Chemosphere 53 (2003) 17–27

www.elsevier.com/locate/chemosphere

Modeling VOCs adsorption onto activated carbon

C.L. Chuang a, P.C. Chiang a,*
, E.E. Chang b

a
Graduate Institute of Environmental Engineering, National Taiwan University, 71 Chou-Shan road, Taipei, Taiwan
b
Department of Biochemistry, Taipei Medical University, Taipei, Taiwan
Received 15 October 2002; received in revised form 26 February 2003; accepted 25 March 2003

Abstract

The activated carbon adsorption process is affected by the characteristics of adsorbent, adsorbate and environmental
conditions. In this study, both adsorption and desorption processes are assumed to occur simultaneously and a nu-
merical model was developed with a non-linear driving force in conjunction with the Langmuir model for predicting the
overall adsorption process. The numerical model provides both adsorption and desorption rate constants and acti-
vation energies. The resultant equilibrium constants are of the same order of magnitude as reported by other studies.
Results show that the model could well predict the adsorption isotherms and breakthrough curves under various
conditions.
Ó 2003 Elsevier Ltd. All rights reserved.

Keywords: Activated carbon; VOCs; Adsorption; Desorption; Kinetics

1. Introduction 1995) considered the activated carbon bed adsorption

process as a first-order reaction and found the Wheeler
Activated carbon is a common adsorbent for the equation to be successful for application to several or-
removal of hazardous pollutants. It has been reported ganic gases.
that factors, such as characteristics of activated carbon, For the column adsorption studies, the most com-
the nature and concentration of adsorbates, and reac- monly used adsorption models include the linear driving
tion conditions including temperature and humidity, can force (LDF) approximation (Gkueckauf, 1955; Critten-
affect the adsorption process. Pressure swing adsorption den and Weber, 1978; Malek and Farooq, 1996, 1997),
and thermal swing adsorption are the most common the empirical method (Yoon and Nelson, 1984), and the
methods used to study the adsorption process (Hwang Wheeler equilibrium approach (Jonas and Rehrmann,
and Lee, 1994). King and Do (1996) used the Fourier 1972; Vahdat, 1997) (Table 1). According to the LDF
transfer-infrared technique to evaluate the effect of method, the reaction rate is governed by the difference
temperature on the adsorption capability of ethane, between the temporal and the equilibrium concentra-
propane, and n-butane onto activated carbon. They re- tions of the gas adsorbate (Gkueckauf, 1955). The LDF
ported that the equilibrium time at high temperatures model was utilized with a linear or non-linear equilib-
was shorter than that at low temperatures. Wood (1992) rium adsorption isotherm for practical consideration
observed a good correlation between the molar polar- (Malek and Farooq, 1996, 1997). Hwang and Lee (1994)
ization of adsorbent and adsorption capacity. Other used the LDF model to successfully describe the ad-
researchers (Jonas and Rehrmann, 1972; Vahdat et al., sorption and desorption behavior of carbon monoxide
and carbon dioxide onto activated carbon. Yoon and
Nelson (1984) developed an empirical model to predict
*
Corresponding author. Fax: +886-2-23661642. the breakthrough curve of activated carbon beds. Based
E-mail address: pcchiang@ntu.edu.tw (P.C. Chiang). on the thermodynamic characteristic of each solute at
0045-6535/03/$ - see front matter Ó 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0045-6535(03)00357-6
18 C.L. Chuang et al. / Chemosphere 53 (2003) 17–27

Table 1
Adsorption models
Column adsorption model Typical model equation Reference
oq
Linear driving force approximation ¼ kðq  qÞ Gkueckauf (1955), Crittenden and Weber (1978),
ot Malek and Farooq (1996, 1997)


1 C
Empirical method t ¼ t1=2 þ 0 ln Yoon and Nelson (1984)
k C0  C
0 kCQ
k ¼
We
t1=2 ¼ t when C ¼ C0 =2


We q Q C0
Wheeler equation t¼ W  B ln Jonas and Rehrmann (1972), Vahdat (1997)
CQ k C

equilibrium, Myers and Prausnitz (1965) developed an adsorbed gas. The treated activated carbon was stored in
ideal adsorbed solution theory to predict the adsorption a N2 -filling chamber (300 K) before experiments.
of organic vapors onto activated carbon. The adsorption VOC was generated by passing a stream of water and
equilibrium of the gas mixture can be predicted from the hydrocarbon-free N2 gas over a series of diffusion tubes
individual single-gas isotherm using the Toch equation containing the pure liquid of interest. The VOC con-
(Myers and Valenzuela, 1986) and/or the Dubinin/ centrations were controlled by the number of diffusion
Radushkevich equation (Dubinin, 1989). The desorption tubes and temperature as well as the carrier gas flow rate
behavior of carbon films was also observed based on the (1.2 l/min at 303 K by a mass flow meter, Instrument
Arrhenius equation and thermal desorption spectro- Inc., Sierra). The temperature of the VOC vapor was
scopy (Pigram et al., 1994). controlled by a thermostat (283–363 K) before entering
The above methods are generally successful in pre- the reaction column (Fig. 1). The concentrations of
dicting the adsorption of volatile organic compounds C6 H6 , CCl4 and CHCl3 studied were in the concentra-
(VOCs) onto activated carbon under specific conditions. tion range of 8–150, 6–300 and 10–220 mmol/m3 , re-
However, the thermodynamic method considers ad- spectively.
sorption as an equilibrium process. The empirical
method lacks a theoretical rigor. In this study, based on 2.2. Adsorption and desorption in activated carbon bed
the basic dynamical reaction (BDR) theory, a thermo-
dynamic model, called BDR model, involving non-linear For VOC adsorption experiments, about 600–2000
driving force was used to predict the kinetics of the mg of treated activated carbon was packed in a small
adsorption of three VOCs onto activated carbon. Im- glass column (height ¼ 9–30 mm and cross-section
portant kinetic parameters, such as adsorption rate area ¼ 1.76 cm2 ). The column temperature was main-
constant and activation energy were determined to ob- tained at the range from 283 to 363 K for 3 h before
tain the adsorption isotherm and evaluate the adsorp- testing. During adsorption experiments the outlet gas
tion behavior under different operational conditions. was continuously monitored for VOC with a portable
This non-linear reaction kinetic approach can also be photoionization air monitor (Perkinelmer, Photovac
used to predict the adsorption rate of VOCs in column Model 2020). The measurements were calibrated by
operations. gas chromatography (HP 5890A; column: Supel Co.,
VOCOLe, #9354-04A) with a flame ionization detector.
After the termination of the adsorption experiment, the
2. Experimental method inlet gas was changed to the N2 carried gas and the
outlet gas was continuously monitored with the same
2.1. Materials portable photoionization air monitor for obtaining de-
sorption data. The detection limits of C6 H6 , CCl4 and
Benzene (C6 H6 ), carbon tetrachloride (CCl4 ) and CHCl3 were 4.0
106 , 1.2
106 and 1.2
106 mmol/m3 ,
trichloromethane (CHCl3 ) were the adsorbent VOCs respectively.
studied. The activated carbon was of pellet form with a
diameter of 3 mm and a density of 880 kg/m3 (Sorbon-
orit 3, Norit, the Netherlands). The activated carbon 3. Mathematical model
was ground first then sieved to a size fraction from 0.35
to 0.50 mm (or 35 to 45 mesh) and heated to 573 K in a The concentration change of the contaminant in the
N2 atmosphere (purity 99.99%) over 24 h to remove any column can be described by the following equation:
 C.L. Chuang et al. / Chemosphere 53 (2003) 17–27 19

OutletGas

Temperature Mass Flow

Controller Controller
Air monitor

Dehydrocarbon

Carbon Dehydration
bed Diffusion
tubes
N2 Gas
VOC Generator
Gas flow direction

Fig. 1. Experimental apparatus.

oC o2 C oC 1  e oq where S (g/m3 ) is a conversion factor between solid and

 DL 2 þ U þ q ¼0 ð1Þ
ot oL oL e ot gas phase, which can be expressed by the following re-
lation:
where DL , C, L, U , e, q, and q are the dispersion coef-
ficient (m2 /s), the concentration of VOC in the gas phase q0 W
(mol/m3 ), the length of activated carbon bed (m), the gas S¼ ð6Þ
LA  W =q
velocity (m/s), the bed void fraction, the density of ac-
tivated carbon (g/m3 ), and the adsorption capacity (g/g), It is known that the flow rate, Q (m3 /s), has the following
respectively. By assuming the process as a non-equilib- expression:
rium surface reaction and since the column is relatively 1
short, the diffusion term can be negligible. Base on W
q A
thermodynamic consideration, the forward reaction rate Q ¼ UA 1  ð7Þ
LA
(Ra , 1/s) is a function of the adsorbent concentration and
free surface site ratio (1  h), i.e.,
It is assumed that the activated carbon bed can be
Ra ¼ ka ð1  hÞC ð2Þ treated as an entity consisting n-series (n ¼ 50) of con-
stant volume units, i.e., the physicochemical conditions
where ka is the adsorption rate constant, m3 /(s mol) and within each unit are identical, and that the adsorption/
h is the surface coverage, equal to q=q0 , where q0 is the desorption process follows the Langmuir model. Let Dt
maximum adsorptive capacity at the monolayer level (g/g). (s) and DV (m3 ) be the reaction time and the gas volume
Likewise, the backward reaction rate (Rd ) is closely re- in each unit. At the ith unit and experimental time, t, the
lated to h: concentration change in the solid phase and the gas
Rd ¼ kd h ð3Þ phase can be determined by a finite difference method:
In the solid phase from Eq. (4), one has
where kd is the desorption rate constant (1/s). The net
reaction rate (R, 1/s) is equal to the subtraction of Rd htþDt
i ¼ hti þ ½ka ð1  hti ÞCit  kd hti Dt ð8Þ
from Ra :
In the gas phase from Eq. (5), it yields
oh
R¼ ¼ ka ð1  hÞC  kd h ð4Þ
ot QDt t
CitþDt ¼ Cit þ t
ðCi  Ci1 Þ þ ½ka ð1  hti ÞCit þ kd hti SDt
DV
In Eq. (1), ð1  eÞ=e means the ratio of solid phase
ð9Þ
volume to gas phase volume. In a packed activated
carbon bed, the solid phase volume equal to W =q and Accordingly, DV can be expressed by the following
the gas phase volume equal to (LA  W =q), where W is equation:
the weight of activated carbon (g), A the cross-sectional
area of the adsorption bed (m2 ). Eq. (1) can be rewritten LA  Wq
as: DV ¼ ð10Þ
n
oC oC oh Fig. 2 shows the conceptual presentation of the model-
¼ U  S ð5Þ
ot oL ot ing approach.
20 C.L. Chuang et al. / Chemosphere 53 (2003) 17–27

VOC
C0 Cross Section
Q Area: A

i=1 C ti=1 θ ti=1

C ti-1
i=2
i-th unit
i=3
• In solid phase
•
C i-1 θ ti-1
t
θit +∆ t = θ ti + R it, θ∆t Eq. 8
•
L i C ti θ ti In gas phase

• C it+t∆ = C ti +
Q∆t t
( )
C i − C ti −1 + R ti ,c ∆t
• ∆V
• Eq. 9
i = n-1

i=n C ti=n θ ti=n Cti

C tb
Q
Outlet gas

Fig. 2. Conceptual presentation of the modeling approach.

By definition, the equilibrium adsorption mass, We , 4. Results and discussion

can be calculated by the following equation:
Z 4.1. Determination reaction rate constant and activation
1
energy (ka , kd , Ea , and Ed )
We ¼ ðC0  Cb ÞQ dt ð11Þ
0
Table 2 shows the results of column operation under
where C0 and Cb are the inlet and outlet organic vapor various experimental conditions. Note that the retention
concentrations, respectively, and We is the equilibrium time, s, is defined as ðLA  W =qÞ=Q. Clearly, the BDR
adsorption mass. From We , it is possible to calculate the model can reasonably predict the outlet VOC concen-
parameters of Langmuir isotherm, i.e., q0 and KL . In this tration (d ¼ 0:01–0.07). As would be expected, both
study a SYSTAT software package was used. adsorption and desorption rates increased with temper-
To perform the numerical modeling computation, a ature. For example, the desorption rate constant of
series of known variables (q, A, C0 , W , L, Q and T ) and CHCl3 increased from 4.7
104 to 1.6
101 1/s as the
estimated parameters (ka and kd ) were input into Eqs. (8) temperature increased from 283 to 363 K with the cor-
and (9) followed by iteration to obtain the breakthrough responding adsorption rate constant 1.5
101 and 1.05
curves, and the reaction rates (Ra , Rd , and R) were well m3 /(s mol). It is also noted that the increase in the de-
known, too. The best fit of ka and kd was verified by sorption rate constant was greater than that of the ad-
evaluating the root mean square error, d: sorption rate constant. Consequently, the equilibrium
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi constant (K ¼ ka =kd ) decreased by two orders of mag-
u nitude (319–6.5 m3 /mol), indicating the adsorption being
1 u 1 X N
d¼ t ðCb  C 0 Þ2 ð12Þ an exothermic process (DH < 0). Results also show that
C0 N n¼1 CCl4 exhibited the highest equilibrium adsorption con-
stant and C6 H6 possessed the lowest.
where Cb and C 0 are the observed and calculated con- Table 3 presents the equilibrium constants calculated
centrations at breakthrough, respectively, and N is the from the Langmuir isotherm, LDF and BDR models
number measured in each test. which indicate that there is no significant difference in
Table 2
Adsorption conditions, parameters and statistics
VOC Adsorption conditions and parameters Unit Run 1 Run 2 Run 3 Run 4 Run 5
283 K 303 K 323 K 343 K 363 K
C6 H6 Reaction conditions Weight of carbon, W g 2.001 2.005 2.001 2.002 2.001
Length of carbon bed, L cm 3 3.1 3 3 3
Influent concentration, C0 mmol/m3 116 132 108 103 88
Saturation concentration, Csat mol/m3 2.6 6.3 13.5 25.8 45.1
Flow rate, Q l/min 1.14 1.2 1.3 1.38 1.46
Retention time, s s 1.60
101 1.59
101 1.41
101 1.33
101 1.25
101
Reaction parameters Adsorption rate constant, ka m3 /(s mol) 1.3
102 1.6
102 1.9
102 4.0
102 6.0
102
Desorption rate constant, kd 1/s 1.0
104 4.4
104 1.6
103 5.0
103 1.1
102
Equilibrium constant, K ¼ ka =kd m3 /mol 130 37 12 8.0 5.4

C.L. Chuang et al. / Chemosphere 53 (2003) 17–27

Measured equilibrium constant, KL m3 /mol 112 40 23 11 6.3

Statistics Measured number in each test, N – 27 29 41 32 26

Root mean square error, d – 0.02 0.03 0.03 0.05 0.04

Arrhenius equations – lnðka Þ ¼ 2000=T þ 2:6, R2 ¼ 0:90, Ea ¼ 4:0 kcal/mol, a ¼ 13:0 m3 /(s mol)
– lnðkd Þ ¼ 6100=T þ 13, R2 ¼ 0:99, Ed ¼ 12:2 kcal/mol, a ¼ 2:6
105 1/s
– lnðKÞ ¼ 4100=T  9:9, R2 ¼ 0:96, E ¼ 8:2 kcal/mol, a ¼ 1:9
104 m3 /mol

CCl4 Reaction conditions Weight of carbon, W g 0.600 0.600 0.600 0.600 0.600
Length of carbon bed, L cm 0.9 0.9 0.9 0.9 0.9
Influent concentration, C0 mmol/m3 26.8 23.7 22.5 22.2 20.6
Saturation concentration, Csat mol/m3 3.2 7.5 15.5 28.9 49.5
Flow rate, Q l/min 1.15 1.21 1.31 1.39 1.47
Retention time, s s 4.70
102 4.46
102 4.14
102 3.89
102 3.68
102
Reaction parameters Adsorption rate constant, ka m3 /(s mol) 7.5
102 1.0
101 1.5
101 2.0
101 6.0
101
Desorption rate constant, kd 1/s 4.0
105 1.1
103 3.0
103 4.8
103 3.1
102
Equilibrium constant, K ¼ ka =kd m3 /mol 1900 90 50 42 19
Measured equilibrium constant, KL m3 /mol 2200 100 78 40 17
Statistics Measured number in each test, N – 41 29 24 22 19
Root mean square error, d – 0.04 0.07 0.02 0.02 0.03
Arrhenius equations – lnðka Þ ¼ 2400=T þ 5:8, R2 ¼ 0:87, Ea ¼ 4:8 kcal/mol, a ¼ 340 m3 /(s mol)
– lnðkd Þ ¼ 7700=T þ 18, R2 ¼ 0:93, Ed ¼ 15:3 kcal/mol, a ¼ 4:6
107 1/s
– lnðKÞ ¼ 5200=T  12, R2 ¼ 0:83, E ¼ 10:4 kcal/mol, a ¼ 1:4
105 m3 /mol

CHCl3 Reaction conditions Weight of carbon, W g 0.600 0.600 0.600 0.600 0.600
Length of carbon bed, L cm 0.9 0.9 0.9 0.9 0.9
Influent concentration, C0 mmol/m3 32.6 83.3 71.8 69.1 64.9
Saturation concentration, Csat mol/m3 5.64 12.84 25.81 46.97 78.93

21
22 C.L. Chuang et al. / Chemosphere 53 (2003) 17–27

Table 3
Comparison of equilibrium constants from different models

lnðka Þ ¼ 2600=T þ 7:1, R2 ¼ 0:98, Ea ¼ 5:2 kcal/mol, a ¼ 1:4
103 m3 /(s mol)
3.68
102
1.1
100
1.6
101

lnðKÞ ¼ 5400=T  13, R2 ¼ 0:97, E ¼ 10:6 kcal/mol, a ¼ 4:3
105 m3 /mol

VOC Model Temperature (K)
Run 5
363 K

lnðkd Þ ¼ 8000=T þ 20, R2 ¼ 0:98, Ed ¼ 15:8 kcal/mol, a ¼ 6:4
108 1/s

1.47

0.01
283 303 323 343 363

6.5
7.3
31

Experimental conditions: density ¼ 880 kg/m3 , bed cross-section area ¼ 1.76 cm2 and q0 460 mg/g for C6 H6 , 600 mg/g for CCl4 , and 297 mg/g for CHCl3 at 303 K.
C6 H 6 Langmuir 112 40 23 11 6.3
isotherm (KL )
LDFa (KLDF )
3.92
102

150 37 12 8.2 5.5

6.5
101
6.1
102
BDR (this 130 37 12 8.0 5.4
Run 4
343 K

study) (ka =kd )

1.38

0.01
11
12
30 CCl4 Langmuir 2200 100 78 40 17
isotherm (KL )
LDFa (KLDF ) 2000 95 50 42 19
4.16
102
4.0
101
1.9
102

BDR (this 1900 90 50 42 19

Run 3
323 K

study) (ka =kd )

1.30

0.01
21
25
33

CHCl3 Langmuir 530 100 25 12 7.3

isotherm (KL )
LDFa (KLDF ) 450 170 22 11 6.5
4.44
102
2.0
101
1.4
103

BDR (this 320 140 21 11 6.5

study) (ka =kd )
Run 2
303 K
1.22

0.01
140
100

a
LDF with Langmuir isotherm.
42
4.57
102
1.5
101
4.7
104

those equilibrium constants. In general, the equilibrium

Run 1
283 K

constants calculated from the BDR model were located

1.14

0.02
320
530
55

between those calculated from Langmuir isotherm and

LDF model with the exception of CCl4 . The equilibrium
constants of CCl4 were higher than that of C6 H6 and
CHCl3 which implied that CCl4 was a stronger adsor-
m3 /(s mol)

bate than C6 H6 and CHCl3 .

m3 /mol
m3 /mol

Based on the Arrhenius equation, the activation en-

l/min
Unit

1/s

ergy and frequency coefficient can be calculated from the

–
–
–
–
–
s

corresponding models and shown in Table 2. Fig. 3

shows the correlation of the reaction constants (ln k) and
Measured equilibrium constant, KL
Measured number in each test, N

the temperature (1=T ), in which the determination of the

Equilibrium constant, K ¼ ka =kd

relevant activation energy associated with the selected

Adsorption rate constant, ka
Desorption rate constant, kd

Root mean square error, d

VOCs were revealed. For instance, the Ea of C6 H6 , CCl4

and CHCl3 were 4.0, 4.8 and 5.1 kcal/mol, respectively;
Arrhenius equations

and the respective Ed values were 12.2, 15.3 and 15.8

Retention time, s

kcal/mol. For the Langmuir model, the aa of C6 H6 , CCl4

Adsorption conditions and parameters

and CHCl3 were 13, 340 and 1240 m3 /(s mol), respec-
Flow rate, Q

tively; the respective ad values were 2.6
105 , 4.6
107

and 6.4
108 1/s. Consequently, the heat of adsorption
of C6 H6 , CCl4 and CHCl3 can be determined; the values
are )8.2, )10.4 and )10.7 kcal/mol, respectively, and
Reaction parameters

these values are similar to those reported by Chiang et al.

(2001). Summary, CHCl3 has a larger adsorption rate
constant, desorption rate constant, adsorption activa-
tion energy, and desorption activation energy but a
Statistics

lower heat of adsorption and adsorption capacity than

Table 2 (continued)

that of C6 H6 and CCl4 .

Table 4 presents the model (LDF and BDR) com-
parison from the aspects of principle and applications.
In principle, LDF assumes the adsorption rate should be
VOC

linearly proportional to a driving force which is deter-

mined by the difference between the surface concentra-
 C.L. Chuang et al. / Chemosphere 53 (2003) 17–27 23

0 model for particle adsorption (Malek and Farooq, 1997)

which could be applicable to the non-isothermal, multi-
component mixture dynamic-column adsorption pro-
cess. The BDR model used adsorption reaction constant
-4 (ka ) and desorption reaction constant (kd ) to describe the
adsorption process.

4.2. Effect of temperature and concentration

-8 ka
In this study, both C6 H6 and CCl4 are non-polar
kd
molecules with a zero dipole moment, whereas CHCl3 is
(a) C6H6 k a polar molecule with a dipole moment of 1.1 D. Breysse
et al. (1987) studied the adsorption ability of three di-
chlorobenzene isomers on activated carbon and reported
that dipole moment does not have significant effects on
the adsorption of VOC for D < 2:5. Fig. 4 shows the
adsorption capacity as a function of ‘‘relative concen-
-4
tration’’ where Csat is the saturation concentration. For
ideal gas, C0 =Csat is identical to P0 =Psat at a given tem-
ln (k)

perature, where P0 is inlet partial pressure (atm) and Psat

is the saturation pressure (atm). As expected, the ad-
-8 sorption capacity increased with the VOC concentra-
tion and/or partial pressure. The Langmuir model was
(b) CCl4 chosen in subsequent numerical model analysis of
the adsorption/desorption process. Table 2 shows the
Langmuir constant (KL ). At 303 K, the adsorptive ca-
pacity (q0 ) of C6 H6 , CCl4 and CHCl3 are 460, 600 and
297 mg/g, respectively, and the KL of C6 H6 , CCl4 and
0 CHCl3 are 40, 100 and 100 m3 /mol, respectively. The
values of q0 and KL are in the same order of magnitude
as those reported by Vahdat (1997). The results also
-4 indicate that CCl4 has high adsorptive capacity and
adsorption constant, which implies that CCl4 is a
stronger adsorbate towards activated carbon than C6 H6
-8 and CHCl3 .
(c) CHCl3
Fig. 5 presents the relationship between the model
fitted equilibrium constant, ka =kd , and measured con-
-12 stant, KL , as reported by Vahdat (1997), Chiang et al.
0.0025 0.003 0.0035 0.004
(2001) and this study. The plot indicates a linear corre-
lation between KL and ka =kd . The dimensionless plots in
Fig. 6 show the predicted breakthrough curves as a
Fig. 3. Relationship between rate constants (ka , kd , and k) and function of temperature (the reaction conditions are
temperature. shown in Table 2). The results indicated that, at higher
temperatures, an early breakthrough occurs due to re-
duced adsorption capacity (lower equilibrium con-
stants). Furthermore, the breakthrough curves were
tion (q ) and the average adsorbed-phase concentration steeper at higher temperatures due in part to the effects
(q) (Gkueckauf, 1955). The surface concentration can be of ka and kd .
expressed either by Langmuir or Freundlich isotherms. In this study, the adsorption and desorption reaction
The BDR model is based on the basic dynamical reac- constants determined at different temperatures (from
tion theory and assumes the adsorption and desorption 283 to 363 K) were used to predict the desorption con-
rate should be a function of adsorbent concentration centration changes at 343 and 363 K. The desorption
and free surface site ratio (1  h) and desorption of efficiency, i.e., total adsorption weight/total desorption
coverage ratio (h), respectively, which are consistent weight, ranged from 85% to 101%. Results (Fig. 7) show
with the Langmuir isotherm. In addition, the LDF is a that the numerical model could well predict the de-
lumped-parameter (adsorbed-phase transfer coefficient, k) sorption process for CHCl3 . For the other two VOCs,
24 C.L. Chuang et al. / Chemosphere 53 (2003) 17–27

Table 4
Comparison between LDF and BDR model
Model LDF BDR
Principle Similarities oC o2 C oC 1  e oq
 DL 2 þ U þ q ¼0
ot oL oL e ot
oq  oh
Dissimilarities ¼ kðq  qÞ ¼ ka ð1  hÞC  kd h
ot ot
The uptake is linearly proportional to a driving Based on the Langmuir hypothesis, the ad-
force, defined as the difference between the sorption rate is a function of adsorbent con-
surface concentration (q ) and the average centration (C) and free surface site ratio
adsorbed-phase concentration (q) (1  h), and desorption rate is a function of
coverage ratio (h)

Application Similarities 1. Applied at non-isothermal process

2. Applied at multi-component mixture system
Dissimilarities 1. Associated with several isotherms, such as 1. Direct associated with Langmuir isotherm
Langmuir and Freundlich isotherm 2. Both adsorption and desorption reaction
2. Only net reaction rate was known rates were known
3. Applied at both particle and 3. Applied at dynamic-column adsorption
dynamic-column adsorption 4. Applied at both adsorption and desorption
4. Usually applied at adsorption process process

the results show that the measured outlet concentrations 1  h (from 1 to 1  he ) cause the adsorption rate ex-
were lower than the predicted values. It is also observed hibiting a maximum value and then gradually de-
that the Cout =C0 values at higher temperatures decrease creases.
more rapidly than that those at lower temperatures. The case of CCl4 show the adsorption reaction rates,
the desorption reaction rates and net reaction rates are
4.3. Adsorption and desorption rate the function of temperatures, and when adsorption
process approaches equilibrium (i.e., Ra ¼ Rd ), the ad-
In the present study, the numerical model provides sorption rate (or desorption rate) at high reaction tem-
the adsorption and the desorption rate constants, and perature is higher than that at low temperatures. For the
these parameters can be used to predict the desorption other two VOCs, the plots show the same trend as in the
process. Further, the numerical model provides the case of CCl4 .
change in both the adsorption and the desorption rates Fig. 9 presents the changes of adsorption and de-
as a function of temperature and time (e.g., CCl4 from sorption rates of CCl4 during the desorption process at
283 to 363 K in Fig. 8). The pattern in Fig. 8 plots is of two temperatures (343 and 363 K). The results show
course affected by temperature. At higher adsorption that the desorption rates are slightly higher than the
temperatures (323, 343 and 363 K; Fig. 8 top), the adsorption rates. At the beginning of desorption ex-
adsorption rate appears to be slightly higher than the periments, both the adsorption and desorption rates at
desorption rate. Both the adsorption reaction rate, Ra , the higher temperature were higher than that at the
and the desorption reaction rate, Rd , increase with time lower temperature; after that, the adsorption and de-
and eventually reach equilibrium. The net reaction rate, sorption rates at the higher temperature decreased rap-
R, reaches a maximum value before approaching zero. idly than that at the lower temperature, and the net
At low adsorption temperature (283 and 303 K; Fig. 8 reaction rate indicates a max value and a right-trail
bottom), the adsorption rate is much higher than the curve.
desorption rate. The desorption and net reaction rates In this study, the experimental data showed that the
exhibit the same trend with those at higher adsorption adsorption process is an exothermic process (DH < 0);
temperatures. On the other hand, the adsorption re- a higher temperature will decrease the adsorption ca-
action rates show a different trend. It reached a maxi- pacity (q=q0 ). As expected, high temperature will in-
mum value and then declined to the level equivalent to crease both the adsorption and desorption reaction rate
the desorption rate. Since the adsorption rate is equal constants, high rate constants will cause high reaction
ka ð1  hÞC, the 1  h value at lower temperatures rate, and high reaction rate will cause the reaction easy
changes appreciably than that at higher temperatures. reach the equilibrium state and reduce the reaction
The increase of C (from 0 to C0 ) and the decrease of time.
 C.L. Chuang et al. / Chemosphere 53 (2003) 17–27 25

Fig. 4. Prediction of adsorption isotherms (points are experi-

mental data and lines are model predicted).

Fig. 6. Prediction of breakthrough curve of adsorption process

as a function of temperature (reaction conditions: see Table 2).

5. Conclusions

In this study, a BDR model was used to assess the

effect of temperature on the adsorption and desorption
of three VOCs. Compare with Langmuir isotherm and
LDF (with Langmuir) model, the BDR model shown
Fig. 5. Relationship of equilibrium constants of Langmuir and the same trend with both of them, such as CCl4 has a
adsorption/desorption models. larger ka =kd (BDR model) and KL (Langmuir isotherm)
26 C.L. Chuang et al. / Chemosphere 53 (2003) 17–27

Fig. 8. Adsorption and desorption rate change on adsorption

process for CCl4 (reaction conditions: see Table 2).

Fig. 7. Prediction of desorption concentration as a function of

temperature (reaction conditions: see Table 2).

Fig. 9. Adsorption and desorption rate change on desorption

and k (LDF with Langmuir) than C6 H6 and CHCl3 , and process for CCl4 (reaction conditions: see Table 2).
the ka =kd values of CCl4 and C6 H6 have the same order
of magnitude as those reported by Vahdat (1997) and
Chiang et al. (2001). Based on these parameters, it is adsorption process is an exothermic process (DH < 0);
possible to predict the adsorption isotherms and high temperature will decrease the adsorption capacity
breakthrough curves under various operational condi- (q=q0 ). From the model prediction, the adsorption rates
tions. Both experimental and model-predicted data in- were slightly higher than the desorption rates at the
dicate that both the adsorption and desorption rate adsorption process; on the other hand, the adsorption
constants increase whereas equilibrium adsorption con- rates were slightly lower than the desorption rates dur-
stants decreased under high reaction temperatures. The ing the desorption process.
 C.L. Chuang et al. / Chemosphere 53 (2003) 17–27 27

References adsorber FT-IR technique. Chem. Eng. Sci. 51 (3), 423–

439.
Breysse, P.N., Cappabianca, A.M., Hall, T.A., Risby, T., 1987. Malek, A., Farooq, S., 1996. Comparison of isotherm models
Effect of polarity on the adsorption of dichlorobenzene for hydrocarbon adsorption on activated carbon. AIChE J.
isomers. Carbon 25 (6), 803–808. 42 (11), 3191–3201.
Chiang, Y.C., Chiang, P.C., Chang, E.E., 2001. Effects of Malek, A., Farooq, S., 1997. Kinetics of hydrocarbon adsorp-
surface characteristics of activated carbons on VOC ad- tion on activated carbon and silica gel. AIChE J. 43 (3),
sorption. J. Environ. Eng. 127 (1), 54–62. 761–776.
Crittenden, J.C., Weber, W.J., 1978. Predictive model for Myers, A.L., Prausnitz, J.M., 1965. Thermodynamics of mixed-
design of fixed-bed adsorbers: single-component model gas adsorption. AIChE J. 11 (1), 121–126.
verification. J. Environ. Eng. 104 (2), 433–443. Myers, A.L., Valenzuela, D.P., 1986. Computer algorithm and
Dubinin, M.M., 1989. Fundamentals of the theory of adsorp- graphical method for calculating adsorption equilibria of
tion in micropores of carbon adsorbents: characteristics of gas mixtures. J. Chem. Eng. Jpn. 19 (5), 392–396.
their adsorption properties and microporous structures. Pigram, P.J., Lamb, R.N., Hibbert, D.B., Collins, R.E., 1994.
Carbon 27 (3), 457–467. Modeling of the desorption behavior of microporous amor-
Gkueckauf, E., 1955. Theory of chromatography. Part 10. phous hydrogenated carbon films. Langmuir 10 (1), 142–147.
Formulae for diffusion into spheres and their application to Vahdat, N., 1997. Theoretical study of the performance of
chromatography. Trans. Faraday Soc. 51 (11), 1540–1551. activated carbon: in the presence of binary vapor mixtures.
Hwang, K.S., Lee, W.K., 1994. The adsorption and desorption Carbon 35 (10), 1545–1557.
breakthrough behavior of carbon monoxide and carbon Vahdat, N., Swearengen, P.M., Johnson, J.S., Priante, S.,
dioxide on activated carbon. Effect of total pressure and Mathews, K., Neihart, A., 1995. Adsorption capacity and
pressure-dependent mass transfer coefficients. Separ. Sci. thermal desorption efficiency of selected adsorbents. Am.
Technol. 29 (14), 1857–1891. Ind. Hyg. Assoc. J. 56 (1), 32–38.
Jonas, L.A., Rehrmann, J.A., 1972. Kinetics of adsorption of Wood, G.O., 1992. Activated carbon adsorption capacities for
organo-phosphorus vapors from air mixtures by activated vapors. Carbon 30 (4), 593–599.
carbon. Carbon 10 (6), 657–663. Yoon, Y.H., Nelson, J.H., 1984. Application of gas adsorption
King, B., Do, D.D., 1996. Measurement of multicomponent kinetics: I. A theoretical model for respirator cartridge
adsorption kinetics of gases in activated carbon by a batch service life. Am. Ind. Hyg. Assoc. J. 45 (8), 509–516.