Sei sulla pagina 1di 675

Proceedings of the Third Pacific Basin Conference on

Adsorption Science
and Technology
This page intentionally left blank
Proceedings of the Third Pacific Basin Conference on

Adsorption Science
and Technology
Kyongju, Korea May 25-29,2003

Editor

Chang-Ha lee
Yonsei University, Korea

r heWorld Scientific .
New Jersey London Singapore Hong Kong
Published by
World Scientific Publishing Co. Re. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: Suite 202, 1060 Main Street, River Edge, NJ 07661
UK oflce: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-Publication Data


A catalogue record for this book is available from the British Library.

ADSORPTION SCIENCE AND TECHNOLOGY


Proceedings of the Third Pacifc Basin Conference
Copyright 0 2003 by World Scientific Publishing Co. Re. Ltd.
All rights reserved. This book, or parts thereof; may not be reproduced in any form or by any means,
electronic or mechanical, includingphotocopying, recording or any informationstorage and retrieval
system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923,USA.In this case permission to
photocopy is not required from the publisher.

ISBN 981-238-349-2

Printed in Singapore
Preface

The Third Pacific Basin Conference on Adsorption Science and Technology was held
fiom May 25 to May 29, 2003 in Kyongju, Korea. The theme for this conference was
thinking about adsorption at a splendid, enjoyable, and sound conference.
It was the first time that an international conference on adsorption was ever held in
Korea. Since the previous conferences organized by Professor K. Kaneko (1997) and
Professor D. D. Do (2000) were very successful, I was very excited as well as very
nervous when I was asked to organize this conference for I wanted to make this one as
successful as the previous ones.
The main purpose of this conference was to encourage the development of new
adsorption science and technology as well as to reflect the growth of this area. The
conference covered a variety of adsorption-related fields from fundamentals to
applications. The conference consisted of plenary and invited sessions, oral sessions and
poster sessions. And the conference areas were as follows: Fundamentals of Adsorption
and lon Exchange, New Materials, Adsorption Characterization, Novel Processes, Energy
and Environmental Processes.
I was very happy to see many contributions from 16 countries, with more than 120
papers. The plenary lectures of Professors D. D. Do (Univ. of Queensland, Australia), K.
E. Gubbins (North Carolina State Univ., USA), K. Kaneko (Chiba Univ., Japan), M.
Morbidelli (ETH Ziirich, Switzerland), A. L. Myers (Univ. of Pennsylvania, USA), D. M.
Ruthven (Univ. of Maine, USA), R. Ryoo (KAIST, Korea), S. Sircar (Leigh Univ. USA),
M. Suzuki (United Nations Univ., Japan), and R. T. Yang (Univ. of Michigan, USA) set
the tone for the theme of the conference. Also, I would like to thank Professors G.Baron,
A. Neimark, and L. Zhou for their contribution as invited speakers. Plenary speakers
presented an in-depth overview of key research areas: Materials, Characterization,
Molecular Simulations, Equilibria, Kinetics, and Processes. Furthermore, many
contributed papers were of high standard. 1 hope that this conference was a worthwhile
and memorable one for all the participants.
I would like to thank all the participants for all the contributions to the conference. I
would like to take this opportunity to thank all the reviewers for their efforts to review
papers within a very short period of time. Thanks should also go to the members of the
organizing committee and secretary, the advisory and scientific committee and session
chairs for their input and assistance. Also, special thanks goes to Professor J. Ritter (Univ.
of South Carolina, USA) for the organization of the US side. And I would like to express
my gratitude to Yeong-Joo Park, my wife, for all of her support and I thank my graduate
students for all of their hard work to make this conference work.
The conference would not have been possible without the generous financial support
from many organizations such as the National Science Foundation, Yonsei University,
Korean Institute of Chemical Engineers, KOSEF, KRF, Daesung Sanso Co., Research
Institute of New Energy and Environmental Systems at Yonsei Univ., Yonsei Center for
Clean Technology, Chonnam National University, Chungnam National University, NRL
for Themophysical Properties, ERC for the Advanced Bioseparation Technology, NRL
for Separation Process, NRL for Environmental Materials & Process.

Professor Chang-Ha Lee


Chairman of the Third Pacific Basin Conference on Adsorption Science and Technology
Department of Chemical Engineering, Yonsei University, Korea

V
Sponsors
Yonsei University
Korean Institute of Chemical Engineers
The Korean Federation of Science and Technology Societies
Korea Science and Engineering Foundation
Korea Research Foundation
Daesung Sanso Co.
National Science Foundation (USA)

Co-Sponsors
Research Institute of New Energy and Environmental Systems
at Yonsei University
Yonsei Center for Clean Technology
Chonnam National University
Chungnam National University
ERC for the Advanced Bioseparation Technology
NRL for Themophysical Properties
NRL for Separation Process
NRL for Environmental Materials & Process

vi
Conference Chair
Chang-Ha Lee (Yonsei Univ., Korea)

Conference Advisory Committee


B. H. Ha (Hanyang Univ.)
H. Lee (Yonsei Univ.)
H. K. Rhee (Seoul National Univ.)

Organizing Committee
S. H. Cho (KIER)
D. K. Choi (KIST)
S. H. Hyun (Yonsei Univ.)
J. W. Jang (SK Eng. & Construction)
H. Kim (Seoul National Univ.)
Y. M. Koo (Inha Univ.)
C. S. Lee (Korea Univ.)
H. Moon (Chonnam Univ.)
D. S. Park (Daesung Sanso Co.)
S. K. Ryu (Chungnam Univ.)
Y. G. Shul (Yonsei Univ.)
J. E. Sohn (Dong-A Univ.)
J. Yi (Seoul National Univ.)

International Advisory Committee


A. S. T. Chiang Wational Central Univ., Taiwan)
D. D. Do (Univ. of Queensland, Australia)
K. E. Gubbins (North Carolina State Univ., USA)
K. Kaneko (Chiba Univ., Japan)
2. Li (South China Univ. of Technology, China)
M. Morbidelli (ETH Zurich, Switzerland)
A. L. Myers (Univ. of Pennsylvania, USA)
J. Ritter (Univ. of South Carolina, USA)
J. L. Riccardo (UNSL, Argentina)
D. M. Ruthven (Univ. of Maine, USA)
H. Tamon (Kyoto Univ., Japan)
R. T. Yang (Univ. of Michigan, USA)
H. Yoshida (Osaka Prefectural Univ., Japan)
L. Zhou (Tianjin Univ., China)

Scientific Advisory committee


T. Bandosz (The City College of New York, USA)
G . Baron (Vrije Universiteit Brussel, Belgium)
M. Bulow (The BOC Group, Inc., USA)
G. Carta (Univ. of Virginia, USA)
K. Chihara (Meiji Univ., Japan)
C. B. Ching (The National Univ. of Singapore, Singapore)
J. Izumi (Mitsubishi Heavy Chem. lnd., Ltd., Japan)
M. Jaroniec (Kent State Univ., USA)

Vii
J. U. Keller (Univ. of Siegen, Germany)
J. M. Lee (KRICT, Korea)
K. H. Lee (Pohang Univ., Korea)
N. Lemcoff (The BOC Group, Inc., USA)
M. D. LeVan (Vanderbilt Univ., USA)
M. Mazzotti (ETH ZUrich, Switzerland)
G. McKay (Hong Kong Univ. of Sci. and Tech., Hong Kong)
F. Meunier (Laboratoire du Froid CNAM, France)
J. K. Moon (KAERI, Korea)
A. Neimark (TRI/Princeton, USA)
Y. D. Park (FUST, Korea)
A. E. Rodrigues (Univ. of d Porto, Portugal)
W. Rudzinski (Maria Curie-Sklodowsk Univ., Poland)
R. Ryoo (KAIST, Korea)
A. Sakoda (Univ. of Tokyo, Japan)
N. Seaton (Univ, of Edinburgh, UK)
S. Sircar (USA)
W. A. Steele (Pennsylvania State Univ., USA)
M. Suzuki (United Nations Univ., Japan)
0. Talu (Cleveland State Univ., USA)
Y. Teraoka (Kyushu Univ., Japan)
Y. Xie (Peking Univ., China)

Organizing Secretary
C. H. Cho (Youngdong Univ.)
S. S. Han (KIER)
K. T. Lee (Yonsei Univ.)
J. Y. Yang (SK Eng. & Construction)

...
Vlll
Contents

Preface V

Plenary Papers
Adsorption Equilibria of Sub-critical and Super-critical Fluids in
Carbonaceous Materials
D. D. Do and H. D. Do 1
FreezingMelting in Porous Carbons
E R. Hung, R. Radhakrishnan, E Beguin, M. Sliwinska-Bartkowiak
and K. E. Gubbins 9
Measurement of Diffusion in Microporous Solids
D. M. Ruthven 17
Ordered Mesoporous Carbons with New Opportunities for Adsorption
Studies
R. Ryoo and S. H. Joo 27
Quantum Micropore Filling and its Application Possibility
T Tanaka, I: Hattori, K. Murata, T. Kodaira, M. Yudasaka,
S. I@ma and K. Kaneko 35
Adsorption in Microporous Materials: Analytical Equations for TYPE I
Isotherms at High Pressure
A. L. Myers 44
New Sorbents for Desulfurization of Transportation Fuels
R. T. Yang, A. Hernandez-Maldonado, A. Takahashi and E H. Yang 51
Optimization of Continuous Chromatography Separations
2. Y. Zhang, M. Mauottiand M. Morbidelli 64
Adsorption Technology for Gas Separation
S. Sircar 72
Carbon Composite Membranes
M. Suzuki, A. Sakoda, S.-D. Bae, T Nomura and Y.-Y. Li 79
Invited Papers
On the Dominant Role of Adsorption Effects in Heterogeneous
Catalysis
J. E Denayer, G. V Baron, D. Devos, J. A. Martens
and R A. Jacobs 87

iX
Supercritical Adsorption: Paradox, Problems, and Insights
L. Zhou 91
Contributed Papers
Microwave Drying for Preparation of Mesoporous Carbon
H. Tamon, T. Yamamoto, T. Suzuki and S. R. Mukai 99
Computer Simulation of Transport in Cylindrical Mesopores
S. K. Bhatia and D.Nicholson 104
Multicomponent Mass Transfer Diffusion Model for the
Adsorption of Acid Dyes on Activated Carbon
K. K. H. Choy, J. E Porter and G. McKay 109
Sorption Thermodynamics of Nitrous OxideLSX Zeolite Systems
M. Biilow, D. Shen and S. R. Jale 114
Activated Carbon Membrane with Carbon Whisker
S.-D. Bae and A. Sakoda 121
Mesoporous Silica with Local MFI Structure
S. P. Naik, A. S. T. Chiang, R. W Thompson, E C. Huang
and H.-M.Kao 126
Infinite Dilution Selectivity Measurements by Gas Chromatography
S. Gumma and 0. Talu 131
Adsorption Properties of Colloid-Imprinted Carbons
M. Jaroniec and 2.-J. Li 136
On the Role of Water in the Process of Methyl Mercaptan
Adsorption on Activated Carbons
S. Bashkova, A. Bagreev and T. J. Bandosz 141
Studies on the Adsorption Properties of Ion-Exchanged Low Silica
X Zeolite
H. Jiang, W Tang, J. P. Zhang, B. I! Zhao and I! C. Xie 147
Carbonization of Organic Wastes Using Super-Heated Water Vapor
and Their Adsorption Properties
H. Yoshida, N. Miyagami and M. Terashima 152
Further Successful Applications of the New Theoretical
Description of Adsorptioflesorption Kinetics Based on the
Statistical Rate Theory
#? Rudzinski and T. Panczyk 157
Characterization and Ethylene Adsorption Properties of
Silver-Loaded FER Zeolite Potentially Used as Trap Material
of Cold-Start Hydrocarbon Emission from Vehicles
!I Teraoka, H. Onoue, H. Furukawa, I. Moriguchi, H. Ogawa
and M. Nakuno 162

X
Pressure-Dependent Models for Adsorption Kinetics on a CMS
Z-S.Bae, I!-K. Ryu and C.-H. Lee 167
Preparative Enantioseparation of Fluoxetine by Simulated
Moving Bed
H.-W Yu and C. B. Ching 172
Optimization Based Adaptive Control of Simulated Moving Beds
G. Erdem, S. Abel, M. Mauotti, M. Morari and M. Morbidelli 177
Mono-Methyl Paraffin Adsorptive Separation Process
S. Kulprathipanja, J. Rekoske, M. Gutter and S. Sohn 182
Chromium (VI) and (111) Species Adsorption from Aqueous Solutions
by Activated Carbon Fibers
0. Astachkina, A. Lyssenko and 0. Muhina 189
Treatment of Complex Wastewaters by Biosorption and Activated
Carbon: Batch Studies
C. Gerente, 2. Reddad, Z: Andres, C. Faur-Brasquet
and I? le Cloirec 194
Adsorption Characteristics of Protein-Based Ligand for Heavy Metals
M. Terashima, N. O h ,T. Sei, K. Shibata and H. Yoshida 199
Preparative Chromatography at Supercritical Conditions
A. Rajendran, M. Mauotti and M. Morbidelli 204
Adsorptive Separation of Oligosaccharides: Influence of
Crosslinking of Cation Exchange Resins
J. A. Vente, H. Bosch, A. B. de Haan and P. J. T. Bussmann 209
Identification and Predictive Control of a Simulated Moving
Bed Process
I.-H. Song, H.-K. Rhee and M. Mazzotti 214
Quick and Compact Ozonation Using Siliceous Zeolite
H. Fujita, T. Fujii, A. Sakoda and J. Izumi 219
Time Resolved Multicomponent Sorption of Linear and Branched
Alkane Isomers on Zeolites, Using NIR Spectroscopy
A. F: P. Ferreira, M. Mittelmeijer, M. Schenk, A. Bliek
and B. Smit 224
Pore Size Effects in the Liquid Phase Adsorption of Alkanes
in Zeolites
J. F: M. Denayer, K. de Meyer, J. A. Martens and G. K Baron 229
Detection of Freezing Point Elevation in Slit Nanospace by
Atomic Force Microscopy
M. Miyahara, M. Sakamoto, H. Kanda and K. Higashitani 234

xi
Modeling of High-pressure Equilibrium Adsorption of Supercritical
Gases on Activated Carbons. Determination of Pore Size
Distribution Using a Combined DFT and EOS
E. A. Ustinov and D. D. Do 239
On the Peculiarity of the Minimum of N-Hexane Permeability in
Activated Carbon
J.-S. Bae and D. D. Do 244
Simplified Experimental Method to Analyse Intra-Activated
Carbon Particle Diffusion Based on Parallel Diffusion Model
I: Miura, I: Otake, H. T. Chang, N. Khalili, S. Iwasawa
and E. G. Furuya 249
In-Situ Characterization of Ion Adsorption at Biomimetic
Airwater Interfaces
T. E Kim, G. S. Lee and D. J. Ahn 254
Single and Multi Component Adsorption of Volatile Organic
Compounds onto High Silica Zeolites - Discussion of
Adsorbed Solution Theory
I? Monneyron, M. -H. Manero and J. -N. Foussard 259
Influence of VOCs Molecular Characteristics on Exothermicity
of Adsorption onto Activated Carbon
l? Pre, C. Faur-Brasquet and I? le Cloirec 264
The Influence of Ar and He on the Rate of Adsorption and on the
Adsorption Equilibrium of Alkanes in Zeolites
M. C,Mittelmeijer-Hazeleger,A. E l? Ferreira and A. Bliek 270
Modeling the Discharge Behavior of Metal Hydride Hydrogen
Storage Systems
S. A. Gadre, A. D. Ebner, S. A. Al-Muhtaseb and J. A. Ritter 276
The Advanced Modeling Technology for Periodic Adsorption
Process: Direct Determination of Cyclic Steady State
J.-H. Yun, A. C. Stawarz and E 0. Jegede 28 1
Adsorption and Desorption Characteristics of Zeolite Impregnated
Ceramic Honeycomb for VOC Abatement
H.-S. Kim, l!-J. Yoo, E-S.Ahn, M.-K. Park, K.-T. Chue
and M.-H. Han 286
Reverse Flow Adsorption Technology for the Recycling of
Homogeneous Catalysts: Selection of Suitable Adsorbents
J. Dunnewijk, H. Bosch and A. B. de Haan 29 1
Molecular Simulation of Gas Separation by Adsorption Processes
J. I? B. Mota 296

xii
Metal-Doped Sodium Aluminium Hydride as a Reversible
Hydrogen Storage Material
J. Wang, A. D. Ebner, K. R. Edison, J. A. Ritter and R. Zidan 30 1
Synthesis and Dehumidification Behaviors of Monodisperse Spherical
Silica Gels with Different Pore and Chemical Structures
C,H. Cho, Y. J. Yoo, J. S. Kim, H. S. Kim, Y. S. Ahn
and M. H. Han 306
Production of Hard Carbons for Lithium Ion Storage by the
Co-Carbonization of Phenolic Resin Precursors
S. R. Mukai, T. Tanigawa, T Harada, T. Masuda and H. Tamon 313
Novel Bioactivite Carbomineral Sorbents, Including Cluster
and Carbon Nanotubes for Superselective Purification of Biodiesel
Fuel - Liquid Hydrocarbons and Carbonhydrate from
Sulfur Containing Impurities
D. I. Shvets 318
Titanosilicate ETS-10: Synthesis, Characterization and Adsorption for
Heavy Metal Ions
G. X. S. Zhao, J. L. Lee and f? A. Chia 324
Ordered Macroporous Materials Structurally Templated by Colloidal
Microspheres
Z. Zhou, ?C.
-I ? and G. X. S. Zhao
Ong 329
Adsorption of Nitrogen, Oxygen and Argon in Transition and Rare
Earth Ion Exchanged Zeolites A and X
R. K Jasra, J. Sebastian and C. D. Chudasama 334
Adsorption of Methylene Blue from Water onto Activated Carbon
Prepared from Coir Pith, an Agricultural Solid Waste
C. Namasivayam and D. Kavitha 339
Separation of Oxygen-Argon Mixture by Pressure Swing Adsorption
X . Jin and S. Farooq 344
Dual Reflux Pressure Swing Adsorption Cycle for Gas Separation
and Purification
A. D. Ebner and J. A. Ritter 349
Simulation of a Coupled MembranePSA Process for Gas Separation
I. A. A. C. Esteves and J. I! B. Mota 354
I3CO and l2C0 Separation on Na-LSX using Pressure-Swing Adsorption
at Low Temperatures
J. Izumi, N. Fukuda, N. Tomonaga, H. Tsutaya, A. Yasutake,
A. Kinugasa and H. Saiki 359

xiii
High Purity Oxygen Generation PSA Process by Using Carbon
Molecular Sieve
J.-G. Jee, T-H. Kwon and C.-H. Lee 365
A Study on the Preparation of Deodorizing Fibers by Coating TiOz
S. W Oh, H. J. Kim and S. M. Park 370
Composite Adsorbents for the Removal of Cs and Sr Ions in
Acidic Solutions
J. K. Moon, C. H. Jung, S. H. Lee, E. H. Lee, H. T. Kim
and I! G. Shul 375
Dehumidification Behavior of Metal(Ti, Al, Mg) Silicates
Impregnated Ceramic Fiber Sheets
I! S. Ahn, C. H. Cho, I! J. Yoo, J. S. Kim, H. S. Kim
and M. H. Han 38 1
Synthesis of Zirconia Colloids from Aqueous Salt Solutions
and Their Applications
K. Lee, P. W Carr and A. V McCormick 387
Comparison of Nano-Sized Amphiphilic Polyurethane (APU)
Particles with SDS, an Anionic Surfactant for the Soil Sorption
and the Extraction of Phenanthrene from Soil
I . 3 . Ahn, H.-S. Choi and J.-Z Kim 392
Synthesis of Mesoporous Activated Carbon with Iron Ion-Aided
Activation
Z Seida, K. Watanabe and Y: Nakano 398
Separation of Peptides from Human Blood by RP-HPLC
S.-K. Lee, I! Polyakova and K.-H. Row 403
Separation of Acanthoside-D in Acanthopanax Senticosus by
Preparative Recycle Chromatography
S . 2 Hong, D.-X. Wang and K.-H. Row 408
Use of Various Forms of &aft Lignin for Toxic Metal Uptake
D. R. Crist, R. H. Crist and J. R. Martin 413
Removal of Uranium Ions in Sludge Waste by Electrosorption
Process
C.-H. Jung, J.-K. Moon, S.-H. Lee, Z-G.Shul and W-Z. Oh 417
Ion Exchange Characteristics of Palladium from a Simulated
Radioactive Liquid Waste
S.-H.Lee, C.-H. Jung, J.-K. Moon, J. H. Kim and H. Chung 422
Application of Characterization Procedure for Complex Mixture
Adsorption in Water and Wastewater Treatment
S.-H.Kim, T.-W Kim, D.-L. Cho, D.-H.Lee and H. Moon 427

xiv
Surface Characteristics of MCM-41 on Cr(1II) and Cr(V1)
Adsorption Behaviors
S. J. Park, B. R. Jun and M. Han 432
Influence of Anodic Oxidation of Activated Carbon Fibers
on the Removal of Heavy Metal in Aqueous Solution
S. J. Park, !I M. Kim and J. R. Lee 437
Kinetics and Diffusion Processes for Reactive Dye Adsorption
by Dolomite
S. J. Allen, G. M. Walker, L. Hansen and J.-A. Hanna 442
Permeate Flux Behavior During Microfiltration of Protein-Adsorbed
Microspheres in Stirred Cell
I: Chang, S.-W Choi, T-G. Lee, S. Haam and W-S. Kim 447
Surface Fractional Dimensions of the Adsorbents from Industrial
Sludge
J. H. You, H. M. Wu and 2. X. Fang 452
Adsorption of Acidic Peptide on Crosslinked Chitosan Fiber:
Equilibria
N. Kishimoto and H. Yoshida 458
Removal of Salt and Organic Acids from Solution used to Season
Salted Japanese Apricots (Ume) by Combining Electrodialysis
and Adsorption
W Takatsujiand H. Yoshida 463
Studies on the One-Column Analogue of a Four-Zone SMB
Y. S. Kim, C. H. Lee, Y. M. Koo and I? C. Wankat 468
Advanced' Flue Gas Treatment by Novel de-SOX Technology over
Active Carbon Fibers
M.-A. Yoshikawa, A. Yasutake and I. Mochida 474
Adsorption of Natural Gas Components on Activated Carbon for
Gas Storage Applications
I. A. A. C. Esteves, M. S. S. Lopes, I? M. C. Nunes,
M. E J. Eusibio, A. Paiva and J. I? B. Mota 479
Prediction of Breakthrough Curves for Toluene and
Trichloroethylene onto Activated Carbon Fiber
J.-M? Park, S.-S. Lee, X-W Lee and D.-K. Choi 484
Catalytic Reduction Mechanism of Nitric Oxide over
ACFslCopper Catalyst
S. J. Park, B. J. Kim and X S. Jang 489

xv
NO Removal of Activated Carbon Fibers Treated by Cu
Electroplating
S. J. Park, J. S. Shin and J. R. Lee 494
The Appliance Study of 02-PSA in the Oxygen Activated
Sludge Process
S. H. Lee, P. S. Yong, H. M. Moon and D. S. Park 499
Application of Solid Adsorbent for VOC Monitoring Sensor
0. J. Joung and E H. Kim 504
PSA for Solvent Recovery with USY-Type Zeolite; an
Experimental and a Simulation Study
K. Chihara, T. Kaneko, T. Aikou and S. O h 509
Azeotropic Adsorption of Organic Solvent Vapor Mixture on
High Silica Zeolite, Experimental & Simulation
K. Chihara, K. Hijikata, H. Yamaguchi, H. Suzuki
and E Takeuchi 514
VOC Enrichment by a VSA Process with Carbon Beds
J. Yang, M. Park, J.-W Chang and C.-H. Lee 519
Isobutane Purification by Pressure Swing Adsorption
S.4. Han, J.-H. Park, J.-N. Kim and S.-H. Cho 524
Characteristics of Gas Separation by Using Organic Ternplating
Silica Membrane
J. Moon, S. Hyun and C.-H. Lee 529
Adsorber Dynamics of Binary and Ternary Hydrogen Mixture
in Activated Carbon and Zeolite 5A Beds
M.-B. Kim, J.-S. Kim, C.-H. Cho and C.-H. Lee 534
Temperature Programmed Adsorption ( P A ) of Various
Hydrocarbons on Adsorbers of Honeycomb Type
D. J. Kim, J. E. fie, E S. Oh and J. M. Kim 539
Non-Isothermal Dynamic Adsorption and Reaction in
Hydrocarbon Adsorber System
D. J. Kim, W. G. Shim, J. E. Yie and H. Moon 544
Sorption of U(V1) onto Granite: Kinetics and Reversibility
M. H. Baik and P. S. Hahn 549
Adsorption of Uranium (VI) on Kaolinite: Speciation
and Mechanism
M. J. Kang, B. E. Han and P. S. Hahn 554
High-Temperature Adsorption of Hazardous Metal Chlorides
Using Activated Kaolinite
H. C. Yang, J. S. Yun, E J. Cho and J. H. Kim 559

xvi
Probing the Cut-Off for Intracrystalline Adsorption on Zeolites:
Pore Mouth Adsorption
R. Ocakoglu, J. E M. Denayer, J. A. Martens, G. B. Marin
and G. I! Baron 564
The Low-Temperature Sorption Behaviour of Cryosorbent Materials
C. Day and I! Hauer 569
Thermal and Surface Mechanism Studies on
Adsorption-Temperature Programmed Desorption of Nitrogen
Oxides over Chemically Activated Carbon Fiber
H.-J. Kim, X-W Lee, E. Lee and D.-K. Choi 574
Surface Properties of Activated Carbons Containing Basic
Hydroxide Ions and NOx Adsorption-Desorption Process
X W Lee, H. J. Kim, D. K. Choi, J. W Park C. H. Lee
and B. K. Na 579
Adsorption Characteristics of Nitrogen Compounds on Silica Surface
H. J. Kim, C.-H. Lee, X G. Shul and W S. Min 584
Adsorption Characteristics of VOCs on Mesoporous Sorbents
W G. Shim, M. S. Yang, J. W Lee, S. H. Suh and H. Moon 589
Molecular Simulation for Adsorption of Halocarbons in Zeolites
K. Chihara, T. Sasaki, S. Miyamoto, M. Watanabe,
C. E Mellot-Draznieks and A. K. Cheetham 595
Adsorption of BTX on MSC in Supercritical C02,a Chromatographic
Study
K. Chihara, N. Omi, E Znoue, T Yoshida and T Kaneko 600
Porous Alumina with Bimodal Pore Size Distribution as an
Organic Adsorbent
I: Kim, C.Kim, P Kim, .IC. Park and J. yi
. 605
Storage and Selectivity of Methane and Ethane into
Single-Walled Carbon Nanotubes
X-G. Seo, B. H. Kim and N. A. Seaton 610
Hydrotalcites for Carbon Dioxide Adsorbents at High Temperature
J. I. Yang, M. H. Jung, S.-H. Cho and J.-N. Kim 615
Effect of Polarity of Polymeric Adsorbents on Desorption of
VOCs under Microwave Field
X. Li, 2. Li, H. X. Xi and H. Wang 620
Mixed-Gas Adsorption on Heterogeneous Substrates in the Presence
of Lateral AD-AD Interactions
A. J. Ramirez-Pastor, E M. Bulnes and J. L. Riccardo 625

xvii
Adsorption on Correlated Disordered Substrates
R. H. Lopez, E M. Bulnes, E Rojas, J. L. Riccardo
and G. Zgrablich 630
Temperature Effects on the Scaling Properties of Adsorption
on Bivariate Heterogeneous Surfaces
E Rom*, E Bulnes, A. J. Ramirez-Pastor and G. Zgrablich 635
Adsorption of Polyatomic Species: An Approach from
Quantum Fractional Statistics
J. L. Riccardo, A. J. Ramirez-Pastor and E Rom 640
Multilayer Adsorption with Multisite-Occupancy
E Rom, A. J. Ramirez-Pastor and J. L.. Riccardo 645

xviii
Plenary Papers
This page intentionally left blank
ADSORPTION EQUILIBRIA OF SUB-CRITICAL AND SUPER-CRITICAL
FLUIDS IN CARBONACEOUS MATERIALS

D.D.DO AND H. D.DO


Department of Chemical Engineering, University of Queensland, St. Lucia, Qld 4072, Australia
E-mail: duongd@cheque.uq.edu.au

In this paper, we present an overview of a number of techniques used to characterize the adsorption
equilibria of sub and super-critical fluids in nonporous carbon black and porous activated carbon.
Tools such as the grand canonical Monte Carlo simulation (GCMC), Density Functional Theory
(DFT), Molecular Layer Structure Theory (MLST) of Ustinov and Do, and enhanced molecular
layering and pore filling proposed by Do and his co-workers will be discussed. Although the GCMC
provides the brute force calculation of molecular interactions and its results have been used as a
benchmark for other techniques to compare with, its extremely time consuming computation makes
the other techniques more appealing to engineers and experimentalists. Despite the ever-increasing
computing power of today personal computer, the advantage of simpler methods is warranted for
their role in solving practical problems. Furthermore, added to this advantage is the good
perfarmance in terms of the prediction power of the simpler methods. AN these tools will be
discussed in this paper regarding their applications to adsorption of super and sub-critical fluids in
carbonaceous materials, such as graphitid thermal carbon black and activated carbon.

1 Introduction

Adsorption equilibria and kinetics are important for the proper design of adsorption
processes. The equilibria information of adsorption isotherm. is clearly the first hand
information that one needs to approximately size the adsorber. Since the adsorption
affinity can vary by many orders of magnitude. Ranging from very low for weakly
adsorbing gases to very high for strongly-adsorbing hydrocarbon vapours, it is very
important that we know the value of this adsorption affinity. Experimentally this
information can be obtained from careful experimentation of adsorption isotherm
measured from very low pressure (where adsorption affinity can be calculated) to very
high pressure where saturation capacity can be determined. Alternatively, the adsorption
affinity can be determined from some appropriate theories or computer simulation. With
the advances of high speed computer and the development of modem tools to deal with
inhomogeneous fluids in confined space such as micropores, the second approach is
gaining ground and new theories are constantly developed allowing engineers and
scientists to calculate adsorption isotherm from minimum amount of information.
Nevertheless, that does not mean to say that we can make do without experimental data.
We still have some ground to cover before that stage can be reached. At the present time,
careful and reliable experimental data are still required for validation of theories and even
confirmation of molecular simulations.
In this paper we will discuss some modem tools for studying adsorption equilibria of
super and sub-critical fluids on non-porous surface and in porous carbons. In particular,
the tools of grand canonical Monte Carlo (GCMC) simulation [l, 21, Density Functional
Theory [3], Molecular Layer Structyre Theory (MLST) [4, 51, and the enhanced
molecular layering and pore filling of Do, first developed in 1998 [6]and later applied in
a number of practical systems [7-101, will be discussed. Their applications to
experimental systems are illustrated to highlight the advantages and disadvantages of
these modem tools of equilibria characterization.

1
2 Tools of characterisation

2. I Grand canonical Monte Carlo simulation


In the molecular simulation of adsorption in confined space such as pores of adsorbent,
the most widely used and successful ensemble is the grand canonical Monte Carlo
simulation. In this ensemble, we specify the chemical potential of the fluid,p, that the
candidate pore is immersed in, the size of the pore, and the temperature.
In the GCMC, the simulation can be carried out in the same procedure suggested by
Metropolis at al. [ 111, and the density distribution (hence average pore density) can be
obtained as a direct result of the simulation. The GCMC simulation is usually carried out
by starting with a very low value of chemical potential. Once the simulation
corresponding to this chemical potential has been completed, the chemical potential is
increased incrementally and the particle configuration of the last run is used as the initial
configuration and the new simulation is then carried out. This process is repeated until
the final chemical potential is performed. In the simulation, the simulation box is
constraint by the width of the pore and the L, and L, in the x and y directions. The box
lengths L, and L, are chosen large enough, and usually chosen as twice the cut-off
distance. In our work we choose r, = 5ofi and periodic boundary conditions are applied
in the x and y directions. The procedure of the GCMC involves three basic moves:
1. DisDlacement of Darticle: A particle is selected in random and is given a new
conformation (translation and rotation). The move is accepted with a probability
p = min[l, exp(-AU / kT)]
where AU is the difference between the new and old configuration energies. In this
move the displacement step and the rotation angle are chosen in such a manner that
the acceptance rate is between 25 and 50%.
2. Insertion of Darticle: The second move is the particle insertion move. A particle is
generated at a random position. Its acceptance must satisfy the probability:
- -
p =min 1,
1 A~(N+~)
exp& - U(N + 1)+ U(N)]/ kT)

where V is the volume of the simulation box and A is the thermal de Broglie
wavelength. Usually in the simulation we supply the activity, z = A- exp(p/ kT) ,
instead of the chemical potential. For bulk gas phase which behaves as an ideal gas,
this activity is the density of the gas phase.
3. Removal of Darticle: The third move in the GCMC simulation is the removal of a
particle. A particle is chosen in random and removed. The probability of acceptance
this removal is

The GCMC is successfully applied to many adsorption systems. For temperatures


greater than the pore critical temperature, the adsorption isotherms exhibit a smooth
behaviour. However, for temperatures less than the pore critical temperature, there is
a possibility of transition. When there is transition, the equilibrium point may be
obtained by applying the thermodynamic integration method [ 121.

2
2.2 Density Functional Theory (DFT)
The DFT method was popularized in the sixty and was increasingly modified and applied
to many problems involving inhomogeneousfluids in the 80s. Among the many versions
of DFT, the one proposed by Tarazona and co-workers [3] remains the popular one in
solving adsorption of confined fluid in pores. Their method is the non-local DFT and is
applied on a grand canonical ensemble. The starting point of the method is the grand
potential of the system
= F(p(l)) + jP@)ucxtcr)dx - jP@>P& (la)
where p(rJ is the singlet particle density. The first term on the RHS of the above equation
is the Helmholtz free energy of the system. The Helmholtz free energy is expressed as a
sum of two terms. The first term is that obtained from a reference system (hard sphere
fluid is chosen as one) and the other is the perturbed component (which is due to the
attractive component of the intermolecular fluid-fluid interaction), that is
F(PCr)) = FH(P@N+ FA-( P O
The Helmholtz free energy of the reference hard sphere system, in turn,can be expressed
as two terms. One is due to the ideal gas contribution (accounting for the momenta of all
particles) while the other term is the repulsive interaction among the hard spheres. It is
FH(P(E)) = jP@
[In(A3P(IN - 11dr + j P ( 0 f m G(INdI
where the smoothed density is given by
-
P(I) = I P W w(x-il;p(l))dzI
Assuming a mean field approximation, the perturbed component of the Helmholtz free
energy due to the attractive force is
1
FA-(P(I)) =?j j P 0 P W %(I-d)dd dx (1b)
The equilibrium density profile can then be obtained by finding minimum of the grand
potential as defined in eq.(la). Tarazona used the Carnahan-Starling equation to derive
the excess function f(p(I)) as (here, d is the hard sphere diameter)
f(&)) = (4y - 3 ~ )/(I - y) ; y = (nd 16) &)

2.3 Molecular Layer Structure Layer Theoty (MLST)


The Molecular Layer Structure Theory (MLST) was first developed [4, 51 to study the
vapour liquid equilibrium and surface tension of many substances over a wide range of
temperature. In this method, the fluid is considered as parallel molecular layers, whose
surface densities are different for inhomogeneous
- fluids. Similar to the DFT method, we
define the following grand potential:
R= cp;k(p;)+cpj
i
+urf - jl]

where py is the surface density, Ask)is the intrinsic Helmholtz free energy of the layer
j, pj is the interaction energy between one molecule on the layerj and all surrounding
layers, and u; is the external potential exerted on the layerj. In this method of MLST,
the intrinsic Helmholtz free energy is calculated from the equation of state of
homogeneous fluid, and pi is calculated from the integration of 12-6 Lennard-Jones

3
potential energy. The equilibrium density is calculated by minimizing the grand potential
with respect to density as well as distances between layers [4,5].

2.4 The Do-Method


Recently, Do and co-workers [6-101 have proposed a very simple method but it does
reveal the mechanistic pictures of what are occurring in pores of different size. The
process of adsorption in pore is viewed as follows. Molecules in pore are constantly
in motion but statistically there is a spatial distribution of these molecules due to the
interactive forces between them and the surface atoms. We treat this spatial distribution
as a step function; uniformly high density near the surface and uniformly low density in
the inner core of the pore. Due to the long range interaction of the surface, the pressure
of the fluid in the inner core is not the same as that in the bulk phase. Assuming a
Boltzmann distribution, the pressure of the inner core is related to the bulk fluid as
Pp = P.exp(- a E/ kT) (2)
where E is the average potential of the inner core, which is a function of pore width and
the layer thickness. The adsorption process is basically treated as a molecular layering
process, and it can be described by a layering equation, for example an equation taking
the form of the BET equation. The affinity parameter C, of this equation the same
as the C-parameter for a flat surface. Rather they are related through the following
relation
c,(H) = c.~xP[(Q,(H)-Q)/RT~ (3)
where Q is the heat of adsorption of the flat surface and Qp is the correspondingvalue for
a pore. These values may be taken as the minimum value of the potential energy between
a molecule and a flat surface and that between a molecule and a pore. The layering
equation can be written as
t / t, = f(P, (HI, c, (HI) (4)
where ,t is the thickness of a monolayer. This layering process is followed by a pore
filling process. Here the term pore filling is used in its most general sense, that is pore
is filled with molecules by either two-dimensional condensation in small pores or three
dimensional condensation in large pores. We argue that this general pore filling process
is governed by the equation

H / 2 - t - ~ , =)*
where ~0 is the position at which the solid-fluid potential energy is zero. Although this
form is similar to that of the Kelvin equation, the significant difference rests on the use of
the pore pressure P, in the above equation. If we substitute the pore pressure of eq.(2)
into the above equation, we get
P M
(H/2 - t - 2,)
For small pores, E(H) dominates the RHS of the above equation and hence the pore
filling pressure in small pores is dominated by the strength of the potential field created
by the overlapping of the fields of the two walls. On the other hand, for larger pores
where &H) = 0 , the RHS is dominated by surface tension and this equation reduces to
the well known modified Kelvin equation.

4
Using eqs. (2) to (5) we can readily
obtain the pressure at which'a pore is
completely filled with adsorbate
molecules. This pressure is called the
pore filling pressure. Figure 1 shows
the plot of the reduced pore filling
pressure (P/Po) versus pore width for
nitrogen adsorption in slit pores at 77.3
K and for argon at 87.3 K. The results
of DFT and GCMC are also shown as
symbols, and it is seen that the
0 10 20 30 40 50 agreement between the DFT., GCMC
Pore Width (A) and the Do method is very good, even
Figure 1: Pore tilling p r e ~ s ~versus
n pore width the minimum position of this curve.
for argon at 87.3 K and nitrogen at 77.3 K

3 Results and Discussion

Having presented briefly the working procedures of the various methods, we now would
like to illustrate their applications to adsorption of super and sub-critical fluids on non-
porous carbon surface and in porous carbonaceous solids having slit pores. But first it is
worthwhile to compare the time scales of computation of these methods:
Do-method < MLST <4 DFT < GCMC
(seconds) (minutes) (hours) (day)

3.1.1 Non-porous surfaces


We show in Figures 2 the adsorption isotherm of argon on graphitized thermal carbon
black at a number of tempemtures. The symbols in the figure are experimental data [131.
10

0 5 10
Resslm,(MPa)
Figure 2n: MLST prediction of argon isotherms
15
0
0
-
5 10
WPa)
15

Figare Z b Comparison between GCMC & MLST

The continuous lines in these plots are fiom the results of MLST theory and GCMC. We
see that the agreement between the MLST theory and the data [131 is excellent (Figure
2a). In patticular, the temperature dependence of the isotherms is well described by the

5
MLST model. The results obtained from GCMC are shown in Figure 2b, and it is
surprising that the GCMC results are not as good as the MLST results. It describes the
data well at low pressure but it consistently over-predicts the amount adsorbed at higher
pressures. One probable explanation for this might be the possibility of temperature and
density dependence of the molecular parameters in GCMC simulation for high pressures.
This reason needs some justification because the MLST also uses molecular parameters,
and they are kept constant in the calculations. This illustrates that the molecular
simulation, despite its great potential, is less effective than the MLST for this problem of
high pressure adsorption of super-critical fluids on non-porous surfaces. The MLST is
very effective in the study of the effects of parameters and operating conditions, such as
surface carbon density and solid-fluid interaction energy, on the behaviour of surface
excess at extremely high pressures. This will be discussed in details at the conference.
What we observed for argon are also repeated with Kr, methane, ethylene, propane
and sulphur hexafluoride. Figures 3 show the comparison between GCMC and MLST
for ethylene at 283 JC, which is just greater than the critical temperature.

We see that at this temperature, the surface


excess versus pressure shows a distinct
spike at a pressure very close to the critical
pressure of 5.036 MPa Beyond this
pressure, the surface excess decreases very
sharply and this is due to the large change
in density in the bulk fluid for a small
change in the pressure. It is remarkable
that the MLST describes the data [14] very
0.1 1 10 100
well. The GCMC, on the other hand,
Pmssurs (MPa)
Figure 3: Comparison between GCMC & MI.ST for describes well at low pressures but badly
ethylene adsorption at 283 K predicts at higher pressures.

3.1.2 Slit pores


For slit pores, we compare the results of DFT and GCMC for slit pores having width 8
and 14 A in Figures 4 for argon adsorption at a super-critical temperature of 298K.
0.8 1.o

0.8
0.8

-&
0.4

B
0.2
0.2

0.0 0.0
104 104 10= 10' lo-' 100 10' 10' lW 103 101 10' 16 10' 102
Activity. z d AdMty, zd
Figure 4n: GCMC versus DFT for slit pore H = 8 A Figure 4b:GCMC versus DFT for slit pore. H = 14 A

We fmt note the gradual behaviour of the adsorption isotherms, and this is typical for
super-critical fluid adsorption in pores of all sizes. For sub-critical adsorption which we
will consider later, there is a possibility of phase transition for certain pore widths.

6
We see that the agreement between the DFT and GCMC results is very good for
large micropores of 14 A, while there is a distinct deviation in small micropore of 8 A.
This is due to the mean field approximation of the DFT method (eq. lb). In terms of
computation time, the DFT method is modestly faster than the GCMC (depending on the
error criteria as well as the grid points used in the determination of singlet density
distribution) but not substantially fast enough to warrant it in its use in routine process
applications.

3.2 Sub-criticalfluid

3.2.1 Non-porous surface


Sub-critical fluids are widely applied for surface and pore characterization. Figure 5
shows the surface excess adsorption of argon at 87.3K on graphitized thermal carbon
black. The results are obtained fiom the GCMC and DFT. The subtle difference in the
formation of second and higher layers is observed and this is due to the mean field
approximation. As an illustration of the MLST method, we show in Figure 6 the results
in the prediction of argon adsorption isotherm on graphitised carbon black at 87.3 K [IS].
The theory exhibits a kink formation of the second layer and this deviation from the data
could probably be due to the heterogeneity of the graphene surface. It is interesting to
note that that kink exists in the DFT as well as GCMC simulation (see arrow in Figure 5).

10 1b 18 1@ 101
Gas Density. po -M (Pa)
pigpre 5: GCMC v m DFT for open surfacc for Ar Figure 6: Paformance of MLST for Ar at 87.3 K

3.2.2 Slit pores:


Here we would like to show the difference between the three methods used in pore
characterization: GCMC, DFT and Dos method. Figures 7 show the average pore
density versus pressure for argon adsorption in slit pores having width of 8 and 30A.
Here we see that the difference between the DFT and GCMC for 8 A pore is quite
significant, and not only in terms of the pore filling pressure but also on the behaviour of
the adsorption isotherm curves. The local isotherms obtained from the Do-method for
these pores, although not completely in agreement with the GCMC results, qualitatively
match these results rather well over the complete range of pressure. The fast computation
of the Do-method can be the greatest advantage among the various methods, and it can be
effectively used in process optimizationor in routine pore characterisation.

7
*r-K!f,Pb Qaa Den*, Fa?
Figure 7a: GCMC versus DFT for slit pore H = 8 A Figwe m: G M C vermu, DFC for slit pore H-30 A

4 Conclusions

We have presented in this paper an overall but brief summary of various tools commonly
used in the adsorption equilibria calculations in carbon. These tools rage fmm simple
methods such as the Do-metho& the MLST of Ustinov and Do to advanced methods such
as the DFT and the GCMC simulation. Although the DFT is modestly faster than the
GCMC, they are both computer time-hungry and unsuitable in any applications that
require routine computation of isotherms, especially when parameters are varied. The
simple methods,in particular the Do-method,might provide an answer to this.

5 Acknowledgements

This work is supported by the Australian Research Council. Support b m PBAST3 is


also gratefully acknowledged.

References

1. A&, D. J., Mol. P h y ~29


. (1975) pp. 307-3 11.
F~nkel,D. and Sinit, B., Understanding Molecular Simulation (Acad. Press, 2002)
2.
Tarazona, P., Marconi, U. and Evans, R., Mol. Phys. 60 (1987) pp. 573-595.
3.
Ustinov, E. and Do, D. D., J. Colloid I n f e ~ a cScience
4. e (in press, 2002)
5. Do, D. D., Ustinov, E. and Do, H.D., FluidPhase Equilibria (bpress, 2002)
6. Do, D. D.,presented at The International Symposium on New Trends in Colloid and
Interface Science (14-26 September, 1998, Chiba, Japan).
7. Nguyen, C. andDo, D. D.,Lungmuir 15 (1999) pp. 3608-3615.
8. Do, D. D.,Nguyen, C. and Do, H.D.,Colloids Sur$acces 187-188 (2001)- _pp. _ 51-71.
9. Do, D.D. and Do,H.D.,Lungmuir 18 (2002) pp. 93-99.
10. Do, D.D. and Do, H.D., AppliedSu@ace Science (in pms, 2002)
11. Metropolis, N., Rosenblutb, A,, Rosenbhth, M., Teller, A. and Teller, E., J. Chem.
. (1953) pp. 1087-1092.
P h y ~21
12. Peterson, B. and GubbinS, K.E., Mol. Phys. 62 (1987) pp. 215-226.
13. Specovius, J. and Findenegg, G., Ber. Bun. Phys. Chem. 82 (1983) pp. 174180.
14. Findenegg, G., Funhentaki of A&otption, May 6-1 1 (1983) pp. 207-218.
15. Olivier, J., J. Porous Materials 2 (1997) pp. 9-17.

8
FREEZINGMELTING IN POROUS CARBONS

F. R. HUNG AND K. E. GUBBINS


Department of Chemical Engineering, North Carolina State University,
Raleigh, NC 2 7695, USA

R. RADHAKRISHNAN
Department of Chemistry and Courant Institute of Mathematical Sciences,
New YorS NY 10003, USA

F. BEGUIN
Centre de Recherche sur la Matiire Divisee, UMR-6619, CNRS-Orleans University,
Orlt!ans945071 cedex 02, France

M. SLIWINSKA-BARTKOWIAK
Institute of Physics, Adam Mickiewicz University, Umultowska 8.5, 61-614 Poznan, Poland

We report molecular simulation and experimental results for the freezing/melting behavior of simple
fluids adsorbed in porous carbons having slit-shaped or cylindrical geometry. In the simulations we use
simple models for activated carbon fibers of pore width 1.5 nm and multi-walled carbon nanotubes of
diameter 5 nm. Experimental studies employ dielectric relaxation spectroscopy and nonlinear dielectric
effect measurements. We show that the solid-fluid phase behavior depends on two main parameters,
the pore width H (or diameter D? and a parameter measuring the ratio of the fluid-wall to fluid-fluid
attractive interaction (a).For slit-shaped pores, values of a larger than unity lead to an increase in the
freezing temperature of the confined fluid with respect to that of the bulk, and values of a smaller than
unity produce a decrease in the freezing temperature. Contact layer and hexatic phases were also
observed for some systems. The hexatic phase was found in slit-shaped pores with larger a values, and
is stable over extended temperature ranges. For cylindrical pores, our results for a D = 9 . 7 q multi-
walled carbon nanotube show no formation of regular 3D crystalline structures. Our results also suggest
that the outer layers experience a slight increase in the freezing temperature, while the inner layers
experience a depression in the freezing temperature with respect to the bulk freezing point. These
simulation results are in good agreement with the experimental results.

1. Introduction
An understanding of fieezing phenomena for fluids confined within nano-scale pores is
important in the fabrication of nano-structured materials, and in nanotribology, adhesion,
and characterization of porous materials. However, studies of these phenomena are
plagued by several difficulties. Experimental investigations must address the difficulties
of the lack of well-characterized porous materials with appropriate pore size, the
unambigous determination of the nature of the confined phase, and the prevalence of
long-lived metastable states. Molecular simulation studies do not suffer fiom any of these
difficulties; moreover, it is possible to determine free energies of the confined phases, and
so determine true thermodynamic equilibrium and the order of the phase transition.
Simulations experience other difficulties, however, in particular uncertainties concerning
intermolecular potentials and pore characterization, as well as limitations due to the speed
of current supercomputers. The difficulties found in experiments and simulations make
the two approaches complementary, so that combined experimental-simulation studies
can be rewarding.
Recent studies for pores of simple geometry have shown a rich phase behavior associated
with fieezing in confined systems -I3]. The freezing temperature may be lowered or
raised relative to the bulk fieezing temperature, depending on the nature of the adsorbate
and the porous material. In addition, new surface-driven phases may intervene between

9
the liquid and solid phases in the pore. "Contact layer" phases of various kinds often
occur, in which the layer of adsorbed molecules adjacent to the pore wall has a different
structure fi-om that of the adsorbate molecules in the interior of the pore. These contact
layer phases have been predicted theoretically, and confirmed experimentally for several
systems [4*91. In addition, for some systems in which strong layering of the adsorbate
occurs (e.g. activated carbon fibers), hexatic phases can occur; such phases have quasi-
long-ranged orientational order, but positional disorder, and for quasi-two-dimensional
systems occur over a temperature range between those for the crystal and liquid phases.
These are clearly seen in molecular simulations, and recent experiments provide
convincing evidence for these phases [14]. It has been shown that this apparently complex
phase behavior results ftom a competition between the fluid-wall and fluid-fluid
intermolecular interactions. For a given pore geometry and width, the phase diagrams for
a wide range of adsorbates and porous solids can be classified in terms of a parameter a
that is the ratio of the fluid-wall to fluid-fluid attractive interaction t1,97131.
In this paper we report molecular simulation studies of fteezing in slit-shaped pores and
in multi-walled carbon nanotubes. Simulations make use of the Landau-Ginzburg
formalism for the free energy[I3] and parallel tempering"']. Experimental results from
dielectric relaxation spectroscopy are reported for several fluids in these materials.

2. Methods
2.1. Simulation

We performed grand canonical Monte Carlo (GCMC) simulations of Lennard-Jones (LJ)


fluids adsorbed in pores of slit-shaped and cylindrical geometry. The slit-shaped pore of
width H was modeled using the 10-4-3 Steele potential [I6], whereas the cylindrical pore
of diameter D was modeled using the potential due to Peterson et al."71. Systems of up to
64000 molecules were considered. We calculated the Landau free energy as a function of
an effective bond orientational order parameter @, which is sensitive to the degree of
order in the system. The Landau fi-ee energy approach used in earlier s t ~ d i e s [ ~was
*'~~
extended to incorporate spatial inhomogeneity in the order parameter, and a generalized
Landau-Ginzburg approach was developed to determine the Landau free energy surface
of inhomogeneous fluids"31,with the latter given by:

A[@(r)] = -k,T ln(P[@(r)]+ constant

where k~ is the Boltzmann constant, T is the temperature and P[@(r)]is the probability of
observing a system having an order parameter value between @ and @+&D at the position
given by r. The probability distribution function P[@(r)] was determined during the
simulations using umbrella sampling in pores of slit-like geomerry [I3], and parallel
tempering "'I for pores of cylindrical geometry. We found significant ordering into
distinct molecular layers in the systems considered; therefore it is possible to define the
order parameter @(r)using a two dimensional bond orientational order parameter, defined
as tw.

10
where the fmt sum is over the adsorbed molecular layers and ij is the coordinate of the
plane in which molecules in layerj are most likely to lie on. We expect @(r)=l for two-
dimensional hexagonal crystals and @(r)=O for two-dimensional liquids. YGJ(p)measures
the hexagonal bond order at position p = x ex + y ey within each two-dimensional layer j.
Nb is the number of nearest neighbors of a molecule at position and Bk is the orientation
of each nearest neighbor bond with respect to an arbitrary axis E31. For slit-shaped pores,
these two-dimensional layers lie in the xy plane. For cylindrical pores, a quasi-two
dimensional configuration of each molecular layer can be obtained by cutting each one of
the concentric layers along the axial direction and unrolling it flat. The grand fiee energy
of a particular phase is then related to the Landau fkee energy by [13]:

The probability distribution fhction P[@(r)] was determined during the simulations
using umbrella sampling in pores of slit-like geometry [13], and parallel tempering [I5] for
pores of cylindrical geometry I19]. The method of parallel tempering is a Monte Carlo
scheme that has been derived to achieve good sampling of systems that have a fiee energy
landscape with many local minima [*'I. Yan and de Pablo [''I implemented this scheme in
the grand canonical ensemble, performing an MC simulation in n-systems which differ in
both temperature and chemical potential. In addition to the standard MC trial moves, they
proposed configuration swaps in this method. A swap attempt between configurations i
a n d j is accepted with probability given by

where A P l l k J , - l/kBT,, and AU and are the difference in potential energies and
number of particles, respectively, between configurations i and j. We have used as many
as 50 different configurations in our calculations with the parallel tempering Monte Carlo
scheme, to cover all the phase space of interest and to guarantee frequent swaps between
replicas. We have checked the evolution of the parallel tempering simulation as a function
of Monte Carlo steps and after the equilibration, we verified that each configuration
visited many sets of (T,p) along a single simulation run.The nature of the phases in each
molecular layer was determined by measuring the two-dimensional bond orientational
order parameter and by monitoring the two-dimensional, in-plane positional and
orientational pair correlation b c t i o n s within each layer. The positional pair correlation
function in layerj is given by the familiar radial distribution function g@), measured
within the two-dimensional plane formed by each layer j. The orientational pair
correlation function is given by G,,J( p ) = (Y; (0)VJ( p ) ) .

2.2. Experiments: Dielectric Relaration Spectroscopy


The relative permittivity of a medium, K'=Kr-iK,, is in general a complex quantity whose
real part K~ (also known as the dielectric constant) is associated with the increase in
capacitance due to the introduction of a dielectric[211.The imaginary component K, is
associated with mechanisms that contribute to the energy dissipation in the system, due to
viscous damping of the rotational motion of the molecules in alternating fields; this effect
is frequency dependent. The experimental setup consisted of a parallel plate capacitor of
empty capacitance Co.The capacitance C and the tangent loss, tan(b) of the capacitor
filled with the sample were measured at different frequencies and temperatures, using a
Solartron 1260 Gain Analyzer. For the case of CC4 in multi-walled carbon nanotubes, the
sample was introduced between the capacitor plates as a suspension of the adsorbent
(multi-walled nanotubes with average internal diameter of 5 nm, average external
diameter of 10 nm)in the pure adsorbate (CC14). The relative permittivity is related to the
measured quantities by K,=C/C~,~ = t a n ( S y Kwhere
, 6 is the angle by which current leads
the voltage. Further details of the experimental methods are described elsewhere

3. Results
3. I . Activated Carbon Fibers (Slit-Shaped Pores)
In Figure 1 we show the Landau free energies for LJ CC14 confined in activated carbon
fibers of pore width H*=3, at P 3 3 5 K and T=290 K [I3]; pores of this width can just
accommodate two adsorbed layers of CC4. We also show in this figure the grand free
energy of the phases as a h c t i o n of temperature. The free energies show three local
minima. Examination of the spatial and orientational correlation functions show that these
correspond to liquid-like (L), hexatic (H) and crystalline (C) phases. Thus for the liquid-
like phase g(r) and G6,,(r)both showed short range order, whereas for the crystalline
phase they both showed long-range order; for the hexatic phase g(r) showed short range
order, but G6J(r)showed algebraic decay, indicating quasi-long range bond orientational
order. For E-347K and TG90 K, the thermodynamically stable phases are the liquid and
hexagonal crystalline phases, respectively. Between these two temperatures, the hexatic
phase is the thermodynamically stable phase. The existence of the hexatic phases was
verified using the Landau-Ginzburg formalism and system size scaling analysis on very
large system sizes[14]. Experimental measurements of the nonlinear dielectric effect on
carbon tetrachloride (CCL,) and aniline confined within activated carbon fibers showed
divergences at temperatures close to those found in the simulation^"^^, and the
divergences obeyed the scaling laws predicted by the KTHNY theory, thus providing
strong evidence for the hexatic phase.
, 1 6 .

TJK

@)

Figure 1. (a) Landau free energy in the two molecular layers of LJ CCIJ confined in a graphitic, slit-like pore of
width H=3ojj(1.5 nm),at T=335 K and P290 K. (b) Grand free energy as a function of temperature for liquid,
hexatic and crystalline phases, for the same system. From
A corresponding states a n a l y ~ i s ~ ~
shows
" ~ ] that the shift in the freezing temperature, as
well as the temperature of any other surface-driven phase transitions, should depend on
three reduced variables, namely H*=H/c~fi a = ~ , E ~ O & A /and E ~o-j&fi here f and w
refer to confined fluid and wall, respectively, avis the atom density of the pore wall, and
A is the spacing between the atomic layers that make up the pore wall. For relatively

12
small fluid molecules the size ratio parameter is found to have little effect on the freezing
behavior, except in the molecular sieving regime. Thus it is possible to plot global
freezing diagrams in terms of a, the ratio of fluid-wall to fluid-fluid attractive forces.
Such a diagram is shown in Figure 2 pores of width H*=3, obtained from both simulations
and experiments [I3. The diagrams from simulations and experiments are in qualitative
agreement, and demonstrate that a parameter determines the qualitative nature of the
fieezing behavior. Large values of a lead to an increase of the freezing temperature of the
confined fluid and to the formation of hexatic phases. As the fluid-wall interactions
become weaker, a decrease on the freezing temperature is observed and the temperature
range where the hexatic phase is stable decreases. Similar diagrams for larger pores, e.g.
H*=7.5, using molecular simulations [I show the presence of contact layer phases, in
which the layer of adsorbed molecules adjacent to the pore wall has a different structure
from that of those molecules in the interior of the pore. These simulations show that for
simple fluids the role of a is to determine the nature of the fkeezing behavior, while that of
H* is primarily to determine the magnitude of shifts in the transition temperatures.

.(a) (b)

Figure 2. Global phase diagram of a fluid in a slit pore of width H=303 (I .5 nm) from (a) simulations, and (b)
experiments. The experiments are for various adsorbates confined within activated carbon fibers: 1=H20,
2=C&N& 3=c&NH2, kCH@H, 5=CCL, 6 e . From and references therein.

3.2. Carbon Nanotubes


The freezing behavior of LJ CCl, adsorbed in model multi-walled carbon nanotubes of
internal diameter D=5 nm (D*=9.7)was studied using parallel tempering. The formation
of five concentric layers of adsorbate was observed when the confined fluid solidified. In
Figure 3(a) we show the average values of the two-dimensional, bond orientational order
parameter in the individual layers, as a function of temperature. We observe a
discontinuity in the average order parameters of the contact and second layers around 260
K, suggesting a phase transition in these two layers. The change in the order parameter for
the third layer is less abrupt than in the first two layers and starts around 255 K. In the
fourth layer, the order parameter value increases continuously until it jumps around 215
K, reaching a maximum value of a-0.7 at lower temperatures. Our results suggest that
for this specific pore diameter, the contact and first inner layers freeze at temperatures at
or slightly higher than the bulk freezing point, whereas the inner layers experience a
depression in the freezing temperature when compared to that of the bulk. These
observations are corroborated by our results for the positional and orientational pair
correlation functions in the unwrapped layers. Some of our results for the contact layer
are shown in Figure 4(b). At T=262 K, the isotropic positional pair correlation hnction
and the exponential decay in the orientational pair correlation function are signatures of

13
an isotropic liquid. At T=252 K, the features observed in both pair correlation functions
are characteristic of a two-dimensional hexagonal crystal "". Similar features were found
for the other layers.

0.8

0n
0.6

0.4
e
02

(a) (b)
Figure 3. (a) Two-dimensional,bond orientational order parameter average values in the molecular fluid layers
of IJ CCL confined in a multi-walled carbon nanotube of diameter D = 9 . 7 9 (5 nm). Triangles, squares,
diamonds and circles represent the order parameter values for the contact, second, third and fourth layers,
respectively. The dotted line represents the bulk solid-fluid transition temperature. (b) Positional and
orientational pair correlation functions in the unwrapped contact layer of LJ CC4 confined in a multi-walled
carbon nanotube of diameter D = 9 . 7 0 ~ ( run)
5 showing liquid phase at 7'=262 K and crystal phase at T=252 K.

Our results for the D=9.7ufcarbon nanotube did not show any sign of ordering of the
confined LJ CCl, into regular, three-dimensional crystal structures. This finding is in
agreement with previous simulation [lo] and experimental 110*22-241 studies, where it was
shown that for pore diameters below 200- only partial crystallization occurs. The same
studies found that for silica materials, the lower limit below which there are no 3-D
crystal domains in the system was around D=120- 110*22-241. Our simulations for carbon
nanotubes of smaller diameters show that the freezing temperatures increase as the pore
diameter decreases. We are currently performing free energy calculations to determine the
thermodynamic stability of each of the phases, as well as to determine the exact phase
transition temperatures for the confined layers [''I.

In Figure 4 we show the behavior of the experimentally determined capacity C as a


function of temperature for CCl, confined in multi-walled carbon nanotubes with average
internal pore diameter of 5 nm (average external diameter equal to 10 nm). The sample
was introduced between the capacitor plates as a suspension of CC14-filled carbon
nanotubes; however, the volume of bulk C C 4 was very small compared to that of the
confined CCl4, so that the results represent the confined adsorbate. The results indicate
that melting in the amorphous inner layers starts at about 205 K, and proceeds
continuously up to 234 K, where there is a sharp increase in C associated with the melting,
in good agreement with the simulations. The bulk melting point is 250 K. The decrease
in C for temperatures above 259 K indicates the final melting of remaining solid CCl, in
the pore at 259 K, and may correspond to melting of the contact layer(s). The features
found at T=175 K may be associated with a solid-solid transition, since such a transition
(from monoclinic to rhombohedric crystal) occurs in bulk CCl, at 210 K. The melting

14
temperatures found in the experiments are in qualitative agreement with those found in
our simulations.

Figure 4. Dependence of capacity C on T for CCL in multi-walledcahon nanotubes with average pore diameter
of 5 nm (10 nm average external diameter), from dielectric relaxation spectroscopy. Symbols represent results
obtained at different frequencies: circles, 30 kHz; squares, 100 kHz; and triangles, 6OOkHz. The signals are for
both bulk and confined CCL.

4. Conclusions
Our results in slit pores show that the solid-fluid phase behavior for this case depends on
two main parameters, the reduced pore width H and a parameter measuring the ratio of
the fluid-wall to the fluid-fluid (a)attractive potential. An increase of the freezing
temperature of the confined fluid relative to the bulk is observed for large values of a,
whereas a decrease is observed for small values of a. We found evidence of new surface-
driven confined phases, such as contact layer and hexatic phases. The hexatic phase
was found in slit-shaped pores with larger a values, and it is stable over extended
temperature ranges. Phase diagrams of the transition temperature vs a were depicted for
several LJ fluids confined in activated carbons H*=3. These phase diagrams are in
qualitative agreement with experimental results. For cylindrical pores, our results for a
D=9.7oS multi-walled carbon nanotube show no formation of regular 3D crystalline
structures, in agreement with experimental results. Our results also suggest that the outer
layers experience an increase in the freezing temperature, while the inner layers
experience a depression in the freezing temperature with respect to the bulk freezing
point. Simulations for carbon nanotubes of smaller diameters show that the freezing
temperatures increase as the pore diameter decreases. Preliminary experimental results
show solid-fluid transition temperatures for CCl, in carbon nanotubes that are in
qualitative agreement with those determined in our simulations.
This work was supported by a grant from the National Science Foundation (Grant No.
CTS-9908535). Supercomputer time was provided under a NSF/NRAC grant
(MCA93S011).

15
[I1 For a review to mid-1999 see: L.D. Gelb, K.E. Gubbins, R. Radhakrishnan and M. Sliwinska-
Bartkowiak, Phase Separation in Confined Systems, Rep. Prog. Phys. 62 (1999) pp. 1573-1659
I[' M. Miyahara and K.E. Gubbins, FreezingMelting Phenomena for Lennard-Jones Methane in
Slit Pores: a Monte Carlo Study, J. Chem. Phys. 106 (1997) pp. 2865-2880
r31 R. Radhakrishnan and K.E. Gubbins, Free Energy Studies of Freezing in Slit Pores: An Order-
Parameter Approach Using Monte Carlo Simulation, Mol. Phys. % (1999) pp. 1249-1267
r41 M. Sliwinska-Bartkowiak, J. Gras, R. Sikorski, R. Radhakrishnan, L.D. Gelb and K.E. Gubbins,

Phase Transitions in Pores: Experimental and Simulation Studies of Melting and Freezing,
Langmuir 15 (1999) pp. 6060-6069
['I H. Dominguez, M.P. Allen and R. Evans, Monte Carlo Studies of the Freezing and Condensation
Transitions of Confined Fluids, Mol. Phys. 96 ( 1998) pp. 209-229
[61 K. Kaneko, A. Watanabe, T. Iiyama, R. Radhakrishnan and K.E. Gubbins, A Remarkable Elevation
of Freezing Temperatureof CCL, in Graphitic Micropores, J. Phys. Chem. B, 103 (1999) pp. 7061-
7063
[71 A. Watanabe and K. Kaneko, Melting Temperature Elevation of Benzene Confined in Graphitic
Micropore", Chem. Phys. Lett. 305 (1999) pp. 71-74
[*I R. Radhakrishnan, K.E. Gubbins, A. Watanabe and K. Kaneko, Freezing of Simple Fluids in
MicroporousActivated Carbon Fibers: Comparison of Simulation and Experiment,J. Chem. Phys.
111 ( 1999) pp. 9058-9067
l9] R. Radhakrishnan, K.E. Gubbins and M. Sliwinska-Bartkowiak, Effect of the Fluid-Wall
Interaction on Freezing of Confined Fluids: Towards the Development of a Global Phase Diagram,
J. Chem. Phys. 112 (2000) pp. 11048-1 1057
['I M. Sliwinska-Bartkowiak,G. Dudziak, R Sikorski, R. Gras,R. Radhakrishnan and K.E. Gubbins,
MeltingRreezing Behavior of a Fluid Confined in Porous Glasses and MCM-41: Dielectric
Spectroscopyand Molecular Simulation, J. Chem.Phys. 114 (2001) pp. 950-962
[ I i 1 M. Sliwinska-Bartkowiak,G. Dudziak, R. Sikorski, R. Gras,K.E. Gubbins and R. Radhakrishnan,
Dielectric Studies of Freezing Behavior in Porous Materials: Water and Methanol in Activated
Carbon Fibers, Phys. Chem. Chem. Phys. 3 (2001) pp. 1 179-1184
[*'] M. Sliwinska-Bartkowiak,R. Radhakrishnan and K.E. Gubbins, Effect of Confinement on Melting
in Slit-Shaped Pores: Experimental and Simulation Study of Aniline in Activated Carbon Fibers,
Mol. Sim. 27 (2001) pp. 323-3337
["I R .Radhakrishnan, K.E. Gubbins and M. Sliwinska-Bartkowiak, Global Phase Diagrams for
Freezing in Porous Media, J. Chem. Phys. 116 (2002) pp. 1147-1155
[14] R. Radhakrishnan, K.E. Gubbins and M. Sliwinska-Bartkowiak, On the Existence of a Hexatic
Phase in Confined Systems, Phys. Rev. Lett. 89 (2002) art. 076101
["I Q. Yan and J.J. de Pablo, Hyper-Parallel Tempering Monte Carlo: Application to the Lennard-
Jones Fluid and the Restricted Primitive Model, J. Chem. Phys. 111 (1999) pp. 9509-95 16
[I6] W.A. Steele, Physical Interaction of Gases with Crystalline Solids: 1. Gas-Solid Energies and
Properties of Isolated Adsorbed Atoms, Sut$ Sci. 36 (1973) pp. 3 17-352
[I7] B.K. Peterson, J.P.R.B. Walton and K.E. Gubbins, Fluid Behavior in Narrow Pores, J. Chem.
Soc., Faraday Trans. 2 82 (1986) pp. 1789-1800
[lS1 R.M. Lynden-Bell, J.S. van Duijneveldt and D. Frenkel, Free-Energy Changes on Freezing and
Melting in Ductile Metals, Mol. Phys. 80 (1993) pp. 801-814
[I9] F.R. Hung, M. Sliwinska-Bartkowiakand K.E. Gubbins, in preparation (2002).
I'[ D. Frenkel and B. Smit, Understanding Molecular Simulation: From Algorithms to Applications
2"* Edition (Academic Press, London, 2002)
["I A. Chelkowski, Dielectric Physics (Elsevier, North-Holland, New York 1980)
I'[ D. Morineau, G. Dosseh, C. Alba-Simionesco and P. Llewellyn, Glass Transition, Freezing and
Melting of Liquids Confined in the Mesoporous Silicate MCM-41,
Philos. Mag. B 79 (1999) pp. 1847-1855
["I G. Dosseh, D. Morineau and C. Alba-Simionesco, Benzene Confined in MCM-41 Below its
Melting Point: A Proton NMR Study, J. Phys. IV 10 (2000)pp. 99-102
[241 D. Morineau, F. Casas, C. Alba-Simionesco, A. Grosman, M.-C. Bellisent-Funel and N.
RatovClomanana, A Neutron Scattering Investigation of the Structural Properties of
Glassforming M-Toluidine Confined in MCM-41, J. Phys. IV 10 (2000) pp. 95-98

16
MEASUREMENT OF DIFFUSION IN MICROPOROUS SOLIDS

DOUGLAS M. RUTHVEN
Department of Chemical and Biological Engineering, University of Maine, Orono, ME 04469
USA
E-mail: druthven!dumche.maine.edu

The experimental methods for measurement of transport and self-diffision in zeolite crystals (and in
other microporous materials) are reviewed. Large discrepancies between different techniques are
commonly observed and appear to be related to the scale of the measurements, suggesting that
structural defects may be more important than is generally believed.

1 Introduction

An understanding of diffision in microporous solids is essential for the rational design


and optimization of catalytic processes, adsorptive separation processes and inorganic
membranes yet, despite intensive study, our understanding is still far from complete. A
wide range of different experimental techniques have been developed but, for many
systems, the difisivities determined by different methods show discrepancies of more
than an order of magnitude, making it difficult to establish the true value. This paper
presents a short overview of the major experimental approaches with a more detailed
discussion of one particular technique (the zero length column or ZLC method) which
was developed several years ago in our laboratory [ l , 21 and has been widely applied to
study many different systems.

2 Definition of Diffusivities

Molecular transport and tracer diffision in porous solids are conveniently correlated in
terms of difisivities defined in accordance with Ficks first equation:

J = -D(q). aq- .
9 J* = - D * ( q ) . Y l
az az 4

The self-difisivity defined on the basis of the random walk representation of a diffusive
process:
~

2n t 2n r

is identical to the tracer diffbsivity. The relationship between the transport and tracer
diffusivity is more complex.
Considering the chemical potential gradient rather than the concentration gradient as
the fundamental driving force leads to:

17
where Do = BRT is the corrected or intrinsic transport diffusivity. Do and D
coincide only in the Henrys Law region where d lnp/d Inq = 1.O. For a type 1 isotherm
Eq. 3 implies a strongly increasing trend of di hi vity with loading. Eq. 3 is often
attributed to Darken [3] but this formulation actually goes back to the much earlier work
of Maxwell and Stefan [4, 51. Krishna and coworkers have applied the Maxwell-Stefan
approach to micropore diffusion by considering the unoccupied sites (vacancies) as the
n+l component in the system [6,7]:

where D:. measures the rate at which components i and j can exchange directly and
0; measures the rate of migration to vacant sites, the velocity of which (u,+,) is zero.
For single component diffusion (e,=O, N,=O) this reduces to:

J=--.-.-
DV q dp
(1-8) RT dz
1 = D 1A which yields [8]:
and for self diffision we have J A = - J g and DAB

This formulation suggests that at low loadings ( b 0 ) the corrected and tracer diffisivities
should coincide but at higher loadings D < Do except when D is very large.
These formulations are applicable only to normal systems in which diffusing
molecules can pass each other in the channels. When molecules cannot pass we have
single file or string of pearls diffusion which has entirely different features. This topic
has been well reviewed by Kiirger [9] and will not be considered here.

3 Experimental Methods

A summary of the major experimental techniques that have been applied to study
diffusion in microporous solids is given in Table 1. These can be classified according to
the nature of the measurement (transient or steady state), the scale of the measurement
(micro, meso or macro) and the type of process (transport or self-diffision). Additional
details with complete references are given in recent reviews [lo-121.
Most of the microscopic techniques are restricted to relatively small and rapidly
diffising species since a measurable mean molecular displacement must be achieved
within a short time. In general the microscopic techniques show reasonable consistency
with other micro techniques but, as the length scale of the measurement increases, the
apparent diffisivities decrease and show increasing discrepancies between different
techniques.

18
TABLE 1

Classification of Experimental Techniques for Diffusion Measurements

Transport Diffusion Self-Diffusion

Transient Steady State

Macroscopic Sorption Rate (12) Membrane Tracer methods (12)


Methods IR spectroscopy (16, 17) permeation (14, 15)
Frequency response (20) Effectiveness factor
Chromatography (12) (18919)
ZLC (1 3)
Differential adsorption
bed (21)
TAP reactor (22)
Imaging (23)

Mesoscopic IR microscopy (16, 17) Single crystal Single crystal


Methoa3 membrane (14, 15) membraneltracer

Microscopic Interference microscopy PFG NMR (12)


Methods (24325) NMR-lifetime (12)
Coherent QENS (26) Incoherent QENS

4 The ZLC Technique

The ZLC technique [ 131 depends on measuring the desorption curve for a small sample of
adsorbent pre-equilibrated with sorbate under well defined conditions. The essentials of
the experimental system are shown in figure 1. The column consists of a very small
sample of adsorbent held between two suiter discs in a Swagelock fitting. The
desorption curve is determined by both the adsorption equilibrium and the kinetics but
these two effects can be easily separated by making measurements over a range of flow
rates. In the low flow rate regime intraparticle diffusion is relatively rapid. The
gradient of sorbate concentration through an adsorbent particle is very small and the
desorption rate is determined by convection (the rate of removal of sorbate fiom the
adsorbent surface). In contrast, at high flow rates, the concentration at the external
surface of the particle is always very small and the desorption rate is controlled by
diffusion out of the particle. Thus, by making measurements over a sufficiently wide
range of flow rates both the equilibrium and kinetic parameters may be determined.
This approach is illustrated in fig. 2 which shows a family of ZLC desorption curves
for COz fiom a sample of a carbon monolith, measured over a wide range of flow rates.
The isotherm is linear so at low flow rates (equilibrium control) the response is given by:

19
r 1

when plotted against the volume of purge (Ft) the response should be invariant with flow
rate and linear in coordinates of In (c/co) vs Ft. The slope yields the sum V,+KV,. For
strongly adsorbed species the dead volume (V3 is usually negligible but for weakly held
species (as in this example) the correction may be important. V, is conveniently
determined from a blank experiment with no adsorbent present.
The general response (treating the adsorbent as a parallel sided disk of thickness 21)
is given by: ,
-c- - a 2 L e x p - P 2I
c
co n=l L + p2(1+ y ) + (L - y#)2

where Pn is given by the roots of


2
L - YPn - Pn tan Pn = 0 (9)

and y = VdKV,. The simplification for y + 0 is obvious. Varying the flow rate
changes L and thus provides a simple experimental test for consistency with the
isothermal d i m i o n model. The corresponding expression for spherical particles is also
available [131.
The ZLC method offers advantages of speed and simplicity and requires only a very
small adsorbent sample thus making it useful for characterization of new materials. The
basic experiment using an inert carrier (usually He) measures the limiting transport
diffusivity (Do)at low concentration. A variant of the technique using isotopically
labeled tracers (TZLC) yields the tracer diffisivity and counter diffision in a binary
system may also be studied by this method. To obtain reliable results a number of
preliminary experiments are needed, e.g. varying sample quality, nature of the purge gas,
the flow rate and, if possible, particle size to c o n f m intracrystalline diffiision control.

5 Comparison of ZLCPFG N M R Diffusivities

PFG NMR measurements can be applied over a wide range of diffusivities and the scale
of the measurement can be varied from a few unit cells to perhaps half the crystal or
particle diameter. For a number of systems good agreement between ZLC and PFG
NMR measurements has been demonstrated; examples are shown in fig. 3. The data for
methanol-NaX are particularly interesting as the unusual concentration trend is revealed
by both sets of data and at low concentrations the tracer diffusivities converge to the
limiting value of Do as is to be expected. Unfortunately such agreement is far 6om
universal. Figure 4 shows data for some systems for which large discrepancies are
found, both in absolute values of the diffusivities and in the concentration trends.

6 Comparison of ZLCWrequency Response Diffusivities

Diffusion of aromatic hydrocarbons in silicalite has been widely studied by several


different methods although these systems are not amenable to NMR measurements

20
because of their short relaxation times and relatively low diffisivities. ZLC and
frequency response (FR) diffusivities for benzene are compared in Fig. 5. The data
show reasonable agreement as to the magnitude of the diffisivities but when trends with
loading are examined some striking discrepancies emerge. For both benzene and p-
xylene the TZLC data show the self-diffisivity to be almost independent of loading
whereas the FR data suggest that Dodecreases quite strongly. For benzene the ZLC data
show that tracer and corrected transport diffisivities are almost the same but for p-xylene
the transport diffisivity is smaller. For both species diffusion in ZSMS (A1 rich) is
slower than in silicalite, presumably as a result of blocking by the strong A1 sites.
The diffusional behavior of p-xylene is complicated. The FR measurements reveal
two different diffusivities corresponding to movement through the straight and sinusoidal
channels. The ZLC method increases only the average diffisivity which is similar to the
value for benzene but it is possible that the difference between the self and transport
diffisivity results from the two channel behavior revealed by the FR data [33].

7 Variation of Diffusivity with Chain Length of Linear Alkanes

The extensive diffusivity data for linear alkanes in silicalite have been recently reviewed
by Talu et a]. [151. The values obtained by several different techniques are shown in fig.
6. It is clear that for these species the discrepancies between different measurements
amount to several orders of magnitude. We have measured diffisivities for a series of
linear alkanes in several different zeolites by the ZLC technique. In all cases we see a
continuous but slow decline at higher carbon numbers with no evidence of local maxima
as shown by the membrane data. For offi-etite-erioniteour results agree well with those
of Maghalaes et al. [34] and show no evidence of the window effect originally
suggested by Gorring [35]. It should be noted that the membrane technique measures
diffusion along the long axis of the crystal whereas other methods measure an average
diffisivity. However the long axis diffisivity difisivity is expected to be lower so the
difference between membrane, FR and ZLC data is puzzling.

8 Concluding Remarks

The diffisivities predicted from molecular simulation are generally too high. For small
rapidly diffusing species the values are comparable with the highest experimental values
but for slower (larger) species the simulation values are higher than even the largest of the
experimental values. For example, it is clear from fig. 6 that the simulation results do
not capture the decreasing trend of difiivity with carbon number which is observed by
all experimental techniques.
In general the shorter range measurements yield higher diffisivity values than the
longer range measurements and this difference becomes greater for larger molecules.
The different microscopic methods commonly show fair agreement but differences
between different macroscopic measurements are often very large.
In a recent experimental study of the adsorption of methanol in a large crystal of
CrAPO by interference microscopy, Lehmann et al. [36] observed that, even at
equilibrium, the distribution of sorbate through the crystal is far from uniform. It seems
clear that access is controlled largely by the defect structure and the growth planes of the
crystal. This observation may provide a plausible explanation for the discrepancies
observed between different diffusion measurements. The impact of the defect structure

21
is minimal over short distances (a few unit cells) but becomes important as the length
scale of the measurement is increased. Furthermore, one might expect that the effect of
defects will be more severe for larger sorbates and the defect structure may differ
substantially between different samples, thus explaining the discrepancies between
different microscopic measurements.

Notation

B intrinsic mobility D' see Eq. 7


c concentration (gas or fluid phase) F purge flow rate
c, initial value of c J flux (relative to adsorbent framework)
D difisivity (transport) K Henry's Law constant (dimensionless)
Do corrected diffusivity L F12/KV,D - dimensionless parameter
D tracer or self-diffusivity 1 half thickness of adsorbed slab
n dimensionality of pore system u velocity in Maxwell Stefan
expression (Eq. 4)
p partial pressure of sorbate V,,V, Volume of gas (dead volume),
volume of solid
q adsorbed phase concentration z distance
qs
- saturation limit P parameter in Eq. 8
r mean square displacement Y V&V,
R gasconstant 8 q/qs hctional loading
t time h step length (random walk)
T temperature (K) T dimensionless time M/l2 Eq. 8;
time between molecular jumps Eq. 2

References

1. M. Eic and D.M. Ruthven, Zeolites 8,4045 (1988).


2. D.M. Ruthven and M. Eic, Amer. Chem. SOC.Symp. Ser. 368,361-375 (1988).
3. L.S.Darken, Trans. A.I.M.E. 175,184(1948).
4. J.C. Maxwell, Phil. Trans. Roy SOC.m , 4 9 - 7 9 (1866).
5. J. Stefan, Bet. Akad. Wissen. Wien 65,323-363 (1872).
6. R. Krishna, Chem. Eng. Sci. a,
1779 (1990).
7. R. Krishna, Chem. Eng. Sci. 48,845 (1993).
8. D. Paschek and R. Krishna, Chem. Phys. Letters, 333,278-284.
9. J. Ktirger, Single File Diffusion, Ch. 7 in Molecular Sieves: Science and
Technology, H.G. Karge and J. Weitkamp eds., Wiley-VCH (2002).
10. D.M. Ruthven and M. Post, Diffusion in Zeolite Molecular Sieves, Ch 12 in
Introduction to Zeolite Science and Practice, H. van Bekkum, E.M. Flanigen, P.A.
Jacobs and J.C. Jansen eds. Elsevier, Amsterdam (2001).
11. J. Ktirger and D.M. Ruthven, Diffusion and Adsorption in Porous Solids, Ch. 5 in
Molecular Sieves: Science and Technology, H.G. Karge and J. Weitkamp eds.,
Wiley-VCH (2002).

22
12. J. Kager and D.M. Ruthven, Di#&on in Zeolites and other Microporous Solids,
John Wiley, New York (1992).
13. D.M. Ruthven and S. Brandani in Recent A&. In Gas Sep. by Microporous Ceramic
Membranes,N.K. Kanellapoulos ed., p. 187-212 Elsevier, Amsterdam (2000).
14. D.T. Hayhurst and A. Paravar, Proc. Sixth Internat. Zeolite Con$, p. 217-224,Reno
(1 983)D. Olson and A. Bisio eds. Butterworth, Guildford (1 984).
15. O.Talu,M.S.SunandD.B.Shah,A.I.ChEJl.~,681-694(1998).
16. W. Niessen and H.G. Karge, Microporous Materials L, 1-8(1993).
17. R. Schurnacher and H.G. Karge, Microporous Materials 1,1-8(1993).
18. W.O. Haag, R.M. Lago and P.B.Weisz, Faraday Discussions Chem. SOC. 72, 317-
330 (1982).
19. M.F.M. Post, J. Amstel and H.W. Kouwenhoeven, Proc. Sixrh Internat. Zeolite Con$
p. 5 17-527,Reno, (1983)D. Olsson and A. Bisio eds. Butterworth, Guildford (1 984).
20. L.V.C. Rees and L. Song in Recent Advances in Gas Separationby Microporous
Ceramic Membranes p. 139-186, N.K. Kanellopoulos ed., Elsevier, Amsterdam
(2000).
21. D.D. Do, X. Xu and P.L.J. Mayfield, Gas. Sep. PuriJ 5,35-48(1991).
22. J.R. Ebner and J.T. Gleaves, U.S. Patent 4,626,412(1986).
23. J.T. Gleaves, J. R. Ebner and T.C. Kuechler, Catol. Revs. Sci. Engg. 30, 49-116
(1988).
24. U. Schemmert, J. Kiirger, C. Krause, R.A. Rakoczy and J. Weitkarnp, Europhys.
Letters 46,204-210 (1 999).
25. U. Schernrnert, J. K&ger, J. Weitkarnp, Microporous and Mesoporous Materials 32,
101-110 (1999).
26. H. Jobic in Recent Advances in Gas Separation by Microporous CeramicMembranes
p. 109-137 N.K. Kanellopoulos ed. Elsevier, Amsterdam (2000).
27. D.M. Ruthven, Zeolites 13,594(1993).
28. S. Brandani, D.M. Ruthven and J. Karger, Zeolites l5,494-496(1995).
29. S.Brandani, J.R. Hufion and D.M.Ruthven, Zeolites l5,624-631 (1995).
30. S. Brandani, Z. Xu and D.M. Ruthven, Microporous Materials 3, 2791-2798
(1 998).
3 1. S.Brandani, M. Jama and D.M. Ruthven, Microporous and Mesoporous Mats. 35-36
283-300(2000).
32. D.M. Ruthven, M. Eic and E. Richard, Zeolites ll,647-653(1991).
33. L. Song and L.V.C. Rees, Microporous and Mesoporous Materials 35-36 301-314
(2000).
34. F.D. Magalhaes, R.L. Lawrence and C.W. Conner, AIChE Jl. a,68-86(1996).
35. R.L. Gorring, J. C a t a l s , 13 (1973).
36. E. Lehmann, C. Chmelik, H. Scheidt, S. Vasenkov, B. Staudte, J. Kgrger, F. Kuemer,
G. Zadvozna and J. Kornatowski, J. Am. Chem. SOC.- in press.

23
c

Fig. 1. Schematic diagram of system for vapor phase ZLC or tracer ZLC
measurements.

0.1

0
0
\
0

0.01

0 2 4 6 8 10 12

t (9
Fig. 2. ZLC response curves for C G h m a piece of carbon monolith at 303K. The
four curves obtained at flow rates h m 11.4 Mmin &=l) to I 0 4 d m i n
(L=9.5) are fitted with the m e parameters: D/12 = 0.92 sec-. KV,= 0.24 ml,
Vf-0.14 ml.

24
:eIr

2-
I.,
1i

Fig. 3 Comparison of ZLC and PFGHMR diffosiviticsfor (a) C& and k


(Do and al (271 and (b) CHjOH-NaX (61[ZS].

n
I

m
Y.

Fig. 4 (a) Comparisonof PFG hMR data (325K)and mcer ZLC &fa
(358IQ for propane and propene inNaX oeoiite crystals [29].
(b) Cmpakon of PFG W R , FR and mer ZLC difbivities for
bepane-wax at 468K (201.

25
1 E-07

Benzene

u) l.E-08 .
*-
E x +- F R a . 1 1331
0- D = S . 2 6 ~lo- c--~
b
e 1.E-09 .

0 NZLCTrmsD.
0 TZLCTracera
A NZLC Richud De(1991)
1.E-10
1.9 2.1 2.3 25 2.7 2.9 3.1 33
4ooofr
Fig. 5 krhenius plot showing comparison between ZLC and FR
diffusivitits for bmzene in silicalite. Data from 116 and 31-331

D or o0 i m s-

10%

109
1o-O
lo-
10-

10-3

0 2 4 6 8 10 12 14 16 18
Number of carbon atoms

Fig. 6 Variation of diffisivity with carbon number for n-paraffins in


silicalite at 300K showing comparison between data obtained by
different techniques: (0)M.D.simulations, (e) hierarchical
simulation, (+)
QENS, (V)skgie crystal membrane. (A)PFG
NMR,(A)ZLC [15,26J.

26
ORDERED MESOPOROUS CARBONS WITH NEW OPPORTUNITIES
FOR ADSORPTION STUDIES

R.RYOO AND S . H.JOO


Centerfor Functional Nanomaterials, Department of Chemistry (School of Molecular Science-
BK21),Korea Advanced Institute of Science and Technology,Daejeon. 305 -701, Republic of
Korea.
E-mail:rryoo@mail.kaist.ac.kr

Mesoporous carbons with various structures were available in recent years, due to the development
of the synthesis route using mesoporous silica templates. The synthesis of these mesoporous carbons,
designated as CMK-1-5, was performed by carbonization of sucrose, furfury1 alcohol or other
suitable carbon sources inside the template pore systems. The mesoporous materials were obtained
in a high-purity carbon state after the silica or aluminosilicate templates were removed with NaOH
or HF solution. The structures of these carbons, e.g., hexagonal p6m. cubic Pm3n and cubic 141/u,
correspond to faithful replication of the silica mesopores to carbon frameworks. The ordered
arrangement of the nano-structured carbon frameworks imparts the carbon materials with uniform
mesoporosity. The X-ray diffraction patterns of these carbon materials exhibit several Bragg
diffractions at small angles, similar to those for the MCM-41-type silica templates. The pore
diameters are uniform and tailorable, typically, in the range between 2 and 10 nm. These carbons
show many new possibilities for applications to adsorbents, catalysts and electrode materials.
Particularly for adsorption studies, it is expected that the well-defined mesoporous structures would
be also useful as a reference pore system for the development of characterization methods and
theoretical modeling for the adsorption on carbon surfaces.

1 INTRODUCTION

The microporous (pore diameters less than 2 nm) and mesoporous (2-50 nm) carbons are
widely used in many areas of modem science and technology, including water and air
purification, gas separation, catalysts preparation, chromatography, fuel storage, and
manufacturing of electrochemical devices [ 1,2]. The wide-spread use results from their
high specific surface areas, large pore volumes, ability for hosting catalytic components
(e.g., metal particles), chemical inertness, good mechanical stability, high electrical
conductivity, and excellent affinity with organic species. The technological importance
has led to many scientific researches on the adsorption of guest species, their diffusion
into the pore systems, and the interaction with the carbon frameworks. Many works on
experimental measurements and theoretical modeling are still being actively performed in
order to characterize accurately the adsorption phenomena as a function of the pore
diameters, shapes and pore-wall structures.
For the accurate characterization of the adsorption phenomena, it is necessary to
obtain accurate information on pore structures. However, most of ordinary microporous
carbons and mesoporous carbons are obtained with amorphous structures that are
characterized by irregular arrangements of non-uniform pores. X-ray (or electron)
diffraction (XRD) techniques are not useful for such carbons because there are no well-
defined structural factors to correlate with the adsorption behavior. Moreover, porous
carbons exhibit wide varieties of the surface functional groups and the thickness of the
pore walls, depending on the details of the synthesis conditions. The lack of distinct XRD
lines makes it difficult to distinguish structural differences between samples which causes
many works to depend empiricaIly on specific samples.

27
Recently, we have discovered a synthesis route to highly ordered mesoporous carbon
molecular sieves using mesoporous silica templates with various structures such as MCM-
48 (bicontinuous cubic Zu34, SBA-1 (cage-type cubic Pm3n) and SBA-15 (channels
arranged in a hexagonal p6m structure) [3-71.The mesoporous templates are synthesized
via the synthesis route using surfactants [8-lo]. The carbon synthesis procedure consists
of the impregnation of an organic carbon precursor such as sucrose, furfury1 alcohol or
acetylene gas into template pores, subsequent pyrolysis of the precursor and removal of
the template frameworks. The carbon materials designated as CMK-n are released after
the removal of the silica template with NaOH or HF. CMK stands for carbon
mesostructured by KAIST. Several XRD lines appear below 10 due to the mesoscale
structural regularity corresponding to that of the silica templates.
The synthesis of the CMK-n carbons is controlled to various pore shapes,
-
connectivity, diameters (typically, 1 10 nm in diameter) and pore wall thickness. These
carbons exhibit high specific surface areas (typically, the BET specific surface areas up to
2000 m2g-),uniform pore diameters, large adsorption capacities, and high thermal, acid-
base and mechanical stabilities. The CMK-type carbons are also suitable for the formation
of well-defined nanocomposite with organic polymers, so that the nanopore walls can be
modified with various functional groups. These carbons show new possibilities for various
applications in adsorption, catalysis and electrochemistry.
Particularly for adsorption studies, it is expected that the well-defined mesoporous
structures would be useful as a reference pore system for the development of
characterization methods and theoretical modeling for the adsorption in carbon pores. In
this regard, we briefly review on the synthesis strategy, structure characterization, and
their perspectives.

2 SYNTHESIS METHOD

The principle of the carbon synthesis is shown in Fig. 1. Suitable carbon sources such as
sucrose, furfuryl alcohol, phenol-resin monomers and acetylene gas are converted to
carbon frameworks inside mesoporous silica template by pyrolysis. An effective method
for the restriction of carbonization to inside the template is to incorporate a suitable
catalyst such as Al, Sn and Fe onto the silica pore walls prior to the use as template. The
template after the carbonization is removed using ethanol-water solution of HF or NaOH.
As shown in Fig. 1, rod- or tube-type carbons are obtained depending on the synthesis
conditions. Rod-type carbons are prepared if cross-linkable carbon precursors such as
sucrose and phenol resin are carbonized after the template pores are filled with the carbon
sources [3-51. Carbon deposition outside the template can be prevented under the present
carbonization conditions using catalyst. Moreover, the carbon formation can be controlled
to occur uniformly throughout the entire volume of the template pore system. Normally,
the conversion of the organic compounds leads to a significant decrease in volume, and
this is accommodated by the generation of micropores in the carbon nano-frameworks.
The micropore volume depends on the source of carbon and the details of carbonization
conditions such as vacuum, nitrogen flow and heating rate.

28
Figure 1. Schematic representation of the templated synthesis route using mesoprous silicas.

Tube-type carbons are obtained when organic compounds are carbonized in a thin-
film state on the template pore walls 161. The tube-type carbon can be obtained even after
the entire volume of pores is filled with carbon source, if the excess carbon source is
removed before the carbonization is completed. For example, cylindrical pores are
generated along the center of the carbon frameworks due to the systematic volume
decrease when furfuryl alcohol is pyrolyzed under vacuum after the initial polymerization.
Alternatively, the tube-type carbons can be synthesized as follows: carbonization can be
controlled to occur partially by catalyst at the pore walls at moderate temperatures. The
remaining carbon source is removed by evacuation, and the carbonization is completed by
pyrolysis at high temperature. Chemical vapor deposition on the pore walls can also be
used to produce the tube-type carbon [ 111 as well as the aforementioned rod-type carbons
[12]. The structure of the resultant carbon depends on the thickness of the carbon
deposition.

3 STRUCTURES AND NITROGEN ADSORPTION PROPERTIES


3.1 CMK-I
CMK-1carbon was the first carbon material reported to exhibit well-resolved XRD lines
characteristic of ordered arrays of carbon mesopores [3]. The synthesis of the carbon was
achieved by carbonization of sucrose inside the MCM-48 mesoporous silica. As shown in
Fig. 2, the XRD pattern exhibits a new diffraction line around 1.4. compared with its
MCM-48 template. This change can be explained by the formation of two separate carbon
networks in the bicontinuously mesoporous MCM-48 template. After the separating silica
frameworks are removed, the two carbon networks join together. The joining of the two
carbon networks attributes to the symmetry change from cubic Zu3d to either 14,h or
lower [12]. The new ordered mesoporous structure is indicated by the XRD pattern and
transmission electron microscopic image shown in Fig. 2.

29
2 4 6 8 1 0
XI (deg-1
Figure 2. (a) XRD patterns for MCM-48 silica template and the CMK-1 carbon synthesized using the MCM-48
template. (b) Transmission elechon microscope image of CMK-I . (c) Scanning electron microscope image of
CMK-I .

The synthesis of carbon with MCM-48 was also reported by another research group
using phenol-formaldehyde resin, following the report on CMK-1 [ 131. However, this
carbon exhibited the same XRD pattern as CMK- 1.
As shown in Fig. 3, nitrogen adsorption isotherms of CMK-1 feature well-pronounced
capillary condensation steps similar to those of ordered mesoporous silicas and indicative
of high degree of mesopore size uniformity. The isotherms reveal that the CMK-1 carbon
has high nitrogen BET specific surface area (1500-1800 m2 g-I), and large total pore
volume (0.9-1.2 cm3 g-') [14]. The adsorption capacity is comparable or larger than that of
MCM-48 template. The pore-size analysis (calibrated BJH analysis) shows that typical
CMK-1 has uniform mesopores about 3 nm in size, which is accompanied by a certain
amount of micropores when sucrose is used as the carbon source.

30
0.0 0.2 44 0.6 48 1.0
Relativepressure

Figure 3. Nitrogen adsorption isotherm for CMK-1. (inset) Correspondingpore size distribution obtained from
adsorption branch by calibrated BJH method [14].

3.2 Other Rod-Type CMK Carbons


Other rod-type carbons with various structures have been synthesized with the SBA- 1,
SBA-15 and SBA-16 (large cage-type, cubic Im3m) mesoporous silica templates. These
carbons (designated as CMK-2 [4], CMK-3 [5] and CMK-6 [15, 161, respectively) exhibit
very similar XRD patterns to those for their silica templates, as shown in Fig. 4. This
result indicates that the structures are maintained in the same space groups during the
synthesis of the carbons from silicas, unlike the case of the CMK-1 synthesis. It is
therefore reasonable that these silica templates are composed of 3-dimensional (3-D)
mesoporous networks of the same continuity. The carbon synthesis within the mesoporous
networks of the same continuity gives the CMK-2, CMK-3 and CMK-6 mesoporous
carbons corresponding to faithful replication of the template pore systems.
It is also noteworthy that, despite the apparently same 2-D hexagonal structures, the
SBA-15 and MCM-41 silicas are distinguished by their markedly different pore
connectivity in addition to the difference in the pore diameters. The MCM-41 silica has 1-
D channels that are not interconnected. However, the large 1-D mesoporous channels
(typically, 9 nm in diameter) of the SBA-15 silica are interconnected through the so-called
complementary pores, which are around or less than 3.5 nm in diameters, and randomly
located perpendicular to the 1-D channels. Because of the 3-D channel structure, the
structure of the SBA-15 silica can be converted to the negative carbon replica exhibiting
the same kind of structural symmetry. Accordingly, the structure of the CMK-3 carbon is
composed of a hexagonal arrangement of 1-D carbon rods as shown by the structural
model in Fig. 4(c). On the other hand, replication of the MCM-41 silica with carbon
results in the formation of carbon fibers that do not retain the 2-D hexagonal arrangement
r141.

31
Figure 4. XRD patterns for rod-type carbons: (a) CMK-2, (b) CMK-3. (c) Structural model for CMK-3.

3.3 Tube-Type Carbons


CMK-5 is the first example of the ordered tube-type mesoporous carbons that can be
characterized with well-defined Bragg diffractions by ordinary XRD instrument [6]. The
XRD pattern of the CMK-5 carbon is distinguished from that of CMK-3 by the much
lower intensity of the (100) diffraction. The structure of CMK-5 may be described by the
substitution of the carbon nanorods in CMK-3 with nanopipes. The CMK-5 carbon is
synthesized using SBA-15, similar to CMK-3, but the carbon source and synthesis
condition are somewhat different from those for CMK-3. The synthesis method for the
tube-type carbon can be extended to the SBA-16 mesoporous template. The resultant
CMK-7 carbon has a bicontinuous mesoporous structure [ 151.
It is noteworthy that the pore-size distribution curve obtained by the N2 adsorption
has exhibited two sharp peaks with the maxima corresponding to the inside diameter of
the carbon nanopipes (typically, 5.5 nm) and the pores formed between the adjacent pipes
(4.2 nm), respectively [6]. It is reported that the outside diameter of the nanopipes is
tailored by the pore diameter of the template SBA-15, while the wall thickness of the
carbon nanopipes are also controllable to a certain degree. The specific BET surface area
of the CMK-5 varies from 1500 to 2200 m2g-'depending on wall thickness [ 171.

4 PERSPECTIVES

The methods developed for the synthesis of ordered mesoporous carbons are simple and
cost-efficient, and the pore size can be tailored. The synthesis process can be scaled up for
production in bulk quantities. Recent works on the synthesis of mesoporous silicas
brought about much improvement in the cost-efficient and custom-tailored synthesis of the
templates [18]. The discovery of new mesoporous silicas is also expected to provide
additional promising templates for the synthesis of new mesoporous carbons. The
resulting high-surface-area materials with uniform pores promise to be suitable as

32
adsorbents, catalyst supports, sensors, and materials for other advanced applications.
The presence of the distinct XRD patterns provides us new opportunities for precisely
monitoring various physico-chemical phenomena that take place inside the well-defined
carbon pores or at the pore walls such as adsorption, impregnation, framework changes,
formation of metal clusters and grafted functional groups. The materials constructed with
such well defined and controllable pore diameters are suitable as standards or references
for the characterization of porosity of the carbons, similar to the already well-known case
of the MCM-41silica.
In addition to this significance, the carbon frameworks can be grafted with various
organic and organometallic functional groups on the carbon-pore walls [19]. The carbon
can be used to compose systematic nanostructures between organic polymers and carbons
[20]. In addition, there was a recent report on the synthesis of an ordered microporous
carbon using NaY zeolite [21]. This result shows possibilities to extend the templating
route to other zeolite-type carbons with various structures. Furthermore, the template
synthesis methods have been being fully advanced for the synthesis of ordered
mesoporous carbons exhibiting graphite-like atomic orders [22]. These new CMK-nG
carbons will give many new possibilities in the adsorption science and technology, in
addition to the CMK-type mesoporous carbons composed of disordered atomic
arrangement.

References

1. Bansal C. R., Donnet J.-B. and F. Stoeckli, Active carbon, (Marcel Dekker, New
York, 1988).
2. Foley H. C., Microporous Muter. 4 (1995) pp. 407-433.
3. Ryoo R., Joo S. H. and S. Jun, J. Phys. Chem. B 103 (1999) pp. 7743-7746.
4. Ryoo R., Joo S. H., Kruk M. and Jaroniec M., Adv. Muter. 13 (2001) pp. 677-671.
5. Jun S., loo S. H., Ryoo R., Kruk M., Jaroniec M., Liu Z., Ohsuna T. and Terasaki 0..
J. Am. Chem. Soc. 122 (2000) pp. 10712-10713.
6. Joo S. H., Choi S. J., Oh I., Kwak J., Liu Z., Terasaki 0. and Ryoo R., Nature 412
(2001) pp. 169-172.
7. Ryoo R., Joo S. H., Jun S., Tsubakiyama T. and Terasaki O., Stud. Surf: Sci. Cutul.
135 (2001) p. 150.
8. Kresge C. T., Leonowicz M. E., Roth W. J., Vartuli J. C. and Beck J. S., Nature 359
(1992) p. 710-712.
9. Huo Q., Margolese D. I., Ciesla U., Feng P., Gier T. E., Sieger P., Leon R., Petroff P.
M., Schuth F. and Stucky G. D., Nature 368 (1994) pp. 317-321.
10. Zhao D., Huo Q., Feng J., Chmelka B. F. and Stucky G. D., J. Am. Chem. Soc. 120
(1998) pp. 6024-6036.
11. Bang W.-H., Liang C., Sun H., Shen Z., Guan Y., Ying P. and Li C., Adv. Muter. 14
(2002) pp. 1776- 1779.
12. Kaneda M., Tsubakiyama T., Carlsson A., Sakamoto Y., Ohsuna T., Terasaki O., Joo
S. H. and Ryoo R., J. Phys. Chem. B 106 (2002) pp. 1256-1266.
13. Lee J., Yoon S., Hyeon T., Oh S. M. and Kim K. B., Chem. Commun. (1999) pp.
2177-2178.
14. Kruk M., Jaroniec M., Ryoo R. and Joo S. H., J. Phys. Chem. B 104 (2000) pp. 7960-
7967.
15. Ryoo R. et al., manuscript in preparation.

33
16. Yu C., Stucky G.D. and Zhao D. In Abstracts of 3rdInternational Mesostructured
Materials Symposium, July 8- 11,2002, P A 4 P. 54
17. Ryoo R. et al., unpublished results
18. For example: Kim S. S.,Pauly T. R. and Pinnavaia T. J., Chem. Commun., (2000) pp.
1661- 1662.
19. Jun S., Choi M., Ryu S.,Lee H.-Y. and Ryoo R., Stud. Sulf: Sci. Card. submitted
20. Choi M. and Ryoo R., manuscript in preparation.
21. (a) Ma Z., Kyotani T. and Tomita A., Chem. Commun. (2000) pp. 2365-2366. (b) Ma
Z., Kyotani T., Liu Z., Terasaki 0. and Tomita A., Chem. Muter. 13 (2001) pp. 4413-
4415.
22. Kim T.-W., Park 1 . 4 . and Ryoo R.,manuscript in preparation.

34
QUANTUM MICROPORE FILLING AND ITS APPLICATION POSSIBILITY

T. TANAKA, Y. HA'ITORI', K. MURATAS, T. KODAIRA**, M. YUDASAKA*,


S.IIJIMA*+, AND K. KANEKO
Department of Chemistry,Faculty of Science and Centerfor Frontier Electronics and Photonics,
Chiba University I-33 Yayoi, inage, Chiba 263-8522,Japan.

#)Instituteof Research and Innovation, 1201 Takada, Kashiwa, Chiba 277-0861, Japan

*I Japan Science and Technologv Corporation, c/o NEC Corporation, 34 Miyukigaoka, Tsukuba
305-850 I , Japan.

+) Department of Physics, Meijo Universiv, I501 Shiogamaguchi, Tenpaku7


Nagoya 468-8502, Japan

**) Nanoarchitectonics Research Center, National tnstitute of Advanced Industrial Science and
Technology,AiST Tsukuba Central 5, 1-1-1 Higashi
Tsukuba, Ibaraki 305-8565, Japan
Adsorption isothenns of Ne on AIPO4-5 and of H2 and D2 on single wall carbon nanohom (SWNH)
were measured at 27 K and 20 K, respectively. The comparison of the experimental Ne adsorption
isotherm at 27 K with that calculated from classical density functional theory (DFT)showed the
presence of quantum effect. The comparison of the experimental isotherm of HZon SWNH at 20 K
with the isotherm simulated by classical grand canonical Monte Carlo simulation suggested the
presence of more remarkable quantum effect than Ne on AIP04-5. Adsorption isothenns of HZand
Dz on SWNH at 20 K were compared with each other. Although both isotherms are very similar in
shape, the adsorption amount of 9is larger than that of H2 over the whole PPo range. This result
was ascribed to a marked quantum effect of H2 molecules, indicating the possibility of quantum
molecular sieving by nanocarbons.

1. Introduction

Recent research activities on nanoporous materials have stimulated hdamental studies


on adsorption mechanism in micropores [ 1 - 51. Both of the precise measurement of
high resolution adsorption isotherms fiom the low P/Po region and molecular simulation
showed the presence of monolayer adsorption on the micropore walls and further filling
in the residual spaces after monolayer completion for supermicropores (0.7 nm < pore
width w < 2 nm) ; the contribution by the monolayer to the filling in the residual spaces is
comparable to that by the pore walls [6- 101. Systematic researches on activated carbon
fiber (ACF) having slit-shaped micropores[ 1 1,121 have contributed to elucidation of the
mechanism of micropore filling to develop better adsorbents in adsorption and separation
engineering.
In case of micropore filling in ultramicropores (w < 0.7 nm), the monolayer adsorption
itself is highly enhanced. As this micropore filling in ultramicropores begins from an
extremely low pressure and is sensitive to the molecular size, the ultramicroporous
materials have been widely applied to separation engineering as molecular sieves.
Although it has been believed that the molecular sieve characterisitics arises from the
geometrical fitting of the pore to the molecule, the molecular sieving mechanism is not
necessarily well unveiled[l3 - 151. Accordingly we need to elucidate the molecular

35
sieving effect more to establish a more efficient separation technology. At the same time,
a new possibility of the molecular sieving should be challenged.
Recently nanocarbons such as single wall carbon nanotube (SWNT)[16,171 or single
wall carbon nanohorn (SWNH) have offered an important possibility for hydrogen and
methane storage [19-221. Although reliable adsorption data are not available for
previous SWNT samples[23,24], highly pure samples of SWNH is going to provide a
precise understanding of their storage potential [22]. In particular, nano-scale windows
can be donated on the wall of SWNH and thereby micropore filling in internal and
interstitial nanospaces can be clearly separated experimentally. Hence, SWNH leads to a
better understanding of adsorption properties of nanocarbons. Theoretical researches on
hydrogen storage by SWNT predicted the important contribution of quantum effect due to
the small mass of the hydrogen molecule [23-251. At the same time, these studies
suggested that the quantum effect should be taken into account in general nanoporous
systems and some experimental results on zeolites were discussed from the basis of this
quantum effect [26]. According to these theoretical studies, the zero-point vibration
cannot be negligible for a light molecule and the quantum molecule behaves as if it had a
larger size than the classical molecule. Some papers on quantum effect in nanoporous
systems were published before the above studies. Kaneko et a1 applied a simple
rectangular box model for He adsorption in slit-shaped graphitic nanospaces of ACF,
suggesting the presence of quantum excitation of translational motion of He atoms in the
motion perpendicular to the pore-walls [27]. Beenakker et a1 [28] proposed the possibility
of quantum sieving using hard spheres in a square-well cylindrical tube.
As D2 is a quite important gas in the present industry, an efficient separation of D2 fiom
H2 has been requested. If the quantum effect can be applied to the separation, a new
separation technology should be introduced. Tanaka et a1 examined the temperature
dependence of Ne adsorption on well-crystalline A1P04-5, showing the presence of an
explicit quantum effect [29]. Murata et a1 and Tanaka et al studied hydrogen adsorption on
SWNH over the wide temperature range from the boiling temperature to 303 K [30]. This
paper will review the quantum effect in micropore filling of Ne, H2, and D2 on A1P04-5,
nanocarbons, and ACF.

2. Classical DFT calculation, classical GCMC simulation and effective potential for a
quantum molecule

The molecular potential of a quantum molecule in a model SWNT is described here. For
simplicity, we assumed a homogeneous cylindrical pore for a model of open-ended
SWNT. Thus classical solid-fluid interactions can be calculated using the Lennard-Jones
(LJ) potential integrated over an infmitely long cylinder [3 I]:

&J4=&,E,fQsf
2

[
-
32
63 F(-4.5,-4.5,1.0;P2)
[R*(l-P2)]'o

-3

36
where F( a,p, y ; x ) is a hypergeometric function, R and pr are the radius of the pore
and the density of solid atoms in the pore wall (38.21m-~).The interaction parameters for
H2, ad and Gp/R are 0.3 18nm and 32.1 K, respectively.
The quantum correction to the classical potential can be calculated by Feynmans
procedure, which employs an effective potential [32,33].

where r is the vector between two particle, 1 = h2/(6mk7)and m is the molecular mass of
hydrogen isotopes. Fig. I shows the effective potentials of hydrogen H2 and deuterium D2
inside the (3,6) nanotube (d = 0.62 nm) at 20 K fiom eq. (3). There is a large difference
in effective potentials for H2 and D2.This suggests that hydrogen is easily excluded from
hydrogen isotopes mixture with the (3,6) nanotube at 20 K. Therefore, such a nanotube
with the small diameter is predicted to exhibit very large seectivity and the other porous
materials that have pores corresponding to the size of the (3,6) nanotube should be
suitable for the hydrogen isotope separation.
The Ne adsorption isotherms on model A1P04-5 micropores were calculated fkom thc
Tarazonas version of the nonlocal density functional theory [34,35] which has beer
actually applied to the study on micropore filling [36,37]. The necessary parameters werc
obtained from the adsorption isotherms of Ne on A1P04-5 at 27K and 30K in a low
pressure range.

Fig. 1 Effective potentials of hydrogen isotopes inside SWNT at 20 K for (3,6)SWNT.


The solid line and (+) symbols denotes hydrogen and deuterium, respectively.
The adsorption isotherms of N2at 77 K and classical H2 at 20 K were calculated wi
classical GCMC simulation for the SWNH aggregate. Here we approximated SWNH 1

37
SWNT. GCMC simulation was carried out using the established procedures. We used the
12-6 Lennard-Jones(LJ) potential for the N2-N2 interaction. The used LJ parameters for N2
are E~ I kB= 95.2 K and = 0.375 nm. As to H2, -1 kB= 36.7 K and aff= 0.2959 nm
were used. The classical solid-fluid interactions given by eq. I were used for the
simulation. We used an established technique of the repeated cell determined by the
triangular and square arrays of SWNHs [7,9,38 - 401. The thickness of the graphene wall
was assumed to be 0.34 nm.

3. Adsorption measurement below 30 K

The cryogenic adsorption system was specially developed to measure adsorption


isotherms of H2 and D2. This system is equipped with a closed helium cycle two-stage
Gifford McMahon refiigerator to operate under cryogenic conditions. The adsorption
temperature can be kept constant within f 0.03 K at 20 K. Adsorption isotherms are
obtained by gas adsorption manometry. This method is based on the measurement of the
gas pressure in a calibrated, constant volume, at a known temperature. The dead space
volume was calculated tiom a helium calibration measurement at the temperature of
interest. Thermal transpiration effect was calibrated according to the work by Takaishi
and Sensui [4 11.

4. Quantum effect in Ne on AIP04-5

A1Po4-5 has one- dimensional pores of wider and narrower parts whose widths are
1.137 and 1.002 nm, respectively. Also well-crystalline samples are available to provide
reliable experimental data. The framework of A1Po4-5 consists of alternate tetrahedral
aluminum and phosphorous atoms bridged by oxygen atoms and thus is electricallyneutral.
The pores of AIP04-5 are not interconnected and form one-dimensional channels parallel
to the crystallographic c -axis [42].
The adsorption isotherms were calculated using DFT based on the classical potential
for narrower pore (radius R of structure = 0.501nm: AP-50) and the average size pore
models (R = 0.5347nm: AP-53). The adsorption increased steeply below P/Po = lo4,
being almost saturated above P/Po = The higher the measuring temperature, the
larger the rising PIPo. The DFT isotherms for both models are completely different from
each other. AP-53 model gives the two-step isotherm, whereas AP-50 model leads to the
single step isotherm which coincides with the experimental one, as shown in Fig.2.
Although the average pore model of AP-53 is expected to be fit for description of the
experimental result, the narrower pore model of AP-50 is better after the correction of the
pore volume by Ar adsorption.
The most probable cause for this marked discrepancy from the prediction is the
quantum effect. The quantum effects are included by replacing parameters E and 0 in V,,
with temperature and A-dependent parameters E and d (where A = h/a(m&).)).The
parameters E and d can be estimated by Feynmans effective potential method. For
quantum neon pairs, the location of the minimum of the effective potential is shifted to
26d and the value at the minimum to 4,and the quantum contributions are dla=
1.014 and &I& = 0.947 at 27 K, respectively. Therefore, the quantum effect decreases the
fluid-fluid interaction. That is, quantum Ne can behave as if they had a bigger size and the

38
moelecule-wall interaction becomes weak. Then, quantum Ne can be adsorbed
monolayerly on the pore-wall, as if the classical Ne molecules do in the narrower
micropores.

Fig.2 Experimental adsorption isotherm of neon on AIP04-5 at 27 K ( 0 ) and theoretical


adsorption isotherms: AP-50 model (+ ); AP-53 model (a). The dashed line denotes the
corrected experimental adsorption isotherm.

5. Micropore structures and quantum effectof SWNH

Nitrogen adsorption isotherm on as-grown SWNH was measured at 77 K to


characterize the SWNH assembly structure (Fig. 3). Then, theoretical nitrogen adsorption
isotherms on several SWNT models at 77 K were calculated by GCMC simulations
treated the classical Lennard-Jones systems to understand the experimental data. The
model adsorption spaces are interstices of triangular-packed and square-packed arrays of
SWNTs (the pore width at the nuclear position d = 3 nm) and external surface of an
isolated SWNT (d = 3 nm), respectively. Simulated isotherms on the three models are also
shown in Fig. 3. While the triangular array model cannot describe the experimental
isotherm in the low PDo range, the square array model well coincides with the
experimental data over a wide P/Po range. This suggests that SWNHs are not perfectly
aligned as the triangular (close-packed) model, but they are roughly assembled in square
arrays. However the square array model cannot reproduce the experimental isotherm in a
high P/Po range. This remarkable upward deviation of the experimental data from the
square model calculation should come from the multi-layer adsorption on the external
surfaces of nanohorns.

39
10-10 lo* 106 lo4 10-2 lo0
PRO

Fig. 3. Experimental adsorption isotherm of N2 on SWNH at 77 K (0)and simulated isotherms


of N2on three models at 77 K: triangular array (+ symbols); square array (0);
adsorption on
the external surface of an isolated model ( A ) .

04

03

c 02
.-
c
a.
b
v1
0
U 01

n- -
10-17 10-6 lo-* 10-11 10-9 10-7 10 10-3 10-1

P/ Po

Fig. 4. Experimental adsorption isotherm of H2 on SWNH at 20 K (0)and simulated


isotherms of H2 on two models at 20 K using the classical GCMC simulations: square
array (0);adsorption on the external surface of an isolated SWNH (A).
Fig. 4 shows comparison between the experimental isotherm of hydrogen on the
as-grown SWNH at 20 K and simulated isotherms for the square SWNT array and the
isolated SWNT models from the classical GCMC simulations. Although the simulated
nitrogen adsorption isotherm for the square. orientation model well agrees with the
experimental one in the low P/Po range, the GCMC-simulated adsorption isotherm of the
classical hydrogen does not coincide with the experimental one. This discrepancy should
stem fiom the predominant quantum effect at 20 K.

6. Quantum molecular sieving effect of SWNH

Experimental adsorption isotherm for D2on the as-grown SWNH at 20 K is compared


with that of hydrogen in Fig. 5. While two isotherms are very similar in shape, the
adsorption amount of D2is larger than that of H2 over the whole P/Po range. As the mass
of D2 is a half of H2,the quantum effect of D2should be less marked than that of H2,as
shown in Fig. 1. Even if the extent of the effective expansion of the H2molecule is only
0.01 nm, such effect is quite sensitive to the filling of molecules in the nanospaces.
Accordingly, H2 molecules cannot be adsorbed in narrower nanospaces to which D2
molecules can be accessible. This adsorption difference between H2 and D2 is a
representative of the quantum molecular sieving. It is expected that SWNH having
nanowindows can show a more remarkable quantum effect for H2 and DZ,providing a
hopeful applicant for hydrogen isotope separation.

25
?
n
20

15
W
E
.-0 10

E
2
71 5
4
rl
0 0.2 0.4 0.6 0.8 I

Fig. 5 Experimental adsorption isotherms of H2 and D2on SWNH at 20 K. H2 ( A ) , D2(0)

7. Acknowledgement

This work was funded by Nanocarbon project from the New Energy and Industrial
Technology Development Organization of Japan.

41
References

1. J. S. Beck, J. C. Vartuli, W. J. Roth, M. E. Leonowicz, C. T. Kresge, K. D. Schmitt,


C. T-W. Chu, D. H. Olson, E. W. Sheppared, S. B. McCullen, J. B. Higgins, J. L.
Schlenker, J. Am. Chem. SOC.,114, 10834 (1992).
2. S. Inagaki, Y. Fukushima, K. Kuroda, J. Chem. Soc., Chem. Commun., 680 (1993).
3. M. Kondo, T. Okubo, A. Asami, S. Noro, T. Yoshitomi, S. Kitagawa, T. Ishii, H.
Matsuzaka, K. Seki, Angew. Chem. Inr. Ed. 38, 140 (1999).
4. R. Ryoo, S.H Joo, S. Jun J. Phys. Chem. B 103,7743.(1999).
5 . E. Bekyarova and K. Kaneko, A h . Muter., 12, 1625 (2000).
6. N.A.Seaton, J.P.R.B.Walton, andN.Quirk, Carbon, 27,853 (1991).
7. C. Lastoskie, K. E Gubbins, N. Quirke, J. Phys. Chem., 97,4786 (1993).
8. K. Kaneko, R. F. Cracknell, D. Nicholson, Langmuir, 10,4606 (1994).
9. R. F. Cracknell, K. E. Gubbins, M. Maddox, D. Nicholson, Acc. Chem. Res. 28 281
(1995).
10. T.Ohba, T.Suzuki, and K. Kaneko, Chem. Phys. Lett., 326, 158(2000).
1 1 . K. Kaneko, Carbon, 38,287 (2000).
12. T. Iiyama, T. Ohkubo, K. Kaneko, In Recent Advances in Gas Separation by
Microporous Ceramic Membranes, Ed. N.K. Kanellopouos, Elsevier, pp.35-66
(2000).
13. Y. D. Chen, R. T. Yang, and P. Uawithya, AIChE Journal, 40,577 (1994).
14. C. Nguyen and D. D. Do, Carbon, 33, 1717 (1995).
15. T. Suzuki, R. Kobori and K. Kaneko, Carbon, 38,630 (2000).
16. S. Iijima,Tkhihashi, Nature, 363,603 (1993).
17. D. S. Bethune, C. H. Kiang, M. S. de Vries, G. Gorman, R. Savoy, J. Vazquez, R.
Beyers, Nature, 363,605 (1 993).
18. S. Iijima, M. Yudasaka, R. Yamada, S. Bandow, K. Suenaga, F. Kokai, F.,
K.Takahashi, Chem. Phys. Lett. 309, 165. (1999).
19. K. Murata, K. Kaneko, F. Kokai, K. Takahashi, M. Yudasaka, S. Iijima, Chem. Phys.
Lett., 331, 14 (2000).
20. K. Murata, K. Kaneko, W.A. Steele, F. Kokai, K. Takahashi, K. Kasuya, K. Hirahara,
M. Yudasaka, S. Iijima, J. Phys. Chem. B 105,10210 (2001)
21. E. Bekyarova, K. Kaneko, M. Yudasaka. K. Murata, D. Kasuya, S. Iijima, Adv.
Muter. 14,973 (2002).
22. K. Murata,K. Kaneko, H. Kanoh, D. Kasuya, M. Yudasaka, S. Iijima, J. Phys. Chem.
in press.
23. A.C.Dillon, K. M Jones,. T. A Bekkedahl,. C. H.Kiang D. S.Bethune, M. J.Heben,
Nature, 386,377 (1997) Nauture, 410,734 (2001)
25. S.R. Challa, D.S. Sholl, J.K. Johnson, Phys. Rev. B 63,245419 (2001).
26. Q. Wang, S.R. Challa, D.S. Sholl, J.K. Johnson, Phys. Rev. Lett. 82,956 (1999).
27. K. Kaneko, N. Setoyama, and T. Suzuki, CharacterizationofPorous Solidr tII,J.
Rouquerol, F. Rodriguez-Reinoso, K. S. W. Sing, and K. Unger eds. Elsevier (1994),
p.593.
28. J. J. Beenakkaer, V. D. Bornman, and S.Yu. Krylov. Chem. Phys. Lett. 232,379
(1 995).
29. H. Tanaka, M. El-Merraoui, T. Kodaira, K. Kaneko, Chem. Phys. Lett. 352,334
(2002).

42
30. H. Tanaka, K. Murata,K. H. Kanoh, D. Kasuya, M. Yudasaka, S. Iiljima, Kaneko, J.
Phys. Chem. to be submitted.
31. G. J. Tjatjouls, D. L. Feke, J. A. Mann, Jr., J. Phys. Chem. 92,4006 (1988).
32. R. P. Feynman, Statistical Mechanics, Benjamin, Reading, MA (1972).
33. G . Stan,M. W. Cole, J. Low Temp. Phys. 41,611 (1980).
34. P. Tarazona, U. M. B. Marconi, R. Evans, Mol. Phys.,60,573 (1 987).
35. P. I. Ravikovitch and A. V. Neimark, J. Phys. Chem. 105,6817 (2001).
36. Mustapha El-Merraoui, M. Aoshha, K. Kaneko, Langmuir 16,4300(2000).
37. K. Murata, Mustapha El-Merraoui, K. Kaneko, J. Chem. Phys. 114,4 196 (200 1)
38. K. R. Matranga, A. L. Myers, E. D. Glandt, Chem. Eng. Sci., 47, 1569 (1992).
39. W. A.Steele, M.Bojan, J. A h . Colloid Interface Sci. 77-76, 153 (1998).
40. T. Ohba and K. Kaneko, J. Phys. Chem. 106,7171 (2002).
4 1. T. Takaishi, Y . Sensui, Trans. Furaday SOC.53,2503 (1 963).
42. J. W. Pichardson Jr., J. J. Pluth, J. V. Smith, Acta Crystullogr. C43, 1469 (1987).

43
ADSORPTION IN MICROPOROUS MATERIALS: ANALYTICAL
EQUATIONS FOR TYPE I ISOTHERMS AT HIGH PRESSURE

A. L. MYERS
Department of'Chemical and Biomolecular Engineering
University of Pennsylvania, Philadelphia PA 19104 USA
E-mail: amyers@seas.upenn.edu

Existing analytical equations for Type I isotherms such as the Langmuir equation and
its modifications fail to describe experimental adsorption isotherms at high pressure.
All of these equations predict that the amount adsorbed increases monotonically with
increasing pressure. Experimental adsorption isotherms attain a maximum value in the
amount adsorbed and then fall to zero. At the critical temperature of the gas, the
maximum in the isotherm occurs at a pressure of about 10 bars and the zero occurs at a
much higher pressure of several hundred bars. For subcritical gases, the maximum occurs
at lower pressure and for supercritical gases the maximum occurs at higher pressure.
The observed behavior of high-pressure adsorption isotherms can be predicted from the
low-pressure (subatmospheric) portion by separating the Type I behavior expected for
absolute adsorption from the experimental behavior observed for excess adsorption.

1 Introduction
Tabulated data for experimental adsorption isotherms are fitted with analytical equations
for the calculation of thermodynamic properties by integration or differentiation. These
thermodynamic properties expressed as a function of temperature, pressure, and composition
are input to process simulators of adsorption columns. In addition, analytical equations for
isotherms are useful for interpolation and cautious extrapolation. Obviously, it is desirable
that the isotherm equations agree with experiment within the estimated experimental error.
The same points apply to theoretical isotherms obtained by molecular simulation, with the
requirement that the analytical equations should fit the isotherms within the estimated
statistical error of the molecular simulation.
Type I isotherms [3]are characterized by an asymptotic approach to a saturation capacity
with increasing pressure. This class of isotherms is most commonly observed for gases or
vapors (water is an exception) adsorbed in zeolites or activated carbon. A typical set of
Type I adsorption isotherms is shown in Fig. 1. Several questions may be asked about sets
of isotherms like these. How is the saturation capacity measured? Is the saturation capacity
a constant or does it decrease with temperature as suggested by Fig. l?
Variation of the pressure P with respect to temperature T at constant loading n is given
by the equation [l]:
IE
nI-[ = -E
where is the differential enthalpy of desorption (isosteric heat). This derivative of the
adsorption isotherms in Fig. 1 along a horizontal line becomes ambiguous at high loading
(n).The differential enthalpy for this system measured by a calorimeter is plotted on Fig. 2.

44
For type I1 isotherms with multilayer adsorption, the differential enthalpy for the second and
higher layers approaches the enthalpy of condensation of the liquid. Does the differential
enthalpy for type I isotherms approach some limit at saturation?
The intention of this paper is to seek answers to these questions.

5
30 C
20 c
4

n 70 C
g3
8
-E2

0
0 0.2 0.4 0.6 0.8 1.o
Pressure (bar)
Fig. 1. Adsorption isotherms of C,& on NaX zeolite. The 2OoC isotherm
was measured experimentally [4]. The other two isotherms were calculated
from Eq. (1) using the differential enthalpies in Fig. 2.

45

35
0 1 2 3 4
Amount adsorbed (moykg)
Fig. 2. Ditferential enthalpy of CzH4 adsorbed on NaX at 2OOC. Points are
experimental calorimetric data [4]. Solid line is polynomial fit of data.

45
2 Adsorption isotherms
A very useful equation for type I adsorption isotherms is 141:

where H is the Henry constant, m is the saturation capacity, and the Ci are virial coefficients
terminated after three or four terms, which are usually sufficient to fit adsorption isotherms
over several decades of pressure. Unlike most adsorption equations, this equation is implicit
in the pressure. This disadvantage is more than offset by the direct connection between the
differential enthalpy and the temperature dependence of the virial coefficients. The main
advantage of this equation is that it can be integrated analytically for the grand potential
(spreading pressure), which is needed for mixture calculations as shown in this mixture
section of this paper.
Experimental measurements yield excess adsorption; molecular simulations calculate ab-
solute adsorption. The relationship between the two variables is given by:

ne = n - V,p (3)
V, is the specific pore volume of the material; typical values are 200-400 cm3/kg for zeolites
and up to 1000 cm3/kg for activated carbon. n is the actual number of molecules contained
in the micropores; the excess adsorption ne subtracts from n the number of molecules which
would have been present in the micropores at the bulk density in the absence of adsorption.
The (oversimplified) case when absolute adsorption is described by the Langmuir equation
and the gas obeys the perfect gas law ( p = P / R T ) has been worked out in detail for the
isotherms and thermodynamic functions (enthalpy, entropy, etc.) [2].
The key step in this development is the recognition that a Type I adsorption equation like
Eq. (1) applies to absolute adsorption n. Absolute adsorption refers to the actual number
of molecules present in the micropores and increases monotonically with pressure to an
asymptote called the saturation capacity m. Experimental excess adsorption isotherms pass
through a maximum and then decrease with pressure.
The adsorption isotherms in Fig. 1 may be considered absolute adsorption as a function
of gas-phase fugacity f of ethylene. For P < 1 bar, the term V,p in Eq. (3) is negligible
compared to n. Therefore a fit of these isotherms with Eq. (1) using constants reported
previously [4] provides the absolute amount adsorbed n ( f ) . Given an equation for the
absolute isotherm and the pore volume of faujasite (340 cm3/kg), one can calculate the
excess isotherms at high pressure. Taking the fugacity f as an independent variable, the bulk
properties of gaseous ethylene derived from the SRK equation [5] were used to determine
the pressure P and density p = P / ( z R T ) , where z is the bulk compressibility factor of
gaseous ethylene. The excess functions for three temperatures are plotted on Fig. 3. For
subatmospheric pressure, the values of absolute and excess adsorption coincide. At higher
pressure, the absolute adsorption approaches a saturation value (4.534 mol/kg) while the
excess adsorption passes through a maximum and eventually decreases to negative values at
pressures above 100 bar. The isotherm for -30 C terminates at the dew point of ethylene
vapor, 19.4 bar. The maximum excess adsorption decreases with temperature. At the
pressure where the isotherms intersect, the temperature coefficient of adsorption is zero.
Above this pressure, excess adsorption increases with pressure.
The maximum in the excess adsorption isotherm occurs at the point where the densities
in the micropore and the bulk gas are increasing at the same rate with respect to pressure, so

46
that an increase in pressure has no effect upon the amount adsorbed. Zero excess adsorption
occurs at the pressure where the bulk and micropores densities are equal. Since the pore
density at saturation is p = m/% = (4.534/340)= 0.0133 mol/cm3, zero excess adsorption
corresponds to the same value for the bulk density of ethylene. The pressures corresponding
to this bulk density are 175 and 345 bar at 20C and 70"C, respectively. The density of
0.0133 mol/cm3 inside the micropores is 1.7 times the critical density or 0.7 of the liquid
density of ethylene at its normal boiling point (-110C). It is interesting that the micropore
volume may be calculated from the bulk density at which excess adsorption is zero using
V, = m/p.

0.1
0 . m 1 o.oO01 0.001 0.01 0.1 1 .o 10 100 500
Pressure (bar)
Fig. 3. Adsorption isotherms of CzH4 on NaX.The -30C isotherm termi-
nates at the dewpoint of ethylene vapor (19.4 bar)

The absolute adsorption isotherms for -30C and 70C were calculated from the 20C
isotherm using the integrated form of Eq. (1) and the differential enthalpy plotted on Fig. 2.
The reasonable approximation was made that the differential enthalpy is independent of
temperature. No other assumptions were needed to calculate the excess adsorption isotherms
on Fig. 3.
Examination of Fig. 1 answers the questions raised in the introduction about the deter-
mination of the saturation capacity. The saturation capacity for absolute adsorption cannot
be extracted from supercritical isotherms measured at sub-atmospheric pressure such as the
70C isotherm. Raising the pressure would bring the isotherm into the region where the dif-
ference between absolute and excess adsorption is no longer negligible. The Henry constant
(the slope of the adsorption isotherm at the limit of zero pressure) is difficult to extract from
sub-critical isotherms such as the -30C isotherm because the linear region occurs at very
low pressure. The optimum temperature for the determination of both the Henry constant
and the saturation capacity is near the critical temperature of the gas (9C for ethylene),
which is the 20C isotherm on Fig. 1.

47
3 Gas Mixtures
The principles of phase.equilibrium do not apply to excess adsorption variables at high
pressure where the excess adsorption passes through a maximum. Under these conditions,
the pressure is no longer a singlevalued function of excess adsorption so that ne cannot serve
as an independent variable for the determination of partial molar quantities such as activity
coefficients. Additional complications which arise at high pressure are: (1) the selectivity
for excess adsorption (S1,= (n:/v1)/(n;/~z)) approaches infinity BS $ + 0; and (2) the
differential enthalpy of the ith component has a singularity at the pressure corresponding to
maximum nt. For excess variables, the differential functions are undefined but the integral
functions for enthalpy and entropy are smooth and well-behaved [l].
The principles of solution thermodynamics can be applied to absolute adsorption vari-
ables without any of these complications. For absolute variables, which arise naturally in
molecular simulation, the pressure is a singlevalued function of n, the differential functions
exhibit no singularities, and the selectivity approaches a finite value as P ---t 00. Absolute
adsorption may be determined experimentally by measuring excess adsorption in the usual
way (volumetric or gravimetric method) at sub-atmospheric pressure where the difference
between absolute and excess adsorption is negligible.
The fugacity equations are solved using absolute variables. As discussed previously, the
single-gas isotherms at sub-atmospheric pressure provide the absolute isotherm in the form
f;(np). Given the temperature of the isotherms, the independent variables are P and y1 in
the bulk gas. For a binary mixture the fugacity equations are written:

The fugacity fi = Pyi&(P, yi) is determined from an equation of state (EOS) for the pure
bulk gas. The adsorbed-phase activity coefficients are functions of the grand potential (Q)
and composition, where $ = -R/RT. The form of Eq. (2) allows the grand potential to be
calculated analytically [4]:

The standard-state properties are measured at equal values of the grand potential:

$l(n;) = $z(%) (7)


Given a bulk EOS to determine the fugacity functions and noting that ( q + z z ) = (y1+yz) =
1, the three Eqs. (4), (5), and (7) can be solved for the three unknowns: ni, n:, and 21.
Following methods described previously 141, the absolute total (n)and individual amounts
adsorbed (nl,nz) are calculated from the standard-state values n;l and ni. Finally, using
an EOS to determine the bulk density p(P,yl), the absolute variables are converted to
experimental excess variables by [l]:

nt = ni - V , p y i (8)
Excess isotherms are shown on Figure 4 for binary mixtures of ethylene and ethane
adsorbed on NaX. Both curves are for 20C and a vapor composition of 10 mole percent eth-
ylene. The individual excess isotherm for ethylene is the difference between two isotherms.
Ethylene is preferentially adsorbed and the individual isotherm for ethane reaches a max-
imum at a pressure of 3 bar. The individual isotherms show the same behavior as the
singlegas isotherms shown on Figure 3: a maximum value in the amount adsorbed followed
by a steep decline to zero.

5
total

1 C2H6

0.1
0 .001 0.01 0.1 1 lo 30
Pressure (bar)
Fig. 4. Adsorption of mixtures of C& and c2H6 on NaX at 20 C for
10 mole % ethylene in gas phase. The individual isotherm for CzH4 is the
difference between the total and the individual isotherm piotted for C2H6.
The solid lines on Figure 4 take into account the nonideal behavior of adsorbed mixtures
of ethylene and ethane in NaX. This system is highly nonideal because of the interaction of
the quadrupole moment of ethylene with the sodium cations of NaX. Activity coefficients at
iniinite dilution are unity at the limit of zero pressure and 0.27 at high pressure. The dashed
lines on Figure 4 were calculated for an ideal adsorbed solution (IAS) and the resulting error
in the individual isotherm for ethane at 30 bar is 20%.

4 Discussion
The high pressure adsorption of single gases and mixtures can be predicted from the low
pressure (subatmospheric) data for the same systems. The optimum temperature for mea-
suring the adsorption of single gases is near their critical temperature where both the Henrys
constant and the absolute saturation capacity can be determined accurately.
The absolute adsorption isotherm as a function of gas-phase fugacity is obtained directly
from molecular simulations based on the grand canonical Monte Carlo (GCMC) method.
Since the difference between absolute and excess adsorption is negligible at subatmospheric
pressure, the low-pressure portion of the absolute isotherm can also be determined from
experiment. Eq. (2) is suitable for extrapolating the absolute isotherm from low to high
pressure and Eq. (3) provides the conversion to excess adsorption. Experiments are needed
to test these predictions of adsorption at high pressure.

49
References
[l] Myers, A. L., A.I.Ch.E. J., 2002,48, 145.
[2] Myers, A. L., and Monson, P. A., Adsorption in porous materials at high pressure:
Theory and experiment, Langmuir, 2002,in press.
[3] Ruthven, D.M.,Principles of Adsorption and Adsorption Processes, Wiley Interscience,
New York, 1982.
(41 Siperstein, F.R. and A.L. Myers, A.I.Ch.E. J., 2001,47, 1141.
[5] Smith, J. M., Van Ness, H.C., and Abbott, M.M., Introduction to Chemical Engineering
Themodynamics, 6th Edition, McGraw-Hill, New York, 2001.
NEW SORBENTS FOR DESULFURIZATION OF TRANSPORTATION FUELS

RALPH T.YANG, ARTURO HERNANDEZ-WDONADO, AKIRA TAKAHASHI, AND


FRANCES H.YANG
Department of Chemical Engineering, University of Michigan, Ann Arbor, MI 48109-2136, USA.

New sorbents for desulfurization of liquid transportation fuels were developed using n-complexation.
Vapor phase adsorption isotherms were investigated to understand the interaction between
benzendthiophene and various kinds of sorbents: Ag-Y, Cu-Y, Na-Y, H-USY, Na-ZSM-5, activated
carbon, and modified activated alumina Liquid breakthrough experiments were conducted to test
their true sulfur removal effectiveness and capacities. Compared with Na-Y, Cu-Y and Ag-Y
adsorbed significantly larger amounts of both thiophene and benzene at low pressures, due to
x-complexation with C d and A$. Molecular orbital calculations confirmed the relative strengths of
n-complexation: thiophene > benzene and Cu' > A$. The experimental heats of adsorption for
x-complexation are in excellent agreement with theoretical molecular orbital predictions. The
sorbent capacities for thiophene at the low pressure of 2 . 3 ~ 1 0atm
~ were 0.92 moleculdCu+ and
0.42 moleculdAg', and followed the order: Cu-Y & Ag-Y >> Na-ZSM-5 > activated carbon > Na-Y
> modified alumina & H-USY. For liquid phase experiments, Cu(1)-Y, Ag-Y and Na-Y zeolites
were used to removal low concentration thiophene from mixtures including benzene and/or n-octane,
all at m m temperature and atmospheric pressure. Sulfur-free (i.e., below the detection limit of 4
ppmw sulfur) fuels were obtained with Cu(1)-Y and Ag-Y, but not Na-Y. Breakthrough and
saturation adsorption capacities obtained for an influent concentration of 760 ppmw sulfur (or 2000
ppmw thiophene) in n-octane follow the order Cu(1)-Y > Ag-Y > Na-Y and Cu(1)-Y > Na-Y > Ag-Y,
respectively. Regeneration of the adsorbent was accomplished by using air at 350'C followed by
re-activation in helium at 450'C. The observed adsorption behavior, in general, agrees well with the
studies performed for pure component vapor phase adsorption of thiophene and benzene with the
same adsorbents.

Introduction

Removal of sulfur-containing compounds is an important operation in petroleum refining,


and is achieved by catalytic processes at elevated temperatures and pressures [I]. The
hydrodesulhkation (HDS) process is highly efficient in removing thiols and sulfides, but
less effective for thiophenes and thiophene derivatives. New legislation will require
substantial reductions in the sulfur content of transportation fuels. For example, the new
US.EPA s u l k standards require that the sulfur contents in gasoline and diesel fuels for
on-board transportation will be 30 and 15 ppm, respectively, decreased fiom the current
levels of several hundred ppm. Faced with the severely high costs of compliance, a
surprisingnumber of refiners are seriously considering reducing or eliminating production
of on-board fuels [2]. The new challenge is to use adsorption to selectively remove these
sulfur compounds from liquid fuels. Since adsorption would be accomplished at ambient
temperature and pressure, success in this development would lead to a major advance in
petroleum refming. However, success would depend on the development of a highly
selective sorbent with a high sulfur capacity, because the commercial sorbents are not
desirable for this application. First results on sorbents based on n-complexation for
desulfurization have been reported recently by Yang et al. [3,4] which were shown to be
superior to all previously reported sorbents.
During the last decade, there have been several published accounts on using
adsorption for liquid fuel desulfurization. Commercially available sorbents (i.e., zeolites,
activated carbon and activated alumina) were used in all of these studies. Weitkamp et al.

51
[5]reported that thiophene adsorbed more selectively than benzene on ZSM-5 zeolite.
Based on this study, King et al. [6]studied selective adsorption of thiophene, methyl- and
dimethyl-thiophenes (all with one ring) over toluene and pxylene, also using ZSM-5.
They showed that thiophene was more selectively adsorbed, both based on fixed bed
breakthrough experiments. However, the capacities for thiophene were low (only 1-2%
wt. adsorbed at 1% thiophene concentration). Both vapor phase and liquid phase
breakthrough experiments were done in these studies, and the results from two phases
were consistent. The pore dimensions of ZSM-5 are 5.2-5.6A. Hence organic sulfur
compounds with more than one ring will be sterically hindered or excluded. Zeolites with
larger pores, as well as larger pore volumes, will be more desirable than ZSM-5 as the
selective sorbents. Indeed, results of Salem and Hamid [7]indicated that 13X zeolite as
well as activated carbon had much higher sorption capacities for s u l k compounds.
Based on the data of Salem and Hamid [7],the capacity for sulfur compounds by 13X
zeolite was approximately an order of magnitude higher than that of ZSM-5, when
compared with the data of King et al. [6]extrapolated to the same conditions. Modified
activated alumina (Alcoa Selexsorb), which contains proprietary modifier to provide
optimum adsorption of a number of polar organic compounds, has been used in an
adsorption process by Irvine [8].No direct comparison has been made among these
commercial sorbents. Their experiments were mostly done in fixed bed adsorbers, by
measuring the breakthrough capacities. Based on the literature, the large pores zeolites
(NaX or Nay) are about the same as activated carbon and alumina, in terms of adsorption
of thiophene.
Based on the principles of n-complexation, we have already developed a number of
new sorbents for a number of applications. These include sorbents for: (a)
olefdparaffm separations [9- 121,(b) diene/olefm separation or purification (i.e., removal
of trace amounts of dienes fiom olefins) [ 131,and (c) aromatics/aliphatics separation and
purification (i.e., removal of trace amounts of aromatics fiom aliphatics [141.Throughout
this work, we have used molecular orbital calculations to obtain a basic understanding for
the bonding between the sorbates and sorbent surfaces, and further, to develop a
methodology for predicting and designing n-complexation sorbents for targeted molecules
(e.g. Ref. 11).
First results on n-complexation sorbents for desulfurization with Ag-Y and Cu(1)-Y
zeolites have been reported recently [3,4]. In this work, we included the known
commercial sorbents such as Na-Y, Na-ZSM5, H-USY, activated carbon and activated
alumina (Alcoa Selexsorb) and made a direct comparison with Cu(1)-Y and Ag-Y which
were the sorbents with n-complexation capability. Thiophene and benzene vapors were
used as the model system for desulfiuization. Although most of these studies can be
applied directly to liquid phase problems, Cu-Y (auto-reduced) and Ag-Y zeolites were
also used to separate liquid mixtures of thiophenehenzene, thiopheneh-octane, and
thiophenehenzene/n-octane at room temperature and atmospheric pressure using
fixed-bed adsorptiodbreakthrough techniques. These mixtures were chosen to understand
the adsorption behavior of s u l k compounds present in hydrocarbon liquid mixtures and
to study the performance of the adsorbents in the desulfurization of transportation fuels.
Moreover, a technique for regeneration of the adsorbents was developed in this study [4].

52
Experimental

Adsorbent Preparation
Various kinds of sorbents were investigated in this work. Four as-received sorbents:
Na-type Y-zeolite (Si/A1=2.43, Strem Chemical), H-type ultra-stable Y-zeolite (Si/AI=l95,
TOSOH Corporation), activated carbon (Type PCB, Calgon Carbon Corporation) and
modified activated alumina (Selexsorb CDX, Alcoa Industrial Chemical), were used in
this study. According to the product datasheets, Selexsorb CDX is formulated for
adsorption of sulfur-based molecules, nitrogen-based molecules, and oxygenated
hydrocarbon molecules. Na-Y and H-USY were in powder form (binderless). Since
activated carbon was in granular form and activated alumina was in pellet form, they were
crushed into powder form for evaluation.
Cu(1)-Y was prepared by ion exchange of Na-Y zeolites with Cu(NO3)Z followed by
reduction of Cu2' to Cu'. First, as-received Na-Y was exchanged twice using excess
amounts (10-fold cation-exchange-capacity(CEC) assuming that one Cu" compensates
two aluminum sites) of 0.5 M Cu(NO& at room temperature for 24 hours. After the
exchange, the zeolite suspension was filtered and washed with copious amount of
de-ionized water. The product was dried at 100 "C overnight. Several groups have
reported reduction of Cu" to Cu' in zeolite in He (i.e., auto-reduction). In this study,
reduction of Cu2' to Cu' was carried out in He only at 450 "C for periods in the 1 to 18
hours range. Ag' ion-exchange Y-zeolite (Ag-Y) was prepared at room temperature for 24
h in the same manner as Cu" exchange, using 5-fold excess AgN03 (O.1M). 13X
(Si/Al=1.25, Linde) was used for the preparation of Cu-X (10 fold CEC solution of
Cu(NO&, ion-exchanged at 65 "C for 24 hrs, three times) and Ag-X (5-fold CEC solution
of AgN03, ion-exchanged at RT for 24 hrs, twice). Na-type ZSM-5 (Na-ZSM-5) was
prepared at room temperature by Na'-exchange of Nh-ZSM-5 (Si/Al=lO, ALSI-PENTA
Zeolite GmbH).
Vapor Phase Isotherms and Heat of Adsorption
The objective of this study is to compare the strength of adsorptive interaction between
adsorbents and thiophenehenzene. Extremely low partial pressures at less than l o 5 atm
would be necessary to meet this objective if isotherms were measured at ambient
temperature, because the isotherms at ambient temperature are fairly flat and are difficult
to compare. However, it is very difficult to obtain and control such low partial pressures
experimentally. Therefore, single component isotherms for benzene and thiophene were
measured at 90, 120 and 180 "C using standard gravimetric methods. A Shimadzu
TGA-50 automatic recording microbalance was employed. Isosteric heats of adsorption
were calculated using the Clausius-Clapeyron equation from isotherms at different
temperatures.
Fired Bed A&orptiodBreakthrough Experiments
All adsorption/breakthrough experiments were performed in vertical custom made quartz
adsorbers equipped with a supporting glass frit. Initially, the adsorbents were loaded
inside the adsorber (between 1 or 2 grams), and heated in situ (250 - 450'C) while
flowing either helium or nitrogen upwards. After activation treatment, the zeolite
adsorbent under study was allowed to cool down to room temperature under. Next, a
sulfur-free octane or benzene solution was allowed to flow downwards through the

53
sorbent at a rate of 0.5 cm3/min.After wetting the adsorbent for about 30 minutes, the feed
was changed to a mixture of CgHI8(n-octane) and/or c6H6 (benzene) containing different
concentrationsof C4H4S(thiophene) also at a 0.5 cm3/minrate. Samples were collected at
regular intervals until saturation was achieved, which .depended on the adsorption
dynamics and amount of adsorbent.
All the samples collected during the breakthrough experiments were analyzed using a
Shimadzu GC (Gas Chromatography) unit equipped with a polar column, an automatic
multi-sampler, and a FID detector. The minimum thiophene concentration detection was
around 10 ppmw or 4 ppmw on a sulfur basis.

Molecular Orbital Computational Details

Molecular orbital (MO) studies on the n-complexation bonding for benzene and sorbent
surfaces had been investigated recently [3]. In this work, similar MO studies were
extended to thiophene and zeolites. The Gaussian 94 Program [ 151 in Cerius2 molecular
modeling software [161fiom Molecular Simulation, Inc. was used for all calculations. MO
calculations for thiophene and sorbent surfaces were performed at the Hartree-Fock (HF)
and density functional theory (DFT) level using effective core potentials (ECPs) [17-191.
The LanL2DZ basis set [20] is a double-t; basis set containing ECP representations of
electrons near the nuclei for post-third-row atoms. The reliability of this basis set has been
confirmed by the accuracy of calculation results as compared with experimental data.
Therefore, the LanL2DZ basis set was employed for both geometry optimization and
natural bond orbital (NBO) analysis.
The restricted Hartree-Fock (RHF) theory at the LanL2DZ level basis set was used to
determine the geometries and the bonding energies of thiophene on AgCl and CuCI. The
simplest models with only a single metal chloride interacting with a thiophene molecule
were chosen for n-complexation studies. The optimized structures were then used for
bond energy calculations according to the following expression:
Eads = Eadsorbate -k Eadsorbent - Eadsorbent-adsorbate
where is the total energy of the bare
is the total energy of thiophene, Eadsorbent
adsorbent i.e. the metal chloride and Ea~~t-adsorb.te is the total energy of the
adsorbate/adsorbentsystem. A higher value of Endscorresponds to a stronger adsorption.
Natural Bond Orbital (NBO)
The optimized structures were also used for NBO analysis at the B3LYP/LanL2DZ level.
The B3LYP [21] approach is one of the most useful self-consistent hybrid (SCH)
approaches [22], it is Becks 3-parameter nonlocal exchange functional[23] with nonlocal
correlation functional of Lee, Yang and Parr [24]. The NBO analysis performs population
analysis that pertains to localized wave-function properties. It gives a better description of
the electron distribution in compounds of high ionic character, such as those containing
metal atoms [25]. It is known to be sensitive for calculating localized weak interactions,
such as charge transfer, hydrogen bonding and weak chemisorption. Therefore, the NBO
program [26] was used for studying the electron density distribution of the adsorption
system.

54
Modelsfor Ag-Zeolite (Agz) and Cu-Zeolite (CuZ)
The zeolite models selected for this study are similar to the ones used by Chen and Yang
[33], with the molecular formula of (HO)3Si- 0 -A1(OH)3, and the cation Ag' or Cu'
sits 2 - 3 A above the bridging Oxygen between Si and Al. This is a good cluster model
representing the chemistry of a univalent cation bonded on site I1 (SII) of the faujasite
framework (Z). Once the optimized structures of AgZ and CuZ are obtained at the
B3LYPLanL2DZ level, then a molecule of thiophene (C4&S) or benzene (C6&) is
added onto the cation of the zeolite model, and the resulting structure is hrther optimized
at the B3LYPLanL2DZ level.

Results and Discussion

Vapor Phase Adsorption

BenzenejThiophene Adsorption Isotherms


Figure 1 and Figure 2 show the isotherms of benzene and thiophene on Ag-Y and Cu-Y.
Curves are fitted with Dubinin-Astakhov (solid line) and Langmuir-Freundlich (dotted
line) isotherms. Compared with Figure 3, these sorbents adsorbed significantly more
thiophenehenzenethan Na-Y at pressures below 10" atm, and nearly the Same amounts at
high partial pressures. This result was a clear indication of n-complexation with Ag'and
Cu'; since Na' could not form n-complexation bonds. However, the difference of
thiophenehenzene adsorption amount did not reflect the relative strengths of
a-complexation between Cu' and Ag" because the Cu' exchange was not complete.
Neutron activation analyses of the sorbent samples showed that the Ag' exchange was
100% but the Cu+ exchange was only 46%. According to the EPR analysis, described
elsewhere [171, only a half of the Cu2+was auto-reduced to Cu' after our heat treatment at
450 "C for lhr in He. On a per-cation basis, it is seen that Cu' could adsorb higher
thiophene adsorption amounts. In fact, 0.92 thiophene molecule per Cu' was obtained at
2xlO-' atm at 120C. This amount was due to Cu' since the amount adsorbed by NaY at
the same pressure was negligible. At the same pressure, only 0.42 thiopheneIAg' was
obtained. This result indicated strong a-complexation bonds between both Cu' and Ag",
and that the bond with Cu' was stronger.

Comparison of Thiophene Adsorption on All Adsorbents


Thiophene adsorption isotherms on all sorbents are compared in Figure 4. It is clearly
seen that Ag-Y and Cu-Y could adsorb significantly larger amounts of thiophene even at
very low pressures.

55
1 1
4.0

3.5
1.5 ._.-
3.0

2.5

2.0
1.o
1.5

1.o

--
0.5

0.0
1.E45 1.E.M 1.E- 1. E m l.E-01
(-1

Figure 1. Pure component equilibrium isotherms Figure 2. Pure component equilibrium isotherms of
of benzene and thiophene on Ag-Y (SVAk2.43) at benzene and thiophene on Cu-Y (Si/AI=2.43) at 90
120 OC and 180 "C. Fitted curves are not shown for "C and 120 "C.
benzene adsorption at 180 "C because the artificial
crossovers to the curves for thiophene at 180 OC are
observed.
3.5

3.0

2.5

2.0

1.5

1.0

0.5

0.0
l.wE.05 1.wE.M l.wE.03 1.oQE42 l.wE-01
1.E.05 1.E.M 1.E- 1.E-02 1.E-01
PuUd Pmsun(um) P N W Raun(am)

Figure 3. Pure component equilibrium isotherms Figure 4. Comparison of equilibrium adsorption


of benzene and thiophene on Na-Y(SVAl=2.43) at isotherms ofthiophene at 120 "C.
120 "C and 180 "C.

Heat of Adsorption
Heats of adsorption were calculated using the Clausius-Clapeyron equation fkom
isotherms at two different temperatures, and are shown in Table 1. All the heats of
thiophenehnzene adsorption had the tendency to decease as the loading increased. This
is a common phenomenon for the sorbents such as ion-exchanged zeolites that have
heterogeneous sites. The heats of adsorption on activated carbon, in particular, ranged
widely fkom 23.9 kcaYmol (at OSmmoVg loading) to 8.0 kcaYmol (at 3mmoVg). Ag-Y
and Cu(1)-Y exhibited the higher heats of adsorption than Na-Y for both benzene and
thiophene because of .rr-complexation. More importantly, the heats of adsorption for
thiophene were higher than that of benzene. These experimental results can be explained
by molecular orbital calculation and NBO analysis, which will be discussed shortly.
At the low loading of 0.5 mmoVg, Na-ZSM-5 showed nearly the same heats of
adsorption as Na-Y for both thiophene and benzene. The different pore dimensions
(5.2-5.6 A for ZSM-5 vs. 7.4 A for Na-Y) apparently had no influence on the heats of
adsorption. It is not clear why the amounts adsorbed on Na-Y decreased sharply at very
low pressures while that on Na-ZSM-5 maintained.
Table 1. Heat of Adsorption (kcal/ml)calculated from isothcm at difhent temperatures.

Na-Y Ag-Y Cu-Y BUSY Na-ZSM5 ActivuedC.rbon AaivakdALumi~

(SilAk2.43) (SUAk2.43) (SilAk2.43) (Si/Al=195) (SilAkIO) T y ~ PCB


e Selcxsorb CDX

Thiophc~c 19.1-19.6 21.3-21.5 20.8-22.4 7.9-11.2 18.6-19.2 8.0-23.9 16.1-17.5

(0.5-2.0) (1.5-1.7) (2.0-3.0) (0.1-0.3) (0.45-0.60) (0.5-3.0) (0.2-1.O)

Benzea~ 17.0-18.2 19.0-20.1 19.3-21.8 6.6-13.1 16.5-17.9 13.1-16.1 16.8-19.6

(1.5-2.0) (1.5-1.8) (1.8-2.5) (0.1-0.9 (0.45-0.65) (1.0-3.0) (0.6-1.0)


*) Numbers in parcnthtses indicate the adsorption ~moullls
(mmoVP, for calculation.
**)The hcpts of adsorptiondmeased with loading in all caacs.

Bond Energies, Geometries and NBO Results


The energies of adsorption are summarized in Table 2. The theoretical calculations
indicate that the n-complexation strengths follow the order CuZ > AgZ and more
importantly,thiophene > benzene. This trend is in agreement with the experimental data,
in Table 4. In fact, the molecular orbital results on CuZ and AgZ are in excellent
agreement with the experimental data. Both chloride and zeolite models were used as
the anion in the theoretical calculations, while only zeolite framework was the anion in the
experiment. It is known fiom our previous work that the anion has a large effect on the
n-complexation bonds [11,27]. The bond energies on the zeolites (Z)are significantly
higher than that on the chlorides (Table 2). This result indicates that the zeolite anion is
more electronegative than the chloride anion, which has been already revealed by Chen
and Yang in their ab initio molecular orbital calculations [27]. In the optimized structures
of thiophene-MC1complexes, the distance between the thiophene molecule and Cu ion is
about 0.3 A shorter than that of thiophene and Ag ion for chloride. The NBO analysis
for thiophene adsorption is summarized in Table 3. There is some donation of electron
charges from the n-orbital of thiophene to the vacant s orbital of metals known as ci
donation and, simultaneously, back donation of electron charges fiom the d orbitals of
metals to x* orbital (i.e., anti-bonding x orbital) of thiophene or x back-donation. It
appears that the o donation is more predominant for thiophene and the x back-donation is
more important for benzene (published elsewhere). Comparing the two anions, zeolite
anion and chloride anion, the NBO results show that both ci donation and d z* -
backdonation are significantly stronger with the zeolite anion bonded to Agf or Cu'. The
charge transfer results again confirmed the experimental data that the relative strengths of
the n-complexation bonds follow the order: thiophene > benzene and Cu' > Ag'.

57
Table 2. Summary of energies of adsorption for thiophene and benzene in kcaUmol calculated
from molecular orbital theory (Z denotesZeolite anion)
E,,(Thiophene) E,,(Benzene)
CUCl 13.5 12.4
AgCl 9.0 8.6
CUZ 21.4 20.5
A@ 20.0 19.1

Table 3. Summary of NBO analysis* of n-complexationbetween thiophene and MCVMZ


C+M interaction M+C interaction
(adonation) -
(d x'backdonation) Net Change
Mx 91 e 91+e
CUCl 0.037 -0.022 0.015
AgCl 0.022 -0.014 0.008
CUZ 0.112 -0.063 0.049
A@ 0.101 -0.086 0.015
~ ~~ ~~ ~ ~~

* q I is the amount of electron population change in valence s orbitals of the metal, and e is the
total amount of electron populationdecrease in valence d orbitals of the metal.

Liquid Phase Aakorption

Figures 5 and 6 show breakthrough curves obtained for 2000 ppmw thiophene (760
ppmw s u l k basis), for n-octane as solvent (the breakthrough curve for Cu(1)-Y is shown
in a separate figure for clarity since the abscissas are quite different). All adsorbents
showed remarkable selectivity towards C41&S, indicating that CgHIg adsorption is not
competitive. Saturation adsorption capacities calculated from the breakthrough curves
were 1.05 and 0.90 mmoVg for Na-Y and Ag-Y, respectively. However, for Na-Y, the
breakthrough of thiophene molecules occurred earlier, at about 2.84 cm3/g compared to
22.50 cm3/g in Ag-Y. This was evidence of weak adsorbate-adsorbent interactions on
Na-Y, which did not have the ability for n-complexation as in the case of Ag-Y. This
agrees very well with the pure vapor phase adsorption data reported above. The saturation
adsorption amount in Na-Y was higher than that Ag-Y due to pore volume differences and
difference in the densities of zeolites.
For the same feed conditions described above, Cu(1)-Y showed again the highest
selectivity and capacities among the adsorbents studied. The saturation capacity was 2.55
mmoVg, which was more than twice the amount found for the other adsorbents, indicating
superior interaction with the thiophene molecules. For about 2 grams of Cu(I>Y, it took
more than 300 minutes for the thiophene molecules to break through the adsorbent at a
feed rate of 0.5 cm3/min (refer to figure 6). Saturation was reached after 600 minutes,
which was remarkable for such a small amount of adsorbent. A large amount of Cu2+ions
must have been reduced to Cu'. As mentioned earlier, Takahashi et al. and others [14]
have reported 50% auto-reduction of copper under helium/vacuum atmospheres after just
1-2 hours. The adsorbent used in this part of the work was exposed to helium at 450'C for
no less than 18 hours. Possibly longer activation time increased the amount of reduced
copper ions, while the adsorption behavior already indicates that the auto-reduction
process yields promising results.

58
Figure 7 shows breakthrough curves for Cu(1)-Y for an influent containing 500 ppmw
thiophene (190 ppmw sulfur> in n-octane. The saturation capacity was reduced to 1.28
mmoVg, which was about 50% of the amount obtained previously with the 2000 ppmw
thiophene feed. This indicates that the equilibrium adsorption isotherm was not
"rectangular" in shape at low concentrations and rather showed a noticeable decrease in
adsorbed amount as one decreased the concentration. Despite this, the observed saturated
amount was not low, when taking into account that the thiophene concentration was 75%
less than the case discussed previously (i.e., 2000 ppmw).
Figure 7 also shows breakthrough curves after Cu(1)-Y adsorbent regeneration
(second cycles). Under an atmosphere of nitrogen at 350'C, the regenerated adsorbent did
not recover the original capacity. The new capacity for the adsorbent at saturation was
0.80 mmoVg, which was more than a 30% reduction fkom the original capacity. In fact, the
color of the adsorbent remained black, which indicated the presence of copper thiophene
complexes. Meanwhile, regeneration under air at 350'C followed by reactivation under
helium at 450'C recovered almost all of the original capacity. For this case, the observed
saturation capacity was about 1.20 mmoVg, which was only a 5% reduction from the
original capacity.
1 "
0"

0.8 0 0

0.2

0
0 10 20 30 40 50 60 70 0 20 40 60 80 100 120 140 160 180 200
cmVg (cumdative efauent volumdsorbentweight) cd/g (cumdative effluent volumdsorbent weid

Figure 5. Breakthrough of thiophene in a fixed-bed Figure 6. Breakthrough of thiophene in a


adsorber with Ag-Y 0 or Na-Y (0) adsorbents, fixed-bed adsorber with Cu(1)-Y adsorbent, with a
with a liquid feed containing 2000 ppmw (Ci) of liquid feed containing 2000 ppmw (Ci) of
thiophene in octane, at room temperature. thiophene in n-octane, at room temperature.
0.4

o o
0.2 o . 6

0 50 100 150 200 250 300 350 400 0 5 0 1 0 0 1 5 0 # ) 0 2 5 0


cmVg (cumdative d u e n t vohmdsorben~weight cmV.3 (clumllative dfiucst v o l d s & t weight)

Figure 7. Breakthrough of thiophene in a fixed-bed Figure 8. Breakthrough of thiophene in a


adsorber with fresh (0) and regenerated (0,[7, fixed-bed adsorber with Cu(I)-Y adsorbent, with a
Cum-Y adsorbent, with a liquid feed containing 500 liquid feed containing 500 ppmw (C,) of thiophene,
ppmw (Ci) of thiophene in natane, at room 20 wt?hbenzene and 80 wt% n-octane, at room
temperahue. (a) Adsorbent regenerated in nitrogen temperature.
at 350C followed by re-activation in helium at
450C. 0 Adsorbent regenerated in air at 350C
followed by re-activation in helium at 450C.

For the final part of this study, it was desired to use mixtures with compositions
similar to that of transportation fuels. Gasoline contains about 20-30 wtoh aromatics, many
thiophenic compounds and 70-80% alkanes such as n-hexane and n-octane. The aromatic
contents in diesel and jet fuels are < 20%. Thus, a mixture containing 20 wt% benzene, 80
wt% n-octane, and 200 ppmw sulfiu (ca. 500 ppmw thiophene) was used to simulate
gasoline. Figure 8 shows a breakthrough curve for thiophene in such mixture after
adsorption at room temperature on Cu(1)-Y. The sulfur adsorption capacity was 0.44
mmoVg or about 1.4 wt% sulfur. The results are promising when compared to other
adsorbents used in previous studies. Ma et al. studied fixed-bed adsorption of thiophene
compounds fiom diesel and jet fuels using an undisclosed transition metal compound ( 5
wt% loading) supported in silica gel [28,29]. They obtained a saturation adsorption
capacity of 0.015 g of sulfur per cm3 of adsorbent which can be compared directly to our
results. Assuming that the density of the Cu(1)-Y is close to that of Na-Y (- 1.3 g/cm3),
which is lower than the actual value because sodium is lighter than copper, then the
observed saturation capacity in our case is approximately 0.018 g of sulfur per cm3 of
adsorbent. Ma et al. also showed that breakthrough occurs at about 20 cm3 effluent
volume for about 3.2 cm3 of the metal loaded silica gel compared to 30 cm3 effluent
volume for about 0.75 cm3 volume of Cu(1)-Y. Therefore, our adsorbent is capable of
processing more fuel with very low sulfur streams with less adsorbent material. It should
be mentioned that jet fuel and other fuels contain heavier thiophene compounds such as
benzothiophenes, dimethylthiophenes, and dimethylbenzothiophenes and these should
adsorb strongly in Cu(1)-Y. Breakthrough results on gasoline and other transportation
fuels will be published elsewhere shortly.
Conclusions

In this work, vapor-phase benzene/thiophene adsorption isotherms were investigated to


develop new sorbents for desulfurization. Among the sorbents studied, Cu(I>Y and Ag-Y
exhibited excellent adsorption performance (capacities and separation factors) for
desulfurization. This enhanced performance compared to Na-Y was due to the
x-complexation of thiophene with Cu' and Ag'. Molecular orbital calculations confirmed
the relative strengths of n-complexation: thiophene > benzene and Cu' > Ag'.
This work has also demonstrated that copper (auto-reduced) and silver exchanged
Y-type zeolites are excellent adsorbents for removal of thiophene tiom aromatic and/or
hydrocarbon mixtures, based on fixed-bed adsorption experiments. Both adsorbents were
capable of reducing sulfur content to values < 4 ppmw sulfur for long periods of time.
Cu(I>Y provided the best adsorption capacity both at breakthrough point and at saturation,
surpassing all other adsorbents by more than 50%. Regeneration of the copper based
adsorbents can be accomplished in air at 35072 which recovered almost all of the original
adsorption capacity. More studies will be needed in order to fully understand the effect of
copper loading and to include heavier thiophenes, which are abundant in liquid fuels.
Breakthrough results on gasoline and other transportation fuels will be published
elsewhere shortly.

Acknowledgements

-
Supports from NSF and DOE are acknowledged. A U.S. Patent is pending filed with U.S.
and foreign Patent Offices.

References

1. Farrauto, R. J.; Bartholomew, C . H. Fundamentals of Industrial Catalytic


Processes, Chapman and Hall, New York, 1997.
2. Parkinson, G., Diesel Desulfurization Puts Refiners in a Quandary. Chemical
Engineering, 2001, February issue, 37.
3. Yang, R. T.;Takahashi, A.; Yang, F. H., New Sorbents for Desulfurization of Liquid
Fuels by n-Complexation. Ind Eng. Chem. Res., 2001,40,6236.
4. Yang, R.T.; Takahashi, A.; Yang, F.H.; Hernandez-Maldonado, A. Seldctive Sorbents
for Desulfurization of Liquid Fuels. U.S. and foreign Patent applicationsfiled, 2002.
5. Weitkamp, J.; Schwark, M.;Emest, S. Removal of Thiophene Impurities from
Benzene by Selective Adsorption in Zeolite ZSM-5. J, Chem. SOC.Chem. Commun.,
1991,1133.
6. King, D. L; Faz,C.; Flynn, T. Desulfurization of Gasoline Feedstocks for Application
in Fuel Reforming. SAE Paper 2000-01-0002, SOC.Automotive Engineers, Detroit ,
2000.
7. Salem, A. S. H.; Hamid, H. S. Removal of Sulfur Compounds fiom Naphtha
Solutions Using Solid Adsorbents. Chem. Eng. Tech., 1997,20, 342.
8. Irvine, R L.,Process for Desulfirizing Gasoline and Hydrocarbon Feebtocks, U. S .
Patent 5,730,860(1998).

61
9. Yang, R.T.; Kikkinides, E. S. New Sorbents for Olefin-Param Separations by
Adsorption via p-Complexation AZChE J., 1995, 4I , 509.
10. Rege, S. U.; Padin, J; R; Yang,, R T. Olefm-Paraffin Separations by Adsorption:
Equilibrium Separation by n-complexation vs. Kinetic SeparationAZChE J., 1998,44,
799.
1 1 . Huang, H.Y.; Padin, J.; Yang, R. T. Ab Initio Effective Core Potential Study of
Olefin/Paraffin Separation by Adsorption via n-Complexation: Anion and Cation
Effects on Selective Olefin Adsorption J. P h s . Chem. B., 1999,103,3206.
12. Padin, J.; Yang, R T. New Sorbents for Olefin-Parafin Separations by Adsorption
via n-Complexation: Synthesis and Effects of Substrates Chem. Eng. Sci., 2000, 55,
2607.
13. Jayaraman, A; Yang, R. T.; Munson, C. L.; Chinn, D. Deactivation of
n-Complexation Adsorbents by Hydrogen and Rejuvenation by Oxidation, Znd Eng.
Chem. Res., 2001,40,4370.
14. Takahashi, A.; Yang, R T. New Adsorbents for Purification: Selective Removal of
Aromatics, AIChE J., 48,1457 (2002).
15. Frisch, M. J. et al., Gaussian 94, Revision B.3, Gaussian, Inc., Pittsburgh, PA., 1995
16. Bowie, J. E., Data Visualization in Molecular Science: Tools for Insight and
Innovation; Addison-Wesley Pub. Co.: Reading, Mass., 1995; chapter 9.
17. Hay, P. J.; Wadt, W. R Ab Inifio Effective Core Potential for Molecular
Calculations. Potentials for the Transition Metals Atoms Sc to Hg. J. Chem. Phys.
1985,82,270
18. Wadt, W. R.; Hay, P. J. Ab Inifio Effective Core Potential for Molecular
Calculations: Potentials for Main Group Elements Na to Bi. J. Chem. Phys. 1985, 92,
284
19. Hay, P. J.; Wadt, W. R. Ab Initio Effective Core Potential for Molecular
Calculations: Potentials for K to Au Including the outermost core orbitals. J. Chem.
Phys. 1985,82,299
20. Russo, T. V.; Martin, R. L.; Hay, P. J. Effective Core Potentials for DFT
Calculations. J. Phys. Chem. 1995,99, 17085
21. Becke, A. D. Density Functional Thermochemistry. 111. The Role of Exact
Exchanges. J. Chem. Phys. 1993,98,5648
22. Becke, A. D.; A New Mixing of Hartree-Fock and Local Density-Functional
Theories. J. Chem. Phys. 1993,98, 1372
23. Becke, A. D.; Density-Functional Thermochemistry. 11. The Effect of the
Perdew-Wang generalized-gradient correlation correction. J. Chem. Phys. 1992, 97,
9173
24. Lee, C.; Yang, W.; Parr, R.G. Development of the Colle-Salvetti Correlation-Energy
Formula into a Functional of the Electron Density. Phys. Rev., 1988, B37,785
25. Reed, A. E.; Weinstock, R.B.; Weinhold, F. Natural Population Analysis. J. Chem.
Phys. 1985, 83 (2). 735
26. Glendening, E. D.; Reed, A. E.; Carpenter, J. E., Weinhold, F. NBO Version 3. I ,
1995.
27. Chen, N.; Yang, R. T. Ab Inifio Molecular Orbital Study of Adsorption of Oxygen,
Nitrogen, and Ethylene on Silver-Zeolite and Silver Halides. Ind. Eng. Chem. Res.
1996,35,4020.

62
28. Ma, X.; Sun, L.; Yin, Z.; Song, C. New Approaches to Deep DesulfUrization of
Diesel Fuel, Jet Fuel, and Gasoline by Adsorption for Ultra-Clean Fuels and for Fuel
Cell Applications. Am. Chem. SOC.Div. Fuel. Chem. Prepr. 2001,46,648.
29. Ma, X.; Sprague, M.; Sun, L.; Song, C. Deep Desulfurization of Liquid
Hydrocarbons by Selective Adsorption for Fuel Cell Applications. Am. Chem. Sac.
Div. Pet. Chem. Prepr. 2002,47,48.

63
OPTIMIZATION OF CONTINUOUS CHROMATOGRAPHY SEPARATIONS

ZIYANG ZHANG AND M. MORBIDELLI


Swiss Federal Institute of TechnologyZurich, Laboratoriumfir Technische Chemie/LTC,
ETH-HiinggerbergIHCI, CH-8093 Ziirich, Switzerland
E-mail: morbidelli@ech.chem.ethz.ch

M.MAZzOTrI
ETH Zurich ,Institut f i r Verfarenstechnik, Sonneggstrasse3, CH-8092 Zurich, Switzerland
E-mail: manotti@ivuk mavt.ethz.ch

Two recent developments of the simulated moving bed chromatographic separation units, i.e. the
Varicol and the PowerFeed processes, are addressed. The performances of these three processes are
compared with reference to two chiral separation systems taken from literature, using a multiple
objective optimization technique based on a genetic algorithm. The performance of each process has
been optimized in a wide range of operating conditions, and from their comparison a good assessment
of their relative potential has been made. The optimization results have been discussed in the frame of
equilibrium theory and the N I ~ S of optimal design of the Varicol and PowerFeed processes have beem
discussed.

1 Introduction

Simulated Moving Bed (SMB) is an established technology for performing continuous


chromatographic separations covering all scales of possible interest in applications,
particularly in optical enantiomer separations. The SMB unit has been originally devised as
a practical realization of a true moving bed (TMB) unit where the two phases move
countercurrently. A schematic diagram of a typical 4-section SMB is shown in Figure 1(a).
The countercurrent movement of the solid and the fluid is simulated by moving
synchronously the inlet (feed and eluent) and outlet (raffinate and extract) ports by one
column in the direction of the fluid flow, with a predetermined period or switching time, 5.
The design and optimization of this unit can be done in the frame of equilibrium theory,
using the so called triangle theory [ 1,2] or using more detailed simulation models in
connection with various optimization strategies [3-71.
In order to make SMB units more economically efficient and competitive, several new
operation modes have been introduced. These include supercritical fluid SMB [&lo],
temperature gradient SMB [ 111 and solvent gradient SMB [12-14]. The basic idea is to
change the adsorption strength of the solute on the stationary phase in the different sections
by creating along the unit a gradient of either pressure or temperature or solvent
composition, respectively. Another direction which has been taken to improve SMB
performance is related to the idea of somehow forcing its dynamics. In this context, the SMB
unit is not regarded as an approximationthrough appropriate discretization of the TMB unit,
but is considered as a unit with many degrees of freedom that can be optimized to improve
its performance. The first step in this direction is the Varicol unit [ 151, where the inlet and
outlet ports are shifted asynchronously. An alternative operation has been proposed by
Kloppenburg and Gilles [ 161 and more recently by Zang and Wankat [171, who considered
fluid flow rates changing during the switching period as shown in Figure l(b). This will be
referred to in the following as PowerFeed operation. It is worth noting that, these two
ideas, Varicol and PowerFeed, can be regarded in some sense as having a common root, i.e.

64
changing within the switching period the flow rates, of the solid and the fluid, respectively.
In this sense, as mentioned above, they do not try to better approximate the TMB unit from
which the SMB is derived.

t -
----
SMB
Powdd

sectiw 2, Q s

Figure 1. (a) Operating diagram of a four section SMB unit; (b) Fluid flow rates schemes in the SMB and
PowerFeed operations during one switching period, t,
In this paper, we present and compare the optimal performances of the SMB,Varicol
and PowerFeed operations, using a multiobjective optimization technique, on two chiral
separation systems from the literature. The aim is to provide a clear picture, although
inevitably confined to the cases examined, of the relative potentials of these three operation
modes.

2 Optimization of SMB,Varicol and PowerFeed processes

Optimization of the SMB,Varicol and PowerFeed processes is very complex due to the
large amount of continuous and discrete parameters involved. These parameters include
fluid flow rates, switching time, unit configuration (total number of columns, column
distributionand column dimension), feed conditions, size of the packing particles and so on.
These parameters might have different values in the subintervals of one switching period in
the Varicol and PowerFeed operations. Depending upon the specific application, the best
process performance may be achieved by maximizing the product quality (mostly in terms
of the purity of either extract or raftinate stream or both) under fixed cost and productivity,
or by minimizing the cost and at the same time maximiig the productivity under some
given specification on product quality, or by other combinations of practical interest. Note
that in most cases, two of the objective hctions to be optimized are conflicting, as for
example: productivity increases if the unit operates at higher system fluid flow rates which,
however, imply a decrease in column efficiency, and therefore in the product purities.
Furthermore, various practical constraints on the column or entire unit pressure drop,
switching time, pump flow rates and product quality etc. complicate the optimization
problem.
In our earlier works [6,7],we have developed a new optimizationprocedure based on a
genetic algorithm, i.e. the non-dominated sorting genetic algorithm (NSGA), that allows to
handle these complex optimization problems. A more detailed description of this global
search and Optimizationtechnique is available elsewhere [18, 191.

65
3 Results and discussion

3.1 Comparison of SMB and Varicolfor diflerent total number of columns


An important aspect in comparing SMB and Varicol processes is the effect of the total
number of columns, Ned.It is in hct expected that as N,, increases, the discretizationof the
movement of the solid in the SMB improves and thereforethe main advantage of Varicol, i.e.
the possibility of better tuning the distribution of the columns in the sections by using non
integer values, becomes less effective. We consider in this work a single objective
optimization problem using as a model system the chiral separation reported by Biressi et al.
[3], aiming at investigating the effect of NW1on the performance of these two processes,
according to the following statement of the optimization problem:
Maximize J = PE[QI, ml, m2, m4, XI
Subject to P R 2 90% and APdt 5 70 bar
= ~dp=30 pn, N,,=4
Vsotid=120ml, R=l cm2,F=0.7 ml/min, C T ~ gA, -8
where the extract purity, PE is the objective to be maximized under the following
constraints: minimum 90% raffiate purity, PR,maximum 70 bar pressure drop along the
entire unit, Munit,given total solid volume, Vlolid,given production rate (i.e. faed feed flow
rate, F and total feed concentration, C,) and given packing particle diameter, d,. Having
fixed also the column cross section a,changes in NCdimply changes in column length. The
optimization variables are the flow rate in section 1, Q1,the flow rate ratios, ml,m2 and m4
[2], and the unit configuration represented by the parameter, x. For the SMB this can be
represented in the form of n1/n2/n3/n4where njrepresents the number of columns in section
j. For the Varicol operation the decision variables are the same, but now x can attain a much
larger number of values, which is in fact determined by the number of subintervals
considered in each switching interval. Thus for example, the complete configuration of a
4-subinterval Varicol unit with NC.l=5 can be described by the configuration sequence
2/1/1/1-1/2/1/1- 1/1/2/1-1/1/1/2 fiom the first subinterval to the last.
The equilibrium stage model reported by Zhang et al. [7], which accounts for the
influence of fluid flow rate on the column efficiency, has been used for the optimization
simulations.
The optimization results for a SMB and a 3-subinterval Varicol unit are compared in
Table 1 for various values of NWI.

Table 1. Optimization results of SMB and Varicol processes for various values ofNml

Process I N
, ,L Qi rnl rnz mc x mun
ti PR PE
I (crn) (mumin) (barj % %
I 4 30 30.328 1.550 0.867 0.754 1/1/1/1 52.79 90.02 86.49
5 24 30.747 1.567 0.843 0.738 1/2/1/1 52.14 90.00 91.68
SMB 6 20 35.126 1.512 0.900 0.594 1/2/2/1 60.47 90.07 96.04
7 17.14 34.735 1.584 0.888 0.525 1/3/2/1 57.86 90.02 97.13
8 15 36.595 1.643 0.900 0.444 1/3/3/1 59.86 90.06 97.29
4 30 29.502 1.747 0.837 0.680 * 48.40 90.01 91.55

varicol 5 24 35.761 1.538 0.895 0.663 ** 61.81 90.00 94.22

011~11-012/111-1111111;** 1/1/2/1-112/111-1/2/1/1

66
It is seen that PEincreases with increasing number of columns, particularly at the lower
values of N d . PEincreases in fact by almost 10%from 86.5% for Nc0,=4to 96.0% for NcOl=6.
A further increase in total number of columns has a smaller influence on PE.In addition,
from a practical point of view, it is worth noting that the short switching time values caused
by the short column lengths may lead to difficulties in the operation of the recycle pump.
With respect to column configuration x, the results in Table 1 indicate that all the additional
columns tend to distribute equally between sections 2 and 3. This is because, once the lower
bound on ml and the upper bound on m4 are satisfied, sections 2 and 3 are the most
important in determining PEand PR.The mivalues in Table 1 indicate that indeed in the
cases under examinationboth such constraints are satisfied. In Table 1 the performances of
a 3-subinterval Varicol with 4 and 5 columns are also shown. It is seen that Varicol
outperforms SMB both for NCoI=4,and to a lower extent, for NeoI=5. For Nco1>5, the
optimization technique has not been able to find a Varicol configuration which improves PE
over the corresponding SMB value. This means that, as expected, for increasing total
number of columns, the Varicol configuration is not worthwhile anymore.
It is worth noting that in the case of N,,=4, the following five possible configurations
have been considered for Varicol: 0/1/2/1, 0/2/1/1, 1/2/1/0, 1/1/2/0 and l/l/l/l. The first
four of such configurationshave only three sections, which actually were not considered in
the cases where NmI>4.The configuration 0/1/2/1~0~/1/1-1/1/1/1 was found to be optimal
for the 3-subinterval Varicol, which improves PEby about 5% over the corresponding SMB
with a 1/1/1/1 configuration. It is interesting to note that in the first two subintervals there is
no column in section 1; the fourth column moves in fact from section 3 in the first
subinterval to section 2 in the second and eventually to section 1 in the last subinterval. This
corresponds to the time averaged configuration 0.33/1.33/1.33/1, which compared to the
configuration 1/1/1/1 ofthe SMB, indicates that in this particular case Varicol improves the
performance of the separation by reducing the length of section I and increasing that of
sections 2 and 3.

3.2 Comparison of SMB and Varicolfor direrent product purity requirements


Another situation of interest in applications is one where the product purities are fixed, and
the objectives for optimal process operation are to reduce operating costs and increase
production. Hence, in this case, for a fixed target product purity of both extract and raffiate
streams, we seek the optimal process parameters for a SMB and a 4-subinterval Varicol unit,
which maximize production using minimum amount of eluent, for another chiral separation
system taken fiom literature [6,15]. The optimization problem is represented
mathematically as follows:
Max J I = F CQz, F, D,t XI
Min Jz = D [Q2, F, D,4, XI
Subject to PE= x f 0.002, x = 0.90,0.95,0.99
P R = Xf0.002,~=0.90,0.95,0.99
Q1= 27.5 mYmin, NWI = 5 , Lcol= 0.1 m, NNTP=80
where Q1,the column flow rate in section 1, is set at 27.5 ml/min to fix the maximum column
pressure drop, and the total solid used is also set by fixing total number of columns, NmIand
column length, L,I. The decision variables are the column flow rate in section 2, Q2, feed
flow rate, F,eluent flow rate, D, switching time, t, and column configuration parameter, x.
Note that the two variables F and D appear also in the objective functions. The Pareto

67
optimal solutionsare shown in Figure 2 for purity requirementsfor both extract and rafiinate
streams of 90%, 95% and 99%.

0
P. and PI = M% PrandPm=SB%
.....................................................

....................................................... 8 ....................................................
................................................... 0

.... 5.8

................... ................................
0 vuisol
5.4
.o..............

.T
............ ........

2.2 2.3 2.4 2.5 2.8 2.7 2.0 1.5 1.8 1.7 1.8 1.9 2 0.4 0.8 0.8 1 1.2
C. r U m b F, mVhlln F, mumln

Figure 2. Optimal solutions for the SMB and Varicol processes with 9004 95% and 99% purity requirements.
In all cases, using 5 columns, the optimal column configurationsare found to be 1/2/1/1
and 1/1/1/2-1/1/2/1-1/1/2/1-1/2/1/1 for the SMB and 4-subinterval Varicol unit,
respectively. It can be observed that for fixed purity specifications, both the SMB and the
Varicol processes require to increase the eluent consumption in order to increase the feed
flow rate. Secondly, the Varicol process consumes less eluent, D than the SMB process for
the same feed flow rate, F;or equivalentlyfor the same eluent consumption, D,the Varicol
process can treat more feed, F.However, the extent of improvementdepends on the purity
specifications. The more stringent the purity requirement, the larger the improvement
achieved by Varicol over SMB. For example, at D = 5.6 mymin, the improvement in
production rate, F of Varicol over SMB is lo%, 25% and 127% for a purity requirement
both in the extract and in the d m a t e streams of 90%, 95% and 99%, respectively. Finally,
it is seen fkom Figure 3 (for the case of purity requirement of 95%)
42 - 4.2
SYB VIriCol
4 .............. ............ 0............ ?..........................
~ 4 ..............................................................
. e
3.8 ............................................................. em 3.8 .......................................................

3.6 ............................................................ .......................................................

E 3.4 .............................................................
A A A
3.4 .... .... ~L
...... . . ~..... d ..... .... *----
.

3 2 ...................................................................... 32 ........................................................

f g............0...........................
0 0 0 0 0
3 ............x............ x.............x .....x ..... x... r ......
3 ................... I......
X
2.8 4 1 2.8 A
1.5 1.8 1.7 1.8 1.0 1.5 1.8 1.7 1.8 1.9 2
F, mllmln F. mllmln

Figure 3. Flowate ratio parameters ml to nu for the SMB and Varicol processes with 95%purity requirement.

that the optimal values of the flow rate ratios in sections 2 and 3, m2 and m3, in the SMB and
in the Varicol process are very similar (although the corresponding optimal performances
are different) and change very little as the feed flow rate increases. This is consistent with
the triangle theory, which indicates that the optimal operating point in the (mz, m3) plane is
independent of the feed and eluent flow rates [1,2].

68
3.3 Comparison of SUB, Varicol and PowerFeed
The separation problem examined in this case requires the simultaneous maximization of
the rafiate (PR)and the extract purity (PE)for a given feed flow rate, F,eluent flow rate, D
and fixed configuration of the unit, for the same chiral separation considered in section 3.2.
This optimization problem in the case of PowerFeed operation can be represented
mathematically as follows:
Max Ji = PR[Qz,~,.*., Qz,s, Fi, Fs-I,DI,*-,b - 1 , t XI
MaX Jz = PE[Qz,i, .**, Qz,sy FI, *-) Fs-I,Di,--, Dsi, 4, XI
Subject to PR2 90%
P E 90%
~
Q1=27.5ml/min, N,I=~ or 6, D,,=6.24 ml/min, Fa,, NNTp
where Q2s, Fi and DI represent the flow rate of section 2, the feed flow rate and the eluent
flow rate in the P subinterval,respectively. The average feed and eluent flowrates (Fave and
Dave respectively) are fixed; therefore the flow rates in the last subinterval are not
independent since they are determined by the corresponding values in the previous
subintervals. Note that in this case it has been assumed that the flowrates Q2, F, D and
consequently Q3, Q4,R and E, change in S subintervals. The problem can be simplified by
changing in time less flow streams, e.g., only F,which implies the change of Q3 and R.It is
to be noted that in the problem above both the SMB and the Varicol operation modes have
three decision variables, i.e. Qz,t,and x. The same stage in series model described by Zhang
et al. [6]has been adopted to simulate SMB, Varicol and PowerFeedprocesses, with a slight
revision, which enables the column flowrates to change in time.
A comparison of the SMB, Varicol and PowerFeed operations has been conducted for
two sets of operatingconditions, one with FBve=1.62ml/min and NNTP=80and the other with
F,,,=2.2 ml/min and Nm=60. Two different PowerFeed configurations, one varying F in 4
subintervals ( S 4 ) and the other varying Q2,F and D in 3 subintervals(S=3), are considered
for the two cases, respectively. In the first configurationthe decision variables reduce to Q2,
FlyF2, F3, t,and x, while in the second they are Qz,,, Q ~ J Q , F2, DI, D2, t, and x. The
, ~ JF1,
optimization results of a 5-column PowerFeed unit are compared to those of the
corresponding SMB and Varicol units in Figure 4(a) and 4(b) for the two sets of operating
conditions, respectively. It is interesting to observe that firstly, in both cases for the SMB,
Varicol and PowerFeed units,we obtained Pareto solutions. Secondly, each Pareto has at
least one discontinuity due to change in column configuration x. This is due to the fact that
different purity requirement requires different column configuration. Thirdly, the 5-column
Varicol and PowerFeed processes perform better than the corresponding 5-column S M B
process. Finally, as the difficulty of the separation increases, as shown in Figure 4(b), the
PowerFeed performs better than the Varicol, and is even comparable to the 6-column S M B
at the two ends of the Pareto curve.
A better understanding of these results can be achieved by reasoning in terms of the
flowrate ratio parameter, m as shown in Figure 5, taking as an example the separation case
shown in Figure 4(a). It can be observed from Figure 5(a) that m2 and m3for the 5-column
SMB, Varicol and PowerFeed units are almost constant, although they tend to decreases
slightly as PRincreases. This is in agreementwith triangle theory as discussedby Bang et al.
[6].For the 5-column PowerFeed unit, the m3 values changing in time are shown in Figure
5(b). It is interestingto note that the m3of the SMB process, goes very closely to the average
m3 of the PowerFeed process, m3ave and to a slightly less degree to the m3of the Varicol

69
process due to the column configurationchange of the Varicol operation, as shown in Figure
5(a). Therefore, the optimal design of the Varicol and PowerFeed processes can be replaced
by the optimal design of the corresponding SMB process given by the equilibrium theory,
followed by the search of the optimal ports switching and flow rates variation schemes
within the switching period, respectively.

0.99 --" O O O OD=-

0.98 -. 0.98

0.87 ..
0.96 -- 0.96

2" 0.95 -.
0.94 -. 0.94

0.83 -.
0.92 -. 0.92

0.91 -.

0.94 : : : ~ ; : : i i
. ,i
I 0.9
0.9 0.91 0.92 0.93 0.94 0.85 0% 0.97 0.88 0.99 1 0.9 0.82 0.94 0.96 0.98 1
PRI' PR
Figure 4. Optimal solutionsfor the SMB, Varicol and PowerFeedprocesses (x changes with increasing PR:5-col
SMB and PowerFeed: (a) and (b) 1/2/111-+1/1/2/1; 6-column SMB: (a) 1/212/1+211/2/1, (b) 1/2/2/1; 5-col
Varicol: (a) 111/~1-111/2/1-112/1/1-1/2/111+1/1/211-1/1/2/1-1/1/211-1~/1/1~1/112/1-1/11211-1/11211-2111111,
(b) 1~11211-112/1~1-1/2/111+1/1/2/1-1/1/211-1/2/1/1).
3.9 3.8
(a) comparison of the m2 and m3 (b) m,in lhe 4 subintervals oflhe P o w i f f e e d process
3.7 3.7

3.6
3.5
" w W ' A S ~ $ = M 3.5
E 3.3
U
E" 3.4
(.3.1
E 3.3

2.8
3.2
A SMB, m3 AVarkol n0 rFvwerFeed.ave tr3
2.7 3.1
0 SMB,nQ Varkol nQ FuwerFeed, nQ

2.5 3
0.8 0.91 0.92 0.93 0.94 0.95 0.88 0.97 0.88 0.99 1 0.9 0.91 0.92 0.93 0.94 0.95 0.88 0.97 0.88 0.99 1
PR PR
Figure 5. Comparison of (a) the average m 2 and m3 among the 5-column SMB, the Varicol and the PowerFeed
processes and (b) Values of m3 in the 4 subintervals of the PowerFeed process corresponding to the points in
Figure 4(a).

4 Conclusions

Our results show that the Varicol and the PowerFeed operations improve the performance of
the SMB process, particularlywhen the total number of columns is small and the separations
are difficult. The optimal design of the Varicol and PowerFeed processes can be replaced by
the optimal design of the corresponding SMB process given by the equilibrium theory,
followed by the search of the optimal ports switching and flow rates variation schemes
within the switching period, respectively.

70
References
1. Storti G., Mazzotti M., Morbidelli M. and Carra S., Robust design of binary
countercurrent adsorption separation processes, AICHE J. 39 (1993) pp. 471-492.
2. Mazzotti M., Storti G. and Morbidelli M., Optimal operation of simulated moving bed
units for nonlinear chromatographic separations, J. Chromafogr. A 769 (1997) pp.
3-24.
3. Biressi G.,Ludemann-HombourgerO., Mazzotti M., Nicoud R.M. and Morbidelli M.,
Design and optimization of a simulated moving bed unit: role of deviations from
equilibrium theory, J. Chromatogr. A 876 (2000)pp. 3-15.
4. Klatt K.U., Hanisch F. and Dunnebier G., Model-based control of a simulated moving
bed chromatographic process for the separation of hctose and glucose, J. Process
Contr. 12 (2002) pp. 203-219.
5. Ruthven D.M. and Ching C.B., Counter-current and simulated counter-current
-
adsorption separation processes, Chem. Eng. Sci. 44 (1989) pp. 1011 1038.
6. Zhang Z., Hidajat K., Ray A.K. and Morbidelli M., Multiobjective optimization of
simulated moving bed system and Varicol process for chiral separation, AZChE J.
(2002) in press.
7. Zhang Z., Mazzotti M. and Morbidelli M., Multiobjective optimization of SMB and
Varicol processes using genetic algorithm,J. Chromafogr.A (2002) in press.
8. Mazzotti M., Storti G.and Morbidelli M., Supercritical fluid simulated moving bed
chromatography,J. Chromafogr.A 786 (1997) pp. 309-320.
9. Di Giovanni O., Mazzotti M., Morbidelli M., Denet F., Hauck W. and Nicoud, R.M.,
Supercritical fluid simulated moving bed Chromatography 11. Langmuir isotherm, J.
Chromafogr.A 919 (2001) pp. 1-12.
10. Denet F., Hauck W., Nicoud R.M., Di Giovanni O., Mazzotti M., Jaubert J.N. and
Morbidelli M., Enantioseparation through supercritical fluid simulated moving bed
(SF-SMB) chromatography,Ind. Eng. Chem. Res. 40 (2001) pp. 4603-4609.
11. Migliorini C., Wendlinger M., Mazzotti M. and Morbidelli M., Temperature gradient
operation of a simulated moving bed unit, Ind. Eng. Chem. Res. 40 (2001) pp.
2606-2617.
12. Jensen T.B., Reijns T.G.P.,Billiet H.A.H. and van der Wielen L.A.M., Novel
simulated moving-bed method for reduced solvent consumption, J. Chromafogr.A 873
(2000) pp. 149-162.
13. Antos D.and Seidel-MorgensternA., Application of gradients in the simulated moving
bed process, Chem. Eng. Sci. 56 (2001) pp. 6667-6682.
14. Abel S., Mazzotti M. and Morbidelli M., Solvent gradient operation of simulated
moving beds I. Linear isotherm, J. Chromafogr.A 944 (2002) pp. 23-29.
15. Ludemann-Hombourger O., Nicoud RM. and Bailly M., The Varicol process: a new
multicolumn continuous chromatographic process, Sep. Sci. Technol. 35 (2000) pp.
1829-1862.
16. Kloppenburg E. and Gilles E.D., A new concept for operating simulated moving-bed
processes, Chem. Eng. Technol. 22 (1999) pp. 8 13-817.
17. Zang Y. and Wankat P.C., SMB operation strategy-Partial feed, Znd. Eng. Chem. Res.
41 (2002) pp. 2504-25 1 1.
18. Bhaskar V., Gupta S.K. and Ray A.K., Applications of multi-objective optimization in
chemical engineering,Rev. Chem. Eng. 16 (2000) pp. 1-54.
19. Srinivas N. and Deb K., Multiobjective function optimization using nondominated
sorting genetic algorithms, Evol. Compuf.2 (1 995) pp. 22 1-248.

71
ADSORPTION TECHNOLOGY FOR GAS SEPARATION

SHIVAJI SIRCAR
Department of Chemical Engimering, Lehigh University
11 1 Research Drive, Iacocca Hall, Bethlehem, Pa 18015-4791, USA.
E-mail: sircar@aoi.com

Separation and purification of gas mixtures by selective adsorption on micro-meso porous


solid adsorbents such as zeolites, activated carbons, silica and alumina gels, polymeric
sorbents, etc., has found numerous commercial applications in the chemical,
petrochemical, environmental, medical, and electronic gas industries. Table 1 lists some
of the key uses of this technology [I]. Two generic cyclic process concepts called
Temperature Swing Adsorption (TSA) and Pressure Swing Adsorption (PSA) are
generally employed. Each of these concepts have numerous variations depending on (a)
the product specifications, (b) the energy of separation, (c) the sequences and the modes
of operation of the steps of the process, (d) the types of adsorbent used, etc. The gas
purification applications typically use the TSA processes except for gas drying and
solvent vapor recovery applications where both the TSA and the PSA processes are used.
The bulk gas separation applicationsexclusively use the PSA processes.

Table 1. Key Commercial Applications of Gas Separation and Purification by Adsorption Technology

T 'lut R val
Trace Organicand inorganic Impurity Removal Air Separation (4 and NZfrom Air)
hDrYing Hydrogen and Carbon Dioxide Roduaion from
Air PollutionControl Steam-Methane Reformer W G a s
Nuclear Waste Management Roduction of Ammonia SynthesisGas
Solvent Vapor Recovery Hydrogen Recovery fromRefinery off Gases
Electronic Gas Purification Methane-CarbonDioxide w o n from Landfill Gas
Carbon Monoxide-HydrogenSeparation
-
N o d Isoparaffn Scparahon
rn Alcohol Dehydration

The interest and growth in the research and development of adsorptive separation
processes have been phenomenal. Table 2 lists the number of U.S. patents cited by the
Derwent Chemical Patent index between the years of 1980 and 2000 under the keywords
given in the table [I]:
Table 2. Results of U.S. Patent Search Between 1980 and 2000

Kevwords NumberofP@nts m r d s m a l t s
Gas Separation by Adsorption 3050 Air Separation by PSA 391
Adsorption for Air Pollution 1164 Hydrogen Purification by PSA 185
Ressun swing Adsorption 608 Gas Drying of PSA 32
Temperature Swing A m t i o n 60

Gas adsorption has become the state-of-the-art technology for (a) trace impurity removal,
(b) small to large scale (1 40,000 SCFH) gas drying, (c) small to medium scale (0.0 1-1 00
TPD) production of O2 (90+%) and N2 (99+%) from air, and (d) small to large scale

72
(1-100 MMSCFD) production of high purity H2 (99.999+%) from steam-methane
reformer and refinery off-gases. Research and development on adsorptive processes has
primarily been directed towards (a) producing purer products (single or multiple) from a
feed gas at higher recoveries (b) lowering the capital and operating costs of separation, (c)
designing novel hardware and process control systems, and (d) increasing the scale of
applications. Some of the key achievements include (a) lowering the specific power ( 4 2
KW/T/D) for production of 90% O2 from air below that of conventional cryogenic
distillation air separation route, (b) direct production of high purity N2 ( 4 0 0 ppm 02)
from air, and (c) production of high purity H2with high recovery (90+%). The primary
reasons for this spectacularp w t b in this area are given below [2]:

(a) There is an extra degree of thermodynamic freedom for describing adsorption


systems compared to those for conventional gas separation methods like distillation
and absorption. This introduces an immense flexibility in the design and operation
of adsorptive separation processes.
(b) Numerous families of porous adsorbents are available which offer multiple choices of
core adsorptive properties (equilibria, kinetics, and heats) for a given separation
application.
(c) A successfuladsorptive process is generally a good marriage between the optimum
adsorbent and the efficient process design.
(d) There can be many different paths (combinationsof materials and processes) for
achieving the same separation goals.

The above reasons are also the key driving forces for promoting innovations in this area.

The points discussed above can be demonstrated by the case of simultaneous production
of O2 and N2 enriched gases from ambient air. Air can be fractionated by selectively
(thermodynamic) adsorbing N2 over O2 and Ar on a zeolite [3], or selectively (kinetic)
adsorbing O2over N2 and Ar on a molecular sieve carbon [4,5].

Furthermore, many different process schemes for air separation can be developed using
different zeolites and molecular sieve carbons having different adsorptive properties.
For example, a zeolite like Na-mordenite having a moderate N2 selectivity of -4 over O2
at ambient conditions can be used in a four-step PSA process [3]. The cyclic steps
would include (a) adsorption of N2at near ambient conditions by flowing air through a
packed bed of the zeolite, while producing the O2enriched product gas at feed air pressure
(PA),(b) rinsing the adsorber cocurrently with a stream of essentially pure N2 and venting
the air-like effluent, (c) evacuating the adsorber counter-currently to pressure PD and
withdrawing a N2 rich product gas, a part of which is used in step (b), and finally (d)
counter-currently pressurizing the adsorber from PD to PA with a part of product gas
generated by step (a).

The N2 rinse step is needed to displace the co-adsorbed and the void 0 2 remaining in the
column at the end of step (a) so that the desorbed gas in step (c) is essentially pure N2.
This step can be eliminated if the N2 selectivity over O2by the zeolite is high (say >8) as
in the case of CaX zeolite. The desorbed gas, in this case, can be hctionated in order to

73
reject the N2lean earlier part, and collect the N2rich latter part as the N2product gas [3].
Higher Nz selectivity of an adsorbent allows the desorbed gas fractionation concept to be
practical, which leads to a simpler three-step process scheme, while meeting the product
specificationsof the former four-step cycle.

The cyclic steps of one of the PSA processes using the molecular sieve carbon as the
adsorbent consist of (a) flowing compressed air through a packed bed of the carbon so that
0 2 can diffuse and adsorb into the carbon pores faster than N2and Ar and produce a N2
rich product gas at feed air pressure (PA), (b) pressure equalizing the adsorber with a
companion adsorber, (c) counter-currently depressuring the adsorber to near ambient
pressure to produce the O2 enriched gas, (d) pressure equalizing with another adsorber,
and finally (e) repressurizingthe adsorber to PAwith feed air [4].

Table 3 shows an example of the comparative performances of these three processes. All
of them can produce a 99+%N2enriched product gas. The two zeolite processes also
produce a 85-90% O2 enriched product gas. The O2 product purity of the carbon sieve
process is, however, low. This shows that different adsorbents can be married with
different process schemes to obtain similar product purities but different process
performances (recovery, productivity, product pressure, etc.).
Table 3. ComparativePerformancesof Various Air Fractionation Processes

Process Adsorbent Purity Productivity Recovery Pressure


(%) (SCfh/P) (W (Psi@
(a) Na-Mordenite 84. I 31 63.0 2
(b) Ca-X 90.0 20 24.0 2
(c) Carbon Molecular Sieve 33.8 144 98.1 0

Process Adsorbent Purity Productivity Recovery Pressure


(a) Na-Mordenite 99.0 122 65.0 0
(b) Ca-X 99.0 83 30.0 0
(c) Carbon Molecular Sieve 99.0 92 49.4 104

TRENDS IN FUTURE ADSORPTIVE PROCESS DEVELOPMENTS

Two areas of development have attracted considerable attention in recent years. They are:

(a) Rapid PSA cycles for bulk gas separation


(b) Novel adsorbet designs

RaDid PSA
The concept is to use faster cycle times (seconds) than those used by the conventional
PSA cycles (minutes) in order to obtain a step change in the productivity of the process
(volume of productholume of adsorbenthour). Some of these designs use a
conventional PSA cycle but they are operated faster by appropriate changes in the
mechanical designs [6]. Others, propose novel process schemes in order to
accommodate fast cycle times [7-91. An example of the second case is a RPSA process
for air separation where a single adsorber vessel is packed with two or more (even
numbers) shallow layers of small zeolite particles (-0.5 mm). The layers are separated

74
by screens which act as pressure drop devices. The cycle steps consist of (a)
simultaneouslypressurizing and adsorbing N2from air on one layer of the adsorbent while
producing an O2 enriched product gas, and then (b) simultaneously depressurizing and
back purging the layer with a part of the O2 enriched product gas produced by a
companion layer [9].

Using a 5A zeolite as adsorbent and a total cycle time of 10 seconds (compared to


conventional PSA cycles of 60-240 seconds), the RPSA process could produce a 27.5%
O2enriched gas stream with an O2recovery of -64.1% from feed air at a pressure of 2.22
atm [3]. The 0 2 productivity rate was -2300 sfi?/ft3/hrwhich was an order of magnitude
higher than that of a conventional PSA. The process was found to be suitable for
producing 2340% 02 from air for enhanced combustion applications in cupolas,
metallurgical furnaces, etc. [lo].

Novel Designs
Conventional adsorber designs include vertical (length/diameter >1) or horizontal
(lengtlddiarneter 4)packed beds. The maximum permissible gas flow rates through
these vessels is governed by pressure drop and the possibility of local fluidization and
channeling [ 1 13. Some of these problems are solved by novel designs such as (a) radial
bed and (b) rotary bed adsorbers.

Radial Bed Adsorbers (PSA and TSA)

The adsorbent is placed in an annular section between two co-axial cylinders. The walls
of the cylinders are perforated for gas flow in a radial direction. The entire assembly is
enclosed inside the adsorber vessel with gas inlet and outlet conduits. Many different
designs are patented [ll-131. The adsorbers allow faster cycle times, lower pressure
drops, and higher gas throughputs without fluidization. However, the equipment design
is more complex and costly.

Rotarv Bed Adsorbers (TSA)

The rotary bed adsorber (also called adsorption wheel) provides a truly continuous TSA
system. It uses a shallow wheel-shaped adsorption bed that continuously turns about an
axis inside a fixed supporting frame. A section of the wheel is continuously used for
adsorbing impurities 60m a gas while the other section is continuously regenerated by
heating it with an impurity free gas. The adsorbent is made from a honeycomb-shaped
alumina substrate that can be coated with layers of silica gels, activated carbons, or
zeolites [14]. It has been used for gas dehumidification, solvent vapor recovery, VOC
removal, and deodorization of a gas stream.

NOVEL ADSORPTIVE GAS SEPARATION CONCEPTS

Two novel hybrid concepts for gas separation using adsorption technology have emerged
in recent years. They include (a) adsorbent membranes, and (b) simultaneous adsorption
and reaction.

Adsorbent Membranes

75
They consist of a thin layer ( 4 0 pm) of a nanoporous (3-lOA) carbon filmsupported on a
meso-macroporous inorganic solid (alumina) or on a carbonized polymeric structure [151.
They are produced by pyrolysis of polymeric films. The following two types of
membranes are produced:

Molecular Sieve Carbon (MSC) Membranes

The MSC membranes are produced by carbonization of PAN, polymide, and phenolic
resins. They contain nanopores, which allow some of the molecules of a feed gas
mixture to enter the pore structure at the high pressure side, adsorb, and then diffise to the
low pressure side of the membrane, while excluding the other molecules of the feed gas.
Thus, separation is based on the difference in the molecular sizes of the feed gas
components. The smaller molecules preferentially d i f i e through the MSC membrane
as shown by Table 4 [16,17].
Table 4. Separation Performance of Various MSC Membranes

Selective Surface Flow (SSF) Membranes

The SSF membranes, which are produced by carbonization of PVDC, contain nanopores
that allow all of the molecules of a feed gas mixture to enter the pore structure. However,
the larger and more polar molecules are selectively adsorbed on the carbon pore walls at
the high pressure side, and then they difiiise selectively to the low pressure side. The
smaller molecules are enriched at the high pressure side. These membranes can be used
to enrich H2 from mixtures with CI-C4hydrocarbons or from mixtures with C02 and CH4.
They can also be used to separate C&-H2S and H2S-H2 mixtures. Table 5 compares
performances of SSF carbon and polymeric PTMSP membranes for H2 enrichment from
FCC off gas [15]. Clearly, the SSF membrane is much superior for this application.
Table 5. Separation Performancesof SSF and PTMSP Membranes

Commnent Reiections Hi& Pressure Product Comwsitions (%)


Gas Components SSF PTMSP SSF PTMSP
C3H8 98.2 86.0 1.16 6.33
c3H6 98.8 86.0 I .49 12.20
CZH6 94.1 76.0 2.30 6.33
c
2H4 93.3 72.0 2.03 5.19
CH4 52.0 46.0 41.46 32.73
Hz 40.0 40.0 51.56 36.19

SimultaneousAdsomtion and Reaction

76
The conversion of reactants to products, as well as the rate of product formation, of an
equilibrium controlled reaction can be increased by removing a product from the reaction
zone, according to the Le Chatelier's principle. Adsorption has been used to achieve this
goal by using admixtures of catalysts and adsorbents in packed bed reactors. Process
concepts called "pressure swing reactors" have been proposed when the adsorbent is
periodically regenerated by using the principles of PSA [181.

Recently, a novel scheme called "Sorption Enhanced Reaction Process (SERP)" was
developed for direct production of essentially carbon oxide free ( 4 0 ppm) H2 enriched
gas stream (90+%) containing CH., as the primary impurity by steam-methane reforming
(SMR). The process could be operated at a much lower temperature (400-5OO0C) than
that needed by the conventional SMR reactor (800-900C), and yet produce high
conversion of CH4 to H2 [19-201. The cyclic steps of the SERP consisted of (a)
sorption-reaction, where a gaseous mixture of CH., and H20 is passed through the reactor
and a stream of COXfree H2 enriched gas is directly produced at feed gas pressure (PA),
counter-current depressurization of the reactor to near ambient pressure and venting the
effluent gas, (b) counter-current evacuation of the reactor to pressure PD and purging the
reactor with steam at PD while discarding the C02 rich effluent gas, and finally (c)
counter-currently pressurizing the reactor fiom PD to PA with steam. The reactor is
maintained at 400-5OO0Cthroughout the entire cycle.

Table 6 shows an example of the cyclic steady state performance of the SERP concept
using an admixture of a SMR catalyst (noble metal on alumina) and a C02 chemisorbent
(KzC03promoted hydmtalcite) in the reactor [20]. The reactor temperature was 4 9 O O C .
The feed H20:CH4ratio was 6: 1. The concept can directly produce -95% H2 product
(dry basis) with a CH., to H2 conversion of 73%. The trace impurities in the product gas
contained less than 40 ppm COP The corresponding product gas composition
(thermodynamic limit) of a SMR reactor operated without the C02 chemisorbent will be
-67.2% H2, 15.7% CH.,, 15.9% C02, and 1-2% CO (dry basis), and the CHI to H2
conversion will be only 52%. Thus, the SEW concept may be attractive for direct
production of a CO free H2 stream for fuel cell applications.
Table 6. Performance of the SEW Concept for H2 Production

-
Feed 1-
Press. (psia) Feed Purge Steam Hydrogen Product
26.2 0.60 1.88 0.25

Hvdroeen Product Puritv IDrvl Methane Conv.


H2 (9'0 cH4 (YO CO m ToH dro en 9'0
94.4 5.6 40 Not detected 73

Removal of Bulk COi from a Wet HiPh TemDerature Gas

Another interesting example of using a chemisorbent (Na20 supported on alumina) in a


PSA process is direct removal and recovery of bulk C02 from a wet high temperature feed
gas without pre-drying or cooling the feed [21]. A gas stream at 200 C containing 10
mole % C02 and an inert component (dry basis), and which is saturated with water, can be
treated to simultaneously produce a COz depleted stream (<3% C02) and a COz enriched
gas (99+%)without removing the water. The PSA cycle steps consist of (a) adsorption at

77
-125 Psia, (b) co-current C02 rinse at that pressure with recycle of effluent gas as feed,
-
(c) counter-current evacuation to 2.5 Psia with simultaneous steam purge, and (d)
counter-current pressurization with a part of C02depleted product gas from step (a). The
effluent gas from step (c) is partly used as the purge gas in step (b) after recompression,
and the balance is withdrawn as the recovered C02 product gas. The inert gas and COZ
-
recoveries from the feed gas are, respectively, 100 and 78% [21]. This application may
be attractive for green house gas emission control (C02sequestration from a hot flue gas
or C02 removal fkom combustion gases for power generation).

REFERENCES

1. S. Sircar, Adsorption, 6,359 (2000).


2. S. Sircar, "Applicationsof gas separation by adsorption for the hture," Adsorption
Sci. Tech., in press.
3. S. Sircar, M. B. Rao, and T. C. Golden, "Fractionation of air by zeolites," in
"Adsorption and its applications in industry and environmentalprotection," A.
Dabrowski (ed.), Elsevier, New York, pp. 395423 (1999).
4. K. Knoblauch, H. Heimbach, and B. Harder, U.S.Patent 4,548,799 (1985).
5. N. C. Lemcoff and R. C. Gmelin, U.S. Patent 5,176,722 (1 993).
6. B. G. Keefer and D. G. Doman, WIPO International Publication No. W097/39821
(1997).
7. R. Jones, G. E. Keller, and R. C. Wells, U.S. Patent 4,194,892(1980).
8. S. Sircar, U.S. Patent 5,071,449 (1991).
9. M. Suzuki, T. Suzuki, A. Sakoda, and J. Izumi, Adsorption, 2,111 (1996).
10. S. Sircar, Adsorption, 2,323 (1996).
11. U. Von Gemmingen, Linde reports on Science and Technology v. 54 (1994).
12. M. Poteau and S. Eteve, U.S. Patent 5,232,479 (1993).
13. J. Smolarek, F. W. Leavitt, J. J. Nowobilski, E. Bergsten, and J. H. Fassbaugh, U.S.
Patent 5,759,242 (1998).
14. T. Hirose and T. Kuma, "Honeycomb rotor continuous adsorber for solvent recovery
and dehumidification," 2nd Korea-Japan Symp. on Separation Technology (1 990).
15. S. Sircar and M. B. Rao, "Nanoporous carbon membranes for gas separation," in
Recent advances on gas separation by microporous ceramic membranes," N.
Kanellopoulos (ed.), Elsevier, New York, p. 473 (2000).
16. Carbon Membranes Ltd., Israel, trade literature.
17. C. W. Jones and W. J. Koros, Carbon, 22,1419 (1994).
18. G. G. Vaporciyan and R. H. Kadlec, AJChE J., 33, 1334 (1987).
19. S. Sircar, J. R. Hufion, and S. Nataraj, U.S. Patent 6,103,143 (2000).
20. W. E. Waldron, J. R. Hufton, and S. Sircar, AIChE J., 47, 1477 (2001).
21. S. Sircar and C.M.A. Golden, US.Patent 6,322,612 (2001).

78
CARBON COMPOSITE MEMBRANES

MOTOYUKI SUZUKP
The UnitedNations University,5-53-70Jingumae, Shibuya-ku, Tokyo 150-8925, JAPAN
E-mail: suzuki@q.unu.edu

AKIYOSHI SAKODA, SANG-DAE BAE, TAKESHI NOMURA


Institute of Industrial Science, Universityof Tokyo, 4-61 Komaba, Meguro-Ku, Tokyo 153-8505,
JAPAN.
E-mail: sakoda@iis.u-tokyo.ac.jp

YUAN-YAO LI
Department of Chemical Engineering, National Chung-Cheng University,Chia-Yi 621, TAIWAN,
R 0.C.
E-mail: chmyyl6Jccu.edu.m

A novel membrane named carbon whisker membrane (CWM) is introduced for an advanced
filtration process. The C W M is a tubular ceramic membrane with a layer of vapor-grown carbon
fibers (VGCF) grown on the surface of the membrane. We employed chemical vapor deposition
(CVD) of methane with the presence of catalysts to fabricate the CWM and examined its
performances on filtration processes. It was found that, in the process, the VGCFs on the
membrane acted as buffers for preventing impurities such as micro-particles or bio-materials in
the liquid not to direct contact with the membrane body so that the problem of fouling which
commonly occurred in the conventional membrane separation processes can be solved. As a result,
the life time of the membrane and the flux of filtrate were enhanced. Moreover, the VGCFs layer
has a self-cleaning hnction, that is, the flexible fibers can remove attached particles easily during
the filtration or a simply cleaning process after the filtration. This phenomenon is similar to the lotus
effect known in the environmental nanotechnology. In addition to the CWM, we report another
type of carbon membrane called carbon-coated ceramic membrane. The membrane consists of a
tubular porous ceramic substrate which is made by sintering of granular ceramics and a thin carbon
film was coated on granular ceramic by CVD method. By controlling the thickness of deposited
carbon film, we enabled to control the pore size of the carbon-coated ceramic membrane.
Moreover, the carbon thin film changes the properties of the membrane such as hydrophobicity,
thermal conductivity and electrical conductivity. These membranes were all designed for solving
water treatment issues.

1 Introduction

In the past few decades, membrane technologies have emerged as an important means of
separation in a variety of industries. Membrane separation technology offers the
advantages of, at least, high selectivity, low energy consumption, moderate cost to
performance ratio, and compact and modular design in comparison with other separation
processes. In addition, inorganic membranes such as ceramic and carbon membranes
possess characteristics of high thermal stability, chemical compatibility, and good
mechanical strength over membranes made of organic polymers. It is therefore that
inorganic membranes have come into wide use in the water/ wastewater treatment
industries. However, fouling/cake generation problem is one of the challenges for the
membrane separation technology. The particles, colloids, macromolecules or
bio-materials in the water permeation process will deposit on the membrane surface and

79
generated secondary film. As a result, membrane will be blocked and permeate flux will
diminish with time [l-31.

h t u s effect is one of the significant concepts in environmental nanotechnology. In


nature, fine wax crystals with around 1 nm in diameter coated on the leaves of lotus plants
and the fine hydrophobic structure repels water. As a consequence, the water droplets
bead up and the surfaces stay dry even during a heavy rain. Self-cleaning of lotus leaves
means that, if the surface of leave inclines, the droplets pick up small particles of dirt on
the surface as they roll. The leaves thereby keep clean and fresh. The lotus effect has
now become well known and biomimic materials based on lotus effect have extensively
been developed in the industries [4]. The nanoparticles can be mixed with paints so that
painted objects, say outside wall of houses, have a rough surface with the presence of
nanoparticles. While rain falls, the nanoparticles on the wall act as an interface between
dirt and objects resulting lotus-effect and a self-cleaning function. Aerosol with
nanoparticles can also be sprayed on paper, leather, textiles and masonry to become
advanced materials with a self-cleaningproperty.

A layer of carbon whiskers grown on the surface of objects can also play a self-cleaning
role. This is because that the whiskers avoid dirt or particles directly attach on the
objects and, because of minute contact surface between whiskers and dirt, the flexible
whiskers can easily remove adsorbed dirt or particles fiom the contact surface while
altering the flow rate, direction and form of the flow. Depending on the size and density
of the grown whiskers, the carbon whisker layer is capable of keeping substrate clean from
dirt with different sizes and also bio-materials which was known as an antibacterial
property. If the substrate is a porous membrane and a filtration process is conducted, the
carbon whisker layer will play characters of self-cleaning and fouling reduction. It is
therefore that the permeant flux and separation efficient will be enhanced.

We report two types of membrane namely, carbon whisker membrane (CWM) and
carbon-coated ceramic membrane. The CVD method and iron catalysts were used to
fabricate a layer of carbon whiskers grown on the surface of the tubular ceramic
membrane. This novel membrane is called carbon whisker membrane (CWM). We
demonstrate the formation of the CWMs with various whisker sizes and membrane pore
size as well as their performance on the filtration process using PMMA particles. A
comparison of cleaning process with conventional membranes will also be discussed.
Carbon coated ceramic membrane were also fabricated by CVD method. By controlling
experimental factors such as concentration of carbon source, pyrolysis temperature or
deposition time, we demonstrate that it is possible to adjust the thickness of the carbon
thin film so that the pore size of the membrane can be regulated. It is therefore that the
membranes can be used for microfiltration, ultrafiltration or even nanofiltration. Figure
1 shows the cross-section of the different type of the membrane. Figure 1(A) is ceramic
membrane made by sintered granular ceramics. We used this ceramic membrane as a
substrate for fabrication of C W M or carbon-coated ceramic membrane. Figure 1(B) is
activated carbon membrane. The membrane is made by spin dipping & coating of a
layer of carbon source on the surface of the ceramic membrane. By thermal activation
process, a layer of porous and activation carbon thin film can be formed. Figure 1(C) is
carbon-coated ceramic membrane fabricated by CVD method. Clearly, the membrane is

80
different fiom activated carbon membrane. Figure 1(D) is the carbon whisker membrane
(CWM).

(A) Ceramic Membrane (C) Carbon-Coated


Ceramic Membrane

Cerami; Substrate/ Carbon- Ceramic


Membrane Film Substrate

(B)Activated Carbon Membrane (D) Carbon Whisker Membrane

pore Carbon L,ayer Carbon Whisker

CeramidSubstrate Carbon Ceramic


Film Substrate
Figure 1 Cross-sections of (A) Ceramic Substrate, (B) Activated Carbon Membrane, (C) Carbon-Coated
ceramic Membrane, and (D) Carbon Whisker Membrane

2 Experimental

To grow carbon whiskers on the membranes, we employed our recipes fkom the previous
studies [5, 6]. We used methane as carbon source to make carbon whiskers and nitrogen
as carrier and diluting gases. Femc sulhte was employed as catalyst precursor to make
Fe catalysts. The recipes offer a cost-effective production and safe process. The
experiments were carried out with a tubular quartz reactor and a furnace. A porous
ceramic tube from Kubota Co., Japan was firstly coated with aqueous ferric sulfate
solution using dipping & coating technique and then dried at room temperature overnight
prior to be suspended in the center of the reactor. A designed pipe line allowed to reverse
feed flow in the tubular reactor was employed for the fabrication. The temperature was
then ramp up to 1000C with nitrogen as a purging gas to form Fe catalyst particles fiom
the precursor. After the temperature reaches the steady state, methane was then introduced
into the reactor and started to pyrolysis and deposition. In order to ensure a consistent
carbon deposition profile for a uniform layer of carbon whiskers, the direction of the feed
flow was reversed periodically after a desired deposition time. The CWM can be
therefore fabricated.

Filtrations for CWM were carried out with a cross flow membrane separation system. It
consists of a cross flow membrane module, a feed pump, a recirculation pump and process
pipes. The feed steam contained 1000 ppm PMMA with a mean particle size of 0.8

81
micrometer for the experiments. Backflushing technique to regenerate membrane was
employed for one minute in every 20 minutes filtration process in order to remove
attached particles. For the fabrication of carbon-coated ceramic membrane, the materials,
apparatus and processes were all the same but the presence of catalyst particles. The
measurement of the pore size of the membranes were conducted by gas permeability.
The experiments were carried out at room temperature with different gases such as oxygen,
nitrogen, argon and helium 171.

3 Results and discussions

3.I Carbon whisker membrane


Figure 2 shows the top-view and side-view of the CWMs under the observation of SEM.
As can be seen, the whiskers were grown on the substrate randomly to form a network
structure on the surface of the membrane. The results show that the diameter of carbon
whisker is about 100 nanometers and the thickness of the carbon whisker film is about 2
micrometers. However, the diameter, density and layer thickness of carbon whiskers is
strongly dependant of experimental conditions such as deposition time, deposition
temperature, concentration of methane, feed flow rate and catalyst concentration.
Different experimental or operational conditions will change the characters of CWMs.
Figure 3 shows the CWMs with different size of carbon whiskers. In three CWMs, the
one on the top has a thickest and longest carbon whisker layer but smallest pore size of
membrane. The diameter of visible carbon whiskers is about a few minimeters while the
length is about 1-2 centimeter. It was known that, if the size of carbon whiskers
increases, the pore size of the membrane decrease. This is because that, while the CVD
of methane is in process, carbon species not only deposited on the catalyst particle to grow
carbon whiskers but also deposited on the surface of the substrate to form a carbon film on
the substrate. As the deposition time increased, diameter and length of carbon whisker
increase and the carbon film thicken, which causes the decrease of the pore size of the
membrane. The middle CWMs has carbon whiskers which vaguely can be seen. The
diameter of whiskers is in micrometer scale while the length can reach to a few minimeters.
The last CWM has to be characterizedby SEM observationjust like Figure 1 showed.

(A) Top-view of CWM


Figure 2 (A) Top-view and (B) Side-view of Carbon whisker membrane

82
Figure 3 Carbon whisker membranes with different size of carbon whiskers. The one on the top has a thicker
and longer carbon whisker layer but smaller pore size of membrane compare to the rest of the two CWMs.
Figure 4 shows the filtration performance between carbon whisker membrane (CWM) and
carbon membrane (CM). The results clearly reveal that the permeant flux of CWM is
higher than that of CM even after a few cycles of filtration. Because of the large amount
and network structure of carbon whiskers, particles will be trapped in the carbon whisker
layer before contacting porous membrane substrates during filtration processes. As a
result, the pores of substrate will not be blocked, which will not cause the fouling
phenomenon and not reduce the life-time of the membrane. In addition, if we used pulse
flow for the feedings or changed the direction of the feedings, the flexible whiskers might
remove the attacheaadsorbed particles away as a self-cleaning function. Figure 5 shows
the result of backflushing cleaning process for the membrane after the filtration. We
found that particles attached on the CWM can be easier to remove than that of the CM.
Within the same cleaning time, particles (white color) were still attached on the CM while
no obvious particle was found on the CWM. These results lead to suggest that the CWM
possess a higher flux permeability, a self-cleaning character and a easier regeneration
process of the membrane.
. -

0.013
-
0.012

o.oll
0.010
t
0
PMMA: lOOOppm (0.8 pm),AP 0.1 kgf/cm*
Backflushing :dirtiled water, lmin,2 kgflem'
'
10 20 30 40 50 60 70
I

80
Time Wn.01
Figure 4 Flux versus time for crossflow membrane filtration with backflushing cleaning

83
Figure 5 Observation of PMMA remod on membranes after backflushing cleaning procedue

3.2 Carbon-coatedceramic membranes


Carbon-coated ceramic membranes were made by the CVD of methane without the
presence of the catalysts. Figure 6 shows the ceramic membrane and carbon-coated
ceramic membranes with different pore size. We simply controlled the deposition time
so that the thickness of the carbon film is different resulting the different pore size in each
membrane. As can been seen, the top membrane has a brighter color and, under the
observation of the surface by SEM, it was found to be a dense membrane. The pore size
can be reduced to nano-scale. The second membrane &om the top has submicro pores
which can be used as an ultrafilter. It is no doubt that the deposition time to fabricate
ultrafilter is less than that of the nanofilter. If only a little carbon deposited on the
ceramic membrane, the pore size of the membrane might just change slightly. However,
the properties of the membrane altered dramatically such as hydrophobicity, electrical
conductivity and thermal conductivity because of the nature of carbon. These changes
create a new research field for a better separation process by membrane technology such
as membrane filtration with assistance of electrically conductivity, hydrophobicity or
magnetic field separation [8]. Gas permeability can also be used for examining the pore
size of the membrane. Figure 7 shows nitrogen flux against transmembrane pressure
through Membranes A, B, and C. It is clear that a higher pressure drop increase the
nitrogen flux for all the membranes. Under the same condition of the pressure drop, the
nitrogen flux through Membrane A is the highest. This result suggests Membrane A has
a bigger pore size so that more nitrogen can flow through. Therefore, the pore size of the
membrane A is biggest followed by Membrane B and then Membrane C.

4 Conclusions

We reported our recent developed membranes called carbon whisker membrane (CWM)
and carbon-coated ceramic membrane. The CWM performed a better permeant flux in
the filtration process in comparison with the membrane without whiskers. Also, C W s
have a self-cleaning function which can increase the separation efficiency during the
filtration and increase the lift-time of the membrane. The carbon-coated ceramic
membranes with various pore sizes can be made for the purpose of nanofiltration,

84
ultrafiltration or microfiltration. It was suggested that the unique properties of the
carbon coated ceramic membranes such as good hydrophobicity, thermal conductivity and
electrical conductivity should be taken into account while designing an advanced
membrane separation process.

Figure 6 Carbon-coated ceramic membrane with different pore sizes

0.6

0.5

5 0.4
I r a

B
-
E
' 0.3
a
2 0.2
E
0.1

0
0 25 50 75 100 125 150 175 200
P w u n drop (mmHg)

Figure 7 Nitrogen flux against transmembrane pressure through MembranesA, B, and C

References

1. Flemming, H.C.,Schaule, G.,Schmitt, J. and Tamachkiarowa, A., Biofoulng- the


Achilles heel of membrane processes, Desalination, 1997, 1 13,215-225.
2. Vrouwenvelder, H.S., van Paassen, J.A.M., Folmer, H.C., Hofjnan, J.A.M.H.,
Nederlof, M.M.and van der Kooij, D., Biofouling of membranes for drinking water
production, Desalination, 1998, 1 18, 157- 166.
3. Abd El Aleem, F.A., Al-Sugair, K.A., Alahmad, M.I.,Biofouling problems in
membrane processes for water desalination and reuse in Saudi Arabia, International
biodeterioration& biodegradation, 1998,4 1, 19-23.

85
4. Bemhard, S., Edwin, N., Markus, O., 2002, Geometyrical shaping of surfaces with a
lotus effect, Patent US2002164443, EP1238717, DElOllO589, JP2002321224.
5. Li, Y.Y.,Bae, S.D., Sakoda, A. and Suzuki, M., 2001, Formation of vapor grown
carbon fibers with sulfuric catalyst precursors and nitrogen as carrier gas, Carbon,
39(1), 91-100.
6. Li, Y.Y., Bae, S.D., Sakoda, A. and Suzuki, M., 2000, Fabrication and
characterization of carbon whisker, The 2nd Pacific Basin Conference on Adsorption
Science and Technology, May 14-18,Queensland, Australia, p376-380.
7. Li, Y.Y.,Nomura, T.,Sakoda, A. and Suzuki, M., 2002, Fabrication of carbon coated
ceramic membranes by pyrolysis of methane using a modified chemical vapor
deposition apparatus, Journal of Membrane Science, 197 (1-2), 23-35
8. Li, Y.Y.,Sakoda, A. and Suzuki, M., 2001, A review of fabrication methods of
carbon membranes and applications related to their hydrophobic and electrically
conductive properties, Journal of Institute of Industrial Science, University of Tokyo,
53(3), 70-74.

86
Invited Papers
This page intentionally left blank
ON THE DOMINANT ROLE OF ADSORPTION EFFECTS IN
HETEROGENEOUSCATALYSIS

JOERl F. DENAYER, GIN0 V. BARON


Dienst Chernische Ingenieurstechnieken, Vrije UniversiteitBrussel, Pleinlaan 2, I050 Brussels,
Belgium E-mail: gvbaron@vub.ac.be

DIRK DEVOS, JOHAN A. MARTENS AND PIERRE A. JACOBS


Centrefor Suflace Chemistry and Catalysis, K. U.Leuven, KasteelparkArenberg 23, 3001 Leuven,
Belgium E-mail:pierrejacobs@gr.kuleuven.ac. be

Adsorption effects in heterogeneous liquid phase reactions are seldom considered or recognized as
important. Liquid phase adsorption is also poorly understood for typical reaction mixtures and hardly
any models are available for its description. In this contribution we will briefly review some of the
methods available for measuring gas and liquid phase adsorption (batch tests, gas and liquid
chromatography, various detection methods) and describe a few applications in heterogeneous
catalysis with zeolites. Concentrations in the micropore system of such zeolite catalysts can differ
strongly from the external liquid bulk phase concentrations, and adsorption and diffusion phenomena
essentiallycontrol reaction rate and product selectivity.A good understandingof these phenomena can
help in selecting the best solvent or catalyst support for the reaction, understand deactivation and may
be instrumental in correct modeling of the reaction. Two examples from hydrocarbon conversion and
partial oxidation are presented.

1 Introduction

Physisorption is often included in gas phase reaction models, and its importance well
recognized, but the parameters are often estimated from reaction data, leading to poor fits
and biased parameters, often leading to erroneous mechanisms. Independent measurement
of reactant and product adsorption in near reaction conditions allows for elegant modeling
of complex reaction systems. For a variety of heterogeneously catalysed gas and liquid
phase reaction systems, we have systematically determinedthe adsorption properties of the
catalyst in close to reaction conditions. Given the high temperatures, high pressure,
multi-phase conditions, only a few of the usual techniques are really applicable. These are
mainly gas or liquid chromatography, both in tracer and perturbation mode and batch
adsorption experiments which we currently run in a robotic system as these measurements
are very labor intensive. These data are then used to model quantitatively the reaction
system, and allowing to extract intrinsic rates of reaction, or used to analyse the behaviour of
the reaction when changing zeolite composition, type, pore size or solvent used. Two
examples from our work are briefly described next as an illustration of the methodology.

2 Epoxidation of a-alkenes with Ti-zeolites

Ti-substituted zeolites (Titanium Silicalite-1 or TS-I, Ti+) are truly heterogeneous


catalysts, and they can use H202as the oxidant. As TS-I hardly decomposes any H24,
yields on peroxide basis are usually excellent. This has qualitatively been ascribed to the
hydrophobic nature of TS-I,disfavoring peroxide accumulation in the intraporous volume.
A second important characteristicof TS-1 epoxidations is that they are generally fastest in

87
methanol. This preference has been explained by a model in which methanol coordinateson
the active Ti sites [1,2]. The possible link between physisorption and solvent effects in
olefin epoxidation, however, has not extensively been studied. In the case of alkene
oxidation with TS-I ,the slower reaction of larger alkenes has been attributed to diffusion
limitation of these molecules. These measurements were however performed at high
conversion where deactivation was dominating.
Tracer liquid chromatographic methods were used to study liquid phase adsorption on
TS-1, Ti-Beta and Ti-MCM-41 and completely elucidate the reaction properties [3].
Partition coefficients for alkenes, alkanes, epoxides, and other, polar products are strongly
dependent on the carrier solvent. Linear a-olefins are concentrated inside the TS-1
micropores, particularly when methanol is the solvent. This agrees well with the superior
initial rates of olefin epoxidation with TS-1 in methanol. Sorption also governs the relative
reactivities of olefin substrates, especially in competitive experiments. Thus, in truly initial
conditions, I-hexene is less reactive than I-octene or I-nonene. For the latter substrates,
however, deactivation is fast, especially in methanol. This process is related to the strong
adsorption of higher 1,2-epoxyalkanes in TS-1 in methanol. Deactivation due to
competitive epoxide adsorption is slower in acetone, making this a more suitable solvent
than methanol for 1-nonene epoxidation with TS-1. Physisorption effects play a dominant
role in the small pore TS-1 catalyst, due to the close interaction of substrates such as alkenes
with the pore wall. Wider-pore catalystssuch as Ti-Beta and especially Ti-MCM-41, do not
adsorb olefins as selectively and hence intraporous olefin concentrations and reaction rates
are much lower.

3 Hydroeonversion of n-alkanes

Bifhctional zeolite catalysts such as platinum loaded acid zeolite catalysts are applied
in several petroleum refinery operations, designated as hydroconversion processes:
isomerisation of light naphtha, iso-dewaxingand hydrocrackingof heavy fractions [4]. Most
experimental investigations in academic laboratories are typically performed with pure
model components or simple mixtures thereof as feedstock, and using reaction conditions
under which the hydrocarbon compoundsare in the vapor phase. Industrialhydroconversion
processes are mostly run under three phase, or even in some cases under liquid phase
conditions and with feedstocks that are extremely complex mixtures of large numbers of
different hydrocarbon compounds [4].
In earlier work, adsorption equilibria of a broad range of hydrocarbon components
were determined on ultrastable Y (USY)zeolites under catalytic conditions. It was found
that Henry adsorptionconstants increase exponentiallywith the alkane carbon number [5,6].
This adsorptive discrimination of long alkanes in favor of shorter ones is most pronounced
in the Henry regime and decreases when the pores are more filled up with hydrocarbon
molecules at increasing vapor pressures [7]. USY type zeolite contacted with liquid alkanes
does not show any differences in affinity towards alkanes of different molecular weight [8].
These investigations showed that on USY type zeolite, adsorption selectivity depends
strongly on the loading of the micropores and on the aggregation state of the alkane
contacted with the zeolite.
The conversions of n-alkanes, cyclo-alkanes and their mixtures over WUSY zeolites
under vapor phase reaction were successhlly modeled using independently determined
adsorption equilibria and fundamental reaction networks based on alkylcarbenium ion
chemistry [9-121. In these models, the intrinsic reaction rate of an individual reaction step

88
such as a branching rearrangement via a protonated cyclopropane is the same in all
molecules and at all positions in the carbon chain. Intrinsic reactivity differences between
alkene intermediates are accounted for by one single parameter, reflecting differences in
protonation enthalpy [lo]. Under vapor phase reaction conditions on an W S Y type
catalyst, there is always preferential conversion of the heaviest alkane owing to its
preferentialadsorption. In absence of adsorption selectivityin the liquid phase, it is however
expected that the relative reactivities of the two hydrocarbons reflect the intrinsic reaction
kinetics of vapor phase experiments.
Figure 1 shows the result for gas phase conversion of an equimolar heptane - nonane
mixture over Pt/USY type zeolite catalyst CBV720 (SVAI = 13, Zeolyst), and demonstrates
it is indispensable to use appropriate expressions for the multicomponent adsorption
equilibria. For vapor phase conditions resulting in significant pore filling, the Langmuir +
interaction model is appropriate.

0.9

0.8

0.7

; o.s
P
g
OS
0.4

0.3

0.2

0. t

0
0.0 0.1 02 0.3 0.4 0.5 0.9 0.7 0.8 0.0 1.0
Avenp. convenion

Figure 1 Conversion of heptane and nonane from Figure 2 Conversion of heptane and nonane From
their binary equimolar mixture in vapor phase their binary mixture in liquid phase plotted against
conditions at 230 O C and different spacc times the average conversion (symbols: experimental
(symbols: experimental data points, full curves: data points, full curves: Langmuir adsorption
Langmuir adsorption model, dotted curves: model, dotted curves: no selective adsorption).
Langmuir + Interaction model).

Figure 2 shows the selectivity for the longer alkane is strongly reduced in liquid phase
and that a non-competitive model for adsorption is needed in that case to represent the
results. Using a Langmuir type model based on gas phase data rather approximates the gas
phase results and is completely in error for the prediction of liquid phase conversion.
Both in vapor and in liquid phase reaction conditions, nonane is more reactive than
heptane. The reactivity difference is however much more pronounced in the vapor phase. In
USY zeolite micropores exposed to the vapor of the two n-alkanes, the heaviest alkane
molecule is preferentially adsorbed, resulting in a higher apparent reaction rate. When the
alkane mixture is fed in the liquid phase, the competing alkanes are adsorbed in a
non-selective manner in the micro- and mesopores of USY.Consequently, in liquid phase
conditions the relative reactivity of the n-alkanes corresponds to the relative intrinsic
reactivities.

89
4 Conclusion

These two examples show that adsorption effects can be dominant over the other
phenomena in a spectacular way and need to be taken into account directly through
independent measurement. Understandingthese phenomena can help in selecting, designing
or optimizing catalysts or reaction conditions.

5 Acknowledgements

This research was financially supported by FWO Vlaanderen (G.0127.99). J. Denayer


is gratehl to the F.W.0.-Vlaanderen, for a fellowship as postdoctoral researcher. The
involved teams are participating in the IAP-PA1 programme on Supramolecular Chemistry
and Catalysis, sponsored by the Belgian Federal Government.

References

1. Bellussi, G., Carati, A., Clerici, M.G., Maddinelli, G., and Millini, R., J. Cuful. 133,
220 ( 1992).
2. Clerici, M.G., and Ingallina, P., J. Cuful. 140,71 (1993).
3. Langhendries G., Baron G.V., De Vos D.E. and Jacobs P.A. (1999) J. Catal., 187,
453-463
4. Maxwell, I.E. and Stork, W.H.J, Stud. Surf.Sci. Catal. 137,2001,747.
5. Denayer, J.F.M., Baron, G.V., Jacobs, P.A., Martens, J.A., J. Phys. Chem. B 102 (1 7),
1998,307.
6. Kiselev, A.V., Adv. Chromatogr. 4, 1967, 1 13.
7. Denayer, Joeri F. M.; Baron, Gino V., Proc. Fundamentals of. Adsorption 6., Ed.
Meunier, F., 1998, Elsevier, 99- 104.
8. Denayer, J.F.M., Bouyermaouen,A., Baron, G.V., Ind. Eng. Chem. Res. 37 (9), 1998,
369 1.
9. Denayer, J.F.M., Baron, G.V., Martens, J.A., J. Catal. 190 (2), 2000,469.
10. Thybaut, J. W., Marin, G. B., Martens, J. A., Jacobs, P. A., Baron, G. V., J. Catal. 202,
200 1,324.
11. G. G. Martens, G. B. Marin, J. A. Martens, P. A. Jacobs, G. V. Baron, J. Catal 195,
2000,253.
12. Denayer, J.F.M., Baron, G.V., Martens, J.A., PCCP 2(5), 2000, 1007.
13. Denayer, J.F.M., De Jonckheere, B., Hloch, M, Marin, G.B., Vanbutsele, G., Martens,
J.A., Baron, G.V., J. Catal210,2002,445.
SUPERCRITICAL ADSORPTION:PARADOX, PROBLEMS, AND INSIGHTS

LI ZHOU
High Pressure Adsorption Laboratory, School of Chemical Engineering, Tianjin Universiv,
Tianjin 300072. P R China
E-mail: zhouli@ublic.tpt.Q.cn

Abstract. Paradoxes, problems and ideologies in the study of supercritical adsorption were
discussed. A macroscopic interpretation of supercritical adsorption was presented basing on a
general model that derived at from the Gibbs definition and a straightforward method of determining
absolute adsorption. The model does not include any assumption, but relies on experimental data
and keeps the formal continuity of adsorption theory. It was shown to apply for wide ranges of
temperature and pressure, and bore an impact to the characterizationof adsorbents.

1. Problems of supercritical adsorption

1.I Is supercritical ahorption an artificial topic?


The adsorption of gases at above-critical temperatures is referred to as supercritical
adsorption. The first paradox about supercritical adsorption is it is an artificial topic and
attracts onlyf a v researchers, which was heard at comments on manuscripts devoting to
the topic. It did not, in deed, attract must interest before, but many works were dedicated
to it in recent years following the need of clean fuels, such as natural gas and hydrogen.
Storage of these fuels on-board vehicles constitutes the bottleneck of the technology, and
adsorptive storage seemed promising [I ,2]. The storage temperature of interest is much
higher than the critical temperature of fuel gases, therefore, supercritical adsorption is the
theoretical base of adsorptive storage. It is also the theoretical base of adsorptive
separation processes [3,4]. Besides, it was suggested for the characterization of porous
materials [ 5 ] . Because of the importance of supercritical adsorption, a special session was
organized for it on FOA7.

I .2 Relation and difference between sub- and super-critical adsorption


The adsorption below and above the critical temperature looks different as shown by
experimental isotherms. There are five types of isotherms below the critical temperature
depending on the porous structure of adsorbents [6],however, there is only one type of
isotherms at above-critical temperatures no matter what kind of adsorbents was tested.
The supercritical isotherm shows type-I feature for its initial part, but there must be a
maximum followed by a negative increment as long as pressure is high enough or
temperature is low enough. Each type of isotherm reflects a definite mechanism o f
adsorption, therefore, supercritical adsorption must have just one mechanism. All
adsorption isotherm equations available, for example, the Langmuir or the BET equation,
were derived at from a prescribed mechanism and explained the experimental isotherms
satisfactorily. Therefore, clarification of the unique mechanism of adsorption at above
critical temperatures is necessary for setting up a pertinent model of supercritical
adsorption.

91
I .3 Application ofpotential theory to supercritical adsorption
Facing the fast development of industrial techniques, progress in theoretical studies of
supercritical adsorption is left far behind the need of engineering requirement. The
Polanyi-Dubinin potential theory [7,8] is universal in nature, but cannot be applied
directly to interpret supercritical adsorption data. Theoretically, the amount adsorbed at
above-critical temperatures might be predicted basing on a characteristic curve, which is
independent on temperature and, thus, could be constructed basing on the isotherms at
subcritical temperatures. However, the characteristic curve was defined as the plot of
adsorbed volume versus the adsorption potential. If we use the curve to predict the amount
adsorbed at above-critical temperatures, we must specify the value of adsorption potential.
The adsorption potential was defined as the compression energy of conveying a mole gas
from the gas phase to the adsorbed phase, i.e., A =
c Vdp . The density of the adsorbed

phase, pa,and the pressure corresponding to f i in the integral is, however, not known.
Therefore, there is not a way to spec!& the adsorption potential at above-critical
temperatures and, hence, the knowledge of potential theory available cannot directly solve
the problems encountered at above-critical temperature.

1.4 Classical versus new models of supercritical adsorption


To explore the mechanism of adsorption, a mathematical model was usually
presented for experimental isotherms. If a model fits experimental data well, the
assumption underlined the model would be a tentative "theory". A pertinent model is also
required by process simulation of PSA. Therefore, modeling supercritical isotherms
attracted much research interest and different methodologies were presented. Classical
models of adsorption isotherms, including Langmuir, BET, and the potential theory
models such as Dubinin-Astakhov equation, correlate the adsorption data with adsorption
condition through few (two to three) parameters. Such models are simple in form and have
been applied successfully to subcritical adsorption, therefore, were tried to apply for
supercritical region [8,9]. However, the difference between sub- and super-critical
adsorption was overlooked previously. It was well known that all isotherm equations were
initially derived at for absolute adsorption, but the isotherms obtained experimentally are
excess quantities. The difference between the excess and the absolute adsorption is
negligible for the subcritical condition, therefore, the absolute adsorption models work
well also for the excess isotherms. However, the difference between the excess and the
absolute adsorption becomes large at high pressures and the classical models do not work
well for the excess supercritical isotherms. Typically, they cannot describe isotherms with
maximum. Then new models that can describe isotherms with maximum such as
Ono-Kondo equation [lo], models basing on density function theory [11,12], and the
models derived from equations of state of gases [13,14] were proposed. The DFT
modeling method is similar to GCMC in that the state of adsorbate at the minimum of
grand potential energy was searched and, hence, a lot of computation work is required.
Computational models may describe the excess isothenns well, but deviated off the
classical way of modeling adsorption isotherms. The philosophy underlined the two kinds
of modeling is different. There is a thermodynamically distinct interface between the
adsorbed phase and the bulk gas according to Gibbsian point of view, and the density
profile in the normal direction of solid surface is not continuous. However, such interface

92
was omitted, and a continuous density profile in the normal direction of solid surface was
assumed in the computations basing on minimizing the grand potential energy. Rome is
there, and many ways lead to Rome. The results obtained by different methods may
contribute to better understanding supercritical adsorption, therefore, microscopic or
macroscopic way of study should not expel each other.

1.5 Volumetric versus gravimetric method of adrorption measurement


Most adsorption data were collected by volumetric method until microbalance of
high sensitivity appeared few years ago. It can hardly say which method is superior to the
other, and both methods need the value of the skeleton volume of sample adsorbent. This
volume has to be subtracted fiom the whole volume of the sample container to obtain the
volume of void space, which is used for the calculation of the amount adsorbed. The
skeleton volume of sample adsorbent was directly used in the calculation of buoyancy
correction in gravimetric method. This volume was usually determined by helium
assuming the amount of helium adsorbed was negligible. If, however, helium adsorption
cannot be omitted, error would yield in the skeleton volume and, finally, in the calculated
amount adsorbed. However, the effect of helium adsorption would be much less for
volumetric method if the skeleton volume is considerably less than the volume of void
space, but the volume of void space cannot affect buoyancy correction. In this respect,
helium adsorption would result in less consequenceon volumetric method especially when
the skeleton volume was determined at room temperature and pressures less than 15 MPa.
The skeleton volume (or density) was taken for a parameter in modeling process in some
gravimetric measurements. However, the true value of skeleton volume (or density) can
hardly be more reliable basing on a fitted parameter than on a measured value. Therefore,
one method of measurement cannot expel the other up to now, and the consequence of
helium adsorption in the measured amount adsorbed should be estimated appropriately.

2 A classical description of supercritical adsorption

The Gibbs definition of adsorption is valid not only for subcritical adsorption, but
also for supercritical one:
n = n, - Vapg (1)
The experimentally measured amount adsorbed, n, is an excess quantity concentrated in
the so-called adsorption space over the bulk gas phase, and the so-called absolute
adsorption, n , corresponds to the total quantity of adsorbate in the mentioned space.
According to Gibbs, there is a thermodynamically distinct interface between the bulk gas
and the equilibrium adsorbed phase, the density of which (pa)is remarkably higher than
that of the bulk gas phase @B). V, is the volume of the adsorbed phase, and the product
V& accounts for the difference between n and nt.
All isotherm equations were derived at for absolute adsorption. Therefore, there are
more choices for the expression of nt. However, the initial part of supercritical isotherms
looks like Type-I, and a simple model proposed for this kind of adsorption on
heterogeneous surface [151 was applied
n = n$ - exp(- bp4)l- rapg (2)
Such a model bears prominent merits: 1. It keeps the formal continuity of isotherm

93
equations available to date; 2. It keeps the simplicity of isotherm equation because there
are only three parameters to be determined experimentally: &', b and q. This equation is
similar to the Langmuir-Freundlich equation and so does the parameters. Parameter np is
a saturation quantity of n, since n,' if p=m. Parameter b is related to the energy
change of adsorbate after adsorption, and parameter q is an index of surface heterogeneity.
There are two unknowns in Eq.(l): n, and V,. Either one is determined, the other could be
evaluated from the equation. Previous efforts were to determine n, basing on an assumed
V, [161 or the density of the adsorbed phase (pb) [171. However, a value of nt determined
basing on experimental data would be preferable.

3 A straightforward method of determining absolute adsorption

Because n, is the totai mass confined in the adsorbed phase, it must vary with the
experimental condition, therefore, should be determined as a function of temperature and
pressure. A straightforward method was proposed by the author [ 18-19]. It is known from
Eq.1 that n = n, if Vs, can be neglected comparing to n. Therefore, we can use the
experimental values of n that comply with the constraint to formulate the model of
absolute adrorption. The experimental data experienced twice transformations to reach a
linear plot as was usually done for the establishment of a model for a set of data. The
experimental data were utilized to the utmost in the transformation processes, and the data
that do not comply with the constraint were sifted out. A plot of In[ln(dk)] versus lllnp 0,
in kPa) was thus constructed. Parameter Swas used to adjust the magnitude of n in order
to avoid evaluating the logarithm of negative numbers. Its value could be set at 1, 10 or
100. A model with two parameters were obtained from the linear plots for the absolute
adsorption isotherms:

[(
n, = exp exp a + -
:PI1
(3)

The value of parameter a and p is a function of temperature and, therefore, n, is a function


of temperature and pressure. The volume of the adsorbed phase, V, could be determined
via Eq.( 1) from the measured n and the Eq.(3)-determined n,. So, Eq.(2) could be used to
model an isotherm of supercritical adsorption. It must emphasized that Eq.(2) is the model
of supercritical isotherms, and there are only three parameters since the value of n, (or V a
could be determined the other way if it is available.
Perfect fit was observed at experimental isotherms of different adsorption systems in
large ranges of temperature and pressure [20,21]. Shown in Fig.1 and Fig2 are only
examples. The model for absolute adsorption isotherm was obtained basing on the data at
relatively low pressure, but the model fits the data at high pressure as well. It is concluded
that the adsorption mechanism of supercritical adsorption does not change as pressure
increases although maximum or even negative (excess) adsorption was observed.

4 Continuity of the model and its impact on the characterization of adsorbent

Eq.(2) together with Eq.(3) describe isotherms not only for supercritical, but also for
subcritical if the isotherms show type-Ifeature. It was shown recently [22] that this model
works well for the isotherms distributed densely around the critical temperature as shown

94
in Fig.3. Although 298 K is lower than the critical temperature (304.2 K), maximum is
shown on the isotherm. Therefore, the difference between the excess and the absolute
adsorption cannot always be neglected for subcritical region. The evaluation of absolute
adsorption is especially important for the characterization of adsorbents basing on the
adsorption isotherm of C02at 273 K [23] and the isotherm shows type-I feature. Shown in
Figd is a calibration isotherm of COz at 273 K for a sample of activated carbon, where
dots are the data measured, and the solid line is the absolute adsorption isotherm
generated by Eq.(3). Since 273 K is below the critical temperature (304 K), the density of
COz gas is so low that the product Vfig is much less than the value of n for most
conditions. As consequence, most experimental points locate on the absolute adsorption
isotherm. More points would also locate on the linear plot shown in Fig.5, which renders
the formulation of absolute isotherms more reliable. However, the difference between the
absolute and the excess isotherm becomes considerable as pressure approaches saturation.
This difference is important because both pore volume and the specific surface area are
determined by the highest amount adsorbed. The volume of micropores was usually
estimated by dividing the amount adsorbed at relative pressure of pips = 0.95 by the
density of saturated liquid if the isotherm is type4 [24]. The assumption underlined is the
full-filled pore space with liquid adsorbate. However, the amount of adsorbate determined
as such does not correspond to the total mass fillkd in the space, but only a part of it,
although the major part. It is the absolute adsorption that corresponds to the total mass in
the adsorbed phase, therefore, the volume of pores should be determined basing on it. As
is seen in Fig.4, the absolute isotherm intersects the condensation line when pressure
reaches saturation, which proves a convention that the adsorbed phase has the same state
as saturation liquid. It is thus concluded,pore volume should be evaluated by dividing the
saturated absolute amount by liquid density. For example, the pore volume of the sample
shown in FigA would be 1.52 cm3/g basing on the absolute isotherm compared to 1.27
cm3/g basing on the excess isotherm. The relative difference is as large as 20%. The
surface area was usually determined by the DRK plot if the calibration isotherm is type-I
[24]. The difference in the DRK plot between the excess and the absolute adsorption is
also considerable as shown in Fig.6 for the same sample. The intercept of the two plots is
1.43 and 1.49 respectively yielding a relative difference of 15% in the surface area.

5 Is there a border of supercritical adsorption?

It is well known that the saturation pressure, ps, is the border of subcritical
adsorption, beyond which another phenomenon, condensation, happens. It seems there is
not a similar border for supercritical adsorption because no matter how high the pressure
is, adsorption isotherms could always be recorded. However, such border must exist
considering the cause of adsorption. It is the interaction between gas and solid that causes
the density difference between gas phase and adsorbed phase. The strength of the
intermolecular (atomic) attraction force is limited, therefore, the density difference
resulted must be limited, and the pressure difference corresponds to the density difference
must be limited. So, there is no reason to think of supercritical adsorption (or
high-pressure adsorption as synonym)as borderless.
A proof for the existence of the border can be found in a parameter of the
above-mentioned model of supercritical adsorption isotherms. Parameter np in Eq.(2)
indicates a saturated amount of absolute adsorption, which is the maximum adsorbate

95
contained in the adsorbed phase, therefore, presents the border of adsorption. The linear
plots of isotherms provide another proof. Shown in Fig.7 is such a plot for the adsorption
of hydrogen on activated carbon AX-2 1 for 77- 298 K [181. All plots intersect at one point,
which defmes a limit state of supercritical adsorption. The representation of isotherm in a
system of n versus pgprovides h t h e r proof of existing border of supercritical adsorption.
As Menon [25] pointed out, such isotherm has a linear section after the maximum. This
was proven by the adsorption of C02 on activated carbon for the near-critical region [22]
as shown in Fig.8. Applying Eq.(l) to the linear section of the isotherm, one would
conclude that both V. and n, must be constant if isotherm is linear. It means the value of 4,
V,, and does not change although pressure keeps increase for the region of linearity.
However, such state of the adsorbed phase is not thermodynamically stable. The density
of bulk gas & keeps increase while the density of the adsorbed phase pa keeps constant.
The metastable state would be broken if pgbecomes larger enough than A.As a symbol of
the breakage of the metastable state, the recorded "isotherm" begins to deviate from
linear.
Determination of the border of supercriticaladsorption is very important for the study
of supercritical adsorption. For example, the fugacity at the border can be used to replace
pr to define adsorption potential or used to draw the 4 plot. In fact, most theories
available for subcritical adsorption would be possible to apply for supercritical adsorption
if the border could be properly defined.

Acknowledgements

This work is subsidized by the Special Funds of Major State Basic Research Projects
(G2000026404) and supported by the National Natural Science Foundation of China
(#I29936100).

References

1. Matranga K. R., Mayers A. L. and Glandt E. D. Storage of Natural Gas by


Adsorption on Activated Carbon. Chem. Eng. Sci. 47 (1992) pp. 1569-1579.
2. Noh J. S., Agarwal R. K. and Schwarz J. A. Hydrogen Storage Systems Using
Activated Carbon. lnt. J. Hydrogen Energy 12 (1987) pp. 693-700.
3. Yang R.T. Gas Separation ly Adsorption Processes (Butterworths, London, 1987).
4. Ruthven D. M., Farooq S. and Knaebel K. S. Pressure Swing Adsorption (VCH
Publishers, New York, 1994).
5. Nguyen C. and Do D. D. Adsorption of Supercritical Gases in Porous Media:
Determination of Micropore Size Distribution, A Pbs. Chem. B 103 (1999) pp.
6900-6908.
6. IUPAC Commission on Colloid and Surface Chemistry Including Catalysis, Pure
Appl. Chem. 57 (1985) p. 603.
7. M. Polanyi, Verh. Deut. Physik. Ges. 1914, 16, 1012; 1916, 18,55.
8. Dubinin M. M. and Astakhov V. A. Trans. from lzyestiya Akademii Nauk SSSR,
Seriya Khimicheskaya, 1971, No. 1 : 5- 1 1.
9. Ozawa S., Kusumi S. and Ogino Y., Physical adsorption of gases at high pressure, IV.
An improvement of the Dubinin-Astakhov adsorption equation. J. Colloid &
Interface Sci. 56 (1976) pp.83-91.

96
10. Aranovich G. L. and Donohue M. D. Adsorption of supercritical fluids, J. Colloid
& hte$ Sci. 180 (1996) pp.537-541.
11. Jiang S. Y., Zollweg J. A. and Gubbins K. E. High-pressure adsorption of methane
and ethane in activated carbon and carbon fibers, J. Phys. Chem. 98 (1994)
pp.5709-57 13.
12. Chen J. H.,Wong D. S. H., Tan C. S., Subramanian R., Lira C. T. and Orth M.
Adsorption and desorption of carbon dioxide onto and from activated carbon at high
pressures. Ind Eng. Chem. Res., 36 (1997) pp.2808-28 15.
13. Aranovich G. L. and Donohue M. D. Surface compression in adsorption systems,
Colloi& andSurfaces A, 187-188 (2001) pp.95-108.
14. Usthov E. A., Do D. D., Herbst A., Staudt R. and P. Harting, Modeling of gas
adsorption equilibrium over a wide range of pressure: A thermodynamic approach
based on equation of state, J. Colloid & Inteface Sci. 250 (2002) pp.49-62.
15. Zhou, L., Zhang, J.-Sh. and Zhou, Y.-P. A Simple Isotherm Equation for Modeling
the Adsorption Equilibria on Porous Solids over Wide Temperature Ranges,
Langmuir 17 (2001) pp.5503-5507.
16. Sipersteh F., Talu 0. and Myers A. L. Gas storage: absolute adsorption versus
excess adsorption, Proceedings FOA7 (2001) pp.3 1 1-318.
17. Staudt R., Saller G., Tomalla M. and Keller J. U. A note on gravimetric
measurements of gas-adsorption equilibria. Ber. Bmsenges. Phys. Chem. 97 (1 993)
pp.98-105.
18. Zhou L. and Zhou Y .-P. Linearization of Adsorption Isotherms for High Pressure
Applications, Chem. Eng. Sci. 53 (1998) pp.253 1-2536.
19. Zhou L. and Zhou Y.-P. A Mathematical Method for the Determination of Absolute
Adsorption from Experimental Isotherms of Supercritical Gases, Cn J. Chem. Eng., 9
(2001) pp.110-115.
20. Zhou Y.-P., Bai Sh.-P., Zhou, L. and Yang, B. Studies on the Physical Adsorption
Equilibria of Gases on Porous Solids over a Wide Temperature Range Spanning the
Critical Region-Adsorption on Microporous Activated Carbon, Cn J. Chem. 19
(2001) pp.943-948.
21. Zhou L., Zhou Y.-P., Bai Sb-P. and Yang B. Studies on the Transition Behavior of
Physical Adsorption from the Sub- to the Supercritical Region: Experiments on Silica
Gel, J. Colloid & Inteflace Sci. 2002 (in press).
22, Bai Sh.-P. Studies on the adsorption behavior of C 0 2 for the near-critical region,
Dissertation for PhD degree (2002) School of Chemical Engineering, Tianjin
University, Tianjin, China.
23. Cazorla-Amoros D., Alcaniz-Monge J., De la Casa-Lillo M. A. and Linares-Solano
A. C02As an Adsorptive to Characterize Carbon Molecular Sieves and Activated
Carbons, Langmuir 14 (1998) pp.4589-4596.
24. Gregg S. J. and Sing K. S. W. Adsorption, Surface Area and Porosity (Academic
Press, London, 1982).
25. Menon P. G. Adsorption at High Pressures, Chem. Rev. 68 (1 968) pp.277-294.

97
p.-, , , , , .. ,,

D
PYR

Fig. 1 A comparison of the Fig.2 Model fitting the data Fig3 M,,,.J~~~ the
absolute with the excess of C& on activated carbon. adsorption of co2on
adsorption isotherms of Points: experimental; activated carbon for the
C& on activated carbon Curves: model predicted near-critical region. points:
experimental; Curves:
model predicted.
0

ti1::
I

Fig.6 A comparison of the


Fig.4 A calibration isotherm Fig5 The linear plot of DRK plot between the
of C02at 273 K formulating the absolute
excess and the absolute
adsorption of COzat 273 K
adsorption

-t

P#=+rn*

Fig.7 Linear plots of the Fig.8 Linear section on the


hydrogen adsorption on isotherms of C@ on
activated carbon AX-2 1 activated carbon

98
Contributed Papers
This page intentionally left blank
MICROWAVE DRYING FOR PREPARATION OF MESOPOROUS CARBON

HAJIME TAMON, TETSUO SUZUKI AND SHIN R. MUKAI


Department of Chemical Engineering, Kyoto University, Kyoto 606-8501,Japan
E-mail:tamon@cheme.kyoto-u.acJp

TAKUJI YAMAMOTO
Research Institute for Green Technology,
National Institute of Advanced Industrial Science and Technology,
Tsukuba 305-8565, Japan
E-mail: yamamoto-t@aist.go.jp

Resorcinol-formaldehyde hydrogels were synthesized by sol-gel polycondensation of resorcinol with


formaldehyde in a slightly basic aqueous solution. RF cryogels, RF xerogels, and RF xerogels (MW
gels) were respectively prepared from RF hydrogels by freeze drying, hot air drying, and microwave
drying. Carbon cryogels, carbon xerogels and carbon M W gels were subsequently obtained by
pyrolyzing RF drygels in an inert atmosphere. Freeze drying and microwave drying were effective to
prepare mesoporous RF drygels and carbon gels. RF cryogels and carbon cryogels showed high
mesoporosity over wide ranges of the molar ratio of resorcinol to catalyst (WC) and the ratio of
resorcinol to water (WW) used in sol-gel polycondensation. Although RF xerogels had a few
mesopores, carbon xerogels had no mesopores. RF M W gels and carbon M W gels showed
mesoporosity if appropriate values of WC and W W were selected.

1 Introduction

Organic aerogels (RF aerogels) are prepared by the sol-gel polycondensation of resorcinol
(1,3-dihydroxybemene) (C6H4(OH)2) (R) with formaldehyde (HCHO) (F) and
supercritical drying with carbon dioxide ( C 0 3 [3]. Carbon aerogels are also obtained by
pyrolyzing RF aerogels in an inert atmosphere at 1323 K [4]. RF and carbon aerogels
have high porosity (> 80 %), and high surface areas (400 - 900 m2/g). The authors have
experimentally elucidated the influence of the amounts of resorcinol, water, and basic
catalyst used in the sol-gel polycondensation on porous structures of EW and carbon
aerogels [5, 61. Carbon cryogels are also prepared via sol-gel polycondensation of
resorcinol with formaldehyde, freeze drying, and pyrolysis [7-101.
The cryogels are unique materials with high surface areas and large mesopore
volumes, and they are expected to be used as catalysts, adsorbents, electric double layer
capacitors, and materials for chromatographic separation. The drying time of freeze
drying is, however, extremely long. Hence, the objective of the present work is to study
the applicability of microwave drying to the preparation of mesoporous carbons.

2 Experimental

Resorcinol-formaldehyde (RF) solutions were prepared from resorcinol (Wako Pure


Chemical Industries, Inc., research grade), formaldehyde (Wako Pure Chemical Industries,
lnc., research grade, 37wt??formaldehyde stabilized with 8 wt?? methanol), sodium
carbonate (C) (Wako Pure Chemical Industries, Inc., research grade) and distilled water
after ion exchange (W). The solutions were poured into glass tubes (inner diameter: 4 mm,

99
length: 4 cm), and gelled by curing them at 298 K for 1 day, at 323 K for lday and at 363
K for 3 days to obtain RF hydrogels. The synthesis conditions are listed in Table 1.

Table 1. Porous properties of RF drygels and carbon gels.

carbon cryogel 7.0 674 0.73 0.16


200 0.250 0.5 RF cryogel 6.2 496 1.14 ND
carbon cryogel 6.2 696 1.29 0.14
200 0.375 0.5 RF cryogel 5.5 554 1.45 ND
carbon cryogel 5.5 662 0.94 0.14
200 0.500 0.5 RF cryogel 4.2 472 1.05 ND
carbon cryogel 2.5 609 0.56 0.13
I00 0.I25 0.5 RF cryogel 7.0 492 1.17 ND
carbon cryogel 6.2 754 0.78 0.15
100 0.250 0.5 RF cryogel 6.2 542 1.06 ND
carbon cryogel 3.6 625 0.82 0.11
100 0.375 0.5 RF cryogel 2.4 558 0.79 ND
carbon cryogel 1.9 577 0.35 0.15
100 0.500 0.5 RF cryogel 2.1 548 0.72 ND
-___ carbon cryogel- .- 1.6 599 0.31 0.13
200 0.125 0.5 RF xerogel 1.8 I52 0.27 ND
carbon xerogel 1.8 49 0.07 ND
200 0.375 0.5 RF xerogel 1.8 255 0.31 ND
carbon xerogel - 1.6 121 0.06 0.03
200 0.125 0.5 RF MW gel 2.1 361 0.58 0.I4
carbon MW gel 1.6 626 0.47 0.I4
200 0.250 0.5 RF MW gel 2.2 323 0.47 0.21
carbon MW gel 2.I 582 0.48 0.41
200 0.375 0.5 RF MW gel 2.4 422 0.56 0.33
carbon MW gel 1.8 549 0.48 0.09
200 0.500 0.5 RF MW gel 2.6 465 0.68 0.25
carbon MW gel 2.0 608 0.47 0.14
100 0.250 0.5 RF MW gel 1.8 407 0.42 ND
carbon MW gel 1.5 407 0.14 0.15
WC:mole ratio of resorcinol to basic catalyst, R/W: ratio of resorcinol to water,
WF: mole ratio of resorcinol to formaldehyde, ND: not detected.

RF hydrogels were immersed into 10 times volume of t-butanol, and RF cryogels


were prepared by drying RF hydrogels at 263 K for 24 hours after pre-freezing at 243 K
for 6 hours [7]. RF xerogels were prepared by drying RF hydrogels in an oven kept at 323
K for 48 hours. RF MW gels were prepared by drying RF hydrogels in a microwave range
(SANYO Electric Co., Ltd.; EM-LAl) for 10 minutes. The changes of the weight of RF
hydrogels during drying were measured and the drying time of 10 minutes is enough to
obtain RF drygels.
Carbon gels were prepared by pyrolyzing RF drygels in a conventional furnace [3].
Nitrogen gas flowed through a quartz reactor containing RF drygels set in the h a c e at
2.0 x lo4 m3/min during the pyrolysis. The h a c e was heated to 523 K at 250 wh and
kept at 523 K for 2 hours. Then the furnace was heated to 1273 K at 250 wh and
maintained at 1273 K for 4 hours. After the pyrolysis, the reactor was cooled to a room
temperature with its own thermal mass.
The porous properties of RF drygels and carbon gels were determined by nitrogen
adsorption. The adsorption and desorption isotherms were measured at 77 K using an

100
adsorption apparatus (BEL Japan, Inc.; BELSORP28). BET surface area, SBET, mesopore
size distribution, and kcroporosity were evaluated. The pore size distribution and the
mesopore volume, V-, were determined by applying the Dollimore-Heal method [ 11 to
the desorption isotherm, and the micropore volume, V,,,i,, was evaluated by the t-plot
method [2]. The cross sections of RF drygels and carbon gels were observed using a
scanning electron microscope (JEOL, Ltd.; JSM-6340FS).

3 Results and Discussion

3. I Porous Structure of RF Drygels


Figure 1 shows the pore size distributions of RF drygels, and Table 1 also shows the
porous properties of RF @gels. r w denotes the peak value of pore size distribution. One
can see the development of mesoporosity in RF cryogels, and the values of rw and V,,
of RF cryogel decrease as decreasing WC or as increasing WW. On the other hand, SBET
shows no obvious dependence on R/C and WW. The above dependencies are considered
to reflect the structure of RF hydrogels. This is because solvent inside the pores is
removed retaining the porous structure of RF hydrogels. From Fig. 1 (b) and Table 1, it
can be seen that pore size distributions of RF xerogels have sharp peaks at around 2 nm,
and SBET and Vmes are much smaller than those of RF cryogels. Figure 1 (c) and Table 1
also show the mesoporosity of RF MW gels. The pore size distributions of RF MW gels
have sharp peaks at around 2 nm.Although it is difficult to detect obvious depencence of
the porous properties on the value of WW,the values of S ~ E Tand V,,, are respectively
estimated to be 65 - 85 % and 38 - 67 % of those of RF cryogels.

R/W=OJ25 ---D-- R/W=O350 ____ &-.. R/W=O375 - - V - . R / W = m

Figure 1. Pore size distributions of (a) RF cryogels, (b) RF xerogels and (c) RF MW gels synthesized under
the condition of WC = 200.

From the above results, it can be concluded that freeze drying is strongly
recommended to obtain mesoporous RF drygels among the drying methods examined in
this work. It is also possible to prepare mesoporous RF drygels by microwave drying.
However, it is difficult to prepare mesoporous RF drygels by hot air drying.
RF aerogels have previously been prepared over wide ranges of WC (12.5 IWC I
800) and R/W (0.125 I R/W I0.500) [6]. By comparing the drying methods from the
viewpoint of the ranges of R/C and R/W to obtain mesoporous RF drygels, supercritical

101
drying is considered to cover the widest ranges, followed by freeze drying, microwave
drying and hot air dtyiig.

3.2 Porous Structure of Carbon Gels


From Figure 2 (a) and Table 1, one can see that the values of V,, of carbon cryogels are
slightly smaller than those of RF cryogels and also see that micropores are formed on
carbon cryogels, which is caused by the shrinkage of mesopores during pyrolysis.
Interestingly, SeErincreases to some extent because of micropore formation during
pyrolysis. Figure 2 (b) and Table 1 show that carbon xerogels scarcely have mesopores
because mesopores of RF xerogels have been collapsed by pyrolysis. Figure 2 (c) and
Table 1 also indicate that the mesopores of RF MW gels have shrunk by pyrolysis. It
should be noted that carbon MW gels prepared under the condition of R/C = 200 also
show mesoporosity although rpcd and V,,, are smaller than those of carbon cryogels.
These results suggest that appropriate values of R/C and R/W should be selected to obtain
mesoporous carbon MW gels. The values of& and V,,, of carbon MW gels synthesized
under the condition of R/C = 200 are confirmed to be 83 - 99 % and 37 - 84 % of those of
carbon cryogels respectively.
The cross sections of RF drygels and carbon gels were observed by SEM and the
following results were obtained. Carbon cryogels are also composed of primary particles.
On the other hand, primary particles composing RF xerogels and RF MW gels are melted
by pyrolysis. Carbon xerogels have smooth surface and no mesopores, which fact supports
the results of nitrogen adsorption. Although the porous structure of carbon MW gels are
partially melted, the cross-linked structure is partially maintained and retaining
mesoporosity.
From the above results, freeze drying and microwave drying are recommended to
prepare mesoporous carbon gels. It is confirmed that freeze drying is available to prepare
mesoporous carbon cryogels over wide ranges of R/C (50 I R/C I 200) and R/W (0.125 I
R/W I0.500).When microwave drying is used, the value of R/C should be kept close on
200 to obtain mesoporous carbon MW gels.

- M
8
-M
8
-M
8

;1'
I
n'
I 5
I

4
I

3 4 3- 4
-
a

-
b
M
-
M
-
Q
>
Q
a 0
3a o
>
4 0
1 10 1 10 1 10

Figure 2. Pore size distributions of (a) carbon cryogels, (b) carbon xerogels and (c) carbon MW gels
synthesized under the condition of WC = 200.

102
4 Conclusion

1. Freeze drying is effective to retain porous structure of RF hydrogels. It is possible to


prepare mesoporous RF drygels and carbon gels over wide ranges of R/C and R/W.
2. It is difficult to obtain mesoporous RF drygels by hot air drying. The mesopores of
RF xerogels are collapsed by pyrolysis.
3. Microwave drying is also available to prepare mesoporous RF drygels. If the
appropriate values of R/C and R/W are selected, it is possible to obtain mesoporous
carbon gels.

5 Acknowledgements

This research was partially supported by Japan Society for the Promotion of Science,
Grant-in-Aid for Scientific Research (B), No. 14350416 (2002), Industrial Technology
Research Grant Program in '01 from New Energy and Industrial Technology
Development Organization (NEDO) of Japan, and Hosokawa Powder Technology
Foundation (2000).

References

1. Dollimore, D. and Heal, G.R., An improved method for the calculation of pore-size
distribution fiom adsorption data, J. Appl. Chem., 14 ( 1964) pp. 109- 1 14.
2. Lippens, B.C. and de Boer, J.H., Pore system in catalyst V. The t- method, J. Cutul.,
4 (1965) pp. 3 19-323.
3. Pekala, R.W., Alviso, C.T., Kong, F.M. and Hulsey, S.S., Aerogels derived fkom
multifunctional organic monomers, J. Non-Crysr. Solids,145 (1992) pp. 90-98.
4. Pekala, R.W. and Alviso, C.T., Carbon aerogels and xerogels, Muter. Res. SOC.Proc.,
270 (1992) pp. 3-14.
5. Tamon, H., Ishizaka, H., Mikami, M. and Okazaki, M.,Porous structure of organic
and carbon aerogels synthesized by sol-gel polycondensation of resorcinol with
formaldehyde, Carbon, 35 (1 997) pp. 79 1-796.
6. Tamon, H., Ishizaka, H., Araki, T. and Okazaki, M., Control of mesoporous structure
of organic and carbon aerogels, Carbon, 36 (1998) pp. 1257-1262.
7. Tamon, H., Ishizaka, H., Yamamoto, T. and Suzuki, T., Preparation of mesoporous
carbon by fieeze drying, Carbon, 37, (1 999) pp. 2049-2055.
8. Tamon, H., Ishizaka, H., Yamamoto, T. and Suzuki, T., Influence of freeze-drying
conditions on the mesoporosity of organic gels as carbon precursors, Carbon, 38
(2000) pp. 1099-1105.
9. Yamamoto, T.,Nishimura, T., Suzuki, T. and Tamon, H., Control of mesoporosity of
carbon gels prepared by sol-gel polycondensation and freeze drying, J. Non-Clyst.
Solids,288 (2001) pp. 46-55.
10. Yamamoto, T., Sugimoto, T., Suzuki, T., Mukai S.R. and Tamon, H., Preparation
and characterization of carbon cryogel microspheres, Carbon, 40 (2002) pp.
1345-1351.

103
COMPUTER SIMULATION OF TRANSPORT IN CYLINDRICAL MESOPORES

S.K.BHATIA AND D.NICHOLSON


Department of Chemical Engineering, The Universityof Queensland, Brisbane QLD 4072,
Australia

The transport of a sub-critical Lennard-Jones fluid in a cylindrical mesopore is investigated here,


using a combination of equilibrium and nonequilibrium as well as dual control volume grand
canonical molecular dynamics methods. It is shown that all three techniques yield the same value
of the transport coefficient for diffiscly reflecting pore walls, even in the presence of viscous
transport. It is also demonstrated that the classical Knudsen mechanism is not manifested, and that
a combination of viscous flow and momentum exchange at the pore wall govern the transport over a
wide range of densities.

1 Introduction

lntense world-wide activity in applicationsof newly developed templated porous materials,


carbon nanotubes and and a variety of related materials [ 1,2], has led to renewed interest
in the long-standing problem of modeling transport in nanopores. Generally this is
represented in a diffusional m e w o r k with a concentration dependent diffusivity [3-51;
however, there is much controversy regarding the underlying mechanisms. It has for
some time been considered that transport in confined spaces, such as in nanopores,
comprises of a purely diffusive component, as well as a hydrodynamic or viscous
component, [6,7]. The diffusive component is often considered dominant and arbitrarily
treated as activated surface flow [8], Knudsen diffusion [4,6,7,8] or slip flow [9],
particularly at low coverage, and in pores of near-molecular width, where the
hydrodynamic approaches [lo] predict a vanishing transport coefficient contrary to
experiment or simulations [3-51.
It has been argued [ 1 13 that the transport coefficient obtained by dual control volume
grand canonical (DCV-GCMD) simulation, in which the flux in a finite capillary under the
action of a chemical potential gradient is measured, represents the combined effects of
diffusive and viscous flow, while nonequilibrium molecular dynamics (NEMD), in which
the steady state flux in an infinite capillary under the action of a constant force is
measured, should yield only the viscous component. On the other hand it has also been
surmised [5] that equilibrium molecular dynamics (EMD) should yield only the diffusive
component since bulk flow is absent in this method. More recently it has been found
[ 121 that all three techniques yield the same transport coefficient in micropores, where
viscous flow is considered negligible. This therefore still leaves open the question
regarding the differences between the methods in larger pores where viscous transport is
significant which, along with elucidation of the transport mechanism, forms the subject of
the present study. A more detailed account of this work is reported elsewhere [13,14].

2 Molecular Dynamics Simulations

The simulations conducted model the flow of Lennard-Jones (LJ) methane at 150 K and
170 K in a cylindrical silica pore of radius 1.919 nm, having infinitely thick pore walls
comprising spherical LJ sites. For methane we use the established LJ parameter values

104
&jkB= 148.2 K,q =0.381 m. For the solid W parameters we use 4 k B = 290 K, a, =
0.29 nm, obtained by fitting argon isotherms at 87 K in MCM-41 of various pore
diameters [15], using grand canonical Monte Carlo (GCMC) simulation. The
Loren&-Berthelot d e s are used to estimate solid-fluid LJ interaction parameters. A
cut-off separation of 1.5 nm corresponding to about 3.94q is used in computing
fluid-fluid potentials.
In the molecular dynamics calculations the trajectories of methane molecules in the
pore are followed using the equation of motion with appropriate temperature control. A
diffuse reflection condition is applied at the pore wall. For the EMD simulations a
collective transport coefficient obtained from autocorrelation of the fluctuating axial
streaming velocity via a Green-Kubo relation [5]
Dm = N lirnC<
r-hw
u,(O)u,(t) > dt

where udt) = C&id. For the NEMD simulations a constant axial acceleration is applied
to the particles, and a transport coefficient computed from the measured flux as
D, = k,Tg&nr, where 1 is the axial number flux and is the methane density in the
pore. For the DCV-GCMD simulations a three-zone method is used [5,16], with the two
end wnes maintained at different chemical potentials. An effective Fickian transport
coefficient is then computed h m the measured flux through D,,a = -jUAp where I!
is the length of the central gradient zone and Apthe applied density difference.

3 Comparison of Methods

The different transport coefficient obtained at


160 M
A A
A% various densities from the three MD techniques
were essentially identical. Figure 1 [13] depicts
the transport coefficients from EMD and NEMD
at the two temperatures, showing essentially no
difference. Clear evidence of an asymptotic
non-zero transport coefficient at low densities is
also seen, with only weak density dependence in
this region. The inset depicts the comparison of a
01 . . . ' . NEYD- inKI purely
A
I

below,
no-slip viscous theory, to be discussed
with the NEMD results at 150 K, showing
o 2 4 6 a 1012
good correspondence at moderate and high
adsorbate density (nm") densities, and failure at low densities where the
Figure 1. Variation of transport
latter predicts a vanishing transport coefficient.
coefficient with density. Inset Despite the latter deviation the general trends are
depicts the comparison of a purely similar, in particular a nearly constant transport
no-slip viscous theory with the coefficient below a density of about 4 nm-3 (after
NEMD Esllitq 1 so K. an initial increase tkom a value of zero for the
theory), which corresponds to the monolayer
region. The good agreement at high density and similarity in Lend does suggest the role
of viscous effects in the transport, though an additional mechanism is also signified by the
quantitative disagreement at low densities.

105
Our results clearly indicate that the EMD
transport coefficient does include the viscous
part, despite the absence of imposed bulk flow,
N<
'O0

.
-
640
0
-
contradicting earlier assertions [5,11] discussed 0
2e 480
above. This is rationalised if cross-sectional
equilibrium of the streaming velocity and density 5
320
profiles is attained. Evidence of the latter is g
provided in the inset in Figure 2 for a pore $ 160
density of 5.95 nm-3at 150 K. E
A large number of DCV-GCMC runs were 0
also conducted at 150 K, for various pair of 0 160 320 480 640 800
density values in the two end sections, and the predicted D, x l
Rn Og (m2/s)
effective Fickian transport Coefficient, D,,,
obtained. A similar coefficient is also estimated Figure 2. Comparison of measured
from h e EMD and NEMD values of Dm,that C O e f i c i e n ~obtained using
correspond to a chemical potential gradient DCV-GCMD, with those predicted
based on coefficients obtained using
driving force, following
EMD ( 0 ) and NEMD (0). Inset
jt 1 a hf shows that equilibrium density profiles
Dt,g-= --4 = -
4 ~ ~ D ~ ~' l( npp) (T dp
- ) (2) are attained in EMD and NEMD.

which matches that from DCV-GCMD, as seen in Figure 2 [ 131, over the wide range of
densities covered. Herefis the bulk hgacity of methane in equilibrium with adsorbed
density p, and the associated derivative is estimated from isotherms obtained by GCMC
simulation. Thus, it is clear that neither NEMD nor DCV-GCMD probes any new
mechanism beyond that captured by EMD simulations, even for mesopores.

4 Transport Model

Our theoretical calculations indicated viscous flow to be dominant at high densities, as


depicted by the agreement of the theory in the inset in Figure 1. A viscous flow model
may be used over length scales larger than the mean free path, which is largely satisfied
for mesopores. To obtain the theoretical transport coefficient we solved the Navier
Stokes equation

assuming cross-sectional equilibrium with a no-slip boundary condition, which leads to

Here R, represents the pore radius measured from the center of the surface atoms, and
is the mean pore density. The radial density profile p(r) is obtained from simulations,
while the local viscosity is evaluated using the method of Chung et al. [ 171, at a density
locally averaged over a sphere of radius of/2 [ 181. The radius r, in eq. (4) represents
the position of the minimum of the fluid-solid potential, which is essentially the location

106
of the reflection boundary.
While both theory and simulation yield agreement at high densities, as seen in Figure
1, the measured non-vanishing transport coefficient at low densities does suggest a
significant degree of slip not considered in the theory. To capture this we consider the
surface boundary condition
duL
kpouo= -r;l- at r = r,
dr
where po is the local density at the potential minimum, u, is the slip velocity and k a
friction coefficient. Solution of eq. (4) with the above boundary condition yields the
transport coefficient

in place of eq. (4), which has the important feature that it predicts an asymptotic
non-vanishing transport coefficient in the low-density
region. The variation of the ffiction coefficient k
with local density p, at the surface of friction (the
location of the potential minimum) can be obtained by
substituting the values of D,,and density profile p(r)
obtained from GCMC simulation at density into
eq. (6). Figure 3 [13] depicts the results, showing a
relatively constant value of the friction constant k of
about 1.2- 1.6 N.sec.mole-' at I50 K and 1.3- 1.7
N.sec.mole-' at 177 K, for local density ,q, below a
0 20 40 80 80 100
critical value (about 85 m-3at 150 K, and 70 run3 at
density at potential minimum (nm? 177 K). Subsequently it increases steeply to very
Figure 3. Variation of friction large values, approaching the no-slip condition at high
factor with density at potential densities.
minimum, obtained at temperatures The constancy of the friction coefficient is a
of I50 K and 177 K. Inset depicts strong indicator of the importance of slip flow at the
predicted and measured streaming
pore wall in rnicropores. Further, calculations
velocity profile at adsorbed density
of 5.95 nm-3 and temperature of showed that the above values of the friction constant
150 K. are closely consistent with momentum transfer
arguments. For this we consider the frictional force
as arising from the momentum loss ondiffise reflection at the wall leading to
kp,u, = mu ,
Z (7)

where Z ( = p, V / 4) is the collision frequency under conditions of local equilibrium.


Substitution of the kinetic theory result V :=-/, yields k = -/,
providing the estimates k = 1.78 N.sec.mole-' at 150 K and 1.94 N.sec.mole-' at 177 K, in
remarkably good agreement with the values obtained from the simulation results.
Further support is obtained from comparison of the predicted streaming velocity profile
from solution of eq. (4) with that obtained from NEMD simulation. The inset in Figure 3
depicts one such comparison, obtained for an adsorbed density of 5.95 nmJ at 150 K and

107
imposed acceleration of 0.07 nm/ps*, showing excellent agreement except for some
deviation in the inner low density region where the flux is negligible.
The abrupt increase of the wall fiction coefficient to very large values leading to the
no-slip condition at a critical density is most likely due to the high frequency of
interparticle collisions in this region. Diffuse wall reflection then dissipates the collective
axial momentum in this region, and the no-slip condition is approached as seen in figure 3.

5 Conclusion

It is clear from the above results that all three simulation techniques yield identical
results for the transport coefficient in pores with diffusely reflecting walls. Further, a
combination of momentum transfer at the wall and viscous transport in the fluid suffices to
explain the transport behavior of pure component fluids in mesopores.

6 Acknowledgements

This research has been supported by a grant from the Australian Research Council under
the Large Grants Scheme.

7 References

1. M.E. Davis, Nature 417,813 (2002).


2. S . Ijima, Nature 354,56 (1991).
3. S.K. Bhatia, Proc. R SOC.Lond A446,15-37 (1994).
4. S.K. Bhatia, in Advances in Transport Processes LY,edited by AS. Mujumdar and
R.A. Mashelkar (Elsevier, Amsterdam, 1993).
5. D. Nicholson and K. Travis, in Recent Advances in Gas Separation by
Microporous Membranes, edited by N. Kanellopoulos (Elsevier, Amsterdam,
2000).
6. E.A. Mason, A.P. Malinauskas and R.B.Evans, J. Chem.Phys. 46,3 199 (1967).
7. R. Jackson, Transport in Porous Catalysts (Elsevier, Amsterdam, 1977).
8. J. Ktirger, and D.M. Ruthven, Dimion in zeolites and other Microporous solids
(John Wiley, New York,1992).
9. D. Nicholson, JMemb Sci. 129,209-219 (1997).
10. J.H. Petropoulos, and G.K. Papadopoulos, J. Memb. Sci 101,127 (1995).
11. K. D. Travis and K.E. Gubbins, Molecular simulation 25,209 (2000).
12. G. Arya, H-C. Chang. And E. Maginn, J. Chem. Phys. 115,8112 (2001).
13. S.K.Bhatia and D. Nicholson, Phys.Rev. Lett., accepted (2002).
14. S.K.Bhatia and D. Nicholson, J. Chem. Phys., submitted (2002).
15. M. Kruk. and M. Jaroniec, Chem. Muter. 12,222 (2000).
16. R.F. Cracknell, D. Nicholson and N. Quirke, Pbs. Rev. Lett. 74,2463 (1995).
17. T.H. Chug, M. Ajlan, L.L. Lee and K.E. Starling, Ind Eng. Chem. Res. 27,671
(1988).
18. 1. Bitsanis, T.K. Vanderlick, M. Tirrell and H.T. Davis, J. Chem. Phys. 89, 3152
(1 988).

108
MULTICOMPONENT MASS TRANSFER DIFFUSION MODEL FOR THE
ADSORPTION OF ACID DYES ON ACTIVATED CARBON

K.K.H. CHOY, J.F. PORTER AND G. MCKAY


Department of Chemical Engiwering, Hong Kong University of Science and Technologx Clear
WaferBay, Kowloon, Hong Kong.
E-mail: kemckayg@st.hk

The ability of activated carbon to adsorb two acid dyes, namely, Acid Blue 80 (ABSO) and Acid
Yellow (AY117) from wastewater has been studied in both single component and multicomponent
systems. Two single component systems and one binary system have been studied experimentally by
measuring the equilibrium isotherm and the concentration versus time decay curves in batch kinetic
systems. For the equilibrium isotherm studies, two single component isotherms (AB80 & AYI17)
were analysed using the Redlich-Peterson (RP) equation. For the batch kinetic studies. the effects of
initial dye concentration and activated carbon mass on the rate of Acid Blue 80 and Acid Yellow 117
removal have been investigated. A two resistance mass transport model based on film and surface
diffusion control, Homogeneous Surface Diffusion Model (HSDM), has been applied to model the
concentration decay curves in single component system. A multicomponent HSDM was developed to
predict the binary component system decay curves from the single component HSDM and the
IAST-RP model.

1 Introduction

Modeling in multicomponent adsorption systems is an extension to that of single


component adsorption. Many models have been reported in the literature for the prediction
of concentration versus time decay curves in single component batch adsorption system.
However, there are very few research papers on the topic of multicomponentmass transport
studies for liquid phase adsorption, therefore, it is a valuable contribution and novel
development to adsorption research.

2 Methods

A multicomponent HSDM for acid dyelcarbon adsorption has been developed based on
the ideal adsorbed solution theory (IAST) and the homogeneous surface diffision model (H
SDM)to predict the concentration versus time decay curves. The IAST with the Redlich-P
eterson equation is used to determine the pair of liquid phase concentrations, Cs,land C.s,2,
fiom the corresponding pair of solid phase concentrations,q5.1and qs.2,at the surface of the
carbon particle in the binary component.
The concentration of solute inside the typical particle at distance r from
the centre and at time t is defined by qi (ri, t) and the variation of 9 with distance and time
is governed by the diffusion equation for each componentj:

A Dimensionless variables are applied by defining q = Xfli, D t and ,y = L in


=s
R2 R
equation (1) to form equation (2) with following boundary condition (3) and (4).

109
do,r )= 0 = ul(x,o)(31,
The material balance equation for the multicomponentsystem and the rate of change of
solid phase concentrationof each componentj are:
' 1 =',.I - ( $ ) 9 J- (5)y E('J)= 3fpJ(xJ9rJ)xja J (6)

The mathematical problem presented by the model requires the simultaneoussolution of


equations (2) to (7). The starting point is a semi-analytical solution of equation (8) [ 11,
previously only applied to single component systems, that satisfiesthe boundary conditions

By considering a short time interval, ro.rl, r2 rn,in the integration of equation (8) forj
,

component and it is assumed that the time intervals (rI- T , - ~ ) are sufficiently short that
f(?) is effectively constant in each interval to solve equation (8) by a numerical method.

The IAS theory dictates that the spreading pressure should be constant for each
component in a given system (v,= vZ= ...= y,,)[2].

For a given pair of qe,,and qe,2,the values of sI,s2 amd qT cam be calculated.
1 s (12)
-=l++ s
9r 9z.1 4r.2
To calculate the spreadingpressure, numerical computer integration is used to solve the
' equation (1 3).

For a given (guess values) pair of qe,1 and qe,2,the values of S1,S2 and qTcan be calculated.
By applying the Redlich-Peterson equation (14) (Redlich & Peterson, 1959) in the equation
(12), it gives equation (15).
1 S. S,

Since 'y, = 'y, ,a numerical computer program is used to optimize the ct1value until
the spreadingpressures of component 1 and 2 are equal. Hence, the value of C,, and Ce,2can
be calculated from the c;., and c,q2by equations (16) and (17) while q I= c:,~ and
c8,*= c;,,.Consequently:

110
3 Results and Discussion
3. I Equilibrium Isotherm Studies
The equilibrium isotherms were measured for Acid Blue 80 (AB80) and Acid Yellow
1 17 (AY 1 17) on Activated Carbon F400.The Redlich-Peterson isotherm (equation 14) is
used to relate equilibrium concentrationsbetween liquid phase and solid phase loading. The
values of KR, aR andpare 28.32dm3/g, 103.6 (dm3/mmole)0.965, 0.965 for AB8O on activated
carbon and 55.40dm3/g, 200.2 (dm3/mmole)0.910, 0.910 for AY117 on activated carbon.
Those isotherm parameters have been used in the multicomponent HSDM to correlate the
concentration decay curve.

3.2 Batch Kinetic Studies


The effects of initial solute concentration and adsorbent mass have been studied on the
diffusion processes for the multicomponentHSDM. The HSDM was used to correlate two
acid dyes (AB80 & AY 117) in the single component system in order to determine the
relation between the single component and multicomponent systems. A comparison of the
theoretical and experimental data for the AY 1 17 in ABSO+AY 1 17 binary systems with
different carbon masses is illustrated in Figure 1. The correlated results are good for all
adsorbent mass variations for the AB80+AY 1 17 system. There is some deviation in the first
30 minutes for the three binary component systems. This is due to the single component
external mass transfer coefficient, k j l and kJ2, (see Tables 1&2) which are used in the
multicomponent HSDM model; indicating that the values of the external mass transfer
coefficient in the multicomponent system are slightly different to those of the single
component system. The Ds,iand SSE values for AB80+AY I I7 binary system with different
carbon masses are shown in Table 1. It was found that parameters D.,I and Ds,2are almost
constant for different carbon mass systems. These results are consistent with the findings in
the single component HSDM and the values of the surface diffusion coefficient in the binary
system of the acid dyes are different to those values in the single component systems (shown
in Table 2).

Figure 1: Effect of Activated


Carbon Masses on the Adsorption
of AY117 in AB8WAY117 using
MulticomponentHSDM (IASRP).

0.5 4 1
0 200 400 600 800

Time (min)
eO.6 WO.88 A I . 2 g 0 1 . 7 g X2.2g

111
Figure 2: Effect of Initial Dye
.oo
1
Concentrations on the Adsorption
of AB80 in ABSO+AY117 using
0.90
MulticomponentHSDM (IASRP).
uo 0.80
3 0.70
0.60

0.50 1 1 1

For the AB80+AY 1 17 system, the D, values of AB80 in the single component system are
smaller than the Ds,fvalues of AB80 in the AB8O+AY 1 17 binary system while the D, values
of AY 117 produce an opposite trend, the D, values of AY 1 17 in the single component
system are greater than the Ds,2values of AY 1I7 in the AB80+AY 1I7 binary system. The
D, values of AB80 (DS,Jare greater than the D,,values of AY 117(Ds,2)in the AB8O+AY 1 17
system. This may be because the AB80 dye has a higher molecular mobility in the
AB80+AY 1 17 mixture. This implies a higher adsorption affinity towards the surface of the
activated carbon than the AY 117 dye.

The effect of different initial dye concentrations is shown in Figures 2 for a wide range of
initial dye concentrations for AB80 in ABSO+AY1 17 binary system. A good correlation
between experimental and theoretical results is also obtained in the binary systems. The
fitted parameters, kl, k2,D , , and Ds,2are given in Table 1. According to the table, the
surface diffusion coefficient is dependent on the initial dye concentration of each system, as
in the single component system, parameters Ds,, and Ds,2are strongly dependent on the
initial dye concentration. The Higashi expression [3] has been used to correlate the
relationship between the multicomponent surface diffisivities, 0,.I and Ds,2, and the
fractional coverage, 0" and BB, of the acid dyes in the binary adsorption system.
4.OE-I0 Figure 3: The Plots of Surface Diffusivity against
Fractional Coverage Expression for AY117 in
3.OE-I0
Single Component System and Binary System
(AMWAY1 17).
- 2.OdlO
a*
I.OE-10

112
The expressionsare shown in equations 18 References.
and 19 and the plots are shown
- in Figure 3: [l] McKay, G., Application of Surface
Diffusion Model to the Adsorption of
Dyes on Bagasse Pith, Adsorption 4
(1998) pp. 361-372.
[2] Radke, C.J., Prausnitz, J.M.,
Thermodynamics of multi-solute
where D,and 0 2 are the self-diffiivities adsorption from dilute liquid solutions.
for the component 1 and 2 in the mixture. It AlChE Journal 18(4) (1972) pp.
was found that the surface diffusivitiesare 76 1-768.
lineirly proportion to the M o n a 1
coverage expressions and are similar in [3] Higashi, K. Ito, H. and Oishi, J. Surface
trend with the Diffusion Phenomena in Gaseous
findings in the single component HDSM Diffusion: I, Suflace Di@iuion of Pure
(see Table 3). GUS5 (1963) pp. 846-853.
Table 1: Extend Masa T r a d e r and Solid-phuc Diffusion
Coeffidentafor the Adaorplioa of ABM)cAY117 011 Aetiv.
ted Carboa using tbe Multieomment HSDM (IASRP).
SOP

Tabk 2: External Mass Traoafer md Sdiipiuse Diffusion


Cocfiieienta for tbe Adaorplh of A W and AY117 on Act
ivated Carbon in Siagk Component System using tbe HSD

i.
1.7
1.7 I f

100
I00
I 7.36xlC"
934x10'" I :::: I
1.69xIO-~~
2.26~10-'~
9.11~10'
9.35xIQ"
9.34~10'"
6.002
7.749
3.363
3.070
6.471 '
1 I
1 . 2 3 ~ 1 0 ' ' ~ 1.498
1 . 7 1 ~ 1 0 ~ ' ~ 3.512
2 . 5 8 ~ 1 0 ' ~ ~ 6.002
3 . 5 5 ~ 1 0 " ~ 7.749
1.86~10' 3.363
1 . 8 9 ~ 1 0 " ~ 3.070
1 . 7 9 ~ 1 0 " ~ 6.471
1.7 100 9.34x10'" 3.512 1.71~10''~ 3.512

Table J:Slope and y-Intercept Valucs of the Correlated Li


nee in the Plots of tbe Surface Diffwivity against Fnetiona
I Coverage Erprrssioa for Smgk Component and Binary S

some y-intcmpt
ABW (sin&) 2.646~10' 2.686~10'
AdBhK80
, ABW+AY117
BW+AY117 1.449xI1T'~
1.449~IO.'~ 1.524~10"~
1.524~10"~
MYdlow AYll7(riagk) 2.702~10' 2.8%x109
117 ABW+AYI17 2.525~10~ 2.566xlO9

113
SORPTION THERMODYNAMICS OF NITROUS OXIDE / LSX ZEOLITE
SYSTEMS

MARTIN BULOW, WNGMIN SHEN AND SUDHAKAR R.JALE


BOC PGS Technology, 100 Mountain Ave., Murray Hill, NJ07974, USA
E-mail: martin.bulow@us.gtc.boccorn

Sorption thermodynamic functions of nitrous oxide, N20, are described for zeolites NaLSX and
CaLSX in shepes of clay-bound beads. They were determined by the Sorption Isosteric Method (SM)
over complete ranges of sorption-phase concentration and compand with those fix carbon dioxide,
CQ,RpoltedearliaforthesameNaLSXsorbent.

1 Introduction

Nitrous oxide is a greenhouse gas. Its earth-atmospheric concentration (currently =2ppm)


increases steadily by c (0.2-0.3)% p.u. Ihis is caused mainly by antmpogenic activities and
emissions limn chemical pmsses, eg., synthesis of adipic acid for Nylon-66, automotive power
genedon and wastewater tmtment N@ is very stable in &,its lifetime amounts to c. 150 years.
Remod of N o knn air streams in tiOnt of airseparation units (ASV) [I] is critical as that of
HzOand (2%. Excess of N@ in ASU may lead to plugging of tubes and heat exchangers and to
-
3
4
c onof products, specificallyof noble gases. As the Concentration of N a in air increases
fhther, the current regime of air-pqmilication units (PPV)in h t of ASU may become
inadequate, since N o cannotbe removed easily by existing technologies. Bmkduwgh curves for
NzO in a CaA-zeolite bed at 8,OOO-ppmN20and variouS C q concenlmtions in air [2] indicate that
C Q displaces NzO as the mass-tmnsfm h t progresses along the adsoher bed. This finding
questions the usage of CaA as single s o h t for N a removal. Although N@ and C Q have
identical molecular weights, soxption interaction of NzO with zeoliteNaX modifications, especially
NaLSX [3], deviates h m that of CQ due to d-i in specific Properties of these gases such
as dipole and quadnrpole moments. It is importslnt to undemtand so@m thmodynamia of NzO
on PPU sorben& and to assess difkences in Sorption properties between N a and COz. Such
knowledge will allow for development of novel materials and @on processes to purify ASU-
fdgas.

2 Sorption Isosteric Method and Sorbents

The Sorption Isosteric Method (SIM)was used to determine thermodynamic data of N2O
sorption by zeolites NaLSX and CaLSX. This technique and its utilization for an
investigation of sorption-thermodynamic properties of COz on identical NaLSX beads
were described in detail in refs. [4-81.
Sorbentsused representbeads of NaLSX (98.99% Na) and CaLSX (97.29 % 0, both of
(8 x 12) mesh. CaLSX is prepared by Na vs. Ca2+exchange of NaLSX beads with c. 12 %
attapulgite binder. Ihe appamt XRD crystallinity ofNaLSX beads amounts to c. (78+5)% as
amparedto arefenmce sample without binder. Dry weights ofNaLSX and CaLSX beads utilized
by SIM amount to 92727 g (bulk densii of NaLSX: 0.67 and 8.7033 g, respectively. Prior
to the z e o b rlre a c t i care~v
~ in ti^ S I M - ~ ~ ~ sample
~ W Wc e at ~ <1 0 - ~ t o ~
and 673K, during 4 8 h m ibr NaLSX (beads wm nearly dry at loading), and over 7days for
CaLSX (fully hydrated beads).

114
3 Results and Discussion

3. I Sorption Isosteres of N 2 0 on NaLSX and CaLSX Zeolites


&@on isosteresmeasured for the N2GNaLSXand N2ocaLsX systems over complete ranges
of sopion-phase concentmtiom,n,are presentedin Figs. 1 and 2, respectively. Slopes
- i ' i ' i ' i ' i ' l ' i ' l ' i ' i ' l
n. mom:
rn 0.0592 7
0 0.1167 :
A 0.2200 :
v 0.4599 -
+ 0.7225 .
1' x 1.0304
f 1.3083 -
1.7314 :
-
0
L
0 2.1365
0
.cI
D
A 2 . 7 ~
D
v 3.0571
Q
0 3 . m
W
0
4.2140
4.7400
-:
4 5.2701 -
I 5.5905 -
0 6.20M
D 6.9337.
I 7.7254
E 8.6133 Y
.
i '
'.
i ' i ' i ' i ' l ' i ' l ' i ' l ' i ' i ~
3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4

1000/(T, K)
Figure 1. Sorption isosteres for NzO on NaLSX zeolite (clay-bound beads).
of these isosteres change with n if the sorption heat depends on concentration. Except for
a few isosteres for concentrations close to saturation capacity, the isosteres determined at
actual experimental conditions appear to be liiear, indicating that no sorption-phase transition
takes place. Therefore, they are characteristicof well-defined sorption-equilibriumprocesses
of N20on NaLSX and CaLSX zeolites. They are attributed to the correspon-ding sorption-
phase concentrations set via dosing procedures. Since the boiling point, 184.67 K, and the
triple point, 182.33 K, of an N20-bulkphase at 1 at, are close to each other, the related phase
transitions can be. observed h m isosteres as N20 concentmtion approaches and exceeds the
sorbent-samation capacity. This feature is obvious h m r.ks. isosteres in Figs. 1 and 2. From
particular slopes of two segments of the isostere for highest n, 8.6133 molikg, shown in Fig.
1, the latent heats of vaporization, 16.55 kl/mol at 184.67 K, and fusion, 6.54 kJ/mol at
182.33 K, are obtained. These quantities agree well with handbook data [9].
Values of isosteric sorption enthalpy, -AH,standard sorption entropy, AS3 and
standard Gibbs fiee sorption energy, AGO, are calculated as dependences on n, c j , [4-81.
To calculate AGO, the boiling-point temperature of N20 is chosen as reference state. This
choice provides a check for thermodynamic consistency of experimental data since AGO

115
should become zero as n exceeds the saturation capacity at boiling-point temperature.
Values of AGO at any other temperature can be calculated from those of -AHand AS
using the Gibbs equation. Thereafter, equilibrium mass distributions, i.e., pressures that
correspond to the quantities n, -AH,AYand dG3 can be determined.

1000/(T,K)
Figure 2. Sorption isosteresfor N20on CaLSX zeolite (clay-bound beads).

3.2 Sorption Thermo&namics of N20 and C02 on N a N and C a mZeolites


Sorption thermodynamic data for N 2 0 on NaLSX and CaLSX obtained h m isosteres as
functions of n, are compared with those for C02 on NaLSX, cf.,Figs. 3-5. For NaLSX, the
-AH vs. n dependence for C02 proceeds always above that for N20. As concentration
exceeds the saturation capacities for the two gases, the enthalpy for C02 sublimation, viz.,
- 25.23 kJ/mol, and the enthalpy for N2O evaporation, - 16.55 kJ/mol, are derived from
the corresponding isosteres. This finding characterizes two bulk-phase transitions that take
place and confirms correctness of experimental data. At low n, the -AH values for N 2 0
decrease sharply with increasing n, unlike for C02, for which the plot -AHvs. n shows a
plateau at n o 1.5 m o m . Sorption thermodynamics suggests that interaction between N20
and NaLSX at low coverage is much weaker than that for C02, although the molecular
weights of the two gases are identical, their quadrupole moments differ slightly only
(3.65 x e.s.u. and 4.3 x e.s.u. for N2O and C02, respectively [lo]), and N2O has
a dipole moment, 0.167 D.
Standard sorption entropy values, ds, that refer to a gas-phase pressure,
p = 760 torr, as standard state, show wave-like concentration dependences for the two
gases on NaLSX, cf.,Fig. 4. Compared to their standard gas phases, a remarkable entropy
loss takes place over the entire ranges of n. Particularly, C02 experiences less freedom
than N20, due to its stronger interaction with sorption sites and, probably, denser
sorption-phase packing.

116
c

;
=i
Y
i
?

lo
o
1
0
~ .
1
l .
2
l
3
. l
4
. l
5
. l
8
.
7
l .
8
, .
I
9
~ . {
n, mollkg
Figure 3. Isosteric sorption heats of N20and C02on CaLSX and NaLSX.

40- . I . I . 1 . 1 . 1 . 1 . 1 . 1 .

60 - -a- N,O I CaLSX


-0- CO, I NaLSX

Y
e--o

-120 -

-1401
0
n I
1
v

2
I - I
3
= I
4
. I
5
- I
6
* I
7
I
8
*
0
n, moMkg
Figure 4. Standard sorptionentropies of N20 and C02on CaLSX and NaLSX
Dependences dG" vs. n of N20 and C02 for NaLSX, as r e f e d to both the boiling
ternperatwe of each gas and the gas pressure, p o = 760 ton; as standard state are shown
in Fig. 5, together with related data for the N20-caLSX system. Values Mi0 for all
systems change k n negative ones to zefo as n increases, approaches and exceeds the
micropore-saturation capacities.This proves thennodynamic consistency and correctness of
the SIM data,

117
0

-5

-2 5

-30

- 3 5 : . 1 . I . , . , . I . I . , . I . I
0 1 2 3 4 5 6 7 8 9
n , mol/kg
Figure 5. Standard Gibbs free sorption energies of NzO and Cot on CaLSX and NaLSX,
referred to the boiling temperatures as standard.

A large difference in AGO values exists between the two gases for n c c. 2 molkg on
NaLSX, which disappears as n increases. This proves that COz is sorbed preferentially over
NzO. Asp and, thus, n increase, this selectivity disappears. If the quantities AGO for NzOand
COZare referred to one and the same temperature, e.g., 298 K, sorption isotherms can be
obtained directly and compared to each other at this temperature.
Sorption thermodynamics for NzO-CaLSX is characterized by quantities -AH,AS" and
AGO as functions of n, which are shown in Figs. 3-5 for comparison with those for the
NaLSX systems. Fig. 3 proves that CaLSX compared to NaLSX, provides sorption sites
for NzO, which exhibit stronger sorption interaction at n < 1.5 mol/kg. This indicates
that NzO-Caz' interactions are considerably stronger than the NzO-Na+ ones, in LSX
supercages. As n increases fiom ca. 1 to 5 molkg, the isosteric sorption heats of NzO
on NaLSX and CaLSX zeolites become close to each other. Most probably, non-
specific van der Wads-type interactions between N 2 0 and the lattices prevail, and they
are nearly the same for NaLSX and CaLSX. Fig. 4 shows AS" vs. n dependences for NzO
on CaLSX and NaLSX. Interestingly, these curves differ significantly for the two cationic
modifications of LSX. For NaLSX, NzOprobes at least two different (cationic) sorption sites.
For CaLSX, the curve shape expresses a more smooth distribution of energetically
heterogeneous sorption sites, which reminds qualitatively that for the Nz-CaA system [6,7]. The
shape of the B" vs. n dependence for COz on NaLSX is reminiscent of that for the NzO-
NaLSX system and witnesses preferred sorption of Cot. Standard Gibbs free sorption
energies as basic sorption-equilibrium quantities of NzO and COZ, are referred to
184.67 K and 194.65 K, respectively. The "more negative" values for NzO on CaLSX if
compared to those for NaLSX at n<2molkg, indicate that CaLSX attracts N 2 0
significantly stronger. At n > 2 molkg, values -AGO of N20 for the two zeolites become
close to each other because non-specific sorption interaction between NzO and the LSX
lattice is similar in the two materials. This situation is reflected by the isotherm courses for
NzO on the two sorbents, cJ,Fig. 6, which were derived from SIM data.
Figs. 3-6 elucidate differences in sorption thermodynamics of systems, NzCX~LSX,NzO-

118
NaLSXand CQ-NaLSX. Generally, the -AH data obey a sequence, C&NaLSX >N2G
CaLSX >N2o-NaLsx, over almost the entire ranges of n. However, the cormponding -0
sequence varies withe which can be seen fiom Fig. 5. At n o 1.5 molkg,the absolute values
8 . ......., .......

-0- CO, I NaLSX


-0- N,O I NaLSX

m
0 4- -
5
i 3- -
-
-
0 ......... .................... ,,.,..I ...............
n
10" 10' loo 10'
10' lo3 10' 105
p, torr
Figure 6. Sorption isotherms of N 2 0 and C02 on CaLSX and NaLSX at 298.15 K.
of dG" obey the order N20aLSX > CGNaLSX > N20-NaLSX. This indicates that sorption of
N20 by CaLSX is stronger than that of C Q by NaLSX, in that @on. In the mge(2 o n
4.5) m o m the ditkmces for the three systems dkqmr, but they reoccur in a partly r e v 4
order at n o 4.5 m o b : N20aLSX > N20-NaLSX > COTNaLSX.
From these dependences A G O vs. n, sorption isotherms can be obtained for any tempemure
and pressure values that are meaningM physically, as exemplified in Fig. 6. Isotherms therein do
not only illustrate differences in sorption properties for the thm systems, NzO-C~LSX,N20-
NaLSX and CQ-NUX, but make it obvious that - at low equilibrium pressure - the sorption
capacity of N20 on CaLSX is comparable with that of C02on NaLSX. These capacity values are
much higher than those for N20on NaLSX. Moreover, at very low equilibrium pressure, < 0.5 torr,
the sorption capacity of CaLSX for N20 is higher than that of NaLSX for C02. This particular
feature makes the CaLSX mIite a preferred N20-selective sorbent for the removal of trace
amounts of N@ h m air in h n t of ASU [lo]. For none of the systems investigated at cryogenic
tem- a Henry region could be identified h m the sorption isostereS, which is due to strong
sorption interactions and m&odical reasonsas well.
A quantitative comparison of sorption-thermodynamicproperties h r N20 as described here
with l i t e m data for a CaA-tYpe zeolite [2] meets difficulties since no such thermodynamic data
could be derived reliably h m the latter source. Prefmntial interaction of e m - h e w o r k Ca"
cations with C@ compared to that with N20as identified in [2] agrees qualitatively with the results
of this investigation.

4 Conclusions

Equilibrium sorption thermodynamic properties of N20 sorbed by NaLSX and CaLSX are

119
investigatedby SIM. These are the first comprehensive data sets ever measured for N20. They
are compared with those for the C&NaLSX system. Soption of N20 on NaLSX is much
weaker than that of CQ on the same zeolite. At low pressures, < lotorr, NaLSX has a very
high sorption capcity for CQ, but has a low capacity for N@. CaLSX has a higher sorption
capacity for N20at pressure less than 0.5 torr, compared with that of NaLSX for CQ, and it has
nearly the same sorption capacity at high pressures. Highly exchanged CaLSX zeolites are
recommended for use as N20-SeleCtive sorbents. By combining NaLSX and CaLSX in certain
adsorber-bed layer anangements for TSA PPU proceses, CQ and N20 could be removed
subsequently and completely h m air streams.

5 Acknowledgements

The authors are grateful to Drs. F.R. Fitch and A.F. Ojo, Murray Hill, for helpful
discussions. They thank the BOC Group for the permission to publish this work.

6 References

[l] Golden T.C., Taylor, F.W., Johnson, L.M., Malik, N.H., and Raiswell, C.J.,
Purification of Air, US Patent No. 6,106,593; August, 22,2000.
[2] Mayinger, F., and Egg&-Steger, R, J Energy. Heat & Mars Traqfiw, 15 (1993) 165.
[3] Ojo, A.F., Fitch, F.R., and Biilow, M., Removal of Carbon Dioxidefiom Gas Streams,
US Patent No. 5,531,808; July 02, 1996.
[4] Biilow, M., and Shen, D., in Fundamentals of Adsorption-6 (Ed.: F. Meunier),
Elsevier, Paris, 1998, p. 87.
[5] Shen, D., and Biilow, M., Micropor. Mesopor, Materials, 22 (1998) 237.
[6] Shen, D., Engelhard, M., Siperstein, F.,Myers, A.L., and BUlow, M., in Adsorption
Science h Technology (Ed.: D.D. Do), Proc. 2d Pacific Bash C o d Adsorption Science
Technol., World Scientific, Singapore, 2000, p. 106.
[7] Shen, D., BUlow, M., Jale, S.R., Fitch, F.R., and Ojo, A.F., Micropor. Mesopor.
Materials, 48 (2001) 2 1 1.
[8] Biilow, M., Shen, D., and Jale, S.R., Appl. Surf Sci., 196 (2002) 157.
[9] The Mathewn Company, Matheson Gar Data Book 4* Edn., East Rutherford, 1%6, p. 83
(CW; P. 387 ( N 2 0 ) .
Eyring, H., Henderson, D., and Jost, W., Physical Chemistry: An Advanced Treatise,
Vol. 1V:Molecular Properties, Academic Press New York, 1970, p. 381.
Shen, D., Huggahalli, M., Biilow, M., Jale, S.R., and Kumar R., TSA Process for the
Removal of Dinitrogen Oxide, Hydrocarbons and Other Trace Impuritiesfrom Air,
US Patent No. 6,391,092; May 21,2002.

120
ACTIVATED CARBON MEMBRANE WITH CARBON WHISKER

SANG-DAE BAE AND AKIYOSHI SAKODA*


Institute of Industrial Science, University of Tokyo
4-6-1 Komaba,Meguro-ku, Tokyo,Japan
E-mail: sakoda@iis.u-tokyo.ac.jp
A novel composite membrane (an activated carbon membrane with carbon whiskers) applicable to water
treatment was fabricated by a combination of conventional carbonization and thermal deposition methods. The
carbonization was performed at 1050-1100 c using poly-vinylydenchloride (PVdc) and ply-vinylchloride
-
(PVC) microsphens, and the thermal deposition was accomplished after femc sulfate (Fe2(S04)j nH20) was
applied as a coat on and within a ceramic support. The membrane consists of three parts, each with a different
function for water treatments: a carbon whisker layer, a carbon layer, and an activated carbon layer. Those
structural characteristics were confirmed with a scanning electron microscope (SEM) and by measuring
electrical resistance and water and gas permeability. Water treatment experiments, using phenol and
polyethylmethacrylate(PMMA) as model pollutants, indicated that this membrane was able to remove dissolved
organics with low molecular weights and suspended solids simultaneously. Also, the carbon whisker layer
prevented the fouling of fine particles on the membrane effectively with performing the frequent back washing.
These results indicate that the novel composite membranes can clarify the water within a relatively short time,
and are usefbl for small scale distributed drinking watedwaste water treatments.

1. INTRODUCTION

Watedwastewater treatment usually consists of several component operations such as


biological treatment, coagulation, sand filtration, membrane filtration, and activated
carbon adsorption. Raw water usually contains solid particles ranging in size from
nanometer to millimeter, as well as dissolved pollutants. This means that any single unit
operation cannot remove all pollutants in one stage. On the other hand, simultaneous
removal of particulate and dissolved pollutants can be achieved if we can combine the
physical filtration and adsorption processes. From this standpoint, we first developed an
activated carbon membrane (Sakoda er ul., 1996, 1998). However, during the membrane
filtration operation, we cannot prevent the deposition and accumulation of fine particles
on and within the membrane, resulting in a cake layer. As a promising method to solve
this fouling problem, we then developed a novel membrane called the carbon whisker
membrane (Li er ul., 200 1a, 200 1b, 2002a), which basically contains vapor-grown carbon
fibers (whiskers) on the external surface of a carbon-coated ceramic support. The
filtration experiments using polyethylmethacrylate (PMMA) as model particles indicated
that the carbon whiskers can prevent direct contact between the membrane body and
PMMA particles, and that fouling would occur with greater dificulty than on as
membranes without carbon whiskers (Li,submitted). Unfortunately, however, the water
flux of this carbon whisker membrane was not large enough for practical use, for the
following reason. To solve this fatal problem, we propose a new methodology for
preparing a new activated carbon membrane with carbon whiskers, as shown in Figure I.
We also characterizethe membrane and the water treatments that can employ it.
For further details of this preceding, see this paper. (Bae ef ul., 2002)

2.EXPERIMENTAL

Ceramic tubing (inner diameter = 9 mm, outer diameter = 13 mm, pore size = 2.3
pm) was kindly provided by Kubota Co., Ltd., Japan, and used as the membrane support.
The polymer latex containingpoly-vinylydenchloride(PVdC) and poly-vinylchloride

121
Carboaiaed and aggregated polymer
microspheres with micropores for
PerRntatioa adsorption
(particle siae=O. 1p mPore size=0.7nrn)
Figure 1 Structure of activated carbon membrane with carbon whiskers

(PVC) microspheres of 0.10-0.15 pm in diameter (Asahi Kasei Co. Ltd., Japan) was used
as a pre-coat reagent and precursor of the activated carbon (Sakoda et ul., 1996).
Preliminary experiments revealed that thermal decomposition occurs around 300 c, and
that a temperature above 700 C is needed for the complete carbonization of the polymer
latex employed. The ferric sulfate (Fe2(S04), nH20) was used as the catalytic
precursor in order to make the carbon whiskers (Li et ul. 2001a, b).
A ceramic tube, with one end plugged by a polymer paste, was slowly dipped into
polymer latex and taken out while being rotated at 600 rpm. By this dipping method, an
aggregate of the polymer micro spheres was formed at the surface and subsurface parts of
the ceramic tube (Sakoda et ul., 1996). The sample was dried at room temperature for
30 minutes. The dried sample was slowly dipped into 0.5 M ferric sulfate solution and
dried at room temperature for 8 hours (Li et ul., 2001 a, b).
A 15-cm-long ceramic sample was put in a quartz tube with an i.d. diameter of 25cm,
placed in an electric oven, then dried at 300 c for 10 c Smin- and kept at 300 c for
30 minutes. The temperature was then raised again to 1000-1100 C at a rate of
10 C-min-. The temperature was reach at 1000-1100 c, we started to input the
nitrogedmethane = (80/20) volume % mixture gas at 500 ml*min-,25C in flow rate, in
order to form carbon whiskers on the activated carbon layer. The hydrocarbon
deposition time and temperature were set at 20 minutes and 1050 or 1100 C,respectively.
Mem. Component of Dipping time CVD time CVD
polymer in polymer (min.) temperature
la tex(w t%1 latexkec.) (c 1
PVdC : PVC
A 45 : 55 20 20 1100
B 70 : 30 20 20 1050
C 7 0 : 30 10 20 1050
D - - 20 1100

Table 1 Preparation conditions for the MembranesA, B and C


Note: Membrane D is the carbon whiskers membrane reported in the previous work
(Li, et ul., 200 1a,b)

122
In this study, we made three novel membranes, labeled Membrane A, B, and C, with
different formation conditions. Table 1 lists the preparation conditions of the
membranes. Membrane D is a carbon whisker reported in Li et al. (2001), and is used
for comparison.
The morphology of each membrane surface was observed using a scanning electron
microscope (SEM).
Pure water flux (PWF) measurements were carried out using a 15-cm-long
membrane in a cross-flow apparatus consisting of a feed reservoir, a membrane module, a
back washing reservoir, and a regeneration reservoir, at 25 'c . Seven gases (hydrogen,
helium, methane, nitrogen, oxygen, argon, and carbon dioxide) were used for the gas
permeability measurements. Gas was introduced to the module, and permeation flux
was measured under a 100 mmHg pressure drop by a bubble flow meter at room
temperature.
Filtration experiments were carried out using this prepared membrane in a cross-flow
apparatus, which was the same as the PWF measurement apparatus. PMMA particles
(1000 mg-l-', 0.8 pm in diameter) were used as model particle pollutants. When the
PMMA solution was introduced by the feed magnet gear pump, the flux of the same
object declined to a certain level. In this time, membranes needed regeneration, we used
the technique of back washing. To carry out the back washing, we changed the flow
direction by niddle valve and used pure water at a pressure of 2 kg*cm'*for 1 minute.
The membrane developed in this work and a commercial activated carbon F-400
(Calgon), which is widely used in various water treatments, were used for adsorption
experiments. The adsorption isotherms of phenol were measured according to the
following batch adsorption method. The sealed vial was placed in a constant-
temperature water bath kept at 25 "c. The samples were well stirred, using a magnetic
stirrer overnight (Sakoda er ul., 1991).

3. RESULTS AND DISCUSSION

Figure 2 shows the SEM pictures of the prepared Membranes A, B, and C.


Membrane D is also shown. It is obvious that Membranes A, B, and C have finer carbon
whiskers, with diameters of approximately 0.6, 0.4, and 0.2 pm, respectively, while the
whiskers of membrane D are larger, 2 pm (Li er ul., 2001). As the SEM pictures show,
after carbonization of polymer latex and CVD of methane, the novel membranes had thin
and dense carbon whiskers compared with the conventional ones.
The results of gas permeability, hydrogen, helium, methane, nitrogen, oxygen, argon,
and carbon dioxide permeation under pressure of 100 mmHg with Membranes A to C.
For all samples, the permeation rates were found to be proportional to the molecular
weight powered to -0.5. This means that for all membrane samples, the flow in the
pores become a Knudsen-type flow and each membrane sample had a linear plot,
indicating the absence of pinholes.
In the pure water flux (PWF) of Membranes A to D that these membranes have
different ranges of pore size. The PWF of Membranes, A, B, C, and D were 0.7,2.4,2.8,
and 0.03 [m day-' atm-'1, respectively. These data indicate that the PWF of the
membranes developed in this work were much better than those of carbon whisker
membranes repeated in the previous works. The resultant PWF over 0.5 [m day-'
atm-'1 at the pressure of 50 [kPa] is large enough for practical applications, since
commercial membranes have PWF typically of 2 [m day-' atm-'1.

123
2um

Figure 2 SEM pictures of membranes prepared in this work, A, B and C


Note: Membrane D is the carbon whiskers membrane reported in the previous work (Li, e
/a/.,2001a,b)

In the virgin state, Membrane C had the highest flux of all membranes. Meanwhile,
Membrane A had a better cleaning recovery rate than did membranes B and C.
Moreover, in membrane A, the first cleaning process had a greater 99% higher cleaning
recovery rate than that of the second cleaning process. This suggests that the flux of a
membrane, as well as the diameter and density of carbon whiskers, can be controlled by
manufacturing conditions.
Thus we concluded that the carbon whisker layer with high surface density and
narrow space between the carbon
whiskers has the practical effect of 1000
preventing fouling.
Figure 3 shows the adsorption
isotherms of the activated carbon
membrane and activated carbon, F-400.
The adsorption capacity of the activated 100
carbon membrane for thus particular 8
adsorbate was similar to that of F-400,
suggesting the possibility that the o Membrane A
adsorption capacity is increased by a Membrane B
A F-400
somehow developing larger micropores 10
within the carbonized microspheres. In 1 10 100
spite of their different compounds of
C|mg/I|
polymer latex, Membranes A and B had
almost the same amounts of adsorption. Figure 3 Adsorption isotherms of phenol
to Membranes A and B, and F-400 at 251:

124
4. CONCLUSION

This study described the manufacturing of a novel composite membrane (an


activated carbon membrane with carbon whiskers) and characterized several variations of
the membranes. The resultant PWF is large enough for practical applications. The
introduction of the polymer latex in the manufacturing membrane process, not only makes
use of femc sulfate solutions preventing of the subsurface, but also acts as a precursor to
the activated carbon layer. The use of the polymer latex allowed us to improve the fine
and high density of carbon whiskers to the extent that they can prevent blocking and/or
fouling on or in the membrane, while allowing for easy cleaning with the back-washing
regeneration technique. Also, this novel membrane increased the adsorptivity of
dissolved organics through the carbonization of the polymer latex. Considering these
results, we are convinced that the novel composite membrane can be used widely in the
practical and commercial fields of water treatments.

-
Acknowledgements Ceramic tubing and polymer latexes used in this work were kindly
provided by Kubota Co. Ltd. and Asahi Kasei Co. Ltd., respectively.

REFERENCES

Bae, S.D.,Masaki, S., Sakoda, A. and Suzuki, M. (2002), Activated Carbon Membran
e with Carbon Wiskers and its Application to Water Treatments, Water Research
(submitted)
Endo M., Takeuchi K., Kobori K., Takahashi K., Kroto H.W. and Sarkar A. (1995),
Pyrolytic carbon nanotubes from vapor-grown carbon fibers, Carbon, 33,873-88 I.
Hagg May-Britt (1998), MEMBRANES IN CHEMICAL PROCESSMCj A Review of
Applications and Novel Developments, SEPARATION AND PURIFICATION
METHODS, 27(1), 51-168.
Iijima S. (1991), Helical microtubules of graphitic carbon, Nature, 354, pp 56-58.
Li, Y.Y.,Bae, S.D., Sakoda, A., Suzuki,- M. (2001a), Formation of vapor grow carbon
fibers with sulfuric catalyst precursors and nitrogen as carrier gas, Carbon, 39,9 1- 100.
Li, Y.Y.,Bae, S.D., Nomura, T., Sakoda, A. and Suzuki, M. (2001b), Preparation of
Custom-Tailored Carbon Whisker Membrane by Chemical Vapor Deposition, in: K.
Kaneko (Ed.), Fundamentals of Adsorption, 7,279-286.
Li, Y.Y., Bae, S.D, Nomura, T., Sakoda, A., Suzuki, M (2002a), Carbon Whisker
Membrane, Adsorption (Submitted) .
Li, Y.Y.,Nomura, T., Sakoda, A., Suzuki, M. (2002b), Fabrication of carbon coated
ceramic membranes by pyrolysis of methane using a modified chemical vapor
deposition apparatus; Journal of Membrane Science, 197,23-35.
Pierson Hugh O., HANDBOOK OF CHEMICAL VAPOR DEPOSITION (1992);
Principles, Technology and Application, Noyes Publications, New Jersey.
Sakoda, A., Suzuki M., Hirai R.,and Kawazoe K. (1991), Trihalomethane adsorption on
the activated carbon fibers, Water Research, 25(2), 2 19-225.
Sakoda, A., Nomura T. and Suzuki M. (1996), Application of Activated Carbon
Membrane to Water Treatments: Decolorization of Coke Furnace Wastewater,
Ahorption, 3(1), 93-98.
V. Kuberkar, P. Czekaj, R.Davis (1998), Flux Enhancement for Membrane Filtration of
Bacterial Suspensions Using High-Frequency Back washing, Biothchnology and
Bioengineering, 60,77-87.

125
MESOPOROUS SILICA WITH LOCAL MFI STRUCTURE

SAJO P. NAIK('), ANTHONY S.T. CHIANG(')*,ROBERT W. THOMPSON'", F. C. HUANG"'


AND HSIEN-MING KAO(~)

("Department of Chemical and Material Engineering, National Central University, Chung-


Li, Taiwan ROC 320.
(2'Department of Chemical Engineering, WorcesterPolytechnic Institute,Worcester,MA
USA 01609.
'3)GraduateInstitute of Environment Engineering,National Central University,Chung-Li,
Taiwan ROC 320.
")Department of Chemistty, National Central University, Chung-Li, Taiwan ROC 320

A new type of mesoporous silica has been prepared via a dual-template, three-step hydrothermal-
flocculation-steaming (HFS) synthesis procedure. This material showed 780 m'/g of BET surface
area and 0.6 mug of primary mesopores narrowly distributed around 4.2 nm. More importantly
however, is that it showed short-range MFI zeolite crystallinity as demonstrated by FTIR and XRD
analysis, and hydrophobicity as demonstrated by water and n-hexane adsorption.

1 Introduction

When the MCM-41s type ordered mesoporous materials were first introduced,'' .*. it was
anticipated that they could function as catalysts for bulky molecules. Subsequent studies
however revealed that their mild acidity and inferior hydrothermal stability could not
filfill such promises. The stability and acidity of these materials may be improved if
zeolite-like order, such as the MFI zeolite structure of ZSMJ and silicalite, could be
introduced into the mesopore walls by using simultaneously the large surfactant as
template for mesopores and the small organic templates as that for zeolitic micropores.
This dual template strategy has been tested'. ' 4. and shown to exhibit enhanced acidity
and steam stability compared to the corresponding MCM-41. However, a composite
material of MCM/MFI instead of a homogeneous one was obtained. Other approaches

surfactants?
achieved.
-'
include the post-treat with structure directing cations under dry condition for creation of
zeolitic micro ores: *a or the hydrothermal treatment of zeolite seeds after adding
In most case, better hydrothermal stability and stronger acidity could be

However, in the post-treatment approach, the surface area decreases quickly with the
formation of zeolite phase. The mesoporous material eventually becomes an aggregate of
zeolite nanocrystals. The zeolite seed hydrothermal ap roach, on the other hand, produces
materials with only indirect hints of zeolitic nature! Neither the FTIR nor the XRD
spectrum showed strong evidence for the formation of zeolite structure.
From our earlier experiences: we have learned that the zeolite seed hydrothermal
approach is rather difficult to reproduce. Consequently, a three-step-synthesis procedure
involving the preparation of zeolite nanoprecursors (NPs) by a short hydrothermal step,
the flocculation of these NPs using a surfactant, and the steaming of the NPs/surfactant
composite to produce the final material was developed. We have recently"' demonstrated
that aggregates of less than 30 nm silicalite nanocrystals can be prepared from this
procedure. We further discovered that the nature of the as-collected NPs was very much
dependent on the stirring time of NPsKTAMeBr flocculants. Under identical steaming
condition, the 3 h-stirred NPs were converted into nanocrystals of silicalite-1, whereas,

126
*. 11% the 36 h-stirred NPs were converted
into a mesoporous structure having
very high surface area.

2 Experimentals
* 47%
A sol with molar composition 0.25
-70 -80 -90 -190 -110 -1u) -130 -140
TPAOH/ITEOS/80H20 was prepared
as per the procedures described
-'*
prevoius~y,'~* which was then
m-1 ~ ~ ~ ~ R r p c e b o l n o l ( ~ ) - t ~ m m ) ) . R rsealed in a polypropylene (PP) bottle
BpldnL.
and heated in an air oven at 80 "C for
18 h. After cooling the sol to room
temperature, the NPs were harvested
by slowly adding an ethanolic
surfactant solution under stirring.
The white flocculants resulted were
filtered after stirred at room
.-
b temperature for 36 h, dried at room
{ temperature for 24 h, then at 70 "C
I

9 for further 3 h. The dried powder was


:.
t
- pressed into pellets and steamed at
I50 "C for 24 h in a Teflon lined 190
ml autoclave, where a ceramic
honeycomb is placed with 0.19 g of
water at the bottom. The as collected
and steamed samples were then
5 Ill 15 m 25 34 calcined in flowing air with a heating
2 eldegree rate of 2 "C/min and a holding time
Figare2XRDprttaosaf(.)uEdledcdNPh(b).Itadirrct~~.(ioa, Of at 550 OC. The samples were
(c) rtamed and calcined then examined by XRD (Shimadzu
LAB-X-700), FTIR (Jasco-4lo), Nitrogen adsorption (Micromeritics ASAP 20 10) and
29
Si MAS-NMR (Bruker DSX-300). The room temperature adsorption of water and n-
hexane were measured with a G-Cahn-200 microbalance.

3 Results and discussions

Showed in Figure 1 are the NMR spectra of the surfactant-collected precursor before and
after 150C steaming. For both samples, there was approximately an equal distribution of
Q3 and Q4 silicon environment. The steaming produced only a small increase of Q2
species, suggesting the hydration of surface silica species. The NMR spectra are very
similar to that obtained by Kremer et al.'" recently.
As showed in Figures 2, the as-collected NPs and the directly calcined one showed
only a XRD band in the 20-25" 20 range. This band is consistent with the XRD pattern of
nanometer-sized MFI crystallites recently simulated by Schlenker and Peterson,14' where
a broad band was predicted to result from the broadening of the 12 distinct peaks in this
range. The steamed sample, on the other hand, showed distinct MFI XRD peaks.

127
For nanometer size MFI zeolite,
XRD may not be the best technique to
determine its crystallinity. It has been
reported that IR crystallinity appeared
before XRD peaks were observed"
. For example, the peak at 550-570
cm-' is known to appear due the
presence of five-member rings present
in MFI zeolitic structures."'
The as collected NPs showed a
weaker hump at 550-590 cm-' and a
stronger absorption at 620 cm" which
could come from the Q2 silica
observed in the solid state NMR The
absorption in 550-590 cm-' range
disappeared if the NPs were directly
calcined, but became stronger if the
calcination was done after steaming.
This highlights the importance of the
steaming step in stabilizing the
zeolitic structure.
The nitrogen adsorption isotherms provide M e r evidence of the type of material
obtained by the steaming of NPs. As observed in Figure 4, both the direct calcined and
the steamed and calcined samples were mesoporous materials. For comparison, the
isotherm of a typical MCM-41 sample with 3.0 nm mesopores was also given. The
steamed sample showed a rather sharp uptake at -0.4P, corresponding to the existence of
uniform mesopores about 4.2 nm in size (from KJS modified BJH calculations" ). The
total pore volume, as well as the mesopore size of this sample, is obviously larger than
that of MCM-41. This suggests that the mesopores of the steamed sample are different
fiom that of MCM-41, and most likely come from the inter-particle void between nano-
crystals. For direct calcined NPs, the amount adsorbed below 0.3P0is higher than steamed
one, but a large hysteresis due to the texture mesopores existed above 0.45P0. However,
according to the BJH calculation, the primary mesopores are about 2.5 nm in size, which
is smaller than that of MCM-41.
The micropore volume, mesopore volume and the surface area of the mesopores can
be calculated from the isotherms. The results of these calculations are also summarized in
Table 1. It is known that well prepared MCM-41 material is free of any micropore. The
mesoporous silica obtained from NPs, on the other hand, do show small but noticeable
micropore volume. The larger mesopore size and higher mesopore volume compared to
MCM-41 confirm that obtained mesoporous silica has it unique structure.
The water and n-hexane adsorption isotherms of the zeolitic mesoporous materials
obtained are compared to that of a 450 nm colloidal silicalite-1 in Figure 5. The water
adsorption isotherms are distinctively type Ill, whereas the n-hexane isotherms are type 1.
The lowest water isotherm was for the colloidal silicalite-I, where the first point
measured for the n-hexane isotherm was already at 80 mg/g. The amount of n-heme
adsorbed reached 250 mg/g at high pressure, which roughly corresponds to the filling of
silicalite-1 micropores.
The steamed NPs is more hydrophobic then the directly calcined one, but slightly
inferior to the colloidal silicalite-1. However, due to its huge mesopore volume, the
loading capacity for hexane is much higher than that of colloidal silicalite-1. It could

128
accommodate more than 350 mg/g of n-hexane at -0.3P0, which is high even compared to
activated carbon.

Table 1 Texture properties from nitrogen adsorption analysis

PR. PRO
F w 4 Adsorption isotherms for nitrgcn at 77JK on Figure 5 Room tempenbre adsorption*otbcrms of water
surf8et.ntcoouce~NPa a R a (a) direct cd&ubiw, (filkdsymbols)and beune ( o w symbols).
(b)steaming .nd calcination,(c) MCM41 mfrrmce materid.

4 Conclusions

We have successfully synthesized a zeolitic material having high surface area, a narrow
mesopore size distribution, and XRD crystallinity of the MFI structure. The size of the
mesopore was -1 nm larger than that of MCM41 materials synthesized from the same
surfactant. This excluded the suspicion of a heterogeneous MCM4l/MFI mixture. We
have further found that the stirring time after the addition of surfactant and the humidity
under which the steaming is carried out have profound effect on the final product. By
indirect proof, it was proposed that the material obtained might be similar to the
delaminated zeolite obtained in other zeolite systems. In any case, the three-step
procedure seemed to give better control over the process of forming this material.

5 References

K.D. Schmitt,
1. J.S. Beck, J.C. Vartuli, W.H. Roth, M.E. Leonowicz, C.T. Kresge,
C.T-W Chu, D.H. Olson, E.W.Sheppard, S.B. McCullen, J.B. Higgins, J.L.
Schlenker, J. Am. Chem. SOC.114 (1992) 10834.

129
2. C. Kresge, M. Leonwicz, W. Roth,J. Vartuli, US patent 5098684 (1992)
3. L.M. Huang, W.P. Guo, P. Deng, Z.Y. Xue, Q.Z. Li, J Phys. Chem. B 104 (2000)
2817.
4. A. Karlsson, M. Stocker, R. Schmidt, Micropor. Mesopor. Muter. 27 (1 999) 18 1.
5. D.T. On,S. Kaliaguine, Angew Chem h t Ed. 40 (2001) 3248.
6. K.R. Kloetstra, H. van Bekkum, J.C. Jansen, Chem Commun, (1997) 2281.
7. S.S. Chen, Y.W. Chen, A.S.T. Chiang, Proc.of Pacific Basin Conf. On Ads. Sci. and
Tech., Ed. D.D. Do, World Scientific Inc., Australia (2000) 130.
8. Y. Liu and T. J. Pinnavaia, Chem. Muter. 14 (2002) 3 .
9. J. Liu, X. Zhang, Y. Han, F.S.Xiao, Chem. Muter. 14 (2002) 2536.
10. S. P. Naik, J. C. Chen and A. S . T. Chiang, Micropor. Mesopor. Muter. 54 (2002)
293.
1 1. S.P. Naik, A.S.T. Chiang, R.W. Thompson, F.C. Huang, submitted to Chem. Muter.
(2002).
12. S.S. Chen, A.S.T. Chiang, presented in 7* International C o d On Fundamental of
Adsorption, Nagasaki, Japan, May 20-25,2001.
13. S.P.B Kremer, C.E.A. Kischhock, M. Tialen, F. Collignon, P.J. Grobet, P.A. Jacobs,
J.A. Martens, A&. Func. Muter. 12 (2002) 286.
14. J.L Schlenker, B.K. Peterson, J. Appl. Cryst. 29 (1996) 178.
15. T. Armaroli, M. Trombetta, A. Alejandre, J. Solis and G Busca, Topics in Cutuhsis
15 (2001) 63.
16. Y.S. Lin, N. Yamamoto, Y. Choi, T. Yamaguchi, T. Okubo, S.1. Nakao, Micropor.
Mesopor. Muter. 38 (2000) 207.
17. R. Ravishankar, C.E.A. Kirschhock, B.J. Schoeman, P. Vannopen, P.J. Grobet, S.
Storck, W.F. Maier, LA. Martens, F.C.De Schryver, P.A. Jacobs, J. Phys. Chem. B
102 (1998) 2633.
18. M. Kruk, M. Jaroniec, ASayari, Lungmuir, 13 (1997) 6267.

130
INFINITE DILUTION SELECTIVITY MEASUREMENTS BY
GAS CHROMATOGRAPHY

SASIDHAR GUMMA AND ORHAN TALU


Department of Chemical Engineering, I960 E. 24IhStreet,Stilwell Hall 455
Cleveland State Vniversiw, Cleveland,OH 441 15, USA
E-mail: s.gumma@csuohio.edu
o.talu@csuohio.edu

Adsorption is a widely used separation process in the chemical process industry. Selectivity is a key
property that determines the ease and efficiency of any separation. Although, selectivity is a function
of both composition and pressure, most of the design calculations are carried out using a Langmuirian
approach with constant selectivity, using only pure component data The conventional techniques to
characterizethe selectivitybehavior by measuring mixture equilibrium data require a large investment
of time and elaborate experimental setup. Moreover selectivity is very sensitive to experimentalerrors.
In this article we present an experimental method to measure the selectivity in binary gas
adsorption under infinite dilution conditions over a range of pressures and thereby characterize its
behavior with respect to both pressure and composition. The method is f k t , efficient and robust. We
use the methane-ethane binary gas mixture on silicalite as a demonstration. The experimental results
obtained using this method are compared with the predictions from two models to check their validity.

1 Introduction

Adsorption equilibrium has an extra degree of freedom compared to conventional


vapor-liquid equilibrium. This extra degree of freedom increases the difficulty in
experimental measurements. It is difficult to find enough experimental data on binary
equilibria in the adsorption literature. On the other hand, pure component isotherm
measurement is so common that commercial push-button systems are available in the market
for over a decade [11.
The variation of total and partial amount of ethane adsorbed as a function of
composition at 270 kPa for a methane-ethane mixture is shown in Fig. 1. Two models,
Langmuir for mixtures using Innes and Rowley correlation [2] and IAST [3] are used to
predict the data from pure component isotherms. Both the models do reasonably well in
predicting both the partial and total amount adsorbed. The pure component methane and
ethane isotherms fix the end points of total amount adsorbed. The partial amount of ethane is
also restricted between its pure component value and zero. The two models simply predict
the curves in between the end points fixed by these thermodynamicrestrictions.
Selectivity is a key variable that affects the adsorption process and is essential for
design. The variation of selectivity of ethane with pressure and composition is shown in a
3D graph in Fig. 2. Pure component data yields only the line AB at zero pressure, which is
the ratio of Henrys constants. Using only this information it is not possible to accurately
estimate the variation in selectivity. The two models differ substantially with respect to
selectivity predictions. In the Langmuirian approach the selectivity is constant and is given
by the ratio of Henrys constants (along a horizontal plane through AB). Selectivityby IAST
approaches the same limit at zero pressure but rapidly decreases with pressure.

131
2 Conventional Measurement Techniques

There are several experimental techniques to characterize binary adsorption. The


conventional techniques like volumetric or volumetric-gravimetric are full information
techniques that yield both the total amount adsorbed and surface composition and hence the
selectivity. In these techniques the error in selectivity goes to infinity

6
0 01 01 03 04 RS 01 07 08 08 ?

Yol

Figure 1. Partial and total amounts adsorbed for CH&&, mixture at 308 K and 270 kPa

Figure 2. Selectivity of CzH6 from CH4-C2&, mixture at 305 K

as the composition approaches either of the ends. In the combined volumetric- gravimetric
technique an additional error source is introduced if the molecular weights of the gases are

132
close to one another. Both these techniques are cumbersome, time-consuming and require
very specialized equipment and expertise.

3 Gas Chromatographic Techniques

There are several reports in the literature that measure binary adsorption equilibria using gas
chromatography [4,5,6]. In GC techniques the adsorbent is equilibrated with a continuous
flow of carrier gas (gas 1). Then a pulse of gas 2 is injected at the column inlet. A peak of the
gas 2 is eluted at the exit of the column after some time. Net retention time (or volume) is
calculated from the first moment of the peak after correcting for void volume (by measuring
the retention time of a non-adsorbing species). If the canier gas is inert (i.e. helium) the net
retention time is related to the pure component Henry's constant. Typical binary
measurements reported so far use a mixture of the two gases as carrier and introduce a small
perturbation in composition. The net retention volume is related to the thermodynamic
properties by [4]
V D dn dn

In the above equation PIand p2 are the partial pressures of the gases, nl and n2 are partial
amounts adsorbed and m is mass of the adsorbent. Thus, the experiments yield the
derivative information of the partial isotherms at various conditions. Some mathematical
technique is needed to integrate the experimentaldata and obtain the partial isotherm [5,6].

Although GC techniques are very fast compared to the conventionalmethods, any error
in the experimental data is magnified and propagated in the calculationsdue to integration.
There are also problems associatedwith polynomial fitting for the retention volume in terms
of composition in order to obtain binary isotherms [5,7].

4 Experimental

We used a typical GC setup and a mass spectroscopic detector. The only modification
involves controllingcolumn pressure between 20 - 1000 kPa. The following table lists the
main features of the experimental system used. All previous attempts to use the GC
technique for binary measurementswere conducted at constant (atmospheric)pressure. The
pure component isotherms are obtained form a conventionalvolumetric technique.
Table 1. Experimental system and conditions used
~

Column 5.35 mm ID, 65 mm long


Adsorbent Silicalite(HISIV 3000, UOP),1.6mm pellets
Adsorbent activation temperature 573 K
Experimental temperature 305 K
Exaerimental aressure 20-233kPa

133
5 Results

When the adsorbent is equilibrated with gas 1 and an infinitely small pulse of gas 2 is
injected into the carrier gas stream, Eq.( 1) can be written as [7]

The selectivity of 2 ( 2,1) at these conditions is given by Eq.(3). The quantity n l ( P } in the
above equation is the pure componentamount adsorbed for gas 1 at total column pressure P.
Experimentalmeasurementsare required for V ,(obtained fiom the infinite dilution system)
and data for pure component isotherm (obtained independently using a volumetric
technique) to calculate selectivity (LHS of Eq.3). A similar equation can be written for the
infinite dilution of gas 1.
Figure 3 shows the variation of selectivity with pressure at infinite dilution conditions
for both the gases. The broken horizontal line is the result from the Langmuir model which
is same at infinite dilution of either gas. The selectivitiespredicted by IAST are different for
the two gases as shown by the solid lines. The points are experimental data.
In contrast to the previous attempts, our GC technique does not involve polynomial fits
and integrations. It is especially developed to measure selectivity at infinite dilution of one
of the components as a h c t i o n of pressure.
This data is also shown in figure 2 as filled circles. As already stated, the solid line AB
is obtained fiom the pure Component data only. After measuring the pure component data,
instead of going through cumbersome measurements of binary selectivity at finite
composition using existing techniques, we characterized it by measuring selectivity at
infinite dilution conditions in a fast and efficient way using the proposed method. Given the
line AB and the infinite dilution curves as the limits, most models will be able to estimate the
3D mesh with reasonable accuracy.
We needed less than two days to measure the data shown. This is much shorter
compared to any conventional technique, which requires almost a day per point for finite
binary compositions. Moreover, the infinite dilution extremes can not be measured using the
conventional techniques.

6 Conclusions

We proposed a novel method which can be used to for fast measurements of infmite dilution
selectivity in binary gas adsorption. This experimental system can handle a wide range of
pressure variation. Using the data obtained one can perform a quick characterization for
selectivity. From the design point of view this method can be effectively employed to check
validity of the model being used.
The major limitation of this technique is that it may not be suitable to use it for highly
selective systems like methane-butane mixture on silicalite. For these systems the retention
times may be long and a significant dispersion in the peak might occur.

134
15

14
0 50 I DO 150 100 I50 D

P (kPa)

Figure 3. Infinite dilution selectivity variation with pressure at 305 K

7 Acknowledgements

We thank the Cleveland State Universitys Doctoral Dissertation Research Expense Award
Program (DDREAP) for partial support of this work.

References

1. Talu,O., Needs,status,techniques and problems with binary gas adsorption


experiments, Advances in Colloid and Interface Science 76-77 (1998) pp.227-269
2. Innes, W.B. and Rowley,H.H. Journal of Physical Chemistry, 51( 1947), pp. 1 154
3. Myers,A.L. and Prausnitz,J.M., Thermodynamics of Mixed Gas Adsorption, AIChE
J , 11 (1965), pp. 121-127
4. Van Der Vlist, E. and Van Der Meijden, J., Determinationof adsorption isotherms of
the components of binary gas mixtures by gas chromatography, Journal of
Chromatography,79( 1973), pp. 1 13 -
5. Hyun,S.H. and Danner, R.P., Determination of gas adsorption equilibria by the
concentration pulse technique, AIChE Symp. Ser., 34( 1982),pp. 1861-1877
6. Tezel, F.H., Tezel, H.O. and Ruthven,D.M., Determination of Pure and Binary
Isotherms for Nitrogen and Krypton, Journal of Colloid and Interface Science, 149
(1992), pp. 197-207
7. Harlick, P.J.E. and Tezel, F.H., A Novel Solution Method for Interpreting Binary
Adsorption Isotherms from Concentration Pulse ChromatographyData, Adsorption, 6
(2000), pp.293-309
8. Kohl,S., Measurement of the effect of pressure on infinite dilution adsorbed phase
activity coe%cients by gas chromatographic techniques, Diplomarbeit, University of
Seigen (1996)

135
ADSORPTION PROPERTIES OF COLLOID-IMPRINTEDCARBONS

MIETEK JARONIEC AND ZUOJIANG LI


Department of Chcmishy.Kent State University, Kent OH 44242, USA
E-mail: Jaroniec@kent.edu

This work is focused on the synthesis and adsorption properties of colloid-imprinted carbons (CIC)
exhibiting bimodal pore size distribution. It is shown that the p @ e s such as the specific surface
area, pore volume. pore openings and the d e p of graphitization can be tailored by selecting the
proper size of the colloidal silica and by adjusting the synthesis conditions. In addition, their surface
properties can be adjusted by proper post-synthesis modification. Nitrogen adsorption isotherms
wen measured for a colloid-irnprintcdcarbon with bimodal distribution of spherical pores. Analysis
of adsorption data shows that the imprinting of the mesophase pitch particles with colloidal silica
affords carbons with uniform spherical mesopores and negligible amount of micropores.

1 Introduction
Since the discovery of the supramolecular self-assembly of silica and surfactant species [11,
an impressive progress has been achieved in the area of novel mesoporous materials, their
synthesis, characterization and applications [2-4]. The supramolecular self-assembly of
various inorganic and organic species into ordered mesostructures became a powerful
method for the synthesis of mesoporousmolecular sieves of tailored -work composition,
pore structure, pore size and desired surface functionality for advanced applications in such
areas as separations, adsorption, catalysis, environmental cleanup and nanotechnology.
While the self-assemblyprocess can be successfully used to obtain various organic-inorganic
nanocomposites such as ordered silica-surfactant nanostructures, it is inherently unsuitable
for the synthesis of ordered mesoporous carbons and carbon composites. Although porous
carbons have a very long history and have become adsorbents of great industrial importance,
some attempts to obtain ordered carbon mesostructures have appeared just recently [5-14).
One of such attempts explored by Ryoo and co-workers employs the ordered mesoporous
silicas (OMS),e.g., MCM-48 and SBA-15, as templates [9,11-141. The templating method
involves the filling of the ordered porous structure of a silica material with a carbon
precursor followed by carbonization and silica dissolution, and affords ordered carbons of
pore sizes below 6-7 nm. Another templating method, which is more appropriate for
preparation of macroporous carbons, employs siliceous colloidal crystals [5-81. The colloidal
crystal templating, which affords carbons with ordered macropores of size greater than 50
nm, involves the formation of colloidal crystals, infiltration of the crystal's interstitial space
with a fluid-type carbon precursor-and its solidificationfollowed by removal of the template.
Although the both templating methods are attractive and worthy of further
exploration, they do not allow for the synthesis of carbons with uniform and tailored pore
sizes in the range from 7 to 50 nm. Recently, a new approach to the synthesis of
mesoporous carbons, which employs the idea of colloidal imprinting, was proposed [ 151.
This approach seems to be very promising for the synthesis of mesoporous carbons with
uniform spherical pores, high pore volume and relatively high surface area. The current
work refers to the recently reported synthesis and adsorption properties of colloid-
imprinted carbons (CIC) [15,16] and focuses on the preparation and adsorption properties
of the CIC samples with bimodal pore size distribution. It is shown that the carbons with
spherical mesopores of two different sizes can be synthesized by using a suitable binary
mixture of colloidal silicas.

136
2 Experimental

2.1 Synthesis of colloid-imprintedcarbom


The colloids used in the synthesis of the CIC sample with bimodal pore size distribution
were Ludox AS40 from Aldrich and Bindzil30/360 from Eka Chemicals. The synthetic
mesophase pitch AR (softening point 237 "C) from Mitsubishi was used as carbon
precursor. The carbon sample with bimodal pore size distribution was prepared by
adapting the previously reported procedure (see Figure 1) [HI, which involves the
grinding of the pitch precursor, pretreating it in silicone oil for 30 min at 250 "C followed
by washing with toluene and acetone. Next, the mesophase pitch particles were dispersed
in ethanol and the resulting mixture was added gradually under vigorous stirring to a
binary mixture (wt. ratio -1:l) containing the Ludox AS40 and Bindzil30/360 colloidal
silicas. The stirring was continued for five hours to allow a complete evaporation of the
solvent. The pore imprinting procedure was canied out at 260 "C in nitrogen atmosphere
for 30 minutes, followed by heating at a rate of 2 "C/min to 600 "C and 5 "Chin to 900
"C, and an isothermal treatment at 900 "C for 120 minutes in nitrogen. The carbonized
sample was treated with 3M sodium hydroxide solution (95 "C) until silica particles were
totally dissolved.

Heating and Silica


imprinting dissolution
Mixture of silica
colloids and pitch _____,
particles Carbonization

Figure 1. Scheme illustrating the synthesis of mesopomus carbons by colloid imprinting process.

2.2 Nitrogen aalsorption measurements


Nitrogen adsorption/desorption isotherm for the CIC sample was measured by using a
volumetric gas adsorption analyzer from Micrornetrics (Norcross, GA) at -196 "C. Prior
adsorption measurements the carbon sample was degassed under vacuum at 200 "C for
two hours. The BET specific surface area was calculated using the standard Brunauer-
Emmett-Teller (BET)method in the relative pressure range of 0.04-0.15. The total pore
volume was evaluated from the amount adsorbed at a relative pressure of 0.99. The t-plot
method was used to estimate the amount of micropores. The pore size distribution (PSD)
was calculated from the adsorption branch by the Barrett-Joyner-Halenda (BJH) method,
in which we used the statistical film thickness (t-curve) evaluated from nitrogen
adsorption isotherm on the BP 280 carbon black by fitting it to the multilayer range of the
t-curve calibrated for the MCM-41 samples [ 171.

3 Results and Discussion

One of the advantages of the colloidal imprinting method is the ease of controlling the pore
size as well as pore size distributionsby choosing proper colloids [HI.
As shown in [I61 the

137
high temperature treatment of the colloid-imprinted carbons (so-called graphitization) did
not reduce the accessibilityof porous structure but even caused its improvement evidenced
by increasing the pore size openings. The current work shows that the use of a binary
mixture of colloidal silicas afford mesoporous carbons with bimodal pore size distribution.
The CIC sample synthesized by using a 1:l mixture of Ludox AS40 and Bindzil30/360
colloidal silicas had the BET specific surface area of 425 m2/gand the total pore volume of
1.08 cm3/g. The nitrogen adsorptioddesorption isotherm and the corresponding pore size
distribution for this CIC sample are shown in Figures 2 and 3, respectively.

0.0 0.2 0.4 0.6 0.8 1.0


Relative Pressure

Figure 2. Nitrogen adsorption isotherm for the CIC sample synthesized by using 1 I and 24 nm silica colloids.

As can be seen from Figure 2 the adsorption branch of this isotherm exhibits two
distinct steps that reflect the capillary condensation inside smaller or larger mesopores at
relative pressures about 0.79 and 0.9, respectively. The condensation in the relative pressure
range of 0.95-0.995 reflects condensation in secondary mesopores or small macropores,
which resulted fiom the imprintingof agglomeratesof colloidal particles. To our knowledge,
this kind of isotherm has not been reported for porous carbon materials. The pore size
distribution for this mesoporous carbon shown in Figure 3 exhibits two distinct peaks
located about 1 1 nm and 24 nm, which correspond to the particle size of Bindzil30/360 and
Ludox AS40 colloidal silicas, respectively.
It should be noted that the t-plot analysis shows a very small amount of micropores in
the sample studied, which is due to the use of the mesophase pitch as the carbon precursor.
It is known that many precursors, e.g., sucrose and polyfurfiuyl alcohol, infiltrate well
siliceous templates but after Carbonization give meso- or macroporous carbons with
complementarymicroporosity. This is not the case for the mesophase pitch, which is used to
synthesize carbon fibers that are nonporous materials.

138
0.06

0.04

0.02

0.00
10 20 30 40 50 60
Pore Sue (nm)

Figure 3. Pore size distribution for the CIC sample synthesized by using I I and 24 nm silica colloids.

Our previous papers [15,16] and the current work show that the imprinting of
mesophase pitch particles with colloidal silica is an efficient technique to prepare
mesoporous carbons with uniform spherical pores as well as carbons with bimodal pore size
distributions. These carbons exhibit negligible amount of micropores, which can be further
eliminated during graphitization process. If micropores are need, they can be created by
controlled oxidation analogous to that used in the preparation of activated carbon fibers.
The possibility of tailoring the size of uniform spherical mesopores is of great importance
for catalysis, adsorption and other advanced applications such as the manufacture of
high-quality electrochemical double-layer capacitors, fuel cells and lithium batteries.

4 Acknowledgements

The authors gratefully acknowledge the donors of the Petroleum Research Fund
administered by the American Chemical Society for the partial support of this research.

References

1. Kresge C. T., Leonowicz M. E., Roth W.J., Vartuli J. C. and Beck J.S., Ordered
mesoporous molecular-sieves synthesized by a liquid-crystal template mechanism.
Nufure359 (1992) pp. 710-712.
2. Pang J. B., Qiu K. Y. and Wei Y ., Recent progress in research on mesoporous materials.
J. Znorg. Mufer. 17 (2002) pp. 407-414.
3. Ying J. Y., Mehnert C. P. and Wong M. S., Synthesis and applications of
supramolecular-templated mesoporous materials. Angew. Chem. Inf. Edit. 38 (1 999)
pp. 56-77.
4. Ciesla U. and Schuth F., Ordered mesoporous materials. Microporous Mesoporous
-
Mafer. 27 (1 999) pp. 13 1 149.
5. Zakhidov A. A., Baughman R.H., Iqbal Z., Cui C., Khayrullin I., Dantas, S. O., Marti J.
and Ralchenko V. G., Carbon structures with three-dimensional periodicity at optical

139
wavelengths. Science 282 (1 998) pp. 897-90 1.
6. Yu J. S., Yoon S. B. and Chai G.S., Ordered uniform porous carbon by carbonization
of sugars. Carbon 39 (2001) pp. 1442-1446.
7. Lei Z., Bang Y., Wang H., Ke Y.,Li J., Li F. and Xing J., Fabrication of well-ordered
macroporous active carbon with a microporous framework. J. Muter. Chem. 11 (2001)
pp. 1975-1977.
8. Kang S., Yu J. S., Kruk M. and Jaroniec M., Synthesis of an ordered macroporous
carbon with 62 nm spherical pores that exhibit unique gas adsorption properties. Chem.
Commun. (2002) pp. 1670-1671.
9. Ryoo R., Joo S. H. and S. Jun, Synthesisof highly ordered carbon molecular sieves via
template-mediated structural transformation. J. Pjys. Chem. B 103 (1999) pp.
7743-7746.
10. Lee J., Yoon S., Hyeon T., Oh S. M. and Kim K.B., Synthesis of a new mesoporous
carbon and its applicationto electrochemical double-layer capacitors. Chem. Commun.
(1 999) pp. 2 177-2178.
1 1. Jun S., Joo S. H., Ryoo R., Kruk M., Jaroniec M., Liu Z., Ohsuna T. and Terasaki O.,
Synthesis of new nanoporous carbon with hexagonally ordered mesostructure. J. Am.
Chem. SOC.122 (2000) pp. 10712-10713.
12. Ryoo R., Joo S. H., Kruk M. and Jaroniec M., Ordered mesoporous carbons, A h .
Muter. 13 (200 1) pp. 677-68 1.
13. Joo S. H., Choi S. J., Oh I., Kwak J., Liu Z., Terasaki 0. and Ryoo R., Ordered
nanoporous arrays of carbon supporting high dispersions of platinum nanoparticles.
Nuture 412 (2001) pp. 169-172.
14. Lee J. S., Joo S. H. and Ryoo R., Synthesis of mesoporous silicas of controlled pore
wall thickness and their replication to ordered nanoporous carbons with various pore
diameters. J. Am. Chem. SOC.124 (2001) pp. 1 156-1157.
15. Li Z. and Jaroniec M., Colloidal imprinting: a novel approach to the synthesis of
mesoporous carbons. J. Am. Chem. SOC.123 (2001) pp. 9208-9209.
16. Li Z., Jaroniec M., Lee Y. J. and Radovic L. R., High surface area graphitized carbon
with uniform mesopores synthesised by colloidal imprinting method. Chem. Commun.
(2002) pp. 1346-1347.
17. Choma J., Jaroniec M. and Kloske M., Improved pore-size analysis of carbonaceous
adsorbents. Adsorption Sci. & Technol. 20 (2002) pp. 307-3 15.

140
ON THE ROLE OF WATER IN THE PROCESS OF METHYL MERCAPTAN
ADSORPTION ON ACTIVATED CARBONS

S.BASHKOVA, A. BAGREEV AND T.J. BANDOSZ


Department of Chemistry City College of New York and Graduate School of CUNY
138kStreet and Convent Ave, New York IVY 10031, USA
E-mail:tbandosz@ccny.cuny.edu

Adsorption of methyl mercaptan in moist conditions was performed on numerous samples of


activated carbons of various origins. Methyl mercaptan adsorption was tested by a dynamic method.
The amount of products of surface reaction was evaluated using thermal analysis. The results
revealed that the main product of oxidation, dimethyl disulfide, is adsorbed in pores smaller than 50
A. There is apparent competition for adsorption sites between water (moist conditions) and
dimethyl disulfide. The competition is won by the latter molecule due to its strong adsorption in the
carbon pore system. Although dimethyl disulfide has to compete with water for the adsorption sites
it can not be formed in a significant quantity without water. Water facilitates dissociation of methyl
mercaptan and thus ensures the efficient removal process.

Introduction

The amount of methyl mercaptan (MM) adsorbed (and converted to dimethyl disulfides
(DMDS)) depends on the surface pH [1,2], and the presence of various impregnants, such
as potassium iodide, potassium iodite, potassium carbonate or ammonia [3, 41. It has also
been pointed out in the literature that different functional groups on the carbon surface
or/and metal ions such as iron can catalyze oxidation of mercaptans to disulfides [3-61. As
we have found recently, there is an indication of a competition for high-energy adsorption
sites between dimethyl disulfide and water molecules when adsorption occurs in the
presence of moisture [l, 21. This happens as a result of big differences between water and
DMDS in the strength of adsorption forces and their incompatibility (DMDS has very low
solubility in water) [7].
An objective of this paper it to describe the results of our further investigation of the
competition for adsorption sites between water and dimethyl disulfide molecules during
methyl mercaptan adsorption on activated carbons. Moreover, we attempt to indicate the
apparent borderlines between the conditions of adsorption processes leading to different
adsorptiodoxidation paths. Those working conditions have a significant effect on the
feasibility of methyl mercaptan removal.

Methods

Materials: Adsorption of methyl mercaptan was performed on numerous samples of


activated carbons of various origins. Among the carbons studied were BAX-I500 (wood
based -Westvaco), S208 (coconut shell - Waterlink Bamabey and Sutcliffe), Centaur
- -
(catalytic carbon Calgon), BPL (bituminous coal Calgon), PCB (coconut shell-Calgon),
Maxsorb (mesophase pitch-Kansai), and polymeric based synthetic carbon, SCN [8]. To
broaden the spectrum of pH and surface properties, carbons were oxidized, acidified with
HCl, impregnated with NaOH, and modified with urea [9]. We do not refer to any specific
sample because our objective here is to show the general behavior of the system under

141
study.
CHSH aborption: The carbons were studied as adsorbents of methyl mercaptan in the
dynamic tests described elsewhere [l]. Briefly, 6 cm3 of adsorbent samples were ground
(1-2 mm particle size) and packed into a glass column (bed depth 80 mm, diameter 9 nun)
and prehumidified with moist air (relative humidity 80 % at 25 "C) for one hour. Dry or
moist air (relative humidity 80 % at 25 "C) containing 0.3 % (3,000 ppm) CH3SH was
then passed through the adsorbent bed at 0.5 L/min. The breakthrough of CH3SH was
monitored using a Micromax monitoring system (Lumidor) with an electrochemical sensor,
previously calibrated for MM. The test was stopped at the breakthrough concentration of
50 ppm. The adsorption capacities of each adsorbent in terms of mg of CH3SH per g of
carbon were calculated by integration of the area above the breakthrough curves, and from
the CH3SH concentration in the inlet gas, flow rate, breakthrough time, and mass of
adsorbent. For each sample the breakthrough test was repeated at least twice. The
determined capacities agreed within 4 %.
p H of carbon surface: 0.4 g of carbon powder was placed in 20 mL of water and
equilibrated during night. Then the pH of suspension was measured.
Sorption of nitrogen: Nitrogen isotherms were measured using a ASAP 2010
(Micromeritics) at -196 OC. Before the experiment the samples were heated at 120 "C
and then outgassed overnight at this temperature under a vacuum of l o 5 Torr to constant
pressure. The isotherms were used to calculate the surface area and pore (DFT [lo]) and
characteristic energy of adsorption, (Eo) (Dubinin-Radushkevichmethod [ 1 I]).
Thermal analysis:Thermai analysis was carried out using TA Instruments Thermal
Analyzer. The instrument settings were: heating rate 10 deg/min in nitrogen atmosphere
with 100 mL/min flow rate.

Discussion

0.8
E
0
B
g 0.6
.-
P
3RI 0.4
P
g 0.2
v)

0
2 4 6 8 1 0 1 2
pH of carbon surface

Figure 1. Dependence of the amount of MM adsorbed on the surface pH.

The dependence of the amount of methyl mercaptan adsorbed normalized to the pore
volume of carbons studied on the surface pH is presented in Figure 1. In the case of

142
microporous carbons the volume of micropores is used whereas for carbons with a large
contribution of mesopores the volume in pores smaller than 50 A was chosen. It was done
assuming that only those pores are active in the adsorption process. This choice was
justified knowing that: 1) at a small concentration of adsorbate in a gas mixture it is likely
that only micropores and small mesopores are active in the adsorption process [12]; 2)
molecular simulation studies of water adsorption suggested that at 80% humidity (our
conditions) 50 A and smaller pores are filled with water molecules [13]; 3) if the
normalization is done using only the volume of micropores the density of adsorbed
species is higher than liquid density of DMDS. Moreover, a certain pore volume has to
be a limiting factor for the adsorption capacity since the products of adsorptiodoxidation
should be stored there [ 1,2, 141. Although the data is scattered due to the complexity of
the system the maximum boundary of the adsorption capacity can be noticed. Moreover,
when experiments were done at dry air the capacity was almost constant. Similar
phenomenon was described for adsorption of hydrogen sulfide on activated carbons 1141.
In the case of MM adsorption on activated carbon the presence of water facilitates its
oxidation to DMDS owing to adsorption and oxidation of thiolate ions [9]. It was found
that for the process to be efficient the pH of the carbon surface should be greater than 7.5.
1.o
@T
E
-.-s 0.8

c)

F 0.6
$
n
m
.S2 0.4
3Q.
1: 0.2
0

ZI 0 . O l - L L F b Y
0.0 0.2 0.4 0.6 0.8 1.0
H20specific adsorption [g/cmT

Figure 2. Dependence of the DMDS adsorption on the amount of water adsorbed on


carbons.
As mentioned elsewhere a typical DTG plot for exhausted carbon after MM
adsorption consists of two peaks [l, 2, 91. One, low temperature, at about 80 "C,
represents desorption of water, and second, with maximum at about 200 "C,represents
desorption of dimethyl disulfide. Following the assumption that either H2O or DMDS are
adsorbed only in pores smaller than 50 A, the data was normalized based on that volume.
Figure 2 shows the relationship between the normalized amount of DMDS and water. The
correlation coefficient and slope are equal to 0.89 and - 0.99, respectively. The slope
represents the density of DMDS (1.06 g/cm3). The small discrepancy is likely related to
the fact that not all pores are filled by oxidation products owing to the existence of some
physical hindrances (blocked pore entrances). The thin line represents theoretical limit of
adsorption assuming real density of DMDS and H 2 0 . The fact that almost all points are
located below this line validates our hypothesis about the "active" pore volume. It is
important to mention here that all points represent equilibrium data. If equilibrium

143
conditions, for instance for adsorption of water, are not fblfilled the amount of DMDS is
usually small and the point "moves" from the established dependence line.
How the adsorption occurs can be evaluated based on Dubin-Radushkevich approach
[ 1 11. The amount of vapor, A, adsorbed at pressure P and temperature T can be calculated
from DR equation [ 1 I]:

where V, E, ..dL and PS are the volume of micropores, characteristic energy of


adsorption of standard gas (benzene), the affinity coefficient for vapor adsorbed, density
of liquid adsorptive and its saturation pressure at temperature T, respectively. The
required for our calculation parameters of adsorbents were determined from nitrogen
adsorption isotherms at - 193 "C [9]. The affinity coefficients were estimated from
parachor values [151. They are equal to 0.580 for MM and 1.018 for DMDS.
Interesting data for BAX and S208 is collected in Figure 3 where the predicted
amounts of MM, DMDS and water adsorbed are plotted. Prediction of water adsorption
was done following the approach of Lodewyckx and Vansant [ 161 based on application of
Dubinin-Astakhov (DA) equation using affinity coefficient equal to 0.063 for activated
carbons with low content of the oxygen surface groups. For both carbons at relative
pressure P/P,s equal to 0.8 (our experimental conditions) the amount of water, which can
be adsorbed is close to the amounts of adsorbed MM and DMDS. What is even more
interesting, the amount of DMDS adsorbed at very low relative pressure f/P,S = 0.05
corresponding to inlet concentration of MM (3000 ppm) is practically the same as the
amount of water adsorbed at P/Ps = 0.8. This can be considered as another favorable
factor to promote competition for adsorption sites, which results in replacement of water
with much stronger adsorbed DMDS molecules. Our calculations based on statistical
molecular approach showed that, however the initial heat of DMDS adsorption on
activated carbons is 3 times higher that that for water (60 kl/mol compared to 20 kl/mol),
the Henry constant is about seven orders of magnitude greater for DMDS than that for
water molecule (4.61 x 10" compared to 3.83 x I d ) [9].
I I i I I I I 1
t I
I
1 I
1 i
1

0.8

0.e
s
B 0.4
B
B
02

0
o 02 0.4 0.e 0.8 I

Figure 3. The predicted isotherms (with relative pressure units) for MM, DMS and HzO
on two carbons, BAX (A) and S208 (B).

144
The results presented in this paper indicate the dual role of water in the process of
methyl mercaptan removal on activated carbons. It is demonstrated that DMDS,formed
during surface oxidation of MM, has to compete with water for adsorption sites. However,
the competition exists, DMDS is always the "winner" owing to its strong adsorption on
carbons. On the other hand, the formation of significant amount of DMDS would not be
possible without the presence of water in the system. Water facilitates dissociation of
methyl mercaptan Ieading to its oxidation by oxygen, mainly from air. Since dissociation
is an important step, the pH of the carbon surface has to be greater than the estimated
threshold (about 7.5).

Acknowledgements

Acknowledgement is made to the donors of the Petroleum Research Fund, administered


by the ACS, for support of this research (Grant # ACS-PRF#35449-AC5). The authors are
gratefbl to Ms. Thiri 00for her experimental help.

References

1. Bashkova, S., Bagreev, A. and Bandosz, T.J., Adsorption of methyl mercaptan on


activated carbons. Environ. Sci. Technol. 36 (2002) 2777.
2. Bashkova, S., Bagreev, A. and Bandosz, T.J.,Effect of surface characteristics on
adsorption of methyl mercaptan on activated carbons. Ind Eng. Chem. Res. 41 (2002)
4346.
3. Turk, A., Sakalis, E., Lessuck, J., Karamitsos, H. and Rago, O., Ammonia injection
enhances capacity of activated carbon for hydrogen sulfide and methyl mercaptan.
Environ. Sci. Technol. 23 (1989) 1242.
4. Shin, C.S., Kim, K.H., Yu,S.H.and Ryu, S.K.,Adsorption of methyl mercaptan and
hydrogen sulfide on impregnated activated carbon. Presented at 7~ International
Conference on Fundamentals of Adsorption, Nagasaki, Japan, May 20-25,2001.
5. Tanada, S.,Boki, K. and Matsumoto, K., Adsorption properties of methyl sulfide and
methyl disulfide on activated carbon, zeolite, and silicate, and their porous structure.
Chem. Pharm. Bull. 26 (1978) 1527.
6. Dalai, A.K., Tollefkon, E.L., Yang, A. and Sasaoka, E., Oxidation of methyl
mercaptan over an activated carbon in a futed bed reactor. Ind. Eng. Chem. Res. 36
(1 997) 4726.
7. Nuzzo, R. G., Zegarski, R. B. and Dubois, L.H., Fundamental studies of the
chemisorption of organosulfur compounds on Au (111). Implications for molecular
self-assembly on gold surfaces. J. Am. Chem. SOC.109 (1987) 733.
8. Lahaye, J., Nanse, G., Bagreev, A. and Strelko, V., Porous structure and surface
chemistry of nitrogen containing carbons from polymers. Carbon 37 (1999), 585.
9. Bagreev, A., Bashkova, S. and Bandosz. T. J., Dual role of water in the process of
methyl mercaptan adsorption on activated carbons. Langmuir, in press.
10. Lastoskie, C.M., Gubbins, K.E. and Quirke, N.J.,Pore size distribution analysis of
microporous carbons: a density functional theory approach. J Phys. Chem. 97 (1993)
4786. Olivier, J.P., Modeling physical adsorption on porous and nonporous solids
using density functional theory. J. Porous Materials 2 (I 995) 9.

145
1 1. Dubinin, M.M. and Radushkevich, L.V., Equation of the characteristic curve of
activated charcoal. Proc. Akud Sci. USSR 55 (1947) 33 1.
12. Everett, D. H.and Powl, J.C., Adsorption in slit-like and cylindrical micropores in the
Henrys law region. J. Chem. SOC.,Faradqy Trans., 1.72 (1976) 619.
13. McCallum, C.L.,Bandosz, T.J., McGrother, S.C., Muller, E.A. and Gubbins, K.E., A
molecular model for adsorption of water on activated carbon: comparison of
simulation and experiment. Lungmuir 15 (I 999) 533.
14. Adib, F., Bagreev, A. and Bandosz, T.J., Analysis of the relationship between H2S
removal capacity and surface properties of unimpregnated activated carbons.
Environ. Sci. Technol.34 (2000) 686.
15. Dubinin, M.M., Porous structure and adsorption properties of active carbons. In
Chemistry and Physics of Carbon, P. J . Walker, Ed., M. Dekker: New York, 2 (1 966)
51.
16. Lodewyckx, P. and Vansant, E.F., Water isotherms of activated carbons with small
amounts of surface oxygen. Carbon 37 (1999) 1647.

146
STUDIES ON THE ADSORPTION PROPERTIES OF ION-EXCHANGED LOW
SILICA X ZEOLITE

H. JIANG, W. TANG. J. P.ZHANG, B. Y.ZHAO AND Y . C . XIE'


College of Chemistry and Molecular Engineering. Peking University. Beijing 100871. China
E-mail: yxie 63Dku.edu.cn

It was found that two kinds of water exerted different influence on the nitrogen adsorption capacity
of IiLSX. One was so called "residual water", which were left in zeolite after heating hydrated
LiLSX the other was so called "adsorbed watef'. which were introduced from outside into fully
dehydrated LiLSX. In the "adsorbed water" samples, the nitrogen adsorption capacity was duced
not so much as that in the "residual water" samples. This can be explained by the heterogeneity of
adsorption sites in IiLSX. The influence of Ii+and Cak exchange degree on the nitrogen adsorption
capacity of IiNaLSX and CaNaLSX were compared. After Li+ exchange degree reached a thnshold
value at about 2/3, the nitrogen adsorption capacity of LiNaLSX started to increase rapidly; while in
the case of CaNaLSX, the threshold appead at about 1M. This is due to that the active adsorption
sites in LiNaLSX are the 32 Li+ at SIII sites, which account 1/3 of the total Li+; while the active
adsorption sites in CaNaLSX are the 32 Caz+at SII sites, which account 2/3 of the total Ca".

1 Introduction

Oxygen is an important gas in industry. It has numerous application such as steel and
nonferrous metallurgy industry, chemical industry, pulp and paper industry, glass melting
furnace, welding, wastewater treatment and city garbage burn out etc.. The traditionally
method of oxygen production is by cryogenic distillation of air. This method needs low
temperature and high pressure equipments which is complicate and high cost. The
cryogenic equipment starts up and shuts down very slowly (several days). Pressure Swing
Adsorption (PSA) as an alternative source of oxygen production first became a viable
option in the early 1970's. Now PSA is more economic and convenient for small and
middle size oxygen production, especially when high purity oxygen is not required [ 11.
Synthetic zeolites 5A and faujasite X (13X
or CaX)have been typically used as adsorbents,
which adsorb NZ more strongly than 0 2 . The
extraframework cations in the zeolites play a
major role in the adsorption selectivity. Because
the quadruple moment of NZ is about 3 times
more than that of Oz,the interaction between the
extraframework cations and nitrogen is much
stronger.
In 1989, Chao [2] reported that LiLSX
(lithium ion-exchanged low silica X zeolite,
having a Si/Al ratio close to 1.0) showed an
unexpected high capacitiy and selectivity for
nitrogen over oxygen. He found a Li' exchange
threshold value in LiNaLSX at about 2/3. Below
Figure 1. Structure of faujasite zeolite and
that threshold value, the adsorption capacity of
the distribution of cation sites

To whom correspondenceshould be addressed.

147
N2 was rather small; while above that, it increased drastically.
To explain this point, the site population of Li' in LiNaLSX and LiLSX was
investigated by neutron dihction and 'Li MAS NMR [3,4,5]. LSX has faujasite
structure (shown in Fig.l), in which five kinds of cation site (SI,SI', SII, SII' and SIII)
exist. G.Engelhardt et.al. [3] reported that in fully exchanged LiLSX, SI', SII and SIII
sites were occupied each by 32 Li' per unit cell, among which only Li' at SIII sites can
adsorb N2. In the case of LiNaLSX, only SI' and SII sites are occupied by Li' up to about
213 Li' exchange. At Li' exchange between 2/3 and 1, SI' and SII sites have already each
been fully occupied by 32 Li', so that the population of SIII sites increase, resulting in the
rapid increase of N2 adsorption.
RT.Yang et.al. [6] studied the influence of residual water on the adsorption
properties of LSSX. They removed water out of hydrated LiLSX by heating at different
temperatures. Their results showed that very small amounts of water in LiLSX have a
significant effect on the adsorptive capacity. The nitrogen adsorption capacity dropped
from -17.4 N2 mo1eculedu.c. (per unit cell) for the hlly dehydrated LiLSX to <2 N2
molecules/u.c. when the sample contained about 32 water molecules/u.c.. They pointed
out that this also could be well correlated to the Li' population of SIII sites in LiLSX. Li'
at SIII sites preferred to attract H 2 0 molecules over N2 molecules, because of the much
stronger interaction between H 2 0 and Li'. When H 2 0 molecules were adsorbed on Li' at
SIII sites, they would block the adsorption of other molecules. This explains why it took
only 32 H 2 0 molecules to significantlydiminish the N2 adsorption capacity.
In PSA oxygen production, dehydrated LiLSX loaded in adsorbent beds might absorb
water h m air to be deactivated gradually. Does this kind of adsorbed water influences
the adsorption properties of LiLSX in the same way as the residual water does? Our
experimentalresults show that the two kinds of water have different effect.
We found that CaNaLSX reported by C.G. Coe et.al. [7] also had a Ca2' exchange
threshold but at a different value at about 1/3. We tentatively explain the difference of
threshold in the view of the different site population of Li' and Ca2' in LiNaLSX and
CaNaLSX.

2 Experimental

Sodium, potassium LSX (NaKLSX) was synthesized following the procedure reported by
Kuhl [8]. On the basis of Kuhl's work, the synthesis conditions such as the ageing
temperature, reactants composition etc. were optimized further so that it was easy to
repeatedly prepare NaKLSX with crystallinity near 100% in our lab [9]. The product was
characterized by XRD and elemental analysis. XRD pattern indicated that the synthesized
NaKLSX was pure faujasite-type zeolite without detectable crystalline impurities or
amorphous materials. Elemental analysis showed that the synthesized NaKLSX had SUAl
molar ratio close to 1.O.
LiLSX was prepared by five static ion exchange of the synthesized NaKLSX at 100
C with a 5.6-fold equivalent excess of 2.0 M LiCl [lo]. After ion exchange, LiLSX with
Li/Al greater than 98% was obtained.
The adsorption isotherms of N2 and O2were measured by constant volume method at
ambient temperature. The adsorption properties deduced from the isotherms further
confirmed the high crystallinityand Li' exchange degree of LiLSX prepared in our lab.

148
3 Results and Discussions

3. I Influence of water ahorption on LiLSX


Influence of residual water on the adsorption properties of LiLSX has been reported by
RT.Yang et.al. [6]. We repeated the experiment in our lab. Fig.2 (a) shows the N2
capacity versus the residual water in LiLSX. At first, as the water content goes up, the N2
capacity drops sharply from 30mYg to 3.8mVg, then it descends much more slowly. The
curve reveals a turn-point at about 32 H20per unit cell. This result is in good agreement
with that of R.T.Yang.
In the industrial application, dehydrated LiLSX is loaded in the beds of PSA
equipment and water in the air can be adsorbed gradually on LiLSX, resulting in the
decrease of its adsorption performance. The adsorbed water may exert different
influence on the adsorption properties of LiLSX fiom the residual water. So we
introduced water into fully dried LiLSX at ambient temperature step by step and measure
the change of Nz adsorption capacity. Seen fiom Fig2 (b), when water content was below
about 7 molecules per unit cell, the nitrogen capacities were almost the same for the two
kinds of samples. Yet above that water content, the nitrogen capacity was always higher
for the samples with adsorbed water. After heating the samples with adsorbed water
in sealed tube, the nitrogen capacity dropped again to almost the same amount as that of
the samples with residual water, as shown in Fig2 (c).

o P 40 60 80 im izo 140 160


y\glsrcalent(mdeduC)
Figure 2. Nitrogen adsorption capacity versus water content. (a) m samples with residual water
(b) V samples with adsMbed water (c) A samples with adsorbed water, heated in sealed tube
This can be explained by the heterogeneity of adsorption sites and the migration of
the adsorbed water in LiLSX. As mentioned above, Li+ at SIII sites are the strongest
adsorption sites in LiLSX, but there are many other site which can adsorb water also.
TPD results obtained by B. Hunger et.al. [I I] showed four desorption peaks for water
adsorbed on L U X , the desorption energy of which were 50, 58, 69, 85 KJ/mol,
respectively. The water content corresponding to the maximum desorption energy (85
KJ/mol) peak was about 29 moIecules/u.c., quite near 32 molecu1edu.c.. Though the
authors didnt make designation to this peak, according to the above results, we

149
tentatively attribute it to water adsorbed on Lit at SIII site inside the supercages. When
water molecules were introduced into a dehydrated LiLSX particle, from the view of
thermodynamics, it was most favorable for them to be adsorbed on the strongest
adsorption sites, namely Li at SIII sites, thus caused the sharp decrease of nitrogen
capacity. But from the view of kinetics, part of the water molecules might adsorb on the
weaker sites before they could diffuse into the inner part of the particle. Therefore, some
Li at SIT1 sites located in the inner part of a LiLSX paricle remained intact, leading to the
higher nitrogen capacity. Heating the samples with adsorbed water caused the
migration of water molecules from the weaker adsorption sites to Li at SIII sites, so that
the nitrogen capacity decreased again to that of the samples with residual water. The
adsorbed water introduced from outside influences nitrogen adsorption not so seriously as
the residual water does. This point provides for us a round understanding of the water
effect in the 4 PSA industrial application.

3.2 Relationship between N2 arlsorption capacity and Cd content in CaNaLSX


As mentioned above, Li content in LiNaLSX showed a threshold value at about 2/3,
above which the nitrogen adsorption capacity increased rapidly [lo], as shown in Fig.3
(a). Ca2 exchanged LSX can also be a good adsorbent for N2 [7].We noticed that N2
adsorption capacity of CaNaLSX as a function of Ca2 exchange degree gives a threshold
value at 3 1% (about 113) [7],different from that of LiNaLSX, as shown in Fig.3 (b).
1 (b) /

..
j 31
o! , , .i/ , . , , , . ,
0 P 40 eQ 80 im
Exchangeoeg- (%I
Figure 3. Comparison of threshold value between LiNaLSX and CaNaLSX.
This difference of threshold values is owing to the difference of their cation
population and distribution. K. Cheetham et.al. [I21 employed X-ray and neutron
dithction to study the structure of CaLSX. They observed that SI and SI1 sites were
respectively occupied by nearly 16 and 32 Ca2 per unit cell. Applying this result to the
N2 adsorption on CaLSX, we can draw the following conclusions. Ca2 at S1 sites which
are located in the double six-ring units, are inaccessible to N2 molecules. Ca2 at SII sites,
which are on the wall of the supercage, are located above the six-membered oxygen ring,
so that they can contaet and adsorb Nz molecules. In the case of CaNaLSX, when Ca2 are
exchanged into LSX, the first 16 Ca will occupy SI sites, because the energy of Ca at
S1 site is lower. The following exchanged 32 Ca2 are located at SII sites. Therefore,
when Ca2 content exceeds 16 Ca2 per unit cell (i.e. 113 of the total Ca in CaLSX), the
adsorption capacity of N2 increases rapidly. Thus, the site population of Ca2 is in
accordance with the threshold value.

150
4 Acknowledgement

We thank the Major State Basic Research Development Program (G2000077503) in


China for the fmancial support.

References

1. Kumar R., VSA process for oxygen production-a historical perspective. Separation
Science and Technology31 (1996) pp. 877-893.
2. Chao C. C.,Process for separating nitrogen from mixtures thereof with less polar
substances. US Pat. 48592 17 (1 989)
3. Feuerstein M., Engelhardt G., McDaniel P. L., MacDougall J. E. and Gaffney T. R.,
Solid state nuclear magnetic resonance investigation of cation siting in LiNaLSX
zeolites. Microporous and Mesoporous Materials 26 (1998) pp. 27-35
4. Plevert J., Di R e m F., Fajuia F. and Chiari G., Structure of dehydrated zeolite Li-
LSX by neutron diffiction: Evidence for a low-temperature orthorhombic faujasite.
J. Phys. Chem. B 101 (1997) pp. 10340-10346
5. Feuerstein M. and Lob0 R. F., Characterization of Li cations in zeolite LiX by solid-
state NMR spectroscopy and neutron diffraction. Chem. Muter. 10 (1 998) pp, 2197-
2204
6. Hutson N. D., Zajic S. C. and Yang R. T., Influence of residual water on the
adsorption of atmospheric gases in LiX zeolite: experiment and simulation. Ind Eng.
Chern. Res. 39 (2000) pp. 1775- 1780
7. Coe C. G., Kirner J. F., Pierantozzi R. and White T. R., Nitrogen adsorption with a
Ca andor Sr exchanged LiX zeolite. US Pat. 5 152813 (1992)
8. Kuhl G. H., Crystallizationof low silica faujasite. Zeolites 7 (1987) pp. 451-457
9. Jiang H., Tang W.,Zhao B. Y. and Xie Y. C., Studies on synthesis of Maximun
Aluminum X zeolite. ChemicalJ of Chinese Universities 23 (2002) pp. 772-776
10. Kirner J. F., Nitrogen adsorption with highly Li exchanged X zeolites with low Si/AI
ratio. U S Pat. 5268023 (1993)
11. Hunger B., Klepel O., Kirschhock C., Heuchel M., Toufar H. and Fuess H.,
Interaction of water with alkali-metal cation-exchanged X type zeolites: TPD and
XRD study. Langmur, 15 (1999) pp. 5937-5941
12. Vitale G., Bull L. M., Moms R. E. and Cheetham A. K., Combined neutron and X-
ray powder diffraction study of zeolite CaLSX and a *H NMR study of its complex
with benzene. J. Phys. Chem. 99 (1995) pp. 16087-16092

151
CARBONIZATION OF ORGANIC WASTES USING SUPER-HEATED WATER
VAPOR AND THEIR ADSORPTION PROPERTIES

HIROYUKI YOSHIDA, NAOYA MIYAGAMI,AND MASAAKI TERASHIMA


Department of Chemical Engineering, Graduate School of Engineering, Osaka Prefecture
University, 1-1, Gakuen-cho,Saki City, JAPAN

E-mail: yoshida@chemeng.osakfi-u.acJp

Various organic wastes, such as waste wood chip, sake lees, used tealeaves and so on, were
-
carbonized with the super-heated water vapor at 623 K. B y the 30 90 minutes treatment, the
organic wastes lost about 50 - 90 % of their original weight. The capability of gas adsorption has
been evaluated. The surface areas determined by nitrogen adsorption for the carbonized materials
were much smaller than that of the activated carbon Granular Shirasagi. The surface area
determined by carbon dioxide adsorption, on the other hand, of the carbonated materials were almost
the same order of magnitude to that of the activated carbons. These results show that the
carbonized materials have micro-pores whose diameter is less than 50 nm.
We have found that the waste wood and used tealeaves showed high adsorption capabilities for
ammonia gas in low equilibrium pressure (< 13.3 kPa). The amount of the adsorbed ammonia for
these carbonized materials were much higher that of the activated carbon Granular Shirasagi GS3 x
4/6. These results suggest that the carbonized materials from organic wastes could be utilized as
adsorbents for ammonia

Introduction

Conversion of organic wastes to usehl resources is a challenging task for the


engineers to contribute to the realization of the sustainable development in the 2 1st century.
The Japanese government, for example, plans to reduce the amount of the wastes
processed by landfill to 370,000,000 todyear in the year of 2010 [ 13. Since about 80%
of the wastes are organic wastes such as sludge, paper, wood chip, fiber, and excreta of
animals and so on, the conversion of the organic wastes to the useful resources is a key
factor to achieve the government plan. Production of methane and compost from the
waste foods and the animal excreta has been studied by the many researchers. Charcoal
production fiom the waste wood chips has also studied to utilized the organic wastes [2].
Super-heated water vapor has been widely used in many industrial processes such as
heat-exchange process and drying, and has also been used in the activation process for
activated carbon production. Recently, the super-heated water vapor has been utilized in
food industry for production of instant food and drying of vegetables and tea leafs. The
characteristics of the super-heated water vapor [3] are (1) it can heat the materials without
oxidation because it does not contain oxygen and carbon dioxide, (2) drying speed
becomes much faster than super-heated air due to heat emission of water molecules, and
(3) waste gas is easily recovered by condensing.
In this work, the various organic wastes were carbonized with the super-heated water
vapor at 623 K. The specific surface areas and adsorption characteristics of the
carbonized materials have been studied.

152
2 Methods

Preparation of carbonized materials


Various organic wastes were carbonized with super-heated water vapor using a rotary
drum super-heated water vapor generator (SJH-IOM, Johnson Boiler, Japan) and a rotary
kiln (JBT-I OM, Johnson Boiler, Japan). The organic wastes were processed at 623 K for
30 - 90 min. Process data and properties of various carbonized materials are
summarized in Table 1. An activated carbon (Granular Shirasagi GS3 x 416, Takeda
Chemical Industries Ltd., Japan) and an activated carbon prepared for alkaline gas
adsorption (GAH 4-8, Cataler Corp., Japan) were used as controls.
Measurements of specific surface area and micro-pore volume
Specific surface areas of various carbonized materials were measured by nitrogen gas
adsorption with BET methods using an automated surface area analyzer (micro-track type
4200, Nikkiso, Japan). For mesopores whose diameter were less than 50 nm, the surface
areas and pore volumes were measured by carbon dioxide adsorption. The carbon
dioxide adsorption at 298 K was measured with Bellsorp 28 (BEL Japan). The pore
volume was determined using Dubinin-Radushkevich equation [4], and the surface area
was determined by Medek's method [5].
Measurement of ammonia gas adsorption
Adsorption capacities of the carbonized materials for ammonia gas were anaIyzed by
batch-wise equilibrium experiment using a micro-balance. The experimental methods
were described in the previous paper in detail [6].
Table 1 Carbonized wastes and their process data

I Garbage 1 10.0 I 1.O I 90 J

3 Results

Surface areas determined by nitrogen gas adsorption (SNZ),and those determined by


carbon dioxide (SCo2)for the various carbonized materials are summarized in Table 2.
The carbonated materials showed much smaller SN2 than the activated carbon Granular
Shirasagi. The Sco2, on the other hand, of the carbonated materials were almost the same
order of magnitude to that of the activated carbon. These results show that the
carbonized materials have mesopores whose diameter is less than 50 nm.

153
Table 2. Surface areas of carbonized materials measured by nitrogen, carbon dioxide,

Figure 1 shows a logarithmic plot of the amount of adsorbed ammonia against the
equilibrium pressure. The adsorption isotherms were well correlated by Freundlich
equation. The waste wood and used tealeaves showed high adsorption capabilities in
low equilibrium pressure (< 13.3 Wa). The amount of the adsorbed ammonia for these
carbonized materials were much higher than that of the activated carbon Granular
Shirasagi GS3 x 416. In the case of the activated carbon GAH 4-8, specialized for
alkaline gas adsorption, the straight line bent at the equilibrium pressure 13.3 kPa. This
result suggests that ammonia adsorbed on the mesopores of the activated carbon GAH 4-8
under low equilibrium pressure and then adsorbed on the macro-pores under high
equilibrium pressure. In the cases of the carbonized materials, ammonia gas adsorbed on
the mesopores, because the surface areas measured by nitrogen gas adsorption were very
small. in order to confirm these points, the adsorption data were analyzed by
Dubinin-Radushkevichequation [4]:
W=W&xp[-(RT/E)( In(PdP))] (1)
where W, Wo,E, PO, P, R, T are adsorption volume [cm3/g-adsorbent], saturated
adsorption volume [cm3/g-adsorbent], characteristic energy for adsorption FJ/mol],
saturated pressure Fpa], equilibrium pressure [kPa], gas constant, and temperature [K].
Figure 2 shows a plot of logW against (log(Pdp)) for the activated carbons. The
experimental data deviated from the straight line in the high-pressure range. These
results suggest that the ammonia adsorbed mainly on the macro pore in the high-pressure
range, and adsorbed on the mesopores in the low-pressure range. The similar plots for
various carbonized materials are shown in Figure 3. All the experimental data were well
correlated by straight lines, suggesting that ammonia adsorbed on the mesopores. The
micro-surface areas determined from ammonia adsorption are also listed in Table 2.
Elementary analysis would be useful to characterize unique carbons developed in this
work.

154
Pressure [ma]
Figure 1.Ammonia adsorption on various carbonized waste materials
+
0 : Used tealeaves :
0:Scrapwood 'I:

-1
-1
3
-
M
0
-1.5 B
8
-2 -1.5

Figure 2 Logarithmic plot for


activated carbon

Figure 3 Log plot for carbonized


wastes (right)

4 Discussion

Various organic wastes could be successfully carbonized by the super-heated water vapor
-
at 623 K. As summarized in Table 1, the organic wastes lost 50 90 'YOof their original
-
weight by 30 90 min treatment. Observation by scanning electron microscope (data

155
not shown) revealed that the macroscopic structures of the carbonized materials strongly
depend on the original structures. The measurements of specific surface area showed
that the diameters of the mesopores of the carbonized materials are smaller than nitrogen
molecule (about 50 nm), and are probably about 20-30 nm.
It should be noted that the adsorption capacities of the waste wood and the used tealeaves
for ammonium gas in the low equilibrium pressure are much higher than that of Granular
Shirasagi, and are compatible to that of GAH 8-4 which is specialized for alkaline gas
adsorption. These carbonized materials probably can be used as the substituted materials
for the activated carbon.
The carbonization of the organic wastes by the super-heated water vapor is an inexpensive
method to convert the organic wastes to useful resources. Effects of carbonization
temperature, heating rate, and other conditions on the characteristics of the carbonized
materials are extensively studying in our laboratory.

Acknowledgement
A part of this research was financially supported by 21" Century COE Program (24403,
E-1) from Ministry of Education, Culture, Sports, Science and Technology of Japan.

References

1. Notification No. 34 of the ministry of the Environment (2001)


2. Nakazaki K., Recycle of organic sludge by conversion into industrial raw materials,
In SOC.Chem. Eng. Japan (ed.), Processing of Wastes (in Japanese) (2001), pp.
65-74.
3. Iyota H., Nishimura N., and Nomura T., Reverse process of super heated steam
drying from condensation to evaporation, Nihon Kikai Gakkai Ronbunshu Part B (in
Japanese) 66 (2000) pp. 2681-2688.
4. Dubunin M. M., Adsorption in micropores, J. Colloid Interjhce Sci. 23 (1967) pp.
489-499
5. Medek J., Possibility of micropore analysis of coal and coke from the carbon dioxide
isotherm, Fuel 56 (1977) pp. 13 1-133
6. Yoshida H., Oehlenschraeger, Water vapor adsorption on basic anion exchangers, Ind.
Eng. Chem. Res., 40 (2001) pp. 4850-4856.

156
FURTHER SUCCESSFUL APPLICATIONS OF THE NEW THEORETICAL
DESCRIPTION OF ADSORPTION/DESORPTIONKINETICS BASED ON THE
STATISTICAL RATE THEORY

WLADYSLAW RUDZINSKI
Department of Theoretical Chemistry, Faculty of Chemistry UMCSpl. Marii Curie-Sklodowskiej3,
Lublin, 20-031, POLANDph.: +48 81 5375633;fa: +48 81 5375685
e-mail: rud-insk@hermes.umcs.lub1in.d

TOMASZ PANCZYK
Groupfor Theoretical Problems of Adsorption, Institute of Catalysis and Surjiace Chemistry,
Polish Academy of Sciences, ul. Niezapominajek8, 30-239 Krakow, POUND

It is shown how the new approach to adsorptioddesorption kinetics based on the Statistical Rate
Theory can be successfully applied to describe the kinetics of dissociative gas chemisorption on
solids.

1 Introduction

Since I9 16 when Langmuir published his fundamental paper on adsorption, the Theory of
Activated AdsorptiodDesorption Kinetics (TAAD) has, almost exclusively, been used for
the interpretation of adsorptioddesorption kinetics. However, contrary to the success of
Langmuir equation to represent the adsorption equilibria, a dramatic failure of TAAD was
observed to represent by the Langmuir kinetic equation and its further modifications, [1-
31 the monitored adsorptioddesorptionkinetics.
So, it became more and more obvious that the theoretical hndamentals of
adsorptioddesorptionkinetics must be reconsidered.
The breakthrough came at the beginning of the eighties of the 20 century with
the new theoretical approach called the Statistical Rate Theory (SRT), linking the rate of
adsorptioddesorptionkinetics to the chemical potentials of the molecules in the bulk and
the adsorbed phases. [4]
The Statistical Rate Theory (SRT) is based on considering the quantum-mechanical
transition probability in an isolated many particle system. Assuming that the transport of
molecules between the phases at the thermal equilibrium results primarily fiom single
molecular events, the expression for the rate of molecular transport between the two
phases I and 2, RI2,was developed by using the first-order perturbation analysis of
the Schrodinger equation and the Boltzmann definition of entropy.

In eq. (I), p1 and p2 are the chemical potentials of the molecules in the phases I
and 2, respectively, and R, is the exchange rate at equilibrium.
The SRT approach was applied first to describe the kinetics of adsorption in the
systems with well-defined solid surfaces. [5-91 However, in spite of the demonstrated
impressive success, that new SRT approach did not meet a general easy acceptance. This
was probably due to some inertia caused by using the classical TAAD approach during
the previous 70 years.

157
Recently the new SRT approach has been generalized fiuther to take into
account the energetic heterogeneity of the actual adsorption systems and the possible role
of the interactions between the adsorbed molecules. 110-20) Most recently, the authors
have shown, that the SRT approach can be successfully applied to describe the multi-site-
occupancy adsorption of molecules which do not dissociate after being adsorbed. [14]
Compact simple analytical expressions were developed, and next used successfully to
correlate experimental data for adsorptioddesorption kinetics in various gadsolid systems.
The purpose of this presentation is to show, that the new SRT approach can be also
applied to represent the kinetics of dissociative gas adsorption on solids. That kinetics is
of a crucial importance in a variety of catalytic reactions occurring on solid surfaces.

2 Theory

Quantum-mechanical calculations of gas-solid interactions, and analysis of experimental


kinetic data suggest existence of a certain precursor state which is a weakly adsorbed
molecular species. In a next step the precursor species dissociates into adsorbed atoms.
Let us assume that the rate determining step of this kinetic process is the transition from
the precursor state to the adsorbed atoms.
We consider the adsorbed species passing the dissociation barrier as atoms, having
potential p*.They are in equilibrium with the bulk phase but not with the adsorbed atoms.
Then, pa # p* = % pg,where pa is the chemical potential of the adsorbed atoms, and pB is
the chemical potential of molecules in the gas phase. The SRT equation takes then the
form,

While assuming the Langmuir model for adsorption of atoms, and the ideal gas
expression for pg,from eq. (2), we obtain,
r 1

where K is a temperature dependent constant. At equilibrium, when de/dt = 0, eq. (3)


yields the following isotherm equation,
I &

- _
1+ ( ~ p ( ~ ) ) s e k T
where the superscript (e) refers to the equilibrium conditions.
The explicit form of the expression K', for the equilibrium exchange rate will
depend on the path of the dissociative adsorption in a particular adsorption system under
consideration. Namely it will depend on the character of the first step of that dissociative
adsorption. In general, that first step (molecular adsorption) may not yet be the precursor
state. By a fast surface diffusion, the adsorbed molecules may migrate to another state

158
from which they dissociate. Because K', - p"' we have to consider K', as the
following expression,
K', = K,p(e)(l-6g))" (5)
where (1 - 6:) is the fixtion of the adsorption sites available for molecular adsorption
in the' first adsorbed state, and n is the number of these adsorption sites occupied by the
adsorbing molecule consisting of s atoms. To solve the differential eq. (3), el,")has to be
related to 6. That relation will depend on the particular nature of chemisorption system
under consideration. For the purpose of illustration we will take into consideration the
dissociative adsorption of hydrogen molecules on the Fe(100) crystal face. For this
particular system, Christmann e t d . [21J reported kinetic data measured at different non-
equilibrium pressures, whereas the related quantum mechanical calculations can be found
in the recent paper by Kresse. [22] These calculations show that the most favorable
reaction path is when the center of hydrogen molecule approaching the surface is located
over the top site, and the molecule is oriented in such a way that after dissociation the
atoms go into two neighboring bridge sites. Thus 6, can be identified with the occupancy
of the top sites, because one adsorbed atom makes (eliminates) adsorption over one
neighboringtop site impossible. Thus 6, can be identified with 6 under the condition that
n=l.
So, we put n=l in eq. (5). Equation ( 5 ) can be then integrated analytically to yield,
r _
1 E_l -I -&

"I
sgn l-(Kp)2ekT (Kp)2ekT 2E
ln[[Kpeg - 1 b 2 -ZKpekT6+KpekT
- +C(6)
t(0) =

L J

When 6(t=O)=O, the integration constant C takes the form,

Using eq. (6) we have been able to fit successfully the kinetic isotherms reported by
Christmann, e t d . [21]. The results are shown in the Figure 1.
So far we have not considered yet the more complicated problem of the kinetics of
dissociative adsorption of asymmetric molecules like CO. The new level of complexity
here lies in the necessity of considering of the mutual blocking of the adsorption centers
for different atoms after dissociation of adsorbed molecule. Nevertheless we continue our
efforts along these lines, and we hope to publish some solution for that problem in our
future publications.
- 0.4
- CD
- - 0.3 e!ao,-
- a>
- 0.2 8
- .-> Q)

- 0.1 9 CI

- h
0
0.01 0.1 1 10 100
time, min
figure 1. The bat-fit of the experimental hetic isotherms of hydrogen adsorption on Ni(100). measud by
christmann et.al. at p~7.10-'~ Ton, 2.5 .lo4 Tom and 7.5 .lo* Tom by applying the SRT eq. (6). and the
following best-fit parameters: K"exp(&") = 7.52.103Todn. &=2.58.10' Ton-'min". and N,= 16.02a.u.
where a.u. axe the arbitrary units used by christmanner.al. [21]to represent the adsorbed amount.

References

1. Rudzinski W. and Panczyk T., The langmuirian adsorption kinetics revised: A


farewell to the XX-th century theories?, Adsorption, 8 (2002) pp.23-34.
2. Rudzinski W. and Panczyk T., Phenomenological Kinetics of Real Gas-Adsorption-
Systems: Isothermal Adsorption, Journal of Non-Equilibrium Thermodynamics, 27
(2002) pp.149-204.
3. Rudzinski W. and Panczyk T.,Remarks on the current state of adsorption kinetic
theories for heterogeneous solid surfaces: A comparison of the ART and the SRT
approaches, Lungmuir, 18 (2002) pp.439-449.
4. Ward C.A. and Findlay R.D., Statistical rate theory of interfacial transport. IV.
Predicted rate of dissociative adsorption, J. Chem Phys. 76 (1982) pp.5624-5631.
5. Ward C.A. and Elmoseli M.B., Molecular adsorption at a well defined gas-solid
interphase: Statisticalrate theory approach, Sur$ Sci. 176 (1986) pp.457-475.
6. Elliott J.A.W. and Ward C.A., Chemical potential of adsorbed molecules from a
quantum statistical formulation,Lungmuir 13 (1997) pp.951-960.
7. Elliott J.A.W. and Ward C.A., Statistical Rate Theory and the Material Properties
Controlling Adsorption Kinetics. In RLJDZINSKI W., STEJXE W.A., ZGRABLICH
G, (eds.) Equilibria and Dynamics of Gas Adsorption on Heterogeneous Solid
Surfaces (Elsevier: New York, 1997), pp.285-333.
8. Elliott J.A.W. and Ward C.A., Temperature programmed desorption: A statistical
rate theory approach, J. Chem. Phys. 106 (1997) pp.5677-5684.
9. Elliott J.A.W. and Ward C.A., Statistical rate theory description of beam-dosing
adsorption kinetics, J. Chem Phys. 106 (1997) pp.5667-5676.

160
10. Rudzinski W. and Panczyk T. Surface Heterogeneity Effects on Adsorption
Equilibria and Kinetics: Rationalisation of Elovich Equation. In SCHWARZ J. and
CONTESCU C. (eds.) Su$aces of Nanoparticles and Porous Materials (Marcel
Dekker, 1999), pp.355-390.
11. Rudzinski W. and Panczyk T., The Kinetics Of Isothermal Adsorption On
Energetically Heterogeneous Solid Surfaces: A New Theoretical Description Based
On The Statistical Rate Theory Of Interfacial Transport, J. Phys. Chem. B. 104
(2000) pp.9149-9 162.
12. Rudzinski W. and Panczyk T., Kinetics of Gas Adsorption in Activated Carbons,
Studied by Applying the Statistical Rate Theory of Interfacial Transport, J. Phys.
Chem. B., 105 (2001) pp.6858-6866.
13. Rudzinski W. and Panczyk T., Remarks on the current state of adsorption kinetic
theories for heterogeneous solid surfaces: A comparison of the ART and the SRT
approaches, Langmuir, 18 (2002) pp.439-449.
14. Panczyk T. and Rudzinski W., Kinetics of Multi-Site-Occupancy Adsorption on
Heterogeneous Solid Surfaces: A Statistical Rate Theory Approach, J. Phys. Chem.,
106 (2002) pp.7846-7851.
15. Rudzinski W., Borowiecki T., Dominko A., and Panczyk T., New Method of
Estimating the Solid Surface Energetic Heterogeneity from TPD Spectra Based on
the Statistical Rate Theory of Interfacial Transport, Langmuir, 13 (1997) pp.3445-
3453.
16. Rudzinski W., Borowiecki T., Dominko A., and Panczyk T. and Gryglicki, J., On the
Quantitative Estimation of Surface Energetic Heterogeneity of Adsorbents and
Catalysts Surfaces from TPD Spectra Based on the Statistical Rate Theory of
Interfacial Transport: Variable Heating Rates as a Promisinig Way to Establish the
Values of All Unknown Parameters of Interest, Polish J. Chem., 72 (1998) pp.2103-
21 14.
17. Rudzinski W., Borowiecki T., Dominko A., and Panczyk T., A New Quantitative
Interpretation of TPD Spectra from Heterogeneous Solid Surfaces, Based on
Statistical Rate Ttheory of Interfacial Transport: the Effects of Simultaneous
Readsorption, Langmuir, 15 (1999) pp.6386-6394.
18. Rudzinski W., Borowiecki T., Panczyk T., and Dominko A., On the Applicability of
Anhenius Plot Methods to Determine Surface Energetic Heterogeneity of Adsorbents
and Catalysts Surfaces fiom Experimental TPD Spectra, A h . Coff.Znteface Sci., 84
(2000) pp. 1-26.
19. Rudzinski W., Borowiecki T., Panczyk T., and Dominko A. A quantitative approach
to calculating the energetic heterogeneity of solid surfaces fiom an analysis of TPD
peaks: Comparison of the results obtained using the absolute rate theory and the
statistical rate theory of interfacial transport, J. Phys. Chem. B, 104 (2000) pp. 1984-
1997.
20. Rudzinski W., Borowiecki T., Panczyk T., and Dominko A., Theory Of
Thermodesorption From Energetically Heterogeneous Surfaces: The Combined
Effects Of Surface Heterogeneity, Re-Adsorption, And Interactions Between The
Adsorbed Molecules, Lungmuir, 16 (2000) pp.8037-8049.
21. Christmann K., Schober O., Ertl G., and Neuman M., Adsorption of hydrogen on
nickel single crystal surfaces, J. Phys. Chem., 60 (1974) pp.4528-4540.
22. Kresse G., Dissociation and sticking of H-2 On the Ni(l1 l), (loo), and (1 10)
substrate, Phys. Rev. B, 62 (2000) pp.8295-8305.

161
CHARACTERIZATION AND ETHYLENE ADSORPTION PROPERTIES OF
SILVER-LOADED FER ZEOLITE POTENTIALLY USED AS TRAP MATERIAL
OF COLD-START HYDROCARBON EMISSION FROM VEHICLES

Y.TERAOKA
Department of Molecular and Material Sciences, InterdisciplinaryGraduate School
of EngineeringSciences, Kyushu University,Kasuga, Fukuoka 816-8580, Japan
E-mail: teraoka@mm.kyushu-u.ac.jp

H. ONOUE, H. FURUKAWA, 1. MORIGUCHI


Department of Applied Chemistry, Faculty of Engineering, Nagasaki Universiy,
Nagasaki 852-8521, Japan

H.OGAWA AND M. NAKANO


Nanyo Research Laboratoy, Tosoh Corporation, Shinnanyo, 746-8501. Japan

The adsorption properties of silver- and cupper-loaded zeolites for C2 and C3 hydrocarbons were
investigated to explore excellent materials for cold-start hydrocarbon trap. The adsorption property
and the stability of the adsorbents depended significantlyon the metal species and host zeolites. It
has turned out that silver-loaded femerite zeolite is the promising material with excellent olefin
selectivity, high adsorption capacity,desirable storage ability and hydrothermal stability.

1 Introduction

The air pollution by vehicle emissions especially in urban area is a serious problem to be
urgently solved, and regulation of the emission has been becoming more and more
tightened with respect particularly to NOx and non-methane hydrocarbons (NMHC). As
for the NMHC emission control, it is well known that the so-called cold-start hydrocarbon
emission, which contributes about 80% of the total emission, should be reduced. Just
after the ignition of an engine, the exhaust gas is too cold for the three-way catalyst
(TWC) to be active for removing CO, NOx and HC. The HC emission before the exhaust
gas heats up to the working temperature of the TWC is called as the cold-start HC
emission. One, and probably the most promising, countermeasure to reduce the cold-start
HC is to install the cold-stat HC trap on the upper stream to or on the same honeycomb
as TWC [I-51. Some of important requirements for trap materials (adsorbents) are as
follows:
0 They should have high adsorption capacity of HC during the cold condition.
0 They should store the adsorbed HC until the exhaust temperature becomes high
enough for TWC to work (200 C or higher).
0 They should not be deteriorated even under the hydrothermal condition in humid gas
stream at elevated temperatures like 850 C.
This paper reports adsorption property for C2-C3hydrocarbons and characterization
of Ag- and Cu-loaded zeolites to explore excellent materials for the cold-start HC trap.

162
2 Experimental
Three kinds of zeolites, which were kindly supplied by Tosoh Corporation, were used in
this study: W-ferrierite (FER) with SiOdA1203 molar ratio of 60.2, W - M F I with
SiOz/&o3=39.5 and NH4-b (BEA) with Si02/A1203=37. Ag- and Cu-loaded zeolites
were prepared by an impregnation method or an ion-exchange method using aqueous
solutions of metal nitrates, followed by the calcination in an open air at 500 "C for 1 h
(fresh sample). The hydrothermal treatment was carried out by exposing fresh samples to
moist air (10% HzO)at 850 "Cfor 5 h (aged sample). The metal loadings were expressed
on the basis of wt%; for example, 2.0wt% Ag-loaded FER(60.2) is expressed as Ag(2.0)-
FER(60.2).
The adsorption property was measured by a static method at 30 "C with a
conventional volumetric apparatus as well as by the temperature programmed desorption
(TPD) method. The details of the pretreatment and adsorption procedures were shown in
Results and Discussion section. Metal-loaded zeolite samples were characterized by
XRD, diffuse reflectance UV-Vis spectroscopy @RS) and electron spin resonance (ESR).

3 Results and Discussion


3.1 Adsorption of Cz and C, hydrocarbons on Ag-FER
Figure 1 shows adsorption isotherms of CzH.,. C2& and C& on the fresh Ag(2.0)-
FJZR(60.2). The sample was pre-evacuated at 500 "C for 0.5 h, and after cooling down to
30 "C,the first adsorption isotherm was obtained. After evacuation at the same

5 2.0
E
E
1 1.5
L
C

g 3
1.0

'$
0.5
$
2 0.0
0 0.1 0.2 0.3 0.4 0.5 0.6
Relative pressure, p/p,
-b 2.0
-
-in 2.0 -
E
. E 1.5 E 1.5
\
c
j
c)

1.0

0.5
e
8
-0
a 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Relative pressure, p/po Relative pressure, p/po
Figure 1. Adsorption isotherms of (a) CZ& (b) Cz& and (c) C3& on fresh Ag(2.0)-FER(60.2)preevacuated
at 500 "C for 0.5 h. Open and closed circles cormpond to the Ist and 2nd adsorption measurements (see text
for the details).

163
temperature for 0.5 h, the second adsorption measurement was carried out. The adsorbed
amount in the second isotherm corresponds to the amount desorbed during the evacuation
at 30 "C or that of weakly (reversibly) adsorbed species (V,,,),and the difference between
the first and second isotherms does to the amount of strongly adsorbed species (V,) which
did not desorb by the evacuation at 30 "C; the amount in the first isotherm thus
corresponds to the total amount adsorbed (V,=V,+V,). The first and second isotherms of
C2Hs completely overlapped each other. This indicates that parafiic C2& reversibly
adsorbs at 30 "C and no strong adsorbed species is formed. For C2& and C&, on the
other hand, the discrepancy between the 1st and 2nd isotherms was observed, showing
that strongly adsorbed species is formed in addition to weakly adsorbed species. These
result clearly show that Ag(2.0)-FER(60.2) is an olefin-selective adsorbent.

0.0 0.1 0.2 0.3


Ag loading I mmol g-1
Figure 2. The amount of C2H4 adsorbed on fresh Ag-FER(60.2) (at p/p0=0.15) as a function of Ag loading.
The Ag loading is shown in unit of mmol g-'.

0 100 200 300 400 500


Temperature I "C
Figure TPD chromatograms of C2I% from Ag- and H-FER(60.2). The aged sample was subje :d to the
hydrothermal treatment by exposing to moist air (1 0%HzO)at 850 "C for 5 h.

The adsorption of C2)4 was measured on fresh Ag-FER(60.2) with various Ag


loadings and the adsorbed amounts at p/po=O. 15 were plotted as a function of the Ag
loading (Figure 2). As can be seen from Figure l(a), the 1st and 2nd isotherms are nearly
parallel above p/po=O.15, which shows that the V, is almost constant above p/po=O. 15. The
amount of the weakly adsorbed species (V,) was roughly independent on the Ag loading,
only showing a tendency to slightly decrease with an increase in the Ag loading. The
amount of strongly adsorbed species (VJ,on the other hand, increased almost linearly

I64
with increasing the Ag loading, and the values of V, were in the same order of the Ag
loading: the ratios of VJAg ranged between 0.68 and 0.86 except for the Ag-FER(60.2)
with the lowest Ag loading (VJAg d.43). It can thus be concluded that Ag loaded idon
the FER zeolite is responsible for the formation of strongly adsorbed CZ& species.
Desorption of C2& as a function of temperature was measured by means of TPD
technique (Figure 3). A sample was preevacuated at 500 "C for 0.5 h, cooled down to
room temperature in vacuo, and exposed to C2& (100 torr) for 0.5 h. After evacuation
for 0.5 h and replacing the atmosphere with a helium flow, the sample was heated at a rate
of 10 "C min-' and the desorbed CZ& was monitored by TCD. The static adsorption
experiment showed that H-FER(60.2) adsorbed a substantial amount of C2& strongly at
30 "C, but a majority of them desorbed below 250 "C as shown in Figure 3. As for fresh
Ag(2.0)-FER(60.2), on the other hand, a majority of the strongly adsorbed C2& desorbed
above 200 "C.
Another characteristic feature of Ag-FER as a C2& adsorbent is that the adsorption
property depends only slightly on the pretreatments and the existing state of Ag. The
amount (Table 1) and temperature behavior (TPD, not shown) of strongly adsorbed C2&
is roughly the same after the evacuation (fresh) and oxidation at 500 "C;combined use of
DRS and ESR indicated that Ag idon the FER zeolite was present as cluster ions (Ag,,,?
[6] and metallic silver (Ag? after the evacuation and Ag,* and Ag2O after the oxidation.
More importantly, it has tuned out that Ag-FER is tolerant against the hydrothermal
treatment at 850 "C. Although the amount of the strongly adsorbed species somewhat
decreased, the desorption behavior did not so much change as can be seen from Figure 3
and Table 1.

Table 1. Adsorbed amounts of Czfi on various metal-loaded zeolite adsorbents (30 "C, p/@.15)')
Adsorbed amount I
Sample S:)l
Pretreatment mmol g-' V,per
[metal content I mmol g-'I m2 g-'
V. V..
Fresh 0.16 0.81 0.86 252
Ag(2.0)-FER(60-2) Ox-5002) 0.15 0.79 0.81 ----
[0:185]
Aging3' 0.13 0.73 0.70 256
Cu(0.6)-FER(60.2) Fresh 0.10 0.91 1.10 ----
[0.090] Aging" 0.01 0.82 0.1 1 ----
Ag(5.4)-MFI(39.5) Fresh 0.35 1.10 0.70 347
[0.499] Aging3' 0.16 0.54 0.32 216
Ag( 1.8)-BEA(37.0) Fresh 0.15 1.21 0.9 1 487
[O. 1641 Aging3' 0.06 0.20 0.37 98
1) All the samdes were evacuated at 500 "C for 0.5 h before. the adsorption measurements.
2) Exposure to 0,(100 torr) at 500 "C for 0.5 h. 3) Hydrothermal treatment by exposing fresh samples
to moist air (10% HzO) at 850 "C for 5 h. 4) VJmetal content. 5) Specific surface area.

3.2 Effect of metal species


Adsorption properties for C2& of Ag(2.0)- and Cu(0.6)-FER(60.2) were compared. In
the fresh state, Cu-FER strongly adsorbs a substantial amount of Cz& with the VJCu ratio
close to unity and the strongly adsorbed species desorbed between 200 "C and 500 "C,
indicating that Cu in FER zeolite in a fresh state also serves as an active site for C2&
adsorption with the desirable capacity and storage ability. After the hydrothermal (aging)
treatment, however, the capability of strongly adsorbing CZ& was almost disappeared.

165
There results clearly indicate that Ag is superior to Cu as the active metal species for C2H4
adsorption with respect to the hydrothermal stability.

3.3 Eflect of host zeolites


In the fresh state, Ag(5.4>MFI(39.5), Ag( 1.8)-BEA(37.0) and Ag(2.O>FER(60.2) showed
the C2I& adsorption property characteristic to Ag species idon zeolites, namely, forming
strongly adsorbed species with the VJAg ratios between 0.7 and 0.9 and desorbing it
above 200 "C. As described above, the impact of the hydrothermal treatment on Ag(2.0)-
FER(60.2) was not so severe. For the other two samples of MFI and BEA zeolites, on the
other hand, the adsorption capacity of both strongly and weakly adsorbed species
considerably decreased. Powder X-ray diffraction measurements showed that the
hydrothermal treatment caused the partial destruction of the zeolite structure for MFI and
BEA systems, but not for the FER system. Such structural stability was also recognized
from the change of the specific surface area, S, (Table I). The S, values of fresh and aged
Ag-FER was almost the same, while those of Ag-MFI and Ag-BEA decreased
considerably after the hydrothermal treatment.

3.4 Characteristics of Ag-FER as cold-HC trap material


The results reported in this paper clearly show that Ag-FER(60.2) is the promising
material for cold-HC trap with the following characteristics.
(1) superior selectivity to olefins (C2H4, C3H6)over a parafin (C2H6)
(2) high adsorption capacity with the VJAg ratio in the range of roughly 0.7-0.9
(at p/po=O.15)
(3) desirable storage ability desorbing the strongly adsorbed species above 200 "C
(4) high hydrothermal stability
Selectivity (l), capacity (2) and storage ability (3) originate from the characteristics
of Ag itself, while the stability (4) of Ag-FER(60.2) is contributed by the structure
stability of the FER(60.2) and insensitivity of the adsorption property of Ag to its existing
states. The higher SiO2/AI2O3ratio of FER(60.2) is the most probable reason for the
zeolite-structure stability. The insensitivity of Ag is a favorable property as -a stable HC
trap material, and its origin will be elucidated by a future study.

References

1. Burk P. L., Hockmuth J. K., Anderson, D. R., Sung, S, Punke, A, Dahle, U, Tauster
S. J., et al., Stud. S u Sci.
~ Catal., 96 (1995) pp.9 13-939.
2. Nishizawa K., J. of Soc. of Automotive Engineers of Jpn. (Japanese), 50 (1996) pp.
61-65.
3. Ballinger, T. H., Manning W. A., Lafyatis, D. S., SAE Paper (1997) pp.27-31
(970741).
4. Lyfyatis, D. S., Ansell, G. P., Bennett, S. C.,Frost, J.C., Millington P. J., Rajaram, R.
R., Walker A. P., Ballinger, T.H., Appl. Catal. B, 18 (1998) pp.123-135.
5. Czaplewski, K. F., Reits T. L., Kim, Y. J., Snurr, R. Q., Microporous and
Mesoporuos Materials, 56 (2002) pp.55-64.
6. Bogdanchikova, N. E., Dulin, M. N., Toktarev, A. V., Shevnia, G. B., Kolomiichuk,
V. N., Zailovskii V. I., Petranovskii, V. P.,Stud Surf: Sci. Catal., 84 (1994) pp. 1067-
1074.

166
PRESSURE-DEPENDENT MODELS FOR ADSORPTION KINETICS ON A CMS

Youn-Sang Bae, YoungKi Ryu, and Chang-Ha Lee'


Dept. of Chem. Eng., Yonsei University, Seoul, Korea
Tel.: +82-2-2 123-2762, Fax: +82-2-3 12-6401, E-mail: leechk2vonsei.ac.h

An adsorption kinetic model was developed to evaluate the adsorption rates of five pure gases (Nz,
02, Ar, CO, and Ch) on a Takeda-3A CMS over a wide range of pressures up to 15atm. The kinetic
characteristics of adsorptionon the CMS were studied by using the adsorption equilibrium of five pure
gases measured at three different temperatures and their physical properties. Since the diffisional time
constants of all the components showed much stronger dependence of pressure than those expected by
the traditional Darken relation, a structural diffusion model was applied to predict the strong pressure
dependence. The proposed model successfully predicted the diffisional time constant up to high
pressure on the CMS.

1 Introduction

Carbon molecular sieve (CMS) is useful in air separation processes because of its
ability to selectively discriminate on the basis of molecular size and hence adsorb the
smaller oxygen molecule over nitrogen. The difference in the adsorption kinetics of
various gases aIlows the separation of gas mixtures into pure components using pressure
swing adsorption (PSA).
Generally, the kinetic rate constants on adsorbents increase with increasing surface
coverage. The reason is probably related to surface diffusion (Reid and Thomas, 1999).
Ruthven (1992) pointed out that the diffisivity of oxygen increased with adsorbate
loading on CMS more or less in terms of Darken's equation. The pressure-dependencesof
D/? are generally predicted by Darken-relation, but in some cases the pressure-
dependeces are so strong that it cannot be predicted by traditional Darken-relation. Hence,
in these cases, the model that can predict these strong pressure-dependencesis needed.
In this paper, the isotherms and diffisivities of five pure gases @I2, 02,Ar, CO, and
CH4)in CMS were studied in the range of 293-313K, 0-15atm.

2 Experimentals

The volumetric method was used to obtain the data of the adsorption equilibrium and
the adsorption kinetics. The adsorbent used in this study was Carbon Molecular Sieve
(Takeda Co.) and has an average pore size of 3A. The adsorbates were 99.99%-purity
gases. Prior to the measurements, the adsorbent was regenerated by evacuation at 423K
during 12hr. The CMS is loaded with an adsorbate in a stepwise procedure; equilibrium
and kinetic data are obtained in each step. In determining the size of each step, we
considered the pressure range in which linear isotherm can be applied.

167
3 Mathematical Models

3. I Equilibrium Models

bP
Langmuir isotherm: c, = c,
l+bP
b PI'"
L F isotherm: c, = c,
1 + b P""
bP
Toth isotherm: c, = c,
(1 + (bP)" ) I / "

D-R isotherm: P
C , = C, -exp[- a' x(In(-))2 ] (4)
P
O

3.2 Pressure-dependent Modelsfor Adsorption Kinetics


The exponential increases of the effective diffusional time constant might be explained
by the following Darken-relation and this relation was derived under the assumption that
the chemical potential gradient is the driving force of the diffusion (Ruthven, 1992).

Darken-relation :

If above Darken-relation (Eq.(5)) is combined with the Langmuir, L-F, Toth, and D-R
isotherms, the resulted models are as follows:
t

Darken-Langmuir: D, = D,, ( 1 + bP ) (6)


I

Darken-LF: D,= D,, [ 1 + bP"" ] (7)

Darken-Toth:

I
Da rken-DR. D, = Dfl
ln(4 / P)

168
4 ResultsandDiscussion

4.1 Adsorption Equilibrium


Adsorption isotherms of five gases at 293K for C M S determined from the volumetric
experiments are shown in Fig. 1. Adsorption capacities of each gas are as follows: Nz,4,
Ar < co << CH4.

- 4

8E CH, --- Langmuir


E3 .... ........ loth
01
c

0
3

5 2
D

i
$ 1

0
0 2 4 6 8 10 12 14 16 18
Pressure [atm]

Figure 1. Adsorplion isotherms of five pun, gases at 293K


and h e lib of the Langmuir, LF. and Tom lsomenns

4.2 Adsorprion Kinetics


In Figure 2, dimensionless pressure histories of NZ and 02 at the same temperature and
pressure condition were shown and they were compared with prediction results by
Brandani model (Brandani, 1998). The values of Dl? were obtained by nonlinear
regression using the least-square method.

I i I

3:L
0.2

Dodng d l P. (Simu1fi.d)

-
0.1 j
- Ad&wpIion cdl P. (Exp.)
0.0
0 200 400 em sm lorn 1200 14m 1-
[-I
Figure 2. Comparisonsof experinentai d~nensionless pressure histories and Me
predicad results by Brandani model at the case of & and 4adsorptions.

169
In the adsorption rates of all the components on the CMS,the diffusional time constants
showed strong dependence of pressure. Reid and Thomas (1 999) suggest that the reason
for these increases in the adsorption rate with surface coverage is probably related to the
surface diffusion. These exponential increases of the effective diffusional time constant
might be explained by above Darken-based models.
As shown in Figures 3 and 4, the large deviations between experimental and predicted
results by Darken-based models (Eqs. (6)-(9)) were observed in the experimental range of
pressure. At each case, the experimental diffusional time constant showed much stronger
pressure dependence than the results predicted by the Darken-based models.

0.008 , 0.0025
Darkenlangmuir N2 Ar
c1

y
0.004 .
0.003-
Darken-LF
Darken-Toth
Darken-DR
-'2 0.0020

0.0015
Darken-Langmuir
DarkEn-LF
Darken-Toth
0
a 0.0010 0

a O.wO5 0

0
0 o o
O.oo00

1 10
1 10
p 11-
P 1aMI
Figure 3.Presswedependena,Ot D/?fW N, adwrptlons Figwe 4. Pressuredependenceof W? for Ar adswptim
and the predictions by Damen-based madsls (Eqs. (6)-(9)) and the prediions by Darkekbasedmodels (Eqs. (6)-(9))

Do (1 996) proposed a structural diffusion model, and derived the following relation for
the surface diffusivity on the assumption that the gradient of the isotherm is large in the
initial stage.
A
Do-rela tion : D, =
dC, i d P

Here, A is a constant.
Above four isotherms were applied to Do-relation (Eq. (lo)), and the following
relations were derived.

*
Do-Langmuir: D, = D,, ( 1 + bP )'

DO-LF:
* (1 + b P y
Djl = Djlo I -n
-
~

P"
Do-Toth: * I+n

D,= Dfl ( 1 + (bP)")"

Do-DR D,=Dd
* (-P).expEa'(InP)'-2dInP-lnP,]
* (1 4)
In(P/P,)

170
In Figures 5 and 6, the prediction results by Do-based models (Eqs. (1 1)-(14)) were
shown for N2 and Ar. The similar prediction results were obtained at the other gas
adsorptions. From these results, it is concluded that Do-based models shows better
prediction results than Darken-based models.

0.006

0.0025

-
0.005 N, 303K-expt
0 Ar
-
y
L
0.004

0.003
0
- 0.0020

b 0.0015
Do-Langmuir
--- Do-LF
_ - OO-TOth
Do-DR
r 0.002 0 L 0
a % 0.0010
a O.OOO5
0.001
0 oooo
0.000
0.OoM)

I
1 10
1 10
p
P [am1
Figure 5. Pressuredependenceof Dlr' for N, adsorptions Figure6. Pressuredependenceof D/r' fw Ar adswptions
and the pmdidions by Do-basedmodels (Eqs. (11)-(14)) and the predidionsby Do-based models (Eqs. (11)-(14))

5 Conclusions

Adsorption kinetics was a hnction of atomic/molecular size, interaction (vertical or


lateral), atomic/molecular shape, polarity, pressure, and so forth. The
pressure-dependences of D/?could not be predicted by Darken-based models, but could
be well predicted by Do-based models.

6 Acknowledgements

The fmancial support of Korea Research Foundation (KRF-2001-005-E0031) is


gratefully acknowledged.

References

1. Ruthven, D.M., Diffusion of Oxygen and Nitrogen in Carbon Molecular Sieve, Chem.
Eng. Sci., 47,4305 (1992).
2. Brandani, S., Analysis of the Piezometric Method for the Study of Diffusion in
Microporous Solids: Isothermal Case, Adsorption, 4, 17 (1998).
3. Bae, Y.S.,* Kim K.I., and Lee C.H., Sorption Equilibrium and Kinetics in CMS by
Piezometric Method, Theory and Applications of Chemical Engineering, 7( l), 2422
(2001).
4. Do, D.D, A model for Surface Diffusion of Ethane and Propane in Activated Carbon,
Chem. Eng. Sci., 51,4145, (1996).
5 . Reid, C.R., and Thomas K.M., Adsorption of Gases on a Carbon Molecular Sieve
Used for Air Separation: Linear Adsorptives as Probes for Kinetic Selectivity, Langmuir,
15,3206 (1999).

171
PREPARATIVE ENANTIOSEPARATION OF FLUOXETINE
BY SIMULATED MOVING BED
H.W.YU AND C.B. CHING
Chemical &Environmental Engineering Department, National University of Singapore,
Singapore, 119260; Engp9001@nus.edu.sg

Based on the triangle method, a shortcut method was developed to establish the possible optimal
operating condition of a SMB unit because the theoretical optimal operating condition was diverted
from the actual complete separation region by too many unexpected disturbances. A new CSP was
used in the simulated moving bed for the enantioseparation of fluoxetine. Good separation results
were obtained. The effects of the difference between m2 and my on performance parameters were
studied. The results show that this method is useful for establishing the operating conditions. It can be
concluded that the new CSP is efficient for the enantioseparationof chiral drugs.

1 Introduction

When compared to the batchwise preparative chromatography, Simulated moving bed


(SMB) units exhibit a number of advantages. These advantages are primarily because of
the continuous nature of the operation and the efficient use of the stationary and mobile
phases, which allows a decrease in desorbent requirement and an improvement of the
productivity per unit time and per unit mass of stationary phase. In addition, high
performances can be achieved even at rather low values of selectivity and with a
relatively small number of theoretical plates. Due to these positive features, SMB is
particularly attractive in the case of enantiomer separations, since it is difficult to separate
enantiomers by conventional techniques. More recent applications related to chiral
technology were reported [l-31.
In this study, the enantioseparation of fluoxetine racemic mixture will be reported
using SMB technique. Fluoxetine is an antidepressant drug. (S)-fluoxetine is highly
desirable because it is a potent antidepressant and appetite suppressant and is free of
many undesirable side effects found with the racemic mixture. A high yield route to
enantiomerically pure (S)-fluoxetine, which could be used in the place of the racemic
mixture, may have commercial value.
The efficiency of chiral stationary phase (CSP) is crucial in chromatographic
technique. Recently, a new P-cyclodextrin phenyl isocyanate bonded chiral stationary
phase (CSP) was developed. This CSP is quite stable and can be used in most of HPLC
solvents. Many drug enantiomers that do not have enantioseparation effect on native f3-
cyclodextrin column in reversed phase were separated very well on this new CSP.
Design and optimization of operating conditions is very important in the operation
of SMB. Based on the equilibrium theory, triangle method was developed to design
operating conditions of SMB under non-linear condition in the past ten years [4-51. Due
to the uncertainty of the theoretical modeling parameters and the disturbance of operation,
the theoretical optimal operating condition is not robust in actual application using this
method. A shortcut method based on the triangle method is developed in this work to
establish possible robust optimal operating condition.

172
2 Experimental

2.1 Introduction of Q new CSP

This CSP was prepared with a pre-derived procedure. Perfunctionalised cyclodextrins


were first synthesized, purified and characterized and then chemically anchored on the
surface of aminized silica gel via the hydrolytically stable urethane linkage.

2.2 Chemicals
The eluent used was methanol and buffer (40:60). The concentration of triethylamine
acetate in buffer is 2% and the pH of buffer is 5.0 adjusted by addition of glacial acetic
acid. The feed was prepared by dissolving racemic fluoxetine in the mobile phase.
Fluoxetine was extracted from Prozace capsule, providbd by Lilly Company.

2.3 Instrument

The SMB system consists of 8 columns (250mmx1Omm) packed by the new CSP. The
columns are fed with either the feed or the eluent (inlet and outlet) via 5 or 8 port rotary
valves (VICI). The configuration tested is 2222 (two columns per zone. The
concentrations of the extract and the raffinate streams were analyzed for each stage using
a standard analytical chromatographic system. An analytical column (250mmx4.6mm)
packed by 5pm CSP was used to analyze the concentration of these samples. The
adsorbance wavelength was set at 225nm.

2.4 Adsorption Isotherm

In the study, Langmuir model was assumed to fit the adsorption equilibrium relationship
of fluoxetine on the column packed by the new CSP and the following approximated
extended Langmuir model was used to describe the adsorption behavior:

- 5.94cA
- 1 + 0 . 7 1 5 ~+ ~0 . 2 1 4 ~ ~
5.18~~
qB = 1 + 0 . 7 1 5 ~+~0 . 2 1 4 ~ ~

where A is (S)-fluoxetine and B is (R)-fluoxetine.

3 Results and Discussions

Now, let us design the operating conditions following the above process. Several runs
were carried out and the position of every operating point on the m2-1~13plane was shown
in Figure 1. Every sample was collected after the system was stable.
Operating point S is the theoretical optimal operating condition derived from
triangle method [5]. The purity of the extract and the rafinate at this experimental point is
83.4% and 99.8%, respectively. Experimental results show that the theoretical optimal
operation condition is not robust or distorts from actual complete separation region;
Points A and C are close to the actual complete separation region and point B is in the
actual complete separation region, When PACand PBAare extended to intersect at WO,a
triangle W ~ A PisB formed. This is assumed to be actual complete separation region. It
must be mentioned that A and C could be in the theoretical complete separation region
after considering the operational disturbance [6].

4) 41 6d 6.6 u u

mt

Figure 1. Schematic diagram on establishingoptimal robust operating condition

174
To improve the robustness of the operation, the difference between m 3 and m2 for
point B is decreased to point H.Due to the increase in eluent flow-rate, the pressure of the
system is increased to an extent that the system cannot tolerate. Thus, the stability of the
operation is damaged, and good separation results are not achieved. This means that
although decreasing the difference between m3 and mz could improve the robustness of
the operation, the operation performance parameters will be sacrificed, sometimes even
the SMB system cannot be run smoothly. Unstable operation should be avoided in real
applications. Runs I, J and K are carried out to find the possible optimal operating
condition by increasing the difference between m3 and m2. At point K, the robustness of
the operation cannot hinder the disturbance of the operation, so the purity of the extract
and the raffinate are all decreased. Enrichment of the extract and the raffinate is increased
when the difference between m3 and m2 is increased. In other words, the solvent
consumption is decreased. Because of the decreasing of the robustness, these three
operation parameters all decreased slightly. With reference to the operation parameters of
B, I, J, K, operating point J can be considered to be the possible optimal operation point.

4 Conclusions

In this work, a new CSP was developed to separate a chiral drug using the
chromatographic technique, and a pragmatic method was developed to establish the
operating conditions of the SMB system based on the theoretical triangle method. A
triangle region determined from the experimental results was assumed to be an actual
complete separation region. In this region, possible an optimal operating condition of the
SMB was found. Good separation results can be obtained. It is shown that this method is
available to establish operating conditions and reduces the experimental efforts
effectively. It can be concluded that the new CSP is efficient for enantioseparation of
chiral drugs.

5 Acknowledgements

Authors express our appreciation to National University of Singapore in financial support


of this project.

175
References

1. Kiister E.,Gerber G. and Antia F.G., Enantioseparation of A Chiral Epoxide by


Simulated Moving Bed Chromatography Using Chiralcel-OD,
Chromatographia,40, (1995) 718, pp.387-393.
2. Cavoy E., Deltent M.-F., Lehoucq S. and Miggiano D., Laboratory-
developed Simulated Moving Bed for Chiral Drug Separationsand Design of the
System and Separation of Tramadol Enantiomers, J. Chromatogr. A 769, (1 997)
pp.49-57.
3. Strube J., Altenhoner U., Meurer M., Schmidt-Traub, H. and Schulte M.,
Dynamic Simulation of Simulated Moving-bed Chromatographic Processes for
the Optimization of Chiral Separations, J. Chromafogr.A, 769, (1997) pp.8 1-92.
4. Storti G., Mazzotti M., Morbidelli M. and Carra s., Robust Design of Binary
Countercurrent Adsorption Separation Processes, AZCHE J., 39(3), ( 1993) pp.
471-492.
5. Mazzotti M., Storti G. and Morbidelli M., Optimal Operation of Simulated
Moving Bed Units for Nonlinear Chromatographic Separations,J. Chromatogr.
A 769, (1 997) pp.3-24.
6. Yu H.W.and Ching C.B., Design and Optimization of a Simulated Moving Bed
Based on an Approximated Model Using a Novel Shortcut Method, AICHE
Journal vol. 48, (2002) No.10.2240-2246.

176
OPTIMIZATION BASED ADAPTIVE CONTROL OF SIMULATED MOVING
BEDS

S. ABEL AND M . MAZZOTTI


ETH Zurich, Institute of Process Engineering, Sonneggstrasse 3, 8092 Zurich, Switzerland
E-mail: abel@ivuk mavt.ethz.ch; mazzotti@,ivukmavt.ethz.ch

G. ERDEM AND M. M O W
ETH Zurich, Automatic Control Laboratory, Physikstrasse 3, 8092 Zurich, Switzerland
E-mail: erdem@aut.ee.ethz.ch; morariamt.ee.ethz.ch

M.MORBIDELLI
ETH Zurich, Laboratory of Chemical Engineering and Industrial Chemistry,8093 Zurich,
Switzerland
E-mail: morbidelli@.tech.chem.etk.ch

In the recent years Simulated Moving Bed (SMB) technology has become more and more attractive for
complex separation tasks. To ensure the compliance with product specifications, a robust control is
required. In this work a new optimization based adaptive control strategy for the SMB is proposed: A
linearized reduced order model, which accounts for the periodic nature of the SMB process is used for
online optimization and control purposes. Concentration measurements at the raffinate and extract
outlets are used as the feedback information together with a periodic Kalman filter to remove model
emors and to handle disturbances. The state estimate from the periodic Kalman filter is then used for
the prediction of the outlet concentrations over a predefined time horizon. Predicted outlet
concentrations constitute the basis for the calculation of the optimal input adjustments, which
maximize the productivity and minimize the desorbent consumption subject to constraints on product
purities.

1 Introduction

SMB technology is based on the concept of a continuous chromatographic countercurrent


process. The real countercurrent is called the True Moving Bed (TMB), which is technically
not feasible without damaging the solid adsorbent particles. SMB technology overcomes
this problem by using fixed bed columns and simulating the solid flow by periodically
switchingthe inlet and outlet ports of the unit in the same direction as the fluid flow. Due to
this periodic process dynamicsthe S M B never reaches a real equilibrium, but a cyclic steady
state with a time dependent concentration profile inside the unit, which challenges the
control strategy. The goal of this work is to develop an automatic control for SMBs, which
can both find the optimal operating conditions for the plant on-line and run it at these
conditions despite possible disturbances or any kind of unexpected deviations of the
predicted behaviour. This will make it possible to exploit the full economical potential of
SMB technology.

The challenges in SMB control are not only the complex nonlinear process dynamics, but
also the long delays of the effect of disturbances. The required control strategy has to be able
to handle multivariable dynamics with time-delays and hard constraints. Model Predictive
Control (MPC) has been proven to be the most effective control strategy for this type of
problems [1,2]. Only recently, a few scientific publications have addressed the automatic

177
control of SMB by adopting different strategies [3,4]. In this work we will follow a new
model-based predictive control method that was introduced by Lee and Natarjan [5,6]. This
is called Repetitive Model Predictive Control (RMPC), and combines the concepts of both
MPC and Repetitive Control (RC).

2 Control concept

The suggested control strategy uses a simplified SMB model in order to optimize the
operating conditions of the plant on-line. The basic concept of the adopted approach is
illustrated in figure 1.
Objective FuncUon
+ Product Specifications

SMB Plant
puw,
Internal Flowrates Productivity,
_____*
Solvent
Consumption
~

I
I I
Concentration
Measurements

Figure 1. Scheme of the OptimizingControl concept.


The controller receives the on-line composition measurementof the product outlets (extract
and rafinate) as feedback data from the plant. These measurements are filtered through a
periodic Kalman filter and used together with the simplified SMB model results to estimate
the state of the system and to remove the possible model errors. The formulation of RMPC
is based on the assumption that possible errors or disturbances are likely to repeat and will
have a periodic effect on the output, which is the most likely correlation between
disturbances and output in a SMB unit. The estimated future concentration profile in the
SMB is used to optimize the future behaviour of the plant over a predefined prediction
horizon. The controller implements the calculated optimal plant input by changing the
external flow rates in order to control the internal flow rates, which are the manipulated
variables. Time lags, e.g. between online concentration measurements and optimizer or
between optimizer and SMB plant, are insignificant relative to the process dynamics and
sampling time for the planned scheme.

2. I Simplified SMB model


It is not possible to apply a detailed SMB model for on-line use together with the controller
without facing severe computational problems. Hence a simplified model is needed, which

178
is simple enough for efficient computation, but still captures the characteristic process
dynamics of an SMB. This model is then used, according to the control scheme shown
above, to estimate the internal concentration profiles and to calculate optimal operating
conditions as input for the plant. To obtain the simplified model from the equilibrium
dispersive SMB model, the model equations are linearized around a compositionprofile (in
space) calculated at cyclic steady state. Since at cyclic steady state concentration fronts
propagate along the columns and the composition profile (in space) also changes, different
composition profiles are used to linearize the model equations at different points in time
during one SMB cycle. It is worth noting here that one cycle corresponds to a number of
time periods between inlet-outlet switches equal to the number of columns in the SMB unit.
Finally the model order is reduced by using balanced model reduction as proposed by Lee
161. The obtained simplified model is then used to predict and to optimize the future
behaviour of the SMB plant.

2.2 Optimizationproblem
The idea of the control concept introduced in this work is to optimize the process over a
predefined prediction horizon, which has been chosen as 2 cycles. The control horizon in
which possible flow rate changes are calculated, is 1 cycle, though. The optimization
problem of a SMB unit with a constant switch time can be formulated as maximizing the
feed flow rate and minimizing the desorbent consumption (cost function). In practice,
economical considerations or others will guide the choice of the relative importance
attributed to these two optimization goals. Of course, this problem is subjected to certain
constraints. First of all the purity requirements in the extract and rafkinate stream has to be
fulfilled. In addition to this the system has a number of physical limitations such as
non-negative flow rates. The cost function together with the defmed constraints constitute a
Linear Program (LP) to be solved at each time step based on the new measurements
available. Calculated optimal internal flow rates, which fi~lfillthe constraints and optimize
the performance of the SMB unit, are implemented on the unit via changes of the external
flow rates.

3 Performance assessment of the controller

In order to evaluate the performance of the controller, various scenarios have been
simulated on a virtual platform. This means that instead of a real plant a SMB model based
on the equilibrium dispersive model is used [7]. In the following an example is given to
show the flexibility and performance of the controller.

3. I VirtualSMB system
An 8 column SMB with a 2-2-2-2closed-loop configurationis considered. Available for the
controller are the on-line concentration measurements at the two outlets of extract and
raffinate. The substances to be separated are A and B, where A adsorbs stronger than B.
Linear adsorptionbehaviour has been assumed (qi=Hiq). The Henry constants Hi have been
chosen as HA=4and H B = ~The . controller is able to adjust the external flow rates in order to
change the internal flow rates in the four sections of the SMB, which are the manipulated
variables. Since all 4 internal flow rates cannot be directly controlled via the external flow
rates, one internal flow rate has to be directly controlled by an additional pump, which

179
controls e.g. the recycle flow. The switch time is kept constant during the process.

3.2 Simulation result


To assess the performance of the developed controller the evolution of the purity over time,
in comparison to the uncontrolled case, is monitored. A minimum product purity of 99% is
given as product specifications. The controller is switched on after the plant has reached the
cyclic steady-state at the chosen start-up operating point, which is the operating point used
for linearization (ml=4.0, m2=2.1, m3=3.9,m4=2.1; see [7]for the definitions) and leads to
poor product purities. The chosen example shows the case of a sudden, rather large step
change in Henry constants. The controller is switchedon at cycle 20. Then it brings the plant
to fulfill specificationsand the plant is run at its optimal operatingconditions. A step change
in the characteristicadsorption parameters occurs at cycle 60, when HAand HBincrease by
10% and 15%, respectively. It can be observed, that the purity of both extract and raffmate
outlet decreases down to 85% for the uncontrolled plant. On the contrary the controller
reacts and recovers the purities within 5 to 8 cycles, then driving the plant to the new
optimum operating conditions.

0.91

0.86

--- extract
O.8ZL I
10 20 30 40 50 60 70 80 90 100 110
Cycles

Figure 2. Comparison of the SMB outlet purities in the controlled and uncontrolled case. The controller is
switched on after reaching steady state and a step disturbance in Henry's constants takes place at cycle 60.
(AHA+~OYO,A Hs=+15%).

4 Conclusion

An optimizing adaptive control strategy for the automatic control of the SMB process has

180
been proposed. It enables to operate the SMB unit at its optimal conditions. The developed
controller has been tested on a virtual platform where a linear adsorption isotherm was used.
It has been shown that the designed controller fulfills the process specificationsand at the
same time it optimizes the productivity and solvent consumption. It can adapt the operating
conditions to the changes caused by disturbances (e.g. caused by temperature changes),
operate under extreme model mismatch conditions and cope with irregularities andor aging
of the chromatographic system. One of the most important features of the controller is its
ability to find the real optimum of the process and adapt the operating conditions if
necessary.

5 Acknowledgements

The authors are gratefiil to Prof. Jay H. Lee and Prof. Hyun-Kun Rhee for helphl
discussions. The support of ETH Zurich through grant TH-23YOO-1 is gratefully
acknowledged.

References

1. Garcia C.E., Prett D.M. and Morari M., Model Predictive Contol - Theory and Practice
- A survey, Automatica 25 (1989) pp. 335-348.
2. Bemporad A. and Morari M., Model Predictive Contol: A survey, Lecture Notes in
Control and Information Sciences 245 (1999) pp. 207-226.
3. Kloppenburg E. and Gilles E.D., Automatic control of the simulated moving bed
process for C8 aromatics separation using asymptotically exact
input/output-linearization.A of Process Control 9 (2000) pp. 4 1-50.
4. Klatt K.U., Hanisch F. and Dlinnebier G., Model-based control of a simulated moving
bed chromatographicprocess for the separation of fructose and glucose. J. of Process
Control 12 (2002) pp. 203-219.
5. Natarajan S. and Lee J.H., Repetitive model predictive control applied to a simulated
moving bed chromatography system, Computers and Chem. Eng. 24 (2000) pp.
1127-1133.
6. Lee J.H., Natarajan S. and Lee K.S., A model-based predictive control approach to
repetitive control of continuous processes with periodic operations. Journal of Process
Control 11 (2001) pp. 195-207.
7. Migliorini C., Gentilini A., Mazzotti M. and Morbidelli M., Design of simulated
moving bed units under non-ideal conditions, Ind. Eng. Chem. Res. 38 (1999) pp.
2400-24 10.

181
MONO-METHYL PARAFFIN ADSORPTIVE SEPARATION PROCESS

SANTI KULPRATHIPANJA, JAMES REKOSKE, MICHAEL GATTER, AND STEPHEN SOHN


UOP LLC, 25 E. Algonquin Rd, Des Plaines IL 6001 7, USA

A process for the recovery and purification of mono-methyl branched C~IJ- CISparaffins From
kerosene or n-paraffin depleted kerosene through an adsorptive separation has boen developed. Two
complimentary adsorptive separation-operatingmodes were demonstrated for efficient mono-methyl
paraffin production. The first method uses simulated moving-bed adsorptive separation with a
silicalite adsorbent and a mixture of CSand/or C6 n-paraffin and is0 and/or cyclo-parafin as the
desorbent. The second method, which provides enhanced mono-methyl paraffin recovery, uses the
pre-pulse technique in combination with the first operating method. Pulse tests (chromatographic
separation) with both techniques showed good performance in the laboratory. Both methods were
demonstrated successfi~llyin a continuous counter-current chromatographic separation pilot plant
using commercial n - p d i n depleted kerosene feedstock. Better than 90% normal and
mono-methyl paraffin purity with greater than 70% recovery of normal and mono-methyl paraffins
was achieved from the pilot plant. Detailed information on the process development including feed
composition, desorbent composition, temperature, diffusion rates, and model simulations will be
discussed.

1. Introduction

Surfactants are the most critical part of a detergent formulation. Surfactants with enhanced
water solubility relative to existing commercial surfactants are highly desirable for the
formulation of improved household detergents and cleaning products. Many existing
commercial detergents are composed of anionic surfactants derived from linear alkyl
benzene. It has been demonstrated that surfactants that are formulated from benzene
alkylated with mono-methyl parafin intermediates have a greater water solubility than
those formulated fiom benzene alkylated n-paraffin intermediates. To create these
enhanced water-soluble surfactants in commercially significant quantities, a raw material
source for Clo to CI3 mono-methyl parafin is required. This range of mono-methyl
paraffm is typically found in kerosene and n-paraffin depleted kerosene. N-Paraffin
depleted kerosene (Molex Raffinate) is a by-product obtained fiom the UOP Molex
process and is typically used in Jet Fuel Pool blending. Both kerosene and n-paraffin
depleted kerosene are composed of a mixture of n-paraffin, mono-methyl parafin,
di-/tri-branched paraffin, naphthenes, and aromatics. Previously, no commercial process
was available to separate mono-methyl paraffin from kerosene or n-paraffin depleted
kerosene. UOP has developed an adsorptive separation process for mono-methyl
parafin. This process utilizes the simulated moving bed system and is called MMP
SorbexTM.

2. MMP Sorbexm Development

2. I Silicalite Adsorbent
Early in the development of the process, we determined that silicalite, a MFI structure
type adsorbent, is suitable for the process. Silicalite is crystalline silica, which has a
novel topologic type of tetrahedral h e w o r k similar to aluminosilicate molecular sieves.

182
Oxygen rings containing 10-members define the straight channels along the b-axis with an
elliptical cross session of 5.7-5.8 Angstrom by 5.1-5.2Angstrom. These channels are
interconnected by zigzag channels along the a-axis defined also by 10-membered oxygen
rings with a nearly circular cross section of 5.4 Angstrom. Based on the pore openings
above, one would expect n-paraffin, mono-methyl paraffi, and low molecular weight
aromatics and naphthenes enter into the channels of silicalite.

2.2 Molecular Simulation


To confirm the silicalite pore fitting predictions above, diffusion simulations was carried
out using the Solids Diffusion module in the Accelrys Insight I1 MoIecular Modeling
package. (Accelrysis a wholly owned subsidiary of Pharmacopeia Inc., headquartered in
San Diego, CA.)

Table 1. Energetics and Predictions from Modeling Molecular Diffusion in Silicalite.

1-butyl-3-propylcyclopentane 14 -13 slow


2,ddimethyloctane 15 -17 slow
n-hexylcyclohexane 27 -15 very slow

3,3,5-trimethylheptane 45 +35 excluded


1,2,3-trimethybenzene 80 +50 excluded
2-ethyl-1,3-dimethylbenzene 80 +50 excluded
trans-l,2-di~ro~ylcvclohexane 125 +loo excluded
150 +I07 excluded I
The simulated diffusion model involves advancing an probe molecule through the
zeolites main channel and using molecular mechanics methods to calculate the energy of
the probe molecule at regular intervals as it interacts with the atoms in the sieve wall.
The energy difference between the low and high energy positions of the probe molecule in
the channel is defined as the energy barrier which can be used to predict the diffusion ease
of the probe molecule in the zeolite channel - the larger the barrier, the more restrictive
the diffusion. Another parameter that assists in predicting diffisivity is the sign and
magnitude of the maximum energy (Emax),which is a measure of how much the probe
molecule is stabilized (negative values) or destabilized (positive values) at the bottleneck
positions in the zeolite channel.
.The modeling results fiom simulating the diffision of pertinent molecules in silicalite
are summarized in Table 1. Note that the predicted diffusivity of the molecules, based
on the magnitude of the energy barrier, agrees with experimental data (see pulse test
session). The n-decane and 2-methylnonane, with energy barriers less than 10 kcal, have
good diffisivity whereas 2,ddimethyloctane and small naphthenes, with energy barriers

183
between 10 and 30 kcals, have limited difisivity in silicalite. Tri-substituted
paraffins/aromaticsand the bulkier naphthenes, which have energy barriers greater than 40
kcal and positiveLF,,values, are found to be too large to fit in silicalite. These results
suggests that, to a great extent, normal and mono-methyl paraffins will be extracted by
applying silicalite in an adsorptive separation, while naphthenes and aromatics will be
largely rejected.

2.3. Pulse Test


Pulse test is a laboratory technique used to evaluate the suitability of adsorbent and
desorbent for adsorptive separation process. This is a means to optimizing the
aforementioned factors. The pulse test procedure (1) begins with an injection of
feedstock into a desorbent stream that is flowing through a packed adsorbent column at
constant flow rate and temperature. To determine the on-stream column effluent as a
function of time or volume of desorbent passed, each feed component is monitored as it
emerges from the column using gas chromatographyor liquid chromatography.
For mono-methyl paraffin separation, two pulse test techniques, one with and one
without iso-octane pre-pulse, were developed (2,3). In each test the feed was a mixture
containing equal volumes of 3,3,5-trimethyl heptane, 2,ddimethyl octane, 2-methyl
nonane, n-decane, and 1,3,5-trimethyl benzene. The pulse test column had a volume of
70 cc and was held at a temperature of 12OOC in the experiments shown. The flow rate
through the column was 1.2 ml/min. The adsorbent was silicalite and the desorbent was
a 50/50 volume % mixture of n-hexane/cyclohexane. Test 1 was run without a pre-pulse
and test 2 was run with a pre-pulse of 40 ml of iso-octane injected into the test loop
immediately before the feed mixture was injected. Iso-octane pre-pulse diluted the
n-hexane concentration at the adsorption zone and increased the adsorbent selectivity for
mono-methyl paraffin.

A graphical representation of the results of this comparison test run is shown in Figures 1
and 2. Figure 1 shows a plot of the relative concentrations of the components versus
volume of effluent. The improved separation using the pre-pulse is also shows
graphically in Figure 2. A comparison of the two plots shows that use of a pre-pulse
resulted in a much better separation of mono-methyl paraffin (2-methyl nonane) from the
balance of feed components.

184
30
20
10
n
"20 30 40 50- 60 70 80 90 100 110 120
Retention Volume (mi)
Figure 1. Chromatographicseparation (Pulse Test) of mono-methylparaffin using Silicalite
adsorbent, 50150 n-43 /cyclohexane desorbent

_r.
80
70 -
60- 3,3,5-TM-C,

50
- I2 ,QDM-C,

Retention Volume (ml)

Figure 2. Chromatographicseparation (Pulse Test) of mono-methylparatlin using silicalite adsorbent, 50150


n-Cb /cycIohexane desorbent and with iso-octane
pre pulse.

A third pulse test with pre-pulse of iso-octane was carried out using a feed representative
of that which would be processed in a commercial operation, n-parafin depleted kerosene
(Molex Raffinate). The feed contains mono-, di-, and tri-branched paraffins, naphthenes,
-
aromatics, and a trace (2 4 wt.%) of n-paraffins. Figure 3 shows a gas chromatogramof
the Molex Raffinate. As one can see, the feed chromatogram is rather complex and
contains at least several hundred compounds. The high number of compounds and its
complexity prevent the identification of individual compounds. The effectiveness of the
separation was therefore determined by collecting fractions of the pulse test effluent and
analyzing the normal and mono-methyl paraffin content of each fraction. Figure 4 shows
the pulse test of the relative concentration of all components present in the feed versus
effluent volume. The separation performance obtained from the Molex Rafinate feed is

185
similar the test 2 above. The highly branched paraffins, naphthenes, and aromatics were
rejected and the desired mono-methyl paraffins were extracted. The gas chromatogram -
mass spectrophotometry of the extracted mono-methyl paraffins is shown in Figure 5.
As expected from the diffusion simulations, trace quantities of naphthenes and other small
molecular compounds are present in the extracted product along with the desired product,
normal and mono-methyl paraffins.

Time I_,

Figure 3. Gas Chromatogram of N-Paraffin depleted kerosene (UOP Molex Railinate)

Retention Volume

Figure 4. Chromatographic separation of mono-methylparaffin from N-Parafin depleted Kerosene


using silicalite adsorbent, 50/50 nC6 kyclohexane desorbent, and iso-octane pre-pulse

186
27 29 31 33 35 37 39 41

Time L

-
Figure 5. Gas Chromatogram Mass Spectrophotometryof extract product from Pulse Test

2.4. Sorbex Pilot Plant Demonstration


The process was demonstrated in a simulated continuous counter-current chromatographic
separation pilot plant. Both the primary method of operation and the pre-pulse technique
were demonstrated, with the pre-pulse technique showing improved recovery. Using
commercial n-paraffin depleted kerosene (Molex Raffinate) feedstock we routinely
demonstrated the ability to achieve better than 90% mono-methyl and normal paraffin
purity with greater than 70% recovery of mono-methyl paraffins.

3 Conclusion

Two adsorption techniques have been developed for the separation of mono-methyl
paraffins from kerosene or n-paraffin depleted kerosene using silicalite adsorbent. The
first one is using a mixture of C5and/or C6 n-paraffin and cyclo- and or iso-paraffin
desorbent. The second one is an enhanced mono-methyl parafin recovering process using
the pre-pulse technique in combination with the previous one. Pulse tests
(chromatographic separation) with both techniques showed good performance in the
laboratory. The process was demonstrated successfully in a simulated continuous
counter-current chromatographic separation pilot plant using commercial n-parafin
depieted kerosene feedstock. Better than 90% mono-methyl and normal paraffin purity
with greater than 70% recovery of mono-methyl paraffins was achieved from the pilot

187
plant. The simulated diffusion model involves advancing an organic molecule through
the zeolites main channel was investigated. Results confirmed the experimental
findings.
References:
1. R.W.Neuzil, U.S. Patent 5,382,747, 1995
2. Santi Kulprathipanja,Monomethyl paraffin adsorptive separation process, U.S.A.
Patent 6,222,088 B 1,200 1
3. Santi Kulprathipanja, Process for monomethyl acyclic hydrocarbon adsorptive
separation, U.S.A. Patent 6,252,127 B1,2001
4. Santi Kulprathipanja and James Johnson Liquid Separation,Handbook of Porous
Solids, Edited by F. Schuth, K. Sing, J. Weitkamp, To be Published by Wiley-VCH,
Germany in 2002

188
CHROMIUM (VI) AND (111) SPECIES ADSORPTION FROM AQUEOUS
SOLUTIONS BY ACTIVATED CARBON FIBERS

0. ASTACHKINA, A. LYSSENKO, 0. MUHINA


Saint-Petersburg State University of Technology and Design,
St. B. Morskaya, 18,Saint-Petersbwg, Russia, I91186
E-mail: LYSSENKO@UTD.RU

Prepared from rayon several activated carbon fiben (ACF) with different pore structure were used to remove
Cr(VI) and/or Cr(1Il) species from solutions. The adsorption experiments were carried out to determine the
influence of ACFlsolution contact time, pH, temperature, initial Cr(V1) and Cr(l1l) concentration on the
efficiency of chromium ions removal by ACF. It was found that for ACF with total pore volume more than
O.bm-/gthe porous texture has no great importance for the amount of chromium retained. For all non-oxidized
ACF the amounts of Cr species adsorbed and Cr(V1) reduced to Cr(l1l) after 48 h of ACF/solution contact are
very close. At the beginning phase of ACF/solution contact the latent period of Cr(VI) to Cr(I1l) reduction
was observed. Oxidized ACF has lower adsorption capacity to Cr(V1) species and higher to Cr(ll1) ions in
respect to non-oxidized ACF. The increase of initial Cr(V1) concentration increases the chromium species
removal but the increase of pH and temperature decreases it.
Key words - adsorption from solution, activated carbon fibers, chromium removal.

1 Introduction

Chromium compounds are widely used in many industries: metal finishing and
electroplating, leather tanning, pigments manufacturing, photography and catalysts
production [I]. The presence of chromium species in wastewater of all of these industries
is a problem because of the affect onto the human physiology. Chromium removal from
wastewaters by adsorption onto activated charcoals is an important process in the
environmental protection [2]. On the other hand the chromium species adsorption from
aqueous solution is one of the processes for chromium catalysts supported on activated
carbons production [3].
Usually the activated carbon granules are employed for chromium removal or for
catalysts preparation. However in some previous reports the perspectives and advantages
of activated carbon fibers (ACF) utilization for the same employment have been
documented [2,4]. It seems from the analysis of the articles that the use of ACF or
activated carbon cloth has a great potential.
In the present work the adsorption of Cr(V1) and Cr(II1) on ACF with different pore
structure as well as on the oxidized form of ACF was studied.

2 Materials and methods

The ACF were characterized by N2 adsorption at 77 K, methylene blue (MB), I2 and


benzene adsorption at 25 C.
The table 1 includes some preparation conditions and some properties of ACF.
Sample ACF-2 was oxidized with HN03 to obtain sample ACF-2ox. The Cr(V1) and
Cr(II1) solutions were prepared from K2Cr207and Cr(N03)3respectively. The retention of
chromium was determined by spectrophotometer method [5].

189
For adsorption 0.1 g of ACF (previously washed up to pH 7.0 and dried) were
added to 100 ml of solution. The experimental variables were: time of contact between
ACF and solutions, initial concentration of chromium ions, pH, and temperature.
Table 1. Preparation conditions and some properties of ACF

SbN2 - specific surface area, measured by the N2 adsorption.


W,- volume adsorbed at a relative pressure - 0.95, corresponds to the (microporetmesopore)volume
Vd - microporevolume
,V - mesopore volume
M B a - Methylene Blue adsorption
12 - 12 adsorption

3 Results and discussion

3.1 Kinetics of Cr (VI) removal


To obtain the experimental data the suspensions of ACF in the solutions of chromium(V1)
with the initial concentrations 200 mg/l were shaken at 22-25C. As the reduction of
Cr(V1) to Cr(II1) was shown in some papers [6] in our investigation the variation of both
chromium ions concentration during adsorption was determined.
Table 2 presents the abatement of Cr(V1) concentration and the rate of all
chromium species adsorptions to wit - the adsorption capacity (AC) of different ACF,
when fig.l show the growth of Cr(II1) concentration. In all cases the initial pH was 4.7,
the final - 6.4. The experiment carried out showed that for ACF- 1, ACF-2, ACF-3 pore
structure and the value of surface area has no great importance.
On the other hand in the case of ACF-4 when the total pore volume is around 0.3
cm3/gand Sh,=500 - 600 m2/g during the initial phase of adsorption (time less then 2 h)
the AC decreases twice. At the same time for all non oxidized ACF the final (reached
after 48 h) amounts of chromium species adsorbed and Cr (VI) reduced to Cr(l11) are very
close. Oxidized sample ACF-~OX, possessing the similar texture as ACF-3, has notably
lower adsorption and reduction capacity in respect to non-oxidized sorbents.
It is important to mention that at the beginning phase of ACF contact with
solutions the latent period of Cr(IV) to Cr(1II) reduction can be distinguished (fig. I).
This phenomena becomes more visible in the range ACF-4>ACF-3 >ACF-2 >ACF-1 and
can be attributed to the self-catalicale reduction-oxidation reactions between carbon and
Cr(V1) species and/or oxygen-containing groups on carbon (-OH, -COH) and Cr(V1) ions.
The similar effect was described in our previous paper for the platinum and gold species
adsorption on ACF [7].
The data presented in table 3 and fig2 show the abatement of Cr(V1) concentration,
AC of ACF and the growth of Cr(II1) concentration at pH 2.5-2.9. A large increase in the
amount of Cr(V1) transformed to Cr(II1) and in the amount of chromium removed is
produced as the pH decreases. Furthermore the amount of Cr(1ll) (fig.2) rises up to

190
maximum in the initial phase of ACF/solution contact (time up to 2 h) and than falls to
the minimum (time up to 48 h). It indicates that the Cr(II1) species could be adsorbed onto
ACF when the ACF are oxidized during the adsorptiodreduction of Cr(V1) ions. The
presence of oxidized form of ACF is expected to facilitate the Cr(1II) elimination but they
are not effective for Cr(V1) removal.
The ACF-2ox/Cr(VI) solution contact confirm low capacity of oxidized fibers for
chromium (VI) ions adsorption.
Table 2. chromium (vi) species adsorption at initial pH - 4.7 as a function of time.

A. Cr(V1) concentration, mg/l


B. Total mount of chromium (VI) and (Ill) removed - AC, mg/g

r
X a

0 0,s 1 1.5 2 24 48
a h
-
Figure 1. Chromium (111) concentrationvariation as a function of time. Initial pH 4.7; temperature 20 "C.
Table 3. chromium (vi) species adsorption at initial PH - 2.5 as a function time.

A. Cr (VI).concent&on, mg/l .
B. Total mount of chromium (VI) and (Ill) removed - AC, mg/g

0
A
m
x ax
X X

0 0.5 I 1.5 2 24 48
81me. h
-
Figure 2. Chromium (III) concentrationvariation as a function of time. Initial pH 2.5; temperature 20 'C.

191
3.2 Kinetics of Cr(II4 removal
As from the previous experiments it was impossible to determine the adsorption capacity
in regard only to Cr(1II) ions theirs adsorption onto different ACF was examined from
Cr@IO3), solution with initial concentration of Cr(II1) - 100 mg/l. Fig.3 shows the
amount of Cr(II1) species removed at pH 3.2 - 3.9 temperature 20-25 "C and at different
ACF/solution contact time.

tACF-4
+ACF-3
t ACF-2
JC-ACF-1
Jlt ACF-ZOX

0 1 2 3 4 5
time, h

Figure 3. Chromium (111) removal by ACF at 20 "C as a hnction of time.

According to the experiments carried out the maximum adsorption capacity to Cr(lI1)
ions can be affirmed for oxidized sample ACF-2ox. The Cr(lI1) removal by non-oxidized
ACF is more than twice less. The adsorption equilibrium for all ACF was reached in a
period of time close to 3 h.

3.3 The influence of temperature on Cr(VI) species adsorption


Figure 4 compares the amount of total chromium removed by ACF-1 at different
temperatures for saturation time 48 h, pH=2 and initial concentration of Cr(V1) in solution
200 mg/l.

0 # ) 4 0 6 0 8 0
T
Figure 4. Chromium (VI) removal by ACF-I as a hnction of temperature (T, "C).

The inverse relationship between the amount of chromium ions removed and the
temperature was found. This fact can be explained in the consideration that the amount of
Cr(V1) reduced to Cr(II1) increases with the increase of the temperature and that the
adsorption of Cr(II1) ions on ACF-1 is relatively low. So, more the Cr(lI1) ions appears in
the solution during ACF and Cr(V1) species contact less the total amount of chromium
removed.
Additionally it can be mentioned that the three states of chromium (Cr20;2, Cr203,
C i 3 ) on surface of ACF were detected by X-ray photoelectron spectroscopy. According

192
to the data obtained the direct adsorption of Cr(V1) and Cr(Il1) species (if they are both in
the solution) as well as Cr(V1) reduction to Cr203 and to Cr (111) ion with the subsequent
adsorption can be supposed.

3.4 The influence of concentration


The relationship between the initial concentration of Cr (VI) species in the solutions and
adsorption capacity of ACF-1, ACF-2 and ACF-3 at pH 2.5, temperature 20 'C, after 48 h
is presented in figure 5.

2 0 200 400 600 800 lo00


initial concentration,mgh

Figure 5. Chromium (VI) adsorptionon ACF-las a function of initial concentration

The amount of chromium increases with the increase of initial concentration of


Cr(V1). The adsorption capacity of 800 mg/g can be reached when the initial Cr(V1)
concentration is 1000mg/l. The adsorption capacity of oxidized sample ACF-2ox at the
same conditions was not more than 310 mg/g. The amount of Cr(II1) removed by ACF-
20x from the solution with Cr(II1) initial concentration 1000 mg/l was only 140 mg/g.
Thus, the factors which affect to Cr(V1) and Cr(II1) removal by ACF are not only
temperature and initial concentration but also the oxidationheduction ability of fibers and
theirs oxidation state before and during adsorption. The ACF used in the study can be
useful to remove Cr(V1) and Cr(II1) from aqueous solution. To ameliorate the rate of
chromium removal it seems better to use at the same time the oxidized and non-oxidized
ACF. The most usefid porous texture of ACF is when the total pore volume is more than
0.4 and less 0.5 cm3/gand contains 70-80 % of micropores.

References
1 . Lalvani S.B., Wiltowsk T., Hubner A., Weston A., Mandich N., Carbon 36 (1998) p.
1219.
2. Morozova A. A., Zh. Prikl. Khimii 68 (1995) p. 770.
3. Brown P.N., Jayson G.G., Thomson G., Wilkinson M.C., Carbon 27 (1989) p. 821.
4. Lyssenko A., Carbon-based media for water purification, The international magazine
for technical textile users 38 N24 (2000) p. 33.
5. Chand S., Agarwal V.K., Kumar P., Indian J., Enviror HLTH 36 Ne3 (1994) p. 15 1.
6. Aggarwal D., Goyal M.,Bansal R.C., Carbon 37 (1 999) p. 1989.
7. Lyssenko A., Simanova S., Abstracts, At 7th International conference of
Fundamentals of Adsorption, Nagasaki, Japan, May (200 1) p. 38.

193
TREATMENT OF COMPLEX WASTEWATERS BY BIOSORPTION AND
ACTIVATED CARBON : BATCH STUDIES

C. GERENTE, Z. REDDAD, Y. ANDRES, C. FAUR-BRASQUET, AND P. LE CLOIREC


Ecole des Mines de Nantes, GEPEA, UMR CNRS 6144, BP 20722, 44300 Nantes Cedex 03,
France

E-mail: Claire.Gerente@emn.fr

In order to treat effluents characterized by metallic and organic pollutants, an association of two
different adsorbents, sugar beet pulp and granular activated carbon, is investigated. In a first step,
equilibrium data are determined for each adsorbent and mono-component solutions. Then. multi-
metallic and organic-metal solutions are tested to determine some inhibitions or special selectivities.
Finally, it is shown that the association of sugar beet pulp for metal removal, and activated carbon
for organic elimination, is efficient to treat complex wastewaters.

1 Introduction

Industrial wastewaters are often complex aqueous mixtures containing various pollutants:
heavy metal ions, organic molecules, dyes, etc. and some of them are toxic or undesirable
to many living species.
Previous investigations were mainly focused on the use of low-cost sorbents [ I ] as a
replacement for costly methods of removing heavy metals from solution. Commonly, it
concerns inorganic materials like fly-ash [2] as well as chitosan [3,4], biomass [5,6],
sewage sludges [7], peat [8] ... For these latter of biological origin, the term of
biosorption is used to encompass contaminant uptake via physico-chemical mechanisms
such as adsorption or ion exchange. The low-cost biosorbent used in this paper is a by-
product of the sugar industry: the sugar beet pulp. This material is very cheap (1 00 per
metric tonne) and its production reaches 14.106 tomes of dry matter each year in the
European Community [9]. Sugar beet pulp is a natural polysaccharide and composed of
20 % and more than 40 % of cellulosic and pectic substances respectively. These latter
contain polygalacturonic acids which cany carboxyl functions and consequently exhibit
good capacities to retain metal ions [10,1 I]. Activated carbon, within its different forms,
is commonly used in water treatment for organic pollutant removal [12,13,14].
Nevertheless, some applicationshave showed its ability in metal removal [ 151.
In this work, the treatment of a synthetic wastewater composed by metals ions (Cu2+,
Ni2+ and Pb2+) and organic molecules (benzoic acid, benzaldehyde and phenol) is
investigated with a mixture of two sorbents, sugar beet pulp and granular activated carbon.
In a first step, equilibrium data are determined in a batch reactor for each adsorbent and
mono-component solutions. Then, the pollutants are mixed to treat binary and ternary
systems of metal ions, or a combination of Cu2+with organics. A second part focuses on
an association of the activated carbon with the polysaccharide for the treatment of a
solution containing Cu2 and phenol.

194
2 Materiels and Methods

2.1 The adsorbents


Raw sugar beet pulp was provided by Lyven (France). Its preparation has been previously
described by GBrente et al. [ 101 and the fixation of several metals have been studied and
published in [11,16]. The granular activated carbon, Pica NC60 from Pica Co. (France),
presents a high specific surface area (1200 mz.g-') coupled with a large microporosity
(94.5% VOI.).

2.2 Sorption experiments


Experiments were performed in batch reactors at 21 f 1C with a continuous stirring at
500 rpm and some ratio solidsolution fixed at 2.26 g.L-' for dry pulp and 0.5 g.L-' for
activated carbon. A pre-hydration of 90 min of the pulp was necessary and the pH of the
solution was stabilized at 5.5. Equilibrium times were deduced from the kinetics. The
mixed metallic solutions had equimolar initial concentrations (8.1 O4 mol.L-'). The
influence of benzaldehyde, benzoic acid and phenol on the fixation of CU" onto the pulp
was conducted using 100 mg.L-' (expressed in TOC) of organic compounds. The
adsorption on the mixture of sorbents of phenol and Cu2' ions was carried out with 50
mg.L-' of each components.

3 Results and Discussions

3.1 Fixation of Cu", Ni2' and Pb2' in mono and multi-metallicsolutions

--
alone CuZ'
............. &!. ...... ......................................................

"
-12% - 5 0 % -42% -10%-40% -43% -56% -47% -61%
% of decrease

Figure 1. Ion competition experiments, conducted at initial equimolar ionic concentrations (8.1O4 mo1.L-I)

In the case of single metal ions adsorption, it has been seen (Fig.1) that the adsorption
capacities order following Pb" > Cu2' > Ni2' [ 161. Smith and Martell [171 and Makridou
et al. [181 have shown that Pb2'presents a higher stability constant with galacturonic acid
than Cu". Moreover, no value are featured for Ni2'and Makridou et al. [ 181 assert that a
complex between this metal and galacturonic acid does not exist. These results confirm

195
the order obtained above. Some experiments are then performed to study the competition
of adsorption of these metallic species. Experiments conducted with two different cations
show that Ni2' ion exhibits the lowest competitive effect: whereas its presence induces
decreases of adsorption capacities of 12 % for Pb" and 10% for Cu", these last two
cations prevent the Ni2'fixation at 47 and 56 % height respectively. Pb" seems to have a
higher effect than Cu". In a mixed solution of copper and lead, the influence of the two
metals is slightly different: Pb2' induces 40 % decrease in Cu2+fixation, and vice-versa
the percentage reaches 50 %. As Pb2' is supposed to have the main preference to the pulp,
this value should be abnormally great, especially as the addition of the third metal induces
a lower decrease (- 42 %). The clear sorption preference for Cu2' and Pb2+is always
marked in the three metal solution: the addition of Ni2' seems to have no effect on the
other metals fixation.

3.2 Influence of organic compounds on Cu" fwation onto raw sugar pulp
The selected organic compounds were simple aromatic molecules (phenol,
benzaldehyde, benzoic acid) which could simulate moieties present in natural organic
matters. Preliminary, removal kinetics of these organic compounds have been performed
using the same concentrations; they showed that no fixation occurred (data not presented).
Secondly, isotherms of Cu2' ions sorption have been performed with a high constant
organic charge and they are plotted on Figure 2. In comparison with the fixation of copper
alone in solution, the presence of benzoic acid, at a concentration of 180 mg.L-', induces a
reduction of about 30% on the fixation capacity. The two other compounds decrease it
lower but the experiment does not enable to separate their influence. An explanation about
this special decrease in the presence of benzoic acid could be a complexation in solution
between Cu2+ions and benzoic acid, which would prevent the cation to be fixed on the
pulp. The stability constant given by Smith and Martell [ 171 between these two species
has a value of 1.6, which is close to that obtained with the complex Cu2'-galacturonic acid
(1.8). The effect of competition is obvious and these results confirm a strong affinity of
the carboxylic functions towards Cu2+ ions. Taking into account that the organic
concentrations are relatively important, it can be supposed that less concentrations would
have little effect on copper elimination.

3.3 Association ofpulp and activated carbonfor the removal of Cu2"and phenol
Preliminary experiments were carried out on each sorbent. As a little part of the organic
content of sugar beet pulp was soluble in water (close to 35 mg.L-' expressed in TOC for a
ratio pulp/water of 2.26 g.L-'), experiments with activated carbon were carried out in a
liquid medium of pulp-water. It was verified that this specific soluble organic matter,
probably composed of great molecules, was not adsorbed on NC60. The results are
presented in Table 1. In a first approach, it was confirmed that phenol was not removed by
sugar beet pulp at this concentration. The removal of copper was efficient with and
without the presence of phenol since the removal percentages reached a value close to
60 % in both cases. This interesting result confirms those obtained above and shows that
copper ions could be treated with pulp, even with a moderate organic charge. As far as
activated carbon is concerned, on one hand the efficiency of this kind of material is
verified towards the organic molecule since NC 60 exhibits a high removal percentage for
phenol (73 %) and low for Cu" ions (14 %). On the other hand, the presence of metal

1%
decreases around 20 % the fixation of phenol and the presence of phenol slightly affects
the copper fixation. The mechanisms of adsorption must be different, favoring a
chemisorption based on an attraction with the surface functions for metal fixation and a
physisorption, highly influenced by steric hindrance when another pollutants, for example
hydrated cations, are present in solution.
0.3

0.25

ECD
3 0.2
3
2
20
0.15
-ca
-ild
E 0.1 %
Cu alone
0.05 0 Cu + 130 mg.L-1 of benzaldehyde
4 Cu + 180 mg.L-1 of benzoic acid
X Cu + 140 mg.L-I of phenol
0

Figure 2 Organic influence of high concentrationon Cu2+fixation on pulp

Table 1. Removal percentage of copper and phenol, on pulp, NC 60 and a mixture of the two adsorbents
(Initial concentration of copper and/or phenol 50 mg.L-'; pulp ratio 2.26 g.L'; NC 60 ratio 0.5 g.L").

Pulp NC60 Pulp+NC60


cu2+ 63 14 67
Cu2'(with presence of phenol) 60 11 66
Phenol 0 73 74
Phenol (with presence of Cu2+) 0 54 53

When these two different adsorbents are used together, the results show that their
respective properties are conserved. In other terms, the sugar beet pulp exhibits a high
affinity with copper ions (67 YO),even if phenol is present (66 %), and NC 60 keeps its
efficiency with phenol (74 %) and is greatly influenced by the presence of copper.

To conclude, these preliminary results have shown the important ability of a low-cost
sorbent, the sugar beet pulp, to remove metal ions from aqueous solution. When several
metals are present in solution, a selectivity can be highlighted. The polysaccharide
exhibits high affinities towards Cu" and Pb2+whereas organic molecules are not retained
on it. The influence of organic matter on metal fixation occurs at high concentration or if a
complexation between species is possible. Finally, if the effluent contains organic and
metallic pollutants, the association of two kinds of adsorbent, namely sugar beet pulp and
activated carbon, seems to be efficient since they would keep their respective properties in

197
terms of adsorption. Further investigations would confirm these results in a dynamic pilot
unit and with real industrial wastewaters.

References
,
1. Bailey S. E., O h T. J., Bricka R. M. and Adrian D. D., A review of potentially low-
cost sorbents for heavy metals, Wat. Res. 33 (1999) pp. 2469-2479.
2. Ricou P., Lecuyer I. and Le Clouec P., Removal of heavy metallic cations by fly ash
in aqueous solution, Environ. Technol. 19 (1998) pp. 1005-1016.
3. Guibal E., Milot C. and Tobin J. M., Metal-anion sorption by chitosan beads:
equilibrium and kinetic studies, Ind Eng. Chem. Res. 37 (1 998) pp. 1454-1463.
4. Gerente C., Andres Y. and Le Cloirec P., Uranium removal onto chitosan:
competition with organic substances. Environ. Technol.,20 (1 999) pp. 5 15-521.
5. Volesky B., Advances in biosorption of metals: selection of biomass types. FEMS
Microbiol. Rev., 14 (1994) pp. 291-302.
6. Texier A. C., An&& Y. and Le Cloirec P., Selective biosorption of Lanthanide (La,
Eu, Yb) ions by Pseudomonas aeruginosa. Environ. Sci. Technol. 33 (1999) pp. 489-
495.
7. Solari P., Zouboulis A. I., Matis K. A. and Stalidis G. A. Removal of toxic metals by
biosorption onto nonliving sewage sludge, Sep. Sci. Technol., 31 (1996) pp. 1075-
1092.
8. Ho Y. S. and McKay G., The sorption of lead(l1) ions on peat, Wat. Rex, 33 (1999)
pp. 578-584.
9. Dronnet V. M., Renard C. M. G. C., Axelos M. A. V. and Thibault J.-F., Binding of
divalent metal cations by sugar-beet pulp, Carbolydr. Polym., 37 (1997) pp. 73-82.
10. Gerente C., Couspel du Mesnil P., Andres Y., Thibault J.-F. and Le Cloirec P.,
Removal of metal ions fiom aqueous solution on low cost natural polyssacharides:
sorption mechanism approach, React. Funct. Polym., 46 (2000) pp. 135-144 .
11. Reddad Z., Gerente C., Andres Y. et Le Cloirec P., Ni(I1) and Cu(I1) binding
properties of native and modified sugar beet pulp, Carbohydrate Polymers, 49 (2002)
pp. 23-3 1.
12. Crittenden B., Thomas WJ., Adsorption technology and design (1998), Butterworth-
Heinemann, Boston
13. Economy J., Lin R.Y., Adsorption characteristics of activated carbon fibers, Applied
Polym. Symposium, 29 ( 1976) pp. 199-21 1.
14. Faur-Brasquet C., Metivier-Pignon H., Le Cloirec P., Activated carbon cloths in
water and wastewater treatments, Res. A h . in Water Res., 2 (2002) pp. I - 19.
15. Faur-Brasquet C, Reddad Z, Kadirvelu K, Le Cloirec P, Modelling the adsorption of
metal ions (Cu2',Ni2',Pb2+) onto activated carbon cloths using surface complexation
models, Applied Surface Science, 196 (2002) pp. 356-365.
16. Reddad Z., GBrente C., Andres Y. and Le Cloirec P., Adsorption of several metal ions
onto a low-cost biosorbent : kinetic and equilibrium studies, Environ. Sci. h Technol.,
36 (2002) pp. 2067-2073.
17. Smith R. M. and Martell A. E. Critical Stability Constants (Plenum Press, New York,
1989).
18. Makridou C., Cromer-Morin M. and Schatff J.-P., Complexation de quelques ions
metalliques par les acides galacturonique et glucuronique, Bull. Soc. Chim. Fr.,
(1977) pp. 59-63.

198
ADSORPTION CHARACTERISTICS OF PROTEIN-BASED LIGAND FOR
HEAVY METALS

MASAAKI TERASHIMA, NORIYUKI OKA, TAKAMASA SEI, KAZUYA SHIBATA, AND


HIROYUKI YOSHIDA
Department of Chemical Engineering, Graduate School of Engineering, Osaka Preficture
Universi& 1-1, Gakuen-cho.S a k i CiQ, JAPAN

E-mail: terasimaachemeng. osakfu-u.ac.jp

A fusion protein was engineered from maltose binding protein (pmal) and human metallothionein
(MT). The recombinant protein (pmal-MT) expressed in E. coli was purified, and immobilized on
ChitopearlTMresin. As expected from a tertiary structure of metallothionein, the prnal-MT ligand
adsorbed 12.1 cadmium molecules per one molecule of the ligand at pH 5.2. We have found that
the prnal-MT ligand also bound 26.6 gallium molecules per one molecule of the ligand at pH 6.5.
Adsorption isotherms for the both ions were correlated by Langmuir-type equation. Two types of
binding sites have been elucidated based on HSAB (hard and soft acid and base) theory: gallium ion
specifically binds to amino acid residues containing oxygen and nitrogen atoms, while cadmium ion
binds to specific binding sites formed by multiple cysteine residues. The pmal-MT protein bound
these metals in the concentration range of 0.2 - 1.O mM, and the bound metal ions could be eluted
under relatively mild condition (pH 2.0). The pmal-MT ChitopearlTMresin was stable and could be
used repeatedly without loss of binding activity. Thus, this new protein-based ligand would be
useful for recovery of toxic heavy metals and/or valuable metal ions from various aqueous solutions.

Introduction

Recently, the recovery and reuse of valuable metal ions such as rare earth metals,
from process waste water of electronic industries and waste electronic devices, is strongly
desired for saving precious resources and for achieving sustainable development. While
synthetic ligands or chelators are widely studied, biosorbents prepared from biomass of
bacteria, fungi, and algae have several advantages over synthetic chemical ligands. The
biosorbents, for example, show high selectivity to various ions depending on their tertiary
structures, and require only relatively mild conditions for adsorption and desorption [ 11.
Peptides and proteins could be efficient metal binding ligands, because they have the
functional groups for metal binding in their amino acid residues, and they can be produced
at low cost by recombinant technologies. While many peptides and proteins are known
to work as metal transport proteins in biological systems, metallothioneins (cysteine rich
proteins with molecular weight of ca. 7 kDa) have attracted researchers attention for
decades because they bind heavy metals in vivo [2]. The metallothioneins are considered
to be involved in detoxication and metabolism of heavy metals.
In this work, a fusion protein has been engineered from maltose binding protein
(pmal) and human metallothionein (MT). The fusion protein (pmal-MT) has been
expressed in E. coli, and purified with an amylose column. The purified fusion protein
was immobilized on a solid matrix, and its characteristics as metal binding ligand have
been studied. We have found that the pmal-MT ligand efficiently binds gallium ion, one
of the valuable rare metals desired to be recovered from aqueous solution [3]. Different
binding mechanisms for two metal ions have been elucidated based on HSAB (hard and
soft acids and bases) theory [4].

199
Methods

Preparation of p-ma1 MT protein, and immobilization of the gmal MT protein on


ChitopearlTMresin were described in detail in the previous work [3]. Chitopearlm resin
inmobilizing p-ma1 MT protein (pmal-MT ChitopearlTM),ChitopearlTMimmobilizing
p-ma1 protein (pmal ChitopearlTM),and ChitopearlTMresins have been prepared. The
latter two resins are prepared as negative controls.
The ChitopearlTMresin was packed in a glass column (inner diameter 1. 4 cm, bed
height 4.3 cm). The column was first equilibrated with a 20 mM MES buffer containing
20 mM NaCl and 10 mM 2-mercaptoethanol. Then 60 ml of the MES buffer containing
metal ion was applied at the flow rate of 0.5 ml/min. Adsorption capabilities of the
ligands were examined for cadmium, gallium, cupric, zinc, or nickel ion. Afier the
column was washed with MES buffer, the adsorbed metal ion was eluted with the MES
buffer (pH of which was adjusted to pH 2.0). In order to examine effects of pH on the
adsorption, pH of the MES buffer was varied from pH 5 to pH 9. The eluted solution
was collected as several fractions of 10 ml each, and the metal concentration of each
fraction was determined with atomic adsorption analysis (SAS 7500A, Seiko Instruments,
Japan). Total amount of the eluted metal ion was defmed as the adsorbed metal ion on
the resin. The total amount of the adsorbed metal ion was divided by the total amount of
immobilized protein to calculate the number of metal molecules bound to one mole of the
protein. The adsorption experiments were carried out multiple times, and the maximum
experimental error was 25%.

Results

Amounts of the protein immobilized on the ChitopearlTMresin were 3.55 (mg/g-wet


resin) for the pmal and 1.51 (mg/g-wet resin) for the pmal-MT. The optimal pH for
cadmium binding was pH 5.2 (data not shown). Figure 1 shows an adsorption isotherm
at 298 K for cadmium adsorption on the pmal-MT ChitopearlTMresin at pH 5.2. Neither
the pmal ChitopearlTMresin nor the ChitopearlTMresin adsorbed cadmium ion under the
employed experimental condition. These results clearly show that cadmium ion binds to
the metallothionein moiety of the pmal-MT ligand. The adsorption equilibrium was
correlated by a Langumuir-type equation. The equilibrium constant K, and adsorption
capacity for cadmium binding Q were 15.74 [mM-'] and 3.76 x 1 0 ' [mol/g-wet resin],
respectively. The maximum amount of adsorbed cadmium ion per metallothionein
molecule is 12.1 (mol cadmium/mol metallothionein), which is relatively close to a
theoretical value 7 confirmed by NMR [ 5 ] . These results strongly suggest that the
methallothionein moiety of the h i o n protein bind cadmium as it works in vivo. Figure 2
shows effect of NaCl concentration in the metal solution applied to the column on the
cadmium adsorption. The amount of adsorbed cadmium ion drastically decreased at
NaCl concentration about 45 mM, suggesting that the tertiary structure of the metal
binding site probably change at this salt concentration, and thus the ligand lose its binding
ability for cadmium ion. The binding ability, however, was easily recovered by washing
the column with the MES buffer (pH5.2).

200
0.2 a4 0.6 QI) 1 0 50 100 150 m
Cadmium eweem. fmMl NaCI ancen. [mMl
Figure 1 Adsorption isotherm Figure 2 Effect of NaCl on
for cadmium ion cadmium adsorption

We have found in this work that the pmal-MT ligand also binds a valuable rare metal
gallium ion. Adsorption characteristics of the pmal-MT for gallium ion, however, were
different from those for cadmium ion. An optimal pH for gallium ion adsorption was pH
6.5, which was different fiom that for cadmium ion (pH 5.2). This result suggests that
.the conformation of metallothionein suitable for binding of cadmium ion is not preferable
to gallium ion, and vise versa. An adsorption isotherm for gallium ion at 298 K is shown
in Figure 3. Unlike the case of cadmium ion, the pmal ChitopearlTMresin and the
ChitopearlTMresin adsorbed gallium ion about 4.0 x lom7(mol/g-wet resin) at pH 6.5.
Since these results suggest that gallium ion bind to the proteins non-specifically,
non-specific binding of gallium ion to the proteins was examined by comparing the
adsorption of gallium ion on BSA. The numbers of gallium ions adsorbed on BSA (mol
gallium iodmol BSA) were 0.0335 at pH 5.2 and 0.114 at pH 6.5, while those of
cadmium ions (mol cadmium iodmod BSA) were 0.392 at pH 5.2 and 0.293 at pH 6.5.
These results showed that BSA did not adsorb cadmium ion and gallium ion under the
employed experimental condition, and suggest that both ions do not bind to proteins by
simple ion-exchange effects. In order to evaluate the adsorption on metallothionein
moiety, the amount of gallium ion adsorbed on the base matrix was subtracted from the
experimental data for the pmal-MT ChitopearlTM resin. The corrected result is shown in
Figure 4, and adsorption equilibrium is correlated by Langmuir-type equation. The
equilibrium constant K, and adsorption capacity for cadmium binding Q were 5.2 1 [mM']
and 9.09 x [moVg-wet resin], respectively. It should be noted that the maximum
amount of adsorbed gallium ion per metallothionein molecule reached 26.6 (mol gallium
iodmol metallothionein), which is much higher than that of cadmium ion. Effect of
NaCl concentration in the metal solution on the gallium adsorption is shown in Figure 5.
The drastic decrease in metal binding at NaCl45 mM, which was seen for cadmium ion,
was not observed in the case of gallium ion. These results also strongly suggest that the
binding mechanism of gallium ion is different from that of cadmium ion. The amounts of
adsorbed metal ion on the pmal-MT ChitopearlTMfor various ions are summarized in
Table 1. These results show that the metal binding on metallothionein is highly
selective.

201
I,

0 .

0 0.2 0.4 0.6 OS I 0 0.2 0.4 0.6 M 1


Gallium C O I I C ~ . [mMl
Gallium cancea. [mMl
Figure 3 Adsorption of gallium ion Figure 4 Adsorption isotherm
for gallium ion

Table 1 Amount of adsorbed metal ion for


various ions

Metals Adsorbed Ionic radius


amount [x 1P m]
Cia 26 62 Hard
Ni 0.3 69 Little hard
Zn 0.2 74 Little soil
Cd 12.1 91 Soil

0 50 100 150 200


NaCI canem. [mMl
Figure 5 Effect of NaCl on gallium
adsorption

Discussion

We have found that metallothionein, which selectively binds cadmium ion in vivo,
binds gallium ion. The number of the gallium ion bound to one molecule of
metallothionein, 26.6, was about twice as large as that of cadmium ion. The results of
adsorption experiments to BSA, and the adsorption of various metal ions show that
metallothionein selectively binds gallium ion. The specificity is very high because other
metals which as similar ionic radius did not bound to the metallothinein as shown in Table
1.
A binding mechanism of metallothionein for cadmium ion and gallium ion is
elucidated as follows. According to HSAB (hard and soft acid and base) theory, metal
ions are classified into hard acid (ion) and soft acid (ion). Cadmium ion is classified as
soft acid that has strong binding affinity to S and P atoms. On the other hand, gallium
ion classified as hard acid that has strong binding affinity to N and 0 atoms. Therefore,
cadmium ions bind to S atom of cysteine residues of the metallothionein as indicated by
NMR. The conformation of binding sites formed by multiple cysteine residues should be

202
strongly affected by the change in tertiary structure of the metallothinein. Drastic change
of the amount of adsorbed cadmium ion by the increase of NaCl concentration (Figure 2)
suggests that the large conformation change of the cadmium binding sites of
metallothionein. On the other hand, gallium ion should bind to N and 0 atoms of amino
acid residues of the metallothionein molecule. Human methallothionein posses 3 Asp, 1
Glu, 7 Lys, 8 Ser, 2 Thr,1 Gln,1 Asn residues per molecule. Some of these negatively
charged residues are probably located the outer surface of metallothinein might form
binding sites.
As elucidated fiom the adsorption experiment on BSA, adsorption of gallium ion by
simple ion-exchange effect is negligible. The characteristicsof metal ions such as ionic
radius and hardness/sobess, and the conformation of the metallothionein probably affect
the selectivity of metal adsorption. The understanding of the mutual interactions among
those factors would be a key factor in designing the protein-based ligand suitable for a
specific metal ion.

References

1. Gutnick, D. L., and Bach, H., Engineering bacterial biopolymers for the biosorption
of heavy metals; new products and novel formulations,Appl Microbiol Biotechnol. 54
-
(2000) pp. 45 1 460
2. Romero-Isart, N. and Vasak, M., Advances in the structure and chemistry of
metallothioneins,J. Inorg. Biochem. 88 (2002) pp.388-396
3. Terashima, M., Oka,N., Sei, T. and Yoshida, H., Adsorption of cadmium ion and
gallium ion to immobilized metallothionein fusion protein, Biotechnol. Progress, in
press (2002)
4. Pearson, R. G., Hard and soft acids and bases, J. Am. Chem. SOC.85 (1963) pp.
3533-3539
5. Messerle, B. A., Schaffer, A., Vasak, M., Kagi, J. H., and Wuthrich, K.
Three-dimensional structure of human [ 1 13Cd7]metallothionein-2 in solution
determined by nuclear magnetic resonance spectroscopy, J Mol Biol. 214 (1990) pp.
765-779

203
PREPARATIVE CHROMATOGRAPHY AT SUPERCRITICAL CONDITIONS

ARVIND RAJENDRAN, MARC0 MAZZOTTI AND MASSIMO MORBIDELLI


ETH Swiss Federal Institute of Technology, CH 8092 Zurich, Switzerland
E-mail: mazzofti@ivukmavt.ethz.ch

The basic issues in packed bed supercritical fluid chromatography (SFC) are: i) characterizing
adsorption equilibrium; ii) charracterizing mass transfer dynamics; iii) predicting column dynamics
and iv) the effect of polar modifiers. The first three issues have been addressed through experiments.
Experiments have been performed on a preparative column with significantpressure drop using pure
COz as an eluent and phenanthrene dissolved in toluene as solute. Parameters relating to retention
behaviour and mass transfer characteristics were measured. The observations are reported and the
deviations with respect to HPLC and GC behaviour are highlighted and discussed.

1 Introduction

Preparative chromatography is a proven technology for the separation of specialty


chemicals mainly in food and pharmaceutical industries, particularly the
enantioseparation of chiral compounds on chiral stationary phases. The potential of
preparative chromatographic systems were further increased by the development of
continuous chromatographic processes like the simulated moving bed (SMB) process.
Compared to the batch column chromatography, the SMB process offers better
performance in terms of productivity and solvent consumption [2].
Supercritical fluids poses properties that lie in between those of liquids and gases.
These properties, which are functions of their density, can be tuned according to the
process requirements. Supercritical fluid chromatography (SFC) is a proven analytical
tool which allows one to exploit the advantages of GC (faster and better separation) for
non-volatile substances which are usually analysed using HPLC. The solvent power of the
supercritical fluid is a function of its density. Hence, when a solute dissolved in a
supercritical fluid is in contact with a stationary phase, i.e. a solid adsorbent, the affinity
of the solute to the stationary phase, which under linear conditions is characterized by the
Henrys constant Hi, depends on the density of the supercritical fluid. At higher densities,
the supercritical fluid has a higher solvent power and hence the solute has a lower Hi
value and vice versa. This property allows for the establishment of a solvent gradient in a
SMB unit, thereby enhancing separation performance. In fact it was shown that the
supercritical fluid simulated moving bed (SF-SMB) process operated under the pressure
gradient mode (where the pressure in each section is regulated by a back pressure
regulator) offers a productivity improvement by a factor of 3 compared to the isocratic
mode (without any pressure regulation) [l]. In this context, it is therefore important to
study and understand the fundamentals of SFC under conditions where the pressure drop
along the column is significant, because these are the conditions in preparative
applications, particularly in SF-SMB.
The pressure gradient in the column causes substantial density gradient, which in turn
leads to a gradient in velocity. Since the mass transfer properties are a function of density,
they change at every point along the column. Moreover, unlike HPLC and GC systems,
where the Henrys coefficient, Hi, does not depend on the pressure, in SFC systems the
retention characteristics are a function of pressure (density) [l]. Hence, the dynamics of
an injected pulse is governed by several factors. This establishes the need to develop a

204
model, which could be used to simulate the pulse response of a packed column under
supercritical conditions with substantial pressure drop. A dynamic model is especially
useful, since it can be easily extended to simulate the SF-SMB process.
In the present study, experiments have been performed on a preparative column using
pure COz as the mobile phase. Experiments have been performed at four different back
pressure levels. For each back pressure setting, runs were performed both at low flow
rates, i.e. where the pressure drop was negligible and at high flow rates. The main process
aspects, i.e. pressure drop, mass transfer and retention have been experimentally
evaluated and analysed.

2 Experimental setup and procedure

Carbon dioxide (99.995% pure, obtained from PanGas, Switzerland) was used as the
mobile phase. Phenathrene @urity>97%) dissolved in Toluene (purity>99.7%, both

100

80 - +- BP =I80 bar
n
L
+-BP =210 bar
lu
a
LI
60-

0
E!
a
2 40-
v)
v)
2
n
20 -

0 I I

I I I I
0 I 2 3 4x1Om2
Mass flow [g/s]

Figure 1. Pressure drop characteristicsof the SFC column at different back pressure levels.
obtained fiom Fluka, Switzerland) was used as a solute. A Lichrospher 100 RP-18
column (Merck, Darmstadt) 125x4 mm, with an average particle size of 5pm, was used
for the experiments.
The experimental set-up consists of a syringe pump (Isco 260D) capable of
producing a continuous flow of COz which flows through an injection valve (Valco

205
C14W) that has an internal loop volume of 60 nL. The mobile phase then enters the
column at the end of which is a UV detector (Jasco W-1570).The pressure in the system
is determined by a back pressure regulator (Jasco BP1580-81) which is located
downstream of the U V detector. The column and the injection valve are housed in a
temperature controlled water bath. Upstream and downstream pressures are measured
using pressure transducers (Trafag 8891).
The experiments are performed by setting the back pressure regulator at the desired
level and programming the syringe pump to operate at a given flow rate. The system is
then allowed to reach a steady state. Once the pressure profile in the system is established,
a mixture of phenanthrene in toluene (2% w/w) is injected into the column through the
injection valve and the data acquisition is started simultaneously. For each setting, the
experiment is repeated more than three times to ensure reproducibility.

3 Experimental Results

Both high flow rate and low flow rate experiments were performed at 4 different
back pressure levels, namely 130, 150, 180 and 210 bar. Some low flow rate experiments
were also perfomed at intermediate back pressure settings. The operating temperature for

0 BP=150 bar
A B P 4 8 0 bar
V BPe210bar

0 200 400 600 800


(mass flow ratej[(g/sj]

Figure 2. Retention time against the inverse of mass flow rate at different back pressure levels.

206
all the runs was 64C.The pressures upstream and downstream of the column and the UV
signal were measured.
From the UV detector readings, the retention time and HETP values were calculated.
The HETP was calculated using the formula
N = 5.45(t,/~)~
HETP = L/N
where t R is the retention time, w the width of the peak at half peak height, N the number of
plates and L the length of the column. These equations have been used to describe the
HETP behaviour though they have the limitation that they assume the velocity to be
constant along the column.
The pressure drop characteristicsfor the four different sets of experiments are plotted
against the mass flow-rate in Fig. 1. The mass flow rate is used as the independent
variable since it is the only variable which remains constant throughout the column. At
higher flow rates there is a deviation from linearity. For a given mass flow rate, the runs
at a low back pressure setting show a larger pressure drop.
The measured retention times are plotted against the inverse of mass flow rate in Fig.
2. It can be seen that the points corresponding to a particular back pressure setting fall on
a straight line and the slope of the line depends on the back pressure setting. The line
corresponding to a lower back pressure setting has a larger slope than the one at a higher
back pressure setting. This is in contrast to HPLC where for a given temperature, under
linear conditions, all points, irrespective of the back pressure setting, will fall on one
straight line whose slope is proportional to Hi. This shows that in the case of SFC, Hi,is a

6o
50 -

-
E3.
Y
40 -

n 30-
F
w
I
20 -

10 -

0
0 10 20 30 40
0
r
50x10"
Mass Flow [glsJ
Figure 3.HETP values at different back pressure levels.

207
function of density. Further at high flow rates, there is a velocity gradient and a density
gradient in the system and these affect the retention time. Hence, the observed, or the
apparent, Hi is a combined effect of these two gradients.
The mass transfer characteristics, which are described by the HETP, are shown in
Fig. 3, where the HETP is plotted against the mass flow rates for different back pressure
levels. In general, the HETP curve has the typical shape of the well-known van Deemter
plot. Though plotting the HETP in this fashion (i.e. grouping runs with the same back
pressure setting) does not offer the provision to extract the mass transfer parameters 6om
the van Deemter equation, it nevertheless offers a qualitative picture of the mass transfer
kinetics. In general, at low flow rates, the H E P values fall with increasing flow rates,
reach a minimum, and gradually rise. Let us focus on the part of the curve after the
respective minima. The curve is flat for the runs with a back pressure of 130 bar
compared to those corresponding to 150 and 180 bar. The slopes of the later part of the
curves increase with increasing back pressure. It can also be seen that the curves show
cross over at larger flow rates. For the curve corresponding to 210 bar, it was however,
not possible to perform high flow rate experiments as the upstream pressures rose beyond
the maximum allowable pressure of the syringe pump.

4 Conclusion

Experiments have been performed on a preparative SFC system using pure CO2 as
the mobile phase under significant pressure drop. The retention times, pressure drop
characteristics and the mass transfer behaviour were studied. The trends observed differ
6om the behaviour of HPLC systems. These trends also emphasize the complexity
involved in analyzing the data for SFC measurements, which imply in turn greater
complexity of the SFC model as compared to standard liquid chromatography model.

Reference

1. Denet, F., Hauck, W., Nicoud, R. M., Di Giovanni, O., Mazzotti, M., Jaubert, J. N.
and Morbidelli, M., Enantioseparation through supercritical fluid simulated moving
bed (SF-SMB) chromatography, Ind. Eng. Chem. Res. 40 (2001) pp. 4603-4609.
2. Juza, M., Mauotti, M. and Morbidelli, M., Simulated moving-bed chromatography
and its application to chirotechnology, Tren& Biorechnof.18 (2000) pp. 108- 1 18.

208
ADSORPTIVE SEPARATIONOF OLIGOSACCHARIDES INFLUENCE OF
CROSSLINKING OF CATION EXCHANGE RESlNS

JOHANA. VENTE'.~,HANS BOSCH', ANDRE B. DE HAAN', PAUL J.T. BUSSMANN~

I Separation Technology Group, Faculty of Chemical Technology, Universiw of Twente, P. 0.Box


21 7, 7500 AE Enschede, The Netherlands

TNO-MEP, P.O. Box 342, 7300A H Apeldoorn, The Netherlands

The influence of crosslinking on the sorption propetties of poly(styrene-co-divinylbenzene) (PS-


DVB) strong acid ion exchanger in K ' or Ca2+form was studied. Isotherms of sugars were deter-
mined for resin containing 2, 4 and 8% DVB. Sorption was strongly influenced by the degree of
crosslinking. At increasing crosslinking the amount of sorption by (non-selective) distribution
decreases and complexation is the dominant sorption mechanism. Resin with 2% DVB was hardly
selective, but resin with 8% DVB was very selective. A selectivity of 7.0 for fructose/sucrose and
Ca" loaded resin was obtained due to a combination of size exclusion and complexation. However,
sorption capacity decreared with increasing crosslinking. Chromatograms of fructo- and
galactooligo-saccharides (0s)showed that the separation of the monosaccharidesfrom the 0s was
best for 8% DVB but the fractionationof the 0s was best for 2%DVB.

1 Iatroductioa

Oligosaccharides (0s)are applied as functional food and feed ingredients. Fructo-OS


(FOS) are made by partial hydrolysis of inulin and galacto-0s (GOS) by trans-
galactosylation of lactose. They are produced as mixtures of different types of carbohy-
drates and further separation is required for most applications. Examples of desired sepa-
rations are the removal of monosaccharides in order to decrease the amount of calories
and the sweet taste. A technique to perform the desired separation is chromatography. The
adsorbent used in a chromatographic separation has a large influence on the performance
of the separation process. Fructose/glucose is separated on process scale with Ca" loaded
crosslinked poly(styrene-co-divinylbenzene) (PS-DVB) strong acid ion exchange resins.
To investigate the influence of resin properties on the separation, we selected PS-DVB
resins. There are two chemical properties of the PS-DVB resin material to optimise: (1)
the type of cation and (2) the degree of crosslinking. Earlier work focussed on the choice
of the cation. K ' was selected as the cation for GOS separation and Ca" for FOS
separation [1,21. The work presented here focussed on the degree of crosslinking.
The degree of crosslinking of the PS-DVB resin can be varied by the amount of DVB
used during the synthesis of the resin. Crosslinking decreases the elasticity of the resin
and thereby the swelling and equilibrium water content [3]. Saccharides can be sorbed
into the resin, either by distribution of the saccharides between the liquid inside the resin
and the liquid outside the resin or complexation [4]. To be able to distribute, the
saccharide molecules have to be smaller than the size of the interstices of the resin. The
size of hydrated monosaccharides is in the same order of magnitude as the size of the pore
diameter of a resin [5]. The separation of monosaccharides from 0 s may be improved by
choosing a DVB content such that disaccharides and 0s are excluded, but the
monosaccharides are still able to be sorbed by the resin. Higher water content increases
the sorption of the saccharides by distribution. The amount of cations per volume unit
water increases with increasing DVB content. Therefore, complexation driven sorption

209
might improve relative to the sorption by distribution with increasing DVB content. The
DVB content not only influences the equilibrium sorption properties of the resin, but also
the sorption kinetics and mechanical properties such as elasticity and attrition resistance.
Published data illustrate the influence of the degree of crosslinking of PS-DVB resins
on the separation of sugar alcohols [6], monosaccharides [7],and malto-OS [8]. Resins
with DVB contents between 3% and 8% are suitable for the separation of two hexoses
[7]. For sugar alcohols the optimal DVB content is 7% DVB for Ca2' form resins [6].
The separation of glucose from maltose improves with increasing DVB content. For
malto-OS and DVB contents between 2% and 6%, the separation improves with
increasing DVB, but is completely lost for a DVB content of 8% due to size exclusion of
the larger 0s by the resin. Resin with 6% DVB was optimal for the separation of malto-
0s with a degree of polymerisation of 1 to 7.It can be concluded from these literature
data that crosslinking has a large effect on the sorption and separation properties of PS-
DVB resin and that the effects are dependent on the type of sugar. However, no
information could be found on the separation of FOS or GOS. The goal of the work
presented in this paper was to determine the influence of the crosslinking on sorption
properties of PS-DVB cation exchange resins for FOS and GOS separation. Improved
understanding of the interactions of saccharides with sorbents may lead to the
development of highly selective sorbents for low cost separations.

2 Materials and methods

Gel type strong acid PS-DVB cation exchange resin, Dowex 50W (particle diameter 38-
74 pm), was used with different DVB contents. The resin was ion exchanged into the
desired cation form. Table 1 summarizes the properties of the columns (internal diameter
0.160 m) packed with resin. Porosity was calculated from the retention time of Dextran
T2000 (Pharmacia, Sweden). Dry substance content was measured by air-drying of
filtrated resin at 105C until constant weight. Isotherms and chromatograms were
measured at 60C in a column set-up, as was described earlier [2]. Glucose (0-300 gll),
galactose (0-10 a),
fructose (0-300 a),
sucrose (0-300 a),
lactose (0-200 and a)
mixtures of FOS (Raftilose@60, Orafti, Belgium), containing fructose, sucrose and 0s
and GOS (Elixor@259, Borculo Domo Food Ingredients, The Netherlands), containing
glucose, galactose, lactose and 0s were used. The isotherms were correlated with
q=uz+bc, with q the saccharide concentration in the resin, a and b fitparamters and c the
concentration in the liquid phase. The chromatograms were plotted as a fimction of
dimensionlesstime, defined as: (r-r,mcer)/r,mce,., with rrmer the retention time of dextran and
r the time. The selectivity of component i relative to component j , was calculated as:
(qh)4@$

Table 1: Properties of columns packed with strong acid PS-DVl3 cation exchange resin (i.e.=ionexchange).

210
3 Results and discussion

3. I Isotherms of saccharides on cation exchange resins


Fig. 1 shows the single sugar isotherms for K+ and Ca2' loaded resins with different
degrees of crosslinking. High sugar concentrations were applied, because concentrated
sugar solutions are used in commercial processes. The isotherm data were correlated with
the equations in Table 2. A stronger crosslinked resin, resulted in less sorption of sugars.
This result can be explained by the decreased elasticity and swelling of the resin with an
increased crosslinking. This results in lower water content of the resin (see also Table 1)
and decreasing sorption of saccharides by distribution. Moreover, crosslinking had a
larger effect on sorption than the type of cation [I, 21. For K
' loaded resin, the observed
order of adsorption was galactose2fructose>glucose>lactose>sucrose. The sorption order
of the sugars was for Ca" loaded resin the same as for K" loaded resin, except that
fructose2galactose. Fructose forms a complex with a Cat+ ion [9], which explains that
fructose sorbed better than glucose or galactose in highly crosslinked Ca" loaded resins.
For low crosslinking however, the monosaccharides sorbed almost to the same extent.
The explanation for the loss of selectivity may be that at low crosslinking the resin
contains more water and the amount of cations per volume unit resin is lower.
Consequently, more distribution may occur and the amount of fructose sorbed due to
complexation may decrease relative to the amount sorbed by distribution. In that case, the
sorption of non-complexing monosaccharides such as glucose or galactose is favoured
relative to fructose, as was observed experimentally.
At equal degree of crosslinking, the disacharides lactose and sucrose sorbed less than
the monosaccharides due to their larger size and hence the restricted accessibility to the
resin interstices. The sorption of the disaccharides decreased strongly with increasing
crosslinking. For 8% DVB the sorption was almost vanished. At increasing crosslinking
less water is available in the resin for distribution. In addition, part of the water is
hydration water of the cations [lo]. Apparently, for 8% DVB almost no water was avail-
able for distribution of sugar. The effect of crosslinking was stronger for sucrose than for
lactose. Lactose (0-PD-galactopyranosyl-(1,4)-D-gIucopyranose)and sucrose (0-a-D-
glucopyranosyl-(1,2)-PD-fructofuranoside)differ in the constituent monosaccharides and
the bond between the monosaccharides. Although sucrose contains a fructose unit, it is
not able to complex with Ca" in the same way as the monosaccharide fructose, because
the glycosidic bond of sucrose occupies the complexation site of the fructose unit. How-
ever, the differences in sorption may be the result of differences in the structure of the
molecules. Lactose is able to convert via the open chain form to another anomeric form,
whereas sucrose does not. Sucrose exhibits two interresidue intermolecular hydrogen
bonds in aqueous solution [I 13. These structural differences may result in a larger effec-
tive size of sucrose compared to lactose, hence increased size exclusion and less sorption.

Table 2: Isotherm correlations for 2,4 and 8% crosslinking of PS-DVBresin loaded with K' or Ca*' at 60C.

211
0 1w 200 300 0 100 m 300 400

Fig. 1: Isotherms of glucosc for 2%,4%and 8% DVB on PSDVB rcsin with K


' (left) or Caz+(right) at 60C

3.2 Selectivity of cation exchange resins


Table 3 lists selectivities for sugar concentrations at 200 g/l, which were calculated from
the single sugar isotherms in Table 2. Besides the fiuctose/glucoseselectivity, Table 3 in-
cludes the selectivities for glucose/lactose and fructose/sucrose, which are indicative for
the separation of glucose from GOS and fructose from FOS, respectively. All selectivities
increased with increasing crosslinking. The increase was already explained for fructose/
glucose in the previous section. The selectivity of glucose/lactose increased with increa-
sing crosslinking due to the larger size of lactose compared to glucose. The selectivity of
fructose/sucrosereflects a combination of complexation of fructose with Ca2' cations and
size exclusion of sucrose and resulted therefore in the highest selectivity. Selectivities as
high as up to 2.5 for glucosellactoseand up to 7.0 for fhctoselsucrose were obtained.

Table 3: Selectivityof resin with P?,


4%and 8% crosslinking at an individual sugar concentration of 200 gll.

1.1 1.3 1.5 1.1 1.3 2.5


1.2 1.7 3.8 1.4 2.4 7.0

3.3 Chromatograms
Fig. 2 presents chromatograms of GOS on K ' resin and FOS on Ca" resin. Due to
the use of a short column and a relatively large resin particle diameter, baseline separation
was not achieved. However, qualitative effects of different DVB content of resins could
be obtained from the chromatograms. It appeared that with decreasing DVB content of
the resin, saccharides eluted over a longer period and the peaks became wider. The
components in the mixtures that were not retained, eluted first and with increasing time,
more retained components eluted. Glucose was the last eluting peak in the chromatogram
of GOS and fructose the last peak for FOS. From injection of pure components it
appeared that the retention times of the saccharides, which were determined only by
equilibrium effects, were in agreement with the measured isotherms [I]. Fig. 2 clearly
shows that, due to increased sorption capacity, the elution times increased with decreasing
crossliiking. Separation of fructose was in particular good for 8% DVB, as was expected
from the high selectivity of fructose/sucrose. Furthermore, Fig. 2 shows that 4% and 2%
DVB improved the hctionation of OS, which resulted in more peaks on the
chromatogram. Although 0s eluted over a longer period, the peaks on the chromatogram

212
for 2% DVB were wider than the peaks on the chromatogram for 4% DVB. Especially in
the case of GOS this resulted in poor separation. The long elution times lead to excessive
product dilution, which is for large-scale applications an economical disadvantage.

GOS m 4% DVB
nrinmK+fan 1 GOS m 2% DVB
rcsin m K' fam

0.0 0.5 1.0 1.5 20


dk*nsM...nktbnk.(J
25 0.0 0.5 1.0
dlnm-
1.5
- 20
-13
25 0.0 0.5 1.0 1.5
dkmnimhu-mC)

Fig.2: Chromatograms of G O S (K' resins) and FOS (Ca resins) with different DVB content (8?4 4%, and 2%
*+
20 25 :

DVB), injection volume 1 ml, concentration 100 g synrp/l, flow rate 0.545 ml/min, temperature 60C.

4 Conclusions

The isotherms of sugars on PS-DVB resin showed that increased crosslinking resulted in
more selective sorption of sugars at the cost of sorption capacity. The effect of cross-
linking on capacity and selectivity is larger than the effect of cation type [2]. Non-
selective sorption of sugars decreases with increasing crosslinking, due to a decrease in
available space in the resin for distribution. Also, with increasing crosslinking, the
relative contribution of complexation to the sorbed mount increases and as a result the
selectivity increases. The chromatogramsof FOS and GOS showed that a DVB content of
8% is better than 4% DVB for monosaccharide removal. A PS-DVB resin with 8% DVB
is best for the removal of fructose from FOS.However, if the goal is to separate the FOS
in several fractions with different degree of crosslinking, then it is better to use 4% DVB.
For separations including molecules with different size, it is recommended to select first
the optimal degree of crosslinkingbefore optimising the cation type.

References

11 J. A. Vente, et al., to be submitted, 2002.


2!I J. A. Vente. et al. in: AIChEAnnual Meeting - Symposium on industria/App[icationsof
I4,dkorption and Ion Exchange. 2001. Reno (NV), USA: AIChE, p. 786.

3 J. Tiihonen, et al., Journal of Applied Polymer Science, 2001.82: p. 1256.


4 S. Adachi, et al., Bioscience Biotechnology and Biochemistry, 1997.61( 10): p. 1626.
1

5 M.Saska et al., Journal of Chromatography, 1992.590: p. 147.


6 H. Caruel, et al., Journal of Chromatography, 1992.594: p. 125.
7 S. Adachi, et al., Journal of Chemical Engineering of Japan, 1999.32(5): p. 678.
a S. Adachi, et al., Agricultural and Biological Chemistry, 1989.53(12): p. 3193.
J R. W. Goulding, Journal of Chromatography, 1975.103: p- 229.
1 9 R. S. D. Toteja et d.,Langmuir, 1997. 13(11): p. 2980.
1 I] S. Immel and F. W. Lichtenthaler, Liebigs Annalen, 1995(1 I): p. 1925.

213
IDENTIFICATION AND PREDICTIVE CONTROL OF A SIMULATED
MOVING BED PROCESS

IN-HYOUP SONG AND HYUN-KU W E E


School of Chemical Engineering h Institute of Chemical Processes
Seoul National Universiw, Kwanak-ku, Seoul, 151 742, Korea -
E-mail: hkrhee@snu.ac.kr

MARC0 MAZZOTTI
Institute of Process Engineering, ETH Zurich, CH-8092 Zurich, Switzerland
E-mail: Mazzotti@ivuk mavt.ethz.ch

Identification technique is applied to a simulated moving bed (SMB)process for chiral separation of
enantiomers of TrOgers base and an advanced predictive controller is designed on the basis of the
identified model. To obtain the identified model, an artificial continuous system is constructed by
keeping the discrete events such as the switching time and the number of columns to be switched
constant. The SMB process is identified as an inputloutput data-based prediction model, which is
then used to design a linear predictive controller. In this study the internal flow rate ratios are chosen
as the input variables whereas the pair of product purities are taken as the controlled outputs. It is
demonstrated by simulation studies that the designed predictive controller performs satisfactorily for
the disturbance rejection as well as for the setpoint tracking in the SMB process.

1 Introduction

In the chemical industry, chromatographic separation process is an emerging


technology for the separation of pharmaceutical products, food and fine chemicals. To
improve the economic viability, a continuous countercurrent operation is often desirable
but the actual movement of the solid leads to a serious operating problem. Therefore, the
simulated moving bed (SMB) process is an interesting alternative option.
In recent years, several researchers have applied some advanced control strategies to
simulated moving bed units to treat the dynamic operation of SMB processes, ranging
from the nonlinear control strategies such as the input-output linearizing control [4] to the
repetitive model predictive control [S]. One of the shortcomings of these control strategies
is that they use the first principles model. Although these controllers may be effective to
treat various control problems, it may be difficult to implement them to actual SMB
processes because of their heavy computational load and complex design procedure.
To overcome these drawbacks of the controller based on first principles model,
various identification techniques were applied. Neural network based model predictive
control was used for the dynamic control of SMB unit [ 11, however, its implementation to
actual process can be very difficult because of the complexity of identified neural net
model.
In this study we identify an SMB process using the subspace identification method.
The well-known inputloutput data-based prediction model is also used to obtain a
prediction equation which is indispensable for the design of a predictive controller. The
discrete variables such as the switching time are kept constant to construct the artificial
continuous input-output mapping. With the proposed predictive controller we perform
simulation studies for the control of the SMB process and demonstrate that the controller
performs quite satisfactorilyfor both the disturbance rejection and the setpoint tracking.

214
2 Brief description for SMB process

In this study, the SMB process is divided into four sections, each of which consists of
2 columns of chromatographyplaying a specific role in the separation. Ethanol solution of
the racemic Troger's base is taken as the feed stream and unsupported microcrystalline
cellulose triacetate(CTA) bead is used as the stationary phase. The separation is carried
out in the two central sections. For the reference conditions of simulation study, one may
refer to the previous work[7].
The principle of operation can be best described with reference to the equivalent true
countercurrently moving bed (TCC) configuration. Since the two configurations are
equivalent, i.e., they achieve the same separation performance provided the geometric and
kinematic conversion rules are fulfilled, the simpler model of the equivalent TCC unit can
be used to predict the steady state separation performances of SMB units, in particular, for
design purposes.
The first principles model of the SMB unit is constructed with reference to the
previous works[4,6] and considered to be the actual plant.

3 Subspace identification of SMB process

Figurel. Validation of identified model.

In order to solve the first principles model, finite difference method or finite element
method can be used but the number of states increases exponentially when these methods
are used to solve the problem. Lee et d [ 8 ] used the model reduction technique to reslove
the size problem. However, the information on the concentration distribution is scarce and
the physical meaning of the reduced state is hard to be interpreted. Therefore, we intend to
construct the input/output data mapping. Because the conventional linear identification
method cannot be applied to a hybrid SMB process, we construct the artificial continuous
inpudoutput mapping by keeping the discrete inputs such as the switching time constant.
The averaged concentrations of rich component in raffinate and extract are selected as the
output variables while the flow rate ratios in sections 2 and 3 are selected as the input
variables. Since these output variables are directly correlated with the product purities, the
control of product purities is also accomplished.

215
We adopt the inputloutput data-based prediction model using the subspace
identification technique. To find the correlation between the inputs and outputs, we need
to obtain the input and output data. On the basis of the triangle theory[6], the optimal feed
flow rate ratios at steady state are calculated. Then, the pseudo random binary input signal
is generated on the basis of this optimal value. Figure 1 compares the output from the
identified model (dot) with that from the first principles model (solid curve). Clearly, we
observe that the identified model based on the subspace identification method shows an
excellent prediction performance. The variance accounted for (VAF) indices for both
outputs are higher than 99%. The detailed identification procedure can be founded in the
literature [3,5,9,10].

4 Predictive control for SMB process

The inputloutput data-based predictive controller based on the identified model is


designed and applied to a MlMO control problem for the SMB process. We use the
input/output data-based prediction model in the MPC algorithm. The QP method is used
to obtain the control input u,by minimizing the objective function defined as

minJ(u f ) = ( L p
"f
+ L,wp - r f ) T@Lu,u, + L , W ~- r,)+u;Ru,
where w p is defined as OPT u ~with ~ values of inputs up and output yp , and rj
~ the) past
denotes the set-point trajectory. Here, Q and R are the weighting matrices for the output
and input, respectively and the controller parameters L,, and L," are determined during the
identification procedure. It is to be noted that there is no need to explicitly calculate the
state estimate or the state space model. The complete separation region in the triangle
theory is considered as the input constraint, and the output constraint comes from the
requirement for the purity. One may refer to our previous work[9] for the details of the
controller design procedure.
Here we shall treat two typical control problems of practical interest; one is the
disturbance rejection and the other is the setpoint tracking. First, we assume that the feed
pump stops during 40 minutes after 40th switching. These may be considered as
unmeasured disturbances introduced to the process. Figure 2 shows that the controller
successfully rejects the unmeasured distu rbance.

216
B
j
g .r
t
B

Figure 2. Disturbance rejection performance.

Figure 3. Tracking control performance.

In the second case the setpoints for the average concentrations of A at extract and
that of B at raffiate are changed simultaneously after 20 switching times as shown in
Figure 3. For the control purpose, the prediction and control horizons are set equal to 5
and 2 switching periods, respectively. The weighting matrices are tuned by the trial and
error method. Here it is noticed that the control inputs act predictively to bring the control
output to their new respective setpoints. It is clearly seen that the control performance is
quite satisfactorily.

217
5 Conclusions

An SMB process is identified by using the subspace identification method. The


inputloutput data-based prediction model is used to obtain the prediction model. The
identified model exhibits an excellent prediction performance. The inputloutput
data-based predictive controller based on the identified model is designed and applied to
MIMO control problems for the SMB process under the presence of the input and output
constraints. The simulation results demonstrate that the controller proposed in this study
shows an excellent control performance not only for the di'sturbance rejection but also for
the setpoint tracking.

References

1. Wang, C.; Engell, S.; Hanisch. F.; Triennial World Congress, Barcelona, Spain
2002,1145-1 150.
2. Dunnebier, G.; Fricke, J.; Klatt, Karsten-Ulrich.; Ind. Eng. Chem. Res. 2000,39,
2290-2304.
3. Favoreel, W.;De Moor, B.; Van Overschee, P.; Gevers, M., Proceedings of the
American Control Conference 1999,3372-3381.
4. Kloppenbwg, E.; Gills, E.D. Journal ofProcess Control 1999,9,41-50.
5 . K.-Y. Yoo; H.-K. Rhee, AZChE J. 2002,48(9), 1981-1990.
6. Migliorini, G.; Mazzotti, M.; Morbidelli, M., Journal of Chromatography 1998, A,
827, 161-173.
7. Pedefem, M.; Zenoni, G.; Mazzotti, M.; Morbidelli, M., Chem. Eng. Sci., 1999, 54,
3735-3747.
8. Natarajan, S.; J. H. Lee, Computers and Chemical Engineering 2000,24, I 127-1 133.
9. Song, I.-H.; K.-Y. Yoo; H.-K. Rhee; Ind. Eng. Chem. Res. 2001,40,4292-4301.
10. Verhaegen, M.; Dewilde, P. Inr. J. Control 1992,5, 1187-1210.

218
QUICK AND COMPACT OZONATION USING SILICEOUS ZEOLITE

Hirotaka Fujita*, Akiyoshi S&da*, Taka0 Fujii* and Jun Inmi**


*Institute of Industrial Science, University of Tokyo
4-6-1 Komaba,Meguro-ku, Tokyo, Japan
**Nagasaki R&D Center, Mitubishi Heavy Industries, Ltd.
5-717-1 Fukahori-cho. Nagasaki, Japan
We developed a novel omnation using high silica mIites as an adsorptive concentrator of
ozone and processing organics, resulting in a significant increase in reaction rate. Throgh
TCE degradation test, it was found that the ozone reaction toward TCE was significantly
increased.

Key words: ozone. adsorption, high silica zeolite

Introduction
Drinking water resoufrces are inmasingly contmimted with chlorinated pollutants
such as hichloroethene (ICE), etc. ozonation is an oxidation process extensively applied to
water treatments for eliminating such pollutants. However, single omnation isn't always effective
in terms of time required for technical pcess. To enhance the omnation effectiveness, many
studies have focused on AOP ( a d v d oxidation pmcesm) such as omnation combined with
H&[ Beltran et al., 1998 I, etc, in which OH radicals, a much more reactive specie than omne
itself, play a main role. 'Ihe present study performs a new trial aiming at the enhancement of the
omne mction,designing a novel oxidationprocess in which the omne reaction rateis sigruficantly
increased through the use of an omne &orbent. We found in our pvious work [ Fujita et al,
ux)2] that high silica zeolitespcwxseda m n g abiity to adsorb omne. In addition to that,it has
been reposed that dissolved organics were also highly adsorbed onto these zeolites [ Giaya et al.,
20001. The above findings will raise the possibility that a considerable increase of reaction rate can
be achieved due to the adsorptive enrichment of omne and target Organics inside the zeolites. The
objectives of this work are to investigate the effectivenessof the novel omnation process proposed
and to elucidate the fundamental phenonena throughTCE degradation tests using a tubular flow
mtor..

Materials and methods lsMelIligb~zeolitesUsed


Adsorbents
employed are listed in Table
TCE was selected as a
model substance to be
decomposed and its adsorption property was examined by batch and brealahrough tests. In order to
investigate TCE degradation, the experimental appmlus shown in figure 1 was employed. Ozone
solution was produced by bubbling gaseous omne generated by an ozone generator (POX-10, Fuji
Electric Co., Ltd). A constant concenhation of TCE solution was prepad, miXing hued water
and high comntrationof TCE solution fed by a syringe type microfeeder (KDS-100, Kd Scientific
Inc.,). From different paths, Ozone solution and TCE solution were pumped into the mixing
vessel before the inlet to the fixed bed, using a dual plunger pump without pulsating c m n t

219
(NP-KX-120, Nihon Seinritsu Kagaku Co., Ltd), and this mixed solution was fed into the glass
column. TCE concentrations at the inlet and oudet of die column at a steady state were detected,
and die conversion of TCE, x, at a steady state was
given.
For die determination of TCE, samples
were withdrawn into an aqueous solution phase of
saturated SQs2 tor die destruction of residual
1
ozone, simultaneously extracting TCE in die
aqueous phase into hexane phase, and analyzed on
a gas chromatograph equipped with an electron
capture detector. For die analysis of chloride ion,
samples were mixed with a little amounts of
phenol solution for die destruction of residual
ozone, and analyzed using an ion chromatograph
(PIA-1000, Shimadzu GLC Ltd) equipped with a
Shim-pack IC-A3 column (Shimadzu GLC Ltd)
Fig. 1 Schematic diagram of
and a conductivity detector. The transfer phase was experimental apparatus employed for
the test of TCE deradation (J) Oxygen
an aqueous solution of 8mM p-Hydroxy benzoic tank @Ozonizer(DOzone stock solution
acid (Wako) and 3.2mM Bis (2-hydroxcyediyl) distilled water Mixing vessel Dual
plunger pump without pulsating
iminotris (hydroxymethyl) methane (Wako). The current syringe pump with high
concentration of TCE solution Packed
concentration of aqueous ozone was monitored column with an adsorbent Column
packed with wet activated carbon for the
with an UV spectrometer (UV-1600, Shimadzu degradation of exhausted ozone
Co., Ltd) at 258 run.

Result and Discussion


/. Adsorption Property of TCE
Adsorption isotherms of TCE are presented in Figure 2. The adsorption performances
were strongly influenced by SiQ/AljQ ratio and the pore structure of die zeolites. Amount
adsorbed increases in
loop
the following order
ZSM-5(SiOj/Al2O5ratio:3000)
Mordenite 100 ZSM-5(SiOj/Al2O3ratio:30)
ratio: 10) < A Mordenile(SiOj/Al2O3ratio:90)

Mordenite (SiQ/AlA
10 O Mordenitc(SiOj/Al2O3 ratio: 20)
D MordenitefSiOj/AljOjratio:10)
ratio: 20) < ZSM-5
ratio: 30)
Mordenite 0.1

ratio: 90) OrjP


0.01
3SM-5 (SiCVAlA 0.1 1 10

ratio: 3000). This order Equilibrium concentration [mg/L]

corresponded closely
Fig..2 Adsorption isotherms of TCE

220
with that of ozone [ Fujita et al., 2002 1.
2. Dep&&n of TCE
F d y , the ratio of
Chloride ion to TCE disappeared
was examined. Almost the
Stoichiornehic release of chloride
ion was found to occuc when
using ZSM-5(SiO#d2@ ratio
3ooo) , which pvided the
evidence that TCE was
decomposed in the column and
the steady state was achieved. D
TCE decomposition behaviors in Reaction time [s]
the column with and without the
Fig. 3 TCE degradation with and without siliceouszeoItte
&orbent were examined. As Wow rate:lO mL/mtn. Particle d i a m e k 5 0 0 - 5 9 0 ~ .
AqueousTCE concentratlon:1.27-3.0 mg/L)
shown in figure 3, much fastex
TCE decomposition was observed
in the column packed with ZSM-5 (SiWAl2Q ratio: 3000) in comparison to bulk reaction. The
d e w o n behavim when using different adsorbents were examined. TCE conversion increased
with higher adsorption performance of ozone and 'ICE, which indicates that the incmse of
reaction rate is likely due to the high enrichment of ozone and TCE inside the adsorbent With the
further increase of adsorption perfomance, TCE conversion converged to a certain value,
independent of high silica zeolite species. In other words, a maximum limit of the increase of
reaction rate existed. Apparent TCE demmposition behavior in the packed column would be
expressed in following Equation(l), which was based on simple mass balanceequation :
r,, = u -
d[TCE]
dz
[O, 1" [TCElp
=4 , (1)

For the determination of these unknown parameters, a , B , Kob,, the dependence of


decompsition upon ozone concentration and n=E concentration was examined using ZSM-5
(SiOJAlfi ratio : 3000) which gives the above-mentioned maximum limit. A semi-logarithmic
plot of fICEll4TCE]~vs z was found to be approximately b, which meatls a is
approximately equal to 1.0. As shown in f i p 4, ozone concentration has little effect on TCE
conversion, h m which we can determine pis approximately equal to 0. As a result, Equation(1)
is consequently changed into Equation(2).
x = 1-exp(- k,a,.r) (2)
Intereshgly, this results suggests that n=E conversion, x, was apparently i n f l u e d only by
reaction time r , not by ozone concentration, which differs h m the bulk reaction directly
influenced by ozone concenttation moigne et d, 1983 I. This suggests that there will be
conclusive factors that have a strongly effect on the apparent reaction behavior, independent of the
Charactenstl
' 'csof the adsorbent natu~. Thus,the effect of other factorssuch as particle size (biier

221
@cle size), flow rate
should be investigated in
order to elucidate ihese
phenomena with smallea
@cle size, hi*
conveasions~fwndto
-
be attained. 'Ihe effect of 0.2

the particle size on the


apparent reaction rate is ' 0 2 4 6 8 1 0

likely due to the Reaction time, r [sl


mass-transfer resistance. Fig. 4 El& of ozone concentrationon TCE degradation
Fkhemme,theincrease (Flow rate:loml/min.particle diameter: 5 0 0 - 5 9 0 ~ .
TCE concentratlon: 1.27-3.Omg/L)
of flow rate gave higher
TCE conversion, which
raised the possibility that film transfer resistance between the adsorbent surface and current w i l l
strongly influence the apparent d o n . The limitation of reaction enhancement was likely due to
the mass transfer limitation of tila Taking these into consideration, a kinetic TCE destruction
model was propod as follows. Apparent reaction rate r-hcan be described in Equation(3)
When ~ 1 is equal
, to 0, r-hcan give the maximum reaction rate, r h , , as described in
Equation(4)
r, = - h a , ([TCa - [TCEJ,) (3)
r,- =-k,a,[TCEI (4)
?his equation can be transformed kto the following Equation (5). k can be guessed by the
equation proposed by carbery et al[Caheny et aL, lW].and that proposed by W h n W h n et
al., 1%6]., while a,can be given by Fiquation(6).
x = 1-exp(- k,a,T) (5)

a, =- 3P,
%PP
We assumed that this equation would give the maximumvalue of TCE conversionobserved in the
experimental results when the adsorption performanas for TCE and ozone were high enough.
Close concordances were exjmxsed between the experimental results and the kinetic model. .As a
result, it can be deduced the adsohen& wexe kept almost in a virgin state. Aqueous ozone
mncentration has no effect on TCE conversion under our experimental condition as observed in
the experimental results due to the relatively fast reaction in the pores in comparison to TCE film
diffusion. However,such acase will be very specific. In the case that the substance to be degraded
is more Unreactive or is adwrbed less, or in the case that ozone CoIlCentration is much lower, other
states would occur.

222
CONCLUSION
Remarkable increase in ozone reaction rate toward TCE was observed in the presence
of high silica zeolite in comparison to bulk reaction. TCE conversion increased with the increase in
adsorption of ozone and TCE, which provided the evidence that this significant increase in reaction
rate is likely due to the high enrichment of ozone and TCE inside the adsorbent With further
increase of adsorption, TCE conversion converged to a certain value, which indicated a maximum
limit of reaction improvement exists. The kinetic model provides an accurate description of the
experimental results, which suggested the limit was due to the limitation of mass transfer in film.
This suggested that ZSM-5 (SiO/AlAratio: 3000) were kept almost in virgin state due to the fest
ozone reaction with TCE inside the adsorbent

NOMENCLATURE
w:Linear flow rate[ ms"1 ], L:Column lengthf m], r :Reaction time defined as L/u,[ s ], (TCE]:TCE
Concentration [ molm"3, or mgL"1 ], (TCE]i,:lhfluent TCE Concentration[ molm"3, or mg/L ],
[TCEL^Effluent TCE Concentration[ molm"3, or mg/L ], gLength[ m ], k^: apparent
Reaction rate constant[m3mor1s"1],[O3]:Ozone Concentration [ molm"3, or mg/L ], a,: Surface area
of adsorbent per unit volume[ nfari3 ], kL :Mass transfer coefficient between liquid and solidf ms"1 ],
rah,: Apparent reaction rate[ molL's"1 ], Dm :Difrusion coefficient of TCE in water[ mV ]

Reference
Beltran FJ., Encinar. J. M., Alonso M. A. (1998) A Kinetic Model for Advanced Oxidation
Processes of Aromatic Hydrocarbons in Water Application to Phenanthrene and Nitrobenzene, bid.
Eng. Chem. Res., 37,32-40
Carberry J. J,.(1960) A boundary-layer model of fluid-particle mass transfer in fixed bed. AIChE J.,
6,460-463
Fujita H., Sakoda A., Fujii T., Izumi J., (2002) Adsorption and decomposition of water-dissolved
ozone on high silica zeolite. Wat Res., submitted
Hoigne J., Bader H. Wat Res. (1983) Rate constants of reactions of ozone with organic and
inorganic compounds in water-1 non-dissociating organic compounds. Wat Res. 17,173-183
Wilke C. R., P. Chang. (1955) Correlation of diffusion coefficients in dilute solutioa AIChE J., 1,
264-270
Wilson E. J., Geankoplis CJ. (1966) Liquid mass transfer at very low Reynolds numbers in packed
beds. I&EC Fundamentals, 5(1), 9-14

223
TIME RESOLVED MULTICOMPONENTSORPTION OF LINEAR AND
BRANCHED ALKANE ISOMERS ON ZEOLITES, USING NIR SPECTROSCOPY

ALEXANDRE F. P. FERREIRA, M. MITTELMEIJER, M. SCHENK, A. BLIEK AND B. SMIT.


University of Amsterdam, Department of Chemical Engineering, Nieuwe Achtergracht 166, 1018
WV Amsterdam, The Netherlad
E-mail: a.ferreira@cience. uva.nl

Snap shot oE

- butane on MFI - iso-butane on MFI


Pressure Swing Adsorption (PSA) unit is a dynamic separation process. In order to create a precise
model of the process and thus an accurate design, it is necessary to have a good knowledge of the
mixtures adsorption behaviour. Consequently, the diffusion rates in the adsorbent particles and the
mixture isotherms are extremely vital data. This article intends to present a new approach to study
the adsorption behaviour of isomer mixtures on zeolites. In a combined simulation and experimental
project we set out to assess the sorption properties of a series of zeolites. The simulations are based
on the configurational-bias Monte Carlo technique. The sorption data are measured in a volumetric
set-up coupled with an online Near Infta-Red (NIR) spectroscopy, to monitor the bulk composition.
Single component isotherms of butane and iso-butane were measured to validate the equipment, and
transient volumetric up-take experiments were also performed to access the adsorption kinetics.

1 Introduction

Branched hydrocarbons are preferred to linear hydrocarbons as ingredients in petrol


because they enhance the fuel octane number. By catalytic isomerisation linear
hydrocarbons are converted into mono and di branched hydrocarbons, and it becomes
necessary to separate the mixture. A variety of zeolites may be used for this purpose,
either on the basis of sorption thermodynamics or on the basis of sorption kinetics. Such
data are relevant to the development of sorption based separation methods, but also they
provide key information regarding the catalytic isomerisation over zeolites themselves.
Zeolites are crystalline materials with a well-defined system of micro-pores. Zeolitic
materials are used in a variety of applications, one of the majors is in the area of
separation processes because of their unique porous properties. The size and structure of
the pores as well as the molecular structure of the adsorbate determines the adsorption
capacity and selectivity [S].
Simulation data on mixture adsorption can be used to screen zeolites as adsorbents,
but experimental data are necessary to validate the simulations and to accurately design
the separation process. The first step of the process design is to obtain such data. However,
the experimental assessment of multi-component adsorption equilibria and kinetics is not
straightforward and is highly time-consuming. As a result, some theories have been
developed that predict adsorption behaviour for a mixture based on the pure component
equilibria [1,3]. The isotherm data have to be correlated before their use in a design model

224
for easier handling. Therefore, the experimental systems have to be measured accurately
over a wide range of pressures and temperatures.
Gravimetric or manometric techniques have been used to establish adsorption data of
gases on zeolites. Both techniques present problems, manometric equipment has an
accumulation of the error; and data obtained by the gravimetric method are influenced by
effects associated with flow patterns, bypassing, and buoyancy. In the mixtures
adsorption behaviour, isomers mixtures have the highest degree of difficulty to study.
Isomers can not be differentiated in standard commercial adsorption equipment. This
problem has been solved in this study by coupling a manometric apparatus with an NIR
spectrometer, which allows us to measure the gas phase composition (in time, if
necessary). In this paper we report this new approach to study the adsorption of mixtures
of butane and iso-butane.

2 ExperimentalSection

In a combined simulation and experimental project we set out to assess the adsorption
properties of a series of zeolites. In the present work the adsorption properties of n-butane
and iso-butane on MFI are being studied. The experimental part consists in the validation
of the molecular simulation model, by confirming its results. The experiments were
performed in a constructed in-house manometric apparatus coupled with a NIR
spectrometer (Perkin Elmer, FT-IR system, GX Spectrum). Figure 1 is a scheme of the
experimental set-up.
Liquidpump

Figure 1. Scheme of the manometric apparatus coupled with the Figure 2. SEM picture of MFI.
NIR spectrometer.
A sample of commercial MFI from ZeoIist was used, with a silicon aluminium ratio
(SYAl) of 100, the crystals do not present a regular shape. On figure 2 Scanning Electron
Micrograph (SEM) of MFI crystals is presented. The sample was calcined for 6h at 873K.
The adsorptive gases used were n-butane with a 99.5% purity, and i-butane with
99.95% purity from Praxair. No further treatment was performed on them before their
admittance into the experimental set-up.
For single component and mixtures calibration curves, the NIR spectra of pure gases
are recorded at several pressures and the data gathered are treated as explained in the
result section.
To measure a point of the single component isotherm, pure gas is admitted to the set-
up with valve 2 closed (figure 1). By closing valve 1 and opening valve 2 the pressure

225
starts to drop, and the final value is measured after equilibrium is reached, and the loading
can be then calculated.
Transient volumetric up-take experiments are performed following similar procedure
of measuring one point of the isotherm, but before opening valve 2, NIR spectra are
recorded in regular intervals. After a short period valve 2 is opened to start the adsorption.
Spectra are recorded until the equilibrium is reached.

3 Results and discussion

Experimental data on single component adsorption isotherms of normal-butane and iso-


butane, on MFI zeolite, at 373K, for a pressure range of OSkPa to 200kPa,were obtained.
1.8 1.o

=-
2?
1.6
1.4
1.2
-
0.8
s0.7
0.9

5
;
1.0
0.8
3 0.6
> 0.4
E
v
-
E 0.6
0.5
0.4
3
0.3
0.2
0.2 0.1
on . 0.0
1 .OE-05 1 .OE-03 1.OE-0 1 1 .OEM 1 1.OE+03 1 .OE-05 1 .OE-03 1 .OE-01 1.OEM1 1 .OEM3
P(kPa) P(kPa)

Figure 3. nButane and i-butane isotherms on MFl at 373K. (- Mol. Simulations, 0 Exp. Results, - - - Exp.
Results ref [ 6 ] )
In the figure 3 we present data on single component adsorption isotherms and
simulation results. Data obtained from the literature [6] are included for comparison. The
increase of the loading observed on the high pressures region can be explained by
capillary condensation in the exterior secondary pore system, in particular between the
crystals. This has been observed also by other authors [5].
Using the presents used manometric set-up coupled with a NIR spectrometer
produced results that are in agreement with literature data, and in agreement with
simulations ( simulations details are provided on ref [4]).
Near Infra-Red spectroscopy is a non-intrusive technique that allows to monitor the
composition of the gas phase (differentiate isomers) and its changes in a time resolved
manner. In the NIR spectra (figure 4) some small differences between iso-butane and n-
butane can be observed. We can also observe that mixtures exhibit a spectral behaviour
that is a linear composition of the pure component spectra. It is necessary to quantify the
spectral differences of the two isomers, so that the composition of the mixtures can be
determined by NIR spectroscopy.
Spectra of pure n-butane and pure i-butane are recorded at several pressures. The
single component calibration curves can be calculated by integrating the spectra in a
certain wave length interval obtaining the total spectral intensity (T.S.I.) for that same
interval (figure 5). After some preliminary studies we conclude that the interval of
[6700,7350]+[8200,8700] cm-' is the best to use for the calibration, since it presents the
lowest degree of non-linearity. We did correct all spectra for baseline drift and water
bands. In the figure 5 we present the single component curves that were obtained.

226
0.8
h

v
0.6
0

0.4
e
w 0.2
9
0
5000 7000 9000
Wave lengh (cm-1)
Figure 4. n-Butane, i-butane and 50/50% mixture NIR spectra.

2 7 1
n-bu tnae i-butane
-1.5
:
- 1
-!
m
& 0.5 0 Experimental
-Calibration
-Cahbratio n
0
0 60 120 180 240 0 60 120 180 240
p Wa) P (kPa)

Figure 5. Single component calibration curve.

0.6 0.6
h n-butane h

j 0.5 4 0.5
f0.4
s
0.3
>
m 0.2
2 0.1
0
0 60 180 240 0 60 120 180 240
P (kPa)

Figure 6. Multi-componentcalibration curves - NAS value.


To calculate the amount of each component in one mixture we use the Net Analyte
Signal (NAS) theory [2]. In figure 6 we can see the NAS calibration curves for n and i-
butane.
In figure 7 we present the volumetric up-take experiment results. The time-base spectra
for n-butane are recorded. We can see a overlapping of spectra on the higher region, that
correspond to the initial part before opening valve 1, so the pressure is kept constant for
150s. On the lower part we see high number of spectra overlapping, that correspond to the
equilibrium. These two regions can also be seen on the total up-take curve, the line at
OmmoVg for the initial part between 0 and 15Os, and the flat part at 0.96moI/g for the
equilibrium, reached after 1000s.
Figure 8 shows adsorption data for n-butane and i-butane mixture. The total loading is
represented by the grey symbols, the black symbols represent the partial n-butane loading
and the open symbols the partial i-butane loading. We can observe that the separation
between the two isomers increases with the pressure, but we need to measure more points
in the high pressures region.

227
0.21 1.2 ~

Total Uptake -final pressure 3.90kPa


c?.
I
t = loo to 700s ,
3 0.14
0 v

P 0.07
n
3> 0.4
< 0.2 - .
0.00 0 7
6500 7500 8500 0 500 1000
Wave Lenght (cm-1) time (s)
Figure 7. nButane time-based spectra and total volumetric uptake versus time.

h
1.6 0.6
I -
Fartial Uptake h a 1 p s u r e 10.6kpa 1
f
E 0.8
v
U
0
$ 0.4
p 0.2
>
0.0 0
l.E-03 l.E-01 l.E+01 l.E+03 0 400 time (s) 800 1200
P(kPa)
Figure 8. Mmnm equilibrium data and partial uptake for mixture components versus time.

4 Conclusions

We conclude that with a manometric set-up coupled to a NIR spectrometer it turns out to
be possible to measure equilibrium and kinetic data for single components and mixtures.

5 Acknowledgements

This research was carried out within the project CW/STW 349-5203. The authors thank
the Stichting Technische Wetenschappen for their financial support.

References

1. Krishna R., Diffusion of Binary Mixtures in Zeolites: Molecular Dynamics


Simulations versus Maxwell-StefanTheory. Chem. Phys. Lr. 326 (2000) pp 477-484
2. Lorber A., Faber K., Kowalski R., Net Analyte Signal Calculation in Multivariate
Calibration. Anal. Chem. 69 (1997) pp 1620-1626
3. Ruthven D. M., Past Progress and Future Challenges in Adsorption Research. I n d
Eng. Chem. Res. 39, (2000) pp2 127-2131
4. Schenk M. et al., Sep. of alkane isomers by exploiting entropy effects during adsorp-
tion on silicalite-1: a CBMC simulation study. Lungmuir 17 (2001) pp 1558-1570
5. Stach H., Lohse U., Thamm H. and Schirmer W., Adsorption Equilibria of
Hydrocarbons on Highly Dealuminated Zeolites. Zeolites 6 (1986) pp 74-90
6. Zhu W. et d.,Adsorption of Light Alkanes in Silicalite-1: Reconciliation of Exp.
Data and Mol. Simulations. Phys. Chem. Chen Phys., 2 (2000) pp 1989-1995

228
PORE SIZE EFFECTS IN THE LIQUID PHASE ADSORPTION OF ALKANES
IN ZEOLITES

JOERI F.M. DENAYER', KURT DE MEYER', JOHAN A. MARTENS' AND GIN0 V. BARON'
'Departmentof Chemical Engineering, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussel,
Belgium. Joeri.deMverO.vub.ac.be Centerfor Surface Science and Catalysis, Katholieke
UniversiteitLeuven, KasteelparkArenberg 23, B-3001 Leuven, Belgium.

The liquid phase adsorption of C5-C22 linear alkanes on ZSM-5 was studied using a batch
adsorption technique. Saturation capacities of the alkanes depend strongly on the chain length. The
CHI packing density in the pores of ZSM-5 increases from pentane to heptane, then decreases steeply
to octane, and subsequently increases gradually to reach a plateau where the theoretically highest
packing density is observed. Binary adsorption experiments show pronounced selectivity effects
between alkanes with different chain length. For most binary mixtures, the longest alkane is adsorbed
preferentially, but for certain binary alkane mixtures, the adsorptionselectivity is inverted.

1 Introduction

The adsorption of the homologous alkane series on zeolites and other adsorbents has been
extensively studied by many research groups. These studies demonstrate the existence of
linear relationships between adsorption enthalpy and entropy, and the carbon number [ 1-
41. Contrarily, there are few studies dealing with the adsorption of alkanes in liquid phase.
This can be explained by the lack of selectivity effects that occur in the adsorption of
alkanes on adsorbents in liquid phase. Indeed, classical stationary phases for HPLC show
no separation of alkane mixtures, as a result of the rather weak interactions between the
molecules and the force field exerted by the surface of the amorphous material.
When the alkane molecules are trapped in the pores of a zeolite, much stronger
energetic interactions occur compared to those on amorphous surfaces. Pulse
chromatographic experiments in liquid phase showed a slight increase of the alkane
retention with carbon number on a column packed with a large pore Y zeolite [5]. When
ZSM-5 was used, a zeolite with much smaller pores than Y, very large differences in
adsorption between short and long alkanes were observed in liquid phase.
Among all zeolites, ZSM-5 is undoubtedly the most studied one, both from the
adsorption and catalysis point of view. Several experimental studies reported the
occurrence of a kink in the pure component adsorption isotherms of certain alkanes in gas
phase on ZSM-5 [6-81, and an inflection point in the adsorption enthalpy and entropy
versus zeolite loading curves [9,10]. Generally, the two-step behavior of these linear and
branched alkanes is interpreted in terms of the ZSM-5 channel geometry. For example, it
has been observed that at low partial pressures, isobutane is adsorbed preferentially in the
channel intersections of ZSM-5, while at higher pressures, these branched molecules are
"pushed" into the channel segments, resulting in a significant loss of entropy and a kink
in the adsorption isotherm [ 1 11. The packing and sitting of alkanes in the pore system of
ZSM-5 has been investigated by computational techniques, such as Configurational-Bias
Monte Carlo simulation [12-14]. These simulations all confirm the importance of entropy
effects in the adsorption of alkanes on ZSM-5.
In the present work, we have investigated the liquid phase adsorption behavior of n-
alkanes on a ZSM-5 zeolite. The saturation capacity of linear C5-C22 alkanes was
determined to study alkane packing effects. Adsorption isotherms of binary mixtures were

229
measured to determine the effect of alkane packing on the competition between short and
long alkanes.

2 Experimental

The H-ZSM-5 zeolite sample used in this study was obtained by deamoniating a NH4-
ZSM-5 from Zeolyst (CBV 8014, SiOZ/A1203 mole ratio of 80) at 673 K in a muffle oven
in presence of air for 48h.
Experimental adsorption isotherms were obtained by means of batch experiments.
Zeolite samples (-I g) were put in 10 ml glass vials and weighed with a balance. After
regeneration overnight at 673 K, the vials were immediately sealed with a cap with
septum in order to avoid water uptake from the air, and were weighed to determine the
regenerated zeolite mass. Mixtures of two linear alkanes or only a pure alkane in solvent
-
were added to the sample (each 20 ml). Iso-octane (99% purity, Acros) was used as a
solvent in the adsorption measurements. Immediately after sealing the cap and weighing
the sample, about 10 ml of the mixture was injected through the septum into the zeolite
containing vials, and another 10 ml was added to a vial without zeolite, to be used as
blank sample. Samples were kept at 277 K, to be sure no compounds could evaporate, and
stirred frequently.
Liquid samples were taken with 2 ml syringes after 24 h and 48 h, to verify if
equilibrium between adsorbed and bulk phase was achieved, and analyzed in a gas
chromatograph with flame ionization detector. For every binary mixture, a calibration line
was obtained by analysis of the blank samples, for which the concentration of the
components is exactly known. For each sample the amount adsorbed (qwhte) at
equilibrium was obtained by calculation of the mass balance.

3 Results and Discussion

Fig. 1 shows the uptake of pure hexane and decane from their mixture with iso-octane on
ZSM-5. The amounts adsorbed in the zeolite remain constant after 20 hours, indicating
that the experiments were performed under equilibrium conditions. In order to verify that
iso-octane can be used as inert solvent, a comparative experiment was performed in
which the binary adsorption isotherm of hexane and decane was determined using iso-
octane and 1,3,5-trimethylbenzene as respective solvents. The same adsorption isotherms
are obtained with both solvent (Fig. 1b), demonstrating that iso-octane does not interfere
with the adsorption of the linear alkanes, and that the relative adsorption of the short and
long n-alkanes is not influenced by the nature of the solvent, given that the solvent is not
able to enter the pore system.
Fig. 2a gives the saturation capacity of the C5-C22 n-alkanes, expressed in number of
molecules adsorbed per unit cell. Fig. 2b gives the total number of carbon atoms per unit
cell. For pentane and hexane, about 7.7 molecules are adsorbed per unit cell of ZSM-5.
This corresponds to 39 C atoms for pentane and 45 C atoms for hexane. Although the
number of adsorbed carbon atoms is the same for heptane as compared to hexane, only
about 6.4 heptane molecules are adsorbed per unit cell. A sudden drop is observed
between heptane and octane: only 4.5 octane molecules are adsorbed per unit cell. From
octane on, the number of C atoms adsorbed per unit cell increases steadily, to reach a
plateau of about 55 C atoms adsorbed unit cell.

230
0 20 40 m w l W l 2 0 1 4 0
LO 02 04 LB 08 10
U r n lh) xw

Fig 1a: uptake of hexane and decane in Fig 1b: Binary adsorption isotherm of C6
ZSM-5 fiom their mixture with iso-octane and C10. Open symbols: mesitylene
solvent; closed symbols: iso-octane solvent

. ....................... ..........

t i 1............................................................. 1
a 4

e 2

5 6 7 8 9 10 11 12 13 14 15 18 20 P 5 8 7 0 9 10 11 12 13 14 15 18 20 P
C.rbmnumbm Urbnn m b r

Fig 2a: Saturation capacities of n-alkanes on Fig 2b: Saturation capacities of n-alkanes
ZSM-5 (molecules/UC) on ZSM-5 (C-atoms/UC)

Obviously, the saturation capacity depends both on the dimensions of the adsorbing
molecules, and the geometry of the pore system. The pore system of ZSM-5 is constituted
of linear channels, with a fiee pore diameter of 5.6 x 5.3 A, intersecting with sinusoidal
channels, with a free diameter of 5.5 x 5.1 A (see Fig. 5a). The length of the linear
channel segments between two intersections is equal to 4.5 A, while the length of the
sinusoidal channel segments between two intersections equals 6.65 A. Each unit cell
contains 4 intersections, 4 sinusoidal segments and 4 linear segments. An octane molecule
has a length of 11.1 A, exceeding the length of a linear channel segment and an
intersection, but is also longer than a sinusoidal channel segment. As a result, not all
channel segments can be occupied by C8 molecules, explaining the lower saturation
capacity compared to C5-C7. Even with these shorter molecules, no use is made of all the
available space for adsorption. This is shown in Fig. 2b, where the total number of C-
atoms adsorbed per unit cell is plotted as a function of the carbon number. Only from C 13
on, a plateau is reached. When the number of adsorbed molecules per unit cell is
multiplied by their molecular length, a plateau value of about 70 A is obtained. This is
longer than the total length of the intersections, linear and sinusoidal channels (66.2 A),
which seems contradictory at fmt sight. However, it should be considered that the kinetic
diameter of the n-alkanes is 4.3 A, leaving an additional 1.1 A at every intersection for
adsorption. This gives a maximal length available for adsorption of 66.2 A + 4* 1.1 A =
70.6 A, which corresponds very nicely to the plateau value. This 100 % occupancy of the
ZSM-5 pores is only possible if the long molecules are able to bend, and occupy both
linear and sinusoidal channels.

231
1

0.8

0.6
E
0.4

02

0
0.0 0.2 0.4 0.6 0.0 1.O 0 02 0.4 0.8 0.8 1
XCO XCO

Fig 3a: Binary adsorption isotherm of CB Fig 3b: Selectivity diagram for the
and C 12 on Z S M J competitive adsorption of C8 and C 12
1.4
1

..
ao 02 0.4 0.8 08 10
0.0 0.2 0.4 0.0 0.8 1.0
XQ)
XQ)

Fig 4a: Binary adsorption isotherm of C6 Fig 4b: Selectivity diagram for the
and C8 on ZSM-5 competitive adsorption of C6 and C8

In Fig 3%the binary adsorption isotherms of a CB/C12 mixture is shown. C12 is adsorbed
in a very selective way from the mixture, as can be seen in the selectivity diagram (Fig
3b), in which x and y represent the molar firactions in the liquid and adsorbed phases
respectively. This selectivity for the longer chain can be explained by its higher
interaction with the zeolite.

Fig 5a: Pore structure of ZSM-5 Fig 5b: Packing of C6 and CS molecules
from their mixture in the pores of z5m-5

232
A peculiar behavior is observed with a C6K8 mixture (Fig 4), where, at low C6
concentrations (thus a high C8 concentration), the lightest component is adsorbed
preferentially over the heavier alkane. This adsorption selectivity reversal is again
explained by the packing of the molecules in the pores. Since octane can only adsorb in a
!+action of the channel segments, vacancies remain available in which C6 can adsorb, as
is shown in Fig. Sb, where an example is given how the octane molecules can pack in the
linear channels, leaving open space in the sinusoidal channels for hexane. Selectivity
inversion has been observed with a range of binary mixtures, but these data will be treated
elsewhere.

4 Conclusions

The adsorption of linear alkanes on ZSM-5 is governed by geometric and packing effects,
which result in pronounced selectivity effects, and even in an inversion of the normal
adsorption selectivity. These effects are certainly important with respect to catalytic and
separative applications, and will be studied in further work.

Acknowledgements
This research was financially supported by FWO Vlaanderen (G.0127.99). J. Denayer is
grateful to the F.W.0.-Vlaanderen, for a fellowship as postdoctoral researcher.

5 References

1. Bond, C.G., Keane, M.A., Kral, H., Lercher, J.A., Catal. Rev. - Sci. Eng., 42(3), 323-
383,2000.
2. Ruthven, D.M., Kaul, B.K., Adsorption, 4,269-273, 1998.
3. Eder, F.; Lercher, J.A. J Phys. Chem. B 1997,101, 1273-1278.
4. Hampson, J.A.; Jam, R.V.; Rees, L.V.C., Characterization of porous solids I1
Rodriguez, F. et al, Amsterdam 1991, 509,517
5. Denayer, J.F.M., Bouyermaouen, A., Baron, G.V., Ind. Eng. Chem. Res., 37 (9),
1998,3691.
6. Richards, R.E., Rees, L.V.C., Langmuir, 1987,3,335-340.
7. Zhu, W.,van der Graaf, J.M., van den Broeke, L.J.P., Kapteijn, F., Moulijn, J.A., Ind.
Eng. Chem. Res., 1998,37,1934-1942.
8. Sun,M. S., Talu, 0. Shah, D. B., J. Phys. Chem., 1996, 100(43), 17276-17280.
9. Yang, Y., Rees, L.V.C., Microporous materials, 12, 1997, 117-122.
10. Millot, B., Methivier, A., and Jobic, H., J. Phys. Chem. B, 1998, 102(17), 3210-
3215.
11. Zhu, W.,Kapteijn, F., Mouiijn, J.A., PCCP, 2000,2, 1989-1995.
12. Maginn, E.J., Bell, A.T., Theodorou, D.N., J. Phys. Chem., 1995,99,2057-2079.
13. Smit, B., Siepmann, J.I., J. Phys. Chem, 1994,98, 8442-8452.
14. Krishna, R, Paschek, D., PCCP, 2001,3,453-462.

233
DETECTION OF FREEZING POINT ELEVATION IN SLIT NANOSPACE BY
ATOMIC FORCE MICROSCOPY

M. MIYAHARA, M. SAKAMOTO,H. KANDA AND K . HIGASHITANI


Department of Chemical Engineering, Kyoto University, Kyoto 606-8501, Japan
E-mail: miyahara@cheme.kyoto-u.ac.jp

An experimental trial for finding the freezing point elevation phenomena was conducted, employing
the so-called colloidal-probe Atomic Force Microscopy. The elevated freezing point had been
predicted in the earlier molecular simulation work by the authors, which is thought to be caused by
the attractive potential energy from pore wall. To make up a strongly attractive nanospace, a
carbon microparticle was attached to the top of the cantilever tip, and its interaction force with
cleaved graphite was measured within a liquid cell filled with organic liquid, controlled at a desired
temperature above the bulk freezing point of the liquid. The two surfaces will form a slit-shaped
nanospace because the radius of the particle is far larger than the separation distance concerned.
For two kinds of liquids, freezing behavior has been detected above the bulk freezing point.
Though the extent of the elevation itself was rather small, the finding of the definite existence of the
elevation, not only in the micropores but also in a nanospace with the size of a few nanometers,
would be of much importance in the research field of the phase behavior in nanopores.

1 Introduction

Solid-liquid phase transition (fieezing) in confined space, which is of importance not only
in adsorption but in the nanomaterial fabrication, nanotribology and in the pore size
determination, has recently been explored by several research groups including ours. In
contrast to the long-believed phenomena of freezing point depression in pores[ 11, some
molecular simulation studies have clarified that the freezing point in slit nanospace in
equilibrium with saturated vapor or liquid in bulk can be higher than the bulk, depending
on the potential energy of the confining walls [2,3]. Note that the pore shape other than
slit geometry, and the equilibrium vapor-phase pressure less than the saturated one would
bring depressing effect in freezing point [4,5]. The present understanding of the freezing
in nanospace is roughly reviewed below.
Our study on the first point [2] clarified the following. Depending on the strength of
the attractive potential energy from pore walls, fluid in a slit pore in equilibrium with
saturated vapor showed freezing point elevation as well as depression, and the critical
strength to divide these two cases was the potential energy exerted by the fluids solid state.
The excess attraction relative to the critical one was considered to bring the confined
liquid to a higher-density state that resembles a compressed state, which would result in
the elevated freezing point.
The above result is in accord with other recent studies. Dominguez et al. [6]
examined freezing of LJ fluid in slit pores of purely repulsive and weakly attractive walls,
employing thermodynamic integral technique to determine true equilibrium points. The
freezing points showed significant downward shift, relative to the bulk, in purely repulsive
walls, while the downward shift was much smaller in magnitude for weakly attractive
walls. Further, Radhakrishnan and Gubbins [3] used a different approach of
determination of freezing point in slit pores employing the Landau free energy calculation,
and simple fluid in strongly attractive slit pore was shown to exhibit elevated freezing
points.

234
Now the research effort goes toward experimental verification of the elevation
phenomena in the simplest geometry, a slit. Our main interest is in the range of a few to
several nanomenters. Some experimental studies have already reported fieezing point
elevation in slit pores [7-91,but the materials used were activated carbon fibers (ACFs),
which have only micropores less than 2nm. In such small pores the first layer adjacent to
the attractive pore wall, which is known to form an ordered phase at a temperature well
above the bulk freezing point, will occupy most of the pore spaces, and the freezing
behavior in the interior of the pore space is difficult to be detected. Further, there may
still remain some controversy if a liquid confmed in a larger nanopore would exhibit
elevation unless an experimental verification is made for such a pore.
The employed technique for this purpose was the so-called colloidal-probe AFM
(Atomic Force Microscopy). The results clearly demonstrate that cyclohexane in the
nanoscale slit space between the carbonaceous solids freezes at a temperature above the
bulk freezing point. Though the extent of the elevation itself might look rather small, we
believe that the finding of the definite existence of the elevation would be of much
importance in the research field of the phase behavior in nanopores.

2 Experimental

An AFM apparatus, PicoSPM manufactured by Molecular Imaging (MI), which is


schematically shown in Fig.1, was used. The temperature of the liquid cell can easily be
controlled from the bottom side of the cell, onto which Peltier cooling device is equipped.
MI guarantees temperature control with 0.1 C accuracy.
A slit-like nanospace can be made up by applying so-called the colloidal probe
A F M technique, which was in its origin developed to measure the force between a solid
surface and a particle of micrometer size. The particle was glued on the cantilever tip,
and this colloidal probe is to be used instead of the usual cantilever. The vertical
scanning of the cantilever controls the distance between the particle surface and a solid
surface, and the usual manner of detecting the bending of cantilever gives the

AFy Piezo synner/Laser unit f Laser f

figure 1. Schm@ic drawing of AFM apparatus and image of the nanospace formed between surfaces.

235
forcedistance relation. If we look at the system with the separation distance of single
nanometers, the space between the two surfaces would be almost slit geometry because the
radius of the particle (order of 10' pn) is far larger than the separation distance.
A graphite plate (HOPG) and carbon particles, which were made from phenolic resin
by pyrolyzing above 2000"C, were used to form a nanospace with carbonaceous surfaces.
The graphite plate was freshly cleaved before measurement by using Scotch tape.
Cyclohexane, and octamethylcyclotetrasiloxane(OMCTS), whose bulk freezing point
were 6.4"C and 18C respectively, were used for the examination of freezing behavior.
The reasons of the choice were: i)aBity to carbonaceous surface, ii)freezing point near
ambient temperature, and iii)almost spherical molecular shape. Methanol, with bulk
6eezing point of -98"C, was also used as a reference.

3 Results and discussion

3. I Ultrahighsensitivity of vdWforce on temperature and its trick


Figure 2 illustrates the force curves between the carbon particle and graphite plate
immersed in cyclohexane at various temperatures above the bulk freezing point. (Note that
the distances shown in the figures are those detrmined from usual mannar of AFM
measurement, and is not the distances between the center of nuclei of surface atoms,
which are often used for simulation results.) As is expected the force at higher
temperature, e.g., at 18.6"C, exhibits typical van der Waals (vdW) force that acts mainly
in the single nanometer range. On the other hand what was NOT expected is its ultrahigh
sensitivity on the temperature. Only about 10 degree C of temperature change brought
multi-fold variation in the force. From theoretical point of view the effect of temperature
on the vdW force would be only from the variation in density of solids and liquid, which
at most would affect only a few percent in this small range of the temperature examined.

I '

0.3 0 17.0"C
c.l
E + ll.0"C
z 0.2
-s
A 10.0"C
E 9.0T
0.1 0 8.0"C

.
3

0-0.1
0

f.4 -0.2

-0.3I I I I I I I
0 10 20 0 10 20 30
Distance [nm] Distance [nm]
Figure 2. Force curves between carbon-graphite Figure 3. Force curves between carbon-graphite
surfaces immersed in cyclohexane for various surfaces immersed in methanol
temperatures above freezing point (6.4"C)
As a matter of fact, the same measurement in methanol gives the results shown in Fig.
3: Quite naturally the force curves do not show any detectable difference against the
temperature variation. We should, then, not take the abnormal temperature dependence

236
as is apparently shown, but should interpret it on the basis of the insensitivity of vdW
force on temperature.
The observed abnormality can be reasonably interpreted if we look at the trick in the
determination of the separation distance in the AFM measurement, which follows below.
Unlike the surface force apparatus, the AFM measurement does not have a direct
method to detect the distance between the surfaces. Instead, one will take the linear
signal of cantilever deflection against sample displacement as the origin of the surface
distance: The linear signal results because the two surfaces are in contact and move
together. This manner of wall detection usually works well. However, what would
result if the freezing point elevation is the case?
Suppose that the liquid between the two surfaces may freeze when the two surfaces
come close to a certain distance, at which the superposition of the potential energy from
each wall exceeds a critical strength that would be needed for the liquid to freeze. The
two surfaces, then, can never go any closer because of the steric repulsion of the frozen
phase (Fig. 4). The system replies with a linear signal as if it reached the situation of real
contact. The observed force curve would then represent a part of the real vdW force cut
out at this distance, taking this point as the origin of the surface distance, A weak
force thus appears.

0.3
CI

-
ul
a
o.2
0.1

.
3
0
8
0
-0.1
0
-0.2

-0.3
0 10 20
Distance [nm]

Figure 4. Schematic illustration of contact Figure 5. Shifted force curves and emerged
with frozen phase between surfaces. walls for cyclohexane.

3.2 Force curves against real separation distance: fieezing point vs. distance
relation
Now we try constructing the real force curves plotted against real separation distance.
The key assumption for this trial would be the insensitivity of the vdW force on the
temperature. The attractive force before the freezing or the apparent contact should stay
almost unchanged. Thus each force curve was slid laterally to exhibit a best fit in the
region of attractive force with long separation distance. The results are shown in Fig. 5.
Each force curve converges into a single curve in longer distance region, which is in
line with the insensitivity of the vdW force on temperature and stands for the validity of
the above assumed interpretation of the apparent phenomena. Then the walls standing
at certain distances for lower temperatures would be a direct reflection of the freezing at a
temperature above the bulk freezing point. Namely, the graphiticlcarbonaceous surfaces

237
in their nature exert strongly attractive potential energy to cyclohexane confined within,
and the superposition of the two potential energies grows with decreasing distance
between them. At the position where the wall stands in Fig. 5 , the potential energy in
the nanospace exceeds the critical value for the confined liquid to freeze. Thus the
distance of the wall gives the pore size needed for the freezing at each elevated
temperature. Quantitative discussion will further be presented in the conference.
We observed similar walls also for the case with OMCTS, though the page
limitation does not allow us to show the results here.
The superposition of the potential energy is typically the case for the micropores.
Note that, however, superposition of potential energy in some larger pores of a few
nanometers would still be sufficient to cause detectable elevation in the freezing point
because the potential energy of carbon surfaces itself shows quite a large value in the unit
of temperature when converted with Boltzmanns constant: It will be more than hundreds
of Kelvins in the vicinity of the surface and can easily be tens of Kelvins even with a
distance of a few nanometers.

4 Conclusion

An experimental trial for finding the freezing point elevation phenomena was conducted,
employing the so-called colloidal-probe Atomic Force Microscopy. A carbonaceous
nanospace with slit geometry was successfully made up by this technique.
The results demonstrated that cyclohexane in the nanoscale slit space between the
carbonaceous solids freezes when the distance comes down to 4 nm, even at 8.4C,which
is above the bulk freezing point of 6.4C. Measurements with octamethyl-
cyclotetrasiloxane, bulk freezing point being 18C, also detected the freezing behavior at
an elevated temperature of 22C. The detected elevation of a few degree C might be felt
as a matter of less significance, but we believe that the finding of the definite existence of
the elevation would be of much importance in the research field of the phase behavior in
nanospaces.

References

1. e.g., Patric W. A. and W. A. Kemper, J. Chem. Phys., 42 (1938) 369; Durn J. A., N.
J. Wilkinson, H. M. Fretwell, M. A. Alam and R. Evans, J. Phys. Cond. Matter, 7
(1995) L713.
2. Miyahara M. and K. E. Gubbins, J. Chem. Phys., 106 (1997) 2865.
3. Radhakrishnan R. and K. E. Gubbins, Mol. Phys., 96 (1999) 1249.
4. Kanda H., M. Miyahara and K. Higashitani, Langmuir, 16, (2000) 8529.
5. Miyahara M., H. Kanda, M. Shibao and K. Higashitani, J. Chem. Phys., 112 (2000)
9909.
6. Dominguez H., M. P. Allen, and R. Evans, Mol. Phys., 96 (1999) 209.
7. Kaneko K., A. Watanabe, T. Iiyama, R. Radhakrishnan and K. E. Gubbins, J. Phys.
Chem. B, 103 (1999) 7061.
8. Watanabe A. and K. Kaneko, Chem. Phys. Lett., 305 (1999) 71.
9. Sliwinska-Bartkowiak M., R. Radhakrishnan and K. E. Gubbins, Mol. Simul., 27
(2001) 323.

238
MODELJNG OF HIGH-PRESSURE EQUILIBRIUM ADSORPTION OF
SUPERCRITICAL GASES ON ACTIVATED CARBONS. DETERMINATION OF
PORE SIZE DISTRIBUTION USING A COMBINED DlV AND EOS
E. A. USTINOV
St. Petersburg State Technological Institute (Technical University). 26 Moskovsky prospect,
St. Petersburg, 198013 Russia
E-mail: eustinov@mail.wplus.net

D.D.DO
Department of Chemical Engineering, University of QueensIand,St Lucia, QLD 4072, Australia
E-mail: duongd@cheque.uq.edu.au

The density functional theory modified by including the Bender equation of state is applied to the
analysis of high-pressure (up to 50 MPa) adsorption of argon, nitrogen, methane and helium in
activated carbon Norit R1 in temperature range from 298 to 343 K. The approach allows us to
determine the pore size distribution (PSD) and the adsorbent density without recourse to helium
experiments. The PSD obtained with nitrogen at 77.35 K and those obtained with the high-pressure
argon, nitrogen and methane are in fair agreement with each other. However in the latter case the
PSD shows small pores of about 0.3 nm (physical pore width), which do not appear in the 77.35 K-
PSD due to diffusional limitations. It was shown that the adsorption of helium is not negligible at
room temperature and it cannot be used as an inert gas to determine the adsorbent density. The
approach can be easily extended to near critical gases such as carbon dioxide and ethane.

1 Introduction

Adsorption of supercritical gases at high-pressure has some features, one of which is that
the adsorption excess vs pressure exhibits a maximum. There are some simplified theories
involving Langmuir, Langmuir - Freundlich, Toth, Dubinin - Astakhov equations for the
absolute adsorption isotherm and 2-Dequation of state (EOS) to the adsorbed phase.
Some approaches were based on Ono - Kondo equations for lattice gas and the simplified
local density theory that utilizes the van der Waals EOS and the Peng - Robinson EOS
modified for the case of confined slit pores. Brief review of these models was presented
earlier [ 13. There are some investigations of supercritical and near to critical temperature
adsorption of gases on the basis of the density functional theory (DFT) [2,3] and the
grand canonical Monte Carlo (GCMC)simulations [4-71. In both cases the PSD obtained
using high temperature and high-pressure gas adsorption and that determined with
nitrogen at 77 K for the same sample of activated carbon was shown to be markedly
different. A group of small pores of the PSD obtained with near critical carbon dioxide
and supercritical methane is not revealed with the low temperature nitrogen adsorption,
which is attributed to diffusion limitations or even blocking these pores by frozen
nitrogen 161. By this reason the PSDs obtained with C02are considered to be more robust
and reliable than those obtained with N2at 77 K. Thus it might suggest that supercritical
adsorption is a suitable means to derive PSD. But there exists one obstacle to this
approach. Since the capillary condensation does not occur in supercritical adsorption the
dependence of amount adsorbed expressed in terms of excess per unit surface area on
pressure is nearly the same for sufficientlywide pores (wider than 6 collision diameters in
the case of methane adsorption [4]). It does not allow us to reliably determine the PSD in
the region of large pore widths and even the total pore volume. This problem may be

239
overcome by increasing the pressure range for experimental isotherms and specifying the
total pore volume, which could be evaluated from N2 adsorption at 77 K.
In this paper we use a modified DFT approach to the data on Ar, N2, CH4 and He
adsorption on activated carbon Norit R1 at pressures up to 50 MPa and four temperatures:
298, 313, 328 and 343 K. Experiments were carried out at Institute of Non-classical
Chemistry of University of Leipzig, Germany, using a magnetic suspended gravimetric
system (Rubotherm PrtlzisionsmesstechnikGmbH, Bochum, Germany) [11. Our aim was
to develope a method of the PSD analysis from high-pressure adsorption data without
recourse to helium adsorption to determine the skeletal density of the adsorbent.

2 Model

The widely used non-local density functional theory [8-101 is known to involve a number
of assumptions. The Helmholtz free energy of a fluid confined in a pore is divided by two
parts, which account for separately attractive and repulsive forces. The former is usually
calculated by integrating the Weeks - Chandler - Andersen (WCA) fluid - fluid potential
over the confined pore on the basis the mean field approximation (MFA). The latter is
modeled by the equivalent hard sphere fluid using the Camahan - Starling equation of
state (CS EOS). The Helmholtz free energy of the hard sphere fluid is supposed to be a
function of a smoothed density, which is defined as a weighted average according to the
T a w n a prescription [el. The solution for a given pore at a specified temperature and a
bulk phase pressure is obtained by minimizing the grand thermodynamic potential with
respect to the density profile across the pore. Mathematically the problem is reduced to a
system of highly nonlinear equations with respect to the set of densities at different
distances from the pore wall, which can be solved by an iteration scheme. In the sub
critical region one may obtain two or more density profiles depending on initial
conditions. In such a case the true density distribution corresponds to the minimal value
of the grand potential. In the case of supercritical gas adsorption only one density profile
is obtained at each bulk phase pressure. The parameters for the fluid - fluid interaction
(the potential well depth m, the collision diameter aff and the equivalent hard sphere
diameter dHS)are chosen in such a manner that the agreement between the bulk phase
properties (dependence of the saturation pressure and densities of coexisting phases on
temperature) and the surface tension and their respective experimental values is
reasonably well. The solid - fluid parameters are calculated by the empirical Berthelot
mixing rule, which sometimes can be corrected using the data in the Henry law region.
For the solid - fluid potential the 10 - 4 - 3 Steele potential is usually used.
Application of the NLDFT to the case of high-pressure adsorption, being principally
the same as in the case of sub critical adsorption, is however accompanied by an
additional difficulty. The problem is that the bulk phase density is comparable with that of
the adsorbate phase and must be described as a function of pressure very precisely. The
Carnahan - Starling EOS does describe this dependence for Ar, Nz and CH4 reasonably
well but its accuracy is not enough to lead to reliable results, especially in the PSD
analysis, which is known to be very sensitive to small experimental errors and hence to
small changes in the governing equations. Moreover, the bulk phase properties near to
critical gases such as COz and C2& the CS EOS cannot fairly correlate at all. The
question is whether the bulk gas could be described by any more accurate equation than
the CS EOS. In the case of high-pressure adsorption the contribution of the central part of
a sufficiently wide slit pore to the excess must tend to zero due to the solid - fluid
interaction becomes very weak, Obviously, it is possible only in the case when the same

240
EOS is applied to the adsorbed phase and the bulk phase. It suggests the following
scheme of a modification of the NLDFT. The Carnahan - Starling EOS is replaced by the
much more accurate Bender EOS [ 111 and then the corresponding molar Helmholtz free
energy is split into three parts: (i) the ideal component, (ii) the excess fiee energy
associating with only repulsive forces and (iii) the component accounting for only
attractive forces. The latter is assumed to be described by the WCA potential in just the
same way done in the original NLDFT. Then the excess free energy is obtained by
subtracting the ideal part and the attractive part (defined for homogeneous fluid) from the
molar Helmholtz fiee energy corresponding to the Bender EOS. At a specified
temperature the excess free energy is considered to be the single-valued function of the
smoothed density, which is defined according to the Tarazona prescription in the case of
inhomogeneous fluid.
The Bender EOS is given by
p = pRT + Bp2 + Cp' + Dp4 + Ep' + Fp6 + (G + Hp2)p3exp(-a,,p2) (1)
where p and p are the pressure and density, respectively. The constants B, ..., H are
temperature-dependent. Employing this equation, after subtracting the ideal component
and that accounting for fluid - fluid attractive interactions for the molar excess Helmholtz
free energy one can obtain:
F,(p) = Bp + Cp212 + Dp3 /3 + Ep4 / 4 + FpS15
(2)
+(Z/2)[G +ZH-(G +ZH+HpZ)exp(-a,,p2)]+C,p
where Z = l/azo,CO= (2'"16/9)lre&, and NA is the Avogadro number. The term --Cop is
the contribution of attractive interactions in the case of homogeneous fluid. In order to
determine the potential well depth Q and the collision diameter uffthe CS EOS was used
as a first approximation:
p = RTdl +(4<-2G2)/(1 -i)3]-C,oiNAp2, 6 = 7tdiSp/6 (3)
We fitted two parameters in this equation, namely dHSand the product q,u$ to achieve
the best agreement with the Bender EOS in the pressure range up to 300 MPa, which
approximately corresponds to the maximal smoothed density in the pore of I nm width at
50 MPa in the bulk phase. In the case of supercritical fluids it is not possible to evaluate
w and urn separately because we cannot invoke additional information such as the surface
tension. By this reason the hard sphere diameter was taken equal to the collision diameter.
The parameters fitted for 298.15 K are listed in Table I .
TaMe 1. Parameters for the Camahan - Starling equation of state at 298.15 K

Nz Ar CH4 He
dHS, nm 0.34764 0.32124 0.36177 0.21592
OK q&, nm3 K 4.1203 3.7780 6.9560 0.048202
M B 7 K 98.07 113.97 146.91 4.79

3 Discussion

Figure 1 gives a representation on primary experimental data in the form of dependence


of measured weight change on pressure in the case of methane adsorption. Buoyancy
contribution is very high and strongly depends on the activated carbon density, which was
considered as a fitting parameter. The experimental data shown in Figure 1 was treated by

241
the modified NLDFT approach and the PSD was determined using Tikhonov
regularization method. All isotherms at different temperatures were treated
simultaneously (in order to diminish sensitivity of the PSD to experimental errors) using
databases for local isotherms generated for these temperatures. The total pore volume
(0.58 cm3/g)was specified and taken from low temperature nitrogen adsorption. Once the
PSD and the skeletal density had been determined the excess isotherms were plotted.
Figure 2 shows excess argon isotherms at four temperatures.

0 1 0 2 0 3 0 4 0 5 0 0 10 20 30 40 50
Pressure (MPa) Pressure (MPa)

Figure 1. Methane adsorption on Norit R1. Figure 2. Argon isotherms on Norit RI.
Temperature, 'C: ( 0 ) 25; (m) 40; (A)55. Temperature, 'C: ( 0 )25; (m) 40; (A)55; (+) 70.
Solid lines in Figures 1 and 2 are calculated for the same PSD. The fitting is excellent,
which means that having determined the PSD from the analysis of one isotherm one can
predict isotherms of other temperatures with high accuracy.
Figure 3 illustratesthe PSD obtained with Ar, N2 and CH4.
Figure 3. Pore size distribution of Norit RI obtained
with different gases: (-) Ar, (-) N1, ( - x - ) CH+ The
pore width is physical, that is H - A, where A is the
separation of adjacent carbon layers in graphite.

It is seen that there are three groups of


pores centered at 0.3, 0.7 and 1.6 nm. The
PSDs obtained with different gases are
reproduced fairly well. The PSD analysis
of low temperature nitrogen adsorption
confirms the existence of the second and
third pore groups and does not reveal the
1 10
population of small pores. This is most
likely due to the diffusional limitations.
Pore width (nm)
Determination of the PSD with
helium is not possible due to small
amount adsorbed and low curvature of the isotherm. By this reason we used the PSD
determined with argon to calculate helium adsorption by the modified NLDFT approach.
In this case the only fitting parameter was the activated carbon density. Figure 4 depicts
the helium excess versus the bulk phase pressure. One can see that the excess amount is
not negligible at 25'C.

242
Figure 4. Helium excess isotherm on Norit RI at
25C. (Points) experiment; (line) calculation with
PSD determined with argon.
The contribution of helium adsorption
in the measured weight change is rather
small and at 50 MPa amounts to 4 percents
compared to that of Archimedes buoyancy
force, which makes the description of
helium adsorption isotherm not reliable.
Probably this is the reason of relatively
3 I/
o,o . I . 2oI . 3o . , , 5oI
I 1
large deviations between experimental
data and calculated curve. Anyway
Pressure (MPa) determination of the helium-solid
density is by no means correct at room
temperatures. The skeletal density
determined from adsorption measurements separately for each gas is 2.006,2.050, 2.1 57
and 2.017 g/cm for Ar,N1,CH4 and He, respectively. The mean value 2.058 g/cm3seems
to be very plausible.

4 Conclusion

A new approach based on a combination of the non-local density fbnctional theory and
the Bender equation of state was successfully applied to high-pressure (up to 50 MPa)
argon, nitrogen, methane and helium adsorption. The approach allows reliably
determining pore size distribution and adsorbent density, and most importantly it does not
require helium experiments to determine the skeletal density.

5 Acknowledgements

Support from the Australian Research Council is gratefully acknowledged. The authors
are gratefbl to Dr.Peter Harting for the experimental data.

6 References

1. Ustinov E. A. and Do D. D. J. Colloid Interface Sci. 250 (2002) pp. 49-62.


2. Quirke N. and Tennison S . R. R. Carbon 34 (1996) pp. 1281-1286.
3. Scaife S., Kluson P. and Quirke N. J. Phys. Chem. B. 104 (2000)pp. 3 13-3 18.
4. Gusev V. Yu., OBrian J. A. and Seaton N. A. Langmuir 13 (1997) pp- 2815-2821.
5. Heuchel M., Davies G. M., Buss E. and Seaton N. A. Langmuir 15 (1999) pp. 8695-
8705.
6. Sweatman M. B. and Quirke N. Langmuir 17 (2001) pp. 501 1-5020.
7. Sweatman M. B. and Quirke N. J. Phys. Chem. B 105 (2001) pp. 1403-141 1.
8. Tarazona P., Marconi U. M.B. and Evans R. Mol. Phys. 60 (1987) pp. 573-595.
9. Lastoskie C., Gubbins K. E. and Quirke N. J. Phys. Chem. 97 (1 993) pp. 4786-4796.
10. Olivier J. P.J. Porous Muter. 2 (1995) pp. 9-17.
1 1. Platzer B.; Maurer, G . Fluid Phase Equilibria 84 (1993) pp. 79-1 10.

243
ON THE PECULIARITY OF THE MINIMUM OF N-HEXANE PERMEABILITY
IN ACTIVATED CARBON

JUN-SEOK BAE AND DUONG D. DO


Department of Chemical Engineering, University of Queensland. St. Lucia, Qld 4072, Australia
E-mail: duongdacheque.uq.edu.au

For the tortuous and irregular capillaries of porous media, it has been reported theoretically and
experimentally that a minimum in the permeability of adsorbates at low pressures is not expected to
appear. In our study of n-hexane in activated carbon, however, a minimum was consistently observed
for n-hexane at a relative pressure of about 0.03, while benzene and CCIJ show a monotonically
increasing behavior of the permeability versus pressure. Such an observation suggests that the
existence of the minimum depends on the properties of permeating vapors as well as the porous
medium. In this paper a permeation model is presented to describe the minimum with an introduction
of a collision-reflection factor. Surface diffusion permeability is found to increase sharply at very
low pressure, then decrease modestly with an increase in pressure. As a result, the appearance of a
minimum in permeability was found to be controlled by the interplay between Knudsen diffusion and
surface diffusion for each adsorbate at low pressures.

1 Introduction

In diffusion and flow of adsorbing gases or vapors through capillaries, a minimum has
been reported in a plot of flow rate versus pressure when the mean free path (k) is
comparable to the capillary radius (r) [1,3,8]. For the tortuous and irregular capillaries of
porous media, on the other hand, the minimum is not expected to appear as reported
theoretically and experimentally [3,9,10]. In our permeation study of n-hexane through
activated carbon, however, a minimum in permeability was observed consistently at
relative pressure of about 0.03 [2], while for benzene and carbon tetrachloride no
evidence of the minimum appearance was found. Such an observation suggests that the
existence of the minimum depends on the properties of permeating vapors as well as the
porous medium. it has been known that the diffusion and flow of sub-critical vapors
through activated carbon is controlled by Knudsen diffusion, gaseous viscous flow and
surface diffusion [4,6]. We argue that the interplay of these three mechanisms may
determine the appearance of the minimum in total permeability, and present iqthis paper
a permeation model which is modified from the Do and Dos model [6].This permeation
model introduces a collision-reflection factor, which takes into account the adsorption
effect on Knudsen permeability at low pressures. In the model surface diffusion is treated
in such a way that the mobility of molecules on the surface increases with surface
concentration due to the reduction of energy barrier with the progress of adsorption. This
paper will present a better description of the three diffusion mechanisms and feasible
reasons for the minimum appearance in the total permeability of n-hexane through
activated carbon.

2 Problem Formulation on permeability

For a given pressure gradient across a porous medium, the mass balance equation can be
described as follows, provided that Knudsen diffusion, viscous flow and surface diffusion
are additive to the total flux.

244
where E is the porosity of the particle, P is the intraparticle pressure and C, is the
adsorbed phase concentration. At steady state the terms in the square bracket on the RHS
of Eq. 1 are corresponding to Knudsen diffusion permeability (Bd, viscous flow
permeability (By) and surface d i m i o n permeability (BJ,respectively.
The pore diffusivity (D,), which is assumed to follow Knudsen mechanism, takes Eq.
2 for non-overlapping cylindrical pore of radius r.

L(1
Dk 2r 8RgT 2 - f
D,=-=-zk 3rk - - f (2)
where f is the fraction of molecules undergoing collision to the solid surface over
reflection From the surface, which is usually taken to be unity in most work in the
literature. Since we have found that the factor f decreases with an increase of pressure
(Carman [3] observed similar behavior in the transient period), we propose the following
pressure dependence form for the factorf for strongly adsorbing vapors:
f = f, +(fb -f,>.exP[-a(P@o)I (3)
wherefo and fm are the values at the limit of zero pressure and at high pressure (close to
vapor pressure), respectively, which will be obtained directly from experiments. The
viscous flow parameter (Bo) for cylindrical pore is equal to ?/(8rv). The tortuosity factors,
'5k and r, are for diffusive and viscous flow, respectively, which will be calculated from
the permeability of non-adsorbing gases such as helium.
The surface diffisivity (D,) is known to be a function of adsorbed phase
concentration and is equal to the corrected diffisivity (D*$ multiplied by a
thermodynamic correction factor (8lnP/i?lnC,). Assuming that the driving force for
surface diffision is the gradient of the chemical potential and that the mobility constant is
a function of adsorbed phase concentration, Do and Do [6] have introduced the following
form for the corrected surface dihivity:

where Eo, is the activation energies for surface diffusion at the limit of zero pressure.
Dollmand Do, are the corrected surface diffisivities at the limit of zero pressure and at the
infmite temperature and at the permeation temperature, respectively. As the pressure
increases, adsorption occurs on progressively lower energy sites, resulting in the
reduction of energy barrier for diffusion. This is described by the term (I-&/( l+pC,)) in
Eq. 4. The larger the parameter p is, the faster E, decreases with adsorbed concentration.
When p=O, Eq. 4 reduces to the classical Darken equation [5].

3 Experimental

A differential permeation method was used to determine diffusion kinetics of strongly


adsorbing vapors through an Ajax activated carbon (type 976) (whose physical properties
[7]: particle density of 733 kg/m3, micropore porosity 0.40, macropore porosity 0.3 1 and
mean macropore radius 0.8 pm). An activated carbon pellet was carefully mounted in a
copper block, separating two reservoirs. One reservoir is much larger in volume than the

245
other, and is used as the supply reservoir. Thus the supply reservoir was maintained
practically at constant pressure during the c o m e of diffusion and adsorption.
The flux of each adsorbate through the porous solid was measured with respect to
time. Knowing the receiving reservoir volume (V), the cross-sectional area (A) and the
length (L) of the particle, we can determine the total permeability (BJ from Eq. 5 which
gives us a straight line passing through the origin with a slope of BdA$TNL) in a plot of
I~[(P,-P~O)/(P~-P~)] versus time.

where P1 is the pressure in the supply reservoir, and P2(t) and P; are the pressure in the
receiving reservoir at any t and at PO, respectively. The detailed experimental procedure
and the set-up of permeation measurement can be found elsewhere [2,6].

4 Results and Discussions

The total permeabilities of n-hexane, benzene and carbon tetrachloride through activated
carbon were obtained from Eq. 5 and shown in Figure la. At very low pressures a sharp
increase in total permeability was observed for all adsorbates studied. The adsorbed
concentration of these adsorbates [2] increases drastically over this very low range of
pressure, indicating that surface diffusion must play a significantrole in its contribution to
the total permeability since the contribution of gas phase diffusion is expected to be
relatively small. Interestingly, it can be clearly seen in Figure la that only n-hexane
exhibits a minimum in total permeability at reduced pressure of around 0.03. This
phenomenon was consistently observed at three different temperatures. In this range of
pressure, Knudsen diffusion and surface diffusion are expected to dominate the transport
of molecules since gaseous viscous flow is insignificant.
To elaborate the contribution of surface diffusion to total permeability with our
model, several parameters are required. Firstly, tortuosity factors for Knudsen diffusion
(~k) and viscous flow (T,) were found to be 5.66 and 3.10, respectively by comparing the
total permeability of adsorbates with that of inert gases whose flow is governed by
Knudsen d i h i o n and gaseous viscous flow. Secondly, the collision-reflectionfactorfo at
the limit of zero pressure has to be 1.79 for the theoretical Knudsen diffisivity (Eq. 2) to
agree with experimental data. Thirdly, the equilibrium isotherms of the three adsorbates
were well described by the Toth equation. With these parameters, the mass balance
equation (Eq. 1) was solved for receiving reservoir pressure with respect to time by the
combination of the orthogonal collocation method and a standard integration method. All
parameters in Eqs. 3 and 4 at 303K were obtained by matching the simulated results
against permeation results and listed in Table 1.
Table 1. Optimal parameters for n-hexane, benzene and carbon tetrachloride at 303K

n-Hexane Benzene Carbon tetrachloride


1.79 1.79 1.79
c
3

a
1.03
2.15 lo3
1.22
2.05 x 10'
1.15
3.16~10'
Dopo(m%) 5.86 x 7.36 1 0 - l ~ 1.10 x 1 0 - l ~
p (m3/mol) 2.30 x lo4 1.75 x lo4 2.55 x lo4
Eo,, (J/mol) 46900 50100 50000

246
One would physically expect that as pressure increases the solid surface may get
smoother due to the filling of small pores and cavities with adsorbed molecules, and as a
result the reflection time of gas phase molecules from the surface may become shorter.
The values off, in Table 1 are close to unity as expected and they are in an increasing
order of n-hexane, carbon tetrachloride and benzene. On the other hand, the parameter a
for n-hexane is much higher than that of the others. Since the parameter a in Eq. 3
represents how fast the Knudsen diffusivity increases with pressure, one would expect a
substantial contribution of the Knudsen diffusion for n-hexane to the total permeability at
very low pressures. Also the parameter j3 is a measure of how fast the activation energy
for surface diffusion decreases with adsorbed concentration. As Table 1 indicates, the
surface diffusion permeabilities of n-hexane and carbon tetrachloride are expected to
increase more sharply than that of benzene.

0.00 OlDp 0.01 0.08 0- 0.10

w w
I Id)
Carbon Taaachkrida

P,.",. [P.]

Figure 1. The permeabilities of strongly adsorbing vapors in activated carbon

Based on the optimal parameters in Table 1, the three permeabilities (BbB, and B,,)
are plotted in Figure 1 (b, c and d for n-hexane, benzene and carbon tetrachloride,
respectively) at 303K.The theoretical permeabilities calculated from 4 s . 1-4 for the
three adsorbates show very good agreement with those obtained from experiments. As
shown in Figure 1, at very low pressures the contribution of surface diffusion to total
permeability is considerable for benzene and carbon tetrachloride while Knudsen
diffusion is more dominant for n-hexane. Additionally it is interesting to note that the
surface diffusion permeability for the adsorbates studied is found to increase drastically at
very low pressures and then decrease moderately with an increase in pressure, which is in
agreement with out earlier theoretical study [4]. The fractional contribution of surface
diffusion to total permeability is the same order as the maximum adsorbed concentration,
confirming the fact that surface diffusivity is a function of adsorbed concentration.

247
Consequently, it can be clearly seen in Figure 1 that the interplay between Knudsen
diffusion and surface diffusion play an essential role in the appearance of a minimum in
total permeability.

5 Conclusions

The Knudsen diffusion, viscous flow and surface diffusion for strongly adsorbing vapors
are well described at low range of pressures in this paper. The collision-reflection factor
for Knudsen diffusion is found to be not constant but exhibit a modest increase with an
increase in pressure. The dependence of the Knudsen diffision for n-hexane on pressure
is stronger than that of the other vapors. Moreover the activation energy for the surface
diffusion of n-hexane exhibits a faster decreasing behavior in comparison with the others.
Conclusively, the reason for the minimum appearance in the total permeability of n-
hexane can be attributed by the interplay between the Knudsen diffusion and surface
diffision.

6 Acknowledgements

Support from the Australian Research Council is gratehlly acknowledged.

References

I. Adzurni H.,Studies on the flow of gaseous mixtures through capillaries: 111. The flow
of gaseous mixtures at medium pressures. Bull. Chem. SOC.Jap. 12 (1937) pp. 292-
305.
2. Bae J . 4 . and Do D. D., Study on diffusion and flow of benzene, n-hexane and C C 4 in
activated carbon by a differential permeation method. Chem. Eng. Sci. in press (2002).
3. Carman P. C., Diffusion and flow of gases and vapours through micropores I. Slip
flow and molecular streaming. Proc.Roy.Soc.(London) A203 (I 950) pp. 55-74.
4. Do H. D. and Do D. D., A new diffusion and flow theory for activated carbon from
low pressure to capillary condensation range. Chemical Engineering Journal 84
(2001) pp. 295-308.
5. Do H. D., Do D. D. and Prasetyo I., On the surface diffision of hydrocarbons in
microporous activated carbon. Chemical Engineering Science 56 (200 1) pp. 435 1-
4368.
6. Do H. D., Do D. D. and Prasetyo I., Surface diffusion and adsorption of hydrocarbons
in activated carbon. AIChE Journal 47 (2001) pp. 2515-2525.
7. Gray P. G. and Do D. D., Dynamics of carbon dioxide sorption on activated-carbon
particles. AZChE J 37 (1991) pp. 1027-34.
8. Knudsen M., Die Gesetze der Molekularstr6mungund der inneren Reibungsstr6mung
der Gase durch R6hren. Annalen der Physik (Leipzig) 28 (1909) pp. 75.
9. Pollard W. G. and Present R. D., On gaseous self-diffusion in long capillary tubes.
Phy~.Rev. 73 (1948) pp. 762-774.
10. Wicke E. and Vollmer W.,Flow of gases through micropores. Chem. Eng. Sci. 1
(1952) pp. 282-291.

248
SIMPLIFIED EXPERIMENTAL METHOD TO ANALYSE INTRA-ACTIVATED
CARBON PARTICLE DIFFUSION BASED ON PARALLEL DIFFUSION
MODEL

Y. MIURA, Y. OTAKE,S.IWASAWA AND E. G. FURUYA


Dep. lnd. Chem.,Meqi Universiy, Tama-Ku, Kawasaki, 214-8571, Japan
E-mail: egfuruyac@isc.meiji.ac.jp

H. T. CHANG AND N. KHALILI,


Dep. Environ, Chem. Eng., Illinois Inst. Tech., Chicago, 6160-37936, IL, USA
E-mail: khalili@iit.edu

The rate of contaminant adsorption onto activated carbon particles is controlled by two parallel
diffusion mechanisms of pore and surface diffusion, which operate in different manners and extents
depending upon adsorption temperature and adsorbate concentration. The present study showed
that two mechanisms are separated successfully using a stepwise linearization technique
incorporated with adsorption diffusion model. Surface and pore diffisivities were obtained based
on kinetic data in two types of adsorben and isothermal data attained from batch bottle technique.
Furthermore, intraparticle difisivities onto activated carbon particles were estimated by traditional
breakthrough curve method and final results were compared with those obtained by more rigorous
stepwise linearization technique.

Introduction

Granular activated carbon (GAC) has been used widely in environmental pollution
control due to its capability to absorb a broad range of organic and inorganic compounds.
To properly design a GAC column, information regarding accurate kinetic data of
intraparticle diffusion rate parameters is required. Intraparticle diffusion is, however, a
complicated process that includes at least two parallel mechanisms: pore diffision and
surface diffusion [7].
In this paper, we will present a procedure to separate pore diffusion and surface
diffusion in GAC adsorption by use of a stepwise linearization. Furthermore, a simplified
method to estimate concentration dependency of apparent diffusivities from breakthrough
curves will be proposed.

Theoretical Considerations

An intraparticle diffusion model including surface and pore diffusion has


been presented by other researchers [2,3,4,61. The mathematical equation for
parallel diffusion model is expressed as

On the other hand, a model for apparent intraparticle diffusion 111 is


(2)

This model was developed for its simplicity to avoid the need to determine

249
surface and pore djffusivity separately.
The relationship between apparent diffusivity (D3,surface diffusivity
(Dh and pore diffusivity (D,)can be derived by comparing (1) and (2).

The limitation of Equation (3) is that the isotherm slope @/& is a h c t i o n of


solution concentration; therefore it may not be a constant over a range of concentration
normally encountered in kinetic tests. It is also noted that D,and D,are a function of solid
concentration q, while D,is independent of adsorbate concentration.
Neretnieks 151 studied the effects of temperature and concentration on
surface diffusion, using a pseudo-linear technique to approximate the slope of
isotherm as depicted in Figure 1. An equilibrium point (c-2 and q=qe2) was
connected to an origin (c=O and q=O)by a straight line in order to approximate
actual slope of the isotherm at ce2. I t is obvious that his pseudo-linear
technique offers poor approximation for highly nonlinear isotherms, which
are often found in industrial applications.

Y
Liquid Concentration, c

Figure 1. Comparison Neretnieks and Stepwise Isotherm Linealization Technique


Using a stepwise linearization technique developed in this study can minimize the
problem associated with the Neretnieks technique. This technique uses a straight-line
section by connecting two isotherm points [(c,,, q,,) and (C,~,qe2)]to approximate the
actual isotherm as presented in equation (4).

The slope obtained by the stepwise method is, therefore, more representative of the actual
slope than that by Neretnieks technique. As seen in Figure 1, the slope (f3) in the
adsorption isotherm curve becomes gentle in high concentration range. Consequently, DS
becomes negligible in comparison with the pore-diffision term Dp/Pppaccording to
Equation (3).

Materials and Experimental Methods

The adsorbent employed in this study was granular activated carbon CAL
(GAC-CAL) manufactured by Toyo Calgon Ltd. The adsorbates used in this study
were pnitrophenol (PNP) and pchlorophenol (PCP), both of which were obtained fiom
WAKO Chemical Co. The concentration of the adsorbate solution was determined by UV

250
absorbance using an WNIS spectrophotometer (Shimazu UV-240). Three types of
adsorbers were used in the present study. For rigorous discussion, Differential Reactor
(DR, shallow bed) and Completely Mixed Batch Reactor (CMBR) were employed. With
aid of curve-fitting procedure, apparent diffusivities (D,)were obtained from uptake
curves in DR and from concentration decay curves in CMBR. Long Column Bed reactor
(LCB), commonly used in industrial application, was also studied to determine the
diffusivitiesby use of Mass Transfer Zone (MTZ)method from breakthrough curves.

Results and Discussions

The correlationsbetween D, and 1@ for PNP and PCP obtained in DR and CMBR are
presented in Figure 2, indicating that the apparent diffisivity is a strong function of the
adsorbate molecule and temperature.

XlW' ~10-7
25 I I , I I , #

20 . :g
DRCMBR
.
i
. . . .I
i
.
!
...i .....-i .......i ....i
.
5,
8

i.
joeg<.O i
* .i* . i .

XI
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

'18 Wm3 1/B Wcm?


(a) PNP - CAL systems (b) PCP - CAL systems
Figure 2. Effective Diffusivity as Function of i@.

Figure 3 also includes the correlation between D, and l/P for PNP in of traditional
long column bed (LCB). It is noted that D, in LCB was determined by use of the MTZ
method.

0.30 I I I I

0.25 -................................................
oi
.-.-.W.i.-.-.- 1-.-.a '.-_-.-. +.-._..
Dflm = 0.235

0.20 _. Dflm 0.22 ( D W M B R ) i


................................................

0.15 .............-...;.--.
_. PNP -
i LCB 0
: DWMBR .---*-

0.10 I I I I
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

'4 Is/cm3 Temperature plcl

Figure 3. Effective Diffusivity method Figure 4. Ratio of Pore Diffusivity to for LCB
Molecular Diffisivity

251
The slopes in Figures 2 and 3 were used in order to calculate the pore diffusivities (D,),
which are independent of adsorbate concentration, by use of Equation (3).
The ratios of Dp to molecular difisivities Dm are plotted against adsorption
temperatures in Figure. 4. The Dflm values are expected to be constant for a particular
adsorbent, irrespective of types of adsorbates and adsorbers over a range of temperatures.
From Equation (3), the value of the surface diffisivity Ds were calculated by
subtracting the pore-diffusivity term Dp/Pppfrom the apparent difisivity D,. The value of
D@, is then plotted against D, for DR and CMBR methods (Figure 5-a) and for LCB
method (Figure 5-b). It can be pointed out from Figure 5-a that that (i) DR and CMBR
methods yield similar estimate of D, and that (ii) intraparticle diffusion of PCP into GAC
is dominated by surface diffusion compared to PNP.

..........
1 !\AA..... ;
....................

-
) ....................
g 0.4 ......,.....
5? I..

n
Len-. i py- OR CUBR ffiB
... ......... i.......
. i ....
. i...:...:..:..:
. . . .....
soa. 0
0.2
.* . . . .. .. .. .. m- .
. . A
a m
AO A
318% 0 : : i i i i i 318%eO
0.0 x10l8 0.0 I I I I l l 1 XlOd
3 4 5 6 8 1 0 M M 3 4 5 6 8 1 0 2 0 M
4 lcn+4 4 ldkl
(a) DR and CMBR Method (b) Traditional LCB Method
Figure 5. Contributionof Surface Diffisivity to Molecular Diffisivity

It is also seen fiom Figure 5-b that the extents of Ds contribution to D, evaluated on
LCB method almost coincides with those for DR and CMBR, implying that LCB method
can be applied for the determination of concentration dependency of D,.

Conclusions

The kinetic adsorption studies in different types of adsorbers were performed with
two phenolic compounds of PNP and PCP on activated carbon. A technique of isotherm
stepwise linearization has been proposed and applied to approximate nonlinear isotherms
for GAC adsorption. The results showed that pore and surface diffisivity are estimated
satisfactorily using this stepwise linearization technique. This study also showed that the
apparent difisivity (D,),which possesses concentration dependence, could be estimated
on LCB by applying the technique in high-adsorption region.

References

1. Bird, R. B., Stewart, W. E., and Lightfoot, E. N. (1960). Transport phenomenon.


John Wiley, New York, N.Y., 542-544.
2. Crittenden, J. C., et al. (1993). Water Res., 27,725.
3. Fleck, R. D., Kirwan, D. K., and Hall, K. R. (1973). Ind. Eng. Chem. Fundam.. 12,

252
95.
4. Mansour, A. D., von Rosenberg, D. U., and Sylvester, N. D. (1982). AlChE J., 28,
765.
5. Neretnieks, I. (1976). Chem. Engrg. Sci.. 31, 1029.
6. Noll, K. E., Gounaris, V.. and Hou, W. S. (1992). Adsorption technology for air and
water pollution control. Lewis Publishers, Chelsea, Mich., 1 - 20.
7. Weber, W. J. Jr, ed. (1972). Physicochemical processes John Wiley, New York, N.Y.,
199-259.

253
IN-SITU CHARACTERIZATIONOF ION ADSORPTION
AT BIOMIMETICAIRlWATER INTERFACES

T.Y.KIM,G. S.LEE,ANDD. J . A H "


Department of Chemical and Biological Engineering,
College of Engineering, Korea University,Seoul 136- 701, Korea
E-mail: ahn@korea.ac.kr

Biomimetic fatty acid Langmuir monolayers at the aidwater interface were investigated both by an
in-situ near-normal external reflection and by an ex-situ transmission FTIR methods. The
monolayers consisted of reactive fatty acids and nonreactive fatty alcohols to cadmium ions, and the
relative amount of each component was varied. For the monolayers interacting with pure water
showed the composition same for both methods regardless of the amount of reactive acid moieties.
However, when the cadmium ions at a concentration of ImM in the aqueous subphase (pH=5.5)
adsorbed to the monolayers, the in-situ method revealed that ion adsorption was significantly less
than full coverage relative to available reactive moieties in case of the amount of nonreactive ones
exceeding 20%. On the contrary, the ex-situ method showed nearly full coverage of cadmium
adsorption regardless of the monolayers' composition. This discrepancy occurs most probably
because the ex-situ analyses requires the sampled monolayers to be neutralized upon drying by an
additional adsorption of cadmium ions, which does not reflect the state of the monolayer at the
&/water interface. In-situ analyses showed that the adsorption of cadmium ions requiring 2:l
covalent complexation became to be restricted as the nonreactive alcohol molecules neighbored with
the reactive sites.

1 Introduction

Adsorption of ions at surfhces has been one of the classical fields of interest in regard
of catalysis and separation. Among various surfaces, Langmuir monolayers mimic the
structure of cell membranes and hence these biomimetic surfaces are a model system that
enables one to investigate surface phenomena analogous to the membrane surfaces [1 21.
Adsorptivity of various ions to Langmuir monolayers has been known to be quite
different fiom that in the bulk phase after extensive studies done by theoretical models and
empirical techniques involving X P S , neutron scattering, FTIR, etc [3, 41. Most of the
empirical studies were done ex-situ because there exists difficulties in dealing with the
monolayers at the aidwater interface in-situ. Recently, much progress has been reported
especially on the in-situ FTIR spectroscopy for the aidwater interfaces [5,6].
In the present study, we report new results on in-situ near-normal external reflection
FTIR on the monolayers containiig both reactive and inactive moieties to adsorption of
ions. Unlike the conventional ex-situ analysis, it was capable of observing suppressing
effects of mixed monolayer surfaces to ion adsorption.

2 Experimental

Compounds used to form Langmuir monolayers were stearic acid (>99.9%) and
stearyl alcohol (>99.9%) purchased fiom Aldrich. Stearic acid known to bind with
cadmium ions (CdC12, >99.9%, Aldrich) were mixed with stearyl alcohol at varying
compositions fiom 1OO:O to 50:50. Each solution was at the concentration of 1mM in
chloroform (>99%, Fluka).

254
Each mixed solution was spread using a micro syringe on a KSV minitrough
(Finland) containing deionized (DI) water or 1mM aqueous CdCl2 solution both at pH of
5.5. After evaporating chloroform for 30min., the molecules were compressed to form the
close-packed monolayer at the surface pressure of 2OmN/m. After waiting 3Omin to reach
the equiIibrium, the monolayer was deposited onto a clean CaFz substrate for the ex-situ
FTIR analyses by the skimming method to make sure the sampled film to represent the
original monolayer at the aidwater interface. The sample was blown dry by the nitrogen
gas stream, and then scanned 512 times with a FTIR spectrophometer (Perkin-Elmer
Spectrum GX) equipped with a liquid-nitrogen cooled MCT detector at the transmission
mode with an unpolarized i n h e d beam.
For the in-situ FTIR analyses, mixed solutions were spread by the same fashion onto
a house-made minitrough that fitted into the sample chamber of the FTIR instrument. For
every in-situ experiment, the close-packed monolayer was initially formed on DI water
that was then replaced with lmh4 CdC12 solution by careful circulation of the subphase
using a peristaltic pump. The background spectrum was taken for a bare water surface
when the sample was the monolayer on pure water, and was taken for a close-packed
monolayer on pure water when the sample was the one on the cadmium-containing
subphase. The latter technique enhanced the carboxylate peak relating to ion adsorption.
All the experiments were done at room temperature.

3 Results and Discussion

The ex-situ transmission FTIR results of the mixed monolayers spread on pure water
@H=5.5) are shown in Fig. la. The spectra are adjusted so that methylene stretching
bands (v,CHz, 2918cm-' and vSCH2, 285Ocm-') have the same intensity height. The
carbonyl band (vC=O, 1705-173Ocm~')representing the acidic state of the stearic acid is
not interfered with peaks arising fiom the stearyl alcohol. As the relative amount of the
alcoholic moiety increases the peak area decreases as expected fiom the monolayer
composition, as shown in Fig. lb. It is also noted that the carbonyl band shifts fiom
1705cm" to 173Ocm-', which indicates that lateral hydrogen bonding among carboxylic
head groups becomes to be weaken as the stearyl alcohol molecules are mixed and
inserted among the stearic acid molecules. On another set of experiments of surface
isotherms (of which results not shown), the surface Gibbs fiee energy implies the two
components to be well-mixed at the surface pressure of 2OmN/m at room temperature.

255
S Fnction of steuyl llcohol b 8 c t i v e molecules)
0 i o m a o ~ 0 ~ 8 0 m m o o t m
Tbeoretic.1 c.1cUhtion
0 vc-0

0.2 -
iw m 80 70 80 50 H) so m 10 o
0 Fraction of siearic acid (reactive molecules)

(a) (b)
Figure 1. Transmission FTIR results on the mixed monolayers spread on pure water with varying monolayer
composition (stearic acid : strearyl alcohol) (a) and the peak area of the carbonyl band at various composition
(b). The peak areas were normalized with respect to that of pure straric acid case (1OO:O).

The ex-situ transmission results on the mixed monolayers spread on the subphase
containing 1mM CdClz (pH=5.5) are shown in Fig. 2a. At this high concentration of
cadmium ions, every acidic molecule is bound with cadmium ions as indicated by the
complete disappearance of the carbonyl band and the existence of the asymmetric
carboxylate band (v,COO, 1542cm-') at the same time. In general, the peak intensity of
the asymmetric carboxylate band decreases as the relative amount of the reactive stearic
acid decreases, as observed in the case of Fig. 1.
The in-situ near-normal external reflection results on adsorption of cadmium ions to
mixed monolayers are given in Fig. 2b. Since the background spectra are taken for the
close-packed mixed monolayers on pure water at each composition, the methylene
stretching bands are nearly cancelled out. By contrast, the asymmetric carboxylate band
appears with reasonable S/N ratios even without a polarizer-modulator set-up [7].It is
noted that the near-normal incidence of the i n h e d beam resulted in negative signals on
the contrary to that of the grazing-angle incidence.
It is interesting in the in-situ investigation (Fig. 2b) that as the relative amount of the
inactive stearyl alcohols to cadmium ion exceeds 20% the asymmetric carboxylate band
suddenly decreases in its intensity and remains about the same afterwards. In Fig. 2c, it is
clearly seen that the adsorbed amount of cadmium ions at 80% of stearic acid is reduced
by more than half of those at larger compositions of the acid. By contrast, the ex-situ
results showed a linear decrease in cadmium adsorption with the monolayer composition
as well as a certain deviation fiom the theoretical calculation that is different to the case
without adsorption of cadmium ions.
(c)
Figure 2. Ex-situ transmission FTDR results (a), in-situ near-normal external reflection FTIR results (b) on the
mixed monolayers spread on aqueous cadmium solution ( I d , pH4.5) with varying monolayer composition
(stearic acid : streayl alcohol), and the peak area of the asymmetric carboxylate band at various composition (c).
The peak areas were normalized with respect to that of pure strearic acid case (1OO:O).

It is strongly suggested that introduction of more than 20% of inactive alcohol


molecules into acid monolayers hinders significantly the 2: 1 covalent complexation
among cadmium ions and carboxylic head groups. When the monolayer is sampled on the
solid substrate and dried, additional neutralization of carboxylic groups with cadmium
ions are enforced, which results in higher amount of adsorption complexation that is yet
less than the complete adsorption coverage as calculated.

4 Conclusions

We investigated the adsorption of cadmium ions on mixed monolayers at the aidwater


interface using both ex-situ and in-situ FTIR spectroscopic methods.
The ex-situ investigation with the sampled monolayers did not always reflect the
original monolayers at the aidwater interface in analyzing adsorption of cadmium ions. It
was newly observed by the in-situ method that the adsorptivity of cadmium ions to acidic
monolayers could be abruptly reduced when the inactive alcohols were mixed into at more
than 20% in composition.
The present results offer usehl information on further characterization of binding of
biologically active ions to cell membranes and also on designing sensors and catalysts at

257
the molecular level.

5 Acknowledgements

This work was supported financially by the Korea Science and Engineering Foundation.

References

1. Vijendra K.A.,Langmuir-Blodgettfilms, Phys. Today (1988) pp. 40-46.


2. Swalen J.D., Allara K.L., Andrade E.A., Chandross E. A., Garoff S., Israelachvili J.,
McCarthy T.J., Murray R., Pease R.F. and Rabolt J.F., Molecular Monolayers and
Films, Langmuir 3 (1987) pp. 932-950.
3. Ahn D.J. and Franses E.I., Ion adsorption and ion exchange in ultrathin films of fatty
acids, AIChE 40 (1994) pp. 1046-1054.
4. Kim T.Y., Lee G.S. and Ahn D.J., Ion separation of binary aqueous solutions at
acidic Langmuir monolayer surfaces, Korean J. Chem. Eng. 18 (2001) pp. 977-985.
5. Geriche A., Huhnerfm H., In situ investigation of saturated long-chain fatty acids at
the adwater interface by external infrared reflection-absorption spectrometry,J. Phys.
Chem. 97 (1993) pp. 12899-12908.
6. Ren Y., Hossain M.M., Iimura K.I. and Kato T., CH3(CH2)COOH/Cd2+system on the
aqueous cadmium acetate solution investigated in situ by polarization modulation
infiared spectroscopy,J. Ply. Chem. B 105 (2001) pp. 7723-7729.
7. Blaudez D., Buffeteau T., Cornut J.C., Desbat B., Escafre N., Pezolet M. and Turlet
J.M., Polarization-modulated FT-IR spectroscopy of a spread monolayer at the
aidwater Interface, Appi. Spectroscopy 47 ( 1993) pp. 869-874.

258
SINGLE AND MULTI COMPONENT ADSORPTION OF VOLATILE ORGANIC
COMPOUNDS ONTO HIGH SILICA ZEOLITES DISCUSSION OF ADSORBED -
SOLUTION THEORY

P. MONNEYRON, M.-H. MANERO AND J.-N. FOUSSARD


LIPE, Dpt GPI, Institut National des Sciences Appliquees, 135 avenue de Rangueil
31077 Toulouse. France
E-mail: Pierre.monneyron@insa-tlse.jL,manero@ch. iut-tlse3.jL

Pure and binary component adsorption isotherms were determined for three volatile organic
compounds daily used as industrial solvents (methyl ethyl ketone (MEK), toluene (TOL),and 1,4-
dioxan (DIO))on two high silica zeolites, a desaluminated faujasite Y (Fau Y) and a silidite (Sil Z).
Apart from steric exclusion taking place with TOL-containing mixtures on Sil Z, adsor