Activated Carbon Fiber Filter Media For Proton Exchange Membrane Fuel Cell 
 
by 
 
Wei Liu 
 
 
 
 
A dissertation submitted to the Graduate Faculty of 
Auburn University 
in partial fulfillment of the 
requirements for the Degree of 
Doctor of Philosophy 
 
Auburn, Alabama 
May 14, 2010 
 
 
 
 
Keywords: Proton exchange membrane fuel cell, activated carbon fibers, sulfur dioxide 
removal  
 
 
Copyright 2010 by Wei Liu 
 
Approved by 
 
Sabit Adanur, Chair, Professor of Polymer and Fiber Engineering 
Bruce Tatarchuk, Professor of Chemical Engineering 
Peter Schwartz, Professor of Polymer and Fiber Engineering 
Ann B. Presley, Associate Professor of Consumer Affairs 
 
 
 
 
 
 
 
ii 
 
 
 
 
 
 
Abstract 
 
 
 Sulfur dioxide, a major air pollutant, has been found to be poisonous to proton 
exchange membrane fuel cells and causes degradation of fuel cell systems. Currently 
used industrial desulfurization systems, such as wet scrubbing, cannot address this 
problem because they cannot completely clean sulfur dioxide and SO2 with the 
concentration lower than 30 ppmv escapes. A fuel cell is more sensitive to poisoning than 
human?s lung and even a trace of sulfur dioxide will cause the irreversible degradation of 
a fuel cell. Activated carbon fibers (ACFs), one of relatively new adsorbents, are 
considered to remove sulfur dioxide from air before it is fed into fuel cell systems since 
ACFs have advantages in adsorption rate in comparison with granular activated carbon 
due to easy access of adsorbate to adsorption sites in micropores in activated carbon 
fibers.  
Activated carbon nanofibers were manufactured in the lab by electrospinning of 
polyacrylonitrile followed by chemical activation with potassium hydroxide. Micropores 
were shown to be dominant in the electrospun activated carbon nanofibers. However, 
their fragility and high manufacturing cost limit their applications in filtration.  
Pore structures and capabilities of removing sulfur dioxide of novoloid-based and 
rayon-based activated carbon fibers were analyzed. Surface areas were found to have 
correlation with pore volume in fibers. The sulfur dioxide capacities of ACFs were found 
to be proportional to the percentage of micropores in fibers. 
 iii 
To improve sulfur dioxide removal efficiency of activated carbon fibers, alkali, metal 
and oxidants were applied onto ACFs as promoters. KOH (potassium hydroxide), K2CO3 
(potassium carbonate), KClO3 (potassium chlorate) and KMnO4 (potassium 
permanganate) increased the desulfurization efficiency of ACFs greatly. Only KMnO4 
modified ACFs are suitable for the wet laying technique with which ACF paper was 
made for filter device in the current study since the other three promoters can be removed 
easily by exposure to water. The optimum potassium permanganate weight ratio in 
ACF15 was found to be in the range of 43~45 wt% which exhibited the maximum sulfur 
dioxide capacity. The effect of carrier gas on desulfurization was studied which showed 
that the co-presence of water and oxygen would be preferred. The advantage of KMnO4 
modified ACF15 is that the modification process does not need high temperature 
calcination in inert gas and thus is energy saving.  
Activated carbon fiber paper (K-ACFP) was manufactured from KMnO4 modified 
ACF15 and the binder fiber, copolyester (CoPET). The strongest K-ACFP was obtained 
when the ratio of KMnO4/ACF15 to CoPET is 7:3.  Sulfur dioxide breakthrough capacity 
of the resulting K-ACFP was found to be similar to KMnO4/ACF15 fiber before the wet 
laying process. Produced K-ACFP can be used as the filter media for fuel cell protection 
from sulfur dioxide.  
The desulfurization mechanisms of KMnO4 modified ACFs with different carrier gases 
were studied. Considering the diffusion of adsorbate within packed bed and single 
activated carbon fiber and reaction over fiber surfaces, models were developed for 
different compositions of the challenge gas to predict breakthrough characteristics of 
KMnO4/ACFs. The predicted curves agree with the experimental results reasonably well. 
 iv
The dependency of sulfur dioxide removal efficiency of KMnO4/ACFs on the location in 
the packed bed and the elapsed time are predicted with the developed models as well.  
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 v
 
 
 
 
 
 
Acknowledgments 
 
 
    The author would like to express her special thanks to her advisor Dr. Sabit Adanur for 
his commitment to this project, technical advice and support through the duration of this 
study. It would have been impossible for me complete this work without his support. I 
also appreciate his support not only in academic work but also in my life. He is more than 
an advisor and like a friend in life. 
    The author wishes to thank Dr. Bruce Tatarchuk, Dr. Peter Schwartz and Dr. Ann B. 
Presley for serving on her committee. Special thanks are extended to Dr. Tatarchuk for 
his financial support and making his top quality labs and research group available for my 
studies, which is appreciated.  
    The author wishes to thank Dr. Roy Broughton for valuable suggestions and help for 
my research and thank the Department of Polymer and Fiber Engineering for providing 
support and a friendly environment. 
    Special acknowledgement and heartfelt gratitude must be given to my parents, husband 
and parents-in-law and friends whose love and vision for me continues to motivate, 
inspire and guide me. 
 
                 
 
 
 
 vi
 
 
 
 
 
 
Table of Contents 
 
 
Abstract ............................................................................................................................... ii 
Acknowledgments................................................................................................................v 
List of Tables ...................................................................................................................... x  
List of Figures ................................................................................................................... xii 
List of Abbreviations ...................................................................................................... xvii 
Chapter 1 Introduction and Literature Review ....................................................................1 
     1.1 Introduction to fuel cells    .............................................................................2 
     1.2 Working mechanism of proton exchange membrane fuel cell (PEMFC)   ....3 
                1.3 Poisoning sources and their effect on PEMFC  .............................................6 
                1.4 Effects of contaminants on PEMFC performance  ........................................8 
                     1.4.1 Contaminants on proton exchange membranes .....................................9 
1.4.2 Contaminants for anode .........................................................................9 
1.4.3 Contaminants on cathode .....................................................................10 
1.5 Need of PEMFC cathode filter  ...................................................................16 
1.6 Industry method for sulfur dioxide purification  .........................................16 
1.7 Porous adsorbents in sulfur dioxide removal ..............................................17 
1.8 Advantages of ACFs in comparison with GAC and PAC ...........................18 
1.9 Manufacturing of activated carbon fibers ...................................................22 
1.10 Sulfur dioxide adsorption over activated carbon fibers .............................24 
 vii
1.10.1 Sulfur dioxide removal on bare activated carbon fibers ...................24 
           1.10.2 Treatment for promotion of sulfur dioxide removal ........................25 
1.11 Objectives of the research .........................................................................29 
References  ........................................................................................................30 
 Chapter 2 Manufacturing of Activated Carbon Fibers via Electrospinning .....................38 
2.1 Introduction .................................................................................................38 
2.2 Materials and experimental procedures .......................................................40 
2.3 Results and discussion .................................................................................41 
2.3.1 Influence of electrospinning parameters on fiber diameters ..............41 
           2.3.2 Physical and mechanical properties of electrospun PAN membranes47 
           2.3.3 Properties of electrospun PAN based activated carbon nanofibers ...52 
2.4 Conclusions .................................................................................................56 
References  ........................................................................................................57 
Chapter 3 Sulfur Dioxide Removal Over Activated Carbon Fibers and Modified 
Activated Carbon Fibers ....................................................................................60 
3.1 Experimental ...............................................................................................60 
3.1.1 Materials .............................................................................................60 
3.1.2 Pore structures and element analysis ..................................................62 
            3.1.3 Sulfur dioxide adsorption testing .......................................................63 
            3.1.4 Modification of activated carbon fibers .............................................65 
            3.1.5 ACFs surface morphology analysis ....................................................67 
            3.1.6 Thermogravimetric analysis (TGA) ...................................................67 
            3.1.7 X-Ray diffraction analysis ..................................................................67 
 viii 
            3.1.8 Manufacturing of activated carbon fiber paper ..................................67 
3.2 Results and discussions ...............................................................................71 
3.2.1 Pore structures and surface morphology of ACFs ..............................71 
3.2.2 Element analysis .................................................................................76 
3.2.3 Sulfur dioxide adsorption properties of activated carbon fibers and 
modified activated carbon fibers   ......................................................76 
3.3 SEM imaging analysis  ................................................................................96 
3.4 Activated carbon fiber papers (ACFP) and their SO2 adsorption capacities99 
3.5 XRD analysis on KMnO4/ACF15 .............................................................105 
3.6 Energy dispersive X-ray analysis ..............................................................108 
References  ......................................................................................................111 
Chapter 4 Mechanism Analysis of Sulfur Dioxide Removal by ACFs and Model 
Development ...................................................................................................113 
4.1 Introduction   .............................................................................................113 
4.2 Mechanism one ..........................................................................................116 
4.2.1 Mechanism study   ............................................................................116 
4.2.2 Model development   ........................................................................121 
          4.2.3 Prediction of breakthrough curve with the developed model ...........132 
4.3 Mechanism two  ........................................................................................134 
4.4 Mechanism three  ......................................................................................139 
           4.4.1 Mechanism analysis .........................................................................139 
           4.4.2 Model development .........................................................................144 
4.5 Mechanism four .........................................................................................148 
4.6 Model system ............................................................................................153 
 ix
References  ......................................................................................................155 
Chapter 5 Conclusions and Recommendations ................................................................158 
5.1 Conclusions ...............................................................................................158 
5.2 Recommendations  ....................................................................................160 
References  ......................................................................................................161 
Appendix 1 Matlab? code for Yoon?s model ..................................................................162 
Appendix 2 Matlab? code for prediction of breakthrough curves for mechanism one and 
two ................................................................................................................163 
Appendix 3  Matlab? code for prediction of breakthrough curves for mechanism three 
and four .........................................................................................................167 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 x
 
 
 
 
List of Tables 
 
 
Table 1.1 Types of fuel cells and their major features  ........................................................3 
Table 1.2 Contaminants in the operation of PEM fuel cells  ...............................................6 
Table 1.3 National ambient air quality standards ................................................................8 
Table 1.4 Contaminants and their concentrations in battlefield  .........................................8 
Table 2.1 Physical properties of electrospun PAN membrane and stabilized PAN                   
membrane  .........................................................................................................51 
Table 2.2 Pore structures of electrospun PAN and activated carbon nanofibers ...............55 
Table 3.1 Compositions of activated carbon fiber papers ..................................................70 
Table 3.2 BET surface area, pore volume and pore size of ACFs and PAC .....................72 
Table 3.3 Pore volume distribution of ACFs and PAC .....................................................74 
Table 3.4 Number of adsorption sites of activated carbon fibers and PAC .......................75 
Table 3.5 Element components of activated carbon fibers  ...............................................76 
Table 3.6 Sulfur dioxide capacities of bare activated carbon fibers in the absence of  
water ..................................................................................................................79 
Table 3.7 Sulfur dioxide capacities of bare activated carbon fibers with water present  ..81 
Table 3.8 Sulfur dioxide capacities of bare activated carbon fibers with initial SO2 
concentration of 50ppm and with humidified air as the carrier gas  .................83 
Table 3.9 Sulfur dioxide capacities of metal/ACF20.........................................................89 
Table 3.10 Element contents of KMnO4/ACF15 before and after adsorption of SO2 in 
humidified air  ................................................................................................109 
Table 4.1 90% saturation capacities  ................................................................................119 
Table 4.2 4% breakthrough capacities .............................................................................144 
 xi
Table 4.3 Parameters in the model system for mechanisms 1~4 .....................................154 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 xii
 
 
 
 
List of Figures 
 
 
Figure 1.1 Schematic of a proton exchange membrane fuel cell structure  .........................4 
Figure 1.2 Polarization curves with different NOx concentrations  ...................................12 
Figure 1.3 Periodic test of the reversibility of cell performances  .....................................12 
Figure 1.4 Cell current after exposure to 2.5 and 5 ppm SO2 in air  ..................................14 
Figure 1.5 Cyclic voltammetry spectra after cathode exposure to 2.5ppm SO2 in air  ......15 
Figure 1.6 Durability of PEMFC with exposure to 5ppm SO2 in air .................................15 
Figure 1.7 Schematic of pore structures of a) ACF and b) GAC  ......................................21 
Figure 1.8 Pore size distribution of ACF and GAC  ..........................................................21 
Figure 2.1 SEM images of electrospun PAN from different concentrations  ....................43 
Figure 2.2 Fiber diameters with different PAN concentrations .........................................43 
Figure 2.3 The frequency distribution of nanofiber diameters for different solution 
concentrations   ................................................................................................44 
Figure 2.4 Fiber diameters at different voltages ................................................................45 
Figure 2.5 Fiber diameters at different flow rates  .............................................................46 
Figure 2.6 Fiber diameters at different need-tip distances .................................................47 
Figure 2.7 Variation of tensile strength with the change of electrospun PAN membrane 
thickness ...........................................................................................................47 
Figure 2.8 Variation of tear strength with the change of membrane thickness  ................49 
Figure 2.9 Variation of burst strength with the change of membrane thickness  ..............49 
Figure 2.10 Air permeability as a function of membrane thickness ..................................50 
 xiii 
Figure 2.11 Tensile strength variation of stabilized electrospun PAN with the increase in 
thickness  ........................................................................................................52 
Figure 2.12 SEM images of (a) electrospun PAN membrane and (b) activated carbon 
nanofibers .......................................................................................................53 
Figure 2.13 FTIR spectrum for (a) electrospun PAN, (b) stabilized PAN, and (c) activated 
carbon nanofibers  ..........................................................................................54 
Figure 2.14 DA pore size distribution................................................................................56 
Figure 2.15 Generalized Halsey V-t plot ...........................................................................56 
Figure 3.1 Structure of KynolTM novoloid fibers ...............................................................61 
Figure 3.2 Structure of carbonized novoloid fiber with 95% carbon ................................61 
Figure 3.3 Schematic and picture of SO2 adsorption test bench ........................................63 
Figure 3.4 Correlation between surface area and pore volume  ........................................73 
Figure 3.5 Pore size distributions of ACFs and ACP  .......................................................74 
Figure 3.6 Surface morphology of ACFs 1) ?9500, 2) ?250,000 .....................................75 
Figure 3.7 Effect of surface area of ACFs on sulfur dioxide breakthrough capacity ........79 
Figure 3.8 Breakthrough profiles of ACFs with absence of water in gas stream  .............80 
Figure 3.9 Breakthrough profiles of ACFs with the presence of water in the carrier  
gas ....................................................................................................................82 
Figure 3.10 Mechanism of sulfur dioxide removal over activated carbon surface with the 
presence of oxygen and water  .........................................................................82 
Figure 3.11 Breakthrough profiles of ACFs in SO2/humidified air  ..................................84 
Figure 3.12 Breakthrough curves of KOH/ACF20 exposed to 20ppm SO2 carried by dry 
air ....................................................................................................................86 
Figure 3.13 SO2 adsorption capacities and utilization of KOH/ACF20 and  
K2CO3/ACF20 ................................................................................................87 
Figure 3.14 SO2 adsorption capacities of KClO3/ACF15 before and after exposure to 
water ...............................................................................................................90 
 xiv
Figure 3.15 SO2 breakthrough capacity of KMnO4 loaded ACF10, ACF15, ACF20 and 
PAC ................................................................................................................92 
Figure 3.16 SO2 breakthrough capacity of ACF15 loaded with different ratios of   
KMnO4    .........................................................................................................93 
 
Figure 3.17 Breakthrough profiles of KMnO4/ACF15 before and after exposure to      
water ...............................................................................................................93 
Figure 3.18 Breakthrough profiles of KMnO4/ACF15 after exposure to water under 
different adsorption conditions .......................................................................94 
Figure 3.19 Breakthrough profiles of KMnO4/ACF15 prepared from 2wt% KMnO4 
solutions ..........................................................................................................95 
Figure 3.20 Sulfur dioxide adsorption properties of humidified KMnO4/ACF15 in 
comparison with dried samples ......................................................................96 
Figure 3.21 Surface morphology of KMnO4/ACF15: 1) 6.55wt% KMnO4 2) 23.16wt% 
KMnO4, 3) 43.34wt% KMnO4 and 4) 43.34wt% KMnO4 after exposure to 
water ...............................................................................................................97 
Figure 3.22 KMnO4/ACF15 surface morphology after exposure to 1) SO2+N2, 2) 
SO2+air, 3) SO2+H2O and 4) SO2+air+H2O  ..................................................98 
Figure 3.23 Thickness, tensile strength and coefficient of thickness variation (CV) of 
ACFP samples corresponding to Table 3.1 ....................................................99 
Figure 3.24 DSC analysis of CoPET  ..............................................................................101 
Figure 3.25 SEM images of ACFP composed of 70wt% ACF20 and 30wt% CoPET with 
a fusing temperature  of 200?C for 20min (?500)  .......................................102 
Figure 3.26 SEM images of K-ACFP: (1) before and (2) after sulfur dioxide adsorption 
tests (?2000)  ..............................................................................................104 
Figure 3.27 Breakthrough profiles of ACFP in comparison with KMnO4/ACF15 with 50 
ppm SO2 challenge gas carried by humidified air  .......................................105 
Figure 3.28 XRD patterns of ACF15, KMnO4/ACF15 and spent KMnO4/ACF15 after 
exposure to sulfur dioxide carried by humidified air  ..................................106 
Figure 3.29 XRD patterns of KMnO4/ACF15 before and after adsorption of sulfur 
dioxide carried by humidified nitrogen ........................................................107 
 xv
Figure 3.30 XRD patterns of KMnO4/ACF15 before and after adsorption of sulfur 
dioxide carried by dry air .............................................................................108 
Figure 3.31 EDS element signals (1) before and (2) after adsorption in the challenge gas 
of SO2+H2O+air  ..........................................................................................110 
Figure 4.1 Prediction of breakthrough curve of ACF20 according to Yoon?s model .....115 
Figure 4.2 Prediction of breakthrough curve of KMnO4/ACF15 according to Yoon?s 
model..............................................................................................................115 
Figure 4.3 TGA curves of KMnO4/ACF15 before and after adsorption tests in 50ppm SO2 
carried by nitrogen .........................................................................................117 
Figure 4.4 Breakthrough curves of ACF15, KMnO4/ACF15 before exposed to water,  
KMnO4/ACF15 and KMnO4 in the challenge gas of SO2+N2  ......................118 
Figure 4.5 Breakthrough curve of KMnO4/ACF15 exposed to 50ppm sulfur dioxide 
carried by nitrogen  ........................................................................................133 
Figure 4.6 Prediction of sulfur dioxide removal efficiency with the variation of location 
and time  .........................................................................................................133 
Figure 4.7 Breakthrough profiles of KMnO4/ACF15 with exposure to sulfur dioxide 
challenge gas carried by air and nitrogen ......................................................135 
Figure 4.8 Weight change of KMnO4/ACF15 before and after adsorption tests in the 
challenge gas of SO2 carried by air with increasing temperatures  ................136 
Figure 4.9 Prediction of breakthrough curve of KMnO4/ACF15 with challenge gas of 
SO2+air ...........................................................................................................138 
Figure 4.10 Prediction of sulfur dioxide removal efficiency at different locations and 
times with the challenge gas of SO2+air   ......................................................139 
Figure 4.11 Weight changes of ACF15, KMnO4/ACF15 before and after SO2 adsorption 
test with the presence of water in the carrier gas with increasing   temperatures 
........................................................................................................................141 
Figure 4.12 Effect of temperature on weight of ACF15 and KMnO4/ACF15.................141 
Figure 4.13 Breakthrough profiles of KMnO4 with different challenge gases ................143 
 xvi
Figure 4.14 Breakthrough profiles of KMnO4/ACF15 in comparison with KMnO4 in the 
gas stream of sulfur dioxide and humidified nitrogen  ..................................143 
Figure 4.15 Prediction of breakthrough profile of KMnO4/ACF15 in humidified     
nitrogen  .........................................................................................................147 
Figure 4.16 Dependence of sulfur dioxide removal efficiency on time and location in the 
packed bed with the carrier gas of humidified nitrogen  ..............................148 
Figure 4.17 Breakthrough profile of KMnO4/ACF15 with exposure to sulfur dioxide 
carried by humidified nitrogen and air .........................................................149 
Figure 4.18 TGA curves of KMnO4/ACF15before and after exposure to challenge gas of 
SO2+H2O+air and SO2+H2O+N2 ..................................................................151 
Figure 4.19 Prediction of breakthrough profile of KMnO4/ACF15 exposed to SO2 
+H2O+air   ....................................................................................................152 
Figure 4.20 SO2 removal efficiency of KMnO4/ACF15 at various times and locations 
with the challenge gas of SO2 +H2O+air  .....................................................153 
Figure 4.21 Prediction of breakthrough curve following Mechanism One with the power 
law model system   .......................................................................................154 
Figure 4.22 Prediction of breakthrough curve following Mechanism Two with the power 
law model system   .......................................................................................155 
 
 
 
 
 
 
 
 
 
 
 
 xvii
 
 
 
 
 
 
List of Abbreviations 
 
 
C Sulfur dioxide concentration in the main gas stream (mol/m3) 
Cr  Gas phase sulfur dioxide concentration within the fiber (mol/m3)  
F
sC       Solid phase sulfur dioxide concentration on fiber surface (mol/m
3) 
Cs Average sulfur dioxide concentration within fibers including gas phase and solid 
phase which is adsorbed on fiber surfaces (mol/m3) 
Crs Solid phase sulfur dioxide concentration adsorbed on fiber surface (mol/m3) 
rC         Average gas phase sulfur dioxide concentration within a fiber (mol/m
3)  
rsC  Average of the solid phase adsorbed sulfur dioxide concentration on the fiber 
surface (mol/m3)  
D Diffusivity (m2/s) 
De  Effective diffusivity (m2/s)  
df  Fiber diameter (m) 
dre Diameter of the reactor 
Dk Knudsen diffusivity (m2/s) 
Dm Molecular diffusivity (m2/s) 
Dz Axial diffusivity (m2/s) 
F Gas phase sulfur dioxide concentration within the fiber used in the modeling code 
(mol/m3) 
 xviii 
FR Volumetric flow rate of gas stream (mL/min) 
?
?
?= 1f  
h Step size in z direction of the model
 
L Length of the packed ACFs bed (m)  
k Step size of time direction in the model 
kB Boltzmann?s constant.  
kg  Mass transfer coefficient   
M Number of segments of the packed bed for the model 
Mg Molecular weight of gas (g/mol) 
  
m Index of time direction in the model 
N Number of the segments of time for the model 
n Index of z direction in the model 
r           Radius of an infinitesimal unit from the central line of a fiber 
R Ideal gas constant (8.3 m3.Pa/K?mol) 
Re Reynolds number 
rf          Fiber radius (m) 
rp  Pore radius (m) 
Sc Schmidt number  
t           Time (s) 
T Temperature (K) 
u Face velocity of the main gas stream (m/s)  
Vcarbon Volume occupied by the solid carbon in an activated carbon fiber (mL) 
 xix
VmodifierVolume of the modification agents (such as potassium hydroxide and potassium 
permanganate) 
Vtpore Total pore volume within the fiber (mL) 
z Distance from the beginning of the packed ACFs bed (m)  
rf  Fiber radius (m) 
? Intrafiber void fraction or the intrafiber porosity 
? Modifier load with respect to the total sample weight 
? Porosity of the packed ACFs bed 
?  Tortuosity factor of pores in porous fibers 
? Mean free path 
?carbon Solid density of ACF (cm3/g) 
?modifier Density of the modification agent (cm3/g) 
?g Density of fluid (cm3/g) 
? Viscosity of the gas stream (cP?s) 
?AB Collision diameter (angstrom) 
?AB  Maximum attractive energy between two molecules 
?AB  Collision integral, a function of kBT/?AB  
1 
Chapter 1 Introduction and Literature Review 
1.1 Introduction to fuel cells  
Fuel cells are electrochemical devices that convert chemical energy into electrical energy 
directly [1]. They are considered to be a viable alternative to the internal combustion 
engines, since they have a higher efficiency. Combined with H2, as the fuel, they enable 
CO2 emission free power conversion [2]. 
     Because the intermediate steps of producing heat and mechanical work of most 
conventional power generation methods are avoided, fuel cells are not limited by 
thermodynamic limitations of heat engines such as the Carnot efficiency. In addition, fuel 
cells produce power with minimal pollutant since combustion is avoided. However, 
unlike batteries the reductant and oxidant in fuel cells must be continuously replenished 
to allow continuous operation.  
    Fuel cell systems are assembled with fuel preparation, air supply, thermal 
management, water management and electric power conditioning equipment. Fuel cells 
can be classified by the type of electrolyte used in the cell as listed in Table 1.1[3]. The 
various fuel cell types offer the strengths in different ways. The proton exchange 
membrane fuel cell (PEMFC) capitalizes on the essential simplicity of the fuel cell. The 
electrolyte is solid polymer in which protons are mobile. The chemistry is the same as the 
acid electrolyte fuel cell. With a solid and immobile electrolyte, this type of fuel cell is 
2 
inherently very simple. These cells run at quite low temperatures, so the problem of slow 
reaction rates is addressed by sophisticated catalysts and electrodes. Platinum is the 
catalyst and the development in recent years allow a small amount to reduce the cost. 
However, the storage of hydrogen is not really addressed. Considering the difficulty in 
storing hydrogen fuel, methanol is used in direct methanol fuel cells. Even though these 
fuel cells have very low powers, they have many potential applications in the rapidly 
growing portable electronic equipment. Alkaline fuel cells have advantages in less 
activation overvoltage, higher operating voltages (as high as 0.875bV) and low costs 
whereas the main problem is that air and fuel supplies must be free of CO2 and hydrogen 
must be used.  The phosphoric acid fuel cell (PAFC) was first commercialized and enjoys 
widespread terrestrial use because of reliability and easy or free maintenance. The solid 
oxide fuel cell (SOFC) operates in the region of 600?C to 1000?C allowing a high 
reaction rate without expensive catalysts but the ceramic materials used to make SOFC 
are difficult to handle and manufacture and extra pre-heating and cooling equipment is 
required. Molten carbonate fuel cell needs carbon dioxide in air to work and the high 
reaction rate is achieved by high temperature and the relatively low cost catalyst, nickel.    
3 
Table 1.1 Types of fuel cells and their major features [3] 
Fuel cell type Mobile ion 
Operating 
temperature 
Applications and notes 
Alkaline (AFC) OH- 50~200?C 
Used in space vehicles, e.g. Apollo, 
Shuttle 
Proton exchange 
membrane (PEMFC) 
H+ 30~100?C Vehicles and mobile applications 
Direct methanol 
(DMFC) 
H+ 20~90?C 
Suitable for portable electronic 
systems of low power, running for 
long times 
Phosphoric acid 
(PAFC) 
H+ ~200?C 
Large number of 200 kW CHP 
systems 
Molten carbonate 
(MCFC) 
CO32- ~650?C 
Suitable for medium to large scale 
CHP systems, up to MW capacity 
Solid oxide (SOFC) O2- 500~1000?C 
Suitable for all sizes of CHP systems, 
2 kW to multi-MW 
*CHP refers to combined heat and power 
1.2 Working mechanism of proton exchange membrane fuel cell (PEMFC) 
The basic structure of a fuel cell consists of an electrolyte layer in contact with an anode 
and a cathode as shown in Figure 1.1. Fuel is fed continuously to the anode and an 
oxidant is fed to the cathode. The electrochemical reactions take place at the electrodes to 
produce an electric current through electrolytes, while driving a complementary electric 
current that performs work on the load.  
4 
    In PEMFC, hydrogen is used as fuel at the anode, oxygen (or air) at the cathode as 
oxidants and polymer membrane as the electrolyte. On the anode, protons are produced 
and electrons are released (equation 1.1). In order to react with oxygen at the cathode 
(equation 1.2), protons penetrate electrolyte and electrons pass through an electrical 
circuit causing current.  
?+ +? e4H4H2
2                                                                                                      (1.1) 
OH2H4e4O 22 ?++ +?                                                                                         (1.2) 
Figure 1.1 Schematic of a proton exchange membrane fuel cell structure. (GDL: gas diffusion 
layer, ACL: anode catalyst layer, PEM: proton exchange membrane, CCL: cathode catalyst 
layer and MEA: membrane electrode assembly) [3] 
e-
Load
Positive Ion 
         or  
Negative Ion 
Depleted Fuel and  
Product Gases Out  
Fuel In 
H2O
H2
H2O
Depleted Oxidant and  
Product Gases Out  
Oxidant In 
1/2O2
GDL 
ACL 
PEM 
CCL 
GDL 
MEA
5 
    The electrolyte of PEMFC is an ion exchange polymer membrane requiring proton 
conductivity. Typical electrodes for both anode and cathode are carbon loaded with 
platinum electrocatalysts [1].  
     PEMFC has a wide application potential in vehicles and stationary devices due to its 
advantages.  PEMFC has excellent resistance to gas crossover due to a solid electrolyte. 
The low operating temperature allows rapid start-up. Corrosion problems of PEMFC are 
minimal since the only liquid is water so that the use of exotic materials required in other 
fuel cells is not needed. PEMFC is capable of high current density (over 2 kW/L). 
However, the applications of PEMFC are limited by certain disadvantages. The low 
thermal stability of polymer membranes limits their operating temperature so that a H2-
rich fuel gas with minimal or no CO, which is a catalyst poison at a low temperature, is 
preferred for anode. The hydrogen generation and storage is a barrier to commercialize 
PEMFC. The low and narrow operating temperatures also present challenge for thermal 
management especially at very high current densities. It is difficult to balance ensuring 
sufficient hydration of the electrolyte against flooding the electrolyte. Moreover, 
PEMFCs are sensitive to poisoning by trace levels of contaminants [1]. 
    Although the use of oxygen rather than air will increase the power of a PEM fuel cell 
by about 30%, pure oxygen is mostly used in air-independent systems, such as 
submarines and spacecraft; otherwise, air is used since it is easy and cheap to get. 
However, contaminants in air may degrade the performance of PEM fuel cells.  
6 
1.3 Poisoning sources and their effect on PEMFC  
Contaminants for PEM fuel cells originate from fuel for the anode, air for the cathode and 
other compounds from fuel cell structure components. Major contaminants identified in 
the operation of PEM fuel cells are listed in Table 1.2 [4].   
Table 1.2 Contaminants in the operation of PEM fuel cells 
Contamination sources Typical contaminants
Air N2, NO, NO2, SO2, SO3, NH3 and O3
Hydrogen fuel CO, CO2, H2S, NH3, CH4
Bipolar metal plates Fe3+, Ni2+, Cu2+, Cr3+
Membranes (Nafion?) Na+, Ca2+ , Fe3+, Cu2+, and Mg2+
Sealing gasket Si
Coolants Si, Al, S, K, Fe, Cu, Cl, V, Cr
Battlefield pollutants 
SO2, NO2, CO, propane, benzene, HCN, CNCl, Sarin, 
Sulfur mustard 
Compressor Oils
    For the anode, hydrogen fuel is provided mainly from commercially available sources. 
The reformation of hydrocarbons and oxygenated hydrocarbons including methane from 
natural gases and methanol from biomass is the dominant method for hydrogen 
production. However, the reformation process results in impurities such as CO, H2S, 
NH3, organic sulfur-carbon and carbon-hydrogen compounds. Steam reforming and 
partial oxidation or autothermal reforming are usually used to produce hydrogen-rich gas 
7 
called ?reformate? which contains 40 to 70% hydrogen, 15 to 25% CO2, 1 to 2% CO and 
sulfur compounds [4].  
In the cathode, air is the most practical material feed to fuel cell stack. A variety of 
air pollutants contaminate the fuel cell, degrading MEA by poisoning the catalysts. They 
include nitrogen oxides (NO and NO2), sulfur oxides (SO2 and SO3), carbon oxides (CO 
and CO2), ozone and other organic compounds. The major sources of these contaminants 
are automotive vehicle exhaust and industrial manufacturing processes. The U. S. 
Environmental Protection Agency (EPA) Office of Air Quality Planning and Standards 
(OAQPS) have set the National Ambient Air Quality Standards for CO, NO2, O3 and SO2 
as shown in Table 1.3 [5]. Pollutant type and level varies with working conditions. For 
instance, in the battlefield, there are usually more air pollutants with higher level 
concentrations resulting from burning fuels, volatile chemicals and generated compound 
during the operation of weapons.   
Types of air pollutants in battlefield are included in Table 1.4 based on Moore et al. 
study [6]. CO, SO2 and NO2 are present more than the amount specified in air quality 
standards for public health. They are also suspected to degrade fuel cell performances.  
8 
Table 1.3 National Ambient Air Quality Standards [5] 
Pollutant Level Averaging Time
Carbon  
Monoxide 
9 ppm 8-hour 
35 ppm 1-hour 
Nitrogen 
Dioxide 
0.053 ppm Annual 
Ozone 
0.075 ppm 8-hour 
0.12 ppm 1-hour 
Sulfur  
Dioxide 
0.03 ppm Annual 
0.14 ppm 24-hour
                     * This standard set limits to protect public health, including the health  
                     of "sensitive" populations such as asthmatics, children, and the elderly.  
Table 1.4 Contaminants and their concentrations in battlefield [4] 
Contaminant CO SO2 NO2 CNCl HCN Benzene Propane Sarin 
Sulfur 
mustard 
Concentration 
(ppm) 
20 0.5 0.4 780 1780 50 90 170 15 
1.4 Effects of contaminants on PEMFC performance  
The effects of contaminants on fuel cell are one of the most important issues in fuel cell 
operation and applications. Three major areas have been addressed in the literature: 
theoretical and empirical modeling of contamination to provide a fundamental 
understanding of the mechanisms, experimental observation and validation and 
mitigation of the effects of contaminants [4].  
9 
Contaminants are considered to have influence on MEA (membrane electrode 
assembly) consisting of ACL (anode catalyst layer), CCL (cathode catalyst layer), GDL 
(gas diffusion layer) and PEM (proton exchange membrane).  The impurities from feed 
gases, hydrogen for anode, air for cathode and stack component materials, can directly 
enter the fuel cell, poisoning the catalyst sites, affecting MEA properties such as 
hydrophilcity and hydrophobicity, modifying the proton transportation path and affecting 
water management, thus causing performance degradation. Three effects of impurities 
can be categorized into three major types: kinetic losses due to the poisoning of catalysts 
on both anode and cathode electrodes, ohmic losses due to an increase in the resistance of 
cell components and mass transport losses due to changes in structure and 
hydrophobacity of catalyst layers, PEMs and GDLs [7]. 
1.4.1 Contaminants on proton exchange membranes 
For PEM, the cationic contaminants, such as Ca2+, Fe3+, Cu2+, Na+ and Mg2+, can get into 
the Nafion? membrane to compete with the proton for the sulfonic acid group (SO3H) 
due to the sulfonic acid group?s affinity for cations and thus decrease the water content 
resulting in a reduction in proton conductivity causing fuel cell degradation. Ammonium 
and Cl- are adsorbed to catalyst surface blocking active sites for oxygen reduction 
reaction (ORR) and enhancing the formation of peroxide (H2O2) which catalyzes the 
radical formation reactions. The migration of metal impurities, for example Cr, allows the 
contaminants to spread from one electrode throughout the MEA [4, 8].  
1.4.2 Contaminants for anode 
For the anode, approximately 95% of hydrogen fuel is produced by steam reforming of 
natural gas (CH4), which produces poisoning byproduct gases such as CO, CO2, H2S, 
10 
NH3 and carbon-hydrogen compounds. Carbon monoxide is reported to have the worst 
effect on fuel cell performance due to its strong bond with catalyst Pt resulting in 
reduction of active sites available for hydrogen adsorption and oxidation. Increases in 
operating temperature and humidity reduce CO coverage on the catalyst by promoting 
CO oxidation with an OHads group [9].  At as low as 10 ppm CO in hydrogen stream, 
anode performance was slightly affected while 100 ppm CO caused a remarkable 
increase of anodic overpotential [10]. Carbon dioxide does not affect anode performance 
directly but is catalytically converted to carbon monoxide [4]. H2S and ammonia also 
cause voltage loss by blocking active site on catalysts [4]. After exposure to 50 ppm 
hydrogen sulfide, only partial recovery of membrane electrode assembly properties by the 
electrochemical oxidation is achieved [11].  
1.4.3. Contaminants on cathode 
This research will focus on the effects of air impurities fed into the cathode. SOx and NOx
have been found as the dominant contaminants for cathode.  
1.4.3.1 Effect of nitrogen oxide At low NO2 concentration, 1 ppm, the cell performance 
has been reported to decrease by about 10% after running 100 hour. However, the 
concentration of NOx in the exhaust of internal-combustion engine vehicles (IEV) on the 
road can easily exceed 50 ppm, even reaching up to 300 ppm [12, 13]. According to the 
literature, 85~95% of NOx is NO [13, 14]. Yang et al. simulated the exhaust of IEV with 
mixture of NO and NO2 in the ratio of 9:1 which was mixed with air to obtain 1480, 140 
and 10 ppm NOx in air stream in order to investigate the effect of NOx on fuel cell 
performance [12]. The polarization curves of different NOx concentrations are shown in 
Figure 1.2. Introduction of 10 ppm NOx to the cathode slightly decreased the cell 
11 
performance and the decline increased with increased NOx concentration. Experiments 
have proved that the degradation due to NOx can be recovered by refeeding clean air as 
shown in Figure 1.3. With the introduction of 124 ppm NOx, cell voltage dropped by 
about 42% but recovered to the initial voltage after operating with clean air for 120 min. 
After injecting 124 ppm NOx for four times, the cell could recover approximately by 
94%. The abrupt voltage drop with the introduction of NOx was probably due to the quick 
adsorption of NOx on catalyst surface suppressing the oxygen reduction reaction (ORR). 
But the weak adsorption is reversible and adsorbed NOx was desorbed recovering the cell 
voltage. The absence of any oxidation peaks in cyclic voltammetry spectra reported by 
Mohtadi et al. revealed that the poisoning of NO2 is not catalyst surface poisoning and 
suggested that the ionomer or the catalyst-ionomer interface could be affected by the 
exposure to NO2 [15]. The possible mechanism of NOx effect is that NO is oxidized to 
NO2 (equation 1.3) which becomes electrochemically reduced on the cathode and thus 
competes with O2 for Pt sites forming NH4+ (equation 1.4). NH4+ is an ionomer poison as 
the effect of NH3 on anodes.  
22 NO2ONO2 ?+                                                                                                       (1.3) 
OH2NHe7H8NO 242 +?++ +?+                                                                               (1.4) 
12 
Figure 1.2 Polarization curves with different NOx concentrations [11].
Figure 1.3 Periodic test of the reversibility of cell performance. The fuel cell was 
discharged at 12.5 A, 338 K, ambient pressure and the utilization rates of H2 and air 
were 80 and 40%, respectively [11]. 
13 
1.4.3.2 Effect of sulfur oxide Several researches have studied the effect of SO2 on cell 
performance. Moore et al. did not observe detrimental effect when 500 ppb SO2 was 
present in the air [6]. To observe the accumulation effect of SO2 in a long term, 1 ppm 
SO2/air was fed to the cathode and the cell potential declined by 35% after running 100 
hour [16]. Mohtadi et al. introduced 2.5 and 5 ppm SO2 in the air stream and compared it 
with the same amount of NO2 and found that the poisoning effect of SO2 appear to be 
severe according to the polarization curve data [15]. The degradation of cell performance 
after exposure to SO2 recovered partially with neat air as shown in Figure 1.4 but the 
effect varied with SO2 concentration. Although the SO2 concentration was different, the 
total dosage of SO2 remained the same by controlling exposure time. 23 hour exposure to 
5 ppm SO2 degraded the cell current by about 78% while 53% current decrease was 
observed after exposing to the same dose of 2.5 ppm SO2. Moreover, the recovery of 
current after applying 2.5 ppm SO2 was lower compared to 5 ppm SO2. After cyclic 
voltammetry, cell voltage almost recovered completely [15].  
  
14 
The mechanism of SO2 poisoning and its effect on cell performance was analyzed 
based on cyclic voltammetry by Mohtadi et al. Two oxidation peaks at 0.89 and 1.05 V 
are shown in Figure 1.5 [15]. They correspond to the presence of two sulfur species 
adsorbed on the surface of the Pt catalyst.  SO2 was reduced to sulfur on Pt with the 
formation of SO intermediate as shown in reaction Equations1.5 and 1.6. Figure 1.6 
shows the durability of the cell in 5 ppm SO2/air. Unlike the effect of NOx on cell 
performance, the degradation caused by SO2 cannot be recovered to or close to the intial 
performance by introducing neat air [15].   
OHSOPteHSO Pt 22 22 +????++ ?+                                                                      (1.5) 
                                                             (1.6) OHSPteHSOPt 222 +??++? ?+
Figure 1.4 Cell current after exposure to 2.5 and 5 ppm SO2 in air [15]
15 
where SOPt ?  and SPt ?  represent that SO and S attach to catalyst Pt.
Figure 1.6 Durability of PEMFC when exposure to 5 ppm SO2 in air [15]
Figure 1.5 Cyclic voltammetry spectra after cathode exposure to 2.5 ppm SO2 in air [15]
16 
1.5 Need for PEMFC cathode filter 
As discussed, even trace amounts of impurities in fuel or air stream will severely 
poison the anode or the cathode. Some contaminants can irreversibly be adsorbed onto 
catalyst and thus block the active sites of the catalysts. Most approaches have focused on 
mitigation of contamination on the fuel side, such as pre-treatment of reformate, anode 
oxidant-bleeding and CO tolerant catalyst synthesis. However, developments in 
mitigation of cathode poisons have not been well documented [4]. In order to 
commercialize PEMFC, contamination prevention on the cathode side must be addressed. 
Two strategies to purify the feed air stream have been suggested. One is the use of 
filtration to filter poisons out and the other is the development of contamination tolerance 
MEA components such as catalysts [4]. The later is more challenging due to the variety 
of contaminants. Filtration technology is more feasible and separate filters can be 
developed for either anode or cathode applications. Considering the negative effect of 
contaminants on the cathode, sulfur dioxide has more severe influence on the PEMFC 
performance as discussed above. Therefore, sulfur dioxide will be our first filtration 
target in this study. Kennedy et al. has proposed methodologies for design of fuel cell 
cathode air filters [17, 18].  
1.6 Industry methods for sulfur dioxide purification 
Sulfur dioxide is usually produced by volcanoes and in various industrial processes. The 
main sulfur dioxide emission is from combustion of coal and petroleum, which are 
currently the main energy resources. The most widely used purification technique to 
clean sulfur dioxide before released into atmosphere is called flue gas desulfurization. 
17 
Due to the US EPA?s Acid Rain Program, the America has witnessed a 33% decrease in 
emissions between 1983 and 2002, resulted from flue gas desulfurization [5]. As this 
technology advances and becomes more efficient, there are many variations on how this 
process is executed, such as wet-scrubbing systems, spray-dry scrubbing, sodium sulfite 
solution and gas phase oxidation with ammonia [19]. Most of the coal burning utilities 
have chosen lime-based web scrubbing. However, there are a number of disadvantages of 
this method. Investment cost is usually high to purchase lime-based scrubbing materials. 
Wet scrubbing system consumes a large amount of water and the byproducts of calcium 
sulfate/sulfite have low value [20]. Carbon dioxide released in this system is a monitored 
greenhouse gas [21]. The most important issue is that no industrial process is 100% 
effective and some sulfur dioxide still escapes into the atmosphere (30ppm) [19, 20].  
That will be a problem for high quality air requirement for PEMFC since even a trace 
amount of sulfur dioxide will cause performance degradation as discussed previously. On 
the other hand, activated carbon offers numerous advantages: a small sorbent make-up, 
generation of relatively small amounts of by-products, recovery of sulfur as useful sulfur-
containing products, simultaneous removal of sulfur dioxide and nitrogen oxide and 
elimination of large water consumption [22].  
1.7 Porous adsorbents in sulfur dioxide removal   
Activated carbons or other porous adsorbents, which have been widely used in 
purification or filtration systems, can be considered for complete removal of sulfur 
dioxide through adsorption, either physisorptions or chemisorptions, before air is fed into 
fuel cell system.  
18 
    Some researche has been conducted on desulfurization over porous adsorbents, such as 
alumina, zeolites and activated carbons. They have been treated to enhance sulfur dioxide 
removal efficiency. We are focusing on the most popular adsorbent, activated carbon 
because it can remove nearly every impurity found in the combustion flue gas, including 
particulate material, heavy metals, organic materials and other air toxics [23].   
    The well-known activated carbon can be classified into three categories according to 
the shape: granular activated carbon (GAC), powdered activated carbon (PAC) and 
extruded activated carbon (EAC). Activated carbon fiber (ACF) is a relatively new 
adsorbent for filtration or purification techniques. Currently used carbon-based flue gas 
desulfurization systems have mostly consisted of granular activated carbons [23]. The 
sulfur dioxide capacity of bare activated carbon usually decreases with the increased 
temperatures. In order to address this problem, they are loaded with metals which result 
in good sulfur dioxide adsorbents at both low and relatively high temperatures.   
1.8 Advantages of ACFs in comparison with GAC and PAC 
The desulfurization properties of the four activated carbons mentioned above have been 
compared for material selection to optimize sulfur dioxide purification technique. GAC 
and PAC are widely used in a variety of applications, such as gas purification, water 
purification, medicine manufacturing, sewage treatment, chemical purification, mercury 
scrubbing and filter masks [24].  A large volume of adsorption devices are usually needed 
in applications due to the relatively slow adsorption rate of GAC, which also results in 
slow throughput. Typical size of GAC varies between 200 and 1000 ?m. PAC, which is 
smaller than GAC with the typical particle size between 20 and 200 ?m, is found to have 
19 
faster adsorption rates. However, powdered activated carbons may compact under flow, 
leading to a strong flow resistance [25].  
The adsorption kinetics and capacities of GAC, PAC and ACF on methylene blue were 
investigated by Shmidt et al. [25]. The adsorption rate of ACF is an order of magnitude 
faster than that by PAC and two orders of magnitude faster than that by GAC. The 
adsorption capacity of ACF was three times higher than the adsorption capacity of PAC 
and ten times higher than that of GAC. However, the diffusion coefficient in ACF was 
found to be 2?10-11 cm2/s in comparison with 4.6?10-10 cm2/s for PAC and 3.5?10-8 cm2/s 
for GAC [25].  Although ACFs possess a smaller diffusion coefficient, which may be 
because of the tighter packed carbon layers, compared to ACP and GAC, they exhibit a 
faster adsorption rate than GAC and PAC. This is due to the ratio of macropore size to 
the adsorbent size. For ACFs, the macropores are relatively large as compared to fiber 
diameters and therefore a single macropore may lead directly into the center of the ACF. 
Macropores in PAC and GAC are comparable in size to macropores in ACF but 
macropores in PAC and GAC are much smaller than the diameters of PAC and GAC 
particles. Penetration of adsorbate into the center of PAC and GAC is controlled by 
diffusion through micropore layers that interconnect individual macropores [25]. In other 
words, the higher the ratio of macropore size to the adsorbent size, the faster the 
adsorption rate. The effective macropore surface area was estimated by Shmidt as well 
[25]. ACF was found to have the smallest macropore surface area of 6.3 m2/g, which is 
half of macropore surface area of PAC at 14 m2/g and one fourteenth of that of GAC at 
87 m2/g [25]. However, the influence of effective macropore surface area on the 
adsorption rate is less than that caused by the adsorbent diameters. Application of PAC is 
20 
limited by possible second contamination [26]. In order to improve the contacting 
efficiency, a technique which entraps PAC in microfiber network has been patented [27, 
28].  
ACFs have their own characteristics compared to GAC and PAC [26]. First of all, the 
features of pores in ACFs are different from GAC. Most of the pores in ACFs are 
micropores with the presence of some mesopores and macropores. Due to the small fiber 
diameters, pores can directly reach the center of fibers and therefore the diffusion path of 
adsorbate to reach the adsorption sites is short. The pore schematics and pore size 
distribution of typical GAC and ACF are shown in Figures 1.7 and 1.8, respectively. The 
more open characteristic of ACF may contribute to the less possibility to be blocked. 
GAC and PAC have plenty of micropores and mesopores which adsorbates usually reach 
after passing macropores. The blockage of macropores would close the access of 
adsorbates to smaller pores where adsorption would occur. ACF usually has a narrower 
pore size distribution and a higher surface area. Small fiber diameters also result in a 
larger contact surface area between adsorbents and adsorbates and thus even contact 
would be achieved. All of these characteristics of ACF may explain the faster adsorption 
rate and the higher adsorption capacity in desulfurization process [29]. ACFs are able to 
remove not only high concentration contaminants but also low concentration 
contaminants. For instance, 100 ppm toluene can pass through GAC while the lowest 
toluene concentration ACF can absorb is 10 ppm [26]. ACFs have also been found to 
have lower pressure drop as compared to GAC [29]. ACF has a lower volume density and 
lower filtration resistance, which is about 1/3 of GAC. Therefore, unlike GAC, ACFs are 
able to filter liquids with a higher viscosity and less penalty of energy consumption. 
21 
ACFs offer the fabrication flexibility which GAC does not have, and thus they are easier 
to be made into filter media or devices. ACFs can be made in various forms such as 
fabrics, nonwovens, paper and composites, which make them suitable for handling in 
special devices [29-32]. Zhao et al. pointed to some merits of ACFs in specific 
applications. In water purification systems, ACFs have advantages of less possibility to 
produce ash, less pressure drop and a higher water flow rate.  In air purification systems, 
ACFs show faster adsorption and desorption rate. In ammonia, toluene, sulfur dioxide 
and nitrogen dioxide cleaning systems, ACFs showed the higher removal efficiency and 
adsorption capacities [26].  
Figure 1.8 Pore size distribution of ACF and GAC [24]
Figure 1.7 Schematic of pore structures of a) ACF and b) GAC [24] 
(a) ACF                        (b) GAC
22 
Furthermore, the desulfurization abilities of GAC (made from industrial coke) and 
viscose-based ACF were tested and compared under the same conditions (70?C) with a 
gas stream of 3 L/min by Xu et al [33]. With the presence and absence of water, GACs 
reached their breakthrough points quickly after the test started while ACFs showed 55 
min (with moisture) and 20 min (without moisture) breakthrough time, respectively. 
However, the utilization of ACFs is low with the short breakthrough period, which is due 
to very small internal pore sizes of ACFs, compared to sulfur dioxide molecules (bond 
distance S-O=1.432??0.003? and bond angle OSO=119.53??0.4?). Capillary 
condensation  
takes place within these pores, resulting in large amounts of adsorption [33]. PAN-based 
ACFs have been reported by Mochida et al. to have the largest sulfur dioxide removal 
efficiency among the commercially available active carbons [34].  
1.9 Manufacturing and characterization of activated carbon fibers  
Manufacturing of activated carbon fibers can be basically classified in terms of the 
preparation method: physical activation (or gas activation) and chemical activation (or 
reagent activation). Similar to the preparation of other shapes of activated carbons, there 
are three basic procedures to make activated carbon fibers: stabilization or oxidation, 
carbonization and activation. The stabilization is usually carried out at 200~300?C in air 
for prevention of fibers from melting during the subsequent high temperature 
carbonization.  Stabilization is required for producing ACF from PAN and pitch fibers, 
but not those from phenolic resin and cellulose because of their thermosetting nature [35]. 
In the carbonization process, non-carbon elements, such as oxygen and nitrogen, are 
eliminated through high temperature heat treatment; residual elementary carbon atoms 
23 
group into sheets and form condensed aromatic rings. Porous structure is initiated at this 
stage; however, the adsorption capacity of carbon fibers is still insignificant because most 
of pore mouths are still blocked with some tarry substances [36], which needs the next 
stage called activation. During activation, disorganized carbon groups are rearranged and 
micropores, mesopores and macropores are generated with the help of activating agents. 
The difference between physical and chemical activation lies in the activating agents and 
thus the activation process.   
In chemical activation, raw materials are first immersed in inorganic salt, such as zinc 
chloride, phosphoric acid or potassium sulfide [37]. The most popular activation agent in 
industry was zinc chloride, however, its use has been discontinued due to its damage to 
environment [38]. Lately, ammonia, potassium hydroxide and sodium hydroxide have 
been reported as the reagents to make activated carbon fibers chemically [39-41]. 
Advantages of chemical activation include the short preparation time since carbonization 
and activation are combined into one step making chemical activation a two-step process 
and the lower activation temperature. 
Unlike chemical activation, the processes of carbonization and activation take place in 
two separate steps in physical activation. Several physical processes are involved in 
physical activation including the removal of the tarry products condensed in the pore by 
the flow of activating agents and the increase of volume as a result of diffusion and 
evaporation of the tarry products [42].   
The most widely used precursors for activated carbons fibers include polyacrylonitrile, 
carbon fiber, petroleum based pitch, viscose rayon, and phenolic resin [37, 43-45]. Some 
other fibers have been tried lately to produce ACFs as well, such as polyvinyl chloride, 
24 
Nomex? aramid fibers and polyamide [37, 44, 46, 47]. Pore characteristics and physical 
properties of activated carbon fibers are influenced by a series of factors, including 
chemistry composition of precursors, activating agents, heating temperature, process 
period, and flow rate of environmental gases [12, 37, 38, 40, 48].  
1.10 Sulfur dioxide adsorption over activated carbon fibers 
1.10.1 Sulfur dioxide removal on bare activated carbon fibers 
Desulfurization properties of bare ACFs have been studied by some researchers and are 
found to be dependent on precursors of ACFs, characteristics of ACFs and experimental 
conditions, such as the composition of carrier gas, flow rate of main gas stream, 
temperature, and concentration of challenge gas [33, 49].  
    Compared to pitch-based ACFs, polyacrylonitrile-based activated carbon fibers (PAN-
based ACFs) were found to have higher activity while they were subjected to 1000 ppm 
sulfur dioxide with a flow rate of 100 mL/min at 30 and 50?C. Breakthrough times were 
longer than 60 hours [50].   
    Adsorption capacity of SO2 onto ACFs is dependent not only on textural 
characteristics, but also surface chemistry, quantity and nature of oxygen-containing and 
nitrogen-containing functional groups [33, 49]. ACFs impregnated in ammonia increased 
the nitrogen-containing surface groups (pyidyl  C5H4N- and C4H4N-) leading to increased 
SO2 adsorptive performance [33].  
    Mochida et al. claimed that physical properties of PAN-based ACFs, such as surface 
area, micropore volume and external surface area appear to be less influential on their 
SO2 removal performance [51]. However, Daley et al. obtained different results on 
phenolic-based ACFs. They found out that initial rate of SO2 adsorption was inversely 
25 
related to pore size, while, at longer times, the amount of SO2 adsorbed was dependent on 
both the pore size and pore volume [52]. ACFs may be oxidized with nitric acid/sulfuric 
acid to introduce more oxygen groups on fiber surfaces followed by heat treatment which 
results in decomposition of oxygen groups to develop free sites which was found to be 
proportional to SO2 adsorption [52, 53]. Activation method also was found to have effect 
on the activity of ACFs in sulfur dioxide capacity. Activity of steam activated ACFs 
continuously declines while the potassium hydroxide activated ACFs produced high 
steady stage sulfur dioxide conversion which was probably due to high oxygen content 
on steam activated ACFs according to Martin-Gullon?s research [54].  
1.10.2 Treatment for promotion of sulfur dioxide removal  
Attempts have been made to increase sulfur dioxide removal efficiency with treated 
activated carbons with various approaches: heat treatment, acid pretreatment, loading of 
base and loading of alkali metals. 
1.10.2.1. Heat treatment Heat treatment at high temperatures under while nitrogen 
atmosphere was found to have effect on sulfur dioxide capacity of activated carbon. Pore 
structures of activated carbons were found to be influenced by both acid pretreatment and 
heat treatment: surface area and pore volume decreased a little by oxidation of acid and 
increased again by heat treatment while pore size was barely affected [52]. 
    Both active carbons and PAN-based ACFs were subjected to high temperature heat 
treatment in order to increase SO2 capacity. The capacity of ACFs was increased 
drastically while only a slight increase of capacity was observed for GAC. At 800?C 
heating, SO2 capacity of ACFs was tripled [51].  
26 
    To maximize the SO2 capacity, appropriate temperature of heat treatment is expected 
to vary with ACFs because the number and natures of oxygen groups on the surface are 
strongly subjective to the extent of activation [51]. The mechanism was proposed by 
Mochida and two different adsorption sites were proposed. One is for SO2 adsorption and 
oxidation and the other one is for holding H2SO4. Polar sites can hold H2SO4; however, 
too many polar sites in the unit area may cause repulsion of polar adsorbate or form the 
liquid phase of H2SO4 at the pore mouth, inhibiting the diffusion of H2SO4 as well as SO2
into pores. Therefore, a balanced number of oxygen groups prepared by heating is 
optimal for maximum SO2 adsorption [51].  
    Pretreatment with nitric acid or sulfuric acid has been applied before heat treatment of 
activated carbon. Oxidation of activated carbon by nitric acid could offer oxygen groups 
which would be decomposed with high temperatures leaving free sites behind for 
enhancement of SO2 adsorption [52, 55]. Oxidized ACFs showed a reduction of SO2
capacity at low temperature although it was regained by subsequent heat treatment. 
Heating could not increase the SO2 capacity indefinitely, which was limited by the 
available pore volume [52].  
1.10.2.2. Impregnation with basic chemicals Considering the acid of sulfur dioxide, basic 
chemicals, e. g., K2CO3, KOH and ammonia, can be loaded on activated carbon for 
desulfurization. Two impregnation methods are usually employed to deposit catalysts on 
porous supports: regular impregnation and wetness incipient impregnation. In the former, 
porous materials are simply immersed in desired precursor solutions followed by drying 
to deposit chemicals on supports. The wetness incipient impregnation is a commonly 
used technique for the synthesis of heterogeneous catalysts in which a solution is added 
27 
to a support containing the same pore volume as the volume of solution that was added. 
Capillary action draws the solution into pores. The catalyst can be dried and calcined to 
drive off the volatile components within the solution, depositing chemicals on the catalyst 
surface. However, the maximum loading is limited by the solubility of the precursor in 
the solution. The concentration profile of the impregnated compound depends on the 
mass transfer conditions within the pores during impregnation and drying [56].  
    Activated carbon impregnated with potassium hydroxide was reported to achieve 
simultaneous removal of sulfur dioxide, nitrogen monoxide and nitrogen dioxide by Lee 
et al. [57]. The sulfur dioxide capacity of K2CO3-impregnated activated carbons prepared 
by soaking varies linearly with K2CO3 loading up to 12wt%. Above 12wt%, the capacity 
levels out and then decreases since high loads of K2CO3 may aggregate and block pores 
of activated carbon. However, it was reported that the incipient wetness impregnation 
may cause aggregation but did not block carbon pores [58].  The release of carbon 
monoxide during desulfurization limits the application of K2CO3 in PEMFC systems due 
to poisoning effect of carbon monoxide on the fuel cell unit. Xu et al. impregnated GAC 
and viscose-based ACFs in ammonia. The impregnation of activated carbon fibers in 
ammonia introduced more nitrogen-containing groups (pyidyl  C5H4N- and C4H4N-) on 
carbon surface enhancing SO2 capacity [33].   
1.10.2.3. Activated carbon supported metal catalysts Several kinds of metal were treated 
onto the activated carbons to increase the SO2 removal efficiency or SO2 adsorption 
capacity [29, 55, 59, 60]. Transition metals, such as Fe, Ni, Co, Mn and V were deposited 
on pitch-based activated carbon by immersing the carbon in aqueous solution of metal 
nitrates followed by filtration, washing, drying and calcinations at 800?C in nitrogen 
28 
flow. Properties of metallic derivatives, such as BET surface area, acidic and basic 
surface characteristics, were found to be similar to the bare activated carbon without 
treatment. With the nitrogen as the carrier gas, AC supported with vanadium is the most 
favorable SO2 adsorbent in comparison with other four metal-AC sorbents. Moreover, 
SO2 capacities of V/AC and Fe/AC sorbents were more likely to be promoted by the 
oxygen and water in the carrier gas (185 mg SO2/g V-AC and 172 mg SO2/g Fe-AC in 
the mixture of SO2+O2+H2O for 0.3 wt% vanadium and 0.8 wt% iron, respectively) [60].  
Polyacrylonitrile-based ACFs were loaded with Cu, Co, Mg, Mn and Ni and increased 
SO2 activity was obtained. Among these metal-ACFs, Co/ACF (34wt% of cobalt) 
showed the best activity (93 mg SO2/g carbon) and a constant removal ratio of sulfur 
dioxide above 87% during continuous exposure to the flow of SO2+O2+H2O with the 
flow rate of 125 mL/min for more than 216 hour [29].  
     Copper catalysts supported on acid treated AC were designed and studied by Tseng et 
al., who found out that carbon acted as catalyst support for copper oxide provided a 
higher catalytic activity for sulfur dioxide. Sulfur dioxide was catalytically oxidized to 
sulfur trioxide which was converted be copper sulfate. Acid pretreatment on AC before 
loading copper dioxide provided the surface oxygen content to enhance the activity of 
sulfur dioxide oxidation [55]. The regeneration of spent CuO/AC was achieved by 
heating in inert gases between 260 and 480?C without the need of reducing agents. The 
decomposition of surface oxygen complexes on carbon surfaces resulted in the evolution 
of CO2 and CO which act as the reducing agents in regeneration process [59].  However, 
during regeneration, the over reduction of Cu2+ to Cu0 may occur, which my results in 
sintering of copper reducing the desulfurization capacity of regenerated Cu/carbon. Ma et 
29 
al. developed a regenerable Fe/AC desulfurizer which may have a longer service life than 
Cu/AC [22]. H2SO4 and Fe2(SO4)3 were produced during the desulfurization process and 
the spent Fe/AC desulfurizer was able to be regenerated by NH3 at 350?C to directly form 
solid ammonium-sulfate salts [22].  
1.11 Objectives of the research  
Because of advantages of ACFs in the adsorption rate and the fabrication flexibility, they 
are considered in this study to removal sulfur dioxide from air fed into fuel cell systems. 
According to literature review, modification on ACFs attempting to promote sulfur 
dioxide adsorption has focused on heat treatment and metal load for catalytically 
oxidation of sulfur dioxide to sulfur trioxide. However, both heat treatment and 
impregnation of metal require high temperature and inert gas, which are energy 
consuming. Impregnating of activated carbon fibers with alkali and oxidants will address 
this problem without the need of nitrogen as the protecting gases and high temperature. 
The adsorption mechanism of ACF supported oxidants will be studied. Since the current 
available models in the literature are estimating SO2 adsorption over bare activated 
carbon fibers, models for prediction of adsorption performance of ACFs with promoters 
(potassium permanganate) will be developed for different conditions: the absence of 
water and oxygen, the presence of oxygen, the presence of water and the copresence of 
water and oxygen.  Although model for the prediction of adsorption of sulfur dioxide and 
other adsorbate over bare activated carbon fibers have been reported by the literature, the 
impregnated ACFs or AC are barely described in these models. In the current research, a 
model will be developed to predict the adsorption properties of the modified ACFs. 
30 
References
1. Fuel Cell Handbook (Sixth Edition), U.S. Department Of Energy, West Virginia, 
(2002). 
2. Podatz, P., Onder, C. and Guzzella, L., Air supply system of a PEMFC stack 
dynamic model, Fuel Cells, 5 (1), 126-133, (2005). 
3. Larminie, J. and Dicks, A., Fuel cell systems explained, John Wiley & Sons Ltd, 
West Sussex, England, (2003). 
4. Cheng, X., Shi, Z., Glass, N., Zhang, L., Zhang, J. J., Song, D. T., Liu, Z. S., 
Wang, H. J. and Shen, J., A review of PEM hydrogen fuel cell contamination: 
Impacts, mechanisms, and mitigation, J. Power Sources, 165 (2), 739-756, 
(2007). 
5.  www.wikipedia.org, Jan 2010 accessed. 
6. Moore, J. M., Adcock, P. L., Lakeman, J. B. and Mepsted, G. O., The effects of 
battlefield contaminants on PEMFC performance, Journal of Power Sources, 85 
(2), 254-260, (2000). 
7. Zhang, J. J., Wang, H. J., Wilkinson, D. P., Song, D. T., Shen, J. and Liu, Z. S., 
Model for the contamination of fuel cell anode catalyst in the presence of fuel 
stream impurities, Journal of Power Sources, 147 (1-2), 58-71, (2005). 
8. Collier, A., Wang, H. J., Yuan, X. Z., Zhang, J. J. and Wilkinson, D. P., 
Degradation of polymer electrolyte membranes, International Journal of 
Hydrogen Energy, 31 (13), 1838-1854, (2006). 
9. Jiang, R., Kunz, H. R. and Fenton, J. M., Electrochemical oxidation of H2 and 
H2/CO mixtures in higher temperature (Tcell>100
?C) proton exchange membrane 
31 
fuel cells: electrochemical impedance spectroscopy, Journal of The 
Electrochemical Society, 152 (7), A1329-A1340, (2005). 
10. Han, J. N., Park, G. G., Yoon, Y. G., Yang, T. H., Lee, W. Y. and Kim, C. S., A 
new evaluation method of anode/cathode used for polymer electrolyte membrane 
fuel cell, International Journal of Hydrogen Energy, 28 (6), 609-613, (2003). 
11. Mohtadi, R., Lee, W. K., Cowan, S., Van Zee, J. W. and Murthy, M., Effects of 
hydrogen sulfide on the performance of a PEMFC, Electrochemical and Solid 
State Letters, 6 (12), A272-A274, (2003). 
12. Yang, D. J., Ma, J. X., Xu, L., Wu, M. Z. and Wang, H. J., The effect of nitrogen 
oxides in air on the performance of proton exchange membrane fuel cell, 
Electrochimica Acta, 51 (19), 4039-4044, (2006). 
13. Kandylas, I. P., Haralampous, O. A. and Koltsakis, G. C., Diesel soot oxidation 
with NO2: Engine experiments and simulations, Industrial & Engineering 
Chemistry Research, 41 (22), 5372-5384, (2002). 
14. Betournay, M. C., Bonnell, G., Edwardson, E., Paktunc, D., Kaufman, A. and 
Lomma, A. T., The effects of mine conditions on the performance of a PEM fuel 
cell, Journal of Power Sources, 134 (1), 80-87, (2004). 
15. Mohtadi, R., Lee, W. K. and Van Zee, J. W., Assessing durability of cathodes 
exposed to common air impurities, Journal of Power Sources, 138 (1-2), 216-225, 
(2004). 
16. Jing, F. N., Hou, M., Shi, W. Y., Fu, J., Yu, H. M., Ming, P. W. and Yi, B. L., 
The effect of ambient contamination on PEMFC performance, Journal of Power 
Sources, 166 (1), 172-176, (2007). 
32 
17. Kennedy, D. M., Cahela, D. R., Zhu, W. H., Westrom, K. C., Nelms, R. M. and 
Tatarchuk, B. J., Fuel cell cathode air filters: Methodologies for design and 
optimization, Journal of Power Sources, 168 (2), 391-399, (2007). 
18. Kennedy, D., Fuel cell cathode air filters: methodologies, Master Thesis. Auburn, 
Auburn University, (2007). 
19.  www.ehow.com, Jan 2010 accessed. 
20. Huda, M., High-efficiency SO2 removal using activated carbon fibers, (2007).  
21. Grutzeck, M., SO2 removal from flue gases using utility synthesized zeolites, 
Report, The Pennsylvania State University, (1999). 
22. Ma, J. R., Liu, Z. Y., Liu, S. J. and Zhu, Z. P., A regenerable Fe/AC desulfurizer 
for SO2 adsorption at low temperatures, Applied Catalysis B-Environmental, 45 
(4), 301-309, (2003). 
23. Debarr, J. A., Lizzio, A. A. and Daley, M. A., Adsorption of SO2 on bituminous 
coal char and activated carbon fiber, Energy & Fuels, 11 (2), 267-271, (1997). 
24. Wey, M. Y., Fu, C. H., Tseng, H. H. and Chen, K. H., Catalytic oxidization of 
SO2 from incineration flue gas over bimetallic Cu-Ce catalysts supported on pre-
oxidized activated carbon, Fuel, 82 (18), 2285-2290, (2003). 
25. Shmidt, J. L., Pimenov, A. V., Lieberman, A. I. and Cheh, H. Y., Kinetics of 
adsorption with granular, powdered, and fibrous activated carbon, Sep. Sci. 
Technol., 32 (13), 2105-2114, (1997). 
26. Zhao, L., Huang, X., Gu, Q. and Di, Y., The performance comparison of ganular 
activated carbon filter and fiber filter of activated carbon, Contamination Control 
& Air-conditioning Technology, (2), (2004). 
33 
27. Tatarchuk, B., Chang, B. K., Lu, Y., Luna, E. and Cahela, D.,  Auburn University, 
Microfibrous entrapment of small reactive particulates and fibers for high 
contacting efficiency removal of contaminants for gaseous or liquid streams, 
patent 7501012, USA,(2009). 
28. Tatarchuk, B., Chang, B. K., Lu, Y., Chen, L., Luna, E. and Cahela, D.,  Auburn 
University, Microfibrous entrapment of small reactive particulates and fibers for 
high contacting efficiency removal of contaminants from gaseous or liquid 
streams, patent 20050169820, USA,(2005). 
29. Wang, J. Y., Zhao, F. Y., Hu, Y. Q., Zhao, R. H. and Liu, R. J., Modification of 
activated carbon fiber by loading metals and their performance on SO2 removal, 
Chin. J. Chem. Eng., 14 (4), 478-485, (2006). 
30. Edens, R. L., Gadsby, E. D., Lindsay, J. D., Pike, R. D. and Mangun, C.,  
Kimberly-Clark Worldwide, Inc. , Flexible activated carbon substrates, patent 
20040121688, USA,(2004). 
31. Hsieh, S.-J.,  Clothing material structure of fiber cloth containing PAN series 
activated carbon, patent 20040092188,(2004). 
32. Giglia, R. and Battistelli, E.,  American Cyanamid Company Non-woven 
activated carbon fabric, patent 4904343, USA,(1990). 
33. Xu, L., Guo, J., Jin, F. and Zeng, H., Removal of SO2 from O2-containing flue gas 
by activated carbon fiber (ACF) impregnated with NH3, Chemosphere, 62, 823-
826, (2006). 
34 
34. Mochida, I., Hirayama, T., Kisamori, S., Kawano, S. and Fujitsu, H., Marked 
increase of SO2 removal ability of polyacrylonitrile-based activated carbon fiber 
by heat treatment at elevated temperatures, Langmuir, 8, 2290-2294, (1992). 
35. Worasuwannarak, N., Hatori, S., Nakagawa, H. and Miura, K., Effect of oxidation 
pre-treatment at 220 to 270 degrees C on the carbonization and activation 
behavior of phenolic resin fiber, Carbon, 41 (5), 933-944, (2003). 
36. Gaur, V., Preparation and characterization of activated cabron fabric for the 
control of BTX, SO2 and NOx emissions, Indian Institute of Technology Kanpur, 
(2005). 
37. Wu, Q. L., Zhang, Z. H., Gu, S. Y. and Gong, J. H., Macroporous activated 
carbon fibers from rayon precursors impregnated with phosphoric acid, Journal of 
Macromolecular Science Part A - Pure and Applied Chemistry, 43 (11), 1787-
1792, (2006). 
38. Su, C. I. and Wang, C. L., Optimum manufacturing conditions of activated carbon 
fiber absorbents. I. Effect of flame retardant reagent concentration, Fibers and 
Polymers, 8 (5), 477-481, (2007). 
39. Shim, J. W., Park, S. J. and Ryu, S. K., Effect of modification with HNO3 and 
NaOH on metal adsorption by pitch-based activated carbon fibers, Carbon, 39 
(11), 1635-1642, (2001). 
40. Lee, Y. J., Kim, J. H., Kim, J., Lee, D. B., Lee, J. C., Chung, Y. J. and Lim, Y. S., 
Fabrication of activated carbon fibers from stabilized PAN-based fibers by KOH, 
Designing, Processing and Properties of Advanced Engineering Materials, Pts 1 
and 2, 449-4, 217-220, (2004). 
35 
41. Li, K., Ling, L., Liu, L., Zhang, B. and Liu, Z., Desulfurization of active carbon 
fibers activated with ammonia water, Acta Scientiae Circumstantiae, 21 (1), 74-
78, (2001). 
42. Gaur, V., Preparation and characterization of activated carbon fabric for the 
control of BTX, SO2 and NOx emission (Dissertation), Inidian Institute of 
Technology Kanpur, (2005). 
43. Moon, S. Y., Kim, M. S., Hahm, H. S. and Lim, Y. S., Preparation of activated 
carbon fibers by chemical activation method with hydroxides, Eco-Materials 
Processing & Design VII, 510-511, 750-753, (2006). 
44. Qiao, W. M., Yoon, S. H., Korai, Y., Mochida, I., Inoue, S., Sakurai, T. and 
Shimohara, T., Preparation of activated carbon fibers from polyvinyl chloride, 
Carbon, 42 (7), 1327-1331, (2004). 
45. Liu, C. L., Wen, Y. H., Cheng, J., Guo, Q. G., Cao, G. P., Liu, L. and Yang, Y. S., 
Influence of pore structure of phenolic resin based activated carbon fibers on the 
electrochemical performance of electrical double-layer capacitors, Acta Physico-
Chimica Sinica, 21 (7), 786-791, (2005). 
46. Suarez-Garcia, F., Martinez-Alonso, A. and Tascon, J. M. D., Beneficial effects 
of phosphoric acid as an additive in the preparation of activated carbon fibers 
from Nomex aramid fibers by physical activation, Fuel Process. Technol., 77, 
237-244, (2002). 
47. Villar-Rodil, S., Suarez-Garcia, F., Paredes, J. I., Martinez-Alonso, A. and 
Tascon, J. M. D., Activated carbon materials of uniform porosity from polyaramid 
fibers, Chem. Mater., 17 (24), 5893-5908, (2005). 
36 
48. Park, S. J., Kim, K. D. and Lee, J. R., Influence of activation temperature on 
surface and adsorption properties of PAN-based activated carbon fibers/phenolic 
resin matrix composites, Polymer-Korea, 24 (1), 97-104, (2000). 
49. Li, K., Ling, L., Lu, C., Liu, Z., Liu, L. and Mochida, I., Influence of CO-
evolving groups on the activity of activated carbon fiber for SO2 removal, Fuel 
Process. Technol., 70 (3), 151-158, (2001). 
50. Kisamori, S., Kuroda, K., Kawano, S., Mochida, I., Matsumura, Y. and 
Yoshikawa, M., Oxidative Removal of SO2 and Recovery of H2SO4 over 
Poly(Acrylonitrile)-Based Active-Carbon Fiber, Energy & Fuels, 8 (6), 1337-
1340, (1994). 
51. Mochida, I., Hirayama, T., Kisamori, S., Kawano, S. and Fujitsu, H., Marked 
Increase of SO2 Removal Ability of Poly(Acrylonitrile)-Based Active-Carbon 
Fiber by Heat-Treatment at Elevated-Temperatures, Langmuir, 8 (9), 2290-2294, 
(1992). 
52. Daley, M. A., Mangun, C. L., Debarr, J. A., Riha, S., Lizzio, A. A., Donnals, G. 
L. and Economy, J., Adsorption of SO2 onto oxidized and heat-treated activated 
carbon fibers (ACFS), Carbon, 35 (3), 411-417, (1997). 
53. Lizzio, A. and Debarr, J., Mechanism of SO2 removal by carbon, Energy & Fuels, 
11, 284-291, (1997). 
54. Martin-Gullon, I., Jagtoyen, M., Kimber, G. and Derbyshire, F.,  Activated carbon 
fibers from PAN. II. Catalytic activity for SO2 oxidation, The American Carbon 
Society 1997 Proceedings - 23rd Biennial Conference,  The Pennsylvania State 
University, University Park, PA,(1997). 
37 
55. Tseng, H. H., Wey, M. Y. and Fu, C. H., Carbon materials as catalyst supports for 
SO2 oxidation: catalytic activity of CuO-AC, Carbon, 41 (1), 139-149, (2003). 
56.  http://en.wikipedia.org/wiki/Incipient_Wetness_Impregnation, Jan 2010 
accessed. 
57. Lee, Y. W., Kim, H. J., Park, J. W., Choi, B. U., Choi, D. K. and Park, J. W., 
Adsorption and reaction behavior for the simultaneous adsorption of NO-NO2 and 
SO2 on activated carbon impregnated with KOH, Carbon, 41 (10), 1881-1888, 
(2003). 
58. Fortier, H., Zelenietz, C., Dahn, T. R., Westreich, P., Stevens, D. A. and Dahn, J. 
R., SO2 adsorption capacity of K2CO3-impregnated activated carbon as a function 
of K2CO3 content loaded by soaking and incipient wetness, Appl. Surf. Sci., 253 
(6), 3201-3207, (2007). 
59. Tseng, H. H. and Wey, M. Y., Study of SO2 adsorption and thermal regeneration 
over activated carbon-supported copper oxide catalysts, Carbon, 42 (11), 2269-
2278, (2004). 
60. Davini, P., The effect of certain metallic derivatives on the adsorption of sulphur 
dioxide on active carbon, Carbon, 39 (3), 419-424, (2001). 
  
38
Chapter 2 Manufacturing of Activated Carbon Fibers via
Electrospinning 
2.1 Introduction 
Electrospinning, a simple approach to make very fine fibers ranging from nano to micro 
scales, is attracting more attention due to high surface area, high porosity and high 
surface area to volume ratio of electrospun fibers. These properties make them suitable 
for manufacturing carbon and graphitic nanofibers, tissue scaffold and drug delivery, 
filtration and reinforced nanocomposites, etc. Dependence of fiber diameter and 
electrospun nonwoven structure on polymer viscoelasticity and electrostatic field formed 
between two electrodes has been reported by many researchers. Although composites 
filled with nanofibers have shown enhanced mechanical properties with respect to 
unfilled composites, compared to cast films, electrospun membranes had a reduction in 
peak tensile strength [1]. However, not much information is available on mechanical 
properties of electrospun nanofiber nonwoven membranes. 
    Activated carbon fiber, a relatively new adsorbent, offers wonderful adsorption 
properties, such as faster adsorption rate, longer life and lower pressure drop, in 
comparison with traditional granular activated carbons. Therefore, they are widely used 
in separation of gases, recovery of solvents, removal of organic pollutants from drinking 
water and catalyst support [2, 3]. Rayon, pitch and PAN fibers are mainly used as 
39
precursors for activated carbon fibers. The advantage of finer fibers to make activated 
carbon fibers has been studied by Park et al. The electropun ultrafine pitch fibers have 
faster activation rate constant than melt-spun pitch fibers due to high surface area [4]. 
Kim et al., activated stabilized electrospun PAN with steam and three different surface 
areas, namely 850 m2/g, 1000 m2/g and 1230 m2/g were obtained at various activation 
temperatures [5, 6]. They also tried poly (amic acid) as precursor to make activated 
carbon nanofibers [7]. These studies mentioned above employed physical activation to 
create pores in precursor fibers. Compared to physical activation, chemical activation 
process has important advantages including low heat treatment temperature, short period 
of time, large surface area and high carbon yield [8]; however, there has been no work 
reported for chemically activated carbon nanofibers.  
    In our work, the effects of main electrospinning parameters on resulting nanofiber 
diameters were studied. Mechanical properties, such as tensile, tear and burst strength of 
electrospun PAN nonwoven membrane structures were measured and the quantitative 
relationships between membrane thickness and membrane properties were established. 
Other physical properties, such as air permeability, pore size and porosity were studied as 
well. Activated carbon nanofibers were produced from electrospun PAN by chemical 
activation with potassium hydroxide (KOH) as the activating agent. They were 
characterized by morphology, Fourier transform infrared (FTIR) spectrum, Brunauer-
Emmeet-Teller (BET) surface area, total pore volume and pore size distribution.  
40
2.2 Materials and experimental procedures 
Polyacrylonitrile (PAN, Mw=150,000g/mol from Sigma-Aldrich?) was dissolved in N,N-
dimethylformamide (DMF) (from Sigma-Aldrich?) and stirred for 24 hours. PAN/DMF 
solution was stored in BD 10mL syringe and delivered by KD? scientific syringe pump 
through BD 18G 1? precision glide needle (outside diameter 1.27 mm+/-0.0127 mm; 
inner diameter 0.711 mm+/-0.0127 mm). Under the accelerating voltage obtained from 
Gamma high voltage power supply (0-30 kV DC), polymer jets were stretched by the 
electrostatic force and columbic force from charged polymer. After the DMF solvent 
evaporated during stretching, nanofibers were deposited on aluminum foil (the negative 
electrode) and the electrospun PAN nonwoven membranes were formed.  
    Activated carbon nanofibers were obtained from electrospun PAN nanofibers with 
stabilization and activation processes. Electrospun PAN nanofibers were stabilized in air 
by raising temperature from 20?C to 280 ?C at 1 ?C /min and holding at 280?C for 1 hour, 
which prevented fibers from sticking to each other for further treatment at higher 
temperature. The stabilized PAN was immersed in KOH (Spectrum?) solution for 12 
hours followed by drying up at 85. The weight ratio of KOH to stabilized PAN was 1, 
at which the highest surface area was reported [8]. The stabilized electrospun PAN 
containing KOH was activated at 800 for 40 min under nitrogen atmosphere to 
generate micropores within fibers. The residual KOH on fibers was neutralized with 
hydrogen chloride acid (36.5-38% concentration, from J.T. Baker) and the extra HCl was 
rinsed with distilled water.    
    Morphology of electrospun PAN nonwoven was observed under Scanning Electron 
Microscope (SEM, Zeiss? EVO 50) and fiber diameters were measured with SEM built-
41
in software. The electrospun PAN membranes were cut into 25.4?101.6 mm rectangle for 
tensile strength tests and into 50.8?76.2 mm for tear strength tests on Instron tester 
(Model 1122). Burst strength test was carried out on Mullen burst tester (B.F. Perkins & 
Son. Inc.). Air permeability was measured on Gurley Densometer. Interfiber pore size 
within membrane was tested with Capillary Flow Porometer (Porous Materials Inc.) 
using Galwick with the surface tension of 15.9 dynes/cm as the solvent for wet curve. 
Activated carbon nanofibers were too brittle for these physical property measurements. 
BET surface area, average pore size and pore volume of electrospun PAN and activated 
carbon nanofibers were analyzed on Autosorb-1 (Quantachrome? Instruments) using 
nitrogen as absorbate at 77 K. FTIR (Thermo?) was used for analyzing the structure 
evolution during stabilization and activation process. Pore size distribution analysis was 
carried out on AS1 WIN software (Quantachrome? Instruments). Dubinin-Astakhov 
(DA) equation is good for analyzing activated carbon nanofibers and the generalized 
Halsey equation is more suitable for determination of micropore volume of electrospun 
PAN in which mesopores might be present [9] . Both analysis methods were employed 
for comparison.
2.3 Results and discussion 
2.3.1 Influence of electrospinning parameters on fiber diameters  
To study the impact of individual electrospinning parameters on fiber diameters, a series 
of experiments were performed. One parameter was changed as all others were fixed 
during any specific experiment.   
2.3.1.1 Solution concentration. At 20 kV voltage, the needle tip-collector distance was set 
at 20 cm, the flow rate at 1 mL/h and PAN/DMF solutions with concentrations of 8, 10, 
42
12 and 14 wt% were electrospun. The observed morphology of resultant PAN nanofiber 
nonwoven is illustrated in Figure 2.1. Fiber diameters vary depending on solution 
concentrations. The average fiber diameters for 8, 10, 12 and 1 4 wt % were 170 nm, 287 
nm, 438 nm and 519 nm, respectively. The relationship between polymer solution 
concentration and fiber diameters is shown in Figure 2.2. The steady increase of fiber 
diameter is observed as the solution concentration increases. The increasing solution 
concentration contributes to increasing molecular entanglement and higher viscoelasticity 
resulting in difficulty in streching fibers, which leads to thicker fibers. The numerical 
relation is estimated by the linear equation 4305x959y .. ?=  with an R2 of 0.9881 as 
shown in Figure 2.2. The statistical frequency distribution of nanofiber diameters for each 
concentration is represented in Figure 2.3. Nanofibers spun from 8 wt% PAN/DMF 
solution have 66.1% of fibers with diameters ranging from 100 nm to 200 nm, which is 
the biggest frequency among the nanofibers from four different concentrations. 
Correspondingly, the biggest fraction of nanofiber diameters from 10 wt% is 58.3% 
which is in the range of 200-300 nm. This number becomes smaller, 52.5% in 400-500 
nm range and 47.8% in 500-600 nm range for fibers from 12 and 14 wt% solutions, 
respectively. Therefore, fibers spun from low concentration have more narrow 
distribution than those from higher concentration. 
43
Figure 2.2. Fiber diameters with different PAN concentrations (20 kV 
voltage, 20 cm distance and 1 mL/h flow rate). 
Figure 2.1. SEM images of electrospun PAN from different concentrations 
(a) 8%  (b)10%  (c)12%  (d)14% (20 kV voltage, 20 cm tip-collector 
distance and 1 mL/h flow rate). 
a b
c d
44
2.3.1.2 Applied voltage. PAN/DMF solution was electrospun at 5, 10, 15, 20, 24 and 29 
kV as solution concentration, flow rate and needle tip-collector distance remained at 10 
wt%, 2 mL/h and 20 cm, respectively. The resulting average fiber diameters are plotted 
as the function of applied voltage as shown in Figure 2.4. The increasing voltage results 
in thicker fibers and their relation fits the equation 05.2260995.7 += xy with an R2 of 
0.7805. The biggest diameter difference of 156nm happens between 10 kV (263 nm) and 
29 kV (419 nm). Nanofibers are drawn by the electrostatic force between two electrodes. 
Figure 2.3 The frequency distribution of nanofiber diameters for 
different solution concentrations. 
45
The increase in applied voltage leads to stronger electrostatic forces and therefore 
accelerates polymer jet toward target collector which was an aluminum foil. Higher jet 
speed might not allow fibers to be stretched enough with less traveling time from needle 
to the collector, resulting in bigger fiber diameter.  Similar results were obtained by Yang 
et al. [10]. The voltages below 5 kV generated too low electrostatic force to overcome 
solution surface tension and no polymer jet was formed. 
2.3.1.3 Flow rate. To study the influence of flow rate on the electrospun fiber diameters, 
10 wt% PAN/DMF was electrospun keeping the needle tip-collector distance at 20 cm 
and applied voltage at 20 kV. The flow rate was set 0.5, 1, 1.5, 2, 2.5 and 3 mL/h. The 
resulting fiber diameters are shown in Figure 2.5. No strong relation between flow rate 
and fiber diameter is observed. Compared to diameter variation caused by PAN/DMF 
concentration and applied voltage, the difference between the thickest and the thinnest 
fiber caused by varying flow rate is smaller, only 82 nm. Thus it can be concluded that 
Figure 2.4. Fiber diameters at different voltages (10% PAN/DMF, 
20 cm tip-collector distance and 2 mL/h flow rate).
46
flow rate has less influence than polymer concentration and applied voltage on fiber 
diameter.  
2.3.1.4 Needle tip-collector distance. The increase of needle tip-collector distance is 
considered to decrease the electrostatic force between the collector and the needle and 
therefore the thicker fibers are expected. However, longer distance provides more time 
for stretching resulting in smaller fiber diameters which may balance the impact of 
electrostatic force. In the experiment, when the distance was raised from 10 cm to 30 cm, 
randomly scattered fiber diameters were observed as shown in Figure 2.6. The finest fiber 
of 366 nm was obtained when the distance was set at 24 cm while the thickest fiber of 
401 nm was at 15 cm. Other fibers diameters vary between them. It is interesting to note 
that the trend of the data points on Figures 2.5 and 2.6 are somewhat similar. 
Figure 2.5. Fiber diameters at different flow rates
  (10% PAN/DMF, 20 cm distance and 20 kV voltage).  
47
2.3.2 Physical and mechanical properties of electrospun PAN membranes  
Generally, the increase of thickness causes an increment of strength with the loss of air 
permeability as the penalty. In this part, experiments were designed to investigate the 
numerical relation between strength (tensile, tear and burst) and electrospun PAN 
membrane thickness. All membranes in this part were electrospun from 10 wt% 
PAN/DMF at 20 kV/ 20cm with a delivery rate of 1 mL/h.  
Figure 2.6. Fiber diameters at different needle-tip distances
 (10% PAN/DMF, 20 kV voltage and 2 mL/h flow rate).
Figure 2.7. Variation of tensile strength with the change of 
electrospun PAN membrane thickness. 
48
 2.3.2.1 Tensile, tear and burst strength. The variations of tensile, tear and burst strength 
with the increase of thickness of electrospun PAN are shown in Figures 2.7, 2.8 and 2.9. 
As expected, the thicker the membrane, the higher the tensile and tear strength. Tensile-
thickness relation fits the linear equation 28630x65650y .. +=  with an R2 of 0.7962 and 
tear-thickness relation fits the equation 6858.3106.81 += xy  with an R 2 of 0.9602. In 
tensile strength test, the membranes with different thickness exhibited different breaking 
behaviour. Thinner membranes had a uniform fracture, which was mostly due to 
breakage of individual fibers simultaneously while the thicker membranes had transverse 
contraction. When the load was applied to a thick membrane, it deformed in order to 
align fibers in the stretching direction and then fiber slippage took place after overcoming 
the friction due to fiber entanglement [11].  
    Figure 2.9 shows the burst strength of membranes at various thicknesses. For the 
membranes with the thickness of 0.34 mm or higher, burst strength varies between 0.096 
and 0.102 Pa which was not a big change. The relation between the thickness and burst 
strength of membranes could be best expressed by the equation 1126.0105.0 xy =  with an 
R2 of 0.7705. For comparison, the burst strength of 1mm thick regular print paper was 
measured as 0.11 Pa, which was higher than electrospun PAN membrane by 7.8%.   
                                                                                                                          
49
2.3.2.2 Air Permeability. Gain of strength with the increasing thickness of membrane 
accompanies the loss of air permeability. As shown in Figure 2.10, the increase of 
thickness results in an exponential decrease in air permeability that could be represented 
by the equation 9245.05551.9 ?= xy  with an R 2 of 0.9792.  
Figure 2.8. Variation of tear strength with the change of 
membrane thickness.  
Figure 2.9. Variation of burst strength with the change of membrane thickness. 
50
2.3.2.3 Porosity and pore size. Porosity of a membrane is calculated as,
%100)1(%100)1( ???=??= At
m
V
V
t
F ??                                                                        (2.1) 
where ? is calculated porosity, VF is the fiber volume in the membrane, Vt is the 
membrane total volume, m is the weight, ? is PAN density (1.17 g/cm3), t is the 
membrane thickness and A is the membrane area [11]. The porosity and interfiber pore 
size within electrospun PAN nonwoven made from various solution concentrations are 
summarized in Table 2.1. As discussed above, different solution concentrations result in 
different fiber diameters if all other electrospinning parameters are fixed. Since the 
applied voltage, flow rate and tip-collector distance are all fixed, higher solution 
concentration represents thicker fibers. Regardless of nanofiber diameters, the porosity of 
the electrospun membrane did not show much variation.  The electrospun PAN 
nonwoven with small pore size ranging from 0.801 to 1.3548 ?m could catch very small 
particles. The high porosity and small interfiber pore size of electrospun membranes 
Figure 2.10. Air permeability as a function of membrane 
thickness. 
51
make them a good candidate for tissue scaffold by providing enough space for cell 
accommodation and an easy passage for nutrient intake and metabolic waste exchange [1, 
12]. 
Table2.1 Physical properties of electrospun PAN membrane and stabilized PAN membrane. 
Concentration (wt%) 8 10 12 14
Interfiber pore size (?m) 0.801 1.0245 0.8727 1.3547 
Porosity (%) 96.12 96.75 96.34 95.89
Porosity after stabilization (%) 96 96.05 96.7 96.19
2.3.2.4 Physical properties of stabilized PAN nanofiber membrane.  Electrospun PAN 
membrane made from 8, 10, 12 and 14 wt% PAN/DMF solutions were stabilized at 
280?C for 1 hour. The stabilized PAN showed different breaking behaviour from the 
electrospun PAN without stabilization as shown in Figure 2.11. For stabilized PAN made 
from 8 and 10 wt% solutions, tensile strength showed an obvious reduction with the 
increase in thickness. However, stabilized PAN from the other two solutions did not 
show strong relation between thickness and tensile strength. More uniform breaking 
fracture was observed on the stabilized electrospun PAN membranes than those before 
stabilization.  
52
   Porosity of stabilized electrospun PAN nanofibers was compared with those without 
stabilization as shown in Table 2.1. Although some shrinkage was observed after 
stabilization at 280 ?C, it did not show significant influence on the porosity of membrane. 
The stabilized PAN nanofibers were too brittle to carry out the pore size test on 
porometer.
2.3.3 Properties of electrospun PAN based activated carbon nanofibers  
Since the thinnest nanofibers were made from 8 wt% PAN/DMF solution based on the 
study above, they were used as precursors to make activated carbon nanofibers.  
2.3.3.1 Morphology. The structures of electrospun PAN and activated PAN nanofibers 
are shown in Figure 2.12. A denser looking structure was obtained after activation, which 
was probably due to shrinkage and crimping of fibers resulting from the cyclization 
during stabilization and activation. The average nanofiber diameter decreased slightly 
from 287 nm before activation to 270 nm afterward. 
Figure 2.11. Tensile strength variations of stabilized electrospun PAN with the 
increase in thickness (20 kV voltage, 20 cm distance and 1.5 mL/h flow rate). 
53
       
2.3.3.2 FTIR spectrum. Figure 2.13 shows the FTIR spectrum of electrospun PAN, 
stabilized PAN and activated PAN (activated carbon nanofibers). The remarkable change 
during the conversion is the decrease of CN triple bond. The FTIR spectrum for 
electrospun PAN has the peak at the wavenumber of 2242 cm-1 to represent CN triple 
bond in the PAN (Figure 2.13a). In the stabilization process, the CN triple bond was 
converted to CN double bond and the cycle including carbon and nitrogen atoms was 
formed so that the peak of CN triple bond became very small. In the spectrum for 
activated PAN shown in Figure 2.13(c), CN triple bond disappeared and no CH or CH2
bond at 2936 cm-1 was obtained. During activation process at high temperature, most of 
non-carbon atoms, such as nitrogen and hydrogen, were removed and micropores grew at 
the same time.  
2.3.3.3 Pore structure of activated carbon nanofiber. BET surface area, average pore 
diameter, total pore volume and DA micropore volume of electrospun PAN and activated 
carbon nanofibers are compared in Table 2.2. The surface area of activated carbon 
nanofiber is 27.2 times larger than that of electrospun PAN nanofibers before activation. 
Figure 2.12. SEM images of (a) electrospun PAN membrane and (b) activated carbon 
nanofibers  (8 wt% PAN/DMF, 20 kV voltage, 20 cm distance and 1mL/h flow rate).
a b
54
The activated carbon nanofibers have 0.59 cc/g micropore volume, which is 14.75 times 
larger than that of 0.04 cc/g before activation. Compared to conventional fibers with 
diameters in the scale of micron or bigger, electrospun PAN has relatively high surface 
area of 34.4 m2/g due to very small fiber diameter in nano scale. However, activation 
process created micropores and mesopores within fibers, which contributes to the 
dramatic increase in surface area that is one of the dominant factors to determine 
adsorption capacity of adsorbents. 
   
Figure 2.13. FTIR spectrum for (a) electrospun PAN, (b) stabilized PAN and (c) activated carbon 
nanofibers. 
55
    Pore size distribution was analyzed based on the DA equation and the generalized 
Halsey equation. In the DA pore size distribution curve presented in Figure 2.14, 
activated carbon nanofiber has a high peak between 1nm and 4nm pore diameter while 
electrospun PAN before activation has a small peak, which indicates that activated 
carbon nanofibers contain much more micropores than electrospun PAN. Based on the 
generalized Halsey method, the pore volume is plotted as the function of Halsey 
statistical thickness, as shown in Figure 2.15. The electrospun PAN line can be 
extrapolated to the y axis close to the origin, which is the evidence of small amount of 
micropores; the V-t plot of activated carbon nanofibers has a large positive intercept on Y 
axis (312.73), which means they contain a big amount of micropores [9]. Micropores and 
mesopores with an average pore diameter of 3nm could trap small molecules so that it 
could be used for water purification, gas separation and chemical protection.  
                                              
Table 2.2. Pore structures of electrospun PAN and activated carbon nanofibers. 
Surface area        
(m2/g) 
Average pore 
diameter (?) 
Total pore volume  
(cc/g) 
DA micropore 
volume (cc/g) 
Electrospun PAN 34.4 147 0.127 0.04
Activated carbon 
nanofibers 936.2 30 0.723 0.59 
56
2.4 Conclusions  
Electrospun polyacrylonitrile fiber diameters were mainly determined by the solution 
concentration and applied voltage rather than flow rate and needle tip-collector distance. 
Tensile and tear strengths increased with increasing thickness of the electrospun PAN 
membrane, which was accompanied by a decrease of air permeability. Burst strength was 
Figure 2.15 Generalized Halsey V-t plot  
Figure 2.14 DA pore size distribution 
57
not affected much by thickness when the membrane is 0.34 mm or thicker. After 
stabilization, tensile strength of electrospun nanofiber membranes made from 8 and 10 
wt% solution decreased with increasing thickness. Electrospun PAN had a very high 
porosity, which was not affected by the solution concentration and stabilization process. 
Electrospun polyacrylonitrile nanofibers were chemically activated to obtain activated 
carbon nanofibers using KOH as the activating agent. The disappearance of CN triple 
bond in the FTIR spectrum for the activated carbon nanofibers and a dramatically 
increased surface area and micropore volume after activation proved the successful 
conversion from PAN to activated carbon nanofibers. In V-t plot based on the generalized 
Halsey equation, activated carbon nanofiber line has a large positive intercept reflecting 
the presence of a large amount of micropores; however, the intercept for electrospun 
PAN line is very small reflecting small micropore volume. The same result is concluded 
from the DA pore size distribution analysis. Most of the generated pores in activated 
carbon nanofibers are micropores and mesopores with an average pore diameter of 3 nm.  
References 
1. Subbiah, T., Bhat, G. S., Tock, R. W., Pararneswaran, S. and Ramkumar, S. S., 
Electrospinning of nanofibers, Journal of Applied Polymer Science, 96 (2), 557-
569, (2005). 
2. Lawrence L, W., Norman C. Pereira, Yang-Tse Hung, Air Pollustion Control 
Engineering, Humana Press, Totowa, NJ, (2004). 
58
3. Moon, S. Y. K., M.; Hahm, H.; Lim, Y., Preparation of Activated Carbon Fibres 
by Chemical Activation Method with Hydroxides, Materials Science Forum, 510-
511, (2006). 
4. Park, S. H., Kim, C., Jeong, Y. I., Lim, D. Y., Lee, Y. E. and Yang, K. S., 
Activation behaviors of isotropic pitch-based carbon fibers from electrospinning 
and meltspinning, Synthetic Metals, 146 (2), 207-212, (2004). 
5. Kim, C. and Yang, K. S., Electrochemical properties of carbon nanofiber web as 
an electrode for supercapacitor prepared by electrospinning, Applied Physics 
Letters, 83 (6), 1216-1218, (2003). 
6. Kim, C., Yang, K. S. and Lee, W. J., The use of carbon nanofiber electrodes 
prepared by electrospinning for electrochemical supercapacitors, Electrochemical 
and Solid State Letters, 7 (11), A397-A399, (2004). 
7. Kim, C., Choi, Y. O., Lee, W. J. and Yang, K. S., Supercapacitor performances of 
activated carbon fiber webs prepared by electrospinning of PMDA-ODA 
poly(amic acid) solutions, Electrochimica Acta, 50 (2-3), 883-887, (2004). 
8. Lee, Y. J., Kim, J. H., Kim, J., Lee, D. B., Lee, J. C., Chung, Y. J. and Lim, Y. S., 
Fabrication of activated carbon fibers from stabilized PAN-based fibers by KOH, 
Designing, Processing and Properties of Advanced Engineering Materials, Pts 1 
and 2, 449-4, 217-220, (2004). 
9. AS1 win Version 1.51, Operating Manual, (2005). 
10. Yang, Y., Jia, Z. D., Li, Q. and Guan, Z. C., Experimental investigation of the 
governing parameters in the electrospinning of polyethylene oxide solution, IEEE 
Transactions on Dielectrics and Electrical Insulation, 13 (3), 580-585, (2006). 
59
11. Wang, Y., A method for tensile test of geotextile with confining pressure, Journal 
of Industrial Textiles, 30 (4), 289-302, (2001). 
12. Thorvaldsson, A., Stenhamre, H., Gatenholm, P. and Walkenstrom, P., 
Electrospinning of highly porous scaffolds for cartilage regeneration, 
Biomacromolecules, 9 (3), 1044-1049, (2008). 
g3
g3
g3
60 
Chapter 3 Sulfur Dioxide Removal over Activated Carbon 
Fibers and Modified Activated Carbon Fibers 
Activated carbon fibers (ACFs) have been considered as one of the most promising 
sorbents for gas and liquid purification, such as water purification, flue gas 
desulfurization and denitrification, and removal of volatile organic compounds. As for 
desulfurization, their ability to absorb sulfur dioxide was discussed in Chapter 1.  
     In this work, activated carbon fibers manufactured from various raw materials are 
studied and their pore structures and sulfur dioxide removal performance under dry and 
humid conditions are investigated. Furthermore, they are modified by various promoters 
with the purpose of improving capability of desulfurization; the corresponding sulfur 
dioxide removal efficiencies are studied. 
3.1 Experimental 
3.1.1 Materials 
Five types of activated carbon fibers are studied in this research, namely ACF10, ACF15, 
ACF20, CR and SY. They are classified based on the precursors. ACF10, ACF15 and 
ACF20 were made from KynolTM novoloid fibers by American Kynol Inc. Novoloid 
fibers, whose structure is shown in Figure 3.1, containing 75.8% carbon are transformed 
to 95% carbon (Figure 3.2) at 700?C followed by activation with activating agents, such 
as CO2 or steam [1]. ACF10, ACF15 and ACF20 are differentiated by distinct surface 
areas, pore sizes and pore volumes due to the extent of activation. CR (purchased from 
61 
Carbon Resources, Inc.) is rayon-based ACFs in the form of felt.  SY is rayon-based 
ACFs provided by Senyou Carbon Fiber Co., Ltd (China). For the purpose of comparison, 
powdered activated carbon is studied in parallel. Activated carbon powder (60~140 mesh) 
is manufactured from coconut shell by PICA Inc.  
                   
    
                Figure 3.2 Structure of carbonized novoloid fiber with 95% carbon [1] 
    ACFs were treated with different promoters to improve their capability in 
desulfurization. Promoters, potassium hydroxide (KOH), potassium carbonate (K2CO3), 
potassium chlorate (KClO3) and potassium permanganate (KMnO4) were dissolved in 
water with certain concentrations and two methods were used for treatment. Incipient 
wetness impregnation (IWI) was used first. The volume of solutions which were injected 
Figure 3.1 Structure of KynolTM novoloid fibers [1]
62 
to ACFs was equal to the pore volume in ACFs. ACFs were also simply immersed in 
solutions so that a variety of promoter loads could be achieved. Drying at 120?C followed 
to remove extra water. In order to prepare metal/ACFs, a certain amount of solution, 
based on the incipient wetness impregnation, was injected onto ACFs, followed by drying 
in air and calcinations at high temperature either in nitrogen or air depending on metal 
precursors. The detailed modification process is further explained in Section 3.1.4. 
3.1.2  Pore structure and element analysis 
Pore characteristics of activated carbon fibers, such as BET surface area, average pore 
size and pore volume, were analyzed based on isotherm obtained on Autosorb-1 
(Quantachrome? Instruments) using nitrogen as adsorbate at 77 K. For BET surface area 
analysis, multipoints with a partial pressure (P/P0) between 0.05 and 0.3 are considered to 
be accurate in predicting an adsorption process. The number of adsorption sites was 
estimated based on the BET testing results using equation 3.1. 
Nad= WmM *N= StA
cs
                                                                                                     (3.1) 
where Nad is the number of adsorption sites, Wm is the sample weight in gram, M is the 
molecular weight of the adsorbate (nitrogen) in g/mol, N is the Avogadro?s number 
(6.023?1023 molecules per mol), St is the specific surface area (m2/g), and Acs is the 
cross-sectional area of the adsorbate (16.2 ?2 for nitrogen).  
    Element analysis for nitrogen, carbon and sulfur was conducted on Vario Macro 
Elementar by analyzing the released carbon dioxide, nitrogen oxide and sulfur oxide 
during burning of raw materials. Other elements were analyzed on VARIAN Vista-MPX 
CCD simultaneous ICP-OES. Both of them were performed in the Soil Testing 
Laboratory at Auburn University.  
63 
3.1.3 Sulfur dioxide adsorption testing 
Dynamic sulfur dioxide adsorption characteristics over ACFs and PAC were studied by 
analyzing the breakthrough curves which were obtained through adsorption tests 
performed in a fixed tubular column reactor (15.9 mm in diameter) made of glass. The 
experimental test bench was designed based on the ASTM D6646-01 and is shown in 
Figure 3.3.   
    High concentration sulfur dioxide (2000 ppmv) purchased from Airgas was diluted by 
air such that a certain concentration of sulfur dioxide could be obtained for adsorption 
11 
6 
5 
1. SO2 tank                                   
2. Compressed air                                
3. Mass flow 
controller or flow 
meter                   
4. Gas bubbler 
5. Gas mixer 
6. Hygrometer 
7. Three way valve       
8.  Reactor 
9. Sample 
10. Glass wool 
11. Gas monitor 
12. NO2 gas tank 
13. H2S gas tank  
                    
3 
7 
9 8
10 
12  13 
1 2 
4 
Figure 3.3 Schematic (top) and picture (bottom) of SO2 adsorption test bench
64 
tests. Before mixing with sulfur dioxide, compressed air can pass through the gas bubble 
filled with water to be humidified. The relative humidity was monitored with RH 390 
Precision Psychrometer (Extech Instruments). In order to obtain a desired relative 
humidity which determines the water concentration in the main gas stream, air with high 
humidity was mixed with dry air.  
The challenge gas for adsorption tests (one of the mixtures of SO2+N2, SO2+air, 
SO2+H2O+N2 or SO2+H2O+air) was fed into the glass reactor where the adsorbent 
sample was supported by glass wool and SO2 adsorption would occur. Before starting the 
adsorption test, the initial sulfur dioxide concentration was monitored and recorded on a 
sulfur dioxide single gas detector (BW Gasalert Extrem). As the experiment started, the 
transient concentration of sulfur dioxide leaking from the reactor was measured. The 
simultaneous removal of nitrogen dioxide and hydrogen sulfide over adsorbents can be 
performed on the same set up by adding gas tanks. Breakthrough times and breakthrough 
curves were obtained during each test and the corresponding sulfur dioxide capacities 
were calculated. The time from the beginning of the test until SO2 can be detected at the 
exit of the reactor is called the breakthrough time or breakthrough period, beyond which 
the desulfurization efficiency starts decreasing. The breakthrough curve or breakthrough 
profile, which is the transient response of the adsorbent bed to adsorbates, is a reflection 
of the adsorbents performance under dynamic conditions. A relatively larger 
breakthrough time and gradual increase in the concentration following breakthrough are 
desirable [2].  In order to study the effect of oxygen on SO2 adsorption, air was replaced 
by nitrogen. The utilization of samples refers to the ratio of breakthrough time to 
saturation time.  
65 
3.1.4 Modification of activated carbon fibers 
3.1.4.1 ACFs supported metals  In the literature, copper, nickel and iron have been 
reported to be better than other promoters for  activated carbon and activated carbon 
fibers in enhancing sulfur dioxide removal efficiency [3-8]. In this study, they were 
supported by ACF20 and their sulfur dioxide capacities were measured and compared 
with each other and with bare ACF20 to investigate their potential to be good desulfurizer. 
For Cu/ACF20 and Ni/ACF20, the metal precursors are copper nitrate with trihydrate 
(Cu(NO3)2?3H2O) and nickel nitrate with hexahydrate (Ni(NO3)2?6H2O), respectively.  
    Before the treatment, ACF20s were dried at 110?C. Cu/ACF20 and Ni/ACF20 were 
prepared with the incipient wetness impregnation with metal loads of 5 wt%. Since the 
total pore volume of ACF20 was 1.46 cc/g, there was 2.92 mL pore volume available for 
the solutions. However, a little more solution was used to account the loss of solution 
induced by the operation. Thus, 3.2 mL 12.9 wt% copper nitrate and 16.96 wt% nickel 
nitrate solutions were injected to 2 g ACF20 with syringe following 110?C drying for 6 
hours for the designed 5 wt% load of metal. Cu(NO3)2/ACF20 and Ni(NO3)2/ACF20 
were calcined at 300?C for 1 hour. The ratio of iron to ACF20 was 3.5 wt% since this 
ratio was found to give the maximum sulfur dioxide capacity in the literature. The 
precursor of iron was Fe(NO3)3g9429H2O, which was made to 17.34 wt% solution. The same 
amount of iron nitrate solution (3.2 mL) was injected to 2 g ACF20 following drying at 
110?C and calcinations at 250?C for 3 hours in nitrogen.   
    Unlike the incipient wetness impregnation, ACF20 was impregnated in both 12.9 wt% 
and 2 wt% Cu(NO3)2 solutions followed by washing, drying and 1 hour calcination at 
300?C in nitrogen such that different weight ratios of copper on ACFs could be obtained. 
66 
The washing process was expected to clean extra Cu(NO3)2 which was not trapped in the 
pores. Pretreatments of ACF20 with HCl and HNO3 were also employed, respectively, 
before loading copper to investigate their influences. ACF20s were immersed in 36.8% 
HCl and HNO3 at 25?C, respectively, and then were washed with distilled water until the 
pH value of water reached 7 followed by drying at 110?C.  
3.1.4.2 ACFs impregnated with oxidants  Sulfur dioxide has the potential to be oxidized 
by stronger oxidants except oxygen. If oxidants are treated onto fiber surfaces, they are 
expected to oxide sulfur dioxide and thus clean it from the main gas stream. In the current 
research, potassium chlorate and potassium permanganate were treated onto porous 
activated carbon fibers through IWI (incipient wetness impregnation) and impregnation 
with a variety of oxidant loads. Their effects on sulfur dioxide removal efficiency of 
ACFs were analyzed. A certain amount of potassium chlorate and potassium 
permanganate solutions was ejected to activated carbon fibers followed by drying. In 
theory, if IWI was used, the maximum load of KMnO4 on ACF10, ACF15 and ACF20 
would be 2.9, 5 and 8 wt%, respectively, since KMnO4/H2O solution gets saturated at 6 
wt% at 20?C. Since the total pore volumes of ACF10, ACF15 and ACF20 were 0.51, 
0.88 and 1.46 mL/g, respectively, 1, 1.8 and 3 mL 6 wt% KMnO4 solutions were sprayed 
onto them, respectively, followed by drying at 110?C and 2.01, 5.54 and 7.24 wt% of 
KMnO4 were obtained on ACF10, ACF15 and ACF20, respectively. The KMnO4 and 
KClO3 solutions were increased gradually such that a series of weight loads of promoters 
over ACFs were obtained. This process avoided the high temperature calcination in inert 
gases (usually nitrogen is chosen) which is energy consuming. Considering wet-laying 
method for formation of filter media where ACFs are exposed to a large amount of water, 
67 
KClO3/ACFs or KMnO4/ACFs needed to be washed with distilled water. The effect of 
washing on the sulfur dioxide capacity was investigated.  
3. 1.5 ACFs surface morphology analysis 
Surface characteristics of bare ACFs, modified ACFs prior to and after SO2 adsorption 
were observed on scanning electron microscope (SEM, Zeiss? EVO 50 or JEOL 7000 
FESEM). Diameters of ACFs were measured on SEM. In order to elucidate pores on 
fiber surfaces, Matlab? was used to simulate SEM images and show micropores. 
Diameters of PAC were measured on Spencer optical microscope made by American 
Optical Company.  
3.1.6 Thermogravimetric analysis (TGA) 
The thermal stability of bare ACFs, modified ACFs and ACFs saturated with sulfur 
dioxide were analyzed using TGA (TA instruments) which can help to identify 
compounds  on fiber surfaces. 
3.1.7 X-Ray diffraction analysis 
For species identification on fiber surfaces, X-Ray diffraction (XRD) was carried out on 
X-ray diffractometer (D8 Discover manufactured by Bruker-axs). Samples were 
supported by a glass slide which was put on the platform of the machine. The scanning 
was performed with a scan rate of 0.01?/s and 2? for 5? to 90? at 40 kV and 25 mA.  
3.1.8 Manufacturing of activated carbon fiber paper (ACFP) 
In comparison with particular adsorbents, ACFs offer advantages in formation of a mat 
with the wet lay technology in two aspects:  formation and retention.  The density 
difference between the fibers and other stock components causes an imbalance in 
formation with respect to dispersion of polymeric fibers across the wet lay mat and thus 
68 
additives are usually used to provide even dispersion in the stock to help formation. In 
our research, the density difference between ACFs and binder fibers is less in comparison 
with that between metal fibers and adsorbents.  Retention is usually a problem if 
particular adsorbents are employed since the loss of particulates would increase the cost; 
therefore, additives are included to help entrap particles. The usage of ACFs as 
adsorbents replacing the particulates addresses the problem of retention. Although ACFs 
may be broken during preparation of paper slurry, they still keep the length of 1~3 mm 
and are more likely to form well-dispersed slurry.   
    For the application of ACFP in filtration, strong, uniform and high contaminant 
removal efficiency is desired.  Physical entanglement along with hydrogen bonding 
between fibers in regular paper made from cellulose fibers offers the major force that 
binds fibers together. After formation of paper, extra water needs to be removed to 
decrease the distance between fibers to form hydrogen bonding which only occur while 
hydrogen and oxygen atoms are close enough. The subsequent pressing and drying can 
achieve this goal. Due to the lack of hydrogen and oxygen in ACFs, it is impossible to 
form a strong ACFP by hydrogen bonding. Therefore, in the current research, cellulose 
fibers acting as binding agents were mixed with ACF20. Other binders, for example, 
polyester and bicomponent polyester (CoPET), which would provide bonding by melting, 
were tried as well in order to optimize the physical and mechanical performance of ACFP, 
such as tensile strength and uniformity. To determine the fusing temperature for bonding 
formation in ACFP, melting points of these bonding agents were tested on a differential 
scanning calorimeter (DSC TA instrument).  
69 
    To make ACFP with the wet-laying process, ACFs were mixed with bonding agents in 
1000 mL water in a blender for 1min to make ACF slurry; then ACFP was formed in a 
hand sheet former. The hand sheet former consists of a metal column (15 cm in diameter) 
called headbox, a forming fabric or wire mesh and a drainage system.  After the stock 
suspension was poured into the headbox, ACFs and bonding agents were dispersed 
thoroughly using a hand held mixer, called a dasher, followed by draining by 
gravitational forces. A wet mat was formed on the forming fabric followed by pressing 
twice with the same pressure (560 KPa) for 5 min and 2 min, respectively. The relatively 
dry wet mat was transferred to a metal plate and covered with another plate to provide 
160 Pa pressure during drying/fusing, which completed the bonding between fibers in 
ACFP. 
     In order to increase the uniformity of ACFP, a surfactant was added into the slurry to 
help disperse ACFs and binder fibers. The strength of ACFP was expected to enhance 
accordingly due to improvement of uniformity.  The compositions of ACFP are listed in 
Table 3.1. 
70 
Table 3.1 Compositions of activated carbon fiber papers 
Sample Composition 
A Cellulose/ACF20=50/50 no heating
B Cellulose/ACF20=30/70 no heating 
C Cellulose/CoPET/ACF20=15/15/70 135?C 
D Cellulose/CoPET/ACF20=15/15/70 165?C 
E Cellulose/CoPET/ACF20=15/15/70 200?C 
F CoPET/ACF20=30/70 122?C 
G CoPET/ACF20=30/70 132?C 
H CoPET/ACF20=30/70 139?C 
I CoPET/ACF20=30/70 180?C 
G CoPET/ACF20=30/70 200?C 
K CoPET/ACF20=30/70 220?C 
L CoPET/ACF20=30/70 260?C 
M CoPET/ACF20=30/70 200?C + 2 drops of surfactant 
    It should be noted that CR is in the form of felt and therefore does not need to be made 
into paper. However, CR felt is too weak to be used in the filtration application due to 
short ACF fibers which cannot provide strong interaction. They may be mixed with 
binder fibers to enhance their strength. Most ACFP involved in this dissertation are made 
from individual ACFs rather than from ACF felt. 
71 
3.2 Results and Discussion  
3.2.1. Pore structure and surface morphology of ACFs 
According to International Union of Pure and Applied Chemistry (IUPAC), pores are 
classified into three categories, namely micropore, mesopore, and macropore with pore 
sizes less than 2 nm, between 2 and 50 nm, and larger than 50 nm, respectively. Surface 
area, total pore volume, average pore diameter and micropore diameter (calculated 
according to Dubinin-Astakhov equation) of ACFs and ACP are summarized in Table 3.2. 
On the Autosorb-1, surface area was estimated based on Brunauer-Emmett-Teller (BET) 
equation by assuming that a monolayer of nitrogen molecules covers the adsorbent 
surfaces. The total pore volume was derived from the amount of vapor adsorbed at a 
relative pressure close to unity, by assuming the pores are filled with liquid adsorbates. 
The average pore diameter was evaluated from the total pore volume. To characterize 
micropores, the Dubinin-Astakhov (DA) equation is more appropriate for strongly 
activated carbon [9-11], which is used in this research:   
g1849 g3404g1849g2868g1857g1876g1868g3429g3398g4678g2879g3019g3021g3039g3041g4666
g3265
g3265g3116g4667
g3006 g4679
g3041
g3433g3                                    (3.2) 
where, W is the weight adsorbed at P/P0 and T, W0 is the total weight adsorbed, E is the 
characteristic energy and n is a non-integer value (typically between 1 and 3) [11]. 
Characteristic energy (E) is the measure of the mean value of adsorption potential which 
is a measure of the affinity of the surface for the adsorbate molecules for microporous 
adsorbents. The values of component n and characteristic energy E were found to be 
related to pore size distribution. High value of n and low value of E result in narrow pore 
size distribution [9]. 
72 
    ACF10, ACF15 and ACF20 were made from the same precursors. However, they 
exhibit different pore properties, which is due to the variation of activation conditions. 
ACF20 with a higher surface area, more pore volume and bigger pore size was activated 
at a higher temperature. Regardless of precursors, total pore volume shows linear relation 
with the surface area: higher surface area corresponds to higher total pore volume, as 
shown in Figure 3.4. Since the majority of pores is micropores (Table 3.3), a similar trend 
is shown between surface area and micropore volume. Although the average pore 
diameter fluctuates in a narrower range of 24.28 and 26.82 regardless of precursors and 
activation conditions, DA micropore diameter is dependent on surface area: the higher 
the surface area, the bigger the micropore diameter is.   
Table 3.2 BET surface area, pore volume and pore size of ACFs and PAC 
ACF10 ACF15 ACF20 CR SY PAC
BET surface area (m2/g) 784.2 1370 2247 1763 806 913 
Total pore volume (cc/g) 0.51 0.88 1.462 1.09 0.515 0.6 
Average pore diameter (?) 24.28 25.79 26.03 24.88 25.5 26.82 
DA micropore width (?) 11.97 15 17.69 16 14 14.2 
73 
Pore size distribution (PSD) of ACFs and PAC based on DA equation is illustrated in 
Table 3.3 and Figure 3.5. Among ACFs made from novoloid fibers (Table 3.3), the 
values of n for three of them are close to each other while ACF10 and ACF20 have the 
lowest and highest characteristic energy, respectively, indicating a narrower PSD of 
ACF10 which agrees with Figure 3.5. As the ACF20 was more activated than ACF10, 
separated micropores in ACF10 might be expanded and thus connect with neighboring 
micropores resulting in the formation of macropores. However, the value of n and E 
cannot generally represent the shape of pore size distribution for all ACFs and PAC 
discussed in this research.  
Figure 3.4 Relation between surface area and pore volume 
74 
Table 3.3 Pore volume distributions of ACFs and PAC
DA pore analysis ACF10 ACF15 ACF20 PAC CR SY 
Micropore (cc/g) 0.376 0.611 0.966 0.427 0.665 0.385
Mesopore (cc/g) 0.051 0.12 0.314 0.138 0.262 0.039
Macropore (?10-4 cc/g) 1.92 2.81 12.7 18.5 13.9 0.323
Total pore volume (cc/g) 0.427 0.731 1.28 0.567 0.928 0.424
Micropore percentage (%) 88 83.55 75 75.33 71.63 90.73
Mesopore percentage (%) 11.98 16.41 24.53 24.35 28.22 9.26
Macropore percentage (%) 0.045 0.038 0.099 0.033 0.015 0.008
E (kJ/mol) 8.59 6.6 5.67 7.11 5.36 7.41
n 1.9 2.1 1.9 1.4 1.8 2.5
Figure 3.5 Pore size distributions of ACFs and ACP
75 
    The number of adsorption sites of ACFs and PAC, which is considered to be 
responsible for the adsorption of adsorbate over adsorbents, are summarized in Table 3.4.  
Table 3.4 Number of adsorption sites of activated carbon fibers and PAC 
   The surface morphology of ACF15 is shown in Figure 3.6. At a low magnification, a 
smooth fiber surface was observed without the appearance of pores. As the magnification 
was increased to 250,000, a macropore with a diameter of about 100nm and some 
mesopores showed up but micropores are still too small to see under SEM.  
    
     PAC was found to be in an irregular shape rather than sphere as assumed by most 
research. The average PAC diameter is 290 ?m with a coefficient of variation of 25.08%.  
Samples ACF10 ACF15 ACF20 CR SY PAC
Number of adsorption sites
(?1020) 
2.99 5.22 8.56 6.72 3.07 3.48 
1 2
Figure 3.6 Surface morphology of ACF15 1) ?9500 and 2) ?250,000 
76 
3.2.2. Element analysis  
Nitrogen and carbon weight percentage of as received various activated carbon fibers 
were measured with burning tests and oxygen weight percent was calculated by weight 
difference; the results are shown in Table 3.5. Nitrogen content and the ratio of nitrogen 
to carbon in ACFs manufactured from the same precursor are found to be related to 
surface area although carbon and oxygen contents are less dependent on the surface area. 
The higher the surface area whose extent of activation may be higher, the more nitrogen 
detected.  It can been seen that the elements in fibers are more dependent on the type of 
precursors of activated carbon fibers. Rayon based ACFs, CR and SY contain more 
nitrogen than novoloid based ACFs. The heat treatment at high temperatures burned non-
carbon elements, nitrogen and oxygen.  
Table 3.5 Element components of activated carbon fibers 
  ACF10 ACF15 ACF20 CR SY 
ACF20-
500 
ACF20-
700 
N % 0.8 0.83 0.96 2.1 1.28 0.66 0.85
C % 97.7 87 94.5 87.1 87.6 96.9 96.9
O % 5.5 12.17 4.54 10.8 11.12 2.44 2.25
N/C % 0.849 0.958 1.015 2.407 1.463 0.686 0.874
        * ACF20-500 and ACF20-700 refer to ACF20 after heat treatment at 500 and 700?C,       
respectively. 
3.2.3 Sulfur dioxide adsorption properties of activated carbon fibers and modified 
activated carbon fibers  
3.2.3.1 Sulfur dioxide adsorption over bare activated carbon fibers The activated carbons 
were subjected to 20 ppm sulfur dioxide with a face velocity of 33cm/s with air as the 
77 
balance at room temperature. The water was firstly excluded in the main gas stream, 
where the sulfur dioxide is expected to be adsorbed onto free sites on fiber surfaces 
where they are oxidized to sulfur trioxide.  The resulting breakthrough time, 
breakthrough capacity, saturation capacity and utilization of ACFs and PAC are listed in 
Table 3.6.  
    Adsorption capacity of porous adsorbents is usually dependent on their surface area 
since more free sites are available in adsorbents with a higher surface area. However, 
there is no general relation between surface area and SO2 breakthrough capacity of ACFs 
as shown in Figure 3.7.  
Samples are classified according to precursors, namely novoloid-based ACFs and 
rayon-based ACFs. For novoloid-based ACFs, ACF10 with 88% micropores exhibited 
the highest breakthrough capacity of 1.72 while ACF20 with the least micropore ratio 
gave the lowest breakthrough capacity of 0.45. In the rayon-based ACF group, SY with 
90.73% micropores is more capable in desulfurization than CR whose micropore ratio is 
much smaller, 71.63%. In other words, SO2 breakthrough capacity of ACFs in dry 
condition is proportionately dependent on micropore ratio or on pore size distribution for 
each group: the higher the micropore ratio, the higher the breakthrough capacity. More 
micropores in adsorbents indicates a lower ratio of mesopores and macropores through 
which adsorbates have to pass before reaching adsorption sites in micropores where 
adsorption usually occurs. Blockage is more likely to happen if there are more 
macropores in adsorbents. Furthermore, SO2 capacity for each group of ACFs was 
inversely proportional to micropore pore size, pore volume and surface area, which can 
be explained by the concept proposed by Marsh et al., where adsorbents with smaller 
78 
pores adsorb contaminants better at lower concentrations due to a higher overlap in 
potential between the pore and walls [10, 12]. 
Breakthrough profiles of ACFs and PAC are shown in Figure 3.8 when water is 
excluded in the carrier gas. ACF10, ACF15, ACF20 and CR became saturated by sulfur 
dioxide quickly after breakthrough in comparison with SY and PAC which have lower 
slopes in breakthrough curves. SY exhibited the highest utilization of 22.69% (Table 3.6) 
which is even higher than PAC which has the highest breakthrough and saturation 
capacity.  
Although PAC has a longer breakthrough time and thus higher breakthrough capacity 
as shown in Figure 3.8, the breakthrough curve of PAC has lower slope which reflects the 
slower adsorption rate, probably due to harder access of adsorbates to free sites in pore, 
in comparison with ACFs.  As discussed in Chapter 1, micropores in ACFs allow 
adsorbate to reach the fiber center directly while, in PAC, adsorbates have to pass 
through macropores in order to reach micropores where adsorption usually happens. The 
diameter of PAC was measured to be 290 ?m, which is much thicker than ACFs with a 
diameter of about 12 ?m. Although ACF20 and PAC have very close micropore ratio, 
micropores in PAC may be imbedded in the middle and therefore it is hard to reach for 
adsorbates.  
79 
Table 3.6 Sulfur dioxide capacities of bare activated carbon fibers in the absence of water 
ACF10 ACF15 ACF20 CR SY PAC
BT (min) 4.67 3.57 1.4 7.67 14.75 19
BC (mg SO2/g sample) 1.72 1.32 0.45 2.85 5.05 7.07
SC (mg SO2/g sample) 4.26 1.98 1.99 5.52 10.68 16.39
Utilization (%) 14.15 16.23 5.83 10.22 22.69 19.79
*Sample weight is about 0.5 g, the initial sulfur dioxide concentration is 20 ppm, and the face 
velocity is about 35 cm/s at the temperature of 20?C.  BT is the breakthrough time and BC is 
breakthrough capacity. 
Figure 3.7 Effect of surface area of ACFs on sulfur dioxide breakthrough capacity 
80 
Without the presence of water in the carrier gas, SY exhibits the highest breakthrough 
capacity with the longest period before breakthrough among all ACFs (Figure 3.8 and 
Table 3.6) although it is still lower than that of PAC. The spent SY felts were regenerated 
both in air (110?C for 2 hour) and nitrogen (600?C for 1 hour). Air-regenerated SY at 
110?C gave a breakthrough capacity of 1.98 mg SO2/g sample, which is 39.21% of the 
original capacity, while the nitrogen-regenerated SY shows a breakthrough capacity of 
6.5 mg SO2/g sample, which is 50% of the initial capacity, in dry air. The saturation 
capacity of nitrogen-regenerated SY is 4.93 mg SO2/g sample, which is 46.16% of the 
fresh sample. Furthermore, the utilization of N2-regenerated SY (27.08%) is a little 
higher than the fresh one (22.69%) which is due to decreased saturation time.  
  As shown in Table 3.7 and Figure 3.10, the presence of water in the main gas stream 
greatly increased the sulfur dioxide capacities of all ACFs and PAC while their 
utilizations were affected by water differently. Utilizations of ACF10, ACF15,ACF20 
Figure 3.8 Breakthrough profiles of ACFs with the absence of water in gas stream 
(Cin is 20 ppm, velocity is 33 cm/s and sample weight is 0.5 g.) 
81 
and CR were improved since their breakthrough times were increased greatly while those 
SY and PAC were reduced because they needed a lot longer time to achieve saturation as 
shown in breakthrough curves in Figure 3.9. For activated carbons fibers, the outlet SO2
concentration first increased quickly after breakthrough; and then the rate of increase 
slowed down, causing longer time to achieve saturation. The increased capacity is 
attributed to the water which eventually promoted the continuous removal of sulfur 
dioxide through a series of procedures which was proposed by Mochida et al., and can be 
described by Figure 3.10 along with equations 3.3 - 3.8 [13]. The oxidized sulfur dioxide, 
sulfur trioxide, is hydrated by water producing sulfuric acid which is attached to fiber 
surfaces occupying the adsorption sites. If extra water is available, adsorbed sulfuric acid 
is eluted to aqueous from fiber surface releasing active adsorption sites for coming sulfur 
dioxide, preceding the continuous removal of sulfur dioxide. 
Table 3.7 Sulfur dioxide capacities of bare activated carbon fibers with water present 
ACF10 ACF15 ACF20 CR SY PAC
BT (min) 25 18.5 14.67 24 19.75 24
BC (mg SO2/g sample) 7.92 6.84 5.46 8.73 6.99 8.86
80% SC (mg SO2/g sample) 13.89 20.1 37.32 32.23 44.94 74.81
Utilization (%) 21 25.26 10.55 16.78 7.68 5.8
*Sample weight is about 0.5 g, the initial sulfur dioxide concentration is 20 ppm, and the face 
velocity is about 35 cm/s at the relative humidity of about 100% and temperature of 20?C.   BT is 
the breakthrough time and BC is breakthrough capacity. 
82 
g3    
SOg2870 ?SOg2870g4666g2911g2914g2929g4667                                                                                                     (3.3) 
Hg2870O ?Hg2870Og4666g2911g2914g2929g4667                                                                                                   (3.4) 
Og2870 ? 2Og4666g2911g2914g2929g4667                                                                                                        (3.5) 
SOg2870g4666g2911g2914g2929g4667 g3397Og4666g2911g2914g2929g4667 ?SOg2871g4666g2911g2914g2929g4667                                                                               (3.6) 
0
20
40
60
80
100
0 50 100 150 200 250 300 350 400
Co
ut/
Ci
n (
%)
Time (min)
ACF10
ACF15
ACF20
SY
PAC
CR
Figure 3.9 Breakthrough profiles of ACFs with the presence of water in the carrier gas (Initial 
sulfur dioxide concentration was 20 ppm, face velocity was 33 cm/s and sample weight was 0.5 
g). 
Figure 3.10 Mechanism of sulfur dioxide removal over activated carbon surface with the 
presence of oxygen and water [13]
83 
g22g18g2871g4666g2911g2914g2929g4667 g3397g11g2870g18g4666g2911g2914g2929g4667 g1372 g11g2870g22g18g2872g4666g2911g2914g2929g4667                                                                       (3.7) 
g11g2870g22g18g2872g4666g2911g2914g2929g4667 g3397g144g11g2870g18 g1372 g11g2870g22g18g2872 g942 g144g11g2870g18g4666g2911g2914g2929g4667                                                            (3.8) 
When water was present in the carrier gas, the breakthrough capacity of novoloid-
based ACF is still proportionally related to micropore ratio of ACFs, which is the same 
trend with those with the absence of water. However, their saturation capacities increased 
with the surface area and pore volume, as shown in Table 3.7. In comparison with fibers, 
PAC has a higher breakthrough capacity if water is absent in the carrier gas. If the air was 
humidified before mixing with sulfur dioxide, breakthrough capacity of CR is close to 
PAC but saturation capacity of PAC is higher.  
Moreover, under humid conditions, both breakthrough capacity and saturation capacity 
of ACFs are observed to be impacted by nitrogen content in fibers. The higher the 
nitrogen content, the higher the breakthrough and saturation capacity of ACFs.   
Considering the long experiment period, the input sulfur dioxide concentration was 
increased to 50 ppm, the resulting SO2 capacities and breakthrough curves over ACF10, 
ACF15, ACF20, CR and PAC in the presence of water are shown in Table 3.8 and Figure 
3.11, respectively.  
Table 3.8 Sulfur dioxide capacities of bare ACFs with initial SO2 concentration of 50 ppm and 
humidified air as the carrier gas 
ACF10 ACF15 ACF20 CR PAC
BT (min) 3.48 1.7 0 4.67 3.15
BC (mg SO2/g sample) 2.63 1.3 0 3.53 2.405
80% SC (mg SO2/g sample) 12.66 14.77 28.94 11.64 36.6
80% Saturation Utilization (%) 13.92 5.89 - 28.3 4.505
84 
    Compared to low concentration sulfur dioxide challenge, breakthrough times and 
capacities of ACFs and PAC decreased drastically when exposed to a higher 
concentration SO2. ACF20 did not show any capacity at breakthrough. The same 
phenomenon of   low utilization of PAC was observed regardless of challenge 
concentration of sulfur dioxide though higher saturation capacities could be reached.  
3.2.3.2. Sulfur dioxide adsorption over alkali/ACF ACF20 has the largest surface area 
and pore volume among all ACFs available in the lab and thus is expected to 
accommodate more solutions in pores, ACF20 was immersed in KOH and K2CO3
solutions to apply them onto ACF20. A variety of weight percentages of KOH or K2CO3
0
20
40
60
80
100
0 10 20 30 40 50 60 70 80
C o
ut/
C in
(%
)
Time (min)
ACF10
ACF15
ACF20
CR-15
PAC
Figure 3.11 Breakthrough profiles of ACFs in SO2/humidified air (Initial SO2
concentration was 50 ppm, the face velocity was 33 cm/s and sample weight was 0.5 g) 
85 
were loaded on ACF20: 5, 8.67, 11, 29.3 wt% KOH and 8.67, 19.3 and 28 wt% K2CO3. 
Their sulfur dioxide capacities were evaluated while exposed to 20 ppm sulfur dioxide 
carried by dry air with a face velocity of 33 cm/s at room temperature (20?C). 
Breakthrough behaviors of these samples are compared in Figure 3.12. It can be seen that 
even a small amount of KOH (5 wt%) on ACF20 increased the breakthrough time from 
1.4 min to 21min, and increased the breakthrough capacity from 0.45 to 7.81 mg SO2/g 
sample. However, the increased capacity by alkali sacrificed the adsorption rate of 
ACF20, which is elucidated by less steep breakthrough curves of alkali/ACF20 compared 
to bare ACF20 in Figure 3.12. However, the utilization of ACF20 was increased. Both 
SO2 breakthrough capacity and saturation capacity are increased with the increased KOH 
and K2CO3 content (up to 29 wt%) on ACF20, as shown in Figure 3.13. The influences of 
KOH load on breakthrough and saturation capacities are different. On the one hand, at 
low KOH weight ratio on AF20, breakthrough capacity increased quickly but the increase 
in speed slowed down when 8.67 wt% or more KOH were loaded. On the other hand, the 
saturation capacity of KOH/ACF20 gradually increased with the increased KOH amount 
on ACF20. The similar trend was observed for K2CO3/ACF20 as well. Utilizations of 
alkali/ACF20 are less dependent on alkali weight ratio: the increase was from 12.96% at 
5 wt% KOH to 21.18% at 8.67 wt% KOH; the fluctuation was in the range of 
19.72~22.15% for other alkali loads. 
86 
Although potassium carbonate is weaker alkali in comparison with potassium 
hydroxide, with the same load (8.67 wt%) on ACF20, their capacities are close and even 
the breakthrough curve shapes are very similar. However, as their load increased, their 
effect on sulfur dioxide capacity became distinct:  KOH turns out to be a better promoter 
for sulfur dioxide removal.  
It should be noted that one of the by-products of the desulfurization over 
K2CO3/ACF20, carbon dioxide, causes a second contamination on fuel cell system during 
the test, which is not desirable.  
Figure 3.12 Breakthrough curves of KOH/ACF20 exposed to 20 ppm SO2 carried by dry air
(Cin=20 ppm, Face velocity=33 cm/s, Sample=0.5 g) 
87 
 
 
 
 
 
 
 
Figure 3.13 SO2 adsorption capacity and utilization of KOH/ACF20 and K2CO3/ACF20 
 
3.2.3.3 Sulfur dioxide adsorption over activated carbon fibers supported metal catalysts 
Copper, nickel and iron were supported on ACF20 in order to enhance sulfur dioxide 
removal abilities of ACFs because these three metal derivatives were reported to have 
higher activity in sulfur dioxide removal. Metal/ACF20 was exposed to the challenge gas 
of 50ppm sulfur dioxide carried by humidified air with a face velocity of 33 cm/s. 
Treatment parameters and resulting breakthrough capacities are shown in Table 3.9. In 
comparison with bare ACF20 under the same conditions, the loads of metals of copper, 
nickel and iron enhanced the sulfur dioxide breakthrough capacity at different levels. 
0
5
10
15
20
25
0
10
20
30
40
50
60
70
0 5 10 15 20 25
Ut
iliz
ati
on
 (%
)
SO
2c
ap
aci
ty 
(m
g S
O 2
/g 
sam
ple
)
Alkali load (wt%)
KOH/ACF20 BC KOH/ACF20 SC K2CO3/ACF20 SC
K2CO3/ACF20 BC KOH/ACF20 utilization K2CO3/ACF20 utilization
88 
When copper and nickel were loaded on fibers with the incipient wetness method, the 
former had higher sulfur dioxide capacity. However, when they were loaded with 
immersing ACFs in metal nitrate solutions, the capacities were found to be much lower, 
about 1. 3mg SO2/g sample. Pretreatment with HCl before modification with copper 
nitrate by the incipient wetness method increased the sulfur dioxide breakthrough 
capacity from 3.55 to 9.28 mg SO2/g sample. 
89 
Table 3.9 Sulfur dioxide capacities of metal/ACF20 
ACFs 
Metal 
precursors 
Con1
(wt%) 
Load    
(wt%) 
TM2
BT 
(min) 
BC
(mgSO2/g 
sample) 
ACF20 Cu(NO3)2g9423H2O 12.9 5 IWI 4.5 3.55 
HCl-
ACF203
Cu(NO3)2g9423H2O 12.9 5 IWI 11.8 9.28 
HNO3-
ACF203
Cu(NO3)2g9423H2O 12.9 5 IWI 2.16 1.78 
ACF20 Ni(NO3)2g9426H2O 16.96 5 IWI 3.34 2.62 
ACF20 Fe(NO3)3g9429H2O 17.34 3.5 IWI 11.67 9.25 
ACF20 Cu(NO3)2g9423H2O 12.9 51.15 IM4 1.667 1.305 
ACF20 Cu(NO3)2g9423H2O 2 7.9 IM4 1.667 1.309 
       1. Concentration of precursor solutions; 2.  Treatment method; 3. HCl-ACF20 and HNO3-
ACF20 represent ACF20 was pretreated with HCl and HNO3, respectively; 4. Immersion. 
3.2.3.4 Sulfur dioxide adsorption over activated carbon fiber supported oxidants Two 
different oxidants were used to oxidize sulfur dioxide in the current research: potassium 
chlorate and potassium permanganate. They were treated on ACF15 by spraying a certain 
amount of KClO3 or KMnO4 solutions over ACF15 followed by washing and drying.  
The prepared oxidant/ACF15 fibers were exposed to 50 ppm SO2 stream carried by either 
dry or humidified air with a face velocity of 33 cm/g at room temperature of 20?C.  
Capacities of ACF15 loaded with KClO3 by impregnation are shown in Figure 3.14.  It 
can be concluded that breakthrough capacity is promoted almost linearly by loading 
KClO3 on the fibers. The maximum SO2 capacity is expected with a certain amount of 
90 
KClO3 considering that overloading may result in pore blockage. However, KClO3 is 
easily removed by exposure to water and thus it is not suitable for wet-laying which is a 
flexible method for forming activated carbon fiber paper. When the same amount of 
KClO3 and KMnO4 (for example, 15 wt%) was applied to ACF15 separately, KMnO4 
showed a higher capacity, about 40 mg SO2/g sample,  compared to about 27 mg SO2/g 
sample of KClO3 as shown in Figure 3.15. Another issue of potassium chlorate is that the 
mixture of sulfur and KClO3 is prone to spontaneous deflagration and sulfur-containing 
acids can cause spontaneous ignition [14].  
Different weight ratios of KMnO4 were loaded on ACF10, ACF15, ACF20 and PAC, 
to investigate the best support. In Figure 3.15, it can be seen that, as KMnO4 content 
increases on ACF10, ACF15, ACF20 and PAC, the breakthrough capacities increase. 
With the same amount of KMnO4, PAC exhibited the lowest sulfur dioxide breakthrough 
capacity and ACF10 showed the best capacity to remove SO2 among the three different 
y = 1.5869x + 3.1853
R? = 0.8154
0
10
20
30
40
50
0 5 10 15 20 25
SO
2b
rea
kth
rou
gh
 ca
pa
cit
y 
(m
g S
O 2
/g 
sam
ple
)
KClO3 load (wt%)
KClO3/ACF15 KClO3/ACF15 after exposure to water
Figure 3.14 SO2 breakthrough capacities of KClO3/ACF15 before and 
after exposure to water 
91 
ACFs. Although ACF20 contains the largest pore volume and thus was expected to 
accommodate more KMnO4 solutions than the other two, KMnO4/ACF20 exhibits the 
lowest capacity for sulfur dioxide.  
When KMnO4 solution was injected onto ACFs, it was expected to be diffused into the 
pores without forming aqueous solutions. However, when the solution volume was 
greater than 5 mL, which resulted in 10 wt% KMnO4 over ACF10, extra aqueous solution 
was observed with ACF10. Thus, ACF10 was suspected to hold more solution with 
increased KMnO4 weight ratio. Therefore, ACF15 was chosen for the rest of the research 
work. 
 As discussed previously, KOH, K2CO3 and KClO3 are not eligible for the wet-laying 
method to form activated carbon fiber paper (ACFP). Capability of KMnO4/ACF in SO2
purification was slightly affected by exposure to water as shown in Figure 3.16. The 
maximum SO2 capacity of KMnO4/ACF was obtained with 43.34 wt% KMnO4 load. 
Furthermore, the effect is more at a higher content of KMnO4 than those with a lower 
ratio. 
The breakthrough curves of KMnO4/ACF15 prior to and after exposure to water are 
shown in Figure 3.17. They are similar to each other except the slightly increased slope 
of the washed sample, which is probably due to the higher contact efficiency caused by 
redistribution of KMnO4 on ACF15 during washing. Although the breakthrough capacity 
of washed sample is higher, its saturation capacity of 210 mg SO2/g sample is lower than 
that of 317 mg SO2/g sample before washing.  
It has been reported that experimental parameters affect desulfurization by carbon 
including gas component, temperature, humidity and other properties. KMnO4/ACF15 
92 
was exposed to different challenge gases, namely, SO2, SO2+air, SO2+H2O and 
SO2+H2O+air and their breakthrough profile characteristics are compared in Figure 3.18, 
which exhibits drastically different SO2 capacities and breakthrough curve patterns. Both 
the presence of oxygen and water promote the ability of desulfurization of 
KMnO4/ACF15 (43.34 wt%) while co-presence of them enhances the ability further. 
Furthermore, the improvement of sulfur dioxide capacity caused by water seems to be 
more than that resulted from oxygen.  
    For the same amount of KOH, K2CO3 and KMnO4 on ACF20, such as 22 wt%, the 
breakthrough capacities were 19.69, 16.75 and 27.95 mg SO2/g sample, respectively, 
indicating that KMnO4 promoted the sulfur dioxide capacity more than the other two.  
Figure 3.15 SO2 breakthrough capacity of KMnO4 loaded ACF10, ACF15, ACF20 and PAC 
(50 ppm input sulfur dioxide, 33 cm/s face velocity and 0.5 g sample for adsorption test). 
93 
0
20
40
60
80
100
0 100 200 300 400 500 600
Time (min)
Before exposure to water
After 
exposure to 
water
Figure 3.16 SO2 breakthrough capacity of ACF15 loaded with different 
ratios of KMnO4 (NW refers to non-washing).  
Figure 3.17 Breakthrough profiles of KMnO4/ACF15 before and after exposure to water 
94 
    The potassium permanganate solutions used for KMnO4/ACFs preparation were all 6 
wt% so far. It is possible that the desulfurization capability was influenced by the 
precursor concentration. A lower KMnO4 concentration (2 wt%) was used to treat ACF15 
and different weight ratios of KMnO4 were obtained. For the same KMnO4 load (43 wt%) 
on ACF15, samples prepared from 2wt% and 6wt% solutions showed similar 
breakthrough curves and saturation process; however, ACF15 impregnated with 6wt% 
concentration of KMnO4 solution exhibited a longer breakthrough period as shown in 
Figure 3.19. Their breakthrough capacities are 34 and 44.3 mg SO2/g sample, 
respectively.  
    The adsorption properties of ACF15 with 20, 33 and 43 wt% KMnO4 prepared from  
2wt% solutions are illustrated in Figure 3.19. Their breakthrough capacities were 
calculated to be 34, 59.7 and 60.8 mg SO2/g sample, respectively. It can be seen that, 
Figure 3.18 Breakthrough profiles of KMnO4/ACF15 after exposure to 
water under different adsorption conditions 
SO2+N2
95 
with the increase of KMnO4 content, sulfur dioxide saturation capacity increased but the 
breakthrough capacity increased first and then leveled off.  
   
  It was reported that the pre-humidification before activated carbons were exposed to 
sulfur dioxide improved the sulfur dioxide removal efficiency by forming a water film 
over carbon surface where the conversion of sulfur trioxide to sulfuric acid and elution of 
aqueous sulfuric acid from carbon surfaces were promoted. KMnO4 treated ACF15 was 
exposed to humidified air (91% relative humidity) for one hour before starting the sulfur 
dioxide adsorption test and a film of water was expected to form. However, the 
breakthrough period was shortened in comparison with dry KMnO4/ACF15s (Figure 
3.20) and the adsorption rate was initially similar which is indicated by a similar slope of 
the breakthrough curve. After 80% sulfur dioxide released from the reactor, the 
adsorption rate on the prehumidified sample slowed down and became lower than that of 
0
20
40
60
80
100
0 100 200 300 400 500
Co
ut/
Ci
n (
%)
Time (min)
20%
33%
43%
43% - 6*
Figure 3.19 Breakthrough profiles of KMnO4/ACF15 prepared from 2 wt% KMnO4
solutions (* 43%-6 represents the sample prepared from 6wt% KMnO4 solution)
96 
dry samples. On the other hand, the saturation capacity (85%) of prehumidified 
KMnO4/ACF15 of 215 mg SO2/g sample is higher than that of dry samples of 205 mg 
SO2/g. Furthermore, as the adsorption process continues, a higher saturation capacity 
would be obtained if samples were prehumidified.   
3.3 SEM imaging analysis  
SEM images of KMnO4/ACF15 with different chemical weight ratios are shown in 
Figure 3.21. At a low KMnO4 load (6.55 wt%), the distribution of KMnO4 on fiber 
surfaces are not uniform: some areas are smooth while KMnO4 is dense on some other 
areas causing coarse fiber surfaces. As the weight ratio increased to 23.16 wt%, the 
uniformity of KMnO4 was enhanced; it seems that fiber surfaces were covered by 
KMnO4 at the weight percentage of 43.34 wt%. After exposure to water, KMnO4 seemed 
to be redistributed uniformly, which may contribute to efficient contact between the 
0
20
40
60
80
100
0 100 200 300 400 500
C o
ut/
C in
(%
)
Time (min)
Dried
Prehumidified
Figure 3.20 Sulfur dioxide adsorption properties of humidified KMnO4/ACF15 in 
comparison with dried sample (initial SO2 is 50 ppm, face velocity is 33 cm/s, 
and sample weight is 0.5 g.) 
97 
adsorbate and oxidant during adsorption tests, which is consistent with the sharper 
breakthrough curve as discussed in the previous section. 
1 2
3 4
Figure 3.21 Surface morphology of KMnO4/ACF15 1) 6.55wt% KMnO4 2) 23.16 
wt% KMnO4, 3) 43.34 wt% KMnO4 and 4) 43.34 wt% KMnO4 after exposure to 
water
98 
    Since 34.34 wt% KMnO4 over ACF15 showed the highest sulfur dioxide capacity, they 
were subjected to different challenge gases: SO2+N2, SO2+H2O+N2, SO2+air and 
SO2+H2O+air and the spent samples were observed under SEM, which are shown in 
Figure 3.22. As KMnO4/ACF15 was exposed to SO2 only, the fiber surfaces were similar 
to the fresh sample but the challenge gas mixture of SO2 and O2 without water resulted in 
rougher surfaces. The coarse surfaces are considered to result from incomplete reaction 
between sulfur dioxide and potassium permanganate.  However, the other two spent 
samples, which were exposed to the water-containing challenge gases, exhibited much 
smoother surface with few particles attached, which may be due KMnO4 being consumed 
Figure 3.22 KMnO4/ACF15 surface morphology after exposure to 1) SO2+N2,  2) 
SO2+air, 3) SO2+H2O+N2 and 4) SO2+H2O+air
43
1 2
by sulfur dioxide during the reaction. 
remove sulfur dioxide by
3.4 Activated carbon 
For the purpose of optimizing paper strength, bare 
were used. Thickness, tensile strength and
thirteen ACFPs are shown in Figure 
Table 3.1.   
Figure 3.23 Thickness, tensile strength and coeffic
samples corresponding to Table 3.1 
    As shown in Figure 
contributed greatly to the paper strength 
As the ratio of cellulose fibers w
decreased. 15% CoPET was inclu
the strength of ACFP A and B is mainly provided by 
entanglement between cellulose fibers. Among all 
0
0.5
1
1.5
2
2.5
3
A B C
Thickness (mm)
Tensile strength (N/mm)
CV of thickness*10 (%)
99 
From the SEM images, it seems that a
 potassium permanganate.  
fiber papers (ACFP) and their SO2 adsorption capacities
ACFs rather than KMnO4/ACF15 
coefficient of variation
3.23. The composition of each sample is listed in 
ient of thickness variation (CV) of ACFP       
3.23, cellulose fibers in the manufactured ACFP A and B 
(the wet mat was dried in air without heating
as decreased from 50% to 15%, the strength of ACFP 
ded for sample C which was fused. It can be seen th
the hydrogen bonding and 
the samples, ACFPs contain
D E F G H I G
 key factor to 
 of thickness (CV) of 
). 
at 
ing 
K L M
100 
ACF20 and cellulose show better uniformity than most of ACFPs containing CoPET 
probably due to easy dispersion of cellulose in the slurry. The hydrophobic feature of 
CoPET causes it to be less likely dispersed in water than the hydrophilic cellulose fibers.  
However, a high ratio of ACF20 is preferred for the purpose of sulfur dioxide removal 
and thus the weight ratio of ACF20 in ACFP was chosen 70% for the rest of samples.  
The cellulose and CoPET were mixed together as the bonding agents and papers were 
fused at high temperature for bonding from CoPET. As the fusing temperature increased 
from 135 to 200?C, the strength of ACFP also increased. However, the fusing 
temperature was limited by cellulose fibers which decompose above 200?C.  
    A large group of ACFP samples were made from ACFs (70 wt%) with CoPET (30 
wt%) as the binder with a series of fusing temperatures. As the fusing temperature 
increased, the strength increased first and then decreased. As shown in DSC curves of 
CoPET (Figure 3.24), CoPET has two melting temperatures due to its core/sheath 
structure. The low melting point is attributed to the sheath and the high melting point is 
due to the core polyester. As the wet mat from the wet lay process was heated above the 
low melting point (76?C), the sheath of CoPET started melting and thus bonded ACF20 
with the unmelted CoPET core which provided the structure of paper as shown in Figure 
3.25. The higher the fusing temperature, the more polyester melted. However, while too 
much binder fibers were melted, less polyester was left, which thus resulted in weaker 
ACFPs. The optimum fusing temperature was found to be 200?C which gave the 
strongest ACFP (referring to sample G).   Surfactant was used to enhance the affinity 
between CoPET and water and thus to produce a more uniform dispersion for the wet lay 
process, which corresponded to sample M. Other parameters were the same with the 
101 
sample G.  Compared to G without surfactant, the low coefficient of variation of 
thickness in M indicates an increase in uniformity which accordingly may have 
contributed to the improved paper strength.  
    
Figure 3.24 DSC analysis of CoPET 
102 
   
  KMnO4/ACF15 paper (K-ACFP) was manufactured with optimum parameters obtained 
with the bare ACFs: the wet mat consisting of 70 wt% KMnO4/ACF15 and 30 wt% 
CoPET was heated at 200?C for 20 min. The surface morphology of K-ACFP was 
observed with SEM as shown in Figure 3.26 (1). Fused polyester formed strong 
interaction in the K-ACFP but some parts of ACFs were covered with the fused binder 
fibers and thus the effective surface area for sulfur dioxide removal reduced in theory.   
The ability of resulting K-ACFP in sulfur dioxide removal was tested and the 
breakthrough properties are shown in Figure 3.27. K-ACFP showed shorter breakthrough 
period because since KMnO4/ACF15 was included because 30 wt% bicompoent 
polyester in ACFP. However, as the capacity was calculated with respect to 
KMnO4/ACF15 rather than the sample weight of ACFP, the estimated breakthrough 
capacity of ACFP (76 mg SO2/g modified ACF15) is close to that of fibers (74.78 mg 
Figure 3.25 SEM images of ACFP composed of 70 wt% ACF20 and 30 wt% CoPET 
with a fusing temperature of 200?C for 20 min (?500) 
103 
SO2/g modified ACF15) while the saturation capacity of ACFP (186.75 mg SO2/g 
modified ACF15) is less than that of fibers (210.83 mg SO2/g modified ACF15). ACFP 
exhibited quicker saturation illustrated by the steeper breakthrough curve (Figure 3.27), 
which is probably due to more efficient contact when activated carbon fibers are diluted 
by binders. As shown in Figure 3.26 (2), fiber surface after sulfur dioxide adsorption test 
became smoother with some scale-like substances which are probably the adsorption 
products.   
It is possible that the desulfurization capability of KMnO4/ACF15 is reduced due to the 
existence of joint points where fused binder fibers blocked fiber surface exhibiting the 
reaction. To investigate the effects of joints, the formed wet mat was dried naturally in air 
rather than heat at high temperature. Thus, binder fibers remained as they were and there 
were not joints in the fiber network. Fibers in unfused ACFP were individual fibers 
which were subjected to sulfur dioxide under the same conditions with those fused ACFP. 
The resulting 90% saturation capacity was 217 mg SO2/g modified ACF15, which is 
slightly higher than those fused ACFP. It can be concluded that the joints formed by 
fusing only occupies a small part of reactive sites on fiber surfaces and thus barely 
influence sulfur dioxide capacity.    
104 
1
2
Figure 3.26 SEM images of K-ACFP: (1) before and (2) after 
sulfur dioxide adsorption tests (?2000). 
105 
3.5 XRD analysis on KMnO4/ACF15 
To investigate the changes on ACFs surface caused by the adsorption of sulfur dioxide 
with KMnO4/ACF15, X-ray diffraction analysis was used. Bare ACF15, KMnO4/ACF15 
and spent KMnO4/ACF15 after exposed to sulfur dioxide carried by dry air, dry nitrogen, 
humidified nitrogen and humidified air were subjected to X-ray scanning.  
XRD patterns of ACF15, KMnO4/ACF15 and spent KMnO4/ACF15 after exposure to 
sulfur dioxide carried by humidified air are shown in Figure 3.28. Peaks at 14.5?, 28? and 
34? indicate the crystals of MnO2, which are only observed in the spent sample and 
therefore MnO2 is considered to be one of the products during the adsorption test. The 
peak at 17? and 19? come from potassium permanganate. The expected product MnSO4
during the sulfur dioxide adsorption was not detected probably due to the small amount. 
Figure 3.27 Breakthrough profiles of ACFP in comparison with KMnO4/ACF15 
with 50 ppm SO2 challenging gas carried by humidified air (ACFP contains 
0.35g KMnO4/ACF15 while KMnO4/ACF15 is 0.5 g). 
106 
MnO2 was also detected by XRD after KMnO4/ACF15 was exposed to the challenge gas 
of SO2+H2O+N2 as shown in Figure 3.29. Figure 3.30 illustrates the XRD pattern of 
KMnO4/ACF15 before and after adsorption in SO2+air. There is no appearance of new 
peaks observed like in Figures 3.28 and 3.29. The same XRD pattern was obtained if 
adsorption was performed in SO2+N2. In other words, if water was absent from the 
challenge gas, reaction between sulfur dioxide and potassium permanganate on fiber 
surfaces would be less likely to occur, which is consistent with the previous discussion. 
2g2016
Figure 3.28 XRD patterns of ACF15, KMnO4/ACF15 and spent 
KMnO4/ACF15 after exposure to sulfur dioxide carried by humidified air 
KMnO4/ACF15
ACF15
Spent KMnO4/ACF15
107 
Figure 3.29 XRD patterns of KMnO4/ACF15 before and after adsorption of sulfur 
dioxide carried by humidified nitrogen 
KMnO4/ACF15
KMnO4/ACF15 after 
adsorption
108 
3.6 Energy Dispersive X-ray analysis
Energy Dispersive X-ray (EDS) analysis was performed on a SEM/EDS system to 
analyze the element composition on KMnO4/ACF15 fiber surfaces both before and after 
sulfur dioxide adsorption tests with the carrier gas of humidified air. Weight percentages 
and atom percentages of K, Mn, S and C are listed in Table 3.10. Sulfur content on 
KMnO4/ACF15 is very low, 0.045 wt% and 0.02 at%, but increased to 9.47 wt% and 4.3 
wt%. Sulfur species were caught by adsorbents effectively and the goal of desulfurization 
was achieved. Changes of other elements are because of the change of sulfur ratio. As 
shown in Figure 3.31, before sulfur dioxide adsorption, the intensity of potassium and 
permanganate signals are very strong and sulfur signal is barely observed. However, after 
KMnO4/ACF15 after adsorption of SO2
carried by dry air  
KMnO4/ACF15
Figure 3.30 XRD patterns of KMnO4/ACF15 before and after 
adsorption of sulfur dioxide carried by dry air 
109 
adsorption, the sulfur signal intensity was strengthened greatly although the percentage of 
sulfur was still far lower than the  carbon element.    
Table 3.10 Element contents of KMnO4/ACF15 before and after adsorption of SO2 in humidified 
air 
K Mn S C
wt% at% wt% at% wt% at% wt% at%
Before 11.77 4.855 21.42 6.29 0.045 0.02 66.75 88.83
After 8.22 3.1 7.34 1.96 9.47 4.30 74.94 90.62
*wt% and at% represent weight percentage and atom percentage, respectively. 
110 
1
2
Figure 3.31 EDS element signals (1) before and (2) after adsorption in the 
challenging gas of SO2+H2O+air 
111 
References 
1. Hayes, J., Activated carbon fiber textiles for filtration and separation, Brochure of 
American Kynol, Inc.  
2. Das, D., Gaur, V. and Verma, N., Removal of volatile organic compound by 
activated carbon fiber, Carbon, 42 (14), 2949-2962, (2004). 
3. Tseng, H. H. and Wey, M. Y., Study of SO2 adsorption and thermal regeneration 
over activated carbon-supported copper oxide catalysts, Carbon, 42 (11), 2269-
2278, (2004). 
4. Davini, P., The effect of certain metallic derivatives on the adsorption of sulphur 
dioxide on active carbon, Carbon, 39 (3), 419-424, (2001). 
5. Tseng, H. H., Wey, M. Y. and Fu, C. H., Carbon materials as catalyst supports for 
SO2 oxidation: catalytic activity of CuO-AC, Carbon, 41 (1), 139-149, (2003). 
6. Ma, J. R., Liu, Z. Y., Liu, S. J. and Zhu, Z. P., A regenerable Fe/AC desulfurizer 
for SO2 adsorption at low temperatures, Applied Catalysis B-Environmental, 45 
(4), 301-309, (2003). 
7. Wang, J. Y., Zhao, F. Y., Hu, Y. Q., Zhao, R. H. and Liu, R. J., Modification of 
activated carbon fiber by loading metals and their performance on SO2 removal, 
Chin. J. Chem. Eng., 14 (4), 478-485, (2006). 
8. Gaur, V., Sharma, A. and Verma, N., Removal of SO2 by activated carbon fibre 
impregnated with transition metals, The Canadian Journal of Chemical 
Engineering, 85, 188-198, (2007). 
112 
9. Burevski, D., The application of the Dubinin-Astakhov equation to the 
characterization of microporous carbons, Colloid and Polymer Science, 260 (6), 
623-627, (1982). 
10. Marsh, H. and Rand, B., The characterization of microporous carbon by means of 
the dubinin-radushkevich equation, Journal of Colloid and Interface Science, 33, 
101, (1970). 
11. Autosorb-1 AS1 Win Version 1.51 operating manual, Boynton Beach, Florida 
USA, (2005). 
12. Daley, M. A., Mangun, C. L., Debarr, J. A., Riha, S., Lizzio, A. A., Donnals, G. L. 
and Economy, J., Adsorption of SO2 onto oxidized and heat-treated activated 
carbon fibers (ACFS), Carbon, 35 (3), 411-417, (1997). 
13. Mochida, I., Korai, Y., Shirahama, M., Kawano, S., Hada, T., Seo, Y., Yoshikawa, 
M. and Yasutake, A., Removal of SOx and NOx over activated carbon fibers, 
Carbon, 38 (2), 227-239, (2000). 
14.  http://en.wikipedia.org/wiki/Potassium_chlorate, Jan 2010 accessed. 
  
113 
Chapter 4 Mechanism Analysis of Sulfur Dioxide 
Removal by ACFs and Model Development  
4.1 Introduction 
In 1946, Klotz proposed the concept of adsorption wave to understand the variation 
in concentration of a gas effluent from a bed of adsorbent and equations were 
derived to predict canister life [1]. Amundson supplied a complete mathematical 
solution of one particular problem according to the material balance over an 
elemental thickness of bed with the assumption of the local rate of removal made by 
Bohart and Adams and gave an equation to estimate the dependence of effluent 
concentration on time and location [2, 3]. Amundson?s equation has been applied to 
practical adsorption over porous adsorbents to describe the breakthrough 
characteristics [4]; however, only the diffusion of adsorbates in the packed bed was 
considered. Later, Yang included the adsorption isotherm, mass and heat balance in 
the interpellet gas phase and mass and heat balance inside each pellet in models for 
an adsorption bed [5]. His concept was widely used in prediction of adsorption over 
activated carbon fibers. For example, Cheng et al., predicted the breakthrough 
curves of adsorption of volatile organic
114 
 compound (VOC) on ACFs in fixed beds [6]. Some researchers combined the 
diffusion of adsorbates in the adsorbent bed, the diffusion of adsorbates within single 
pore and adsorption isotherm or reaction on the pore walls. For example, Gupta et al., 
analyzed the breakthrough characteristics of sulfur dioxide over zeolites considering 
mass balances and adsorption isotherm [7]. Das et al., predicted the adsorption 
procedure of volatile organic compounds over activated carbon fibers with the same 
concept of Gupta et al. [8]. The removal of sulfur dioxide by ACFs in the presence of 
oxygen and water was modeled with this concept including the mass balance in the 
packed bed reactor, the mass balance within pores and reactions occurred on fiber 
surfaces [9]. Adapa et al., employed the same concept with that of sulfur dioxide 
removal on the catalytic oxidation of nitrogen monoxide by ACFs [10]. Only the 
reaction over fiber surface converting sulfur dioxide to sulfuric acid was considered 
by Mochida et al., in analyzing the kinetics of continuous removal of sulfur oxide 
using polyacrylonitrile-based ACFs; diffusions of adsorbates were ignored [11]. Yoon 
and his colleagues proposed a theoretical model for respirator cartridge service life by 
addressing the adsorption and breakthrough of contaminant vapors with the 
assumption that the rate of change in the breakthrough concentration is proportional to 
breakthrough concentration and to the number of reactive sites at a specific moment 
and that the rate of decrease in the probability of adsorption is directly proportional to 
the contaminant concentration and flow rate and inversely proportional to the weight 
of activated carbon charcoal. The model was applied to toluene and methyl acetate 
and predicted breakthrough curves agreed with the experimental data well [12].  
115 
     In the present work, Yoon?s model was applied to ACF20 and KMnO4/ACF15, and 
the results were compared with the experimental results as shown in Figures 4.1 and 
4.2. The predicted data deviate from the experimental data and deviations are 
unacceptable. Program code for prediction of breakthrough curves is listed in 
Appendix 1.  
Figure 4.1 Prediction of breakthrough curve of ACF20 according to Yoon?s model 
Figure 4.2 Prediction of breakthrough curve of KMnO4/ACF15 according to Yoon?s model 
0 20 40 60 80 100 120 140 160 180 2000
10
20
30
40
50
60
70
80
90
100
Time (min)
Co
ut/
Ci
n (
%
)
Yoon's model
ACF20-14 in 90%RH air
k=0.036, tau=8.18
0 50 100 150 200 250 300 350 4000
10
20
30
40
50
60
70
80
90
100
Time (min)
Co
ut/
Ci
n (
%
)
Yoon's model
KMnO4-ACF in 90% nitrogen
k=0.0213, tau=154
116 
Most models available in the literature are for adsorption over bare adsorbents. In 
this research, mechanisms of removal of sulfur dioxide over potassium permanganate 
modified ACF15 (KMnO4/ACF15) are studied. As discussed in Chapter 3, the
removal efficiencies in different conditions are different and thus their mechanisms 
were regarded to be different. Four challenging gases are considered, SO2+N2, 
SO2+Air+N2, SO2+H2O and SO2+H2O+air. It should be noted that KMnO4/ACF15 in 
this Chapter refers to the sample which was obtained after 43wt% KMnO4 on ACF15 
was subjected to water, since it had the highest adsorption capacity. 
4.2 Mechanism one 
In the case of combination of SO2 and N2 as the challenging gas, the adsorption test of 
KMnO4/ACF15 was performed without the presence of either oxygen or water vapor.  
4.2.1 Mechanism study  
In order to understand the mechanism of SO2 adsorption over KMnO4/ACF15 without 
presence of oxygen and water in the carrier gas, the products of adsorption need to be 
identified. TGA and XRD techniques were used for this purpose (equipment and 
experimental conditions were included in Chapter 3). Using TGA, the decomposition 
of products on fiber surfaces was determined as shown in Figure 4.3. The TGA curve 
of KMnO4/ACF15 after the adsorption test with the challenging gas of sulfur dioxide 
carried by nitrogen exhibits a similar shape with the one before the test. The 
KMnO4/ACF15 after the adsorption test lost a small amount of weight at about 550?C 
117 
and 770?C. The weight loss at 550?C may be attributed to the decomposition of 
MnO2 whose decomposition temperature is 535?C. The fresh sample has a similar 
weight loss at 770?C with the spent sample. Therefore, MnO2 is postulated to be the 
only product of the reaction between sulfur dioxide and KMnO4. This result agrees 
with the theory that in the neutral and basic condition, sulfur dioxide is oxidized to 
MnO2 or K2MnO4. Since K2MnO4 is not stable, the product must be MnO2.  
In order to analyze the mechanism of adsorption of sulfur dioxide over 
KMnO4/ACF15, the breakthrough curves of pure ACF15, KMnO4/ACF15 without 
washing, KMnO4/ACF15 and pure potassium permanganate are compared in Figure 
4.4. Sulfur dioxide penetrated into the packed potassium permanganate (1gram) in the 
reactor immediately after starting the adsorption test without showing any adsorption 
After adsorption test Before adsorption test
546.56?C
0.05875%/?C 766.97?C0.05626%/?C
MnO2
0.0
0.1
0.2
0.3
0.4
0.5
De
riv
. W
eig
ht 
(%
/?C)
40
60
80
100
W
eig
ht 
(%
)
50 250 450 650
Temperature (?C) Universal V4.3A TA Instruments
Figure 4.3 TGA curves of KMnO4/ACF15 before and after adsorption tests in 50ppm 
SO2 carried by nitrogen 
118 
capacity. In comparison with KMnO4, the removal of SO2 over ACF15 is more 
efficient, which is indicated by the higher saturation capacity (90%) of ACF15 as 
shown in Table 4.1 and breakthrough curves in Figure 4.4. It can be included that the 
physical adsorption caused by pure activated carbon fibers is more than the chemical 
adsorption caused by pure potassium permanganate under this condition.  
After KMnO4 was distributed on activated carbon fiber surfaces, KMnO4/ACF15  
before washing shows a relatively higher SO2 saturation capacity with less sharp 
breakthrough curve, which may be attributed to the increase of contact surface 
between sulfur dioxide and KMnO4.  
0
20
40
60
80
100
0 10 20 30 40
C o
ut/
C in
(%
)
Time (min)
KMnO4
ACF15
KMnO4/ACF15-NW
KMnO4/ACF15
Figure 4.4 Breakthrough curves of ACF15, KMnO4/ACF15 before exposed to water, 
KMnO4/ACF15 and KMnO4 in the challenging gas of SO2+N2 (KMnO4 weight was 1 g 
and other samples were about 0.5 g.) 
119 
Table 4.1 90% Saturation capacities 
Sample ACF15 KMnO4* KMnO4/ACF15 -NW KMnO4/ACF15 
90% Saturation 
capacity 
 (mg SO2/g sample) 
2.927 0.086 6.832 37.15 
*Sample weight was 1 g for KMnO4 while other samples were about 0.5 g, initial sulfur 
dioxide concentration was 50 ppm and face velocity of the gas stream was 33 cm/s. NW 
refers to not washing.  
    Although KMnO4 does not show any SO2 capacity on its own, its capacity was 
greatly improved when it was distributed over ACFs, which probably was due to the 
increased contact area between KMnO4 and SO2. 90% saturation capacity of ACF15 
is lower than that of KMnO4/ACF15 that was not exposed to water even though 
breakthrough period of ACF15 is longer. Compared to the other three samples in 
Figure 4.4, KMnO4/ACF15 gave the longest breakthrough time and the highest 
saturation capacity. The great increase in SO2 capacity was probably either caused by 
the increased available adsorption sites on ACF15 regenerated by exposure to water, 
or by further increase of contact surface between KMnO4 and SO2, providing more 
chances for reaction. It is postulated that some of SO2 was physically adsorbed onto 
free adsorption sites on fibers and some others were transferred to other compounds 
by reactions. KMnO4 may react with either gas phase sulfur dioxide or adsorbed 
sulfur dioxide. We assumed that adsorbed sulfur dioxide was dominant in reacting 
with potassium permanganate. Thus, the mechanism of SO2 removal over 
KMnO4/ACF15 can be described by equations (4.1)~(4.3).  
v2v2 FSOFSO ??+                                                                                             (4.1)     
120 
v3v42v2v2v4 FSO2FSOKFMnO2FSO3FKMnO2 ?+?+???+?                         (4.2)  
v3v3 FSOFSO +??                                                                                              (4.3) 
Fv denotes free adsorption sites. X?Fv represents the substance that is attached or 
adsorbed on fibers. The resulting SO3 may desorb from fiber surfaces offering more 
vacant sites for incoming sulfur dioxide.   
As shown in Table 4.1, by loading 43% KMnO4 on ACF15, the 90% saturation 
capacity of ACF15 is increased by 133%; after exposure to water, the 90% saturation 
capacity was increased by 444%. In other words, the improvement of SO2 adsorption 
capacity caused by KMnO4 distribution on ACF15 surfaces is less than that caused by 
increased vacant sites. Therefore, equation (4.1) is assumed to be the rate-determining 
step.  
Sulfur dioxide adsorption rate can be expressed as follows.  
FvFvdFvSOa
FSO CkCkCCk
dt
dC
v ****)1(***
2]2[
]2[ ??? ???=?                           (4.4) 
where, ak  is the adsorption rate constant and dk  is the desorption rate constant for 
equation (4.1), ?  is the fraction of surface sites that are occupied by adsorbed 
molecules, 2k is the reaction rate constant of equation (4.2), and CFv is the total active 
sites on fiber sample.  
At the equilibrium, dtd vFSO ]2[ ? =0 
22
2
**
*
kkCk
Ck
dSOa
SOa
++=?                                                                                         (4.5)  
Therefore,  
121 
)1** *(***
**)1(***
2222
2
2
2
2
?+++++=
+??=
kkCk
k
kkCk
CkCCk
CkCCkdtdC
dSOa
d
dSOa
SOa
FvSOa
FvdFvSOa
SO ??
                        (4.6) 
                      
 4.2.2 Model development 
Some assumptions are made for model developments as follows.  
1) The pressure drop of the fixed bed is very small; therefore isobaric conditions 
are assumed throughout the tests.  
2) The system is isothermal throughout the fixed bed during the whole adsorption 
process. 
3) The fluid velocity throughout the bed is constant. 
4) The main gas stream and sulfur dioxide are considered as ideal gases. 
5) Pore diffusion and surface diffusion occurs simultaneously.  
6) The fixed bed is subjected to axial (longitudinal) dispersion, external mass 
transfer and intra-fiber diffusion of adsorbing species. 
The model development is based on three governing equations: 
1) mass balance of the adsorbing components in the bed. 
2) mass balance of the component inside a single activated carbon fiber. 
            3)  adsorption and reaction on the pore-walls of the fibers. 
4.2.2.1 The mass balance of the adsorbing components in the packed bed  
The mass balance of the adsorbing components in the packed bed is described as 
follows. 
122 
t
C
ft
C
z
Cu
z
CD s
Z ?
??=
?
?+
?
?+
?
?? 1
2
2
                                                                          (4.7)     
where, Dz is the axial diffusivity, C the sulfur dioxide concentration in the main gas 
stream, z is the distance from the beginning of the packed ACFs, u is the face velocity 
of the main gas stream, ???= 1f , ? is the porosity of the packed ACFs bed and sC is 
the average SO2 concentration within pores including gas phase and solid phase which 
is adsorbed onto fiber surfaces. tCs?? is determined on the basis of the steady 
boundary film theory as follows [13]. 
)(22 FSgfsf CCLkrtCLr ?=?? pipi                                                                               (4.8)      
where, FsC is the solid phase SO2 concentration on the fiber surface. 
Therefore, )(2 Fsg
f
s CCk
rt
C ?=
?
?                                                                                 (4.9)      
Combining (4.7) and (4.9), the following equation for expression of the mass 
balance of adsorbates in packed ACFs is obtained:  
)(212
2
F
sg
f
Z CCkrft
C
z
Cu
z
CD ??=
?
?+
?
?+
?
??                                                          (4.10)    
4.2.2.2 The mass balance within a single porous ACF In an individual ACF, the flux 
flowing into an infinitesimal unit volume with the thickness of dr is rCrLD re ??? pi2 , 
where r is the radius of the infinitesimal unit from the central line of the fiber. The 
flux flowing out of the unit is drr r
CrLD
r
CrLD
r
e
r
e ?
?
???
+???
)2(
2
pi
pi .  
123 
Therefore, the mass balance within a fiber is described as : 
t
CrLdr
t
CrLdrdr
r
r
CrLD
r
CrLD
r
CrLD rsr
r
e
r
e
r
e ?
??+
?
?=
?
?
???
+??????? pi?pi?
pi
pipi 2)1(2)
)2(
2(2
                                                                                                                            (4.11)            
The right side of equation (4.11) represents the sulfur dioxide accumulated in the 
unit both gas phase and solid phase on the fiber surface.  
Equation (4.11) can be rearranged as:   
t
Cr
t
Cr
r
Cr
rD
rsrr
e ?
??+
?
?=
?
?
?
? )1()( ??                                                                (4.12)       
where, De is the effective diffusivity, ? is the intrafiber porosity, Cr is the gas phase 
sulfur dioxide concentration within the fiber, and Crs is the solid phase sulfur dioxide 
concentration adsorbed on the fiber surface. To simplify the calculation process, Cr
and Crs are averaged within a single fiber, which reduces the second order partial 
differential equation (PDE) to first order PDE with only t direction. This approach has 
been used for adsorption on porous sorbents, such as Ca-sorbent, zeolites and 
activated carbon fibers, by other research groups as well [7, 9]. The average gas phase 
sulfur dioxide concentration within a fiber ( rC ) and the average of the solid phase 
adsorbed sulfur dioxide concentration on the fiber surface ( rsC ) are derived as follows.  
Integrating the both side of equation (4.12) with respect to r at the rage of (0, rf) 
where rf is the fiber radius,  
drtrCdrtrCdrrCrrDf f fr r r rsrre? ? ? ???+??=????
0 0 0
)()1()()( ??                                 (4.13)   
Rearranging equation (4.13),  
124 
)2()1()2()(2
0 20 22
drrCrtdrrCrtrCrDr ff
f
r
rs
f
r
r
f
rr
r
e
f
?? ???+??=?? = ??                    (4.14)     
rC  and rsC  are expressed as: 
2
0
2
0
22
f
r
r
f
r
r
r r
drrC
Lr
drrLC
C
ff ??
== pi
pi
                                                                            (4.15)   
2
0
2
0
22
r
drrC
Lr
drrLC
C
ff r
rs
f
r
rs
rs
?? ==
pi
pi
                                                                         (4.16)    
Substituting (4.15) and (4.16) into equation (4.14), the mass balance within a fiber is 
obtained: 
)()1()()(2 rsrrrre
f
CtCtrCDr
f ?
??+
?
?=
?
?
= ??                                                         (4.17)   
Equations (4.10) and (4.17) are coupled by continuity at the fiber surface [14]: 
( )Fsg
r
r CCk
r
CDe
f
?=?
?
??
?
?
?
?                                                                                      (4.18)                  
g1829g3020g3007 is determined by equating mass flow across the external gas film to that on fiber 
surfaces and is dependent on variables rC and C [15]. 
b
dDCCkC fergF
S
5+=
                                                                                        (4.19) 
where 
f
e
g d
Dkb 5+= and k
g is the mass transfer coefficient and df is the fiber diameter.                                          
Similar to Equation (4.6),  
)1** *(***
22
2 ?
+++++=?
?=
?
?
kkCk
k
kkCk
CkCCk
t
C
t
C
dra
d
dra
ra
Fvra
SOrs           (4.20) 
The governing equations (4.10), (4.17) and (4.20) describe the mass balance of 
adsorbates in           t he packed bed,  t he mass  balance within a fiber and adsorption and 
reaction on fiber surface, respectively. They are the basis of the model development 
125 
for predicting the ACFs performance in the reactor under dynamic conditions. These 
equations are coupled partial differential equations with time (t), location of 
adsorbents (z) and radial direction (r) as independent variables.  Initial and boundary 
conditions are  
g1872 g3404 g882g482g1829 g3404 g882g481 g1829g3045g3364g3364g3364=0 
        g1872 g3408 g882g481 g1878 g3404 g882g481g1829 g3404 g1829g3036g3041   
        g1878 g3404 g1864g482g3g3105g3004g3293g3364g3364g3364g3105g3053 g3404 g882
By solving them, the time is related to the sulfur dioxide concentration on the 
main stream and breakthrough profile can be predicted. The dependency of sulfur 
dioxide removal efficiency on time and location can be estimated as well.  
4.2.2.3 Derivation of equations for the modeling code with the finite difference 
methods The equations in the model are non-linear partial differential equations which 
usually require some numerical methods. A finite-difference method was used and the 
explicit central difference and backward difference themes were employed.  
The mass balance between the main gas stream and the ACFs in the packed bed is 
given by Equation (4.10) 
By substituting equation (4.20) into (4.10), the mass transfer in the packed bed is 
obtained: 
)/5(212
2
b
dDC
b
CkCk
rfz
Cu
z
CD
t
C ferg
g
f
Z ????
??
?
?=
?
?                                     (4.21)                                              
For the prediction of the breakthrough curves, the fixed bed is divided into M 
equal sections along the reactor axial direction. The time direction is divided into N 
126 
segments. The corresponding step sizes for each direction are denoted as k and h, 
respectively.  
According to the explicit finite difference themes [16],  
22
2 )1,1(),1(2)1,1(
h
mnCmnCmnC
z
C ??+??+?=
?
?                (4.22)   
h
mnCmnC
z
C )1,1(),1( ????=
?
?                                                                          (4.23)    
k
mnCmnC
t
C ),1(),( ??=
?
?                                                                                    (4.24)   
where the indices m and n refer to the spatial location and time in the sample bed, 
respectively. Assuming that the sample bed is divided into m intervals with M+1 grip 
points with m=1 at Z=0 and m=M+1 at the exit of the gas which is equal to the bed 
depth.  
    Substituting equations (4.22), (4.23) and (4.24) into equation (4.12), the following 
discretised equations for diffusion within the sample bed can be obtained:  
),1(*4)1,1(*3)1,1(*2),1(*1),( mnFpmnCpmnCpmnCpmnC ?+??++?+?=     
                                                             g884 g3409 g1865 g3409 g1839                                              (4.25) 
C(n,M+1)=C(n,M) 
where, )1(* **2***211 2 bkrf kkh kuh kDp g
f
gz ????=
2
*2
h
kDp z=
h
ku
h
kDp z **3
2 +=
ff
eg
dbrf
kDkp
***
***104 =
127 
The intrafiber mass balance is expressed by combining equations (4.17), (4.18) 
and (4.20).  
2
2
*
****1
***
***10)1(
*
**2
kkCk
CCkk
bdr
CDk
b
k
r
Ck
t
C
dra
Fvra
ff
regg
f
gr
++
?+??=
?
?
?
?
??
ra
r
da
a
rad
rda
r
ff
geg
f
gr
Ck
Ckk kCkk CkkCdrb kDCbkrktC
**1
**1)*(* ***)1(**** **10*)1(***2
2
?
?
?
?
?
?
??
??
+
?+
+
?+??=
?
?
                (4.26)    
k
mnFmnF
t
Cr ),1(),( ??=
?
?
                                                                                                                        (4.27)   
Therefore, the following discretized equations for the intrafiber diffusion can be 
obtained.  
)1(* ),1(***2
),1(*
*),1(****1
***
),1(****10),1(),(
2
2
b
k
r
mnCkk
kkmnFk
CmnFkkk
bdr
mnFkDkmnFmnF
g
f
g
da
Fva
ff
eg
??+
++?
??+???=
?
?
?
?
                                                                                                                                (4.28) 
2),1(*
),1(*7),1(*6),1(*5),1(),(
kkmnFk
mnFpmnFpmnCpmnFmnF
da ++?
?+???+?=
where, )1(* **25 bkr kkp g
f
g ?=
?
bdr
kDekgp
ff ***
***106
?=
kCkkp Fva ****17 2?? ?=
128 
Equations (4.25) and (4.28) are the two main equations for the modeling code. 
4.2.2.4 Parameter evaluations  
(1) Effective diffusivity ( eD ) 
The actual diffusion path of the adsorbate within a porous activated carbon fiber 
is not equal to the distance in the radial direction, but will be a quite tortuous path, 
which depends on the pore structure. Taking into account that the volume occupied by 
the solid is not available for diffusion, the value of diffusivity (D) is usually converted 
to the effective diffusivity (De) for the fact that diffusion only occurs in the pore space 
[14].   
?
?DD
e =                                                                                                                (4.29)         
where, ? is the tortuosity factor of pores in porous fibers. Tortuosity is essentially a 
geometric factor which is independent of either temperature or the nature of the 
diffusion species. A tortuosity factor is defined as the ratio between the actual 
diffusion path length and the net distance in the direction of flux or the radial distance.    
    Diffusion in gas phase is determined by pore diameter and the mean free path of 
molecules. Knudsen diffusion is assumed if the mass free path is greater than ten pore 
diameters and collisions between pore walls and molecules are important. On the 
contrary, if pore diameter is ten times greater than the mean free path, then collisions 
between molecules dominate, which is called molecular diffusion. The two 
mechanisms for diffusion are both important when the mean free path and pore 
diameter are of the same order of magnitude. The diffusivity is calculated as follows:  
129 
km DD
D 11 1
+
=                                                                                                   (4.30)   
where Dm is molecular diffusivity and Dk is Kudsen diffusivity.    
The mean free path 22 1pi?? n= , where, n is the gas number density in 
molecules per volume, and ? is collision diameter in the experiments. Since over 95% 
air is present in the main gas stream, the mean free path is estimated to be 0.2 ?m 
which is the mean free path for air at 1 atm and 300 K. Pore diameter of ACFs 
included in this research is in the range of 2.4~2.7 nm, which is 80 times smaller than 
the mean free path (0.2 ?m). Therefore, Knudsen diffusion dominates. 
5.03 )(**10*7.9
M
TrDD
fk ==                                                                              (4.31) 
where, rp (cm) is the pore radius, T is temperature (K), M is molecular weight of gas 
(g/mol), and Dk is Knudsen diffusivity (cm2/s).  
The molecular weight of gas depends on the experimental conditions, such as the 
relative humidity, carrier gas (N2 or air) and sulfur dioxide concentration. At 20
, the 
water vapor volume fraction in the air is 2.0851% at 90% relative humidity and 1.62% 
at 70% relative humidity.  
The tortuosity factor of activated carbon fibers was reported to be smaller than 
that of granular activated carbon. Leyva-Romos et al., found the tortuosity factor of 
activated carbon fibers to be less than 2.7 [17]. ? is estimated to be 2.6 in his study. 
The intrafiber porosity is calculated using the following relationship: 
ifierfibertpore
ifiercarbon
VVV
VV
mod
mod1
++
+?=?                                                                                 (4.32)     
130 
where, Vcarbon is the volume occupied by the solid ACF, Vmodifier is the volume of the 
modification agents (such as potassium hydroxide and potassium permanganate) and 
Vtpore is the total pore volume within the fiber. If the modifier load is < with respect to 
the total sample weight, the intrafiber porosity can be obtained as follows:  
ifiercarbon
tpore
ifiercarbon
V
mod
mod
1
1
1
?
?
?
?
?
?
?
?
?
+?+
+?
?=                                                                            (4.33) 
where, ?carbon is the solid density of ACF, and ?moidifier is the density of the 
modification agents. The solid density of ACF is reported to be 2.16 g/cm3 which was 
determined by the helium displacement method using a Helium pycnometer [17]. 
Densities of potassium hydroxide and potassium permanganate are 2.044 and 2.703 
g/cm3, respectively.  
A simple approximate estimation of tortuosity is given by Lua et al. [18] as follows.  
g2028 g3404 g4666g884 g3398 g2009g4667
g2870
g2009
    For KMnO4/ACF15, the calculated tortuosity is 2.6873 with the intrafiber porosity 
of 0.6641, which agrees with Leyva-Ramos? report [17].  
(2) Axial diffusivity (Dz)  
The axial diffusivity is given by [14]: 
?/)5.020( ecmZ RSDD +=                                                                                    (4.34)        
where, Sc is the Schmidt number and Re is the Reynolds number, which are 
determined by the following equations:  
DS gc ?
?=                                                                                                               (4.35)        
131 
?
? reg
e
duR ??=                                                                                                       (4.36)       
where, ? is the viscosity of the fluid, ?g is the density of the fluid and dre is the 
diameter of the reactor. The viscosity of air at 18?C (291 K) is 18.27?10-6 sPa ? . The 
fluid density can be obtained by TRPM f , where P is pressure (Pa), Mf is the average 
molecular weight of the fluid, T is the temperature and R is the ideal gas constant.  
The molecular diffusivity for a binary gas mixture can be evaluated by the 
Chapman-Enskog equation [14]: 
ABAB
2/1
BA
2/3
m P
)M/1M/1(T0018583.0D
??
+?=                                                          (4.37)        
where, MA and MB are molecular weights of A (sulfur dioxide) and B (air or nitrogen), 
respectively, Dm is molecular diffusivity (cm2/s), P is pressure (atm), ?AB and ?AB are 
constants in the Lennard-Jones potential energy function for the pair AB. ?AB is the 
collision diameter in angstrom and ?AB is the maximum attractive energy between two 
molecules. ?AB is the collision integral, a function of kBT/?AB, where kB is 
Boltzmann?s constant.  
If the carrier gas is air [19],  
8215.3)617.3026.4(21)(21)(21 air2SOBAAB =+=+=+= ????? ?  
where, ?SO2 is sulfur dioxide molecular diameter and ?air is the average molecular 
diameter of air. 
K65.18797363kkkkk
B
air
B
2SO
B
B
B
A
B
AB =?=?=?= ?????
132 
56.165.187293kTTk
BABAB
B ===
??  if the temperature is 20
.  
?AB is 1.183 if the carrier gas is air and temperature is 20
.  
The porosity of the packed ACF bed is calculated with the equation:  
Lr
/m1
V
V1
2
r
fiberfiber
bed
fiber
pi
?? ?=?=                                                                              (4.38)     
where mfiber is the fiber mass used in the adsorption test, ?fiber is the fiber apparent 
density, rr is the radius of the reactor and L is the packed bed depth.  
(3) The mass transfer coefficient, kg  
For 3<Re<104, DdkRS1.12S fg6.0e3/1ch =+= , where df is the fiber diameter [20]. 
Therefore, the mass transfer coefficient is determined by  
Dd RS1.12d DSk
f
6.0
e
3/1
c
f
h
g ?
?+=?=                                                                   (4.39)    
4.2.3 Prediction of breakthrough curve with the developed model  
As discussed earlier, the adsorption of sulfur dioxide carried by nitrogen over 
KMnO4/ACF15 consists of physical adsorption and chemical adsorption. The 
breakthrough curve is predicted using the model developed above and the resulting 
predicted curve is shown in Figure 4.5. The predicted breakthrough curve agrees with 
the experimental result reasonably well. The Matlab? code for model prediction is 
listed in Appendix 2. 
133 
Figure 4.6 Prediction of sulfur dioxide removal efficiency with the variation of 
location and time  
Figure 4.5 Breakthrough curve of KMnO4/ACF15 exposed to 50ppm sulfur dioxide 
carried by nitrogen    
134 
With the developed method, SO2 removal efficiency profile within the reactor 
during the adsorption can be predicted as well, which is shown in Figure 4.6. In the 
packed fiber depth direction, fibers close to gas entrance got saturated faster than 
those closer to gas exit. The fibers in the reactor were saturated at the end of the test 
after a certain period of reaction time.  
4.3 Mechanism Two  
During the sulfur dioxide adsorption tests, if oxygen is included in the main gas 
stream to feed the reactor, sulfur dioxide removal mechanism is expected to be 
different from the first mechanism, in which oxygen was not included.  
The breakthrough profiles of KMnO4/ACF15 with and without the presence of 
oxygen under very low relative humidity (4~5%) are shown in Figure 4.7. For both air 
and nitrogen as the carrier gas, the breakthrough curves are consistent. It can be 
concluded that the presence of oxygen in the main gas stream does not affect sulfur 
dioxide breakthrough profiles and therefore does not contribute to improve the sulfur 
dioxide removal.  
Figure 4.7 Breakthrough profiles of KMnO
gas carried by air and nitrogen
KMnO4/ACF15 samples before and after adsorption tests were subj
to observe the reaction products of sulfur dioxide 
weight on temperature in nitrogen atmosphere is sho
adsorption test, potassium permanganate 
Potassium permanganate may work as the activating ag
carbon fibers, which cause
When the adsorption test was completed, a part of p
consumed to oxidize sulfur dioxide; the reaction pr
permanganate would be present on fiber surfaces sim
135 
4/ACF15 with exposure to sulfur dioxide challenge
removal. The dependence of sample 
wn in Figu
was distributed on fiber surfaces 
ent and reactivate the activated 
s continuous loss of weight with increa
otassium permanganate 
oducts and unconsumed potassium 
ultaneously. 
ected to TGA 
re 4.8. Before the 
uniformly. 
sed temperatures. 
has been 
136 
As shown in TGA curves, KMnO4/ACF15s gave similar trend before and after 
SO2 adsorption tests, which demonstrates that potassium permanganate is present after 
adsorption. However, after adsorbing SO2, KMnO4/ACF15 is showing one quick 
weight loss between 500 and 600?C, indicating the presence of manganese dioxide. It 
is worthy to note that the TGA pattern of spent KMnO4/ACF15 in dry air is similar to 
the one when dry nitrogen was the carrier gas in the main gas stream which was 
explained in Mechanism One. Theoretically, the oxygen in air will oxidize sulfur 
dioxide to sulfur trioxide (SO3) especially with the help of a catalyst (activated 
carbon), and both of them will be adsorbed onto the catalyst. However, desorption of 
adsorbed sulfur trioxide is not detected by TGA probably due to low amount of SO3 or 
gradual rather than sudden weight loss with the change of temperature.  
0.0
0.8
De
riv
. W
eig
ht 
(%
/?C
)
0
20
40
60
80
100
W
eig
ht 
(%
)
50 250 450 650 850
Temperature (?C)
                  After adsorption test in dry air-20.001???????
                  After adsorption test in dry introgen-50.001? ? ? ?
                  Before adsorption test-38.001????? ?
Universal V4.3A TA Instruments
Figure 4.8 Weight change of KMnO4/ACF15 before and after adsorption tests in the 
challenging gas of SO2 carried by air with increasing temperatures 
137 
 The SO2 adsorption over KMnO4/ACF15 in dry air may include the diffusion of 
air to the fiber surfaces, diffusion of the gas into pores within fibers, physical 
adsorption of SO2 onto vacant sites of fibers, oxidation of SO2 by oxygen and 
formation of SO3 which stays on the fiber surfaces and oxidation of SO2 by the 
oxidant, potassium permanganate. This proposed process is expressed with the 
following equations:  
v2v2 FSOFSO ??+                                                                                              (4.1)      
v32v2 FSOO2
1FSO ??+?                                                                                    (4.40)   
v3v42v2v2v4 FSO2FSOKFMnO2FSO3FKMnO2 ?+?+???+?                          (4.2) 
v3v3 FSOFSO +??                                                                                              (4.3) 
    KMnO4 is excessive for the reaction (4.2), therefore its reaction rate constant is 
determined by the concentration of solid SO2 on fiber surfaces.  
Compared to the Mechanism One, one more reaction may occur involving oxygen 
in air. However, both TGA results and breakthrough curves did not show much 
change from the first mechanism. Oxygen was assumed not contributing too much in 
sulfur dioxide removal. Therefore, the model developed for the first mechanism can 
be used for this mechanism to predict the breakthrough properties of KMnO4/ACF15 
in the challenging gas of sulfur dioxide carried by air. The predicted breakthrough 
profile agrees with the experimental results reasonably well as shown in Figure 4.9.  
138 
   The sulfur dioxide removal efficiency in the reactor dependens on both time and 
distance as discussed in Mechanism One. The developed model can predict the sulfur 
dioxide removal efficiency at different locations and reaction times as shown in 
Figure 4.10.   
Figure 4.9 Prediction of breakthrough curve of KMnO4/ACF15 with the 
challenging gas of SO2+ air 
139 
4.4 Mechanism Three 
4.4.1 Mechanism analysis  
 In Mechanism Three, the sulfur dioxide removal capability of KMnO4/ACF15 was 
analyzed with the presence of water vapor and absence of oxygen in the main gas 
stream. There are two aspects of sulfur dioxide removal by KMnO4/ACF15 in this 
case: direct adsorption of sulfur dioxide onto the vacant active sites on fiber surfaces 
and oxidation of sulfur dioxide by potassium permanganate. However, the presence of 
water will further convert the sulfur dioxide to sulfuric acid which is similar to the 
adsorption of sulfur dioxide with bare ACFs as discussed in Chapter 3. Like 
Figure 4.10 Prediction of sulfur dioxide removal efficiency at 
different locations and times with the challenging gas of SO2+ air 
140 
Mechanisms One and Two, KMnO4/ACF15 before and after adsorption tests were 
subjected to increasing temperatures and weight changes are illustrated in Figure 4.11. 
Before the SO2 adsorption test, KMnO4/ACF15 lost its weight gradually with 
increasing temperatures. However, after exposure to SO2 carried by humidified 
nitrogen until the sample got saturated, KMnO4/ACF15 showed a sudden weight loss 
at about 550?C, corresponding to decomposition of MnO2, which is similar to the 
previous two mechanisms. XRD results confirmed the product of MnO2 which was
discussed in Chapter 3. 
The gradual weight loss of KMnO4/ACF15 under increased temperatures probably 
is attributed to reactivation of activated carbon fibers, which may generate more pores 
or expand current pores to a bigger size. In order to confirm this proposal, 
KMnO4/ACF15 and bare ACF15 were heated to 800?C and held at 800?C for 1 hour. 
The weight changes were recorded as shown in Figure 4.12. As the temperature was 
increased to 800?C from room temperature, both ACF15 and KMnO4/ACF15 lost 
weight continuously while the latter exhibited quicker weight loss. At 800?C, the 
weight of ACFs remained constant but the weight of KMnO4/ACF15 continued 
dropping which was due to the activation caused by the activating agent, potassium 
permanganate. Therefore, it is possible that more pores were formed or the existing 
pores were expanded.  
  
141 
  
Figure 4.11 Weight changes of ACF15, KMnO4/ACF15 before and after SO2 adsorption 
test with the presence of water in the carrier gas with increasing temperatures. 
20
40
60
80
100
W
eig
ht 
(%
)
50 250 450 650
Temperature (?C)
                  AFC15???????
                  KMnO4-ACF15 after SO2 adsorption? ? ? ?
                  KMnO4-ACF15 before SO2 adsorption????? ?
Universal V4.3A TA Instruments
Figure 4.12 Effect of temperature on weight of ACF15 and 
KMnO4/ACF15  
. 
142 
    Potassium permanganate may react with either gas phase or adsorbed SO2. In order 
to determine the dominant process, pure solid potassium permanganate was packed in 
the reactor for sulfur dioxide removal under different conditions: dry air, humidified 
nitrogen and humidified air. In this case, there were no adsorption sites for sulfur 
dioxide to be adsorbed and only gas phase sulfur dioxide was available. 1g KMnO4
rather than 0.5g was used because 0.5g KMnO4 could not cover the cross section of 
the reactor and thus sulfur dioxide would escape from the reactor skipping KMnO4
samples. The resulting breakthrough profiles are shown in Figure 4.13 Without the 
presence of water vapor, sulfur dioxide penetrated the potassium permanganate 
packed bed immediately as discussed in Mechanism One and thus it barely reacted 
with gas phase sulfur dioxide. When the water was included, sulfur dioxide removal 
efficiency of KMnO4 was promoted greatly. Oxygen also contributed to the reaction 
as indicated by less steep breakthrough curve for KMnO4 in the humidified air. It can 
be concluded that gas phase sulfur dioxide can react with potassium permanganate 
with the presence of water in the main gas stream. In order to investigate the 
contribution the activated carbon fibers, the breakthrough profile of KMnO4/ACF15 is 
compared with that of the pure KMnO4 with the same challenging gas: SO2+H2O+N2, 
as shown in Figure 4.14. In spite of more KMnO4 (1g) in the reactor, sulfur dioxide 
breakthrough occurs much quicker than that of KMnO4/ACF15 (0.5g). The sulfur 
dioxide 4% breakthrough capacities in humidified nitrogen were calculated and are 
shown in Table 4.2. Breakthrough capacity of KMnO4 was increased from 19.23 to 
80.89 mgSO2/g sample after distributing on fiber surfaces. Activated carbon fibers 
143 
greatly improved the sulfur dioxide removal ability due to the increase of contact 
surface between KMnO4 and SO2. Therefore, sulfur dioxide had more chance to be 
oxidized and reaction was accelerated, resulting in a longer breakthrough time as 
shown in Figure 4.15. Therefore, the dominant reaction can be assumed to happen 
between gas phase SO2 and KMnO4 on fiber surfaces. 
0
20
40
60
80
100
0 100 200 300 400 500
C o
ut/
C in
(%
)
Time (min)
SO2+H2O+air
SO2+H2O+N2
SO2+air
Figure 4.13 Breakthrough profiles of KMnO4 with different challenging gases 
Figure 4.14 Breakthrough profiles of KMnO4/ACF15 in comparison with 
KMnO4 in the gas stream of sulfur dioxide and humidified nitrogen 
144 
Table 4.2 4% breakthrough capacities 
    Based on the analysis above, the sulfur dioxide removal process is described as 
follows. 
v2v2 FSOFSO ??+                                                                                                  (4.1)
v42v42v22v2v4 FSOH2FSOKFMnO2OH2FSO3FKMnO2 ?+?+??+?+?                                             
                                                                                                                                  (4.41)
v42v42v222v4 FSOH2FSOKFMnO2OH2SO3FKMnO2 ?+?+??++?                (4.42)                                                   
v422v42 FSOHOHFSOH +?+?                                                                            (4.43)                                                   
4.4.2 Model development 
Mochida et al., applied power law to describe the reaction and adsorption of sulfur 
dioxide over polyacrilonitrile-based ACFs and reasonable agreement between 
prediction and experiments was obtained [11]. The power-law model is used in this 
study to fit one apparent rate constant rather than using separate rate constants for 
each step. Equation (4.44) demonstrates this model:  
c
]O2H[
b
]2SO[
a
]Fv4KMnO[
2SO C*C*C*k
dt
d
?=?                                                          (4.44) 
    If only 90% relative humidity is considered, water concentration is constant and 
Samples 
4% breakthrough capacity 
(SO2 mg/g sample) 
4% breakthrough capacity 
(SO2 mg/g KMnO4) 
KMnO4/ACF15 in humidified nitrogen 80.8992 109 
KMnO4 in  humidified nitrogen   19.23 - 
KMnO4 in  humidified air  130.1975 - 
145 
thus Equation (4.44) can be rewritten as: 
b
]2SO[
a
]Fv4KMnO[
2SO C*C'*k
dt
d
?=?                                                                          (4.45) 
where g1863g4593 g3404 g1863 g1499 c]O2H[C
Since the consumed KMnO4 is V*C*32 ]2SO[  and the unreacted KMnO4 is 
( V*C*32ttotalamoun ]2SO[? ), which is rewritten as follows: 
tFRChtsampleweigVChtsampleweig SOSO ***10*7.6*10*72.2**32158/4337.0* ]2[43]2[ ?? ?=?
where, FR is the flow rate of gas stream (mL/min) and t is elapsed time in minute .  
Therefore, 
a
]2SO[
43b
]2SO[
b
]2SO[
a
]Fv4KMnO[
]2SO[
)t*FR*C*10*7.6htsampleweig*10*72.2(*C"*k
CC'kdtdC
??
?
?=
=?                         
                                                                                                                             (4.46) 
    The mass balance of adsorbate diffusion within the packed bed and the mass 
balance of intrafiber diffusion within a fiber are expressed by equation (4.10) and 
(4.17), respectively.  They are coupled by continuity at the fiber surface through 
equation (4.18) [14]:  
     Equations (4.10), (4.17) and (4.46) are combined to develop model for this 
mechanism. The gas phase sulfur dioxide concentration in the packed bed is related to 
time and location of the samples with the equation:  
)b d/DC5bCkC(kr2f1zCuzCDtC fergg
f
2
2
Z ????
??
?
?=
?
?                                     (4.47) 
    The average sulfur dioxide concentration within fibers is related to time as follows. 
146 
ar43b
r
r
ff
egg
f
gr
)t*FR*C*10*7.6mass*10*72.2(*C"*k*1
C*b*d*r* D*k*10C*)bk1(*r*k*2tC
?? ??+
??=??
?
?
??                             
                                                                                                                          (4.48)   
The same analysis process with Mechanism One is used to obtain the expressions 
for iteration in the computer program. Gas phase sulfur dioxide concentration in the 
main gas stream in the packed bed is:  
)m,1n(F*4p)1m,1n(C*3p)1m,1n(C*2p)m,1n(C*1p)m,n(C ?+??++?+?=
                                                                                                                                  (4.49)
where, )bk1(r*f k*k*2h k*uh k*D*211p g
f
g
2
z ????=
2
z
h
k*D2p =
h
k*u
h
k*D3p
2
z +=
ff
eg
d*b*r*f
k*D*k*104p =
The average sulfur dioxide concentration within fibers is: 
                                                   
                                                                                                                                (4.50) 
where, g1832g4666g1866g481g1865g4667 denotes g1829g3045g3364g3364g3364 in equation (4.48) 
       )bk1(*r* k*2*k5p g
f
g ?=
?
       b*d*r* D*k*10*k16p
ff
eg
??=
       "k*k*17p ?? ?=
a43b )t*FR*)m,1n(F*10*7.6mass*10*72.2(*)m,1n(F*7p
)m,1n(F*6p)m,1n(C*5p)m,n(F
???+
?+?=
??
147 
The iterations of g1829g4666g1866g481g1865g4667andg3g1832g4666g1866g481g1865g4667 are programmed with Matlab?; the code is 
attached in Appendix 3.  
    The breakthrough profile of KMnO4/ACF15 in humidified nitrogen is predicted 
with this model, which agrees with the experimental results reasonably as shown in 
Figure 4.15. Furthermore, the dependence of sulfur dioxide removal efficiency in the 
packed bed on time and location is estimated based on the developed model as shown 
in Figure 4.15.  At different depths of the bed, the time when the material got 
saturated was different. It took longer time for fibers which were close to the end of 
the bed to get saturated with sulfur dioxide. Eventually, the whole bed was saturated 
which is denoted with the color of dark red.  
0 50 100 150 200 250 300 350 400 450 5000
10
20
30
40
50
60
70
80
90
100
Time (min)
Co
ut/
Ci
n (
%
)
Predicted-313
Experimental-313
Figure 4.15 Prediction of breakthrough profile of KMnO4/ACF15 in humidified nitrogen 
148 
4.5 Mechanism Four  
In this mechanism, the challenging gas for adsorption tests consists of sulfur dioxide, 
water and air. To investigate the influence of water on adsorption properties of 
KMnO4/ACF15, breakthrough curves with the humidified nitrogen and humidified air 
as the carrier gases are shown in Figure 4.17. It took longer time to breakthrough if 
humidified air is the carrier gas although the period to get saturated was close for both 
carrier gases. The saturation capacity for the air carried challenging gas is higher than 
that carried by humidified nitrogen. In other words, oxygen in air promoted the sulfur 
dioxide removal when the water was present. In the co-presence of both water and 
oxygen, the sulfur dioxide removal mechanism can be divided into two components. 
Figure 4.16 Dependence of sulfur dioxide removal efficiency on time and 
location in the packed bed with the carrier gas of humidified nitrogen 
149 
One is the oxidation of sulfur dioxide by potassium permanganate and the other one is 
the conversion of sulfur dioxide to sulfur trioxide by oxygen in air. These two 
oxidations occur simultaneously in the reactor. 
    The oxidative removal of sulfur dioxide over KMnO4/ACF15 was discussed in 
Chapter 3. The water hydrated the sulfur trioxide to sulfuric acid. The high relative 
humidity, 90%, provided additional water to elute the aqueous sulfuric acid from fiber 
surfaces and thus more vacant active sites were released for further reaction. This 
oxidation process was described with equations (4.1), (4.40), (4.49) and (4.13). 
Reaction products were identified using TGA and XRD. KMnO4/ACF15 before and 
after adsorption tests were subjected to increasing temperatures on TGA and X-ray 
scanning, as explained in Chapter 3. The TGA analysis results are shown in Figure 
4.18. After sulfur dioxide adsorption tests in both presence and absence of oxygen, 
0
20
40
60
80
100
0 100 200 300 400
C o
ut/
C in
(%
)
Time (min)
Cin=50ppm
Face velocity=33cm/s 
Sample=0.5g
Humidified air
Humidified nitrogen
Figure 4.17 Breakthrough profiles of KMnO4/ACF15 with exposure to sulfur 
dioxide carried by humidified nitrogen and humidified air  
150 
their TGA curves gave similar patterns and thus the oxidation products of sulfur 
dioxide by KMnO4 are assumed to be the same. As analyzed in the section above, 
KMnO4 might be reduced to MnO2 and MnSO4. Weight losses of spent 
KMnO4/ACF15 at around 550?C and 850?C correspond to decomposition of MnO2
and MnSO4, respectively. The XRD analysis in Chapter 3 confirmed the existence of 
MnO2 on the spent KMnO4/ACF15 but MnSO4 was not detected by XRD probably
due to the small amount. Both adsorbed SO2 and gas phase SO2 can react with 
potassium permanganate as discussed in Mechanism Three. Sulfuric acid produced 
during the oxidation reactions caused by oxygen and potassium permanganate 
oxidation occupies vacant sites of activated carbon fibers. The additional water 
accelerates the elution of aqueous sulfuric acid from fibers and release active sites for 
further oxidation of coming sulfur dioxide.  Therefore, the mechanism of SO2 removal 
over KMnO4/ACF15 with the presence of oxygen and water is described with 
following equations.  
v2v2 FSOFSO ??+                                                                                                 (4.1) 
v32v2 FSOOFSO ??+?                                                                                         (4.40) 
v422v3 FSOHOHFSO ??+?                                                                                   (4.49) 
v42v42v42v2v4 FSOH2FSOKFMnSO2OH2FSO5FKMnO2 ?+?+??+?+?        (4.41) 
v42v42v422v4 FSOH2FSOKFMnSO2OH2SO5FKMnO2 ?+?+??++?             (4.42) 
v42v42v22v2v4 FSOH2FSOKFMnO2OH2FSO3FKMnO2 ?+?+??+?+?       (4.50) 
v42v42v222v4 FSOH2FSOKFMnO2OH2SO3FKMnO2 ?+?+??++?                (4.51) 
v422v42 FSOHOHFSOH +?+?                                                                           (4.43) 
151 
    Compared to the third mechanism, the presence of oxygen in this mechanism 
complicates the reaction involving oxidation by both oxygen and potassium 
permanganate. It is hard to determine the rate-determining step and dominant reaction. 
Therefore, power law is employed to describe the reaction kinetics as follows 
d
]2O[
c
]O2H[
b
]2SO[
a
]Fv4KMnO[
2SO CC*C*m*k
dt
d
?=?                                              (4.52)
    In the experiments, oxygen was carried in air and the variation of its concentration 
is small and thus neglected. Therefore, the oxygen concentration is assumed to be 
0.0
0.2
0.4
0.6
0.8
De
riv
. W
eig
ht 
(%
/?C
)
20
40
60
80
100
W
eig
ht 
(%
)
50 250 450 650 850
Temperature (?C)
                  After SO2 Adsorption in 90%RH air.001???????
                  After SO2 Adsorption in 90%RH nitrogen.001? ? ? ?
                  KMnO4-ACF15 before SO2 adsorption.001????? ?
Universal V4.3A TA Instruments
Figure 4.18 TGA curves of KMnO4/ACF15 before and after exposure to 
the challenge gas of SO2 +H2O+air and SO2 +H2O+ N2
152 
constant. Moreover, the relative humidity remained at 90%. Thus, water and oxygen 
concentrations can be combined with the apparent rate constant, k. The following 
equation is obtained: 
b
]2SO[
a
]Fv4KMnO[
2SO C*m'*k
dt
d
?=?                                                                         (4.53)
This reaction kinetic expression is the same with Mechanism Three. Therefore, the 
model developed for Mechanism Three is applicable for this mechanism. The 
predicted breakthrough profile using the developed model is shown in Figure 4.18 in 
comparison with the experimental results. The predicted sulfur dioxide removal 
efficiency agrees with the experimental results very well. Since the gas stream is 
flowing through the reactor continuously, the sulfur dioxide removal is in unsteady 
state. The dependence of sulfur dioxide removal efficiency on time is different at 
different locations of the packed bed in the reactor and is expressed with the curve 
shown in Figure 4.19.  
0 50 100 150 200 250 300 350 400 450 5000
10
20
30
40
50
60
70
80
90
100
Time (min)
Co
ut/
Ci
n (
%
)
Predicted
Experimental
Figure 4.19 Prediction of breakthrough profile of KMnO4/ACF15 exposed 
to SO2 +H2O+air    
153 
4.6 Model system 
Mechanisms One and Two can be predicted with the power law as well and therefore 
all four mechanisms are eligible for the power law model. The corresponding 
conpoents (a and b) and the apparent reaction rate constant (k?) can be located in 
different ranges for each mechanism. These ranges and the best parameter set are 
listed in Table 4.3. The predicted breakthrough profiles of Mechanisms One and Two 
using the power-law model are compared with the experimental data and the models 
agree with the experimental data reasonably well as shown in Figuers 4.21 and 4.22. 
Figure 4.20 SO2 removal efficiency of KMnO4/ACF15 at various 
times and locations with the challenging gas of SO2 +H2O+air 
154 
Table 4.3 Parameters in the model system for mechanisms 1~4 
 a b k? 
Mechanism One 1.5~1.6 -0.4~-0.55 10~30 
Best 1.55 -0.5 20 
Mechanism Two 1.75~1.9 -0.55~-0.7 25~50 
Best 1.8 -0.6 35 
Mechanism Three 1.2~1.9 -0.01~-0.7 70~120 
Best 1.5 -0.3 90 
Mechanism Four 1.46~1.65 -0.1~-0.26 150~300 
Best 1.55 -0.2 200 
0 20 40 60 80 100 120 140 1600
10
20
30
40
50
60
70
80
90
100
Time (min)
Co
ut/
Ci
n (
%
)
Predicted
Mechnism1-324
Figure 4.21 Prediction of breakthrough curve following Mechanism One with 
the power-law model system 
155 
References 
1. Klotz, I. M., The adsorption wave, Chemical Reviews, 39 (2), 241-268, (1946). 
2. Bohart, G. S. and Adams, E. Q., Some aspects of the behavior of charcoal with 
respect to chlorine, Journal of American Chemical Society, 42, 523-544, 
(1920). 
3. Amundson, N. R., A note on the mathematics of adsorption in beds, Journal of 
Physical and Colloid Chemistry, 52 (7), 1153-1157, (1948). 
4. Kennedy, D., Fuel cell cathode air filters: methodologies, Master Thesis. 
Auburn, Auburn University, (2007). 
0 10 20 30 40 50 60 70 80 90 1000
10
20
30
40
50
60
70
80
90
100
Time (min)
Co
ut/
Ci
n (
%
)
Predicted
Mechanism2-319
Figure 4.22 Prediction of breakthrough curve following 
Mechanism Two with the power-law model 
156 
5. Yang, R. T., Gas separtion by adsorption processes, Imperial College Press, 
London, (1997). 
6. Cheng, T. B., Jiang, Y., Zhang, Y. P. and Liu, S. Q., Prediction of breakthrough 
curves for adsorption on activated carbon fibers in a fixed bed, Carbon, 42 
(15), 3081-3085, (2004). 
7. Gupta, A., Gaur, V. and Verma, N., Breakthrough analysis for adsorption of 
sulfur-dioxide over zeolites, Chemical Engineering and Processing, 43 (1), 9-
22, (2004). 
8. Das, D., Gaur, V. and Verma, N., Removal of volatile organic compound by 
activated carbon fiber, Carbon, 42 (14), 2949-2962, (2004). 
9. Gaur, V., Asthana, R. and Verma, N., Removal of SO2 by activated carbon 
fibers in the presence of O2 and H2O, Carbon, 44 (1), 46-60, (2006). 
10. Adapa, S., Gaur, V. and Verma, N., Catalytic oxidation of NO by activated 
carbon fiber (ACF), Chemical Engineering Journal, 116 (1), 25-37, (2006). 
11. Mochida, I., Kuroda, K., Kawano, S., Matsumura, Y., Yoshikawa, M., Grulke, 
E. and Andrews, R., Kinetic study of the continuous removal of SOx using 
polyacrylonitrile-based activated carbon fibres 2: Kinetic model, Fuel, 76 (6), 
537-541, (1997). 
12. Yoon, Y. H. and Nelson, J. H., Application of Gas-Adsorption Kinetics 1. A 
Theoretical-Model for Respirator Cartridge Service Life, American Industrial 
Hygiene Association Journal, 45 (8), 509-516, (1984). 
13. Coulson, J. M. and Richardson, J. F., Chemical Engineering, Pergamon Press, 
157 
(1990). 
14. Yang, R. T., Gas Separation by Adsorption Processes, Imperial College Press, 
London, (1997). 
15. Gupta, V. K. and Verma, N., Removal of volatile organic compounds by 
cryogenic condensation followed by adsorption, Chemical Engineering 
Science, 57 (14), 2679-2696, (2002). 
16. Smith, G. D., Numerical Solution of Partial Differential Equations: Finite 
Difference Methods, Oxford University Press, New York, (1985). 
17. Leyva-Ramos, R., Diaz-Flores, P. E., Leyva-Ramos, J. and Femat-Flores, R. 
A., Kinetic modeling of pentachlorophenol adsorption from aqueous solution 
on activated carbon fibers, Carbon, 45 (11), 2280-2289, (2007). 
18. Lua, A. C. and Yang, T., Theoretical Analysis and Experimental Study on SO2 
Adsorption onto Pistachio-nut-shell Activated Carbon, Aiche Journal, 55 (2), 
423-433, (2009). 
19. Bird, R. B., Stewart, W. E. and Lightfoot, E. N., Transport Phenomena, John 
Wiley & Sons, Inc., New York, (2002). 
20. Coulson, J. M. and Richardson, J. F., Chemical Engineering, Butterworth 
Heinemann, Oxford, (2002). 
158
Chapter 5 Conclusions and Recommendations 
5.1 Conclusions 
Proton exchange membrane fuel cell, one of the promising alternative energy sources, is 
prone to sulfur dioxide poisoning which decreased the cell performance of PEMFC and it 
cannot be recovered after purging fresh air. Therefore, there is a need to develop a filter 
media for sulfur dioxide removal, which is compatible with fuel cell systems. Since 
activated carbon fibers have been reported to have advantages in adsorption rate in 
comparison with granular activated carbons, they are chosen for sulfur dioxide removal 
in this research.  
     In order to manufacture activated carbon fibers, polyacrilontirile (PAN) was 
electrospun to nanofibers followed by chemical activation with potassium hydroxide. The 
resulting PAN-based activated carbon nanofibers have small micropore volume with an 
average pore diameter of 3 nm. However, the low productivity and fragility limit its 
application in filtration.  
Commercialized rayon-based and novoloid-based ACFs were studied for sulfur 
dioxide removal. Micropores are dominant in ACFs with an average pore diameter of 
about 2.5 nm. The surface area was found to be related to the pore volume of ACFs but 
was not correlated with sulfur dioxide capacity. However, sulfur dioxide adsorption 
capacity was affected by the micropore percentages: the higher the ratio of micropores in 
159
fibers, the higher the adsorption capacity. Different promoters were treated onto ACFs to 
enhance sulfur dioxide capacity, such as metal, alkali and oxidants. KOH, K2CO3 and 
KClO3 can enhance the capability of fibers to purify sulfur dioxide in the gas stream. 
However, they are not suitable for the wet laying process through which modified ACF 
paper is manufactured due to the loss of sulfur dioxide capacity. On the other hand, 
KMnO4 modified ACFs showed high sulfur dioxide breakthrough capacity and can 
withstand the wet-laying process without loss of sulfur dioxide capacity. The optimum 
weight ratio of potassium permanganate over ACFs is in the range of 43~45 wt% with 
the highest sulfur dioxide capacity. The co-presence of oxygen and water in the carrier 
gas is preferred for effective desulfurization.  
Activated carbon fiber paper (K-ACFP) was manufactured from KMnO4 modified 
ACF15 and the binder fiber, copolyester (CoPET). The strongest K-ACFP was obtained 
when the ratio of KMnO4/ACF15 to CoPET is 7:3.  Sulfur dioxide breakthrough capacity 
of the resulting K-ACFP was found to be similar to KMnO4/ACF15 fiber before the wet 
laying process. Produced K-ACFP can be used as the filter media for fuel cell protection 
from sulfur dioxide.  
Although the removal mechanisms of sulfur dioxide over bare activated carbon or 
activated carbon fibers have been studied to some extent, desulfurization mechanisms of 
modified ACFs were barely studied. Mechanisms of removal of SO2 with KMnO4/ACF 
under different conditions, namely SO2+N2, SO2+air, SO2+H2O+N2 and SO2+H2O+air, 
were analyzed in this research. It is found that KMnO4 was reduced to MnO2 while the 
sulfur dioxide was oxidized to SO42-. Considering diffusions of adsorbates in the packed 
bed and within a single fiber and adsorption and reactions on fibers surface, models were 
160
developed for the prediction of their breakthrough properties. The predicted breakthrough 
curves agree with experimental results reasonably well. With the developed models, the 
dependency of sulfur dioxide removal efficiency on the location in the packed bed and 
the elapsed time are estimated as well.  
5.2 Recommendations 
The KMnO4/ACF filter media developed in the current research can be used to 
manufacture a filter device which can be installed in a fuel cell system. They can be 
packed in a tube, wrapped over a tube or packed in a cartridge to structure the filter 
devices. Different designs of filter devices may result in distinct efficiency to remove 
sulfur dioxide [1], which may need to be further studied to maximize their efficiency.  
    Since nitrogen oxide also degrades fuel cell systems, simultaneous removal of nitrogen 
oxide and sulfur dioxide would work more efficiently in protection of fuel cells. 
Effectively developed filter media on nitrogen oxide removal needs to be studied.  
The ability to be regenerated is an important feature for adsorbents which will decrease 
the investment cost. KMnO4/ACFs need to be studied in this respect.  
The ratio of KMnO4/ACF to CoPET in activated carbon fibers paper was set at 7:3 in 
term of the tensile strength. However, this ratio may not be the optimal for sulfur dioxide 
removal since both the dilution caused by binder fibers [2, 3] and the number of joint 
points, which would influence the available adsorption sites, may influence their sulfur 
dioxide capacity. Therefore, further research is needed to find out the optimal ratio of 
KMnO4/ACF to CoPET for both strength and sulfur dioxide adsorption capacity.  
161
References 
1. Shmidt, J. L., Pimenov, A. V., Lieberman, A. I. and Cheh, H. Y., Kinetics of 
adsorption with granular, powdered, and fibrous activated carbon, Separation 
Science and Technology, 32 (13), 2105-2114, (1997). 
2. Kalluri, R., Microfibrous entrapped catalysts and sorbents: microstructured 
heterogeneous contacting systmes with enhanced efficiency, Dissertation. Auburn, 
Auburn Unviersity, (2008). 
3. Kalluri, R. R., Cahela, D. R. and Tatarchuk, B. J., Microfibrous entrapped small 
particle adsorbents for high efficiency heterogeneous contacting, Separation and 
Purification Technology, 62 (2), 304-316, (2008). 
162g3
g3
Appendix 1 Matlab? code for Yoon?s model 
Function modelyoon  % Code for Yoon?s model  
%For #14 
k=0.036;  t=[0:0.1667:477.93];  
a=length(t);  A=ones(1,a);  tau=8.18*ones(1,a); 
c=A+exp(k*(tau-t)); 
P=(1./c)*100;  b=length(c); 
hold on 
plot(t,P,'m') 
  
%For #293 
k=0.0213;  t=[0:0.1667:477.93]; 
a=length(t);  tau=154*ones(1,a); 
c=A+exp(k*(tau-t)); 
 P=(1./c)*100; 
 b=length(c) 
 hold on 
plot(t,P,'m') 
163g3
g3
Appendix 2 Matlab? code for prediction of breakthrough curves for Mechanisms 
One and Two 
Function model(s,dep,r,theta,ka,kd,k2,beta,mass,Q,RT)   
h=dep/s;   %dep is the bed depth 
k=r*h^2;  
C=ones(theta*s,s+1);  
F=ones(theta*s,s+1);   
C(1,1:end)=0; 
F(1,1:end)=0;  
C(2:end,1)=2.079*10^(-3);   
rf=5*10^(-6); 
df=rf*2; 
rre=0.7937*10^-2;   %the reactor radius (m) 
u=0.04*Q/(60*pi*1.5875^2); %m/s, only for the reactor with diameter of 5/8". 
rhoc=1.99; 
rhom=2.703; 
  Vtp=0.88;  
alpha=1-((1-beta)/rhoc+beta/rhom)/(Vtp+(1-beta)/rhoc+beta/rhom); 
rhof=alpha*1.23+(1-alpha)*(beta*rhom-(1-beta)*rhoc); 
epsi=1-mass/(rhof*10^6)/(pi*rre^2*dep); 
% To calculate Dz,  
%To calculate D (cm^2/s) 
rp=1.3*10^-9;  %pore radius (m) 
164g3
g3
        T=293;        
Mw=28.97;    
D=9.7*10^3*rp*10^2*(T/Mw)^0.5;   
%To calculate Re and Sc    
R=8.314;  
mu=18.27*10^-6; %the viscosity of fluid (air+SO2)(Pa.s). 
rhoa=101325*Mw/(T*R);  
Sc=mu/(rhoa*D*10^-4);   %D is in cm^2/s 
 Re=rhoa*u*df/mu; 
      
%To calculate Dm(molecular diffusivity, cm^2/s),  
Ma=64;    Mb=Mw;  P=1;       
delta=1/2*(4.026+3.617);     
omega=(363*97)^0.5;   
Dm=0.0018583*T^1.5*(1/Ma+1/Mb)^0.5/(P*delta*omega);  
%To calculate Dz 
Dz=Dm*10^-4*(20+0.5*Sc*Re)/epsi; %m^2/s 
  
%kg (mass transfer coefficient) 
Sh=2+1.1*Sc^(1/3)*Re^0.6; %For 3<Re<10^4 
kg=Sh*D*10^-4/df; 
%De  
tau=2.6;  %tortuosity factor of ACFs 
165g3
g3
De=alpha*D*10^-4/tau; 
f=epsi/(1-epsi); 
b=kg+5*De/df; 
p1=1-2*Dz*k/h^2-u*k/h-2*kg*k*(1-kg/b)/(f*rf); 
p2=Dz*k/h^2; 
p3=(Dz*k/h^2)+(u*k/h); 
p4=10*kg*De*k/(f*rf*b*df); 
p5=2*kg*k*(1-kg/b)/(alpha*rf); 
p6=10*kg*De*k/(alpha*rf*df*b); 
p7=((1-alpha)/alpha)*k2*ka*k;  
for n=2:theta*s 
    for m=2:s  
    C(n,m)=p1*C(n-1,m)+p2*C(n-1,m+1)+p3*C(n-1,m-1)+p4*F(n-1,m); 
  F(n,m)=F(n-1,m)-p6*F(n-1,m)+p7*F(n-1,m)/(ka*F(n-1,m)+kd+k2)+p5*C(n-1,m); 
    end  
    F(n,s+1)=F(n,s); 
    C(n,s+1)=C(n,s); 
end 
B=C(1:end,s)/0.002079*100; 
for n=1:theta*s 
        if B(n)>0.5 
            ST=n;  
            B(n) 
166g3
g3
            break; 
          else 
            ST=-1;  % controller  
        end 
end  
if ST==-1 
    msgbox('No 0.5% breakthrough time is found.'); 
else 
TS=RT/ST;            
BT=TS*ST; 
t=[0:TS:(theta*s-1)*TS]; 
d=[0:h:s*h]; 
plot(t,(C(1:end,s)/0.002079)*100,'k:') 
axis([0,100,0,100]) 
xlabel('Time (min)'); 
ylabel('Cout/Cin (%)'); 
end 
167g3
g3
Appendix 3  Matlab? code for prediction of breakthrough curves for Mechanisms Three 
and Four 
Function model(s,dep,r,theta,x,y,k1,beta,mass,Q,RT)   
h=dep/s; % the z direction is divided into s parts and h is stepsize, L is the bed depth 
k=r*h^2; % k is the stepsize in the time direction 
C=ones(theta*s,s+1);  
F=ones(theta*s,s+1);   
C(1,1:end)=0; 
F(1,1:end)=0;  
C(2:end,1)=2.079*10^(-3); %50ppm initial sulfur dioxide 
rf=5*10^(-6); 
df=rf*2; 
rre=0.7937*10^-2; %reactor radius (m) 
u=0.04*Q/(60*pi*1.5875^2); %m/s, only for the reactor with diameter of 5/8". 
rhoc=1.99; 
rhom=2.703; 
Vtp=0.88;  
alpha=1-((1-beta)/rhoc+beta/rhom)/(Vtp+(1-beta)/rhoc+beta/rhom); 
  rhof=alpha*1.23+(1-alpha)*(beta*rhom-(1-beta)*rhoc); 
epsi=1-mass/(rhof*10^6)/(pi*rre^2*dep); 
    
%To calculate D (cm^2/s) 
     rp=1.3*10^-9; %pore radius (m) 
168g3
g3
 T=293;  
     Mw=28.97;  
D=9.7*10^3*rp*10^2*(T/Mw)^0.5;  
%To calculate Re and Sc    
R=8.314;  
mu=18.27*10^-6;  
rhoa=101325*Mw/(T*R);  
     Sc=mu/(rhoa*D*10^-4); 
     Re=rhoa*u*df/mu; 
Ma=64;  
Mb=Mw;   
P=1; 
delta=1/2*(4.026+3.617); %(SO2+air) angstrom 
omega=(363*97)^0.5;  %First note P4 
Dm=0.0018583*T^1.5*(1/Ma+1/Mb)^0.5/(P*delta*omega); %cm^2/s 
%To calculate Dz 
Dz=Dm*10^-4*(20+0.5*Sc*Re)/epsi; %m^2/s 
 tau=2.6;  
De=alpha*D*10^-4/tau; 
f=epsi/(1-epsi); 
b=kg+5*De/df; 
p1=1-2*Dz*k/h^2-u*k/h-2*kg*k*(1-kg/b)/(f*rf); 
p2=Dz*k/h^2; 
p3=(Dz*k/h^2)+(u*k/h); 
169g3
g3
p4=10*kg*De*k/(f*rf*b*df); 
p5=2*k*kg*(1-kg/b)/(alpha*rf); 
p6=1-k*10*kg*De/(alpha*rf*df*b); 
p7=(alpha-1)*k*k1/alpha; %k1 is the k'' in the equation   
for n=2:theta*s 
for m=2:s  
 C(n,m)=p1*C(n-1,m)+p2*C(n-1,m+1)+p3*C(n-1,m-1)+p4*F(n-1,m); 
    F(n,m)=p5*C(n-1,m)+p6*F(n-1,m)+p7*F(n-1,m)^x*(2.73*10^(-3)*mass-6.7*10^(-
4)*F(n-1,m)*Q*k*n)^y; 
    end  
    F(n,s+1)=F(n,s); 
    C(n,s+1)=C(n,s); 
end 
B=C(1:end,s)/0.002079*100; 
for n=1:theta*s 
        if B(n)>5    
            ST=n  
            B(n); 
            break; 
        else 
            ST=-1;   
        end 
end  
if ST==-1 
    msgbox('no 5% breakthrough time is found.'); 
170g3
g3
else 
TS=RT/ST;   
BT=TS*ST;            
t=[0:TS:(theta*s-1)*TS]; 
d=[0:h:s*h]; 
figure,mesh(d,t,(C./0.002079*100)) 
axis([0,0.011,0,200,0,100]) 
xlabel('Depth (m)'); 
ylabel('Time (min)'); 
zlabel('Cout/Cin (%)'); 
colorbar; 
end