Anda di halaman 1dari 14

This article was downloaded by: [Universidad Autónoma del Estado de México]

On: 27 November 2014, At: 12:43


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Chemical Engineering Communications


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/gcec20

A Mathematical Model for Prediction of Pore Size


Distribution Development during Activated Carbon
Preparation
a a a a
Amir Hassan Faramarzi , Tahereh Kaghazchi , Habib Ale Ebrahim & Ali Afshar Ebrahimi
a
Petrochemical Center of Excellence, Chemical Engineering Department , Amirkabir
University , Tehran , Iran
Accepted author version posted online: 24 Apr 2014.Published online: 15 Sep 2014.

To cite this article: Amir Hassan Faramarzi , Tahereh Kaghazchi , Habib Ale Ebrahim & Ali Afshar Ebrahimi (2015) A
Mathematical Model for Prediction of Pore Size Distribution Development during Activated Carbon Preparation, Chemical
Engineering Communications, 202:2, 131-143, DOI: 10.1080/00986445.2013.830609

To link to this article: http://dx.doi.org/10.1080/00986445.2013.830609

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Chemical Engineering Communications, 202:131–143, 2015
Copyright # Taylor & Francis Group, LLC
ISSN: 0098-6445 print/1563-5201 online
DOI: 10.1080/00986445.2013.830609

A Mathematical Model for Prediction of Pore Size


Distribution Development during Activated
Carbon Preparation
AMIR HASSAN FARAMARZI, TAHEREH KAGHAZCHI, HABIB ALE EBRAHIM, and ALI AFSHAR EBRAHIMI
Petrochemical Center of Excellence, Chemical Engineering Department, Amirkabir University, Tehran, Iran
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

A new modeling approach was introduced for the prediction of pore size distribution development during activated carbon
preparation. The mathematical model is based on the modification of a single-pore model for pore growth rate estimation using
a population balance and applying a variable structural parameter random-pore model. The model predictions were compared with
experimental pore size distributions and conversions at various times for pistachio shell char activation by steam between 800 and
950 C, and the kinetic parameters were estimated.
Keywords: Activated carbon; Kinetics; Mathematical modeling; Population balance modeling; Pyrolysis; Simulation

Introduction process, adsorption of heavy metal ions from water, and


as a filter for gas masks.
Noncatalytic gas-solid reactions are very important in the In the ANG process, a high surface area (over 2500 m2=g)
chemical and metallurgical industries. Some of the applica- and a predefined PSD for AC is necessary (Biloé et al., 2002;
tions of these reactions are the adsorption of SO2 by lime, Castello et al., 2002a; Gelzard and Fierro, 2005; MacDonald
reduction of metal oxides, roasting of metal sulfides, catalyst and Quinn, 1998). This tailor-made PSD is essentially
regeneration, and activated carbon (AC) preparation required for maximum adsorption of methane molecules in
(Ramachandran and Doraiswamy, 1982). Mathematical the AC pores (Castello at al., 2002b).
modeling of gas-solid reactions is essential for kinetic para- On the other hand, the PSA process for gas separation
meter estimation from experimental data, design of pilot requires a suitable pre-designed PSD for its AC. For
plants and industrial reactors, and simulation of existing example, in the hydrogen purification after steam reforming
units. Gas-solid reaction models include the sharp interface and high-temperature water-gas shift reaction, AC with a
(Wen, 1968), volume reaction (Duduković and Lamba, pore size suitable for adsorption of CO and CO2 is required
1978), modified grain (Georgakis et al., 1979), nucleation (Ribeiro et al., 2009). In the PSA process for purification of
(Sohn, 1978), single-pore (Ramachandran and Smith, hydrogen from refinery gases, the main impurities are CH4
1977), and random-pore models (Bhatia and Perlmutter, and C2H6; therefore, another AC with its PSD equal to
1980, 1981a, 1981b). The last is the most sophisticated the corresponding size of light hydrocarbons must be used
gas-solid reaction model that has been applied to predict (Malek and Farooq, 1997, 1998; Zhou et al., 2002).
the conversion-time profiles in some environmental pol- The majority of previous works about the evaluation of
lution control reactions with diminishing porosity in recent PSD propagation during AC preparation are experimental.
works (Ale Ebrahim, 2010; Grasa et al., 2009). The change The design of a certain PSD for the final AC has been
of pore size distribution (PSD) of a porous solid due to accomplished using a tedious trial-and-error process. For
the reaction has not been considered by the random-pore example, experimental char activation rates versus burn-off
model until now. with a maximum of about 50% conversion have been
AC is one of the most important chemical materials. reported for CO2 and O2=N2 atmospheres (Dutta and
Some of the applications of AC are the adsorbed natural Wen, 1977; Dutta et al., 1977). A similar experimental work
gas (ANG) process, the pressure swing adsorption (PSA) on CO2 activation for reaching a maximum of BET
(Brunauer-Emmett-Teller) surface area was reported
(Hashimoto et al., 1979). Experimental PSD profiles for
physical and chemical char activation have been presented
Address correspondence to Habib Ale Ebrahim, Petrochemical (Mendoza et al., 2006; Navarro et al., 2006). Finally, a com-
Center of Excellence, Chemical Engineering Department, plete set of experimental PSDs of CO2 activation of char at
Amirkabir University, Tehran 15875-4413, Iran. E-mail: various conversions has been reported (Balci et al., 1987).
alebrm@aut.ac.ir
132 A. H. Faramarzi et al.

Some of the earliest works on pore structure during growth rate has been used for prediction of PSD develop-
activation for prediction of BET surface area and total pore ment of char reaction with CO2 at various activation times,
volume versus conversion and comparison with experimen- with the agreement between experiments and predictions still
tal data are based on the distributed pore model and popu- poor (Navarro et al., 2007).
lation balance approach (Hashimoto and Silveston, 1973a, In this work, a comprehensive mathematical model was
1973b). Another early work is based on the random capillary developed for the prediction of PSD changes versus time
model for the prediction of conversion-time and also during AC preparation. This model is a combination of
rate-conversion profiles in the gasification reaction (Gavalas, the single-pore model concept, a variable-property ran-
1980). A probability model has been presented for conver- dom-pore model, and a population balance approach
sion time and BET surface area and total pore volume versus with its pore growth rate as a function of pore radius. The
conversion profile prediction (Miura and Hashimoto, 1984). results of modeling equations were compared with the
Applications of the random-pore model or its extensions experimental PSD curves and other characteristics of pis-
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

to the char activation process have been reported in several tachio shell char activation by steam with fair agreement.
articles. For example, a maximum surface area was pre- Therefore, this modeling framework can be used as a design
dicted by this model in CO2 activation of char (Balal and tool for PSD adjustment of AC for a predefined engineering
Zygourakis, 1987; Gupta and Bhatia, 2000). This model application.
was also used for prediction of conversion-time and
rate-conversion profiles of char activation (Liu et al., 2003; Mathematical Model
Morimoto et al., 2006; Raghunathan and Yang, 1989).
Mathematical modeling studies for the prediction of PSD The modeling of pore growth rate is based on the single-pore
during activation are very scarce. A discrete random-pore model concept applied to an existing initial PSD in the char.
model was applied for theoretical evolution of PSD after The experimentally defined PSD of char as the initial con-
activation, but its development during activation and also dition and applying the model to all single pores individually
comparison with experimental data were not considered lead to the prediction of PSD for produced ACs. The process
(Bhatia and Vartak, 1996). A comparison between evaluated of physical activation consists of a gas-solid reaction with
PSD from population balance equations and experimental a negligible solid product layer depending on mineral
data was presented for only the pyrolysis (not activation) contaminants of solid reactant remains as ash. Therefore,
process (Klose and Schinkel, 2002). Another theoretical in this case, a porous structure develops with time. Figure 1
work for the activation process showed good comparisons shows different pore sizes that exist in a porous solid before
for BET surface area and total pore volume between model- and after reaction. The dotted lines represent the position of
ing and experiments, but its agreement for PSD profiles at cylindrical pores at the beginning, and solid lines show them
47% and 90% burn-offs between predictions and experi- after some reaction. Since the solid product during AC prep-
mental data was very poor (Junpirom et al., 2005). Finally, aration is negligible, a significant increase in pore sizes will
a simple population balance approach with constant pore occur. It is clear that this increase is more predominant at

Fig. 1. Schematic view of pore widening for three different initial pore sizes (a: medium, b: large, and c: small pore) clarifying vari-
ous parameters.
Prediction of PSD during AC Preparation 133

the pore mouth, which is due to the gaseous reactant concen- The dimensionless form of Equation (7) is:
tration gradient along the pore longitudinal axis.  

d  2 dC 2l 2 BiC 
The assumptions for mathematical modeling are as ð1 þ d 2 Þ ¼ ð1 þ d2 Þ ð8Þ
follows: dx dx r2
1. The pseudo-steady-state approximation for gas conser- Inserting Equation (6) in Equation (8) yields:
vation is valid.  

2. The activation system is isothermal. d  2 dC 2l 2 BiC 
ð1 þ BiC sÞ ¼ ð1 þ BiC  sÞ ð9Þ
3. The reaction is nearly first order and irreversible. dx dx r2
4. There is excess steam versus char, thus bulk gas concen-
tration is constant. At time zero, this equation reduces to the following form
5. The pistachio shell char pellets are slab-like. for a slab pellet:
6. The char slabs remain as their initial size during the d 2C L2 Bi 
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

reaction. ¼ C ¼ m2 C  ð10Þ
7. For low-thickness char slabs, mass transfer resistance dx2 18e0 r2
is negligible. The boundary conditions for a cylindrical pore are:
8. The diffusion in small pores is of Knudsen type.
x ¼ 0 C  ¼ 1 ð11Þ
Char activation can be considered as the following
gas-solid reaction: x ¼ 1 dC  =dx ¼ 0 ð12Þ
AðgÞ þ n B BðsÞ ! CðgÞ ð1Þ The solution for Equation (10) with related boundary
From stoichiometry of reaction (1), we have: conditions is:
cosh½mð1  x Þ
RB ¼ n B RA ð2Þ C ¼ ð13Þ
coshðmÞ
Inserting the values from the single-pore model (without
product layer and Z ¼ 0) into Equation (2) for an arbitrary Averaging the above gas concentration along the pore
initial cylindrical pore radius r yields the following length yields:
(Ramachandran and Smith, 1977): tghðmÞ

  Cav ¼ ð14Þ
dd2 q m
2pðr þ d2 Þ Dx B ¼ n B ½2pðr þ d2 ÞDxkC ð3Þ
dt MB For an increment of time Ds, from Equation (9) we have
the following approximate equation:
As shown in Figure 1, the term d2 is the distance between
reaction interface and initial pore radius position (perpen- d 2C L2 Bi
¼   C ¼ M 2C ð15Þ
dicular to pore wall). Since there is a gas concentration dx 2 2 tghðmÞ
18e0 r 1 þ BiDs m
gradient due to diffusion in the pore, d2 in the pore mouth
is greater than in other interior positions. Consequently,
The solution of Equation (15) is:
a conic pore-mouth opening is gradually produced due to
the gasification reaction that is the reason for pore growth. cosh½Mð1  x Þ
Simplification of Equation (3) yields: C ¼ ð16Þ
coshðMÞ
dd2 n B MB Now the average pore radius at s ¼ Ds can be determined
¼ kC ð4Þ
dt qB by a combination of Equations (6) and (16) as:
After converting Equation (4) into dimensionless form,
rav ¼ r þ rd2;av ¼ r þ r
we have:
Z1  
dd2 cosh½Mð1  x Þ tghðMÞ ð17Þ
¼ BiC  ð5Þ BiDsdx ¼ r 1 þ BiDs
ds coshðMÞ M
0
Integration of the above equation with zero initial
Thus, an equivalent cylindrical pore radius is calculated
condition yields:
by Equation (17) for an actual conic pore shape.
d2 ¼ BiC  s ð6Þ Assuming that the tortuosity factor equals the inverse
of porosity, we can compute effective pore diffusion, Biot
The equality for reaction and diffusion of gas into a single number, and modulus m as follows:
cylindrical pore yields:
sffiffiffiffiffiffiffiffiffiffiffiffi
  2 8Rg T
d dC D ¼ e20 r ð18Þ
D pðr þ d2 Þ2 Dx ¼ kC½2pðr þ d2 ÞDx ð7Þ 3 pMA
dx dx
134 A. H. Faramarzi et al.
sffiffiffiffiffiffiffiffiffiffiffiffi
3k pMA Therefore, we have:
Bi ¼ 2 ð19Þ pffiffiffiffi
2 e0 8Rg T ni 6 e0 ð1  e0 ÞqB vi vi
n0i ¼ ¼ 2
¼ K3 2 ð26Þ
Ari p ri ri
sffiffiffiffiffiffiffisffiffiffiffiffiffiffiffiffiffiffiffi
L k 4 pMA Inserting Equations (23) and (26) into (24) yields:
m¼ ð20Þ  
2e0 r 3e0 8Rg T    
d rv2i d K2 vi
i
¼ K1 ri tgh ð27Þ
The pore growth is now evaluated as: dt dri ri r2i
Differentiation of Equation (27) reduces to:
qffiffiffiffiffiffiffiffiffi  
pMA
r 32 ek2 8R  
rav  r gT d rv2i
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

¼ qffiffiffiffiffiqffiffiffiffiffiffiffiffi0ffisffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i
K1 K2 K1 K2 K1 K2
Ds L k 4 pMA qffiffiffiffiffiffi
1 ffi ¼ 3 vi þ 2 vi tgh  3 vi
2e r
0 3e 8R T
0 g
dt ri ri ri ri ð28Þ
1þ32 k2
pMA tghðmÞ
8Rg T Ds m
   
e
0 ð21Þ K2 K1 K2 dvi
2 sffiffiffiffiffiffiffisffiffiffiffiffiffiffiffiffiffiffiffivffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3 tgh2  tgh
ri ri ri dri
6 L k 4 pMA u u 1 7
tgh4 t qffiffiffiffiffiffiffiffiffi 5 Now, a simple differentiation of the left side of Equation
2e0 r 3e0 8Rg T 1 þ 3 k pMA Ds tghðmÞ
2 e20 8Rg T m (28) in combination with Equation (23) introduces changes
of volumetric PSD with time with a first-order differential
Equation (21) can be used at each time, but there is not any equation as:
analytical method for this general case. On the other hand,      
dvi K2 K2 dvi
the initial pore growth rate is evaluated for Ds ! 0 as ¼ K1 K2 ri vi 1  tgh2 þ K1 r2i tgh 3vi  ri
a simplification approximation versus dimensionless time dt ri ri dri
in the step-wise solution technique as: ð29Þ

Therefore, the change of PSD by activation time can be


pffiffiffi 2 sffiffiffiffisffiffiffiffiffiffiffiffiffiffiffiffi
dr 3 3r k 4 pMA computed from Equation (29), while the last term on the
G0 ¼ ¼ tgh right-hand side contains the slope of present PSD at each
ds L e0 8Rg T
" sffiffiffiffiffiffiffisffiffiffiffiffiffiffiffiffiffiffiffi# ð22Þ pore radius. Moreover, the variation of pore radius ri versus
L k 4 pMA time is calculated from Equation (23). Consequently, the
PSD development during activation is estimated by this
2e0 r 3e0 8Rg T
theoretical step-wise procedure. The initial PSD (initial con-
dition of Equation (29)) was determined experimentally on
The pore growth rate versus actual time can be the initial char before activation.
computed as: In this work, a fourth-order Runge-Kutta method was
applied for solving Equation (23) for all points of pore sizes
pffiffiffi sffiffiffiffisffiffiffiffiffiffiffiffiffiffiffiffi of the initial condition (initial experimental PSD). The calcu-
dri 2 3e20 ri n B MB C0 k 4 8Rg T lated radii of growing pores within one step of the predefined
G¼ ¼
dt LqB e0 pMA time interval were used for solving Equation (29) to compute
" sffiffiffiffiffiffiffisffiffiffiffiffiffiffiffiffiffiffiffi#   ð23Þ the new pore volume distributions. Applying this procedure
L k 4 pMA K2 for the next steps of time intervals predicts the whole PSD at
tgh ¼ K1 ri tgh
2e0 ri 3e0 8Rg T ri several activation times.
Now the concept of random-pore model is applied for
Therefore, the pore growth rate is a function of pore evaluation of conversion-time profiles. Therefore, the vari-
radius ri, as Equation (23) shows. Now the population able parameters of the random-pore model can be estimated
balance for pores without birth and death rates can be from computed PSD at each time (from Equation (29)) by
presented as: the following relations (Bhatia and Perlmutter, 1980):
Z1
@n0 ðr; tÞ @½n0 ðr; tÞG VP ¼ vi dri ð30Þ
þ ¼0 ð24Þ 0
@t @r
VP qB
If vi is volumetric PSD function, n0 i or number distri- e¼ ð31Þ
bution of pore size can be determined as: 1 þ VP qB

Z1
1
lpr2i ni pr2i ni r ¼ vi ri dri ð32Þ
r i vi ¼ ¼ pffiffiffiffi ð25Þ VP
ALð1  e0 ÞqB 6A e0 ð1  e0 ÞqB 0
Prediction of PSD during AC Preparation 135
Z1 the reaction from a diffusion control regime at initial
2 vi
S¼ dri ð33Þ activation times into a mixed diffusion-kinetic control
VP ri at final activation times. The effect of this phenomenon
0
was incorporated in the random-pore model by using
SBET ¼ SVP ð34Þ variable properties computed from estimated PSD at each
time and from integral Equations (30)–(36).
Z1 A suitable rate constant was determined from a comparison
1 vi of experimental conversion-time profiles with random-pore
LP ¼ dri ð35Þ
pVP r2i model predictions (from Equations (37) and (38)). The same
0
rate constant (without any other fitting parameter) was used
simultaneously in Equations (23) and (29) for the prediction
4pLP ð1  eÞ
W¼ ð36Þ of PSD development during activation. As will be shown in
S2
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

the results section, the predicted PSD profiles are in reasonable


For the prediction of the conversion-time profile, coupled agreement with the experimental curves. Moreover, the
partial differential equations of the random-pore model predicted BET surface area profiles compare well with the
(with variable parameters) must be solved numerically. experimental values.
These coupled partial differential equations are (Bhatia and
Perlmutter, 1980, 1981b): Experimental Section
  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Pyrolysis and Activation Tests
@ @a /2 ab 1  w ln b
d ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð37Þ Pistachio shells were crushed and sieved into 4–6 mesh
@y @y 1 þ bZ
w ½ 1  w ln b  1
sizes (3.4–4.8 mm). Then these materials were washed with
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi distilled water and dried at 120 C in an oven for 24 h.
@b ab 1  w ln b Pyrolysis experiments were carried out in a tubular reactor
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð38Þ
@h 1 þ bZ
w ½ 1  w ln b  1
of 2.5 cm diameter and 70 cm length under an inert atmo-
sphere of nitrogen (100 mL=min). The reactor was placed
However, for pistachio shell char activation with only 1% in a vertical tubular electrical furnace with a maximum
ash, Z ¼ 0 can be a reasonable approximation and there is working temperature of 1200 C. The upper side of the
not any product layer resistance. In this work, a finite reactor was connected to a water-cooled condenser in order
element approach was used for solving Equations (37) and to collect the tar of carbonization and excess water vapor
(38). The details of this finite element method have been during the activation stage.
described in our previous works (Ebrahimi et al., 2008, A schematic view of the experimental setup is shown
2009). The growth of pore radius (pore opening) converts in Figure 2.

Fig. 2. Schematic view of experimental setup applied for pyrolysis and activation reaction.
136 A. H. Faramarzi et al.

A K-type thermocouple was applied to monitor and con-


trol the reaction zone temperature. In this system, the sensor
was placed at the center of a tubular reactor among the solid
reactant packed bed. About 40 cm of reactor height from the
bottom was filled with small pieces of ceramic for preheating
of the inert or reactant gaseous streams to the desired set
point. Above this ceramic preheater, about 50 g of pistachio
shells were placed on a support structure provided by the
thermocouple in the middle of the reactor for pyrolysis tests
and heated under a flow of nitrogen with an adjusted heating
rate (20 C=min) to the desired temperature. After 2 h of hold
time with the adjusted temperature, the carbonization step
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

was completed.
The activation experiments were carried out in the same
reactor used for carbonization. In this step, the effects of
temperature and hold time under an excess flow of steam
(5 g=min) were studied. The flow of the activating agent
was controlled more than the stoichiometric value to obtain
an AC with a constant property along the reactor height.
The reactor was heated to the desired temperature (800 ,
850 , 900 , and 950 C) under an inert gas stream. The Fig. 3. Nitrogen adsorption isotherms at 77.4 K of various
activating agent was connected after reaching a temperature prepared samples.
to the set point and stopped after hold time completion. For
example, the hold times were chosen as 5, 10, 15, and 25 min
for consideration of experimental PSD development during Results and Discussion
activation at 850 C. The experimental conversions were
calculated by determining the weight losses of ACs versus First, the conversion-time profile at 850 C is considered. The
initial char. Moreover, the produced ACs at various comparison between model predictions and experimental
temperatures and activation times were experimentally conversions is given in Figure 4 for a value of 4.3 
characterized for BET surface area (7 points) and PSD 107 cm=s as rate constant at 850 C. The conversion value
profiles (55 points). of the first experiment at 300 s is a little higher than the
calculated one, showing more weight loss due to steam
stripping of tar and heavy components remaining at the char
Characterization of Chars and ACs
structure from the pyrolysis section. The quality of these
The quality and properties of pistachio shell chars (PSC) materials is mainly related to the pyrolysis temperature.
and ACs were determined by BET surface area and PSD After completion of pyrolysis, the condensation of these
curves using nitrogen adsorption at 77 K at 55 points melts starts by cooling down, and finally they remain in
with an Autosorb 1-MP (Quantachrome) surface area and the char structure and affect the porosity available for
pore size analyzer with relative pressures between 1  105 activation. The existence of such materials can be confirmed
and 0.99 (Figure 3). The BET surface area of products was by scanning electron microscopy images reported in the
determined applying 7 points of nitrogen isotherms in the literature (Jia and Lua, 2008; Lua et al., 2004).
P=P0 range of 0.05 to 0.3 into the BET equation. As mentioned before, a variable-property random-
Different methods for estimation of PSD of the porous pore model was used in this work for prediction of the
materials from nitrogen adsorption isotherms such as the conversion-time profile. At 850 C, the Thiele modulus
HK (Horvath-Kawazoe) and SF (Saito-Foley) methods for begins from 3.363 at initial activation and reaches to about
micropores and the BJH (Barrett, Joyner, Halenda) method 0.383 at the final stage. Therefore, the system moves from
for mesopores have been reported in the literature. In recent a diffusion-controlled regime at initial activation to kinetic
years other approaches such as nonlocal density functional control at the end. The structural random-pore model para-
theory (NLDFT) have been proposed to characterize porous meter (w) variation is from 1.420 at the beginning to about
solids (Albero et al., 2012; Jia and Lua, 2008; Lua et al., 0.077 at the end.
2004). However, according to new findings in disordered Applying the derived value of the reaction rate constant,
carbon materials such as active carbons, the NLDFT the PSDs of ACs at various times were predicted; they are
method has some artificial false gaps in the micropore given in Figure 5. In Figure 5, initial PSD for modeling
region (Gauden et al., 2004; Neimark et al., 2009; Ustinov was obtained from a very low exposure time between char
et al., 2006). and steam only for stripping the heavy tar from the char
Therefore, a combination of HK (for micropores) surface and pores. Therefore, the birth of small pores was
with the NLDFT method (for mesopores) was selected for accomplished, and it is assumed in the modeling that after
estimation of experimental and theoretical PSD results in this period there is a growth of these small pores. Figure 5
this work. shows fair agreement between the model-predicted PSDs
Prediction of PSD during AC Preparation 137
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

Fig. 4. Comparison of conversion-time behavior from mathematical model and experiments for activation at 850 C.

Fig. 5. Pore size distributions at various times from mathematical model and comparison with PSDs from adsorption isotherm
for activation at 850 C.
138 A. H. Faramarzi et al.
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

Fig. 6. BET surface area propagation at various times from mathematical model and comparison with experimental data for
activation at 850 C.

and PSDs from the adsorption isotherm of AC at 850 C and Figure 9 is a comparison between the model predicted
various activation times. Some faster pore growth predic- PSDs from the adsorption isotherm at 900 C and at various
tions are due to model assumptions and the approximate sol- activation times. In modeling, one of the assumptions
ution in the pore growth rate estimation, but the PSD shape belongs to the population balance section, considering no
and maximum peaks are identical between PSDs from the birth and no death within the pores. Therefore, the pore
adsorption isotherm and model predictions. growth rate of the mathematical model is higher than from
A comparable quality of porous materials is BET surface the adsorption isotherm, which causes the computed PSDs
area. By using Equations (30), (33), and (34) it is possible to to locate on the somewhat larger pores. But in reality, grow-
predict BET surface area from the computed PSDs. Figure 6 ing pores within the reaction are combined with a generation
indicates the propagation of calculated BET surface area of some new pores. Consequently, the PSDs from the
with activation time at 850 C and also the experimental
BET of ACs. This figure shows a very good prediction of
experimental BET surface area propagation by the suggested
mathematical model.
Another interesting subject is the behavior of total pore
surface area. As Bhatia and Perlmutter (1980, 1983) men-
tioned for the random pore model, for w > 2, the pore sur-
face increases monotonically for Zm  Z < 1. Thus, for
w > 2, a maximum in the pore surface area exists only for
Z < Zm. Similarly, for w < 2 a maximum exists only for
Z > Zm, the pore surface monotonically increasing if
1 < Z  Zm and decreasing if Z < 1. With regard to the neg-
ligible amounts of remaining ash in activation reactions, the
Z value is equal to zero. Also, the reported value of the
structural parameter (w ¼ 1.420) predicts a decreasing trend
for pore surface area, which is fully in agreement with
experimental values presented in Figure 7. It is noted that
the limiting value for Z that can occur at a maximum is
determined as follows (Bhatia and Perlmutter, 1980, 1983):

e1=2
Zm ¼ ð39Þ
ð1  e0 Þ

Now the reaction behavior and model validation at 900 C


is considered by the same procedure. Figure 8 shows the Fig. 7. Total pore surface area propagation at various times
predicted and experimental conversion-time profiles with the from mathematical model and from adsorption isotherm for
value of reaction rate constant at 900 C as 6.3  107 cm=s. activation at 850 C.
Prediction of PSD during AC Preparation 139
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

Fig. 8. Comparison of conversion-time behavior from mathematical model and experiments for activation at 900 C.

adsorption isotherm seem to be developed at a relatively experimental BET surface area and predicted BETs from
slower rate than model predictions. the above-mentioned mathematical model is observed.
Finally, Figure 10 shows the propagation of the predicted Detailed error analysis of this work is presented in Table I
and experimental BET surface area with activation time and Table II. As this table shows, the highest error is for the
at 900 C. In this figure, very good agreement between location of the most probable peak radius in the predicted

Fig. 9. Pore size distributions at various times from mathematical model and comparison with PSDs from adsorption isotherm
for activation at 900 C.
140 A. H. Faramarzi et al.
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

Fig. 10. BET surface area propagation at various times from mathematical model and comparison with experimental data for
activation at 900 C.

Table I. Detailed error analysis of this work at 850 C

Error analysis for ACs at 850 C

Maximum Dv of micropores Most probable micropore


(cm3=g  Å) radius (Å) BET surface area (m2=g)

Values from Predicted Error Values from Predicted Error Values from Predicted Error
ads. isotherms values (%) ads. isotherms values (%) ads. isotherms values (%)

t ¼ 5 min 0.07132 0.07005 1.78 2.738 2.916 6.52 1017 1093 7.48
t ¼ 10 min 0.07987 0.08032 0.56 2.788 3.160 13.35 1273 1539 20.91
t ¼ 15 min 0.08636 0.09241 7.01 2.713 3.505 29.23 1620 1799 11.08
t ¼ 25 min 0.119355 0.12227 2.44 2.763 4.180 51.32 1880 2298 22.21

Differential pore volume size distribution.

PSD. One of the reasons is using initial pore growth rate and pore radius overlapping at the population balance equa-
from Equation (22). Moreover, the mesopore birth mech- tion in future work.
anism has been neglected in the population balance equation The computed rate constants at 800 , 850 , 900 , and
(24). Since this work is just an initial try to quantitatively 950 C by this method are used in Figure 11 for the
predict PSD development during activation, one of the most Arrhenius plot. As Figure 11 shows, the activation energy
important modifications is to include mesopore birth rates for pistachio shell char activation by steam is 155 kJ=mol.

Table II. Detailed error analysis of this work at 900 C

Error analysis for ACs at 900 C

Maximum Dv of micropores


(cm3=g  Å) Most probable micropore radius (Å) BET surface area (m2=g)

Values from ads. Predicted Error Values from ads. Predicted Error Values from ads. Predicted Error
isotherms values (%) isotherms values (%) isotherms values (%)

t ¼ 5 min 0.06808 0.06537 3.98 2.688 3.091 15.01 990 1046 5.62
t ¼ 10 min 0.08465 0.08751 3.37 2.788 3.729 33.76 1258 1296 3.03
t ¼ 15 min 0.11612 0.11594 0.16 2.613 4.462 70.80 1605 1602 0.17

Differential pore volume size distribution.
Prediction of PSD during AC Preparation 141
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

Fig. 11. Arrhenius plot for reaction rate constants of pistachio shell char activation by steam at 800 , 850 , 900 , and 950 C.

This value is in the usual range for AC preparation from D gaseous effective diffusivity in a pellet of
different raw materials existing in the literature (Encinar average pore radius r , m2=s
et al., 2001; Klose and Wolki, 2005). 0
D initial gaseous diffusivity in a pellet of
average pore radius r, m2=s
Conclusion G pore growth rate versus time, m=s
G0 pore growth rate versus dimensionless
In this work, a comprehensive mathematical model was time, m
developed for the prediction of conversion time, BET k surface rate constant, m=s
surface area, and PSD development at various times of K1 constant in Equation (23)
pistachio shell char activation by steam. This model is a K2 constant in Equation (23)
combination of a single-pore model, a variable-property K3 constant in Equation (26)
random-pore model, and a population balance approach. pffiffiffiffi
l ¼ L=6 e0 single pore length, m
The predictions of modeling equations were compared with L thickness of the slab pellet, m
the experimental characteristics of ACs at various time inter- LP total pore length per unit volume, 1=m2
vals with fair agreement. Ly position from center of the slab pellet, m
m modulus defined in Equation (10)
Nomenclature M modulus defined in Equation (15)
a ¼ C=C0 dimensionless gas concentration in the MA molecular weight of gaseous reactant,
random-pore model kg=kmol
A external surface area of a slab pellet, m2 MB molecular weight of solid reactant,
b ¼ CB=CB0 dimensionless solid concentration in the kg=kmol
random-pore model MD molecular weight of solid product,
Bi ¼ kr=D Biot number for a pore kg=kmol
C gaseous reactant concentration in the ni number of pores with ri radius
pore, kmol=m3 n0 i number pore size distribution per pellet
C0 bulk gas concentration, kmol=m3 area, 1=m3
C ¼ C=C0 dimensionless gas concentration in the r initial radius of a single pore of char, m
single-pore model rav average pore radius after some
Cav average dimensionless gas concentration gasification, m
in the pore ri radius of pore size distribution, m
CB solid concentration, kmol=m3 r average radius of a pore size distribution,
CB0 initial solid concentration, kmol=m3 m
D gaseous effective diffusivity in a pore with RA reaction rate for gaseous reactant, kmol=s
radius r, m2=s RB reaction rate for solid reactant, kmol=s
DP gaseous effective diffusivity in the product Rg gas constant, J=kmol  K
layer, m2=s S total pore surface per unit volume, 1=m
142 A. H. Faramarzi et al.

SBET BET surface area per unit mass, m2=kg Biloé, S., Goetz, V., and Guillot, A. (2002). Optimal design of an
activated carbon for an adsorbed natural gas system, Carbon, 40,
t time, s
1295–1308.
T absolute temperature, K Castello, D. L., Amoros, D. C., and Solano, A. L. (2002a). Powdered
vi volumetric pore size distribution, m2=kg activated carbons and activated carbon fibers for methane storage:
VP total pore volume per unit mass, m3=kg A comparative study, Energy Fuels, 16, 1321–1328.
x pore length axis, m Castello, D. L., Amoros, D. C., Solano, A. L., and Quinn, D. F.
x ¼ x=l dimensionless pore length axis (2002b). Influence of pore size distribution on methane storage at
X solid conversion relatively low pressure: Preparation of activated carbon with opti-
mum pore size, Carbon, 40, 989–1002.
y ¼ 2Ly=L dimensionless position in the pellet
Duduković, M. P., and Lamba, H. S. (1978). Solution of moving
Z ¼ n D qB MD = molar volume ratio of solid product to boundary problems for gas-solid noncatalytic reactions by orthog-
n B qD MB solid reactant onal collocation, Chem. Eng. Sci., 33, 303–314.
Dutta, S., and Wen, C. Y. (1977). Reactivity of coal and char. 2. In
Greek letters
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

oxygen-nitrogen atmosphere, Ind. Eng. Chem. Process Des. Dev.,


16, 31–37.
b ¼ 2k(1  e)= product layer resistance parameter Dutta, S., Wen, C. Y., and Belt, R. J. (1977). Reactivity of coal and
(n BDPS) char. 1. In carbon dioxide atmosphere, Ind. Eng. Chem. Process
d¼D  =D
0 variation ratio of average pore diffusion Des. Dev., 16, 20–30.
d2 distance between reaction interface and Ebrahimi, A. A., Ale Ebrahim, H., Hatam, M., and Jamshidi, E.
(2008). Finite element solution for gas-solid reactions: Application
initial pore radius position, m
to the moving boundary problems, Chem. Eng. J., 144, 110–118.
d2 ¼ d2 =r dimensionless of d2 Ebrahimi, A. A., Ale Ebrahim, H., Hatam, M., and Jamshidi, E.
e pellet porosity (2009). Finite element solution of the fluid-solid reaction equations
e0 initial pellet porosity with structural changes, Chem. Eng. J., 148, 533–538.
h ¼ kSCt= dimensionless time for the random-pore Encinar, J. M., Gonzalez, J. F., Rodriguez, J. J., and Ramiro, M. J.
CB0(1  e) model (2001). Catalysed and uncatalysed steam gasification of
eucalyptus char: Influence of variables and kinetic study, Fuel, 80,
nB stoichiometric coefficient of solid reactant
2025–2036.
nD stoichiometric coefficient of solid product Gauden, P. A., Terzyk, A. P., Furmaniak, S., Wesolowski, P. P.,
qB true density of solid reactant, kg=m3 Kowalczyk, P., and Garbacz, J. K. (2004). Impact of an adsorbed
qD true density of solid product, kg=m3 phase nonideality in the calculation of the filling pressure of carbon
s ¼ n BMBC0Dt= dimensionless time for the single-pore slit-like micropores, Carbon, 42, 573–583.
qBr2 model Gavalas, G. R. (1980). A random capillary model with application to
/ ¼ ðL=2ÞðkS= Thiele modulus of the random-pore model char gasification at chemically controlled rates, AIChE J., 26,
577–585.
 ÞÞ1=2
ðn B D Gelzard, A., and Fierro, V. (2005). Preparing a suitable material
W random-pore model parameter designed for methane storage: A comprehensive report, Energy Fuel,
19, 573–583.
Georgakis, C., Chang, C. W., and Szekely, J. (1979). A changing
grain size model for gas-solid reactions, Chem. Eng. Sci., 34,
References 1072–1075.
Albero, S. J., Albero, S. A., Reinoso, R. F., and Thommes, M. (2012). Grasa, G., Murillo, R., Alonso, M., and Abanades, C. (2009). Appli-
Physical characterization of activated carbons with narrow cation of random pore model to the carbonation cyclic reaction,
microporosity by nitrogen, carbon dioxide and argon adsorption AIChE J., 55, 1246–1255.
in combination with immersion calorimetry, Carbon, 50, 3128–3133. Gupta, J. S., and Bhatia, S. K. (2000). A modified discrete random pore
Ale Ebrahim, H. (2010). Application of random pore model to SO2 model allowing for different initial surface reactivity, Carbon, 38,
capture by lime, Ind. Eng. Chem. Res., 49, 117–122. 47–58.
Balal, G., and Zygourakis, K. (1987). Evolution of pore surface area Hashimoto, K., and Silveston, P. L. (1973a). Gasification: Part I.
during non-catalytic gas-solid reactions: 1. Model development, Isothermal kinetic control model for a solid with a pore size distri-
Ind. Eng. Chem. Res., 26, 911–921. bution, AIChE J., 19, 259–268.
Balci, S., Dogu, G., and Dogu, T. (1987). Structural variations and a Hashimoto, K., and Silveston, P. L. (1973b). Gasification: Part II.
deactivation model for gasification of coal, Ind. Eng. Chem. Res., Extension to diffusion control, AIChE J., 19, 268–277.
26, 1454–1458. Hashimoto, K., Miura, K., Yoshikawa, F., and Imai, I. (1979). Change
Bhatia, S. K., and Perlmutter, D. D. (1980). A random pore model for in pore structure of carbonaceous materials during activation and
fluid-solid reactions: I. Isothermal kinetic control, AIChE J., 26, adsorption performance of activated carbon, Ind. Eng. Chem.
379–386. Process Des. Dev., 18, 72–80.
Bhatia, S. K., and Perlmutter, D. D. (1981a). A random pore model for Jia, Q., and Lua, A. C. (2008). Effects of pyrolysis conditions on the
fluid-solid reactions: Application to the SO2-lime reaction, AIChE physical characterization of oil-palm-shell activated carbons used
J., 27, 226–234. in aqueous phase phenol adsorption, J. Anal. Appl. Pyrolysis, 83,
Bhatia, S. K., and Perlmutter, D. D. (1981b). A random pore model for 175–179.
fluid-solid reactions: II. Diffusion and transport effects, AIChE J., Junpirom, S., Do, D. D., Tangsathitkulchai, C., and Tangsathitkulchai,
27, 247–254. M. (2005). A carbon activation model with application to longan
Bhatia, S. K., and Perlmutter, D. D. (1983). Unified treatment of seed char gasification, Carbon, 43, 1936–1943.
structural effects in fluid-solid reactions, AIChE J., 29, 281–289. Klose, W., and Schinkel, A. (2002). Measurement and modeling of the
Bhatia, S. K., and Vartak, B. J. (1996). Reaction of microporous solids: development of pore size distribution of wood during pyrolysis, Fuel
The discrete random pore model, Carbon, 34, 1383–1391. Process. Technol., 77–78, 459–466.
Prediction of PSD during AC Preparation 143
Klose, W., and Wolki, M. (2005). On the intrinsic reaction rate of network connectivity of a porous carbon during activation, Carbon,
biomass char gasification with carbon dioxide and steam, Fuel, 84, 44, 2281–2288.
885–892. Navarro, M. V., Seaton, N. A., Mastral, A. M., and Murillo, R. (2007).
Liu, H., Luo, C., Kaneko, M., Kato, S., and Kojima, T. (2003). Assessment of the development of the pore size distribution during
Unification of gasification kinetics of char in CO2 at elevated tem- carbon activation: A population balance approach, Stud. Surf. Sci.
peratures with a modified random pore model, Energy Fuels, 17, Catal., 160, 551–558.
961–970. Neimark, A. V., Lin, Y., Ravikovitch, P. I., and Thommes, M. (2009).
Lua, A. C., Yang, T., and Guo, J. (2004). Effects of pyrolysis con- Quenched solid density functional theory and pore size analysis of
ditions on the properties of activated carbons prepared from micro-mesoporous carbons, Carbon, 47, 1617–1628.
pistachio-nut shells, J. Anal. Appl. Pyrolysis, 72, 279–287. Raghunathan, K., and Yang, R. Y. K. (1989). Unification of coal gasi-
MacDonald, J. A. F., and Quinn, D. F. (1998). Carbon adsorbents for fication data and its applications, Ind. Eng. Chem. Res., 28, 518–523.
natural gas storage, Fuel, 77, 61–64. Ramachandran, P. A., and Doraiswamy, L. K. (1982). Modeling of
Malek, A., and Farooq, S. (1997). Study of a six-bed pressure swing noncatalytic gas-solid reactions, AIChE J., 28, 881–900.
adsorption process, AIChE J., 43, 2509–2521. Ramachandran, P. A., and Smith, J. M. (1977). A single pore model for
Downloaded by [Universidad Autónoma del Estado de México] at 12:43 27 November 2014

Malek, A., and Farooq, S. (1998). Hydrogen purification from refinery gas-solid noncatalytic reactions, AIChE J., 23, 353–361.
fuel gas by pressure swing adsorption, AIChE J., 44, 1985–1992. Ribeiro, A. M., Grande, C. A., Lopes, F. V. S., Loureiro, J. M., and
Mendoza, M. P., Schumacher, C., Garcia, F. S., Almazan, M. C. A., Rodrigues, A. E. (2009). Four beds pressure swing adsorption for
Garcia, M. D., Garzon, F. J. L., and Seaton, N. A. (2006). Analysis hydrogen purification: Case of humid feed and activated carbon
of the microporous texture of a glossy carbon by adsorption mea- beds, AIChE J., 55, 2292–2302.
surements and Monte Carlo simulation: Evaluation with chemical Sohn, H. Y. (1978). The law of additive reaction times in fluid-solid
and physical activation, Carbon, 44, 638–645. reactions, Metall. Trans. B, 9, 89–96.
Miura, K., and Hashimoto, K. (1984). A model representing the change Ustinov, E. A., Do, D. D., and Fenelonov, V. B. (2006). Pore size
of pore structure during the activation of carbonaceous materials, distribution analysis of activated carbons: Application of density
Ind. Eng. Chem. Process Des. Dev., 23, 138–145. functional theory using nongraphitized carbon black as a reference
Morimoto, T., Ochiai, T., Wasaka, S., and Oda, H. (2006). Modeling system, Carbon, 44, 653–663.
on pore variation of coal chars during CO2 gasification associated Wen, C. Y. (1968). Non-catalytic heterogeneous solid-fluid reaction
with their sub-micropores and closed pores, Energy Fuels, 20, models, Ind. Eng. Chem., 60, 34–54.
353–358. Zhou, L., Lu, C. Z., Bian, S. J., and Zhou, Y. P. (2002). Pure hydrogen
Navarro, M. V., Seaton, N. A., Mastral, A. M., and Murillo, R. (2006). from the dry gas of refineries via a novel pressure swing adsorption
Analysis of the evaluation of the pore size distribution and the pore process, Ind. Eng. Chem. Res., 41, 5290–5297.

Anda mungkin juga menyukai