Você está na página 1de 11

Chemical Engineering Science 61 (2006) 4762 – 4772

www.elsevier.com/locate/ces

Modeling the breakthrough behavior of an activated carbon fiber monolith in


n-butane adsorption from diluted streams
Gregorio Marbán ∗ , Teresa Valdés-Solís, Antonio B. Fuertes
Instituto Nacional del Carbón, CSIC, Francisco Pintado Fe, 26, 33011 Oviedo, Spain

Received 8 August 2005; received in revised form 31 January 2006; accepted 6 March 2006
Available online 13 March 2006

Abstract
A complete breakthrough model is proposed and solved in this work to describe the removal of volatile organic compounds from a diluted
single-component gas stream passing through a microporous activated carbon fiber monolith (ACFM) under isothermal conditions. All significant
parameters are considered in the model, including type of adsorption isotherm, resistance to external gas diffusion, non-instantaneous adsorption
kinetics at the external surface and type of gas diffusion within the pore system. n-Butane was employed as the test compound for obtaining
the experimental breakthrough curves. The model including non-instantaneous adsorption at the external surface and surface pore diffusion
describes properly the experimental adsorption results.
䉷 2006 Elsevier Ltd. All rights reserved.

Keywords: Activated carbon; Carbon fibers; Volatile organic compounds; Adsorption; Mass transfer; Diffusion; Simulation; Modeling

1. Introduction VOCs (Centeno et al., 2003; Fuertes et al., 2003). However, the
use of granular packed beds has serious drawbacks, such as the
Low-density activated carbon fiber monoliths (ACFMs) have high pressure drop associated with the flow of gas through the
been developed recently (Burchell et al., 2000; Burchell and packed media, attrition of the granular material, channeling, gas
Judkins, 1997; Lee et al., 1997; Marbán et al., 2000; Vila- bypassing, etc. These disadvantages can be avoided with the
plana-Ortego et al., 2002) for several gas–solid applications utilization of ACFM that, in addition, exhibits a narrow poros-
due to their properties such as light-weight, high mechanical ity in the micropore range (< 2 nm). Micropores are primarily
strength and fast adsorption kinetics. The monolithic structure responsible for adsorption at low VOCs concentrations in am-
makes them easier to handle than packed beds and produces a bient air environments due to the overlapping of the attractive
low resistance to bulk gas flow. Thus, among the possible appli- forces of opposite pore walls (Derbyshire et al., 2001; Foster
cations, ACFMs are of interest for the adsorption and recovery et al., 1992). In many cases n-butane has been employed as
of organic vapors (Fuertes et al., 2003; Marbán and Fuertes, a reference molecule to compare the VOC adsorption capac-
2004), CO2 adsorption (Burchell et al., 1997; Burchell and Jud- ity of various materials (DeLiso et al., 1997; Gadkaree, 2001;
kins, 1997), as catalysts or catalyst support for the removal of Valdés-Solís et al., 2004).
SOx and NOx from flue gas (Marbán et al., 2003), water treat- The adsorption of VOCs in carbonaceous materials has
ment (Suzuki, 1991) and for CH4 storage (Muto et al., 2005). received increasing attention in recent years, and several at-
The removal of volatile organic compounds (VOCs), com- tempts to model the process have been published for granular
monly performed by adsorption, is of great interest for the air active carbon (Linders et al., 2001, 2003), ACF (Cheng et al.,
quality control. At a low-concentration level, adsorption on ac- 2004; Das et al., 2004) and carbon-coated ceramic monoliths
tive carbon is the most employed method for the removal of (Valdés-Solís et al., 2004). In general, the models found in liter-
ature include several simplifications; such as considering only
∗ Corresponding author. Tel.: +34 985 11 90 90; fax:+34 985 29 76 62. gas diffusion in the pores or assuming instantaneous adsorption
E-mail address: greca@incar.csic.es (G. Marbán). kinetics (Cheng et al., 2004; Linders et al., 2001, Valdés-Solís
0009-2509/$ - see front matter 䉷 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2006.03.008
G. Marbán et al. / Chemical Engineering Science 61 (2006) 4762 – 4772 4763

et al., 2004), in order to get substantial savings in computation Table 1


time. Then, every approximate solution has its own restricted Textural properties of the adsorbent material (ACFM) and bulk parameters
used during the dynamic adsorption experiments
scope of validity which should be accurately defined.
The objective of this work is to propose and solve a math- Textural properties
ematical breakthrough model for low-concentration gas ad- SBET (m2 /g)a 1306
N2
sorption in ACF monoliths which takes into consideration all Vpore (cm3 /g)b 0.46(5)
CO2
significant parameters (non-linear adsorption equilibrium, Vpore (cm3 /g)c 0.44
axial dispersion, finite mass transfer resistance, type of pore L0 (nm)d 0.93
Adsorbed n-butane (mmol/g) at C0 = 106 ppm(30 ◦ C) 4.55
diffusion, etc). Thus, a systematic approach is performed,
Adsorbed n-butane (mmol/g) at C0 = 1500 ppm(30 ◦ C) 2.02
starting from a simplified and unrealistic model (gas phase Adsorbent characteristic energy; E0 (kJ/mol) 20.3
diffusion in the micropores with instantaneous adsorption), (Dubinin–Radushkevich)
going through a more complex and realistic model that permits
Fiber dimensions
to analyze the effect of the equilibrium adsorption isotherm
(Dubinin–Radushkevich, Freundlich, Toth, etc) on the surface Diameter DF (m) 11.6
diffusion-controlled process, to finally include the possibility Bulk parameters
of non-instantaneous adsorption kinetics at the external surface Wcarbon (mg) 48
of the fibers, thanks to which the simulation results permit to Diameter D (mm) 5.5
Height H (mm) 10.6
eventually match the experimental breakthrough curves.
Aspect ratio (H/D) 1.9
Bulk carbon density b (g/cm3 ) 0.19
2. Experimental Bulk porosity (b ) 0.8
a BET specific surface area obtained by N adsorption at −196 ◦ C
2
b Pore volume obtained by N adsorption at −196 ◦ C (percentage of meso-
2.1. Adsorbent material 2
pore volume in brackets).
c Micropore volume obtained by CO adsorption at 20 ◦ C
2
The preparation of activated monolith (ACFM) is reported d Mean micropore width obtained from n-butane adsorption experiments as
elsewhere (Marbán et al., 2000). Briefly, rejects of Nomex姠 reported in (Centeno et al., 2003).
fibers (DuPont) were carbonized in N2 at 850 ◦ C (heating rate:
5 ◦ C min−1 ; soaking time: 1 h). Afterwards, they were milled cess of the gas through the carbonaceous samples. With these
by conventional blade-mills and the fraction of size 0.1–0.4 mm conditions, and considering a fast (adsorbent fully loaded in
was dry-mixed with powdered phenolic resin (Novolak) in a 10 s) and linear adsorption rate, and negligible gas bypassing,
mass ratio of 3/1. The mixture was moulded, cured (in air at the value of the adsorption rate was estimated be below 1.2%
180 ◦ C) and carbonized (in N2 at 700 ◦ C). The carbonized com- the value of the n-butane molar flow rate, which clearly estab-
posite with a bulk density of around 0.25 g cm−3 was activated lished differential regime. However, it was determined that a
at 700 ◦ C with a stream of water (∼ 25 vol% in N2 ) until acti- high degree of gas bypassing through the annulus between the
vation degree (burnoff) of 42 wt% was reached. crucible and the TGA chamber impeded to attain strict differ-
ential conditions, and therefore the TGA results are discussed
2.2. Adsorbent characterization in the corresponding section from a qualitative perspective.

The textural properties (specific surface area and pore vol- 2.4. Dynamic adsorption experiments
ume) of the adsorbent material were evaluated by means of N2
adsorption isotherms (−196 ◦ C) obtained with a Micromerit- The adsorption chamber used for obtaining the breakthrough
ics ASAP 2010 analyzer, and by CO2 adsorption at 20 ◦ C with curves was a cylindrical pipe (7 mm i.d.) made of quartz. The
a thermogravimetric analyzer (TG CI Electronics). The results cylindrical ACFM was tightly fitted to the internal wall of the
obtained are presented in Table 1. Data on n-butane adsorption chamber by means of Teflon strips and vacuum grease (ther-
experiments performed with the TG system, including parame- mally stable up to 210 ◦ C). In this way gas bypassing was com-
ters for the Dubinin–Radushkevich isotherm (i.e., E0 ) are also pletely avoided. The bulk parameters of the ACFM used are
indicated in Table 1. shown in Table 1.
For the analysis of dynamic n-butane adsorption, the gas
2.3. TGA experiments mixture (500–2000 ppm in He; flow rate = 150 mL/ min) was
led into the adsorption chamber by means of a set of mass flow
A differential adsorption experiment was attempted in the controllers. The composition of the gases exiting the adsorp-
thermogravimetric analyzer at 30 ◦ C. In order to approach dif- tion chamber was analyzed continuously by a mass spectrom-
ferential conditions a very low sample weight (1.475 mg) and a eter (Baltzers, Omnistar model 300O) using mass fragments
high value of n-butane gas concentration (100 vol%) were used 18 for water and 58 for n-butane. The adsorbent temperature
at the maximum flow rate allowed by the experimental setup was measured by a thermocouple inserted into the adsorption
(82 mL/min, 1 atm). The crucible used for these experiments chamber. Prior to the adsorption experiments the adsorbent was
was specially fabricated with a wired mesh to assure fast ac- heated in He at 160 ◦ C for 30 min.
4764 G. Marbán et al. / Chemical Engineering Science 61 (2006) 4762 – 4772

3. Model where e is the dimensionless amount of adsorbate in equi-


librium at Cp , and is given by the adsorption isotherm at Cp .
For modeling the dynamic behavior of ACF monoliths under Rearranging:
isothermal conditions, axially dispersed plug flow is assumed  
(DF /D  1), while mass transfer to the external surface is 4t0 Deff
implemented by a Sherwood relationship coupled with a non- 2  2 
jy Deqv j y 2 jy
instantaneous rate of adsorption. The gas phase mass balance of =   + . (6)
j qe (C0 ) je j2  j
the single-adsorptive (n-butane) in the monolithic column can p +
be calculated, in dimensionless form as (see notation below): C0 jy
  je /jy can be calculated as the slope of the adsorption
jx 1 j2 x 1 jx 1 − b qe (C0 ) j
= − − , (1) isotherm. For a D–R (Dubinin–Radushkevich) isotherm a cor-
j P ep j2  j b C0 j
rection must be made at the range of the small reduced pressures
where below the inflection point of the curve (y/ys < 1.4 × 10−16 ),
 1 in which the Henry regime must be considered to ensure ther-
=3 2 d. (2) modynamic correctness. Thus, two equations can be derived
0 considering the two zones as included below:

     
y 1 y 1
Henry regime: < exp − 2 D.R regime:  exp − 2
ys 2 ys 2

   2
   ys
1 y e = s exp −  ln , (8)
e = s exp , (7) y
42 ys

     2
    je 2 y ys
je s 1 = 2
s
ln
s
exp −  ln . (10)
= exp , (9) jy y y y
jy ys 42

The particle Peclet number for Re < 1 (experimental value:


The effective diffusivity in the pore system can be calcu-
Re ∼ 0.02) can be estimated as (Linders, 1999; Ruthven, 1984):
lated by means of the Wakao and Smith equation, in which the
a tortuosity factor was approximated to 1/p (Wakao and Smith,
Pep = , a = 0.3.0.4. (3)
ReSc 1962):
Depending on the regime and type of adsorbent, the value
2p
of (j/j) will be calculated by different expressions. Two Deff = , (11)
−1 −1
different regimes, described below, are considered: DM + DK
R1. Gas phase diffusion inside the micropores with instan-
taneous equilibrium adsorption: It is generally accepted that where the Knudsen diffusivity (DK ) is a function of the average
surface diffusion plays an important role when dealing with pore size (Doraiswamy and Sharma, 1984):
adsorption on narrow micropores (Ruthven, 1984), being how-  0.5
−8 Tads
ever less important for meso/macroporous systems (i.e., active DK = 4.85 × 10 × L(nm) × . (12)
carbon), for which gas phase diffusion inside the pore system is MW
thought to be the dominating diffusion process. Nevertheless, For the continuity between the axial concentration profile
gas phase diffusion will be considered here as a possible option (Eq. (1)) and the intraparticle concentration profile (Eq. (6)) the
for checking and comparative purposes. The appropriate form following equations are considered:
of the Fickian diffusion equation for a gas diffusion-controlled
system may be obtained from a differential mass balance on a   
j 6b Sh C0
spherical shell element: = (x − y=1 ), (13)
j Bo2 qe (C0 )
    
qe (C0 ) j jy 4t0 Deff j2 y 2 jy  
+ p = + . (4) jy DM Sh
C0 j j 2
Deqv j2  j = (x − y=1 ). (14)
j =1 Deff 2
Assuming equilibrium adsorption within the micropores:
Thus, Eq. (14) is the border equation for solving (6) and getting
j de jy the value of y=1 . With this value, the axial concentration profile
= , (5)
j dy j can be evaluated by Eqs. (13) and (1).
G. Marbán et al. / Chemical Engineering Science 61 (2006) 4762 – 4772 4765

The Sherwood number (Sh) is estimated here by the stan- In a similar way, the Darken equation can be applied to any
dard Ranz and Marshall correlation (Bird et al., 2002; Ruthven, kind of isotherm (e.g. Langmuir, Toth, etc.). Since adsorption
1984): is instantaneous:
  
Sh = 2 + 0.6Re1/2 Sc1/3 . (15) j 6b Sh C0
= (x − y=1 )
j Bo2 qe (C0 )
Although this correlation is valid for Re > 2, we have decided
= kext (x − y=1 ). (21)
to employ it because, according to Ruthven (Ruthven, 1984),
most correlations for Re < 2 usually underestimate the Sher- Combining (1) and (21):
wood number, and therefore it seems advisable to use a correla-
tion that yields the value of Sh=2 for the limiting case of Re=0 jx 1 j2 x 1 jx 1 − b
= − −
(infinite stagnant fluid around an isolated spherical particle). j P ep j2  j b
 
R2. Adsorption on highly dispersed microporous particles qe (C0 )
(i.e., fibrous monoliths): In this kind of systems it is assumed × kext (x − y=1 ). (22)
C0
that diffusion inside the fibers occurs via surface migration
along their small micropores. Different situations can be found: Coupling (17) and (21) we can obtain the border equation
that expresses the continuity between the axial and the radial
• R2.1: External mass transfer towards the external surface of profile:
the fibers, with instantaneous adsorption at the surface;    
j kext
• R2.2: Resistance to adsorption at the external surface of the = (x − y=1 )
j =1 3
fibers (non-instantaneous adsorption).   
DM Sh C0
= (x − y=1 ). (23)
It is desirable to use an expression for j/j that consider Dso 2 qe (C0 )
all the above resistances, following a similar formalism to that R2.2. Non-instantaneous surface adsorption: In this case the
used when modeling chemical processes. axial profile is expressed as in the former case by Eq. (22),
R2.1. Instantaneous surface adsorption: In this case, only the radial profile is calculated by the mass balance of Eq. (16)
external mass transfer and intraparticle diffusion control the and the external mass transfer resistance is given by Eq. (21),
process. The intraparticle diffusion resistance can be expressed whereas the resistance to external surface adsorption can be
by the following mass balance: accounted by the following equation, which assumes Lang-
 
j  j j muir kinetics and resembles that used in the old Thomas model
= 2 2 (Thomas, 1944, 1948):
j  j j
 2     
j  2 1 j j j j =1
=  + + . (16) = ka t0 C0 y=1 (s − =1 ) − . (24)
j2  j j j j Ka
Similar to Eq. (2) we can integrate (16) to obtain By considering the equilibrium isotherm, constant Ka can
  be removed from Eq. (24), which turns into
j j
= 3 . (17)  
j j =1 j =1
= ka t0 C0 s y=1 1 − , (25)
j e,=1
The dimensionless surface diffusivity now depends on the
instantaneous adsorption and can be evaluated by the thermo- where e,=1 is the dimensionless adsorbed amount in equi-
dynamic relation expressed by the Darken equation: librium with y=1 . By equaling (21) and (25) and removing
d Ln(x) y=1
= . (18)
d Ln() j kext ka t0 C0 s (1 − =1 /e,=1 )x
= . (26)
Thus, application of the Darken equation to the D–R isotherm j kext + ka t0 C0 s (1 − =1 /e,=1 )
gives

     
 1  1
Henry regime: < exp − 2 D.R regime:  exp − 2
s 4 s 4
1
= 1, = ,
2[Ln(s /)]1/2
j j 1
= 0, (19) = [Ln(s /)]−3/2 . (20)
j j 4
4766 G. Marbán et al. / Chemical Engineering Science 61 (2006) 4762 – 4772

The value of e,=1 for a given value of y=1 can be evaluated


for the D–R isotherm as follows:
   
y=1 1 y=1 1
Henry regime: < exp − 2 D.R regime:  exp − 2
ys 2 ys 2

   2
  e,=1 ys
e,=1 y=1 1 = exp −  ln . (28)
= exp , (27) s y=1
s ys 42

Again, for a more simple resolution, one can assume j


=0 for  = 0. (39)
y=1 j
e,=1 = . (29)

Resolution of the system of equations for regimes R1 and R2


when
= 1, this equation takes into consideration the low was performed by means of an ad hoc-designed Visual Basic
adsorptive concentration (dilute gas stream) and thus assumes macro included in an Excel datasheet. For discretizing the ra-
a Henry regime at the external surface, qe = KC, in which the dial profile a three-point-based generalized three-level scheme
Henry constant is was used, whereas the program allowed the procedure for dis-
cretizing the axial profile to be selected among several implicit
qe (Cs ) schemes (backwards difference, Crank Nicholson, three level
K= exp{−[ ln(Cs /C0 )]2 }. (30)
C0 schemes) by using either three-point or five-point derivatives.
In this way, the D–R isotherm is considered also at the exter- For solving the set of non-linear coupled equations the Thomas
nal surface when the breakthrough profile is fully developed. algorithm was used together with a Newton–Raphson iterative
The D–R isotherm can also be considered for any value of y=1 procedure.
by calculating the appropriate value of
in any iteration step.
Coupling (17) and (26) the border equation that expresses 4. Results and discussion
the continuity between the axial and the radial profile can be
obtained as: 4.1. Differential adsorption
 
j kext ka t0 C0 s (1 − =1 /e,=1 )x Fig. 1 shows the experimental variation with time of the rel-
= . (31)
j =1 3[kext + ka t0 C0 s (1 − =1 /e,=1 )] ative mass increase () for ACFM obtained in the TGA at near
differential conditions (see experimental section). As compari-
This general expression can be simplified by using Eq. (29). son, the results obtained by other authors (Linders et al., 2001)
In this way, Eqs. (21), (25) and (29) permit to obtain the fol- with a commercial active carbon are also included (Kureha ac-
lowing expression for calculating y=1 : tive carbon (10.8 mg), 3.7 vol% n-butane/He, F =200 mL/ min,
kext x + ka t0 C0 s (
=1 ) 1 atm, 30 ◦ C, experiment performed in a TEOM).
y=1 = , (32) According to the results displayed in the figure, both the ac-
kext + ka t0 C0 s tive carbon pellets and the ACFM seem to attain equilibrium at
with this, Eq. (31) becomes a similar rate. However, as commented before, in the case of
  the ACFM, the data were obtained with a TGA that had specific
j kext ka t0 C0 s (x −
=1 ) structural features that made it impossible the establishment of
= . (33)
j =1 3[kext + ka t0 C0 s ] strict differential conditions. Therefore, under strict differential
conditions it should be expected that the ACFM adsorbent pre-
It can be noticed that this equation becomes Eq. (23) when sented a faster rate of n-butane adsorption than that shown in
ka = ∞. The rest of border equations are the following: Fig. 1 (experimental results) and consequently also faster than
x=0 for  < 0 and for all , (34) that reported for Kureha AC. As will be seen now, the results
of the model confirm this expectation.
x = x0 for 0 and  = 0[or x = x0 × (t/ti ) for t < ti By applying the model at differential conditions with the ki-
for linear increase of feed concentration], (35) netic parameters that better fit the dynamic breakthrough pro-
files for the ACFM (see Section 4.2 Breakthrough profiles)
=0 for  < 0 and for all , (36) curves A (obtained with values of gas flow and concentration as
jx those used for differential experiment with ACFM) and B (ob-
=0 for  = 0, (37) tained with values of gas flow and concentration as those used
j
for differential experiment with Kureha active carbon (Linders
jx et al., 2001)) in Fig. 1 were obtained (surface diffusion, Darken
=0 for  = 1, (38)
j equation applied to D–R isotherm, Dso = 1.5 × 10−12 m2 s−1 ,
G. Marbán et al. / Chemical Engineering Science 61 (2006) 4762 – 4772 4767

C Table 3
1.0 Expressions of evaluated by the Darken equation for different isotherms
A,D
Isotherms (Ds /Dso )
0.8 B
θ [x for curve A]

Constant diffusivity 1 (19)


0.6
1
Dubinin–Radushkevich (20)
0.4 2[Ln((s /)]1/2
(D–R) isotherm (Eq. (9))
ACFM
0.2
Kureha active carbon (Linders et al., 2001)
1
Langmuir isotherm (40)
(1 − /s )
0.0
0 5 10 15 20
time (s) 1
Chen and Yang for diffusion (41)
(1 − (1 − )/s )
Fig. 1. Differential adsorption experiments performed with the ACFM in
100 vol% n-butane flow. Data for Kureha active carbon were taken from (Chen and Yang, 1992)
(Linders et al., 2001). Also included the results of the model applied to
ACFM at differential conditions (curves A to D). ( = 0.5 for calculations)

1
Freundlich isotherm (42)
Table 2 n(1 − /s )
Parameter values of simulation curves plotted in Fig. 1 (square n-butane (n = 0.6 for calculations)
concentration steps)

Curve C0 Flow rate Diffusion Diffusivity ka (m3 mol−1 s−1 ) 1


(ppm) (mL/min) type (m2 s−1 ) Toth isotherm (43)
1 − (/s )t
A 106 82 Surface (D–R) 1.5 × 10−12 20
(t = 2 for calculations)
B 37000 200 Surface (D–R) 1.5 × 10−12 20
C 106 82 Gas 3.0 × 10−8 ∞
D 106 82 Surface (D–R) 1.5 × 10−12 ∞

centration in the gas phase used for evaluating the  curve at


ka = 20 m3 mol−1 s−1 ). This procedure is contrary to that gen- differential conditions (C0 = 106 ppm).
erally followed, which consists of firstly evaluating the intra- Application of the model at differential conditions can of-
particle diffusion parameters via fitting the experimental curves fer us two practical conclusions for the modeling of the break-
obtained at differential conditions and then predicting the dy- through profiles. First of all, the surface diffusion-controlled
namic profiles with the so-obtained parameters. However, it is regime can be analyzed by considering different adsorption
followed in this work because of the impossibility to use dif- isotherms which will relate in different ways the surface dif-
ferential conditions with the available TGA (Table 2). fusivity with the amount of n-butane adsorbed at any moment
Curves A and B clearly show that ACFM adsorbent be- inside the pore system. This relationship is performed through
comes fully loaded at differential conditions in less than 3 s, the Darken equation (18). In this work, six expressions of
which is an indication of the high adsorption rate attained corresponding to different situations and isotherms listed in
with this microporous material. Only when considering in the Table 3 have been analyzed.
model gas phase diffusion inside the pores a faster adsorp- In principle, all these expressions of should yield different
tion rate is obtained (curve C). Obviously the latter is not trends in the resulting plot of  versus time for a given initial
an operative regime in the microporous adsorbent, as will be value of surface diffusivity (Dso ). This is clearly observed in
demonstrated with the simulated breakthrough profiles obtained Fig. 2a. Except for the case of constant diffusivity, all expres-
under gas phase diffusion conditions (Section 4.2 Breakthrough sions of predict an increasing value of Dso with the local
profiles). amount of n-butane adsorbed (). This is correspondingly re-
It is also interesting to note that at differential conditions flected in the trends observed in Fig. 2a, for which the slow-
there seems to be no influence of the external adsorption re- est rate to attain equilibrium belongs precisely to the case of
sistance (ka ) on the results, as deduced from the overlapping constant diffusivity. The fastest rate is obtained when the D–R
of curves A (ka = 20 m3 mol−1 s−1 ) and D (ka = ∞; Fig. 1), isotherm applies, as is the case for highly microporous adsor-
although a clear influence was observed under dynamic con- bents such as ACF monoliths. On the other hand, from a practi-
ditions on the breakthrough profiles (see Section 4.2 Break- cal point of view it is not clear whether the differences observed
through profiles). As will be proved at the end of this section, in Fig. 2a are an important modeling issue or not, since all
this is just a consequence of the high value of n-butane con- isotherms seem to produce converging results for specific values
4768 G. Marbán et al. / Chemical Engineering Science 61 (2006) 4762 – 4772

6 -12
Co= 1x10 Co= 1x106 Dso= 1.50x10 ka = 0.1
1.0 1.0 Dso= 1.50x10-12
ka ≥ 10

0.8 0.8

0.6 0.6
D.R.

θ
Co= 1x106 Co= 1x106
θ

Langmuir 0.4 Dso= 1.50x10-13 Dso= 1.50x10


-13
0.4
Freundlich ka ≥ 10 ka = 0.1
Square step Co= 500 Co= 500
Toth 0.2 Dso= 1.50x10-12 Dso= 1.50x10
-12
0.2 Dso=1.5×10-12 ka = 20
Yang ka = ∞ Square step
(ka=∞) D-R
Constant 0.0
0.0 0 1 2 3 4 5
0 1 2 3 4 5 time (s)
(a) time (s)
Fig. 3. Application of the model at differential conditions for different values
of C0 , Dso and ka .

1.0

Square step
0.8 ka . As observed, at high values of gas concentration (106 ppm),
(ka=∞)
the resistance to external surface adsorption exerts no effect
on the model results for values of ka over 10 m3 mol−1 s−1 ,
0.6
Dso=1.50×10-12 whereas for low values of gas concentration (500 ppm) clear
D.R.
differences are observed between the curves corresponding to
θ

Dso=2.00×10-12
0.4 Langmuir
-12 ka = 20 m3 mol−1 s−1 and ka = ∞ (dashed curves in Fig. 3).
Freundlich Dso=2.75×10
Dso=3.50×10-12
This fact has a clear mathematic explanation since parameter
Toth
0.2 Co=1×106 ppm C0 is multiplying ka in the equation that expresses the adsorp-
Yang Dso=9.50×10-12
Co=500 ppm
tion resistance (Eq. (33)). Due to the microporous character of
Constant Dso=1.60×10-12
0.0
ACFM adsorbents, they are thought to have excellent perfor-
0.0 0.5 1.0 1.5 2.0 2.5 3.0 mance in the removal of contaminants at low concentration.
(b) time (s) Therefore, according to the results shown in Fig. 3, any possible
external adsorption resistance will be detrimental at these spe-
Fig. 2. Application of the model at differential conditions for: (a) different cific conditions, whereas it can be unnoticed in environments
adsorption isotherms at a given value of Dso (C0 = 106 ppm); (b) different with a high concentration of the adsorptive.
adsorption isotherms at different values of Dso and C0 .
Since adsorption at the external surface and surface diffusion
inside the pores of the fibers are serial processes, the external
of Dso . This is appreciated in Fig. 2b, in which values of Dso adsorption resistance has a less marked effect on the model
were found in such a way that all expressions of yielded an results for lower values of surface diffusivity, as can be clearly
almost identical curve (solid lines for C0 = 106 ppm). However, observed in Fig. 3. Finally, from the modeling point of view,
when using a much lower gas concentration (C0 =500 ppm), the it is not possible to find a combination of [Dso , ka ] values that
same values of Dso produced very different curves, as indicated produces the same curve as that obtained under instantaneous
by dashed lines in Fig. 2b. Under dynamic conditions, before adsorption and, therefore, the existence of non-instantaneous
the breakthrough is attained, the gas phase inside the monolith adsorption is a clear modeling issue that can eventually be used
may present values from C = 0 to C0 . Therefore, it is obvious to improve the curve fitting of the breakthrough profile.
that the breakthrough profiles will be affected by the associated
differences in adsorption at varying values of gas concentration 4.2. Breakthrough profiles
(Fig. 2b) which permits to conclude that selection of the cor-
rect expression for the isotherm to be employed in the model is Fig. 4 shows the results of applying the model to ACFM
a key modeling issue. In this work the D–R isotherm was used (C0 = 500 ppm) under two different gas diffusion models. As
for fitting the experimental breakthrough profiles. The values of expected (see previous section), the curve corresponding to
the parameters of the D–R equation (Table 1) were determined gas phase diffusion inside the micropore system (regimen R1)
gravimetrically in specific experiments of n-butane adsorption. cannot reproduce the experimental results. Changing the effec-
A second observation, already pointed when commenting tive diffusivity to lower values, apart from lacking of physical
curves A and D of Fig. 1, refers to the effect of external adsorp- significance since Eq. (11) is not accomplished any more, nei-
tion resistance in the shape of the plot of  versus time. Fig. 3 ther permitted to simulate the actual shape of the experimental
shows the simulated curves obtained by applying the model un- breakthrough profile. The other simulated curve displayed
der differential conditions for different values of C0 , Dso and in Fig. 4 was obtained by considering surface diffusion
G. Marbán et al. / Chemical Engineering Science 61 (2006) 4762 – 4772 4769

Surface diffusion (D-R)


1.0 Dso=1.5x10-12m2s-1 ka
1.0 ka= 20 m3mol-1s-1
ka=∞
3
ka= 20 m mol s -1 -1
0.8
0.8 ka= ∞
0.6
xξ=1

Gas phase diffusion

xξ=1
0.6 Co= 500 ppm
0.4 Deff=1.38x10-8m2s-1
ka=∞ Dso=1.5×10-12m2s-1
0.4
0.2 s
Co=500 ppm Experimental
0.0 0.2 Experimental
700 900 1100 1300 1500 Model
0.0
time (s)
700 900 1100 1300 1500
Fig. 4. Experimental and simulated breakthrough profiles obtained for different (a) time (s)
diffusion models (gas phase diffusion and surface diffusion).

1.2

Dso ka= 20 m3mol-1s-1 ka


1.0
Dso=7.0×10-13m2s-1 ka= ∞
1.0
Dso =1.0×10-12m2s-1 0.8
0.8 Dso=1.5×10-12m2s-1
xξ=1

0.6 Co= 2000 ppm


0.6
xξ=1

Dso=1.5×10-12m2s-1
Co=500 ppm 0.4
0.4
s ka=∞
0.2 ka Experimental
0.2 Experimental
Model
Model
0.0
0.0
250 300 350 400 450 500
700 900 1100 1300 1500
(b) time (s)
(a) time (s)
Fig. 6. Experimental and simulated breakthrough profiles obtained at different
values of ka : (a) for C0 = 500 ppm; and (b) for C0 = 2000 ppm.
Dso=7.0×10-13m2s-1 Dso
1.0
Dso =1.0×10-12m2s-1
deal with the high-concentration gradient produced at the ex-
0.8 Dso=1.5×10-12m2s-1
ternal surface in such a way that the whole breakthrough curve
could be resolved in a reasonable time. Added to this, D–R
0.6
xξ=1

adsorption should only be considered inside the pore system


Co=2000 ppm of the particles and not at their external surface.
0.4
ka=∞ A value of initial surface diffusivity of 1.5 × 10−12 m2 /s
produced a simulated breakthrough profile that matched the
0.2 Experimental
Dso first zone of the experimental curve (up to x=1 ∼ 0.5) with
Model a significant degree of agreement (Fig. 4). This value of sur-
0.0
250 300 350 400 450 500
face diffusivity is higher than those reported for n-butane
time (s)
adsorption in molecular sieves (zeolites and molecular sieve
(b)
carbons) which are around 1 × 10−14 m2 /s (Ruthven, 1984).
Fig. 5. Experimental and simulated breakthrough profiles obtained at different Nevertheless, the average pore size of the ACFM is also
values of Dso : (a) for C0 = 500 ppm; and (b) for C0 = 2000 ppm. somewhat higher (∼ 1 nm) than that existing in typical molec-
ular sieves (∼ 0.5 nm) so that a higher surface diffusion rate is
expected for the former.
(D–R)-controlled regime, with Henry regime at the external However, there is a patent discrepancy between the simulated
surface (Eq. (29)), and instantaneous adsorption at the exter- breakthrough profile and the experimental curve for values of
nal surface (ka = ∞, regime R2.1). When instantaneous D–R x=1 over ∼ 0.5 (Fig. 4). Apart from Dso and ka , the rest of the
adsorption was considered at the external surface (Eq. (9)) the model parameters are determined by the adsorption conditions,
constant step resolution used in the VB program avoided to the adsorbent characteristics and few standard correlations
4770 G. Marbán et al. / Chemical Engineering Science 61 (2006) 4762 – 4772

Experimental 5. Conclusions
Model
1.0
In this work a model to predict the breakthrough profile
0.8 of the adsorption n-butane by microporous fibrous adsorbents
Co = 2000 ppm

Co = 1500 ppm

Co = 1000 ppm
(ACFM) is proposed. The model includes non-instantaneous

Co = 500 ppm
0.6
xξ=1

adsorption at the external surface and surface diffusion inside


0.4 the pores of the fibers. It has been proved that the external
surface resistance to adsorption for the lower concentrations
0.2
cannot be neglected. The selection of the appropriate expres-
(a) 0.0 sion for the isotherm to be employed in the model is a key
Dso =1.5×10-12m2s-1
issue to obtain a suitable fit between the experimental data and
ka= 20 m3mol-1s-1 the model, especially for the low-concentration experiments.
1.2E-02
The model solved for non-instantaneous surface adsorption and
dXξ=1/dt(s-1)

Experimental D–R-based surface diffusivity in the pore system adequately


Model
8.0E-03 simulates the breakthrough profiles for adsorption of n-butane
in diluted streams, since a good agreement between the experi-
mental and the simulated data is obtained for the whole profiles
4.0E-03
at different values of n-butane concentration in the inlet gases.

0.0E+00
0 500 1000 1500 Notation
(b) time (s)
D v
Fig. 7. Experimental and best fitting simulated breakthrough profiles obtained
B0 Re Sc = DeqvM s , Bodenstein number according to
at different values of C0 : (a) Breakthrough profiles; and (b) breakthrough Levenspiel (Levenspiel, 1996)
rates. C adsorptive concentration in the gaseous bulk at
an axial position z in the monolith (0 < z < H ),
(i.e., Sherwood and Peclet numbers). Therefore, it was decided mol mgas−3
to make a survey on Dso and ka in an attempt to produce a C0 adsorptive concentration in the gaseous bulk at
better fitting of the upper zone of the breakthrough profile with −3
z < 0, mol mgas
the following conditions: (1) not to provoke a worsening of the Cp adsorptive concentration in the intraparti-
fitting in the lower zone of the breakthrough profile and (2) to cle gaseous phase at an absolute position
permit simultaneously a reasonable fitting of the breakthrough −3
(r, z)(0 < r < Deqv /2; 0 < z < H ), mol mgas
profiles obtained at higher n-butane concentrations in the inlet Cs saturation adsorptive concentration at
gas (1000, 1500 and 2000 ppm). −3
Tads , mol mgas
Fig. 5a shows that a decrease in the value of initial surface D monolith diameter, m
diffusivity allows the curve fitting to be improved in the upper Dax axial diffusivity, m2 s−1
zone of the breakthrough profile corresponding to the experi- Deff effective diffusivity in the pore system, m2 s−1
ment performed at C0 = 500 ppm, without provoking a signifi- Deqv equivalent diameter of the fibers = 1.5 × DF , m
cant worsening of the fitting for x=1 < 0.5. However, the same DF fiber diameter, m
procedure applied to the profile obtained at C0 = 2000 ppm DK Knudsen diffusivity, m2 s−1
produced a significant deterioration of the curve fitting in both DM molecular diffusivity, m2 s−1
zones of the profile (x=1 < 0.5 and x=1 > 0.5), as clearly ob- Dso surface diffusivity in the pore system at  = 0,
served in Fig. 5b. m2 s−1
Introducing in the model some resistance to n-butane adsorp- E0 characteristic energy of the adsorbent, J mol−1
tion on the external surface of the fibers, or, in other words, H monolith height, m
decreasing the value of ka coefficient, also permitted to im- 3 m −3
K Henry constant, mgas part
prove the curve fitting in the upper zone of the breakthrough
ka external surface adsorption coefficient,
profile obtained at C0 = 500 ppm (Fig. 6a), but in this case the 3 mol−1 s−1
mgas
profiles obtained at higher values of C0 were also reasonably 3 m −2 s−1
well simulated (Fig. 6b). It was found in this way that values kext external mass transfer coefficient, mgas part
for Dso and ka of 1.5 × 10−12 m2 s−1 and 20 m3 mol−1 s−1 , MW adsorptive molecular weight, g mol−1
Deqv vs
respectively, produced a satisfactory fitting of the profiles ob- Pep Dax b , Bodenstein or particle Peclet number
tained at the different values of n-butane gas concentration, as according to CRC Handbook (CRC, 1984). Axial
shown in Fig. 7. It is worth noting that the time derivative of dispersion number according to Levenspiel
x=1 is also well reproduced by the model for the different val- (Levenspiel, 1996)
ues of C0 (Fig. 7b), which expresses the authentic goodness of q instantaneous amount of adsorbate, per adsorbent
−3
the curve fitting. volume, at an absolute position (r, z), mol mpart
G. Marbán et al. / Chemical Engineering Science 61 (2006) 4762 – 4772 4771

qe (Cp ) amount of adsorbate in equilibrium at Cp , per ad-  r


(Deqv /2 ),dimensionless radial position in the fibers
−3
sorbent volume, mol mpart (0 <  < 1)
g gas phase density, g mgas −3
qe (C0 ) amount of adsorbate in equilibrium at C0 , per ad-
−3 4t0 Dso
sorbent volume, mol mgas  D2
, surface diffusion group
eqv
qe (Cs ) amount of adsorbate in equilibrium at Cs , per ad-  t
−3
sorbent volume, mol mpart t0 , dimensionless time
r radial position in the particle, m
R universal gas constant, 8.31441 J K −1 mol−1 Acknowledgments
g vs Deqv
Re , Reynolds number
TVS acknowledges the CSIC-ESF for the award of an I3P
Sc g DM , Schmidt number postdoc contract.
Deqv kext
Sh DM , Sherwood number
H 2 b
References
t0 Deqv vs , time constant, s
t time, s Bird, R.B., Stewart, W.E., Lightfoot, E.N., 2002. Transport Phenomena. Wiley,
Tads adsorption temperature, K New York.
Burchell, T.D., Judkins, R.R., 1997. A novel carbon fiber based material
vs gas superficial velocity, m s−1 and separation technology. Energy Conversion and Management 38,
C
x C0 , dimensionless adsorptive concentration in the S99–S104.
gas phase at z Burchell, T.D., Judkins, R.R., Rogers, M.R., Williams, A.M., 1997. A novel
x=1 dimensionless adsorptive concentration in the gas process and material for the separation of carbon dioxide and hydrogen
sulfide gas mixtures. Carbon 35, 1279–1294.
phase at z = H ( = 1)
Burchell, T.D., Weaver, C.E., Chilcoat, B.R., Derbyshire, F., Jagtoyen, M.,
y=1 dimensionless adsorptive concentration in the in- 2000. Activated carbon fiber composite material and method of making.
traparticle gaseous phase at the external surface of US Patent 6030698, Lockheed Martin Energy Research Corporation, Oak
the fibers ( = 1) Ridge, TN, US.
Cp Centeno, T.A., Marbán, G., Fuertes, A.B., 2003. Importance of micropore
y C0 , dimensionless adsorptive concentration in the size distribution on adsorption at low adsorbate concentrations. Carbon
intraparticle gaseous phase at an absolute position 41, 843–846.
(, ) Chen, Y.D., Yang, R.T., 1992. Predicting binary Fickian diffusivities from pure-
Cs component fickian diffusivities for surface-diffusion. Chemical Engineering
ys C0 , ratio of Cs to C0
Science 47, 3895–3905.
z axial position in the monolith, m Cheng, T.B., Jiang, Y., Zhang, Y.P., Liu, S.Q., 2004. Prediction of breakthrough
curves for adsorption on activated carbon fibers in a fixed bed. Carbon
Greek letters 42, 3081–3085.
Deqv CRC, 1984. CRC Handbook of Chemistry and Physics. 64th edn. CRC Press,
 H , dimensionless size of the monolith Boca Raton, FL.
affinity coefficient of the adsorbate, Das, D., Gaur, V., Verma, N., 2004. Removal of volatile organic compound
RT ads
 E , dimensionless group of the D–R isotherm
by activated carbon fiber. Carbon 42, 2949–2962.
0 DeLiso, E.M., Gadkaree, K.P., Mach, J.F., Streicher, K.P., 1997. Carbon-
Ds ()
Dso , dimensionless surface diffusivity coated inorganic substrates, US Patent 5597617. Corning Inc, Corning,
b bed porosity, NY, US.
p particle porosity, Derbyshire, F., Jagtoyen, M., Andrews, R., Rao, A., Martín-Gullón, I., Grulke,
q E.A., 2001. Carbon Materials in environmental applications. In: Radovic,
 qe (C0 ) , ratio of the amount of adsorbate at an ab- L.R., (Ed.), Chemistry and physics of carbon, vol. 27, Marcel Dekker,
solute position (r, z) to the equilibrium amount of New York, pp. 1–66.
adsorbate in equilibrium at C0 Doraiswamy, L.K., Sharma, M.M., 1984. Heterogeneous Reactions: Analysis,
qe (Cp ) Examples and Reactor Design. Wiley, New York.
e qe (C0 ) , ratio of the amount of adsorbate in equilib- Foster, K.L., Fuerman, R.G., Economy, J., Larson, S.M., Rood, M.J., 1992.
rium at Cp to the equilibrium amount of adsorbate Adsorption characteristics of trace volatile organic compounds in gas
in equilibrium at C0 streams onto activated carbon fibres. Chemistry of Materials 4, 1068–1073.
qe (Cs )
s qe (C0 ) , ratio of the amount of adsorbate in equilib-
Fuertes, A.B., Marbán, G., Nevskaia, D.M., 2003. Adsorption of volatile
organic compounds by means of activated carbon fibre-based monoliths.
rium at Cs to the equilibrium amount of adsorbate
Carbon 41, 87–96.
in equilibrium at C0 Gadkaree, K.P., 2001. Method of making activated carbon bodies having
 dynamic conditions: average value of  at z; dif- improved adsorption properties. US Patent 6187713. Corning Inc., Corning,
ferential conditions: average ratio of the amount of NY, US.
adsorbate in the fibers to the equilibrium amount of Lee, J.C., Lee, B.H., Kim, B.G., Park, M.J., Lee, D.Y., Kuk, I.H., Chung,
H., Kang, H.S., Lee, H.S., Ahn, D.H., 1997. The effect of carbonization
adsorbate in equilibrium at C0 [equivalent to frac- temperature of PAN fiber on the properties of activated carbon fiber
tional mass increase (from 0 to 1) during adsorp- composites. Carbon 35, 1479–1484.
tion] Levenspiel, O., 1996. The Chemical Reactor Omnibook. OSU Book Stores,
gas viscosity, g m−1 s−1 Inc., Corvallis, Oregon.
 z
H , dimensionless axial position in the monolith
Linders, M.J.G., 1999. Prediction of breakthrough curves of activated carbon
based sorption systems: from elementary steps to process design. Ph.D.
(0 <  < 1) Dissertation. Delft University of Technology, Delft, The Netherlands.
4772 G. Marbán et al. / Chemical Engineering Science 61 (2006) 4762 – 4772

Linders, M.J.G., van de Broeke, L.J.P., Nijhuis, T.A., Kapteijn, F., Moulijn, Ruthven, D.M., 1984. Principles of Adsorption and Adsorption Processes.
J.A., 2001. Modelling sorption and diffusion in activated carbon: a novel Wiley, New York.
low pressure pulse-response technique. Carbon 39, 2113–2130. Suzuki, M., 1991. Application of fiber adsorbents in water-treatment. Water
Linders, M.J.G., Mallens, E.P.J., van Bokhoven, J.J.G.M., Kapteijn, F., Science and Technology 23, 1649–1658.
Moulijn, J.A., 2003. Breakthrough of shallow activated carbon beds under Thomas, H.C., 1944. Heterogeneous ion exchange in a flowing system. Journal
constant and pulsating flow. AIHA Journal 64, 173–180. of the American Chemical Society 66, 1664–1666.
Marbán, G., Antuña, R., Fuertes, A.B., 2003. Low-temperature SCR of NOx Thomas, H.C., 1948. Chromatography—a problem in kinetics. Annals of the
with NH3 over activated carbon fiber composite-supported metal oxides. New York Academy of Sciences 49, 161–182.
Applied Catalysis B: Environmental 41, 323–338. Valdés-Solís, T., Linders, M.J.G., Kapteijn, F., Marbán, G., Fuertes, A.B.,
Marbán, G., Fuertes, A.B., 2004. Co-adsorption of n-butane/water vapour 2004. Adsorption and breakthrough performance of carbon-coated ceramic
mixtures on activated carbon fibre-based monoliths. Carbon 42, 71–81. monoliths at low-concentration n-butane. Chemical Engineering Science
Marbán, G., Fuertes, A.B., Nevskaia, D.M., 2000. Dry formation of low- 59, 2780–2791.
density Nomex(TM) rejects-based activated carbon fiber composites. Vilaplana-Ortego, E., Alcañiz-Monge, J., Cazorla-Amorós, D., Linares-Solano,
Carbon 38, 2167–2170. A., 2002. Activated carbon fibre monoliths. Fuel Processing Technology
Muto, A., Bhaskar, T., Tsuneishi, S., Sakata, Y., 2005. Activated carbon 77–78, 445–451.
monoliths from phenol resin and carbonized cotton fiber for methane Wakao, N., Smith, J.M., 1962. Diffusion in catalyst pellets. Chemical
storage. Energy & Fuels 19, 251–257. Engineering Science 17, 825–834.

Você também pode gostar