Raman Spectroscopic Investigation of Gas Interactions with an

Jan 4, 2006 - Christopher Matranga,*,† Bradley Bockrath,† Nitin Chopra,‡ Bruce J. Hinds,‡ and. Rodney Andrews§. National Energy Technology La...
0 downloads 0 Views 307KB Size
Download PDF
Langmuir 2006, 22, 1235-1240

1235

Raman Spectroscopic Investigation of Gas Interactions with an Aligned Multiwalled Carbon Nanotube Membrane Christopher Matranga,*,† Bradley Bockrath,† Nitin Chopra,‡ Bruce J. Hinds,‡ and Rodney Andrews§ National Energy Technology Laboratory, United States Department of Energy, 626 Cochrans Mill Road, Pittsburgh, PennsylVania 15236, Department of Chemical and Materials Engineering, UniVersity of Kentucky, Lexington, Kentucky 40506, and Center for Applied Energy Research, 2540 Research Park DriVe, Lexington, Kentucky 40511 ReceiVed June 20, 2005. In Final Form: NoVember 22, 2005 Raman spectroscopy has been used to investigate ethane, propane, and SF6 interactions with an aligned multiwalled carbon nanotube (MWNT) membrane. Pressures of 7.5-9.3 atm and temperatures of 293-333 K were examined for propane and SF6, whereas slightly lower temperatures (263-293 K) and pressures (6.7-7.5 atm) were used for ethane. Red-shifting and broadening is seen for the C-C stretching vibrations of the two hydrocarbons, as well as for the A1g symmetric vibration (ν1) of SF6. These spectral features indicate that the interaction between the gas and the nanotube membrane is capable of perturbing molecular vibrations and creating red-shifted features. Control experiments done on polycrystalline graphite and a polystyrene blank indicate that this spectral behavior is unique to gases interacting with the nanotubes in the membrane.

Introduction In this paper, we report a Raman spectroscopic investigation of gases interacting with a membrane which was fabricated by incorporating an array of aligned multiwalled carbon nanotubes (MWNTs) into a polystyrene film.1 These measurements establish that gases adsorbed in the pores of the MWNTs are the source of the observed spectral features. The observation of adsorbed gases under these conditions suggests that the nanotube surface should play a role in molecular transport in this membrane. These findings illustrate the potential of these MWNT membranes for use in gas separation applications. The measurement technique itself, as well as our findings regarding how gases interact with this MWNT membrane, should be of broad interest to those working on both basic and applied membrane research projects. Our interest in these membranes stems from the fact that the inner diameters of the MWNT nanotubes used start to approach molecular size scales. Nitrogen porosity and transmission electron microscopy (TEM) data indicate that the inner pores of these MWNTs are ∼4-7 nm in diameter.1 The size of the nanotubes in the membrane can be controlled, in part, by varying the diameter of the catalyst particle, so there is potential for porosity tuning in these membranes.2,3 In fact, functionalization was shown in several studies to drastically affect the flux of various species through this MWNT membrane.1,4,5 This level of control illustrates that it is possible to tune the molecular transport properties of this system in a systematic fashion. * To whom correspondence should be addressed. E-mail: matranga@ netl.doe.gov. † National Energy Technology Laboratory. ‡ University of Kentucky. § Center for Applied Energy Research. (1) Hinds, B. J.; Chopra, N.; Rantell, T.; Andrews, R.; Gavalas, V.; Bachas, L. G. Science 2004, 303, 62. (2) Ago, H.; Komatsu, T.; Ohshima, S.; Kuriki, Y.; Yumura, M. Appl. Phys. Lett. 2000, 77, 79. (3) Kanzow, H.; Lenski, C.; Ding, A. Phys. ReV. B 2001, 63, 125402. (4) Nednoor, P.; Chopra, N.; Gavalas, V.; Bachas, L. G.; Hinds, B. J. Chem. Mater. 2005, 17, 3595. (5) Majumder, M.; Chopra, N.; Hinds, B. J. J. Am. Chem. Soc. 2005, 127, 9062.

Recent computational studies have predicted both high selectivity and flux in carbon nanotube materials which makes them ideal candidates for gas adsorption and separation studies.6-9 In most cases, the transport of light gases such as H2 and CH4 were many orders of magnitude higher than other materials with comparable pore sizes.9 These transport properties result from the smooth potential energy surface which describes the interactions of the gas with the inner endohedral pore of the carbon nanotube.9 It is therefore desirable to understand how gases interact with nanotube materials, particularly when they are incorporated into a functioning separation membrane. Our primary motivation for the current study was to better understand both adsorption and molecular transport in these MWNT membranes. In particular, we would like to clearly establish whether the nanotubes themselves, the polymer binding matrix, or defects were responsible for the properties of this membrane. We would also like to develop simple analytical techniques to aid with in situ membrane characterization. In this regard, we decide to use vibrational spectroscopy because molecular vibrations are sensitive to their local environment. Recent work using transmission Fourier transform infrared spectroscopy (FTIR) has been very successful in exploiting this sensitivity in order to understand the different adsorption sites in heterogeneous nanotube systems.10-21 (6) Ackerman, D. M.; Skoulidas, A. I.; Sholl, D. S.; Johnson, J. K. Mol. Simul. 2003, 29, 677. (7) Sokhan, V. P.; Nicholson, D.; Quirke, N. J. Chem. Phys. 2004, 120, 3855. (8) Chen, H.; Sholl, D. S. J. Am. Chem. Soc. 2004, 126, 7778. (9) Skoulidas, A. I.; Ackerman, D. M.; Johnson, J. K.; Sholl, D. S. Phys. ReV. Lett. 2002, 89, 185901. (10) Matranga, C.; Bockrath, B. J. Phys. Chem. B 2004, 108, 6170. (11) Matranga, C.; Bockrath, B. J. Phys. Chem. B 2004, 109, 4853. (12) Matranga, C.; Bockrath, B. J. Phys. Chem. B 2005, 109, 9209. (13) Matranga, C.; Chen, L.; Bockrath, B.; Johnson, J. K. Phys. ReV. B 2004, 70, 165416. (14) Matranga, C.; Chen, L.; Smith, M.; Bittner, E.; Johnson, J. K.; Bockrath, B. J. Phys. Chem. B 2003, 107, 12930. (15) Yim, W.; Byl, O.; Yates, J. T.; Johnson, J. K. J. Chem. Phys. 2004, 120, 5377. (16) Byl, O.; Kondratyuk, P.; Forth, S.; Fitzgerald, S.; Yates, J. T. J. Am. Chem. Soc. 2003, 125, 5889. (17) Byl, O.; Kondratyuk, P.; Yates, J. T. J. Phys. Chem. B 2003, 107, 4277.

10.1021/la0516577 CCC: $33.50 © 2006 American Chemical Society Published on Web 01/04/2006

1236 Langmuir, Vol. 22, No. 3, 2006

Matranga et al.

We chose Raman microscopy instead of our previously reported FTIR technique10-14 because the large optical density of the MWNT membrane prohibited any type of infrared transmission study. In fact, the Raman technique appears, in several regards, to be superior, since we are able to study MWNT membranes at pressures and temperatures which are realistic for separation applications. Thus, we feel that this technique has potential for concurrent, in situ, studies of membranes being used in gas separation experiments. Experimental Section The MWNT membranes used in this study have been described previously.1 Vertically aligned MWNTs of 5-10 µm length and ∼4-7 nm inner diameters are grown on a quartz substrate. Polystyrene is then spin coated onto the substrate to fill the void spaces between the MWNT cylinders. The quartz substrate is etched with HF acid producing a free-standing composite film with MWNT alignment from top to bottom. The MWNTs are then opened with a H2O plasma-enhanced oxidation. Transmission electron microscopy shows that about 70% of the tips are opened by the plasma oxidation process. More details regarding the membrane fabrication and characterization can be found in the literature.1 Raman spectra were collected with the confocal microscope attachment of a Nicolet Almega dispersive Raman spectrometer with a 532 nm excitation source. The laser power was varied between 15 and 150 mW and focused onto the cell with a 20 X microscope objective (Mitutoyo Brand mPlan Apo SL) having a 30 mm working distance. Typical spectral resolution was ∼2 cm-1 and scan times were 1000-2000 seconds total. Experiments were conducted in a high pressure, stainless steel, optical cell (Reflex Analytical Corporation) with CaF2 windows (13 mm diameter by 6 mm thickness). The path length in the cell can be varied by using stainless steel spacers between the windows. For the current study, we use a path length of ∼5 mm. The cell is capable of working at pressures from 1 to 330 atm and temperatures from 253-473 K. A portion of the membrane (∼2 mm × 2 mm) was mounted in the cell by sandwiching the membrane between 2 pieces of copper foil which had a small hole punched through them to allow for optical access to the MWNT membrane. The foil was sandwiched between the window and the 5 mm spacer in such a way as to allow gas access to the membrane. The size of the copper foil was significantly smaller than the ∼13 mm diameter optical window which allowed us to collect spectra on both the MWNT membrane and the gas itself (see below). Spectra were collected by focusing into the membrane while it was mounted in a pressurized cell. The focus and alignment of the system was optimized by maximizing the signal from the MWNT bands at 1570 cm-1 (G band) and 1340 cm-1 (D band). The observation of these bands ensured that we were in fact seeing signal from the membrane itself. The laser intensity was varied and the D band intensity was monitored for signs of laser damage. Laser damage was not typical, but would occasionally occur in certain regions of the sample for undetermined reasons. Sometimes during long experimental runs (several hours), laser damage would be noted over the course of a scan. All runs involving laser damage were discarded. To get the spectrum of the gas at the same temperature, pressure, and spectrometer settings used for the MWNT membrane run, we simply translate the cell away from the copper foil used to hold the sample and focus the laser into the void space created by the 5 mm spacer. This was done either right before or directly after a MWNT (18) Feng, X.; Irle, S.; Witek, H.; Morokuma, K.; Vidic, R.; Borguet, E. J. Am. Chem. Soc. 2005, 127, 10533. (19) Feng, X.; Matranga, C.; Vidic, R.; Borguet, E. J. Phys. Chem. B 2004, 108, 19949. (20) Ellison, M. D.; Crotty, M. J.; Koh, D.; Spray, R. L.; Tate, K. E. J. Phys. Chem. B 2004, 108, 7938. (21) Ellison, M. D.; Good, A. P.; Kinnaman, C. S.; Padgett, N. E. J. Phys. Chem. B 2005, 109, 10640.

Figure 1. (A) Raman spectra of ethane interacting with a MWNT membrane, polystyrene blank, and graphite sample at 263 K and 6.8 atm. (B) Illustration of the spectral fits discussed in the text for ethane adsorbed in the MWNT membrane. Experimental data is from (A). membrane spectrum was taken. Under these conditions, we see no discernible variations of the gas-phase spectrum by changing the temperature. In all experiments, the cell was charged to 7.5 atm at 293 K, sealed, and placed in the Raman microscope. The temperature was then varied as desired and no new gas was introduced into the cell. This creates minor variations (10-25%) in cell pressure during heating/cooling runs as determined through the ideal gas equation. It does keep the total number of moles of gas in the cell at a constant value (∼0.20 × 10-7 mol). Ethane, propane, and SF6 were chosen since they all have one or more vibrations with fairly large Raman cross sections. The primary vibrations chosen for study occur between ∼750-1000 cm-1. This region of the MWNT Raman spectrum does not have any real structure thus removing problems associated with separating contributions of the adsorbed gas from those of the membrane. Data are also presented for C-H vibrations in the 28003000 cm-1 region. Since the MWNT and graphite spectra are fairly flat in the 7501000 cm-1 region, background subtractions for gases interacting with these materials were done simply by subtracting a constant from the spectrum to set the baseline to zero in this region. A similar subtraction was used in the 2800-3000 cm-1 region. The polystyrene spectrum has structure in the 750-1000 cm-1 which was subtracted out by using a spectrum of polystyrene taken in room air before the cell was charged with gas. The Raman intensity of each peak was then normalized to its own peak intensity for direct comparisons between the gas, graphite, polystyrene, and membrane samples. For runs where the temperature was varied, the Raman intensities were normalized to the MWNT G band peak intensity. No systematic temperature variations were seen with the G band intensity.

Results and Discussion The Raman spectra for ethane are shown in Figure 1. A single peak is seen at 993 cm-1 for the gas-phase species. This peak arises from the C-C stretching mode (ν3) of gas-phase ethane.22 The spectra for ethane with graphite and the polystyrene blank are essentially identical to what is seen for gas-phase ethane. The surface area of these materials is small, so the signal from adsorbed molecules should be negligible. We therefore attribute the feature (22) Herzberg, G. Infrared and Raman Spectra; D. Van Nostrand Company: Princeton, 1960; p 286.

Interactions with an Aligned MWCN Membrane

at 993 cm-1 in the graphite and polystyrene spectra to gas-phase ethane near the surface of these materials. A similar result was seen in Raman studies of H2 interacting with the low surface area highly ordered pyrolytic graphite (HOPG) material.23 The spectrum for ethane interacting with the MWNT membrane shows significant differences from the gas-phase spectrum. A small (∼2 cm-1), but discernible, red-shifting is seen. A slight, asymmetric, broadening also appears to occur on the lower energy side of the peak. We expect that the line shapes from the MWNT membrane most likely include contributions from species inside of the nanotubes. As a simple approximation, we have modeled the spectra from the MWNT membrane using a line shape composed of 2 Lorentzians (Figure 1B). Clearly, more complicated line shapes (i.e., Voigt) could also be employed. For ethane, we find that this simple model does a fairly adequate job of reproducing the observed spectral features. Overall, we can see that most of the red-shifting and broadening seen with respect to the gas-phase spectrum are accounted for. The spectrum in Figure 1B is fit with features at 992 and 989 cm-1. The integrated intensity ratios are approximately 1:2 with full width half-maximums (fwhm) of ∼3 and 6 cm-1, respectively. The feature at 992 cm-1 is barely shifted from the gas-phase value of 993 cm-1 and as such we suspect it most likely arises from a weakly bound species. The component at 989 cm-1 is shifted by 4 cm-1 and most likely arises from a more strongly sorbed species. The relative populations for the ethane associated with the 992 and 989 cm-1 feature can be estimated using the Lorentzian fit data yielding approximately 2 times as many molecules in the site associated with the 989 cm-1 feature as in the site associated with the 992 cm-1 peak. Recent simulations reported for ethane adsorption in a 2.719 nm (20, 20) and 4.077 nm (40, 40) diameter nanotubes can shed some light on the spectra in Figure 1.24 The nanotube diameters used in these simulations are close to the diameter range of MWNTs present in our membranes. In these simulations, the local density profiles for ethane at 1 atm and 300 K in both the (20, 20) and (40,40) nanotubes indicate that the molecule forms a monolayer at the wall. Both methyl groups are found to be lying predominately on the wall indicating that relatively few ethane molecules have their bond axis pointed toward the center of the nanotube. A less ordered, gaslike phase exists between this monolayer and the center of the tube. It therefore seems plausible that the red-shifted spectral features seen in Figure 1 are related to this monolayer and less ordered gas phase. If we blindly assume that the degree of interaction with the wall is proportional to the magnitude of the red-shift, then we would most likely associate the peak at 989 cm-1 with the adsorbed monolayer and the 992 cm-1 peak with the less ordered, gaslike phase. More insight into ethane behavior in nanotubes can be derived from recent classical molecular dynamics simulations and density functional theory calculations on nanotubes ranging from 1.3 to 2.2 nm in diameter. The largest tubes used in these simulations approach the smallest tube diameters used in our MWNT membrane. The calculation predicts that ethane aligns its molecular axis with the C-C bonds in the tube wall.25 Diffusion down the length of the tube maintains the C-C bond alignment between the nanotube and the ethane molecule. This adsorption and diffusion geometry creates a spiral diffusion path along the inner surface of the nanotube.25 The alignment of the ethane (23) Williams, K. A.; Bhabendra, K. P.; Eklund, P. C.; Kostov, M. K.; Cole, M. W. Phys. ReV. Lett. 2002, 88, 165502. (24) Zhang, X.; Wang, W. Phys. Chem. Chem. Phys. 2002, 4, 3048. (25) Mao, Z.; Sinnott, S. B. Phys. ReV. Lett. 2002, 89, 278301.

Langmuir, Vol. 22, No. 3, 2006 1237

Figure 2. (A) Raman spectra of propane interacting with a MWNT membrane, polystyrene blank, and graphite sample at 293 K and 7.5 atm. (B) Spectral fits discussed in the text for propane adsorbed in the MWNT membrane. The experimental Raman spectrum is from (A).

C-C bond axis with the nanotube wall agrees, qualitatively, with the simulations mentioned above, which also find that ethane lies on the nanotube wall.24 The Raman spectra for propane are displayed in Figure 2. A spectral feature for gas-phase propane is seen at 869 cm-1 which is assignable to the known ν8 C-C stretching vibration reported, previously.22 Much like ethane, propane interacting with the MWNT membrane produces red-shifted spectra. This shifting is a clear indication that propane is interacting with the MWNTs in the membrane. We would expect for many aspects of propane adsorption to mimic what is seen in the molecular simulations for ethane. In particular, we would expect for adsorption to occur near the nanotube walls and for a more gaslike phase to occur toward the center of the nanotube. The fits shown in Figure 2B, in part, support this similarity. We are capable of fitting the spectra to a 2 Lorentzian line shape with features at 867 and 863 cm-1. The intensity ratio is approximately 1:1.5. The fwhms are ∼4 cm-1 for the features at 867 and ∼8 cm-1 for the feature at 863 cm-1. From a spectroscopy standpoint, we can clearly see that many similarities occur between ethane and propane adsorption in these MWNT membranes. Figure 3 shows the results for Raman experiments involving SF6. The Raman peak seen at ∼774 cm-1 for gas-phase SF6 is easily assignable to the totally symmetric ν1(a1g) fundamental vibration.22 For the MWNT membrane, much of this feature is red-shifted from the gas-phase value. The fits to the spectra (Figure 3B) are achieved with features at 772 and 769 cm-1 having an integrated intensity ratio of ∼1:1.6. The fwhms are approximately 3 and 7 cm-1 for the 772 and 769 cm-1 features, respectively. As for the case of ethane and propane, we attribute the features at 772 and 769 cm-1 as occurring because of SF6 inside of the MWNT nanotubes. The red-shifting seen in Figures 1-3 is a common spectral behavior when molecules adsorb in porous carbons. This type of behavior has been reported in FTIR studies of CO2,10,12-15 CO,11 NO,17 CF4,16 SF6,12 NH3,18,20 and H2O21 in single-walled carbon nanotube (SWNT) systems. It has been associated with physisorption in the endohedral, interstitial, and groove/external

1238 Langmuir, Vol. 22, No. 3, 2006

Matranga et al.

Figure 4. Effects of laser intensity on the Raman spectra of both the gas phase and MWNT sample.

Figure 3. (A) Raman spectra of SF6 interacting with a MWNT membrane, polystyrene blank, and graphite sample at 293 K and 7.5 atm. (B) Illustration of the spectral fits discussed in the text for ethane adsorbed in the MWNT membrane. Experimental data is from (A).

surface sites in these SWNT bundles. Red-shifting is also seen when gases such as CO2 interact with different types of coal.26 Red-shifting and broadening has also been observed for CO2 which is permanently trapped within the endohedral and interstitial pores of SWNT bundles at temperatures from 5 to 700 K.10,12,14,19 At low temperatures, molecular simulations show that this confined CO2 is mostly adsorbed in a solidlike phase near the nanotube wall.14 As the temperature is increased, the molecules desorb and form a gaslike phase with a uniform density across the tube.14 The experimental observation that this spectral redshifting occurs even at temperatures where simulations show that the trapped molecules are in a gaslike phase inside of the nanotube clearly indicates that the confinement effect alone can be enough to perturb these molecular vibrations. This illustrates that the presence of a clear-cut, adsorbed phase, is not necessary for red-shifting and broadening to occur in a vibrational spectrum. In Raman experiments studying H2 physisorption in SWNT bundles, both red and blue-shifted vibrational peaks are found.23 The shifts seen are typically less than ∼2 cm-1 and are comparable in magnitude to the shifts seen in Figures 1-3. The blue-shifted peaks in the previous H2/SWNT study were assigned to adsorption on the inner and outer surfaces of the SWNTs.23 The red-shifted peaks were assigned to H2 adsorbed in interstitial spaces.23 These assignments were made by using perturbation theory to calculate the frequency shift associated with each adsorption site which yielded surprisingly good quantitative agreement with the experimental values.23 In a semiclassical molecular dynamics calculation for H2 in SWNT bundles, only red-shifted features of ∼16 cm-1 are found for adsorbed species in 2.0 nm nanotubes.27 The magnitude of the red-shifting increases with decreasing nanotube diameter.27 In first-principles molecular dynamics simulations, large (∼40-60 cm-1) red-shifted features are found.28 These features are attributed to H2 with long residence (26) Goodman, A. L.; Campus, L. A.; Schroeder, K. T. Energy Fuels 2005, 19, 471. (27) Frankland, S. J. V.; Brenner, D. W. Chem. Phys. Lett. 2001, 334, 18. (28) Canto, G.; Ordejon, P.; Cheng, H.; Cooper, A. C.; Pez, G. P. New J. Phys. 2003, 5, 124.

times on the nanotube wall.28 Smaller magnitude red- and blueshifting (less than 10 cm-1) was found to occur from molecules which were colliding with the nantoube walls.28 The shifts in both the semiclassical27 and first-principles28 molecular dynamics studies are much larger in magnitude than anything reported experimentally for molecular adsorption of hydrogen in nanotubes,23 amorphous carbons,29 or metal-organic frameworks,30 but we expect that the general trends of spectral shifting, the effects of nanotube size, and system temperature should be valid. To check that laser heating does not cause the spectral features seen in Figures 1-3, we have varied the laser intensity during Raman experiments. A representative example is shown in Figure 4. Changing the laser power from 150 to 15 mW does not have any discernible effect on the spectra. This clearly illustrates that direct heating of the gas is not occurring. It also rules out an indirect process involving laser heating of the MWNTs which in turn heat the gas inside the nanotubes. We have also examined the temperature dependence of the Raman bands associated with ethane, propane, and SF6 adsorption (Figure 5). In all cases, there is a decrease in the normalized Raman intensity with temperature which is consistent with a physisorption type interaction. This finding supports our interpretation that red-shifted features seen in Figures 1-3 occur because of gases physisorbed inside the MWNT pores. For each gas, the normalized Raman intensity decreases by about a factor of 3 over the temperature ranges studied. The Raman intensity should be directly proportional to the number of molecules adsorbed in the sample so this indicates a similar change in physisorption population. A population change of this magnitude is typical for ethane in porous materials at these temperatures and pressures.31 During these intensity changes, the pressure in our sealed cell only changes by 10-25% over the two temperature ranges investigated so we anticipate that most of the intensity change occurs solely because of desorption. Cooling the cell after a heating run returns the integrated intensity to within ∼1020% of its original value indicating that the spectral changes are reversible. The intensity variations seen in Figure 5 clearly point to a physisorption type interaction between the gases and the MWNTs. This observation is consistent with the physisorption predicted for ethane in nanotubes of this size at comparable temperatures and pressures.24 Figure 5 also shows that the adsorption is not associated with any type of activated process which could potentially be expected if dissolution in the polymer binding matrix or intercalation between graphitic layers in the MWNT were occurring. If an activated process were occurring, we would (29) Centrone, A.; Siberio-Perez, D. Y.; Millward, A. R.; Yaghi, O.; Matzger, A. J.; Zerbi, G. Chem. Phys. Lett. 2005, 411, 516. (30) Centrone, A.; Brambilla, L.; Zerbi, G. Phys. ReV. B 2005, 71, 245406. (31) Chaudhary, V. R.; Mayadevi, S. Zeolites 1996, 17, 501.

Interactions with an Aligned MWCN Membrane

Figure 5. Normalized Raman intensity as a function of temperature for (A) ethane, (B) propane, and (C) SF6. The spectra were normalized to the MWNT G-Band at ∼1578 cm-1 (see text).

expect more Raman intensity with increasing temperatures reflecting the expected increase in molecules crossing the activation barrier. In addition to ethane, propane, and SF6, several other gases were surveyed at similar temperatures and pressures. For CF4, CH4, CO, CO2, ethylene, and propylene, we do not see any appreciable red-shifting or broadening of the Raman active bands at 7.5 atm and 293 K. This seems to indicate that the vibrations of these molecules are not as strongly perturbed by the MWNT pores as is the case with ethane, propane and SF6. Our inability to detect red-shifting and broadening for physisorbed molecules could result from a smaller Raman cross section, lower coverage, or a combination of both factors. The adsorption geometry of the molecule and the effect this has on perturbing the molecular vibration should also have a role in the relative magnitude seen for red-shifted features. Binding energy data supports our hypothesis that the difficulty in detecting red-shifted Raman features in some of these lighter molecules is largely a result of lower coverage for these adsorbates. Experimentally, the isosteric heats of adsorption for the first monolayer of propane32 and CH433 on planar graphite are 30 and 12 kJ/mol, respectively. SF6 yields isosteric heats ∼25 kJ/mol34 on various nonporous, powdered, graphitic carbons. CO2 on graphitized carbon black is found to have an isosteric (32) Zhao, X.; Kwon, S.; Vidic, R.; Borguet, E.; Johnson, J. K. J. Chem. Phys. 2002, 117, 7719. (33) Weber, S. E.; Talapatra, S.; Journet, C.; Zambano, A. Z.; Migone, A. D. Phys. ReV. B 2000, 61, 13150.

Langmuir, Vol. 22, No. 3, 2006 1239

heat of 14.4 kJ/mol in the zero coverage limit.35 The groundstate energy of CF4 on graphite is estimated to be about 12.4 kJ/mol in a recent computational study.36 Based on the data presented from simple, nonporous, graphite surfaces, we would expect lower coverages for lighter molecules such as CH4, CF4, and CO2 than heavier molecules such as SF6 and propane under the same conditions of temperature and pressure. For ethylene, we see an interesting case where neither binding energies nor Raman cross-sections can easily explain why we do not easily see red-shifted Raman features. On graphitized carbon black, ethylene37-39 and ethane39 are both found to have zero coverage isosteric heats of ∼18 kJ/mol in simulations and experiments. Previous experiments using a 488 nm laser excitation source find that the differential Raman cross section (∂σ/∂Ω) for the CdC stretching mode (ν2) of gas-phase ethylene is 80.7 × 10-36 square meters per steradian (m2/sr) which converts to 64.2 × 10-36 m2/sr for a 514.5 nm excitation source.40 From ethane experiments with a 514.5 nm excitation source, the C-C stretching mode (ν3) is found to have a ∂σ/∂Ω of 64.4 × 10-36 m2/sr.41 It is convenient to also consider the mean derivative of the molecular polarizability tensor (∂R/∂qi) which can be estimated using41 ∂σ/∂Ω ) (π2/202)(ν0 - νi)4[1- exp(-hcνi/kT)]-1(∂R/ ∂qi)2, where h and k are Plank’s and Boltzman’s constants, T is temperature, ν0 and νi are the wavenumber values of the exciting radiation and vibrational band frequency (respectively), and 0 is the permittivity of vacuum (8.8542 × 10-12 Cm2/V). Through this relation and the differential Raman cross sections,40,41 we estimate ∂R/∂qi to be roughly 0.10 × 10-40 Cm2/V2 for both vibrational modes at 300 K. All of these estimates indicate that for the same coverage we should expect a similar magnitude of Raman signal for both molecules. As illustrated above, one would expect ethane and ethylene to behave similarly in these Raman experiments, yet within the resolution of our experiment, we only observe red-shifting for ethane. We do not fully understand the lack of observable redshifted features for ethylene and at present can only speculate that it must be related, in part, to how adsorption in the MWNT perturbs a particular vibration. This concept is supported, in part, by results from our current investigation. For ethane and propane, we are able to observe the red-shifted features associated with the C-C stretching modes for physisorbed species, but we do not see any convincing evidence of red-shifting in the C-H stretching vibrations around 2800-3000 cm-1 (Figure 6). The gas phase ∂σ/∂Ω values for these C-H vibrations are ∼ 5-8 times larger40,41 than what is reported for the C-C and CdC vibrations displayed in Figures 1 and 2 so the cross sections should not be responsible for our inability to observe red-shifting. For SF6, the only Raman active band with a large cross section22 is the ν1 band at ∼775 cm-1 noted in Figure 3 so a similar comparison with different vibrations is not possible for this adsorbate. The data presented in Figure 6 illustrates that coverage and Raman cross sections are not the only factors which dictate whether red-shifted features will be observed for physisorbed (34) Pribylov, A. A.; Kalinnikova, I. A.; Regen, N. I. Russ. Chem. Bull. 2003, 52, 882. (35) Bottani, E. J.; Ismail, I. M. K.; Bojan, M. J.; Steele, W. A. Langmuir 1994, 10, 3805. (36) Stan, G.; Bojan, M. J.; Curtarolo, S.; Gatica, S. M.; Cole, M. W. Phys. ReV. B 2000, 62, 2173. (37) Battezzati, L.; Pisani, C.; Ricca, F. J. Chem. Soc., Faraday Trans. 1975, 71, 1629. (38) Kalashnikova, E. V.; Kiselev, A. V.; Petrova, R. S.; Shcherbakova, K. D. Chromatographia 1971, 4, 495. (39) Do, D. D.; Do, H. D. Langmuir 2004, 20, 10889. (40) Orduna, M. F.; del Olmo, A.; Domingo, C.; Montero, S. J. Mol. Struct. 1986, 142, 201. (41) Gough, K. M.; Murphy, W. F. J. Chem. Phys. 1986, 85, 4290.

1240 Langmuir, Vol. 22, No. 3, 2006

Matranga et al.

Overall, the data presented here shed light on how gases interact with MWNT membranes. Our results clearly indicate that gas physisorption inside of the MWNTs is responsible for the observed spectral features for ethane, propane, and SF6. Even for larger tubes where size exclusion cannot occur, surface adsorption can be advantageous for high selectivity and flux in membrane structures.9 Our findings illustrate that adsorption on the nanotube surfaces occurs under the experimental conditions used in this paper. This suggests that the nanotube surfaces should play a significant role in any molecular transport occurring through these membrane structures under these conditions. It also illustrates that adsorption significantly perturbs the vibrations of these molecules. Any computational modeling of diffusion in these systems must be able to capture the role of the MWNT surface in order to accurately reproduce experimental findings.

Summary

Figure 6. (A) Raman spectra of the C-H stretching region for ethane interacting with a MWNT membrane at 293 K and 7.5 atm. (B) Raman spectra of the C-H stretching region for ethane interacting with a MWNT membrane at 293 K and 7.5 atm.

species. It also appears that the adsorption geometry and the role this plays on perturbing the molecular vibration must also be considered.

A Raman spectroscopic study of the interactions of several gases with an aligned multiwalled carbon nanotube membrane was conducted. Red-shifting is seen for ethane, propane, and SF6 during adsorption. By using previously published computational results, we are able to attribute the spectral changes to gases interacting with the inner pores of the MWNT membrane. Our results clearly show that the nanotubes in this composite structure are responsible for its adsorption and transport properties. Acknowledgment. We thank Ed Bittner, Milton Smith, David Sholl, and Karl Johnson for useful discussions. Reference in this work to any specific commercial product is to facilitate understanding and does not necessarily imply endorsement by the United States Department of Energy. LA0516577