressure (psi) Figure 4-15. Conversion profile comparison of Cargill and pure lactic acid <9 9: g , . I ‘ 22 _ .. I ’ g 40% 4 . _ . I ' 3 I —0-Cargiu(60 o) - o- JT Baker 20% i —II-Cargiu(100 C) - ‘1 JTBaker -I-Cargm(120 C) - I- JTBaker 0% r r r r 200 400 600 800 1000 1200 Pressure (Psig) Figure 4-16. Selectivity profile comparison of Cargill lactic acid and regular lactic acid 108 J. T. Baker lactic acid hydrogenation was conducted first because we know from former experiments that it does not deactivate the catalyst. Following that Cargill lactic acid was used. The temperature order we used was 100, 80 and 120°C and the pressure order was 1200 to 200psi in 200psi increments. Both refined lactic acid and Cargill lactic acid show a trend of increase in conversion with pressure and temperature. However, compared to refined lactic acid, the conversion of the Cargill lactic acid is lower at most reaction conditions. Apparently, either impurities or catalyst deactivation are responsible for the lower reactivity. The yield difference at 100°C and high pressure (1000 and 1200psi, these are the first two experiments with Cargill lactic acid) is barely distinguishable, but the difference increases as the experiments progress. Results of the selectivity in Figure 4-16 are a little surprising because we failed to enhance the selectivity by adding sulfur (N 323) as shown in Section 4.2.8. At most reaction conditions, Cargill lactic acid has higher selectivity than refined lactic acid. The most exciting result is at 80 °C; the observed selectivity of 95% is the highest we have achieved in the trickle bed reactor. In addition, like conversion, the selectivity difference at 100 °C is much smaller than at the other temperatures, likely because of the catalyst selective deactivation over the course of the run. 4.3.2.2. Catalyst deactivation from impurities in Cargill lactic acid The most crucial issue is catalyst deactivation from the impurities. After finishing the investigation of temperature and pressure, the catalyst Charge was treated (reduced) at 150 °C and 300 psi hydrogen for 12 hours. We typically conduct this treatment every 15~20 hours online to ensure that the catalyst is not oxidizing; in all cases the catalyst 109 performance is the same or slightly better following the treatment. Then, a confirmation experiment was done to verify the change in catalyst activity. This experiment used standard reaction conditions (100 °C and 1200 psi); other parameters are same as the experiments in Section 4.3.2.1. The results are shown in Figure 4-17. 1 00% 90% 80% 70% A----A-"" o--.-.---- Time=0 before deactivation : 30% ‘ --A-—CargiII-Conversion - i' JK Baker-Conversion 20% ~ —O—Cargill-Yield - 0- JK Baker-Yield 10% _ —Cl—Cargill-Selectivity - I- JK Baker-Selectivity 0% I I I I I ) 42 44 46 48 50 52 Reaction time (hour) Conversion, selectivity and yield or O a? Figure 4-17. Catalyst deactivation (Switch from Cargill to pure lactic acid at 6 hr) After comparing results at the start and after the temperature and pressure investigations, it is clear that the catalyst partially deactivates during the reaction. After 40-hour of reaction with Cargill lactic acid, which is equivalent to passing the entire 960- ml original Cargill sample (50%) over the catalyst, lactic acid conversion decreases from 95 to 70% and selectivity increases from 80% to about 90%. Apparently, this deactivation is selective, in that selectivity increases somewhat. Thus, by-products formation is reduced more rapidly than PG formation. From a process standpoint, deactivation would require progressively more recycle of lactic acid, but would result in overall higher yields. 110 43. refit Refi 1W0 chro addil other acid For 6 after show Conv., $9190. and Yield 4.3.2.3. The effect of lactic acid sources The purpose of this experiment is to investigate the reactivity difference between refined lactic acid (J. T. Baker), Purac lactic acid and, Cargill unrefined lactic acid. Refined lactic acid comes from chemical synthesis with no optical activity and the other two are L (+) lactic acid. The impurities in the Cargill lactic acid do not appear in HPLC chromatography, so at least we know that no other organic acids are in the impurities. In addition, identical concentrations, temperature, and pressure were used to eliminate any other effects. The experiments were conducted at 100°C and 1200psi. Hydrogen to lactic acid molar ratio was fixed at 4:1 and all the lactic acids were diluted to 9.3% by weight. For each lactic acid, total reaction time was five hours, and the first sample was taken after two hours stabilization and analyzing product every hour after that. The results are shown in Figure 4-18. 100% 90% - .—-.-+' W 2 80% ~ Purac Car ill 5 70% _ J. K. Baker 9 g 60% ‘ W 53:33 3 50% ~ (705 40% . >7 30% ‘ +JK Baker-Conversion +JK-Baker-Selectlvlty +JK Baker-Yield C 20% q +Purac-Converslon —D—Purac-Selectlvlty —o—Purac-Yleld 8 1 09/ +Cargill-Converslon —D—CargilI-Selectivity -O—CargilI-Yleld o - 0% 1 1 0 5 Time (hour) 10 15 Figure 4-18. Conversion, selectivity and yield of three lactic acids 111 From these results, we notice that the reactivity of racemic lactic acid is slightly lower than L+ lactic acid, although we did not found any difference in batch reactor and the confirmation test (Figure 4-17). The reason is unclear yet; but the catalyst was left in reactor without pressure for a week before this run. Likely air leaked in the reactor and some oxidation happened. The Purac and Cargill lactic acid are very similar, but a 3% decrease in conversion and yield after five hours of feeding Cargill lactic acid is detected. That means that deactivation does happen during the five-hour reaction. This is consistent with the average decrease of 25% conversion over the 40-hour Cargill lactic acid feeding. The gas phase composition is monitored by a mass spectrometer (Figure 4-19). As shown in Figure 4-18, the first five hours was racemic lactic acid feed, then five hours of Purac lactic acid, and finally five hours of Cargill lactic acid. It is apparent that the impurities affect the gas by-product composition. The most significant change is the relative amount of methane and ethane, although the total gas phase product yield are very close for Cargill and Purac lactic acid. 0.10 . . 0.09 - — CH4 —C2H6 ------ 03H8 - - -Total % 0.06 4 g k .. fl¢ - . as 0.07 - ' '-..-- ---"' -- --- % 0.06 1 "---__ Purac _ f-E 0.05 - J. K. Baker Carg'" g 0.041 M g g 0.03 d — -_l E 0.02 - _ _—I _ 0.01 ~ '-------.-.. ----~...-.. .-......---........- 0.00 0 5 Ttime (hour) 10 15 Figure 4—19. Gas phase composition changes with reaction time and feed 112 4.3.3. Summary Identical concentration of Baker, Purac and Cargill lactic acids were hydrogenated at 80~120°C and 200~1200psi in a trickle bed reactor on 5% Ru/CG6M. The comparison of three lactic acids shows that the impurities in Cargill lactic acid slowly deactivate the catalyst. After passing 1000 ml of the unrefined Cargill sample, the conversion decreased by 25% and at the same time the selectivity increased from 80% to 90%. Therefore, the deactivation process is selective. In addition, reducing the catalyst in hydrogen at 150°C could not recover the activity. Purac and Cargill lactic acid have very similar liquid products, but the gas phase by-products'are different. 4.4. Conclusion Studies on catalytic hydrogenation of lactic acids at 80~120°C and 200~1200psi in a trickle bed reactor over granular activated carbon supported ruthenium catalysts indicate the following: 1. Lactic acids from variable sources can be continuously converted to propylene glycol in a trickle bed reactor. 2. CG6M is the best catalyst among those we made and as in batch reactor, high dispersion is not connected to high reactivity. 3. The stronger adsorption of hydrogen (as seen from the Intermediate temperature in Table 4-3) correlates with low activity in this hydrogenation 4. Lactic acid conversion increases with temperature at the same pressure and hydrogen to lactic acid ratio shows that intra-particle diffusion or surface reaction also controls the trickle bed hydrogenation beside gas-liquid mass transfer. 113 . L1- or fermented lactic acid has slightly higher reactivity than racemic lactic acid in trickle bed hydrogenation. . The deactivation from sulfur (N623) only decreases the lactic acid conversion and does not change the selectivity at 611. Therefore, sulfur deactivates the catalyst without any preference for the favored reaction. . The impurities in Cargill unrefined lactic acid selectively deactivate the catalyst. After passing 1000-ml Cargill sample, the conversion decreased by 20% and at the same time the selectivity increased by 10%. Further reduction of the catalyst in hydrogen at 150°C cannot recover the activity. . These studies imply that there may exist two kinds of active sites on the catalyst surface and we may be able to selectively poison the catalyst to achieve higher propylene glycol selectivity, provided lower conversions can be accommodated. 114 Chapter 5. Mass transfer, kinetics and modeling In three-phase hydrogenation, gas-liquid (G-L), liquid-solid (L-S) and intra- particle mass transfer will significantly influence the reaction rate. To investigate intrinsic kinetics, mass transfer effects must be eliminated by choosing suitable process parameters (catalyst particle size, stirring speed, catalyst loading, initial concentration, and reaction pressure and temperature). 5.1. H2 solubility measurement H2 solubility or saturation concentration is a very important parameter in mass transfer analysis. However, no H2 solubility data are available in the literature for high temperature and pressure. The measured solubility of H2 at our reaction conditions will be reported in this section. 115 5.1.1. Apparatus The equipment used for solubility measurement is shown in Figure 5-1. It consists of a high-pressure reactor with stirrer (Parr autoclave, 300 ml), 6 burette with two port caps (for gas out and mixture in), and a water-bath with a glass cylinder for measuring the gas volume by water displacement. A needle valve was used in the liquid outlet to control the saturated liquid fluid rate. Coiled steel tubing was used after the needle valve to cool down the saturated liquid to minimize flash vaporization. Saturated IlQUldO \(y I I l _ Liquid weight Measurement Gasvolume I W Figure 5-1. Apparatus for solubility measurement 5.1.2. Experimental steps The first step was to fill the reactor with liquid (water or reactant solution) and seal it. Then the reactor was heated up to the desired temperature and pressurized to the desired pressure at the same time. High stirring speed (1200 rpm) was used to ensure the gas and liquid reached equilibration within 10 minutes. After stopping stirring, another 10 minutes was allowed to let the liquid and gas fully separate. The saturated liquid was then taken out from the dip tube to purge the liquid outlet line. A hydrogen source with 116 constant pressure (pressure regulator and 6 hi gh-pressure cylinder tank) was connected to the reactor to maintain a constant pressure in the reactor. While the burette was empty and the glass cylinder was filled with water, the nwdle valve was carefully opened to let saturated liquid depressurize in the burette. The liquid was collected in the bottom and the gas displaced the water in the glass cylinder. When the liquid level in the burette reached about 20 ml or the gas volume in the glass cylinder was over 100 ml, the needle valve was closed and the liquid in the burette (weight and volume) and the gas volume in the cylinder were recorded. 5.1.3. Calculation and results Hydrogen solubility can be calculated as: wr(8) dl(g/ml) w] (8 ) V9, is the total volume in mL (STP) WI is the liquid weight (g) d, is liquid density (g/mL) V9, (ml) - S (ml/g) = Table 5-1. Solubility in HPLC water (ml (STP)/g) l36atm 1026trn 68atrn 346tm (°C) 13 l 6. 3. 1 100 2.39 1.78 1.26 0.58 130 2.64 1 1.41 0.62 150 2.87 2.35 1.58 0.65 170 3.15 1.78 0.66 Table 5-1 is the hydrogen solubility in HPLC water. The solubility slightly increases with temperature at a given pressure. To verify the measurement, these data Were compared to literature data (65) in Figure 5-2. The comparison shows that this measurement is reliable. 117 O: A 2 5 _ ‘ ‘ 1! a . A ‘ fl ‘ E 2 I g ‘ ’ 5 1735’”; ‘ ‘1 :1" 1.5 ‘ , .6 In a ' E . a"??? fi‘ ‘2" s ’ ’ E 1 “ 1.315? F‘ ’ _ g ,i‘ v - - Literature data (100 °C) '5 5,“ I This measurement (100°C) 0) 0-5 7 66“ Literature(at150°C) A This measurement (150°C) 0 I I I I I I I l I I I 30 40 50 60 70 80 90 100110120130140150 Pressure (atm) Figure 5-2. Comparison of measured solubility and literature data 5.1.4. Solubility of hydrogen in 10% lactic acid In the same way, hydrogen solubility in 10% lactic acid water solution was measured. Table 5-2 shows that H2 solubility in lactic acid has the same trend as in water, but the solubility in lactic acid solution is smaller than that in pure water at all our experiment conditions. The average difference is about 12%. At 100°C, the solubility linearly increases with pressure and can be represented by S = 0.0012 * P - 0.165 . At 1200psi, solubility is 1.28 ml/g or 5.7x10'5moze/mL. Table 5-2. Solubility of hydrogen in 10% lactic acid (ml(STP)lg) 0 Pressure (psi) “ C) 500 1000 1500 2000 100 0.46 0.99 1.63 2.24 130 0.57 1.31 1.93 2.46 150 0.62 1.47 2.05 2.57 170 0.65 1.57 2.17 2.66 118 5.2. Characterization of mass transfer in the batch reactor (autoclave) Understanding mass transfer in the batch reactor should be relative simple because the fast stirring rate most likely eliminates gas-liquid and liquid—solid mass transfer, and the powder catalyst also makes the intra-particle mass transfer unimportant. Experiments and calculations from literature correlations were used to investigate mass transfer effects in the autoclave reactor. From comparison of the reaction rate with mass transfer rate, one can figure out the influence of mass transfer on conversion of lactic acid to PG. 5.2.1. Suspension of catalyst Catalyst suspension is not directly related to mass transfer, but the assumption for mass transfer study is that catalyst powder is evenly distributed in the liquid. Therefore, it is necessary to verify that the entire solid catalyst is suspended, or that no catalyst settles at the bottom of reactor. The minimum stirring speed requirement in the autoclave was given by Zwietering (66) (1959). = 2(dR ldT )1.33d:.2fll0.lgo.45 (pp _ pL)0.45w0,13 055 0.85 pr. d7 N where d3 and d;- are the reactor diameter and the stirrer diameter respectively and w is the catalyst loading (g/ 100g). NIn is the minimum speed needed to suspend all catalyst. Minimum speed requirement for different catalyst loadings is shown in Figure 5-3. It is clear that the stirring speeds we used (200~1200rpm) are much larger than the minimum suspension speed. 119 O I I I 0 2 4 6 8 10 Catalyst loading (g/ 100g solution) Figure 5-3. Minimum stirring speed for catalyst suspension 5.2.2. Maximum reaction rate and pseudo first order rate constant in batch reactor The highest reaction rate is needed for investigating the mass transfer effects. Reaction rate is defined as converted lactic acid mole per weight of catalyst and time. The fastest reaction happens at high temperature and high pressure. Therefore, the reaction rates are calculated only for the reaction at 150°C and 2000psi (matrix 2, see Chapter 3). The initial reaction rate is the fastest because it is at the highest reactant concentration and with fresh catalyst. The maximum rate is about 0.081 mole/hr when the catalyst loading is 3g per 100g solution as shown in Table 5-1. mole dC d (1 — x) C 1.40 (mole/ ml) dx Ru = M = Cuo = In" gear prdt pLWdt “'(gca: [10081116101 )pL (guquu /ml) d’ 120 Table 5-3. Reaction rate for three catalysts loading at 150°C and 2000psi 3 mole/ml l 1 Solution . 1.1 1.1 Conversion % 1 hr 53 hr 69 81 3 hr 84 hr 97 dx/dt (at t=0) 49 rate mole/hr) rate (mole/hr Table 5-4. Pseudo first order constant Loading Time (hr) Con % -ln(1-x)/wp k(llhr) 1 o r 21 022 0.22 a 121100: 2 39 0.45 0.22 0-3- wow /I a 3 52 0.66 0.22 £06 “2:894 .°H : 1 40 0.23 0.23 $0.4 - / 3 6:? *0 2g/100g 2 69 0.53 0.27 60.2 . Y’= 5 3 84 0.63 0.26 0 +""° E 1 53 0.23 0.23 0 i, 4 3g/100g 2 81 0.50 0.25 Reactiontlmemour) 3 94 0.65 0.28 k=0.25 (llhr) from fitting all data If the reaction is assumed as first order for lactic acid, we can calculate the rate constant. The calculation is shown in Table 5-4. The data from all loadings fall on the same line. Therefore, the same rate constant is obtained for all loading. dCu (1CLA CW -r =— =kC =>——= kdt=>l — = kt =>ln1-x =- kt (.4 prdt LA C Paw ‘4 C ] pr ( ) pr LA LA 5.2.3. Gas-liquid mass transfer A simple method was used to estimate the magnitude of the hydrogen—water mass transfer coefficient in batch reactor. The principle is to measure the pressure change in a scaled reactor after the beginning of stirring. From the pressure drop rate, the mass 121 transfer coefficient can be calculated. Because the limit of the precision of pressure measurement and the speed of time recording, this method is only for the estimation of the magnitude of the mass transfer coefficient. 5.2.3.1. Principle and procedure First, a certain amount of liquid (water) was charged into the sealed reactor; then the reactor was heated with stirring to specified temperature (25 or 100°C). When the temperature stabilized, the stirring motor was stopped. The reactor was pressurized carefully to a desired pressure with hydrogen. To minimize the mass transfer during the pressurization, the hydrogen was introduced from the gas phase (not from dip tube). The mass transfer during pressurizing is assumed small enough to be neglected; this was confirmed by the very slow pressure drop observed without stirring. The stirrer was then turned on; reactor pressure drop with time was recorded immediately after the beginning of stirring at specified speed. 5.2.3.2. Data analysis Let So be the hydrogen solubility at temperature T and pressure P, S (ml/g) the hydrogen concentration in the liquid at time t, and kLa the gas (G) liquid (L) mass transfer coefficient. Then, the rate of change of hydrogen concentration will be: dS ET = (S0 —S)kLa Hydrogen solubility So will change during the experimental process because the pressure P will change. In the low pressure range, we can estimated the change with Henry’s law So=H*P (at 150°C, H=0.0218 mllg.atm, from the measurement in Section 122 5.1). From the hydrogen mass balance, the instant hydrogen concentration in liquid phase 8 can be calculated S= 22400"'VG(P0 -P) (ml/g) R*T*WQ where WL is the liquid weight in reactor,VG is gas phase volume at room temperature, P0 is the initial pressure in reactor, and P is the pressure at time t. 22KD*V¢ LetB=—— thenS=B(P -P) R*T*“Q ° dS -+MP 67:77: .6606) whereSo = H*PandS =B(Po -P) dS/dt is obtained by fitting S vs. t data to a fourth order polynomial and differentiating the polynomial. Plotting dS/dt vs. So-S and forcing a line to pass through the origin, the mass transfer coefficient kLa can be calculated from the slope. 5.2.3.3. Results and comparison with literature The measured mass transfer coefficients at different liquid loading in the reactor are given in Table 5-5. The regression coefficients R2 in most of the regressions are 0.90~0.98, with the worst case of 0.8. That is reasonably good considering the very simple experimental equipment and method. It is seen that kLa sharply increases after the stirring speed reaches 800rpm. Due to the limitation of equipment and experimental method, the reactor had to be filled to 73~87% of capacity to ensure measurable gas pressure drops. This leads to a very bad flow pattern in the reactor, because the stirrer blades are on the bottom of the reactor. Therefore, the coefficients we measured are the mass transfer coefficient at the worst conditions, and represent an estimate of the lower 123 bound of gas-liquid mass transfer. Visual experiments were conducted to show the effects of stirring speed. Only the glass linear was used in visual experiments, so the gas liquid interaction can be seen from the wall of glass liner. The experiments showed that gas bubbles were formed when the stirring speed reaching 300rpm, and after 500rpm, no clear liquid phase could be seen. This partially explains the mass transfer coefficient change with stirring speed. Table 5-5. Summary of mass transfer coefficient Water:218 (25 °C) Water=256 ( 25 °C) Water=260 (100 °C) Rrrpm) kLa(S') 1.72 min“) Rrrpm) 1,66") 1...: min") R (rpm) 1,66") 1,6 (m6) 347 0.0055 0.33 82 0.0034 0.2 216 0.0033 0.2 600 0.0096 0.58 600 0.013 0.75 615 0.01 0.61 805 0.02 1.19 734 0.018 1.1 891 0.03 1.78 847 0.023 1.4 869 0.025 1.5 913 0.036 2.2 870 0.028 1.7 1023 0.054 3.2 933 0.037 2.2 1075 0.068 4.1 1186 0.098 5.9 1280 0.14 8.2 1337 0.139 8.3 The comparison with literature data (67) is shown in Figure 5-4. The literature data come from a large (2L) Parr autoclave and the mass transfer coefficient is measured for hydrogen in methanol. The density of methanol is close to water and the hydrogen solubility in methanol is 1.4 cc/g at measurement conditions, so it is very close to our system. The higher solubility of H2 in water than in methanol and the bad flow pattern measurement are possibly canceled out. Therefore, the mass transfer coefficients are very close to our measurement at low stirring speeds. However, the liquid and gas contact pattern is much different in high stirring speed and fast mass transfer cannot be achieved with large liquid loading. This comparison verifies that this result is reasonably good. 124 _: 7 4 Awater=256g (86% E liquid) is 6 * OWater=260g(87% 4 8 5 - liquid) 6:: 9g 4 2 .Waterz218g (73% . o quuld) 8 3 - . . ‘5; ltherature g 2‘ .c‘ 8 1 - l . 2 . I 5 '1] 0 '447 T r r a r O 0 200 400 600 800 1000 1200 1400 Stirring Speed (rpm) Figure 5-4. Hydrogen—water mass transfer coefficient in the autoclave 5.2.3.4. Comparing G-L mass transfer with reaction rate At a stirring speed of 1000rpm, the G-L coefficient is about 3 (l/min), which is equal to 180 (llhr). The solubility of hydrogen in 10% lactic acid at 2000psi and 150°C is 2.57cc/g (0.115 mole/L). Therefore, the maximum gas-liquid mass transfer rate (when hydrogen concentration in solution is zero) will be 2.1 mole/hr (100g solution-)0.1L), which is much faster than the maximum observed reaction rate (0.081 mole/hr, see Section 5.2.2. ). Therefore, G-L mass transfer in batch reactor is negligible. Alternatively, the hydrogen concentration in liquid can be estimated as R6 (mole/hr) = KLa(So — 3) *V1 :5 i =1_ R6 1 0.081 = — = 0.96 ->l So KLaVLSo 180-0.1-0.115 That means that the liquid hydrogen concentration is 96% of the solubility limit and gas liquid mass transfer will not control the autoclave hydrogenation of lactic acid to PG. 125 5.2.3.5. Comparing with literature correlation Gas- liquid mass transfer in mechanically stirred tank reactors has been investigated by a number of workers. Ramachandran (64) (1983) has given an extensive review. Among these investigators, Bern er al (68) (1976) correlation is relatively reliable, for it used data from different size reactors (include commercial reactor). kLa = 1.099x10-2 N1.16d11.979u:.32VL-0.521 N is stirring speed (rpm) d1 is the diameter of impeller (cm) u8 is superficial gas velocity (cm/s) based on the reactor diameter V1, is the liquid volume in reactor (ml) With a fixed superficial gas velocity (0.001 cm/s), which was estimated from actual hydrogen consumption rate, the mass transfer coefficient was calculated at different stirring speeds. Compared to measurements in Section 5.2.1 (Figure 5-5), the predicted value is very close to the measured value at both low and high speeds. However, the deviation in medium stirring speed is significant. This calculation also shows the measurement is consistent with published data. 126 7 J ”Bern’s Equation A water=256g (86% liquid) /' 6 1 O water=218g (73% liquid) ” I water=260g (87% liquid) G-L Mass transfer coefficient 1/min . I 0 200 400 600 800 1000 1200 1400 Stirring Speed (rpm) Figure 5-5. Comparison of Bern’s correlation and measurement 5.2.3.6. Verification by investigating the stirring speed effect As shown in Section 5.2.3.4 , the maximum rate for G-L mass transfer calculated using the measured coefficient is much larger than the maximum reaction rate, so we conclude that the G—L mass transfer resistance is negligible. This conclusion also is supported by the following lactic acid reactions at different stirring speeds. These reactions were conducted at exactly the same temperature, pressure, catalyst loading and pre-reduction conditions [150°C, 2000psi, 1-gram 5% Ru/C powder (Moisture 52.6% Lot 325980 England), 30 minutes reduction at 150°C]. The experimental results are summarized in Table 5-6. Virtually no relationship can be seen between the conversion and stining speeds. That means the mass transfer does not control this reaction even at a stirrer speed of 200. The maximum difference in conversion is about 10%; this deviation may come from the uneven catalyst reduction. The later experiments showed that one-half hour was not enough time to completely 127 reduce powder catalyst, because the powder catalyst did not have good contact with hydrogen gas during the reduction process. Table 5-6. Stirring speed effects at hour 58% 600 800 1000 1200 5.2.4. Liquid-solid mass transfer It is well known that the coefficient of liquid-solid (L-S) mass transfer is very large compared to gas liquid mass transfer, so no actual measurement was conducted. Only a creditable correlation was used. The commonly used correlation is from Sano er al. (1974)“. k d ‘ 3 1/4 1/3 —’ " =2+0. —ed"f‘ —”" e= P 01% Ill. 2.0 prl Fc is the shape factor of particles (=1) a ___6w(g)p.(mL/g) P 100(ml)-dp(cm) external area of particles per unit volume of solution pL Liquid density (1.1 g/ml) P power consumption (watt) We do not know the exact power consumption P, but it should be around 20~300 watt. The particle size used is dp=0.05¢m as an upper limit. The L-S mass transfer coefficient of hydrogen ksap was calculated and shown in Figure 5-6. The 128 calculation shows that liquid-solid mass transfer is much faster than that at the gas-liquid interface. Therefore, its resistance also can be neglected. 700 600~ 50° ‘ _P=300 Watt ' ' - P=20 watt 4002 300~ 200 - L-S mass transfer ksap(l/min) 100- 0 I I I I 0 2 4 6 8 10 Catalyst loading (g/ 100g) Figure 5-6. L-S mass transfer coefficient from Sano’s correlation Another recommended correlation is from Boon-long et al (73). Their equation does not need the power consumption, but it needs stirring speed. Figure 5-7 shows results from this correlation. Comparing with Sano’s correlation, the stirring power consumption of our autoclave at 1200 rpm is around 20W. 129 16 . Boon long equation 12~ _L O 1 N-hOJCD 11 L-S mass transfer ksap(1/min) O l I l I T I 0 200 400 600 800 1 000 1 200 Stirring speed (rpm) Figure 5-7. Mass transfer coefficient from Boon—long’s equation 5.2.5. Intra-particle mass transfer Because we do not known the reaction order, the observable modulus (W 6152 - Prater criterion) was calculated to estimate the effect of intra-particle mass transfer. = (—RG)L2 2 77¢ p-CADe where (-RG) is observed reaction rate (mole/gcat.sec) The diffusivity of hydrogen and lactic acid are calculated from correlation (Appendix-l) and catalyst true density (p) is 0.8 g/ml forCGSP as measured in incipient wetness. Lactic acid consumption rate (mole/geat.sec) was taken from Table 5-4. The modulus Characteristic length of catalyst is obtained by L=Dp/6, where Dp is the catalyst diameter. The effective diffusivity De: 8; D is 3.3x10'5 cm2/s at 100°C for hydrogen in water. Figure 5-8 shows that if the catalyst particle size is smaller than 0.02 mm, the observable modulus will be less that 0.1, and then mass transfer can be neglected. 130 9.09.0.0 801888 0.15 4 0.10 ‘ Observable modulus 0.05 . A————’— ‘ / 0.00 - 0.00 0.01 0.02 0.03 0.04 Catalyst diameter (cm) Figure 5-8. Observable modulus changes with catalyst diameter (10% lactic acid) 5.2.6. Summary of mass transfer in the batch reactor The powder catalyst used in batch reactor is less than 0.01cm diameter; therefore, the intra-particle mass transfer is negligible. The slowest mass transfer is gas-liquid. The maximum G-L mass transfer is 50 times large than the maximum reaction rate. Therefore, the calculations and experiments show that mass transfer in autoclave is unimportant, and the intrinsic reaction kinetics can be determined. 5.2.7. Batch reactor macro kinetics Since G-L, LS and intra-particle mass transfer all can be neglected in batch reactor for lactic acid hydrogenation, then the observed rate in batch should be the intrinsic rate. The data used in the following analysis are from Matrix 2 and Matrix 3 as Shown in Chapter 3. Two temperatures (130 and 150°C), three pressures (1000, 1500, 20OOpsi) and three different catalyst loading (1, 2 and 3 gram/100g solution) were the Variable parameters. The same catalyst and pre-reduction conditions were used. To avoid 131 complications from possible active metal leaching and deactivation, only initial rates will be considered here. 5.2.7.1. Initial reaction rate The initial reaction rate was obtained by fitting the lactic acid concentration profile (Curt) to a fourth order polynomial, differentiating the polynomial and setting the time to zero to get the initial reaction rate. This method is shown in Figure 5-9 and Figure 5-10 for experiment 99-15 (1 gram catalyst, 1500psi and 150°C). All initial rates are summarized in Figure 5-11. 0.8 - g 0.6 ~ '8' E 0.4 - U 0.2 - y = -ll~:-06x‘ - 0.000413 + 0.0219X2 - 0.2384x + 1.0066 R2 : 0.9967 0 I I I I I 0 1 2 3 4 5 6 Reaction Time(hour) Figure 5-9. Fit concentration curve to 4th order polynomial 132 Initial reacton rate(mole/L.hr) 0.3 0.25 y = -0.05x + 0.25 E 0.2 - R2 = 1.00 S E 0.15 - g 0.1 — A A '? Initial rate 0.25 0.05 4 A 0 1 i i l . 0 l 2 3 4 5 Reaction Time(hour) Figure 5-10. Get initial reaction rate from extrapolating the rate curve 0.9 0.8 d R... :iC_I-1-| a 0.7 - W dt ‘=° 0.6 ~ 6, 0.5 . 5 -I- 1,000,130 0.4 4 _ + 1,500,130 0.3 .., " ‘ +2,000,130 0.2 g: M#‘ -3- 1,000,150 // -0— 1,500,150 0'1 -9— 2,000,150 0 . r 6 l . i 1 1.5 2 2.5 3 3.5 Catalyst loading g/ 100mL Figure 5-11. Initial reaction rates with catalyst loading 133 5.2.7.2. Activation energy From regressing the initial reaction rate by _ E m R- - - = k ex — P" mztzal 0 P RT H 2 One can get the macro activation energy, catalyst loading effect “m” and hydrogen pressure effect “11”. First, the equation was rewritten as lnRW, = lnko —;—f+mlnw+nlnpm Then the multiple variable regressions were used to get energy E and constants m and n. The result is shown in Table 5-7. Then initial rate expression is Rinitial H 2 =1 95x10” exp —'96000 ”66160-3 RT The activation energy (96kJ/mole) shows that this is a chemical reaction control process, which also verifies that the mass transfer is negligible in the batch reactor (consistent with mass transfer analysis in Section 5.2.6. ). This regression shows hydrogen pressure only slightly affects the initial rate (m=0.3), which indicates that lactic acid hydrogenation is not a simple surface reaction. The comparison of initial rate and predicted initial rate by the regressed expression is given in Figure 5-12. 134 hl -1.-1.h1.~(.L(4-(4~CJpal-gl- Table 5-7. Regression results rate mole/Lhr) Cat 1 ln(rate) Statistics 0. 103 - R 0. l 1 -2.21 0.95 0. 12 -2. 1 0.061 -2. Error 0.1 -2. Coefficients 23. 0.178 Variable 1 Variable 2 Variable 3 -11 Results 1.95E+1 0.1 0.135 0.2 0.37 0.51 0.2 0.33 0.43 HWN—WNHWNI—th-t y—s U1 1 1 l 1 1 1 1 l 0.67 mic—butc- A T=150C A T=13OC Predicted reaction rate 0 # 0.2 - X=Y 0.1 4 0 I I I I 0 0.2 0.4 0.6 0.8 1 Actual reaction rate Figure 5-12. Comparison of experiment and predicted rates 135 5.2.8. Kinetics model Even thought the main reaction is simple, without the knowledge of surface reaction, it is still very difficult to get a reasonable kinetic model. From Chapter 6, the main reaction path for PG formation is LA to propane-1,1,2-triol to propene-1,2-diol (or 2-Hydroxy-propionaldehyde) to propylene glycol. Based on the main reaction mechanism, a H-W model is derived and fit to the low temperature reaction data. 5.2.8.1. Model derivation The Hougen-Watson (H-W) model will be used to get a workable reaction model. First, we assume that all species molecularly adsorb on to single sites. The reaction consists of six steps, hydrogen and lactic acid adsorption, formation and dehydration of - propane-1,1,3-triol and the formation of PG. Hydrogen and lactic acid adsorption are not likely the controlling steps beCause fast adsorption is seen from the literature and our experiments. The other surface reactions are the possible control step. S is active site; K is equilibrium constant and k is rate constant. Cv is the vacant active site concentration and CT is the total active site concentration. LA( lactic acid) 21» Propane - 1,1,2 - triol (PT) U Propylene glycol (PG) 4: {2 - Hydroxy - propronaldehyde}(PD) Propene -1,2 - dlol 1. H2 +S<—"L—)S ~H2 Hydrogen adsorption 2. LA +S<——"L—)S 1 LA Lactic acid adsorption 3. S-H2 +S-LA(—KI—)S-PT+S Form propane-1,1,3-triol 4. S-P'I‘ <——"‘—-)S - PD + H20 Dehydration 5. S-PD+S-H2(—"1—)S-PG+S FormPG 6. S-PG(—K'—)PG+S PG desorption 136 We assume the first hydrogen addition (reaction 3) is the irreversible rate controling step and all other reactions are in equilibrium. Also we assume that total active site density is constant and neglect the water adsorption. C K,-—‘&— —+ CSH2=KlPH2C, - Pflzcv K2 =—CSA T") CS-LA = chqu CMCV K4 = CS") ‘9 Csr'r =Csm/K4 = CPGCV Cs” K6K1K4Pllz K6 = CPGCV __) CSPG _ CPGCv Caro K6 CPGC, CSJ’G CV CSPG CV K6 V CPG CV 5 =—_ —) CS.PD = 2 = = CS-PDCS.H2 ' C5112 K1PH2CV K6K1P 11, R ‘- k3Cs.HZCu CT = C, + Csflz + Csu + (33,,G + Cs," + C s. m CT is total active site concentration = C, +K1PH C, + KZCMC, + CPGC” + CPGC" + CPGC" 1 K, KGKlKJ’Hz K6K1Pu, C = CT " Cm Cm Cm 1+ Km“2 + K2CM + + + K6 KrsKlK4Pli2 KrsKan2 0,2 r = kscsnzcsu = K1PH2K2CM 2 1+ Km“: + KZCM + C” + C” + C” K, KGKlK‘Pnz K6K1PH2 “ocupn, 2 [1+alPfl2 +a2CM +a3CPG +a‘ ii] “2 PG addition experiments (Section 6.1.4) showed the PG concentration only slightly affects the hydrogenation rate; therefore, the PG concentration term can be neglected for simplification (equal to very large K5 or very low surface PG concentration). The simplified expression is: 137 R = aOCMI’H2 /[1+a,PH2 +azCM +a3CPG +a‘ To fit the constants, rewrite the rate express as P J 1/«/--=———l +01 H’ +02 C“ J aOCIAP 11, “0Cyl “op 11, 5.2.8.2. Fitting the data of the reactions at 130°C 2 CK; _ “ocular, 19H (”6,6,1 Mac”)2 2 The data used for fitting the parameters are from Matrix 3 (see Chapter 3). We only use the data from low temperature reactions to fit the model because the reaction rates are low and mass transfer is negligible at these conditions and conversions. The catalyst loading is 1,2, or 3 gram and the pressure was 1000, 1500, or 2000psi. First, the conversion curve was fitted by third order polynomial and then the relation of rate and time was obtained by differentiating the polynomial. For each experiment, 12 points (30 minutes interval, 5 hours) were used. The total numbers of data points were 108. The fitted parameters are shown in Table 5-8 and the comparison of original data and model is given Figure 5-13. Table 5-8. Regression results for 130°C in autoclave 60 0.021 0.0216326, 62 10.3 Multiple R 0.85 R = al 000% (1+ 0.006619“, +1030“)2 P in psi and Cur in mole/L 138 0.009 c 0.008 - A Predicted I .C g 0'007 ‘ Rate K 5 0.006 - y 8 0.005 1 .3 0.004 .. .3 0.003 . A? .‘3 0.002 - 14‘, A, E 0.001 _ %,M 0 49"“ . . . 0 0.002 0.004 0.006 0.006 Original rate data (mole/gcat.hr) Figure 5-13. Comparison of model prediction (130°C) In the point of view of mathematics, the regression result is not good, but this is the only model that gives all positive kinetic constants. Using only initial data to fit the model did not give any improvement in regression quality, which implies that no significant catalyst deactivation and leaching happen during the batch hydrogenation of lactic acid. Other possible models are dehydration control (reaction 4) and second hydrogen addition (reaction 5), but neither gives positive rate constants. Actually, modeling of this process only from the reaction data is very difficult and inaccurate because of the side reaction. The difficulty is that side reactions consume larger amounts of hydrogen than the main reaction. Two moles of hydrogen are needed for PG formation, but six moles will be consumed for methane formation for every mole of lactic acid. That means that 10% of lactic acid converted in the side reaction will use 30% of the hydrogen consumed. Therefore, the poor regression may come from side reactions and the first hydrogen addition (reaction 3) is most likely the rate-controlling step. 139 pf th. 1161 10‘. 0Pe liqu 5.3. Continuous reactor (trickle bed) In trickle bed reactor, the relative velocities of gas, liquid, and solid are very low compared to the stirred autoclave; therefore, mass transfer most likely controls the hydrogenation process. Experiments and literature correlation will be used to investigate the trickle bed kinetics. 5.3.1. Dynamic Liquid holdup Liquid holdup (dynamic or free draining and static or residual) is a very important parameter in the trickle bed reactor. Static holdup is the fraction of liquid that remains in the catalyst pore after it has been completely wetted and drained. Dynamic liquid holdup is the fraction of liquid that is drained out after a sudden shut-off of liquid feeding and is a measurement of residence time in the trickle bed. The dynamic liquid holdup was measured in a trickle bed filled with 48 grams of CG6M Ru/carbon catalyst (61cm height) using HPLC water as liquid. Because the viscosity of water is 5% less than that of 10% lactic acid, the measured liquid holdup is a little smaller than that during actual operation. Figure 5-14 shows that the liquid holdups at 100 and 150°C are very close because the viscosity change with temperature is very slow. Liquid holdup increases with liquid flow at low flow rate, leveling out at higher liquid flow rate. 140 Liquid holdup (HZO mllmL bed) Residence time (minutes) 16% 14% - 12% ~ 10% - 8% 4 -A-T=100 C 6% q -I-T=23 C -A-— T=150 C 4% . 2% n 0% I T l I l 0 1 2 3 4 5 Liquid flow rate (ml/min) Figure 5-14. Liquid holdup at different liquid flow rates 14 12 A -A—T=100 C 10 ~ +r=23 C 8 —A— T=150 C 6 '1 4 .. 2 -l 0 W I f I I 0 1 2 3 4 5 Liquid flow rate (ml/min) Figure 5-15. Liquid residence time in the trickle bed 141 The residence time, which is defined as the ratio of holdup volume to liquid flow rate, is given in Figure 5-15. Except for one point at 23°C, 611 other data form a slowly decreasing curve. That means that the residence times almost are independent of liquid flow rate and temperature. The average residence time is about 5 minutes when the liquid flow rate ranges from 1 to 3 mllmin. 5.3.2. Residence time distribution (species adsorption on catalyst) The liquid residence time distribution is the reflection of trickle bed flow pattern and species adsorption on the catalyst. The experiment was done with a mixture of ethanol (3.3%), lactic acid (4.2%) and PG (2.3%) in water at room temperature (to ensure no reaction) with the reactor not pressurized. The liquid flow rate was fixed at 2 mllmin. According to the literature, gas flow rate does not affect the liquid flow very much, so it was fixed at 25 mllmin. After switching the liquid feed from pure water to the feed solution, the liquid composition in the outlet was monitored by HPLC. After the outlet concentration was fully stabilized after 2 hours, the desorption profile was recorded by switching back to pure water. 142 4.5% “4.0%~ ,..---"""' §35%6 .' ____‘_,__. ° 3.0%J -' -" "" a - X 5 2.5% , a—- 113 2.0% . . .LA 8 15% ‘ +PG o O S 1-0/° ‘ —A- -Etoh 0 0.5% - 0.0% , . 100 125 150 Time (after switching water to solution) minutes Figure 5-16. Outlet concentration change with time aftr switch to solution 4.5% 4.0% 9" *3 0 ‘ - . ILA 3.5/o “N‘ 0 3.0% 6 ’54:. +PG Concentration at outlet .. Time (after switching back to water) minutes Figure 5-17. Outlet concentration change with time after switching back to water 143 Figure 5-16 and Figure 5-17 are the adsorption and desorption profiles. The “residence times” of all three species in this experiment are over 50 minutes, while actual residence time is only 5 minutes from Section 5.3.1. Therefore, the actual information shown here reflects species adsorption on the catalyst. A simple calculation shows where these adsorbed species go (Table 5-9). First, the quantities of adsorbed species (collected from the water during desorption) are much larger than in the static holdup (calculated from the porosity of catalyst) and the dynamic holdup (measured from last section). If all adsorbed species adsorbed on ruthenium (10% dispersion), then each ruthenium atom would have to adsorb 53 lactic acid (or ethanol, or PG) molecules! Therefore, the carbon support is clearly the main adsorbent for the reactant and products. The desorption and adsorption curves measured are actually 6 characterization of the carbon support. As seen from the adsorption and desorption curves, the liquid was well distributed, with no apparent liquid by-pass, and the carbon adsorbs all three species without apparent preference. Only the desorption of ethanol is slightly different from that of lactic acid and PG, as seen in desorption curves (Figure 5-17). The species balance of adsorption and desorption in Table 5-10 shows that water flow can strip off most adsorbed species. 144 Table 59. Residual species after switching to water Lactic PG Ethanol Total acid Species desorbed into water (g) 4.7 2.6 4.5 11.8 Maximum possible quantity in dynamic holdup“ (g) 0.64 0.36 0.51 1.5 Maximum possible quantity in catalyst pore ** (g) 1.7 0.97 1.4 4.0 Excess adsorbed on catalyst (g) 2.3 1.3 2.7 6.3 Excess adsorbed on catalyst (mole) 0.026 0.017 0.058 0,1 Total Ruthenium (on surface)(mole) 0,0019 Species (mole)/Surface Ru (mole) 53 " Total dynamic holdup 15ml ** Total catalyst pore volume 40111] Table 5-10. Balance of desorption and adsorption Lactic acid PG Ethanol Total Species adsorbed from solution (g) 5.4 2.9 4.4 12.7 Species desorbed into water (g) 4.8 2.6 4.5 11.9 Difference 0.6 0.3 -0.1 0.8 5.3.3. Reaction rate and pseudo first order constant in the trickle bed To account for the effects of mass transfer, reaction rate has to be first calculated. For simple isothermal Operation, a constant H2 concentration in gas phase and constant overall effectiveness factor is assumed. The mass balance in control volume is: dFLCLA W]? = _rob5p VR is the bed volume, FL is liquid feeding flow rate (ml/min). p is catalyst bulk density (g/ml). The observed reaction rate rob, is defined as: rm=k'CM dFL= dcLA =-k'pc =ch= dv, dVR/FL ”‘ CM k'is pseudo first order reaction constant. 40,0de IFL After integration, we get the relation between conversion and “constant” k. 145 Cu C1041 p=0.42glmLand 1=L = 2mL/min VR =ll8mL ln(—] = k'pVR/FL =5 ln(l -x) = - k'pVR/FL => k' = — FL ln(l - x) Va P Table 5-11 gives two sets of data from two catalysts. The hydrogen to lactic acid mole ratio is 4: 1. Feed lactic acid concentration is 10 W%. The maximum rate constant is calculated to be 0.12, which come from the CG6M catalyst. Table 5-11. Pseudo first order reaction constants in trickle bed In flow rate [P(psi) [Con% | k (ngwmin) [Rate(R (mole/kg.hr) 4W CG6M catalyst (1 18ml) 2.0 200 16% 0.0067 0.045 2.0 400 22% 0.0096 0.064 .0 2.0 600 26% 0.013 0.06 3 2.0 800 35% 0.017 0.11 2.0 1000 42% 0.021 0.14 2.0 1200 49% 0.026 0.17 2.0 200 51% 0.027 0.18 2.0 400 66% 0.042 0.28 9 2.0 600 80% 0.062 0.41 § 2.0 800 88% 0.082 054 2.0 1000 92% 0.097 0.65 2.0 1200 95% 0.12 0.77 2.0 200 54% 0.030 0.20 2.0 400 74% 0.052 0.35 3’ 2.0 600 86% 0.076 0.50 «3 2.0 800 92% 0.097 0.65 2.0 1000 95% 0.12 0.77 2.0 1200 96% 0.12 0.83 30 (7 11111) CGSP catalyst 0.5 1200 57% 0.054 0.36 S) 1 1200 54% 0.050 0.33 § 2.0 1200 39% 0.032 0.21 3.0 1200 26% 0.021 0.14 146 5.3.4. Mass transfer coefficients in the trickle bed Hydrogen needs three steps (fiom gas to liquid, liquid to catalyst and diffusion in the catalyst pore) to reach the active sites. In this section, gas to liquid (G-L) and liquid to solid (L-S) mass transfer coefficients will be calculated by credible literature correlations and the intra-particle mass transfer will be investigated by both calculation and experiment. For the coefficients of G-L and L-S mass transfer, which are weak function of gas flow rate (‘9), Goto & Smith 00) recommend two dimensionless correlations. The G-L mass transfer coefficient at 100°C vs. liquid flow rate is calculated in Figure 5-18. Kw-“ i” .11..” D L p pD K La Gas-liquid mass transfer coefficient (llsec) D Diffusivity (cmzlsec) GL Liquid mass flow rate (g/cmzsec) p Liquid density (g/ml) )1 liquid viscosity (cp) For catalyst CuO.ZuO (0.54mm), 011:7 .8 and nL=0.39. 147 0.8 0.7 - 0.6 - 0.5 - KLa Q4 4 (1/min)0-3 ‘ 0.2 * O I I I I 0 1 2 3 4 Liquid flow rate (mL/min) Figure 5-18. G-L mass transfer coefficient vs. flow rate By comparing the maximum G-L mass transfer rate with average lactic acid consumption rate 0") r0,” , we can estimate the importance of G-L mass transfer. To make the comparison possible, we need to know the liquid holdup hd as measured in Section 5.3.1. The units of observed reaction rate are mollkgcathr. The maximum mass transfer rate calculated from kLa Ck is in mole/hr.L , which needs to be transformed to the same units by using dynamic liquid holdup and catalyst bulk density. RM = kLa duh, / p p = 0.42g/cm3 The maximum H2 concentration is its solubility. Ten typical reaction rates were picked from Table 5-11 and the comparison of G-L mass transfer rate with average lactic acid consumption rate is given in Table 5-12. For all ten runs, the maximum mass- transfer rate of hydrogen and average lactic acid consumption rate are of the same order of magnitude. The ratio of maximum hydrogen G-L mass transfer rate and maximum lactic acid consumption rate is less than two (the stoichiometry) for most of the trickle bed runs. Therefore, gas-liquid mass transfer of hydrogen limits the trickle bed reaction. 148 For H2 L-S mass transfer, Goto & Smith conelation is 0.52 ~28 k,6, =1.46uL %L— —‘“— 6(1-6)/d,, e = 0.41 II [)0 ksa, is liquid-solid mass transfer coefficient (llsec) and dp is catalyst particle size. The mass transfer coefficient calculated for hydrogen and lactic acid at 100°C are shown in Figure 5-19. Like batch reactor, L-S mass transfer coefficient is an order of magnitude larger than that of the G-L mass transfer. The maximum mass transfer rate can be calculated as RH, = k,a CLzhd lp The maximum L-S mass-transfer rates are calculated and shown in Table 5-12. The maximum L-S mass-transfer rate is a magnitude larger than lactic acid consumption rate; therefore, resistance of Liquid-solid mass-transfer is almost negligible. However, the actually hydrogen concentration in liquid is much less than its solubility because the G-L mass transfer limitation, therefore, L-S may be part of H2 mass transfer resistance at some conditions. For lactic acid L-S mass transfer, the lactic acid concentration is much larger than that of hydrogen, so L-S mass transfer will not a problem unless the conversion is very high. ‘ Calculated from liquid flow rate, lactic acid conversion X and the total catalyst weight by FLCMoX/Ww 149 Table 5-12. Comparisons of H2 G-L mass transfer and observed reaction rate F1.(m1/min) kLa P(psi) Con% Obs?“ m" Rat-(6'1" RL-S (1766) (m0 “8"") (mole/kghr) (mole/kw) 48gram CG6M catalyst (118ml) 2.0 0.57 200 51% 0.18 0.07 3 2.0 0.57 400 66% 0.28 0.14 6 2.0 0.57 600 80% 0.41 0.21 10 2.0 0.57 800 88% 0.54 0.28 13 g 2.0 0.57 1000 92% 0.65 0.34 16 ~ 2.0 0.57 1200 95% 0.77 0.41 19 30gram (71ml) CGSP catalyst 0.5 0.33 1200 57% 0.36 0.24 10 g 1 0.43 1200 54% 0.33 0.31 14 2.0 0.57 1200 39% 0.21 0.41 19 3.0 0.66 1200 28% 0.14 0.48 23 * Average lactic acid consumption rate over entire trickle bed reactor (1/min)25 A 20 9 15 ~ 10 9 5 _ 0 6 6 6 l . 0 1 2 3 4 5 6 Liquid flow rate (mUmin) Figure 5-19. L-S mass transfer coefficient vs. liquid flow rate 150 5.3.5. Intra-particle mass transfer in the trickle bed Inna-particle mass transfer in trickle bed reactor can be estimated by the observable modulus (W eisz -Prater criterion). _ 2 mil2 = g—R-glg— where (-R0) is obsered reaction rate (mole/gcat.sec) A c The diffusivity of hydrogen and lactic acid are calculated from correlation (Appendix-1) and catalyst true density is 0.8 g/ml as measured in incipient wetness. Maximum lactic acid consumption rate (0.77 mole/kghr) was taken from Table 5-12. The modulus for hydrogen and lactic acid vs. catalyst particle size (Figure 5-20) shows that hydrogen intra-particle is 6 control factor in trickle bed because the catalyst diameter we used is 0.05cm and 771112 larger than 0.1. 0.70 ° 0.60 - 0.50 a 0.40 2 0.30 - 0.20 - 0.10 ~ 0.00 observable modulus 0 0.01 0.02 0.03 0.04 0.05 Catalyst diameter (cm) Figure 5-20. Observable modulus in trickle bed (100°C and 1200psi) 151 5.3.6. Trickle bed kinetics The trickle bed experiments analyzed in this section are from the trickle bed reactor with 48-g (118-ml) CG6M catalyst. Liquid feed was fixed at 2 mllmin of 10% lactic acid, and H2 to lactic acid molar ratio was fixed at 4: 1. These reactions were conducted at three different temperatures and six different pressures. 5.3.6.1. Macro kinetics in trickle bed In Section 5.3.3, the pressure effect was lumped into rate constant for simplification and comparison with mass transfer. In this section, reaction was assumed as first order with respect to both hydrogen pressure and lactic acid concentration. The rate expression used in Section 5.2.7.2 is not used here for the simplified case. From the mass (mole) balance of a differential volume across the reactor, one can relate the rate constant with outlet concentration by a differential equation. After integrating the differential equation, lactic acid conversion can be related to reaction hydrogen pressure. FLdCu (disappearance) = -Rp,dVR (reacted lactic acid) R = k,PmCLA R :mole/(g.cat) - min, FL : ml / min, CLA :mole/ mL FLdCLA = - k, pBPmCMdVR C V V lnF‘: = —F—:p,,k,P"2 =>ln(1— x) = “1?:- p,,k,PH2 Plotting - ln(l- x) vs. Pm, and forcing the line to pass through the origin (set intercept to zero), then the pseudo first order constant k" in trickle bed can be obtained from the slope of the line. The calculations at three temperatures given in Table 5—13 show that the linear relationship is reasonable good. With k, at three different temperatures, The Arrhenius plot (Figure 5-21) gives macro activation (slope = -E./R) energy of Ea=48kJ/mole. Therefore, it is not a chemical reaction controlled process. As a 152 comparison, in batch reactor, the macro activation energy is 96kJ/mole, atypical number for chemical reaction. From here, we can preliminarily conclude that in the trickle bed, lactic acid hydrogenation is a mass transfer controlled process. Table 5-13. Pseudo first order constant at different temperature Temp(C) P (psi) Con % -ln(Com/Cin) 81* kv (ml/psi.sec.gcat) 200 21.0 0.24 8.75E-04 24313-05 400 33.5 0.41 O 600 43.3 0.57 °° 800 50.7 0.71 1000 56.0 0.87 / 1200 63.2 1.00 200 52.0 0.73 2. 49E- 031 6. 92E- 05 400 70.7 1.23 8 600 61.5 1.69 / "' 800 87.6 2.09 1000 91.1 2.42 P 1200 93.6 2.75 200 77.4 1.49 4. 65B 0311.291; 04 400 86.9 2.03 (:31 600 94.3 2.86 — 800 97.7 3.77 1000 98.7 4.34 1200 99.6 5.62 * SI=VRkav /FL 1 1.0 y = 5818.54x - 5.90 10.5 — 2 R = 0.99 g 10.0 E ' 9.5 ~ 9.0 ~ 8.5 I l I 2.5E-03 2.6E-03 2.7E-03 2.8E-03 2.9E-03 1/T Figure 5-21. Arrhenius plot for pseudo first order reaction 153 5.3.6.2. The limiting reactant in liquid phase The above sections show that gas-liquid mass transfer plays an important role, but it can not be the solely controlling factor because the G-L mass transfer limits the hydrogen liquid concentration, which also limits the L-S mass transfer rate. Therefor, L-S mass transfer may also limit the process at some conditions and locations in the trickle bed. The limiting reactants in the liquid can be identified by calculating the D*C term (diffusivity and concentration) of hydrogen and lactic acid. From their ratio in the range of operating conditions of interest (Doraisewamy and Sharma (71)), we can see the limiting reactant. DMCM (v y: H2 +0.42LA—)PG+H20 ’ 0.42DH2C H2 The ratio is indicative of the relative availability of the species at the reaction site. Thus, a value y>>1 implies a gaseous reactant limitation, while y< — _4 2 0.1 ~ 80 O.42D,,2C,,2 0.42*l.3x10 *(O.23 ~ 1.3)/ 22.4 7 Therefore, in the inlet region of trickle bed reactor, hydrogen is the limiting reactant, but close to the outlet, the limiting reactant may switch to lactic acid. ' The stoichiometry comes from considering main reaction and side reactions see Appendix-A 154 In conclusion, G-L, L-S mass transfer and chemical reaction will play a role at some conditions and at some positions of the trickle bed reactor. In the next section, we give a mathematical model for all these resistances and give concentration profiles along the trickle bed. 5.4. Trickle bed modeling Modeling all details of the three-phase trickle bed flow and reaction is impossible at this stage, however for our reactor radial diffusion can be neglected. The trickle bed can thus be seen as a plug flow or one-dimensional reactor. 5.4.1. Model equations and boundary conditions From a steady mass balance for hydrogen (A) and lactic acid (B) in liquid, we can relate bulk and catalyst surface concentrations of hydrogen and lactic acid (A ”BL, A5 ,BS) to the kinetic parameters and reaction conditions. In a control volume (a slice of cylinder), the mass balance for hydrogen (A) is: H2 from Axial Diffusion , , H2 transfer from bulk reaction + H2 from quurd flow = . . = . , , , quurd to catalyst surface consumption of H2 + H2 drffuse from gas to quurd Dynamic liquid hold up (hd) is used to relate catalyst volume to liquid phase volume. Reaction rate (lactic acid mole/geat.min) is defined as R = klAg'Bg'. Intra- particle diffusion effects were combined into the reaction rate constant k1. Therefore, it is n e o e e v actually not a real constant and Will change wrth reactron condrtrons. The mathematical expression for the hydrogen mass balance is d274, _ dA DA? “L $149.an —AL)=kSa,,,, (AL _AS):2Rp/hd 155 The boundary conditions are -DA%=uL(Au—AL) atx=0 fl: at x=L For B (lactic acid), the mass balance is similar to hydrogen, but no gas to liquid mass-transfer term. LA from Axial Diffusion _ LA diffuse from bulk liquid _ reaction LA from Liquid flow _ to catalyst surface - consumption of LA 2 D, %-ML 11%— : ksamw,‘ —BS) = 2Rplhd For B (lactic acid) The boundary conditions are "Du d5: =uL(Bu—BL) atx=0 (1:: = at x=L D A , D, Hydrogen and lactic acid diffusivity (mzls) AL , BL Hydrogen and lactic acid concentration in bulk liquid (mole/m3) Au , Bu Hydrogen and lactic acid concentration in liquid in inlet (at x=0) (mole/m3) As , BS Surface concentration of hydrogen and lactic acid at catalyst surface (mole/m3) A' Saturation concentration (hydrogen solubility) (mole/m3) K L Overall gas to liquid mass transfer coefficient (m/s) u ,_ Liquid flow velocity (m/s) a, a 5 Gas liquid interfacial area per unit volume of reactor (m2/m3) m, 11 Reaction order respecting to A and B p Catalyst bulk density (0.44 g/ml) 156 5.4.2. Analysis of the model Table 5-14. Model equations and boundary conditions for trickle bed reactor dZA dA . dzB dB DAEzi—uL-EL+KLa(A -AL)= Day-“LEE: It’d,” (AL “‘45): 295114153? ”'4 ksaPlA(BL -85) = “1ng [ha -DA%=uL(Au—AL) atx=0 —DB%=uL(Bu—BL) atx=0 fl=0 at x=L 2l‘-'-"() at x=L dx dx The four-coupled equations (two differential and two algebraic) are summarized in Table 5-14. Bulk hydrogen concentration AL , liquid lactic acid concentration BL , and surface concentrations of hydrogen and lactic acid (AS and BS ) change along reactor length (x-axis). This is a standard two-point boundary value problem in mathematics and is very difficult to solve, even with numerical methods. Only if the surface reaction is pseudo zero order in either A or B and first order for the other, the two differential equations are de-coupled and can be solved analytically. For example, if lactic acid surface concentration is constant and the reaction is first order with respect to hydrogen, then reaction rate can be simplified as R = klAs (motel gcat. min). The de-coupled equations (for hydrogen) can be simplified as d’A dA . DAETL-_uLIL+KLa(A _AL)=kraPH(AL-AS)=2flclAS/hd M (M —DA—dxi=uL(Au —AL) atx=0 i=0 at x=L With introduction of dimensionless variables, x . . . z = I Drmensronless drstance 157 A . . . CL = A]; Drmensronless concentratron A . . . C3 = 5 Drmensronless surface concentration A . . . . Cu = A“ D1mens1onless inlet concentration am = mDimensionless gas -liquid mass transfer coefficient “1. kSaPL . . . . . . a” = -— Drmensronless quurd solrd mass transfer coefficrent “I. a, = “'1' Dimensionless reaction rate constant "1. d u LL u LL . . . . PeA = F, Pe, = D Drmensronless Irqurd phase peclet number for H2 and LA A B The dimensionless form is 1 d2C dC PeA dzzL - dzL +001. (1- CL) = a” (CL - C5) = 2a,CS l dCL=CL-Cu at z=0 dC" =0 at z=1 PeA dz The surface concentration Cs from the algebraic equation is found a C a C -C =ZaC =>C =—”—"—= C LS ( L S) r S S a” + 20, £5 L Then the differential equation is simplified as 1 d2C dC PeA dzzL - dzL -(aGL +55 )CL =—acr. £=0 at 2 =1, dC’“ = Pe(C,_ -Cu) dz dz With given numerical values of am, ,65 , Pe and C u , this equation can be solved in Mathematica. For example when aGL =1, 65 =1, PeA = 2 and Cu = l , the solution is 1 2(3+J§)exp(2J§ + (1—J5)z) (J5 -1)exp(1+J§)z) CL = - 1+ + 2 2-2J§+4(2+J'5')exp(2J§) JE-3+(3+J§)exp(2J§) If Fe 11‘ > ea or neglecting the diffusion term, then it become a plug flow, the differential equation is simplified as: dCL + (0'61, + .35 )CL = “or. With at Z = O’CL = C“) ’ Z The analytical solution is 158 _ exp(_Z(aSL + flu ))(CU - l + exp(z(agL + flu ))agL + CU fllJ agL + flu For PeA = 200 , PeA = 0.01 and Fee, = 1, the hydrogen concentration profile is CL shown in Figure 5-22, which shows that it is almost a plug flow at PeA = 200. In our trickle bed reactor, PeA = “LL = 0'02“" l-fx621cm DA 1.3x10 cm ls = 9384 for hydrogen at 100°C; therefore our system is almost a true plug flow. In the same way, we can verify this is true for lactic acid (lactic acid diffusivity is smaller than that of hydrogen). In the following treatment, axial dispersion (diffusion) terms will be neglected. 1 Q C 3 0.95 2 \ _ _ _ P6=1 (U _ g 0'9 ..\ ---- Pe=100 g 0-35 ‘ [:1 ~----— Pe=0.01 8 0-3 ‘ .‘El :1 Plug g 0.75 -‘~-.~ . “13 a 07‘ ..-.‘--:E~. '5' 0.65 - 131-é - -- - -. . . '2 0.6 « a0, =1,;3, =1, and Cu =1 “El-"E,” o f] g 0.55 - D 0.5 r l T I 0 0.2 0.4 0.6 0.3 1 Dimensionless bed length Figure 5-22. H2 in bulk liquid with constant liquid reactant concentration 5.4.3. Modeling of trickle bed After neglecting diffusion terms, the reaction order for both hydrogen and lactic acid surface concentrations was set to one. This is the only form we can solve at this stage, and then the trickle bed system will be simplified as the following. 159 —uL %+ KLa(A‘ - A) = ksaPH (A - A5) = zap/1535 /h,, A=Au atx=0 dB ““1. 2: = kSaPlA(B— BS) : klpASBS H14 BzBu atx=0 As and Bs were obtained by solving the algebraic equation in the term of AL and BL in Mathematica (neglecting the negative roots). Then As and Bs were plugged into the two differential equations, which were then solved in Mathematica. 5.4.4. Model parameters The parameters needed for this model are listed in Table 5-15. G-L and L-S coefficients were calculated from correlation as listed in Table 5-16. Table 5-15. Model parameters at 100°C Symbol Name Unit Source Value “L Superficial liquid flow rate Cm/min Calculation 1.03 KL“ Gas-liquid mass transfer (H2) llmin Correlation 0.57 Ksam Liquid - solid mass transfer (H2) llmin Correlation 9.0 Ksapu Liquid - solid mass transfer (LA) llmin Correlation 4.8 Bu LA concentration in inlet Mole/ml Measurement 0.001 1 Au H2 concentration in inlet Mole/ml Saturation Solubility k1 Surface reaction constant 1.2/mole. gcat.sec The unknown p Catalyst bulk density g/ml Measurement 0.44 hd Dynamics holdup MUml Measurement 0.1 160 Table 5-16. L-S mass transfer coefficients for hydrogen and lactic acid Lap" ( llsec) for hydrogen FL(m1/min) T=80 (KS-2) T=100 (KS-2) T=120 (KS-2) 1 4.2 6 6.6 2 7.2 9 10.2 3 9 12 13.2 4 10.8 14.4 16.2 La,“ ( llsec) (lactic acid) T=80 (KS-2) T=100 (KS-2) T=120 (KS-2) 1 2.9 3.5 3.7 2 4 4.8 5.2 3 4.9 5.9 6.3 4 5.5 6.8 7.2 Table 5-17. Simulation results for k; P(psi) Conversion A" mole/ml Itl K La ksa (LA) ks“ (Hz) 200 52.0% 1.0505 N/A 0.57 9 4.8 00 400 70.7% 2.01305 171 0.57 9 4.8 8 600 81.5% 3.01305 57 0.57 9 4.8 "‘ 800 87.6% 4.01305 39 0.57 9 4.8 1000 91.1% 5.01505 31 0.57 9 4.3 1200 93.6% 5913-05 25.5 0.57 9 4.8 5.4.5. Simulation results at 100°C The principle of this simulation is to match lactic acid conversion by choosing a suitable RI. The results are shown in Table 5-17, which shows that It; increases with pressure decreasing. At p=200psi, the maximum conversion is 21% (set k1=oo), that is the maximum mass transfer rate and so this model fails at this condition. The bulk and surface concentration profiles are shown in Figure 5-23. 161 P: l 200psi C(surface LA)/Bin~ Reactor length P: 1 200psi Reactor length y P=400psi C(Bulk H2)/A*~Reactor length P: 1 200psi P=400psi Reactor length Figure 5-23. Simulation results at 100°C and 400~1200psi P=1200psi CLA/Bin ~Reactor length P=l200psi P=400psi leBin~reactor length Figure 5-24. Comparison of bulk liquid and surface concentration 162 With only one adjustable parameter, all concentration profiles along the trickle bed can be calculated. The results are consistent with the kinetic analysis. Figure 5-24 shows that lactic acid bulk concentration and surface concentration are almost identical (at P=1200 and P=400psi). However, for hydrogen the catalyst surface concentration is always lower than bulk concentration, which indicates the L-S mass transfer also limits the process. The decrease in k1 with increasing reaction pressure can be explained by the partially wetted catalyst. At low hydrogen pressure, G-L mass transfer is very slow, and most hydrogen directly comes from gas phase and the model does not include this part. At high pressure, the main path for hydrogen mass transfer is gas-liquid, which is what this model is based on. Therefore, to match the conversion, k1 has to become larger and larger to deplete the bulk hydrogen to reach high mass transfer rates with pressure decreases. As in all trickle bed modeling, the partially wetted catalyst is always a problem. 5.5. Summary Hydrogen solubility, a very important parameter in kinetics investigation, is measured at our reaction conditions. The comparison with literature shows our measurement is very reliable. The investigation of residence time distribution in the trickle bed shows that the carbon support strongly adsorbs both reactant and product, which makes the hydrogenation process even more complicated. Trickle bed dynamic liquid holdup is another very important parameter and is a bridge between liquid volume and catalyst volume in trickle bed modeling. Most trickle bed modeling in literature has missed this issue. 163 The calculation with creditable literature correlations and experiments shows that G-L, L-S and intra-particle mass transfer can be neglected in the batch reactor at our reaction conditions. The intrinsic kinetics is analyzed and the activation energy is 96kJ/mole. The initial reaction rate for 10% lactic acid hydrogen with Ru/C powder Catalyst is well represented by: Rm,“ = 1.95X1010 ex fl 0.661323 RT 2 At 130°C, an H-W model was derived and the parameter fitting is reasonably good considering the existence of side reactions. This is the only H—W that gives all positive constants. Therefore, the first hydrogen addition to lactic acid is most likely the rate controlling step. _ 0.021CMPH2 6+ 0.008819“: +10.3C,,,)2 Calculations and experiments have verified that G-L mass transfer is the major resistance in trickle bed reactor. Macro kinetic analysis shows that the activation energy is only 48kJ/mole in trickle bed, which indicates that mass-transfer is the rate controlling step. That further confirms the mass transfer calculations. Finally, a one-dimensional trickle bed model is derived. This model consists of two differential and two algebraic equations. In this model, dynamic liquid holdup was used to relate catalyst and liquid volume. Mathematically, it is a typical two-point boundary value problem and is very difficult to solve. With reasonable simplification (surface reaction is first order with respect to hydrogen and lactic acid), the model was solved in Mathematica. Bulk hydrogen and lactic acid, surface hydrogen and lactic acid concentration profiles are plotted. 164 Chapter 6. Mechanistic insight The focus of this chapter is to investigate the mechanism of lactic acid hydrogenation, which is helpful to further enhance the performance of the lactic acid to PG. By combining product distribution under data from different reaction conditions with different catalysts, specially designed control reactions (propionic acid, propylene glycol, ethanol and methanol hydrogenation), and the knowledge of organic chemistry, it should be possible to propose reasonable pathways for propylene glycol formation and side reactions. 6.1. Control experiments for mechanism elucidation Several hydrogenation reactions of substrates other than lactic were studied to probe the possible reaction pathways. These substrates chosen represent possible 165 intermediates or by-products and were subjected to the conditions as for lactic acid hydrogenation. 6.1.1. PG hydrogenation (M47, M48) Propylene glycol (PG) is the desired product. PG hydrogenation is used to investigate the possible deep hydrogenation or further reaction of PG at reaction conditions. Two runs were conducted, one at the standard condition (150°C and 2000psi) and another at higher temperature (170°C). The first one was used to identify the source of gas by—products and the second one was for the liquid by-products identification. In run M48 (1 gram Ru/C new PMC at 150°C and 2000psi), a 10% propylene glycol water solution was hydrogenated for 6.2 hours. 6.4% of the PG was converted and no detectable liquid products were detected The final gas phase analysis (Figure 6-1) shows methane and very limited amount of ethane are the products from PG hydrogenation. For lactic acid hydrogenation at the same conditions and catalyst loading (See Chapter 3, carbon balance), 4.5% of the lactic acid was converted to gas and 40% was converted to PG after 4.7 hours hydrogenation. The results from lactic acid and PG hydrogenation at the same conditions show PG deep hydrogenation is the major side reaction at standard reaction conditions because about the same quality of gas is formed . As seen in Figure 6-1, the ethane peaks are very small compared with the methane peaks at both temperatures’. Therefore, PG deep hydrogenation could not be the major source of ethane formation. ’ The response factors of urethane, ethane, propane, CO and C0; are in the same magnitude 166 1 M48-PG-Hydrogenation (150C) fl methane M47-PG-Hvdrogenation ( 170C) l ethane and propane l l A! L1} ,.- '4“4Af“9:v Y'fvva/xgr/Vg -- 1 Y rTwV'vv 4A- 4 ML M ‘r'vvv 10 20 30 40 50 10 20 30 40 50 Figure 6-1. Gas products from propylene glycol hydrogenation at 150 and 170°C Table 6-1. Product distribution for PG hydrogenation at 170°C Liquid (mole) Gas (mole) Sum PG Ethanol 2—propanol CH4 C2H6 Before 0.066 O 0 0 0 0.066 After Mole 0.055 0.0015 0.001 0.007 0.0014 0.065 % of initial PG 83.3 2.3 1.5 10.6 2.1 99.85 In experiment M47, increasing the temperature to 170°C and fixing the other reaction conditions as in M48, shows that 17% PG was consumed after 6.2 hours. PG was converted to methane, ethane, 2-propanol and ethanol. The gas and liquid analyses were summarized in Table 6-1; as in run M48, the major by-product gas was methane. The liquid analysis (Figure 6-2) shows that only 2-propanol and ethanol comes from PG hydrogenolysis. These two runs also imply that the majority of ethane, l-propanol, methanol and propane must directly arise from the lactic acid hydrogenation and not from PG deep reactions, or that they are quickly further converted to other by-products if they are formed. 167 l. 2-propanol ethanol J” L l l ‘10 15 20 25 30 35 h.- Figure 6-2. Liquid products from PG hydrogenation at 170°C 6.1.2. Ethanol and methanol hydrogenation on Ru/C catalyst (M49) Because ethanol and methanol were found in high temperature lactic acid hydrogenation, it is reasonable to assume that they are intermediate. The idea is thus to investigate their survivability at standard reaction conditions. In this experiment, a mixture of 6.4% ethanol and 5.5% methanol in water solution was hydrogenated under the same conditions as standard lactic acid hydrogenation (1% PMC Ru/C catalysts, 150°C and 2000psi). After 4 hours of reaction, 24% methanol and 28% ethanol were converted (Figure 6-3) and no liquid by-products were found. In gas products, ethane is barely detectable and methane is almost the only product from hydrogenation of methanol and ethanol (Figure 6-4). This experiment shows that ethanol and methanol can be converted to methane and the reaction rate is much faster than that of PG at the same conditions. Therefore, ethanol, methanol and PG (from last section) are possible intermediates for methane formation. This run also implies that ethane does not come from ethanol hydrogenation. 168 30% e 25% ‘ Amethanol Oethanol ' A g 20% . .E ‘ g 15% - C O 0 10% - e 5% - ‘ 0% x I I I I 1 2 3 4 5 Time(hr) Figure 6-3. Conversion of methanol and ethanol hydrogenation at 150°C 5.x :7 4.1% Gas analysis of methanol and ethanol hyth‘ogenation (M49) 34; ’ Methane I 32.0. 1.0 ethane propane r , ..... A ,......‘*.£AC";’: ........ .fi.‘.‘.‘...r.j m an ' 30 an an Figure 64. Gas phase analysis of methanol and ethanol hydrogenation at 150°C 169 6.1.3. Propanoic acid hydrogenation Propanoic acid was not found in the liquid by—products of lactic acid hydrogenation. The purpose of studying propanoic acid was simply to compare its reactivity with that of lactic acid. A 10% propanoic acid solution in water was hydrogenated with 2 gram Ru/C PMC (3310, the third batch catalyst from PMC) at 2000psi and 150~320°C in the Parr autoclave. At the standard reaction temperature (150°C), propanoic acid hydrogenation is much slower than that of lactic acid (Figure 6-5). For propanoic acid hydrogenation, no detectable l-propanol was found in the liquid products after 6 hours at 150~320°C. All reacted propanoic acid was converted to methane and ethane. Only a trace amount of propane was formed. It needs to be pointed out here that propanoic acid hydrogenation needs much higher temperature (almost 300°C) to achieve the same rate as lactic acid hydrogenation and the primary ending product is gases (methane and ethane) rather than l-propanol. From this experiment, we can conclude that propanoic acid is not an intermediate product in the lactic acid hydrogenation because of its high stability in the lactic acid hydrogenation environment. 170 100% A ILactic acid I 80% ~ A Propanoic acid I c I Lactic acid: 0 a Tag 60% T=150°C g I A ............._.. c _ propanoic acid: 8 40% A 0~2 hr 150°C I 2~3 hr 170°C 20% - A A 3~4hr 200°C 4~5 hr 250 °C A 5-5 hr 320 °c 00/0 I I 1 l l T O 1 2 3 4 5 6 7 Reaction time (hour) Figure 6-5. Comparison of lactic acid and propanoic acid hydrogenation 6.1.4. Adding PG in lactic acid hydrogenation Four experiments were designed to investigate the product (PG) effect on lactic acid hydrogenation. The initial solutions were 10% lactic acid and 0~20% PG. All of these reactions were conducted at 150°C and 1500psi with 2-gram Ru/C new PMC catalyst. The conversion profiles are shown in Figure 6-6. These results show that with the high concentration of PG, lactic acid still can be converted in a rate close to pure lactic acid hydrogenation. Therefore, the lactic acid hydrogenation is not reversible, in agreement with thermodynamic calculation. It is also clear that lactic acid and PG are not competing for the same reactive sites. The slight decrease in reaction rates with increasing PG concentration may come from the slower desorption of the PG product from the catalyst surface because of the high PG concentration in solution. 171 O O . I 5 80% . , I .. . a A o I. = 60% ~ 3 ’1’ 0994’ $407. - 00 l PQ=5% 8 g! e pg=10% 20% . I Apg=20% 0°/o ‘3 r 1 l 1 1 0 1 2 3 4 5 6 Reaction time (hour) Figure 6-6. Conversion profiles for adding different PG concentration Figure 6-7 shows the conversion and “macro” selectivity (4») after 4 hours hydrogenation. Because of the PG hydrogenolysis, the “macro” selectivity became worse and worse as the PG concentration increased. One explanation is that PG does not disturb lactic acid hydrogenation and the PG deep reactions lower the macro selectivity. From these runs, it is reasonable to assume that two kinds of active sites exist on the catalyst. One is good for lactic acid and another is good for PG deep hydrogenation. ‘ In here macro selectivity is defined as the molar ratio of PG increased to converted lactic acid. So, this selectivity is not the real selectivity of lactic acid hydrogenation because of PG deep hydrogenation. 172 I con% [1 Selectivity% 40% ~ 20% - LA conversion and PG selectivity o § 1 20% 10% 5% 0% Initial PG concentration Figure 6-7. PG addition effect lactic acid hydrogenation at 150°C and 1500psi 6.2. Gas by-product information at different reaction conditions The change of product gas composition with reaction conditions may give some clues about the side reactions. Catalyst loading, pressure and temperature effects will be shown in this section. Figure 6-8 to Figure 6-10 are gas analyses for Matrices 2 and 3 (see Chapter 3). As shown in Chapter 2, the Mass 15 and Mass 16 peaks are for methane, the Mass 28 peak is ethane, and Mass 29 & the peaks around Mass 40 come from propane. Because these spectra are final gas analyses and the reaction times are not exactly same, only a qualitative analysis will be given here. At a given temperature and pressure, increased catalyst loading increases both methane and ethane formation. Low temperature and low pressure tend to favor ethane and propane formation. One reasonable explanation is that the severe reaction conditions (high T and P) quickly cleave the formed ethane and propane to methane before they leave the catalyst surface. 173 1ngc . WW1. WW..- LWJLAAW .11 '1' a 10 20 30 40 10 20 30 40 Figure 6-8. Gas product composition change with catalyst loading at 1000psi and 150°C 1 I III 3g Ru/C lg Ru/C l l ‘ .Ml‘ilfll -- -JA-__ Mai NA“ -rj‘fl- 10 20 30 '40 10 20 I 30 . ' 40 Figure 6-9. Gas product composition change with catalyst loading at 2000psi and 150°C W l ,,,,,,,,,, Milli JKA-M - will, fin - Am 10 20 30 40 50 10 20 30 40 50 Figure 6-10. Gas product composition change with catalyst loading at 1000psi and 130°C 1 ’1 I i 3g Ru/C 1 lg Ru/C 11! | . “.1111, . . . - -. .LJ 1.. ”/va Al h , lam-xmma- g, MMI v v I 710 20 V 30 40 10 V '20" 30 40 50 Figure 6-1 1. Gas product composition change with catalyst loading at 2000psi and 130°C 174 6.2.1. Gas composition for low pressure hydrogenation (330psi) The gas phase analysis of products from low-pressure hydrogenation of lactic acid in the autoclave is shown in Figure 6—12. Low hydrogen pressure presumably leads to low hydrogen concentration on the catalyst surface, which is not favorable for deep hydrogenation (we assume that high surface hydrogen concentration will lead to more deep hydrogenation). Therefore, the ethane and propane concentrations have the same magnitude as methane in the gas phase. The selectivity is lower than seen at high pressure, but the difference is only 10%. That means that the reaction mechanism does not change dramatically at such a low pressure. 3.0: :15 Low pressure hydrogenation 2 gram RulC 330psi ‘-‘~ 150 c 12 hours || conversion 70% l selectivity 78% .I. .2. I Methane .eo ethane 4°. mm 40 Figure 6—12. Gas phase spectrum after 12 hours at 330psi and 150°C 6.2.2. Gas product distribution in trickle bed reactor Because different catalysts were used in the trickle bed and autoclave reactors, it is not appropriate to compare them directly. For two catalysts (CG5P and CG6M), the slightly different carbon supports gave very similar gas product distributions (Figure 6-13 175 and F1 gure 6-14). In the autoclave, propane was the major gaseous product instead of methane as in the autoclave. If we assume these two catalysts are similar to the powder Ru/C (PMC) catalyst used in autoclave reactor, then the different gas product distribution may come from the difference in mass transfer between trickle bed and autoclave reactors. In the trickle bed reactor, hydrogen availability to catalyst is limited due to slow diffusion. In addition, low reaction temperature should not favor deep hydrogenation because propane has more chance to leave the catalyst rather than being hydrogenated to methane. Therefore, propane is the dominant by-product. However, the deficit of hydrogen on the catalyst surface in the trickle bed reactor only inhibits the methane formation but not propane formation, and the selectivity in the trickle bed reactor is not as high as in the obtained in autoclave runs. CGSM-1OOC-1200psi-tr22 2728291516423839434144 Mass Figure 6-13. Gas phase analysis for CG6M catalyst in trickle bed 176 CGSP 1000 1200 psi 2728291516393241424344 Mass Figure 6-14. Gas phase analysis for CGSP catalyst in trickle bed 6.2.3. Catalyst deactivation by unrefined Cargill lactic acid samples H2 adsorption was used to investigate the change in active site density after catalyst use. After passing 4000ml of 10% Cargill lactic acid over the CG6M catalyst, the conversion decreased from 95 to 70% and the selectivity increased from 80% to 90%. Therefore, deactivation by impurities in the Cargill lactic acid is a selective process. The double peaks in fresh catalyst desorption support the idea that two kinds of active sites must exist on the catalyst surface. After deactivation, the second sharp peak (higher temperature) obviously was reduced. Therefore, it is highly possible that this peak represents the active sites used for PG deep hydrogenation. 6.3. Rationalization of the reaction paths For rationalizing the reaction paths, AB initio energies for related molecules were calculated (Table 6-3) and some related bond energy, which is the energy need to break the bond, is also listed in Table 6-2. Combining the information from above sections and molecules calculation, the main reaction and side reaction pathways will be discussed in this section. 177 adsorption Desorption Fresh catalyst Tenperamre After Deactivation Figure 6-15. H2 adsorption and desorption profiles change after deactivation Table 6-2. Average bond energy Bond C—C C—O C=C C=O Average energy (Kl/mole) 348 358 614 799 178 Table 6-3. AB initio energy (Kcal/mole) for different structures 0') HO O HO OH HO OH Structure CH—C\ CH—CH /\3 (b H ”30 OH “30 H H3C Name Lactic acid Propane-1,1,2-triol Propene-1,2-diol 13mg -21 1905 -212345 -165246 HO //0 H30 H2C\ Structure CH—CH =CH2 \C=CH2 H30/ HO Name 2-Hydroxy—propionaldehyde Propen-2-ol Propa-1,2-diene Energy 465252 -1 18914 -71706 H30 H30 ch (E) H H H- Name Propyne Propylene glycol (PG) Propen-l-ol Energ -71816 466002 -118910 Hac\c H30 H30 Structure / =CH2 H):-CH2 CH—CH3 "9 - H HO Name Propen-2-ol 1-propanol 2-propanol Enery -1l8914 -1l9669 -119673 H20 CI-L OH Structure ”ac—CH2 Name H20 Methane Ethanol Energy 47039.96 -24928.7 -95463.2 H3C‘OH H2 H30—CH3 Structure ,0 H30 \CHa Name Methanol Propane Ethane Energ -7l215.4 -73346.6 49137.4 H20=CH2 3 H2 Structure Hac’ Fem, Name Ethene Propene H2 ‘ Ener 48362.9 -72575.3 0 "‘ Calculated by SPARTAN, product of Wavefunction Inc, with HF-STO—3G single point energy 179 The first step of lactic acid hydrogenation should be the formation of propane-1, 1,2-trio] by the addition of H2 because no propanoic acid is formed during hydrogenation and the AB initio energy of propane-l, 1,2-trio] is less than that of lactic acid. “No propanoic” implies that the first step of lactic acid hydrogenation does not take off hydroxyl group. The second step is the dehydration of propane-l, 1,2-triol. Two possible products are propene-1,2-diol and 2-hydroxy-propiona1dehyde (Table 6-3), both products have higher energy than that of propane-l, 1,2-trio] (Table 64). Table 64. AB initio energy change during reaction Compound Energy KJ/mole Lactic acid -211905 Propane-1, 1.2-triol -212345 2-Hydroxy-propionaldehyde-1-H20 -212292 Propene-1,2-diol+H20 -212286 PG+H20 -213042 The AB initio energy of propene-1,2-diol is a little higher than that of 2-hydroxy- propionaldehyde, but the difference is so small (6 KJ/mole) that we can not exclude anyone to be the intermediate for PG formation. At low temperature, propene-1, 2-diol may be dominant intermediate because propene-1,2-diolcan can be further hydrogenated to PG with syn stereochernistry (Both hydrogen add to the double bond from the same face). This assumption is supported by the optically active product at low temperature from lactic acid hydrogenation reported by Antons (63). Either propene-1,2-diol or 2- hydroxy-propionaldehyde can be hydrogenated to PG because the low energy of PG and the favorable hydrogenation environment (ruthenium catalyst and hydrogen) for C=O and C=C bonds (Figure 6-16). 180 H3C HC on HaC 0H H3C on \ CO// +H2 \ / .1120 \:-.--/ +112 \ I CH—C —> CH— CH —CH —> CH—CH2 / \O / \ / HO H HO OH HO HO lactic acid Propylene glycol H3C o H3C on H3C H c on \ // +112 \ / .1120 \ (I; +H2 /CH—C\——> /CH—CH\——> /CH—CH ———> CH—CH2 HO OH HO OH HO HO lactic acid Propylene glycol Figure 6-16. PG formation path 6.3.1. Side reaction 1(methane formation) In PG, ethanol, and methanol hydrogenation, the major gas product is methane and only a trace of ethane was produced. So the most likely the first step of PG deep hydrogenation will be the cleavage of the C2-CB bond to form ethanol and methanol, which will be further converted to methane quickly as partially verified in Section 6.1.2. Although this path is not supported by AB initio calculation (the energy of methanol plus ethanol is higher than PG), this analysis is supported by the high temperature hydrogenation of PG (170°C, a trace amount of ethanol was found in the liquid, Section 6.1.1) and no CO was found in gas phase at any reaction condition. The comparison of the conversion of PG hydrogenation (6.4% after 6.2 hour) at 150°and 2000psi and the lactic acid hydrogenation at same conditions shows that most methane comes from PG hydrogenolysis. The mechanism of methane formation from PG is shown in Figure 6-17. 181 cu—cn2 J2_. H2 . "36/ "0 pG HO/ —HZ—> on, + H20 H2 OH H2 CH4 + H3C Figure 6-17. Scheme of PG hydrolysis ”3 HC OH H36 H30 C -H20 (\2H—C//o_+H'L—: \CH—CH/ -H20 ——>\C=CH ——'> H =0" l-l/ \OH HO/ \m Ho/_H \on +2.21 \H 0 H30 H30 H;C\-H2 H \: CH—CH; #— c=c|n|2 H —CH2 HO HO/ H 2-Propanol l-Propanol Figure 6-18. Propanol formation scheme 6.3.2. Side reaction 2 (propanol formation) 2-Propanol was detected in high temperature PG hydrogenation (Figure 6-2), so part of 2-propanol may come from PG hydrogenation. However, l-propanol comes directly from lactic acid hydrogenation. The following scheme is derived fi'om by- product information, bond energies, and common chemistry knowledge. The key point here is that propene-1,2-diol dehydrates to propen-l-ol or propen-2-ol when hydrogen is not available for hydrogenation. In addition, adding hydrogen to a C=C double bond is easier than adding H2 across a C-OH bond. Figure 6-18 shows the l-propanol and 2- propanol formation scheme from lactic acid hydrogenation. 182 6.3.3. Side reaction 3 (propane formation) Propane may be the product of hydrogenation of l-propanol and 2-propanol, but it does not come from PG hydrogenation. at most reaction conditions, propane formation is very limited in autoclave, so the formation of propanol is not favorable. Figure 6-19 is the propane formation scheme. In trickle bed reactions, the gas product almost is exclusively propane. The reason could be that the intermediate propene-1,2-diol (figure 6-18) tends to further dehydrate into propen-Z-ol and propen-l—ol because the low pressure and mass transfer effect lead to low surface hydrogen concentration. H30 H30 Ho; +120 \: CH—CH3 —?> H —CH3<——- H —CH2 HO H 2-Propanol l-Propanol Figure 6-19. Propane formation scheme - o ._ l-Propanol H Figure 6-20. Ethane formation scheme 6.3.4. Side reaction 4 (ethane formation) The only clue about ethane formation is that it always comes out with propane. That means it comes from the same path as propane. We already know the ethanol can not be converted to ethane. Therefore, the only possible path is propanol hydrogenation. Figure 6-20 is the possible scheme for ethane formation. 183 6.4. Summary The following conclusions can be obtained from the above analyses: 0 The first step of lactic acid hydrogenation is the formation of propane-1,1,2- triol, which dehydrates to propene-1,2-diol or 2-hydroxy-propionaldehyde, which will be hydrogenated to PG when enough hydrogen is available propene-1,2-diol. This is the main reaction. 0 If the reaction pressure is low or mass transfer limits the catalyst surface hydrogen concentration, propene-1,2-diol may be further dehydrated to propen-l-ol and propen-2-ol, which are the precursors of 1-propanol and 2- propanol. o Forming ethanol and methanol from breaking of the C-C bond is the pathway of PG deep hydrogenolysis. Ethanol and methanol can quickly undergo further hydrogenated into methane. o Ethane and propane both arise via the l-propanol or 2-propanol hydrogenation, which is formed when the hydrogen supply is limited on catalyst surface at high temperature and low pressure. With this knowledge, one has the potential to suppress the side reactions and enhance the selectivity for PG. For example, modifying the catalyst surface to decrease the highly active sites may decrease PG deep hydrogenolysis and thus enhance the selectivity. 184 Chapter 7. Summary and Recommendations Important conclusions from above chapters will be summarized in this chapter and some recommendations for the further research also is outlined. 7 .1. Summary Lactic acid obtained from fermentation was converted to propylene glycol in a stirred tank reactor (autoclave) and a continuous trickle bed reactor with supported ruthenium catalyst at mild reaction conditions. 7.1.1. Hydrogenation of lactic acid in autoclave Lactic acid water solution (5%~30%) was hydrogenated to propylene glycol in a D stirred tank reactor with powder catalyst. Only supported ruthenium was active enough to give reasonable conversion at our conditions. The carbon-supported catalyst is good in separating catalyst and liquid. When reaction temperature is below 170°C, the only liquid 185 product is propylene glycol; therefore, the product purification is very simple. The major by-products are methane, ethane and propane; the relative amount of gas by-products depends on reaction conditions. As high as 90% PG yields and over 95% lactic acid conversion were achieved at optimal conditions. The optimal temperature is around 150°C. Higher pressures are always good for reaction rate and propylene glycol yield, but 1500~2000psi is sufficient; these are very mild reaction conditions compared to prior carboxylic acid hydrogenation studies in the literature. The calculations and experiments show that gas-liquid, liquid-solid and intra- particle mass transfer are negligible in the autoclave at our reaction conditions. The intrinsic kinetics has been analyzed and the activation energy is 96KJ/mole. The initial reaction rate for 10% lactic acid hydrogen with Ru/C powder catalyst is well represented by: Rm = 1.95X 1010 CX{——fi—)Wo'“P:: This equation indicates that initial reaction is very sensitive to reaction temperature and catalyst loading but less so to hydrogen pressure. Direct hydrogenation of lactate salts was impossible in aqueous phase for our catalyst and reaction conditions. Low metal ion (K or Ca) concentration does not affect the lactic acid hydrogenation reaction. Acidified calcium lactate can be hydrogenated to propylene glycol at the same rate as pure lactic acid if the precipitated calcium sulfate is filtered before hydrogenation. An H-W model was given and the kinetic parameters were fit by using the data at 130°C. The fitting results were reasonably good considering the existence of side reactions. This is the only model that gives all positive constants. Therefore, the first 186 addition of H2 to lactic acid to form Propane-1,1,2-triol is most likely the rate controlling step. 7.1.2. Hydrogenation reaction in trickle bed reactor A trickle bed reactor with laboratory prepared 5% Ru on active carbon catalysts was used to continuously convert lactic acid to propylene glycol at even mild reaction conditions with conversion over 90% and selectivity over 95%. The reaction temperature can be as low as 100 °C and pressure can be as low as 800psi without significant sacrifice of the PG yield. Experiments and calculation show that gas-liquid, liquid-solid and intra-particle mass transfers and surface chemical reaction together control the lactic acid hydrogenation reaction in trickle bed reactor. Gas to liquid mass transfer is the major resistance. lactic acid conversion increases with temperature at the same pressure and hydrogen to lactic acid molar ratio. Like the reaction in autoclave, propylene glycol selectivity increases with hydrogen pressure. 7.1.3. Catalyst characterization and deactivation The catalysts prepared from different carbon supports have different hydrogen desorption profiles. Stronger hydrogen adsorption on the catalyst relates to the lower reactivity toward lactic acid hydrogenation. Two kinds of active sites may exist on supported ruthenium catalyst: both sites for lactic acid hydrogenation and only one kind of sites for PG deep hydrogenolysis. Carbon support adsorbs both reactant and product. Pure lactic acid did not deactivate the catalyst over a 100-hour reaction. The impurities in unrefined sample slowly deactivated the catalyst. The deactivation is 187 selective because with the lactic acid conversion decreasing from 95% to 70% yet the PG yield increasing from 80% to 90% after 4000ml of sample was reacted. However, addition of sulfur (Na2S) to pure lactic acid deactivates the catalyst without any enhancement in selectivity, only decreasing the lactic acid conversion. The loss in activity could not be recovered by high temperature hydrogen reduction. 7.1.4. Kinetic parameter measurement and trickle bed modeling Hydrogen solubility was measured at our reaction conditions. The comparison with literature shows our measurement is very reliable. Trickle bed dynamic liquid holdup, which is a bridge between liquid volume and catalyst volume in trickle bed, was measured in our trickle bed reactor. Gas-liquid mass transfer coefficient was measured in the autoclave and the results are consistent with published data and correlation. A one-dimensional trickle bed model was derived. This model consists of two differential and two algebraic, equations and forms a typical two-point boundary value problem. With simplification, the model was solved in Mathematica. Bulk hydrogen and lactic acid, surface hydrogen and lactic acid concentration profiles were plotted. The results give us more information about the role of gas-liquid and liquid-solid mass transfers. 7.1.5. Reaction pathway A simple surface reaction scheme was given. The first step of lactic acid hydrogenation is the formation of propane-1, 1, 2-triol, which dehydrates to 2-hydroxy- propionaldehyde or propene-1,2-diol, which will be hydrogenated to PG when enough hydrogen is available. This is the main reaction path. If the reaction pressure is low or 188 mass transfer limits the catalyst surface hydrogen concentration, propene-1, 2—diol may be further dehydrated to propen-l-ol and propen-2-ol, which are the precursors of 1- propanol and 2-propanol. Forming ethanol and methanol from breaking the C-C bond is the pathway of PG deep hydrogenolysis. Ethanol and methanol can be quickly further hydrogenated into methane. Ethane and propane all come from the 1(2)—propanol hydrogenation, which is formed when hydrogen supply is limited on catalyst surface at high temperature and low pressure. This scheme is supported by specially designed reactions and theoretical analysis. 7.2. Recommendations The conversion and yield achieved is unbelievable good without the knowledge of reaction pathway and catalytic mechanism. To further enhance the performance, surface reaction pathways, the role of catalyst support and the changes in optical property of product with reaction conditions need to be extensively investigated. 7.2.1. Surface reaction pathway investigation Although a simple surface reaction pathway was given, it is mostly from inference and deduction. We did not do anything to verify and apply our theory. It is very important to verify and correct this scheme by advanced surface analysis techniques and specially designed reactions. Another important aspect for this process is the role of catalyst support. Activated carbon support strongly adsorbs both reactant and product, and different carbon supports show different reactivity. Therefore, the carbon support must play a very important role in lactic acid hydrogenation. This can be done by systematically investigating the catalyst 189 activity for different carbon supports; the variable parameters of the carbon support may include the source of carbon (coal, wood...) and its physical properties (BET are, pore size distribution). 7.2.2. Selective deactivation of the catalyst and yield enhancement Another way to enhance the propylene glycol yield is via selective deactivation of the catalyst. The possibility comes fiom the reaction of Cargill unrefined sample in trickle bed reactor, although we do not know the composition of the impurities. We are sure there exist some compounds that can deactivate the high activity sites to suppress the propylene glycol deep hydrogenolysis. Searching for this type of compound may provide another way to enhance the propylene glycol yield. 7.2.3. Production of optically active propylene glycol Combining the results of this work and the patent of Antons (63), lactic acid from fermentation (L+) can be continuously converted to optically active propylene glycol in a trickle bed reactor. The very mild reaction temperature (<100°C) ensures very high ee efficiency. This process provides an economical process to produce optically active propylene glycol because the pressure used in our process is much lower than that shown in the patent. The change in ee efficiency of this reaction with catalyst and reaction conditions may be studied to find the optimal conditions. 190 Appendix A. Parameters calculation and physical data A.l. Lactic acid consumption rate and hydrogen consumption rate Two moles hydrogen is needed to hydrogenate one mole lactic acid to PG, but six moles will be consumed for methane formation for every mole of lactic acid. If we assume 10% lactic acid is converted to methane then side reaction will use 30% hydrogen. C3H603 + 2H2 <—) C,Irl,,02 + H20 C3H603 +6H2 (—)3CH‘ +3H20 C311,,03 +0.9x2H2 +0.1x6H2 <———> 0.9C3H,,02 +0.1x3CH, + (O.9+O.1x3)H20 C3H603 + 2.4H2 <——> 0.9C3H,,02 + 0.301, + 1.21120 Therefore, the hydrogen consumption rate (mole/hr) is 2.4 time of the rate of lactic acid consumption if side reactions are included. A.2. Bulk density of catalyst For CG6M, the trickle bed height (H) is 61 cm; the reactor diameter (d) is 1.57 cm, the thermocouple diameter (do) is 0.125 cm and total catalyst weight is 48 gram. Therefore the catalyst bulk density is: W 48 = 2 2 = 2 2 =0.4Zg/cm3 Hfl(d -do)/4 61120.57 -0.125 )/4 p For CG6P, the bulk density is 0.47g/ml. 191 A.3. External porosity For both CGSP and CG6M, the average particle diameter (dp) is about 0.05cm, and the ratio of dp to trickle bed diameter is 0.03. From the figure 5—68 in Perry’s handbook (74), we find the external porosity e. is 0.41. A.4. Internal porosity and catalyst density Combining catalyst bulk density and the incipient wetness measurement, one can calculate the internal porosity by 8 _ v _ 5m! ‘ w/ p 4.4g/0.47g/ml Catalyst density can be calculated as p = 0'42 =0.79g/ml for CG6M. 1-s 1—o.41 Table A-1. Catalyst density and porosity = 0.48 for CGSP Pr: Catalyst Bulk density yfml Catalyst (g) Add water (ml) Internal porosity Catalyst density gm] CGSP 0.42 4.4 5 0.48 0.80 CG6M 0.47 6.3 9.7 0.65 0.71 A5. Diffusion coefficients H2 and lactic acid diffusion coefficients were calculated from the equation qu39 in Perry’s Handbook 04). That also is called Wilke and Chang (1955) equation. _ 7.4x10*r(0M,)°~’ D ”ngs where the mole volume of H2 at normal boiling point, v2.4, is 14.3 cm3/mole; 0 is a constant, for water solution it is 2.6; The viscosity of water at 100 °c is 0.28 Ns/m2 192 From this data we can calculate the hydrogen diffusion coefficient is 1.3E4 cmzlsec. For lactic acid diffusion in water, the molar volume is calculated from its density at its normal boiling temperature, which comes from extrapolating the data at 0~80 °C and 0~88% lactic acid water solution. The mole volume of lactic acid is 70 cm3/mole. The diffusion coefficients are summarized in Table A-2. Table A-2. Diffusivity (cmzlsec) 80°C 100 °C 120 °C 140 °C 160 °C H2 1.02E-4 1.29E-4 1.3E4 1.07E-4 8.02E-5 Lactic acid 3.96E—5 5.01E-5 5.05E—5 4.15E-5 3.11E-5 A.6. Basic physical data used in kinetic calculation Table A-3. Thermodynamic properties for LA and PG acid PG Lactic acid PG C C C No 598-82-3 57-55-6 90 76 heat KJ/rnol -682.0 485.7 ai NIA 371 °C heat -l356.0 -l838.2 190 188 J/mol.k 210.5 190.8 c 254 255 99 50 60 16.8 -60 A.6.1. Lactic acid Lactic acid is, when pure and anhydrous, actually a white crystalline solid with a low melting point. However, because of the physical properties and the difficulties in the preparation of the pure and anhydrous acid this material is rare. Lactic acid appears generally in the form of more or less concentrated aqueous solutions. Lactic acid undergoes intermolecular esterification spontaneously, resulting in the formation of lactoyllactic acid and chain polyesters containing more lactic acid units in the molecule. All data come from the book “Lactic Acid” (Holten, CH, 1971) unless specified. 193 A.6.1.1.Density of aqueous solution of lactic acid Table A4. Densities of aqueous solutions of lactic acid Temperature °C LAW% 20 25 30 35 4o 50 60 70 80 0.00 0.998 0.997 0.996 0.994 0.992 0.988 0.983 0.978 0.972 6.29 1.012 1.008 9.16 1.020 1.018 1.016 1.011 1.007 1.001 0.995 0.989 12.19 1.025 1.022 24.35 1.057 1.054 1.052 1.047 1.041 1.035 1.030 1.023 25.02 1.057 1.053 37.30 1.086 1.081 45.48 1.110 1.105 1.102 1.094 1.087 1.079 1.072 1.064 54.94 1.130 1.124 64.89 1.155 1.152 1.147 1.140 1.132 1.124 1.115 1.108 75.33 1.179 1.175 1.170 1.161 1.153 1.143 1.134 1.125 85.32 1.199 1.195 1.190 1.181 1.172 1.163 1.153 1.144 88.60 1.201 1.192 A.6.1.2.Viscosity of aqueous solution of lactic acid Table A-5. Viscosities as a function of concentration and temperature (cp) . . Temperature °C Lam Ac'dw % 25 30 35 4o 50 60 70 80 0.8937 0.801 0.7225 0.656 0.5494 0.4688 0.4061 0.3165 6.29 1.042 0.838 9.16 1.15 1.03 0.809 0.671 0.571 0.473 0.416 12.19 1.21 0.961 24.35 1.67 1.46 1.13 0.918 0.746 0.632 0.532 25.02 1.725 1.328 37.3 2.45 1.857 45.48 3.09 2.74 2.03 1.59 1.26 1.02 0.843 54.94 4.68 3.38 64.89 6.96 6.01 4.22 3.12 2.38 1.85 1.47 75.33 13.03 10.55 7.08 4.98 3.57 2.73 2.08 85.32 28.5 22.6 13.91 9.4 6.4 4.59 3.4 194 A.6.2. Saturation pressures of lactic acid and propylene glycol co 8 Vapor pressure (mmHg) -* N 01 \1 o s 8 8° 8 8 8 s O 50 100 150 200 Temperature( C) Figure A-l. Saturation pressures of lactic acid and propylene glycol A.6.3. Equilibrium of lactic acid hydrogenation to PG The Gibbs free energies and reaction heats (Kcal/mol) of lactic acid and, propylene glycol, water and hydrogen are listed in following. Compound Formation heat (kcal/mole) Gibbs free energy (kcal/mole) LA -148 -123.1 PG -100.6 -72.6 Water -57.7 -54.6 H2 0 0 Reaction heat AH(kcal/mole) -10.3 A G (kcal/mole) 4.1 K m = exp(%;—;) = ape—1.9220298) = 1017 In K, =-AHr(_1___1_)=10300(i__1__)= 5184 K2,, R T 298 1.987 T 298 T -l7.4 ln(KT) = ln(1016.6) + Elf—4- —l7.4 = -10.47 +3—ng 195 CPG _ Cuo - Cur Cu 1 CMPHZ CMPHZ CMO KTPHZ +1 CW-cLA = xxi Kr Lactic acid conversion Cw, 1+ K,I",,2 8 as § 8 o\° uilibrium conversion 320% 0% l I l l l O 200 400 600 800 1000 Hydrogen pressure (psi) Figure A-2. LA equilibrium conversion 196 REFERENCE 1. Bennina, H “A histroy of lactic acid making”, Kluwer Academic Publishers, (1990) 2. Holten, C.H, “Lactic Acid”, p3, 1971 3. Chemical Market Reporter - 01-Mar-1999 4. Kroschwitz, J. 1., M. Encyclopedia of Chemical Technology 4th ed., 13, 1043, John Wiley & Sons, 1992 5. Levy, N and Scaife, C. W., J. Chem. Soc., 1946,1100. 6. British Patent 1,123,147,1968 7. Kirk, R. E., et a1 Encyclopedia of Chemical Technology 2nd ed., 8, 167, Inter science Encyclopedia, 1952 8. Monteaguo, J. M., L Rodriguez, J. Rincon, J. Fuertes, J. Chem. Tech. and Biotech, 68, p271, 1997 9. Urbas, B., U. S. Patent # 4,444,881, 1984 10. Giomo, L., P. Spicka, E. Drioli, Separation Sci. Tech., 31, p2159, 1996 11. Rincon, J ., J. et al, Solvent Extraction and Ion Exchange, 15, p329, 1997 12. Boniardi, N., R. Rota, G. Nano, B. Mazza, J. Appl. Electrochemistry, 27, p125, 1997 13. Chemical Market Reporter - June 26, 1991 14. Walkup, P. C., C. A. Rohrmann, R. T. Hallen, D. E. Eakin, U. S. Patent # 5,071,754, 1991 15. Barrie] et a1, United States Patent, 5510526, 1996 16. McCoy, M., Ed, Chemical Market Reporter - July 14, 1997 197 l7. Paparizos, C., S. R. Dolkyj, W. G. Shaw, U. S. Patent # 4,786,756, 1988 18. Ghose, T. K., Ed, Bioprocess Engineering - The First Generation, 367, Ellis Horwood Limited, 1989 19. Chemical Market Reporter - December 13, 1999 20. Chemical Market Reporter - July 20, 1998 21. Chemical Market Reporter - 08-May-00 22. Man, T. M Dissertation, Michigan state University,1998 23. Broadbent, H S, et al. J. Org.Chem. 24, p1847-1854,1959 24. Stefan Antons, Leverkusen, US patent 5731479,Mar,24,l998 25. Kida, K., S. Morimura, Y. Sonoda, Journal Fermentation and Bioengineering, 75, p213,l993 26. Hiltunen, K. J. V. Seppala, M. Harkonen, Macromolecules, 30, p373, 1997 27. Lipinsky, E. S., R. G. Sinclair, Chem. Engr. Prog., 82, p26. 1986 28. Hiljanen-Vainio, M., J. Kylrna, K. Hiltunen, J. V., Seppala, J. Appl. Poly. Sci., 63, pl335,1997 29. Gruber; Patrick Richard, US patent #05981694,Nov. 9,1999 30. Sawicki, R. A., U. S. Patent # 4,729,978, 1988 31. Mok, W. S., M. J. Antal, In, M. Jones, Jr., J. Org. Chem., 54, M597, 1989 32. Lira, C. T., P. J. McCrackin, Ind. Eng. Chem. Res., 32, 2608, 1993 33. Gunter, G. C., Ph.D. Dissertation Michigan State University, 1994 34. Gunter, G. C., J. E. Jackson, D. J. Miller, J. Catal., 148, 252, 1994 35. James, B. R. Homogeneous hydrogenation, John Wiley &sons, 1973 198 36. Mcquillin, F. J. Homogeneous hydrogenation in organic chemistry, D. Reidel publishing company, 1976 37. Everett Brown and H. Adkins, et al., J. Am. Chem. Soc., vol. 56, pp. p3689-691, 1934 38. Adkins, H. et al., J. Am. Chem. Soc., vol. 69, p3039-3041, 1947. 39. Carnahan, J .E. et al, J. Am. Chem. Soc., vol. 77, p3766-3768, 1955. 40. Ford, T. A. US Patents 2607807, Aug. 1952 41. Novotny; Miroslav, patent, US4273947: Hydrogenation of fluorine-containing carboxylic acids, 1981 42. Velenyi, L. J ., S. R. Dolhyj, U. S. Patent # 4,663,479, 1987 43. Paul N, Rylander, hydrogenation methods, 1985, p78 44. Harada, K., Chem, Commum., p1071, 1970 45. Calvin, M., Trans. Faraday Soc., 34,p1181, 1938 46. Iguchi, M.,J. Chem. Soc. Jap., 60, p1287. 1939 47. Halpem, J., J. Am., Chem. Soc., 83, p753, 1961 48. Halpem, J., J. Am., Chem. Soc., 88, p5150,l966 49. M,Ando, and S.Emoto, Bull, chem. Soc. Jpn,47, p501, 1974 50. Adams, P. T.,R. E. Selff, and B. M. Tolbert, J. Am. Chem.Soc, 74, p2416, 1952 51. Richardson, A. S. and J. E. Taylor, US Patents 2340344,2340687 to 2340691, 1959 52. Stephen, H. and T. Stephen “the solubility of inorganic and organic compounds”, The Macmillan company, New York, 1963. 53. Hofmann,H. P., chem. Eng. Tech., 47 , p823-868, 1975 199 54. Yuanxin Wu, et al., Ind. Eng. Chem. Res., 35, p397-405, 1996 55. Henry, H. C. and J. B. Gilbert, Ind. Eng. Chem. Proc. Des. Dev. 12,328,1973 56. Sato, Y. et al. “Flow pattern and pulsation properties of co-current downflow inn packed beds, J. Chem. Eng. Japan 6,315, 1973 57. Kim et al , J. Chem, Eng. Jpn,l4,p311, 1981 58. Shigeo goto and J. M. Smith, AIChE Journal (vol. 21,No4), p706-720, 1975 59. Shigeo goto, et al, The Canadian Journal of Chemical Engineering Vol. 54, p551- 555, 1976 60. Dick Stegeman, et a1, Ind. Eng. Chem. Res. 35, P378-385, 1996 61. Wammes, W. J. A. et al. AIChE Journal (vol. 37,Nol2), p1849-1862, 1991 62. Both hydrogen add to the double bond from the same face 63. Antons Stefan, US patent 5731479, March 24, 1998 64. Ramachandran, P. A., and R. V. Chaudhai, “Three-phase catalytic reactors”, Gordon and Breach science 65. Stephen H and Stephen T., “Solubility of inorganic and organic compounds”, Volume 1, The Macmillan Company, New York, 1963 66. Zwietering, T. N. “Suspending of solid particles in liquid by agitators. Chemical Engineering Science 8, 244.1958 67. Vasant R, et a1 Ind. Eng. Chem. Res. 1998, 37, 3879-3887 68. Bern, L., Lidefelt, J. O. and Shoon, N. H., J. Am, Oil Chem. Soc. 53, 463(1976) 69. Sherwood, T, K and F A L Holloway, “performance of packed towers”, Chem. Eng.Sci.,28,559(1973) 70. Shigeo Goto and J. M. Smith, AIChE J. (V 01, 21) N0 4,1975,p706 200 71. Doraisewamy, L K. Heterogeneous Reactions; John Wiley and Sons; New York, 1984;Vol.I 72. Sano, Y. et al. Mass transfer from suspended particles in agitated vessels. J. Chem. Eng. Japan, 255 (1974) 73. Boon-long, S., et al. “Mass transfer fiom suspended solids to a liquid in agitated vessels”. Chem. Eng. Sci, 33, 813 (1978) 74. Perry's Chemical Engineering Handbook, 6th ed 75. Zrang 2., “Reactive Properties of Hydrogen Adsorbed on Carbon Surfaces” Thesis, Michigan state University, 1998 201