702 Langmuir 2006, 22, 702-708

Modeling Water Adsorption in Carbon Micropores: Study of Water

in Carbon Molecular Sieves
S. W. Rutherford*
Los Alamos National Laboratory, Engineering Sciences and Applications DiVision, MS E581,
Los Alamos, New Mexico 87545

ReceiVed July 6, 2005. In Final Form: September 16, 2005

Measurements of water adsorption equilibrium in a carbon molecular sieve are undertaken in order to gain insight
into the nature of water adsorption in carbon micropores. The measurements are taken at low concentrations to
emphasize the role of oxygen-containing functional groups in the adsorption of water. Comparisons are made with
previously published water adsorption data at higher concentrations to provide a data set spanning a wide range of
loading. The assembled data set provides an opportunity for comparison of various theories for prediction of water
adsorption in carbon micropores. Shortcomings of current theories are outlined, and an analytical theory that is free
of these deficiencies is proposed in this investigation. With the consideration of micropore volume and pore size
distribution, the experimental data and proposed isotherm model are consistent with previous studies of Takeda carbon
molecular sieves. Also investigated is the uptake kinetics of water, which is characterized by a Fickian diffusion
mechanism. The Maxwell-Stefan formulation is applied to characterize the dependence of the diffusional mobility
upon loading.

1. Introduction obtained in other investigations in order to provide an isotherm

model that can characterize equilibrium over a wide range of
Microporous carbon is a material that plays a key role in many
concentration and that is consistent with previous studies of
commercially important technologies ranging from catalyst
Takeda carbon molecular sieves.
support, adsorbent, membrane, electrolytic material, abrasion-
resistant material in magnetic storage media, and mechanical 2. Background
reinforcement material in composites. Recently, new forms of
Carbon molecular sieves are useful materials in the gas
microporous carbon have been derived, including specialty fibers,
separation and purification industry because they have finely
nanotubes, nanohorns, and buckminsterfullerenes.1 The potential
tuned pore size distributions that allow them to selectively adsorb
impact of these microporous materials on existing and emerging
molecules of size smaller than the pore width while excluding
technologies provides an enormous opportunity for research and
molecules of larger size.5,6 Molecular sieves used for the
development. Some research efforts have been directed toward
separation of air can exclude nitrogen molecules (of kinetic size
an understanding of the interaction of these carbon-based materials
0.364 nm) from the micropores while allowing oxygen molecules
with their gaseous environments because this interaction can
(of kinetic size 0.346 nm) to penetrate. It has been proposed that
affect performance. Water, in particular, is a common component
a pore mouth barrier is responsible for the size exclusion effect.7
that interacts with microporous carbon in many of these
In the case of CMS used in air separation, known as CMS 3A,
applications. However, the interaction of water molecules and
the size of this barrier is expected to be on the order of 3 Å.
carbon surfaces is a phenomenon that is still not well understood.2,3
According to Juntgen et al.,8 hydrocarbon cracking is the crucial
For example, many theories have been proposed to characterize
step in the manufacture of these materials that yields the pore
the adsorption equilibrium of water in microporous carbon, but
mouth barrier. The carbon that is deposited from this cracking
shortcomings are evident when applied to experimental data.4 It
step is assumed to reside at the entrance of the micropore and
is the purpose of this investigation to highlight the strengths and
not significantly penetrate the micropore itself.7 Under this
limitations of these theories and to offer a model that is free of
assumption, the pore size distribution of the carbon would be
the proposed deficiencies. The model is applied to experimental
essentially unaltered by the cracking treatment. A reduction may
water adsorption data measured on carbon molecular sieves
be evident in the measured total micropore volume per gram of
(CMS) manufactured by the Takeda chemical company. This
sample because the deposited carbon adds additional weight to
material is chosen because it is a well-studied microporous carbon
the sample. However, this has not been widely investigated
with a well-defined pore size distribution (PSD). The measure-
because of the difficulty posed by CMS 3A in applying the
ments are taken at low pressure to emphasize the role of oxygen-
standard technique of pore size determination using nitrogen
containing functional groups in the adsorption of water.
adsorption at 77 K. At 77 K, nitrogen molecules fail to penetrate
Comparisons are made to higher-pressure water adsorption data
the material on a reasonable time scale, thereby rendering the
technique unusable. Nevertheless, Takeda CMS 3A and CMS
* E-mail: stevenr@[Link]. Phone: (505) 6640812. Fax: (505) 6640815.
(1) Carbon Materials for AdVanced Technologies; Burchell, T. D., Ed.;
Pergamon: New York, 1999. (5) Ruthven, D. M. Principles of Adsorption and Adsorption Processes;
(2) Gregg, S. J.; Sing, K. S. W. Adsorption, Surface Area, and Porosity, 2nd Wiley: New York, 1984.
ed.; Academic Press: New York, 1982. (6) Yang, R. T. Adsorbents: Fundamentals and Applications, Wiley: New
(3) Rouquerol, F.; Rouquerol, J.; Sing, K. Adsorption by Powders and Porous York, 2003.
Solids; Academic Press: London, 1999. (7) Karger, J.; Ruthven, D. M. Diffusion in Zeolites and Other Microporous
(4) Brennan, J. K.; Bandosz, T. J.; Thomson, K. T.; Gubbins, K. E. Colloids Solids; Wiley: New York, 1992.
Surf., A 2001, 187-188, 539-568. (8) Juntgen, H.; Knoblauch, K.; Harder, K. Fuel 1981, 60, 817.

10.1021/la051826n CCC: $33.50 © 2006 American Chemical Society

Published on Web 12/15/2005
Study of Water in Carbon Molecular SieVes Langmuir, Vol. 22, No. 2, 2006 703

carbon are oxygen-containing functional groups that may include

carboxylic groups (as considered by Turner et al.15), but there
are also significant carbonyl and hydroxyl groups in coconut
shell-derived carbon.16 The functional groups are believed to be
located at the edges of the graphene layers that stack to form the
micropore structure in the CMS.17 Hence, these sites provide an
opportunity for water to bond outside of the stacked graphene
structure. Additionally, it has been shown by molecular simulation
that water may also adsorb within the micropores formed from
the spacing between the stacked graphene layers that offer
attraction for the water molecules. In fact, it is shown that small
micropores offer a larger attraction for water than do larger
micropores.18,19 Molecular simulations also show that a threshold
of 1 nm exists where pores smaller than this size can accommodate
only a single layer of water molecules.18,19 Applied to Takeda
3A and 5A CMS, this would imply that monolayer adsorption
of water occurs within these materials. Due to the similar PSD
for Takeda 3A and 5A and the expectation of simple monolayer
Figure 1. Corrected Dubinin-Radushkevich plot of the adsorption water formation, the molecular sieves provide useful materials
of carbon dioxide in Takeda 3A and Takeda 5A at 20 °C. to experimentally and theoretically study the phenomenon of
water adsorption in micropores. Water molecules are also useful
5A, which are employed for the separation of larger molecules
probes in assessing ultramicroporosity in carbon molecular sieves
on the order of 5 Å, have been studied by other techniques and
due to small molecular size that allow them to penetrate
appear to display differing total micropore volumes. Horvath
micropores without interference from the pore mouth barrier of
and Kawazoe9 calculated the PSD for Takeda 5A from measure-
CMS.
ments of nitrogen adsorption. The PSD was found to have a
modal value at around 0.5 nm and negligible micro/mesopores
3. Water Adsorption in Takeda 5A
of size greater than 1 nm, a result later verified by methyl chloride
adsorption.10 Rutherford et al.11 calculated a similar PSD for Water adsorption in Takeda 5A samples has been measured
Takeda 3A from methane adsorption data and obtained a modal by volumetric adsorption conducted differentially as is outlined
value at 0.5 nm and negligible micro/mesopores of size greater in a previous investigation.20 The Takeda 5A samples have the
than 1 nm. Additionally, Cazorla-Amoros et al.12 obtained carbon following characteristics: the pellets are 0.3 cm in diameter and
dioxide isotherms for Takeda 5A and 3A and showed a larger 0.3 cm in length and have been characterized by mercury
saturation capacity for Takeda 5A at around 6.9 mmol/g compared porosimetry yielding a total macroporosity at 28% and an average
to 4.2 mmol/g for Takeda 3A. This corresponds to a total macropore size on the order of 0.5 micron.13
micropore volume of 0.31 (cm3 of micropore volume)/g for The measurement of the adsorption equilibrium of water in
Takeda 5A and 0.19 (cm3 of micropore volume)/gram for Takeda these pellets was undertaken at 20 °C from six independent
3A (using a carbon dioxide density of 1.023 g/cm3 12). Using experiments separated by an outgassing period of 72 h under
these values, the data for carbon dioxide adsorption in Takeda ultrahigh vacuum at 20 °C. The results are shown in Figure 2
3A and 5A obtained by Rutherford and Do13,14 can be recalculated for relative pressures up to 0.4, a range that can allow elucidation
on the basis of the amount adsorbed per cubic centimeter of of the role of oxygen containing functional groups in the
micropore volume (Cµ). The Dubinin-Radushkevich plot, which adsorption of water. The isotherm plot in Figure 2 shows an
is commonly employed to obtain micropore size distributions, observable rise at very low loadings and could be considered to
presents the recalculated data in Figure 1. It can be seen that the be of type IV in the BDDT classification scheme.21 Evans22
data appear to overlay to provide a single curve, thereby generating proposes that this rise at low loadings is difficult to observe
further evidence that the micropore size distributions for the two unless careful measurement of data at low pressure is made. This
materials are similar. is due to the upturned nature of the isotherm which leads to much
2.1. Surface Chemistry of Takeda CMS. Technical infor- larger amounts adsorbed at higher pressures, allowing the low-
mation from the manufacturer indicates that both Takeda 3A and pressure data to be obscured due to scaling. Additionally, Evans22
5A CMS are derived from coconut shell, which provides a useful attributes the behavior at low loading to a Langmuir-type binding
precursor material for CMS manufacture. On a molecular level, of water molecules to the functional groups within the carbon.
coconut shell-derived carbon has graphite-like pores that have Also evident from Figure 2 is the fact that the equilibrium
surface corregations. Turner, Pikunic, and Gubbins15 supply a measurements were conducted in adsorption and desorption
snapshot of the structure from an atomistic standpoint and derive modes. It appears that there is a negligible difference between
a PSD very similar to that obtained for both Takeda 3A and 5A points obtained from adsorption and desorption in the range
(i.e., micropores smaller than 1 nm and a modal value of 0.5 nm).
Also included in the molecular model of coconut shell-derived (16) Albers, P. W.; Pietsch, J.; Krauter, J.; Parker, S. F. Phys. Chem. Chem.
Phys. 2003, 5, 1941-1949.
(9) Horvath, G.; Kawazoe, K. J. Chem. Eng. Jpn. 1983, 16, 470-475. (17) Marsh, H., Ed. Introduction to Carbon Science; Butterworths: Cornwall,
(10) Mariwala, R. K.; Foley, H. C. IEC Res. 1994, 33, 2314-2321. England, 1989.
(11) Rutherford, S. W.; Nguyen, C.; Coons, J. E.; Do, D. D. Langmuir 2003, (18) Striolo, A.; Chialvo, A. A.; Cummings, P. T.; Gubbins, K. E. Langmuir
19, 8335-8342. 2003, 19, 8583-8591.
(12) Cazorla-Amoros, D.; Alcaniz-Monge, J.; de la Casa-Lilli, M. A.; Linares- (19) Striolo, A.; Gubbins, K. E.; Chialvo, A. A.; Cummings, P. T. Mol. Phys.
Solano, A. Langmuir 1998, 14, 4589-496. 2004, 102, 243-251.
(13) Rutherford, S. W.; Do, D. D. Carbon 2000, 38, 1339-1350. (20) Rutherford, S. W.; Coons, J. E. Langmuir 2004, 20, 8681-8687.
(14) Rutherford, S. W.; Do, D. D. Langmuir 2000, 16, 7245-7254. (21) Brunauer, S.; Deming, L. S.; Deming, W. E.; Teller, E. J. Am. Chem. Soc.
(15) Turner, C. H.; Pikunic, J.; Gubbins, K. E. Mol. Phys. 2001, 99, 1991- 1940, 62, 1723.
2001. (22) Evans, M. J. B. Carbon 1987, 25, 81-83.
704 Langmuir, Vol. 22, No. 2, 2006 Rutherford

Figure 4. Water adsorption equilibria on Takeda 3A and Takeda

Figure 2. Water adsorption and desorption equilibria of Takeda 5A 5A together with the fit of the composite Langmuir-Ising model,
CMS at 20 °C. represented by eq 3. The contribution of the Ising equation is also
shown.

Takeda 3A and 5A appears to form a single curve, as was the

case for carbon dioxide adsorption. This result indicates that
consistent behavior is observed for both water adsorption and
carbon dioxide adsorption in Takeda CMS. Furthermore, these
data can be coupled with the data obtained by Alcaniz-Monge
et al.23 at higher pressures and at a higher temperature as shown
in Figure 4. From a theoretical standpoint, the temperature
dependence of the isotherm is expected; however, we note that
the saturation capacity and therefore the data at loadings
approaching saturation should be approximately independent of
temperature over a range of 5 °C. Additionally, the data presented
by Alcaniz-Monge23 at low loading is obscured by a large number
of measurements for other carbon materials and the Takeda CMS
data cannot be accurately reproduced at low loading. For these
reasons, only the data at higher loading are included in Figure
4. It can be seen that there is some deviation between the data
Figure 3. Water adsorption equilibria of Takeda 3A and Takeda
5A at 20 °C. of this investigation and that of Alcaniz-Monge et al.23 at mid
to high loading of water. Such differences are ascribed to
variations in outgassing conditions that can have a significant
studied, indicating negligible hysteresis. The lack of observable effect on the water adsorption isotherm according to Alcaniz-
hysteresis is significant because it precludes the intrusion of Monge et al.23 Nevertheless, the combination of data from various
capillary condensation, which represents a different mechanism sources provides a data set spanning almost a 1000-fold change
to the molecular-level phenomena investigated in this study. A in relative pressure. This data set provides an opportunity for a
similar observation was made for water adsorption in Takeda 3A comparison of various theories for prediction of water adsorption
by Rutherford and Coons.20 In contrast to this result are simulation in carbon micropores. These theories are considered in the
studies that propose that hysteresis should be observed even in following section.
pores of size 1 nm.18,19 The discrepancy may be attributable to
the difference between the experimental and simulation condi- 4. Water Adsorption Isotherm Equations
tions. The simulation studies18,19 involved loading of water into
Water adsorption in microporous carbon is known to display
the micropores to complete filling where hysteresis effects would
complex behavior because of the low energy of interaction with
be more pronounced. The experiments of this investigation involve
the carbon surface coupled with the high potential for association
water loading to less than approximately one-third of the saturation
of water molecules. This complex behavior leads to a variety of
capacity. Under these conditions, the effects of molecular packing
observed isotherm types ranging from type I to type V of the
that are assumed to be responsible for hysteresis may be less
BDDT21 classification scheme. Obviously, a robust model for
pronounced. Further simulation studies that involve partial loading
water adsorption must be capable of characterizing this widely
and unloading of the micropores are required to resolve the
varying behavior. Specifically, the following criteria for a robust
apparent discrepancy.
equilibrium isotherm model are proposed:
As a measure of validation, the data obtained in this
(1) Water adsorption on carbon surfaces with a relatively low
investigation is compared to the data obtained for Takeda 3A in
affinity for water, for example, graphite surfaces, display type
a previous investigation.20 Figure 3 shows the amount of water
adsorbed, expressed on the basis of the amount adsorbed per (23) Alcaniz-Monge, J.; Lozano-Castello, D. Adsorpt. Sci. Technol. 2003, 21,
cubic centimeter of micropore. The water adsorption data for 841-848.
Study of Water in Carbon Molecular SieVes Langmuir, Vol. 22, No. 2, 2006 705

III/V behavior. Water adsorption in microporous carbons with However, the Darcy-Watt model suffers the same limitation as
measurable oxygen content appear to display significant amounts the Dubinin-Serpinski isothermsit cannot represent a clear
adsorbed at very low pressure (<10% relative pressure) and saturation capacity in type II, IV, and V isotherms and therefore
possess type IV isotherms. Type IV behavior has been observed cannot accurately represent water adsorption in microporous
in several studies.24-26 Carbons with a moderate density of carbon.
oxygenated functional groups display type II behavior, which Methods based on Dubinin’s theory of volume filling of
has been observed in several studies27-29 of water adsorption micropores (including the Dubinin-Astakhov equation) have
in commercially derived carbon samples. Microporous carbons the feature of displaying finite saturation capacity and can
with a very high density of functional groups would be capable represent type I, III (assigning an infinite saturation capacity),
of displaying type I behavior. Therefore, a valid isotherm and V isotherms. However, these methods do not have correct
model must be capable of representing type I, II, III, IV, and V thermodynamic behavior at low concentrations,30,31 fail to
isotherms. represent Henry’s law, and therefore fail to meet the second
(2) It is a thermodynamic mandate that a water adsorption criterion.
isotherm display linear behavior at low pressure.30,31 That is, a The Talu-Meunier method can represent type I, III (assign-
valid isotherm model must be capable of representing Henry’s ing an infinite saturation capacity), and V isotherms and has
law at low concentrations. correct thermodynamic behavior at low pressure. However, it
(3) Recent molecular simulation studies show that water can neither represent type II isotherms nor characterize type IV
adsorption in purely graphitic carbon nanopores without functional isotherms.
groups is initiated via water attraction to the carbon surface.18,19 The method proposed by Do and Do can represent type I, II,
At low pressures, monomeric water can bond to the carbon surface III, IV, and V isotherms and has correct thermodynamic behavior
with a range of interaction energy that is dependent upon the at low pressure. However, crucial to the development of this
pore width. At higher pressures, clustering of water molecules theory is the assumption that water molecules cannot exist in the
occurs, and the formation of dimers, trimers, and so forth is micropore as monomers, dimers, or trimers. Recent molecular
generated by hydrogen bonding. A valid isotherm model must simulation studies show that water can exist in these forms within
be capable of representing the formation of monomers, dimers, the carbon micropore.18,19,39 Although this theory may provide
trimers, and so forth within the micropores formed between a mathematical representation of the various forms of water
graphene layers. adsorption, it fails to meet the third criterion.
Several theories have been proposed to characterize water The CMMS theory can represent type I, II, III, and V isotherms
adsorption equilibrium in microporous carbon. These theories and has correct thermodynamic behavior at low pressure.
have been reviewed by Brennan et al.4 At present, the following However, it suffers from the inability to characterize type IV
quantitative analytical theories have been proposed: isotherms. It is the purpose of this investigation to extend the
(1) the Dubinin-Serpinski32,33 method; CMMS theory to account for type IV isotherms, thereby making
(2) the Darcy-Watt method34 used by Evans;20 it the only considered analytical model that can satisfy all of the
(3) the Dubinin-Astakhov equation;35 above-mentioned criteria.
(4) the Talu-Meunier method;36 4.1. Extension of the CMMS Equation. CMMS theory models
(5) the method proposed by Do and Do;37 and the adsorption of molecules on primary adsorption sites and
(6) CMMS theory applied to water adsorption in carbon.38 includes a cooperative effect in which the sorbed molecules
The Dubinin-Serpinski isotherm equation is capable of promote the adsorption of additional molecules onto adjacent
representing Henry’s law at low pressures and can represent type sites.40 The model considers a unit triad where a central sorption
I and III isotherms. However, it suffers the limitation that it can- site promotes sorption on either side of the central site. Secondary
not represent a clear saturation capacity in type II, IV, and V iso- interactions are also incorporated by the inclusion of side
therms37 and therefore cannot accurately represent water adsorp- associates that form from the primary surface unit triad and allow
tion in microporous carbon over a wide range of concentrations. for the formation of dimers, trimers, and so forth. The model
does not consider phenomena beyond the molecular level such
The Darcy-Watt method represents an improvement of the
as capillary condensation that occurs in mesopores at high relative
Dubinin-Serpinski isotherm in the respect that it can account
pressure. As a result, the application of CMMS theory to
for type I and type III isotherms, has correct limiting behavior
adsorption in mesoporous materials at high relative pressures is
at low pressure, and is better at characterizing low-pressure data
not valid. However, Takeda 3A and 5A contain micropores
of water adsorption in carbon with measurable oxygen content.
smaller than 1 nm, and the application of CMMS theory to high
(24) Barton, S. S.; Evans, M. J. B.; Holland, J.; Koresh, J. E. Carbon 1984, relative pressures is accepted for these materials. The CMMS
22, 265-272. equilibrium isotherm equation is represented by the following
(25) Harding, A. W.; Foley, N. J.; Norman, P. R.; Francis, D. C.; Thomas, K. analytical form
M. Langmuir 1998, 14, 3858-3864.
(26) OKoye, I. P.; Benham, M.; Thomas, K. M. Langmuir 1997, 13, 4054-
4059.
C K0a
(27) Bandosz, T. J.; Jagiello, J.; Schwarz, J. A.; Krzyzanowski, A. Langmuir θ) ) (1a)
1996, 12, 6480-6486. Csat (1 - K a)[K a + w2(1 - K a)]
(28) Salame, I. I.; Bagreev, A.; Bandosz, T. J. J. Phys. Chem. B 1999, 103, as 0 as
3877-3884.
(29) Salame, I. I.; Bandosz, T. J. Langmuir 2000, 16, 5435-5440. where θ is the fractional loading, C is the concentration of sorbate
(30) Kapoor, A.; Ritter, J. A.; Yang, R. T. Langmuir 1989, 5, 1118-1121.
(31) Do, D. D. Adsorption Analysis: Equilibria and Kinetics; Imperial College in the adsorbent that is in equilibrium with the gas phase, Csat
Press: London, 1998. is the adsorption capacity, a is the relative pressure (i.e., the ratio
(32) Dubinin, M. M. Carbon 1980, 18, 355-364.
(33) Dubinin, M. M.; Serpinsky, V. V. Carbon 1981, 19, 402-403. of gas-phase pressure to saturated vapor pressure), K0 is the
(34) Darcy, R. L.; Watt, I. C. Trans. Faraday Soc. 1970, 66, 1236-1245.
(35) Dubinin, M. M.; Astakhov, V. A. AdV. Chem. Ser. 1971, 102, 69. (39) Ohba, T.; Kanoh, H.; Kaneko, K. J. Phys. Chem. B 2004, 108, 14964-
(36) Talu, O.; Meunier, F. AIChE J. 1996, 42, 809-819. 14969.
(37) Do, D. D.; Do, H. D. Carbon 2000, 38, 767-773. (40) Malakhov, A. O.; Volkov, V. V. J. Polym. Sci., Ser. A 2000, 42, 1120-
(38) Rutherford, S. W. Carbon 2003, 41, 622-625. 1126.
706 Langmuir, Vol. 22, No. 2, 2006 Rutherford

equilibrium constant for sorption of the central unit on the primary Table 1. Parameter Values Evaluated from Isotherm and
site, and Kas is the equilibrium constant for sorption of the side Kinetic Data Fitting
associate. The parameter w is given by parameter value

( )
K1 1.9

x( )
1 K1a K1a 2 4K0a K0 7.0 × 10-2
w) 1- + 1- + (1b) Csat (mol/cc micropore) 4.8 × 10-2
2 1 - Kasa 1 - Kasa 1 - Kasa CsatL (mol/cc micropore) 5.2 × 10-4
bL 1.0 × 102
where K1 is the equilibrium constant for sorption of the side unit Dµ0/Rµ2 (/s) 1.75 × 10-3
on the primary site.
Many prior studies have shown that there are two modes of of attraction is affected by pore size and site density.41-44
water adsorption in carbon. The first mode considers water within According to Jorge et al.,44 we may anticipate values for the
the micropores formed from the spacing between graphene layers Langmuir affinity constant (bL) from 10 to 10 000 depending on
and the second mode considers water bound to the functional the functional group chemistry, pore width, and site density.
groups located at the edge of the graphene layers.4,22,37 A general 4.2. Application to Water Adsorption in Takeda 5A and
composite isotherm is proposed to account for these two modes Takeda 3A. By accounting for the two modes of adsorption, a
of adsorption new composite isotherm model has been proposed in the previous
section. This isotherm model satisfies all of the previously
CsatM1K0M1a proposed criteria. The isotherm equation can be applied to
Ctotal ) + characterize the data set assembled for water adsorption in Takeda
(1 - KasM1a)[K0M1a + (wM1)2(1 - KasM1a)] 3A and 5A. For this purpose, the Langmuir-Ising model
CsatM2K0M2a represented by eq 3 was fitted to the data set with five parameters
(2) (K1, K0, bL, Csat, CsatL) evaluated from the fitting procedure.
(1 - KasM2a)[K0M2a + (wM2)2(1 - KasM2a)] These parameters are summarized in Table 1. In agreement with
the earlier discussion, the fitted value for the equilibrium constant
where the superscript M1 denotes the first mode of adsorption for sorption of the side unit on the primary site (K1) is larger than
and the superscript M2 denotes the second mode of adsorption. the equilibrium constant for sorption of the central unit on the
To simplify eq 2 for application to water adsorption data in primary site (K0) on the graphene surface. Additionally, the value
CMS, we note that Takeda CMS contains pores smaller than 1 for the Langmuir affinity parameter is within the expected range
nm. According to molecular simulation,18,19 only single layers according to simulation.44 Furthermore, the density of water
of water can form in these pores, and side association of water obtained from a comparison of the adsorption capacity for water
molecules is prevented. Hence, a zero value for the side association (Csat) to that for carbon dioxide is evaluated to be around 0.9
parameter is assigned to eq 2 (i.e., KasM1 ) KasM2 ) 0). A further g/cm3. This value is close to that obtained by Alcaniz-Monge et
simplification can be made to eq 2 by noting that the second al. (0.92 g cm-3)45,46 for water adsorption in micropores of
mode of adsorption is of the Langmuir type, according to Evans.20 activated carbon. It must also be noted that the Langmuir saturation
Therefore, as a further simplification, K0M2 ) K1M2 is additionally capacity (CsatL), which is indicative of the oxygenated functional
assigned to eq 2. This results in a composite isotherm with group concentration within the carbon, should exhibit values
contributions from the Ising equation, representing the first mode comparable with the concentration of oxygen-containing species
of adsorption, and the Langmuir equation, representing the second in other carbons. For comparison, Tsunoda47 obtained concentra-
mode of water adsorption, as follows tions from fractions of 1 mmol/g for carbons with low primary
site density to above 5 mmol/g for carbons with high primary
CsatK0a CsatLbLa site density. The values obtained in this investigation are
Ctotal ) + (3a)
(K0a + wIsing2) (1 + bLa) comparable with the low primary site density carbons studied
by Tsunoda.47
where wIsing is given by The resulting fit of the isotherm model, with the parameters
specified in Table 1, is shown in Figure 4. It can be seen that
wIsing ) (1 - K1a + x(1 - K1a)2 + 4K0a)
1 the model can successfully characterize the observed type IV
(3b)
2 data by application of the composite Langmuir-Ising equation.
Also shown in Figure 4 is the contribution of the amount adsorbed
To conform with standard nomenclature, some nomenclature
within the micropores (formed from the spacing between graphene
changes have been made in the derivation of eq 3. Specifically,
layers) as represented by the Ising equation. It can be seen that
the following parameters are introduced: Langmuir affinity
at low loading the contribution of the Ising equation to the total
parameter bL, where bL ) K0M2 ) K1M2, and Langmuir saturation
amount adsorbed is small, implying that the adsorbed water is
capacity CsatL, where CsatL ) CsatM2. Additionally, the superscript
predominantly bound to the functional groups. At higher loading,
M1 has been removed.
the contribution increases until around 20% of the saturation
The first mode of adsorption is characterized by water
interaction with the carbon surface followed by cooperative (41) Muller, E. A.; Rull, L. F.; Vega, L. F.; Gubbins, K. E. J. Phys. Chem.
interaction of the adsorbed water with other water molecules. It 1996, 100, 1189-1196.
would therefore be anticipated that the value for the equilibrium (42) McCallum, C. L.; Bandosz, T. J.; McGrother, S. C.; Muller, E. A.; Gubbins,
K. E. Langmuir 1999, 15, 533-544.
constant for sorption of the side unit on the primary site (K1) (43) Brennan, J. K.; Thomsom, K. T.; Gubbins, K. E. Langmuir 2002, 18,
would be larger than the equilibrium constant for sorption of the 5438-5447.
(44) Jorge, M.; Schumaker, C.; Seaton, N. A. Langmuir 2002, 18, 9296-
central unit on the primary site (K0) because of the higher affinity 9306.
for water molecules to other water molecules than for water to (45) Alcaniz-Monge, J.; Linares-Solano, A.; Rand, B. J. Phys. Chem. B 2002,
the graphene surface. The second mode of adsorption that occurs 106, 3209-3216.
(46) Alcaniz-Monge, J.; Linares-Solano, A.; Rand, B. J. Phys. Chem. B 2001,
via interaction of water with oxygenated functional groups is 105, 7998-8006.
characterized by a high degree of attraction. However, the degree (47) Tsunoda, R. J. Colloid Interface Sci. 1999, 218, 269-274.
Study of Water in Carbon Molecular SieVes Langmuir, Vol. 22, No. 2, 2006 707

pressure is reached. At this point and beyond, the contribution

of the Ising equation dominates, implying that the majority of
the adsorbed water is stored between the graphene layers.

5. Diffusion of Water in Takeda 5A

The diffusional uptake of water in CMS is known to display
a widely varying time scale, which can be up to 100 times slower
at higher loading than the time scale at lower loading.48 As a
result, it is particularly important to monitor the uptake kinetics
to ensure accuracy in the measured equilibrium isotherm. The
fractional uptake of water (F) can be expressed by Fick’s equation
applied to the spherical micrograins of CMS, resulting in the
micropore diffusion equation5,7

F)1-
6 ∞

∑
π2 n)1 n2
1
exp -
( Dµn2π2t

Rµ2 ) (4)

where Dµ is the diffusivity of the adsorbed species in the micropore Figure 5. Dependence of the mobility parameter for water adsorption
and Rµ is the radius of the grain. Often the grain radius is unknown, at 20 °C in Takeda 5A CMS together with the fit of the Maxwell-
and it is convenient to define the mobility parameter, Dµ/Rµ2, Stefan formulation represented as a solid line.
which is used for analysis.5 Application of the Fickian diffusion
equation (eq 4) assumes the following: influence of the pore mouth barrier in water uptake is therefore
(1) The diffusivity is held constant throughout each increment. disregarded. The influence of macropore diffusion can be assessed
To validate this assumption, the batch adsorption experiment is using the procedure suggested by Ruthven5 in which the ratio
conducted differentially as outlined in Rutherford and Do.14 The of macropore to micropore mobility is calculated. Using the
increments in pressure were taken small enough such that an results of this investigation for water adsorption in Takeda 5A
approximately linear relationship exists between the change in CMS, the ratio is calculated to be less than 0.01. On the basis
the gas- and adsorbed-phase concentrations. Also, small changes of this calculation, a consideration of macropore diffusion for
in pressure allow the approximation of a constant boundary water transport is not included, and the rate-limiting process of
condition. To ensure that the adsorbent temperature remains water transport in CMS is attributed to micropore diffusion.
essentially constant, increments of less than 0.2 mmol/g are taken Equation 3 is fitted to the uptake data at each measured relative
in this investigation, similar to the condition imposed on the pressure, and the results are shown in Figure 5. The mobility
uptake of water in Takeda 3A CMS.20 parameter initially appears to increase at low pressures and then
(2) No water is adsorbed on the walls of the dosing and sample decrease at higher relative pressures. This behavior has been
chambers. To validate the assumption, separate experiments have observed in several studies.25,58 The Maxwell-Stefan (MS)
been performed that involve the expansion of water from the formulation7,58 has been applied to characterize water diffusion
vapor source into the dosing chamber. The pressure is monitored in Takeda 3A CMS21 and can be expressed as
continuously over the time scales of water adsorption in the
CMS. The pressure change over this period is negligible in Dµ Dµ0 d(ln a)
comparison to the pressure changes during the exposure of water 2
) (5)
to Takeda CMS, and we therefore ignore the influence of water Rµ Rµ2 d(ln Ctotal)
adsorption on the chamber walls.
(3) Micropore diffusion is assumed to control mass transfer. where Dµ0/Rµ2 represents the zero-loading mobility parameter.
However, in the single gas batch adsorption experiment, the The MS formulation has been applied to the data taken in this
process of mass transfer in carbon molecular sieves may be rate investigation and requires a mathematical form for the equilibrium
limited by macropore diffusion, micropore diffusion, or transport isotherm. Equation 3, with the five parameters evaluated
through the pore mouth barrier. In many cases, the first two previously from equilibrium data fitting (and presented in Table
mechanisms can control mass transfer.5 However, large molecules 1), is used for this purpose. Numerical differentiation of eq 3 is
transported in CMS may be restricted from entry into the required, and one fitting parameter, the zero-loading mobility,
micropore by a barrier at the pore mouth. Characteristic of this is evaluated from the data fitting. The fit is shown in Figure 5,
is a kinetic uptake that adheres to a non-Fickian linear driving and it is clear that this formulation can capture the behavior at
force (LDF) model.49-57 However Fickian diffusion, as evidenced low loading that is characterized by an increasing mobility
by conformity to eq 4, is observed in this investigation. The parameter. At higher loading, a decrease in the mobility parameter
is evident, a characteristic captured by the MS formulation. From
(48) Cossarutto, L.; Zimny, T.; Kaczmarczyk, J.; Siemieniewska, T.; Bimer, an observation of Figure 5, it is evident that the MS formulation
J.; Weber, J. V. Carbon 2001, 39, 2339-2346.
(49) Koresh, J.; Soffer, A. J. Chem. Soc., Faraday Trans. 1981, 77, 3005- can successfully characterize the dependence of the mobility
3018. parameter upon loading.
(50) LaCava, A. I.; Koss, V. A.; Wickens, D. Gas Sep. Purif. 1989, 3, 180-
186.
(51) Chagger, H. K.; Ndaji, F. E.; Sykes, M. L.; Thomas, K. M. Carbon 1995, (54) Srinivasan, R.; Auvil, S. R.; Schork, J. M. Chem. Eng. J. 1995, 57, 137-
33, 1411. 144.
(52) Liu, H.; Ruthven, D. M. In Proceedings of the 5th International Conference (55) Reid, C. R.; Thomas, K. M. Langmuir 1999, 15, 3206-3218.
of the Fundamentals of Adsorption; Le Van, M. D., Ed; Kluwer Press: Boston, (56) Reid, C. R.; Thomas, K. M. J. Phys. Chem. B 2001, 10619-10629.
1996. (57) Foley, N. J.; Thomas, K. M.; Forshaw, P. L.; Stanton, D.; Norman, P. R.
(53) Loughlin, K. F.; Hassan, M. M.; Fatehi, A. I.; Zahur, M. Gas Sep. Purif. Langmuir 1997, 13, 2083-2089.
1993, 7, 264. (58) Krishna, R.; Wesselingh, J. A. Chem. Eng. Sci. 1997, 52, 861-911.
708 Langmuir, Vol. 22, No. 2, 2006 Rutherford

6. Conclusions proposed CMMS theory are overcome by extension of this theory

Water molecules are useful probes in assessing ultrami- to incorporate the binding of water to functional groups. By
croporosity in carbon molecular sieves because of their small comparison with a variety of other analytical theories for water
molecular size that allows them to penetrate micropores without adsorption in carbon, the extended CMMS theory is shown to
interference from the pore mouth barrier. A comparison of Takeda be the only considered isotherm model that can fulfill a variety
3A and 5A, which appear to differ only in the size of the pore of theoretical criteria.
mouth barrier and not in pore size distribution, reveals that these
materials are characterized by a type IV isotherm that is evident
Acknowledgment. Los Alamos National Laboratory, an
only by close scrutiny of low-pressure data. This isotherm is
affirmative action/equal opportunity employer, is operated by
attributed to the binding of water to functional groups, which
the University of California for the United States Department of
dominates at low pressure, coupled with filling of the micropores
Energy under contract W-7405-ENG-36.
formed from spacing between graphene layers, a process that
dominates at higher pressures. Shortcomings of the previously LA051826N