Separation and Purification Technology 39 (2004) 23–37

Comparative study of removal of volatile organic compounds by

cryogenic condensation and adsorption by activated carbon fiber
P. Dwivedi, V. Gaur, A. Sharma, N. Verma∗
Department of Chemical Engineering, Indian Institute of Technology Kanpur, Kanpur 208016, India

Abstract

This paper describes studies on two abatement techniques, namely: adsorption by activated carbon fiber (ACF) and conden-
sation at cryogenic temperatures for controlling emissions of volatile organic compounds (VOCs). The breakthrough analysis
was carried out on commercially ACF packed in a tubular reactor under dynamic adsorption conditions. The breakthrough
time was found to significantly decrease with increase in VOC concentration from 5000 to 50,000 ppm. The regeneration of
ACF was carried out by electrical (dc ∼50 V) heating. A temperature range of 120–150 ◦ C and regeneration time of 45–60 min
were typically required for the complete regeneration of ACF pre-equilibrated with 8000 ppm of VOC. The VOC removal by
condensation was carried out in a vertical tubular condenser using liquid nitrogen (LN) as a coolant medium. The studies showed
that adsorption was effective if VOCs concentrations in the pollutants laden gas were in parts per million (ppm) levels, whereas
the removal by cryogenic condensation was found to be effective at the relatively higher VOCs concentrations levels (>1%).
The simulation results from the mathematical models developed to understand the separation process in each of two methods
studied were found to be in good agreement with the experimental data.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Volatile organic compounds (VOC); Adsorption; Cryogenic condensation; Activated carbon fiber (ACF); Breakthrough analysis

1. Introduction especially in industrialized countries have forced in-

Volatile organic compounds (VOCs) are primarily
emitted by chemical process industries (CPI) dealing
with the manufacture, storage, handling and distri-
bution of paints, lubricants and liquid fuels. With
continued global industrialization, the atmospheric
concentrations of VOCs due to primary as well as
fugitive emissions have constantly been on rise. It is
only in recent times that stringent environmental laws,
∗ Corresponding author. Tel.: +91-512-597629;
fax: +91-512-590104.
E-mail address: nishith@iitk.ac.in (N. Verma).
dustries to search for effective abatement technologies
that can comply with the existing laws. The control
techniques for VOCs emissions are broadly based on
the “end-of-pipe” approach due to the inherent pres-
ence of these species in the raw materials and in the
final products.
A number of VOC control techniques are presently
employed world-wide. These are condensation, ther-
mal oxidation, catalytic oxidation, absorption and ad-
sorption. Ruddy and Carrol [1] have summarized the
comparative performances of these techniques, their
advantages, limitations and typical fixed and operat-
ing costs. While condensation using N2 at cryogenic

1383-5866/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.seppur.2003.12.016
24 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37

Nomenclature j component index

a external surface area per unit volume of max maximum
the pellet (m−1 ) p pore of particle
s gas liquid interface, surface of adsorbent
al BET area per unit volume of sorbent
(m−1 )
Superscripts
C concentration (mol/m3 )
0 pure component
Cp heat capacity (J/mol K); gas phase molar
S the particle outer surface
concentration inside pores (mol/m3 )
– volume average quantities inside the pores
CS surface concentration inside pores
of particle
(mol/m2 s)
* non-dimensionalized variables
Dz axial dispersion coefficient (m2 /s)
G molar flowrate of gas (mol/s)
h heat transfer coefficient (W/m2 s)
k mass transfer coefficient (m/s); thermal temperatures was recognized as the most effective and
conductivity (W/m s) commercially suitable technique, the authors termed
ka rate constant for adsorption (m/s) “adsorption” as the “state-of-art” technology having
kd rate constant for desorption (s−1 ) immense potential for VOC control. The removal of
VOCs by adsorption has not drawn much attention in
K Sips isotherm parameter (m3 /mol)
the past. This is probably because the concentration
Km mass transfer coefficient between bulk
ranges of VOCs emitted into the atmosphere are usu-
gas and particle surface (m/s)
ally within the parts per million (ppm) and sub-ppm
l length along the tube (m)
levels. With increasing environmental awareness and
L length of the reactor (m)
stringent regulations of pollutants emission, adsorp-
N Sips isotherm parameter
tion offers an effective means to control VOCs emis-
p partial pressure (Pa)
sions in low concentration levels. In a recent study,
P pressure (Pa)
cryogenic condensation followed by adsorption was
Q flowrate (slpm)
theoretically shown to be more effective in controlling
r radial coordinate
VOCs than the condensation or adsorption alone, if the
R tube outside radius (m)
VOC concentrations varied over a large range [2]. Fur-
Rp particle or fabric radius (m)
thermore, there have been a numerous studies carried
t time (s)
out in the past few years on the development of acti-
T temperature (K)
vated carbon fibers (ACF), which have demonstrated
u velocity (m/s)
that ACF has significant adsorption characteristics for
V vapor molar flowrate (kg/s)
capturing the common atmospheric air pollutants, in-
W coolant flowrate (mol/s)
cluding VOCs, NOx , and SO2 . The following section
X condensate mol fraction
briefly describes various studies carried out on these
Z axial distance along the bed (m)
two control techniques.
Industrial condensation has been in practice for the
Greek letters
past several years. Several widely used unit operations
α porosity of Particle or fiber
in industry, such as distillation, concentration and de-
ε porosity of bed
salination make use of condensation [3]. The recov-
λ latent heat of vaporization (J/mol)
ery of vapors from certain hydrocarbon effluents by
direct condensation has also been carried out as early
Subscripts
as in 1973, when the new hydrocarbon effluent stan-
c coolant
dards came within the statutory regulation in the US
g gas
and European countries [4]. Literature is replete to
in inlet
a great extent with the process design of condensers
 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37
for condensing a pure component, for example satu-
rated steam. However, not much has been published
in literature concerning condensation of a single or
multi-components vapor mixture accompanied by a
non-condensable gas, for example air, although the lat-
ter scenario has practical industrial significance. Un-
der this category, the studies by Krishna and Panchal
[5], Burghardt and Berezowski [6,7] and Mazzarotta
and Sebastiani [8] deal with the process design of a
condenser for multi-components system in the pres-
ence of an inert (non-condensable) species. Barring a
few differences in the assumptions made with regard
to the diffusional interactions between the condens-
ing components in the vapor phase, all of these stud-
ies incorporate the simultaneous effects of heat and
mass transfer within the liquid condensate and vapor
phase.
Although a number of studies have been reported
in literature on the removal of VOCs by porous car-
bon (activated charcoal), the studies on the adsorp-
tion of VOC by ACF have started to emerge only in
the past few years. The adsorption studies on VOC
by activated charcoal range from determining various
appropriate isotherms under equilibrium conditions to
breakthrough analysis in a fixed bed of adsorbents
under dynamic conditions [9–12]. Among the stud-
ies on ACF, Mochida et al. [13] have carried out a
comprehensive theoretical and experimental investiga-
tion over ACF. In this work, the authors related the
extent of adsorption of VOC with different types of
precursor used to manufacture ACF, such as pitched
or rayon-based fibers. Mangun et al. [14] have deter-
mined the effects of fibers pores size on the adsorption
of various hydrocarbons by phenolic-based activated
carbon fibers and concluded that it was necessary to
have a full understanding of the adsorbent structure
for successful design of the adsorbents. A number of
studies have also been reported in literature where the
focus has primarily been on the development of ACF
of relatively larger pore-size and higher BET area by
improving upon the intermediate steps involved in the
manufacture of ACF, such as stabilization, carboniza-
tion, and activation [15–18]. A few studies have also
been reported on the regeneration of ACFs by electri-
cal heating [19].
This paper describes a comparative study carried
out on the removal of VOC by condensation us-
ing liquid nitrogen (LN) and adsorption by ACF.
25
The condensation experiments were carried out in
a vertical tubular condenser. The adsorption exper-
iments were carried out in a vertical fixed packed
bed reactor. The regeneration of ACF saturated with
VOC was carried out by electrical heating. A lim-
ited number of experiments were also performed on
activated charcoals to determine the relative perfor-
mance of these materials with that of ACF. For each
of the separation process (condensation or adsorp-
tion), theoretical analysis were carried out to charact-
erize the mechanism of the removal of VOC and
develop corresponding mathematical models to predict
the VOC removal efficiency under varying operating
conditions.
2. Theoretical analysis
2.1. Condensation
Fig. 1a is a schematic representation of various
resistances to heat and mass transfer that may exist
during condensation of a vapor in the presence of an
inert species on a vertical surface. The mathematical
model for predicting the steady-state VOC concentra-
tion profiles in a tubular condenser takes into account
the simultaneous effects of heat and mass transfer
of the diffusing vapor into the condensate film. The
model also allows for the sensible heat carried by the
diffusing vapors across the film. Here, for the sake
of brevity, we have outlined the major mathematical
steps leading to the development of the governing
equations in the model. The detailed mathematical
steps and various thermodynamics assumptions made
in this model are described elsewhere [2]. Assuming
that equilibrium conditions prevail on the vapor-liquid
interface throughout the condenser tube length, and
the resistance to mass and heat transfer is confined
within the vapor film, as postulated in the film theory
[20], the governing equations required for determin-
ing the temperature and concentration profiles of
VOCs under steady-state conditions in the condenser
can be obtained as follows.
2.1.1. Condensate side
Differential mass balance: A species balance for
condensing species, j at a location l in the tube of
outside radius R is given as
26 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37

Fig. 1. (a) Temperature and concentration profiles during condensation in a vertical tube. (b) Adsorption in a packed bed.

dVj 2.1.2. Coolant side

= −2πRkg,j (pg,j − ps,j ) (1) During condensation, the temperature of the coolant
dl
increases as the sensible and latent heats are conducted
where, kg,j is the vapor side gas film mass transfer through the condensing film, tube wall and coolant
coefficient and pg,j , and ps,j the partial pressures of the film. Similar to the condensate side, the heat balance
condensing species j in the bulk and on the interface, on coolant side will result in the following equation:
respectively. Here, it is assumed that the condensing dTc
vapor species is on the outside surfaces of the tube Wc Cpc = −2πRhoc (Ts − Tc ) (3)
dl
and the coolant flows inside the tube in the counter
current direction. Here, hoc is the overall heat transfer coefficient.
Differential heat balance: The corresponding sensi- Overall heat balance across the film: Taking into
ble heat balance is given as: account the sensible heat as well as the latent heat of
condensation of the diffusing vapors across the film,
dTg −2πRhg ε the following overall heat balance is obtained under
GCp,g = ε (2)
dl (e − 1)(Tg − Ts ) steady-state:
hg ε
Here, the term ε/(1 − e−ε ) in Eq. (2) is a correction hoc (Ts − Tc ) = (Tg − Ts )
factor for the film heat transfer coefficient, hg , required (1 − e−ε )

to account for the sensible heat carried by diffusing + kg,j (pg,j − pg,j )λj (4)
vapors across the film. For multi-components vapor
condensation, ε is the fraction of the total
sensible heat Thermodynamic relations: Since equilibrium is as-
carried by diffusing vapors and equals kg,j (pg,j − sumed to exist at the vapor-liquid interface, the vapor
ps,j )Cp,j / hg . partial pressure, ps,j is related with the vapor pressure
 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37 27

of the pure component, j, corresponding to the inter- of adsorption/desorption of species at trace levels, un-
facial temperature, Ts as given by Raoult’s law: like in the case of bulk separation where concentra-
tion levels are typically higher, has also been justified
ps,j = Xj Pj0 (Ts ) (5)
elsewhere [2,21].
Eqs. (1)–(5) complete the governing equations re-
quired to predict the temperature and concentration 2.2.2. Species balance inside the pores of the
profiles, including the exit VOC concentration under spherical pellet
steady-state conditions in a vertical tubular condenser. Assuming parabolic concentration profiles within
The required boundary conditions for solving the dif- the pore-volume and using volume average concentra-
ferential Eqs. (1)–(3) are determined by the prevailing tion, the following simplified expression for the aver-
operating conditions (temperature and flowrate) at the age concentration of the species j, Cp,j in the pores is
respective inlet ends of the individual stream (gas and obtained:
coolant). In the proposed model, the thermal resis- ∂C̄p,j 3Km (Cg,j − Cp,j
S )
al ∂C̄s,j
tance of the tube walls has been neglected. In addition, − + =0 (7)
∂t Rp α∂t
the diffusional interactions within the vapor film were
neglected considering that the VOC concentrations in In deriving the above simplified expression for Cp,j ,
the typical industrial emissions are small (<10–30%). we have also made use of the well-known linear driv-
ing force approximation (LDF) [21–23]. The terms in
2.2. Adsorption the equation are unsteady-state, diffusive flux into the
adsorbents and the rate of change in the surface con-
The mathematical model for predicting the break- centration of the adsorbed species due to adsorption/
through characteristics (unsteady-state concentrations) desorption, respectively. The detailed mathematical
of an adsorbed species in a fixed bed packed with steps leading to the formulation of the above equa-
adsorbents essentially constitutes of three governing tion are described elsewhere [2,21,24]. Essentially in
equations for concentration balance: (1) in the bulk this approach, the radial concentration profiles in the
gas phase in the bed, (2) in the gas phase within the pores are averaged and the concentration at the pore
pore volume of the adsorbents, and (3) on the surface mouth of the adsorbents Cp,j
S is related with the bulk
of the adsorbents within the pore volume. Referring gas phase concentration, Cg,j , and average gas phase
to Fig. 1b, the primary objective is to determine the concentration in the pores, Cp,j through gas film
time dependent exit gas concentration of the bed dur- transfer coefficient, Km and effective pore diffusivity,
ing dynamic adsorption and desorption. The model De as follows:
equations incorporate the axial dispersion in the bed,
pore diffusion, and the dynamic adsorption/desorption Km Cg + (5Dp /Rp C̄p )
CpS =
rates on the surface of the adsorbents. Km + (5Dp /Rp )
The salient advantage as a consequence of this ap-
2.2.1. Species balance for the bulk gas phase in the
proximation is the reduction of second order PDEs
packed bed
with variation in z and r directions to first order PDE
∂Cg,j ∂2 Cg,j ∂ with variation only in z direction.
− Dz + (uCg,j )
∂t ∂Z 2 ∂Z
1−ε 2.2.3. Adsorption/desorption rate
+ Km a(Cg,j − Cp,j
S
)=0 (6) The rate of change of surface concentration of the
ε
adsorbed species, j, ∂Cs,j /∂t is determined from the
The terms in the equation are unsteady-state, axial dis- individual surface rates of adsorption and desorption:
persion, convection, and the diffusive mass flux into  N   N
the adsorbents, respectively. Since the VOC concen- ∂C̄s,j C̄s,j C̄s,j
 1−  − kd
trations were in ppm levels, the exothermic heat of ad- ∂t = ka C̄p,j
C̄max,j C̄max,j
sorption was negligible and isothermal condition was
assumed. The assumption of isothermality in the case (8)
28 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37
On the inspection of the above rate expression it can
be inferred that at equilibrium, the adsorbent phase
surface concentrations are related with the gas phase
equation through Sips isotherm:
 1/N
KC̄p,j ka
C̄s,j = C̄max , where K =
1 + KC̄p,j kd
(9)
Eqs. (6)–(8) complete the required governing equa-
tions for predicting the gas phase concentrations in
the bed, within the pores and on the surface of the
adsorbents during adsorption/desorption of VOC. The
different parameters such as ka , kd , K, N, Cmax for
charcoal-toluene system were obtained from the pub-
lished works [9,10]. Similarly, the other quantities
such as Dz , Dp , and Km were calculated from vari-
ous correlations based on the prevailing operating con-
ditions in the adsorber bed (refer Appendices A–C).
Since the pore size of granular activated carbon (GAC)
typically varies over a wide range due to the amor-
phous nature of the materials, the effective pore diffu-
sivity was used as an adjustable model parameter. Rp
in Eq. (7) is the radius of the particle or that of the
fiber, depending upon whether the bed is packed with
the granular activated carbons assumed to be spheri-
cal in shape, or with the ACF assumed to be made of
several fibers of uniform diameter.
2.3. Solution technique
The governing Eqs. (1)–(5) of the model for con-
densation and those Eqs. (6)–(8) of the model for ad-
sorption were numerically solved using NAG Fortran
library. The ordinary differential Eqs. (1)–(3) together
with the algebraic Eqs. (4) and (5) were solved si-
multaneously using fourth order Runga–Kutta method,
while the partial differential Eqs. (6)–(8) were solved
with a centrally implicit finite difference scheme [25].
2.4. Model verification
The simulation results obtained from the models
developed in the present work were verified with the
published work [2]. A reasonably good agreement was
observed between the reported experimental data and
the model predictions. The model validation was also
done with the experimental data obtained in this work,
as shown later in this paper.
3. Experimental
Fig. 2 is the schematic of the experimental set-up
designed and used in this work for carrying out sep-
arately the condensation and adsorption experiments.
As described in the figure, the entire set-up can be
assumed to consist of three sections: a gas mixing
section, a separator or reactor section, and an ana-
lytical section. In the gas mixing section, the VOC
laden gas of known concentration was prepared by
bubbling nitrogen at a constant flowrate in the liq-
uid VOC (e.g. toluene, benzene, etc.) contained in a
bottle. The temperature of the bottle was varied with
the help of a refrigeration unit and a PID temperature
controller. Thus, the concentration of VOC in the gas
stream was determined by the vapor pressure of the
vaporizing species corresponding to the temperature
of the liquid. The level of the liquid in the bottle was
adjusted such that the gas bubbles (the size of which
is determined by the perforated holes in the tube) had
sufficient residence time in the bubbler and as a con-
sequence, the exit gas was assumed to be saturated
with VOC at the concentration corresponding to the
temperature of the bubbler.
In the reactor section, two types of reactor were
used depending upon the types of the experiments con-
ducted. For the experiment on the packed bed of ad-
sorbents, a vertical tubular reactor (L = 0.3 m, i.d.:
0.02 m) made of SS was used. SS mesh filters were
used at both ends of the tubes to support the adsorbents
or prevent them from getting carried with the influ-
ent gas. In other type of experiment, the fabric cloth
was wrapped on a vertical reactor (L = 0.3 m, i.d.:
0.02 m) made of Teflon. The shell of the reactor was
perforated and the top end was closed. The reactor was
encased in a SS tube with provisions made for the gas
inlet and outlet connections. The VOC laden gas was
fed to the reactor at a measured flowrate with the aid
of an electronic mass flow controller (Bronkhorst, The
Netherlands). There was a provision of bypassing the
gaseous mixture directly to the analytical section for
calibration as well as for the measurement of the inlet
concentration of the gaseous mixture entering the re-
actor (test section). For the condensation experiments,
a vertical double pipe heat exchanger made of SS was
used. The VOC laden gas was allowed to flow in the
outer tube, while the saturated vapor-liquid mixture of
LN was pumped counter-currently through the inner
 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37
Fig. 2. Schematic diagram of experimental set-up for cryogenic condensation and adsorption studies.
tube by a vacuum pump. The VOC condensed on the
outer surface of the tube at temperature correspond-
ing to the saturation pressure. The entire condenser
was thermally insulated to prevent any heat leakage
from the atmosphere to the condenser. The temper-
ature measurements were made at various locations
along the condenser tube and interconnecting pipes
with the help of temperature sensors, as schematically
shown in Fig. 2.
The concentration of the exit gas from the reactor
(in the case of adsorption/desorption experiment) or
condenser (in the case of cryogenic condensation) was
determined by a gas chromatography (Nucon Co.).
The poropak-Q column (13 cm long × 0.2 cm i.d.)
was used for separating the gaseous mixture. Flame
ionization detector (FID) was used for detecting and
measuring VOC. The operating experimental variables
common in the two studies (adsorption and condensa-
tion) were the gas flowrate, VOC concentrations and
the types of VOC. For the adsorption/desorption ex-
periments, the additional variables were the amount of
the adsorbents (ACF or charcoal) and the temperature
of the adsorbents bed. It is important to point out that
under the experimental conditions, the gaseous stream
29
containing the binary mixture of VOC was kept dry
and the adsorbing materials (charcoals or ACF) were
also activated in a vacuum oven prior to the start of
the adsorption experiments for 24 h at 100 ◦ C to re-
move any adsorbed moisture from the materials. The
vacuum (100 mmHg) was created with the help of a
rotary vacuum pump (Lawrence & Mayo, New Delhi).
The presence of moisture in the gaseous stream or on
the surfaces of the adsorbents is likely to impact the
adsorption performance of the adsorbent materials, as
moisture may occupy some of the active sites for ad-
sorption.
Fig. 3 describes the arrangement employed in this
work for the regeneration of ACF wrapped over the
perforated Teflon reactor by electrical heating. The
regeneration of the ACF saturated with VOC was
carried out in-situ at a constant temperature of around
150 ◦ C and under small N2 flowrate (∼0.05 slpm).
The temperature of the fiber was fixed by the dc
voltage (35–70 V) applied across the longitudinal
direction of the fiber with the help of a dc power
supply. The voltage and current in the wrapped fiber
were measured by a voltmeter and an ammeter, re-
spectively, as depicted in the schematic drawing. The
30 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37
Fig. 3. Experimental set-up for regeneration of ACF by electrical heating.
temperature was monitored by a thermocouple placed
at the center of the shell and adjacent to the ACF
sample. During regeneration, the exit VOC concen-
tration was continuously monitored and measured by
the gas chromatography.
4. Results and discussion
4.1. Condensation
A number of experiments were carried out to deter-
mine the effects of operating conditions such as the
flowrates of the gas and coolant, types of VOC, the
inlet gas VOC concentration and temperature on the
efficiency of VOC removal by cryogenic condensa-
tion. For the sake of brevity only representative re-
sults are discussed here. Fig. 4 describes the effects of
the flowrate of toluene laden N2 gas on the transient
toluene concentrations at the condenser outlet, as con-
densation progresses. The condensation was carried
out using liquid N2 at a flowrate of 5 slpm, while the
toluene concentration in the inlet gas was maintained
constant at 3000 ppm (0.3%) corresponding to the tem-
perature of the bubbler held constant at −15 ◦ C. As
observed from the figure, the higher is the N2 flowrate,
the time to reach the steady-state concentration is also
longer (though marginal) under identical operating
conditions, indicating the overall decrease in the sep-
aration rate at higher flowrate. Under steady-state, the
exit toluene concentration in each case was, however,
approximately 1000 ppm, i.e. only three times reduc-
tion in the toluene concentration. The experiments
were also conducted at relatively higher inlet toluene
concentration in N2 , i.e. 50,000 ppm (5%) correspond-
ing to the bubbler temperature of 30 ◦ C. The remain-
ing operating conditions including gas flowrates were
kept the same as before. Fig. 5 describes the experi-
mental results under these conditions. There are two
distinct observations that can be made from Fig. 5.
First, the differences between the rates of decrease
in the condenser exit toluene concentrations cor-
responding to different gas flowrates (0.1, 0.2 and
0.5 slpm) are comparatively significant. The time to
reach the final steady-state concentration in the case
of 0.1 slpm N2 flowrate is observed to be 15 min as
against 35 and 45 min at the gas flowrates of 0.2
and 0.5 slpm, respectively. Second, comparing with
the corresponding times (approximately 50 min) in
the first case for low toluene concentration (0.3%),
the rates of decrease in toluene concentrations in the
second case are also observed to be considerably
higher. It may also be pointed out that in the latter
case (higher toluene concentration), the liquid N2
flowrate had to be kept marginally lower at 5 slpm
as against 6 slpm in the first case (refer the sub-titles
under each figure). This was found essential under the
existing experimental conditions because at higher
coolant flowrate, decrease in the toluene concentration
was rapid and as a consequence, the measurements
by the gas chromatography could not be tracked at
times.
 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37
Fig. 4. Effect of gas flowrate on toluene exit concentration at low concentration levels (L = 1 m, i.d.: 0.02 m, QLN = 6 slpm,
Cinlet = 3000 ppm).
From these results it was, therefore, concluded
that the removal of VOC is significantly enhanced at
higher concentration (typically >1%) due to higher
mass transfer rate from the bulk vapor phase to the
liquid condensate. The simulation results based on
the model developed in this work also showed that at
low VOC concentrations, relatively larger amount of
coolant, which were found to be sometimes impracti-
cal, was required to reduce the emitted concentrations
below the permitted levels. For example, in a con-
denser of identical dimensions as in the experimental
conditions, the required coolant (LN) flowrate to
reduce the VOC concentration in the pollutant gas
(1 slpm flowrate) from 3000 ppm to less than 100 ppm
was as high as 200 slpm. On the other hand, only
15 slpm flowrate of LN was required if the VOC con-
centration was 30,000 ppm (3%). Thus, in the present
study we also experimentally corroborated the the-
oretical finding of a recent study with regard to the
suitability of the condensation method in controlling
VOC emissions, if the concentration levels are low
31
[2]. Fig. 6 compares the experimental data with the
model predicted results.
As mentioned earlier in the section on the experi-
mental set-up, temperature measurements were made
along the tube length: near the entrance, at the mid-
dle and near the exit. As observed from the figure, a
reasonably good agreement is observed between the
data and model predictions in each case (varying N2
flowrates). The change in the temperature profiles in
the condenser tube due to change in the gas flowrate is
consistent with the mechanism of VOC condensation
described in the theoretical analysis.
4.2. Adsorption
Figs. 7–10 describe the results of the adsorption
study carried out in this work. Fig. 7 compares the
experimental breakthrough curves of toluene during
adsorption by ACF and GAC under identical operat-
ing conditions. In either case, the reactor was packed
with 10 g of the adsorbents (ACF or GAC). Due to
32 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37
Fig. 5. Effect of gas flowrate on toluene exit concentration at high concentration levels (L = 1 m, i.d.: 0.02 m, QLN = 5 slpm,
Cinlet = 50,000 ppm).
difference in their bulk density and therefore, in the
length of the bed, the bed porosity was calculated
to be 0.68 and 0.37 for ACF and GAC, respectively.
The temperature of the packed bed was held constant
at 45 ◦ C. The toluene concentration in the inlet gas
was 4000 ppm, which corresponded to approximately
−10 ◦ C of the bottle temperature. As observed from
the figure, the breakthrough times in the case of AFC
are larger than those in the case of GAC under three
N2 flowrates (0.2, 0.5 and 1 slpm), indicating superior
performance of the former than the latter. At higher
flowrate (i.e. 1 slpm), the breakthrough time in the
case of GAC was observed to be less than 5 min in
comparison with 35 min for ACF. Similarly, the times
to asymptotically reach the VOC inlet concentration
were approximately 5 and 40 min, respectively, under
these conditions. The improved performance of ACF
in comparison with GAC is also reported elsewhere
[26]. The superiority of ACF over GAC in capturing
VOC is attributed to a large number of micropores
on their surfaces, responsible for the gas adsorption.
The model developed in this study was also applied to
simulate the experimental breakthrough curves. Fig. 7
also compares the model results with the adsorption
data of toluene for both ACF and GAC. The model
predictions are observed to be in good agreement with
the data. As pointed out in the theoretical analysis the
values of the adsorption/desorption constants required
in the model were obtained from a previous work and
the remaining parameters were calculated based on
the existing operating conditions. Appendix D sum-
marizes the various physical properties of ACF and
GAC required in the model for predicting the VOC
breakthrough characteristics.
The adsorption experiments were also carried out at
relatively higher concentration, i.e. 50,000 ppm (5%)
to determine the relative performances of ACF and
GAC in adsorbing the organic pollutant. This concen-
tration corresponds to the liquid toluene temperature
of 30 ◦ C. As observed from the breakthrough curves
drawn on Fig. 8 for varying N2 flowrates (0.2, 0.5, and
1 slpm) the breakthrough times for ACF were higher
 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37
Fig. 6. Steady-state temperature profiles along tube length: comparison between data and model predictions (L = 1 m, i.d.: 0.02 m,
QLN = 5 slpm, Cinlet = 50,000 ppm).
than the respective times for GAC under identical
experimental operating conditions. The similar break-
through characteristics were consistently observed
under several other experimental conditions used in
this study. Based on these results, it was concluded
that ACF exhibited enhanced adsorption capacity for
VOC in comparison with GAC for the same amount
of the materials.
One of the primary objectives in the present study
was to demonstrate the efficiency of the two methods,
condensation and adsorption, in capturing the gaseous
organic pollutant in two emission levels: high (∼>1%)
and low (∼<1%). In previous section, it was shown
that the condensation rate appreciably decreased with
decrease in the VOC concentration levels, resulting
in relatively larger time to reach the steady-state
level. The corresponding coolant (LN) requirement
was also higher. To determine the effectiveness of
adsorption in removing the VOC, experiments were
conducted over a wide range of the concentration lev-
33
els (4000–50,000 ppm or 0.4–5%). Fig. 9 describes
the results for varying toluene concentrations. In this
set of the experiments, ACF clothes in fixed quantity
(6.2 g) were wrapped over the perforated Teflon tubu-
lar reactor. The use of Teflon reactor instead of SS
reactor was necessitated due to the electrical heating
required for the regeneration of ACF, as shown in
the next section. As observed from Fig. 9, the break-
through time considerably decreased from approxi-
mately 45 min to 15 min as the toluene concentration
in the carrier gas increased from 4000 to 50,000 ppm,
indicating an early saturation of the adsorbents at high
concentration levels. However, there was marginal
difference between the times corresponding to 4000
and 8000 ppm of concentration levels, as also ob-
served from the figure. The trend in the respective
times to asymptotically reach the final concentration
(inlet gas concentration) is also similar, i.e. the higher
the concentration level (closer to 1%), the significant
increase in the adsorbent saturation time. These data
34 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37

Fig. 7. Breakthrough characteristics of toluene: data and model predictions (W = 10 g, Tbed = 45 ◦ C, Cinlet = 4000 ppm).

Fig. 8. Comparative adsorption performance of GAC and ACF (W = 10 g, Tbed = 45 ◦ C, Cinlet = 50,000 ppm).
 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37 35

Fig. 9. Effect of toluene concentration on adsorption by ACF wrapped over perforated teflon reactor (W = 6.2 g, Tbed = 45 ◦ C,
QN2 = 0.5 slpm).

Fig. 10. Effect of regeneration time on adsorption of toluene by ACF (adsorption conditions: Tbed = 45 ◦ C, Cinlet = 8000 ppm,
QN2 = 0.2 slpm; regeneration conditions: T = 150 ◦ C, QN2 = 0.1 slpm, voltage = 40 V, I = 3 A).
36 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37
confirm the finding of a theoretical study that ad-
sorption is an effective means for controlling VOC if
the concentration levels are typically lower than 1%,
while the condensation is mostly suited for higher
concentration levels [2].
In the present study, electrical heating was em-
ployed to regenerate the fibers saturated with VOC
during adsorption. The details of the arrangement
were previously described in the section on the ex-
periment. A series of the test runs were carried out
to determine the optimum regeneration conditions,
including regeneration time and dc voltage. For each
case the regenerated fiber cloth was re-subjected to
the adsorption test to determine its adsorption perfor-
mance following regeneration. Fig. 10 describes one
of such results obtained on the breakthrough of the
ACF regenerated for different regeneration times. The
adsorbent was equilibrated with 8000 ppm of toluene
prior to regeneration. The voltage was maintained
constant at 40 V. The corresponding temperature of the
cloth was measured to be 150 ◦ C. As observed from
the figure, there is insignificant difference between
the adsorption profiles of the fiber cloth regenerated
for 15 and 30 min. The adsorption characteristics of
the fiber clothes which were regenerated for at least
45 min are, however, found to improve, resulting in
increase in the breakthrough time. The adsorption
profiles of ACF regenerated for 60 min are observed
to be approximately identical to those of a fresh ACF.
Therefore, in this study a regeneration time of 60 min
was chosen to be optimum for the maximum removal
of the adsorbed VOC. During the regeneration period,
the current gradually increased from an initial value
of 1–3 A, indicating decrease in the electrical resis-
tance of the fiber from 40 # to approximately 13 #,
as the adsorbed VOC was gradually removed from
the fabric due to heating.
5. Conclusion
The removal of VOCs by cryogenic condensation
and adsorption has been studied experimentally and
theoretically. The condensation rate of VOC was
shown to decrease with decrease in the effluent con-
centration suggesting difficulty in the separation at
low concentration levels (<1%). The commercially
obtained activated carbon fiber was shown to exhibit
larger adsorption for VOC than granular activated
carbon under identical operating conditions. The
breakthrough time during adsorption was observed to
considerably decrease with increase in the concentra-
tion suggesting the utility of the adsorption method for
controlling VOC from gaseous effluents at ppm and
sub-ppm levels. The regeneration of ACF equilibrated
with VOC was carried out by dc electrical heating.
The regeneration time of 60 min and temperature of
150 ◦ C were determined to be effective in completely
removing the adsorbate. The model predictions for the
steady-state concentration and temperature profiles in
the condenser and the unsteady-state breakthrough
responses in the fixed-bed adsorber were found to
be in reasonable agreement with the experimental
data. With the help of the mathematical models, the
extent of VOC removal by these two methods was
quantitatively predicted.
Acknowledgements
The authors acknowledge the partial financial sup-
port (Grant No.: 01(1599)/99/EMR-II) from Central
Scientific Industrial Research (CSIR), New Delhi to
conduct this study. The authors also thankfully ac-
knowledge HEG Ltd., Bhopal (India) for supplying
the samples of ACF for conducting the present study.
Appendix A
The axial dispersion coefficient in the packed bed,
Dz is expressed in terms of Peclet number defined as,
Pe = 2Rp µ/Dz ρ. The Peclet numbers are correlated
with particles Reynolds and Schmidt numbers through
one of the empirical correlations as follows [21]:
1 0.3 0.5
= +
Pe Re × Sc 1 + 3.8/(Re × Sc)
for 0.008 < Re < 400 and 0.28 < Sc < 2.2
Appendix B
The mass transfer coefficient Km in the packed bed
is calculated using the following correlation for Sher-
wood number defined as, Sh = 2Km Rp /Dm [21]
Sh = 2.0 + 1.1(Sc)0.33 (Re)0.6
 P. Dwivedi et al. / Separation and Purification Technology 39 (2004) 23–37 37

Appendix C [7] A. Burghardt, M. Berezowski, Comparison of various mass

transport models in multicomponent condensation, Chem.
Eng. Process. 31 (1992) 161.
The effective diffusivity inside the pores is deter- [8] B. Mazzarotta, E. Sebastiani, Progress design of condensers
mined using the formula for vapor mixtures in the presence of non-condensable gases,
CJChE 73 (1995) 456.
Dα
Dp = [9] J.H. Yun, D.K. Choi, S.H. Kim, Equilibria and dynamics for
τ mixed vapor of BTX in an activated carbon bed, AICHE J.
45 (1999) 751.
The combined diffusivity, D inside the macro-pores of [10] J.H. Yun, D.K. Choi, S.H. Kim, Adsorption equilibria of
the zeolites pellet is given by a combination of both chlorinated organic solvents onto activated carbon, Ind. Eng.
Knudsen (DK ) and molecular (Dm ) diffusivities as: Chem. Res. 37 (1998) 1422.
[11] A. Kapoor, R.T. Yang, Roll-up in fixed bed, multicomponent
1 1 1
= + adsorption under pore diffusion limitation, AICHE J. 33
D DK Dm (1987) 1215.
[12] S. Farooq, D.M. Ruthven, Dynamics of kinetically controlled
where Knudesn diffusivity is given as, binary adsorption in a fixed bed, AICHE J. 37 (1991)

299.
Temp 1/2 [13] I. Mochida, Y. Korai, M. Shirahama, S. Kawano, T. Hada,
DK (m /s) = 97 × Rpore (m)
2
M Y. Seo, M. Yoshikawa, A. Yasutake, Removal of SOx
and NOx over activated carbon fibers, Carbon 38 (2000)
227.
[14] C.L. Mangun, M.A. Daley, R.D. Braatz, J. Economy, Effect
Appendix D
of pore size on adsorption of hydrocarbons in phenolic-based
activated carbon fibers, Carbon 36 (1995) 123.
Physical properties of activated carbon fibers and [15] J.C. Lee, B.H. Lee, B.G. Kim, M.J. Park, D.Y. Lee, I.H.
granular activated carbons [25]. Kuk, H. Chung, H.S. Kang, H.S. Lee, D.H. Ahn, The effect
of carbonization temperature of PAN fiber on the properties
ACF GAC of activated carbon fiber composites, Carbon 35 (1997)
1479.
Fiber or particle size 10–20 ␮m 1–3 mm [16] P.J.M. Carrot, J.M.V. Nabais, M.M.L.R. Carrott, J.A. Pajares,
Specific surface area 700–2500 900–1200 Preparation of activated carbon fibers from acrylic textile
(m2 /g) fibers, Carbon 39 (2001) 1543.
[17] S. Park, Ki.-D. Kim, Influence of activation temperature
Physical surface area 0.2–2.0 ∼0.001 on adsorption characteristics of activated carbon fiber
(m2 /g) composites, Carbon 39 (2001) 1741.
Average diameter of <40 NA [18] S.Y. You, Y.H. Park, Preparation and properties of activated
pores (nm) carbon fabric from acrylic fabric waste, Carbon 38 (2000)
1453.
[19] A. Subrenat, J.N. Baleo, P.L. Cloirec, P.E. Blanc, Electrical
behaviour of carbon cloth heated by the joule effect:
References desorption application, Carbon 39 (2001) 701.
[20] A.P. Colburn, T.B. Drew, The condensation of nixed vapors,
[1] E.N. Ruddy, L.A. Carroll, Select the best VOC control Trans. Am. Inst. Chem. Eng. 33 (1937) 197.
strategy, Chem. Eng. Prog. 89 (1993) 28. [21] R.T. Yang, Gas Separation by Adsorption Processes, Imperial
[2] V. Gupta, N. Verma, Removal of volatile organic compounds College Press, UK, 1997.
by cryogenic condensation followed by adsorption, Chem. [22] C.W. Tien, Adsorption Calculations and Modeling,
Eng. Sci. 57 (2002) 97. Butterworth-Heinemann, US, 2001.
[3] D.R. Webb, J.M. McNaught, Heat Exchanger Technology, [23] D.D. Duong, Adsorption Analysis Equilibria and Kinetics,
first ed., Applied Science, Barking, UK, 1980. Imperial College Press, UK, 1998.
[4] C. Ibbetson, At ease with VOCs, Chem. Eng. 92 (1996) 33. [24] K. Dasgupta, K.N. Rai, N. Verma, SO2 breakthrough and
[5] R. Krishan, C.B. Panchal, Condensation of binary vapor sulfate conversion analysis during sorption by Ca-based
mixture in presence of an inert gas, Chem. Eng. Sci. 32 sorbents, Can. J. Chem. Eng. 81 (1) (2003) 53.
(1977) 741. [25] J.H. Ferziger, Numerical Methods for Engineering
[6] A. Burghardt, M. Berezowski, A computational design Applications, Wiley, New York, 1981.
method for multicomponent condensation, Chem. Eng. [26] M. Inagaki, New Carbons: Control of Structure and Functions,
Process. 24 (1988) 189. Elsevier, Amsterdam, 2000.