Effect of Surface Oxygen and Temperature on External and Micropore

Jun 19, 2009 - ... and Microscopy Group, Oak Ridge National Laboratory, 1 Bethel Valley Road, Building 4515, MS 6064, Oak Ridge, Tennessee 37831-6064...
0 downloads 0 Views 1MB Size
J. Phys. Chem. C 2009, 113, 12109–12117

12109

Effect of Surface Oxygen and Temperature on External and Micropore Adsorption of Water in Single-Walled Carbon Nanotubes by Gravimetric and Spectroscopic Experiments Pyoungchung Kim,† Harry M. Meyer,‡ and Sandeep Agnihotri*,† EnVironmental Engineering, Department of CiVil and EnVironmental Engineering, UniVersity of Tennessee, KnoxVille, Tennessee 37996-2010, and Microscopy Group, Oak Ridge National Laboratory, 1 Bethel Valley Road, Building 4515, MS 6064, Oak Ridge, Tennessee 37831-6064 ReceiVed: January 20, 2009; ReVised Manuscript ReceiVed: May 19, 2009

Gravimetric water adsorption experiments (T ) 5, 20, and 35 °C and 0 < P/Po < 0.95) were performed on several chemically and structurally distinct samples of single-walled carbon nanotubes including two activated carbon samples. The isotherms followed the type V curve and were fitted to a semiempirical model which allowed the adsorptive contributions of primary sites and micropores (referred to here as pseudoexperimental isotherms) to be distinguished with statistical confidence. The isosteric heats of water adsorption calculated from experimental isotherms ranged between 46 and 58 kJ/mol. The same calculations were performed on the separated adsorptive components (functional groups and micropore isotherms) and were found to be 0.5-16 kJ/mol and 1-8.6 kJ/mol, respectively. These values are similar to those available in the current literature reportedly estimated by calorimetric and molecular simulation techniques. From semiempirical modeling, we were also able to qualitatively estimate temperature sensitive water specific sample properties such as the concentration of primary sites (found directly related to % O) and the size of water clusters aggregating on primary sites (found inversely related to % O) and those filling micropores (found directly related to the dominant pore size) and adsorption equilibrium constants. We believe that our approach is useful in interpreting experimental water adsorption thus aiding purely simulation based methods of studying the behavior of water in nanocarbons. 1. Introduction The adsorption of water in nanoporous carbons is an important process affecting a variety of scientific fields including electrochemistry, chromatography, catalysis, and membrane separation.1,2 Single-walled carbon nanotubes (SWNTs) provide a well-defined nanoscaled cylindrical geometry of high porosity, interstitial wedges and a large external surface for enhanced adsorption.3 The well-ordered SWNT structures have been widely investigated for water adsorption using molecular simulations and experimental techniques. Molecular dynamic (MD) simulation studies of water in SWNTs have predicted mechanisms of filling and emptying of nanotubes via sequential addition and removal of water molecules.4 Research has explored the effects of pore size and adsorption temperature,5 and revealed the formation of ice-like structures of water confined in SWNT pores.6 Some experimental studies have validated the MD prediction of water-nanotube interactions. The ice formation in SWNTs has been observed by proton nuclear magnetic resonance (NMR)7 and neutron diffraction studies.8 Using experimental vibrational spectroscopy, it was suggested that ring-like clusters of water molecules arrange themselves to create a hydrogen-bonded stacked configuration of water confined in nanotubes.9 Recently, an NMR study of preadsorbed water in SWNTs suggested that with a decrease in adsorption temperature the properties of nanoconfined water in hydrophobic carbon pores shift from being highly hydrophobic to slightly hydrophilic.10 * Corresponding author. E-mail: [email protected]. Phone: (865) 9747728. Fax: (865) 974-2669. † University of Tennessee. ‡ Oak Ridge National Laboratory.

SWNTs have a geometric configuration that is vastly different from that of typical activated carbons; while the presence of micropore and mesopore volumes and their capacities for common adsorbates is similar to those of microporous activated carbons.11-15 It is known that water adsorption in carbon pores is mainly affected by the presence of surface functional groups and by pore size. The surface functional groups facilitate adsorption of water at low relative pressures by forming hydrogen bonds after which the aggregation of water molecules into ring-like clusters occurs. Micropore filling is facilitated by the adsorption of water clusters.16 Pore filling is known to occur without the formation of monolayer unlike that observed for simple fluids.17 Using experimental and molecular simulation methods, Kaneko and co-workers,18 suggested that water clusters formed on the functional groups located at the pore entry of hydrophobic carbon pores exhibited a shift from hydrophilic to hydrophobic after entering the pore. These authors also showed19 that the number of molecules that form a cluster, i.e., the cluster size, increases with larger pore size or a decreased concentration of surface functional groups. Estimating the thermodynamic properties of water adsorbed in SWNTs is as important as measuring adsorption capacities and kinetics. Recently, experimental methods and molecular simulation were employed to study the temperature dependency of water adsorption isotherm in carbon nanopores. Striolo et al.20 suggested that for carbon slit pores and carbon nanotubes an increase of temperature causes the isotherm to shift rightwards, i.e., toward an increasing relative pressure, while a decreasing temperature exhibits a narrower hysteresis. Kimura et al.21 used microcalorimetric and X-ray diffraction techniques to estimate heat of water adsorption in carbon nanopores and concluded that at higher vapor pressures several small clusters

10.1021/jp900609a CCC: $40.75  2009 American Chemical Society Published on Web 06/19/2009

12110

J. Phys. Chem. C, Vol. 113, No. 28, 2009

combine to form a large highly ordered cluster that fills the micropore. Some related studies have developed new equations for calculating the heat of water adsorption in carbon pores.22,23 Furmaniak et al.24 derived an equation by incorporating temperature dependency in the fitting parameters of the Do & Do equation.16 These methods may describe the thermodynamic behavior of water in the porous carbon; however, they may be limited in application to the samples tested in the particular study. The objective of this study was to collect water isotherms on multiple SWNT samples and apply a previously modified version25 of the Do & Do water-isotherm model16 in order to extract from experimental data the heat of water adsorption in SWNTs, the size of adsorbing water clusters, the concentration of primary sites, and the limiting pore volume. In our previous study,26 we tested the applicability of several common water isotherm-equations to experimental isotherms collected for SWNTs and activated carbons at T ) 20 °C. We concluded that a model suggested by Marban et al.25 (Cluster Formation Induced Micropore Filling, CIMF model) is the most suitable equation for interpreting the experimental water-SWNT isotherms. Using this model we were able to deconvolute experimental isotherms into micropore and functional group fractions and estimate cluster sizes, the concentration of functional groups, and the micropore volume. In this publication, we are applying this equation to data collected at multiple temperatures and have estimated the total and separated heats of water adsorption on functional groups and in micropores of several SWNT samples. The heats of adsorption (46-58 kJ/mol) were found to be similar to those reported in literature by theoretical calculations27,28 as well as calorimetric measurements.29 In micropores, the heat of adsorption was found to be inversely related to the size of water clusters. As posited, the value of parameters representing the concentration of surface functional groups and micropore volumes decreased with an increasing temperature and was found to relate to total % O and pore widths independently determined by spectroscopic characterization of samples. 2. Experimental Section 2.1. Sample Information and Characterization. SWNT samples selected for this study were commercially produced and were purified by the manufacturer. The sample description, morphologies and characterization details were provided in our previous works.3,11,15,26,30-33 Briefly, the SWNT samples are labeled as SWNT1 (>95% pure, MER Corp.); SWNT2 (>90% pure, Carbon Nanotechnologies Inc.) and SWNT3 and SWNT4 (70% and 80% pure, Carbon Solution Inc.). Samples of activated carbon were also tested as a benchmark. They were labeled as AC (Calgon activated carbon) and ACF10 (nonwoven activated carbon fiber from American Kynol). The majority of experimental analysis was performed on as-received samples, without additional treatment. Adsorptive characterization of all samples was performed by standard N2 adsorption at 77 K (Autosorb 1-c by Quantachrome Instruments). Twenty to 30 mg of each sample was outgassed at 140 °C and 10-3 mbar for 24 h after which N2 adsorption isotherms were obtained volumetrically at relative pressure (P/Po) of 10-6 < P/Po< 0.99, where P is actual pressure and Po is the saturation pressure of N2 at 77 K. For each sample, the surface area, pore volumes and pore size distributions (cylindrical geometry) were extracted from the N2 isotherm data by applying respective numerical models built into the Autosorb 1 operating software. The morphology of SWNT samples were also characterized by Raman spectroscopy (λexcitation ) 532 nm,

Kim et al. T6400 Raman research system by JY Horiba). Those experimental details can be found in our previous work.32 The dominant diameter was calculated using the relation ωRBM ) 10.0 + 234.0/D (nm), where ωRBM is the radial breathing mode (RBM) Raman shift (cm-1).32 Spectra are present in Figure 1. Table 1 provides a summary of the characterization data. The relevant surface chemistry of samples was measured as total atomic surface oxygen (% O) by X-ray photoelectron spectroscopy (Thermo Scientific K-Alpha XPS). The % O peak is typically used to estimate the total concentration of oxygenated functionality.34 Spectra were recorded using monochromatic Al K-R radiation at pass energy of 25 eV and power of 300 W. The base pressure of the analysis chamber was 2 × 10-9 mbar and the operating pressure was maintained at approximately 2 × 10-7 mbar while the charge compensation (a combination of low energy Ar ions and low energy electrons) system was active. The atomic % O peak was calculated from the areas of oxygen 1s (525-545 eV) peaks using Thermo Scientific’s Avantage software (v3.85) (Figure 2). 2.2. Measurement of Water Adsorption Isotherms. The equipment used in collecting the adsorption isotherms included a gravimetric balance (Digital Recording Balance, DRB-200 by Thermo Cahn; limit of detection ) 0.1 µg), a gas generation system and a data acquisition system. The details of experimental setup and its operation are provided in our previous work33 and are also available in Supporting Information. The following is a brief description of the methodology. The experiments were conducted in an open-configuration, in equilibrium with the atmospheric pressure, so that the vapor carrying gas was continuously flowing through the balance while the sample mass was being monitored in real-time during adsorption. The carrier gas for all experiments was ultrahigh purity (UHP) N2 at a total flow rate of 200 sccm. Prior to each experiment, the balance was zeroed and calibrated. This step obviated the need for buoyancy corrections that are typically needed in volumetric-gravimetric measurements. In our experiments 3-5 mg of the adsorbent sample was placed on the sample pan and its initial weight was measured after reaching equilibrium in dry carrier gas at T ) 140 °C. The sample was then cooled to the operating temperature (T ) 5, 20, or 35 °C). The environmental chamber in which the balance was placed had also been set at the operating temperature. The chamber humidity was maintained at 50 ( 1%. These precautions were taken to minimize signal fluctuation and to allow gravimetric measurements with µg-level accuracy. Water vapor, measured as relative humidity (RH ) 100 × P/Po, 0 < P/Po < 0.95), was generated by purging the carrier gas through double-distilled water using a fritted glass bubbler followed by mixing with dry carrier gas. The carrier gas containing desired water vapor concentration was introduced into the balance. The sample adsorbed some water which resulted in an increase in the sample mass. Equilibrium was noted when no increase in sample mass ( 0.95), the fitting parameters obtained from the CIMF model25 were the most compatible to independent measurements of sample properties: N2 adsorption (77 K) for micropore volume, surface-chemistry from Raman scattering and relation of pore widths to cluster sizes. For this study, eq 1 was applied to the water adsorption isotherms collected at 5, 20, and 35 °C. Equation 1 was found to fit extremely well with the experimental data regardless of the temperature and sample type (R2 > 0.995, presented here as smooth lines in Figure 3). The accuracy of the fitting parameters was tested by relating them with sample characterization: So was compared to the % O from the XPS spectra of samples, and Cµs was compared to the N2 adsorption (77 K) micropore volume. Representative fitting parameters at T ) 5 °C with complementary sample properties are presented in Table 2. A detailed analysis of the fitting parameters is presented further on in this publication. It can be observed that the So sampleto-sample trends followed those of % O trends. This is also consistent with the trends in the ID/IG ratio of Raman spectra of the samples performed in our previous work.26 The sample-tosample trends in Cµs were similar to those for the N2 micropore volume, Cµs-N2. It was observed that Cµs < Cµs-N2 for all SWNTs and ACF10, which strengthens the accuracy of Cµs values as N2 is expected to fill the pores more effectively than waterclusters.35 The only exception to this secondary constraint was found in sample AC (Cµs ) 0.2 cm3/g, Cµs-N2 ) 0.18 cm3/g) because of a higher degree of structural heterogeneity in activated carbons. 3.2. Analysis of Pseudoexperimental Water Isotherms of SWNT Functional Groups and Micropores: Effect of % O and T. The deconvolution of experimental isotherms of samples SWNT1, 2, 3, and 4 at T ) 5, 20, and 35 °C is presented in Figure 4. For every sample, the first term of eq 1 (eq 1a) is plotted in the left figures and the second term (eq 1b) is plotted in the corresponding right figures. The sum of two terms is plotted as smooth lines in Figure 3, where a high correlation coefficient (R2 > 0.99) with experimental isotherms can be observed. We suggest that the individual terms can be interpreted as pseudoexperimental isotherms36 on functional groups (eq 1a) and in micropores (eq 1b) of SWNTs. This is because a purely experimental method to determine these two adsorptive components cannot be traced in present literature. Therefore, as reported in our recent work,26 a semiempirical approach of fitting experimental isotherms to eq 1 may be the most advanced method available to date.

J. Phys. Chem. C, Vol. 113, No. 28, 2009 12113 The effect of total surface-oxygen on water adsorption onto sample’s functional groups was apparent from the plots of eq 1a (Figure 4, left). It can be observed that as the total surface oxygen progressively increased from 1.8% O in SWNT1 to 13.9% O in SWNT4, the amount of water adsorbed onto the functional groups at a given relative pressure increased as well. For example, at 20 °C and P/Po ≈ 0.6, SWNT1, 2, 3, and 4 adsorbed 0.1, 1.2, 2.0, and 2.8 mmol/g of water on functional groups, respectively. The isotherms exhibited a type III shape that is typical for hydrophobic materials. The effect of surface oxygen, however, was not directly apparent on the micropore component of total adsorption (Figure 4, right). This is because micropore adsorption is also affected by the pore volume and pore sizes, both of which were sample specific (Table 1). The pseudoexperimental isotherms suggest that at a given relative pressure water should simultaneously adsorb on to functional groups and into micropores. Increasing the temperature reduced both adsorptive contributions. Even at low P/Po adsorption on functional groups was affected by temperature. This was most obvious for sample SWNT4 which, in comparison to other samples, exhibited the highest adsorption loading on functional groups. This sample adsorbed 1.3 mmol/g of water at T ) 5 °C but only 0.7 mmol/g at T ) 35 °C at P/Po ) 0.1. For this sample, noticeable adsorption in micropores could be observed at low P/Po < 0.2. This is unexpected because morphologically this sample is similar to SWNT3. We believe that a significant microporous adsorption and a leftward shift in the isotherm are due to much higher % O in this sample. This further suggests that a majority of surface-oxygen may be located at the pore entry.37 As expected from the total adsorption isotherm, all samples exhibited a decreasing adsorption loading and a pseudoisotherm shift toward increasing P/Po with an increasing temperature. 3.3. Heat of Adsorption. The isosteric heat of water adsorption is a differential molar quantity typically derived from the temperature dependency of an isotherm. In the case of water, it should describe the energy released from H2O-primary site interaction and H2O-H2O interaction. Here, the isosteric heat of adsorption, Qst(kJ/mol), was calculated by applying the Clausius-Clapeyron equation, eq 2.38 It was calculated from the experimental data pertaining to the total isotherms (Figure 3) as well as from the pseudoexperimental isotherms simulated from eq 1 (Figure 4).

Qst ) Lst + qst | isotherm ) Lst + RT 2

dln(P/Po) dT

(2)

where latent heat of water condensation (Lst) is 45 kJ/mol;39 qst|isotherm is the net isosteric heat of adsorption (kJ/mol) calculated from experimental isotherm; P/Po are the values of relative pressures at different temperature T (K) which corresponded to the same amount of water adsorbed; R is the gas constant (8.314 J/mol · K). At a specific water adsorption loading, P/Po values are determined at different T. The slope in the plot of ln (P/Po) versus 1/T equals the ratio -qst/R.40 This procedure was repeated for several equinumerical adsorption capacities to yield the dependency of qst|isotherm on the amount adsorbed (mmol/g).40 The heat of adsorption calculated from experimental isotherms (qst|isotherm), and pseudoexperimental isotherms on functional groups (qst|func) and in micropores (qst|micro) is presented in Figure 5. A representative isoster figure is available in the Supporting Information.

12114

J. Phys. Chem. C, Vol. 113, No. 28, 2009

Kim et al.

Figure 4. Effect of % O and temperature on pseudoexperimental isotherms of water adsorption on functional groups (eq 1a, left) and in micropores (eq 1b, right) of SWNT samples at T ) 5 (a), 20 (b), and 35 °C (c). Note that adding the y axis values of functional group and micropore isotherms will yield the total isotherm presented in Figure 3.

3.3.1. Total (Qst). Figure 5a shows the isosteric heat of water adsorption, Qst, calculated from experimental isotherms of samples SWNT1, 2, 3, 4, AC, and ACF10. The values of Qst ranged from 46 to 58 kJ/mol, specific to each sample and the mmol/g water adsorbed. The exceptionally high Qst value of 58 kJ/mol was exhibited by sample SWNT4 in the low coverage region. This sample has a relatively high concentration of surface oxygen, a likely cause for its high heat of adsorption. Using microcalorimetric technique, Terzyk et al.22 have also reported Qst as high as 60-70 kJ/mol at low adsorption capacities (equivalent to P/Po < 0.1) for certain acid-treated carbons. Initially, the Qst values drop with increasing mmol/g of water because adsorption onto the functional groups is no longer the dominating mechanism.23,41 Qst should approach the latent heat of condensation, Lst to denote water binding to preadsorbed water molecules. A local minimum is achieved where the first point of inflection is observed in the isotherm beyond which Qst g Lst depending

upon the proximity of P/Po to micropore filling. Each sample exhibited increasing Qst values with increasing amounts of water adsorbed with the exception of SWNT4. At lower water adsorption capacities (prior to the first point of inflection, P/Po < 0.2), Qst could not be calculated for other samples as the temperature dependency of their isotherms could not be established from the experimental data. This suggests that the H2O-primary site interaction may be stronger than the H2O-H2O interaction, and therefore is not easily affected by increasing T from 5 to 35 °C. At adsorption corresponding to 0.2 < P/Po < 0.5, where water molecules simultaneously adsorb onto functional groups and in micropores, Qst ≈ 45-50 kJ/mol for samples SWNT1, 2, 3, AC, and ACF10. This value of Qst is close to the heat of condensation of water (45 kJ/mol) and is consistent with the current literature.27-29 Sample SWNT4, however, exhibited high Qst value of 58 kJ/mol at 0.15 P/Po which decreased to 51 kJ/mol at 0.3 to 0.4 P/Po and then steeply increased to 57 kJ/mol. All samples

Adsorption of Water in Single-Walled Carbon Nanotubes

Figure 5. Isosteric heat of adsorption for water adsorbed on SWNTs and activated carbons as estimated from experimental isotherms (a) and those calculated from the CIMF model (eq 1) using only the functional group term (b) and the micropore term (c).

exhibited gradually increasing Qst with mmol/g of water adsorbed in regions closer to the micropore filling. Such trends are consistent with those previously reported by experiments21 and molecular simulations.20 3.3.2. Functional Groups (qst|func) and Micropores (qst|micro). The heat of adsorption calculated from fitting eq 2 to the pseudoexperimental isotherms on functional groups and micropores are presented in Figure 5, panels b and c, respectively. It should be noted that Lst is not added to either of these quantities in order to avoid data misrepresentation. This is because one cannot determine if Lst is released solely upon adsorption on to functional groups and not in micropores or vice versa. qst|func for SWNT1 could not be calculated due to very low adsorption (Figure 4a). The qst|func ranged from as high as 16 kJ/mol to as low as 0.5 kJ/mol, specific to sample chemistry and mmol/g of water adsorbed (Figure 5b). The highest of this value was exhibited by the most oxygenated sample, SWNT4. The qst|funcfor samples

J. Phys. Chem. C, Vol. 113, No. 28, 2009 12115 SWNT1, 2, 3, and ACF10 were much lower, between 5 and 0.5 kJ/mol. In general qst|func decreased with increasing adsorption because after saturating the surface oxygen, water molecules continue to bind to the preadsorbed water. The qst|micro was comparatively lower than qst|func (Figure 5c). It varied between 1 and 3 kJ/mol for most of the isotherms, corresponding to the threshold values of P/Po where micropore adsorption takes place prior to complete filling. These values are much lower than the heat of condensation.23,42 A sharp increase in qst|micro was observed closer to saturation which was most likely due to micropore filling at P/Po corresponding to the second point of inflection in the isotherm. These trends are consistent with those found by the microcalorimetric technique21 and molecular simulations.20 Both, Kimura et al.21 and Striolo et al.20 have reported increasing heat of adsorption with mmol/g in slit pores and SWNTs during micropore filling. They hypothesized that it is due to the stabilization of water clusters upon stacking and their dispersive interaction with the pore walls. 3.4. Analysis of Fitting Parameters. Equation 1 was fitted to experimental isotherms measured at T ) 5, 20, and 35 °C. All fitting parameters for SWNTs are presented in Figure 6, where their dependency upon temperature is apparent and the verity of their values can be evaluated by comparison with results from independent sample characterization. First, the numerical value of the fitting parameter So, which delineates the concentration of hydrophilic functional groups, was found related to the total % O of sample (Figure 6a). SWNT1 with the least 1.8% O had the lowest So values (0.4-0.05 mmol/g) while SWNT4 with the most 13.9% O had the highest So values (1.4-1.2 mmol/g). Furthermore, in all samples increasing temperature caused So to drop in value. Since the temperatures in this study are not high enough to permanently remove surface oxygen, the lowering of So suggests water adsorbing on fewer primary sites. Therefore, in essence, So does not denote the actual concentration of hydrophilic functional groups but only that which is being “detected” by water at any given thermodynamic state. Some studies have indicated that at T ≈ 50 °C functional groups can interact with each other which lower the concentration available for water to bind.43 Second, the numerical value of Cµs, which is a measure of the maximum micropore volume occupied by water, was found to relate to the sample’s N2 micropore volume, Cµs-N2 (Figure 6b). SWNT1 with maximum Cµs-N2 ) 0.16 cm3/g also had the highest Cµs of 0.1 to 0.8 cm3/g. However, it was noted that although SWNT4 has zero N2 microporosity, it had significant Cµs. We believe that this is because in this sample the extremely high surface oxygen was blocking the pore opening completely. This implies that electrically neutral molecules such as N2 will not be able to penetrate the pore volume due to the physical restriction at the pore entry. Water molecules, however, will hydrogen-bond with the same physical restriction; therefore, growth of water clusters and their penetration of micropores are expected. In our previous work,26 we reported that heat treatment of SWNT4 significantly enhances the N2 micropore volume, which supports the notion of pore blockage due to functional groups as opposed to nanotubes being uncapped. Increasing temperature reduced the Cµs values regardless of the sample type, which is expected of a physisorption type adsorption behavior of water in the micropores. Third, the numerical value of the fitting parameter n, which describes the maximum number of water molecules per cluster bonded to primary sites, was also found to relate to the % O of a sample; however, this parameter seems to be inversely related to % O (Figure 6c). The least oxygenated sample, SWNT1

12116

J. Phys. Chem. C, Vol. 113, No. 28, 2009

Kim et al.

Figure 6. Analysis of the numerical value of fitting parameters obtained by applying eq 1 to gravimetric water isotherms measured at T ) 5, 20, and 35 °C. (a) Concentration of primary sites, Sο. (b) Micropore volume by water, Cµs. (c) Maximum number of molecules comprising water clusters that grow onto the functional groups, n. (d) Average number of molecules per cluster in water clusters that fill micropores, m. (e) Equilibrium constant for water adsorption onto primary sites followed by growth of n size clusters, Kf, and (f) equilibrium constant for m size water clusters filling into micropores.

had the highest n values (15-17) whereas the most oxygenated SWNT4 sample had the lowest n values (5-6). This trend indicates that a high degree of oxygenation results in formation of smaller aggregates of water molecules on primary adsorption sites. The average gap between neighboring functional groups is expected to be smaller in highly oxidized samples. This should lead to smaller size of water clusters (i.e., lower n in more oxidized samples) per functional group. It was also observed that n is a function of T. For all samples, increasing temperature caused their n value to increase. We believe that when temperature is raised adsorption is restricted to a lesser number of functional groups: only those that have activation energy equal to or higher than those limited by T. This would make the SWNTs appear less hydrophilic thus allowing clusters to grow slightly larger in size. Fourth, the numerical value of the fitting parameter m, which represents the average number of water molecules per cluster migrating into micropores, was discovered to relate directly to

the pore width (Figure 6d). These values were in general agreement with previously published sizes of water clusters in hydrophobic carbon fibers.18,19 It is emphasized that the pore width presented here was extracted from the Raman spectra of the samples (experimental details are presented in our previous work32). The pore width plotted here is the SWNT diameter that corresponded to that value of cm-1 which exhibited the tallest RBM peak (Figure 1). This method of determining SWNTs’ pore sizes is far more accurate than adsorption based methods.3,31,32 Only the tallest peak was considered for the following comparative analysis as the effects of largest-diameter nanotube on water-cluster size should be the most obvious. It was noticed that the sample with the widest nanotubes, SWNT1 (1.5 nm) had the highest m values (11-12). The smallest pore width sample, SWNT2 (0.9 nm) defied expectation of the lowest m ranking (m ) 3.7-4.89) appearing above the most oxidized sample, SWNT4 (1.1 nm). The lowest m values exhibited by SWNT4 indicate that in addition to geometric restriction, high

Adsorption of Water in Single-Walled Carbon Nanotubes concentration of functional groups can also influence the size of clusters entering the pores.25 Fifth, the fitting parameter, Kf, is the equilibrium constant for water adsorption onto functional groups. It expresses the ratio of equilibrium constants for H2O-primary site bonding versus H2O-H2O bonding around the functional groups.25 We found that Kf was directly related to the % O of the sample (Figure 6e). The least oxygenated sample, SWNT1 had the lowest Kf values (0.1-0.4) while SWNT4 with the most 13.9% O had the highest Kf values (11-24). Furthermore, with the exception of SWNT4, increasing temperature caused Kf to increase in all samples. This is due to the fact that H2O-primary site interaction is stronger than H2O-H2O interaction. Increasing temperature, within the limits of these experiments, will affect H2O-H2O bonding far more than H2O-primary site bonding. Last, the numerical value of the fitting parameter Kµ, which represents the equilibrium constant for water cluster insertion into micropores, was found directly related to the sample pore width and inversely related to the sample % O (Figure 6f). The least oxygenated sample which also had the largest nanotubes, sample SWNT1, had exceptionally high values of Kµ (4200-7000) while the most oxygenated sample SWNT4 had one of the lowest values (Kµ ) 31-112). For all samples, increasing temperature lowered Kµ values. This is typical in a physisorption process. 4. Conclusions We collected water adsorption isotherms (0 < P/Po < 0.95) on several samples of single-walled carbon nanotubes (SWNTs) at three isotherms conditions of T ) 5, 20, and 35 °C. The samples were characterized for surface oxygen, micropore volume and pore sizes. The isotherms were fitted to a semiempirical model.25 This data fitting was employed as a methodology to calculate the heat of water adsorption in SWNT micropores and on functional groups, and to determine the size of water clusters, the concentration of functional groups and the limiting pore volumes. The validity of fitting parameters was determined by comparison of their values with sample characterization for total % O by XPS, micropore volume by N2 adsorption at 77 K and pore sizes by Raman spectroscopy. The total heat of adsorption (46-58 kJ/mol) calculated for several chemically and morphologically different SWNTs was found to be similar to those reported in literature derived from other methods such as calorimetry and molecular simulations. Individual heats of adsorption on functional groups (0.5-16 kJ/mol) and micropores (1-8.6 kJ/mol) were extracted from pseudoexperimental isotherms on functional groups and micropores. They were found to be inversely related to the size of water clusters. The equilibrium constant on functional groups increases with increasing temperature while equilibrium constant of micropore filling decreases. As expected, the concentration of surface functional groups and micropore volumes decreased with an increase of temperature. Acknowledgment. The work was partially supported by the National Science Foundation (award number: CBET-0836365). A Portion of this research was conducted at the SHaRE UserFacility, which is sponsored by the Division of Scientific User Facilities, Office of Basic Energy Sciences, U.S. Department of Energy. Supporting Information Available: Schematic of gravimetric setup with operational details. This material is available free of charge via the Internet at http://pubs.acs.org.

J. Phys. Chem. C, Vol. 113, No. 28, 2009 12117 References and Notes (1) Rasaiah, J. C.; Garde, S.; Hummer, G. Annu. ReV. Phys. Chem. 2008, 59, 713. (2) Collins, P. G.; Bradley, K.; Ishigami, M.; Zettl, A. Science 2000, 287, 1801. (3) Agnihotri, S.; Mota, J. P. B.; Rostam-Abadi, M.; Rood, M. J. Carbon 2006, 44, 2376. (4) Waghe, A.; Rasaiah, J. C.; Hummer, G. J. Chem. Phys. 2002, 117, 10789. (5) Striolo, A.; Chialvo, A. A.; Gubbins, K. E.; Cummings, P. T. J. Chem. Phys. 2005, 122, 234712. (6) Koga, K.; Gao, G. T.; Tanaka, H.; Zeng, X. C. Nature 2001, 412, 802. (7) Ghosh, S.; Ramanathan, K. V.; Sood, A. K. Europhys. Lett. 2004, 65, 678. (8) Kolesnikov, A. I.; Zanotti, J. M.; Loong, C. K.; Thiyagarajan, P.; Moravsky, A. P.; Loutfy, R. O.; Burnham, C. J. Phys. ReV. Lett. 2004, 93, 4. (9) Byl, O.; Liu, J. C.; Wang, Y.; Yim, W. L.; Johnson, J. K.; Yates, J. T. J. Am. Chem. Soc. 2006, 128, 12090. (10) Wang, H. J.; Xi, X. K.; Kleinhammes, A.; Wu, Y. Science 2008, 322, 80. (11) Agnihotri, S.; Rood, M. J.; Rostam-Abadi, M. Carbon 2005, 43, 2379. (12) Lafi, L.; Cossement, D.; Chahine, R. Carbon 2005, 43, 1347. (13) Shi, W.; Johnson, J. K. Phys. ReV. Lett. 2003, 91, 4. (14) Long, R. Q.; Yang, R. T. Ind. Eng. Chem. Res. 2001, 40, 4288. (15) Agnihotri, S.; Rostam-Abadi, M.; Rood, M. J. Carbon 2004, 42, 2699. (16) Do, D. D.; Do, H. D. Carbon 2000, 38, 767. (17) Striolo, A.; Gubbins, K. E.; Chialvo, A. A.; Cummings, P. T. Mol. Phys. 2004, 102, 243. (18) Ohba, T.; Kanoh, H.; Kaneko, K. J. Am. Chem. Soc. 2004, 126, 1560. (19) Ohba, T.; Kaneko, K. J. Phys. Chem. C 2007, 111, 6207. (20) Striolo, A.; Gubbins, K. E.; Gruszkiewicz, M. S.; Cole, D. R.; Simonson, J. M.; Chialvo, A. A. Langmuir 2005, 21, 9457. (21) Kimura, T.; Kanoh, H.; Kanda, T.; Ohkubo, T.; Hattori, Y.; Higaonna, Y.; Denoyel, R.; Kaneko, K. J. Phys. Chem. B 2004, 108, 14043. (22) Terzyk, A. P.; Gauden, P. A.; Rychlicki, G. Colloid Surf. APhysicochem. Eng. Asp. 1999, 148, 271. (23) Rutherford, S. W. Langmuir 2006, 22, 9967. (24) Furmaniak, S.; Gauden, P. A.; Terzyk, A. P.; Rychlicki, G.; Wesolowski, R. P.; Kowalczyk, P. J. Colloid Interface Sci. 2005, 290, 1. (25) Marban, G.; Fuertes, A. B. Carbon 2004, 42, 71. (26) Kim, P.; Agnihotri, S. J. Colloid Interface Sci. 2008, 325, 64. (27) Velikov, A. A.; Grigor’ev, S. V.; Chuikin, A. V. Russ. J. Phys. Chem. 2006, 80, 2047. (28) Hassan, N. M.; Ghosh, T. K.; Hines, A. L.; Loyalka, S. K. Carbon 1991, 29, 681. (29) Terzyk, A. P.; Rychlicki, G.; Cxiertnia, M. S.; Gauden, P. A.; Kowalczyk, P. Langmuir 2005, 21, 12257. (30) Agnihotri, S.; Mota, J. P. B.; Rostam-Abadi, M.; Rood, M. J. J. Phys. Chem. B 2006, 110, 7640. (31) Agnihotri, S.; Mota, J. P. B.; Rostam-Abadi, M.; Rood, M. J. Langmuir 2005, 21, 896. (32) Agnihotri, S.; Zheng, Y. J.; Mota, J. P. B.; Ivanov, I.; Kim, P. C. J. Phys. Chem. C 2007, 111, 13747. (33) Kim, P.; Zheng, Y. J.; Agnihotri, S. Ind. Eng. Chem. Res. 2008, 47, 3170. (34) Yang, C. M.; Kaneko, K.; Yudasaka, M.; Iijima, S. Physica B-Condensed Matter 2002, 323, 140. (35) Alcaniz-Monge, J.; Linares-Solano, A.; Rand, B. J. Phys. Chem. B 2002, 106, 3209. (36) Walton, K. S.; Snurr, R. Q. J. Am. Chem. Soc. 2007, 129, 8552. (37) Mao, S.; Kleinhammes, A.; Wu, Y. Chem. Phys. Lett. 2006, 421, 513. (38) Alcaniz-Monge, J.; Linares-Solano, A.; Rand, B. J. Phys. Chem. B 2001, 105, 7998. (39) Ruthven, D. M. Principles of Adsorption and Adsorption Processes; Wiley: New York, 1984. (40) Jayendra Kumar, A.; Singh, R. R. B.; Patil, G. R.; Patel, A. A. LWT - Food Science and Technology 2005, 38, 303. (41) Brennan, J. K.; Bandosz, T. J.; Thomson, K. T.; Gubbins, K. E. Colloid Surf. A-Physicochem. Eng. Asp. 2001, 187, 539. (42) Ohba, T.; Kaneko, K. Mol. Phys. 2007, 105, 139. (43) Garcia-Perez, J. V.; Carcel, J. A.; Clemente, G.; Mulet, A. LWTFood Sci. Technol. 2008, 41, 18.

JP900609A