13.07.2015 Views

the use of nanoporous carbon membranes for high temperature ...

the use of nanoporous carbon membranes for high temperature ...

the use of nanoporous carbon membranes for high temperature ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

THE USE OF NANOPOROUS CARBON MEMBRANES FORHIGH TEMPERATURE CARBON DIOXIDE SEPARATIONCLARE JULIA ANDERSONSUBMITTED IN TOTAL FULFILMENT FOR THE REQUIREMENTS OFTHE DEGREE OF DOCTOR OF PHILOSOPHYMAY, 2009DEPARTMENT OF CHEMICAL AND BIOMOLECULAR ENGINEERINGTHE UNIVERSITY OF MELBOURNE


DEDICATIONThis <strong>the</strong>sis is dedicated to my two grandmo<strong>the</strong>rs, Julia (deceased) and Vonda and <strong>the</strong>irunfaltering belief in <strong>the</strong> importance <strong>of</strong> education.ii


ABSTRACTTechnology designed to capture and store <strong>carbon</strong> dioxide (CO 2 ) will play a significant role in<strong>the</strong> near-term reduction <strong>of</strong> CO 2 emissions and is considered necessary to slow globalwarming.Nanoporous <strong>carbon</strong> (NPC) <strong>membranes</strong>, made from <strong>the</strong> pyrolysis <strong>of</strong> a polymer, may be <strong>the</strong>new generation <strong>of</strong> gas separation <strong>membranes</strong> suitable <strong>for</strong> CO 2 capture. The research workpresented in this <strong>the</strong>sis has assessed <strong>the</strong> suitability <strong>of</strong> NPC <strong>membranes</strong> <strong>for</strong> CO 2 capture fromnatural gas (methane, CH 4 ) and air-blown syn<strong>the</strong>sis gas purification (nitrogen, N 2 ) at <strong>high</strong>operating <strong>temperature</strong>s and in <strong>the</strong> presence <strong>of</strong> condensable and non-condensable impurities.Three different types <strong>of</strong> NPC <strong>membranes</strong>, namely supported, modified-supported andunsupported, have been manufactured. The elimination <strong>of</strong> cracking within <strong>the</strong> NPC<strong>membranes</strong>, which is a significant concern <strong>for</strong> <strong>the</strong> commercialisation <strong>of</strong> <strong>the</strong>se <strong>membranes</strong>,was most successful <strong>for</strong> supported NPC <strong>membranes</strong> made from polyfurfuryl alcohol (PFA).Characterisation results <strong>of</strong> <strong>the</strong> produced NPC <strong>membranes</strong> showed a pore size range suitable<strong>for</strong> gas separation via molecular sieving (3 - 5Å). The characterisation results also indicated asignificant increase in porosity with pyrolysis <strong>temperature</strong>. The membrane per<strong>for</strong>manceresults supported <strong>the</strong>se findings with a significant increase in permeance observed withincreasing pyrolysis <strong>temperature</strong>.Mixed gas per<strong>for</strong>mance measurements revealed an increase in <strong>the</strong> CO 2 /CH 4 selectivity over<strong>the</strong> pure gas CO 2 /CH 4 permselectivity due to <strong>the</strong> competitive adsorption <strong>of</strong> <strong>the</strong> more polarCO 2 .Mixed gas per<strong>for</strong>mance measurements also showed an increase in CH 4 permeance as <strong>the</strong>operating <strong>temperature</strong> was increased from 35 to 200 o C, which can be related to an increase in<strong>the</strong> rate <strong>of</strong> diffusion. However, <strong>the</strong> selectivity decreased with increasing operating<strong>temperature</strong> due to <strong>the</strong> smaller changes in <strong>the</strong> CO 2 permeance. These smaller changes in CO 2permeance are related to <strong>the</strong> stronger adsorption <strong>of</strong> this gas to <strong>the</strong> <strong>carbon</strong> surface at loweroperating <strong>temperature</strong>s as confirmed by gas adsorption experiments and molecular modellingsimulation.The exposure <strong>of</strong> <strong>the</strong> NPC <strong>membranes</strong> to condensable and non-condensable impuritiestypically found in natural gas and syn<strong>the</strong>sis gas saw a small reduction in permeance and aminimal reduction in selectivity, particularly at elevated operating <strong>temperature</strong>s.iii


Overall, NPC <strong>membranes</strong> have good potential as a technology suitable <strong>for</strong> CO 2 capture as<strong>the</strong>y traditionally have a better per<strong>for</strong>mance than <strong>the</strong>ir polymeric counterparts and can beoperated at <strong>high</strong>er <strong>temperature</strong>s with only a slight impact on per<strong>for</strong>mance in <strong>the</strong> presence <strong>of</strong>impurities. However, <strong>the</strong> challenges <strong>of</strong> cracking and aging are considerable and have not asyet been fully resolved.The very recent emergence <strong>of</strong> new membrane technologies that combine <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> aNPC membrane with <strong>the</strong> durability <strong>of</strong> a polymeric membrane are certainly now attractingmore attention and may well reveal <strong>the</strong>mselves as <strong>the</strong> best <strong>of</strong> both worlds.iv


PREFACESome <strong>of</strong> <strong>the</strong> findings <strong>of</strong> this <strong>the</strong>sis have been published elsewhere as follows.Journal PublicationsC. J. Anderson. S. J Pas, G. Arora, S. E. Kentish, A. J. Hill, S. I. Sandler and G. W. Stevens,“Effect <strong>of</strong> Pyrolysis Temperature and Operating Temperature on <strong>the</strong> Per<strong>for</strong>mance <strong>of</strong>Nanoporous Carbon Membranes,” Journal <strong>of</strong> Membrane Science 322 (2008) 19-27.Peer-Reviewed Conference PublicationsC. J Anderson. G. Arora, S. E. Kentish, S. I. Sandler and G. W. Stevens, “Nanoporous CarbonMembranes <strong>for</strong> CO 2 Capture,” in Chemeca 2007 – 35th Australasian Chemical EngineeringConference. 2007. Melbourne, Australia.Conference ProceedingsC. J Anderson. G. Arora, S. E. Kentish, S. I. Sandler and G. W. Stevens, “Nanoporous CarbonMembranes <strong>for</strong> CO 2 Capture,” in North American Membrane Society Annual Meeting. 2007.Orlando, United States <strong>of</strong> America.C. J Anderson, “Bridging <strong>the</strong> Gap from Fossil Fuels to Sustainable Energy Using CarbonDioxide Sequestration,” in Youth Symposium <strong>of</strong> <strong>the</strong> 19th World Energy Congress. 2004.Sydney, Australia.v


ACKNOWLEDGEMENTSThe research work <strong>for</strong> this PhD <strong>the</strong>sis was carried out during <strong>the</strong> very busy time in which mytwo daughters, Estelle and Elise, were born. I would never have been able to complete suchan undertaking as a PhD without <strong>the</strong> support and understanding <strong>of</strong> several people whom Iwould like to acknowledge here.Firstly, to my primary supervisor A/Pr<strong>of</strong>. Sandra Kentish (<strong>the</strong> University <strong>of</strong> Melbourne) <strong>for</strong>her dedication and ability to provide <strong>the</strong> most comprehensive support despite her numerouscommitments to o<strong>the</strong>r students and to <strong>the</strong> Department <strong>of</strong> Chemical and BiomolecularEngineering.Secondly to my o<strong>the</strong>r supervisors, Pr<strong>of</strong>. Ge<strong>of</strong>f Stevens (University <strong>of</strong> Melbourne) and Pr<strong>of</strong>.Stan Sandler (University <strong>of</strong> Delaware) <strong>for</strong> setting aside time in <strong>the</strong>ir busy schedules and <strong>for</strong>providing access to invaluable resources.Thanks to Barry Hooper and <strong>the</strong> Cooperative Research Centre <strong>for</strong> Greenho<strong>use</strong> GasTechnologies (CO2CRC) <strong>for</strong> <strong>the</strong> funding and also enabling me <strong>the</strong> opportunity to becomeinvolved in assisting on such important matters as reducing greenho<strong>use</strong> gas emissions. Iwould also like to acknowledge <strong>the</strong> University <strong>of</strong> Melbourne, <strong>the</strong> Particulate FluidsProcessing Centre (PFPC) and <strong>the</strong> Department <strong>of</strong> Chemical and Biomolecular Engineering <strong>for</strong>financial assistance related to visiting <strong>the</strong> University <strong>of</strong> Delaware in <strong>the</strong> USA.To Wendy Tan, <strong>the</strong> CO2CRC research assistant, <strong>for</strong> all your hard work in <strong>the</strong> lab especially inrunning experiments and handling chemicals whilst I was pregnant and on maternity leave.Your diligence and tenacity (especially in <strong>the</strong> face <strong>of</strong> cracked <strong>membranes</strong> and brokenequipment) is exceptional.Thank you also to Drs Steven Pas and Anita Hill (CSIRO) <strong>for</strong> providing <strong>the</strong> PALS analysispresented in this <strong>the</strong>sis. It has been a pleasure working with you both. Additionally, thank youto Roger Curtin (University <strong>of</strong> Melbourne) <strong>for</strong> <strong>the</strong> SEM analysis and Dr Peng Ling (RMIT)<strong>for</strong> <strong>the</strong> HRTEM analysis presented in this <strong>the</strong>sis.To Drs Gaurav Arora and Jianwen Jiang (<strong>for</strong>merly <strong>of</strong> <strong>the</strong> University <strong>of</strong> Delaware) <strong>for</strong>allowing access to <strong>the</strong>ir molecular simulation models and <strong>for</strong> <strong>the</strong>ir patience and support withhelping to get <strong>the</strong> models up and running in Australia. Additionally, thanks to Dr DaltonHarvey (University <strong>of</strong> Melbourne) <strong>for</strong> providing access to better computing power and <strong>for</strong>teaching me <strong>the</strong> basics <strong>of</strong> <strong>the</strong> Unix Language.vi


To <strong>the</strong> Department <strong>of</strong> Chemical and Biomolecular Engineering, it has been lovely to be backin <strong>the</strong> building after my eight year break in industry following undergraduate studies. It is avery welcoming and supportive work place. Additionally, thank you to Kevin Smeaton and<strong>the</strong> workshop team <strong>for</strong> making <strong>the</strong> seemingly impossible possible.To my parents, Kate and Kevin, thank you both <strong>for</strong> my upbringing. Being a parent myselfnow, I realise how much work goes in <strong>of</strong>ten without acknowledgement. So thank you. I feelproud <strong>of</strong> <strong>the</strong> values you have installed along with <strong>the</strong> unconditional support <strong>for</strong> all <strong>of</strong> my lifechoices.Great thanks also to our next door neighbour, Nancy Vella, <strong>for</strong> providing a loving home <strong>for</strong>Estelle, which enabled me some time to complete this <strong>the</strong>sis.Finally I would like express my gratitude to my husband, Darren Millard. Darren, you are <strong>the</strong>most amazing person I know and <strong>the</strong> love <strong>of</strong> my life. Your untiring support <strong>of</strong> both my studiesand in <strong>the</strong> home with our children is beyond words.vii


DECLARATIONThis is to certify that:The <strong>the</strong>sis comprises only my original work.Due acknowledgement has been made in <strong>the</strong> text to all o<strong>the</strong>r material <strong>use</strong>d.The <strong>the</strong>sis is less than 100,000 words in length, exclusive <strong>of</strong> tables, maps, bibliographies.Clare AndersonMay 2009viii


TABLE OF CONTENTSDEDICATION ........................................................................................................................ IIABSTRACT ...........................................................................................................................IIIPREFACE ................................................................................................................................VJOURNAL PUBLICATIONS ....................................................................................................... VPEER-REVIEWED CONFERENCE PUBLICATIONS..................................................................... VCONFERENCE PROCEEDINGS.................................................................................................. VACKNOWLEDGEMENTS .................................................................................................. VIDECLARATION ................................................................................................................ VIIITABLE OF CONTENTS ...................................................................................................... IXLIST OF FIGURES..............................................................................................................XVLIST OF TABLES.............................................................................................................XXIINOMENCLATURE ........................................................................................................ XXVINON-ENGLISH CHARACTERS .......................................................................................... XXVIISUBSCRIPTS.................................................................................................................... XXVIII1 INTRODUCTION.............................................................................................................11.1 GLOBAL WARMING......................................................................................................11.2 CO 2 CAPTURE AND SEQUESTRATION...........................................................................21.3 CAPTURE TECHNOLOGIES ............................................................................................32 LITERATURE REVIEW.................................................................................................72.1 INTRODUCTION.............................................................................................................72.2 GAS SEPARATION MEMBRANES...................................................................................72.2.1 Per<strong>for</strong>mance Indicators........................................................................................72.2.1.1 Permeance and Selectivity ................................................................................72.2.1.2 Diffusivity and Heat <strong>of</strong> Adsorption...................................................................92.2.2 Per<strong>for</strong>mance Measurement.................................................................................102.2.2.1 Permeance and Selectivity ..............................................................................102.2.2.2 Diffusivity .......................................................................................................122.2.2.3 Heat <strong>of</strong> Adsorption..........................................................................................132.2.3 Materials and Transport Mechanisms................................................................152.2.3.1 Microporous Membranes ................................................................................15ix


2.2.3.2 Nanoporous Membranes .................................................................................172.2.3.3 Nonporous Membranes ...................................................................................182.2.3.4 Membranes <strong>of</strong> Mixed Materials......................................................................192.2.3.5 Summary .........................................................................................................192.3 MANUFACTURE OF NPC MEMBRANES ......................................................................202.3.1 Supported NPC Membranes ...............................................................................212.3.1.1 Polymer ...........................................................................................................212.3.1.2 Porous Support................................................................................................242.3.1.3 Modified Porous Support ................................................................................242.3.1.4 Deposition Method..........................................................................................252.3.1.5 Pyrolysis..........................................................................................................252.3.1.6 Number <strong>of</strong> Coats .............................................................................................292.3.2 Unsupported NPC Membranes...........................................................................292.3.2.1 Polymer ...........................................................................................................292.3.2.2 Pyrolysis Temperature ....................................................................................302.3.3 Characterisation.................................................................................................312.3.3.1 Membrane Thickness ......................................................................................312.3.3.2 Surface Structure.............................................................................................312.3.3.3 Pore Size Distribution .....................................................................................312.4 PERFORMANCE OF NPC MEMBRANES .......................................................................342.4.1 Supported NPC Membranes ...............................................................................342.4.2 Unsupported NPC Membranes...........................................................................352.5 GAS TRANSPORT MECHANISMS OF NPC MEMBRANES .............................................362.5.1 Experimental Observations ................................................................................362.5.2 Molecular Modelling ..........................................................................................372.5.2.1 Monte Carlo Simulation..................................................................................372.5.2.2 Molecular Dynamics Simulation.....................................................................392.5.2.3 Published Simulation Studies <strong>of</strong> NPC.............................................................402.6 HIGH TEMPERATURE OPERATION OF NPC MEMBRANES ..........................................432.6.1 Arrhenius Relationships......................................................................................432.7 EFFECT OF IMPURITIES ON THE PERFORMANCE NPC MEMBRANES ..........................472.7.1 Water ..................................................................................................................472.7.2 Air .......................................................................................................................492.7.3 Condensable Impurities......................................................................................512.7.4 Non-Condensable Impurities ..............................................................................532.8 SCOPE OF THIS RESEARCH WORK ..............................................................................55x


3 EXPERIMENTAL MATERIALS, APPARATUS AND METHODS........................573.1 INTRODUCTION...........................................................................................................573.2 MEMBRANE MATERIALS, MANUFACTURING EQUIPMENT AND TECHNIQUES ...........573.2.1 Materials.............................................................................................................573.2.2 Manufacturing Equipment and Techniques........................................................593.2.2.1 Supported NPC Membranes............................................................................593.2.2.2 Modified Supported NPC Membranes............................................................643.2.2.3 Unsupported NPC Membranes .......................................................................663.3 MEMBRANE CHARACTERISATION TECHNIQUES ........................................................673.3.1 Surface Structure ................................................................................................673.3.2 Pore Size.............................................................................................................683.3.2.1 PALS...............................................................................................................683.3.2.2 WAXD ............................................................................................................693.4 MEASUREMENT OF MEMBRANE PERMEANCE AND SELECTIVITY..............................703.4.1 Pure Gas .............................................................................................................703.4.2 Mixed Gas...........................................................................................................723.4.3 At Elevated Temperatures...................................................................................753.4.4 Exposure to Impurities........................................................................................763.4.4.1 Condensable Impurities...................................................................................763.4.4.2 Non-Condensable Impurities ..........................................................................793.5 MEASUREMENT OF MEMBRANE ADSORPTION...........................................................823.5.1 Theoretical Adsorption .......................................................................................823.5.1.1 Model Development........................................................................................823.5.1.2 Model Description...........................................................................................823.5.1.3 Fitting <strong>the</strong> Adsorption Iso<strong>the</strong>rms ....................................................................833.5.1.4 Calculating <strong>the</strong> Adsorption Separation Factor ................................................833.5.2 Experimental Adsorption ....................................................................................843.5.2.1 Buoyancy Correction ......................................................................................853.5.2.2 Fitting <strong>the</strong> Adsorption Iso<strong>the</strong>rms ....................................................................874 MANUFACTURE AND HIGH TEMPERATURE OPERATION OF NPCMEMBRANES........................................................................................................................894.1 INTRODUCTION...........................................................................................................894.2 SUPPORTED NPC MEMBRANES..................................................................................894.2.1 Manufacture........................................................................................................894.2.2 Characterisation.................................................................................................914.2.2.1 Surface Structure.............................................................................................91xi


4.2.2.2 Pore Size .........................................................................................................934.2.3 Pure Gas Per<strong>for</strong>mance .......................................................................................974.2.4 Mixed Gas Per<strong>for</strong>mance ...................................................................................1024.2.4.1 Effect <strong>of</strong> Feed Pressure .................................................................................1034.2.5 High Temperature Per<strong>for</strong>mance.......................................................................1054.2.5.1 Effect on Permeance .....................................................................................1054.2.5.2 Effect on Selectivity......................................................................................1074.2.5.3 Effect <strong>of</strong> Thermal Cycling ............................................................................1094.2.6 Concluding Remarks on Supported NPC Membranes......................................1104.3 MODIFIED-SUPPORTED NPC MEMBRANES..............................................................1124.3.1 Manufacture and Characterisation ..................................................................1124.3.2 Membrane Per<strong>for</strong>mance ...................................................................................1144.3.3 Concluding Remarks on Modified-Supported NPC Membranes ......................1165 UNDERSTANDING THE HIGH TEMPERATURE PERFORMANCE OF NPCMEMBRANES THROUGH GAS ADSORPTION MOLECULAR SIMULATION ANDEXPERIMENTS...................................................................................................................1175.1 INTRODUCTION.........................................................................................................1175.2 MOLECULAR SIMULATION .......................................................................................1175.2.1 Pure Gases........................................................................................................1175.2.1.1 Heats <strong>of</strong> Adsorption ......................................................................................1195.2.2 Mixed Gases......................................................................................................1215.2.2.1 Carbon Dioxide / Methane............................................................................1215.2.2.2 Carbon Dioxide / Nitrogen............................................................................1235.3 GAS ADSORPTION EXPERIMENTS.............................................................................1255.3.1 Pure Gases........................................................................................................1255.3.1.1 Carbon Dioxide.............................................................................................1255.3.1.2 Nitrogen ........................................................................................................1285.3.1.3 Heats <strong>of</strong> Adsorption ......................................................................................1305.3.1.4 Relating <strong>the</strong> CO 2 Heat <strong>of</strong> Adsorption to <strong>the</strong> CO 2 Permeance .......................1315.4 CONCLUDING REMARKS ON UNDERSTANDING THE HIGH TEMPERATUREPERFORMANCE OF NPC MEMBRANES THROUGH GAS ADSORPTION MOLECULARSIMULATION AND EXPERIMENTS ........................................................................................1326 PERFORMANCE OF NPC MEMBRANES IN THE PRESENCE OFCONDENSABLE AND NON-CONDENSABLE IMPURITIES .....................................1336.1 INTRODUCTION.........................................................................................................133xii


6.2 CONDENSABLE IMPURITIES......................................................................................1336.2.1 Water ................................................................................................................1346.2.2 Hexane ..............................................................................................................1376.2.3 Toluene .............................................................................................................1406.2.4 Comparison <strong>of</strong> Condensable Impurities ...........................................................1436.3 NON-CONDENSABLE IMPURITIES.............................................................................1446.3.1 Hydrogen Sulphide ...........................................................................................1446.3.2 Carbon Monoxide .............................................................................................1466.3.3 Comparison <strong>of</strong> Non-Condensable Impurities ...................................................1496.4 CONCLUDING REMARKS ON THE PERFORMANCE OF NPC MEMBRANES IN THEPRESENCE OF IMPURITIES ...................................................................................................1507 CONCLUSIONS AND RECOMMENDATIONS ......................................................1537.1 CONCLUSIONS ..........................................................................................................1537.2 RECOMMENDATIONS FOR FURTHER WORK .............................................................1568 REFERENCES..............................................................................................................159APPENDICES ......................................................................................................................175APPENDIX A – RELEVANT BACKGROUND INFORMATION ................................177A.1 GLOBAL WARMING..................................................................................................177A.2 INTERNATIONAL RECOGNITION OF CLIMATE CHANGE............................................178A.3 AUSTRALIAN RECOGNITION OF CLIMATE CHANGE.................................................179A.4 REDUCING CO 2 EMISSIONS FROM AUSTRALIA’S ENERGY PRODUCTION ................180A.5 OPPORTUNITIES FOR CO 2 SEQUESTRATION IN AUSTRALIA .....................................181A.6 CHALLENGES FOR CO 2 SEQUESTRATION IN AUSTRALIA.........................................182A.7 CO 2 CAPTURE TECHNOLOGIES ................................................................................183A.7.1 Absorption.........................................................................................................183A.7.2 Adsorption.........................................................................................................183A.7.3 Cryogenics ........................................................................................................184A.7.4 Gas Cryogenics.................................................................................................184A.7.5 Gas Separation Membranes .............................................................................184APPENDIX B – STANDARD OPERATING PROCEDURES ........................................187B.1 STANDARD OPERATING PROCEDURE FOR THE CPVV (MIXED GAS) APPARATUS....187B.2 STANDARD OPERATING PROCEDURE FOR THE CPVV (MIXED GAS) APPARATUS INTHE PRESENCE OF CONDENSABLE IMPURITIES ....................................................................199xiii


B.3 STANDARD OPERATING PROCEDURE FOR THE CPVV (MIXED GAS) APPARATUS INTHE PRESENCE OF NON-CONDENSABLE IMPURITIES............................................................211B.4 RISK ASSESSMENT FOR THE OPERATION OF THE CPVV (MIXED GAS) APPARATUS INTHE PRESENCE OF NON-CONDENSABLE IMPURITIES............................................................223APPENDIX C – EXPERIMENTAL METHODS..............................................................231C.1 PURE GAS PERMEANCE MEASUREMENT..................................................................231C.1.1 Error Analysis...................................................................................................231C.1.2 Sample <strong>of</strong> Calculated Results ...........................................................................234C.2 MIXED GAS PERMEANCE MEASUREMENT ...............................................................236C.2.1 Error Analysis...................................................................................................236C.2.2 Sample <strong>of</strong> Calculated Results ...........................................................................238C.3 ADSORPTION MEASUREMENT ..................................................................................239C.3.1 Error Analysis...................................................................................................239C.3.2 Sample <strong>of</strong> Calculated Results ...........................................................................241C.4 MOLECULAR MODELLING........................................................................................242C.4.1 Sample Input File..............................................................................................242C.4.2 Sample Output File...........................................................................................243C.4.3 Sample Iso<strong>the</strong>rm Data ......................................................................................244APPENDIX D – UNSUPPORTED NPC MEMBRANES .................................................245D.1 MANUFACTURE AND CHARACTERISATION ..............................................................245D.1.1 Matrimid Polyimide..........................................................................................245D.1.2 Kapton Polyimide .............................................................................................247D.2 PERFORMANCE .........................................................................................................248D.2.1 Matrimid Polyimide..........................................................................................248D.2.2 Kapton Polyimide .............................................................................................249D.3 CONCLUDING REMARKS ON UNSUPPORTED NPC MEMBRANES .............................250xiv


LIST OF FIGURESFigure 1-1 : CO 2 capture and sequestration process. Diagram provided by <strong>the</strong> CooperativeResearch Centre <strong>for</strong> Greenho<strong>use</strong> Gas Technologies (CO2CRC) [5].......................2Figure 1-2 : Separation <strong>of</strong> CO 2 from (a) natural gas and (b) air-blown syn<strong>the</strong>sis gas .............4Figure 2-1 : Membrane separation unit.....................................................................................7Figure 2-2 : Equipment <strong>use</strong>d <strong>for</strong> permeance or permeability measurement; (a) constantpressure variable volume apparatus (CPVV) <strong>for</strong> <strong>high</strong>er flow <strong>membranes</strong> and (b)constant volume variable pressure (CVVP) <strong>for</strong> lower flow <strong>membranes</strong> ................10Figure 2-3 : An extension <strong>of</strong> <strong>the</strong> constant pressure variable volume apparatus (CPVV) <strong>use</strong>d<strong>for</strong> permeance and selectivity measurement using a mixed gas feed. ....................11Figure 2-4 : Calculation <strong>of</strong> <strong>the</strong> diffusion coefficient using <strong>the</strong> time-lag method, once <strong>the</strong>gradient is constant and steady state flow through <strong>the</strong> membrane has beenreached, an extrapolation <strong>of</strong> <strong>the</strong> steady state flow line back to <strong>the</strong> x-axis where <strong>the</strong>gas flow is 0 reveals <strong>the</strong> value <strong>of</strong> <strong>the</strong> time lag (θ). .................................................12Figure 2-5 : Summary <strong>of</strong> transport mechanisms indicating as <strong>the</strong> size <strong>of</strong> <strong>the</strong> membrane poresdecrease, transport changes from bulk flow to molecular sieving. ........................19Figure 2-6 : Structure <strong>of</strong> a typical fullerene (<strong>the</strong> Buckminster Fullerene C60) discovered in1985 by Kroto et al [56].........................................................................................20Figure 2-7 : Structure <strong>of</strong> a hexagonal sheet, which when randomly bonded to o<strong>the</strong>r sheets<strong>for</strong>ms <strong>the</strong> curved <strong>nanoporous</strong> <strong>carbon</strong> structure. Figure adapted from [61]. .........21Figure 2-8 : Typical furnace <strong>temperature</strong> pr<strong>of</strong>ile <strong>for</strong> <strong>the</strong> pyrolysis <strong>of</strong> a polymer to produce<strong>nanoporous</strong> <strong>carbon</strong>.................................................................................................26Figure 2-9 : Per<strong>for</strong>mance <strong>of</strong> unsupported NPC <strong>membranes</strong> [17] with respect to <strong>the</strong> RobesonUpper Bound [38] <strong>for</strong> polymeric <strong>membranes</strong>, which shows that <strong>the</strong> <strong>carbon</strong>isation<strong>of</strong> a Matrimid polymeric membrane increases <strong>the</strong> permeability and selectivity suchthat it breaks <strong>the</strong> upper bound................................................................................35Figure 2-10 : The process that is <strong>use</strong>d by molecular simulation to create flux and separationfactors <strong>for</strong> <strong>nanoporous</strong> <strong>carbon</strong>. The two simulation methods <strong>use</strong>d are (1) GrandCanonical Monte Carlo (GCMC) simulation <strong>for</strong> determining <strong>the</strong> adsorptionproperties and (2) molecular dynamic (MD) simulation <strong>for</strong> determining <strong>the</strong> selfdiffusivities.Combining <strong>the</strong> results from both methods produces a predicted fluxand separation factor. ............................................................................................40Figure 2-11 : Structure <strong>of</strong> C168 Schwarzite <strong>use</strong>d to represent <strong>nanoporous</strong> <strong>carbon</strong> [106]......40Figure 3-1 : Chemical structure <strong>of</strong> Matrimid polyimide (Vantico, Switzerland) .....................59Figure 3-2 : Chemical structure <strong>of</strong> Kapton polyimide (Dupont, USA) ....................................59xv


Figure 3-3 : (a) Photograph and (b) schematic diagram <strong>of</strong> <strong>the</strong> spin coating apparatus. Theporous support is held onto a spinning chuck using a vacuum pump and a sealpurge gas. The PFA solution is sprayed onto <strong>the</strong> spinning porous support using apneumatic spray nozzle...........................................................................................60Figure 3-4 : (a) Photograph and (b) schematic diagram <strong>of</strong> <strong>the</strong> quartz tube furnace with acontinuous purge <strong>of</strong> ultra <strong>high</strong> purity argon. .........................................................61Figure 3-5 : Quartz tube furnace typical <strong>temperature</strong> pr<strong>of</strong>ile with time during pyrolysis <strong>of</strong>supported NPC <strong>membranes</strong>....................................................................................62Figure 3-6 : The impact <strong>of</strong> oxidising <strong>the</strong> first coat <strong>of</strong> <strong>carbon</strong> <strong>of</strong> <strong>membranes</strong> manufactured at550°C. The increase in permeance is 1 to 2 orders <strong>of</strong> magnitude. This result issimilar to that achieved from <strong>the</strong> addition <strong>of</strong> silica particles [28] [27].................63Figure 3-7 : Apparatus <strong>for</strong> creating <strong>the</strong> colloidal solution <strong>of</strong> silica nanoparticles in which asolution containing ethanol and ammonia was mixed with a solution containingethanol and TEOS and reacted at a constant <strong>temperature</strong>.....................................65Figure 3-8 : Millipore filtration cell <strong>use</strong>d <strong>for</strong> filtering diluted silica nanoparticle colloidsolution through <strong>the</strong> porous stainless steel support. ..............................................66Figure 3-9 : Quartz tube furnace typical <strong>temperature</strong> pr<strong>of</strong>ile with time during pyrolysis <strong>of</strong>unsupported NPC <strong>membranes</strong>................................................................................67Figure 3-10 : (a) Photograph and (b) process flow diagram <strong>of</strong> <strong>the</strong> constant volume, variablepressure (CVVP) apparatus <strong>for</strong> measuring pure gas permeance andpermselectivity........................................................................................................70Figure 3-11 : Photograph <strong>of</strong> <strong>the</strong> CVVP membrane cell showing one inlet connection <strong>for</strong> <strong>the</strong>feed gas and one outlet connection <strong>for</strong> <strong>the</strong> permeate gas.......................................71Figure 3-12 : Process flow diagram <strong>of</strong> <strong>the</strong> constant pressure, variable volume apparatus(CPVV) <strong>for</strong> measuring mixed gas permeance and selectivity. ...............................72Figure 3-13 : Photograph <strong>of</strong> <strong>the</strong> CPVV membrane cell showing two inlet connections <strong>for</strong> <strong>the</strong>feed and sweep gases and two outlet connections <strong>for</strong> <strong>the</strong> retentate and permeatestreams....................................................................................................................73Figure 3-14 : Photograph <strong>of</strong> <strong>the</strong> vessel <strong>use</strong>d to saturate <strong>the</strong> feed gas <strong>of</strong> <strong>the</strong> constant pressurevariable volume (CPVV) apparatus <strong>for</strong> measuring mixed gas permeance andpermselectivity in <strong>the</strong> presence <strong>of</strong> condensable impurities.....................................76Figure 3-15 : Process flow diagrams <strong>of</strong> (a) <strong>the</strong> constant pressure, variable volume (CPVV)apparatus <strong>for</strong> measuring mixed gas permeance and selectivity in <strong>the</strong> presence <strong>of</strong>condensable impurities (b) vessel <strong>for</strong> contacting feed gas with condensableimpurities................................................................................................................77xvi


Figure 3-16 : Process flow diagram <strong>of</strong> <strong>the</strong> constant pressure, variable volume apparatus(CPVV) <strong>for</strong> measuring mixed gas permeance and selectivity in <strong>the</strong> presence <strong>of</strong>non-condensable impurities....................................................................................79Figure 3-17 : Photograph <strong>of</strong> <strong>the</strong> gas extraction system installed around <strong>the</strong> gas bottlecontaining H 2 S or CO impurities............................................................................81Figure 3-18 : Structure <strong>of</strong> C 168 Schwarzite from <strong>the</strong> side (left) and rotated at 20°C (right)....83Figure 3-19 : (a) Photograph and (b) schematic diagram <strong>of</strong> <strong>the</strong> microbalance <strong>use</strong>d <strong>for</strong>measuring gas adsorption iso<strong>the</strong>rms at elevated <strong>temperature</strong>s. ............................84Figure 4-1 : The creation <strong>of</strong> a NPC membrane within 3 coats <strong>of</strong> <strong>the</strong> polymeric precursor.....89Figure 4-2 : The cumulative <strong>carbon</strong> density per coat <strong>for</strong> NPC <strong>membranes</strong> manufactured atpyrolysis <strong>temperature</strong>s between 400 and 600°C. Also depicted in this figure is <strong>the</strong>cumulative <strong>carbon</strong> density <strong>of</strong> an NPC membrane made by Rajagopalan and Foleyat 600°C [82] with a final O 2 /N 2 permselectivity <strong>of</strong> 5............................................91Figure 4-3 : (a) SEM pictures showing (clockwise from top left) <strong>the</strong> uncoated stainless steelsupport, <strong>the</strong> <strong>carbon</strong> membrane after three coats with no defects, macroscopiccracking in <strong>the</strong> <strong>carbon</strong> surface and microscopic cracking in <strong>the</strong> <strong>carbon</strong> surface,(b) HRTEM pictures shown (from left) <strong>nanoporous</strong> <strong>carbon</strong> and enlarged photo toshow pore size detail. .............................................................................................92Figure 4-4 : Pore size distribution <strong>of</strong> NPC produced at 600°C obtained from <strong>the</strong> HRTEMpicture using <strong>the</strong> image processing program Image J. The distribution range isfrom 2.7 to 5.2 Å centred around 4.0 Å. .................................................................93Figure 4-5 : Characterisation <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> material manufactured at pyrolysis<strong>temperature</strong>s from 400 to 600°C using positron annihilation lifetime spectroscopy(PALS). The primary x-axis (τ 2 ) is <strong>the</strong> lifetime <strong>of</strong> position annihilating in <strong>the</strong> pores,which indicates a decrease in pore size with increasing pyrolysis <strong>temperature</strong>. Thesecondary y-axis (I 2 ) is <strong>the</strong> number <strong>of</strong> positrons annihilating in <strong>the</strong> pores, whichindicates an increase in <strong>the</strong> number <strong>of</strong> pores with increasing pyrolysis<strong>temperature</strong>.............................................................................................................94Figure 4-6 : Characterisation <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> material manufactured at pyrolysis<strong>temperature</strong>s <strong>of</strong> 400, 600 and 800°C using wide angle x-ray diffraction (WAXD).The primary x-axis is <strong>the</strong> diffraction angle whilst number <strong>of</strong> counts or intensity isgiven on <strong>the</strong> primary y-axis. The values <strong>of</strong> 2θ are converted to d-spacing usingBragg’s Law. ..........................................................................................................95Figure 4-7 : The effect <strong>of</strong> pyrolysis <strong>temperature</strong> on permeance showing a trend <strong>of</strong> increasingpermeance with increasing pyrolysis <strong>temperature</strong>.................................................98Figure 4-8 : The effect <strong>of</strong> pyrolysis <strong>temperature</strong> on permselectivity showing no clear trendbetween selectivity and pyrolysis <strong>temperature</strong>. ......................................................99xvii


Figure 4-9 : Change in permeance and selectivity with number <strong>of</strong> coats <strong>for</strong> NPC membrane Dmade at 400°C. Knudsen diffusion dominates until a selective membrane is <strong>for</strong>medwith <strong>the</strong> third coat. However, <strong>the</strong> final 4 th coat results in cracking <strong>of</strong> <strong>the</strong> membranelayer and a subsequent decrease in selectivity and increase in permeance.........100Figure 4-10 : Change in (a) CO 2 /CH 4 permselectivity and (b) CO 2 / N 2 permselectivity withamount <strong>of</strong> <strong>carbon</strong> deposited <strong>for</strong> selected <strong>membranes</strong> manufactured at pyrolysis<strong>temperature</strong>s <strong>of</strong> 400 and 550 °C. The decrease in selectivity after a maximumvalue has been reached represents <strong>the</strong> point at which cracking occurs. For <strong>the</strong><strong>membranes</strong> manufactured at 550°C, this critical point is at ~ 6 mg/cm 2 , whereas<strong>for</strong> <strong>the</strong> <strong>membranes</strong> manufactured at 400°C, this critical point is at ~ 8 mg/cm 2 . 101Figure 4-11 : Change in CO 2 permeance with increasing feed pressure. Increasing <strong>the</strong> feedpressure reduces <strong>the</strong> permeance...........................................................................104Figure 4-12 : Change in CO 2 /N 2 selectivity with increasing feed pressure. Increasing <strong>the</strong> feedpressure has <strong>the</strong> slightly increasing <strong>the</strong> selectivity...............................................104Figure 4-13 : Effect <strong>of</strong> operating <strong>temperature</strong> on mixed gas permeance. The permeanceincreases with increasing operating <strong>temperature</strong> due to <strong>the</strong> increased diffusion.The membrane manufactured at 550°C maintains a <strong>high</strong>er permeance across <strong>the</strong>operating <strong>temperature</strong> range................................................................................105Figure 4-14 : Arrhenius plot <strong>of</strong> permeance and operating <strong>temperature</strong>. The gradient <strong>of</strong> <strong>the</strong>fitted lines indicate <strong>the</strong> activation energies <strong>of</strong> permeance. The fitted lines aresteeper <strong>for</strong> CH 4 resulting in <strong>high</strong>er activation energies as presented in Table 4-10...............................................................................................................................106Figure 4-15 : Effect <strong>of</strong> operating <strong>temperature</strong> on mixed gas selectivity <strong>for</strong> NPC <strong>membranes</strong>manufactured at pyrolysis <strong>temperature</strong>s <strong>of</strong> 400 and 550°C. Increasing <strong>the</strong>operating <strong>temperature</strong> has <strong>the</strong> effect <strong>of</strong> decreasing <strong>the</strong> selectivity due to <strong>the</strong>disproportional increase in <strong>the</strong> CH 4 permeance. This is due <strong>the</strong> significantreduction in <strong>the</strong> CO 2 adsorption at <strong>the</strong> <strong>high</strong>er <strong>temperature</strong>s. ..............................108Figure 4-16 : Extrapolated effect <strong>of</strong> operating <strong>temperature</strong> on mixed gas selectivity <strong>for</strong> NPC<strong>membranes</strong> manufactured at pyrolysis <strong>temperature</strong>s <strong>of</strong> 400 and 550°C.Extrapolating <strong>the</strong> CO 2 /CH 4 selectivity relationship with operating <strong>temperature</strong>using an exponential fit shows that <strong>the</strong> selectivity will approach <strong>the</strong> Knudsen value<strong>of</strong> 0.6 at an operating <strong>temperature</strong> between 400 and 500°C................................108Figure 4-17 : Effect <strong>of</strong> <strong>the</strong>rmal cycling on an NPC membrane. After eleven regenerationcycles <strong>the</strong> CO 2 and CH 4 permeance increased dramatically whilst <strong>the</strong> selectivitydropped to 1, indicating <strong>the</strong> presence <strong>of</strong> significant cracks. ................................109Figure 4-18 : Silica nanoparticles, with a size less than 300 nm produced from <strong>the</strong> reaction <strong>of</strong>ethanol, ammonia and TEOS at a <strong>temperature</strong> <strong>of</strong> 50 °C <strong>for</strong> 45 mins...................112xviii


Figure 4-19 : Effect <strong>of</strong> <strong>the</strong> addition <strong>of</strong> silica nanoparticles on reducing <strong>the</strong> permeance <strong>of</strong> <strong>the</strong>stainless steel support suggesting some reduction in pore size and porosity.......113Figure 4-20 : Effect <strong>of</strong> <strong>the</strong> addition <strong>of</strong> <strong>carbon</strong> on permeance to supports filled with silicananoparticles. The membrane layer begins to <strong>for</strong>m around 1.5 – 2 mg/cm 2 <strong>of</strong><strong>carbon</strong> but cracks at 3.5 mg/cm 2 . .........................................................................115Figure 4-21 : Effect <strong>of</strong> <strong>the</strong> addition <strong>of</strong> <strong>carbon</strong> on selectivity to supports filled with silicananoparticles. The selectivity results agree with <strong>the</strong> permeance results, whichsuggest cracking <strong>of</strong> <strong>the</strong> <strong>for</strong>med membrane layer occurs around 3.5 mg/cm 2 .......115Figure 5-1 : Simulated adsorption iso<strong>the</strong>rms <strong>of</strong> pure CO 2 onto <strong>nanoporous</strong> <strong>carbon</strong>.Increasing <strong>the</strong> <strong>temperature</strong> decreases <strong>the</strong> amount <strong>of</strong> CO 2 adsorbed. ..................118Figure 5-2 : Simulated adsorption iso<strong>the</strong>rms <strong>of</strong> pure CH 4 onto <strong>nanoporous</strong> <strong>carbon</strong>.Increasing <strong>the</strong> <strong>temperature</strong> decreases <strong>the</strong> amount <strong>of</strong> CH 4 adsorbed. ..................118Figure 5-3 : Simulated adsorption iso<strong>the</strong>rms <strong>of</strong> pure N 2 onto <strong>nanoporous</strong> <strong>carbon</strong>. Increasing<strong>the</strong> <strong>temperature</strong> decreases <strong>the</strong> amount <strong>of</strong> N 2 adsorbed. .......................................119Figure 5-4 : Straight line plot <strong>of</strong> 1/T versus <strong>the</strong> simulated Ln(b) <strong>for</strong> pure gases <strong>for</strong>determining <strong>the</strong> heat <strong>of</strong> adsorption (∆H ads ). .........................................................120Figure 5-5 : Adsorption iso<strong>the</strong>rms <strong>of</strong> mixed CO 2 /CH 4 onto <strong>nanoporous</strong> <strong>carbon</strong>. Increasing <strong>the</strong><strong>temperature</strong> decreases <strong>the</strong> amount <strong>of</strong> CO 2 and CH 4 adsorbed.............................121Figure 5-6 : CO 2 /CH 4 mixed gas adsorption separation factor as a function <strong>of</strong> pressure <strong>for</strong>system <strong>temperature</strong>s <strong>of</strong> 35 and 200°C..................................................................122Figure 5-7 : Adsorption iso<strong>the</strong>rms <strong>of</strong> mixed CO 2 /N 2 onto <strong>nanoporous</strong> <strong>carbon</strong>. Increasing <strong>the</strong><strong>temperature</strong> decreases <strong>the</strong> amount <strong>of</strong> CO 2 and N 2 adsorbed. ..............................123Figure 5-8 : CO 2 /N 2 mixed gas adsorption separation factor as a function <strong>of</strong> pressure <strong>for</strong>system <strong>temperature</strong>s <strong>of</strong> 35 and 200°C..................................................................124Figure 5-9 : Experimental adsorption iso<strong>the</strong>rms <strong>of</strong> pure CO 2 onto <strong>nanoporous</strong> <strong>carbon</strong>.......125Figure 5-10 : Experimental and simulated adsorption iso<strong>the</strong>rms <strong>of</strong> pure CO 2 onto NPC.....126Figure 5-11 : Experimental adsorption iso<strong>the</strong>rms <strong>of</strong> pure N 2 onto <strong>nanoporous</strong> <strong>carbon</strong>. ......128Figure 5-12 : Experimental and simulated adsorption iso<strong>the</strong>rms <strong>of</strong> pure N 2 onto <strong>nanoporous</strong><strong>carbon</strong>. The large discrepancy at 35°C is attributed to not achieving equilibriumin <strong>the</strong> adsorption experiments...............................................................................129Figure 5-13 : Straight line plot <strong>of</strong> 1/T versus <strong>the</strong> simulated Ln(b) <strong>for</strong> pure gases <strong>for</strong>determining <strong>the</strong> heat <strong>of</strong> adsorption (∆H ads ). .........................................................130Figure 6-1 : Normalised effect <strong>of</strong> water concentration on <strong>the</strong> reduction in CO 2 permeance <strong>of</strong> aNPC membrane manufactured at 550°C as a function <strong>of</strong> time. ...........................134Figure 6-2 : Normalised effect <strong>of</strong> water concentration on <strong>the</strong> CO 2 /CH 4 selectivity <strong>of</strong> a NPCmembrane manufactured at 550°C as a function <strong>of</strong> time. ....................................135xix


Figure 6-3 : Normalised effect <strong>of</strong> water concentration on <strong>the</strong> permeance <strong>of</strong> a NPC membranemanufactured at 550°C. The effect appears to be more pronounced on CO 2permeance and at lower operating <strong>temperature</strong>s.................................................136Figure 6-4 : Normalised effect <strong>of</strong> water concentration on <strong>the</strong> selectivity <strong>of</strong> a NPC membranemanufactured at 550°C. The effect is minimal. ....................................................136Figure 6-5 : Normalised effect <strong>of</strong> hexane concentration on <strong>the</strong> reduction in CO 2 permeance <strong>of</strong>a NPC membrane manufactured at 550°C as a function <strong>of</strong> time. ........................137Figure 6-6 : Normalised effect <strong>of</strong> hexane concentration on <strong>the</strong> CO 2 /CH 4 permeance <strong>of</strong> a NPCmembrane manufactured at 550°C as a function <strong>of</strong> time. ....................................138Figure 6-7: Normalised effect <strong>of</strong> hexane concentration on <strong>the</strong> permeance <strong>of</strong> a NPC membranemanufactured at 550°C. The effect appears to be more pronounced on <strong>the</strong> CO 2permeance and at lower operating <strong>temperature</strong>s.................................................139Figure 6-8 : Normalised effect <strong>of</strong> hexane concentration on <strong>the</strong> CO 2 /CH 4 selectivity <strong>of</strong> a NPCmembrane manufactured at 550°C. The impact is minimal. ................................139Figure 6-9 : Normalised effect <strong>of</strong> toluene concentration on <strong>the</strong> reduction in CO 2 permeance <strong>of</strong>a NPC membrane manufactured at 550°C as a function <strong>of</strong> time. ........................140Figure 6-10 : Normalised effect <strong>of</strong> toluene concentration on <strong>the</strong> CO 2 /CH 4 permeance <strong>of</strong> aNPC membrane manufactured at 550°C as a function <strong>of</strong> time. ...........................141Figure 6-11 : Normalised effect <strong>of</strong> toluene concentration on <strong>the</strong> permeance <strong>of</strong> a NPCmembrane manufactured at 550°C. The effect is more pronounced on <strong>the</strong> CO 2permeance and at lower operating <strong>temperature</strong>s.................................................142Figure 6-12 : Normalised effect <strong>of</strong> toluene concentration on <strong>the</strong> selectivity <strong>of</strong> a NPCmembrane manufactured at 550°C. The effect is minimal. ..................................142Figure 6-13 : Normalised effect <strong>of</strong> H 2 S on <strong>the</strong> reduction in CO 2 permeance <strong>of</strong> an NPCmembrane manufactured at 550°C. The effect is minimal. ..................................145Figure 6-14 : Normalised effect <strong>of</strong> H 2 S on <strong>the</strong> reduction in CO 2 /N 2 selectivity <strong>of</strong> an NPCmembrane manufactured at 550°C. The effect is minimal. ..................................145Figure 6-15 : Normalised effect <strong>of</strong> CO on <strong>the</strong> reduction in CO 2 permeance <strong>of</strong> an NPCmembrane manufactured at 550°C. The effect is greatest at <strong>the</strong> lowest pressureand <strong>temperature</strong>....................................................................................................147Figure 6-16 : Normalised effect <strong>of</strong> CO on <strong>the</strong> reduction in CO 2 /N 2 selectivity <strong>of</strong> an NPCmembrane manufactured at 550°C. The effect is minimal. ..................................147Figure A-1 : Correlation between <strong>the</strong> Earth’s <strong>temperature</strong> and atmospheric <strong>carbon</strong> dioxide(data sourced from [163])....................................................................................177Figure A-2 : Predicted greenho<strong>use</strong> gas emissions from by sector <strong>for</strong> 1990 and <strong>the</strong> predicted2008 to 2012 period (data sourced from [166])...................................................180xx


Figure A-3 : Australia’s large stationary sources <strong>of</strong> <strong>carbon</strong> dioxide by industry (data courtesy<strong>of</strong> <strong>the</strong> CRC <strong>for</strong> Greenho<strong>use</strong> Gas Technologies) ...................................................182Figure C-1 : Permeate pressure rise <strong>for</strong> pure gases CO 2 , N 2 and CH 4 through MembraneNPC550B. The respective permeances were calculated from <strong>the</strong> steady stategradients...............................................................................................................234Figure C-2 : Input file <strong>use</strong>d <strong>for</strong> <strong>the</strong> molecular simulation <strong>of</strong> <strong>the</strong> adsorption <strong>of</strong> a CO 2 :CH 4mixture (10:90 vol%) at a system <strong>temperature</strong> <strong>of</strong> 300 K and pressure <strong>of</strong> 100 kPa...............................................................................................................................242Figure C-3 : Output file produced from <strong>the</strong> molecular simulation <strong>of</strong> <strong>the</strong> adsorption <strong>of</strong> aCO 2 :CH 4 mixture (10:90 vol%) at a system <strong>temperature</strong> <strong>of</strong> 300 K and pressure <strong>of</strong>100 kPa.................................................................................................................243Figure D-1 : Weight loss <strong>of</strong> unsupported NPC <strong>membranes</strong> made from Matrimid polyimidewith increasing pyrolysis <strong>temperature</strong>. The transition from polymer to <strong>carbon</strong> iswithin 400 and 600°C...........................................................................................246Figure D-2 : SEM photographs <strong>of</strong> Matrimid (a) after pyrolysis at 420°C and (b) afterpyrolysis at 430°C. ...............................................................................................247Figure D-3 : Weight loss <strong>of</strong> unsupported NPC <strong>membranes</strong> made from Kapton polyimide withincreasing <strong>temperature</strong>. The transition from to <strong>carbon</strong> is between 500 and 700°C...............................................................................................................................247Figure D-4 : SEM photograph <strong>of</strong> Kapton after pyrolysis at 450°C. The surface was coatedwith a gold layer to enable SEM imaging suggesting <strong>the</strong> absence <strong>of</strong> <strong>carbon</strong> and <strong>the</strong>presence <strong>of</strong> polymer..............................................................................................248xxi


LIST OF TABLESTable 1-1 : Operating conditions <strong>of</strong> CO 2 separation from natural gas and syn<strong>the</strong>sis gas.........5Table 2-1 : Kinetic diameter, molecular weight and Knudsen selectivities <strong>of</strong> gas molecules <strong>of</strong>interest. ...................................................................................................................16Table 2-2 : H 2 /N 2 permselectivity <strong>for</strong> NPC <strong>membranes</strong> made from PFA and polyimide. TheNPC <strong>membranes</strong> made from PFA have a <strong>high</strong>er selectivity but lower permeancethan those made from polyimide.............................................................................23Table 2-3 : Permeance and permselectivity <strong>of</strong> NPC <strong>membranes</strong> made from PFA at differentpyrolysis <strong>temperature</strong>s [75]. Increasing <strong>the</strong> <strong>temperature</strong> <strong>of</strong> pyrolysis has asignificant effect upon increasing <strong>the</strong> permeance. However, <strong>the</strong> permselectivitygoes through a maximum at 450 °C. ......................................................................28Table 2-4 : PALS data <strong>for</strong> NPC <strong>membranes</strong> manufactured from <strong>the</strong> pyrolysis <strong>of</strong> <strong>the</strong> polyimideP84 and Matrimid [96]. The lifetime <strong>of</strong> ortho-positronium annihilating in <strong>the</strong>pores (τ 2 ) indicates <strong>the</strong> size <strong>of</strong> <strong>the</strong> pores. The number or intensity <strong>of</strong> orthopositroniumannihilating in <strong>the</strong> pores (I 2 ) indicates <strong>the</strong> number <strong>of</strong> pores or <strong>the</strong>porosity...................................................................................................................32Table 2-5 : WAXD data <strong>for</strong> NPC <strong>membranes</strong> manufactured from <strong>the</strong> pyrolysis <strong>of</strong> <strong>the</strong>polyimide Matrimid and P84 [17, 96, 97] along with powder x-ray diffraction data<strong>for</strong> bulk <strong>nanoporous</strong> <strong>carbon</strong> manufactured from PFA [71]. ..................................33Table 2-6 : Permeance results <strong>of</strong> various gas molecules <strong>for</strong> <strong>membranes</strong> made from PFA [19].The increase in permeance with decreasing molecular size is a clear indicator <strong>of</strong> amolecular sieving mechanism.................................................................................36Table 2-7 : Activation energy <strong>of</strong> permeance <strong>of</strong> various gas molecules <strong>for</strong> NPC <strong>membranes</strong>made from polyamic acid. The activation energy <strong>for</strong> CO 2 is 0 indicating that <strong>the</strong>CO 2 permeance will remain constant with increasing <strong>temperature</strong>.......................43Table 2-8 : Activation energies <strong>of</strong> diffusion and heats <strong>of</strong> adsorption <strong>of</strong> various gas molecules<strong>for</strong> NPC <strong>membranes</strong> made from PFA and Cellulose..............................................44Table 2-9 : Activation energy <strong>of</strong> permeance <strong>of</strong> various gas molecules <strong>for</strong> NPC <strong>membranes</strong>made from PFA and Cellulose obtained by <strong>the</strong> addition <strong>of</strong> <strong>the</strong> diffusion andactivation energies presented in Table 2-8. Interestingly, <strong>for</strong> CO 2 <strong>the</strong> activationenergy <strong>of</strong> permeance is negative beca<strong>use</strong> adsorption is such a large contributor to<strong>the</strong> permeance.........................................................................................................45Table 3-1 : Materials <strong>use</strong>d to manufacture supported NPC <strong>membranes</strong>. ................................57Table 3-2 : Extra materials <strong>use</strong>d to manufacture modified supported NPC <strong>membranes</strong>.........58Table 3-3 : Materials <strong>use</strong>d to manufacture unsupported NPC <strong>membranes</strong>. ............................58xxii


Table 3-4 : The composition <strong>of</strong> <strong>the</strong> two solutions prepared <strong>for</strong> <strong>the</strong> manufacture <strong>of</strong> <strong>the</strong> silicananoparticles <strong>use</strong>d <strong>for</strong> filling <strong>the</strong> porous support prior to coating with PFA. .......65Table 3-5 : Mixed gas testing conditions..................................................................................75Table 3-6 : The concentration <strong>of</strong> condensable impurities in <strong>the</strong> feed gas based on <strong>the</strong>operating <strong>temperature</strong> <strong>of</strong> <strong>the</strong> saturating vessel containing <strong>the</strong> liquid impurity. ....78Table 3-7 : Details <strong>of</strong> <strong>the</strong> adsorbents, <strong>temperature</strong> and pressures <strong>use</strong>d <strong>for</strong> obtaining simulatedadsorption iso<strong>the</strong>rms <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> using GCMC simulation..................82Table 3-8 : Details <strong>of</strong> <strong>the</strong> adsorbents, <strong>temperature</strong>s and pressures <strong>use</strong>d <strong>for</strong> obtainingexperimental adsorption iso<strong>the</strong>rms <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong>....................................85Table 3-9 : Equipment mass and densities <strong>use</strong>d to calculate buoyed volume. The equipmentmasses and densities were provided by <strong>the</strong> supplier (VTI). The density <strong>of</strong><strong>nanoporous</strong> <strong>carbon</strong> was assumed to be <strong>the</strong> standard value <strong>of</strong> 1.6 g/cm 3 [79].......87Table 4-1 : Volumes <strong>of</strong> polymer solution (40 wt% PFA in acetone) added to each coat viaspin-coating at 1000 rpm and <strong>the</strong> resulting <strong>carbon</strong> mass added after pyrolysis....89Table 4-2 : Manufacture results <strong>of</strong> <strong>the</strong> <strong>membranes</strong> made from <strong>the</strong> spin-coating <strong>of</strong> PFA ontostainless steel supports and <strong>the</strong>n pyrolysised between 400 and 600°C..................90Table 4-3 : Basic Statistics <strong>of</strong> <strong>the</strong> pore size distribution <strong>of</strong> NPC produced at 600°C obtainedfrom <strong>the</strong> HRTEM picture using <strong>the</strong> image processing program Image J...............93Table 4-4 : PALS data <strong>for</strong> NPC <strong>membranes</strong> made from Matrimid and P84 [96] and PFA.Values <strong>of</strong> τ 2 are similar, which indicates a comparable pore size range...............96Table 4-5 : WAXD data <strong>for</strong> NPC <strong>membranes</strong> made from Matrimid and P84 [17, 96, 97] andPFA. The values <strong>of</strong> <strong>the</strong> d-spacing indicate that pore size decreases with increasing<strong>temperature</strong>.............................................................................................................96Table 4-6 : Pure gas per<strong>for</strong>mance results <strong>for</strong> <strong>the</strong> final coats <strong>of</strong> <strong>membranes</strong> made between 400and 600°C...............................................................................................................97Table 4-7 : Pure gas per<strong>for</strong>mance results <strong>for</strong> each coat <strong>for</strong> <strong>membranes</strong> at 400 and 550°C..100Table 4-8 : Pure gas vs. mixed gas results <strong>for</strong> <strong>membranes</strong> manufactured at 400 and 550°C.An increased CO 2 /CH 4 selectivity is seen with <strong>the</strong> mixed gas testing. .................103Table 4-9 : Activation Energy (E p ) <strong>of</strong> CO 2 and CH 4 Permeance (kJ/mol) <strong>for</strong> NPC <strong>membranes</strong>made at 400 and 550°C. Activation energies were calculated from <strong>the</strong> permeancedata over <strong>the</strong> range <strong>of</strong> operating <strong>temperature</strong>s using Equation 2-38. Data fromo<strong>the</strong>r authors is included <strong>for</strong> comparison.............................................................106Table 4-10 : Effect <strong>of</strong> varying reaction <strong>temperature</strong> and time on <strong>the</strong> size <strong>of</strong> <strong>the</strong> produced silicananoparticles. The fourth sample with a reaction <strong>temperature</strong> <strong>of</strong> 50°C and time <strong>of</strong>45 mins gave monodisperse silica nanoparticles with an average size less than 300nm. ........................................................................................................................112xxiii


Table 4-11 : Effect <strong>of</strong> <strong>the</strong> addition <strong>of</strong> silica nanoparticles on <strong>the</strong> permeance and selectivity <strong>of</strong><strong>the</strong> porous stainless steel support. The error in <strong>the</strong> permeance is ± 1.38% and <strong>the</strong>error in <strong>the</strong> permselectivity is 1.95%....................................................................113Table 4-12 : Addition <strong>of</strong> <strong>carbon</strong> to supports modified with silica nanoparticles. The error in<strong>the</strong> permeance is ± 1.38% and <strong>the</strong> error in <strong>the</strong> permselectivity is 1.95%. ...........115Table 5-1 : Values <strong>of</strong> <strong>the</strong> Langmuir constants (b and V m ) <strong>for</strong> <strong>the</strong> simulated adsorptioniso<strong>the</strong>rms <strong>of</strong> pure CO 2 , CH 4 and N 2 on NPC with increasing <strong>temperature</strong>s. .......119Table 5-2 : Simulated values <strong>of</strong> <strong>the</strong> heat <strong>of</strong> adsorption (∆H ads ) <strong>of</strong> pure gases on <strong>nanoporous</strong><strong>carbon</strong> compared with experimental values reported in <strong>the</strong> literature. ...............120Table 5-3 : Experimental values <strong>of</strong> <strong>the</strong> Langmuir constants (b and V m ) <strong>for</strong> <strong>the</strong> adsorptioniso<strong>the</strong>rms <strong>of</strong> pure CO 2 on <strong>nanoporous</strong> <strong>carbon</strong>.....................................................125Table 5-4 : Buoyed volume <strong>for</strong> CO 2 adsorption onto NPC at 35°C and 200°C calculated from<strong>the</strong> methods <strong>of</strong> (1) helium adsorption and (2) equipment masses. .......................127Table 5-5 : Experimental values <strong>of</strong> <strong>the</strong> Langmuir constants (b and V m ) <strong>for</strong> <strong>the</strong> adsorptioniso<strong>the</strong>rms <strong>of</strong> pure N 2 on <strong>nanoporous</strong> <strong>carbon</strong>........................................................128Table 5-6 : Values <strong>of</strong> <strong>the</strong> experimental and simulated heats <strong>of</strong> adsorption <strong>of</strong> pure gases on<strong>nanoporous</strong> <strong>carbon</strong> compared with values reported in <strong>the</strong> literature..................131Table 5-7 : A comparison <strong>of</strong> <strong>the</strong> experimental CO 2 heat <strong>of</strong> adsorption (∆H ads ) with <strong>the</strong>activation energy <strong>of</strong> permeance (E p ) <strong>for</strong> a NPC membrane manufactured at 550°C(refer Section 4.2.5) to calculate <strong>the</strong> activation energy <strong>of</strong> diffusion (E d ) viaEquation 2-38. ......................................................................................................131Table 6-1 : Normalised stabilised reduction in CO 2 permeance with impurity presence. .....143Table 6-2 : Normalised stabilised reduction in CO 2 /CH 4 selectivity with impurity presence...............................................................................................................................143Table 6-3 : Normalised stabilised reduction in CO 2 permeance as a function <strong>of</strong> impurity....149Table 6-4 : Normalised stabilised reduction in CO 2 /N 2 selectivity as a function <strong>of</strong> impurity...............................................................................................................................149Table C-1 : Error Analysis Relationships [175] ....................................................................231Table C-2 : Error analysis <strong>of</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> pure gas CO 2 CH4 permselectivity fromduplicate measurements <strong>of</strong> Membrane NPC400C taken 6 days apart. ................232Table C-3 : Error analysis <strong>of</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> pure gas CO 2 permeance <strong>for</strong> MembraneNPC550B. The error analysis is independent <strong>of</strong> <strong>the</strong> absolute values <strong>of</strong> <strong>the</strong>variables as <strong>the</strong> error is a percentage <strong>of</strong> <strong>the</strong> absolute value with <strong>the</strong> exception <strong>of</strong><strong>the</strong> values <strong>of</strong> <strong>the</strong> membrane area. As <strong>the</strong> membrane area reading is <strong>the</strong> same <strong>for</strong><strong>the</strong> CH 4 permeance, <strong>the</strong> error analysis <strong>for</strong> <strong>the</strong> CH 4 permeance gives <strong>the</strong> sameresult.....................................................................................................................233xxiv


Table C-4 : Summary <strong>of</strong> calculated pure gas permeances and permselectivities <strong>of</strong> MembraneNPC550B based upon <strong>the</strong> steady state gradients from <strong>the</strong> permeate pressure risecurves. The system constants were permeate volume (V p ) = 525 cm 3 , <strong>temperature</strong>(T) = 306 K and membrane area = 9.62 cm 2 . ......................................................235Table C-5 : Error analysis <strong>of</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> mixed gas CO 2 /CH 4 permselectivity fromduplicate measurements <strong>of</strong> Membrane NPC550B................................................236Table C-6 : Error analysis <strong>of</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> CO 2 permeance <strong>for</strong> Membrane NPC550B.The error analysis is independent <strong>of</strong> <strong>the</strong> absolute values <strong>of</strong> <strong>the</strong> variables as <strong>the</strong>error is a percentage <strong>of</strong> <strong>the</strong> absolute value with <strong>the</strong> exception <strong>of</strong> <strong>the</strong> values <strong>of</strong> <strong>the</strong>permeate flow and membrane area. As <strong>the</strong> permeate flow and membrane areareadings are <strong>the</strong> same <strong>for</strong> <strong>the</strong> CH 4 permeance, <strong>the</strong> error analysis <strong>for</strong> <strong>the</strong> CH 4permeance gives <strong>the</strong> same result. .........................................................................237Table C-7 : Sample <strong>of</strong> <strong>the</strong> calculated mixed gas per<strong>for</strong>mance <strong>of</strong> Membrane NPC550B. Thesystem constants were <strong>temperature</strong> = 35 °, feed gas composition = 10:90(CO 2 :CH 4 vol%), feed pressure (p f ) = 1200 kPag, retentate pressure = 1150 kPag,retentate flow = 100 ml/min, sweep gas flow = 20 ml/min, permeate pressure = 0kPag and membrane area (A) = 9.62 cm 2 . ...........................................................238Table C-8 : Error analysis <strong>of</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> adsorbed volume <strong>of</strong> CO 2 onto <strong>nanoporous</strong><strong>carbon</strong> 303 Torr (40 kPa) at 35°C. ......................................................................240Table C-9 : Sample <strong>of</strong> <strong>the</strong> calculated results <strong>for</strong> <strong>the</strong> adsorption <strong>of</strong> CO 2 on <strong>nanoporous</strong> <strong>carbon</strong>at 35°C..................................................................................................................241Table C-10 : Data produced <strong>for</strong> <strong>the</strong> 300 K iso<strong>the</strong>rm from <strong>the</strong> molecular simulation <strong>of</strong> <strong>the</strong>adsorption <strong>of</strong> CO 2 :CH 4 (10:90 vol%) on <strong>nanoporous</strong> <strong>carbon</strong> represented by C 168Schwartize.............................................................................................................244Table D-1 : Per<strong>for</strong>mance results <strong>of</strong> raw Matrimid and partially pyrolysised NPC <strong>membranes</strong>made from Matrimid at 420 and 430°C. The results suggest significant cracking in<strong>the</strong> partially pyrolysised NPC <strong>membranes</strong>. ..........................................................249Table D-2 : Per<strong>for</strong>mance results <strong>of</strong> raw Matrimid and partially pyrolysised NPC <strong>membranes</strong>made from Kapton at 420°C. The results suggest some improvement inper<strong>for</strong>mance with partial pyrolysis.......................................................................250xxv


NOMENCLATUREA Membrane Surface Area (m 2 )bCDdLangmuir ConstantConcentrationDiffusion Coefficient (m/s)D Spacing (nm)d m Diameter <strong>of</strong> <strong>the</strong> Membrane (m) in Equation 2-23d p Diameter <strong>of</strong> <strong>the</strong> Particle (m) in Equation 2-17dEEEf∆HJDimensionality <strong>of</strong> <strong>the</strong> System in MD SimulationActivation Energy (kJ/mol)Energy <strong>of</strong> <strong>the</strong> System in GCMC SimulationError in Variable A in Appendix CFugacity (Pa)Heat <strong>of</strong> Adsorption (kJ/mol)Flux (sm 3 /m 2 .s)k BBoltzmann ConstantMWNMolecular WeightNumber <strong>of</strong> Molecules in GCMC and MD Simulationn Integer in Equation 2-28PPermeability (sm 3 .m/m 2 .s.Pa) or (mol/m 2 .s.Pa)P’ Permeance (mol/m.s.Pa)pPressure (Pa)xxvi


pp∆pPartial Pressure (Pa)Pressure Drop (Pa)p 0 Driving-Force Pressure (Pa) in Equation 2-27QRVolumetric flow rate (m 3 /s)Ideal Gas Constant = 8.314 (J/mol.K)r Separation Distance in Equation 2-32SStTuVSeparation FactorSolubility Coefficient (mol/m 3 .Pa)Time (s)Temperature (K)Interaction PotentialVelocity (m/s)V Volume (m 3 )V m Molar Volume at Standard Temp and Pressure (mol/m 3 ) in Equation 2-6V m Langmuir Constant in Section 2.2.2.3x∆xyConcentration <strong>of</strong> a Component on <strong>the</strong> Feed Side (mol fraction)Membrane Thickness (m)Concentration <strong>of</strong> a Component on <strong>the</strong> Permeate Side (mol fraction)Non-English Charactersα A/BSelectivityεδWell DepthEffective Membrane Thickness (m)xxvii


λ Specific Radiation Wavelength (nm) in Equation 2-28λ Coefficient <strong>of</strong> Laminar or Turbulent Flow in Equation 2-23λ De Broglie Wavelength in Equation 2-31ρ Density (kg/m 3 )π Ma<strong>the</strong>matical Constant (~3.14159)φFugacity Coefficientθ Angle <strong>of</strong> Diffraction (°C) in Equation 2-28θ Time Lag (s) in Equation 2-11ФστPartition Function in GCMC SimulationSite Collision DiameterTime Axis Intercept <strong>of</strong> <strong>the</strong> Integral Molar Flux (s)µ Chemical Potential in GCMC SimulationSubscriptsAadsBComponent AAdsorptionComponent BC Convective Flow in Equations 2-25fIMPKMpSFeed SideImpurityKnudsenMolecular SievingPermeate SideSurface Diffusionxxviii


SDSSSolution DiffusionSteady State0 Initial or at Zero Concentrationxxix


xxx


1 INTRODUCTION1.1 Global WarmingThe energy demands <strong>of</strong> our modern society are primarily met by burning fossil fuels, whichproduce large quantities <strong>of</strong> <strong>the</strong> greenho<strong>use</strong> gas <strong>carbon</strong> dioxide (CO 2 ). CO 2 is termed agreenho<strong>use</strong> gas beca<strong>use</strong> it absorbs infrared radiation and <strong>the</strong>reby reduces <strong>the</strong> energy flow out<strong>of</strong> <strong>the</strong> atmosphere, which in turn leads to an increase in <strong>the</strong> earth’s <strong>temperature</strong>s. Thisphenomenon is known as “Global Warming”.Definitive changes in <strong>the</strong> Earth’s global climate, such as increases in <strong>the</strong> average air andocean <strong>temperature</strong>s, increased melting <strong>of</strong> snow and ice leading to rising sea levels, havealready been observed. Additionally, <strong>the</strong> global emissions <strong>of</strong> greenho<strong>use</strong> gases, in particularCO 2 , are projected to grow over <strong>the</strong> coming decades leading to even greater changes in <strong>the</strong>Earth’s climate [1].Locally, Sou<strong>the</strong>rn Australia is experiencing <strong>the</strong> worst drought on record [2]. Whilst <strong>the</strong>re havebeen o<strong>the</strong>r impacts <strong>of</strong> climate change such as coral bleaching <strong>of</strong> <strong>the</strong> Great Barrier Reef [3], itis <strong>the</strong> relentlessness, severity and ultimately personal impact <strong>of</strong> <strong>the</strong> drought which has led tosome shift in <strong>the</strong> attitude towards climate change <strong>of</strong> <strong>the</strong> Australian People.In <strong>the</strong> 2007 federal election, <strong>the</strong> Australian Labor Party led by Kevin Rudd was elected <strong>for</strong> <strong>the</strong>first time in eleven years. The more progressive environmental policies, including signing <strong>the</strong>Kyoto Protocol, are said to have been a contributor to Rudd’s landslide win [4].However, <strong>the</strong> cost impact <strong>of</strong> reducing Australia’s greenho<strong>use</strong> gas emissions is a major barrier<strong>for</strong> fur<strong>the</strong>r significant support from <strong>the</strong> Australian people. The distress ca<strong>use</strong>d by recentincreases in petrol prices due to <strong>the</strong> increasing oil prices unrelated to climate change are anindication <strong>of</strong> attitudes toward paying more <strong>for</strong> energy. This is particularly true <strong>for</strong> <strong>the</strong>proportion <strong>of</strong> <strong>the</strong> population living away from <strong>the</strong> major centres where energy pricessignificantly influence <strong>the</strong> transportation costs <strong>of</strong> essential goods.Australia has also enjoyed inexpensive electricity due to plentiful black and brown coalreserves. In fact, <strong>the</strong> presence <strong>of</strong> cheap energy and o<strong>the</strong>r substantial mineral reserves are <strong>the</strong>biggest contributors to Australia’s economy. This was <strong>the</strong> principal argument <strong>for</strong> <strong>the</strong> lack <strong>of</strong>action on climate change by <strong>the</strong> previous Australian government.1


1.2 CO 2 Capture and SequestrationOne promising process <strong>for</strong> significantly reducing greenho<strong>use</strong> gas emissions in <strong>the</strong> shorterterm with potential <strong>for</strong> lower cost is CO 2 capture and sequestration. Whilst <strong>the</strong> energy orelectricity is still produced from fossil fuels, <strong>the</strong> CO 2 produced as part <strong>of</strong> <strong>the</strong> process iscaptured and stored underground preventing release into <strong>the</strong> atmosphere and fur<strong>the</strong>rcontribution to global warming.The Australian Cooperative Research Centre <strong>for</strong> Greenho<strong>use</strong> Gas Technologies (CO2CRC)was established in 2003 and is a collaboration <strong>of</strong> research participants, government andindustry partners working towards commercialisation <strong>of</strong> CO 2 capture and sequestration inAustralia.The CO 2 capture and sequestration process as shown in Figure 1-1 involves (i) capturing CO 2from <strong>the</strong> source, (ii) pressurising CO 2 so that it <strong>for</strong>ms a dense phase, (iii) transportation <strong>of</strong> <strong>the</strong>CO 2 to a storage site (via pipeline) and finally (iv) storing <strong>the</strong> CO 2 underground in an existinggeological <strong>for</strong>mation such as an oil and gas or saline reservoir.Figure 1-1 : CO 2 capture and sequestration process. Diagram provided by <strong>the</strong>Cooperative Research Centre <strong>for</strong> Greenho<strong>use</strong> Gas Technologies (CO2CRC) [5].Looking at CO 2 capture in more detail, <strong>the</strong>re are three major points within stationary energyproduction where CO 2 is produced and <strong>the</strong>n emitted to <strong>the</strong> atmosphere. The first is in <strong>the</strong>production <strong>of</strong> natural gas (primarily methane, CH 4 ) from an underground reservoir. In thiscase, <strong>the</strong> CO 2 coexists with methane in <strong>the</strong> reservoir. The CO 2 must be removed from <strong>the</strong>natural gas be<strong>for</strong>e it can be <strong>use</strong>d in a commercial or domestic application. In Australia, this is2


usually done via an onshore gas processing plant. The second emission source comes from <strong>the</strong>production <strong>of</strong> syn<strong>the</strong>sis gas (hydrogen, H 2 ) using fossil fuels such as natural gas, oil and coal.CO 2 capture from this emission source is <strong>of</strong>ten referred to as pre-combustion capture. Thefinal emission source is <strong>the</strong> flue gas from electricity power stations that are from directcombustion <strong>of</strong> fossil fuels, such as natural gas, oil or coal. CO 2 capture from this emissionsource is <strong>of</strong>ten referred to as post-combustion capture.The present cost <strong>of</strong> CO 2 sequestration is estimated by <strong>the</strong> International Energy Agency (IEA)to be between $40 and $60 US per tonne <strong>of</strong> CO 2 emissions avoided [6, 7] <strong>of</strong> which capturecosts contribute approximately 90 %. There<strong>for</strong>e, significant improvements to <strong>the</strong> capturetechnology remain a focus <strong>for</strong> major reduction in <strong>the</strong> cost <strong>of</strong> CO 2 sequestration process.1.3 Capture TechnologiesTechnologies that exist to separate CO 2 from o<strong>the</strong>r gases include gas absorption, gasadsorption, cryogenics, gas hydrates and gas separation <strong>membranes</strong>. The focus <strong>of</strong> currentresearch is now on improving <strong>the</strong>se technologies with new materials to create more efficientand hence less expensive processes.Nanoporous <strong>carbon</strong> is one such new material that has attracted a lot <strong>of</strong> interest in recent years.Made from <strong>the</strong> pyrolysis (heating) <strong>of</strong> a polymer at <strong>high</strong> <strong>temperature</strong>s, <strong>nanoporous</strong> <strong>carbon</strong> is astructure related to <strong>the</strong> fullerene group <strong>of</strong> <strong>carbon</strong> structures (<strong>the</strong> Buckyball, nanotubes etc.),which contains nanosized pores.Initial testing <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> (NPC) <strong>membranes</strong> revealed superior per<strong>for</strong>mance over<strong>the</strong> traditional polymeric <strong>membranes</strong>. High per<strong>for</strong>mance NPC <strong>membranes</strong> have beenmanufactured from <strong>the</strong> pyrolysis <strong>of</strong> cellulose [8-13], polyvinylindene chloride (PVDC) [14],phenolic resin [15], polyimide [16, 17], polyamic acid [18] and polyfurfuryl alcohol (PFA)[19]. The CO 2 /CH 4 separation factors <strong>of</strong> <strong>the</strong>se NPC <strong>membranes</strong> range from 10 to as <strong>high</strong> as200 whilst retaining a good permeance or flow through <strong>the</strong> membrane.The <strong>high</strong> values <strong>of</strong> per<strong>for</strong>mance published sparked a flurry <strong>of</strong> research work on NPC<strong>membranes</strong>. The majority <strong>of</strong> <strong>the</strong> research work was foc<strong>use</strong>d on <strong>the</strong> transport mechanism [8,15, 20-26] and fur<strong>the</strong>r development <strong>of</strong> <strong>high</strong> per<strong>for</strong>ming NPC membrane [27, 28]. While initialstudies utilised an NPC layer supported on a porous support, thinner unsupported NPC<strong>membranes</strong> were also produced from thin polymer sheets to increase <strong>the</strong> flow. However,difficulties with brittleness and cracking made working with such NPC <strong>membranes</strong> verydifficult [29]. In fact, cracking still remains an impediment to <strong>the</strong> fur<strong>the</strong>r development <strong>of</strong>supported and unsupported <strong>membranes</strong> alike [30].3


Following on from this, <strong>the</strong>re have been several studies related to <strong>the</strong> effect <strong>of</strong> oxygen, waterand condensable impurities (such as heavier hydro<strong>carbon</strong>s) on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC<strong>membranes</strong>. Oxygen has been shown to react with <strong>the</strong> <strong>carbon</strong> creating hydrophilic <strong>carbon</strong>yl(C=O) functional groups which attract water from <strong>the</strong> atmosphere resulting in a severereduction in membrane per<strong>for</strong>mance [31-33]. The presence <strong>of</strong> condensable hydro<strong>carbon</strong>s wasalso shown to result in a severe reduction in membrane per<strong>for</strong>mance [34, 35].The objective <strong>of</strong> this research is to follow-on from <strong>the</strong> a<strong>for</strong>ementioned ground breaking workand specifically look at <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> in <strong>the</strong> application <strong>of</strong> CO 2capture. The two emission sources examined <strong>for</strong> <strong>the</strong> application <strong>of</strong> NPC <strong>membranes</strong> are (1)separation <strong>of</strong> CO 2 from natural gas (primarily methane CH 4 ) at <strong>high</strong> <strong>temperature</strong> and in <strong>the</strong>presence <strong>of</strong> <strong>the</strong> impurities – water, hexane, toluene and hydrogen sulphide (H 2 S) and (2) <strong>the</strong>separation <strong>of</strong> CO 2 from air-blown syn<strong>the</strong>sis gas production (nitrogen, N 2 after initialhydrogen, H 2 separation) <strong>for</strong> electricity or hydrogen production at <strong>high</strong> <strong>temperature</strong> and in <strong>the</strong>presence <strong>of</strong> <strong>the</strong> impurities – water, H 2 S and <strong>carbon</strong> monoxide (CO).CO 2 capture from power station flue gases using NPC <strong>membranes</strong> is not considered as part <strong>of</strong>this research work given <strong>the</strong> <strong>high</strong> concentrations <strong>of</strong> both oxygen and water.Details <strong>of</strong> <strong>the</strong> operating conditions <strong>of</strong> natural gas and syn<strong>the</strong>sis gas separation <strong>use</strong>d toevaluate <strong>the</strong> suitability <strong>of</strong> NPC <strong>membranes</strong> <strong>for</strong> <strong>the</strong>se two applications are given in Figure 1-2and Table 1-1. The data was provided by <strong>the</strong> Cooperative Research Centre <strong>for</strong> Greenho<strong>use</strong>Gas Technologies (CO2CRC) [36].(a)(b)CO 2 /CH 4Feed> 2000 kPag> 100 °CH 2 0, H 2 SFocus <strong>of</strong> thisresearch workCooler, Knockout Potand FilterCH 4CO 2H 2 /CO 2 /N 2Feed> 2000 kPag250 °CH 2 0, CO, COSCooler, KnockoutPot and FilterFocus <strong>of</strong> thisresearch workN 2N 2 , CO 2 CO 2H 2 , H 2 OFigure 1-2 : Separation <strong>of</strong> CO 2 from (a) natural gas and (b) air-blown syn<strong>the</strong>sis gas4


Table 1-1 : Operating conditions <strong>of</strong> CO 2 separation from natural gas and syn<strong>the</strong>sis gasNatural GasSyn<strong>the</strong>sis GasPressure kPag >2000 2800 – 5700Temperature °C >100 >250Composition Mol %H 2 0 55 – 56CO 0 2.5 – 2.8CO 2 10 37 – 40N 2 0 0.68 – 3.10O 2 0 0CH 4 90 0.02 – 0H 2 S + COS ppm levels 0.18 – 0.22 (no COS)H 2 O Saturated 0.19 – 0.31Hexane Saturated 0Toluene Saturated 0The structure <strong>of</strong> this <strong>the</strong>sis presents <strong>the</strong> findings <strong>of</strong> <strong>the</strong> research work beginning with <strong>the</strong>manufacture and <strong>high</strong> <strong>temperature</strong> operation <strong>of</strong> NPC <strong>membranes</strong> (Chapter 4). Themechanisms behind <strong>the</strong> transport <strong>of</strong> gases through NPC <strong>membranes</strong> at elevated <strong>temperature</strong>sare <strong>the</strong>n fur<strong>the</strong>r studied using <strong>the</strong> results <strong>of</strong> gas adsorption experiments and molecularmodelling (Chapter 5). Finally, <strong>the</strong> impact <strong>of</strong> condensable and non-condensable impuritiesrelevant to <strong>the</strong> specific application <strong>of</strong> CO 2 capture from <strong>the</strong> two emission sources beingreviewed is assessed (Chapter 6).For fur<strong>the</strong>r details on global warming, <strong>the</strong> CO 2 sequestration process and capture technologiesplease refer to Appendix A.5


2 LITERATURE REVIEW2.1 IntroductionThe review <strong>of</strong> <strong>the</strong> published knowledge related to this research work begins with a generaldiscussion on gas separation <strong>membranes</strong>, which includes typical materials <strong>of</strong> manufacture,per<strong>for</strong>mance measurement and gas transport mechanisms. Following on from this, <strong>the</strong>manufacturing methods, per<strong>for</strong>mance and gas transport mechanisms relevant to <strong>nanoporous</strong><strong>carbon</strong> (NPC) <strong>membranes</strong> is discussed.Relating to <strong>the</strong> specific application <strong>of</strong> this research work, <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC<strong>membranes</strong> at <strong>high</strong> <strong>temperature</strong>s and in <strong>the</strong> presence <strong>of</strong> particular impurities is <strong>the</strong>n presented.2.2 Gas Separation MembranesA membrane separates one component from ano<strong>the</strong>r on <strong>the</strong> basis <strong>of</strong> size, shape or chemicalaffinity. A feed stream <strong>of</strong> mixed components enters a membrane unit where it is separatedinto a retentate and permeate stream as depicted in Figure 2-1 below. The retentate stream istypically <strong>the</strong> purified product stream and <strong>the</strong> permeate stream contains <strong>the</strong> waste component.FeedRetentateMembranePermeateFigure 2-1 : Membrane separation unit2.2.1 Per<strong>for</strong>mance Indicators2.2.1.1 Permeance and SelectivityThe per<strong>for</strong>mance <strong>of</strong> <strong>membranes</strong> is generally described by <strong>the</strong> flux (J) or permeability (P),which is equivalent to <strong>the</strong> throughput <strong>of</strong> <strong>the</strong> membrane, and <strong>the</strong> selectivity, which isequivalent to <strong>the</strong> separation efficiency <strong>of</strong> <strong>the</strong> membrane.Membrane flux is a measure <strong>of</strong> <strong>the</strong> flow through <strong>the</strong> membrane per unit area (mol/(m 2 /s)). Asper Fick’s first law, <strong>for</strong> a component to flow through <strong>the</strong> membrane a driving <strong>for</strong>ce is7


equired. Gas flow or flux through a membrane <strong>of</strong> a certain thickness (∆x) is driven by <strong>the</strong>fugacity differential across <strong>the</strong> membrane (∆f) and is described by <strong>the</strong> following equation.JP= ∆fEquation 2-1∆xFor ideal gases, <strong>the</strong> fugacity differential is simply equal to <strong>the</strong> pressure differential (∆p). Fornon-ideal gases, <strong>the</strong> fugacity differential is related to <strong>the</strong> pressure differential via <strong>the</strong> fugacitycoefficient (φ) as shown in Equation 2-2.∆ f = φ − φEquation 2-21p12p2For <strong>the</strong> range <strong>of</strong> pressures and gas components studied in this research works <strong>the</strong> fugacitycoefficients are all greater than 0.95 and usually close to 1.0. There<strong>for</strong>e, it has been assumedthat fugacity can be approximated by pressure. All <strong>of</strong> <strong>the</strong> following equations and resultspresented in this <strong>the</strong>sis will <strong>use</strong> pressure.The permeability (P) is more commonly <strong>use</strong>d to describe <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> a membranethan flux. This is beca<strong>use</strong> <strong>the</strong> permeability <strong>of</strong> a homogenous stable membrane material isconstant regardless <strong>of</strong> <strong>the</strong> pressure differential or membrane thickness and hence it is easier tocompare <strong>membranes</strong> made from different materials. The permeability is usually given in <strong>the</strong>units “Barrer”, which is equivalent to 10 -10 scm 3 .cm/(cm 2 .s.cmHg) or 3.347 x 10 -13mol.m/(m 2 .s.Pa) in S.I. Units.The term permeance (P’) is more commonly <strong>use</strong>d in <strong>the</strong> literature <strong>for</strong> supported <strong>nanoporous</strong><strong>carbon</strong> (NPC) <strong>membranes</strong>. The permeance is equal to <strong>the</strong> permeability multiplied by <strong>the</strong>membrane thickness and is given <strong>the</strong> S.I. units mol/(m 2 .s.Pa). As stated in <strong>the</strong> ASTM standardtest method <strong>for</strong> determining gas permeability characteristics [37], permeability is onlymeaningful <strong>for</strong> homogenous films where <strong>the</strong> membrane thickness can be accuratelydetermined. Supported NPC <strong>membranes</strong> are not homogenous due to <strong>the</strong> supporting porouslayer and as such <strong>the</strong> term permeance is generally <strong>use</strong>d <strong>for</strong> this research work.Selectivity is a measure <strong>of</strong> <strong>the</strong> ability <strong>of</strong> <strong>the</strong> membrane to preferentially permeate onecomponent over <strong>the</strong> o<strong>the</strong>r. An indication <strong>of</strong> selectivity can be obtained from <strong>the</strong> ratio <strong>of</strong> purecomponent permeances <strong>for</strong> a gas pair (A/B) as shown in Equation 2-3. This term is referred toas <strong>the</strong> ideal selectivity or permselectivity (α A/B ).P'AαA / B=Equation 2-3P'B8


For a binary mixture, <strong>the</strong> ideal selectivity can be calculated from <strong>the</strong> component (A/B) molfractions in <strong>the</strong> permeate stream (y) and in <strong>the</strong> feed stream (x) along with <strong>the</strong> feed (p f ) andpermeate (p p ) stream pressures as per Equation 2-4 below.⎡ pp⎤⎢xB− yB⎥y ⎢⎣pAf ⎥α =⎦A / BEquation 2-4yB⎡ pp⎤⎢xA− yA) ⎥⎢⎣pf ⎥⎦Ano<strong>the</strong>r per<strong>for</strong>mance measurement indicator <strong>use</strong>d in mixed gas experiments is <strong>the</strong> separationfactor (S A/B ). The separation factor is simply <strong>the</strong> ratio <strong>of</strong> components (A/B) in <strong>the</strong> permeatestream (y) and feed stream (x) as shown in Equation 2-5. However, <strong>the</strong> value <strong>of</strong> <strong>the</strong> separationfactor is specifically related to <strong>the</strong> feed composition and pressure drop <strong>use</strong>d in <strong>the</strong> experiment.S =yAyBA / BEquation 2-5xAxBThe permeances and selectivities quoted are usually at 35 °C (308 K) unless o<strong>the</strong>rwise stated.The relationship between <strong>the</strong> permeability (<strong>for</strong> <strong>membranes</strong> with a known thickness) andselectivity, is <strong>of</strong>ten described by Robeson’s <strong>the</strong>ory [38] where <strong>the</strong> selectivity <strong>of</strong> <strong>the</strong> membranedecreases with increasing permeability.2.2.1.2 Diffusivity and Heat <strong>of</strong> AdsorptionO<strong>the</strong>r indicators <strong>of</strong> membrane per<strong>for</strong>mance <strong>of</strong>ten <strong>use</strong>d are <strong>the</strong> diffusivity and heat <strong>of</strong>adsorption. These indicators are <strong>use</strong>d to fur<strong>the</strong>r understand <strong>the</strong> transport mechanisms <strong>of</strong> gasesthrough <strong>membranes</strong> as discussed later in Section 2.2.3.The diffusion coefficient or diffusivity (D) indicates <strong>the</strong> speed <strong>of</strong> <strong>the</strong> gas molecules travellingthrough <strong>the</strong> membrane.The heat <strong>of</strong> adsorption (∆H ads ) is <strong>the</strong> energy released when a gas molecule adsorbed onto <strong>the</strong>surface <strong>of</strong> <strong>the</strong> membrane and thus indicates <strong>the</strong> affinity <strong>of</strong> a particular gas to <strong>the</strong> <strong>membranes</strong>urface.9


2.2.2 Per<strong>for</strong>mance Measurement2.2.2.1 Permeance and SelectivityThere are several techniques that can be <strong>use</strong>d to measure permeance and selectivity. Thesimplest method is to subject <strong>the</strong> membrane to a pressure differential, measure <strong>the</strong> flowthrough <strong>the</strong> membrane and calculate <strong>the</strong> permeance (or permeability) from Equation 2-1. Thismethod works well <strong>for</strong> <strong>membranes</strong> that have a <strong>high</strong> permeance and as such <strong>the</strong> flow caneasily be measured using conventional flow metering techniques such as a rotameter or abubble flow meter. The equipment <strong>use</strong>d <strong>for</strong> this method is referred to as a constant pressurevariable volume (CPVV) apparatus and is shown in Figure 2-2 (a).(a)Flow Meter(b)P fP pP fP pCylinder <strong>of</strong>Known VolumeMembraneMembraneGas CylinderGas CylinderFigure 2-2 : Equipment <strong>use</strong>d <strong>for</strong> permeance or permeability measurement; (a) constantpressure variable volume apparatus (CPVV) <strong>for</strong> <strong>high</strong>er flow <strong>membranes</strong> and (b)constant volume variable pressure (CVVP) <strong>for</strong> lower flow <strong>membranes</strong>However, most <strong>membranes</strong> prepared on a laboratory scale when subjected to a reasonablepressure difference have a flow that is too low to be measured conventionally. In this case, itis more accurate to calculate <strong>the</strong> flow and <strong>the</strong>re<strong>for</strong>e permeance by measuring <strong>the</strong> rise inpressure on <strong>the</strong> permeate side <strong>of</strong> <strong>the</strong> membrane within a known volume. This equipment <strong>use</strong>d<strong>for</strong> this method is referred to as a constant volume, variable pressure (CVVP) apparatus and isshown in Figure 2-2 (b).Feed gas is introduced on top <strong>of</strong> <strong>the</strong> membrane at pressure (p f ). The rise in pressure on <strong>the</strong>permeate (p p ) side <strong>of</strong> <strong>the</strong> membrane within <strong>the</strong> known volume (V p ) is recorded against timeand <strong>use</strong>d to calculate permeance using <strong>the</strong> following equation [39].10


ppVpP ' ∆∆x= P =∆xVm∆t(p − p ) ARTEquation 2-6fpPlotting ∆p p against time (t) gives a line with a slope equal to <strong>the</strong> permeability (P) or gaspermeance (P’) if <strong>the</strong> membrane thickness (∆x) is removed from <strong>the</strong> equation. The term ARTrefers to <strong>the</strong> membrane surface area (A), <strong>the</strong> universal gas constant (R) and <strong>the</strong> experimental<strong>temperature</strong> (T), whilst <strong>the</strong> term V m refers to <strong>the</strong> molar volume at standard <strong>temperature</strong> andpressure.The two methods mentioned above are generally <strong>use</strong>d <strong>for</strong> measuring <strong>the</strong> permeance <strong>of</strong> a puregas and from which permselectivity can be calculated using Equation 2-3.To measure <strong>the</strong> separation per<strong>for</strong>mance <strong>of</strong> a mixed gas stream, an extended version <strong>of</strong>constant pressure variable volume apparatus is <strong>use</strong>d as shown in Figure 2-3.RetentateFlow MeterP rP fMembranePermeateFlow MeterGasChromatographMixed Gas Cylinder(Known Composition)P pFigure 2-3 : An extension <strong>of</strong> <strong>the</strong> constant pressure variable volume apparatus (CPVV)<strong>use</strong>d <strong>for</strong> permeance and selectivity measurement using a mixed gas feed.In <strong>the</strong> mixed gas stream, <strong>the</strong> permeance <strong>of</strong> a component (A <strong>for</strong> example) is calculated from<strong>the</strong> flux (J) and <strong>the</strong> component pressure drop (∆p A ) by rearranging Equation 2-1 <strong>for</strong>permeance as shown in Equation 2-7.PJA'A=Equation 2-7∆pAThe component flux (J A ) is calculated from <strong>the</strong> permeate molar flow rate (Q p ), permeatecomponent mol fraction (y) and membrane surface area (A) as presented in Equation 2-8.11


QpyAJA= Equation 2-8AThe change in component partial pressure is calculated from <strong>the</strong> feed (x) and permeate (y)component mol fraction along with <strong>the</strong> pressure <strong>of</strong> <strong>the</strong> feed (p f ) and permeate streams (p p ) aspresented in Equation 2-9.∆ p = x p − y pEquation 2-9AAfApSubstituting Equations 2-8 and 2-9 into Equation 2-7 yields Equation 2-10 which gives <strong>the</strong>final equation <strong>for</strong> calculating <strong>the</strong> component permeance using a mixed gas feed.P'AJ QpyAA= =Equation 2-10∆pA(x p − y p )AAfAp2.2.2.2 DiffusivityThe diffusivity <strong>of</strong> pure gases through <strong>membranes</strong> can be calculated using <strong>the</strong> time-lag method[40]. A plot <strong>of</strong> <strong>the</strong> gas flow (Q) through <strong>the</strong> membrane versus time (t) reveals an initialtransient permeation followed by steady state permeation. Extending <strong>the</strong> linear section <strong>of</strong> <strong>the</strong>plot back to <strong>the</strong> intersection <strong>of</strong> <strong>the</strong> x-axis gives <strong>the</strong> value <strong>of</strong> <strong>the</strong> time-lag (θ) as shown inFigure 2-4.Gas Flow (Q)Transient PermeationSteady-State PermeationθTime (t)Figure 2-4 : Calculation <strong>of</strong> <strong>the</strong> diffusion coefficient using <strong>the</strong> time-lag method, once <strong>the</strong>gradient is constant and steady state flow through <strong>the</strong> membrane has been reached, anextrapolation <strong>of</strong> <strong>the</strong> steady state flow line back to <strong>the</strong> x-axis where <strong>the</strong> gas flow is 0reveals <strong>the</strong> value <strong>of</strong> <strong>the</strong> time lag (θ).12


The time lag relates to <strong>the</strong> time it takes <strong>for</strong> <strong>the</strong> first gas molecules to travel through <strong>the</strong>membrane and is thus related to <strong>the</strong> diffusivity. The diffusion coefficient (D) (m/s) can becalculated from <strong>the</strong> time-lag and <strong>the</strong> membrane thickness (∆x) as shown in Equation 2-11.This method produces a value <strong>of</strong> <strong>the</strong> diffusivity that is applicable to ideal gases. CO 2 is a nonidealgas with a diffusivity that is concentration dependent. Wang et al [41, 42] have studied<strong>the</strong> permeation <strong>of</strong> CO 2 in <strong>carbon</strong> molecular sieves and have developed an extended version <strong>of</strong><strong>the</strong> time-lag method that allows <strong>for</strong> <strong>the</strong> concentration dependency.D = ∆xEquation 2-11 Adapted from Reference [40]6θ2.2.2.3 Heat <strong>of</strong> AdsorptionThe heat <strong>of</strong> adsorption (∆H ads ) <strong>of</strong> a gas onto <strong>the</strong> surface <strong>of</strong> a membrane is calculated fromadsorption iso<strong>the</strong>rms at various <strong>temperature</strong>s obtained using a gravimetric micro-balance [12].The adsorption behaviour <strong>of</strong> gases on <strong>nanoporous</strong> <strong>carbon</strong> generally falls outside <strong>the</strong> simplestmodel <strong>for</strong> adsorption, Henry’s Law, especially <strong>for</strong> <strong>the</strong> more adsorptive gases like CO 2 [43].Type 1 adsorption [44], where pores are extremely small, is more common <strong>for</strong> materials like<strong>nanoporous</strong> <strong>carbon</strong> and can be fitted well using <strong>the</strong> Langmuir Equation.The Langmuir model assumes that adsorption energy is constant and independent <strong>of</strong> surfacecoverage and that <strong>the</strong> maximum adsorption (V m ) occurs when <strong>the</strong> surface is covered by amonolayer <strong>of</strong> adsorbate. The amount <strong>of</strong> gas adsorbed (V) at a particular pressure (p) isdescribed by <strong>the</strong> Langmuir Equation as shown in Equation 2-12 [42].VVmbP= Equation 2-121 + bPRearranging <strong>the</strong> Langmuir Equation to a linear <strong>for</strong>m as given in Equation 2-13 below gives<strong>the</strong> value <strong>of</strong> b and V m from a straight line plot <strong>of</strong> p/V versus p.pVp= 1 +Equation 2-13V bmV mThe Langmuir rate constant, b, determined at various <strong>temperature</strong>s can <strong>the</strong>n be <strong>use</strong>d tocalculate <strong>the</strong> heat <strong>of</strong> adsorption (∆H ads ) from <strong>the</strong> Arrhenius-type relationship shown inEquation 2-14 below [43].13


⎛ ∆Hads ⎞b = bo exp ⎜ ⎟Equation 2-14⎝ RT ⎠2.2.2.3.1 Variations on <strong>the</strong> Langmuir ModelThe deviation from ideal Langmuir behaviour <strong>of</strong> <strong>the</strong> adsorption <strong>of</strong> a single component gas at<strong>high</strong> pressures can <strong>of</strong>ten be more accurately modelled using a double-layer Langmuir model[45] as shown in Equation 2-15 below.VVm1b1PVm2b2P= +Equation 2-151 + b P 1+b P12For a mixed gas with components A and B, <strong>the</strong> fractional adsorption (θ) <strong>of</strong> one component(A) onto <strong>the</strong> substrate surface from bulk phase can be described using <strong>the</strong> dual site orextended Langmuir model [46, 47] as described in Equation 2-16.bPθ A AA= 1+ bAPA+ bBPEquation 2-16BHowever, <strong>the</strong> extended Langmuir iso<strong>the</strong>rm is most accurate when <strong>the</strong> bulk concentrations <strong>of</strong><strong>the</strong> two components are similar [46], which is not <strong>the</strong> case <strong>for</strong> this modelling work.Ano<strong>the</strong>r approach is to <strong>use</strong> <strong>the</strong> ideal adsorbed solution <strong>the</strong>ory (IAST), which predicts mixtureadsorption behaviour from <strong>the</strong> shape <strong>of</strong> <strong>the</strong> pure gas adsorption iso<strong>the</strong>rms [48]. Chen andSholl [49] have examined <strong>the</strong> accuracy <strong>of</strong> <strong>the</strong> IAST method <strong>for</strong> <strong>the</strong> adsorption <strong>of</strong> CO 2 /CH 4 on<strong>carbon</strong> nanotubes using transition matrix Monte Carlo simulations. The authors have found<strong>the</strong> IAST method based upon pure gas iso<strong>the</strong>rms fitted with <strong>the</strong> dual-site Langmuir model tobe accurate over a large pressure range.Ritter and Yang [50] have examined <strong>the</strong> accuracy <strong>of</strong> several adsorption models <strong>for</strong> <strong>the</strong>competitive adsorption <strong>of</strong> CO 2 /CH 4 onto activated <strong>carbon</strong> and have found IAST to provide abetter fit to <strong>the</strong> experimental data than <strong>the</strong> extended Langmuir iso<strong>the</strong>rm.14


2.2.3 Materials and Transport MechanismsThe sizes <strong>of</strong> <strong>the</strong> pores within a membrane dictate <strong>the</strong> gas transport mechanism and <strong>the</strong>resulting separation per<strong>for</strong>mance.2.2.3.1 Microporous MembranesMicroporous <strong>membranes</strong> are generally manufactured from materials such as silica ceramicsand have pore sizes > 3 nm (30 Å) [40]. The transport <strong>of</strong> gases in microporous <strong>membranes</strong>occurs via Knudsen diffusion [8, 23] or in <strong>the</strong> case <strong>of</strong> very large pores via convective (bulk)diffusionKnudsen diffusion occurs when <strong>the</strong> membrane pores are smaller than <strong>the</strong> mean free path <strong>of</strong><strong>the</strong> gas molecules [40]. In this case, <strong>the</strong> gas molecules collide more frequently with <strong>the</strong> porewalls than with each o<strong>the</strong>r. The Knudsen diffusion coefficient (D K ) <strong>of</strong> a gas (A) is inverselyproportional to <strong>the</strong> square-root <strong>of</strong> <strong>the</strong> molecular weight (MW) as shown in Equation 2-17below, where d p is <strong>the</strong> pore radius <strong>of</strong> <strong>the</strong> substrate. There<strong>for</strong>e, gases with a low molecularweight have a greater ability to diff<strong>use</strong> through <strong>the</strong> membrane than <strong>high</strong>er molecular weightgases and hence separation occurs.DK ( A)d 8RT= Equation 2-17p3 πMW(A)Fick’s law describes <strong>the</strong> relationship between <strong>the</strong> flux (J) and <strong>the</strong> concentration (C) driving<strong>for</strong>ce <strong>of</strong> a gas (A) as shown in Equation 2-18. If it is assumed that ideal gas behaviour applies,Equation 2-18 is integrated to Equation 2-19, where ∆p is <strong>the</strong> pressure drop, A is <strong>the</strong>membrane area and ∆x is <strong>the</strong> thickness.∂C(A)J( A)= −D(A)Equation 2-18∂xD(A)∆p(A)J( A)=Equation 2-19RT∆xSubstituting Equation 2-17 into 2-19, gives <strong>the</strong> following expression <strong>for</strong> <strong>the</strong> flux (J K ) <strong>of</strong> a gas(A) through substrate via <strong>the</strong> Knudsen diffusion mechanism.JK ( A)d ∆p8RT= Equation 2-20p ( A)3RT∆xπMW(A)15


Using Equation 2-1, to relate <strong>the</strong> flux to <strong>the</strong> permeance (P’ K ), <strong>the</strong>n gives <strong>the</strong> followingexpression <strong>for</strong> <strong>the</strong> permeance <strong>of</strong> a gas (A) through substrate via <strong>the</strong> Knudsen diffusionmechanism.P'K ( A)dp 8RT= Equation 2-213RT∆xπMW( A)The Knudsen selectivity <strong>for</strong> a pair <strong>of</strong> gases (A and B) is simply described by <strong>the</strong> ratio <strong>of</strong> <strong>the</strong>irrespective Knudsen diffusion coefficients, which reduces to <strong>the</strong> inverse square-root ratio <strong>of</strong><strong>the</strong> molecular weights as shown in Equation 2-22. The Knudsen selectivities <strong>for</strong> <strong>the</strong> gasmolecules <strong>of</strong> interest in relation to CO2 are presented in Table 2-1.DK( A)MW(B)αK ( A / B)= =Equation 2-22D MWK ( B)( A)Table 2-1 : Kinetic diameter, molecular weight and Knudsen selectivities <strong>of</strong> gasmolecules <strong>of</strong> interest.GasKinetic Diameter Molecular Weight Knudsen Selectivity(nm)(g/mol) In Relation to CO 2CO 2 0.330 44 1.00O 2 0.346 32 0.85H 2 S 0.360 34 0.88N 2 0.364 28 0.80CO 0.376 28 0.80CH 4 0.380 16 0.60When <strong>the</strong> pore radius increases fur<strong>the</strong>r bulk / convective flow will dominate. Under <strong>the</strong>seconditions, <strong>the</strong> gas molecules will collide more frequently with each o<strong>the</strong>r and no separationwill occur.Convective flow through a macroporous medium is described by Darcy’s Law, which is <strong>the</strong>proportional relationship between <strong>the</strong> velocity (v), density (ρ) and pressure drop <strong>of</strong> <strong>the</strong> gas(∆p) as shown in Equation 2-23 [51].Equation 2-23 is also known as <strong>the</strong> Darcy-Weisbach equation, where λ is <strong>the</strong> coefficient <strong>of</strong>laminar or turbulent flow and d m is <strong>the</strong> membrane diameter.16


2x ρ(A)v(A)∆ p(A)= λ ∆Equation 2-23d 2mRelating <strong>the</strong> velocity <strong>of</strong> <strong>the</strong> component to <strong>the</strong> volumetric flow (Q) using Equation 2-24,substituting back into Equation 2-19 and rearranging <strong>for</strong> flow gives Equation 2-25.2⎛ dm( ) ( )2 ⎟ ⎞QA= vAπ ⎜Equation 2-24⎝ ⎠QC ( A)⎛ d 2∆pd= π ⎜Equation 2-25⎝2m ⎞( A)m⎟2 ⎠ ∆xλρ( A)Convective flow selectivity (α C(A/B) ) <strong>for</strong> ideal gas behaviour is equal to unity.2.2.3.2 Nanoporous MembranesNanoporous <strong>membranes</strong> are generally made from <strong>carbon</strong> and zeolites [40]. These <strong>nanoporous</strong><strong>membranes</strong> are usually asymmetric, i.e. <strong>the</strong>y have a selective membrane layer on top <strong>of</strong> asupport. Porous <strong>membranes</strong> can be made without a support but <strong>the</strong>y generally have littlemechanical strength due to brittleness.Nanoporous <strong>carbon</strong> (NPC) <strong>membranes</strong> are porous <strong>membranes</strong> made from <strong>the</strong> pyrolysis(heating) <strong>of</strong> a polymer with or without a porous support.The major advantage <strong>of</strong> <strong>membranes</strong> like NPC <strong>membranes</strong> is an ability to operate at <strong>high</strong><strong>temperature</strong>s. However, unlike polymeric counterparts which are already <strong>use</strong>d commercially<strong>for</strong> natural gas separation, NPC <strong>membranes</strong> are still only at a research level. There are only acouple <strong>of</strong> companies (namely, Carbon Membranes Ltd and UBE Industries) producing NPC<strong>membranes</strong> <strong>for</strong> specific applications [23].In <strong>nanoporous</strong> <strong>membranes</strong>, with pore sizes < 30 Å, o<strong>the</strong>r flow mechanisms dominate overKnudsen diffusion and bulk flow.If <strong>the</strong> pore sizes are relatively large (between 10 and 30 Å), gas separation can occur via <strong>the</strong>partial condensation <strong>of</strong> one component <strong>of</strong> <strong>the</strong> gas mixture. The condensed componentproceeds to diff<strong>use</strong> through <strong>the</strong> membrane as a liquid and separation takes place. Carbonnanotubes have a diameter generally in <strong>the</strong> range <strong>of</strong> 1 to 2 nm (10 to 20 Å). Andrews et al[52] have demonstrated through a packed bed arrangement <strong>of</strong> <strong>carbon</strong> nanotubes that CO 217


selectively condenses in <strong>the</strong> pores <strong>of</strong> <strong>the</strong> nanotubes over N 2 and <strong>the</strong>re<strong>for</strong>e produces <strong>the</strong>separation <strong>of</strong> <strong>the</strong> two gases.As <strong>the</strong> pore sizes decrease below around 10 Å, as is <strong>the</strong> case <strong>for</strong> NPC <strong>membranes</strong>, amechanism <strong>of</strong> surface diffusion becomes important [20]. More polar gas molecules areadsorbed onto <strong>the</strong> surface <strong>of</strong> <strong>the</strong> membrane and move through via surface diffusion.Surface diffusion is <strong>the</strong> mechanism proposed by Rao and Sircar <strong>for</strong> gas flow through <strong>the</strong>irSSF (Surface Selective Flow) membrane manufactured from <strong>the</strong> pyrolysis <strong>of</strong> polyvinylindenechloride (PVDC). In <strong>the</strong> case <strong>of</strong> SSF <strong>membranes</strong>, <strong>the</strong> pore sizes are slightly larger than <strong>the</strong>gas molecules and as such <strong>the</strong> larger and more polar molecules are adsorbed onto <strong>the</strong> porewalls and penetrate through <strong>the</strong> membrane [20]. Surface diffusion follows a transportmechanism activated by <strong>temperature</strong>, which is described in detail in Section 2.6.Gas separation via molecular sieving occurs when <strong>the</strong> pore sizes <strong>of</strong> <strong>the</strong> membrane approach<strong>the</strong> size <strong>of</strong> gas molecules (< 5 Å). Gas molecules smaller than <strong>the</strong> membrane pores will passthrough whilst <strong>the</strong> larger molecules will be retained on <strong>the</strong> <strong>high</strong> pressure side <strong>of</strong> <strong>the</strong>membrane. Like surface adsorption, transport via molecular sieving also follows a<strong>temperature</strong> activated transport mechanism, which is also fur<strong>the</strong>r described in Section 2.6.2.2.3.3 Nonporous MembranesNonporous <strong>membranes</strong> <strong>use</strong>d <strong>for</strong> gas separation are generally made from polymers and can besupported (asymmetric spiral wound sheets) or unsupported (integrally skinned hollowfibres). Membranes made from Palladium <strong>use</strong>d in hydrogen separation are ano<strong>the</strong>r example <strong>of</strong>nonporous <strong>membranes</strong>. These <strong>membranes</strong> are similar to NPC <strong>membranes</strong> in that <strong>the</strong>ycomprise a selective membrane layer upon a porous support [53].For mechanical strength, polymeric <strong>membranes</strong> are usually operated at <strong>temperature</strong>s below<strong>the</strong> glass transition <strong>temperature</strong>, where <strong>the</strong> structure <strong>of</strong> <strong>the</strong> polymer changes is glassy ra<strong>the</strong>rthan rubbery. Additionally, glassy <strong>membranes</strong> have a <strong>high</strong> free volume and are size selectivewhereas rubbery <strong>membranes</strong> select on <strong>the</strong> basis <strong>of</strong> solubility.Common polymers <strong>use</strong>d in gas separation include cellulose acetate and polyimide. Inparticular, polyimide is a popular choice <strong>for</strong> gas separation (particularly CO 2 from natural gas- CH 4 ) beca<strong>use</strong> it has a <strong>high</strong> glass transition <strong>temperature</strong> (300 °C).Gas transport through a nonporous membrane, such as a polymeric membrane, is described bya solution diffusion mechanism where <strong>the</strong> permeability (P SD ) is <strong>the</strong> product <strong>of</strong> <strong>the</strong> diffusivity(D SD ) and <strong>the</strong> solubility (S SD ) as shown in Equation 2-26.18


P = D SEquation 2-26SD( A)SD(A)SD(A)Diffusivity is a kinetic term, which gives an indication <strong>of</strong> how fast a gas molecule travelsthrough a membrane. Solubility is a <strong>the</strong>rmodynamic term that represents <strong>the</strong> amount <strong>of</strong> gasmolecules adsorbed by <strong>the</strong> membrane. This selectivity <strong>of</strong> two gases (α A/B ) is <strong>the</strong>n simply <strong>the</strong>ratio <strong>of</strong> <strong>the</strong> permeances or permeability.Equation 2-26 is also sometimes <strong>use</strong>d to simply model <strong>the</strong> transport <strong>of</strong> gases through NPC<strong>membranes</strong> as being a combination <strong>of</strong> diffusion and adsorption representing <strong>the</strong> molecularsieving and surface diffusion mechanisms.2.2.3.4 Membranes <strong>of</strong> Mixed MaterialsOne example <strong>of</strong> <strong>membranes</strong> comprising <strong>of</strong> different materials are mixed matrix <strong>membranes</strong>.Mixed matrix <strong>membranes</strong> are generally polymeric <strong>membranes</strong> impregnated with molecularsieving materials, such as <strong>nanoporous</strong> <strong>carbon</strong> or zeolites. Mixed matrix <strong>membranes</strong> arepromising materials as <strong>the</strong>y have selectivity from <strong>the</strong> molecular sieving domains [54, 55].2.2.3.5 SummaryA diagrammatic summary <strong>of</strong> <strong>the</strong> transport mechanisms discussed is presented in Figure 2-5below. The transport mechanisms are grouped according to <strong>the</strong> membrane pore size. Typicalmaterials <strong>use</strong>d to manufacture <strong>membranes</strong> <strong>of</strong> various pore sizes are also presented.TypicalMaterialsPolymericMembranesNanoporous CarbonMembranesCarbonNanotubesSilicaCeramicsMacroporousMaterialsTransportMechanismMolecularSievingCapillaryCondensationBulkDiffusionCO 2N 2 or CH 4SolutionDiffusionSurfaceDiffusionKnudsenDiffusionNone 0 5 10 30 100 1000Pore Size (Å)Figure 2-5 : Summary <strong>of</strong> transport mechanisms indicating as <strong>the</strong> size <strong>of</strong> <strong>the</strong> membranepores decrease, transport changes from bulk flow to molecular sieving.19


2.3 Manufacture <strong>of</strong> NPC MembranesThe three known crystalline <strong>for</strong>ms <strong>of</strong> <strong>carbon</strong> are as follows; (i) graphite, (ii) diamond and (iii)fullerenes. The <strong>nanoporous</strong> <strong>carbon</strong> structure is most closely related to <strong>the</strong> fullerene group <strong>of</strong><strong>carbon</strong>. In each <strong>for</strong>m <strong>of</strong> <strong>carbon</strong> <strong>the</strong> atoms are arranged differently leading to alteredproperties. Graphite, <strong>for</strong> example, is one <strong>of</strong> <strong>the</strong> s<strong>of</strong>test known structures whilst diamond is <strong>the</strong>hardest. The <strong>carbon</strong> atoms in graphite have a sp 2 bonding arrangement whereas <strong>the</strong> <strong>carbon</strong>atoms in diamond have a sp 3 bonding arrangement.Fullerenes are closed ball-like structures with <strong>the</strong> <strong>carbon</strong> in a pentagon or hexagon ring<strong>for</strong>mation and were discovered in 1985 by Kroto et al [56]. The bonding arrangement inFullerenes is predominately sp 2 bonding with some sp 3 bonding [57].Figure 2-6 : Structure <strong>of</strong> a typical fullerene (<strong>the</strong> Buckminster Fullerene C60) discoveredin 1985 by Kroto et al [56].A ribbon-like structure consisting <strong>of</strong> <strong>carbon</strong> sheets was originally proposed as <strong>the</strong> structure <strong>of</strong>NPC by Jenkins and Kawamura in 1976 [58]. However, in relation to work on carbogenicmolecular sieves (CMS) manufactured from NPC, Foley et al [59] presented new <strong>the</strong>oriesrelating <strong>the</strong> microstructure <strong>of</strong> NPC to fullerenes and proposed “mixed ring structures as <strong>the</strong>ultimate building blocks”. This proposal comes from <strong>the</strong> postulation that defects in <strong>the</strong> <strong>for</strong>ms<strong>of</strong> rings must be present to enable <strong>the</strong> <strong>high</strong> curvature that explains <strong>the</strong> narrow pore sizedistribution <strong>of</strong> 4 – 5 Å. Building on <strong>the</strong>se findings, Acharya et al [60] have obtained threedimensionalstructures <strong>for</strong> <strong>nanoporous</strong> <strong>carbon</strong> using simulation on <strong>the</strong> basis <strong>of</strong> known keyproperties <strong>of</strong> NPC, such as complete sp 2 <strong>carbon</strong> hybridisation. The model developed simulates<strong>the</strong> real-life pyrolysis process by beginning with basic building blocks to randomly <strong>for</strong>mulatebonds between hexagonal sheets (refer to Figure 2-7), which lead to <strong>the</strong> curvature. Acharya etal [60] have concluded that <strong>the</strong> ribbon-like structure originally proposed by Jenkins andKawamura [58] does not represent <strong>the</strong> curved structures obtained from <strong>the</strong> pyrolysis <strong>of</strong>polymers and is more applicable to glassy <strong>carbon</strong>s.20


Figure 2-7 : Structure <strong>of</strong> a hexagonal sheet, which when randomly bonded to o<strong>the</strong>rsheets <strong>for</strong>ms <strong>the</strong> curved <strong>nanoporous</strong> <strong>carbon</strong> structure. Figure adapted from [61].2.3.1 Supported NPC MembranesThe manufacture <strong>of</strong> a supported NPC membrane is undertaken in a series <strong>of</strong> steps. The initialstep requires <strong>the</strong> selection <strong>of</strong> an appropriate polymer precursor and porous support. In <strong>the</strong> nextstep, <strong>the</strong> polymer is deposited onto <strong>the</strong> porous support using a selected technique. Theresultant polymer-coated porous support is <strong>the</strong>n pyrolysised (heated) under an inertatmosphere and a <strong>nanoporous</strong> <strong>carbon</strong> layer is <strong>for</strong>med on <strong>the</strong> porous support. Finally, <strong>the</strong>polymer deposition and pyrolysis steps are repeated a certain number <strong>of</strong> times in order toincrease <strong>the</strong> amount <strong>of</strong> NPC deposited.2.3.1.1 PolymerA variety <strong>of</strong> polymers can be <strong>use</strong>d to manufacture NPC <strong>membranes</strong> on condition that <strong>the</strong>y are<strong>the</strong>rmosetting and hence undergo cross-linking at <strong>high</strong> <strong>temperature</strong>s and so prevent <strong>the</strong><strong>for</strong>mation <strong>of</strong> graphite layers, which decrease <strong>the</strong> microporosity <strong>of</strong> <strong>the</strong> material [62]. Polymersthat can be <strong>use</strong>d include cellulose, PVDC (polyvinylindene chloride), phenolic resin,polyimide and PFA (polyfurfuryl alcohol).It is difficult to directly compare <strong>the</strong> permeances and selectivities obtained from differentpolymer precursors as deposition techniques, pyrolysis conditions and <strong>the</strong> gas moleculesemployed are <strong>of</strong>ten not <strong>the</strong> same.From a general analysis <strong>of</strong> <strong>the</strong> literature data, NPC <strong>membranes</strong> made from PVDC displayextremely <strong>high</strong> permeances that are several orders <strong>of</strong> magnitude <strong>high</strong>er than those made frompolyimide. Similarly, NPC <strong>membranes</strong> made from <strong>the</strong> pyrolysis <strong>of</strong> phenolic resin or PFAexhibit permeances an order <strong>of</strong> magnitude lower again than those made from polyimide.Selectivities are a little harder to compare beca<strong>use</strong> <strong>the</strong> gas molecules employed <strong>for</strong> measuringpermeances differ due to different targeted commercial applications. The published results as21


shown in Table 2-2 at <strong>the</strong> end <strong>of</strong> this section suggest that <strong>membranes</strong> with lower permeances(such as those made from PFA) display <strong>high</strong>er selectivities as is consistent with Robeson’s<strong>the</strong>ory discussed earlier in Section 2.2.2. Some detail <strong>of</strong> <strong>the</strong> results presented in <strong>the</strong> literature<strong>for</strong> each polymer is presented below.2.3.1.1.1 CelluloseHollow-fibre NPC <strong>membranes</strong> made from pyrolysis cellulose were patented in 1999 [13] withcommercialisation starting soon after by a company named Carbon Membranes Ltd, Thiscompany closed in 2001 [12]. These NPC <strong>membranes</strong> with a CO 2 /CH 4 selectivity around 3 at25°C [8] have been <strong>use</strong>d <strong>for</strong> extensive research into <strong>the</strong> transport mechanism behind <strong>the</strong>separation <strong>of</strong> gas by NPC <strong>membranes</strong> [8-13].2.3.1.1.2 Polyvinylindene Chloride (PVDC)Air Products and Chemicals have commercialised a membrane <strong>for</strong> <strong>the</strong> separation <strong>of</strong> hydrogencalled <strong>the</strong> Surface Selective Flow (SSF TM ) membrane, which is made from <strong>the</strong> pyrolysis <strong>of</strong>PVDC on tubular supports [14, 20, 63-66]. In an initial paper, Rao and Sircar reported H 2 /N 2and CH 4 /H 2 selectivities <strong>of</strong> 1.7 and 5, respectively with extremely <strong>high</strong> correspondingpermeances [20]. In later work, Anand et al reported that <strong>the</strong> SSF membrane can achieve a<strong>high</strong> pressure (retentate side) H 2 recovery <strong>of</strong> 50 % from effluent gas containing methane,ethane, ethylene, propane, propylene and <strong>carbon</strong> dioxide [63].2.3.1.1.3 Phenolic ResinFuertes [21, 67] has also produced a selective flow NPC membrane from <strong>the</strong> pyrolysis <strong>of</strong>phenolic resin. Oxidative treatment following pyrolysis was <strong>use</strong>d to increase <strong>the</strong> pore size to 5– 7 Å. The increased pore size allows larger and more polar molecules such as hydro<strong>carbon</strong>sto selectively adsorb and diff<strong>use</strong> through <strong>the</strong> membrane. Selectivities as <strong>high</strong> as 87 arereported <strong>for</strong> CO 2 /CH 4 [67].2.3.1.1.4 PolyimidePolyimide is already established as a material <strong>for</strong> <strong>use</strong> in non-porous polymeric gas separation<strong>membranes</strong> due to <strong>the</strong> <strong>high</strong> selectivity <strong>for</strong> CO 2 and <strong>the</strong> stability <strong>for</strong> operating <strong>temperature</strong>s upto 300 °C. Pyrolysis <strong>of</strong> polyimide <strong>for</strong> <strong>the</strong> manufacture <strong>of</strong> NPC has also been researched.Kusakabe et al [68] report CO 2 /N 2 selectivities <strong>for</strong> NPC <strong>membranes</strong> manufactured frompolyimide as <strong>high</strong> as 51. Hayashi et al [69] have manufactured NPC <strong>membranes</strong> from <strong>the</strong>pyrolysis <strong>of</strong> polyimide and have decreased <strong>the</strong> pore size even fur<strong>the</strong>r by <strong>the</strong> chemical vapourdeposition (CVD) <strong>of</strong> propylene at 650 °C. CVD was shown to increase <strong>the</strong> CO 2 /N 2 selectivityat 35 °C from 47 to 73.22


2.3.1.1.5 Polyfurfuryl Alcohol (PFA)PFA has also been widely reported in <strong>the</strong> literature <strong>for</strong> <strong>the</strong> manufacture <strong>of</strong> NPC. PFA is afuran polymer and is derived from wood and o<strong>the</strong>r biomass sources. PFA has very <strong>high</strong><strong>carbon</strong> yield <strong>of</strong> 50 % on pyrolysis, which makes it an attractive material <strong>for</strong> <strong>the</strong> manufacture<strong>of</strong> NPC. Bird and Trimm completed <strong>the</strong> early work using PFA <strong>for</strong> gas separation <strong>membranes</strong>but did not have any success with making a membrane without significant cracks [70].Following this, <strong>the</strong>re was fur<strong>the</strong>r research into <strong>the</strong> <strong>use</strong> <strong>of</strong> PFA to manufacture NPC as amolecular sieve material [62, 71, 72].The research <strong>the</strong>n developed to <strong>the</strong> current situation where PFA is usually coated onto aporous support to make an NPC membrane (<strong>of</strong>ten referred to as <strong>the</strong> “Carbon Molecular SieveMembrane”) [19, 25, 43, 73-82]. Such <strong>membranes</strong> are reported with O 2 /N 2 permselectivitiesranging from 2 to 60. Most <strong>of</strong> <strong>the</strong> research published has measured permeances <strong>of</strong> O 2 and N 2through NPC <strong>membranes</strong> at room <strong>temperature</strong> with <strong>the</strong> proposed application being airseparation. There is little data published on CO 2 and CH 4 separation <strong>for</strong> NPC Membranesmade from PFA.As discussed earlier, it is difficult to directly compare permeance and selectivities <strong>of</strong> NPC<strong>membranes</strong> made from different polymers. However, <strong>the</strong> comparison <strong>of</strong> H 2 /N 2permselectivity between NPC <strong>membranes</strong> made from PFA [75] and those made frompolyimide [68] is presented in Table 2-2 below. The pyrolysis <strong>temperature</strong> in both cases was600 °C.Table 2-2 : H 2 /N 2 permselectivity <strong>for</strong> NPC <strong>membranes</strong> made from PFA and polyimide.The NPC <strong>membranes</strong> made from PFA have a <strong>high</strong>er selectivity but lower permeancethan those made from polyimide.PolymerH 2 Permeance N 2 PermeanceH 2 /N 2(mol/m 2 /s/Pax10 10 ) (mol/m 2 /s/Pax10 10 ) PermselectivityPFA [75] 14 0.5 30Polyimide [68] 300 40 8Subsequent discussions on deposition method and pyrolysis <strong>of</strong> supported NPC <strong>membranes</strong>will focus on literature that has <strong>use</strong>d PFA. PFA has been selected <strong>for</strong> this research as <strong>the</strong>appropriate polymer precursor <strong>for</strong> <strong>the</strong> manufacture <strong>of</strong> supported NPC <strong>membranes</strong> on <strong>the</strong> basis<strong>of</strong> <strong>the</strong> <strong>high</strong> selectivity and yield <strong>of</strong> <strong>carbon</strong> produced.23


2.3.1.2 Porous SupportThe porous support must contain pore sizes larger than <strong>the</strong> pore sizes <strong>of</strong> <strong>the</strong> resultant NPCmembrane. Mesoporous materials such as graphite, alumina or porous stainless steel can be<strong>use</strong>d. With <strong>the</strong> porous support providing no contribution to gas separation, <strong>the</strong> choice <strong>of</strong> aporous support is governed by mechanical strength, ease <strong>of</strong> manufacture and cost. Porousstainless steel supports are advantageous beca<strong>use</strong> <strong>of</strong> mechanical strength and reduced costgiven <strong>the</strong>y can be welded to ensure a gas tight fitting suitable <strong>for</strong> <strong>high</strong> <strong>temperature</strong> operation.The Foley research group [73, 74, 79, 80, 82] <strong>use</strong> 2 micron porous stainless steel supports in<strong>the</strong> <strong>for</strong>m <strong>of</strong> disks with a diameter <strong>of</strong> 46 mm and thickness <strong>of</strong> 1 mm. The same research group[43, 75, 77, 78, 83, 84] has also <strong>use</strong>d tubular stainless steel supports to manufacture NPC<strong>membranes</strong>. The tubular support is <strong>use</strong>d <strong>for</strong> creating membrane modules.The porous supports are <strong>of</strong>ten cleaned to remove contaminants such as oxides and greasesprior to coating. Acharya and Foley [74] cleaned <strong>the</strong> porous supports with chlor<strong>of</strong>orm in asonification bath <strong>for</strong> 15 mins followed by air drying, be<strong>for</strong>e depositing <strong>the</strong> polymericprecursor solution.2.3.1.3 Modified Porous SupportDuring <strong>the</strong> pyrolysis process, <strong>the</strong> polymer enters <strong>the</strong> pores <strong>of</strong> <strong>the</strong> support creating a membranethat is thick and heavy with <strong>carbon</strong>, which can lead to cracking <strong>of</strong> <strong>the</strong> NPC membrane surface[27, 28].One method <strong>for</strong> producing ultra-thin crack-free membrane surfaces involves filling <strong>the</strong> pores<strong>of</strong> <strong>the</strong> porous support with nanoparticles prior to adding <strong>the</strong> membrane layer. Thenanoparticles act to prevent <strong>the</strong> polymer entering <strong>the</strong> pores <strong>of</strong> <strong>the</strong> support during pyrolysisand ensure that <strong>carbon</strong> is only <strong>for</strong>med as a thin layer on <strong>the</strong> surface <strong>of</strong> <strong>the</strong> support [27, 28].With a support pore size around 2 micron, Rajagopalan et al [28] have <strong>use</strong>d silicananoparticles ranging from 40 to 500 nm in size to fill <strong>the</strong> support. In this case, <strong>the</strong>modification <strong>of</strong> <strong>the</strong> support has provided a two-order <strong>of</strong> magnitude increase in gas permeance,whilst maintaining selectivity. Merritt et al [27] have <strong>use</strong>d Silica nanoparticles and <strong>carbon</strong>black particles to modify a porous stainless steel support resulting in a two-order <strong>of</strong>magnitude increase in gas permeance.Adding a γ-alumina layer to a porous support prior to <strong>for</strong>ming a <strong>nanoporous</strong> <strong>carbon</strong>membrane is ano<strong>the</strong>r method <strong>for</strong> reducing <strong>the</strong> <strong>carbon</strong> membrane thickness and reduces <strong>the</strong>chance <strong>of</strong> cracking. Sea and Lee [85] have altered tubular α-alumina supports by adding a24


coating <strong>of</strong> γ-alumina. The presence <strong>of</strong> <strong>the</strong> γ-alumina layer on top <strong>of</strong> <strong>the</strong> support reduces <strong>the</strong>overall support pore size from 1 micron to between 5 to 7 nm and thus prevents <strong>the</strong> polymerentering <strong>the</strong> support during pyrolysis. This technique <strong>of</strong> adding a γ-alumina layer has been<strong>use</strong>d by o<strong>the</strong>r researchers <strong>for</strong> reducing <strong>the</strong> pore size <strong>of</strong> a support [86-88].2.3.1.4 Deposition MethodThe first step in depositing <strong>the</strong> polymer on <strong>the</strong> porous support is <strong>the</strong> preparation <strong>of</strong> apolymeric precursor solution. Polymeric precursor solutions containing 25 to 60 wt% PFA inacetone have been reported in <strong>the</strong> literature [75, 78, 83] depending on <strong>the</strong> viscosity required<strong>for</strong> <strong>the</strong> chosen deposition method.Deposition methods include brush-coating, spray-coating or ultrasonic deposition. For eachmethod, <strong>the</strong> porous support is rotated at a constant speed to allow uni<strong>for</strong>m coverage <strong>of</strong> <strong>the</strong>precursor. For <strong>the</strong> disk-shaped porous supports, <strong>the</strong> rotation is achieved using a spin coater.Brush-coating is <strong>the</strong> simplest method <strong>of</strong> deposition using a paint brush to apply <strong>the</strong> polymericprecursor (60 wt% PFA) to <strong>the</strong> porous support. Brush-coating has been <strong>use</strong>d by Acharya et al[73] to produce NPC <strong>membranes</strong> with an O 2 /N 2 selectivity <strong>of</strong> 3.In subsequent work, Acharya and Foley [74, 76] have sprayed <strong>the</strong> polymeric precursor (50wt% PFA) to <strong>the</strong> support using an external mix airbrush (Badger Model 250). The NPC<strong>membranes</strong> produced using this spray-coating method demonstrated O 2 /N 2 selectivitiesranging from 3 to 9.Shiflett and Foley [83] argue that ultrasonic deposition <strong>of</strong> <strong>the</strong> polymeric precursor (30 wt%PFA) provides a greater degree <strong>of</strong> control and <strong>the</strong>re<strong>for</strong>e results in better reproducibly. Usingan automated ultrasonic spray system designed and built by Sono-Tek Corporation (USA),Shiflett and Foley [75] have manufactured 7 <strong>membranes</strong> under <strong>the</strong> same pyrolysis conditions(450 °C and 2 hours). The O 2 /N 2 selectivity produced ranged from 2 to 6 with permeances allwithin <strong>the</strong> same order <strong>of</strong> magnitude. In his patent <strong>for</strong> <strong>the</strong> NPC Membrane using ultrasonicdeposition, Shiflett [77] has quoted O 2 /N 2 selectivities from 2 to as <strong>high</strong> as 60. It is suggestedthat <strong>high</strong>er selectivities are observed using ultrasonic deposition due to <strong>the</strong> thinner and moreuni<strong>for</strong>m coats obtained, which have less cracks and inconsistencies.2.3.1.5 PyrolysisThe generic method <strong>for</strong> <strong>the</strong> pyrolysis referenced in <strong>the</strong> literature [75, 78, 83] is as follows; <strong>the</strong>polymer on <strong>the</strong> porous support is heated under an inert atmosphere increasing at a set rate to apredetermined <strong>temperature</strong> (termed <strong>the</strong> pyrolysis or soak <strong>temperature</strong>) where it is held at that25


<strong>temperature</strong> <strong>for</strong> a predetermined time period (termed <strong>the</strong> pyrolysis or soak time). The newly<strong>for</strong>med NPC membrane is <strong>the</strong>n allowed to cool to room <strong>temperature</strong> under <strong>the</strong> inertatmosphere. An example <strong>of</strong> <strong>the</strong> furnace <strong>temperature</strong> pr<strong>of</strong>ile with time is given in Figure 2-8.Furnace Temperature (°C)800Oven HeatingOven Off (Natural Cooling)700Soak Temperature & Time600500400300200100Solvent Evaporation00 100 200 300 400 500 600 700 800 900 1000Time (min)Figure 2-8 : Typical furnace <strong>temperature</strong> pr<strong>of</strong>ile <strong>for</strong> <strong>the</strong> pyrolysis <strong>of</strong> a polymer toproduce <strong>nanoporous</strong> <strong>carbon</strong>.Pyrolysing PFA under inert conditions liberates H 2 O, CO 2 , CO, CH 4 and H 2 gases, leaving a<strong>nanoporous</strong> <strong>carbon</strong> structure according to <strong>the</strong> following reactions.600°C PFA→ H2O + FA + CO2 + CO + CH4 + H2+ Cwhere PFA = Polyfurfuryl Alcohol FA = Furfuryl AlcoholO CH 2O CH 2OH O CH 2OH26


Pyrolysis <strong>of</strong> PFA is generally per<strong>for</strong>med under a He or Ar atmosphere. Shiflett [89] obtained<strong>the</strong> <strong>high</strong>est O 2 /N 2 selectivity factors when using such an inert atmosphere. Geiszler and Koros[90] have demonstrated whilst making NPC <strong>membranes</strong> from <strong>the</strong> pyrolysis <strong>of</strong> polyimide, thatresidual trace levels <strong>of</strong> O 2 in <strong>the</strong> inert atmosphere (> 3 ppm) will affect <strong>the</strong> structure <strong>of</strong> <strong>the</strong><strong>nanoporous</strong> <strong>carbon</strong> <strong>for</strong>med.The final pyrolysis <strong>temperature</strong> has a significant impact on <strong>the</strong> resulting per<strong>for</strong>mance <strong>of</strong> <strong>the</strong>NPC membrane. The evolution <strong>of</strong> <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong> structure with increasing pyrolysis<strong>temperature</strong> has been studied <strong>for</strong> <strong>carbon</strong> molecular sieves (CMS) derived from PFA [59, 62,71]. Foley et al [59] explain that as <strong>the</strong> pyrolysis <strong>temperature</strong> is increased from 300°C, <strong>the</strong>PFA precursor begins to produce <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong>. At <strong>temperature</strong>s above 500°C, <strong>the</strong>hydroxyl groups disappear and when <strong>the</strong> pyrolysis <strong>temperature</strong> is increased fur<strong>the</strong>r beyond600°C, dehydrogenation occurs and finally at <strong>temperature</strong>s above 1200°C where all <strong>of</strong> <strong>the</strong>hydrogen has been removed, a rearrangement <strong>of</strong> <strong>the</strong> <strong>carbon</strong> atoms occurs that ultimatelyproduces a graphite-like structure.Nuclear magnetic resonance (NMR) spectra <strong>of</strong> CMS samples prepared at 400, 500 and 600°Cby Mariwala and Foley [71] have showed <strong>the</strong> presence <strong>of</strong> oxygen atoms, as are present in <strong>the</strong>original PFA polymer, at pyrolysis <strong>temperature</strong>s up to 500°C whilst at 600°C elementalanalysis reveals <strong>the</strong> presence <strong>of</strong> only <strong>carbon</strong> with less than 0.5% hydrogen. Powder X-raydiffraction on samples prepared at <strong>high</strong>er pyrolysis <strong>temperature</strong>s have revealed a shift in <strong>the</strong>d-spacing from between 3.9 and 5.2Å at 500°C to between 3.7 and 3.9Å at 800°C and <strong>the</strong>n0.36 Å at 1000°C.The narrowing <strong>of</strong> <strong>the</strong> pore size distribution with increasing pyrolysis <strong>temperature</strong> is said to beca<strong>use</strong>d by <strong>the</strong> growth <strong>of</strong> <strong>the</strong> aromatic microdomains which also ca<strong>use</strong> a subsequent increasein porosity. However, <strong>the</strong> removal <strong>of</strong> oxygen and <strong>carbon</strong>-hydrogen bonds with increasingpyrolysis <strong>temperature</strong> leads to a fragile structure and <strong>the</strong> collapse <strong>of</strong> <strong>the</strong> micropores [71].In a more recent study on bulk <strong>nanoporous</strong> <strong>carbon</strong>, Burket et al [61] <strong>use</strong>d a variety <strong>of</strong>characterisation techniques to fur<strong>the</strong>r understand <strong>the</strong> <strong>carbon</strong>isation <strong>of</strong> PFA in <strong>the</strong> range from300°C up to 600°C. It was concluded that between 300 and 400°C, <strong>carbon</strong> and polymer<strong>carbon</strong> chains co-exist resulting in a mesoporous structure. As <strong>the</strong> <strong>temperature</strong> is increasedbeyond 400 to 600°C, with <strong>the</strong> major weight loss at 500 °C, <strong>carbon</strong>isation completes and <strong>the</strong>mesopores collapse leaving a microporous structure.Looking specifically at <strong>membranes</strong>, Shiflett and Foley [75] have produced three NPC<strong>membranes</strong> from PFA at pyrolysis <strong>temperature</strong>s <strong>of</strong> 300, 450 and 600°C using <strong>the</strong> method <strong>of</strong>27


ultrasonic deposition. The results are presented in Table 2-3. Increasing <strong>the</strong> pyrolysis<strong>temperature</strong> from 300 to 600°C increased <strong>the</strong> oxygen and nitrogen permeance by an order <strong>of</strong>magnitude, which <strong>the</strong> authors attribute to <strong>the</strong> increase in porosity.Table 2-3 : Permeance and permselectivity <strong>of</strong> NPC <strong>membranes</strong> made from PFA atdifferent pyrolysis <strong>temperature</strong>s [75]. Increasing <strong>the</strong> <strong>temperature</strong> <strong>of</strong> pyrolysis has asignificant effect upon increasing <strong>the</strong> permeance. However, <strong>the</strong> permselectivity goesthrough a maximum at 450 °C.Pyrolysis Temp (°C)O 2 Permeance(mol/m 2 /s/Pax10 12 )@ 293 KN 2 Permeance(mol/m 2 /s/Pax10 12 )@ 293 KO 2 /N 2 Permselectivity300 0.11 0.01 13450 0.56 0.02 30600 1.27 0.46 3The permselectivity results are perhaps explained by <strong>the</strong> change in proportions <strong>of</strong> smaller tolarger pore sizes as <strong>the</strong> pyrolysis <strong>temperature</strong> is increased. Burns et al [91] explain <strong>the</strong>occurrence <strong>of</strong> a selectivity maximum <strong>for</strong> <strong>carbon</strong> molecular sieving materials as being ca<strong>use</strong>dby <strong>the</strong> subtle changes in <strong>the</strong> distribution <strong>of</strong> selective elements with pyrolysis <strong>temperature</strong>.Interestingly, <strong>the</strong> majority <strong>of</strong> NPC <strong>membranes</strong> that have been manufactured from PFA are atpyrolysis <strong>temperature</strong>s at and below 600°C [73, 75, 78, 83, 84]. Whilst a complete<strong>nanoporous</strong> <strong>carbon</strong> structure is produced at a pyrolysis <strong>temperature</strong> <strong>of</strong> 600°C with acontinuous reduction in pore size and increase in porosity achieved at <strong>temperature</strong>s greaterthan 600°C, <strong>the</strong> apparent fragility <strong>of</strong> <strong>the</strong> structure beyond 600°C makes it difficult to workwith in a membrane <strong>for</strong>m.In his PhD Thesis, Shiflett [89] discusses <strong>the</strong> effect <strong>of</strong> <strong>the</strong> heating ramp rate. Heating ramprates above 10 °C/min can lead to <strong>the</strong> <strong>for</strong>mation <strong>of</strong> bubbles on <strong>the</strong> surface due to insufficienttime <strong>for</strong> <strong>the</strong> evolving gases to escape. A value <strong>of</strong> 5 °C/min is commonly <strong>use</strong>d.Increasing <strong>the</strong> pyrolysis time is generally reported to have a similar impact on <strong>the</strong> structure <strong>of</strong><strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong> as increasing <strong>the</strong> pyrolysis <strong>temperature</strong>. Rajagopalan and Foley [82]have demonstrated that <strong>the</strong> pyrolysis time ca<strong>use</strong>s significant changes in <strong>the</strong> morphology <strong>of</strong> <strong>the</strong><strong>membranes</strong>. As <strong>the</strong> pyrolysis time is increased, <strong>the</strong> size <strong>of</strong> globular domains is increased and<strong>the</strong> structure becomes disordered, reducing <strong>the</strong> selectivity. A pyrolysis time <strong>of</strong> 2 hours alongwith a pyrolysis <strong>temperature</strong> <strong>of</strong> 450 °C <strong>for</strong> NPC membrane manufactured from PFA is mostcommonly reported in <strong>the</strong> literature [24, 73, 75, 77, 78, 83, 84].28


2.3.1.6 Number <strong>of</strong> CoatsThe amount <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> deposited onto <strong>the</strong> stainless steel support can be increasedwithout producing cracks by applying additional thin coats <strong>of</strong> polymer. The pyrolysis processis repeated <strong>for</strong> each additional polymer coat applied.Using a spin-coating deposition method, Rajagopalan and Foley [82] have reviewed <strong>the</strong>relationship between <strong>the</strong> number <strong>of</strong> coats applied, weight gain (<strong>carbon</strong> deposition) andpermeability. For <strong>the</strong> first two coats applied, <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong> fills <strong>the</strong> pores <strong>of</strong> <strong>the</strong>porous support. By <strong>the</strong> third coat, a NPC membrane layer is <strong>for</strong>med. They have also shownthat <strong>the</strong> permeability increases with increasing <strong>carbon</strong> deposition whilst <strong>the</strong> O 2 /N 2permselectivity reaches a maximum <strong>of</strong> 5 after 8 coats with <strong>the</strong> weight gain having reached 6mg/cm 2 . The reduction in selectivity is thought to be due to cracking in <strong>the</strong> initial layer <strong>of</strong> <strong>the</strong>membrane due to a long exposure to <strong>high</strong> <strong>temperature</strong>s.2.3.2 Unsupported NPC MembranesWhilst unsupported NPC <strong>membranes</strong> lack <strong>the</strong> mechanical strength <strong>of</strong> supported NPC<strong>membranes</strong>, <strong>the</strong>y <strong>of</strong>ten display a <strong>high</strong>er permeance due to having a thinner membrane layer.The manufacturing process <strong>for</strong> unsupported NPC <strong>membranes</strong> is much simpler than supportedNPC <strong>membranes</strong>. The polymer is cast as a thin film or extruded as a hollow fibre; <strong>the</strong> solventis <strong>the</strong>n evaporated through drying <strong>the</strong> polymer over several days and <strong>the</strong>n finally pyrolysed toproduce an unsupported <strong>carbon</strong> membrane.2.3.2.1 PolymerA polymer can be <strong>use</strong>d to make an unsupported <strong>carbon</strong> membrane provided it (a) undergoescross-linking at <strong>high</strong> <strong>temperature</strong>s to prevent <strong>the</strong> <strong>for</strong>mation <strong>of</strong> graphite layers as discussed inSection 2.3.1.1 and (b) can be produced as a flat sheet or hollow fibre.By far <strong>the</strong> most commonly <strong>use</strong>d polymer in <strong>the</strong> production <strong>of</strong> unsupported <strong>carbon</strong> <strong>membranes</strong>is polyimide as it is a well established polymer <strong>use</strong>d <strong>for</strong> gas separation <strong>membranes</strong>. Kapton(DuPont, USA) and Matrimid (Ciba Specialty Chemicals, Japan) are commonly <strong>use</strong>d as <strong>the</strong>yare commercially available. NPC <strong>membranes</strong> made from Kapton are generally morepermeable and less selective than those made from Matrimid [16, 17]. More recently, Liu et al[92] produced <strong>high</strong>ly selective and permeable unsupported NPC <strong>membranes</strong> from <strong>the</strong> novellow cost polymer poly(phthalazinone e<strong>the</strong>r sulfone ketone).29


2.3.2.2 Pyrolysis TemperatureThe pyrolysis <strong>temperature</strong> required to produce <strong>carbon</strong> is dependent on <strong>the</strong> original polymer.Barsema et al [29] <strong>carbon</strong>ised Matrimid at <strong>temperature</strong>s between 300 and 525°C and foundthat <strong>carbon</strong>isation did not begin to occur until 425°C. Similarly, Suda and Haraya [93]<strong>carbon</strong>ised Kapton at <strong>temperature</strong>s between 500 and 1000°C and found that <strong>carbon</strong>isation didnot begin until 527°C. As <strong>the</strong> <strong>temperature</strong> was increased up to 1000°C, <strong>the</strong> permeabilitydecreased whilst <strong>the</strong> CO 2 /N 2 permselectivity increased.Suda and Haraya [93] also reported a increase in permeability upon decreasing <strong>the</strong> heatingrate from 1.33 °C/min to 13.3 °C/min. As discussed in Section 2.3.1.5, this same result hasbeen reported <strong>for</strong> supported <strong>membranes</strong>.In a recent study, Su and Lua [94], have examined <strong>the</strong> effect <strong>of</strong> pyrolysis atmosphere on <strong>the</strong><strong>carbon</strong>isation <strong>of</strong> Kapton at 600 and 800°C. It was concluded that <strong>the</strong> permeance <strong>of</strong> all gaseswas lowest under vacuum, whilst O 2 /N 2 permselectivity was <strong>high</strong>est under vacuum andCO 2 /CH 4 permselectivity was <strong>high</strong>est under Argon.Characterisation <strong>of</strong> <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong> produced revealed <strong>the</strong> lowest d-spacing (anindication <strong>of</strong> pore size) under <strong>the</strong> Argon atmosphere, which concurs with <strong>the</strong> <strong>high</strong>permselectivity observed. Likewise, <strong>the</strong> BET surface area was lowest <strong>for</strong> <strong>the</strong> samples madeunder vacuum, which arguably explains <strong>the</strong> low permeance observed <strong>for</strong> <strong>the</strong> <strong>membranes</strong>made under vacuum. Whilst this study was undertaken with unsupported NPC <strong>membranes</strong>,<strong>the</strong> findings apply to <strong>the</strong> structure <strong>of</strong> <strong>the</strong> <strong>carbon</strong> produced and as such are also relevant tosupported NPC <strong>membranes</strong>.Additionally, a clear trend <strong>of</strong> decreasing permeance and increasing permselectivity withincreasing <strong>the</strong> pyrolysis <strong>temperature</strong> from 600 to 800°C was also reported [94].In <strong>the</strong> manufacture <strong>of</strong> NPC <strong>membranes</strong> from poly(phthalazinone e<strong>the</strong>r sulfone ketone) [92], itis argued that <strong>the</strong> <strong>the</strong>rmostabilisation <strong>of</strong> <strong>the</strong> membrane in air between 400 and 500°C prior topyrolysis under argon between 650 and 850°C allows <strong>for</strong>mation <strong>of</strong> oxygen bridges between<strong>the</strong> aromatic molecules that prevent rearrangement during <strong>carbon</strong>isation and hence result in agreater porosity.30


2.3.3 Characterisation2.3.3.1 Membrane ThicknessMembrane thickness is usually calculated by <strong>the</strong> mass <strong>of</strong> <strong>carbon</strong> deposited on <strong>the</strong> poroussupport and <strong>the</strong> known standard density <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> (1.6 g/cm 3 ). Strano and Foley[79] have also calculated an effective membrane thickness (δ) using <strong>the</strong> steady state flux (J ss )and driving <strong>for</strong>ce pressure (p 0 ) <strong>of</strong> Helium (a non-adsorbing gas) from <strong>the</strong> diffusion time-lag(τ) as shown in Equation 2-27, which <strong>the</strong>y have derived from Equation 2-11. The term R is<strong>the</strong> ideal gas constant and T is <strong>the</strong> experimental <strong>temperature</strong>.6J ssτRTδ = Equation 2-27p02.3.3.2 Surface StructureVarious microscopy techniques are <strong>use</strong>d to visually examine <strong>the</strong> surface structure <strong>of</strong> NPC<strong>membranes</strong>. Foley et al. [59] have <strong>use</strong>d High Resolution Tunnelling Electron Microscopy(HRTEM) and Energy Dispersive X-Ray Spectrometry (EDX) to capture and describe <strong>the</strong>changing structure <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> as it is pyrolysised from 400 to 800°C as wasdescribed previously.Shiflett et al. [84] have <strong>use</strong>d Atomic Force Microscopy (AFM), EDX and HRTEM to obtainfur<strong>the</strong>r in<strong>for</strong>mation on <strong>the</strong> membrane surface. AFM was <strong>use</strong>d to obtain in<strong>for</strong>mation on poresizes; however, <strong>the</strong> resolution <strong>of</strong> AFM was not powerful enough to detect <strong>the</strong> poredimensions. Some metallic impurities on <strong>the</strong> surface <strong>of</strong> <strong>the</strong> membrane were detected by EDX.2.3.3.3 Pore Size DistributionVarious porosity techniques are reported in <strong>the</strong> literature <strong>for</strong> characterising <strong>nanoporous</strong><strong>carbon</strong>. Shiflett et al [84] have <strong>use</strong>d methyl chloride porosity to confirm <strong>the</strong> presence <strong>of</strong>nanopores and to determine <strong>the</strong> size distribution from <strong>the</strong> Horvath-Kawazoe (HK) and Kelvinmodels. Adsorption <strong>of</strong> methyl chloride was measured using a microbalance from C. I.Electronics (USA). The results from <strong>the</strong> methyl chloride porosity showed an average poresize <strong>of</strong> 4.5 – 5 Å. Shiflett [89] has compared <strong>the</strong>se results with digitally enhanced HRTEMimages that showed slit shaped pores within <strong>the</strong> range <strong>of</strong> 3 to 6 Å.In <strong>the</strong>ir paper dedicated to <strong>the</strong> <strong>use</strong> <strong>of</strong> methyl chloride adsorption <strong>for</strong> calculating pore sizes,Mariwala and Foley [95] conclude that methyl chloride provides advantages over <strong>the</strong>traditional nitrogen adsorption technique due to <strong>high</strong>er rates <strong>of</strong> diffusion. Similarly, given <strong>the</strong>very slow diffusion rate <strong>of</strong> N 2 , Steel and Koros [17] have <strong>use</strong>d CO 2 adsorption and a31


Micromeritics ASAP 2010 apparatus to obtain a pore size distribution centred around 5 Å <strong>for</strong>NPC <strong>membranes</strong> manufactured from <strong>the</strong> pyrolysis <strong>of</strong> polyimide.A different technique, positron annihilation lifetime spectroscopy (PALS), has recently been<strong>use</strong>d by Tin et al [96] to characterise unsupported <strong>carbon</strong> <strong>membranes</strong> made from <strong>the</strong> pyrolysis<strong>of</strong> <strong>the</strong> polyimides Matrimid and P84 at 800°C. PALS characterises samples by bombarding<strong>the</strong> sample with positrons and <strong>the</strong>n analysing subsequent lifetime and intensity <strong>of</strong> producedortho-positronium. The lifetime <strong>of</strong> ortho-positronium annihilating in <strong>the</strong> pores (τ 2 ) indicates<strong>the</strong> size <strong>of</strong> <strong>the</strong> pores whist <strong>the</strong> number or intensity <strong>of</strong> ortho-positronium annihilating in <strong>the</strong>pores (I 2 ) indicates <strong>the</strong> number <strong>of</strong> pores or <strong>the</strong> porosity. The values <strong>of</strong> τ 2 and I 2 determined <strong>for</strong>NPC <strong>membranes</strong> made by Tin et al [96] are presented in Table 2-4.Table 2-4 : PALS data <strong>for</strong> NPC <strong>membranes</strong> manufactured from <strong>the</strong> pyrolysis <strong>of</strong> <strong>the</strong>polyimide P84 and Matrimid [96]. The lifetime <strong>of</strong> ortho-positronium annihilating in <strong>the</strong>pores (τ 2 ) indicates <strong>the</strong> size <strong>of</strong> <strong>the</strong> pores. The number or intensity <strong>of</strong> ortho-positroniumannihilating in <strong>the</strong> pores (I 2 ) indicates <strong>the</strong> number <strong>of</strong> pores or <strong>the</strong> porosityPolymer Precursor Pyrolysis Temperature (°C) τ 2 (ps) I 2 (%)P84 800 378 70Matrimid 800 381 68In <strong>the</strong> same work, Tin et al [96] have also <strong>use</strong>d wide angle x-ray diffraction (WAXD) todetermine <strong>the</strong> d-spacing <strong>of</strong> <strong>the</strong> NPC membrane. The d-spacing, which can be interpreted as<strong>the</strong> distance between <strong>carbon</strong> molecules or <strong>the</strong> pore size, is determined from WAXD byexposing <strong>the</strong> sample to a specific radiation wavelength (λ), measuring <strong>the</strong> angle <strong>of</strong> diffraction(θ) and relating this to <strong>the</strong> d-spacing (d) using Bragg’s Law as shown in Equation 2-28. Theterm n is an integer related to <strong>the</strong> order <strong>of</strong> reflection.n λ = 2d sinθEquation 2-28In a related study, Tin et al [97] <strong>use</strong> WAXD to specifically review <strong>the</strong> effect <strong>of</strong> pyrolysis<strong>temperature</strong> on <strong>the</strong> morphology <strong>of</strong> NPC <strong>membranes</strong> made from P84. This is a similar study tothat done by Steel and Koros [17] who also <strong>use</strong>d WAXD to characterise NPC <strong>membranes</strong>made from Matrimid at different pyrolysis <strong>temperature</strong>s. The values <strong>of</strong> <strong>the</strong> d-spacingdetermined <strong>for</strong> NPC <strong>membranes</strong> made from <strong>the</strong> different polyimide [17, 97] are presented inTable 2-5. With a pore size around 4 Å, <strong>the</strong>se results are in line with those determined usingadsorption as discussed above. For <strong>the</strong> <strong>high</strong>est pyrolysis <strong>temperature</strong> <strong>use</strong>d to make NPC<strong>membranes</strong> from P84 and Matrimid (800 °C) <strong>the</strong>re is a peak located greater than 40° 2θ or 2.1Å, which <strong>the</strong> authors <strong>of</strong> both papers attribute to <strong>the</strong> <strong>for</strong>mation <strong>of</strong> a more graphitic structure.32


Looking specifically at bulk <strong>nanoporous</strong> <strong>carbon</strong> manufactured from PFA, Powder x-raydiffraction was <strong>use</strong>d by Mariwala and Foley [71] to obtain d-spacing values at increasingpyrolysis <strong>temperature</strong>. These results are also included in Table 2-5 <strong>for</strong> comparison.Table 2-5 : WAXD data <strong>for</strong> NPC <strong>membranes</strong> manufactured from <strong>the</strong> pyrolysis <strong>of</strong> <strong>the</strong>polyimide Matrimid and P84 [17, 96, 97] along with powder x-ray diffraction data <strong>for</strong>bulk <strong>nanoporous</strong> <strong>carbon</strong> manufactured from PFA [71].Polymeric Precursor Pyrolysis Temperature (°C) d (Å)Polyimide – P84 ([97]) 550 4.2Polyimide – P84 ([96, 97]) 800 3.5Polyimide – Matrimid ([17]) 550 3.8Polyimide – Matrimid ([96]) 800 3.7PFA [71] 500 3.9 – 5.2PFA [71] 800 3.7 – 3.9PFA [71] 1000 3.633


2.4 Per<strong>for</strong>mance <strong>of</strong> NPC Membranes2.4.1 Supported NPC MembranesThe permeance <strong>of</strong> NPC <strong>membranes</strong> reported in <strong>the</strong> literature range from less than 1 x 10 -10 tomore than 1000 x 10 -10 mol/(m 2 .s.Pa) depending on <strong>the</strong> flowing gas molecule, polymer <strong>use</strong>dand conditions <strong>of</strong> manufacture. Similarly, <strong>the</strong>re is also a large range <strong>of</strong> reported selectivitiesfrom as low as 1 (no separation achieved) to over 200.Looking specifically at <strong>the</strong> gas molecule pair CO 2 /CH 4 , separation factors between 80 and 90have been reported <strong>for</strong> NPC <strong>membranes</strong> made from <strong>the</strong> pyrolysis <strong>of</strong> <strong>the</strong> polymer PFA byWang et al [19]. Sedigh et al [25] have reported a separation factor <strong>of</strong> 33 <strong>for</strong> a NPCmembrane made from <strong>the</strong> pyrolysis <strong>of</strong> PFA on an alumina tube. Centeno et al [15] have alsoreported a <strong>high</strong> separation <strong>of</strong> 87 <strong>for</strong> NPC <strong>membranes</strong> made from <strong>the</strong> pyrolysis <strong>of</strong> phenolicresin. Ogawa and Nakano [18] have made hollow fibre NPC <strong>membranes</strong> from <strong>the</strong> pyrolysis <strong>of</strong>polyamic acid and report a very <strong>high</strong> CO 2 /CH 4 separation factor <strong>of</strong> 181.Similarly, <strong>for</strong> <strong>the</strong> gas molecule pair <strong>of</strong> CO 2 /N 2 , Wang et al [19] reported separation factors <strong>of</strong>79 and 40 <strong>for</strong> a NPC <strong>membranes</strong> made from <strong>the</strong> pyrolysis <strong>of</strong> <strong>the</strong> polymer polyfurfuryl alcohol(PFA). This compares favourably against a zeolite membrane tested by Bernal et al [98] witha CO 2 /N 2 permselectivity <strong>of</strong> 13.7. Hayashi et al [99] have demonstrated that supported NPC<strong>membranes</strong> manufactured from <strong>the</strong> pyrolysis <strong>of</strong> polyimide have much greater CO 2permeabilities and CO 2 /N 2 permselectivities than polymer <strong>membranes</strong> made from <strong>the</strong> sameprecursor. The sorptivity and diffusivity through <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong> <strong>membranes</strong> wererespectively 2 to 3 and 300 to 1000 times larger than with <strong>the</strong> polyimide polymer <strong>membranes</strong>.As discussed earlier, <strong>the</strong> separation factor from mixed gases provides a more realisticindicator <strong>of</strong> membrane per<strong>for</strong>mance than permselectivity from pure gases. Variousresearchers [18, 25] have obtained permeance measures <strong>for</strong> CO 2 /CH 4 mixtures throughsupported NPC <strong>membranes</strong> and conclude that mixture selectivities are <strong>high</strong>er than idealselectivities due to <strong>the</strong> selective adsorption <strong>of</strong> CO 2 on <strong>the</strong> pore walls, which hinders <strong>the</strong>transport <strong>of</strong> CH 4 .Whilst <strong>the</strong> focus <strong>of</strong> this research is on <strong>the</strong> CO 2 /CH 4 and CO 2 /N 2 gas pairs, it should be notedthat also in <strong>the</strong> area <strong>of</strong> CO 2 separation and capture very promising selectivities have beenreported <strong>for</strong> <strong>the</strong> H 2 /CO 2 gas pair where CO 2 is retained on <strong>the</strong> retentate side [14, 20, 22, 25,29, 63, 64, 100-103] <strong>for</strong> molecular sieving <strong>membranes</strong>. However, <strong>for</strong> <strong>the</strong> SSF <strong>membranes</strong>where <strong>the</strong> pore size is larger and surface diffusion dominates <strong>the</strong> H 2 is actually retained on <strong>the</strong>retentate side [63].34


There is also research work in syn<strong>the</strong>sis gas production foc<strong>use</strong>d on H 2 separation via catalytic<strong>membranes</strong> where <strong>the</strong> catalyst is impregnated into <strong>the</strong> polymer prior to pyrolysis [79].2.4.2 Unsupported NPC MembranesThe per<strong>for</strong>mance <strong>of</strong> unsupported NPC <strong>membranes</strong> is generally reported as permeability as <strong>the</strong>thickness <strong>of</strong> <strong>the</strong> membrane is known. This makes it possible to directly compare <strong>the</strong>per<strong>for</strong>mance <strong>of</strong> unsupported NPC <strong>membranes</strong> with polymeric <strong>membranes</strong>.As discussed earlier in Section 2.2.2, Robeson’s <strong>the</strong>ory relates to <strong>the</strong> relationship betweenpermeability and selectivity. Based upon <strong>the</strong> published data on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> polymeric<strong>membranes</strong>, Robeson [38] developed relationships <strong>for</strong> <strong>the</strong> <strong>the</strong>oretical per<strong>for</strong>mance limits <strong>of</strong>polymeric <strong>membranes</strong> <strong>for</strong> particular gas pairs. The per<strong>for</strong>mance <strong>of</strong> unsupported NPC<strong>membranes</strong> made from <strong>the</strong> pyrolysis <strong>of</strong> polyimide [17] in relation to <strong>the</strong> Robeson upperbound <strong>for</strong> <strong>the</strong> CO 2 /CH 4 gas pair <strong>for</strong> polymeric <strong>membranes</strong> is presented in Figure 2-9.1000Target Per<strong>for</strong>manceHigh Selectivity and PermeabilitySelectivity (CO 2 /CH 4 )10010Robeson Upper BoundRaw Polyimide Polymer MembranePolyimide NPC Membrane11 10 100 1000 10000CO 2 Permeability (Barrer)Figure 2-9 : Per<strong>for</strong>mance <strong>of</strong> unsupported NPC <strong>membranes</strong> [17] with respect to <strong>the</strong>Robeson Upper Bound [38] <strong>for</strong> polymeric <strong>membranes</strong>, which shows that <strong>the</strong><strong>carbon</strong>isation <strong>of</strong> a Matrimid polymeric membrane increases <strong>the</strong> permeability andselectivity such that it breaks <strong>the</strong> upper bound.As can be seen from Figure 2-9, <strong>the</strong> pyrolysis <strong>of</strong> Matrimid to <strong>for</strong>m a NPC membrane [17]increases both <strong>the</strong> permeability and selectivity compared with <strong>the</strong> original Matrimidmembrane resulting in a shift to above <strong>the</strong> Robeson upper bound.35


2.5 Gas Transport Mechanisms <strong>of</strong> NPC MembranesAn ideal NPC membrane will have a pore size distribution that is very narrow and centred at asize just larger than that <strong>of</strong> <strong>the</strong> desired penetrant thus facilitating gas separation via molecularsieving. In this study, <strong>the</strong> desired penetrant is CO 2 , which has a kinetic diameter <strong>of</strong> 3.3 Å andas such a pore size distribution centred around 4-5 Å would result in reasonable molecularsieving ability <strong>for</strong> this molecule. However, in a real manufactured NPC membrane <strong>the</strong>re are<strong>of</strong>ten some pores present within <strong>the</strong> 5 – 10 Å range [84], which leads to <strong>the</strong> surface diffusiontransport mechanism also contributing <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> <strong>the</strong> NPC membrane.2.5.1 Experimental ObservationsThe research work <strong>of</strong> Fuertes [21] provides a good explanation <strong>of</strong> <strong>the</strong> transition betweenmolecular sieving and surface diffusion. Fuertes has oxidised NPC <strong>membranes</strong> produced from<strong>the</strong> pyrolysis <strong>of</strong> phenolic resin to increase <strong>the</strong> pore size <strong>of</strong> <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong> from 4 to 5– 7 Å. Comparing <strong>the</strong> permeances <strong>of</strong> gases through oxidised and non-oxidised <strong>membranes</strong>, itwas noted that in <strong>the</strong> non-oxidised <strong>membranes</strong> <strong>the</strong> permeance <strong>of</strong> a gas increases withdecreasing molecular size, which suggests a molecular sieving mechanism. However, in <strong>the</strong>oxidised <strong>membranes</strong> with <strong>the</strong> larger pore size, permeance was no longer correlated withmolecular size as <strong>the</strong> larger more polar molecules like hydro<strong>carbon</strong>s displayed <strong>high</strong>erpermeance.A molecular sieving trend has been observed by Wang et al [19] <strong>for</strong> a single NPC membranemade from <strong>the</strong> vapour deposition on PFA (with a narrow pore size range <strong>of</strong> 4- 5 Å) onalumina tubes. As shown in Table 2-6, <strong>the</strong> permeance decreases with increasing kineticdiameter.Table 2-6 : Permeance results <strong>of</strong> various gas molecules <strong>for</strong> <strong>membranes</strong> made from PFA[19]. The increase in permeance with decreasing molecular size is a clear indicator <strong>of</strong> amolecular sieving mechanism.Permeance (mol/m 2 /s/Pax10 10 ) @ 293 KH 2 (0.289 nm) CO 2 (0.33 nm) O 2 (0.346 nm) N 2 (0.364 nm) CH 4 (0.38 nm)225.00 58.20 7.75 0.73 0.63The permeance <strong>of</strong> low-adsorbing gases (such as N 2 or CH 4 ) will be constant <strong>for</strong> a molecularsieving membrane over a range <strong>of</strong> driving <strong>for</strong>ce pressure [15, 24]. However, <strong>for</strong> gases thatadsorb well (such as CO 2 ), <strong>the</strong> permeance will be pressure-dependent [104] with <strong>high</strong>estpermeances observed at low pressures.36


If, however, <strong>the</strong>re is insufficient coverage <strong>of</strong> <strong>carbon</strong> or cracking occurs, <strong>the</strong> gas moleculeswill proceed directly through <strong>the</strong> porous support via bulk flow or Knudsen diffusion. Todetermine <strong>the</strong> contributions <strong>of</strong> each mechanism to <strong>the</strong> overall per<strong>for</strong>mance requires a detailedunderstanding <strong>of</strong> <strong>the</strong> pore distribution along with <strong>the</strong> number <strong>of</strong> defects. Several researchers[12, 105] have developed a parallel resistance transport models to account <strong>for</strong> <strong>the</strong> presence <strong>of</strong>defect pores in NPC <strong>membranes</strong>.2.5.2 Molecular ModellingMolecular modelling or simulation can be <strong>use</strong>d to predict <strong>the</strong> adsorption and diffusion <strong>of</strong>gases through a substrate (e.g. <strong>nanoporous</strong> <strong>carbon</strong>). Molecular simulation is a very <strong>use</strong>fulcomputational tool <strong>for</strong> predicting per<strong>for</strong>mance and <strong>for</strong> understanding <strong>the</strong> transportmechanisms behind <strong>the</strong> per<strong>for</strong>mance observed experimentally.There are two methods <strong>of</strong> molecular simulation <strong>use</strong>d <strong>for</strong> predicting <strong>the</strong> adsorption anddiffusion <strong>of</strong> gases through <strong>nanoporous</strong> <strong>carbon</strong>. The first is Monte Carlo simulation which is<strong>use</strong>d <strong>for</strong> simulating equilibrium properties such as adsorption iso<strong>the</strong>rms [106-109]. Thesecond is molecular dynamic simulation which is <strong>use</strong>d <strong>for</strong> simulating <strong>the</strong> movement ordiffusion <strong>of</strong> molecules through <strong>the</strong> substrate. Equilibrium properties calculated from usingMonte Carlo simulation are also <strong>use</strong>d as an input to <strong>the</strong> molecular dynamic simulation [110,111].2.5.2.1 Monte Carlo SimulationThe Monte Carlo (MC) method is a stochastic (nondeterministic) process that <strong>use</strong>s repeatedrandom sampling to compute <strong>the</strong> results.Monte Carlo (MC) simulations, which like molecular dynamic simulations employ computersto solve <strong>the</strong> algorithms, <strong>use</strong> <strong>the</strong> MC method to calculate <strong>the</strong> equilibrium properties <strong>of</strong> asystem. This is achieved by sampling <strong>the</strong> phase space <strong>of</strong> an ensemble <strong>of</strong> state points andcomputing <strong>the</strong> averages <strong>of</strong> mechanical properties over <strong>the</strong> ensemble [112].Adsorption properties and <strong>the</strong>re<strong>for</strong>e adsorption iso<strong>the</strong>rms can be calculated using an ensemble<strong>of</strong> constant chemical potential (µ), volume (V) and <strong>temperature</strong> (T) known as <strong>the</strong> GrandCanonical (GC) ensemble.Each system within <strong>the</strong> GC ensemble is in equilibrium with <strong>the</strong> environment and as such <strong>the</strong>energy (E) and number <strong>of</strong> particles (N) <strong>of</strong> each system is allowed to fluctuate at constantµVT. The partition function (Ф), which describes <strong>the</strong> statistical properties <strong>of</strong> a system in37


equilibrium, <strong>for</strong> <strong>the</strong> GC ensemble is represented in Equation 2-29, where k B is <strong>the</strong> Boltzmannconstant and f is <strong>the</strong> fugacity related directly to <strong>the</strong> chemical potential via equation 2-30.iEi−kBTNiΦ ( f , V , T ) = ∑ f eEquation 2-29µ = k BT ln fEquation 2-30For each GCMC simulation at constant bulk pressure and <strong>temperature</strong>, <strong>the</strong> number <strong>of</strong>particles adsorbed from <strong>the</strong> bulk phase in <strong>the</strong> equilibrium state is determined from byper<strong>for</strong>ming trial moves <strong>of</strong> displacement, rotation, insertion and deletion. Each move isselected at random and has an associated probability function [45].The adsorption iso<strong>the</strong>rms are created by completing several such Grand Canonical MonteCarlo (GCMC) simulations at varying values <strong>of</strong> chemical potential to calculate <strong>the</strong> associatednumber <strong>of</strong> particles adsorbed from <strong>the</strong> bulk phase. The bulk phase pressure is related to <strong>the</strong>chemical potential via Equation 2-31 as shown below, where λ is <strong>the</strong> de Broglie wavelengthand φ is <strong>the</strong> fugacity coefficient <strong>of</strong> <strong>the</strong> fluid in <strong>the</strong> bulk phase [112].1 3 1µ = ln λ + ln( pφ)Equation 2-31k Tk TBBLikewise, adsorption iso<strong>the</strong>rms at different <strong>temperature</strong>s can be created by completingano<strong>the</strong>r family <strong>of</strong> simulations to determine <strong>the</strong> number <strong>of</strong> adsorbed particles at varying bulkpressure (chemical potential).Inputs into <strong>the</strong> GCMC simulation include <strong>the</strong> surface structure <strong>of</strong> <strong>the</strong> substrate (e.g.<strong>nanoporous</strong> <strong>carbon</strong>), <strong>the</strong> initial position <strong>of</strong> <strong>the</strong> gas molecules and potential models to describe<strong>the</strong> interactions within <strong>the</strong> gas molecules and between <strong>the</strong> gas molecules and <strong>the</strong> substrate.The Lennard Jones potential model is <strong>use</strong>d to provide <strong>the</strong> interaction potential (u) between <strong>the</strong>gas molecules and is described by Equation 2-32, where r ij is <strong>the</strong> separation distance betweensites i and j, ε i,j is <strong>the</strong> well depth and σ i,j is <strong>the</strong> site collision diameter [112].126⎡⎛σ⎤i,j⎞ ⎛ σi,j⎞u ⎢⎜⎟ − ⎜ ⎟ ⎥i, j( ri, j) = 4ε Equation 2-32⎢⎥⎣⎝ri, r ⎠ ⎝ ri, r ⎠ ⎦The Steele potentials are <strong>use</strong>d in <strong>the</strong> simulation to provide <strong>the</strong> interaction potential between<strong>the</strong> gas and <strong>carbon</strong> molecules and are based upon adsorption on planar graphite sheets [113].38


Ab-initio potentials can be <strong>use</strong>d to provide <strong>the</strong> interaction potential between gas-<strong>carbon</strong>molecules, which incorporate <strong>the</strong> presence <strong>of</strong> <strong>the</strong> non-hexagonal rings <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong>[113].2.5.2.2 Molecular Dynamics SimulationMolecular dynamic (MD) simulation <strong>use</strong>s numerical methods to calculate <strong>the</strong> properties <strong>of</strong> asystem over a period <strong>of</strong> time. Numerically, MD simulation <strong>use</strong>s an algorithm to integrateNewton’s equations <strong>of</strong> motion from initial positions and velocities to create moleculartrajectories, which can <strong>the</strong>n be analysed to obtain equilibrium and time-dependent properties[112].For <strong>the</strong> study <strong>of</strong> gas adsorption and diffusion through porous substrates, MD simulations arecarried out in <strong>the</strong> Grand Canonical ensemble <strong>of</strong> constant number <strong>of</strong> particles, volume and<strong>temperature</strong> (NVT). Newton’s equations <strong>of</strong> motions are integrated using <strong>the</strong> Verlet velocityalgorithm. The Nose-Hoover <strong>the</strong>rmostat is <strong>use</strong>d to maintain a constant <strong>temperature</strong>[45].The sum <strong>of</strong> kinetic and potential energy functions <strong>of</strong> <strong>the</strong> system is known as <strong>the</strong> Hamiltonian(H). The Hamiltonian <strong>of</strong> a given system is constant and this value can be <strong>use</strong>d to check <strong>the</strong>accuracy <strong>of</strong> MD calculations.To obtain <strong>the</strong> diffusivity <strong>of</strong> a gas particle through <strong>the</strong> porous substrate, <strong>the</strong> Einstein relation is<strong>use</strong>d. The Einstein relation, as shown in Equation 2-33, relates <strong>the</strong> time (t) limit <strong>of</strong> <strong>the</strong> meansquared displacement <strong>of</strong> particles (N) at position r to <strong>the</strong> component diffusivity (D). Thedimensionality <strong>of</strong> <strong>the</strong> system is denoted d [45].1NN2lim ∑[ ri( t)− ri(0)] = 2dDtt→∞i=1Equation 2-33The potential energy models employed to describe <strong>the</strong> interactions within molecules andbetween <strong>the</strong> molecule and <strong>the</strong> substrate are <strong>the</strong> same as discussed <strong>for</strong> GCMC simulation in <strong>the</strong>previous section.The combination <strong>of</strong> <strong>the</strong> adsorption results from GCMC simulation and component diffusivityresults from MD simulation can be to calculate <strong>the</strong> predicted flux and ideal selectivity at agiven <strong>temperature</strong> and pressure [45]. A simplified summary <strong>of</strong> <strong>the</strong> GCMC and MDSimulation processes is shown in Figure 2-10.39


Substrate structure / coordinatesInitial positions and velocities <strong>of</strong>gas moleculesGrand CanonicalMonte CarloSimulationAdsorption propertiesPotential model <strong>for</strong> describinginteractions between gas molecules(e.g. Lennard Jones, Steele or Ab-Initio)• Stochastic / Random• System <strong>of</strong> constant µVT• Equilibrium statedetermined from trial movesEquilibrium propertiesFlux &IdealSelectivitySubstrate structure / coordinatesInitial positions and velocities <strong>of</strong>gas moleculesPotential model <strong>for</strong> describinginteractions between gas molecules(e.g. Lennard Jones, Steel or Ab-Initio)Molecular DynamicSimulation• Numerical• System <strong>of</strong> constant NVT• Solving Newton’s Equations<strong>of</strong> Motion (Nose-HooverThermostat) using <strong>the</strong> VerletAlgorithm <strong>for</strong> obtainingtrajectories• Einstein relation <strong>for</strong> obtainself-diffusivitySelf-diffusivitiesPhenomenologicalcoefficients <strong>for</strong> mixturesFigure 2-10 : The process that is <strong>use</strong>d by molecular simulation to create flux andseparation factors <strong>for</strong> <strong>nanoporous</strong> <strong>carbon</strong>. The two simulation methods <strong>use</strong>d are (1)Grand Canonical Monte Carlo (GCMC) simulation <strong>for</strong> determining <strong>the</strong> adsorptionproperties and (2) molecular dynamic (MD) simulation <strong>for</strong> determining <strong>the</strong> selfdiffusivities.Combining <strong>the</strong> results from both methods produces a predicted flux andseparation factor.2.5.2.3 Published Simulation Studies <strong>of</strong> NPCNanoporous <strong>carbon</strong> is commonly modelled using <strong>the</strong> coordinates <strong>of</strong> C 168 Schwarzite(presented in Figure 2-11), which is a very similar curved-surface <strong>carbon</strong> compound.Figure 2-11 : Structure <strong>of</strong> C168 Schwarzite <strong>use</strong>d to represent <strong>nanoporous</strong> <strong>carbon</strong> [106].40


Klauda et al [113] have specifically reviewed <strong>the</strong> effect <strong>of</strong> <strong>the</strong> curved <strong>carbon</strong> surface on gasadsorption using molecular modelling <strong>of</strong> planar graphite, curved C 60 Fullerene and curvedC 168 Schwarzite. The interaction potential between nitrogen and <strong>the</strong> C 60 and C 168 surfaceswere found to be different than those <strong>for</strong> <strong>the</strong> planar graphite, resulting in <strong>high</strong>er adsorptionenergies with <strong>the</strong> curved surfaces. However, <strong>the</strong> curved surface only had a small effect on <strong>the</strong>interaction potential between oxygen and <strong>the</strong> C 60 and C 168 surfaces. The authors suggest thatthis difference in adsorption energies on <strong>the</strong> curved <strong>carbon</strong> surface may explain <strong>the</strong> <strong>high</strong>O 2 /N 2 selectivity observed experimentally in NPC <strong>membranes</strong> by Shiflett and Foley [75, 83].Using grand canonical Monte Carlo (GCMC) simulation, Jiang et al [106] have modelled <strong>the</strong>adsorption <strong>of</strong> oxygen (O 2 ) and nitrogen (N 2 ) in C 168 Schwarzite. At low pressure <strong>the</strong>adsorption <strong>of</strong> N 2 is favoured but as <strong>the</strong> pressure is increased <strong>the</strong> adsorption <strong>of</strong> O 2 dominatesdue to its smaller size. The simulation iso<strong>the</strong>rms are fitted well to <strong>the</strong> two-site Langmuirmodel. In a later publication, Jiang and Sandler [107] have examined <strong>the</strong> mixed gascompetitive adsorption <strong>of</strong> O 2 and N 2 using both GCMC and Gibbs ensemble Monte Carlosimulation. At pressures above 400 kPa and at 27 °C, <strong>the</strong> O 2 /N 2 selectivity increases above 1to 1.2 at 10,000 kPa. However, increasing <strong>the</strong> <strong>temperature</strong> to 127 °C <strong>the</strong>n reduces <strong>the</strong>selectivity back to 1.Arora and Sandler [110, 111] argue that <strong>the</strong> adsorption O 2 /N 2 selectivity <strong>of</strong> <strong>the</strong> curved C 168Schwarzite surface predicted by molecular modelling can not solely explain <strong>the</strong> experimentalresults observed by Shiflett and Foley [75, 83] and it is necessary to understand <strong>the</strong> diffusion<strong>of</strong> <strong>the</strong>se gases through <strong>nanoporous</strong> <strong>carbon</strong>. In <strong>the</strong> initial publication, <strong>the</strong> authors [110]conclude that using <strong>the</strong> ab initio potential <strong>for</strong> describing <strong>the</strong> interaction between <strong>the</strong> gasmolecules and <strong>the</strong> <strong>carbon</strong> surface better explains <strong>the</strong> <strong>high</strong> selectivity seen experimentally. Tomodel <strong>the</strong> pore size <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong>, <strong>the</strong> authors [111] subsequently per<strong>for</strong>medsimulations using <strong>carbon</strong> nanotubes with constrictions that represented <strong>the</strong> pore size <strong>of</strong><strong>nanoporous</strong> <strong>carbon</strong>, which allow molecular sieving. If <strong>the</strong> correct constriction is present, O 2will pass through <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong> much more readily than N 2 resulting in a muchgreater overall O 2 /N 2 selectivity.With <strong>the</strong> purpose <strong>of</strong> reviewing <strong>the</strong> suitability <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong>s <strong>for</strong> separating CO 2 fromN 2 in a typical power station flue gas, Jiang and Sandler [108] have <strong>use</strong>d a combination <strong>of</strong>quantum mechanics (to obtain <strong>the</strong> ab initio potential) and GCMC (to predict <strong>the</strong> adsorption).The authors also explain that <strong>the</strong> ab initio potential provides a more accurate adsorbateadsorbentinteraction potential <strong>for</strong> <strong>nanoporous</strong> <strong>carbon</strong>, with an adsorption selectivity <strong>of</strong>CO 2 /N 2 around 160 being observed <strong>for</strong> a binary mixture (21:79 mol& CO 2 :N 2 ) at ~ 100 kPa.This work has only considered <strong>the</strong> effect <strong>of</strong> adsorption <strong>for</strong> separating CO 2 and N 2 . If using41


NPC <strong>membranes</strong>, <strong>the</strong> diffusion effect on selectivity must also be considered. Based on <strong>the</strong>observations <strong>of</strong> Arora and Sandler [111] <strong>for</strong> O 2 /N 2 selectivity, it can be suggested that <strong>the</strong>selectivity based on <strong>the</strong> diffusion <strong>of</strong> CO 2 and N 2 provided that correct pore size is present willbe larger due to <strong>the</strong> even greater difference in molecule size. As such it is expected that <strong>the</strong>CO 2 /N 2 selectivity <strong>of</strong> NPC <strong>membranes</strong> would be much larger than that <strong>for</strong> O 2 /N 2 beca<strong>use</strong> <strong>of</strong><strong>the</strong> greater difference in both <strong>the</strong> adsorption and molecule size.Molecular simulations <strong>of</strong> <strong>carbon</strong> nanotubes have been undertaken by various researchers[109, 114-122] and considering that <strong>nanoporous</strong> <strong>carbon</strong> and nanotubes have a similarstructure, <strong>the</strong>se simulation results are also <strong>of</strong> interest. Arora et al [118] have <strong>use</strong>d GCMC andmolecular dynamic simulations to analyse <strong>the</strong> adsorptive and diffusive behaviour <strong>of</strong> N 2 insingle wall <strong>carbon</strong> nanotubes <strong>of</strong> varying diameter. It was found that as <strong>the</strong> diameter <strong>of</strong> <strong>the</strong>nanotube was decreased from 15.7 to 8.6 Å, <strong>the</strong> adsorption <strong>of</strong> N 2 was increased. Similarly, <strong>the</strong>self-diffusivity <strong>of</strong> N 2 increased with decreasing nanotube diameter due to <strong>the</strong> localisation <strong>of</strong><strong>the</strong> N 2 molecules in <strong>the</strong> core <strong>of</strong> <strong>the</strong> nanotube ra<strong>the</strong>r than in bilayers as is seen with nanotubeswith <strong>high</strong>er diameters. However, when considering <strong>the</strong> adsorption <strong>of</strong> a N 2 /O 2 (air) mixture on<strong>carbon</strong> nanotubes, it was discovered through simulation [109, 117] that O 2 adsorptiondisplaces N 2 with increasing pressure beca<strong>use</strong> <strong>of</strong> O 2 being a smaller molecule and being ableto reach <strong>the</strong> smaller cavities.In a recent study, Skoulidas et al [121] have <strong>use</strong>d molecular modelling to understand <strong>the</strong>adsorption and diffusion <strong>of</strong> CO 2 and N 2 in <strong>carbon</strong> nanotubes at room <strong>temperature</strong> as afunction <strong>of</strong> nanotube diameter. Whilst <strong>the</strong> authors have not given separation factors in <strong>the</strong>irpublication, <strong>the</strong>y have observed that <strong>the</strong> diffusion mechanism is not Knudsen and shouldresult in a large CO 2 /N 2 selectivity.In summary, molecular simulations <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong>s reveal <strong>the</strong>oretically <strong>high</strong>adsorptive selectivity due to <strong>the</strong> curved structure <strong>of</strong> <strong>the</strong> <strong>carbon</strong> compared with planar graphitealong with <strong>high</strong> <strong>the</strong>oretical diffusive selectivity provided <strong>the</strong> appropriate sized porerestrictions are present to allow separation via molecular sieving. These results are thus verypromising <strong>for</strong> considering <strong>nanoporous</strong> <strong>carbon</strong>s as a new material <strong>for</strong> gas separation.42


2.6 High Temperature Operation <strong>of</strong> NPC MembranesGiven <strong>the</strong> <strong>high</strong> <strong>temperature</strong>s at which NPC <strong>membranes</strong> are manufactured, <strong>the</strong>y are ideal <strong>for</strong>operation above ambient <strong>temperature</strong> and indeed up to <strong>the</strong> original pyrolysis <strong>temperature</strong>.High <strong>temperature</strong> operation is advantageous <strong>for</strong> CO 2 separation from natural gas or syn<strong>the</strong>sisgas production as it eliminates <strong>the</strong> need <strong>for</strong> pre-cooling, which is required when usingpolymeric <strong>membranes</strong>. There<strong>for</strong>e, understanding how <strong>high</strong> <strong>temperature</strong> operation affects <strong>the</strong>per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> is pertinent.2.6.1 Arrhenius RelationshipsThe relationship between permeance (P’) and operating <strong>temperature</strong> (T) <strong>of</strong> an NPC membranecan be described by an Arrhenius-type relationship as given in Equation 2-34 [22].⎡− ( Ep) ⎤P' = P'0exp⎢⎥Equation 2-34⎣ RT ⎦Rearranging Equation 2-34 to a linear <strong>for</strong>m results in Equation 2-35 shown below.− Ep 1Ln( P')= + Ln(P'0)Equation 2-35R TPlotting Ln (P’) against 1/T reveals a straight line from which <strong>the</strong> permeance activationenergy (E p ) can be calculated from <strong>the</strong> gradient. Examples <strong>of</strong> activation energies calculated<strong>for</strong> NPC Membranes made from polyamic acid [26] are presented in Table 2.7. Relating thisback to <strong>the</strong> experimental data, as <strong>the</strong> operating <strong>temperature</strong> increases, <strong>the</strong> permeance <strong>of</strong>methane will increase at <strong>the</strong> faster rate than N 2 , O 2 or CO 2 as it has <strong>the</strong> <strong>high</strong>est value <strong>of</strong> E p .Table 2-7 : Activation energy <strong>of</strong> permeance <strong>of</strong> various gas molecules <strong>for</strong> NPC<strong>membranes</strong> made from polyamic acid. The activation energy <strong>for</strong> CO 2 is 0 indicating that<strong>the</strong> CO 2 permeance will remain constant with increasing <strong>temperature</strong>.Activation Energy <strong>of</strong> Permeance (E p ) (kJ/mol)NPC MembraneMade fromCO 2 (0.33 nm) O 2 (0.346 nm) N 2 (0.364 nm) CH 4 (0.38 nm)Polyamic Acid [26] 0 3.1 9.8 14.5Looking now in more detail at <strong>the</strong> effect <strong>of</strong> <strong>temperature</strong> on <strong>the</strong> transport mechanisms, <strong>the</strong>permeance <strong>of</strong> a gas (A) via surface diffusion (P’ S ) or molecular sieving (P’ M ) are bothactivated by <strong>temperature</strong>s as presented below in Equations 2-36 and 2-37 [8, 20, 23]. The43


diffusion coefficient <strong>of</strong> surface diffusion is represented by D 0S and <strong>the</strong> diffusion coefficient <strong>of</strong>molecular sieving is represented by D 0M .D0S( A)⎡−( ES( A)− ∆Hads(A)) ⎤P's( A)= exp⎢⎥Equation 2-36RT∆x⎣ RT ⎦P'MD0M( A)⎡−( ED(A)− ∆Hads(A)) ⎤( A)= exp⎢⎥Equation 2-37RT∆x⎣ RT ⎦In <strong>the</strong> case <strong>of</strong> surface diffusion, whe<strong>the</strong>r <strong>the</strong> permeance increases or decreases with<strong>temperature</strong> depends on <strong>the</strong> relative values <strong>of</strong> <strong>the</strong> activation energy <strong>for</strong> diffusion (E S(A) ) andadsorption (∆H ads ). If ∆H ads > E S(A) , <strong>the</strong>n <strong>the</strong> permeance will decrease with increasing<strong>temperature</strong> as is expected <strong>for</strong> strongly adsorbing gases. If ∆H ads < E S(A) <strong>the</strong>n diffusion isdominant and <strong>the</strong> permeance will increase with increasing <strong>temperature</strong>.In <strong>the</strong> case <strong>of</strong> molecular sieving, it is assumed that adsorption is a very small contributor andas such E D(A) >>∆H ads resulting in an increase in permeance with increasing <strong>temperature</strong>.To simplify <strong>the</strong> situation when <strong>the</strong> relative contributions <strong>of</strong> molecular sieving and surfacediffusion are unknown, Suda and Haraya [22] developed a relationship <strong>for</strong> <strong>the</strong> permeanceactivation energy using <strong>the</strong> activation energy <strong>of</strong> diffusion (E d ) to approximate molecularsieving and <strong>the</strong> heat <strong>of</strong> adsorption (∆H ads ) to approximate surface diffusion (Equation 2-38).Ep= E − ∆HEquation 2-38dadsPublished values <strong>for</strong> <strong>the</strong> activation energies <strong>of</strong> diffusion (E d ) and heats <strong>of</strong> adsorption (∆H ads )<strong>for</strong> NPC <strong>membranes</strong> made from PFA and Cellulose are presented in Table 2-8.Table 2-8 : Activation energies <strong>of</strong> diffusion and heats <strong>of</strong> adsorption <strong>of</strong> various gasmolecules <strong>for</strong> NPC <strong>membranes</strong> made from PFA and Cellulose.Activation Energy <strong>of</strong> Diffusion (E d ) (kJ/mol)NPC MembraneMade fromCO 2 (0.33 nm) O 2 (0.346 nm) N 2 (0.364 nm) CH 4 (0.38 nm)PFA [43] 9.2 17.3 24.7 Data N/ACellulose [12] 23.4 21.6 26.5 Data N/AHeat <strong>of</strong> Adsorption (∆H ads ) (kJ/mol)NPC MembraneMade fromCO 2 (0.33 nm) O 2 (0.346 nm) N 2 (0.364 nm) CH 4 (0.38 nm)PFA [43] 10.6 9.3 10.2 Data N/ACellulose [12] 29.6 20.8 10.5 Data N/A44


The activation energies <strong>for</strong> diffusion are positive implying that diffusion increases withincreasing <strong>temperature</strong>, whereas <strong>the</strong> activation energies <strong>for</strong> adsorption are >0 implying thatadsorption increases with decreasing <strong>temperature</strong> as per Equation 2-38. For both <strong>the</strong> NPC<strong>membranes</strong> made from PFA and Cellulose, <strong>the</strong> activation energy <strong>of</strong> diffusion increases withincreasing molecular size suggesting molecular sieving. Additionally, <strong>the</strong> heat <strong>of</strong> adsorption isgreatest <strong>for</strong> <strong>the</strong> most polar molecule, CO 2 , suggesting surface diffusion also occurs.The expected activation <strong>for</strong> permeance can <strong>the</strong>n be obtained from adding <strong>the</strong> two activationenergies <strong>of</strong> diffusion and adsorption as shown in Table 2-9. Like <strong>the</strong> data presented earlier inTable 2-7, <strong>the</strong> permeance activation energy <strong>for</strong> CO 2 is negative, which indicates that <strong>the</strong> value<strong>of</strong> <strong>the</strong> CO 2 permeance will decrease with increasing <strong>temperature</strong>. This is due to <strong>the</strong> increase indiffusion being smaller than <strong>the</strong> decrease in adsorption with increasing <strong>temperature</strong>.Table 2-9 : Activation energy <strong>of</strong> permeance <strong>of</strong> various gas molecules <strong>for</strong> NPC<strong>membranes</strong> made from PFA and Cellulose obtained by <strong>the</strong> addition <strong>of</strong> <strong>the</strong> diffusion andactivation energies presented in Table 2-8. Interestingly, <strong>for</strong> CO 2 <strong>the</strong> activation energy<strong>of</strong> permeance is negative beca<strong>use</strong> adsorption is such a large contributor to <strong>the</strong>permeance.Activation Energy <strong>of</strong> Permeance (E p = E d + ∆H ads ) (kJ/mol)NPC MembraneMade fromCO 2 (0.33 nm) O 2 (0.346 nm) N 2 (0.364 nm) CH 4 (0.38 nm)PFA [43] -1.4 8 14.5 Data N/ACellulose [12] -6.2 0.8 16.0 Data N/AStrano and Foley [24] argue that at low operating <strong>temperature</strong>s adsorption dominates while at<strong>high</strong> <strong>temperature</strong>s gas diffusion dominates. It is suggested that as <strong>the</strong> <strong>temperature</strong> increases,gas diffusion will approach Knudsen diffusion and very little separation will occur. Gilronand S<strong>of</strong>fer [8] also suggest that at <strong>high</strong>er operating <strong>temperature</strong>s <strong>the</strong> permeability andselectivity observed is ca<strong>use</strong>d by Knudsen diffusion despite <strong>the</strong> average pore size <strong>of</strong> <strong>the</strong> NPC<strong>membranes</strong> being within <strong>the</strong> range <strong>for</strong> molecular sieving (less than 5 Å). Knudsen diffusiondoes not follow an Arrhenius relationship ra<strong>the</strong>r <strong>the</strong> permeance is related to operating<strong>temperature</strong> through a square root relationship as presented earlier in Equation 2-17. In thiscase, permeance increases as <strong>temperature</strong> increases.In ano<strong>the</strong>r experimental and modelling study, Sedigh et al [25] have examined <strong>the</strong> effect <strong>of</strong>operating <strong>temperature</strong> on tubular NPC <strong>membranes</strong> made from PFA using mixed gas feeds.For a binary feed mixture <strong>of</strong> CO 2 /CH 4 , <strong>the</strong> selectivity decreases from 33 at 22°C to 5 at140°C. This reduction in selectivity is concluded to be due to <strong>the</strong> increase in CH 4 diffusion45


and <strong>the</strong>re<strong>for</strong>e CH 4 permeance and <strong>the</strong> simultaneous decrease in CO 2 permeance due to <strong>the</strong>loss <strong>of</strong> CO 2 adsorption. Molecular simulations <strong>of</strong> <strong>the</strong> adsorption <strong>of</strong> a binary gas mixture inslit-like <strong>carbon</strong> nanopores have shown that CO 2 is adsorbed more strongly than CH 4 onto <strong>the</strong>pore wall and that adsorption decreases with increasing <strong>temperature</strong>.Fuertes and Centeno have looked specifically at <strong>the</strong> transport <strong>of</strong> gases through NPC<strong>membranes</strong> produced from <strong>the</strong> pyrolysis <strong>of</strong> phenolic resin [15] and polyamic acid [26].Per<strong>for</strong>mance measurements were taken <strong>for</strong> both <strong>membranes</strong> at <strong>temperature</strong>s up to 150 °C andit was found that <strong>the</strong> permeance <strong>of</strong> CH 4 increases with <strong>temperature</strong> suggesting activated gasdiffusion. However, <strong>the</strong> permeance <strong>of</strong> CO 2 does not change with <strong>temperature</strong>. It is againsuggested that this is beca<strong>use</strong> <strong>of</strong> <strong>the</strong> adsorption <strong>of</strong> CO 2 occurring in <strong>the</strong> micropores. As <strong>the</strong><strong>temperature</strong> increases, <strong>the</strong> decrease in <strong>the</strong> adsorption <strong>of</strong> CO 2 is compensated by <strong>the</strong> increase<strong>of</strong> gas diffusion <strong>of</strong> CO 2 resulting in no net change in <strong>the</strong> permeance.In a related study, Fuertes [67] has reviewed <strong>the</strong> separation <strong>of</strong> a non-adsorbable component(N 2 ) from a strongly adsorbable component (n-Butane) using NPC <strong>membranes</strong> also madefrom <strong>the</strong> pyrolysis <strong>of</strong> phenolic resin. For <strong>the</strong> n-Butane/N 2 separation, increasing <strong>the</strong><strong>temperature</strong> from 25 to 150°C reduces <strong>the</strong> selectivity from 65 to 5 beca<strong>use</strong> <strong>of</strong> <strong>the</strong> <strong>high</strong>contribution <strong>of</strong> adsorption to <strong>the</strong> permeance at <strong>the</strong> lower <strong>temperature</strong>s <strong>of</strong> <strong>the</strong> more adsorbablecomponent. In <strong>the</strong> same paper, Fuertes also reviews <strong>the</strong> effect <strong>of</strong> operating <strong>temperature</strong> on <strong>the</strong>separation <strong>of</strong> <strong>the</strong> two non-adsorbable components, N 2 /CH 4 , and shows that <strong>for</strong> <strong>the</strong> same<strong>temperature</strong> range, <strong>the</strong> selectivity reduces from 3 to 2. It is concluded that aside frommolecular sieving by size, <strong>the</strong> relative adsorption <strong>of</strong> components onto NPC <strong>membranes</strong> is asignificant contributor to selectivity at low operating <strong>temperature</strong> whilst not at <strong>high</strong> operating<strong>temperature</strong>s.46


2.7 Effect <strong>of</strong> Impurities on <strong>the</strong> Per<strong>for</strong>mance NPC MembranesFor <strong>the</strong> commercial applications <strong>of</strong> NPC <strong>membranes</strong> being considered as part <strong>of</strong> this researchwork additional minor components are also present in <strong>the</strong> feed stream.In <strong>the</strong> case <strong>of</strong> CO 2 removal from natural gas, saturated amounts <strong>of</strong> water, heavier andaromatic hydro<strong>carbon</strong>s along with trace amounts <strong>of</strong> hydrogen sulphide (H 2 S) are present.In <strong>the</strong> case <strong>of</strong> CO 2 removal from air-blown syn<strong>the</strong>sis gas <strong>for</strong> hydrogen or electricityproduction, trace amounts <strong>of</strong> H 2 S and water are present along with significant quantities <strong>of</strong><strong>carbon</strong> monoxide (CO). Refer to <strong>the</strong> process details presented earlier in Table 1-1 Section 1.3.Assessing <strong>the</strong> robustness <strong>of</strong> NPC <strong>membranes</strong> in <strong>the</strong> presence <strong>of</strong> <strong>the</strong>se impurities is critical inestablishing <strong>the</strong> suitability <strong>of</strong> NPC <strong>membranes</strong> <strong>for</strong> <strong>the</strong> a<strong>for</strong>ementioned commercialapplications. For <strong>the</strong> purpose <strong>of</strong> reviewing <strong>the</strong> available literature, <strong>the</strong> “impurities” have beengrouped as (a) water, (b) condensable impurities (heavier and aromatic hydro<strong>carbon</strong>s) and (c)non-condensable impurities (H 2 S and CO).An additional discussion on <strong>the</strong> published effects <strong>of</strong> exposing <strong>the</strong> NPC <strong>membranes</strong> to air isalso presented after <strong>the</strong> discussion on water. Whilst no air or oxygen is present in applicationsinvestigated in this research, <strong>the</strong> effect <strong>of</strong> air exposure on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC<strong>membranes</strong> is relevant to <strong>the</strong> storage <strong>of</strong> <strong>the</strong> <strong>membranes</strong>.2.7.1 WaterWater is in equilibrium with natural gas when produced from a reservoir. Similarly, waterwill be present in syn<strong>the</strong>sis gas production beca<strong>use</strong> <strong>of</strong> <strong>the</strong> water gas shift reaction.Due to <strong>the</strong> problems if operating equipment (such as polymeric <strong>membranes</strong>, compressors andtransmission pipelines) with a humid feed, gas streams are dehydrated using establishedtechnologies such as glycol dehydration or a solid desiccant [123]. If <strong>the</strong> water tolerance <strong>of</strong><strong>the</strong> equipment (such as <strong>membranes</strong>) can be increased by using <strong>nanoporous</strong> <strong>carbon</strong> instead <strong>of</strong>polyimide, this will decrease <strong>the</strong> amount <strong>of</strong> pre-dehydration required and <strong>the</strong>re<strong>for</strong>e <strong>the</strong> cost.To specifically address <strong>the</strong> impact <strong>of</strong> water vapour on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong>,Jones and Koros [31] have completed a comprehensive study using humidified feeds. Hollowfibre polyimide <strong>membranes</strong> made at 500 to 550°C were tested using a mixed oxygen andnitrogen feed at relative humidities <strong>of</strong> 23, 44 and 85 %. The flux through <strong>the</strong> membrane wasdecreased dramatically by 50, 60 <strong>the</strong>n 70 % with increases in <strong>the</strong> relatively humidity.Interestingly, <strong>the</strong> impact on <strong>the</strong> O 2 /N 2 selectivity was negligible at relative humidities <strong>of</strong> 2347


and 44 % but decreased by 20 % in <strong>the</strong> case <strong>of</strong> 85 % humidity. The authors suggest that <strong>the</strong>initial phase <strong>of</strong> water sorption is rapid upon <strong>the</strong> surface and is <strong>the</strong>n followed by a <strong>the</strong>n slowerpore filling. Gawrys et al [124] also concluded that at low relative humidities water is weaklyadsorbed to <strong>the</strong> polar sites on <strong>carbon</strong> molecular sieves whilst at <strong>high</strong>er relative humiditiesmicropore filling will prevail resulting in a greater uptake <strong>of</strong> water.Both Groszek [125] and Avraham et al [126] have investigated <strong>the</strong> impact <strong>of</strong> gas streamcomposition on <strong>the</strong> adsorption <strong>of</strong> water onto <strong>carbon</strong> molecular sieves. From gas adsorptionexperiments, Groszek [125] concludes that water adsorption is much <strong>high</strong>er from ahumidified methane stream compared with a humidified nitrogen stream due to <strong>the</strong> <strong>for</strong>mation<strong>of</strong> methane-hydrates which occur in <strong>the</strong> micropores. Avraham et al [126] have studiedoxygen/nitrogen adsorption in <strong>the</strong> presence <strong>of</strong> water. At low relative humidities, <strong>the</strong>adsorption kinetics <strong>for</strong> oxygen and nitrogen are faster than water. As <strong>the</strong> humidity isincreased, water fills <strong>the</strong> pores displacing first <strong>the</strong> oxygen and <strong>the</strong>n <strong>the</strong> nitrogen leading to asignificant reduction in <strong>the</strong> adsorption <strong>of</strong> <strong>the</strong>se two gases.In a more recent study upon exposure to a 32.5 % relative humidity feed, Lagorsse et al [127]report a 50 % reduction in <strong>the</strong> permeance <strong>of</strong> NPC <strong>membranes</strong> at 29°C. To fur<strong>the</strong>r understandthis result, <strong>the</strong> authors have obtained fractional uptake curves <strong>for</strong> water, N 2 and CO 2 on <strong>the</strong>membrane at 29 °C. The uptake curves show that whilst <strong>the</strong> <strong>carbon</strong> has a slower uptake <strong>of</strong>water than CO 2 but not N 2 , <strong>the</strong> total amount <strong>of</strong> water absorbed is greater than CO 2 and N 2 .Interestingly, it is concluded that effect <strong>of</strong> humidity exposure should be viewed as a stronglyadsorbing component diffusing slowly through <strong>the</strong> <strong>carbon</strong> ra<strong>the</strong>r than a pore-blockingmechanism as such. This result carries on from an earlier publication where Lagorsse et al[128] conducted a detailed study into <strong>the</strong> nature <strong>of</strong> <strong>the</strong> adsorption iso<strong>the</strong>rms <strong>of</strong> water in NPC<strong>membranes</strong>. In this study, <strong>the</strong> authors developed a model which concludes that water adsorbsinitially on <strong>the</strong> hydrophilic oxygen functional groups, which allows a cluster <strong>of</strong> watermolecules to grow via hydrogen bonding. When <strong>the</strong> number <strong>of</strong> water molecules in <strong>the</strong> clusterreaches a critical size and has enough dispersion energy to be released it is adsorbed directlyonto <strong>the</strong> <strong>carbon</strong> surface.Jones and Koros [31] suggest that a reduction in water adsorption onto a NPC <strong>membranes</strong> canbe achieved by changing <strong>the</strong> feed flow configuration <strong>of</strong> <strong>the</strong> membrane such that <strong>the</strong> feed isfirstly exposed to <strong>the</strong> more porous support. By doing this a hydrophobic guard layer isproduced in <strong>the</strong> porous support resulting in a less dramatic effect on membrane per<strong>for</strong>mance.In a follow-up publication, Jones and Koros [129] have fur<strong>the</strong>r improved <strong>the</strong> stability <strong>of</strong> NPC<strong>membranes</strong> in <strong>the</strong> presence <strong>of</strong> water by coating <strong>the</strong> finished membrane with a very thin layer48


<strong>of</strong> unique polymeric materials which are hydrophobic but do not impact upon <strong>the</strong> permeance<strong>of</strong> o<strong>the</strong>r molecules. The unique polymeric materials were DuPont’s Teflon products AF1600and AF2400 (perfluro-2,2,-dimethyl-1,3-dioxole and tetrafluoroethylene, respectively).2.7.2 AirIn <strong>the</strong> late 1960s and early 1970s, Mattson et al published a series <strong>of</strong> papers exploring <strong>the</strong>surface chemistry <strong>of</strong> <strong>carbon</strong> materials in <strong>the</strong> presence <strong>of</strong> air at room <strong>temperature</strong> [130-132].The authors <strong>use</strong> an infrared spectrophotometer to detect <strong>the</strong> presence <strong>of</strong> functional groups onactivated <strong>carbon</strong>s that have been exposed to air and conclude that <strong>carbon</strong>yl (C=O) functionalgroups are present on <strong>the</strong> surface. The hydrophilic nature <strong>of</strong> <strong>the</strong>se <strong>carbon</strong>yl groups providessites <strong>for</strong> water adsorption, which blocks <strong>the</strong> pores <strong>of</strong> <strong>the</strong> <strong>carbon</strong> as discussed previously inSection 2.7.1. More recently, Petit and Bahaddi [133] demonstrated that <strong>the</strong> uptake <strong>of</strong> water isgreater on oxidised activated <strong>carbon</strong> than on non-oxidised <strong>carbon</strong> confirming <strong>the</strong> importance<strong>of</strong> oxygen in creating a hydrophilic surface.At <strong>high</strong>er <strong>temperature</strong>s (600°C), <strong>carbon</strong> will rapidly oxidise to <strong>carbon</strong> dioxide or <strong>carbon</strong>monoxide resulting in complete disintegration <strong>of</strong> <strong>the</strong> <strong>carbon</strong> [134]. However, as discussedearlier in Section 2.2.3.2, carefully controlled oxidation can be <strong>use</strong>d to open <strong>the</strong> pores <strong>of</strong><strong>nanoporous</strong> <strong>carbon</strong> <strong>membranes</strong>. Hayashi et al [135] have oxidised NPC <strong>membranes</strong> madefrom polyimide at 700 °C in an oxygen-nitrogen stream at 300 °C and found that <strong>the</strong> gaspermeance was increased <strong>for</strong> a range <strong>of</strong> gas molecules due to <strong>the</strong> widening <strong>of</strong> <strong>the</strong> pore sizedistribution. In <strong>the</strong> same paper, Hayashi et al [135] have also tested <strong>the</strong> stability <strong>of</strong> CMS<strong>membranes</strong> when exposed to air at 100 °C over a period <strong>of</strong> 30 days. Over <strong>the</strong> period <strong>of</strong> 30days, <strong>the</strong> membrane lost a total <strong>of</strong> ~ 17 % <strong>of</strong> <strong>the</strong> initial weight due to <strong>the</strong> loss <strong>of</strong> <strong>carbon</strong> from<strong>the</strong> <strong>for</strong>mation <strong>of</strong> <strong>the</strong> <strong>carbon</strong>yl groups. The permeance <strong>of</strong> <strong>the</strong> exposed membrane decreaseddramatically with time due to <strong>the</strong> adsorption <strong>of</strong> water but was largely recovered by a post-heattreatment using nitrogen at 600 °C <strong>for</strong> 4 hours.To specifically review <strong>the</strong> impact <strong>of</strong> <strong>the</strong> storage environment upon <strong>the</strong> aging <strong>of</strong> <strong>the</strong>membrane, Menendez and Fuertes [32] have tested <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> madefrom phenolic resin at 700°C stored under air, nitrogen and propylene over a period <strong>of</strong> time.Upon storage under air, <strong>the</strong> permeance <strong>of</strong> various gases through <strong>the</strong> membrane drops by 60 %over a period <strong>of</strong> one week. With storage under nitrogen, <strong>the</strong> drop in permeance is limited toless than 20 % after <strong>the</strong> same period <strong>of</strong> time, whereas under propylene storage, <strong>the</strong> permeanceis actually increased by up 100 %. The authors suggest that <strong>the</strong> increase in permeance <strong>of</strong> <strong>the</strong>samples stored in propylene is possibly related to <strong>the</strong> expansion <strong>of</strong> <strong>the</strong> pores due tohydro<strong>carbon</strong> exposure. The effect on permselectivity is <strong>the</strong> reverse <strong>of</strong> permeance. In <strong>the</strong> case49


<strong>of</strong> air and nitrogen storage, permselectivity is slightly increased whilst <strong>for</strong> propylene storagepermselectivity is decreased by ~ 50 % <strong>for</strong> <strong>the</strong> CO 2 /CH 4 and CO 2 /N 2 pair over a period <strong>of</strong> oneweek.Fur<strong>the</strong>rmore, Menendez and Fuertes [32] have also studied <strong>the</strong> impact <strong>of</strong> humidity on agingby storing <strong>the</strong> NPC <strong>membranes</strong> under normal laboratory air, 100 % humidity air and dry air.The permeance <strong>of</strong> <strong>the</strong> NPC <strong>membranes</strong> stored under lab air and dry air decreased in a similarmanner to around 10 % <strong>of</strong> <strong>the</strong> original permeance after 3 months. Interesting, <strong>for</strong> <strong>the</strong><strong>membranes</strong> stored in 100 % humid air, <strong>the</strong> permeance was only decreased to 80 % <strong>of</strong> <strong>the</strong>original per<strong>for</strong>mance. The reasoning <strong>for</strong> a less dramatic decrease in <strong>the</strong> per<strong>for</strong>mance <strong>for</strong> <strong>the</strong><strong>membranes</strong> stored under humidity is <strong>the</strong> <strong>for</strong>mation <strong>of</strong> a protective surface layer <strong>of</strong> watermolecules which prevents deeper oxygen attack. Regeneration <strong>of</strong> <strong>membranes</strong> at 120°C undervacuum <strong>for</strong> 3 hours resulted in a partially recovery <strong>of</strong> <strong>the</strong> original per<strong>for</strong>mance.In <strong>the</strong> most recent study, Lagorsse et al [127] conducted a detailed experimental investigationinto <strong>the</strong> effect <strong>of</strong> air exposure on NPC <strong>membranes</strong>. The <strong>membranes</strong> were initially exposed toair <strong>for</strong> several days, <strong>the</strong>n pure nitrogen <strong>for</strong> 3 months, air <strong>for</strong> ano<strong>the</strong>r 40 days, propylene <strong>for</strong>several days and <strong>the</strong>n finally pure oxygen <strong>for</strong> ano<strong>the</strong>r 38 days. The rate <strong>of</strong> reduction inpermeance was greatest during <strong>the</strong> first exposure to oxygen and air. The final exposure tooxygen resulted in a much smaller reduction in permeance due to an apparent stabilisation <strong>of</strong><strong>the</strong> <strong>carbon</strong> surface. The authors believe that <strong>the</strong> reaction <strong>of</strong> oxygen with <strong>carbon</strong> <strong>for</strong>ms oxygensurface groups, which are subsequently gasified to CO 2 and CO. This explains <strong>the</strong> 4 % weightreduction <strong>of</strong> <strong>the</strong> NPC membrane when exposed to oxygen and air <strong>for</strong> 30 days.Exposure to nitrogen and propylene resulted in very low reductions in permeance and in fact<strong>the</strong> CO 2 permeance increased upon exposure to propylene [127].This increase in CO 2permeance observed after exposure to propylene concurs with <strong>the</strong> findings <strong>of</strong> Menendez andFuertes [32] suggesting that propylene could be <strong>use</strong>d as a regeneration agent. However, <strong>the</strong>impact <strong>of</strong> selectivity is unclear. Menendez and Fuertes [32] observed a significant decrease in<strong>the</strong> CO 2 /N 2 selectivity whilst Lagorsse et al [127] observed a slight increase in <strong>the</strong> CO 2 /N 2selectivity.Interestingly, Lagorsse et al [127] employ a ra<strong>the</strong>r extreme regeneration technique to restore<strong>the</strong> membrane per<strong>for</strong>mance after exposure to oxygen or air. The technique involves heattreating<strong>the</strong> membrane at 620°C in a reducing atmosphere (H 2 ) to remove any oxygen surfacegroups. The amount <strong>of</strong> oxygen present in <strong>the</strong> NPC membrane was reduced from 4.2 to 3.3 %.The fractional uptakes <strong>of</strong> N 2 and CO 2 were also improved after <strong>the</strong> heat treatment.50


2.7.3 Condensable ImpuritiesCondensable impurities (heavier and aromatic hydro<strong>carbon</strong>s) are an omnipresent part <strong>of</strong> <strong>the</strong>natural gas processing system. Even reservoirs which are largely comprised <strong>of</strong> methane willhave some heavier and aromatic hydro<strong>carbon</strong>s (usually termed “gas condensate”). Themajority <strong>of</strong> <strong>the</strong> gas condensate will drop-out in one or a series <strong>of</strong> upstream pressure separationvessels but <strong>the</strong>re will invariably be some small carry-over into <strong>the</strong> gas stream. As such, gasprocessing facilities using <strong>the</strong> traditional polymeric <strong>membranes</strong> are equipped with expensivepre-treatment equipment to remove <strong>the</strong>se condensable impurities.With NPC <strong>membranes</strong> being touted as more robust than polymeric <strong>membranes</strong> in <strong>the</strong>presence <strong>of</strong> impurities, <strong>the</strong>re is potential <strong>for</strong> a reduction in <strong>the</strong> pre-treatment required and<strong>the</strong>re<strong>for</strong>e cost. At <strong>the</strong> research level, hexane (C 6 H 14 ) or heptane (C 7 H 16 ) is commonly <strong>use</strong>d tosimulate <strong>the</strong> impact <strong>of</strong> heavier hydro<strong>carbon</strong>s whilst toluene (C 7 H 8 ) is <strong>use</strong>d to simulate <strong>the</strong>impact <strong>of</strong> aromatic hydro<strong>carbon</strong>s.Tanihara et al [103] present an interesting comparison between <strong>the</strong> effect <strong>of</strong> toluene exposureon polymeric polyimide <strong>membranes</strong> compared with NPC <strong>membranes</strong> also made frompolyimide. Using a feed gas comprising H 2 /CH 4 (50:50 vol%), <strong>the</strong> authors demonstrate thatwith increasing Toluene concentration <strong>the</strong> permeance and selectivity fall significantly <strong>for</strong> <strong>the</strong>polymeric <strong>membranes</strong> but not <strong>for</strong> <strong>the</strong> NPC <strong>membranes</strong>. It is suggested that <strong>the</strong> significanteffect on <strong>the</strong> polymeric <strong>membranes</strong> is due to toluene adsorbing in <strong>the</strong> Langmuir domainsalong with irreversible compaction <strong>of</strong> <strong>the</strong> membrane ca<strong>use</strong> by plasticisation. The resistance <strong>of</strong><strong>the</strong> NPC <strong>membranes</strong> to plasticisation is considered to prevent <strong>the</strong> adverse effect onper<strong>for</strong>mance in <strong>the</strong> presence <strong>of</strong> toluene. Regeneration <strong>of</strong> <strong>the</strong> <strong>membranes</strong> using dry feed gas at50 °C resulted in a 50 % recovery <strong>for</strong> <strong>the</strong> polymeric <strong>membranes</strong> and a complete recovery <strong>for</strong><strong>the</strong> NPC <strong>membranes</strong>.Conversely, Jones and Koros [34] found a dramatic decrease in O 2 /N 2 separation per<strong>for</strong>mance<strong>of</strong> NPC <strong>membranes</strong> made from polyimide when exposed to trace amounts <strong>of</strong> hexane andtoluene. It was also observed that with exposure to trace amounts <strong>of</strong> even <strong>high</strong>erhydro<strong>carbon</strong>s (up to C 16 ), <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> <strong>the</strong> <strong>membranes</strong> drops by up to 99%. On <strong>the</strong>o<strong>the</strong>r hand, exposing <strong>the</strong> NPC <strong>membranes</strong> to lighter hydro<strong>carbon</strong>s such as propane andpropylene resulted in a much smaller drop in per<strong>for</strong>mance when compared with hexane,toluene and <strong>the</strong> <strong>high</strong>er hydro<strong>carbon</strong>s. A novel regeneration technique using pure propylene at1000 kPag <strong>for</strong> 2 to 3 hours was developed resulting in complete recovery <strong>of</strong> <strong>the</strong> membraneper<strong>for</strong>mance after hydro<strong>carbon</strong> exposure.51


In a later work, Vu et al [35] have reviewed <strong>the</strong> effect <strong>of</strong> condensable impurities on <strong>the</strong> mixedgas CO 2 /CH 4 separation per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> made from polyimide. In <strong>the</strong>presence <strong>of</strong> ei<strong>the</strong>r heptane or toluene, <strong>the</strong>re is a considerable initial drop in <strong>the</strong> permeancewhich <strong>the</strong>n reaches equilibrium after approximately 1 hour. Interestingly, <strong>the</strong>re is no changein <strong>the</strong> CO 2 /CH 4 selectivity presumably beca<strong>use</strong> <strong>the</strong> CO 2 and CH 4 permeances have reduced ina similar manner. In <strong>the</strong> instance <strong>of</strong> toluene exposure, increasing <strong>the</strong> concentration from 70.4to 295 ppm resulted in a similar % reduction in per<strong>for</strong>mance. In <strong>the</strong> instance <strong>of</strong> hexaneexposure, increasing <strong>the</strong> concentration from 100 to 302 ppm actually resulted in a reducedeffect on <strong>the</strong> membrane per<strong>for</strong>mance. No explanation is given <strong>for</strong> this result. Completerecovery <strong>of</strong> <strong>the</strong> permeance and selectivity is obtained from regeneration at 90°C in flowingnitrogen at 300 kPag <strong>for</strong> at least 12 hours suggesting that <strong>the</strong> impurities are only physicallyadsorbed to <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong>.In a similar paper, Vu et al [136] have exposed mixed matrix <strong>membranes</strong> comprising <strong>carbon</strong>molecular sieve particles within a polyimide matrix (Matrimid) to a low concentration (70ppm) <strong>of</strong> toluene in a CO 2 /CH 4 mixed feed gas. A membrane made from pure Matrimid wasalso exposed to 70 ppm <strong>of</strong> toluene <strong>for</strong> direct comparison. Over a period <strong>of</strong> 60 hours, <strong>the</strong>re isno change in <strong>the</strong> CO 2 permeability <strong>for</strong> <strong>the</strong> mixed matrix membrane and only a slight decreasein <strong>the</strong> CO 2 permeability <strong>for</strong> <strong>the</strong> pure polymer membrane. Selectivity is fairly stable in bothcases. The authors postulate that <strong>the</strong> toluene may only occupy <strong>the</strong> larger non-selective pores<strong>of</strong> <strong>the</strong> mixed matrix membrane and as such <strong>the</strong> effect on <strong>the</strong> per<strong>for</strong>mance is minimal. Thismay well be true <strong>for</strong> very small concentrations <strong>of</strong> toluene, as <strong>use</strong>d in this work.Zeolite <strong>membranes</strong>, like NPC <strong>membranes</strong>, have pores sizes < 1 nm and generally follow amolecular sieving mechanism. Li et al [137] have exposed SAPO-34 zeolite <strong>membranes</strong> toethylene (C 2 H 4 ), propane (C 3 H 8 ) and butane (C 4 H 10 ) to review <strong>the</strong> impact on per<strong>for</strong>mance <strong>of</strong>condensable impurities in a CO 2 /CH 4 mixed gas feed. With a 1 mol% hydro<strong>carbon</strong> impurityin <strong>the</strong> feed, <strong>the</strong> CO 2 permeance and CO 2 /CH 4 selectivity were reduced with <strong>the</strong> greatestreduction in per<strong>for</strong>mance being from exposure to <strong>the</strong> heaviest hydro<strong>carbon</strong> (butane in thiscase). This is explained by <strong>the</strong> increasing heat <strong>of</strong> adsorption <strong>of</strong> hydro<strong>carbon</strong>s on zeolite withincreasing <strong>carbon</strong> number. Whilst increasing <strong>the</strong> partial pressure or concentration <strong>of</strong> <strong>the</strong>hydro<strong>carbon</strong> in <strong>the</strong> feed also increases <strong>the</strong> reduction in per<strong>for</strong>mance, <strong>the</strong> per<strong>for</strong>mance is stableif <strong>the</strong> partial pressure <strong>of</strong> <strong>the</strong> feed gases is increased in <strong>the</strong> same proportion. Regeneration <strong>of</strong><strong>the</strong> zeolite <strong>membranes</strong> was achieved by exposing <strong>the</strong> membrane to feed gas containing noimpurities or by calcination.In summary, condensable impurities do seem to impact upon <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC<strong>membranes</strong> by blocking <strong>the</strong> pores and reducing permeance and selectivity. This effect is more52


pronounced as <strong>the</strong> concentration or molecular weight <strong>of</strong> <strong>the</strong> condensable impurity increases.However, it appears that adsorption <strong>of</strong> <strong>the</strong> impurity onto <strong>the</strong> surface is physical and as such<strong>the</strong> NPC membrane can be regenerated using fairly benign techniques.2.7.4 Non-Condensable ImpuritiesNon-condensable impurities are impurities that remain in a gaseous <strong>for</strong>m upon contact with<strong>the</strong> NPC membrane and may impact on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> <strong>the</strong> membrane by physicaladsorption and pore blocking or by chemically reacting with <strong>the</strong> <strong>carbon</strong>.In <strong>the</strong> case <strong>of</strong> CO 2 removal from natural gas, hydrogen sulphide (H 2 S) is present due tosulphide producing bacteria present in <strong>the</strong> reservoir. In <strong>the</strong> case <strong>of</strong> CO 2 removal fromsyn<strong>the</strong>sis gas, H 2 S is present again from <strong>the</strong> sulphur producing bacteria present in <strong>the</strong> originalhydro<strong>carbon</strong> source (such as coal). Ano<strong>the</strong>r impurity present in syn<strong>the</strong>sis gas, which is known<strong>for</strong> poisoning catalyst materials due to its reactivity, is <strong>carbon</strong> monoxide (CO) produced by<strong>the</strong> water-gas shift reaction <strong>use</strong>d to make <strong>the</strong> syn<strong>the</strong>sis gas (H 2 ).It is generally necessary to remove H 2 S from natural gas to a very low level to meet <strong>the</strong> salesgas specification. This is done using a chemical scavenger or molecular sieve if <strong>the</strong>concentration <strong>of</strong> H 2 S is low. If <strong>the</strong> concentration <strong>of</strong> H 2 S in <strong>the</strong> natural gas is <strong>high</strong> <strong>the</strong>n largerequipment <strong>for</strong> removing H 2 S must be employed. The most commonly <strong>use</strong>d de-sulphurisationprocess is <strong>the</strong> Claus Process, which oxidises H 2 S to produce elemental sulphur (S) [123]. Theequipment required <strong>for</strong> <strong>the</strong> Claus process is extensive due to <strong>the</strong> <strong>high</strong> <strong>temperature</strong>s andcatalysts necessary to obtain a good yield.The traditional polymeric <strong>membranes</strong> are susceptible to plasticisation upon prolongedexposure to H 2 S [138]. It has also been shown that H 2 S <strong>the</strong> reduces <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> Pd/Cu<strong>membranes</strong>, which are traditionally <strong>use</strong>d in syn<strong>the</strong>sis gas purification, due to <strong>the</strong> <strong>for</strong>mation <strong>of</strong>surface sulphides that block H 2 adsorption sites [139]. If NPC <strong>membranes</strong> could be shown tobe resistant to H 2 S whilst also removing some H 2 S into <strong>the</strong> CO 2 stream <strong>for</strong> sequestration,savings could be made in terms <strong>of</strong> <strong>the</strong> H 2 S removal systems required.Whilst it sometimes necessary to remove <strong>the</strong> CO present in syn<strong>the</strong>sis gas as it can poisoncatalysts, it is <strong>of</strong>ten desired to retain <strong>the</strong> CO in <strong>the</strong> syn<strong>the</strong>sis gas if it is being <strong>use</strong>d to makesyn<strong>the</strong>tic fuels (Fisher-Tropsh Process) as shown in <strong>the</strong> reaction below [140].~ 400 °C (2n+1)H 2 + nCO → C n H (2n+2) + nH 2 OStudies <strong>of</strong> polymeric <strong>membranes</strong>, in particular polyimide, show that CO is retained on <strong>the</strong>retentate side as it has a larger kinetic diameter and lower solubility than CO 2 [138].53


Whilst <strong>the</strong>re is little known specifically <strong>of</strong> <strong>the</strong> effect <strong>of</strong> H 2 S and CO on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong>NPC <strong>membranes</strong>, <strong>the</strong>re is considerable literature available on <strong>the</strong> adsorption <strong>of</strong> <strong>the</strong>se gases onactivated <strong>carbon</strong>. This is a good beginning <strong>for</strong> understanding how <strong>the</strong>se non-condensableimpurities may affect <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong>.Ritter and Yang [50] have obtained adsorption iso<strong>the</strong>rms <strong>of</strong> CO and H 2 S along with H 2 , CO 2and CH 4 on activated <strong>carbon</strong> at room <strong>temperature</strong>. For single gas experiments, <strong>the</strong> volume <strong>of</strong>gas adsorbed is as follows; H 2 S>CO 2 >CH 4 >CO>H 2 , with all gases following <strong>the</strong> Langmuirmodel <strong>of</strong> adsorption. For <strong>the</strong> dual gas experiments, CO 2 adsorbed more than CO, H 2 Sadsorbed more than CO 2 and CO 2 adsorbed more than CH 4 . An extended Langmuir Equationincorporating interaction parameters was <strong>use</strong>d to model <strong>the</strong> adsorption <strong>of</strong> mixed gas systems.In a later article, Yan et al [141] have studied <strong>the</strong> mechanism <strong>of</strong> <strong>the</strong> adsorption <strong>of</strong> H 2 S ontoactivated <strong>carbon</strong>. Using kinetic adsorption and characterisation data, it was determined thatH 2 S both chemically and physically adsorbs onto <strong>the</strong> <strong>carbon</strong> surface. The presence <strong>of</strong>potassium hydroxide (KOH) in <strong>the</strong> activated <strong>carbon</strong> reacts with <strong>the</strong> H 2 S to produce potassiumhydrosulphide (KHS) or potassium sulphide (K 2 S).Additionally and more importantly when relating Yan et al’s [141] results to NPC<strong>membranes</strong>, it was found that H 2 S reacts with adsorbed oxygen upon <strong>the</strong> activated <strong>carbon</strong> t<strong>of</strong>orm elemental sulphur, which in <strong>the</strong> presence <strong>of</strong> water can <strong>for</strong>m sulphuric acid. In fact, <strong>the</strong>authors go on to comment that activated <strong>carbon</strong>s can deteriorate through <strong>the</strong> <strong>for</strong>mation <strong>of</strong>sulphuric acid.Activated <strong>carbon</strong>s impregnated with metal salts are <strong>of</strong>ten <strong>use</strong>d <strong>for</strong> <strong>the</strong> removal <strong>of</strong> CO. Tamonet al [142] have impregnated activated <strong>carbon</strong> with palladium chloride (PdCl 2 ) and copperchloride (CuCl) and found that this increases <strong>the</strong> adsorption <strong>of</strong> CO onto <strong>the</strong> activated <strong>carbon</strong>surface by over an order <strong>of</strong> magnitude. Taking a different approach and using a metal oxide,Mohamad et al [143] have impregnated activated <strong>carbon</strong> with tin oxide (SnO 2 ) and found anincrease in <strong>the</strong> adsorption capacity due to CO reacting with <strong>the</strong> SnO 2 to produce CO 2 .Assuming that NPC <strong>membranes</strong> will exhibit <strong>the</strong> same adsorptive behaviour as activated<strong>carbon</strong>, it is anticipated that based on <strong>the</strong> adsorptive component <strong>of</strong> permeance H 2 S shouldtravel through <strong>the</strong> membrane with <strong>the</strong> CO 2 , whilst <strong>the</strong> CO should be retained on <strong>the</strong> retentate.Additionally, if we consider <strong>the</strong> kinetic diameters <strong>of</strong> <strong>the</strong> gas molecules (Table 2-1) in relationto a molecular sieving mechanism, H 2 S should travel through with <strong>the</strong> CO 2 , whilst CO maybe retained on <strong>the</strong> retentate side.54


2.8 Scope <strong>of</strong> this Research WorkBased upon a promising lab-scale per<strong>for</strong>mance, <strong>the</strong>re has been significant research into <strong>the</strong>production <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> <strong>membranes</strong> over <strong>the</strong> last decade.Methods to produce NPC <strong>membranes</strong> from <strong>the</strong> pyrolysis <strong>of</strong> various polymers are welldocumented although <strong>the</strong> problems with cracking and aging <strong>of</strong> <strong>the</strong> membrane surface duringpyrolysis are still to be resolved.The published per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> is frequently measured at ideal conditions,i.e. with a pure gas and at room <strong>temperature</strong>. The per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> in acommercial setting, i.e. with a mixed gas feed, at elevated <strong>temperature</strong>s and in <strong>the</strong> presence <strong>of</strong>impurities, is not as well understood.The focus <strong>of</strong> this research work is to develop a better understanding <strong>of</strong> <strong>the</strong> per<strong>for</strong>mance <strong>of</strong>NPC <strong>membranes</strong> in commercial applications and as such assess <strong>the</strong> suitability <strong>for</strong> <strong>the</strong>separation <strong>of</strong> CO 2 from natural gas and syn<strong>the</strong>sis gas (pre-combustion capture). Specifically,<strong>the</strong> aims <strong>of</strong> <strong>the</strong> research work are <strong>the</strong>re<strong>for</strong>e to obtain a better understanding <strong>of</strong>1. Avoiding cracking in NPC membrane manufacture (Chapter 4),2. Mixed feed gas per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> (Chapter 4),3. High <strong>temperature</strong> operation <strong>of</strong> NPC <strong>membranes</strong> (Chapter 4),4. The transport mechanisms <strong>of</strong> gases through NPC <strong>membranes</strong> at <strong>high</strong> operating<strong>temperature</strong>s (Chapter 5),5. The impact <strong>of</strong> impurities present in natural gas and syn<strong>the</strong>sis gas on <strong>the</strong> per<strong>for</strong>mance<strong>of</strong> NPC <strong>membranes</strong> (Chapter 6).55


3 EXPERIMENTAL MATERIALS, APPARATUS AND METHODS3.1 IntroductionThe reagents, materials and methods <strong>use</strong>d to manufacture supported, modified supported andunsupported NPC <strong>membranes</strong> are presented in this chapter. The techniques employed tophysically characterise <strong>the</strong> NPC <strong>membranes</strong> are <strong>the</strong>n described.Following on from this, <strong>the</strong> experimental equipment <strong>use</strong>d to measure <strong>the</strong> permeance andselectivity <strong>of</strong> <strong>the</strong> manufactured <strong>membranes</strong> is outlined. Finally, <strong>the</strong> equipment and methods<strong>use</strong>d <strong>for</strong> <strong>the</strong> measurement <strong>of</strong> gas adsorption onto <strong>the</strong> NPC <strong>membranes</strong> both experimentallyand <strong>the</strong>oretically via molecular simulation are presented.3.2 Membrane Materials, Manufacturing Equipment and Techniques3.2.1 MaterialsLists <strong>of</strong> <strong>the</strong> materials <strong>use</strong>d to manufacture supported, modified supported and unsupportedNPC <strong>membranes</strong> are presented in Tables 3.1, 3.2 and 3.3, respectively. Please note that <strong>the</strong>materials <strong>use</strong>d to make modified supported NPC <strong>membranes</strong> presented in Table 3.2 are inaddition to those presented in Table 3-1.Table 3-1 : Materials <strong>use</strong>d to manufacture supported NPC <strong>membranes</strong>.Material Use Supplier PurityPolymer, which uponPolyfurfuryl Alcohol100 %pyrolysis becomes Polysciences, USA(PFA)(Note 1)<strong>nanoporous</strong> <strong>carbon</strong>Solvent to makeAcetonepolymer precursorsolutionMerck, Germany 99.8 %Porous Stainless SteelDisks (44 mm Diameter,1 mm thick)Chlor<strong>of</strong>ormNitrogenSupportCleaning agent <strong>for</strong>support disksSpray nozzle andspin-coater purge gasMott MetallurgicalCompany, USAMerck, Germany 99.8 %BOC, AustraliaArgon Pyrolysis purge gas BOC, AustraliaHigh Purity99.99%Ultra High Purity99.999%(Note 2)57


Note 1 – Whilst <strong>the</strong> purity <strong>of</strong> <strong>the</strong> PFA from <strong>the</strong> supplier was documented as 100%, <strong>the</strong>molecular weight varied. The PFA received prior to August 2007 had a <strong>high</strong> molecularweight, which gave a good conversion to <strong>nanoporous</strong> <strong>carbon</strong> upon pyrolysis. All <strong>of</strong> <strong>the</strong>per<strong>for</strong>mance data presented in Chapter 4 and in <strong>the</strong> presence <strong>of</strong> condensable impurities weremanufactured from this lot <strong>of</strong> PFA. Un<strong>for</strong>tunately, manufacturing difficulties experienced by<strong>the</strong> supplier led to cessation <strong>of</strong> supply and <strong>the</strong>n a replacement lower molecular weightpolymer in January 2008. Numerous attempts were made to manufacture NPC <strong>membranes</strong>from this polymer but were un<strong>for</strong>tunately unsuccessful as <strong>the</strong> polymer could not be spincoatedadequately and also did provide a good conversion to <strong>nanoporous</strong> <strong>carbon</strong>. In July2008, <strong>the</strong> supplier provided a <strong>high</strong>er molecular weight polymer, similar to <strong>the</strong> originalpolymer <strong>use</strong>d. NPC <strong>membranes</strong> capable <strong>of</strong> gas separation were made out <strong>of</strong> this polymer.These <strong>membranes</strong> were <strong>use</strong>d <strong>for</strong> testing with non-condensable impurities. The quality <strong>of</strong> <strong>the</strong>PFA has a significant impact on <strong>the</strong> ability to manufacture NPC <strong>membranes</strong> as was alsodiscovered by Sedigh et al [25].Note 2 – It was necessary to <strong>use</strong> ultra <strong>high</strong> purity Argon to limit <strong>the</strong> oxygen content duringpyrolysis. Using Argon with a lower purity (e.g. <strong>high</strong> purity 99.99%) resulted in very poorconversion <strong>of</strong> <strong>the</strong> polymer to <strong>nanoporous</strong> <strong>carbon</strong>.Table 3-2 : Extra materials <strong>use</strong>d to manufacture modified supported NPC <strong>membranes</strong>.Material Use Supplier PurityTetraethylorthosilicate Stoeber syn<strong>the</strong>sis <strong>of</strong>(TEOS)silica particlesFluka 99.5 %Ethanol Stoeber syn<strong>the</strong>sis Chem-Supply 99.5 %Ammonia Stoeber syn<strong>the</strong>sis Merck 25 %Table 3-3 : Materials <strong>use</strong>d to manufacture unsupported NPC <strong>membranes</strong>.Material Use Supplier PurityMatrimid Polyimide Polymer, which uponVantico,Pellets (structure shown pyrolysis becomes100 %Switzerlandin Figure 3-1) <strong>nanoporous</strong> <strong>carbon</strong>DichloromethaneKapton PolyimideFlatsheet 100 HN and200 HN (structureshown in Figure 3-2)Solvent to makeMatrimid precursorsolutionPolymer, which uponpyrolysis becomes<strong>nanoporous</strong> <strong>carbon</strong>Ajax Finechem,Australia99.8 %Dupont, USA 100 %Ultra High Purity Argon Pyrolysis purge gas BOC, Australia 99.999%58


OOONCCCCCNOOCH 3CH 3H 3 CnFigure 3-1 : Chemical structure <strong>of</strong> Matrimid polyimide (Vantico, Switzerland)OOCCNCCNOOOnFigure 3-2 : Chemical structure <strong>of</strong> Kapton polyimide (Dupont, USA)3.2.2 Manufacturing Equipment and Techniques3.2.2.1 Supported NPC MembranesThe polyfurfuryl alcohol (PFA) polymer was diluted in acetone to reduce <strong>the</strong> viscosity.Different concentrations <strong>of</strong> PFA in acetone were trialled but ultimately a solution with 40wt% PFA in acetone gave a solution viscosity similar to water and was thus ideal <strong>for</strong> coating.Porous stainless steel disks with a pore size <strong>of</strong> 0.2 micron were <strong>use</strong>d <strong>for</strong> support. Prior tocoating, <strong>the</strong> disks were sonicated in chlor<strong>of</strong>orm <strong>for</strong> 15 mins to remove any grease, washedtwice with distilled water and <strong>the</strong>n dried in <strong>the</strong> oven at 120 °C overnight.The PFA solution was coated onto <strong>the</strong> porous support using a spin coater (LaurellTechnologies, USA) and a pneumatic spray nozzle (Spraying Systems Co. Pty Ltd, Australia).The set up <strong>of</strong> <strong>the</strong> coating equipment is presented in Figure 3-3. The porous support wasplaced centrally onto <strong>the</strong> spinning chuck and secured by a vacuum line (V4) and seal purgegas (V3). Using an inbuilt controller, <strong>the</strong> support was <strong>the</strong>n accelerated to a constant spinningspeed <strong>of</strong> 1000 rpm.The PFA solution was pipetted into <strong>the</strong> measuring cylinder. The solution was <strong>the</strong>n atomisedand sprayed onto <strong>the</strong> spinning porous support by simultaneously opening valves V1 and V2.The amount <strong>of</strong> PFA solution sprayed onto <strong>the</strong> support was estimated using <strong>the</strong> graduations on<strong>the</strong> measuring cylinder be<strong>for</strong>e and after coating. For <strong>the</strong> initial coat <strong>of</strong> <strong>the</strong> PFA solution, 3-459


ml was sprayed and <strong>for</strong> <strong>the</strong> subsequent coats (after <strong>carbon</strong>isation), 1 -2 ml <strong>of</strong> solution wassprayed.The pressure <strong>of</strong> <strong>the</strong> atomising gas was controlled at a pressure <strong>of</strong> less than 40 kPag to ensure asteady and gradual flow <strong>of</strong> <strong>the</strong> PFA solution through <strong>the</strong> nozzle. The distance between <strong>the</strong>spray nozzle and <strong>the</strong> spinning porous support was optimised at 6 cm as this distanceminimised <strong>the</strong> overall loss <strong>of</strong> PFA solution whilst still giving an even coating.(a)(b)PFA SolutionMeasuringCylinderSpray NozzleAtomising GasPGV1V2N 2PGSeal Purge GasV3Porous SupportSpinning ChuckV4Vacuum PumpN 2Figure 3-3 : (a) Photograph and (b) schematic diagram <strong>of</strong> <strong>the</strong> spin coating apparatus.The porous support is held onto a spinning chuck using a vacuum pump and a sealpurge gas. The PFA solution is sprayed onto <strong>the</strong> spinning porous support using apneumatic spray nozzle.60


Immediately after coating, <strong>the</strong> support and polymer were pyrolysed in a quartz tube furnace(Ceramic Engineering, Australia) (refer to Figure 3-4) at a <strong>temperature</strong> <strong>of</strong> between 400 and600°C <strong>for</strong> 2 hours. The <strong>temperature</strong> was ramped to <strong>the</strong> final pyrolysis <strong>temperature</strong> at 5°C/minwith a pa<strong>use</strong> at 100°C <strong>for</strong> 60 mins to allow evaporation <strong>of</strong> <strong>the</strong> acetone solvent. A typical<strong>temperature</strong> pr<strong>of</strong>ile <strong>of</strong> <strong>the</strong> furnace with time is shown in Figure 3-5.An ultra <strong>high</strong> purity argon flow <strong>of</strong> 200 ml/min was maintained at all times until <strong>the</strong> furnacehad returned to room <strong>temperature</strong>. The argon flow rate <strong>of</strong> 200 ml/min was chosen based uponexperience <strong>of</strong> o<strong>the</strong>r researchers presented in <strong>the</strong> literature. It is generally agreed that <strong>the</strong><strong>high</strong>er <strong>the</strong> inert gas rate, <strong>the</strong> more permeable <strong>the</strong> membrane due to faster sweeping <strong>of</strong> <strong>the</strong>liberated gases upon pyrolysis [90]. Additionally, argon was chosen in this case but o<strong>the</strong>rinert gases such as helium can be <strong>use</strong>d provided <strong>the</strong> oxygen content is limited to less than 3ppm [90, 144].(a)(b)Quartz TubeInsulated FurnaceFlow MeterExhaustPGCoated PorousSupportSupport HolderPurge GasArFigure 3-4 : (a) Photograph and (b) schematic diagram <strong>of</strong> <strong>the</strong> quartz tube furnace with acontinuous purge <strong>of</strong> ultra <strong>high</strong> purity argon.61


Furnace Temperature (°C)800Oven HeatingOven Off (Natural Cooling)7002 hours @ final pyrolysis <strong>temperature</strong>6005004003002001001 hour @ 100°C to remove solvent00 100 200 300 400 500 600 700 800 900 1000Time (min)Figure 3-5 : Quartz tube furnace typical <strong>temperature</strong> pr<strong>of</strong>ile with time during pyrolysis<strong>of</strong> supported NPC <strong>membranes</strong>.The coating and pyrolysis process was repeated up to four times as was necessary to <strong>for</strong>m acomplete membrane layer. The presence <strong>of</strong> a complete membrane layer was establishedvisually from coverage <strong>of</strong> continuous shiny black surface on <strong>the</strong> porous support and viaper<strong>for</strong>mance testing to determine <strong>the</strong> permeance and selectivity.The <strong>membranes</strong> were weighed be<strong>for</strong>e and after coating as well as be<strong>for</strong>e and after pyrolysis.Additionally, <strong>the</strong> <strong>membranes</strong> were stored in a dessicator filled with ultra <strong>high</strong> purity argon tolimit <strong>the</strong> chemisorption <strong>of</strong> oxygen and physisorption <strong>of</strong> water, which reduce <strong>the</strong> per<strong>for</strong>mance<strong>of</strong> <strong>the</strong> membrane as referred to in Section 2.7.2.3.2.2.1.1 Repeatability in Supported NPC Membrane ManufactureObtaining a smooth coating <strong>of</strong> <strong>the</strong> polymer solution was paramount. Any bubbles in <strong>the</strong>solution or an uneven coating led to defeats in <strong>the</strong> <strong>carbon</strong> surface. Bubbles or an unevencoating occurred if <strong>the</strong> spray nozzle became partially blocked and did not spray evenly.Additionally, <strong>the</strong> change in <strong>the</strong> purity and molecular weight <strong>of</strong> <strong>the</strong> PFA by <strong>the</strong> polymersupplier (in January 2008) greatly affected <strong>the</strong> viscosity, <strong>the</strong> dilution required <strong>for</strong> coating andalso <strong>the</strong> ability <strong>of</strong> <strong>the</strong> solution to wet <strong>the</strong> stainless steel support as discussed earlier in Note 1to Table 3-1.Additionally, to obtain a crack-free <strong>carbon</strong> surface during pyrolysis, it was necessary toensure that <strong>the</strong> quartz tube furnace was completely flat to prevent any pooling <strong>of</strong> polymer onone side <strong>of</strong> <strong>the</strong> support and also that <strong>the</strong> furnace was free <strong>of</strong> loose particles. It was alsonecessary to ensure that all equipment related to weighing <strong>the</strong> <strong>carbon</strong> <strong>membranes</strong> be<strong>for</strong>e andafter pyrolysis was free <strong>of</strong> any loose particles.62


3.2.2.1.2 Manufacturing Method Development Enabling Support Pore OpeningDuring <strong>the</strong> initial stages <strong>of</strong> <strong>the</strong> research work, it was noticed that if <strong>the</strong> support was exposedto air during <strong>the</strong> first and second coats, <strong>the</strong> permeance was significantly larger.The following trial was <strong>the</strong>n undertaken. Two NPC <strong>membranes</strong> were manufactured at 550°C.For <strong>the</strong> first membrane, <strong>the</strong> second and third coats <strong>of</strong> <strong>carbon</strong> were applied directly after <strong>the</strong>first coat and thus minimising air contact. The second membrane was manufactured byallowing <strong>the</strong> first coat to sit in atmospheric air <strong>for</strong> approximately one week be<strong>for</strong>e <strong>the</strong> secondand third coats were added.The permeance <strong>of</strong> <strong>the</strong> <strong>membranes</strong> was <strong>the</strong>n measured. The results presented in Figure 3-6,show an increase in <strong>the</strong> permeance <strong>of</strong> <strong>the</strong> oxidated-membrane by 1 to 2 orders <strong>of</strong> magnitude.It should also be noted that <strong>the</strong> oxidation <strong>of</strong> <strong>the</strong> first coat did not compromise <strong>the</strong> selectivityas similar selectivities were observed <strong>for</strong> <strong>the</strong> both <strong>membranes</strong>.The results are presented in terms <strong>of</strong> <strong>the</strong> N 2 permeance <strong>for</strong> comparison with <strong>the</strong> publishedresults <strong>of</strong> Rajagopalan et al [28] and Merritt et al [27] who added silica nanoparticles to <strong>the</strong>porous stainless steel support. The addition <strong>of</strong> silica nanoparticles prevented <strong>nanoporous</strong><strong>carbon</strong> ingress into <strong>the</strong> support, which decreased <strong>the</strong> thickness <strong>of</strong> <strong>the</strong> membrane and henceincreased <strong>the</strong> permeance.1.00E-07N 2 Permeance (mol/m 2 .s.Pa)1.00E-081.00E-091.00E-101.00E-111.00E-120 NPC Membrane 1 NPC 2MembraneNPC 3Membranes4NPC Membranes 5Not Oxidised OxidisedNo Silica Particles With Silica Particles[Rajagapolan et al] [22, 21]Figure 3-6 : The impact <strong>of</strong> oxidising <strong>the</strong> first coat <strong>of</strong> <strong>carbon</strong> <strong>of</strong> <strong>membranes</strong>manufactured at 550°C. The increase in permeance is 1 to 2 orders <strong>of</strong> magnitude. Thisresult is similar to that achieved from <strong>the</strong> addition <strong>of</strong> silica particles [28] [27].63


Interestingly, Hayashi et al [135] reported a widening <strong>of</strong> <strong>the</strong> pore size distribution wi<strong>the</strong>xposure to oxygen at 300°C prior to pyrolysis at a <strong>high</strong>er <strong>temperature</strong>. Whilst it has not beenpossible to confirm <strong>the</strong> mechanism <strong>of</strong> pore opening via pore size characterisation techniquesas part <strong>of</strong> this research work, it is suggested that oxygen adsorption from <strong>the</strong> air onto <strong>the</strong> firstcoat has increased <strong>the</strong> pore size <strong>of</strong> <strong>the</strong> NPC within <strong>the</strong> support during <strong>the</strong> pyrolysis <strong>of</strong> <strong>the</strong>second coat.As this technique was developed in <strong>the</strong> early stages <strong>of</strong> this research work, <strong>the</strong> resultspresented in this <strong>the</strong>sis <strong>for</strong> supported NPC <strong>membranes</strong> were obtained using <strong>membranes</strong> withan oxidised first coat.3.2.2.2 Modified Supported NPC MembranesThe alternative method <strong>of</strong> modifying <strong>the</strong> porous supports with silica particles to reduce <strong>the</strong>overall thickness <strong>of</strong> <strong>the</strong> membrane proposed by Rajagopalan et al [28] and Merritt et al [27]was also trialled.The pore size <strong>of</strong> <strong>the</strong> supports was reduced by <strong>the</strong> addition <strong>of</strong> silica nanoparticles. The silicananoparticles were produced via <strong>the</strong> reaction <strong>of</strong> tetraethylorthosilicate (TEOS) with water inlow-molecular weight alcohols using an ammonia-based catalyst, which is also known as <strong>the</strong>Stoeber syn<strong>the</strong>sis [145].The reaction <strong>of</strong> TEOS (Si(OR) 4, R = C 2 H 4 ) with water in alcohol to produce silica (SiO 2 ) isdescribed by <strong>the</strong> two reactions presented below [146, 147].Si OR)+ H 0 → ( OR)SI(OH ) + ROH(4 23( RO ) Si(OH H O → SiO + ROH3) +22The ultimate size <strong>of</strong> <strong>the</strong> silica particles produced from <strong>the</strong> Stoeber syn<strong>the</strong>sis depends upon (1)<strong>the</strong> concentration <strong>of</strong> TEOS, (2) <strong>the</strong> concentration <strong>of</strong> ammonia, (3) <strong>the</strong> concentration <strong>of</strong> water,(4) <strong>the</strong> reaction <strong>temperature</strong> and (5) <strong>the</strong> alcohol which is <strong>use</strong>d as a solvent [146].Various authors have studied <strong>the</strong> effect <strong>of</strong> <strong>the</strong> above five variables and correlations exist <strong>for</strong>producing silica nanoparticles <strong>of</strong> a particular size [145-148]. However, in this case, <strong>high</strong>lymonodisperse silica nanoparticles were made using a modification to <strong>the</strong> conventional methodproposed by Zhang et al [149]. By diluting <strong>the</strong> TEOS with alcohol prior to <strong>the</strong> reaction, <strong>the</strong>need <strong>for</strong> prior distillation <strong>of</strong> <strong>the</strong> TEOS is eliminated which in turn dramatically decreases <strong>the</strong>complexity <strong>of</strong> <strong>the</strong> manufacture.64


Using <strong>the</strong> method <strong>of</strong> Zhang et al [149], two solutions as shown in Table 3-4 were prepared.Table 3-4 : The composition <strong>of</strong> <strong>the</strong> two solutions prepared <strong>for</strong> <strong>the</strong> manufacture <strong>of</strong> <strong>the</strong>silica nanoparticles <strong>use</strong>d <strong>for</strong> filling <strong>the</strong> porous support prior to coating with PFA.Solution Ethanol Ammonia TEOS1 46 ml 10 ml -2 4 ml - 1 mlThe two solutions were <strong>the</strong>n mixed in a conical flask contained within a water bath controlledat a constant <strong>temperature</strong>. The reaction time and <strong>temperature</strong> <strong>of</strong> <strong>the</strong> water bath were varied toobtained silica particles within <strong>the</strong> desired size range (< 300 nm). The particle sizes <strong>of</strong> <strong>the</strong>produced colloid solution were determined using Scanning Electron Microscopy (SEM) as isdescribed in Section 3.3.A schematic diagram <strong>of</strong> <strong>the</strong> equipment <strong>use</strong>d <strong>for</strong> <strong>the</strong> manufacture <strong>of</strong> <strong>the</strong> silica nanoparticles ispresented in Figure 3-7.Produced SilicaNanoparticlesSolution <strong>of</strong> Ethanol,Ammonia and TEOSWater Bath atConstant TemperatureMagnetic StirrerHot PlateFigure 3-7 : Apparatus <strong>for</strong> creating <strong>the</strong> colloidal solution <strong>of</strong> silica nanoparticles in whicha solution containing ethanol and ammonia was mixed with a solution containingethanol and TEOS and reacted at a constant <strong>temperature</strong>.To impregnate <strong>the</strong> porous support with <strong>the</strong> silica nanoparticles, <strong>the</strong> colloidal solution wasfiltered through <strong>the</strong> support using a filtration cell (Millipore, USA) as shown in Figure 3-8.The porous support was clamped between <strong>the</strong> upper and lower parts <strong>of</strong> <strong>the</strong> filtration cell. Thecolloidal silica solution was <strong>the</strong>n diluted with methanol (1 to 5.5 by volume) and poured ontop <strong>of</strong> <strong>the</strong> porous support. The vacuum line (V1) was <strong>the</strong>n opened to allow a driving <strong>for</strong>ceacross <strong>the</strong> porous support and allow <strong>the</strong> colloidal solution to flow through at a reasonable rate.65


The filtering process was repeated several times until <strong>the</strong> solution contained within <strong>the</strong> bottom<strong>of</strong> <strong>the</strong> filtration cell had become clear indicating <strong>the</strong> presence <strong>of</strong> few silica particles.Colloidal SilicaSolutionPorous StainlessSteel SupportClampV1Vacuum PumpSolution Containing VeryFew Silica ParticlesMillipore Filtration CellFigure 3-8 : Millipore filtration cell <strong>use</strong>d <strong>for</strong> filtering diluted silica nanoparticle colloidsolution through <strong>the</strong> porous stainless steel support.The support containing <strong>the</strong> silica nanoparticles was <strong>the</strong>n dried overnight at 100 °C in avacuum oven to remove <strong>the</strong> liquid. The filtering and drying processes were repeated up to 4times until a white coat was present on <strong>the</strong> surface <strong>of</strong> <strong>the</strong> support indicating saturation <strong>of</strong> <strong>the</strong>support with silica nanoparticles. The supports were weighed be<strong>for</strong>e and after <strong>the</strong> addition <strong>of</strong><strong>the</strong> silica.3.2.2.3 Unsupported NPC MembranesTwo different types <strong>of</strong> polyimide (Matrimid and Kapton) were <strong>use</strong>d to make unsupported<strong>membranes</strong>. Flat sheet polyimide <strong>membranes</strong> were cast from Matrimid pellets. The Matrimidwas dissolved in dichloromethane on a 2 wt% basis and <strong>the</strong>n dispensed into glass rings in <strong>the</strong>fume hood. The <strong>membranes</strong> were be left overnight in <strong>the</strong> fume hood to allow evaporation <strong>of</strong><strong>the</strong> solvent and <strong>the</strong>n dried in <strong>the</strong> vacuum oven <strong>for</strong> 4 days at 150 °C. Kapton polyimide flatsheet (100 HN and 200 HN) was obtained directly from Dupont.Both <strong>the</strong> Matrimid <strong>membranes</strong> and <strong>the</strong> Kapton polyimide flat sheets were <strong>the</strong>n cut to adiameter <strong>of</strong> 47 mm using a punch and placed in <strong>the</strong> quartz tube furnace between two graphiteblocks.Under an UHP argon flow <strong>of</strong> 200 ml/min, <strong>the</strong> polymeric <strong>membranes</strong> were <strong>carbon</strong>ised at afinal pyrolysis <strong>temperature</strong> between 400 and 800°C <strong>for</strong> 2 hours. A typical <strong>temperature</strong> pr<strong>of</strong>ile<strong>of</strong> <strong>the</strong> quartz tube oven is shown in Figure 3-9 below. The argon purge gas was maintained atall times until <strong>the</strong> furnace had returned to room <strong>temperature</strong>.66


800700Oven HeatingOven Off (Natural Cooling)Furnace Temperature (°C)6005004003002001002 hours @ final pyrolysis <strong>temperature</strong>00 100 200 300 400 500 600 700 800 900 1000Time (min)Figure 3-9 : Quartz tube furnace typical <strong>temperature</strong> pr<strong>of</strong>ile with time during pyrolysis<strong>of</strong> unsupported NPC <strong>membranes</strong>.3.3 Membrane Characterisation TechniquesThe supported and unsupported <strong>membranes</strong> were all weighed be<strong>for</strong>e and after pyrolysis tocheck weight loss. Details <strong>of</strong> <strong>the</strong> surface structure and pore size distribution were obtainedusing <strong>the</strong> techniques described in <strong>the</strong> following sections.3.3.1 Surface StructureA Philips XL30 and a FEI Quanta FEG 200 scanning electron microscope (SEM) were <strong>use</strong>dto detect <strong>the</strong> presence <strong>of</strong> <strong>carbon</strong> and also to ensure <strong>the</strong> absence <strong>of</strong> micro-cracking. For <strong>the</strong>partially <strong>carbon</strong>ised unsupported <strong>membranes</strong>, which were not conductive, <strong>the</strong> samples werecoated in gold using a Dynavac mini sputter coater prior to using <strong>the</strong> SEM.High resolution transmission electron microscopy (HRTEM) was <strong>use</strong>d to estimate <strong>the</strong> poresize and shape <strong>of</strong> NPC manufactured at 600°C. Images were recorded using a Jeol-2010exelectron microscope equipped with a Gatan parallel electron energy loss spectrometer(PEELS), STEM and Gatan Imaging Filter (GIF) located at <strong>the</strong> Royal Melbourne Institute <strong>of</strong>Technology (RMIT). The microscope was operated at 200keV and most images wererecorded at 600k magnification. The spherical aberration coefficient <strong>of</strong> <strong>the</strong> ultra-<strong>high</strong>resolution pole-pieces was CS = 1.00 mm and <strong>the</strong> effective image resolution was 0.2 nm.67


3.3.2 Pore SizeAn Image J analysis <strong>of</strong> <strong>the</strong> HRTEM image <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> was undertaken to estimate<strong>the</strong> pore size distribution. Image J is a publicly available java-based image processingcomputer program developed by <strong>the</strong> American National Institutes <strong>of</strong> Health. The pore sizes <strong>of</strong>100 pores were measured individually using <strong>the</strong> measuring tool available in Image J and <strong>the</strong>ncalibrated back to a relative size <strong>of</strong> 1 nm.3.3.2.1 PALSPositron Annihilation Lifetime Spectroscopy (PALS) was also <strong>use</strong>d to indicate porosity and<strong>the</strong> pore size distribution <strong>of</strong> <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong> produced. In order to obtain enough NPC<strong>for</strong> <strong>the</strong> PALS analysis, a crucible <strong>of</strong> PFA was placed in <strong>the</strong> quartz tube furnace whilst making<strong>the</strong> <strong>membranes</strong>.The PALS experiment measures <strong>the</strong> lifetimes <strong>of</strong> positrons by detecting start quanta <strong>of</strong> 1.28MeV associated with <strong>the</strong>ir birth, and stop quanta <strong>of</strong> 0.511 MeV associated with <strong>the</strong>ir death(annihilation radiation) [150]. When a positron is emitted from a radioactive isotope, such as22Na, it is injected into <strong>the</strong> sample under consideration with <strong>high</strong> energy (up to 544 keV).Once in <strong>the</strong> sample, <strong>the</strong> positrons are <strong>the</strong>rmalised via inelastic collisions. Positrons canannihilate with electrons in <strong>the</strong> bulk <strong>of</strong> <strong>the</strong> material. Alternatively, as positrons are repelled by<strong>the</strong> constituent nuclei <strong>of</strong> <strong>the</strong> material and are annihilated by electrons, positrons can belocalised in voids or regions <strong>of</strong> lower than average electron density. As a positron is <strong>of</strong> similarsize as an electron, it is able to access small voids.The lifetime <strong>of</strong> <strong>the</strong> positron is dependent on <strong>the</strong> overlap <strong>of</strong> <strong>the</strong> positron wavefunction with <strong>the</strong>wavefunctions <strong>of</strong> <strong>the</strong> electrons in <strong>the</strong> material. There<strong>for</strong>e, as <strong>the</strong> probability <strong>of</strong> wavefunctionoverlap is decreased with increasing pore size, an increase in <strong>the</strong> average positron lifetime ismeasured [151]. In materials with <strong>high</strong> electron density, such as <strong>carbon</strong> <strong>membranes</strong>, twoseparate positron lifetimes are observed. The first lifetime (τ 1 ) is attributed to positrons beingannihilated in <strong>the</strong> bulk <strong>of</strong> <strong>the</strong> material (~150-280 ps). The second lifetime, τ 2 , is attributed topositrons being annihilated in defects (300-500 ps). The relative number <strong>of</strong> positrons beingannihilated in defects (I 2 ) compared to those annihilating in <strong>the</strong> bulk (I 1 ) gives in<strong>for</strong>mation on<strong>the</strong> number <strong>of</strong> defects in <strong>the</strong> material [150, 152, 153].Positron Annihilation Lifetime Spectroscopy (PALS) experiments were per<strong>for</strong>med using anEG&G Ortec fast-fast coincidence system with fast plastic scintillators and a resolutionfunction <strong>of</strong> 260 ps FWHM (60Co source with <strong>the</strong> energy windows set to 22Na events).Approximately 500 mg <strong>of</strong> each sample was packed on ei<strong>the</strong>r side <strong>of</strong> a 30 µCi 22NaCl foil68


(2.54 µm Ti) source. At least 5 spectra <strong>of</strong> one million integrated counts were collected wi<strong>the</strong>ach spectrum taking about 0.47 h to collect. Data analysis was per<strong>for</strong>med using LT9 [154].The spectra were best fitted with two components. No evidence <strong>of</strong> source irradiation damagewas observed and no source correction was <strong>use</strong>d.3.3.2.2 WAXDWide angle X-ray diffraction (WAXD) was also <strong>use</strong>d to determine <strong>the</strong> pore size distribution.X-ray diffractograms were measured using a Bruker Advance, D8 diffractometer operating in2θ mode, using copper Kα radiation at 40 mA, 40 kV at room <strong>temperature</strong> in air. Theinstrument was equipped with a graphite diffracted beam monochromators. The powderedsample was mounted in a standard Bruker sample holder, giving minimal background scatter.Diffractograms were measured at 0.02° steps over <strong>the</strong> range 2θ = 5° to 65° with a step time <strong>of</strong>3 s. Bragg’s Law (Equation 2-28) was <strong>use</strong>d to determine d-spacing from 2θ values.While not a direct indication <strong>of</strong> pore size, <strong>the</strong> d-spacing values give an indication <strong>of</strong> <strong>the</strong>average spacing between individual layers <strong>of</strong> <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong> structure [155].69


3.4 Measurement <strong>of</strong> Membrane Permeance and SelectivityPure gas per<strong>for</strong>mance measurements were initially made in order to screen <strong>the</strong> <strong>membranes</strong> <strong>for</strong>mixed gas testing and to determine <strong>the</strong> effect <strong>of</strong> pyrolysis <strong>temperature</strong> on per<strong>for</strong>mance. Mixedgas measurements were <strong>use</strong>d to determine <strong>the</strong> effect <strong>of</strong> operating <strong>temperature</strong> and <strong>the</strong>presence <strong>of</strong> impurities on <strong>the</strong> NPC per<strong>for</strong>mance.3.4.1 Pure GasThe permeance <strong>of</strong> pure CH 4 , N 2 and CO 2 (99.99%, BOC Gases, Australia) <strong>of</strong> each membranewas measured using a custom built constant volume, variable pressure (CVVP) apparatus[156] as shown in Figure 3-10 below(a)(b)ExhaustPGV2PTData LoggerV1V3Pure CH 4 , N 2 or CO 2OvenV4PSVV5PTV6Calibrated Volumewithin Water Bathat Constant TemperatureExhaustVacuumPumpFigure 3-10 : (a) Photograph and (b) process flow diagram <strong>of</strong> <strong>the</strong> constant volume,variable pressure (CVVP) apparatus <strong>for</strong> measuring pure gas permeance andpermselectivity70


The NPC membrane was placed in a purpose-built 44 mm membrane holder, which was fittedto <strong>the</strong> CVVP apparatus. The membrane unit (Figure 3-11) was ho<strong>use</strong>d within an oven(Scientific Equipment Manufacturers, Australia) to keep <strong>the</strong> <strong>temperature</strong> constant at 35°C.Similarly, a cylinder <strong>of</strong> calibrated volume (150 or 500 ml Swagelok, USA) was immersed in awater bath maintained at 27°C using a water bath circulator (Bioblock Scientific Polystat,France). Duthie et al [156] has shown that <strong>the</strong> errors associated with <strong>the</strong> <strong>temperature</strong> variationbetween <strong>the</strong> oven and <strong>the</strong> bath are small.Figure 3-11 : Photograph <strong>of</strong> <strong>the</strong> CVVP membrane cell showing one inlet connection <strong>for</strong><strong>the</strong> feed gas and one outlet connection <strong>for</strong> <strong>the</strong> permeate gasAs <strong>the</strong> first step, <strong>the</strong> vacuum connections (V4 and V5) were opened to allow <strong>the</strong> system to beevacuated. In order to draw a deep vacuum <strong>the</strong> system was <strong>of</strong>ten left to evacuate overnight.The vacuum connections were <strong>the</strong>n sealed <strong>of</strong>f and feed gas introduced on top <strong>of</strong> <strong>the</strong>membrane at a pressure (p f ). The pressure <strong>of</strong> <strong>the</strong> feed gas was measured using a 0 to 6985 kPapressure transducer (MKS Baratron, USA) and <strong>the</strong> pressure on <strong>the</strong> permeate side wasmeasured using a 0 to 1.33 kPa pressure transducer (MKS Baratron, USA). During <strong>the</strong>subsequent pressure rise on <strong>the</strong> permeate side, both pressure measurements were recordedevery second using a PC installed with Labview 6.2 (National Instruments, USA). A pressuresafety valve (PSV) was installed directly upstream <strong>of</strong> <strong>the</strong> permeate pressure transducer toprevent overpressure from <strong>the</strong> feed gas in <strong>the</strong> instance <strong>of</strong> a membrane failure.For each gas, <strong>the</strong> rise in pressure within <strong>the</strong> known permeate volume (V p ) was <strong>the</strong>n <strong>use</strong>d tocalculate <strong>the</strong> permeance (P’) as discussed earlier in Section 2.2.2 and as shown in Equation 3-1. It should also be noted that prior to introducing a feed gas, <strong>the</strong> vacuum valves were closedand <strong>the</strong> pressure rise recorded in <strong>the</strong> absence <strong>of</strong> a feed gas in order to calculate <strong>the</strong> leak rate <strong>of</strong><strong>the</strong> system which was <strong>the</strong>n subtracted from <strong>the</strong> pressure rise readings <strong>of</strong> each gas.71


P∆xV pp f− pp0t = P' t = lnEquation 3-1 [84]ART p − pfpThe calculated error in <strong>the</strong> pure gas permeance is +/- 1.4% and in <strong>the</strong> selectivity is +/- 2.0%.Details <strong>of</strong> <strong>the</strong> error analysis and sample calculations can be found in Appendix CC.1– Puregas permeance measurement.3.4.2 Mixed GasThe permeance <strong>of</strong> a mixed CO 2 :CH 4 or CO 2 :N 2 feed (10:90 vol%, BOC Gases Australia) wasmeasured using a custom-built constant pressure variable volume (CPVV) apparatus builtfrom ¼” stainless steel tubing and fittings (Swagelok USA) as shown in Figure 3-12 below.ExhaustPGFeedPGRetentateMF1 1BPC 1MFI 2ExhaustV1V2PGPGCH 4 /CO 2V3V4OvenTGV5V6ExhaustPGSweep GasPGTGPermeateMFC 1 MFI 3V7GC On-line SamplingLegendHeOn/Off ValveNeedle ValveMFCMFIMass Flow ControllerMass Flow IndicatorBPCBack Pressure ControllerFigure 3-12 : Process flow diagram <strong>of</strong> <strong>the</strong> constant pressure, variable volume apparatus(CPVV) <strong>for</strong> measuring mixed gas permeance and selectivity.Similar to <strong>the</strong> CVVP apparatus, <strong>the</strong> NPC membrane was placed in a custom-built 44 mmmembrane holder (Figure 3-13), which was fitted to <strong>the</strong> CPVV apparatus. The membrane unitwas ho<strong>use</strong>d within an oven (Scientific Equipment Manufacturers, Australia) to keep <strong>the</strong><strong>temperature</strong> <strong>of</strong> <strong>the</strong> membrane constant.72


Be<strong>for</strong>e introducing <strong>the</strong> mixed gas feed, <strong>the</strong> feed line was connected to a nitrogen cylinder(BOC Gases Australia) and <strong>the</strong> membrane regenerated overnight at 30 ml/min and 120°C toremove any water that had adsorbed onto <strong>the</strong> membrane from <strong>the</strong> atmosphere during handlingand installation.Figure 3-13 : Photograph <strong>of</strong> <strong>the</strong> CPVV membrane cell showing two inlet connections <strong>for</strong><strong>the</strong> feed and sweep gases and two outlet connections <strong>for</strong> <strong>the</strong> retentate and permeatestreams.The next morning, <strong>the</strong> N 2 was disconnected, valves V2, V3, V4, V5 and V6 were closed and<strong>the</strong> mixed gas feed introduced. Valve V1 was always set to flow towards <strong>the</strong> feed mass flowindicator (MFI 1) (Aalborg, USA) unless <strong>the</strong> feed line needed depressuring as is <strong>the</strong> case inshutdown. The pressure <strong>of</strong> <strong>the</strong> mixed gas feed was increased at <strong>the</strong> regulator to <strong>the</strong> desiredpressure, usually between 1000 and 2000 kPag.If <strong>the</strong> flowrate reading given by MFI 1 was not 0 ml/min, <strong>the</strong> fittings were tightened wherepossible. Valves V2 and V4 were <strong>the</strong>n opened to allow a flow through <strong>the</strong> retentate line. Thesetting <strong>of</strong> <strong>the</strong> retentate back pressure controller (BPC 1) (Aalborg USA) was adjustedaccordingly to give <strong>the</strong> desired retentate flow as shown by <strong>the</strong> retentate mass flow indicator(MFI 2) (Aalborg, USA), usually 100 ml/min.Helium sweep gas (BOC Gases Australia), was introduced by opening valve V5 and <strong>the</strong>nsetting <strong>the</strong> desired flow through <strong>the</strong> sweep gas mass flow controller (MFC 1) (Aalborg, USA).The purpose <strong>of</strong> <strong>the</strong> inert sweep gas was to reduce <strong>the</strong> partial pressure <strong>of</strong> <strong>the</strong> <strong>carbon</strong> dioxide in<strong>the</strong> permeate flow and hence increase <strong>the</strong> driving <strong>for</strong>ce. The sweep gas flow was optimised at20 ml/min. If <strong>the</strong> flow was much <strong>high</strong>er than this, it was difficult to obtain accurate readings73


<strong>of</strong> <strong>the</strong> permeate concentration from <strong>the</strong> gas chromatograph and if <strong>the</strong> flow was lower thanthis, <strong>the</strong>re was inadequate flow to <strong>the</strong> gas chromatograph.Finally, <strong>the</strong> permeate flow was initiated by opening valve V6. Valve V7 was always settowards <strong>the</strong> gas chromatograph unless recording <strong>the</strong> permeate flowrate (MFI 3) (AalborgUSA) in which case V7 was switched toward <strong>the</strong> exhaust system. Doing this preventedexcessive backpressure on MFI 3, which in turn gave a more accurate reading.The system was allowed 2 hours to equilibrate be<strong>for</strong>e <strong>the</strong> first readings were taken. During<strong>the</strong>se initial 2 hours, <strong>the</strong> gas chromatograph (GC) (Varian CP-3800 USA) fitted with both a<strong>the</strong>rmal conductivity detector (TCD) and flame ionisation detector (FID) was brought on-line.The GC was calibrated once per month with CH 4 , N 2 and CO 2 .Once all flow rates were steady and 2 hours had elapsed, <strong>the</strong> first readings were taken. Thiswas done by opening <strong>the</strong> integral sampling valve on <strong>the</strong> GC to allow <strong>the</strong> permeate gas to flowthrough <strong>the</strong> GC columns. A fur<strong>the</strong>r two permeate compositions were <strong>the</strong>n sampled in quicksuccession using <strong>the</strong> GC to ensure consistency <strong>of</strong> readings. Valve V7 was <strong>the</strong>n directed tovent and readings <strong>of</strong> <strong>the</strong> flowrate and pressure <strong>of</strong> each stream recorded. Fur<strong>the</strong>r GC samplesalong with pressure and flowrate readings were <strong>the</strong>n taken approximately every 30 to 60 mins<strong>for</strong> up to 8 hours.As described in Section 2.2.2, <strong>the</strong> permeance <strong>of</strong> each component was calculated from <strong>the</strong> flux(F) and <strong>the</strong> partial pressure drop (∆p). The flux was calculated from <strong>the</strong> permeate flow rate(Q p ), permeate component mol fraction (y) and membrane surface area (A). Similarly, <strong>the</strong>changes in partial pressure <strong>of</strong> each component were calculated from <strong>the</strong> feed (x) and permeate(y) component mol fractions and <strong>the</strong> pressure <strong>of</strong> <strong>the</strong> feed (p f ) and permeate streams (p p ) aspresented in Equation 3-2.P'AJ=∆pAA=A(xAQpfpyA− yApp)Equation 3-2The ideal CO 2 /CH 4 (α CO2/CH4 ) selectivity was calculated from <strong>the</strong> ratio <strong>of</strong> permeances asshown earlier in Section 2.2.1 as Equation 2.3. The CO 2 /CH 4 separation factor (S CO2/CH4 ) <strong>of</strong><strong>the</strong> membrane was calculated as shown earlier in Section 2.2.1 as Equation 2.5.The experimental conditions <strong>for</strong> which <strong>the</strong> NPC <strong>membranes</strong> were tested using a mixed gasfeed are shown in Table 3-5. These conditions were chosen as <strong>the</strong> best representation <strong>of</strong> CO 2separation from natural gas within <strong>the</strong> limits <strong>of</strong> <strong>the</strong> experimental equipment.74


Table 3-5 : Mixed gas testing conditionsStream Property Controlled? ValueFeed Composition Yes 10:90 mol% CO 2 :CH 4Feed Flow Yes 100 ml/minFeed Temperature Yes 35°CFeed Pressure Yes 500 to 2000 kPagSweep Gas Composition Yes 100 mol% HeSweep Gas Flow Yes 20 ml/minSweep Gas Temperature Yes 35°CSweep Gas Pressure Yes ~ 50 kPagRetentate Composition No Determined by system per<strong>for</strong>manceRetentate Flow No Determined by system per<strong>for</strong>manceRetentate Temperature Yes 35°CRetentate Pressure Yes 450 to 1950 kPagPermeate Composition No Determined by system per<strong>for</strong>mancePermeate Flow No Determined by system per<strong>for</strong>mancePermeate Temperature Yes 35°CPermeate Pressure Yes 0 kPagA more detailed operating procedure is given in Appendix BB.1 – The standard operatingprocedure <strong>for</strong> <strong>the</strong> CPVV (Mixed Gas) apparatus.The calculated error in <strong>the</strong> mixed gas permeance is +/- 5.2% and in <strong>the</strong> selectivity is +/- 7.3%.Details <strong>of</strong> <strong>the</strong> error analysis and sample calculations can be found in Appendix CC.2 – Mixedgas permeance measurement.3.4.3 At Elevated TemperaturesThe separation per<strong>for</strong>mance <strong>of</strong> mixed gases at elevated <strong>temperature</strong>s was measured byincreasing <strong>the</strong> oven <strong>temperature</strong>, in which <strong>the</strong> membrane unit was ho<strong>use</strong>d, from 35°C tobetween 75°C and 200°C. After each <strong>temperature</strong> increase, <strong>the</strong> system was allowed 1 hour toequilibrate be<strong>for</strong>e GC samples and readings were taken.Temperatures greater than 200 °C were not exceeded due to <strong>the</strong> pressure-<strong>temperature</strong>limitations <strong>of</strong> <strong>the</strong> piping and fittings <strong>use</strong>d. The <strong>temperature</strong> <strong>of</strong> <strong>the</strong> gas exiting <strong>the</strong> oven wasmeasured using a <strong>the</strong>rmocouple to ensure that <strong>the</strong> gas flowing through <strong>the</strong> mass flow metershad cooled back to room <strong>temperature</strong> again beca<strong>use</strong> <strong>of</strong> <strong>the</strong> <strong>temperature</strong>-pressure limitations <strong>of</strong><strong>the</strong> flow meters.75


3.4.4 Exposure to Impurities3.4.4.1 Condensable ImpuritiesThe permeance <strong>of</strong> a mixed CO 2 :CH 4 or CO 2 :N 2 feed (10:90 vol%, BOC Gases Australia) in<strong>the</strong> presence <strong>of</strong> condensable impurities was measured using <strong>the</strong> CPVV apparatus withmodification as shown in Figures 3-14 and 3-15.The CPVV (mixed gas) apparatus was started using <strong>the</strong> method as described previously inSection 3.4.2 and a baseline per<strong>for</strong>mance reading established. Condensable impurities were<strong>the</strong>n introduced into <strong>the</strong> mixed gas feed by directing <strong>the</strong> feed gas through valve V2 tostainless steel vessel (Swagelok, USA) containing 50 ml <strong>of</strong> liquid impurity (refer to Figure 3-14). The vessel was filled with steel ballotini and was equipped with a sparger to allow <strong>the</strong>gas to become saturated with <strong>the</strong> liquid impurity. The saturated gas was <strong>the</strong>n directed back to<strong>the</strong> membrane cell through a liquid droplet filter (Swagelok, USA) via valve V7. The systemwas allowed 30 mins to allow <strong>the</strong> permeate gas under <strong>the</strong> presence <strong>of</strong> impurities to reach <strong>the</strong>GC. The permeate composition along with <strong>the</strong> stream flowrates and pressures were recordedas described in Section 3.4.2. Measurements were <strong>the</strong>n taken every 30 mins until <strong>the</strong> values <strong>of</strong>permeance and selectivity had stabilised (usually less than 4 hours).Figure 3-14 : Photograph <strong>of</strong> <strong>the</strong> vessel <strong>use</strong>d to saturate <strong>the</strong> feed gas <strong>of</strong> <strong>the</strong> constantpressure variable volume (CPVV) apparatus <strong>for</strong> measuring mixed gas permeance andpermselectivity in <strong>the</strong> presence <strong>of</strong> condensable impurities.76


(a)ExhaustTo Condensable ImpuritiesFrom Condensable ImpuritiesPGFeedPGRetentateV1MFI 1V2V7BPC 1V8MFI 2ExhaustPGPGCH 4 /CO 2V9V10OvenTGV11V12ExhaustPGSweep GasPGTGPermeateMFC 1 MFI 3V13GC On-line SamplingLegendHeOn/Off ValveNeedle ValveMFCMFIMass Flow ControllerMass Flow IndicatorBPCBack Pressure Controller------------------------------------------------------------------------------------------------------(b)PGFilling Line <strong>for</strong> Condensable ImpurityLab Exhaust SystemV3V4GC Sample PtV5From FeedFilterTo Membrane UnitCondensable ImpurityV6LegendOn/Off ValveNeedle ValveCheck ValveMFCMFIBPCMass Flow ControllerMass Flow IndicatorBack Pressure ControllerFigure 3-15 : Process flow diagrams <strong>of</strong> (a) <strong>the</strong> constant pressure, variable volume(CPVV) apparatus <strong>for</strong> measuring mixed gas permeance and selectivity in <strong>the</strong> presence<strong>of</strong> condensable impurities (b) vessel <strong>for</strong> contacting feed gas with condensable impurities77


The impurities tested were water, hexane (99.8 %, Merck Germany) to represent <strong>the</strong> impact <strong>of</strong>heavier hydro<strong>carbon</strong>s and toluene (99.8 %, Merck Germany) to represent <strong>the</strong> impact <strong>of</strong>aromatic hydro<strong>carbon</strong>s. The concentration <strong>of</strong> impurity in <strong>the</strong> feed gas was controlled bysubmerging <strong>the</strong> vessel in a <strong>temperature</strong> controlled solution <strong>of</strong> glycol. The concentration <strong>of</strong>impurity (x IMP ) (mol %) was <strong>the</strong>n calculated from <strong>the</strong> partial pressure <strong>of</strong> <strong>the</strong> impurity (pp IMP )at <strong>the</strong> vessel <strong>temperature</strong> and <strong>the</strong> feed gas pressure (p f ) using Equation 3-3. As discussedpreviously in Section 2.2.1 Partial pressures ra<strong>the</strong>r than fugacities have been <strong>use</strong>d as <strong>the</strong>fugacity coefficients <strong>of</strong> <strong>the</strong> components at <strong>the</strong> vessel <strong>temperature</strong>s and pressure are greaterthan 0.95.ppIMPxIMP= Equation 3-3pfThe partial pressure <strong>of</strong> <strong>the</strong> impurity (pp IMP ) at <strong>the</strong> vessel <strong>temperature</strong> was determined from <strong>the</strong>HYSYS simulation package (Aspentech, USA) using <strong>the</strong> Peng-Robinson equation <strong>of</strong> state.The typical concentrations <strong>of</strong> impurities <strong>use</strong>d are presented in Table 3-6.Table 3-6 : The concentration <strong>of</strong> condensable impurities in <strong>the</strong> feed gas based on <strong>the</strong>operating <strong>temperature</strong> <strong>of</strong> <strong>the</strong> saturating vessel containing <strong>the</strong> liquid impurity.Component Vessel Temp (°C) Vapour Press (kPa) Impurity Concentration (ppm)Water 25 3.1 2390Water 10 1.2 918Water 5 0.9 650Hexane 25 20.2 15538Hexane 10 10.2 7844Hexane 5 7.8 6128Toluene 25 4.1 2549Toluene 10 1.8 1146Toluene 5 1.4 860Regeneration <strong>of</strong> <strong>the</strong> membrane after impurity exposure was completed using nitrogen at a<strong>temperature</strong> between 120 and 150 °C at a flow <strong>of</strong> 30 ml/min and pressure <strong>of</strong> 200 kPag <strong>for</strong>more than 12 hours.A detailed description <strong>of</strong> <strong>the</strong> operating procedure in <strong>the</strong> presence <strong>of</strong> condensable impurities isgiven in Appendix B2 – The standard operating procedure <strong>for</strong> <strong>the</strong> CPVV (mixed gas)apparatus in <strong>the</strong> presence <strong>of</strong> condensable impurities.78


3.4.4.2 Non-Condensable ImpuritiesThe permeance <strong>of</strong> a mixed CO 2 :CH 4 or CO 2 :N 2 feed (10:90 vol%, BOC Gases Australia) in<strong>the</strong> presence <strong>of</strong> condensable impurities was measured using <strong>the</strong> CPVV apparatus withmodification as shown in Figure 3-16.To expose <strong>the</strong> membrane to H 2 S or CO, two speciality bottles were purchased from BOCAustralia. The first bottle contained 500 ppm <strong>of</strong> H 2 S in a CO 2 :N 2 mixture (10:90 vol%) whilst<strong>the</strong> second contained 1000 ppm <strong>of</strong> CO in a CO 2 :N 2 mixture (10:90 vol%).The CPVV (mixed gas) apparatus was started using <strong>the</strong> method as described previously inSection 3.4.2 using a CO 2 :N 2 mixture (10:90 vol%, BOC Gases Australia) and a baselineper<strong>for</strong>mance reading established. Valves V5, V6, V7 and V8 were <strong>the</strong>n closed to keep <strong>the</strong>membrane at pressure. The CO 2 :N 2 mixture (10:90 vol%) cylinder was <strong>the</strong>n closed and <strong>the</strong>feed line up to valve V5 depressured by venting through V2. Valve V3 was <strong>the</strong>n directedtoward valve V2. Locked valve V1 was opened and <strong>the</strong> pressure at <strong>the</strong> regulator on <strong>the</strong>impurity cylinder (containing ei<strong>the</strong>r H 2 S or CO) was slowly increased to <strong>the</strong> desired valve.The valves V4, V5 and V6 were <strong>the</strong>n opened and BPC 1 adjusted so that a flow <strong>of</strong> 100ml/min was seen on MFI 2. Sweep gas was introduced through V7 and <strong>the</strong> permeate flowestablished by opening V8 and directing V9 to <strong>the</strong> GC.ExhaustGlove BoxPGV1 (Locked)V2V3FeedMFI 1PGPGBPC 1V4PGRetentateMFI 2ExhaustCO 2 /N 2 /H 2 SPGV5V6OvenExhaustTGN 2 or CO 2 /N 2V7V8ExhaustPGSweep GasPGTGPermeateMFC 1 MFI 3V9GC On-line SamplingLegendHeOn/Off ValveNeedle ValveMFCMFIMass Flow ControllerMass Flow IndicatorRelief ValveBPCBack Pressure ControllerFigure 3-16 : Process flow diagram <strong>of</strong> <strong>the</strong> constant pressure, variable volume apparatus(CPVV) <strong>for</strong> measuring mixed gas permeance and selectivity in <strong>the</strong> presence <strong>of</strong> noncondensableimpurities.79


The system was allowed 30 mins to allow <strong>the</strong> permeate gas under <strong>the</strong> presence <strong>of</strong> impuritiesto reach <strong>the</strong> GC. The permeate composition along with <strong>the</strong> stream flowrates and pressureswere recorded as also described in Section 3.4.2. Measurements were <strong>the</strong>n taken every 30mins until <strong>the</strong> values <strong>of</strong> permeance and selectivity had stabilised (usually less than 4 hours).Regeneration <strong>of</strong> <strong>the</strong> membrane after impurity exposure was completed using nitrogen at a<strong>temperature</strong> between 120 and 150 °C at a flow <strong>of</strong> 30 ml/min and pressure <strong>of</strong> 200 kPag <strong>for</strong>more than 12 hours. The CO 2 :N 2 mixture (10:90 vol%) without <strong>the</strong> impurity was <strong>the</strong>nreconnected and <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> <strong>the</strong> membrane tested to evaluate <strong>the</strong> effectiveness <strong>of</strong>regeneration.The gases H 2 S and CO are both very toxic and can ca<strong>use</strong> death quickly even when in very lowconcentrations, such as those present in <strong>the</strong> gas bottles <strong>use</strong>d <strong>for</strong> <strong>the</strong> experiments. A detailedrisk assessment was undertaken be<strong>for</strong>e <strong>the</strong>se gases were <strong>use</strong>d in <strong>the</strong> lab. Five scenarios inwhich H 2 S or CO could leak into <strong>the</strong> lab were identified from <strong>the</strong> risk assessment. For eachscenario <strong>the</strong> volumetric rate <strong>of</strong> H 2 S or CO leaked into <strong>the</strong> lab was calculated using <strong>the</strong>universal gas equation <strong>for</strong> flow through an orifice [157]. Assuming ideal mixing, <strong>the</strong>concentration <strong>of</strong> H 2 S or CO in <strong>the</strong> lab with time was calculated. From <strong>the</strong> knownconcentration effects <strong>of</strong> H 2 S and CO, <strong>the</strong> timing to an alarm, unconscious state and fatalitycould be determined.For each scenario, <strong>the</strong> risk <strong>of</strong> each scenario occurring was significantly reduced throughengineering and operator training. Engineering risk reduction measures included installing anextraction system around <strong>the</strong> gas bottle and from <strong>the</strong> oven (refer to Figure 3-17 below), a keylocksystem on <strong>the</strong> gas bottle, replacing all existing nylon seals and gaskets within <strong>the</strong> CVPPapparatus with Kalrez – a H 2 S resistant material and minimising fittings. Operators weretrained using a detailed operating procedure and a register <strong>of</strong> those trained was kept with <strong>the</strong>operating procedure. The operating pressure <strong>of</strong> <strong>the</strong> rig did not exceed 1000 kPag to minimise<strong>the</strong> volume leaked in <strong>the</strong> event <strong>of</strong> a leak. The operators were also trained to return <strong>the</strong> gasbottle containing ei<strong>the</strong>r H 2 S or CO to <strong>the</strong> outdoor gas cage when not in <strong>use</strong> <strong>for</strong> more than 2consecutive days and advise all lab personnel if ei<strong>the</strong>r gas was in <strong>use</strong>. A buddy system wasfollowed at all times.80


Figure 3-17 : Photograph <strong>of</strong> <strong>the</strong> gas extraction system installed around <strong>the</strong> gas bottlecontaining H 2 S or CO impurities.In <strong>the</strong> unlikely event <strong>of</strong> a leak occurring, emergency procedures were also developed and alllab personnel were trained to follow <strong>the</strong>se procedures. These procedures included immediateevacuation <strong>of</strong> <strong>the</strong> lab upon sounding <strong>of</strong> one <strong>of</strong> two alarms <strong>for</strong> detecting H 2 S and CO alongwith availability and training <strong>of</strong> all lab personnel in <strong>the</strong> <strong>use</strong> <strong>of</strong> self-contained breathingapparatus’ (SCBA) <strong>for</strong> <strong>use</strong> when re-entering <strong>the</strong> lab. In addition to two <strong>of</strong> <strong>the</strong> lab personnelbeing trained in first aid, first aid was also available within minutes via <strong>the</strong> security system.A detailed description <strong>of</strong> <strong>the</strong> operating procedure in <strong>the</strong> presence <strong>of</strong> non-condensableimpurities is given in Appendix B.3 – The standard operating procedure <strong>for</strong> <strong>the</strong> CPVV (mixedgas) apparatus in <strong>the</strong> presence <strong>of</strong> non-condensable impurities.The details <strong>of</strong> <strong>the</strong> risk assessment, leakage calculations and safety precautions implementedare presented in Appendix B.4 – Risk assessment <strong>for</strong> <strong>the</strong> operation <strong>of</strong> <strong>the</strong> CPVV (mixed gas)apparatus in <strong>the</strong> presence <strong>of</strong> non-condensable impurities.81


3.5 Measurement <strong>of</strong> Membrane AdsorptionThe adsorption <strong>of</strong> gases onto NPC <strong>membranes</strong> was calculated <strong>the</strong>oretically using a molecularsimulation and measured experimentally using a microbalance.3.5.1 Theoretical AdsorptionMolecular modelling (GCMC simulation) was undertaken to obtain simulated adsorptioniso<strong>the</strong>rms <strong>of</strong> pure CO 2 , N 2 and CH 4 along with <strong>the</strong> mixtures <strong>of</strong> CO 2 :N 2 (10:90 vol%) andCO 2 /CH 4 (10:90 vol%) on <strong>nanoporous</strong> <strong>carbon</strong> at various <strong>temperature</strong>s.3.5.1.1 Model DevelopmentThe molecular model <strong>use</strong>d was developed by Dr Jianwen Jiang and Pr<strong>of</strong>essor Stanley Sandlerat <strong>the</strong> University <strong>of</strong> Delaware, USA.3.5.1.2 Model DescriptionMonte Carlo simulations were carried out in <strong>the</strong> grand canonical ensemble <strong>of</strong> constantchemical potential, volume and <strong>temperature</strong> (µVT) to determine <strong>the</strong> number <strong>of</strong> molecules (N)adsorbed from <strong>the</strong> bulk phase onto <strong>the</strong> substrate.An adsorption iso<strong>the</strong>rm was generated by running <strong>the</strong> model at <strong>the</strong> same <strong>temperature</strong> andcomposition at various pressure points. All pressures quoted in Chapter 5 are absolute (i.e.kPaa).A summary <strong>of</strong> <strong>the</strong> pressure and <strong>temperature</strong> conditions modelled using GCMC simulation ispresented in Table 3-7 below.Table 3-7 : Details <strong>of</strong> <strong>the</strong> adsorbents, <strong>temperature</strong> and pressures <strong>use</strong>d <strong>for</strong> obtainingsimulated adsorption iso<strong>the</strong>rms <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> using GCMC simulation.Condition 1 Condition 2 Pressure PointsAdsorbent Temperature (°C) (kPaa)CO 2 27 0.01N 2 35 0.10CH 4 75 1.00CO 2 /N 2 100 10.00CO 2 /CH 4 125 100.00150 1000.00175 10000.00200 100000.0082


Inputs to <strong>the</strong> model included <strong>the</strong> coordinates <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> (modelled as C 168Schwarzite as shown in Figure 3-18), <strong>the</strong> Lennard-Jones potential model to provide <strong>the</strong> gasgasmolecule interactions and <strong>the</strong> ab-initio potential model to provide <strong>the</strong> gas-<strong>carbon</strong>interactions.Figure 3-18 : Structure <strong>of</strong> C 168 Schwarzite from <strong>the</strong> side (left) and rotated at 20°C (right)The GCMC simulation model developed by Dr Jianwen Jiang was written in FORTRAN 90and was run on a <strong>high</strong> speed UNIX machine using a standard FORTRAN compiler. The errorin <strong>the</strong> statistical average <strong>of</strong> <strong>the</strong> number <strong>of</strong> molecules adsorbed was calculated using <strong>the</strong> blocktrans<strong>for</strong>mation technique [106, 112].Fur<strong>the</strong>r details on <strong>the</strong> inputs to <strong>the</strong> model are presented in Appendix C.4.3.5.1.3 Fitting <strong>the</strong> Adsorption Iso<strong>the</strong>rmsThe simulated adsorption iso<strong>the</strong>rms <strong>of</strong> <strong>the</strong> pure gases were fitted using <strong>the</strong> Langmuir modeland <strong>the</strong> heat <strong>of</strong> adsorption (∆H ads ) was calculated from <strong>the</strong> Langmuir constant, b, as describedpreviously in Section 2.2.2.3.3.5.1.4 Calculating <strong>the</strong> Adsorption Separation FactorThe adsorption separation factor <strong>of</strong> <strong>the</strong> mixed gases at various operating <strong>temperature</strong>s andpressures was calculated from <strong>the</strong> component (A and B) proportions in <strong>the</strong> bulk phase (x) and<strong>the</strong> adsorbed phase (y) as shown in Equation 3-4.SyA/ yB= Equation 3-4(x / xads A / B)AB83


3.5.2 Experimental AdsorptionThe adsorption iso<strong>the</strong>rms <strong>of</strong> pure gases on <strong>nanoporous</strong> <strong>carbon</strong> were measured using aGravimetric High Pressure (GHP-300) analyser fitted with a Cahn microbalance (VTICorporation, USA) as shown in Figure 3-19. The GHP-300 is a gravimetric analyser thatoperates at both <strong>high</strong> and low <strong>temperature</strong>s and pressures. Some measurements were conductedin ho<strong>use</strong>, but due to equipment failure o<strong>the</strong>r results were obtained on a comparable machine at<strong>the</strong> VTI site in <strong>the</strong> USA.(a)(b)MicrobalancePTWater CooledFurnaceV8Reference SideV7MFMVPTVacuumPumpV9V6ConstantTemperatureWater BathSampleV5VentLegendV1 V2 V3 V4Pneumatic ValveVessel <strong>for</strong>OrganicVapourExposurePorts <strong>for</strong> Adsorbate GasesMFMVSolenoid ValveMass Flow Metering ValveFigure 3-19 : (a) Photograph and (b) schematic diagram <strong>of</strong> <strong>the</strong> microbalance <strong>use</strong>d <strong>for</strong>measuring gas adsorption iso<strong>the</strong>rms at elevated <strong>temperature</strong>s.84


Due to time constraints it was only possible to complete pure gas iso<strong>the</strong>rms at two <strong>temperature</strong>sas presented in Table 3-8 below.Table 3-8 : Details <strong>of</strong> <strong>the</strong> adsorbents, <strong>temperature</strong>s and pressures <strong>use</strong>d <strong>for</strong> obtainingexperimental adsorption iso<strong>the</strong>rms <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong>.Condition 1 Condition 2 Pressure Points Point TypeAdsorbent Temperature (°C) (kPaa +/- 5)CO 2 25 (N 2 only) 0 AdsorptionN 2 35 (CO 2 only) 41 Adsorption200 67 Adsorption131 Adsorption264 Adsorption520 Adsorption777 Adsorption968 Adsorption1283 Adsorption1522 Adsorption1280 Desorption1023 Desorption517 Desorption260 Desorption131 DesorptionTo obtain <strong>the</strong> adsorption iso<strong>the</strong>rms <strong>for</strong> this research work, <strong>the</strong> GHP-300 was operated in staticmode. A sample <strong>of</strong> bulk <strong>nanoporous</strong> <strong>carbon</strong> that had been manufactured in a crucible at 550°C<strong>for</strong> a soak time <strong>of</strong> 2 hours under an ultra <strong>high</strong> purity argon purge at 200 ml/min was placed in<strong>the</strong> balance. Due to equipment limitations, <strong>the</strong> maximum sample size allowed was ~100 mg.The sample was evacuated and <strong>the</strong>n heated at 350 °C <strong>for</strong> 3 hours to dry <strong>the</strong> sample as much aspossible. The sample was <strong>the</strong>n cooled to <strong>the</strong> experimental <strong>temperature</strong> and <strong>the</strong> adsorptioniso<strong>the</strong>rm constructed by measuring <strong>the</strong> weight change with increasing pressure. Likewise, <strong>the</strong>desorption iso<strong>the</strong>rms was constructed by measuring <strong>the</strong> weight reduction with decreasingpressure.3.5.2.1 Buoyancy CorrectionDuring adsorption and desorption, <strong>the</strong> sample was buoyed by <strong>the</strong> gas in <strong>the</strong> sample chamber,which led to a reduction in <strong>the</strong> weight change measured. The weight changes measured<strong>the</strong>re<strong>for</strong>e need to be corrected to remove <strong>the</strong> buoyancy effect.85


The buoyancy correction was made using ei<strong>the</strong>r <strong>of</strong> <strong>the</strong> following two methods.3.5.2.1.1 Buoyed Volume from Helium AdsorptionFollowing an adsorption experiment, <strong>the</strong> sample chamber was fully evacuated and dried <strong>for</strong> 30mins at 150°C. The sample was <strong>the</strong>n heated or cooled to <strong>the</strong> experimental <strong>temperature</strong>, whichwas <strong>the</strong> same as <strong>the</strong> preceding adsorption experiment. An adsorption iso<strong>the</strong>rm using helium as<strong>the</strong> adsorptive gas was <strong>the</strong>n obtained. On <strong>the</strong> assumption that helium doesn’t adsorb to <strong>the</strong><strong>nanoporous</strong> <strong>carbon</strong>, <strong>the</strong> weight changes observed were attributed to <strong>the</strong> buoyancy <strong>of</strong> <strong>the</strong> sample.The “helium” adsorption iso<strong>the</strong>rms constructed show a linear decrease in <strong>the</strong> weight withincreasing pressure as <strong>the</strong> sample became more buoyant.The buoyed mass (per unit pressure P) was <strong>the</strong>n calculated from <strong>the</strong> gradient <strong>of</strong> <strong>the</strong> straight linefit <strong>of</strong> <strong>the</strong> helium iso<strong>the</strong>rm. The buoyed mass (m) was <strong>the</strong>n converted to a buoyed volume (V)using <strong>the</strong> ideal gas law (Equation 3-5) at <strong>the</strong> <strong>temperature</strong> (T) <strong>of</strong> <strong>the</strong> helium adsorptionexperiment. The compressibility factor (z) allows <strong>for</strong> non-ideality <strong>of</strong> <strong>the</strong> gas and is particularlyimportant at <strong>high</strong>er <strong>temperature</strong>s and pressures. R is <strong>the</strong> ideal gas constant.mRTV = zEquation 3-5MWPThe buoyed volume was <strong>the</strong>n <strong>use</strong>d to correct <strong>the</strong> original adsorption iso<strong>the</strong>rm by converting <strong>the</strong>weights measured at each pressure to volumes using Equation 3-5 with a compressibility factor(Z), adding <strong>the</strong> buoyed volume and converting back to a final mass again using Equation 3-5.3.5.2.1.2 Buoyed Volume from Equipment MassesThe buoyed volume can also be obtained from <strong>the</strong> difference in <strong>the</strong> weight <strong>of</strong> <strong>the</strong> microbalancesample and reference sides (as shown previously in Figure 3-19) provide <strong>the</strong> two sides are at <strong>the</strong>same <strong>temperature</strong>. The weights <strong>of</strong> each component <strong>of</strong> <strong>the</strong> sample and reference side areconverted to volumes using standard densities as given in Table 3-9. The sum <strong>of</strong> <strong>the</strong> volumes on<strong>the</strong> reference side is <strong>the</strong>n subtracted from <strong>the</strong> sum <strong>of</strong> <strong>the</strong> volume on <strong>the</strong> sample side to give <strong>the</strong>buoyed volume.The buoyed volume is <strong>the</strong>n <strong>use</strong>d to correct <strong>the</strong> original adsorption iso<strong>the</strong>rm as discussedpreviously in Section 3.5.2.1.1.The calculated error in adsorption iso<strong>the</strong>rms is +/- 8.35 %. Details <strong>of</strong> <strong>the</strong> error analysis andsample calculations can be found in Appendix C.3 – Adsorption measurement.86


Table 3-9 : Equipment mass and densities <strong>use</strong>d to calculate buoyed volume. Theequipment masses and densities were provided by <strong>the</strong> supplier (VTI). The density <strong>of</strong><strong>nanoporous</strong> <strong>carbon</strong> was assumed to be <strong>the</strong> standard value <strong>of</strong> 1.6 g/cm 3 [79].Side Equipment MassDensity @35°C(g) (g/cm 3 )Reference SS 316 Fittings 0.6303 7.99Sample SS 316 Hang Downwire 0.1054 7.99Sample SS 316 Holder Bail 0.0119 7.99Sample Aluminium Sample Holder 0.4917 2.72Sample Nanoporous Carbon Varied with Experiment 1.60 [79]3.5.2.2 Fitting <strong>the</strong> Adsorption Iso<strong>the</strong>rmsThe experimental adsorption iso<strong>the</strong>rms were fitted using <strong>the</strong> Langmuir model and <strong>the</strong> heats <strong>of</strong>adsorption (∆H ads ) were calculated from <strong>the</strong> Langmuir constant, b, as described previously inSection 2.2.2.3.87


4 MANUFACTURE AND HIGH TEMPERATURE OPERATION OF NPCMEMBRANES4.1 IntroductionThe manufacture and characterisation <strong>of</strong> NPC <strong>membranes</strong> along with <strong>the</strong> pure gas and mixedgas per<strong>for</strong>mance results at <strong>high</strong> operating <strong>temperature</strong>s will be presented in this chapter. Bothsupported NPC <strong>membranes</strong> and modified-supported NPC <strong>membranes</strong> are considered. Theper<strong>for</strong>mance results <strong>for</strong> unsupported <strong>membranes</strong> are presented in Appendix D.4.2 Supported NPC Membranes4.2.1 ManufactureSupported NPC <strong>membranes</strong> were manufactured from <strong>the</strong> pyrolysis <strong>of</strong> polyfurfuryl alcohol asdescribed in Section 3.2.2.1 at pyrolysis <strong>temperature</strong>s between 400 and 600°C. Between 2 and 4coats <strong>of</strong> <strong>the</strong> polymeric precursor were required to produce a selective membrane. During <strong>the</strong> 1stcoat, <strong>the</strong> polymer filled <strong>the</strong> pores <strong>of</strong> <strong>the</strong> support and <strong>carbon</strong> was not yet seen on <strong>the</strong> surface. Itwas upon <strong>the</strong> 2nd coat that <strong>carbon</strong> appeared on <strong>the</strong> surface. If <strong>the</strong> <strong>carbon</strong> upon <strong>the</strong> surface <strong>of</strong> <strong>the</strong>support had not <strong>for</strong>med a complete layer after <strong>the</strong> 2nd coat <strong>the</strong>n a 3rd and sometimes 4th coatwas required. The creation <strong>of</strong> a membrane is depicted in Figure 4-1. The approximate volume <strong>of</strong>polymer solution added to each coat via spin-coating and <strong>the</strong> resulting <strong>carbon</strong> mass added afterpyrolysis is presented in Table 4-1.Figure 4-1 : The creation <strong>of</strong> a NPC membrane within 3 coats <strong>of</strong> <strong>the</strong> polymeric precursorTable 4-1 : Volumes <strong>of</strong> polymer solution (40 wt% PFA in acetone) added to each coat viaspin-coating at 1000 rpm and <strong>the</strong> resulting <strong>carbon</strong> mass added after pyrolysis.Coat Volume <strong>of</strong> Polymer Solution (ml) Carbon Mass Added After Pyrolysis (mg/cm 2 )1 4 – 9 2.5 – 6.82 2 – 3 1.5 – 2.53 1 0.84 0.5 – 1 0.489


All <strong>of</strong> <strong>the</strong> <strong>membranes</strong> reported in this chapter were manufactured using <strong>the</strong> <strong>high</strong>er molecularweight PFA received from Polysciences. The mass <strong>of</strong> <strong>carbon</strong> added <strong>for</strong> each membrane ispresented below in Table 4-2.Table 4-2 : Manufacture results <strong>of</strong> <strong>the</strong> <strong>membranes</strong> made from <strong>the</strong> spin-coating <strong>of</strong> PFAonto stainless steel supports and <strong>the</strong>n pyrolysised between 400 and 600°C.MembranePyrolysisTemperature# <strong>of</strong>CoatsCumulative Carbon Mass Added Per Coat(mg/cm 2 )°C 1 2 3 4 TOTALA 400 2 6.23 7.04 7.04B 400 2 6.51 7.58 7.58C 400 3 5.56 7.38 7.87 7.87D 400 4 4.69 6.71 7.57 8.77 8.77A 450 2 4.14 5.89 5.89B 450 2 4.88 6.15 6.15C 450 4 3.22 5.44 6.31 6.91 6.91A 500 2 3.64 5.29 5.29B 500 2 3.45 6.10 6.10C 500 3 4.81 5.81 6.13 6.13D 500 3 6.78 7.64 8.12 8.12A 550 2 2.99 5.18 5.18B 550 4 3.05 3.86 4.52 5.19 5.19C 550 3 2.89 4.46 5.22 5.22D 550 4 3.60 5.11 5.15 5.58 5.58E 550 2 4.09 6.07 6.07F 550 3 3.09 5.27 6.88 6.88A 600 2 2.84 4.62 4.62B 600 3 2.79 4.85 5.73 5.73C 600 4 3.11 4.42 5.31 6.28 6.28A graphical representation <strong>of</strong> <strong>the</strong> cumulative <strong>carbon</strong> mass achieved <strong>for</strong> each coat <strong>for</strong> <strong>the</strong><strong>membranes</strong> made is presented in Figure 4-2. As is subsequently discussed in Section 4.2.3, aselective membrane was <strong>for</strong>med when <strong>the</strong> total <strong>carbon</strong> added was between 4 and 8 mg/cm 2 .Membranes made at <strong>high</strong>er pyrolysis <strong>temperature</strong>s were selective with a lower total amount <strong>of</strong><strong>carbon</strong> added whilst <strong>membranes</strong> membrane made at <strong>the</strong> lower pyrolysis <strong>temperature</strong>s wereselective with a <strong>high</strong>er total amount <strong>of</strong> <strong>carbon</strong> added. The effect <strong>of</strong> pyrolysis <strong>temperature</strong> on <strong>the</strong><strong>carbon</strong> structure and resulting per<strong>for</strong>mance <strong>of</strong> <strong>the</strong> NPC <strong>membranes</strong> is fur<strong>the</strong>r discussed in <strong>the</strong>following sections.90


Rajagopalan and Foley @ 600 °C 400°C 450°C 500°C 550°C 600°C10.009.008.00Carbon Mass (mg/cm 2 )7.006.005.004.003.002.001.000.000 1 2 3 4Number <strong>of</strong> CoatsFigure 4-2 : The cumulative <strong>carbon</strong> density per coat <strong>for</strong> NPC <strong>membranes</strong> manufactured atpyrolysis <strong>temperature</strong>s between 400 and 600°C. Also depicted in this figure is <strong>the</strong>cumulative <strong>carbon</strong> density <strong>of</strong> an NPC membrane made by Rajagopalan and Foley at600°C [82] with a final O 2 /N 2 permselectivity <strong>of</strong> 5.4.2.2 Characterisation4.2.2.1 Surface StructureScanning electron microscopy (SEM) was <strong>use</strong>ful as an initial characterisation tool <strong>for</strong>confirming <strong>the</strong> presence <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> on <strong>the</strong> surface <strong>of</strong> <strong>the</strong> support and any cracking.SEM pictures <strong>of</strong> a <strong>nanoporous</strong> <strong>carbon</strong> surface with and without cracks are presented in Figure 4-3 (a).Likewise, <strong>high</strong> resolution transmission electron microscopy HRTEM was very <strong>use</strong>ful in <strong>the</strong>early stages <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> manufacture to identify that <strong>the</strong> correct pore size range toenable molecular sieving was being produced. Figure 4-3 (b) shows a HRTEM picture <strong>of</strong> <strong>the</strong>produced <strong>nanoporous</strong> <strong>carbon</strong> clearly showing <strong>the</strong> presence <strong>of</strong> pores less than 1 nm.91


(a)Uncoated Stainless Steel Support3 Coats <strong>of</strong> NPC on SS SupportMicroscopic CrackingMacroscopic Cracking(b)Larger Detail <strong>of</strong> Nanoporous CarbonPore Size Detail <strong>of</strong> Nanoporous Carbon0 1 2 3 4 nmFigure 4-3 : (a) SEM pictures showing (clockwise from top left) <strong>the</strong> uncoated stainless steelsupport, <strong>the</strong> <strong>carbon</strong> membrane after three coats with no defects, macroscopic cracking in<strong>the</strong> <strong>carbon</strong> surface and microscopic cracking in <strong>the</strong> <strong>carbon</strong> surface, (b) HRTEM picturesshown (from left) <strong>nanoporous</strong> <strong>carbon</strong> and enlarged photo to show pore size detail.92


4.2.2.2 Pore SizeAn estimate <strong>of</strong> <strong>the</strong> pore size distribution was obtained from measuring <strong>the</strong> size <strong>of</strong> 100 poreswithin a random selection <strong>of</strong> <strong>the</strong> HRTEM picture <strong>of</strong> NPC produced at 600°C using <strong>the</strong> imageprocessing program Image J. The estimated pore size distribution range is from 2.7 to 5.2 Åcentred on 4.0Å as shown in Figure 4-4. This result is consistent with <strong>the</strong> findings <strong>of</strong> Shiflett[89] who <strong>use</strong>d digitally enhanced HRTEM images <strong>of</strong> NPC to calculate pores within <strong>the</strong> range <strong>of</strong>3 to 6 Å. The basis statistics <strong>of</strong> <strong>the</strong> pore size distribution are presented in Table 4-3.1410012# <strong>of</strong> PoresCummulative # <strong>of</strong> PoresMean = 4.0 AMedian = 4.0 A9080# <strong>of</strong> Pores108647060504030Cummulative # <strong>of</strong> Pores20210002.00 2.50 3.00 3.50 4.00 4.50 5.00 5.50 6.00Pore Size (A)Figure 4-4 : Pore size distribution <strong>of</strong> NPC produced at 600°C obtained from <strong>the</strong> HRTEMpicture using <strong>the</strong> image processing program Image J. The distribution range is from 2.7 to5.2 Å centred around 4.0 Å.Table 4-3 : Basic Statistics <strong>of</strong> <strong>the</strong> pore size distribution <strong>of</strong> NPC produced at 600°Cobtained from <strong>the</strong> HRTEM picture using <strong>the</strong> image processing program Image J.Statistic Unit Value# <strong>of</strong> Samples 100Minimum Å 2.70Maximum Å 5.20Range Å 2.50Mean Å 3.96Median Å 4.00Mode Å 3.60Standard Deviation Å 0.5693


Two o<strong>the</strong>r techniques, PALS and WAXD, were <strong>use</strong>d to characterise <strong>the</strong> pores <strong>of</strong> <strong>nanoporous</strong><strong>carbon</strong> with increasing pyrolysis <strong>temperature</strong>.The PALS results shown in Figure 4-5 indicate an increase in <strong>the</strong> number <strong>of</strong> pores (increasingnumber <strong>of</strong> positrons annihilating in pores, I 2 ) and decrease in <strong>the</strong> average pore size (decreasinglifetime <strong>of</strong> positrons annihilating in pores, τ 2 ) with increasing pyrolysis <strong>temperature</strong>.Figure 4-5 : Characterisation <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> material manufactured at pyrolysis<strong>temperature</strong>s from 400 to 600°C using positron annihilation lifetime spectroscopy (PALS).The primary x-axis (τ 2 ) is <strong>the</strong> lifetime <strong>of</strong> position annihilating in <strong>the</strong> pores, which indicatesa decrease in pore size with increasing pyrolysis <strong>temperature</strong>. The secondary y-axis (I 2 ) is<strong>the</strong> number <strong>of</strong> positrons annihilating in <strong>the</strong> pores, which indicates an increase in <strong>the</strong>number <strong>of</strong> pores with increasing pyrolysis <strong>temperature</strong>.Likewise, <strong>the</strong> WAXD results shown in Figure 4-6 tell a similar story. In general, <strong>the</strong> threesamples show <strong>high</strong> intensity between 2θ = 12° and 2θ = 27° and low intensity above 2θ = 35°.The lack <strong>of</strong> a single peak suggests a very wide distribution <strong>of</strong> d-spacings <strong>for</strong> <strong>the</strong> 400 and 600°Csamples, whereas at 800°C a more crystalline structure is evidenced by <strong>the</strong> peaks at 2θ = 23°and 40°. The intensity <strong>for</strong> <strong>the</strong> region 2θ = 12° to 35° is greatest <strong>for</strong> <strong>the</strong> sample pyrolysed at800°C and least <strong>for</strong> <strong>the</strong> sample pyrolysed at 400°C. This result suggests increasing poreconcentration with increased pyrolysis <strong>temperature</strong>, as concluded by <strong>the</strong> PALS results. For <strong>the</strong>sample pyrolysed at 400°C <strong>the</strong> peak in intensity occurs at 2θ = 18.8° or a d-spacing <strong>of</strong> 4.7 Å(marked in Figure 4-6). The sample pyrolysed at 600°C has a corresponding intensity peak at ad-spacing <strong>of</strong> 4.1Å while <strong>for</strong> <strong>the</strong> sample pyrolysed at 800°C this peak occurs at a d-spacing <strong>of</strong>94


3.9 Å. The d-spacing is thus decreasing with increasing pyrolysis <strong>temperature</strong>, in agreementwith <strong>the</strong> PALS results and HRTEM analysis.800700400°C 600°C 800°CCounts6005004004.7 A3.9 A4.1 A300200100012 17 22 27 32 37 42 47 52 57 622θ (°)Figure 4-6 : Characterisation <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> material manufactured at pyrolysis<strong>temperature</strong>s <strong>of</strong> 400, 600 and 800°C using wide angle x-ray diffraction (WAXD). Theprimary x-axis is <strong>the</strong> diffraction angle whilst number <strong>of</strong> counts or intensity is given on <strong>the</strong>primary y-axis. The values <strong>of</strong> 2θ are converted to d-spacing using Bragg’s Law.PALS has been <strong>use</strong>d previously by Tin et al [96] to characterise unsupported <strong>carbon</strong> <strong>membranes</strong>made from <strong>the</strong> pyrolysis <strong>of</strong> <strong>the</strong> polyimides Matrimid and P84 at 800°C. Table 4-4 depicts <strong>the</strong>values <strong>of</strong> τ 2 and I 2 determined <strong>for</strong> <strong>the</strong>se <strong>membranes</strong> in comparison with <strong>the</strong> <strong>membranes</strong> <strong>of</strong> thisresearch work. Whilst it is difficult to compare directly as different polymers and pyrolysisconditions have been <strong>use</strong>d, <strong>the</strong> values <strong>of</strong> τ 2 are similar indicating that <strong>the</strong> pore sizes are within<strong>the</strong> range typical <strong>for</strong> <strong>nanoporous</strong> <strong>carbon</strong>s. In <strong>the</strong> same work, <strong>the</strong> authors have also <strong>use</strong>d WAXDto determine d-spacing <strong>of</strong> 3.68 and 3.51 Å, which is slightly lower than our pore size range asshown in Table 4-5 but again within <strong>the</strong> 3 – 6 Å range typical <strong>for</strong> <strong>nanoporous</strong> <strong>carbon</strong> made from<strong>the</strong> pyrolysis <strong>of</strong> PFA [84, 89].In a related study, Tin et al [97] <strong>use</strong> WAXD to specifically review <strong>the</strong> effect <strong>of</strong> pyrolysis<strong>temperature</strong> on <strong>the</strong> morphology <strong>of</strong> NPC <strong>membranes</strong> made from P84. This is a similar study tothat done by Steel and Koros [17] who also <strong>use</strong>d WAXD to characterise NPC <strong>membranes</strong> madefrom Matrimid at different pyrolysis <strong>temperature</strong>s. Although <strong>the</strong> pyrolysis <strong>temperature</strong> range95


<strong>use</strong>d <strong>for</strong> <strong>the</strong> polyimide NPC <strong>membranes</strong> is different to ours, <strong>the</strong> results show a decrease in <strong>the</strong> d-spacing or pore size with increasing pyrolysis <strong>temperature</strong> as shown in Table 4-5. Tin et al [97]also observe an increase in crystallinaty at 800°C, consistent with our results.For <strong>the</strong> <strong>high</strong>est pyrolysis <strong>temperature</strong> <strong>use</strong>d to make NPC <strong>membranes</strong> from P84 and Matrimid(800°C) <strong>the</strong>re is a peak located greater than 40 degrees-2-<strong>the</strong>ta or 2.1 Å, which <strong>the</strong> authors <strong>of</strong>both papers attribute to a transition towards a more graphitic structure. Likewise, <strong>for</strong> oursamples made from PFA pyrolysed at 600°C and 800°C, <strong>the</strong>re is a peak located greater than 40degrees-2-<strong>the</strong>ta or 2.1 Å, which suggests a similar tendency towards a graphitic structure.The WAXD and PALS results are also consistent with those obtained by Burket et al [61] usingmethyl chloride adsorption. These workers find that pyrolysis <strong>of</strong> PFA at 300°C to 400°Cgenerates both mesopores (> 20 Å) and micropores (


Finally, it should be noted that both <strong>the</strong> WAXD and PALS analysis was conducted at ambient<strong>temperature</strong>. In comparing <strong>the</strong>se results to <strong>the</strong> permeation data at elevated <strong>temperature</strong>s below,it should be noted that <strong>the</strong> pore size may vary slightly with <strong>temperature</strong>. Fur<strong>the</strong>rmore, <strong>the</strong>expansion <strong>of</strong> <strong>the</strong> metal substrate may also assist slightly in pore opening at <strong>high</strong>er <strong>temperature</strong>s.This means that <strong>the</strong> pore size <strong>of</strong> a membrane tested at elevated operating <strong>temperature</strong>s might besomewhat larger than <strong>the</strong> values given above.4.2.3 Pure Gas Per<strong>for</strong>manceA summary <strong>of</strong> <strong>the</strong> pure gas per<strong>for</strong>mance results <strong>for</strong> <strong>the</strong> <strong>membranes</strong> made between 400 and600°C is given below in Table 4-6. The pure gas per<strong>for</strong>mance results were <strong>use</strong>d to assess <strong>the</strong>impact <strong>of</strong> pyrolysis <strong>temperature</strong> upon <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> <strong>the</strong> resulting NPC <strong>membranes</strong>.Table 4-6 : Pure gas per<strong>for</strong>mance results <strong>for</strong> <strong>the</strong> final coats <strong>of</strong> <strong>membranes</strong> made between400 and 600°C.Membrane Temp Coats CarbonPermeancemol/m 2 .s.Pa x10 -10Permselectivity°C mg/cm 2 CO 2 N 2 CH 4 CO 2 /N 2 CO 2 /CH 4A 400 2 7.04 0.28 0.05 N/A 5.5 N/AB 400 2 7.58 0.23 0.04 N/A 6.1 N/AC 400 3 7.87 2.69 0.21 0.18 12.6 14.8D 400 4 8.77 3.44 1.80 2.56 1.9 1.3A 450 2 5.89 0.24 0.10 N/A 2.4 N/AB 450 2 6.15 3.20 0.91 N/A 3.5 N/AC 450 4 6.91 0.07 0.01 0.01 9.7 12.0A 500 2 5.29 1.2 1.0 1.3 1.2 0.9B 500 2 6.10 7.3 1.9 2.2 3.9 3.3C 500 3 6.13 25.7 17.7 22.4 1.5 1.1D 500 3 8.12 61.4 53.5 74.0 1.1 0.8A 550 2 5.18 22.6 1.9 1.7 11.7 13.0B 550 4 5.19 45.6 4.1 N/A 11.2 N/AC 550 3 5.22 31.8 2.4 3.6 13.0 8.9D 550 4 5.58 48.9 2.4 N/A 20.6 N/AE 550 2 6.07 22.4 2.0 2.1 11.5 10.6F 550 3 6.88 57.3 16.2 19.2 3.5 3.0A 600 2 4.62 78.6 31.1 40.8 2.5 1.9B 600 3 5.73 334.3 287.2 444.1 1.2 0.8C 600 4 6.28 45.9 24.7 35.6 1.9 1.397


Figure 4-7 shows that <strong>the</strong> pure gas CO 2 permeance is consistently <strong>high</strong>er than that <strong>of</strong> N 2 andCH 4 . This trend indicates that Knudsen diffusion is not dominant, as in this case, <strong>the</strong> relativemolecular weights would imply a lower CO 2 permeance. Ra<strong>the</strong>r, <strong>the</strong> trend is indicative <strong>of</strong> ei<strong>the</strong>rmolecular sieving, as <strong>the</strong> kinetic diameter <strong>of</strong> CO 2 (3.3 Å) is substantially smaller than that <strong>of</strong> N 2(3.64 Å) and CH 4 (3.8 Å), and/or surface diffusion as CO 2 is more condensable than <strong>the</strong> o<strong>the</strong>rgases.For all gases, <strong>the</strong>re is an increase in <strong>the</strong> pure gas permeance with increasing pyrolysis<strong>temperature</strong>. This permeance increase can be directly related to <strong>the</strong> increased in porosity withincreasing pyrolysis <strong>temperature</strong> observed in both <strong>the</strong> PALS and WAXD results as discussed in<strong>the</strong> previous section, which would increase <strong>the</strong> rate <strong>of</strong> diffusion.1.00E-061.00E-07CO2 N2 CH4Expon. (CO2) Expon. (N2) Expon. (CH4)Permeance (mol/m 2 .s.Pa)1.00E-081.00E-091.00E-101.00E-111.00E-121.00E-13300 350 400 450 500 550 600 650Pyrolysis Temperature (°C)Figure 4-7 : The effect <strong>of</strong> pyrolysis <strong>temperature</strong> on permeance showing a trend <strong>of</strong>increasing permeance with increasing pyrolysis <strong>temperature</strong>.However, <strong>the</strong>re was no observed correlation with increasing pyrolysis <strong>temperature</strong> andpermselectivity as shown in Figure 4-8. This result suggests that <strong>the</strong> change in pore size asobserved by PALS and WAXD is possibly too insignificant to have any impact on <strong>the</strong>permselectivity.98


25.00CO2/N220.00CO2/CH4Permselectivity15.0010.005.000.00300 350 400 450 500 550 600 650Pyrolysis Temperature (°C)Figure 4-8 : The effect <strong>of</strong> pyrolysis <strong>temperature</strong> on permselectivity showing no clear trendbetween selectivity and pyrolysis <strong>temperature</strong>.As discussed in <strong>the</strong> following paragraph, <strong>the</strong> amount <strong>of</strong> <strong>carbon</strong> added to each membrane and<strong>the</strong>re<strong>for</strong>e <strong>the</strong> presence <strong>of</strong> microcracking appears to have a greater effect on <strong>the</strong> selectivity than<strong>the</strong> pyrolysis <strong>temperature</strong>.The change in permeance and permselectivity with increasing number <strong>of</strong> coats is presented <strong>for</strong> asingle NPC membrane in Figure 4-9. A selective membrane is not observed until <strong>the</strong> membranelayer is complete upon <strong>the</strong> second or third coat. However, as fur<strong>the</strong>r coats are applied, crackingbegins to occur in <strong>the</strong> membrane layer. If <strong>the</strong> cracks are microscopic, Knudsen diffusion willdominate and <strong>the</strong> permselectivity will again drop below 1. If <strong>the</strong> cracks are significant,convective flow will dominate and permselectivity will be 1 as is <strong>the</strong> case <strong>for</strong> <strong>the</strong> exampleshown in Figure 4-9.The permselectivity results as a function <strong>of</strong> <strong>the</strong> amount <strong>of</strong> <strong>carbon</strong> deposited and <strong>the</strong> pyrolysis<strong>temperature</strong> <strong>for</strong> <strong>membranes</strong> made at 400 and 550°C including previous coats, where data wasavailable, are presented in Table 4-7 and Figure 4-10.These results reveal that <strong>for</strong> <strong>the</strong> lower manufacturing <strong>temperature</strong>, it is necessary to deposit alarger amount <strong>of</strong> <strong>carbon</strong> (~ 8 mg/cm 2 ) be<strong>for</strong>e a selective membrane is produced. Conversely, at<strong>the</strong> <strong>high</strong>er manufacturing <strong>temperature</strong>, a smaller amount <strong>of</strong> <strong>carbon</strong> (~6 mg/cm 2 ) is required toproduce a selective membrane.99


1.00E+0010.00CO 2 Permeance (mol/m 2 .Pa.s)1.00E-011.00E-021.00E-031.00E-041.00E-051.00E-061.00E-071.00E-081.00E-09Carbon filling <strong>the</strong> supportand creating top layerKnudsen DiffusionCO 2 / N 2 Selectivity ~ 0.8Membrane LayerCreatedSelectivity > 1Cracking BeginsSelectivity Permeance 5.000.00-5.00-10.00-15.00CO 2 /N 2 Permselectivity1.00E-101.00E-11-20.000 1 2 3 4# <strong>of</strong> CoatsFigure 4-9 : Change in permeance and selectivity with number <strong>of</strong> coats <strong>for</strong> NPCmembrane D made at 400°C. Knudsen diffusion dominates until a selective membrane is<strong>for</strong>med with <strong>the</strong> third coat. However, <strong>the</strong> final 4 th coat results in cracking <strong>of</strong> <strong>the</strong> membranelayer and a subsequent decrease in selectivity and increase in permeance.Table 4-7 : Pure gas per<strong>for</strong>mance results <strong>for</strong> each coat <strong>for</strong> <strong>membranes</strong> at 400 and 550°C.Membrane Temp Coats CarbonPermeancemol/m 2 .s.Pa x10 -10Permselectivity°C mg/cm 2 CO 2 N 2 CH 4 CO 2 /N 2 CO 2 /CH 4A 400 2 7.04 0.28 0.05 N/A 5.5 N/AB 400 2 7.58 0.23 0.04 N/A 6.1 N/AC 400 2 7.38 3.2 0.91 1.1 3.5 2.9C 400 3 7.87 2.7 0.21 0.18 12.6 14.8D 400 2 6.71 2.2 2.0 N/A 1.1 N/AD 400 3 7.57 1.9 0.27 0.24 6.8 7.8D 400 4 8.77 3.4 1.8 2.6 1.9 1.3A 550 2 5.18 22.6 1.9 1.7 11.7 13.0B 550 4 5.19 45.6 4.1 N/A 11.2 N/AC 550 2 4.46 98.4 52.2 75.2 1.9 1.3C 550 3 5.22 31.8 2.4 3.6 13.0 8.9D 550 4 5.58 48.9 2.4 N/A 20.6 N/AE 550 2 6.07 22.4 2.0 2.1 11.5 10.6F 550 3 6.88 57.3 16.2 19.2 3.5 3.0100


(a)25.0550 °C 400 °C20.0CO 2 /CH 4 Permselectivity15.010.05.00.02.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00Carbon Deposited (mg/cm 2 )(b)25.0550 °C 400 °C20.0CO 2 /N 2 Permselectivity15.010.05.00.02.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00Carbon Deposited (mg/cm 2 )Figure 4-10 : Change in (a) CO 2 /CH 4 permselectivity and (b) CO 2 / N 2 permselectivity withamount <strong>of</strong> <strong>carbon</strong> deposited <strong>for</strong> selected <strong>membranes</strong> manufactured at pyrolysis<strong>temperature</strong>s <strong>of</strong> 400 and 550 °C. The decrease in selectivity after a maximum value hasbeen reached represents <strong>the</strong> point at which cracking occurs. For <strong>the</strong> <strong>membranes</strong>manufactured at 550°C, this critical point is at ~ 6 mg/cm 2 , whereas <strong>for</strong> <strong>the</strong> <strong>membranes</strong>manufactured at 400°C, this critical point is at ~ 8 mg/cm 2 .101


Ano<strong>the</strong>r interesting observation is <strong>the</strong> order <strong>of</strong> permeance <strong>for</strong> <strong>the</strong> three molecules, CO 2 , N 2 andCH 4 as additional <strong>carbon</strong> is added. Under Knudsen diffusion (based upon molecular weight), <strong>the</strong>rate <strong>of</strong> permeance is CH 4 >N 2 > CO 2 and <strong>the</strong>re<strong>for</strong>e 1> CO 2 /N 2 > CO 2 /CH 4 . This is <strong>the</strong> trend thatis observed when <strong>the</strong> amount <strong>of</strong> <strong>carbon</strong> added is too low and a membrane layer has not yet<strong>for</strong>med or too <strong>high</strong> and microscopic cracking has occurred.As <strong>the</strong> amount <strong>of</strong> <strong>carbon</strong> is increased, <strong>the</strong> trend becomes CO 2 > CH 4 > N 2 , suggesting amechanism based upon surface diffusion where CO 2 is <strong>the</strong> most adsorptive component and N 2 is<strong>the</strong> least adsorptive component. In this case <strong>the</strong> selectivities are CO 2 /N 2 > CO 2 /CH 4 > 1 as isapparent <strong>for</strong> <strong>the</strong> <strong>membranes</strong> made at 550°C shown in Figure 4-10.When <strong>the</strong> amount <strong>of</strong> <strong>carbon</strong> is close to <strong>the</strong> critical point, <strong>the</strong> membrane layer is complete andwithout any larger pores or defects. In this case <strong>the</strong> rate <strong>of</strong> permeance is CO 2 >N 2 > CH 4 andhence CO 2 /CH 4 > CO 2 /N 2 >> 1 suggesting a mechanism that is based upon molecular sieving asis apparent <strong>for</strong> <strong>the</strong> <strong>membranes</strong> made at 400°C shown in Figure 4-10.4.2.4 Mixed Gas Per<strong>for</strong>manceMixed gas testing was per<strong>for</strong>med <strong>for</strong> several <strong>of</strong> <strong>the</strong> <strong>high</strong>er per<strong>for</strong>ming supported NPC<strong>membranes</strong>. The conditions <strong>for</strong> testing <strong>the</strong> <strong>membranes</strong> using a mixed gas feed and <strong>the</strong> methods<strong>use</strong>d <strong>for</strong> calculating <strong>the</strong> permeance and selectivity were as previously presented in Section 3.4.2.The differences in <strong>the</strong> results <strong>for</strong> <strong>the</strong> pure gas and mixed gas testing <strong>of</strong> two <strong>membranes</strong>, onemade at a pyrolysis <strong>temperature</strong> <strong>of</strong> 400°C (membrane C from Table 4-2) and one made at apyrolysis <strong>temperature</strong> <strong>of</strong> 550°C (membrane C from Table 4-2) are given in Table 4-8. Themixed gas testing <strong>of</strong> <strong>the</strong> membrane made at 400°C was completed immediately following <strong>the</strong>pure gas testing, whilst <strong>the</strong> testing <strong>of</strong> <strong>the</strong> membrane made at 550°C was completed usingregeneration that removes <strong>the</strong> majority <strong>of</strong> <strong>the</strong> adsorbed water as discussed in Section 3.4.2.The results as shown in Table 4-8, indicate that mixed gas testing results in a <strong>high</strong>er CO 2 /CH 4selectivity. Un<strong>for</strong>tunately, <strong>the</strong> results can not be compared directly beca<strong>use</strong> <strong>of</strong> <strong>the</strong> differing feedpartial pressures <strong>of</strong> CO 2 and CH 4 , which is limited by <strong>the</strong> operating pressure <strong>of</strong> <strong>the</strong> equipment.With CO 2 being a more adsorptive molecule, as is discussed in Chapter 5, it is suggested thatwith a mixed gas feed <strong>the</strong> adsorption <strong>of</strong> CO 2 molecules within <strong>the</strong> porous <strong>carbon</strong> actuallyinhibits <strong>the</strong> transport <strong>of</strong> CH 4 molecules resulting in a <strong>high</strong>er selectivity than with pure gas feedtesting.102


Table 4-8 : Pure gas vs. mixed gas results <strong>for</strong> <strong>membranes</strong> manufactured at 400 and 550°C.An increased CO 2 /CH 4 selectivity is seen with <strong>the</strong> mixed gas testing.PyrolysisTempFeedPressure∆P CO 2 ∆P CH 4°C kPag kPa kPaCO 2Permeancemol/m 2 .s.Pax 10 -10CH 4Permeancemol/m 2 .s.Pax 10 -10CO 2 /CH 4Selectivity400Pure Gas 850 950 950 3.2 ± 0.04 0.2 ± 0.00 16.0 ± 0.31Mixed Gas 1500 150 1350 6.3 ± 0.33 0.3 ± 0.0221.0 ± 1.53(↑ 31 %)550Pure Gas 400 500 500 31.8 ± 0.44 3.6 ± 0.05 8.8 ± 0.17Mixed Gas 1200 120 1080 22.7 ± 1.18 1.5 ± 0.0815.1 ± 1.10(↑ 70 %)4.2.4.1 Effect <strong>of</strong> Feed PressureTwo <strong>membranes</strong> made at 550°C (B and D from Table 4-2) were each tested using a CO 2 /N 2mixed gas at feed pressures <strong>of</strong> 500 and 1000 kPa and at feed <strong>temperature</strong>s <strong>of</strong> 35 and 100°C.There is a reduction in <strong>the</strong> CO 2 permeance with increasing pressure <strong>for</strong> both <strong>membranes</strong> asshown in Figure 4-11. The reduction in <strong>the</strong> permeance is more pronounced at <strong>the</strong> loweroperating <strong>temperature</strong>. The reduction in permeance with increasing pressure is attributed to adecrease in sorptivity as <strong>the</strong> molecular sieving transport mechanism is not dependent on feedpressure [84, 105]. Whilst <strong>the</strong>re is an increase in <strong>the</strong> number <strong>of</strong> molecules adsorbed as <strong>the</strong>pressure is increased as shown in Chapter 5, <strong>the</strong> value <strong>of</strong> <strong>the</strong> sorptivity, which is normalised <strong>for</strong>pressure, decreases.Interestingly, <strong>the</strong> relative reduction in <strong>the</strong> N 2 permeance with increased feed pressure is slightly<strong>high</strong>er than that <strong>for</strong> <strong>the</strong> CO 2 permeance resulting in a slight increase in <strong>the</strong> CO 2 /N 2 selectivitywith increased feed pressure as shown in Figure 4-12. This is a curious result given that N 2molecules are much less adsorptive than CO 2 as discussed in Chapter 5. With <strong>the</strong> exception <strong>of</strong>Membrane D at 35°C, a firm conclusion can not be made on <strong>the</strong> impact <strong>of</strong> increasing feedpressure on <strong>the</strong> CO 2 /N 2 selectivity due to <strong>the</strong> magnitude <strong>of</strong> <strong>the</strong> error margins.103


CO 2 Permeance (mol/m 2 .s.Pa)1.0E-089.0E-098.0E-097.0E-096.0E-095.0E-094.0E-093.0E-092.0E-09Membrane B 35°C Membrane B 100°CMembrane D 35°C Membrane D 100°C1.0E-09400 500 600 700 800 900 1000 1100 1200Feed Pressure (kPag)Figure 4-11 : Change in CO 2 permeance with increasing feed pressure. Increasing <strong>the</strong> feedpressure reduces <strong>the</strong> permeance.30.0Membrane B 35°C Membrane B 100°C25.0Membrane D 35°C Membrane D 100°CCO 2 /N 2 Selectivity20.015.010.05.00.0400 500 600 700 800 900 1000 1100 1200Feed Pressure (kPag)Figure 4-12 : Change in CO 2 /N 2 selectivity with increasing feed pressure. Increasing <strong>the</strong>feed pressure has <strong>the</strong> slightly increasing <strong>the</strong> selectivity.104


4.2.5 High Temperature Per<strong>for</strong>manceThe mixed gas separation per<strong>for</strong>mance <strong>of</strong> <strong>membranes</strong> made at 400°C (membrane C in Table 4-2) and 550°C (membrane A in Table 4-2) was tested at operating <strong>temperature</strong>s up to 200°C.4.2.5.1 Effect on PermeanceIncreasing <strong>the</strong> operating <strong>temperature</strong> increases <strong>the</strong> energy <strong>of</strong> <strong>the</strong> gas particles and <strong>the</strong>re<strong>for</strong>e <strong>the</strong>diffusion <strong>of</strong> <strong>the</strong> gas particles through <strong>the</strong> <strong>carbon</strong> membrane. However, as depicted in Figure 4-13, <strong>the</strong> <strong>membranes</strong> manufactured at <strong>high</strong>er <strong>temperature</strong>s, maintain a <strong>high</strong>er permeance over <strong>the</strong>operating <strong>temperature</strong> range.The results show that permeance increases at a faster rate <strong>for</strong> CH 4 than CO 2 . It is thought thatwhilst <strong>the</strong> diffusion component <strong>of</strong> <strong>the</strong> permeance increases with <strong>temperature</strong> in a similar manner<strong>for</strong> both CH 4 and CO 2 , <strong>the</strong>re is a significant reduction in <strong>the</strong> adsorption component <strong>of</strong> <strong>the</strong> CO 2permeance. There<strong>for</strong>e, increasing <strong>the</strong> operating <strong>temperature</strong> has <strong>the</strong> effect <strong>of</strong> decreasing <strong>the</strong>selectivity due to <strong>the</strong> disproportional increase in <strong>the</strong> methane permeance.Mixed Gas Permeance (mol/m 2 .s.Pa)1.00E-07CH4 - 550 °C CH4 - 400 °CCO2 - 550 °C CO2 - 400 °C1.00E-081.00E-091.00E-101.00E-110 50 100 150 200 250Operating Temperature (°C)Figure 4-13 : Effect <strong>of</strong> operating <strong>temperature</strong> on mixed gas permeance. The permeanceincreases with increasing operating <strong>temperature</strong> due to <strong>the</strong> increased diffusion. Themembrane manufactured at 550°C maintains a <strong>high</strong>er permeance across <strong>the</strong> operating<strong>temperature</strong> range.105


An Arrhenius plot <strong>of</strong> <strong>the</strong> membrane permeance as a function <strong>of</strong> 1/T reveals a straight line (witha gradient equal to <strong>the</strong> activation energy) and is provided in Figure 4-14.Mixed Gas Permeance (mol/m 2 .s.Pa)1.00E-07CH4, 400°C PyrolysisCO2, 400°C PyrolysisCH4, 550°C PyrolysisCO2, 550°C PyrolysisExpon. (CH4, 400°C Pyrolysis) Expon. (CO2, 400°C Pyrolysis)1.00E-08Expon. (CH4, 550°C Pyrolysis) Expon. (CO2, 550°C Pyrolysis)1.00E-091.00E-101.00E-110.0020 0.0022 0.0024 0.0026 0.0028 0.0030 0.0032 0.00341/T (K -1 )Figure 4-14 : Arrhenius plot <strong>of</strong> permeance and operating <strong>temperature</strong>. The gradient <strong>of</strong> <strong>the</strong>fitted lines indicate <strong>the</strong> activation energies <strong>of</strong> permeance. The fitted lines are steeper <strong>for</strong>CH 4 resulting in <strong>high</strong>er activation energies as presented in Table 4-10.The accompanying values <strong>of</strong> <strong>the</strong> activation energy E p as calculated from Equation 2-38 arepresented in Table 4-9. These results are consistent with previously published work [15, 25, 26].Table 4-9 : Activation Energy (E p ) <strong>of</strong> CO 2 and CH 4 Permeance (kJ/mol) <strong>for</strong> NPC<strong>membranes</strong> made at 400 and 550°C. Activation energies were calculated from <strong>the</strong>permeance data over <strong>the</strong> range <strong>of</strong> operating <strong>temperature</strong>s using Equation 2-38. Data fromo<strong>the</strong>r authors is included <strong>for</strong> comparison.Polymer PrecursorPyrolysisTemperatureActivation Energy(E p ) CO 2Activation Energy(E p ) CH 4°C kJ/mol kJ/molPFA (this work) 400 3.9 ± 0.3 9.1 ± 1.3PFA (this work) 550 -0.8 ± 1.2 6.7 ± 0.2BPDA-pPDA polyimide [26] 550 0 14.5Phenolic Resin [15] 700 6.2 17.7106


The steep increase in permeance with <strong>temperature</strong> <strong>for</strong> CH 4 at both pyrolysis <strong>temperature</strong>s isreflected in positive activation energies (E p ) and indicates that molecular sieving is <strong>the</strong> dominantmechanism <strong>of</strong> transport <strong>for</strong> this species. As discussed in Section 4.2.2.2, <strong>the</strong> WAXD resultsindicated a distribution <strong>of</strong> pore sizes <strong>of</strong> less than 4.7 Å and 4.0 Å respectively <strong>for</strong> each pyrolysis<strong>temperature</strong>, which would suggest that a large number <strong>of</strong> <strong>the</strong> pores <strong>of</strong> both samples would becomparable in size to <strong>the</strong> kinetic diameter <strong>of</strong> methane (3.8 Å). If Knudsen diffusion and/orsurface diffusion were dominant <strong>the</strong>n permeance would fall as operating <strong>temperature</strong> increased.While <strong>the</strong> two activation energies <strong>for</strong> CH 4 are comparable, <strong>the</strong> value <strong>of</strong> E p <strong>for</strong> CO 2 fallssignificantly as <strong>the</strong> pyrolysis <strong>temperature</strong> increases from 400 to 550°C. The narrowing <strong>of</strong> poresize predicted by both PALS and WAXD as discussed in Section 4.2.2.2 would be expected toincrease <strong>the</strong> activation energy <strong>for</strong> diffusion (E d ) <strong>for</strong> CO 2 , which has a kinetic diameter <strong>of</strong> 3.3 Å.The fall in <strong>the</strong> total activation energy (E p ) is thus unlikely to result from changes to <strong>the</strong> rate <strong>of</strong>diffusion. Ra<strong>the</strong>r, it would seem that <strong>the</strong> fall in total activation energy is related to an increase in<strong>the</strong> heat <strong>of</strong> adsorption (∆H ads ).Burket et al [61] show that at pyrolysis <strong>temperature</strong>s up to 400°C, <strong>nanoporous</strong> <strong>carbon</strong> generatedfrom PFA retains residual polymer in polyaromatic domains as well as a number <strong>of</strong> oxygen andhydrogen moieties. These hetero-atoms and unreacted polymer are driven out above 400°C,leading to a pure <strong>carbon</strong> structure at 600°C. It would follow that <strong>the</strong> hetero-atoms and unreactedpolymer restrict <strong>the</strong> adsorption <strong>of</strong> CO 2 at <strong>the</strong> lower pyrolysis <strong>temperature</strong>s. Conversely, at550°C, <strong>the</strong> structure is almost entirely <strong>carbon</strong>, which would be expected to adsorb CO 2 strongly,leading to an enhanced ∆H ads and lower E p .4.2.5.2 Effect on SelectivityAs <strong>the</strong> permeance increases at a faster rate <strong>for</strong> CH 4 than CO 2 , increasing <strong>the</strong> operating<strong>temperature</strong> has <strong>the</strong> effect <strong>of</strong> decreasing <strong>the</strong> selectivity. The results as displayed in Figure 4.15show that <strong>the</strong> selectivity declines continuously as <strong>the</strong> operating <strong>temperature</strong> is increased. Thiscan be related to <strong>the</strong> relative importance <strong>of</strong> surface diffusion declining at <strong>the</strong>se <strong>high</strong>er<strong>temperature</strong>s. The decline is slightly faster <strong>for</strong> <strong>the</strong> membrane prepared at 550°C, due to <strong>the</strong><strong>high</strong>er heat <strong>of</strong> adsorption. This means that <strong>the</strong> two curves converge at <strong>high</strong>er <strong>temperature</strong>s.Extrapolation suggests that this selectivity might fall to levels consistent with purely Knudsendiffusion (CO 2 /CH 4 selectivity = 0.6) at operating <strong>temperature</strong>s <strong>of</strong> around 400 to 500°C asshown in Figure 4-16. This is consistent with o<strong>the</strong>r researchers [8, 24] who also indicate thatKnudsen diffusion is <strong>the</strong> major contributor to <strong>the</strong> transport mechanism at <strong>high</strong> <strong>temperature</strong>s.107


98550 °C 400 °CCO 2 /CH 4 Selectivity765432100 50 100 150 200 250Operating Temperature (°C)Figure 4-15 : Effect <strong>of</strong> operating <strong>temperature</strong> on mixed gas selectivity <strong>for</strong> NPC<strong>membranes</strong> manufactured at pyrolysis <strong>temperature</strong>s <strong>of</strong> 400 and 550°C. Increasing <strong>the</strong>operating <strong>temperature</strong> has <strong>the</strong> effect <strong>of</strong> decreasing <strong>the</strong> selectivity due to <strong>the</strong>disproportional increase in <strong>the</strong> CH 4 permeance. This is due <strong>the</strong> significant reduction in <strong>the</strong>CO 2 adsorption at <strong>the</strong> <strong>high</strong>er <strong>temperature</strong>s.CO 2 /CH 4 Selectivity9.008.007.006.005.004.003.002.001.00550 °C400 °CCO2/CH4 Knudsen SelectivityExpon. (550 °C)Expon. (400 °C)0.000 100 200 300 400 500 600Operating Temperature (°C)Figure 4-16 : Extrapolated effect <strong>of</strong> operating <strong>temperature</strong> on mixed gas selectivity <strong>for</strong>NPC <strong>membranes</strong> manufactured at pyrolysis <strong>temperature</strong>s <strong>of</strong> 400 and 550°C.Extrapolating <strong>the</strong> CO 2 /CH 4 selectivity relationship with operating <strong>temperature</strong> using anexponential fit shows that <strong>the</strong> selectivity will approach <strong>the</strong> Knudsen value <strong>of</strong> 0.6 at anoperating <strong>temperature</strong> between 400 and 500°C.108


4.2.5.3 Effect <strong>of</strong> Thermal CyclingThe impact on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> an NPC membrane from <strong>the</strong>rmal cycling is presented inFigure 4-17. This NPC membrane was manufactured at 550°C and underwent 11 cycles <strong>of</strong>regeneration <strong>for</strong> > 12 hours at 150°C.1.00E-05CH4 Permeance CO2 Permeance CO2/CH4 Selectivity20.0RRMembrane Exposed to Impurities15.0Permeance (mol/m 2 .s.Pa)1.00E-061.00E-071.00E-081.00E-09R R RR R R R RR10.05.00.0-5.0-10.0CO 2 /CH 4 Selectivity-15.01.00E-10-20.00 50 100 150 200 250 300 350 400 450 500Time (Hours)Figure 4-17 : Effect <strong>of</strong> <strong>the</strong>rmal cycling on an NPC membrane. After eleven regenerationcycles <strong>the</strong> CO 2 and CH 4 permeance increased dramatically whilst <strong>the</strong> selectivity droppedto 1, indicating <strong>the</strong> presence <strong>of</strong> significant cracks.Following <strong>the</strong> 11 th regeneration, <strong>the</strong>re was a dramatic increase in <strong>the</strong> permeance <strong>of</strong> both CH 4and CO 2 along with a drastic drop <strong>of</strong> <strong>the</strong> CO 2 /CH 4 . This significant drop in per<strong>for</strong>mancesuggests that a crack or several cracks had developed due to <strong>the</strong> continuous cycling <strong>of</strong> <strong>the</strong>membrane <strong>temperature</strong> from operation to regeneration.This result raises obvious concerns about <strong>the</strong> continual regeneration <strong>of</strong> NPC <strong>membranes</strong> in acommercial setting. Adsorbents (such as zeolite molecular sieve) are commonly known <strong>for</strong>having a limited lifetime due to <strong>the</strong> <strong>the</strong>rmal cycling <strong>of</strong> <strong>the</strong> regeneration process [158]. Thisresults in change-out <strong>of</strong> <strong>the</strong> molecular sieve material after a particular time period (depending on<strong>the</strong> adsorbent and <strong>the</strong> severity <strong>of</strong> regeneration process), which is a costly undertaking both dueto <strong>the</strong> new material required and <strong>the</strong> shutdown <strong>of</strong> production <strong>for</strong> <strong>the</strong> change-out if a sparesystem is not available.109


One <strong>of</strong> <strong>the</strong> supposed benefits <strong>of</strong> NPC <strong>membranes</strong> over polymeric <strong>membranes</strong> is <strong>the</strong>ir ability tobe regenerated resulting in lower maintenance and replacement costs. The effect <strong>of</strong> repeated<strong>the</strong>rmal cycling on NPC <strong>membranes</strong> is an area <strong>for</strong> fur<strong>the</strong>r work.4.2.6 Concluding Remarks on Supported NPC MembranesIn summary, <strong>the</strong> results <strong>of</strong> <strong>the</strong> characterisation and pure gas per<strong>for</strong>mance testing conclude that a<strong>high</strong>er permeance is achieved by <strong>the</strong> manufacture <strong>of</strong> <strong>the</strong> NPC <strong>membranes</strong> at a <strong>high</strong>er pyrolysis<strong>temperature</strong> (e.g. 550°C) as was also shown by Shiflett and Foley [75]. The PALS and WAXDcharacterisation results reveal that <strong>the</strong> porosity <strong>of</strong> <strong>the</strong> <strong>carbon</strong> increases with increasing pyrolysis<strong>temperature</strong>, which greatly influences <strong>the</strong> increase in permeance.There was no observed correlation between pyrolysis <strong>temperature</strong> and permselectivity. ThePALS and WAXD characterisation results suggest that <strong>the</strong> pore size change may be too small topresent any change in <strong>the</strong> permselectivity or that <strong>the</strong> presence <strong>of</strong> cracking has a more significantimpact.To produce a selective molecular sieving membrane with a minimal amount <strong>of</strong> cracking at550°C, it is necessary to limit <strong>the</strong> total <strong>carbon</strong> deposited to just under 6 mg/cm 2 due to <strong>the</strong> moreporous and <strong>the</strong>re<strong>for</strong>e fragile nature <strong>of</strong> <strong>the</strong> <strong>carbon</strong> produced at this <strong>high</strong> <strong>temperature</strong>.Increasing <strong>the</strong> pyrolysis <strong>temperature</strong> beyond 550°C did not produce selective <strong>membranes</strong>,possibly due to <strong>the</strong> even more fragile nature <strong>of</strong> <strong>the</strong> <strong>carbon</strong> at 600°C.Testing <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> an NPC membrane with a mixed gas feed provides a more accuraterepresentation <strong>of</strong> how <strong>the</strong> <strong>membranes</strong> would per<strong>for</strong>m in a commercial setting than pure gasper<strong>for</strong>mance testing. This is beca<strong>use</strong> mixed gas testing incorporates <strong>the</strong> interaction effectsbetween <strong>the</strong> different components in <strong>the</strong> feed gas.The results <strong>of</strong> pure and mixed gas testing <strong>for</strong> <strong>the</strong> same <strong>membranes</strong> presented in this chapterrevealed that <strong>the</strong> CO 2 /CH 4 mixed gas selectivity was greater than <strong>the</strong> CO 2 /CH 4 permselectivitydetermined from <strong>the</strong> pure gas testing. This result is probably explained by <strong>the</strong> competitiveadsorption effect <strong>of</strong> <strong>the</strong> more adsorbable component CO 2 “blocking” <strong>the</strong> way <strong>for</strong> CH 4permeance.The <strong>carbon</strong> dioxide/methane (CO 2 /CH 4 ) mixed gas per<strong>for</strong>mance measurements generally showan increase in permeance as <strong>the</strong> operating <strong>temperature</strong> is increased due to <strong>the</strong> dominance <strong>of</strong>molecular sieving. The exception is <strong>for</strong> CO 2 permeance <strong>of</strong> <strong>membranes</strong> prepared at a <strong>high</strong>er110


pyrolysis <strong>temperature</strong> (550°C) where adsorption processes are competitive with molecularsieving effects. This leads to a CO 2 permeance value that is relatively independent <strong>of</strong><strong>temperature</strong>. While this results in a membrane that is slightly more selective at 35°C, <strong>the</strong>membrane selectivities converge at <strong>high</strong>er operating <strong>temperature</strong>s. Indeed, extrapolationsuggests that <strong>the</strong> membrane selectivity would reduce to <strong>the</strong> Knudsen (CO 2 /CH 4 = 0.6) value atan operating <strong>temperature</strong> between 400 and 500°C.Whilst <strong>the</strong> selectivities <strong>of</strong> <strong>the</strong> <strong>membranes</strong> manufactured at different pyrolysis <strong>temperature</strong>sconverge at <strong>high</strong>er operating <strong>temperature</strong>s, <strong>the</strong> permeance is still related <strong>the</strong> original pyrolysis<strong>temperature</strong>. This result implies that it is important to manufacture a membrane with <strong>high</strong>porosity/permeance ra<strong>the</strong>r than <strong>high</strong> adsorption/selectivity when considering <strong>high</strong> operating<strong>temperature</strong> applications such as CO 2 capture from natural gas.In conclusion, <strong>the</strong> results <strong>of</strong> mixed gas <strong>high</strong> <strong>temperature</strong> per<strong>for</strong>mance trials show that CO 2 /CH 4selectivity remains above 1 at operating <strong>temperature</strong> less than ~ 300°C. This is a promisingresult <strong>for</strong> <strong>high</strong> <strong>temperature</strong> applications. However, fur<strong>the</strong>r work needs to be undertaken toevaluate <strong>the</strong> impact <strong>of</strong> <strong>the</strong>rmal cycling on cracking <strong>of</strong> <strong>the</strong> NPC membrane surface.111


4.3 Modified-Supported NPC MembranesBy modifying <strong>the</strong> pore size <strong>of</strong> <strong>the</strong> support and <strong>the</strong>reby reducing <strong>the</strong> weight <strong>of</strong> <strong>carbon</strong> required tocreate <strong>the</strong> membrane layer, <strong>the</strong> chance <strong>of</strong> cracking is reduced as discussed in Section 2.3.1.3.Silica nanoparticles were produced, filtered into <strong>the</strong> stainless steel supports and <strong>the</strong>n coated with2 to 3 layers <strong>of</strong> <strong>carbon</strong> to create <strong>the</strong> modified-supported NPC <strong>membranes</strong>.The manufacture and testing <strong>of</strong> <strong>the</strong> modified supported NPC <strong>membranes</strong> work was completedby Budi Suharto and Yik Koon (Evan) Ang as <strong>the</strong> final year research project <strong>for</strong> <strong>the</strong>irundergraduate engineering degree.4.3.1 Manufacture and CharacterisationThe two precursor solutions, as described earlier in Section 3.2.2.3 Table 3-4, required <strong>for</strong>manufacturing silica nanoparticles were reacted at various <strong>temperature</strong> and times until silicananoparticles with a size less than 300 nm were <strong>for</strong>med. The size <strong>of</strong> <strong>the</strong> silica nanoparticles wasdetected using scanning electron microscopy. A summary <strong>of</strong> <strong>the</strong> manufacturing results ispresented in Table 4-10. Typical SEM pictures <strong>of</strong> <strong>the</strong> nanoparticles are given in Figure 4-18.Table 4-10 : Effect <strong>of</strong> varying reaction <strong>temperature</strong> and time on <strong>the</strong> size <strong>of</strong> <strong>the</strong> producedsilica nanoparticles. The fourth sample with a reaction <strong>temperature</strong> <strong>of</strong> 50°C and time <strong>of</strong> 45mins gave monodisperse silica nanoparticles with an average size less than 300 nm.Method Reaction Temperature Reaction Time Particle Size°C mins nm1 21 120 500 +/- 202 39 30 310 +/- 203 39 60 295 +/- 204 (Selected <strong>for</strong> Filling) 50 45 275 +/- 20Figure 4-18 : Silica nanoparticles, with a size less than 300 nm produced from <strong>the</strong> reaction<strong>of</strong> ethanol, ammonia and TEOS at a <strong>temperature</strong> <strong>of</strong> 50 °C <strong>for</strong> 45 mins.112


The supports were <strong>the</strong>n filled up to 4 times with <strong>the</strong> silica nanoparticles (< 300 nm) until silicawas clearly seen on <strong>the</strong> surface <strong>of</strong> <strong>the</strong> support after drying at 100°C overnight.To determine <strong>the</strong> effect <strong>of</strong> <strong>the</strong> silica nanoparticles on <strong>the</strong> stainless steel supports, pure gasper<strong>for</strong>mance testing was undertaken and <strong>the</strong> results are shown in Table 4-11 and Figure 4-19.The addition <strong>of</strong> <strong>the</strong> silica nanoparticles decreases <strong>the</strong> permeance <strong>of</strong> <strong>the</strong> porous stainless steelsupport by almost an order <strong>of</strong> magnitude due to <strong>the</strong> reduction in <strong>the</strong> pore size and porosity.However, <strong>the</strong> selectivity does not alter significantly indicating that <strong>the</strong> reduction in <strong>the</strong> porositywas insufficient to ca<strong>use</strong> a change in <strong>the</strong> transport mechanism and as such bulk flow was stilldominant.Table 4-11 : Effect <strong>of</strong> <strong>the</strong> addition <strong>of</strong> silica nanoparticles on <strong>the</strong> permeance and selectivity<strong>of</strong> <strong>the</strong> porous stainless steel support. The error in <strong>the</strong> permeance is ± 1.38% and <strong>the</strong> errorin <strong>the</strong> permselectivity is 1.95%.SupportPyrolysisTempSilica Permeance (mol/m 2 .s.Pa) x 10 -5 Permselectivity°C mg/cm 2 CO 2 N 2 CH 4 CO 2 /N 2 CO 2 /CH 4Blank 0 5.9 6.3 5.1 0.9 1.2A 550 0.0526 1.6 1.5 1.3 1.0 1.2B 550 0.3288 1.3 1.3 1.0 1.0 1.3C 550 0.5459 0.9 0.8 0.6 1.1 1.6D 550 2.4728 0.6 0.7 0.6 0.9 1.01.00E-04CO2 N2 CH4Permeance (mol/m 2 .s.Pa)1.00E-051.00E-060 0.5 1 1.5 2 2.5 3Mass <strong>of</strong> Silica Nanoparticles Added (mg/cm 2 )Figure 4-19 : Effect <strong>of</strong> <strong>the</strong> addition <strong>of</strong> silica nanoparticles on reducing <strong>the</strong> permeance <strong>of</strong><strong>the</strong> stainless steel support suggesting some reduction in pore size and porosity.113


4.3.2 Membrane Per<strong>for</strong>manceAs expected <strong>for</strong> <strong>the</strong>se modified supports, less <strong>carbon</strong> was required to create <strong>the</strong> membrane layerthan <strong>the</strong> un-modified supported <strong>membranes</strong> as <strong>the</strong> <strong>carbon</strong> was not filling <strong>the</strong> pores <strong>of</strong> <strong>the</strong>support. Likewise, less <strong>carbon</strong> was required <strong>for</strong> <strong>the</strong> onset <strong>of</strong> cracking. Two silica containingsupports (B and D) were both initially coated with 5 ml <strong>of</strong> PFA solution, which in turn gavegreater than 4 mg/cm 2 <strong>of</strong> <strong>carbon</strong> deposited onto <strong>the</strong> modified-support. Cracking was clearlyvisible on <strong>the</strong> surface <strong>of</strong> <strong>the</strong>se <strong>membranes</strong>. A fur<strong>the</strong>r two silica containing supports (A and C),were coated with 1 – 2 ml <strong>of</strong> PFA solution leading to less than 2 mg/cm 2 <strong>of</strong> <strong>carbon</strong> depositedonto <strong>the</strong> modified support. No cracking was seen on <strong>the</strong> surface <strong>of</strong> <strong>the</strong>se two <strong>membranes</strong> and <strong>the</strong>per<strong>for</strong>mance results showed Knudsen selectivity (CO 2 /N 2 = 0.8 and CO 2 /CH 4 = 0.6), whichindicated that <strong>the</strong> membrane layer was not complete. A second coat was <strong>the</strong>n added to complete<strong>the</strong> membrane layer <strong>for</strong> <strong>the</strong>se two supports.The per<strong>for</strong>mance results as each coat <strong>of</strong> <strong>carbon</strong> was added are presented in Table 4-12. Figures4-20 and 4-21 show <strong>the</strong> effect on permeance and selectivity, respectively, as <strong>the</strong> amount <strong>of</strong><strong>carbon</strong> added to <strong>the</strong> modified support was increased. As <strong>the</strong> amount <strong>of</strong> <strong>carbon</strong> added wasincreased above 1.5 mg/cm 2 , <strong>the</strong> permeance dropped dramatically and <strong>the</strong> selectivity increasedabove 1, which suggested <strong>the</strong> <strong>for</strong>mation <strong>of</strong> a selective layer. However, when <strong>the</strong> amount <strong>of</strong><strong>carbon</strong> added was increased above 4 mg/cm 2 cracking appeared. Comparing this result to <strong>the</strong><strong>membranes</strong> manufactured at 550°C without silica nanoparticles this is a reduction in <strong>the</strong> criticalcracking point from 6 mg/cm 2 to 4 mg/cm 2 .Interestingly, <strong>the</strong> <strong>membranes</strong> manufactured with silica nanoparticles in <strong>the</strong> support showed anorder <strong>of</strong> magnitude lower overall permeance than those manufactured without silicananoparticles which had a similar or <strong>high</strong>er selectivity. This is an interesting result given <strong>the</strong>amount <strong>of</strong> <strong>carbon</strong> added <strong>for</strong> <strong>the</strong> modified-supported <strong>membranes</strong> was lower, suggesting that <strong>the</strong>thickness <strong>of</strong> <strong>the</strong> membrane was lower and <strong>the</strong>re<strong>for</strong>e <strong>the</strong> permeance should be <strong>high</strong>er. However,<strong>the</strong>re was a slight but important difference in <strong>the</strong> manufacturing procedure <strong>of</strong> <strong>the</strong> two differenttypes <strong>of</strong> <strong>membranes</strong>. The membrane manufactured with <strong>the</strong> silica nanoparticles in <strong>the</strong> supportwere coated over 2 consecutive days whereas <strong>the</strong> <strong>membranes</strong> manufactured without <strong>the</strong> silicananoparticles were coated with a 5 day break between <strong>the</strong> first and <strong>the</strong> second coat. During <strong>the</strong> 5day break, <strong>the</strong> first coat was exposed to <strong>the</strong> atmosphere. As was discussed earlier in Section3.2.2.1.2, it is believed that exposing <strong>the</strong> first coat to oxygen opens <strong>the</strong> pores <strong>of</strong> <strong>carbon</strong> that hasfilled <strong>the</strong> pores <strong>of</strong> <strong>the</strong> support such that when <strong>the</strong> second coat is added, it is only this coat that<strong>for</strong>ms <strong>the</strong> membrane layer.114


Table 4-12 : Addition <strong>of</strong> <strong>carbon</strong> to supports modified with silica nanoparticles. The errorin <strong>the</strong> permeance is ± 1.38% and <strong>the</strong> error in <strong>the</strong> permselectivity is 1.95%.Membrane Coat Carbon Permeance (mol/m 2 .s.Pa) x 10 -10 Permselectivitymg/cm 2 CO 2 N 2 CH 4 CO 2 /N 2 CO 2 /CH 4A 1 1.6 908 1010 1370 0.9 0.7A 2 2.1 3.2 1.0 1.3 3.3 2.5B 1 4.2 444 395 504 1.1 0.9C 1 1.6 679 660 991 1.0 0.7C 2 3.2 6.8 2.6 2.8 2.7 2.5D 1 4.5 632 635 737 1.0 0.91.00E-041.00E-05CO2 N2 CH4Permeance (mol/m 2 .s.Pa)1.00E-061.00E-071.00E-081.00E-091.00E-101.00E-110 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5Mass <strong>of</strong> Carbon Added (mg/cm 2 )Figure 4-20 : Effect <strong>of</strong> <strong>the</strong> addition <strong>of</strong> <strong>carbon</strong> on permeance to supports filled with silicananoparticles. The membrane layer begins to <strong>for</strong>m around 1.5 – 2 mg/cm 2 <strong>of</strong> <strong>carbon</strong> butcracks at 3.5 mg/cm 2 .5.04.54.03.5CO2/N2CO2/CH4Selectivity3.02.52.01.51.00.50.00 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5Mass <strong>of</strong> Carbon Added (mg/cm 2 )Figure 4-21 : Effect <strong>of</strong> <strong>the</strong> addition <strong>of</strong> <strong>carbon</strong> on selectivity to supports filled with silicananoparticles. The selectivity results agree with <strong>the</strong> permeance results, which suggestcracking <strong>of</strong> <strong>the</strong> <strong>for</strong>med membrane layer occurs around 3.5 mg/cm 2 .115


4.3.3 Concluding Remarks on Modified-Supported NPC MembranesThe addition <strong>of</strong> silica nanoparticles to <strong>the</strong> porous support prior to coating with PFA andpyrolysis reduced <strong>the</strong> amount <strong>of</strong> <strong>carbon</strong> required to produce a selective layer. The onset <strong>of</strong>cracking occurred at a lower amount <strong>of</strong> <strong>carbon</strong> added making it more difficult to produceuncracked <strong>membranes</strong>.This was a disappointing outcome given <strong>the</strong> promising work published by Rajagopalan [28] etal and Merritt et al [27] who successfully demonstrated a factor <strong>of</strong> two increase in permeancealong with no sacrifice in selectivity upon <strong>the</strong> addition <strong>of</strong> silica nanoparticles.116


5 UNDERSTANDING THE HIGH TEMPERATURE PERFORMANCE OF NPCMEMBRANES THROUGH GAS ADSORPTION MOLECULAR SIMULATIONAND EXPERIMENTS5.1 IntroductionThe adsorption <strong>of</strong> gases onto <strong>nanoporous</strong> <strong>carbon</strong> has been studied using molecular simulationand mass uptake experiments to fur<strong>the</strong>r understand <strong>the</strong> transport mechanism and per<strong>for</strong>mance <strong>of</strong>NPC <strong>membranes</strong>. The results are presented in this chapter.5.2 Molecular SimulationThe adsorption <strong>of</strong> pure and mixed gases on <strong>nanoporous</strong> <strong>carbon</strong> (represented by C 168 Schwarzite)was simulated using a molecular simulation model based upon <strong>the</strong> grand canonical Monte Carlomethod as described in Section 3.5.1.5.2.1 Pure GasesThe adsorption iso<strong>the</strong>rms <strong>of</strong> pure gases (CH 4 , N 2 and CO 2 ) from 27 to 200°C up to 10,000 kPawere simulated in order to calculate <strong>the</strong> individual heats <strong>of</strong> adsorption <strong>of</strong> <strong>the</strong> three gases on<strong>nanoporous</strong> <strong>carbon</strong>.The adsorption iso<strong>the</strong>rms computed <strong>for</strong> <strong>the</strong> adsorption <strong>of</strong> CO 2 , CH 4 and N 2 onto NPC at selected<strong>temperature</strong>s are presented in Figures 5-1, 5-2 and 5-3, respectively. As <strong>the</strong> operating<strong>temperature</strong> <strong>of</strong> <strong>the</strong> system is increased, <strong>the</strong> loading decreases more prominently <strong>for</strong> CO 2 than <strong>for</strong>CH 4 and N 2 . The loading <strong>of</strong> CO 2 also begins at lower pressures than CH 4 and lower again thanN 2 .The adsorption iso<strong>the</strong>rms were <strong>the</strong>n fitted using <strong>the</strong> Langmuir model, as described in Section2.2.2.3. The Langmuir model fitted well <strong>for</strong> CH4 and N2 <strong>for</strong> all pressures, and <strong>for</strong> CO2 itprovided a good fit <strong>for</strong> <strong>the</strong> region where <strong>the</strong> pressure is greater than 0.01 kPa and less than 1000kPa. For CO 2 adsorption at pressures <strong>high</strong>er than 1000 kPa, it becomes necessary to <strong>use</strong> adouble Langmuir model, as it more accurately captures <strong>the</strong> deviation from <strong>the</strong> ideal Langmuirmodel at <strong>high</strong>er densities [45]. However, <strong>the</strong> partial pressure <strong>of</strong> CO 2 in <strong>the</strong> typical mixed gasfeeds studied in this research work is less than 1000 kPa and as such <strong>the</strong> single Langmuir fit <strong>for</strong>CO 2 up to pressures <strong>of</strong> 1000 kPa is deemed sufficient.117


The values <strong>of</strong> <strong>the</strong> Langmuir constants, b and V m <strong>for</strong> <strong>the</strong> adsorption <strong>of</strong> pure CO 2 , CH 4 and N 2 arepresented in Table 5-1. For each gas, <strong>the</strong> value <strong>of</strong> <strong>the</strong> Langmuir rate constant b decreases withincreasing <strong>temperature</strong> indicating <strong>the</strong> reduction <strong>of</strong> adsorption with <strong>temperature</strong>. Likewise <strong>the</strong>constant V m , which indicates <strong>the</strong> maximum number <strong>of</strong> adsorbed particles decreases withincreasing <strong>temperature</strong>.The values <strong>of</strong> <strong>the</strong> Langmuir rate constant b are <strong>high</strong>est <strong>for</strong> CO 2 followed by CH 4 and <strong>the</strong>n N 2 ,reasserting that <strong>the</strong> adsorptive strength <strong>of</strong> <strong>the</strong> pure gases upon <strong>nanoporous</strong> <strong>carbon</strong> is <strong>of</strong> <strong>the</strong> orderCO 2 >>CH 4 >N 2 .35°C 100°C 150°C 200°C8.00Langmuir Fit (35°C) Langmuir Fit (100°C) Langmuir Fit (150°C) Langmuir Fit (200°C)7.00Loading (mmol/gram NPC)6.005.004.003.002.001.000.000 1 10 100 1000 10000Pressure (kPa)Figure 5-1 : Simulated adsorption iso<strong>the</strong>rms <strong>of</strong> pure CO 2 onto <strong>nanoporous</strong> <strong>carbon</strong>.Increasing <strong>the</strong> <strong>temperature</strong> decreases <strong>the</strong> amount <strong>of</strong> CO 2 adsorbed.35°C 100°C 150°C 200°C8.00Langmuir Fit (35°C) Langmuir Fit (100°C) Langmuir Fit (150°C) Langmuir Fit (200°C)7.00Loading (mmol/gram)6.005.004.003.002.001.000.000 1 10 100 1000 10000Pressure (kPa)Figure 5-2 : Simulated adsorption iso<strong>the</strong>rms <strong>of</strong> pure CH 4 onto <strong>nanoporous</strong> <strong>carbon</strong>.Increasing <strong>the</strong> <strong>temperature</strong> decreases <strong>the</strong> amount <strong>of</strong> CH 4 adsorbed.118


35°C 100°C 150°C 200°C8.00Langmuir Fit (35°C) Langmuir Fit (100°C) Langmuir Fit (150°C) Langmuir Fit (200°C)7.00Loading (mmol/gram)6.005.004.003.002.001.000.000 1 10 100 1000 10000Pressure (kPa)Figure 5-3 : Simulated adsorption iso<strong>the</strong>rms <strong>of</strong> pure N 2 onto <strong>nanoporous</strong> <strong>carbon</strong>.Increasing <strong>the</strong> <strong>temperature</strong> decreases <strong>the</strong> amount <strong>of</strong> N 2 adsorbed.Table 5-1 : Values <strong>of</strong> <strong>the</strong> Langmuir constants (b and V m ) <strong>for</strong> <strong>the</strong> simulated adsorptioniso<strong>the</strong>rms <strong>of</strong> pure CO 2 , CH 4 and N 2 on NPC with increasing <strong>temperature</strong>s.Temperature K 300 308 348 373 398 423 448 478°C 27 35 75 100 125 150 175 200CO 2b bar -1 16.4 12.5 3.55 1.71 0.924 0.523 0.323 0.231V m mmol/gram NPC 7.02 6.91 6.56 6.33 6.05 5.75 5.49 5.08CH 4b bar -1 2.02 1.53 0.533 0.295 0.178 0.12 0.091 0.059V m mmol/gram NPC 7.93 7.90 7.83 7.74 7.61 7.32 6.99 6.64N 2b bar -1 0.315 0.252 0.110 0.075 0.057 0.040 0.031 0.026V m mmol/gram NPC 7.62 7.55 6.99 6.57 6.10 5.76 5.48 5.065.2.1.1 Heats <strong>of</strong> AdsorptionBy rearranging Equation 2-14 to provide a linear relationship, as shown as Equation 5-1, <strong>the</strong>heat <strong>of</strong> adsorption can be calculated from <strong>the</strong> gradient <strong>of</strong> <strong>the</strong> plot <strong>of</strong> 1/T versus Ln(b).∆HadsLn( b)= Ln(b0 ) +Equation 5-1RT119


The straight line plots <strong>of</strong> <strong>the</strong> values <strong>of</strong> b determined at various <strong>temperature</strong>s <strong>for</strong> <strong>the</strong> adsorption<strong>of</strong> pure CO 2 , CH 4 and N 2 onto <strong>nanoporous</strong> <strong>carbon</strong> are presented in Figure 5-4.4.03.02.01.0Carbon Dioxide Nitrogen Methaney = 3.5576x - 9.022R 2 = 0.9994y = 2.8649x - 8.8724R 2 = 0.999ln(b)0.0-1.0y = 2.0464x - 8.029R 2 = 0.9977-2.0-3.0-4.0-5.01.0 1.5 2.0 2.5 3.0 3.51000/T (K -1 )Figure 5-4 : Straight line plot <strong>of</strong> 1/T versus <strong>the</strong> simulated Ln(b) <strong>for</strong> pure gases <strong>for</strong>determining <strong>the</strong> heat <strong>of</strong> adsorption (∆H ads ).The values <strong>of</strong> <strong>the</strong> resulting heats <strong>of</strong> adsorption (∆H ads ) are given in Table 5-2. Also presented inTable 5-2, are <strong>the</strong> experimental values <strong>of</strong> ∆H ads as reported in <strong>the</strong> literature <strong>for</strong> graphite [159],<strong>nanoporous</strong> <strong>carbon</strong> from PFA [43] and <strong>nanoporous</strong> <strong>carbon</strong> from cellulose [12].Table 5-2 : Simulated values <strong>of</strong> <strong>the</strong> heat <strong>of</strong> adsorption (∆H ads ) <strong>of</strong> pure gases on <strong>nanoporous</strong><strong>carbon</strong> compared with experimental values reported in <strong>the</strong> literature.Gas CO 2 CH 4 N 2∆H ads kJ/mol kJ/mol kJ/molThis Work (Simulation) 33.2 26.7 19.1Graphite [159] 16.8 Data N/A 8.9NPC from PFA [43] 10.6 Data N/A 10.2NPC from Cellulose [12] 29.6 Data N/A 10.5As expected, <strong>the</strong> heats <strong>of</strong> adsorption <strong>for</strong> <strong>the</strong> pure gases as calculated from <strong>the</strong> molecularmodelling are <strong>high</strong>er than those <strong>for</strong> graphite due to <strong>the</strong> curved nature <strong>of</strong> <strong>the</strong> Schwarzite surface[113]. However, <strong>the</strong> literature experimental values obtained <strong>for</strong> adsorption <strong>of</strong> CO 2 and N 2 on<strong>nanoporous</strong> <strong>carbon</strong> manufactured from PFA ([43]) and Cellulose [12] are also both lower thanpredicted here via simulation. This difference between simulated and experimental results isdiscussed fur<strong>the</strong>r in Section 5.3.120


5.2.2 Mixed GasesThe adsorption behaviour <strong>of</strong> two mixtures was also modelled. The two mixtures were (a)CO 2 :CH 4 (10:90 vol%) representing CO 2 separation from natural gas and (b) CO 2 :N 2 (10:90vol%) representing CO 2 separation from syn<strong>the</strong>sis gas.5.2.2.1 Carbon Dioxide / MethaneThe simulated adsorption iso<strong>the</strong>rms <strong>of</strong> CO 2 and CH 4 from a CO 2 :CH 4 (10:90 vol%) mixed gasat 35 and 200°C are shown in Figure 5-5. As expected from <strong>the</strong> pure gas simulation results, <strong>the</strong>adsorption <strong>of</strong> both components decreases with increasing <strong>temperature</strong>.8.007.0035°C CO2 200°C CO235°C CH4 200°C CH46.00Loading (mmol/gram)5.004.003.002.001.000.000.01 0.1 1 10 100 1000 10000Pressure (kPa)Figure 5-5 : Adsorption iso<strong>the</strong>rms <strong>of</strong> mixed CO 2 /CH 4 onto <strong>nanoporous</strong> <strong>carbon</strong>. Increasing<strong>the</strong> <strong>temperature</strong> decreases <strong>the</strong> amount <strong>of</strong> CO 2 and CH 4 adsorbed.From <strong>the</strong> pure gas Langmuir iso<strong>the</strong>rm fits shown in Figures 5-1 and 5-2, <strong>the</strong> saturation <strong>of</strong> CO 2(<strong>the</strong> point at which additional molecules are no longer adsorbed onto <strong>the</strong> NPC surface) at 35°Coccurs at around 100 to 1000 kPa and saturation <strong>of</strong> CH 4 occurs at a pressure >1000 kPa.Likewise, <strong>for</strong> <strong>the</strong> mixed bulk gas at 35°C, as shown in Figure 5-5, <strong>the</strong> NPC surface is saturatedwith CO 2 molecules at around 100 kPa whilst CH 4 adsorption continues until 10,000 kPa.At a system <strong>temperature</strong> <strong>of</strong> 200°C, adsorption is greatly reduced and as such much <strong>high</strong>erpressures are required <strong>for</strong> surface saturation to occur. From Figures 5-1 and 5-2, saturation <strong>of</strong>CO 2 and CH 4 molecules onto <strong>the</strong> NPC surface at 200°C does not occur at pressures lower than121


10,000 kPa. Similarly, in <strong>the</strong> mixed gas scenario as shown in Figure 5-5, saturation has not beenachieved <strong>for</strong> ei<strong>the</strong>r component <strong>for</strong> <strong>the</strong> pressure range shown.The impact on <strong>the</strong> mixed gas separation factor with increasing pressure <strong>for</strong> system <strong>temperature</strong>s<strong>of</strong> 35 and 200°C is presented in Figure 5-6. In all cases, <strong>the</strong> simulated NPC is selective <strong>for</strong> CO 2with <strong>the</strong> CO 2 /CH 4 separation factor being greater than 1.9.008.00Feed Temperautre = 35°C Feed Temperature = 200°CCO 2 /CH 4 Separation Factor7.006.005.004.003.002.001.000.001 10 100 1000 10000 100000Pressure (kPa)Figure 5-6 : CO 2 /CH 4 mixed gas adsorption separation factor as a function <strong>of</strong> pressure <strong>for</strong>system <strong>temperature</strong>s <strong>of</strong> 35 and 200°C.At 35°C, <strong>the</strong> CO 2 /CH 4 separation factor reduces with increasing pressure beca<strong>use</strong> CO 2saturation has been reached at 100 kPa whereas CH 4 adsorption continues to occur. However, at200°C <strong>the</strong> CO 2 /CH 4 separation factor remains relatively constant <strong>for</strong> <strong>the</strong> pressure range studiedbeca<strong>use</strong> saturation <strong>of</strong> nei<strong>the</strong>r component has occurred and thus <strong>the</strong> relative adsorption <strong>of</strong> CO 2and CH 4 from <strong>the</strong> bulk phase is <strong>the</strong> same.This result has surprising consequences <strong>for</strong> <strong>the</strong> application <strong>of</strong> CO 2 /CH 4 separation in that <strong>the</strong>CO 2 /CH 4 adsorptive selectivity is similar whe<strong>the</strong>r operating at a feed <strong>temperature</strong> <strong>of</strong> 35 or100°C and <strong>for</strong> feed pressures <strong>of</strong> 1000 kPa or greater.122


5.2.2.2 Carbon Dioxide / NitrogenThe simulated adsorption iso<strong>the</strong>rms <strong>of</strong> CO 2 and N 2 from a CO 2 :N 2 (10:90 vol%) mixed gas at 35and 200°C are shown in Figure 5-7. As expected from <strong>the</strong> pure gas simulation results, <strong>the</strong>adsorption <strong>of</strong> both components decreases with increasing operating <strong>temperature</strong>.8735°C CO2 200°C CO235°C N2 200°C N26Loading (mmol/gram)5432200°C CO 2200°C N 2100 0 1 10 100 1000 10000Pressure (kPa)Figure 5-7 : Adsorption iso<strong>the</strong>rms <strong>of</strong> mixed CO 2 /N 2 onto <strong>nanoporous</strong> <strong>carbon</strong>. Increasing<strong>the</strong> <strong>temperature</strong> decreases <strong>the</strong> amount <strong>of</strong> CO 2 and N 2 adsorbed.From Figures 5-1 and 5-3, saturation <strong>of</strong> CO 2 at 35°C occurs at around 100 to 1000 kPa andsaturation <strong>of</strong> N 2 occurs at a pressure >> 10,000 kPa. Similarly, <strong>for</strong> <strong>the</strong> CO 2 /N 2 mixed gasadsorption at 35°C, as shown in Figure 5-7, <strong>the</strong> NPC surface is saturated with CO 2 molecules ataround 1000 kPa whilst N 2 adsorption continues beyond <strong>the</strong> pressure range shown.At a system <strong>temperature</strong> <strong>of</strong> 200°C, adsorption is greatly reduced and as such much <strong>high</strong>erpressures are required <strong>for</strong> surface saturation to occur. From Figures 5-1 and 5-3, saturation <strong>of</strong>CO 2 and N 2 molecules onto <strong>the</strong> NPC surface at 200°C does not occur at pressures lower than10,000 kPa. Likewise, in <strong>the</strong> mixed gas scenario as shown in Figure 5-7, saturation has not beenachieved <strong>for</strong> ei<strong>the</strong>r component <strong>for</strong> <strong>the</strong> pressure range shown.The impact on <strong>the</strong> CO 2 /N 2 mixed gas separation factor with increasing pressure <strong>for</strong> system<strong>temperature</strong>s <strong>of</strong> 35 and 200°C is presented in Figure 5-8. In all cases, <strong>the</strong> simulated NPC isselective <strong>for</strong> CO 2 with <strong>the</strong> CO 2 /N 2 separation factor being greater than 1.123


8070Feed Temperature = 35°C Feed Temperature = 200°CCO 2 /N 2 Separation Factor60504030201001 10 100 1000 10000 100000Pressure (kPa)Figure 5-8 : CO 2 /N 2 mixed gas adsorption separation factor as a function <strong>of</strong> pressure <strong>for</strong>system <strong>temperature</strong>s <strong>of</strong> 35 and 200°C.At both 35 and 200°C, <strong>the</strong> CO 2 /N 2 separation factor is relatively constant <strong>for</strong> <strong>the</strong> pressure rangebeca<strong>use</strong> saturation <strong>of</strong> ei<strong>the</strong>r component has not occurred and thus <strong>the</strong> relative adsorption <strong>of</strong> CO 2and N 2 from <strong>the</strong> bulk phase is <strong>the</strong> same.The absolute values <strong>of</strong> <strong>the</strong> CO 2 /N 2 separation factors are significantly <strong>high</strong>er than <strong>the</strong>corresponding CO 2 /CH 4 separation factors. This result concurs with <strong>the</strong> larger difference in <strong>the</strong>pure gas CO 2 and N 2 heats <strong>of</strong> adsorptions (Section 5.2.1.1). Whilst <strong>the</strong>re is a substantialdecrease in <strong>the</strong> CO 2 /N 2 separation factor when increasing <strong>the</strong> operating <strong>temperature</strong> from 35 to200 °C, it is still a promising result <strong>for</strong> <strong>the</strong> separation <strong>of</strong> CO 2 from N 2 at <strong>high</strong> <strong>temperature</strong>s.124


5.3 Gas Adsorption Experiments5.3.1 Pure Gases5.3.1.1 Carbon DioxideThe adsorption <strong>of</strong> pure CO 2 on <strong>nanoporous</strong> <strong>carbon</strong> measured experimentally at 35 and 200°C ispresented in Figure 5-9 below. The adsorption behaviour follows a Type 1 iso<strong>the</strong>rm with nohysteresis as is expected <strong>for</strong> a material with nano-sized pores [44].CO 2 Uptake (mmol/gram NPC)6.05.04.03.02.01.0Adsorption at 35°C Desorption at 35°C Langmuir Fit at 35°CAdsorption at 200°C Desorption at 200°C Langmuir Fit at 200°C0.00 200 400 600 800 1000 1200 1400 1600 1800Pressure (kPa)Figure 5-9 : Experimental adsorption iso<strong>the</strong>rms <strong>of</strong> pure CO 2 onto <strong>nanoporous</strong> <strong>carbon</strong>.The adsorption iso<strong>the</strong>rms were <strong>the</strong>n fitted using <strong>the</strong> Langmuir Equation, as described in Section2.2.2.3. The values <strong>of</strong> <strong>the</strong> Langmuir constants, b and V m are given in Table 5-3.Table 5-3 : Experimental values <strong>of</strong> <strong>the</strong> Langmuir constants (b and V m ) <strong>for</strong> <strong>the</strong> adsorptioniso<strong>the</strong>rms <strong>of</strong> pure CO 2 on <strong>nanoporous</strong> <strong>carbon</strong>.Operating Temp K 308 478°C 35 200b bar -1 0.47 0.02V m mmol/gram NPC 3.6 5.6125


The values <strong>of</strong> b and V m are lower than those determined via pure gas simulation (as presentedearlier in Table 5-1). However, <strong>the</strong> value <strong>of</strong> b is only a little lower than <strong>the</strong> 1.5 bar -1 <strong>for</strong> CO 2adsorption at 30°C on NPC (from pyrolysis <strong>of</strong> cellulose) published by Lagorsse et al [12].A plot <strong>of</strong> <strong>the</strong> experimental results with <strong>the</strong> simulation results presented in Figure 5-10 <strong>high</strong>lightsthat <strong>the</strong> amount <strong>of</strong> adsorption seen experimentally is much lower than that predicted viasimulation.12Adsorption at 35°C ExperimentalAdsorption at 35°C SimulatedCO 2 Uptake (mmol/gram NPC)108642Adsorption at 200°C ExperimentalAdsorption at 200°C Simulated00 200 400 600 800 1000 1200 1400 1600 1800 2000Pressure (kPa)Figure 5-10 : Experimental and simulated adsorption iso<strong>the</strong>rms <strong>of</strong> pure CO 2 onto NPC.There are a few possible explanations <strong>for</strong> <strong>the</strong> difference between <strong>the</strong> simulated and experimentalresults. The first is that <strong>the</strong> coordinates <strong>of</strong> <strong>the</strong> substrate <strong>use</strong>d in <strong>the</strong> simulation was that <strong>of</strong>Schwarzite (C 168 ) not <strong>nanoporous</strong> <strong>carbon</strong>. It is not possible to model <strong>nanoporous</strong> <strong>carbon</strong> directlybeca<strong>use</strong> <strong>of</strong> <strong>the</strong> amorphous structure but <strong>the</strong> curved surface and sp 2 hybridized bonding <strong>of</strong>Schwarzite is similar to that <strong>of</strong> <strong>nanoporous</strong> <strong>carbon</strong> [45]. However, Schwartize has a larger poresize than <strong>nanoporous</strong> <strong>carbon</strong> providing greater internal access to adsorbing gases and hence agreater total amount <strong>of</strong> adsorption at <strong>the</strong> various system pressures.The second possible explanation is that equilibrium was not reached in our adsorptionexperiments. At 35°C, <strong>the</strong> CO 2 adsorption experiment was run with a maximum equilibriumtime <strong>of</strong> 30 mins. Comparing this to work <strong>of</strong> Lagorsse et al [12] who determined an equilibriumtime <strong>of</strong> 30 min <strong>for</strong> CO 2 adsorption on NPC at 35°C and 200 kPa, it can be concluded thatequilibrium has been achieved at 35°C. The equilibrium time <strong>of</strong> 60 mins <strong>use</strong>d in running <strong>the</strong>CO 2 adsorption experiments at 200°C is <strong>the</strong>re<strong>for</strong>e more than sufficient.126


Ano<strong>the</strong>r possible explanation <strong>for</strong> <strong>the</strong> difference in <strong>the</strong> results is an error within <strong>the</strong> buoyancycalculation. As discussed in Section 3.5.2, a correction is made to <strong>the</strong> adsorbed weights to allow<strong>for</strong> <strong>the</strong> buoyed volume <strong>of</strong> <strong>the</strong> sample.The standard method <strong>for</strong> making this correction <strong>use</strong>s non-adsorbing Helium gas to calculate <strong>the</strong>buoyed weight. Ano<strong>the</strong>r method that can be <strong>use</strong>d calculates <strong>the</strong> buoyed volume using <strong>the</strong>masses and densities <strong>of</strong> <strong>the</strong> sample and reference sides as is also discussed in Section 3.5.2.The buoyed volumes calculated using <strong>the</strong>se two methods <strong>for</strong> <strong>the</strong> adsorption <strong>of</strong> CO 2 at 35 and200°C are presented below in Table 5-4. The NPC sample size <strong>use</strong>d at 35°C was 50 mg and 107mg at 200°C.The error bands shown on <strong>the</strong> graphs in this chapter have been calculated to include this error in<strong>the</strong> buoyed volume (detailed in Appendix C Section C.3.1) and are concluded to be small.Table 5-4 : Buoyed volume <strong>for</strong> CO 2 adsorption onto NPC at 35°C and 200°C calculatedfrom <strong>the</strong> methods <strong>of</strong> (1) helium adsorption and (2) equipment masses.Buoyed Volume ml Difference from He Method35°CHelium Adsorption Method 199 0Equipment Masses Method 185 - 7 %200°CHelium Adsorption Method 216 0Equipment Masses Method 221 + 2 %The final explanation is in <strong>the</strong> “freshness” <strong>of</strong> <strong>the</strong> sample. Oxygen chemisorption and waterphysisorption are well reported problems with <strong>nanoporous</strong> <strong>carbon</strong>, as discussed previously inSection 2.7.2.The drying <strong>of</strong> <strong>the</strong> sample at 350°C prior to adsorption testing would have removed most <strong>of</strong> <strong>the</strong>physisorbed water however very <strong>high</strong> <strong>temperature</strong>s are necessary to remove chemisorbedoxygen. The occupation <strong>of</strong> adsorption sites with oxygen were not accounted <strong>for</strong> in <strong>the</strong> molecularsimulations.127


5.3.1.2 NitrogenThe adsorption <strong>of</strong> pure N 2 on <strong>nanoporous</strong> <strong>carbon</strong> measured experimentally at 35 and 200°C ispresented in Figure 5-12. The adsorption follows a Type 1 iso<strong>the</strong>rm behaviour although <strong>the</strong>re issome hysteresis upon desorption at 35°C. The hysteresis is most likely related to incompleteadsorption at 25°C particularly at <strong>the</strong> lower pressures.2.01.81.6Adsorption at 25°C Desorption at 25°C Langmuir Fit at 25°CAdsorption at 200°C Desorption at 200°C Langmuir Fit at 200°CN 2 Uptake (mmol/gram NPC)1.41.21.00.80.60.40.20.00 200 400 600 800 1000 1200 1400 1600 1800Pressure (kPa)Figure 5-11 : Experimental adsorption iso<strong>the</strong>rms <strong>of</strong> pure N 2 onto <strong>nanoporous</strong> <strong>carbon</strong>.The adsorption iso<strong>the</strong>rms were <strong>the</strong>n fitted using <strong>the</strong> Langmuir Equation, as described in Section2.2.2.3. The values <strong>of</strong> <strong>the</strong> Langmuir constants, b and V m are given in Table 5-5.Table 5-5 : Experimental values <strong>of</strong> <strong>the</strong> Langmuir constants (b and V m ) <strong>for</strong> <strong>the</strong> adsorptioniso<strong>the</strong>rms <strong>of</strong> pure N 2 on <strong>nanoporous</strong> <strong>carbon</strong>.Operating Temp K 298 478°C 25 200b bar -1 0.05 0.01V m mmol/gram NPC 3.2 4.7The value <strong>of</strong> b is lower in comparison with 0.19 bar -1 <strong>for</strong> N 2 adsorption at 30°C on NPC (frompyrolysis <strong>of</strong> cellulose) published by Lagorsse et al [12]. The values <strong>of</strong> b and V m are also lowerthan those determined via pure gas simulation (as presented earlier in Table 5-1).128


A plot <strong>of</strong> <strong>the</strong> experimental results with <strong>the</strong> simulation results presented in Figure 5-13 again<strong>high</strong>lights that <strong>the</strong> amount <strong>of</strong> adsorption seen experimentally is significantly lower than thatpredicted via simulation, particularly at 25°C.N 2 Uptake (mmol/gram NPC)1098765432Adsorption at 25°C ExperimentalAdsorption at 25°C SimulatedAdsorption at 30°C Experimental Literature (Lagorsse et al)Adsorption at 200°C ExperimentalAdsorption at 200°C SimulatedAdsorption at 25°C Simulated10Adsorption at 200°C Simulated0 200 400 600 800 1000 1200 1400 1600 1800 2000Pressure (kPa)Figure 5-12 : Experimental and simulated adsorption iso<strong>the</strong>rms <strong>of</strong> pure N 2 onto<strong>nanoporous</strong> <strong>carbon</strong>. The large discrepancy at 35°C is attributed to not achievingequilibrium in <strong>the</strong> adsorption experiments.As discussed in Section 3.5.2, <strong>the</strong> experimental data was collected using a maximumequilibrium time <strong>of</strong> 60 mins at both 25 and 200°C. This equilibrium time was confirmed to besufficient at 200°C but too short at 25°C. Lagorsse et al [12] state that it takes 14 hours <strong>for</strong> <strong>the</strong>adsorption <strong>of</strong> N 2 at 30°C and 40 kPa, confirming that <strong>the</strong> kinetics <strong>of</strong> N 2 adsorption on NPC arevery slow. This explains <strong>the</strong> lower b value obtained relative to <strong>the</strong> Lagorsse et al [12] results.Like <strong>the</strong> CO 2 adsorption data, <strong>the</strong> discrepancy between <strong>the</strong> experimental and simulated data at200°C is most likely due to <strong>the</strong> different substrate (C 168 Schwarzite) <strong>use</strong>d in <strong>the</strong> simulation and<strong>the</strong> reduction in available NPC sample adsorption sites due to exposure to air (oxygen).129


5.3.1.3 Heats <strong>of</strong> AdsorptionThe straight line plots <strong>of</strong> <strong>the</strong> values <strong>of</strong> b determined experimentally at <strong>the</strong> two tested<strong>temperature</strong>s <strong>for</strong> <strong>the</strong> adsorption <strong>of</strong> pure CO 2 and N 2 onto <strong>nanoporous</strong> <strong>carbon</strong> are presented inFigure 5-14. Also included <strong>for</strong> comparison is <strong>the</strong> pure gas CO 2 and N 2 adsorption data publishedby Lagorsse et al <strong>for</strong> NPC manufactured from cellulose [12] and <strong>the</strong> results <strong>of</strong> <strong>the</strong> molecularsimulation presented in Section 5.2.1.6.04.0CO2 (This Work Experimental) CO2 (This Work Simulation) CO2 (Lagorsse et al)Nitrogen (This Work Experimental) N2 (This Work Simulation) Nitrogen (Lagorsse et al)ln(b)2.00.0∆H = 33 kJ/mol∆H = 19 kJ/mol-2.0∆H = 27 kJ/mol∆H = 10 kJ/mol∆H = 19 kJ/mol∆H = 13 kJ/mol-4.0-6.01.5 2.0 2.5 3.0 3.5 4.01000/T (K -1 )Figure 5-13 : Straight line plot <strong>of</strong> 1/T versus <strong>the</strong> simulated Ln(b) <strong>for</strong> pure gases <strong>for</strong>determining <strong>the</strong> heat <strong>of</strong> adsorption (∆H ads ).As seen from <strong>the</strong> calculated heats <strong>of</strong> adsorption shown in Table 5-6, CO 2 has a greater heat <strong>of</strong>adsorption than N 2 in all cases, which indicates that CO 2 is better adsorbed than N 2 on NPC.This result concurs with <strong>the</strong> simulated mixed gas results presented in Section 5.2.2.2 that show agood CO 2 /N 2 separation factor with a 10:90 vol % CO 2 :N 2 feed.The heats <strong>of</strong> adsorption <strong>for</strong> CO 2 and N 2 predicted via molecular simulation are <strong>high</strong>er than thatseen experimentally. This difference is attributed <strong>the</strong> <strong>use</strong> <strong>of</strong> different substrates and to <strong>the</strong>presence <strong>of</strong> chemisorbed oxygen from <strong>the</strong> atmosphere assuming potential adsorption sites <strong>for</strong>CO 2 and N 2 on <strong>the</strong> experimental NPC samples.Interestingly, despite having greater values <strong>of</strong> <strong>the</strong> Langmuir constants and <strong>the</strong>re<strong>for</strong>e greaterextents <strong>of</strong> adsorption, <strong>the</strong> experimental heat <strong>of</strong> adsorption values <strong>for</strong> CO 2 and N 2 on NPCreported by Lagorsse et al [12] are lower than <strong>the</strong> experimental values reported here. The reason130


<strong>for</strong> this difference is not clear. One possible explanation is a difference in <strong>the</strong> internal structureor <strong>the</strong> proportion <strong>of</strong> surface impurities as <strong>the</strong> two <strong>carbon</strong>s are made from different polymers.Table 5-6 : Values <strong>of</strong> <strong>the</strong> experimental and simulated heats <strong>of</strong> adsorption <strong>of</strong> pure gases on<strong>nanoporous</strong> <strong>carbon</strong> compared with values reported in <strong>the</strong> literature.Gas CO 2 N 2∆H ads kJ/mol kJ/molThis Work (Experimental) 27.0 12.6This Work (Simulated) 32.6 19.0NPC from Cellulose [12] 18.6 10.4Graphite [159] 16.8 8.95.3.1.4 Relating <strong>the</strong> CO 2 Heat <strong>of</strong> Adsorption to <strong>the</strong> CO 2 PermeanceThe activation energy <strong>of</strong> permeance (E p ) <strong>for</strong> CO 2 was presented earlier in Section 4.2.5. UsingEquation 2-38 from Section 2.6.1, it is possible to back calculate <strong>the</strong> activation energy <strong>of</strong>diffusion (E d ) <strong>for</strong> CO 2 from <strong>the</strong> CO 2 heat <strong>of</strong> adsorption (∆H ads ) presented above in Table 5-6.As expected when <strong>the</strong> value <strong>of</strong> E p is close to 0, <strong>the</strong> results shown in Table 5-7 indicate that <strong>the</strong>CO 2 diffusion activation energy is similar to that <strong>of</strong> <strong>the</strong> heat <strong>of</strong> adsorption. At lower operating<strong>temperature</strong>s CO 2 adsorption will dominate and at <strong>high</strong>er operating <strong>temperature</strong>s diffusion withdominate. A similar conclusion was reached by Strano and Foley [43], <strong>the</strong> results <strong>of</strong> which arealso given in Table 5-7 <strong>for</strong> comparison.Table 5-7 : A comparison <strong>of</strong> <strong>the</strong> experimental CO 2 heat <strong>of</strong> adsorption (∆H ads ) with <strong>the</strong>activation energy <strong>of</strong> permeance (E p ) <strong>for</strong> a NPC membrane manufactured at 550°C (referSection 4.2.5) to calculate <strong>the</strong> activation energy <strong>of</strong> diffusion (E d ) via Equation 2-38.GasPermeanceActivation Energy(E p )Heat <strong>of</strong>Adsorption(∆H ads )DiffusionActivation Energy(E d )kJ/mol kJ/mol kJ/molCO 2 (This Work) -0.800 27.0 26.2CO 2 [43] -1.43 10.6 9.17131


5.4 Concluding Remarks on Understanding <strong>the</strong> High Temperature Per<strong>for</strong>mance<strong>of</strong> NPC Membranes through Gas Adsorption Molecular Simulation andExperimentsThe simulated and experimental iso<strong>the</strong>rms <strong>for</strong> <strong>the</strong> adsorption <strong>of</strong> pure gases on NPC revealed <strong>the</strong>value <strong>of</strong> <strong>the</strong> heats <strong>of</strong> adsorption are in order <strong>of</strong> CO 2 > CH 4 > N 2 . Mixed gas simulations <strong>of</strong> <strong>the</strong>adsorption <strong>of</strong> CO 2 :CH 4 (90:10 vol%) and CO 2 : N 2 (90:10 vol%) mixtures, attested that CO 2 isindeed more adsorptive than CH 4 and N 2 with adsorptive separation factors greater than 1. Thegreater difference between <strong>the</strong> values <strong>of</strong> heat <strong>of</strong> adsorption <strong>for</strong> <strong>the</strong> CO 2 : N 2 pair compared with<strong>the</strong> CO 2 :CH 4 pair has resulted in a significantly larger separation factor <strong>for</strong> <strong>the</strong> CO 2 :N 2 pair.Comparing <strong>the</strong> experimental and simulation results <strong>for</strong> CO 2 and N 2 revealed a significantdifference in <strong>the</strong> amount <strong>of</strong> pure gas molecules adsorbed to <strong>the</strong> NPC surface. This difference isattributed to <strong>the</strong> different substrate (C 168 Schwarzite) <strong>use</strong>d in <strong>the</strong> simulation and also to <strong>the</strong>possibility <strong>of</strong> a smaller amount <strong>of</strong> adsorption sites available <strong>for</strong> gas adsorption on <strong>the</strong>experimental samples due to oxygen adsorption from <strong>the</strong> air.The adsorption results indicate that <strong>nanoporous</strong> <strong>carbon</strong> shows good promise as a material <strong>for</strong> <strong>the</strong>separation <strong>of</strong> CO 2 from N 2 at <strong>high</strong> <strong>temperature</strong>s as is <strong>the</strong> case <strong>for</strong> <strong>the</strong> syn<strong>the</strong>sis gas purification.However, <strong>the</strong>y also indicate that any exposure <strong>of</strong> <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong> to air may reduce <strong>the</strong>adsorption capacity.132


6 PERFORMANCE OF NPC MEMBRANES IN THE PRESENCE OFCONDENSABLE AND NON-CONDENSABLE IMPURITIES6.1 IntroductionThis chapter details <strong>the</strong> results <strong>of</strong> experiments undertaken to assess <strong>the</strong> impact <strong>of</strong> <strong>the</strong> presence <strong>of</strong>small quantities <strong>of</strong> condensable (water, hexane and toluene) and non-condensable (H 2 S and CO)on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> supported NPC <strong>membranes</strong> manufactured from <strong>the</strong> pyrolysis <strong>of</strong>polyfurfuryl alcohol at 550°C.6.2 Condensable ImpuritiesThe impact <strong>of</strong> <strong>the</strong> presence <strong>of</strong> condensable impurities on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong>must be understood as part <strong>of</strong> assessing <strong>the</strong> suitability <strong>for</strong> commercial applications.For <strong>the</strong> existing polymeric membrane technology, <strong>the</strong> presence <strong>of</strong> condensable impurities canhave a significant effect on reducing <strong>the</strong> per<strong>for</strong>mance [103, 138], which <strong>of</strong>ten makes itnecessary to remove <strong>the</strong> impurities from <strong>the</strong> feed gas (via expensive pretreatment) prior toexposure to <strong>the</strong> membrane.The supported NPC <strong>membranes</strong> manufactured from PFA were exposed to three differentimpurities, namely water, hexane (representing heavier hydro<strong>carbon</strong>s) and toluene (representingaromatic hydro<strong>carbon</strong>s). Water, heavier and aromatic hydro<strong>carbon</strong>s are present as condensableimpurities in <strong>the</strong> separation <strong>of</strong> CO 2 from natural gas (CH 4 ).However, only water is present as a condensable impurity in <strong>the</strong> separation <strong>of</strong> CO 2 fromsyn<strong>the</strong>sis gas (N 2 ). Assuming that <strong>the</strong> interaction <strong>of</strong> water with <strong>the</strong> NPC membrane will besimilar with a CO 2 /N 2 feed as with a CO 2 /CH 4 feed, water impurity tests have been completedwith only a mixed gas CO 2 /CH 4 feed. The impact <strong>of</strong> non-condensable impurities with a mixedgas CO 2 /N 2 feed is discussed in Section 6.3.The concentrations <strong>of</strong> condensable impurities in natural gas depend upon <strong>the</strong> operatingconditions <strong>of</strong> <strong>the</strong> separators <strong>use</strong>d to stabilise <strong>the</strong> natural gas as it is produced from <strong>the</strong> reservoir.The condensable impurities carried over into <strong>the</strong> natural gas overhead from <strong>the</strong> final stage <strong>of</strong>separation will be at saturated levels.In order to assess <strong>the</strong> impact <strong>of</strong> <strong>the</strong> condensable impurities on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> <strong>the</strong> NPC<strong>membranes</strong>, <strong>the</strong> concentration <strong>of</strong> <strong>the</strong> condensable impurity in <strong>the</strong> feed was increased by133


increasing <strong>the</strong> <strong>temperature</strong> <strong>of</strong> <strong>the</strong> saturating vessel containing <strong>the</strong> impurity at a constant pressure<strong>of</strong> ei<strong>the</strong>r 1500 kPag or 1200 kPag. Details <strong>of</strong> <strong>the</strong> <strong>temperature</strong>s and saturation concentrations<strong>use</strong>d were presented previously in Table 3-6.In this work, <strong>the</strong> NPC <strong>membranes</strong> were regenerated overnight under a nitrogen purge (50ml/min) at 200 kPag. Overnight (12 - 16 hours) was deemed sufficient based upon <strong>the</strong> resultspresented in <strong>the</strong> literature as discussed in Section 2.7.3.6.2.1 WaterThe <strong>temperature</strong> <strong>of</strong> <strong>the</strong> saturating vessel containing water at 1200 kPag, was decreased from25°C to 10°C and <strong>the</strong>n to 5°C, resulting in a concentration drop from 2390 ppm (0.24 mol%) to918 ppm (0.09 mol%) and finally 650 ppm (0.07 mol%).Firstly, looking at <strong>the</strong> impact <strong>of</strong> <strong>the</strong> water on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> a NPC membrane over a period<strong>of</strong> time, it is evident that <strong>the</strong>re is an initial drop in <strong>the</strong> CO 2 permeance as <strong>the</strong> membrane isexposed to water impurity as shown in Figure 6-1. The extent <strong>of</strong> this reduction in permeance isgreater with <strong>the</strong> increased concentration <strong>of</strong> water. This trend is observed whe<strong>the</strong>r <strong>the</strong> data isplotted as a percentage <strong>of</strong> <strong>the</strong> original permeance (as done in Figure 6-1) or on an absolute scale.130%Water 2390 ppm, Membrane at 100 °C% <strong>of</strong> Original CO 2 Permeance120%110%100%90%Water 918 ppm, Membrane at 35 °CWater 918 ppm, Membrane at 100 °CWater 650 ppm, Membrane at 100 °C80%70%0 50 100 150 200 250 300Time (min)Figure 6-1 : Normalised effect <strong>of</strong> water concentration on <strong>the</strong> reduction in CO 2 permeance<strong>of</strong> a NPC membrane manufactured at 550°C as a function <strong>of</strong> time.As time progresses <strong>the</strong> CO 2 permeance is somewhat recovered and stabilises after 100 mins.134


The reduction in <strong>the</strong> CH 4 permeance also drops initially although not to <strong>the</strong> same extent as withCO 2 . This results in an initial drop in <strong>the</strong> CO 2 /CH 4 selectivity as shown in Figure 6-2. However,<strong>the</strong> reduction in <strong>the</strong> CH 4 permeance once <strong>the</strong> system has stabilised is similar to that <strong>of</strong> CO 2resulting in no net change in <strong>the</strong> stabilised CO 2 /CH 4 selectivity as also shown in Figure 6-2.130%Water 2390 ppm, Membrane at 100 °C% <strong>of</strong> Original CO 2 /CH 4 Selectivity120%110%100%90%80%Water 918 ppm, Membrane at 35 °CWater 918 ppm, Membrane at 100 °CWater 650 ppm, Membrane at 100 °C70%0 50 100 150 200 250 300Time (min)Figure 6-2 : Normalised effect <strong>of</strong> water concentration on <strong>the</strong> CO 2 /CH 4 selectivity <strong>of</strong> a NPCmembrane manufactured at 550°C as a function <strong>of</strong> time.The stabilised permeance as a function <strong>of</strong> increasing water concentration is presented in Figure6-3. The reduction in permeance is very small even at <strong>the</strong> <strong>high</strong>est concentration <strong>of</strong> water. Theimpact appears to be even smaller at <strong>the</strong> <strong>high</strong>er operating <strong>temperature</strong> due to <strong>the</strong> loweradsorptive strength <strong>of</strong> water although <strong>the</strong> results at 35 and 100°C are within <strong>the</strong> same margin <strong>of</strong>error so it is not possible to make a firm conclusion.These results are quite different to those <strong>of</strong> Jones and Koros [31], who found a significantreduction in <strong>the</strong> O 2 and N 2 flux <strong>of</strong> NPC <strong>membranes</strong> made from polyimide when exposed tohumidified feeds. In <strong>the</strong>ir case, humidified feeds <strong>of</strong> up to 85% relative humidity were <strong>use</strong>d andflux reductions as <strong>high</strong> as 70% were observed. It is possible that <strong>the</strong> physisorbed water alreadypresent upon our <strong>membranes</strong> from <strong>the</strong> exposure to atmosphere during installation is lessening<strong>the</strong> impact when exposed to more water in <strong>the</strong> feed stream.The impact <strong>of</strong> increasing water concentration upon selectivity is very slight (+/- 10%). Theselectivity change is minimal both at 35 and 100°C as shown in Figure 6-4.135


130%120%CH4 Permeance @ 35 °C CO2 Permeance @ 35 °CCH4 Permeance @ 100 °C CO2 Permeance @ 100 °C% <strong>of</strong> Original Permeance110%100%90%80%70%0 500 1000 1500 2000 2500 3000Water Concentration (ppm)Figure 6-3 : Normalised effect <strong>of</strong> water concentration on <strong>the</strong> permeance <strong>of</strong> a NPCmembrane manufactured at 550°C. The effect appears to be more pronounced on CO 2permeance and at lower operating <strong>temperature</strong>s.120%115%CO2/CH4 Selectivity @ 35 °C CO2/CH4 Selectivity @ 100 °C% <strong>of</strong> Original CO 2 /CH 4 Selectivity110%105%100%95%90%85%80%75%70%0 500 1000 1500 2000 2500 3000Water Concentration (ppm)Figure 6-4 : Normalised effect <strong>of</strong> water concentration on <strong>the</strong> selectivity <strong>of</strong> a NPCmembrane manufactured at 550°C. The effect is minimal.136


6.2.2 HexaneThe <strong>temperature</strong> <strong>of</strong> <strong>the</strong> saturating vessel containing hexane at 1200 kPag, was decreased from25°C to 10°C and <strong>the</strong>n to 5°C, resulting in a concentration drop from 15,538 ppm (1.55 mol%)to 7844 ppm (0.78 mol%) and finally 6128 ppm (0.61 mol%).Firstly, looking at <strong>the</strong> impact <strong>of</strong> <strong>the</strong> hexane on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> <strong>the</strong> NPC membrane over aperiod <strong>of</strong> time, it is evident that <strong>the</strong>re is an initial drop in CO 2 per<strong>for</strong>mance as <strong>the</strong> membrane isexposed to hexane impurity as shown in Figure 6-5.120%110%Hexane 15538 ppm, Membrane at 35 °C Hexane 15538 ppm, Membrane at 100 °CHexane 7844 ppm, Membrane at 35 °C Hexane 7844 ppm, Membrane at 100 °CHexane 6128 ppm, Membrane at 35 °C Hexane 6128 ppm, Membrane at 100 °C% <strong>of</strong> Original CO 2 Permeance100%90%80%70%60%0 50 100 150 200 250Time (min)Figure 6-5 : Normalised effect <strong>of</strong> hexane concentration on <strong>the</strong> reduction in CO 2 permeance<strong>of</strong> a NPC membrane manufactured at 550°C as a function <strong>of</strong> time.The extent <strong>of</strong> <strong>the</strong> reduction in per<strong>for</strong>mance is larger with <strong>the</strong> increased concentration <strong>of</strong> hexanebut not necessarily with <strong>the</strong> decreased operating <strong>temperature</strong> as expected. After 60 mins, <strong>the</strong>permeance is somewhat recovered but continues to <strong>the</strong>n gradually decrease particularly at <strong>the</strong>operating <strong>temperature</strong> <strong>of</strong> 35°C. In <strong>the</strong> longer term, <strong>the</strong> data indicates (albeit within <strong>the</strong> margin <strong>of</strong>error) that <strong>the</strong> impact on <strong>the</strong> reduction in permeance is less at <strong>the</strong> operating <strong>temperature</strong> <strong>of</strong>100°C than 35°C. This is probably due to <strong>the</strong> lower adsorptive strength <strong>of</strong> hexane at <strong>the</strong> <strong>high</strong>eroperating <strong>temperature</strong>.Interestingly, when <strong>the</strong> NPC membrane is exposed to hexane <strong>the</strong>re is little change in <strong>the</strong> initialCH 4 permeance resulting in a sharp drop in <strong>the</strong> CO 2 /CH 4 selectivity as shown in Figure 6-6.However, <strong>the</strong> CH 4 permeance <strong>the</strong>n declines in a similar manner to <strong>the</strong> CO 2 permeance resultingin no net change in <strong>the</strong> stabilised CO 2 /CH 4 selectivity as also shown in Figure 6-6.137


% <strong>of</strong> Original CO 2 /CH 4 Selectivity120%110%100%90%80%70%Hexane 15538 ppm, Membrane at 35 °C Hexane 15538 ppm, Membrane at 100 °CHexane 7844 ppm, Membrane at 35 °C Hexane 7844 ppm, Membrane at 100 °CHexane 6128 ppm, Membrane at 35 °C Hexane 6128 ppm, Membrane at 100 °C60%0 50 100 150 200 250 300 350Time (min)Figure 6-6 : Normalised effect <strong>of</strong> hexane concentration on <strong>the</strong> CO 2 /CH 4 permeance <strong>of</strong> aNPC membrane manufactured at 550°C as a function <strong>of</strong> time.The stabilised permeance as a function <strong>of</strong> hexane concentration is presented in Figure 6-7.There is a trend <strong>of</strong> reducing permeance with increasing concentration at 35°C although this isless obvious at 100°C. Despite <strong>the</strong> uncertainty in <strong>the</strong> permeance measurement, it can beconcluded that up to hexane concentrations <strong>of</strong> ~ 15,600 ppm (1.5 mol%) <strong>the</strong> reduction in <strong>the</strong>CO 2 and CH 4 permeance is limited less than 20 %.These results <strong>for</strong> <strong>the</strong> lowest concentration <strong>of</strong> hexane are similar those <strong>of</strong> Vu et al [35] who alsoinvestigated <strong>the</strong> impact <strong>of</strong> condensable impurities on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> <strong>for</strong><strong>the</strong> separation <strong>of</strong> CO 2 and CH 4 . Whilst Vu et al [35], <strong>use</strong>d n-heptane to represent a straight chainhydro<strong>carbon</strong> ra<strong>the</strong>r than n-hexane, <strong>the</strong>ir results showed a small decrease in permeance with lowconcentrations <strong>of</strong> <strong>the</strong> impurity (100 to 300 ppm). The effect <strong>of</strong> a larger concentration <strong>of</strong> <strong>the</strong>impurity was not investigated in <strong>the</strong>ir publication.The impact <strong>of</strong> <strong>the</strong> presence <strong>of</strong> hexane on <strong>the</strong> CO 2 /CH 4 selectivity is relatively small as presentedin Figure 6-8. Whilst it was commented in <strong>the</strong> previous paragraphs that <strong>the</strong> reduction in <strong>the</strong> CO 2permeance was slightly greater, <strong>the</strong> difference is not large enough to dramatically affect <strong>the</strong>selectivity. Despite <strong>the</strong> uncertainly in <strong>the</strong> permeance measurement as depicted by <strong>the</strong> error band,it can still be concluded that up to hexane concentrations <strong>of</strong> ~ 15600 ppm (1.5 mol%), <strong>the</strong>reduction in <strong>the</strong> selectivity is limited to less than 10%.138


This result is similar to that <strong>of</strong> Vu et al [35] who observed no net change in <strong>the</strong> CO 2 /CH 4selectivity <strong>of</strong> an NPC membrane made from polyimide when exposed to heptane.130%120%CH4 Permeance @ 35 °C CO2 Permeance @ 35 °CCH4 Permeance @ 100 °C CO2 Permeance @ 100 °C% <strong>of</strong> Original Permeance110%100%90%80%70%60%0 2000 4000 6000 8000 10000 12000 14000 16000 18000Hexane Concentration (ppm)Figure 6-7: Normalised effect <strong>of</strong> hexane concentration on <strong>the</strong> permeance <strong>of</strong> a NPCmembrane manufactured at 550°C. The effect appears to be more pronounced on <strong>the</strong> CO 2permeance and at lower operating <strong>temperature</strong>s.130%CO2/CH4 Selectivity @ 35 °C CO2/CH4 Selectivity @ 100 °C% <strong>of</strong> Original CO 2 /CH 4 Selectivity120%110%100%90%80%70%60%0 2000 4000 6000 8000 10000 12000 14000 16000 18000Hexane Concentration (ppm)Figure 6-8 : Normalised effect <strong>of</strong> hexane concentration on <strong>the</strong> CO 2 /CH 4 selectivity <strong>of</strong> aNPC membrane manufactured at 550°C. The impact is minimal.139


6.2.3 TolueneThe <strong>temperature</strong> <strong>of</strong> <strong>the</strong> saturating vessel containing toluene at 1500 kPag, was decreased from25°C to 10°C and <strong>the</strong>n to 5°C, resulting in a concentration drop from 2549 ppm (0.25 mol%) to1146 ppm (0.11 mol%) and finally 860 ppm (0.09 mol%).Firstly, looking at <strong>the</strong> impact <strong>of</strong> <strong>the</strong> toluene on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> <strong>the</strong> NPC membrane over aperiod <strong>of</strong> time, it is evident that <strong>the</strong>re is also a definite initial drop in CO 2 permeance as <strong>the</strong>membrane is exposed to toluene impurity as shown in Figure 6-9.120%110%Toluene 2549 ppm, Membrane at 35 °C Toluene 2549 ppm, Membrane at 100 °CToluene 1146 ppm, Membrane at 35 °C Toluene 1146 ppm, Membrane at 100 °CToluene 860 ppm, Membrane at 35 °C Toluene 860 ppm, Membrane at 100 °C% <strong>of</strong> Original CO 2 Permeance100%90%80%70%60%50%40%0 50 100 150 200 250 300 350Time (min)Figure 6-9 : Normalised effect <strong>of</strong> toluene concentration on <strong>the</strong> reduction in CO 2permeance <strong>of</strong> a NPC membrane manufactured at 550°C as a function <strong>of</strong> time.After 150 mins, <strong>the</strong> permeance is somewhat recovered and stabilised. The impact on <strong>the</strong>reduction in permeance appears to less at <strong>the</strong> lower concentrations <strong>of</strong> toluene and at <strong>the</strong> <strong>high</strong>eroperating <strong>temperature</strong> <strong>of</strong> 100°C. This is due to <strong>the</strong> lower adsorptive strength <strong>of</strong> toluene at <strong>the</strong><strong>high</strong>er operating <strong>temperature</strong>.Like <strong>the</strong> results with hexane exposure, when <strong>the</strong> NPC membrane is exposed to toluene <strong>the</strong>re islittle change in <strong>the</strong> initial CH 4 permeance resulting in a sharp drop in <strong>the</strong> CO 2 /CH 4 selectivity asshown in Figure 6-10. However, <strong>the</strong> CH 4 permeance <strong>the</strong>n declines in a similar manner to <strong>the</strong>CO 2 permeance resulting in no net change in <strong>the</strong> stabilised CO 2 /CH 4 selectivity as also shown inFigure 6-10.140


% <strong>of</strong> Original CO 2 /CH 4 Selectivity140%130%120%110%100%90%80%70%60%50%Toluene 2549, Membrane at 35 °C Toluene 2549 ppm, Membrane at 100 °CToluene 1146 ppm, Membrane at 35 °C Toluene 1146 ppm, Membrane at 100 °CToluene 860 ppm, Membrane at 35 °C Toluene 860 ppm, Membrane @100 °C40%0 50 100 150 200 250 300 350Time (min)Figure 6-10 : Normalised effect <strong>of</strong> toluene concentration on <strong>the</strong> CO 2 /CH 4 permeance <strong>of</strong> aNPC membrane manufactured at 550°C as a function <strong>of</strong> time.The stabilised value permeance as a function <strong>of</strong> toluene concentration is presented in Figure 6-11. There is a more definitive trend <strong>of</strong> reducing permeance with increasing concentration,particularly at 35°C.The impact upon <strong>the</strong> CO 2 permeance is greater than <strong>the</strong> CH 4 permeance at both 35 and 100°Cand more so than was observed with hexane. The greater reduction in CO 2 permeance in <strong>the</strong>presence <strong>of</strong> toluene ra<strong>the</strong>r than hexane is presumably due to <strong>the</strong> ability <strong>of</strong> toluene to assumemore available adsorption sites and reduce CO 2 adsorption.The impact <strong>of</strong> <strong>the</strong> presence <strong>of</strong> toluene on <strong>the</strong> CO 2 /CH 4 selectivity, as presented in Figure 6-12,although small appears to be more pronounced than both water and hexane. It can be concludedthat up to toluene concentrations <strong>of</strong> ~ 2500 ppm (0.25 mol%), <strong>the</strong> reduction in <strong>the</strong> selectivity islimited to less than 15%.This result is similar to that <strong>of</strong> Vu et al [35] who observed no net change in <strong>the</strong> CO 2 /CH 4selectivity <strong>of</strong> an NPC membrane made from polyimide when exposed to toluene. Using mixedmatrix <strong>membranes</strong> made from NPC and polyimide, Vu et al [136] also reported a minimalimpact on <strong>the</strong> per<strong>for</strong>mance in <strong>the</strong> presence <strong>of</strong> toluene although <strong>the</strong> toluene concentrations werevery small (70.4 ppm).141


However, Tanihara et al [103] who have <strong>use</strong>d much <strong>high</strong>er concentrations <strong>of</strong> toluene (7500ppm) reported a minor change in <strong>the</strong> H 2 /CH 4 selectivity <strong>of</strong> an NPC membrane. This is incomparison to <strong>the</strong> significant change in <strong>the</strong> H 2 /CH 4 selectivity <strong>of</strong> un-pyrolysised polyimidemembrane when exposed to toluene [103].130%120%CH4 Permeance 35 °C CO2 Permeance 35 °CCH4 Permeance 100 °C CO2 Permeance 100 °C% <strong>of</strong> Original Permeance110%100%90%80%70%60%0 500 1000 1500 2000 2500Toluene Concentration (ppm)Figure 6-11 : Normalised effect <strong>of</strong> toluene concentration on <strong>the</strong> permeance <strong>of</strong> a NPCmembrane manufactured at 550°C. The effect is more pronounced on <strong>the</strong> CO 2 permeanceand at lower operating <strong>temperature</strong>s.130%120%CO2/CH4 Selectivity 35 °C CO2/CH4 Selecitivity 100 °C% <strong>of</strong> Original Selectivity110%100%90%80%70%60%0 500 1000 1500 2000 2500Toluene Concentration (ppm)Figure 6-12 : Normalised effect <strong>of</strong> toluene concentration on <strong>the</strong> selectivity <strong>of</strong> a NPCmembrane manufactured at 550°C. The effect is minimal.142


6.2.4 Comparison <strong>of</strong> Condensable ImpuritiesAs a summary, <strong>the</strong> reduction in <strong>the</strong> CO 2 permeance and CO 2 /CH 4 selectivity <strong>for</strong> eachcondensable impurity and concentration level tested are presented in Tables 6-1 and 6-2. Thepercentage errors on each data point are +/- 7 %.The reduction in NPC membrane per<strong>for</strong>mance is greatest upon exposure to toluene <strong>the</strong>n hexaneand lowest upon exposure to water. Likewise, it appears that <strong>the</strong> impact <strong>of</strong> <strong>the</strong> impurities isgreatest at <strong>the</strong> lower operating <strong>temperature</strong> where impurity adsorption is <strong>high</strong>est.The data is <strong>of</strong> a similar magnitude as <strong>the</strong> error margin making it difficult to make absoluteconclusions. However, it can be concluded that <strong>the</strong> case with <strong>the</strong> worst reduction in per<strong>for</strong>mance(2549 ppm <strong>of</strong> toluene at 35°C) is still only a 30% reduction on <strong>the</strong> original CO 2 permeance anda 20% reduction on <strong>the</strong> original CO 2 /CH 4 selectivity.Table 6-1 : Normalised stabilised reduction in CO 2 permeance with impurity presence.Impurity Feed Temp. Impurity Conc. Impurity Conc. Impurity Conc.°C ppm ppm ppmWater 650 918 2390Water 100 0 % 2 % 1 %Hexane 4980 6375 12627Hexane 35 6 % 13 % 13 %Hexane 100 0 % 2 % 5 %Toluene 860 1147 2549Toluene 35 7 % 18 % 24 %Toluene 100 6 % 13 % 12 %Table 6-2 : Normalised stabilised reduction in CO 2 /CH 4 selectivity with impurity presence.Impurity Feed Temp. Impurity Conc. Impurity Conc. Impurity Conc.°C ppm ppm ppmWater 650 918 2390Water 100 0 % 0 % 0 %Hexane 4980 6375 12627Hexane 35 2 % 3 % 1 %Hexane 100 2 % 1 % 4 %Toluene 860 1147 2549Toluene 35 6 % 13 % 12 %Toluene 100 5 % 4 % 3 %143


6.3 Non-Condensable ImpuritiesThe impact <strong>of</strong> <strong>the</strong> presence <strong>of</strong> two non-condensable impurities, hydrogen sulphide (H 2 S) and<strong>carbon</strong> monoxide (CO), on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> has been assessed.Hydrogen sulphide (H 2 S) is present as an impurity in CO 2 removal from natural gas and fromsyn<strong>the</strong>sis gas. The concentration <strong>of</strong> H 2 S in natural gas is dependent on <strong>the</strong> amount <strong>of</strong> sulphideproducing bacteria present in <strong>the</strong> reservoir and can range from only a few ppm to a severalmol%. Likewise, <strong>the</strong> concentration <strong>of</strong> H 2 S in syn<strong>the</strong>sis gas is dependent on <strong>the</strong> concentration <strong>of</strong>sulphur in <strong>the</strong> original hydro<strong>carbon</strong> feed stock.Carbon Monoxide (CO) is present in syn<strong>the</strong>sis gas as a product from <strong>the</strong> water gas shiftreaction. The concentration <strong>of</strong> CO in syn<strong>the</strong>sis gas can be as <strong>high</strong> as several mol percent.Both gases are extremely toxic and <strong>of</strong>ten poisonous to catalysts and materials. There<strong>for</strong>e, it is<strong>of</strong>ten necessary to remove or reduce <strong>the</strong>se impurities via pre-treatment. If a <strong>membranes</strong>eparation material, such as <strong>nanoporous</strong> <strong>carbon</strong>, is not adversely affected by <strong>the</strong> presence <strong>of</strong> H 2 Sor CO, a significant reduction in pre-treatment equipment, footprint and cost is possible.6.3.1 Hydrogen SulphideTo reduce <strong>the</strong> risk <strong>of</strong> H 2 S exposure to lab personnel in <strong>the</strong> instance <strong>of</strong> a leak, <strong>the</strong> experimentswere conducted using a pre-purchased bottle <strong>of</strong> CO 2 /N 2 with a specified 500 ppm <strong>of</strong> H 2 S.There<strong>for</strong>e, it was not possible to change <strong>the</strong> gas phase concentration. It is assumed that <strong>the</strong>interaction <strong>of</strong> H 2 S with <strong>the</strong> NPC membrane will be similar with a CO 2 /CH 4 feed as with aCO 2 /N 2 feed.The impact <strong>of</strong> <strong>the</strong> H 2 S on <strong>the</strong> CO 2 permeance <strong>of</strong> an NPC membrane manufactured at 550°Cover time is illustrated in Figure 6-13. The data has been normalised to <strong>the</strong> original value <strong>of</strong>permeance <strong>for</strong> <strong>the</strong> each operating <strong>temperature</strong> and pressure <strong>for</strong> a clearer comparison.There is an initial decline in <strong>the</strong> CO 2 permeance which <strong>the</strong>n stabilises around 100 mins. Thestabilised drop in <strong>the</strong> CO 2 permeance is within 20 % <strong>of</strong> <strong>the</strong> original value at feed operating<strong>temperature</strong>s <strong>of</strong> 35 and 100°C and feed pressures <strong>of</strong> 500 and 1000 kPag. The impact <strong>of</strong> H 2 S on<strong>the</strong> N 2 permeance is very similar to that <strong>of</strong> CO 2 resulting in a small drop (


Whilst <strong>the</strong> reduction in permeance and selectivity <strong>of</strong> <strong>the</strong> membrane is minimal, <strong>the</strong> results dosuggest that H 2 S is adsorbing onto <strong>the</strong> <strong>nanoporous</strong> <strong>carbon</strong> and slightly inhibiting <strong>the</strong> transport <strong>of</strong>CO 2 and N 2 . This selective adsorption <strong>of</strong> H 2 S is in agreement with <strong>the</strong> experimental adsorptionstudies <strong>of</strong> H 2 S on activated <strong>carbon</strong> by Ritter and Yang [50] who observed that H 2 S was moreadsorptive than CO 2 .120%115%35°C 1000 kPa 35°C 500 kPa% <strong>of</strong> Original CO 2 Permeance110%105%100%95%90%85%80%75%100°C 1000 kPa 100°C 500 kPa70%0 50 100 150 200 250 300 350 400Time (min)Figure 6-13 : Normalised effect <strong>of</strong> H 2 S on <strong>the</strong> reduction in CO 2 permeance <strong>of</strong> an NPCmembrane manufactured at 550°C. The effect is minimal.120%% <strong>of</strong> Original CO 2 /N 2 Selectivity115%110%105%100%95%90%85%80%75%35°C 1000 kPa 35°C 500 kPa100°C 1000 kPa 100°C 500 kPa70%0 50 100 150 200 250 300 350Time (min)Figure 6-14 : Normalised effect <strong>of</strong> H 2 S on <strong>the</strong> reduction in CO 2 /N 2 selectivity <strong>of</strong> an NPCmembrane manufactured at 550°C. The effect is minimal.145


The impact <strong>of</strong> feed <strong>temperature</strong> and pressure on <strong>the</strong> reduction in CO 2 permeance with increasingH 2 S partial pressure is not distinctive. It was expected that with increasing feed <strong>temperature</strong>, <strong>the</strong>impact <strong>of</strong> H 2 S would be less as <strong>the</strong> adsorption <strong>of</strong> H 2 S should be lower at <strong>high</strong>er <strong>temperature</strong>s.The NPC membrane per<strong>for</strong>mance results show that <strong>the</strong> impact <strong>of</strong> H 2 S is fairly minimal. This isa very promising result in comparison with <strong>the</strong> significant impact H 2 S has upon reducing <strong>the</strong>per<strong>for</strong>mance <strong>of</strong> Cu/Pd <strong>membranes</strong> [139] and plasticising polymeric <strong>membranes</strong> [138].However, fur<strong>the</strong>r work is required to assess <strong>the</strong> long term stability <strong>of</strong> NPC <strong>membranes</strong> in <strong>the</strong>presence <strong>of</strong> H 2 S and also at <strong>high</strong>er concentrations <strong>of</strong> H 2 S. Studies [141] have shown that H 2 Sadsorption on activated <strong>carbon</strong> in <strong>the</strong> presence <strong>of</strong> adsorbed oxygen and water has lead to <strong>the</strong><strong>for</strong>mation <strong>of</strong> sulphuric acid which degrades <strong>the</strong> <strong>carbon</strong>.As discussed earlier in Section 2.7.2, NPC <strong>membranes</strong> readily adsorb oxygen and water whenexposed to air. If NPC <strong>membranes</strong> are stored or exposed to air prior to installation and <strong>the</strong>nexposed to H 2 S during operation, rapid deterioration <strong>of</strong> <strong>the</strong> <strong>membranes</strong> could occur. Longerexperiments could not be run as part <strong>of</strong> this research work due to <strong>the</strong> associated health andsafety risks.Additionally, it will be necessary to determine <strong>the</strong> permeance <strong>of</strong> H 2 S through NPC <strong>membranes</strong>in order to determine whe<strong>the</strong>r H 2 S will concentrate in <strong>the</strong> retentate or <strong>the</strong> permeate stream. It isexpected that H 2 S will concentrate in <strong>the</strong> permeate stream as it is a smaller and more adsorptivemolecule than CO 2 . This could not be done as part <strong>of</strong> this research work as <strong>the</strong> concentrationswere too low to be measured by <strong>the</strong> Gas Chromatograph and increasing <strong>the</strong> H 2 S concentrationbeyond 500 ppm posed a significant health and safety risk.6.3.2 Carbon MonoxideHealth and safety risks associated with CO did not allow testing with CO levels as <strong>high</strong> as thosefound in syn<strong>the</strong>sis gas (2.8 mol% refer Table 1-1 in Chapter 1). A risk assessment was approvedon <strong>the</strong> basis <strong>of</strong> a feed gas CO concentration <strong>of</strong> 1000 ppm.The impact <strong>of</strong> <strong>the</strong> presence CO on <strong>the</strong> CO 2 permeance <strong>of</strong> an NPC membrane manufactured at550°C over a period <strong>of</strong> time is illustrated in Figure 6-15. As with H 2 S, <strong>the</strong> data has beennormalised to <strong>the</strong> original value <strong>of</strong> permeance <strong>for</strong> <strong>the</strong> particular feed operating <strong>temperature</strong> andpressure in order to clearly identify <strong>the</strong> trends.146


There is an initial decline in <strong>the</strong> CO 2 permeance which <strong>the</strong>n stabilises in around 50 to 100 mins.The stabilised drop in <strong>the</strong> CO 2 permeance is within 30 % <strong>of</strong> <strong>the</strong> original value at feed operating<strong>temperature</strong>s <strong>of</strong> 35 and 100°C and feed pressures <strong>of</strong> 500 and 1000 kPag. The greatest drop inpermeance is observed at feed conditions <strong>of</strong> 35°C and 500 kPag.120%115%35°C 1000 kPa 35°C 500 kPa% <strong>of</strong> Original CO 2 Permeance110%105%100%95%90%85%80%75%100°C 1000 kPa 100°C 500 kPa70%0 50 100 150 200 250 300 350Time (min)Figure 6-15 : Normalised effect <strong>of</strong> CO on <strong>the</strong> reduction in CO 2 permeance <strong>of</strong> an NPCmembrane manufactured at 550°C. The effect is greatest at <strong>the</strong> lowest pressure and<strong>temperature</strong>.The impact <strong>of</strong> CO on <strong>the</strong> N 2 permeance over time is slightly lower than that <strong>of</strong> CO 2 resulting ina small drop (


The adsorption studies <strong>of</strong> H 2 S and CO on activated <strong>carbon</strong> by Ritter and Yang [50] observedthat CO is less adsorptive than CO 2 and much less than H 2 S suggesting that unlike H 2 S, CO willnot be competitively adsorbed over CO 2 . Although in this case, <strong>the</strong> impact on <strong>the</strong> permeance asshown in Figure 6-15 suggests that CO is adsorbing to <strong>the</strong> surface particularly at <strong>the</strong> lower<strong>temperature</strong> and pressure, where adsorption is greater.However, <strong>the</strong> NPC membrane per<strong>for</strong>mance results show that impact <strong>of</strong> CO is fairly minimal.This is a promising result <strong>for</strong> <strong>the</strong> <strong>use</strong> <strong>of</strong> NPC <strong>membranes</strong> in CO 2 capture from syn<strong>the</strong>sis gas.Fur<strong>the</strong>r work is required to assess <strong>the</strong> long term stability <strong>of</strong> NPC <strong>membranes</strong> in <strong>the</strong> presence <strong>of</strong>CO and at <strong>high</strong>er concentrations <strong>of</strong> CO. Longer experiments could not be run as part <strong>of</strong> thisresearch work due to <strong>the</strong> associated health and safety risks.Additionally, it will be necessary to determine <strong>the</strong> permeance <strong>of</strong> CO through NPC <strong>membranes</strong>in order to determine whe<strong>the</strong>r CO will concentrate in <strong>the</strong> retentate as is <strong>the</strong> case <strong>for</strong> polymeric<strong>membranes</strong> [138] or <strong>the</strong> permeate stream. It is expected that CO will concentrate in <strong>the</strong> retentatestream as it is a larger and less adsorptive molecule than CO 2 . This could not be done as part <strong>of</strong>this research work as <strong>the</strong> concentrations were too low to be measured and increasing <strong>the</strong> COconcentration beyond 1000 ppm posed a significant health and safety risk.148


6.3.3 Comparison <strong>of</strong> Non-Condensable ImpuritiesAs a summary, <strong>the</strong> stabilised reduction in <strong>the</strong> CO 2 permeance and CO 2 /N 4 selectivity <strong>for</strong> eachcondensable impurity and concentration level tested are presented in Tables 6-3 and 6-4. Thepercentage errors on each data point are +/- 7 %.The reduction in NPC membrane per<strong>for</strong>mance is slightly greater upon exposure to CO than H 2 S.However, this is most likely due to <strong>the</strong> <strong>high</strong>er feed concentration <strong>of</strong> CO (1000 ppm) comparedwith H 2 S (500 ppm).Table 6-3 : Normalised stabilised reduction in CO 2 permeance as a function <strong>of</strong> impurity.ImpurityOperatingTemperatureImpurityPartial PressureImpurityPartial Pressure°C kPaa kPaaHydrogen Sulphide 0.30 0.55Hydrogen Sulphide 35 7 % 5 %Hydrogen Sulphide 100 5 % 11 %Carbon Monoxide 0.60 1.10Carbon Monoxide 35 22 % 8 %Carbon Monoxide 100 12 % 10 %Table 6-4 : Normalised stabilised reduction in CO 2 /N 2 selectivity as a function <strong>of</strong> impurity.ImpurityOperatingTemperatureImpurityPartial PressureImpurityPartial Pressure°C kPaa kPaaHydrogen Sulphide 0.30 0.55Hydrogen Sulphide 35 2 % 4 %Hydrogen Sulphide 100 1 % 3 %Carbon Monoxide 0.60 1.10Carbon Monoxide 35 4 % - 1%Carbon Monoxide 100 2 % 2 %149


6.4 Concluding Remarks on <strong>the</strong> Per<strong>for</strong>mance <strong>of</strong> NPC Membranes in <strong>the</strong>Presence <strong>of</strong> ImpuritiesThe presence <strong>of</strong> condensable and non-condensable impurities on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> supportedNPC <strong>membranes</strong> made from <strong>the</strong> pyrolysis <strong>of</strong> polyfurfuryl alcohol at 550°C is relatively minor,particularly at elevated <strong>temperature</strong>s. This is a very promising result <strong>for</strong> <strong>the</strong> commercial <strong>use</strong> <strong>of</strong>NPC <strong>membranes</strong> <strong>for</strong> <strong>the</strong> separation <strong>of</strong> CO 2 from CH 4 and N 2 .Upon exposure to condensable impurities, <strong>the</strong> CO 2 permeance <strong>of</strong> <strong>the</strong> NPC <strong>membranes</strong> isreduced by up to 24 % (+/- 7 %) with <strong>the</strong> greatest reduction being observed in <strong>the</strong> presence <strong>of</strong>toluene, followed by hexane and <strong>the</strong>n water. The reduction in <strong>the</strong> CH 4 permeance is not assignificant as that seen <strong>for</strong> CO 2 due to <strong>the</strong> lower dependence <strong>of</strong> <strong>the</strong> CH 4 permeance on surfaceadsorption. However, <strong>the</strong> reduction in <strong>the</strong> CO 2 /CH 4 selectivity is less than 13 % (+/- 7%).The per<strong>for</strong>mance reductions ca<strong>use</strong>d by water are significantly lower than those reported <strong>for</strong>polymeric <strong>membranes</strong> [138]. However, <strong>the</strong> per<strong>for</strong>mance results are also significantly lower thanthose reported <strong>for</strong> NPC <strong>membranes</strong> by Jones and Koros [31]. This suggests that ourregeneration technique may not be removing <strong>the</strong> majority <strong>of</strong> <strong>the</strong> water adsorbed from <strong>the</strong>atmosphere during installation and hence limiting water adsorption from <strong>the</strong> feed stream.The per<strong>for</strong>mance reductions ca<strong>use</strong>d by <strong>the</strong> presence <strong>of</strong> hexane and toluene are similar to thoseobserved by Vu et al [35] even though <strong>high</strong>er impurity concentrations have been in this researchwork. Tanihara et al [103], who have <strong>use</strong>d a <strong>high</strong>er concentration <strong>of</strong> toluene albeit in a H 2 /CH 4feed stream, have reported that <strong>the</strong> impact on per<strong>for</strong>mance <strong>of</strong> a NPC membrane to be minimalespecially in comparison to <strong>the</strong> impact on a polymeric membrane.In <strong>the</strong> case <strong>of</strong> non-condensable impurities, CO seems to have a greater effect on reducing <strong>the</strong>per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> than H 2 S but this is most likely due to <strong>the</strong> <strong>high</strong>er concentration<strong>of</strong> CO <strong>use</strong>d. Upon exposure to <strong>the</strong>se non-condensable impurities, <strong>the</strong> CO 2 permeance <strong>of</strong> <strong>the</strong>NPC <strong>membranes</strong> is reduced by up to 12 % (+/- 7 %). The impact on <strong>the</strong> N 2 permeance is similargiving a reduction in <strong>the</strong> CO 2 /N 2 permeance <strong>of</strong> less than 4 % (+/- 7%).In particular, <strong>the</strong> minor impact on NPC membrane per<strong>for</strong>mance by H 2 S is hopeful considering<strong>the</strong> per<strong>for</strong>mance <strong>of</strong> Cu/Pd <strong>membranes</strong> <strong>of</strong>ten <strong>use</strong>d in syn<strong>the</strong>sis gas purification are adverselyaffected by H 2 S [139].150


Whilst <strong>the</strong>se results are promising <strong>for</strong> <strong>the</strong> <strong>use</strong> <strong>of</strong> NPC <strong>membranes</strong> in <strong>the</strong> separation <strong>of</strong> CO 2 fromnatural gas production and syn<strong>the</strong>sis gas production, fur<strong>the</strong>r work on <strong>the</strong> longer term impact <strong>of</strong>exposure to impurities is required.151


152


7 CONCLUSIONS AND RECOMMENDATIONS7.1 ConclusionsFor <strong>the</strong> last decade, <strong>nanoporous</strong> <strong>carbon</strong> has been touted as a superior material <strong>for</strong> <strong>the</strong> separation<strong>of</strong> gases. Manufactured from <strong>the</strong> pyrolysis <strong>of</strong> polymers at <strong>high</strong> <strong>temperature</strong>, <strong>nanoporous</strong> <strong>carbon</strong>sare structurally related to <strong>the</strong> fullerene group <strong>of</strong> <strong>carbon</strong> and contain nanosized sized poressuitable <strong>for</strong> <strong>the</strong> separation <strong>of</strong> gases based upon molecular size.Membranes can be manufactured from <strong>nanoporous</strong> <strong>carbon</strong> by coating a porous support with apolymeric precursor prior to pyrolysis (supported NPC <strong>membranes</strong>) or from <strong>the</strong> direct pyrolysis<strong>of</strong> a flat sheet or hollow fibre <strong>of</strong> polymer (unsupported NPC <strong>membranes</strong>). The ability to make<strong>membranes</strong> from <strong>nanoporous</strong> <strong>carbon</strong> is advantageous due to reduced equipment size, footprintand potentially cost when compared with o<strong>the</strong>r separation processes such as gas absorption.The separation per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> is well documented as being above <strong>the</strong>Robeson line, which is considered <strong>the</strong> <strong>the</strong>oretical limit <strong>of</strong> per<strong>for</strong>mance <strong>for</strong> <strong>the</strong> traditionalpolymeric <strong>membranes</strong>. Additionally, <strong>the</strong> <strong>high</strong> <strong>temperature</strong>s <strong>use</strong>d <strong>for</strong> <strong>the</strong> manufacture <strong>of</strong> NPC<strong>membranes</strong> makes <strong>the</strong>m ideal candidates <strong>for</strong> <strong>high</strong> <strong>temperature</strong> separation resulting in significantpotential <strong>for</strong> cost reduction by removing pre-cooling requirements.However, NPC <strong>membranes</strong> are not without drawbacks. The brittleness and cracking <strong>of</strong><strong>nanoporous</strong> <strong>carbon</strong> is a significant manufacturing challenge. Likewise, <strong>the</strong> significant reductionin per<strong>for</strong>mance seen when exposed to <strong>the</strong> atmosphere (termed “aging”) makes transportationand storage <strong>of</strong> <strong>the</strong> NPC <strong>membranes</strong> difficult along with <strong>the</strong> inability <strong>of</strong> <strong>the</strong>se <strong>membranes</strong> tooperate in <strong>the</strong> presence <strong>of</strong> oxygen even when present in trace quantities.The objective <strong>of</strong> this research work was to put NPC <strong>membranes</strong> to <strong>the</strong> test and specificallyreview <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC <strong>membranes</strong> in <strong>the</strong> application <strong>of</strong> CO 2 capture from natural gas(methane) and from air-blow syn<strong>the</strong>sis gas production (nitrogen). A critical component <strong>of</strong> <strong>the</strong>research work was to assess <strong>the</strong> per<strong>for</strong>mance under realistic conditions – <strong>high</strong> operating<strong>temperature</strong> and in <strong>the</strong> presence <strong>of</strong> condensable and non-condensable impurities.A substantial amount <strong>of</strong> time and ef<strong>for</strong>t was spent on developing a method to produce selectiveNPC <strong>membranes</strong>. Cracking proved to be <strong>the</strong> most significant obstacle <strong>of</strong> this research work andas such dedicated attempts to eliminate cracking was critical to obtaining working <strong>membranes</strong>that could be <strong>use</strong>d <strong>for</strong> fulfilling <strong>the</strong> objectives <strong>of</strong> <strong>the</strong> research work.153


Of <strong>the</strong> three types <strong>of</strong> NPC <strong>membranes</strong> produced, namely supported, modified-supported andunsupported, <strong>the</strong> most success in eliminating cracking and producing selective <strong>membranes</strong>, wasfound in <strong>the</strong> manufacture <strong>of</strong> supported NPC <strong>membranes</strong> from <strong>the</strong> pyrolysis <strong>of</strong> PFA.The critical mass <strong>of</strong> <strong>carbon</strong> deposited onto <strong>the</strong> support that marked <strong>the</strong> onset <strong>of</strong> cracking wasdetermined to be 8 mg/cm 2 and 6 mg/cm 2 <strong>for</strong> pyrolysis <strong>temperature</strong>s <strong>of</strong> 400 and 550°C,respectively. The characterisation results revealed a significant increase in <strong>the</strong> porosity withincreasing pyrolysis <strong>temperature</strong>, which explains <strong>the</strong> earlier onset <strong>of</strong> cracking at <strong>high</strong>ermanufacturing <strong>temperature</strong>s due to <strong>the</strong> more porous and <strong>the</strong>re<strong>for</strong>e delicate structure.The increase in permeance observed with increasing pyrolysis <strong>temperature</strong> is also explained by<strong>the</strong> increase in porosity. However, <strong>the</strong> pore size change with increasing pyrolysis <strong>temperature</strong>was too small to impact upon <strong>the</strong> permselectivity. In fact <strong>the</strong> pore size range <strong>for</strong> <strong>the</strong> pyrolysis<strong>temperature</strong>s studied stayed within <strong>the</strong> published range <strong>of</strong> 4 – 5Å [61]. The permselectivity wasthus found to be more greatly effected by <strong>the</strong> <strong>for</strong>mation <strong>of</strong> a complete membrane layer and <strong>the</strong>absence <strong>of</strong> cracks than <strong>the</strong> pore size.Testing <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> an NPC membrane with a mixed gas feed provides a more accuraterepresentation <strong>of</strong> how <strong>the</strong> <strong>membranes</strong> would per<strong>for</strong>m in a commercial setting than <strong>the</strong>traditional pure gas per<strong>for</strong>mance testing. The results <strong>of</strong> pure and mixed gas testing <strong>for</strong> <strong>the</strong> sameNPC <strong>membranes</strong> revealed that <strong>the</strong> CO 2 /CH 4 mixed gas selectivity was greater than <strong>the</strong> CO 2 /CH 4permselectivity determined from <strong>the</strong> pure gas testing. This result is explained by <strong>the</strong> competitiveadsorption effect <strong>of</strong> <strong>the</strong> more adsorbable component CO 2 “blocking” <strong>the</strong> way <strong>for</strong> CH 4permeance.The <strong>carbon</strong> dioxide/methane (CO 2 /CH 4 ) mixed gas per<strong>for</strong>mance measurements generally showan increase in permeance as <strong>the</strong> operating <strong>temperature</strong> is increased due to <strong>the</strong> dominance <strong>of</strong>molecular sieving. The exception is <strong>for</strong> <strong>the</strong> CO 2 permeance <strong>of</strong> <strong>membranes</strong> prepared at a <strong>high</strong>erpyrolysis <strong>temperature</strong> (550°C) where adsorption processes are competitive with molecularsieving effects. This leads to a CO 2 permeance value that is relatively independent <strong>of</strong><strong>temperature</strong>. Whilst this leads to a membrane that is slightly more selective at 35°C, <strong>the</strong>membrane selectivities converge at <strong>high</strong>er operating <strong>temperature</strong>s. Indeed, extrapolationsuggests that <strong>the</strong> membrane selectivity would reduce to <strong>the</strong> Knudsen (CO 2 /CH 4 = 0.6) value atan operating <strong>temperature</strong> between 400 and 500°C.Although <strong>the</strong> selectivities <strong>of</strong> <strong>the</strong> <strong>membranes</strong> manufactured at different pyrolysis <strong>temperature</strong>sconverge at <strong>high</strong>er operating <strong>temperature</strong>s, <strong>the</strong> permeance is still related <strong>the</strong> original pyrolysis154


<strong>temperature</strong>. This result implies that it is important to manufacture a membrane with <strong>high</strong>porosity/permeance ra<strong>the</strong>r than <strong>high</strong> adsorption/selectivity when considering <strong>high</strong> operating<strong>temperature</strong> applications such as CO 2 capture from natural gas or syn<strong>the</strong>sis gas.Molecular simulation <strong>of</strong> pure and mixed gas adsorption on <strong>nanoporous</strong> <strong>carbon</strong> (represented byC 168 Schwarzite) confirmed that CO 2 is a more strongly adsorbing molecule than CH 4 and muchmore so than N 2 . Whilst adsorption does decreases with increasing operating <strong>temperature</strong> <strong>the</strong>selectivity <strong>for</strong> CO 2 still remains at <strong>temperature</strong>s as <strong>high</strong> as 200°C.The greater difference between <strong>the</strong> values <strong>of</strong> heat <strong>of</strong> adsorption <strong>for</strong> <strong>the</strong> CO 2 / N 2 pair comparedwith <strong>the</strong> CO 2 /CH 4 pair gives a significantly larger separation factor <strong>for</strong> <strong>the</strong> CO 2 /N 2 pair. This isa particularly promising result <strong>for</strong> <strong>the</strong> separation <strong>of</strong> CO 2 from N 2 at <strong>high</strong> <strong>temperature</strong>s as is <strong>the</strong>case <strong>for</strong> <strong>the</strong> air-blown syn<strong>the</strong>sis gas purification.A significant difference between <strong>the</strong> simulated and experimental adsorption results wasobserved. This difference is attributed to <strong>the</strong> different substrate (C 168 Schwarzite) <strong>use</strong>d in <strong>the</strong>simulation and to a smaller amount <strong>of</strong> adsorption sites available <strong>for</strong> gas adsorption on <strong>the</strong>experimental samples due to oxygen adsorption from <strong>the</strong> air.The presence <strong>of</strong> condensable and non-condensable impurities on <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> NPC<strong>membranes</strong> is small, particularly at elevated operating <strong>temperature</strong>s where <strong>the</strong> impurityadsorption is lower.The small reductions in <strong>the</strong> CH 4 and CO 2 permeance <strong>of</strong> a NPC membrane observed in <strong>the</strong>presence <strong>of</strong> condensable impurities is <strong>of</strong> <strong>the</strong> order toluene > hexane > water. The impact <strong>of</strong> <strong>the</strong>condensable impurities on <strong>the</strong> CH 4 and CO 2 permeances is very similar resulting in no netchange in <strong>the</strong> selectivity.Likewise, small reductions in <strong>the</strong> CH 4 and CO 2 permeance <strong>of</strong> a NPC membrane were observedin <strong>the</strong> presence <strong>of</strong> H 2 S and CO. The impact <strong>of</strong> <strong>the</strong> non-condensable impurities on <strong>the</strong> CH 4 andCO 2 permeances is very similar resulting in no significant change in <strong>the</strong> selectivity.In summary, provided crack-free NPC <strong>membranes</strong> can be easily manufactured and handled suchthat exposure to oxygen is prevented, NPC <strong>membranes</strong> do show some promise <strong>for</strong> <strong>the</strong> <strong>use</strong> in<strong>high</strong> <strong>temperature</strong> CO 2 capture processes such as separation from natural gas or syn<strong>the</strong>sis gas.The supported NPC <strong>membranes</strong> tested as part <strong>of</strong> this research work maintain selectivity above 1<strong>for</strong> operating <strong>temperature</strong>s as <strong>high</strong> as 200°C. Additionally, <strong>the</strong> impact <strong>of</strong> condensable and noncondensableimpurities is relatively small particularly at <strong>high</strong>er operating <strong>temperature</strong>s.155


7.2 Recommendations <strong>for</strong> Fur<strong>the</strong>r WorkThe rapid increase in publications on NPC <strong>membranes</strong> over <strong>the</strong> last decade has slowedsomewhat over <strong>the</strong> last few years and un<strong>for</strong>tunately no industrial success has been reported asyet. The significant problems with cracking and aging appear to have had a great impact uponfur<strong>the</strong>ring this technology to commercialisation.The emerging membrane technologies that are presently attracting a lot <strong>of</strong> attention are thosethat endeavour to achieve <strong>the</strong> per<strong>for</strong>mance <strong>of</strong> a NPC membrane with <strong>the</strong> flexibility <strong>of</strong> apolymeric membrane, such as mixed matrix or partially pyrolysised <strong>membranes</strong>. Park et al [160]recently developed a novel technique utilising <strong>high</strong> manufacturing <strong>temperature</strong>s <strong>for</strong> producingpolymeric <strong>membranes</strong> containing <strong>high</strong>ly tunned pore sizes <strong>for</strong> gas transport. These <strong>membranes</strong>surpass <strong>the</strong> Robeson upper bound whilst maintaining <strong>the</strong> robustness <strong>of</strong> a polymeric membrane.Whilst interest in NPC <strong>membranes</strong> appears to have been overtaken by <strong>the</strong> emerging compositemembrane technologies discussed above, <strong>the</strong> possibility <strong>of</strong> solving <strong>the</strong> problems with crackingand aging <strong>of</strong> NPC <strong>membranes</strong> still remains.The recent work <strong>of</strong> Rajgopalan et al [28] and Merritt et al [27] in regards to <strong>the</strong> addition <strong>of</strong> silicaparticles is promising <strong>for</strong> reducing <strong>the</strong> amount <strong>of</strong> <strong>carbon</strong> added. However, attempts to makeNPC <strong>membranes</strong> upon supports filled with silica nanoparticles as part <strong>of</strong> this research workwere unsuccessful again due to tendency <strong>for</strong> cracking. It is possible that <strong>the</strong> <strong>use</strong> <strong>of</strong> alternativesupports (such as porous ceramics or <strong>carbon</strong>), which have a similar <strong>the</strong>rmal expansion to<strong>nanoporous</strong> <strong>carbon</strong>, could reduce cracking. However, obtaining <strong>high</strong> pressure feed/permeateseal using <strong>the</strong>se supports at <strong>high</strong> <strong>temperature</strong>s is likely to be difficult and costly.The work <strong>of</strong> Jones and Koros [129] on applying a thin layer <strong>of</strong> unique polymeric materials toprotect <strong>the</strong> finished membrane from aging shows potential. However, fur<strong>the</strong>r tests on <strong>the</strong>stability <strong>of</strong> this polymeric layer at <strong>high</strong> operating <strong>temperature</strong>s and in <strong>the</strong> presence <strong>of</strong> impuritieswould need to be undertaken.The impact <strong>of</strong> condensable and non-condensable impurities has been shown here to be fairlyminimal. However, fur<strong>the</strong>r work is required to assess <strong>the</strong> long term impact <strong>of</strong> <strong>the</strong> impurities.Considering <strong>the</strong> health and safety risks associated in particular with H 2 S and CO, <strong>the</strong> mostappropriate way to fur<strong>the</strong>r assess <strong>the</strong> impact <strong>of</strong> <strong>the</strong>se impurities in <strong>the</strong> long term would be CO 2capture trials on existing processing plants where <strong>the</strong>se impurities are ever present andcontrolled within <strong>the</strong> rigorous health and safety systems <strong>of</strong> an operating company.156


As part <strong>of</strong> <strong>the</strong> Victorian Government’s Energy Technology Innovation Strategy (ETIS), variouscapture technologies including membrane systems are being tested using actual syn<strong>the</strong>sis gasproduced by HRL’s coal gasifier in Mulgrave (pre-combustion capture) and actual flue gasproduced by International Power at <strong>the</strong> Hazelwood Power Station in <strong>the</strong> Latrobe Valley (postcombustioncapture). Testing NPC <strong>membranes</strong> as part <strong>of</strong> <strong>the</strong> ETIS scheme, in particular <strong>for</strong> precombustioncapture, is recommended as <strong>the</strong> logical follow-on from this research work.157


158


8 REFERENCES1. IPCC, Climate Change 2007: Syn<strong>the</strong>sis Report. Contribution <strong>of</strong> Working Groups I, IIand III to <strong>the</strong> Fourth Assessment Report <strong>of</strong> <strong>the</strong> Intergovernmental Panel on ClimateChange. IPCC. 2007, Geneva, Switzerland.2. Karoly, D., J. Risbey, and A. Reynolds, Global Warming Contributes to Australia'sWorst Drought, WWF, Editor. 2003.3. Flannery, T., We are <strong>the</strong> Wea<strong>the</strong>r Makers. 2006: Text Publishing.4. Squires, N., Rudd Win Ushers in New Era <strong>for</strong> Australian Politics, in The ChristianScience Monitor. 2007: Boston.5. CO2CRC. www.CO2CRC.com.au. 2009.6. Davison, J. Comparison <strong>of</strong> CO2 Abatement by Use <strong>of</strong> Renewable Energy and CO2Capture and Storage. in 5th International Conference on Greenho<strong>use</strong> Gas ControlTechnologies,. 2000. Cairns, Australia.7. Wallace, D. Capture and Storage <strong>of</strong> CO2 - What Needs to be Done? in COP(Conference <strong>of</strong> <strong>the</strong> Parties) 6,. 2000. The Hague, Holland: IEA.8. Gilron, J. and A. S<strong>of</strong>fer, Knudsen Diffusion in Microporous Carbon Membranes withMolecular Sieving Character. Journal <strong>of</strong> Membrane Science, 2002. 209(2): p. 339-352.9. Koresh, J.E. and A. S<strong>of</strong>fer, The Carbon Molecular Sieve Membranes. GeneralProperties and <strong>the</strong> Permeability <strong>of</strong> Methane/Hydrogen Mixture. Separation Science andTechnology, 1987. 22(2-3): p. 973-82.10. Koresh, J.E. and A. S<strong>of</strong>fer, Mechanism <strong>of</strong> Permeation through Molecular-Sieve CarbonMembrane. Journal <strong>of</strong> <strong>the</strong> Chemical Society - Faraday Trans. 1, 1986. 82: p. 2057-2063.11. Koresh, J.E. and A. S<strong>of</strong>fer, A Molecular Sieve Carbon Membrane <strong>for</strong> ContinuousProcess Gas Separation. Extended Abstracts and Program - Biennial Conference onCarbon, 1983. 16th: p. 367-8.12. Lagorsse, S., F.D. Magalhaes, and A. Mendes, Carbon Molecular Sieve Membranes.Sorption, Kinetic and Structural Characterization. Journal <strong>of</strong> Membrane Science, 2004.241(2): p. 275-287.13. S<strong>of</strong>fer, A., et al., Process <strong>for</strong> <strong>the</strong> Production <strong>of</strong> Hollow Carbon Fiber Membranes. 1999:US Patent 5,925,591.159


14. Rao, M.B. and S. Sircar, Per<strong>for</strong>mance and Pore Characterization <strong>of</strong> Nanoporous CarbonMembranes <strong>for</strong> Gas Separation. Journal <strong>of</strong> Membrane Science, 1996. 110: p. 109-118.15. Centeno, T.A. and A.B. Fuertes, Supported Carbon Molecular Sieve Membranes Basedon a Phenolic Resin. Journal <strong>of</strong> Membrane Science, 1999. 160: p. 201-211.16. Fuertes, A.B., D.M. Nevskaia, and T.A. Centeno, Carbon Composite Membranes fromMatrimid and Kapton Polyimides <strong>for</strong> Gas Separation. Microporous and MesoporousMaterials, 1999. 33(1-3): p. 115-125.17. Steel, K.M. and W.J. Koros, Investigation <strong>of</strong> Porosity <strong>of</strong> Carbon Materials and RelatedEffects on Gas Separation Properties. Carbon, 2003. 41(2): p. 253-266.18. Ogawa, M. and Y. Nakano, Separation <strong>of</strong> CO2/CH4 Mixture Through CarbonizedMembrane Prepared by Gel Modification. Journal <strong>of</strong> Membrane Science, 2000. 173(1):p. 123-132.19. Wang, H., L. Zhang, and G.R. Gavalas, Preparation <strong>of</strong> Supported Carbon Membranesfrom Furfuryl Alcohol by Vapor Deposition Polymerization. Journal <strong>of</strong> MembraneScience, 2000. 177: p. 25-31.20. Rao, M.B. and S. Sircar, Nanoporous Carbon Membranes <strong>for</strong> Separation <strong>of</strong> GasMixtures by Selective Surface Flow. Journal <strong>of</strong> Membrane Science, 1993. 85: p. 253-264.21. Fuertes, A.B., Preparation and Characterization <strong>of</strong> Adsorption-Selective CarbonMembranes <strong>for</strong> Gas Separation. Adsorption, 2001. 7(2): p. 117-129.22. Suda, H. and K. Haraya, Carbon Molecular Sieve Membranes: Preparation,Characterization, and Gas Permeation Properties. ACS Symposium Series, 2000.744(Membrane Formation and Modification): p. 295-313.23. Hagg, M., J.A. Lie, and A. Lindbra<strong>the</strong>n, Carbon Molecular Sieve Membranes: APromising Alternative <strong>for</strong> Selected Industrial Applications. Annals <strong>of</strong> <strong>the</strong> New YorkAcademy <strong>of</strong> Sciences, 2003. 984(Advanced Membrane Technology): p. 329-345.24. Strano, M. and H.C. Foley, Deconvolution <strong>of</strong> Permeance in Supported NanoporousMembranes. American Institute <strong>of</strong> Chemical Engineers, 2000. 46(3): p. 651.25. Sedigh, M.G., et al., Experiments and Simulation <strong>of</strong> Transport and Separation <strong>of</strong> GasMixtures in Carbon Molecular Sieve Membranes. Journal <strong>of</strong> Physical Chemistry A,1998. 102: p. 8580-8589.160


26. Fuertes, A.B. and T.A. Centeno, Preparation <strong>of</strong> Supported Carbon Molecular SieveMembranes. Journal <strong>of</strong> Membrane Science, 1998. 144: p. 105-111.27. Merritt, A., R. Rajagopalan, and H.C. Foley, High Per<strong>for</strong>mance Nanoporous CarbonMembranes <strong>for</strong> Air Separation. Carbon, 2007. 45(6): p. 1267-1278.28. Rajagopalan, R., et al., Modification <strong>of</strong> Macroporous Stainless Steel Supports withSilica Nanoparticles <strong>for</strong> Size Selective Carbon Membranes with Improved Flux.Carbon, 2006. 44(10): p. 2051-2058.29. Barsema, J.N., et al., Intermediate Polymer to Carbon Gas Separation Membranes basedon Matrimid PI. Journal <strong>of</strong> Membrane Science, 2004. 238(1-2): p. 93-102.30. Pereira-Nunes, S. and K.-N. Peinemann, Membrane Technology in <strong>the</strong> ChemicalIndustry, ed. S. Pereira-Nunes and K.-N. Peinemann. 2006: Wiley-VCH.31. Jones, C.W. and W.J. Koros, Characterization <strong>of</strong> Ultramicroporous Carbon Membraneswith Humidified Feeds. Industrial & Engineering Chemistry Research, 1995. 34(1): p.158-63.32. Menendez, I. and A.B. Fuertes, Aging <strong>of</strong> Carbon Membranes under DifferentEnvironments. Carbon, 2001. 39(5): p. 733-740.33. Vu, D.Q., W.J. Koros, and S.J. Miller, Fouling <strong>of</strong> Carbon Molecular Sieve Hollow-Fiber Membranes by Condensable Impurities in Carbon Dioxide-Methane Separations.Industrial & Engineering Chemistry Research, 2003. 42(5): p. 1064-1075.34. Jones, C.W. and W.J. Koros, Carbon Molecular Sieve Gas Separation Membranes 2 -Regeneration Following Organic Exposure. Carbon, 1994. 32(8): p. 1427-32.35. Vu, D.Q. and W.J. Koros, Effect <strong>of</strong> Condensible Impurities in CO2/CH4 Gas Feeds onCarbon Molecular Sieve Hollow-Fiber Membranes. Industrial & Engineering ChemistryResearch, 2003. 42: p. 1064-1075.36. Hooper, B., Investigating Viable CO2 Capture and Sequestration Options. 2004.37. ASTM International, D 1434 -82: Standard Test Method <strong>for</strong> Determining GasPermeability Characterisitics <strong>of</strong> Plastic Film and Sheeting. 2003: United States.38. Robeson, L.M., Correlation <strong>of</strong> Separation Factor versus Permeability <strong>for</strong> PolymericMembranes. Journal <strong>of</strong> Membrane Science, 1991. 62: p. 165-185.39. Bos, A., High Pressure CO2/CH4 Separation with Glassy Polymer Membranes. 1996:Febodruk.161


40. Mulder, M., Basic Principles <strong>of</strong> Membrane Technology. 2nd ed. 1996: KluwerAcademic Publishers.41. Wang, K., H. Suda, and K. Haraya, Permeation Time Lag and <strong>the</strong> ConcentrationDependence <strong>of</strong> <strong>the</strong> Diffusion Coefficient <strong>of</strong> CO2 in a Carbon Molecular SieveMembrane. Industrial & Engineering Chemistry Research, 2001. 40(13): p. 2942-2946.42. Webb, P.A. and C. Orr, Analytical Methods in Fine Particle Technology. 1997,Norcross: Micromeritics Instrumental Corporation.43. Strano, M.S. and H.C. Foley, Temperature and Pressure Dependent Transient Analysis<strong>of</strong> Single Component Permeation through Nanoporous Carbon Membranes. Carbon,2002. 40: p. 1029-1041.44. Brunauer, S., P.H. Emmett, and E. Teller, Adsorption <strong>of</strong> Gases in MultimolecularLayers. Journal <strong>of</strong> <strong>the</strong> American Chemical Society, 1938. 60: p. 309-19.45. Arora, G., Thermodynamic and Transport Properties in Carbon Nanostructures.Chemical Engineering, University <strong>of</strong> Delaware, 2006. PhD46. Li, S., et al., High-Pressure CO2/CH4 Separation Using SAPO-34 Membranes.Industrial & Engineering Chemistry Research, 2005. 44(9): p. 3220-3228.47. Suzuki, M., Adsorption Engineering. 1990, Tokyo: Kodansha Ltd & Elsevier Science.48. Myers, A.L. and J.M. Prausnitz, Thermodynamics <strong>of</strong> Mixed-Gas Adsorption. AIChEJournal, 1965. 11(1): p. 121-127.49. Chen, H. and D.S. Sholl, Examining <strong>the</strong> Accuracy <strong>of</strong> Ideal Adsorbed Solution TheoryWithout Curve-Fitting Using Transition Matrix Monte Carlo Simulations. Langmuir,2007. 23(11): p. 6431-6437.50. Ritter, J.A. and R.T. Yang, Equilibrium Adsorption <strong>of</strong> Multicomponent Gas Mixtures atElevated Pressures. Industrial & Engineering Chemistry Research, 1987. 26(8): p. 1679-86.51. Treybal, R.E., Mass-Transfer Operations. 1981: McGraw-Hill.52. Andrews, R., et al. Separation <strong>of</strong> CO2 from Flue Gases by Carbon-Multiwall CarbonNanotube Membranes. in NASA Conference Publication. 2001.53. Wu, L.Q., N. Xu, and J. Shi, Preparation <strong>of</strong> a Palladium Composite Membrane by anImproved Electroless Plating Technique. Industrial and Engineering ChemistryResearch, 2000. 39(2): p. 342-348.162


54. Villaluenga, J.P.G., et al., Gas Permeation Characteristics <strong>of</strong> Heterogeneous ODPA-BISP Polyimide Membranes at Different Temperatures. Journal <strong>of</strong> Membrane Science,2007. 305(1-2): p. 160-168.55. Koros, W.J. and R. Mahajan, Pushing <strong>the</strong> Limits on Possibilities <strong>for</strong> Large Scale GasSeparation: Which Strategies? Journal <strong>of</strong> Membrane Science, 2000. 175: p. 181-196.56. Kroto, H.W., et al., C60: Buckminsterfullerene. Nature (London, United Kingdom),1985. 318(6042): p. 162-3.57. Dresselhaus, M.S., et al., Introduction to Carbon Materials, in Carbon Nanotubes:Preparation and Properties, T.W. Ebbesen, Editor. 1997, CRC Press: Boca Raton.58. Jenkins, G.M. and K. Kawamura, Polymeric Carbons. 1976, Cambridge: CambridgeUniversity Press.59. Foley, H.C., M.S. Kane, and J.F. Goellner, New Directions with Carbogenic MolecularSieve Materials, in Access in Nanoporous Materials, T.J. Pinnavaia and M.F. Thorpe,Editors. 1995, Plenum Press: New York. p. 39-58.60. Acharya, M., et al., Simulation <strong>of</strong> Nanoporous Carbons: a Chemically ConstrainedStructure. Philosophical Magazine B: Physics <strong>of</strong> Condensed Matter: StatisticalMechanics, Electronic, Optical and Magnetic Properties, 1999. 79(10): p. 1499-1518.61. Burket, C.L., et al., Genesis <strong>of</strong> Porosity in Polyfurfuryl Alcohol Derived NanoporousCarbon. Carbon, 2006. 44(14): p. 2957-2963.62. Foley, H.C., Carbogenic Molecular Sieves: Syn<strong>the</strong>sis, Properties and Applications.Microporous Materials, 1995. 4: p. 407-433.63. Anand, M., et al., Multicomponent Gas Separation by Selective Surface Flow (SSF) andPoly-Trimethylsilylpropyne (PTMSP) Membranes. Journal <strong>of</strong> Membrane Science,1997. 123: p. 17-25.64. Paranjape, M., et al., Separation <strong>of</strong> Bulk Carbon Dioxide and Hydrogen Mixtures bySelective Surface Flow Membranes. Adsorption, 1998. 4(3 & 4): p. 355-360.65. Parrillo, D.J., C. Thaeron, and S. Sircar, Separation <strong>of</strong> Bulk Hydrogen Sulfide -Hydrogen Mixtures by Selective Surface Membrane. Journal <strong>of</strong> <strong>the</strong> American Institute<strong>of</strong> Chemical Engineers, 1997. 43(9): p. 2239-2245.163


66. Sircar, S. and M.B. Rao, Nanoporous Carbon Membranes <strong>for</strong> Gas Separation, in RecentAdvances in Gas Separation by Microporous Ceramic Membranes, N.K. Kanellopoulos,Editor. 2000, Elsevier: Amsterdam. p. 473-496.67. Fuertes, A.B., Adsorption-Selective Carbon Membrane <strong>for</strong> Gas Separation. Journal <strong>of</strong>Membrane Science, 2000. 177(1-2): p. 9-16.68. Kusakabe, K., M. Yamamoto, and S. Morooka, Gas Permeation and MicroporeStructure <strong>of</strong> Carbon Molecular Sieving Membranes Modified by Oxidation. Journal <strong>of</strong>Membrane Science, 1998. 149: p. 59-67.69. Hayashi, J., et al., Pore Size Control <strong>of</strong> Carbonized BPDA-pp'-ODA PolyimideMembrane by Chemical Vapor Deposition <strong>of</strong> Carbon. Journal <strong>of</strong> Membrane Science,1997. 124(2): p. 243-251.70. Bird, A.J. and D.L. Trimm, Carbon Molecular Sieves <strong>use</strong>d in Gas SeparationMembranes. Carbon, 1983. 21(3): p. 177-80.71. Mariwala, R.K. and H.C. Foley, Evolution <strong>of</strong> Ultramicroporous Adsorptive Structure inPFA-Derived Carbongenic Molecular Sieves. Industrial and Engineering ChemistryResearch, 1994. 33: p. 607-615.72. Lafyatis, D.S., J. Tung, and H.C. Foley, Poly(Furfuryl Alcohol)-Derived CarbonMolecular Sieves: Dependence <strong>of</strong> Adsorptive Properties on Carbonization Temperature,Time, and Poly(Ethylene Glycol) Additives. Industrial & Engineering ChemistryResearch, 1991. 30(5): p. 865-73.73. Acharya, M., et al., Metal-Supported Carbogenic Molecular Sieve Membranes:Syn<strong>the</strong>sis and Applications. Industrial and Engineering Chemistry Research, 1997. 36:p. 2924-2930.74. Acharya, M. and H.C. Foley, Spray-Coating <strong>of</strong> Nanoporous Carbon Membranes <strong>for</strong> AirSeparation. Journal <strong>of</strong> Membrane Science, 1999. 161: p. 1-5.75. Shiflett, M.B. and H.C. Foley, Ultrasonic Deposition <strong>of</strong> High Selectivity NanoporousCarbon Membranes. Science, 1999. 285: p. 1902.76. Acharya, M. and H.C. Foley, Transport in Nanoporous Carbon Membranes:Experiments and Analysis. American Institute <strong>of</strong> Chemical Engineers, 2000. 46(5): p.911.77. Shiflett, M.B., Supported Nanoporous Gas Separation Membrane and its Preparation.2000: Patent WO 2000053299.164


78. Shiflett, M.B. and H.C. Foley, Reproducible Production <strong>of</strong> Nanoporous CarbonMembranes. Carbon, 2001. 39: p. 1421-1446.79. Strano, M.S. and H.C. Foley, Syn<strong>the</strong>sis and Characterization <strong>of</strong> Catalytic NanoporousCarbon Membranes. American Institute <strong>of</strong> Chemical Engineers, 2001. 47(1): p. 66-78.80. Foley, H.C., et al., Nanoporous Carbon Catalytic Membranes and Method <strong>for</strong> Making<strong>the</strong> Same. 2002: United States <strong>of</strong> America.81. Vu, D.Q. and W.J. Koros, High Pressure CO2/CH4 Separation Using Carbon MolecularSieve Hollow Fiber Membranes. Industrial and Engineering Chemistry Research, 2002.41: p. 367-380.82. Rajagopalan, R. and H.C. Foley, Study <strong>of</strong> <strong>the</strong> Effect <strong>of</strong> Morphology <strong>of</strong> NanoporousCarbon Membranes on Permselectivity. Materials Research Society, 2003. 752: p. 225-230.83. Shiflett, M.B. and H.C. Foley, On <strong>the</strong> Preparation <strong>of</strong> Supported Nanoporous CarbonMembranes. Journal <strong>of</strong> Membrane Science, 2000. 179: p. 275-282.84. Shiflett, M.B., et al., Characterization <strong>of</strong> Supported Nanoporous Carbon Membranes.Advanced Materials, 2000. 12(1): p. 21-25.85. Sea, B. and K.-H. Lee, Molecular Sieve Silica Membrane Syn<strong>the</strong>sized in Mesoporousg-Alumina Layer. Bulletin <strong>of</strong> <strong>the</strong> Korean Chemical Society, 2001. 22(12): p. 1400-1402.86. Diniz da Costa, J.C., et al., Novel Molecular Sieve Silica (MSS) Membranes:Characterization and Permeation <strong>of</strong> Single-Step and Two-Step Sol-Gel Membranes.Journal <strong>of</strong> Membrane Science, 2002. 198(1): p. 9-21.87. Duke, M.C., et al., Hydro<strong>the</strong>rmal Stable Templated Molecular Sieve Silica (TMSS)Membranes <strong>for</strong> Gas Separation. 2006.88. Leenaars, A.F.M. and A.J. Burggraaf, The Preparation and Characterization <strong>of</strong> AluminaMembranes with Ultrafine Pores. 2. The Formation <strong>of</strong> Supported Membranes. Journal<strong>of</strong> Colloid and Interface Science, 1985. 105(1): p. 27-40.89. Shiflett, M.B., Syn<strong>the</strong>sis, Characterisation and Application <strong>of</strong> Nanoporous CarbonMembranes. Chemical Engineering, University <strong>of</strong> Delaware, 2002. PhD165


90. Geiszler, V.C. and W.J. Koros, Effects <strong>of</strong> Polyimide Pyrolysis Conditions on CarbonMolecular Sieve Membrane Properties. Industrial and Engineering Chemistry Research,1996. 35: p. 2999-3003.91. Burns, R.L., et al., Explanation <strong>of</strong> a Selectivity Maximum as a Function <strong>of</strong> <strong>the</strong> MaterialStructure <strong>for</strong> Organic Gas Separation Membranes. Industrial & Engineering ChemistryResearch, 2004: p. ACS ASAP.92. Liu, S., et al., Gas Permeation Properties <strong>of</strong> Carbon Molecular Sieve MembranesDerived from Novel Poly(phthalazinone e<strong>the</strong>r sulfone ketone). Industrial & EngineeringChemistry Research, 2008: p. ACS ASAP.93. Suda, H. and K. Haraya, Gas Permeation through Micropores <strong>of</strong> Carbon MolecularSieve Membranes Derived from Kapton Polyimide. Journal <strong>of</strong> Physical Chemistry B,1997. 101(20): p. 3988-3994.94. Su, J. and A.C. Lua, Effects <strong>of</strong> Carbonisation Atmosphere on <strong>the</strong> StructuralCharacteristics and Transport Properties <strong>of</strong> Carbon Membranes Prepared from KaptonPolyimide. Journal <strong>of</strong> Membrane Science, 2007. 305(1-2): p. 263-270.95. Mariwala, R.K. and H.C. Foley, Calculation <strong>of</strong> Micropore Sizes in CarbogenicMaterials from <strong>the</strong> Methyl Chloride Adsorption Iso<strong>the</strong>rm. Industrial & EngineeringChemistry Research, 1994. 33(10): p. 2314-21.96. Tin, P.S., T.-S. Chung, and A.J. Hill, Advanced Fabrication <strong>of</strong> Carbon Molecular SieveMembranes by Nonsolvent Pretreatment <strong>of</strong> Precursor Polymers. Industrial &Engineering Chemistry Research, 2004. 43(20): p. 6476-6483.97. Tin, P.S., et al., Separation <strong>of</strong> CO2/CH4 through Carbon Molecular Sieve MembranesDerived from P84 Polyimide. Carbon, 2004. 42(15): p. 3123-3131.98. Bernal, M.P., et al., Separation <strong>of</strong> CO2/N2 Mixtures using MFI-Type ZeoliteMembranes. AIChE Journal, 2004. 50(1): p. 127-135.99. Hayashi, J., et al., Simultaneous Improvement <strong>of</strong> Permeance and Permselectivity <strong>of</strong>3,3',4,4'-Biphenyltetracarboxylic Dianhydride-4,4'-Oxydianiline Polyimide Membraneby Carbonization. Industrial & Engineering Chemistry Research, 1995. 34(12): p. 4364-70.100. Foley, H.C., M.B. Shiflett, and D.R. Corbin, Mixed Matrix Nanoporous CarbonMembranes. 2002: United States <strong>of</strong> America.166


101. Kusakabe, K., S. Gohgi, and S. Morooka, Carbon Molecular Sieving MembranesDerived from Condensed Polynuclear Aromatic (COPNA) Resins <strong>for</strong> Gas Separations.Industrial & Engineering Chemistry Research, 1998. 37(11): p. 4262-4266.102. Li, S., J.L. Falconer, and R.D. Noble, SAPO-34 Membranes <strong>for</strong> CO2/CH4 Separation.Journal <strong>of</strong> Membrane Science, 2004. 241: p. 121-135.103. Tanihara, N., et al., Gas Permeation Properties <strong>of</strong> Asymmetric Carbon Hollow FiberMembranes Prepared from Asymmetric Polyimide Hollow Fiber. Journal <strong>of</strong> MembraneScience, 1999. 160(2): p. 179-186.104. Shelekhin, A.B., A.G. Dixon, and Y.H. Ma, Theory <strong>of</strong> Gas Diffusion and Permeation inInorganic Molecular-Sieve Membranes. AIChE Journal, 1995. 41(1): p. 58-67.105. Strano, M.S. and H.C. Foley, Modeling Ideal Selectivity Variation in NanoporousMembranes. Chemical Engineering Science, 2002. 58: p. 2745-2758.106. Jiang, J., J.B. Klauda, and S. Sandler, Monte Carlo Simulation <strong>of</strong> O2 and N2Adsorption in Nanoporous Carbon (C168 Schwarzite). Langmuir, 2003. 19(8).107. Jiang, J. and S. Sandler, Monte Carlo Simulation <strong>of</strong> O2 and N2 Mixture Adsorption inNanoporous Carbon (C169 Schwartize). Langmuir, 2003. 19: p. 5936-5941.108. Jiang, J. and S. Sandler, Separation <strong>of</strong> CO2 and N2 by Adsorption in C168 Schwarzite:A Combination <strong>of</strong> Quantum Mechanics and Molecular Simulation Study. 2004.109. Jiang, J. and S.I. Sandler, Nitrogen and Oxygen Mixture Adsorption on CarbonNanotube Bundles from Molecular Simulation. Langmuir, 2004: p. ACS ASAP.110. Arora, G. and S.I. Sandler, Mass Transport <strong>of</strong> O2 and N2 in Nanoporous Carbon (C168schwarzite) using a Quantum Mechanical Force Field and Molecular DynamicsSimulations. Langmuir, 2006. 22(10): p. 4620-4628.111. Arora, G. and S.I. Sandler, Nanoporous Carbon Membranes <strong>for</strong> Separation <strong>of</strong> Nitrogenand Oxygen: Insight from Molecular Simulations. Fluid Phase Equilibria, 2007. 259(1):p. 3-8.112. Frenkel, D. and B. Smit, Understanding Molecular Simulation: from Algorithms toApplications. 2002, San Diego: Academic.113. Klauda, J.B., J. Jiang, and S. Sandler, An Ab Initio Study on <strong>the</strong> Effect <strong>of</strong> CarbonSurface Curvature and Ring Structure on N2(O2) - Carbon Intermolecular Potentials.2003.167


114. Agnihotri, S., et al., Structural Characterization <strong>of</strong> Single-Walled Carbon NanotubeBundles by Experiment and Molecular Simulation. Langmuir, 2005: p. ACS ASAP.115. Arora, G. and S.I. Sandler, Molecular Sieving Using Single Wall Carbon Nanotubes.Nano Letters, 2007. 7(3): p. 565-569.116. Arora, G. and S.I. Sandler, Air Separation by Single Wall Carbon Nanotubes: MassTransport and Kinetic Selectivity. Journal <strong>of</strong> Chemical Physics, 2006. 124(8): p.084702/1-084702/11.117. Arora, G. and S.I. Sandler, Air Separation by Single Wall Carbon Nanotubes.Thermodynamics and Adsorptive Selectivity. Journal <strong>of</strong> Chemical Physics, 2005.123(4): p. 044705/1-044705/9.118. Arora, G., J. Wagner, and S. Sandler, Adsorption and Diffusion <strong>of</strong> Molecular Nitrogenin Single Wall Carbon Nanotubes. Langmuir, 2004.119. Chen, H. and D.S. Sholl, Predictions <strong>of</strong> Selectivity and Flux <strong>for</strong> CH4/H2 Separationsusing Single Walled Carbon Nanotubes as Membranes. Journal <strong>of</strong> Membrane Science,2006. 269(1-2): p. 152-160.120. Paraedes, J.I., et al., N2 Physisorption on Carbon Nanotubes: Computer Simulation andExperimental results. Journal <strong>of</strong> Physical Chemistry, 2003. 107: p. 8905-8916.121. Skoulidas, A.I., D.S. Sholl, and J.K. Johnson, Adsorption and Diffusion <strong>of</strong> CarbonDioxide and Nitrogen through Single-Walled Carbon Nanotube Membranes. Journal <strong>of</strong>Chemical Physics, 2006. 124(5): p. 054708/1-054708/7.122. Yin, Y.F., T. Mays, and B. McEnaney, Adsorption <strong>of</strong> Nitrogen in Carbon NanotubeArrays. Langmuir, 1999. 15: p. 8714-8718.123. Gas Processors Suppliers Association, Engineering Data Book. 11 ed. 1998, Oklahoma:GPSA.124. Gawrys, M., et al., Prevention <strong>of</strong> Water Vapour Adsorption by Carbon MolecularSieves in Sampling Humid Gases. Journal <strong>of</strong> Chromatography, A, 2001. 933(1-2): p.107-116.125. Groszek, A.J., Heats <strong>of</strong> Water Adsorption on Microporous Carbons from Nitrogen andMethane Carriers. Carbon, 2001. 39(12): p. 1857-1862.168


126. Avraham, I., A. Danon, and J.E. Koresh, Water Coadsorption Effect on <strong>the</strong> PhysicalAdsorption <strong>of</strong> N2 and O2 at Room Temperature on Carbon Molecular Sieve Fibers.Physical Chemistry Chemical Physics, 1999. 1(3): p. 479-484.127. Lagorsse, S., F.D. Magalhaes, and A. Mendes, Aging Study <strong>of</strong> Carbon Molecular SieveMembranes. Journal <strong>of</strong> Membrane Science, 2008. 310: p. 494-502.128. Lagorsse, S., et al., Water Adsorption on Carbon Molecular Sieve Membranes.Experimental Data and Iso<strong>the</strong>rm model. Carbon, 2005. 43(13): p. 2769-2779.129. Jones, C.W. and W.J. Koros, Carbon Composite Membranes: A Solution to AdverseHumidity Effects. Industrial & Engineering Chemistry Research, 1995. 34(1): p. 164-7.130. Mattson, J.S., H.B. Mark, Jr., and W.J. Weber, Jr., Identification <strong>of</strong> Surface FunctionalGroups on Active Carbon by Infrared Internal-Reflection Spectrometry. AnalyticalChemistry, 1969. 41(2): p. 355-8.131. Mattson, J.S., et al., Surface Oxide <strong>of</strong> Activated Carbon: Internal ReflectanceSpectroscopic Examination <strong>of</strong> Activated Sugar Carbons. Journal <strong>of</strong> Colloid andInterface Science, 1970. 33(2): p. 284-93.132. Mattson, J.S. and H.B. Mark, Jr., Infrared Internal Reflectance SpectroscopicDetermination <strong>of</strong> Surface Functional Groups on Carbon. Journal <strong>of</strong> Colloid andInterface Science, 1969. 31(1): p. 131-44.133. Petit, J.C. and Y. Bahaddi, New Insight on <strong>the</strong> Chemical Role <strong>of</strong> Water Vapor in <strong>the</strong>Aging <strong>of</strong> Activated Carbon. Carbon, 1993. 31(5): p. 821-5.134. Marsh, H. and F. Rodriguez-Reinoso, Activated Carbon. 2006: Elsevier.135. Hayashi, J., et al., Effect <strong>of</strong> Oxidation on Gas Permeation <strong>of</strong> Carbon Molecular SievingMembranes Based on BPDA-pp'ODA Polyimide. Industrial & Engineering ChemistryResearch, 1997. 36(6): p. 2134-2140.136. Vu, D.Q., W.J. Koros, and S.J. Miller, Effect <strong>of</strong> Condensable Impurity in CO2/CH4Gas Feeds on Per<strong>for</strong>mance <strong>of</strong> Mixed Matrix Membranes using Carbon MolecularSieves. Journal <strong>of</strong> Membrane Science, 2003. 221(1-2): p. 233-239.137. Li, S., et al., Effects <strong>of</strong> Impurities on CO2/CH4 Separations through SAPO-34Membranes. Journal <strong>of</strong> Membrane Science, 2005. 251(1-2): p. 59-66.169


138. Scholes, C.A., S.E. Kentish, and G.W. Stevens, Effects <strong>of</strong> Minor Components inCarbon Dioxide Capture Using Polymeric Gas Separation Membranes. Separation andPurification Reviews, 2009. 38: p. 1-44.139. Pomerantz, N. and Y.H. Ma, Effect <strong>of</strong> H2S on <strong>the</strong> Per<strong>for</strong>mance and Long-TermStability <strong>of</strong> Pd-Cu Membranes. Industrial and Engineering Chemistry Research, 2009.48: p. 4030-4039.140. Anderson, R.B., The Fischer-Tropsch Syn<strong>the</strong>sis. 1984, Orlando: Academic Press Inc.141. Yan, R., et al., Kinetics and Mechanisms <strong>of</strong> H2S Adsorption by Alkaline ActivatedCarbon. Environmental Science and Technology, 2002. 36(20): p. 4460-4466.142. Tamon, H., K. Kitamura, and M. Okazaki, Adsorption <strong>of</strong> Carbon Monoxide onActivated Carbon Impregnated with Metal Halide. AIChE Journal, 1996. 42(2): p. 422-30.143. Mohamad, A.B., et al., Adsorption <strong>of</strong> Carbon Monoxide on Activated Carbon-TinLigand. Journal <strong>of</strong> Molecular Structure, 2000. 550-551: p. 511-519.144. Jones, C.W. and W.J. Koros, Carbon Molecular Sieve Gas Separation Membranes 1 -Preparation and Characterization based on Polyimide Precursors. Carbon, 1994. 32(8):p. 1419-25.145. Stoeber, W., A. Fink, and E. Bohn, Controlled Growth <strong>of</strong> Monodisperse Silica Spheresin <strong>the</strong> Micron Size Range. Journal <strong>of</strong> Colloid and Interface Science, 1968. 26(1): p. 62-9.146. Sadasivan, S., et al., Alcoholic Solvent Effect on Silica Syn<strong>the</strong>sis - NMR and DLSInvestigation. Journal <strong>of</strong> Sol-Gel Science and Technology, 1998. 12(1): p. 5-14.147. Green, D.L., et al., Size, Volume Fraction, and Nucleation <strong>of</strong> Stober SilicaNanoparticles. Journal <strong>of</strong> Colloid and Interface Science, 2003. 266(2): p. 346-358.148. Bogush, G.H., M.A. Tracy, and C.F.I. Zukoski, Preparation <strong>of</strong> Monodisperse SilicaParticles: Control <strong>of</strong> Size and Mass Fraction. Journal <strong>of</strong> Non-Crystalline Solids, 1988.104(1): p. 95-106.149. Zhang, J.H., et al., Preparation <strong>of</strong> Monodisperse Silica Particles with Controllable Sizeand Shape. Journal <strong>of</strong> Materials Research, 2003. 18(3): p. 649-653.170


150. Zipper, M.D. and A.J. Hill, The Application <strong>of</strong> Positron Annihilation LifetimeSpectroscopy to <strong>the</strong> Study <strong>of</strong> Glassy and Partially Crystalline Materials. MaterialsForum, 1994. 18: p. 215-233.151. Polarz, S. and B. Smarsly, Nanoporous Materials. Journal <strong>of</strong> Nanoscience andNanotechnology, 2002. 2(6): p. 581-612.152. Schmidt, W., Positron Annihilation Lifetime Spectroscopy, in Handbook <strong>of</strong> PorousSolids, F. Schuth, K.S.W. Sing, and J. Weitkamp, Editors. 2002, Wiley-VCH:Cambridge.153. Yampolskii, Y. and V. Shantarovich, Positron Annihilation Lifetime Spectroscopy andO<strong>the</strong>r Methods <strong>for</strong> Free Volume Evaluation in Polymers, in Materials Science <strong>of</strong>Membranes <strong>for</strong> Gas and Vapor Separation, Y. Yampolskii, I. Pinnau, and B. Freeman,Editors. 2006, John Wiley and Sons: West Sussex.154. Kansy, J., Microcomputer Program <strong>for</strong> Analysis <strong>of</strong> Positron Annihilation LifetimeSpectra. Nuclear Instruments & Methods in Physics Research, Section A: Accelerators,Spectrometers, Detectors, and Associated Equipment, 1996. 374(2): p. 235-244.155. Jenkins, G.M. and K. Kawamura, Polymeric <strong>carbon</strong>s: <strong>carbon</strong> fiber,glass and char. 1976,London: Cambridge University Press.156. Duthie, X., et al., Operating Temperature Effects on <strong>the</strong> Plasticization <strong>of</strong> Polyimide GasSeparation Membranes. Journal <strong>of</strong> Membrane Science, 2007. 294(1-2): p. 40-49.157. Fisher, Control Valve Handbook. 1977, Iowa, USA: Fisher Controls158. de Bruijn, J.N.H., et al., Maximizing Molecular Sieve Per<strong>for</strong>mance in Natural GasProcessing. Annual Convention Proceedings - Gas Processors Association, 2002. 81st:p. 209-226.159. Rao, M.B., R.G. Jenkins, and W.A. Steele, Potential Functions <strong>for</strong> Diffusive Motion inCarbon Molecular Sieves. Langmuir, 1985. 1(1): p. 137-41.160. Park, H.B., et al., Polymers with Cavities Tuned <strong>for</strong> Fast Selective Transport <strong>of</strong> SmallMolecules and Ions. Science (Washington, DC, United States), 2007. 318(5848): p.254-258.161. United Nations, Kyoto Protocol to <strong>the</strong> United Nations Framework Convention onClimate Change. 1997, Kyoto, Japan: United Nations.171


162. UNEP and UNFCCC, Climate Change In<strong>for</strong>mation Kit, ed. M. Williams. 2002,Chatelaine, Switzerland: UNEP and UNFCCC.163. IPCC, Climate Change 2001: The Scientific Basis. 2001, New York: CambridgeUniversity Press.164. IPCC, Introduction to <strong>the</strong> Intergovernmental Panel on Climate Change (IPCC). 2003,Geneva: IPCC.165. United Nations, United Nations Framework Convention on Climate Change. 1992.166. Australian Greenho<strong>use</strong> Office, Tracking to <strong>the</strong> Kyoto Target 2007. 2007, AustralianGovernment: Canberra.167. Saddler, H., M. Diesendorf, and R. Denniss, A Clean Energy Future <strong>for</strong> Australia. 2004,WWF.168. Australian Greenho<strong>use</strong> Office, Renewable Energy Commercialisation in Australia.2003, Canberra: Australian Greenho<strong>use</strong> Office.169. Australian Greenho<strong>use</strong> Office, Renewable Opportunities. 2003, Canberra: AustralianGreenho<strong>use</strong> Office.170. Herzog, H.J., An Introduction to CO2 Separation and Capture Technology. MIT EnergyLaboratory Working Paper, 1999.171. Chapel, D., J. Ernest, and C. Mariz. Recovery <strong>of</strong> CO2 from Flue Gases: CommercialTrends. in First National Conference on Carbon Sequestration. 2001. Washington,United States <strong>of</strong> America.172. Falk-Pedersen, O., et al. Gas Treatment Using Membrane Gas/Liquid Contactors. in 5thInternational Conference on Greenho<strong>use</strong> Gas Control Technologies. 2000. Cairns,Australia.173. IEA Greenho<strong>use</strong> Gas R&D Programme, Carbon Dioxide Capture from Power Stations.1994: IEA.174. Saracco, G., G.F. Versteeg, and W.P.M. van Swaaij, Current Hurdles to <strong>the</strong> Success <strong>of</strong>High-Temperature Membrane Reactors. Journal <strong>of</strong> Membrane Science, 1994. 95: p.105-123.175. Taylor, J.R., An Introduction to Error Analysis: The Study <strong>of</strong> Uncertainties in PhysicalMeasurements. 1982: University Science Books.172


176. Abrutis, A., et al., Preparation <strong>of</strong> Dense, Ultra-Thin MIEC Ceramic Membranes byAtmospheric Spray-Pyrolysis Technique. Journal <strong>of</strong> Membrane Science, 2004. 240(1-2): p. 113-122.177. Chern, R.T., et al., Selective Permeation <strong>of</strong> CO2 and CH4 through Kapton Polyimide:Effects <strong>of</strong> Penetrant Competition and Gas-Phase Nonideality. Journal <strong>of</strong> PolymerScience, Polymer Physics Edition, 1984. 22(6): p. 1061-84.178. Petersen, J., M. Matsuda, and K. Haraya, Capillary Carbon Molecular Sieve MembranesDerived from Kapton <strong>for</strong> High-Temperature Gas Separation. Journal <strong>of</strong> MembraneScience, 1997. 131(1-2): p. 85-94.179. Wang, K., H. Suda, and K. Haraya, The Characterization <strong>of</strong> CO2 Permeation in aCMSM Derived from Polyimide. Separation and Purification Technology, 2003. 31(1):p. 61-69.173


174


APPENDICES175


176


APPENDIX A – RELEVANT BACKGROUND INFORMATIONA.1 Global WarmingGreenho<strong>use</strong> gases absorb infrared radiation and <strong>the</strong>re<strong>for</strong>e directly affect <strong>the</strong> energy flow through<strong>the</strong> Earth’s atmosphere. Increasing <strong>the</strong> quantity <strong>of</strong> greenho<strong>use</strong> gases within <strong>the</strong> Earth’satmosphere leads to a build up in energy and hence an increase in <strong>temperature</strong>. Thisphenomenon is known as global warming.Altering <strong>the</strong> Earth’s climate in such a way is predicted to have devastating consequences, suchas increased ocean levels and changes to wea<strong>the</strong>r patterns which will destroy <strong>the</strong> delicatebalance <strong>for</strong> various ecological systems.Greenho<strong>use</strong> gases are specified in <strong>the</strong> Kyoto Protocol as <strong>carbon</strong> dioxide (CO 2 ), methane (CH 4 ),nitrous oxide (N 2 O), hydr<strong>of</strong>luoro<strong>carbon</strong>s (HFCs), perfluoro<strong>carbon</strong>s (PFCs) and sulphurhexafluoride (SF 6 ) [161]. Activities within <strong>the</strong> global community are increasing <strong>the</strong>concentrations <strong>of</strong> <strong>the</strong>se greenho<strong>use</strong> gases in <strong>the</strong> atmosphere. Carbon dioxide is produced from<strong>the</strong> burning <strong>of</strong> fossil fuels; methane and nitrous oxide are produced through land clearing andagriculture; and industrial processes produce hydr<strong>of</strong>luoro<strong>carbon</strong>s, perfluoro<strong>carbon</strong>s and sulphurhexafluoride [162].375Current Atmospheric Level <strong>of</strong> CO 2= 370 ppmvCarbon Dioxide Concentration (ppmv)350Methane Concentration (ppmv)Temperature over AntarcticaAtmospheric <strong>carbon</strong> dioxide concentrationAtmospheric methane concentrationTemperature relative to present climate (°C)Thousands <strong>of</strong> years be<strong>for</strong>e present (Ky BP)Figure A-1 : Correlation between <strong>the</strong> Earth’s <strong>temperature</strong> and atmospheric <strong>carbon</strong>dioxide (data sourced from [163])177


Whilst providing <strong>the</strong> lowest global warming potential, <strong>carbon</strong> dioxide accounts <strong>for</strong> <strong>the</strong> majority<strong>of</strong> greenho<strong>use</strong> gases emitted to <strong>the</strong> atmosphere through human activity and as such greenho<strong>use</strong>gas emissions are generally quoted in tonnes <strong>of</strong> <strong>carbon</strong> dioxide equivalent. The correlationbetween <strong>the</strong> Earth’s <strong>temperature</strong> and <strong>the</strong> atmospheric concentrations <strong>of</strong> <strong>carbon</strong> dioxide andmethane is presented in Figure A-1.A.2 International Recognition <strong>of</strong> Climate ChangeThe United Nations and <strong>the</strong> World Meteorological Society created <strong>the</strong> Intergovernmental Panelon Climate Change (IPCC) in 1988 in recognition <strong>of</strong> <strong>the</strong> problem <strong>of</strong> global climate change. Therole <strong>of</strong> IPCC is to provide an objective source <strong>of</strong> accepted in<strong>for</strong>mation on global climate change.The IPCC does not conduct research or monitor data, it collates and <strong>the</strong>n presents peer-reviewedin<strong>for</strong>mation on climate change from around <strong>the</strong> world such that a balanced viewpoint isobtained to enable policy making [164].The First Assessment Report published by <strong>the</strong> IPCC in 1990 presented a confirmation <strong>of</strong> <strong>the</strong>impact <strong>of</strong> human activity on global climate change based on scientific data. Not surprisingly thisreport created concern within <strong>the</strong> international community and prompted <strong>the</strong> establishment <strong>of</strong><strong>the</strong> Intergovernmental Negotiating Committee, which adopted <strong>the</strong> United Nations (UN)Framework Convention on Climate Change in 1992 [164]. The objective <strong>of</strong> <strong>the</strong> UN FrameworkConvention on Climate Change is to allow “stabilisation <strong>of</strong> greenho<strong>use</strong> gas concentrations in <strong>the</strong>atmosphere at a level that would prevent dangerous anthropogenic interference with <strong>the</strong> climatesystem” [165].At <strong>the</strong> Earth Summit hosted in Rio de Janeiro in June 1992, <strong>the</strong> UN Framework Convention onClimate Change was open to signature by participating countries. Countries that signed <strong>the</strong>convention were committed to monitoring greenho<strong>use</strong> gas emissions, developing programs <strong>for</strong>mitigating climate change and promoting public awareness <strong>of</strong> climate change. The frameworkprescribes that developed countries take <strong>the</strong> lead on addressing climate change [165].The Kyoto Protocol to <strong>the</strong> UN Framework Convention on Climate Change was established inKyoto, Japan in December 1997. The Second Assessment Report published by <strong>the</strong> IPCC in1996 contributed significantly to <strong>the</strong> negotiations <strong>of</strong> <strong>the</strong> Kyoto Protocol [164]. The primaryobjective <strong>of</strong> <strong>the</strong> Kyoto Protocol is to reduce <strong>the</strong> worldwide greenho<strong>use</strong> gas emissions by 5 %below <strong>the</strong> 1990 levels during <strong>the</strong> commitment period <strong>of</strong> 2008 to 2012. Consequently, eachcountry that is a Party to <strong>the</strong> Kyoto Protocol has been assigned an emissions quota based on1990 levels <strong>for</strong> <strong>the</strong> commitment period [161]. As <strong>of</strong> <strong>the</strong> 16 th <strong>of</strong> February 2005, <strong>the</strong> Kyoto178


Protocol entered into <strong>for</strong>ce and as such 36 (Annex B) developed countries along with <strong>the</strong>European Union are required to reduce emissions <strong>for</strong> <strong>the</strong> commitment period. Of <strong>the</strong> developingnations, including China and India, 137 countries have ratified <strong>the</strong> protocol but are only requiredto monitor and report emissions ra<strong>the</strong>r than reduce emissions. With Russia and Australia havingnow ratified <strong>the</strong> Kyoto Protocol, <strong>the</strong> only major country not to have joined is <strong>the</strong> United States<strong>of</strong> America.The most recent IPCC reports, <strong>the</strong> Third and Fourth Assessment Reports, were published by <strong>the</strong>IPCC in 2001 and 2007, respectively. Both reports provides stronger than ever evidence <strong>of</strong> <strong>the</strong>anthropological impact on climate change [1, 163].The global community is now looking towards what happens after 2012 when <strong>the</strong> commitmentperiod to <strong>the</strong> Kyoto Protocol ends. The United Nations Climate Change Conference held in Bali,Indonesia in December 2007 led to <strong>the</strong> <strong>for</strong>mation <strong>of</strong> <strong>the</strong> “Bali Roadmap” <strong>for</strong> a futureinternational agreement on climate change, which is hoped to be completed by <strong>the</strong> end <strong>of</strong> 2009.A.3 Australian Recognition <strong>of</strong> Climate ChangeAustralia’s economy is dominated by <strong>the</strong> <strong>use</strong> <strong>of</strong> fossil fuels, especially coal, and it is a netexporter <strong>of</strong> energy. None <strong>of</strong> <strong>the</strong> energy generated by Australia is produced using nuclearresources. With <strong>the</strong> change <strong>of</strong> government in November 2007, Australia finally ratified <strong>the</strong>Kyoto Protocol on <strong>the</strong> 3 rd <strong>of</strong> December 2007.Under <strong>the</strong> Kyoto Protocol, Australia was allocated <strong>the</strong> target <strong>of</strong> 108 % <strong>of</strong> 1990 emission levelsover <strong>the</strong> commitment period from 2008 to 2012 (Australia, like some o<strong>the</strong>r nations, has foc<strong>use</strong>don <strong>the</strong> year 2010 as a target date). Australia was one <strong>of</strong> only three Annex B countries allocatedan increased target over <strong>the</strong> 1990 emission levels [161]. This is attributed to Australia’s relianceon fossil fuels and it having a relatively <strong>high</strong> economic growth <strong>for</strong> a developed country.As shown in Figure A-2, <strong>the</strong> Australian Greenho<strong>use</strong> Office has projected that Australia willapproach but not achieve this target as greenho<strong>use</strong> gas emission levels are predicted to reach <strong>the</strong>Kyoto Target (108 %) <strong>of</strong> <strong>the</strong> 1990 level (596 million tonnes <strong>of</strong> <strong>carbon</strong> dioxide equivalent) onaverage <strong>for</strong> <strong>the</strong> period from 2008 to 2012. The projected emission level <strong>of</strong> 108 % is on <strong>the</strong> basisthat greenho<strong>use</strong> gas reduction measures will be introduced. Maintaining business as usual andwithout adopting any greenho<strong>use</strong> gas reduction measures predicts an emission level <strong>of</strong> 124 %<strong>for</strong> <strong>the</strong> 2008 to 2012 period [166].179


Greenho<strong>use</strong> gas measures adopted by <strong>the</strong> Australian Government include <strong>the</strong> implementation <strong>of</strong>a range <strong>of</strong> polices and programs to reduce emissions, establishment <strong>of</strong> new <strong>for</strong>est plantationsand land <strong>use</strong> changes [166]. Under <strong>the</strong> Kyoto accounting rules, Australia is permitted to includeland <strong>use</strong> changes in calculating <strong>the</strong> emissions levels <strong>for</strong> <strong>the</strong> 2008 to 2012 period [161]. Land <strong>use</strong>changes make a significant contribution to <strong>the</strong> reductions.CO2 Equivalent Emissions (Millions Tonnes/Year)7006005004003002001000-1001990 LevelsPredicted 2008 - 2012 Levels2874298893Predicted 596 mT/year (w ith reduction measures) = 108 % <strong>of</strong> 1990 Emissions Level18 1525Energy Agriculture Waste IndustrialProcessesPredicted 685 mT/year (business as usual) = 124 % <strong>of</strong> 1990 Emissions Level 685Kyoto Target 585 mT/year = 108 % <strong>of</strong> 1990 EmissionsLevel380-2113644554Forestry Sinks Land Use Change Total596Figure A-2 : Predicted greenho<strong>use</strong> gas emissions from by sector <strong>for</strong> 1990 and <strong>the</strong> predicted2008 to 2012 period (data sourced from [166])A.4 Reducing CO 2 Emissions from Australia’s Energy ProductionIn 1990, energy production from stationary sources such as power stations accounted <strong>for</strong> 39 %<strong>of</strong> Australia’s total greenho<strong>use</strong> gas emissions (primarily <strong>carbon</strong> dioxide, CO 2 ). Emission levelsfrom stationary sources is increasing and is predicted to account <strong>for</strong> 49 % <strong>of</strong> Australia’s totalgreenho<strong>use</strong> gas emissions <strong>for</strong> <strong>the</strong> 2008 to 2012 period [166].Fur<strong>the</strong>r reductions in greenho<strong>use</strong> gas emissions from energy production can be achieved via (i)increasing energy efficiency and (ii) using less <strong>carbon</strong> intensive energy sources such asrenewable fuels or solar energy supplies [167].Increasing energy efficiency is achieved with both advances in technology and an increasedsocial awareness usually brought about by an increase in energy cost or with a threat to energyavailability. The Australian society does not have an inherent sense <strong>of</strong> urgency <strong>for</strong> increasingenergy efficiency as fossil fuels remain af<strong>for</strong>dable and available.180


Developing an energy efficient society can be achieved through education and governmentpolicy. The introduction <strong>of</strong> a <strong>carbon</strong> tax is one example <strong>of</strong> how policy would promote energyefficiency through <strong>the</strong> <strong>high</strong>er cost <strong>of</strong> fossil fuels. A mix <strong>of</strong> measures, such as <strong>carbon</strong> taxes,levies, en<strong>for</strong>cing efficiency standards and providing incentives <strong>for</strong> market-available low-energyusage equipment, will eventually be required. The Australian Federal Government has not yetintroduced a <strong>carbon</strong> tax or a <strong>carbon</strong> trading market. However, <strong>the</strong> latter is being planned at <strong>the</strong>State and Territory level.Af<strong>for</strong>dable and easily accessible renewable energy will not be available <strong>for</strong> several decadesunless <strong>the</strong>re is a significant discovery in <strong>the</strong> short term. Renewable energy, with a capacity <strong>of</strong>16,000 GWh, presently only accounts <strong>for</strong> 10.5 % <strong>of</strong> Australia’s electricity generation with <strong>the</strong>majority sourced from hydro-electric power stations [168].The previous Australian Government introduced <strong>the</strong> Mandatory Renewable Energy Target(MRET) as method <strong>of</strong> increasing electricity production from renewable energy sources. TheMRET requires <strong>the</strong> production <strong>of</strong> an additional 9,500 GWh <strong>of</strong> renewable energy annually by2010 and will result in a reduction <strong>of</strong> 6.5 million tonnes <strong>of</strong> <strong>carbon</strong> dioxide emitted per annum[169]. While this reduction is modest it has allowed <strong>the</strong> emergence <strong>of</strong> a wind energy industry.Enhanced uptake <strong>of</strong> renewable energy will also occur as <strong>the</strong> cost is reduced with advances intechnology.Alternatively, geological storage (sequestration) <strong>of</strong> <strong>carbon</strong> dioxide has great potential <strong>for</strong> asocially acceptable solution that can be <strong>use</strong>d to significantly reduce greenho<strong>use</strong> gas emissions in<strong>the</strong> interim while sustainable solutions are fur<strong>the</strong>r advanced and implemented. Carbon dioxidesequestration is particularly applicable to Australia’s situation where <strong>the</strong> majority <strong>of</strong> electricityis generated from burning coal.A.5 Opportunities <strong>for</strong> CO 2 Sequestration in AustraliaCO 2 sequestration is best suited to large stationary emitters <strong>of</strong> <strong>carbon</strong> dioxide. Approximatelyhalf <strong>of</strong> Australia’s <strong>carbon</strong> dioxide emissions are from large stationary sources and could becaptured using <strong>carbon</strong> dioxide sequestration. A breakdown <strong>of</strong> Australia’s large stationarysources by industry is presented in Figure A-3.181


80.0%73.3%70.0%60.0%50.0%40.0%30.0%20.0%7.6% 6.0% 5.8% 4.5%10.0%2.8%0.0%PowerStationsSteel IndustryNon-FerrousMetalsPetroleumIndustryO<strong>the</strong>rIndustrialOil RefineriesFigure A-3 : Australia’s large stationary sources <strong>of</strong> <strong>carbon</strong> dioxide by industry (datacourtesy <strong>of</strong> <strong>the</strong> CRC <strong>for</strong> Greenho<strong>use</strong> Gas Technologies)A.6 Challenges <strong>for</strong> CO 2 Sequestration in AustraliaThe major technical challenge <strong>for</strong> CO 2 sequestration is reducing <strong>the</strong> cost. Developing a positivepublic perception <strong>of</strong> <strong>the</strong> geological storage <strong>of</strong> <strong>carbon</strong> dioxide should also be recognised as animportant challenge <strong>for</strong> implementing <strong>carbon</strong> dioxide sequestration.The present cost <strong>of</strong> CO 2 sequestration is estimated by <strong>the</strong> IEA (International Energy Agency) tobe between $40 and $60 US per tonne <strong>of</strong> <strong>carbon</strong> dioxide emissions avoided [6, 7]. In <strong>the</strong> case <strong>of</strong>power generation, this cost translates to an increase <strong>of</strong> 1.5 to 3 US cents per kilowatt hour to <strong>the</strong>price <strong>of</strong> electricity. Using <strong>the</strong> average 1998 electricity price <strong>for</strong> OECD countries, an increase <strong>of</strong>1.5 to 3 US cents per kilowatt hour represents a 11 to 43 % increase in <strong>the</strong> domestic electricityprice [7].From <strong>the</strong> data supplied by <strong>the</strong> IEA [7], it is estimated that capturing and compressing <strong>the</strong> <strong>carbon</strong>dioxide from <strong>the</strong> source represents approximately 90 % <strong>of</strong> <strong>the</strong> sequestration cost. The cost <strong>of</strong>transporting and storing <strong>the</strong> CO 2 is relatively small. There<strong>for</strong>e, significant improvements to <strong>the</strong>capture technology should remain a focus <strong>for</strong> major reduction in <strong>the</strong> cost <strong>of</strong> <strong>carbon</strong> dioxidesequestration process.182


A.7 CO 2 Capture TechnologiesSeparating CO 2 from o<strong>the</strong>r gases is an integral part <strong>of</strong> many existing processes, such as <strong>the</strong>purification <strong>of</strong> natural gas be<strong>for</strong>e <strong>the</strong> liquefaction process. Existing gas separation technologyincludes gas absorption, gas adsorption, cryogenics, gas hydrates and gas separation<strong>membranes</strong>.A.7.1AbsorptionIn <strong>the</strong> gas absorption (“scrubbing”) process, <strong>the</strong> gas is contacted with a liquid (solvent) and oneor more components <strong>of</strong> <strong>the</strong> gas are preferentially absorbed into liquid. The gas-rich liquid is<strong>the</strong>n sent to a regeneration process where <strong>the</strong> separated gas is released from <strong>the</strong> liquid.The gas absorption process is ei<strong>the</strong>r a chemical or physical absorption process. In <strong>the</strong> chemicalabsorption process, <strong>the</strong> <strong>carbon</strong> dioxide is bonded to <strong>the</strong> solvent via a chemical reaction.Commercial chemical solvents are amine or <strong>carbon</strong>ate based, such as monoethanolamine(MEA), diethanolamine (DEA) or hot potassium <strong>carbon</strong>ate as <strong>use</strong>d in <strong>the</strong> Benfield process. In<strong>the</strong> physical absorption process, <strong>the</strong> <strong>carbon</strong> dioxide is physically absorbed into <strong>the</strong> solventaccording to Henry’s law. Commercial physical solvents include Selexol (glycol) and Rectisol(methanol). The physical absorption process is favourable <strong>for</strong> <strong>high</strong>er pressure applications.All commercial <strong>carbon</strong> dioxide capture processes from power station flue gases currently <strong>use</strong>solvent technology [170]. The advantage <strong>of</strong> gas absorption is that it is a mature technology andwell proven <strong>for</strong> <strong>the</strong> application. The disadvantage <strong>of</strong> gas absorption is <strong>the</strong> size and amount <strong>of</strong>equipment required <strong>for</strong> <strong>the</strong> process. The application <strong>of</strong> chemical absorption (Fluor DanielEconamineSM FG Process) has been shown to separate and produce <strong>carbon</strong> dioxide at a cost <strong>of</strong>US$ 29.50 / tonne [171].A promising adaptation <strong>of</strong> <strong>the</strong> gas absorption process is <strong>the</strong> gas adsorption membrane contactordeveloped recently by Kvaerner Oil and Gas. The gas adsorption membrane contactor replaces<strong>the</strong> traditional packed column resulting in a significant reduction in equipment size [172].A.7.2AdsorptionIn <strong>the</strong> gas adsorption process one or more <strong>of</strong> <strong>the</strong> components <strong>of</strong> <strong>the</strong> gas are preferentiallyabsorbed onto a solid desiccant. The solid desiccant is <strong>the</strong>n regenerated via heat or pressureswing to remove <strong>the</strong> absorbed gases. Molecular sieves are a commonly <strong>use</strong>d solid desiccant <strong>for</strong><strong>the</strong> removal <strong>of</strong> water or trace components <strong>of</strong> gases such as hydrogen sulphide.183


Solid desiccants <strong>for</strong> <strong>the</strong> bulk removal <strong>of</strong> <strong>carbon</strong> dioxide in flue gases are not presentlyconsidered <strong>for</strong> commercial <strong>use</strong> due to <strong>the</strong> limited adsorption capacity and low selectivity [173].A.7.3CryogenicsThe flue gases are cooled to cryogenic <strong>temperature</strong>s such that <strong>the</strong> <strong>carbon</strong> dioxide is condensedfrom <strong>the</strong> o<strong>the</strong>r gases. The recovery <strong>of</strong> <strong>carbon</strong> dioxide is dependent on <strong>the</strong> lowest achievable<strong>temperature</strong>. Increasing <strong>the</strong> pressure will also increase <strong>the</strong> recovery.Separating methane from <strong>carbon</strong> dioxide is difficult and requires <strong>the</strong> <strong>use</strong> <strong>of</strong> cryogenicdistillation, such as <strong>the</strong> Ryan-Holmes process, which was developed in <strong>the</strong> 1980s <strong>for</strong> separatingacid gases (such as <strong>carbon</strong> dioxide) from light hydro<strong>carbon</strong> gases [123].Cryogenic separation technology is best suited <strong>for</strong> flue gases that contain <strong>high</strong> levels <strong>of</strong> <strong>carbon</strong>dioxide such as those produced from a pre-combustion or oxy-fuels process. The majordisadvantage <strong>of</strong> cryogenic separation technology is that it is inherently energy intensive, whichwill lead to a large operating cost.A.7.4Gas CryogenicsA hydrate is a crystalline structure that <strong>for</strong>ms from water and a gas, such as <strong>carbon</strong> dioxide, andhas <strong>the</strong> appearance and characteristics <strong>of</strong> ice. If <strong>the</strong> flue gases are carefully cooled with a smallamount <strong>of</strong> water present, <strong>carbon</strong> dioxide hydrates can be selectively <strong>for</strong>med enabling separationfrom <strong>the</strong> o<strong>the</strong>r gases. The <strong>temperature</strong>s and pressures required are less extreme than cryogenicseparationA.7.5Gas Separation MembranesMembrane technology <strong>of</strong>fers exciting promise <strong>for</strong> <strong>use</strong> in <strong>carbon</strong> dioxide capture and <strong>the</strong>re issubstantial research activity in this area. Although <strong>membranes</strong> are not yet proven in <strong>the</strong>application <strong>of</strong> <strong>carbon</strong> dioxide capture from flue gases, <strong>the</strong>y are currently <strong>use</strong>d <strong>for</strong> <strong>the</strong> separation<strong>of</strong> <strong>carbon</strong> dioxide from natural gas in <strong>the</strong> oil and gas industry.Membranes have <strong>the</strong> advantage <strong>of</strong> being very compact in comparison with o<strong>the</strong>r capturetechnologies. If an improvement in separation efficiency can be achieved along with flexibilityto operate in <strong>high</strong>er <strong>temperature</strong>s, membrane technology could <strong>of</strong>fer significant cost advantagesover o<strong>the</strong>r technologies.184


The <strong>membranes</strong> presently <strong>use</strong>d in commercial <strong>carbon</strong> dioxide gas separation are polymer basednon-porous <strong>membranes</strong> such as spiral wound cellulose acetate <strong>membranes</strong> and polyimidehollow fibre <strong>membranes</strong>. These polymeric <strong>membranes</strong> are presently limited to operating<strong>temperature</strong>s below <strong>the</strong> glass transition <strong>temperature</strong>, where <strong>the</strong> structure <strong>of</strong> <strong>the</strong> polymer changesfrom a glassy to a rubbery state and <strong>the</strong> mechanical strength is drastically reduced. Polyimide,with a glass transition <strong>temperature</strong> <strong>of</strong> 300 °C, has one <strong>of</strong> <strong>the</strong> <strong>high</strong>est polymer glass transition<strong>temperature</strong>s [40]. The <strong>temperature</strong> limit <strong>the</strong>re<strong>for</strong>e necessitates cooling <strong>of</strong> flue gases be<strong>for</strong>e<strong>carbon</strong> dioxide separation can take place.Porous <strong>membranes</strong>, from ceramic materials, with pore sizes less than 1 µm are presently beingexplored <strong>for</strong> <strong>use</strong> in gas separation as <strong>the</strong>y <strong>of</strong>fer <strong>the</strong> advantage <strong>of</strong> stability at <strong>high</strong>er operating<strong>temperature</strong>s [174]. Nanoporous <strong>carbon</strong> is an example <strong>of</strong> a material that can be <strong>use</strong>d to makeporous <strong>membranes</strong> suitable <strong>for</strong> operation at <strong>high</strong> <strong>temperature</strong>s.185


186


APPENDIX B – STANDARD OPERATING PROCEDURESB.1 Standard operating procedure <strong>for</strong> <strong>the</strong> CPVV (Mixed Gas) apparatus187


188


189


190


191


192


193


194


195


196


197


198


B.2 Standard operating procedure <strong>for</strong> <strong>the</strong> CPVV (Mixed Gas) apparatus in <strong>the</strong>presence <strong>of</strong> condensable impurities199


200


201


202


203


204


205


206


207


208


209


210


B.3 Standard operating procedure <strong>for</strong> <strong>the</strong> CPVV (Mixed Gas) apparatus in <strong>the</strong>presence <strong>of</strong> non-condensable impurities211


212


213


214


215


216


217


218


219


220


221


222


B.4 Risk Assessment <strong>for</strong> <strong>the</strong> operation <strong>of</strong> <strong>the</strong> CPVV (Mixed Gas) apparatus in<strong>the</strong> presence <strong>of</strong> non-condensable impurities223


224


225


226


227


228


229


230


APPENDIX C – EXPERIMENTAL METHODSC.1 Pure Gas Permeance MeasurementC.1.1Error AnalysisThe permeance <strong>of</strong> a gas (A) using a pure gas feed was calculated via Equation 3-1 Section 3.4.1.The error analysis method <strong>use</strong>d [175] <strong>use</strong>s <strong>the</strong> following relationships as shown in Table C-1 <strong>for</strong>calculating <strong>the</strong> overall error from <strong>the</strong> individual errors. The variable Z depends on <strong>the</strong> variablesA and B. The respective errors are denoted as E Z , E A and E B .Table C-1 : Error Analysis Relationships [175]Relationship between Z, A and B Relationships Between <strong>the</strong> Errors <strong>of</strong> Z, A and B.Z = A +BZ = A − BZ = ABZ =ABZ = ln(A)⎛⎜⎝⎛⎜⎝22( EZ) = ( EA) + ( EB22( EZ) = ( EA) + ( EBEZZEZZ2⎞⎟⎠2⎞⎟⎠⎛= ⎜⎝⎛= ⎜⎝EEAAEAAZ =2⎞⎟⎠2⎞⎟⎠EAA))222⎛ EB⎞+ ⎜ ⎟⎝ B ⎠2⎛ EB⎞+ ⎜ ⎟⎝ B ⎠Using <strong>the</strong>se relationships, <strong>the</strong> error analysis <strong>for</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> permeance is as follows.⎛⎜⎝EP'P'2⎞⎟⎠⎛ E= ⎜⎝ VVpp⎞⎟⎠2⎛ EA+ ⎜⎝ A2⎞⎟⎠⎛ E+ ⎜⎝ TT2⎞⎟⎠⎛ Ep+ ⎜⎝ Pff⎞⎟⎠2⎛ Ep+ ⎜⎝ Ppp⎟ ⎞⎠2Equation C-1The error analysis <strong>for</strong> <strong>the</strong> CO 2 permeance from a typical pure gas experiment (MembraneNPC550B) is shown in Table C-3. Substituting <strong>the</strong> various values <strong>of</strong> E A /A into Equation C-1gives a 1.38% error in <strong>the</strong> CO 2 permeance. The error in <strong>the</strong> CH 4 permeance is also 1.38 % as <strong>the</strong>relative contributing errors are <strong>the</strong> same.The pure gas selectivity or permselectivity was calculated from <strong>the</strong> ratio <strong>of</strong> <strong>the</strong> pure gaspermeances as presented earlier as Equation 2-3 in Section 2.2.2. Following <strong>the</strong> error analysisrelationships presented in Table C-1, <strong>the</strong> error in <strong>the</strong> selectivity is described as shown inEquation C-2.231


2⎛ Eα⎞⎜ ⎟⎝ α ⎠⎛ E= ⎜⎝P'P'CO2CO2⎞⎟⎠2⎛ E+ ⎜⎝P'P'CH 4CH 4⎟ ⎞⎠2Equation C-2Substituting <strong>the</strong> errors in <strong>the</strong> CO 2 and CH 4 permeances (1.38%) gives a 1.95% error in <strong>the</strong>permselectivity. Thus, <strong>for</strong> membrane NPC550B with an initial pure gas CO 2 /CH 4permselectivity <strong>of</strong> 8.9, <strong>the</strong> error band is +/- 0.2.The error method <strong>use</strong>d here was checked using duplicate measurements from <strong>the</strong> samemembrane. The manufacturing variability between <strong>membranes</strong> is so large that comparingmeasurements between different <strong>membranes</strong> would not indicate <strong>the</strong> error in <strong>the</strong> pure gasper<strong>for</strong>mance measurements.As duplicate measurements had not been made <strong>for</strong> membrane NPC550B, this analysis wascompleted using <strong>the</strong> 2 nd coat <strong>of</strong> membrane NPC400C. The CO 2 /CH 4 permselectivities calculatedfrom <strong>the</strong> two measurements taken 6 days apart are shown in Table C-2. The error is a largerthan that calculated via <strong>the</strong> addition <strong>of</strong> <strong>the</strong> errors in <strong>the</strong> instruments. However, <strong>the</strong> time between<strong>the</strong> two measurements can not be ignored as attributing to <strong>the</strong> difference in <strong>the</strong> measurements assome exposure to <strong>the</strong> atmosphere has occurred.With this in mind, <strong>the</strong> errors calculated using <strong>the</strong> addition <strong>of</strong> <strong>the</strong> errors in <strong>the</strong> instruments hasbeen <strong>use</strong>d <strong>for</strong> <strong>the</strong> error bands on <strong>the</strong> pure gas results presented in this <strong>the</strong>sis.Table C-2 : Error analysis <strong>of</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> pure gas CO 2 CH4 permselectivityfrom duplicate measurements <strong>of</strong> Membrane NPC400C taken 6 days apart.Measurement Date CO 2 /CH 4 Permselectivity1 1 st Feb 2007 2.762 7 th Feb 2007 2.95Error 6.88%232


Table C-3 : Error analysis <strong>of</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> pure gas CO 2 permeance <strong>for</strong> Membrane NPC550B. The error analysis is independent <strong>of</strong> <strong>the</strong>absolute values <strong>of</strong> <strong>the</strong> variables as <strong>the</strong> error is a percentage <strong>of</strong> <strong>the</strong> absolute value with <strong>the</strong> exception <strong>of</strong> <strong>the</strong> values <strong>of</strong> <strong>the</strong> membrane area. As <strong>the</strong>membrane area reading is <strong>the</strong> same <strong>for</strong> <strong>the</strong> CH 4 permeance, <strong>the</strong> error analysis <strong>for</strong> <strong>the</strong> CH 4 permeance gives <strong>the</strong> same result.Variable (A) Value (A) Reading Error (E A ) Comments E A /A E A /A(%)Permeate Volume V p cm 3 525.00 5.00 +/- 5 cm 3 error in volume 0.0095 0.95Membrane Area A m 2 9.62 x 10 -4 7.86 x 10 -7 +/- 1 mm error indiameter0.0008 0.08Permeate Temperature T K 306.00 0.85% <strong>of</strong> ValueManufacturerspecification0.0085 0.85Feed Pressure p f kPa(a) 531.14 0.50% <strong>of</strong> ValueManufacturerspecification0.0050 0.50Permeate Pressure p p kPa(a) Increasing 0.13% <strong>of</strong> ValueManufacturerspecification0.0013 0.13233


C.1.2Sample <strong>of</strong> Calculated ResultsThe permeate pressure rise plots <strong>for</strong> <strong>the</strong> gases CO 2 , CH 4 and N 2 <strong>use</strong>d to determine <strong>the</strong> respectivepure gas permeances from <strong>the</strong> steady state gradients as described in Section 3.4.1 is presented inFigure C-1.2018CO2 N2 CH41614CO 2 Steady State Gradient= 0.0636 Torr/secPressure (Torra)1210864CH 4 Steady State Gradient= 0.0070 Torr/sec20N 2 Steady State Gradient= 0.0051 Torr/sec0 100 200 300 400 500 600 700 800Time (secs)Figure C-1 : Permeate pressure rise <strong>for</strong> pure gases CO 2 , N 2 and CH 4 through MembraneNPC550B. The respective permeances were calculated from <strong>the</strong> steady state gradients.A summary <strong>of</strong> <strong>the</strong> calculated permeances and permselectivities based upon <strong>the</strong> steady stategradients as shown in Figure C-1 is given in Table C-4.234


Table C-4 : Summary <strong>of</strong> calculated pure gas permeances and permselectivities <strong>of</strong> Membrane NPC550B based upon <strong>the</strong> steady state gradientsfrom <strong>the</strong> permeate pressure rise curves. The system constants were permeate volume (V p ) = 525 cm 3 , <strong>temperature</strong> (T) = 306 K and membranearea = 9.62 cm 2 .GasUpstreamPressureUpstreamPressureRawSteady State Pressure GradientAdjusted (<strong>for</strong> Leak Rate)Steady State Pressure GradientPermeancePermselectivitywith respect to CO 2kPaa (Torr) (Torr/sec) (Torr/sec)mol/m 2 .s.Pax 10 -10Blank - - 0.0003 - - -CO 2 561.14 81.39 0.0636 0.0633 31.8 1.0N 2 553.60 80.29 0.0051 0.0048 2.5 13.0CH 4 528.53 76.66 0.0070 0.0067 3.6 8.9235


C.2 Mixed Gas Permeance MeasurementC.2.1Error AnalysisThe permeance <strong>of</strong> a gas (A) using a mixed gas feed was calculated using Equation 3-2 shown inSection 3.4.2. The error analysis method <strong>use</strong>d [175] also employed <strong>the</strong> relationships as shown inTable C-1, resulting in Equation C-3 <strong>for</strong> calculation <strong>of</strong> <strong>the</strong> error analysis <strong>for</strong> <strong>the</strong> CO 2 permeance.⎛ Ep⎜⎝ P⎞⎟⎠2⎛ E= ⎜⎝ QQpp⎞⎟⎠2⎛ E+ ⎜⎝yyCO2CO2⎞⎟⎠2⎛ EA+ ⎜⎝ A2⎞⎟⎠⎛ E+ ⎜⎝xxCO2CO2⎞⎟⎠2⎛ E+ ⎜⎝ pp ff⎞⎟⎠2⎛ E+ ⎜⎝ pp pp⎞⎟⎠2Equation C-3The error analysis <strong>for</strong> <strong>the</strong> CO 2 permeance from a typical mixed gas experiment (MembraneNPC550B) is shown in Table C-6. Substituting <strong>the</strong> various values <strong>of</strong> E A /A into Equation C-3gives a 5.16 % error in <strong>the</strong> CO 2 permeance. The error in <strong>the</strong> CH 4 permeance is also 5.16 %.As <strong>the</strong> mixed gas selectivity was calculated from <strong>the</strong> ratio <strong>of</strong> <strong>the</strong> mixed gas permeances, <strong>the</strong>same error analysis method as <strong>use</strong>d <strong>for</strong> <strong>the</strong> permselectivity was employed. Substituting <strong>the</strong>errors in <strong>the</strong> CO 2 and CH 4 permeances gives a 7.30% error in <strong>the</strong> mixed gas selectivity. Thus,<strong>for</strong> membrane NPC550B with an initial mixed gas selectivity <strong>of</strong> 15, <strong>the</strong> error band is +/- 1.1.This error method was checked using duplicate measurements from <strong>the</strong> same membrane. Thevariability between <strong>membranes</strong> is so large that comparing measurements between different<strong>membranes</strong> would not indicate <strong>the</strong> error in <strong>the</strong> pure gas per<strong>for</strong>mance measurements. Duplicatemeasurements <strong>of</strong> <strong>the</strong> CO 2 /CH 4 selectivity <strong>of</strong> NPC550B taken within 1 day <strong>of</strong> each o<strong>the</strong>r arepresented in Table C-5. However, <strong>the</strong> time between <strong>the</strong> two measurements can again not beignored as some exposure to hexane and a subsequent regeneration has occurred.With this in mind, <strong>the</strong> errors calculated using <strong>the</strong> addition <strong>of</strong> <strong>the</strong> errors in <strong>the</strong> instruments hasbeen <strong>use</strong>d <strong>for</strong> <strong>the</strong> error bands on <strong>the</strong> pure gas results presented in this <strong>the</strong>sis.Table C-5 : Error analysis <strong>of</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> mixed gas CO 2 /CH 4 permselectivityfrom duplicate measurements <strong>of</strong> Membrane NPC550B.Measurement Date CO 2 /CH 4 Selectivity1 15 th Jan 2008 15.42 16 th Jan 2008 13.9Error 9.74 %236


Table C-6 : Error analysis <strong>of</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> CO 2 permeance <strong>for</strong> Membrane NPC550B. The error analysis is independent <strong>of</strong> <strong>the</strong> absolutevalues <strong>of</strong> <strong>the</strong> variables as <strong>the</strong> error is a percentage <strong>of</strong> <strong>the</strong> absolute value with <strong>the</strong> exception <strong>of</strong> <strong>the</strong> values <strong>of</strong> <strong>the</strong> permeate flow and membranearea. As <strong>the</strong> permeate flow and membrane area readings are <strong>the</strong> same <strong>for</strong> <strong>the</strong> CH 4 permeance, <strong>the</strong> error analysis <strong>for</strong> <strong>the</strong> CH 4 permeance gives<strong>the</strong> same result.Variable (A) Value (A) Reading Error (E A ) Comments E A /A E A /A(%)Permeate Flow Q p ml/min 20.20 0.50 Reading fluctuation 0.0248 2.48Permeate CO 2ConcentrationMembraneAreaFeed CO 2Concentrationy CO2molfraction0.018 4% <strong>of</strong> ValueError in calibrationregression lineA m 2 9.62 x 10 -4 7.86 x 10 -7 1 mm error in diametermeasurementx CO2molfraction0.10 0.2% <strong>of</strong> ValueFeed Pressure p f kPa(a) 1301.33 1.5% <strong>of</strong> ValuePermeatePressurep p kPa(a) 101.33 1.5% <strong>of</strong> ValueManufacturerspecificationManufacturerspecificationManufacturerspecification0.0400 4.000.0008 0.080.0020 0.200.0150 1.500.0150 1.50237


C.2.2Sample <strong>of</strong> Calculated ResultsA sample <strong>of</strong> <strong>the</strong> calculated mixed gas permeances and selectivity <strong>for</strong> Membrane NPC550B calculated using Equation C-3 is presented in Table C-6.Table C-7 : Sample <strong>of</strong> <strong>the</strong> calculated mixed gas per<strong>for</strong>mance <strong>of</strong> Membrane NPC550B. The system constants were <strong>temperature</strong> = 35 °, feed gascomposition = 10:90 (CO 2 :CH 4 vol%), feed pressure (p f ) = 1200 kPag, retentate pressure = 1150 kPag, retentate flow = 100 ml/min, sweep gasflow = 20 ml/min, permeate pressure = 0 kPag and membrane area (A) = 9.62 cm 2 .Sample #Feed GasFlowPermeateFlowPermeate CompositionCH 4 / CO 2 / HeCH 4 Permeance CO 2 Permeance CO 2 /CH 4Selectivityml/min ml/min mol% mol/m 2 .Pa.s x 10 -10 mol/m 2 .Pa.s x 10 -10150108CHC1 100.2 20.2 1.09 / 1.82 / 97.09 1.46 22.5 15.4160108CHC1 100.2 20.2 1.00 / 1.52 / 97.47 1.34 18.6 13.9238


C.3 Adsorption MeasurementC.3.1Error AnalysisThere are two parts to <strong>the</strong> error in <strong>the</strong> adsorption measurement. Firstly, <strong>the</strong>re is <strong>the</strong> accuracy <strong>of</strong><strong>the</strong> balance and secondly <strong>the</strong>re is <strong>the</strong> error in <strong>the</strong> buoyed mass correction.The accuracy <strong>of</strong> <strong>the</strong> balance is ± 0.1 microgram as specified by <strong>the</strong> equipment manufacturer.The buoyed mass correction was calculated from <strong>the</strong> buoyed volume using <strong>the</strong> ideal gas lawdescribed by Equation 3-5 in Section 3.5.2.1.The error analysis method <strong>use</strong>d [175] also employed <strong>the</strong> relationships as shown in Table C-1,resulting in <strong>the</strong> error analysis <strong>for</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> buoyed mass correction (Equation C-4)as follows. It is assumed that <strong>the</strong> errors in z, R and MW are negligible.⎛⎜⎝Emm2⎞⎟⎠⎛= ⎜⎝EVV2⎞⎟⎠⎛ E+ ⎜⎝ TT2⎞⎟⎠2⎛ EP⎞+ ⎜ ⎟⎝ P ⎠Equation C-4The error analysis <strong>for</strong> <strong>the</strong> adsorbed volume <strong>of</strong> CO 2 on <strong>nanoporous</strong> <strong>carbon</strong> at an examplepressure and <strong>temperature</strong> is shown in Table C-8. Substituting <strong>the</strong> various values <strong>of</strong> E A /A intoEquation C-4 gives an 8.35 % error in <strong>the</strong> buoyed mass. With <strong>the</strong> addition <strong>of</strong> <strong>the</strong> 0.0002 % errorin <strong>the</strong> balance accuracy gives and 8.3502 % error in <strong>the</strong> adsorption measurement.239


Table C-8 : Error analysis <strong>of</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> adsorbed volume <strong>of</strong> CO 2 onto <strong>nanoporous</strong> <strong>carbon</strong> 303 Torr (40 kPa) at 35°C.Variable (A) Value (A) Reading Error (E A ) Comments E A /A E A /A(%)Measured Mass Micrograms 51366 0.1 Balance Accuracy 1.97 x 10 -6 0.0002Buoyed Volume V cm 3 20.20 0.50Maximum difference in twomethods <strong>for</strong> calculating <strong>the</strong>buoyed volume as given in0.07 7.00Section 5.3.1.1, Table 5-7.Experimental Temperature T K 308 0.85 % <strong>of</strong> Value Manufacturer specification 0.0085 0.85Experimental Pressure P Torr 303 7.85E-07 Manufacturer specification 0.0050 0.50240


C.3.2Sample <strong>of</strong> Calculated ResultsA sample <strong>of</strong> <strong>the</strong> calculated results <strong>for</strong> <strong>the</strong> adsorption <strong>of</strong> CO 2 on <strong>nanoporous</strong> <strong>carbon</strong> at 35°C is shown in Table C-9.Table C-9 : Sample <strong>of</strong> <strong>the</strong> calculated results <strong>for</strong> <strong>the</strong> adsorption <strong>of</strong> CO 2 on <strong>nanoporous</strong> <strong>carbon</strong> at 35°C.Temperature Pressure Weight Buoyed Correction Total Uptake°C Torr kPa mg mg Weight Wt.% mmol/gram35.1 -0.2 0.0 49.7 0.0 49.7 0.0 0.034.6 303.3 40.4 51.4 0.1 51.5 3.7 0.834.6 493.5 65.8 51.9 0.2 52.1 4.9 1.134.6 1002.2 133.6 52.6 0.4 53.0 6.8 1.534.2 1971.1 262.8 53.1 0.8 54.0 8.8 2.034.3 3885.8 518.1 53.2 1.7 54.9 10.7 2.434.6 5826.2 776.8 52.8 2.6 55.4 11.9 2.734.7 7258.3 967.7 52.5 3.2 55.7 12.6 2.934.6 9615.0 1281.9 51.7 4.3 56.0 13.4 3.134.6 11514.7 1535.2 50.9 5.3 56.2 14.0 3.2241


C.4 Molecular ModellingFur<strong>the</strong>r experimental details on <strong>the</strong> molecular simulation <strong>use</strong>d to model <strong>the</strong> adsorption <strong>of</strong> pureand mixed gases on <strong>nanoporous</strong> <strong>carbon</strong> are presented in this section.C.4.1Sample Input FileAn example <strong>of</strong> an input file <strong>use</strong>d by <strong>the</strong> FORTRAN compiler <strong>for</strong> <strong>the</strong> molecular modelling ispresented below in Figure C-2. Along with specifying <strong>the</strong> appropriate interaction andcoordinate files, <strong>the</strong> input file specifies <strong>the</strong> number <strong>of</strong> simulation steps (nstep), <strong>the</strong><strong>temperature</strong> (temp) and pressure (partb in kPa) <strong>of</strong> <strong>the</strong> system. In this example, <strong>the</strong> gas is amixed gas <strong>of</strong> 10 vol% CO 2 and 90 vol% CH 4 at a <strong>temperature</strong> <strong>of</strong> 300 K and system pressure<strong>of</strong> 100 kPa.topologyfile sequencefile <strong>for</strong>cefieldfile ze<strong>of</strong>orcefieldfile zeocoordfile'CO2CH4.top' 'CO2CH4.seq' 'ffCO2CH4' 'ff<strong>carbon</strong>' 'C168.xyz'outputfile'1d2'iseed-1nstep nminit nsamp nmove iprint isave n<strong>of</strong>ort n<strong>of</strong>ort220000 1 10000 3000 200 20001 .true. .true.temp numcom LXY3D LXZ3D LYZ3D LCosAng LDL_POLY LUbind300. 2 .false. .false. .false. .false. .false. .false.partb in kPa(1..numcom) <strong>for</strong> GCMC0.1d2 0.9d2 ! (CO2,CH4)linit.true.LNVT LGrand.false. .true.Lexzeo Lzgrid Lnewtab Ltesttab LTube RTube.true. .true. .false. .true. .false. 0.d0pmswotch pmswap pmcb pmregr pmtra pmrot0.05 0.5 0.1 0.1 0.1 0.1stuk Imv Imvtot Iratp Igyra Isample0 200 100000 500 200 1lisotrop iratio iratv.false. 500 500boxlx1 boxly1 boxlz150. 50. 50.boxlx2 boxly2 boxlz250. 50. 50.rmtrax rmtray rmtraz rmzeo0.8d0 0.8d0 0.8d0 0.05d0rmrotx rmroty rmrotz rmvol0.05d0 0.05d0 0.05d0 0.1d0tatra tarot tazeo tavol0.5d0 0.5d0 0.5d0 0.5d0ntrmove rsmallmc150 0.4d0no <strong>of</strong> molecules10 10Figure C-2 : Input file <strong>use</strong>d <strong>for</strong> <strong>the</strong> molecular simulation <strong>of</strong> <strong>the</strong> adsorption <strong>of</strong> a CO 2 :CH 4mixture (10:90 vol%) at a system <strong>temperature</strong> <strong>of</strong> 300 K and pressure <strong>of</strong> 100 kPa.242


C.4.2Sample Output FileAn example <strong>of</strong> an output file produced from <strong>the</strong> molecular modelling is presented below inFigure C-3. This is <strong>the</strong> output file related to <strong>the</strong> input file presented in Figure C-2 and as such,<strong>the</strong> gas is a mixed gas <strong>of</strong> 10 vol% CO 2 and 90 vol% CH 4 at a <strong>temperature</strong> <strong>of</strong> 300 K andsystem pressure <strong>of</strong> 100 kPa.RESULTS OF THE SIMULATION...RADIUS OF GYRATION COMP. 1 BOX 1 : 0.9897545E+00 +/- 0.6922732E-03END-TO-END DISTANCE COMP. 1 BOX 1 : 0.2320872E+01 +/- 0.2232917E-02RADIUS OF GYRATION COMP. 2 BOX 1 : 0.0000000E+00 +/- 0.0000000E+00END-TO-END DISTANCE COMP. 2 BOX 1 : 0.0000000E+00 +/- 0.0000000E+00AVERAGE TOTAL ENERGY BOX 1 : -0.1185288250E+07AVERAGE INTERMOLECULAR LJ ENERGY BOX 1 : -0.7076677914E+05AVERAGE LJ TAIL CORR. ENERGY BOX 1 : -0.1555018982E+04AVERAGE EXTERNAL ENERGY BOX 1 : -0.1154704794E+07AVERAGE TOTAL COULOMBIC ENERGY BOX 1 : -0.2518416855E+04AVERAGE INTRAMOLECULAR LJ ENERGY BOX 1 : 0.0000000000E+00AVERAGE BOND BENDING ENERGY BOX 1 : 0.4425675911E+05AVERAGE TORSIONAL ENERGY BOX 1 : 0.0000000000E+00AVERAGE STRETCHING ENERGY BOX 1 : 0.0000000000E+00AVERAGE VOLUME BOX 1 : 0.8288185600E+05AVERAGE NUMBER OF PARTICLES 1 BOX 1 : 0.1476132E+03 +/- 0.2168395E+01AVERAGE DENSITY OF PARTICLES 1 BOX 1 : 0.1781008E-02 +/- 0.2616248E-04AVERAGE NUMBER OF PARTICLES 2 BOX 1 : 0.2079234E+03 +/- 0.3075098E+01AVERAGE DENSITY OF PARTICLES 2 BOX 1 : 0.2508672E-02 +/- 0.3710218E-04AVERAGE LOADING OF PARTICLES IN [MMOL/GRAM ZEOLITE]VV1 COMPONENT 1VV1 COMPONENT 2AVERAGE LOADING OF PARTICLES IN [MOLECULES/UNIT CELL]VV2 COMPONENT 1VV2 COMPONENT 2************************** Ended Normally **************************Job CPU time:1.066 hours: 0.2288152E+01 +/- 0.3361227E-01: 0.3223019E+01 +/- 0.4766707E-01: 0.1845165E+02 +/- 0.2710493E+00: 0.2599043E+02 +/- 0.3843872E+00Figure C-3 : Output file produced from <strong>the</strong> molecular simulation <strong>of</strong> <strong>the</strong> adsorption <strong>of</strong> aCO 2 :CH 4 mixture (10:90 vol%) at a system <strong>temperature</strong> <strong>of</strong> 300 K and pressure <strong>of</strong> 100kPa.243


C.4.3Sample Iso<strong>the</strong>rm DataThe data <strong>high</strong>lighted in yellow in Figure C-3 in <strong>the</strong> previous section along with <strong>the</strong> resultsproduced <strong>for</strong> o<strong>the</strong>r system pressures was <strong>the</strong>n <strong>use</strong>d to produce <strong>the</strong> adsorption iso<strong>the</strong>rm, wherecomponent 1 is CO 2 and component 2 is CH 4 .The data <strong>for</strong> <strong>the</strong> 300 K iso<strong>the</strong>rm <strong>of</strong> <strong>the</strong> mixed gas adsorption <strong>of</strong> CO 2 :CH 4 (10:90 vol%) ispresented in Table C-10 below.Table C-10 : Data produced <strong>for</strong> <strong>the</strong> 300 K iso<strong>the</strong>rm from <strong>the</strong> molecular simulation <strong>of</strong> <strong>the</strong>adsorption <strong>of</strong> CO 2 :CH 4 (10:90 vol%) on <strong>nanoporous</strong> <strong>carbon</strong> represented by C 168Schwartize.System Pressure CO 2 Loading CH 4 Loading CO 2 /CH 4 Adsorptive Selectivitymmol/gram mmol/gram0.01 0.0021 0.0022 8.630.10 0.0208 0.0222 8.441.00 0.1913 0.2017 8.5410.00 1.1021 1.2149 8.16100.00 2.2882 3.2230 6.391000.00 2.2052 5.0822 3.9110000.00 2.0587 5.8848 3.15100000.00 2.4947 5.9408 3.78244


APPENDIX D – UNSUPPORTED NPC MEMBRANESThe advantage <strong>of</strong> unsupported NPC <strong>membranes</strong> over <strong>the</strong>ir supported counterparts is asignificant decrease in <strong>the</strong> membrane thickness leading to <strong>high</strong>er values permeance and<strong>the</strong>re<strong>for</strong>e flux.However, <strong>the</strong> significant disadvantage <strong>of</strong> unsupported NPC <strong>membranes</strong> is <strong>the</strong>ir intrinsic lack<strong>of</strong> mechanical strength, which makes handling and operation very difficult. The manufacture,characterisation and per<strong>for</strong>mance <strong>of</strong> unsupported NPC <strong>carbon</strong> <strong>membranes</strong> are discussed in <strong>the</strong>following sections.The manufacture and testing <strong>of</strong> <strong>the</strong> unsupported NPC <strong>membranes</strong> work was completed byAmelia Russo and Timothy Carter as <strong>the</strong> final year research project <strong>for</strong> <strong>the</strong>ir undergraduateengineering degree.D.1 Manufacture and CharacterisationUnsupported <strong>carbon</strong> <strong>membranes</strong> were manufactured from as-cast Matrimid polyimide andpre-purchased Kapton polyimide.A significant amount <strong>of</strong> time was spent understanding <strong>the</strong> transition <strong>of</strong> <strong>the</strong> Matrimidpolyimide and Kapton polyimide from polymer to <strong>carbon</strong>. This was undertaken beca<strong>use</strong>initial attempts to make <strong>carbon</strong> <strong>membranes</strong> from <strong>the</strong>se polymers at <strong>the</strong> published <strong>high</strong><strong>temperature</strong>s that were required to produce fully <strong>carbon</strong>ised <strong>membranes</strong> resulted inunworkable <strong>membranes</strong> due to brittleness. Thus, it was decided to focus on <strong>the</strong> transitionregime where a portion <strong>of</strong> polymer still exists resulting in a better mechanical strength.The same analysis <strong>of</strong> <strong>the</strong> transition region from polymer to <strong>carbon</strong> was not undertaken <strong>for</strong>PFA as <strong>the</strong>se <strong>membranes</strong> were supported and did not require any contribution <strong>of</strong> <strong>the</strong> polymerto <strong>the</strong> mechanical strength. There<strong>for</strong>e <strong>the</strong> pyrolysis <strong>temperature</strong> range <strong>use</strong>d <strong>for</strong> <strong>the</strong> supported<strong>membranes</strong> was based solely on <strong>the</strong> literature data <strong>for</strong> <strong>the</strong> optimal manufacture <strong>of</strong> <strong>carbon</strong>from <strong>the</strong> polymer PFA as discussed earlier in Section 2.3.1.5.D.1.1Matrimid PolyimideThe weight loss <strong>of</strong> Matrimid with increasing <strong>temperature</strong> is given in Figure D-1. The weightbegins to drop at 400°C until a steady state is reached around 600°C, which is believed to be<strong>the</strong> point at which <strong>the</strong> Matrimid has fully trans<strong>for</strong>med into <strong>carbon</strong> with a total weight loss <strong>of</strong>245


40 %. This result is similar with that <strong>of</strong> Barsema et al [29] who reported a transition regionwithin <strong>the</strong> same <strong>temperature</strong> range as also shown in Figure D-1.10%Matrimid Matrimid [23]0%Weight Change-10%-20%-30%-40%-50%200 300 400 500 600 700 800Temperature (°C)Figure D-1 : Weight loss <strong>of</strong> unsupported NPC <strong>membranes</strong> made from Matrimidpolyimide with increasing pyrolysis <strong>temperature</strong>. The transition from polymer to <strong>carbon</strong>is within 400 and 600°C.NPC <strong>membranes</strong> that were made from Matrimid at pyrolysis <strong>temperature</strong> beyond 450°C wereextremely brittle and cracked upon testing. This result is similar to that <strong>of</strong> Barsema et al [29]who reported that at <strong>temperature</strong>s at and above 475°C, <strong>the</strong> <strong>membranes</strong> were very brittle anddifficult to handle.Characterisation using SEM <strong>of</strong> <strong>the</strong> surface <strong>of</strong> two selected <strong>membranes</strong>, made at pyrolysis<strong>temperature</strong>s <strong>of</strong> 420 and 430°C, respectively, are shown in Figure D-2 below. These twosamples were chosen <strong>for</strong> SEM analysis as <strong>the</strong> separation per<strong>for</strong>mance <strong>of</strong> <strong>the</strong>se <strong>membranes</strong>was known (as discussed in Section D.2).Interestingly, both samples had to be coated with a thin layer <strong>of</strong> gold be<strong>for</strong>e an image couldbe seen by <strong>the</strong> SEM. This means that <strong>the</strong> surface is not conductive and thus indicates that <strong>the</strong>surface is mostly likely still polymer ra<strong>the</strong>r than <strong>carbon</strong>. However, several defects/cracks in<strong>the</strong> surface were clearly seen even at this early stage in <strong>the</strong> pyrolysis process.246


(a)(b)Figure D-2 : SEM photographs <strong>of</strong> Matrimid (a) after pyrolysis at 420°C and (b) afterpyrolysis at 430°C.D.1.2Kapton PolyimideThe weight loss <strong>of</strong> Kapton polyimide with increasing pyrolysis <strong>temperature</strong> is presented inFigure D-3 below. The weight loss follows a similar pattern to that reported in <strong>the</strong> literature<strong>for</strong> this polymer [93]. Compared with Matrimid, <strong>carbon</strong>isation begins and finishes at much<strong>high</strong>er <strong>temperature</strong>s but with <strong>the</strong> same total weight loss <strong>of</strong> ~ 40 %.10%Kapton Batch 1 - 200HN Kapton Batch 2 - 100HN Kapton 100H [74]0%Weight Change-10%-20%-30%-40%-50%200 300 400 500 600 700 800 900 1000 1100Temperature (°C)Figure D-3 : Weight loss <strong>of</strong> unsupported NPC <strong>membranes</strong> made from Kapton polyimidewith increasing <strong>temperature</strong>. The transition from to <strong>carbon</strong> is between 500 and 700°C.247


Again, NPC <strong>membranes</strong> made from Kapton were extremely brittle and cracked upon testing.In fact, <strong>the</strong> only membrane (from <strong>the</strong> many that were made at <strong>temperature</strong>s up to 800°C) thatcould be tested was one made at 450°C.Characterisation using SEM <strong>of</strong> <strong>the</strong> surface <strong>of</strong> <strong>the</strong> membrane is shown in Figure D-4. Again,this sample had to be coated with a thin layer <strong>of</strong> gold be<strong>for</strong>e an image could be seen by <strong>the</strong>SEM, which means that <strong>the</strong> surface is not conductive and thus indicates that <strong>the</strong> surface ismostly likely still polymer ra<strong>the</strong>r than <strong>carbon</strong>. This result concurs with <strong>the</strong> minimal weightloss result at 450°C <strong>for</strong> Kapton.Figure D-4 : SEM photograph <strong>of</strong> Kapton after pyrolysis at 450°C. The surface wascoated with a gold layer to enable SEM imaging suggesting <strong>the</strong> absence <strong>of</strong> <strong>carbon</strong> and<strong>the</strong> presence <strong>of</strong> polymer.D.2 Per<strong>for</strong>manceAs indicated in <strong>the</strong> previous section, testing <strong>of</strong> <strong>the</strong> NPC <strong>membranes</strong> was extremely difficultdue to <strong>the</strong> brittleness <strong>of</strong> <strong>the</strong> samples. Cracks would appear in <strong>the</strong> surface upon handling,inserting into <strong>the</strong> membrane cell <strong>for</strong> testing or upon testing itself. Despite trying differenttechniques, such as reducing <strong>the</strong> pyrolysis <strong>temperature</strong> and altering <strong>the</strong> supports <strong>use</strong>d in <strong>the</strong>furnace, <strong>the</strong> brittleness <strong>of</strong> <strong>the</strong> <strong>carbon</strong> produced could not be reduced. Eventually, <strong>the</strong> only<strong>membranes</strong> that could be tested were those that were partially pyrolysised.D.2.1Matrimid PolyimideA summary <strong>of</strong> <strong>the</strong> per<strong>for</strong>mance results <strong>for</strong> raw Matrimid and <strong>the</strong> partially pyrolysised NPC<strong>membranes</strong> made from Matrimid at 420 and 430 °C is presented in Table D-1. Note that <strong>the</strong>per<strong>for</strong>mance results are presented as permeability as <strong>the</strong> thicknesses <strong>of</strong> <strong>the</strong> <strong>membranes</strong> couldbe determined using a micrometer. Whilst <strong>the</strong>re was a significant increase in permeability248


with partial pyrolysis, <strong>the</strong>re was also a significant reduction in selectivity. Combining thisresult with <strong>the</strong> surface cracking seen by SEM it is evident that <strong>the</strong> cracking has producedKnudsen diffusion and and/or bulk flow, leading to <strong>the</strong> poor per<strong>for</strong>mance observed.As NPC <strong>membranes</strong> made from unsupported Matrimid polyimide could not be producedwithout cracking, it was not possible to compare <strong>the</strong> results with those <strong>of</strong> Barsema et al [29]who reported that <strong>the</strong> permeability increases exponentially with increasing <strong>temperature</strong> butwith a some sacrifice in permselectivity.Table D-1 : Per<strong>for</strong>mance results <strong>of</strong> raw Matrimid and partially pyrolysised NPC<strong>membranes</strong> made from Matrimid at 420 and 430°C. The results suggest significantcracking in <strong>the</strong> partially pyrolysised NPC <strong>membranes</strong>.MembranePyrolysisTemperature Permeability (Barrer) Permselectivity°C CO 2 N 2 CH 4 CO 2 /N 2 CO 2 /CH 4Raw Matrimid N/A 6.49 0.58 0.49 11.26 13.19NPC420MA 420 198.61 287.95 278.85 0.69 0.86NPC430MB 430 14.64 19.07 14.21 0.77 0.90D.2.2Kapton PolyimideA summary <strong>of</strong> <strong>the</strong> per<strong>for</strong>mance results <strong>for</strong> raw Kapton and <strong>the</strong> partially pyrolysised NPC<strong>membranes</strong> made from Kapton at 450 °C is presented in Table D-2 below. Again, <strong>the</strong>per<strong>for</strong>mance results are given as permeability as <strong>the</strong> thickness <strong>of</strong> <strong>the</strong> membrane could bedetermined using a micrometer. In this case, cracking does not appear to have occurred as <strong>the</strong>selectivity has increased with <strong>the</strong> partial pyrolysis.The published results <strong>for</strong> Kapton polyimide [16, 22, 93, 94, 176-179] do not provideper<strong>for</strong>mance results <strong>for</strong> partial pyrolysis <strong>for</strong> <strong>temperature</strong>s lower than 600°C. Su and Lua [94]report <strong>the</strong> CO 2 permeability <strong>of</strong> an NPC membrane made <strong>for</strong>m <strong>the</strong> pyrolysis <strong>of</strong> Kapton at600°C to be 1820 barrer, which is significantly <strong>high</strong>er than <strong>the</strong> result shown in Table D-2 <strong>for</strong>raw Kapton and <strong>the</strong> partially <strong>carbon</strong>ised membrane made at 450°C. Interesting, Su and Lua[94] also report a significant decrease in permeability and increase in permselectivity withincreasing <strong>the</strong> pyrolysis <strong>temperature</strong> from 600 to 800 and <strong>the</strong>n to 1000°C. The authorsattribute this decrease in permeability to <strong>the</strong> narrowing pore size distribution with increasingpyrolysis <strong>temperature</strong>.It is <strong>the</strong>re<strong>for</strong>e suggested that <strong>carbon</strong>isation <strong>of</strong> Kapton polyimide does provide an increase in<strong>the</strong> permeability and selectivity over <strong>the</strong> original polymeric membrane. The permeability <strong>the</strong>n249


decreases whilst <strong>the</strong> selectivity increases even fur<strong>the</strong>r with increasing <strong>temperature</strong> above600°C.Despite <strong>the</strong> per<strong>for</strong>mance results obtained by Su and Lua [94], <strong>the</strong> problem with brittleness andhandling still remains. To obtain a workable unsupported NPC membrane, <strong>the</strong> optimummanufacturing <strong>temperature</strong> may still be within <strong>the</strong> transition region where <strong>the</strong>re is an increasein <strong>the</strong> permeance even though <strong>the</strong> <strong>high</strong>est possible selectivity has not been obtained.Table D-2 : Per<strong>for</strong>mance results <strong>of</strong> raw Matrimid and partially pyrolysised NPC<strong>membranes</strong> made from Kapton at 420°C. The results suggest some improvement inper<strong>for</strong>mance with partial pyrolysis.MembranePyrolysisTemperature Permeability (Barrer) Permselectivity°C CO 2 N 2 CH 4 CO 2 /N 2 CO 2 /CH 4Raw Kapton N/A 1.32 0.39 0.32 3.38 4.13Kapton 420 2.51 0.61 0.46 4.11 5.46D.3 Concluding Remarks on Unsupported NPC MembranesLimited success was found in making crack-free unsupported NPC <strong>membranes</strong> from <strong>the</strong>pyrolysis <strong>of</strong> Matrimid or Kapton. This was a disappointing result as it was hoped that <strong>the</strong>permeance or permeability <strong>of</strong> <strong>the</strong> NPC <strong>membranes</strong> could be greatly reduces by <strong>the</strong> thinnermembrane layer. Whilst <strong>the</strong> decrease in weight observed with increasing pyrolysis<strong>temperature</strong> was similar to that observed in <strong>the</strong> literature <strong>for</strong> both Matrimid and Kapton, <strong>the</strong>per<strong>for</strong>mance <strong>of</strong> <strong>the</strong> <strong>membranes</strong> could not measured beca<strong>use</strong> <strong>of</strong> <strong>the</strong> brittleness <strong>of</strong> <strong>the</strong> <strong>carbon</strong>.Partial pyrolysis <strong>of</strong> Matrimid and Kapton was hoped to be a good compromise betweenobtaining a superior per<strong>for</strong>mance whilst maintaining <strong>the</strong> flexibility <strong>of</strong> <strong>the</strong> polymer. Some <strong>of</strong><strong>the</strong> Matrimid <strong>membranes</strong> could be tested but cracking had still occurred even at such a lowpyrolysis <strong>temperature</strong> and as such <strong>the</strong> per<strong>for</strong>mance was poor. However, <strong>the</strong> partiallypyrolysised Kapton membrane did show some improvement in both selectivity andpermeability.Due to <strong>the</strong> difficulties in <strong>the</strong> manufacture <strong>of</strong> unsupported NPC <strong>membranes</strong>, <strong>the</strong> per<strong>for</strong>mance<strong>of</strong> <strong>the</strong>se <strong>membranes</strong> at <strong>high</strong>er <strong>temperature</strong> or in <strong>the</strong> presence <strong>of</strong> impurities has not beenexplored fur<strong>the</strong>r. All <strong>of</strong> <strong>the</strong> results reported in <strong>the</strong> subsequent chapters were obtained usingsupported <strong>membranes</strong> manufactured from <strong>the</strong> pyrolysis <strong>of</strong> PFA without silica nanoparticles250

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!