2890
Langmuir 2004, 20, 2890-2899
Near-Critical CO2 in Mesoporous Silica Studied by In Situ FTIR Spectroscopy Michael S. Schneider, Jan-Dierk Grunwaldt, and Alfons Baiker* Institute for Chemical and Bioengineering, ETH Ho¨ nggerberg - HCI, CH-8093 Zu¨ rich, Switzerland Received September 24, 2003. In Final Form: January 16, 2004 Attenuated total reflection Fourier transform infrared spectroscopy was used to correlate the band shift of the ν2 vibrational band of carbon dioxide with the density of the fluid. Upon adsorption of CO2 on mesoporous silica and a nonporous SiO2 film, additional bands were detected due to interactions of CO2 with SiO2. Near the saturation pressure for the porous samples, the absorbance of the ν2 band increased strongly, which was concluded to be caused by liquidlike CO2 inside the pores. Integration of singlebeam-sample-reference spectra between bulk CO2 and CO2 adsorbing on the mesoporous silica coated on one part of the internal reflection element revealed excess adsorption type isotherms with sharp maxima at 21 °C. A flatter curve shape could be observed at 25 °C, which allowed estimating the pore critical temperature. Moreover, the density of the fluid inside and outside the pores could be compared. Over the investigated ranges of pressure, temperature, and pore size, the results evidenced that the CO2 density was always higher in the silica pores than in the bulk, even under supercritical conditions. This has important consequences on the pressure dependence of dissolution power and diffusivity of fluids in mesoporous solids. An overview is given on the influences of fluid phase behavior in the bulk and in the pores at various conditions on solubility and diffusivity.
Introduction In recent years, homogeneously and heterogeneously catalyzed reactions in supercritical fluids (SCFs) have received much attention (cf. refs 1-5 and references therein). The combination of fair and tuneable solvent power, complete miscibility with gases, and advantageous mass transport properties render SCFs very attractive solvents, especially in cases when the conventional process is mass transport limited. One of the most often used SCFs to replace organic solvents is carbon dioxide, which has the additional benefits of being environmentally benign, readily available, and nontoxic.6-8 It is nowadays accepted that careful investigation of the phase behavior of reaction mixtures with solvents above their critical temperature is mandatory, as it may be very complex compared to the pure solvent’s phase behavior.2,9-11 By weighing model accuracy against complexity, it has been shown that the number of components in a mixture can be virtually reduced to two.2,12 The phase behavior of binary mixtures contains all important phen* To whom correspondence should be addressed. E-mail:
[email protected]. Tel: +41 1 632 31 53. Fax: +41 1 632 11 63. (1) Baiker, A. Chem. Rev. 1999, 99, 453. (2) Grunwaldt, J. D.; Wandeler, R.; Baiker, A. Catal. Rev.sSci. Eng. 2003, 45, 1. (3) Noyori, R. Chem. Rev. 1999, 99, 353. (4) Jessop, P. G.; Leitner, W. Chemical Synthesis Using Supercritical Fluids; Wiley-VCH: Weinheim, 1999. (5) Savage, P. E. In Handbook of Heterogeneous Catalysis; Ertl, G., Kno¨zinger, H., Weitkamp, J., Eds.; Wiley-VCH: Weinheim, 1997. (6) Black, H. Environ. Sci. Technol. 1996, 30, 124A. (7) Kaupp, G. Angew. Chem. 1994, 106, 1519. (8) Hyde, J. R.; Licence, P.; Carter, D.; Poliakoff, M. Appl. Catal., A 2001, 222, 119. (9) Ke, J.; Han, B. X.; George, M. W.; Yan, H. K.; Poliakoff, M. J. Am. Chem. Soc. 2001, 123, 3661. (10) Arai, K.; Adschiri, T. Fluid Phase Equilib. 1999, 158-160, 673. (11) Peters, C. J. In Supercritical Fluids: Fundamentals for Application; Kiran, E., Levelt Sengers, J. M. H., Eds.; Kluwer Academic: Dordrecht, 1994. (12) Wandeler, R.; Ku¨nzle, N.; Schneider, M. S.; Mallat, T.; Baiker, A. J. Catal. 2001, 200, 377.
omena which can generally be encountered in multicomponent systems, such as for example composition-dependent phase transition lines, liquid-liquid immiscibility, gas-gas equilibrium, and retrograde condensation.13 However, knowledge of solubility and phase transition lines is not sufficient to fully understand mass transport phenomena of reactions catalyzed by mesoporous solids. The situation is usually simplified by assuming the supercritical (sc) solvent as a continuum and solutes partitioning between the surface and the fluid. But the introduction of wall forces and the competition between fluid-wall and fluid-fluid forces can lead to surface-driven phase changes. Far below the solvent’s bulk critical temperature, adsorption isotherms are known to exhibit discontinuity and adsorption-desorption hysteresis caused by capillary condensation,14 for example, used to estimate the pore size distribution from nitrogen adsorption data using the Kelvin equation. On the basis of the classical theories of pore condensation, the condensation line would extend up to and terminate in the bulk critical point.15 However, theoretical models developed over the last two decades lead to the conjecture that a fluid confined between two parallel semi-infinite walls or in a single cylindrical pore reaches criticality at a temperature lower than the bulk critical one.16-19 Summarizing the recent progress in this field, one distinguishes between a hysteresis critical temperature Tch, where the adsorption hysteresis van(13) Rowlinson, J. S.; Swinton, F. L. Liquids and Liquid Mixtures; Butterworth: London, 1982. (14) Sing, K. S. W.; Everett, D. H.; Haul, R. A. W.; Moscou, L.; Pierotti, R. A.; Rouque´rol, J.; Siemieniewska, T. Pure Appl. Chem. 1985, 57, 603. (15) Thommes, M.; Findenegg, G. H.; Schoen, M. Langmuir 1995, 11, 2137. (16) Evans, R.; Marconi, U. M. B. J. Chem. Soc., Faraday Trans. 1986, 82, 1763. (17) Nakanishi, H.; Fisher, M. E. J. Chem. Phys. 1983, 78, 3279. (18) Burgess, C. G. V.; Everett, D. H.; Nuttall, S. Pure Appl. Chem. 1989, 61, 1845. (19) De, S.; Shapir, Y.; Chimowitz, E. H. J. Chem. Phys. 2003, 119, 1035.
10.1021/la035778n CCC: $27.50 © 2004 American Chemical Society Published on Web 02/21/2004
Near-Critical CO2 in Porous Silica
ishes, and the (higher) pore critical temperature Tcp where the jump in adsorption near the saturation pressure gives way to a continuous increase. Both critical temperatures are below the bulk critical temperature, but the differences from it decrease with increasing pore diameter.20 In other words, up to Tch capillary condensation occurs irreversibly, between Tch and Tcp it is reversible, and above Tcp, no capillary condensation should be observed. Although several measurements were published that confirm these models,18,21-26 there are still few experimental data for the commonly used supercritical solvents at temperatures and pressures of practical interest.27-30 In addition, the question arises of whether the solvent behaves differently within the pores compared to the bulk phase above the critical temperature of the fluid. For example, a recent kinetic Monte Carlo study on diffusion of a solute dissolved in a solvent near its critical point and embedded in a porous structure predicted that the solute flux through a membrane can stop, despite a finite value for the intrinsic solute diffusion coefficient.19 The experimental verification of such queries requires (quasi) simultaneous monitoring of density and concentration inside and outside the pores. As the infrared spectrum of CO2 strongly depends on the density,31-33 it is possible to investigate its phase behavior in porous solids with appropriate IR probing techniques and therefore it should be possible to distinguish between CO2 inside the pores and in the bulk fluid as long as the density is different. Note that performing such studies with pure fluids is still a simplification compared to real catalytic multicomponent systems, where, in addition to the bulk phase behavior features, the composition of the mixture will also vary within the pores, which again affects the solubility and phase behavior. Here, we tackle the question with pure fluids, as the occurring phenomena of near-critical fluids confined to pores are not well understood yet. For this purpose, attenuated total reflection Fourier transform infrared spectroscopy is applied to investigate adsorption and phase behavior of carbon dioxide in porous silica at elevated pressures and near-critical temperatures. Silica samples with defined pore sizes (2-14.5 nm) served as models for technically used mesoporous solids (catalysts, supports, adsorbents) and were compared to an evaporated SiO2 film. To be able to estimate the CO2 density in the pores, a relation between the ν2 vibrational band maximum and the fluid density was determined. Subsequently, isothermal pressure studies at sub- and supercritical conditions and isochoric temperature experiments were performed in order to get information on the (20) Gelb, L. D.; Gubbins, K. E.; Radhakrishnan, R.; SliwinskaBartkowiak, M. Rep. Prog. Phys. 1999, 62, 1573. (21) Morishige, K.; Fujii, H.; Uga, M.; Kinukawa, D. Langmuir 1997, 13, 3494. (22) Morishige, K.; Shikimi, M. J. Chem. Phys. 1998, 108, 7821. (23) Thommes, M.; Findenegg, G. H. Langmuir 1994, 10, 4270. (24) Machin, W. D. Langmuir 1999, 15, 169. (25) Michalski, T.; Benini, A.; Findenegg, G. H. Langmuir 1991, 7, 185. (26) Wong, A. P. Y.; Chan, M. H. W. Phys. Rev. Lett. 1990, 65, 2567. (27) Humayun, R.; Tomasko, D. L. AIChE J. 2000, 46, 2065. (28) Chen, J. H.; Wong, D. S. H.; Tan, C. S.; Subramanian, R.; Lira, C. T.; Orth, M. Ind. Eng. Chem. Res. 1997, 36, 2808. (29) Strubinger, J. R.; Parcher, J. F. Anal. Chem. 1989, 61, 951. (30) vonBehren, J.; Chimowitz, E. H.; Fauchet, P. M. Adv. Mater. 1997, 9, 921. (31) Buback, M.; Schweer, J.; Tups, H. Z. Naturforsch., A: Phys. Sci. 1986, 41, 505. (32) Buback, M.; Schweer, J.; Tups, H. Z. Naturforsch., A: Phys. Sci. 1986, 41, 512. (33) Yagi, Y.; Tsugane, H.; Inomata, H.; Saito, S. J. Supercrit. Fluids 1993, 6, 139.
Langmuir, Vol. 20, No. 7, 2004 2891
Figure 1. Schematic top view of the single-beam-samplereference (SBSR) setup. The movable blinds allow one to select one of the two optical pathways through the IRE; hence only one half is probed at a time. The silica layer is coated on one part of the crystal, and the carbon dioxide is only probed on the top surface.
phase behavior of near-critical CO2 confined to mesopores at conditions typically applied in catalytic reactions using scCO2. Experimental Section All experiments were carried out in a custom-made stainless steel high-pressure view-cell. The volume of the cylindrical reactor can be varied between 19.5 and 67.5 mL by means of a manual screw pump. A sapphire window covering the entire diameter allows visual observation and digital imaging of the bulk phase behavior. An internal reflection element (IRE; made of ZnSe; trapezoidal shape; angle of incidence, 60°; length, 27 mm; height, 2 mm; depth, 10 mm) mounted at the bottom of the hollow cylinder allows attenuated total reflection (ATR) measurements, where the evanescent electric field at the point of total reflection can be used to probe material close to the crystal surface. The penetration depth dp depends on the refractive indices of the IRE material and the sample layer (or bulk phase, respectively), the angle of incidence, and the wavelength.34 For example, assuming a wavenumber of 667 cm-1 and a sample layer refractive index of 1.3, dp is calculated to be 1.5 µm. Fourier transform infrared (FTIR) spectra were recorded with a Bruker IFS-66 spectrometer equipped with a liquid-nitrogen-cooled MCT detector, accumulating 20 scans at a resolution of 2 cm-1. A detailed description of the experimental setup has been published recently.35 For the present study, the ATR part of the cell has been extended in order to be able to perform single-beam-samplereference (SBSR) type measurements.36-38 The basic idea of SBSR is that one single ATR crystal is used to measure two similar but not equal spectra (e.g., cells with two separate compartments can be built, one contains only solvent and the other solvent and solute). Thus either the difference (used throughout this study) or, depending on the type of spectrum measured, the quotient of the two spectra can be calculated. The IRE holder of the viewcell was equipped with two movable blinds that cover one of the two possible pathways for the IR beam. Hence the IR radiation can only propagate through either the front or the back half of the IRE. A computer-controlled electromagnet enables fast switching between the two measurement modes. Figure 1 shows schematically the two possible states of the SBSR setup. If only one side of the IRE is coated with solid material, the other side allows measuring the bulk fluid phase. By calculating the difference of the two spectra, the influence of the bulk phase on (34) Harrick, N. J. Internal Reflection Spectroscopy; Interscience: New York, 1967. (35) Schneider, M. S.; Grunwaldt, J. D.; Bu¨rgi, T.; Baiker, A. Rev. Sci. Instrum. 2003, 74, 4121. (36) Gisler, A.; Bu¨rgi, T.; Baiker, A. Phys. Chem. Chem. Phys. 2003, 5, 3539. (37) Fringeli, U. P. Chimia 1992, 46, 200. (38) Fringeli, U. P.; Baurecht, D.; Siam, M.; Reiter, G.; Schwarzott, M.; Bu¨rgi, T.; Bru¨esch, P. In Handbook of Thin Film Materials; Nalwa, H. S., Ed.; Academic Press: New York, 2001.
2892
Langmuir, Vol. 20, No. 7, 2004
Schneider et al.
Table 1. Structural Properties of the Investigated Silica Materials Determined by Nitrogen Adsorption Measurements name
SBETa/m2 g-1
dpb/nm
Vpc/cm3 g-1
SI1401 SI1402 SI1301
772 436 336
2.1 10.2 14.8
0.4 1.1 1.3
a Total Brunauer-Emmett-Teller (BET) surface area. b Average pore diameter (4V/SBET). c Pore volume.
Figure 3. SEM picture of the silica layer coated on the ZnSe IRE. Numbers and white bars indicate thicknesses of the IRE and the SiO2 layer. The inset shows the silica layer magnified 5 times.
Figure 2. Pore size distributions of the used silicas determined with the BJH model using the desorption branch of nitrogen adsorption experiments; from left to right: SI1401, SI1402, SI1301. the spectrum of the solid layer can be eliminated. When using the SBSR technique, the single spectrum of one IRE side is directly used as background for the measurement of the other side of the IRE. Here, after flushing the cell with nitrogen, a separate background spectrum for each side of the IRE was measured. In this way, spectra can be looked at separately and the subtraction can be done manually later on. Furthermore, even subtle absorption differences of the two modes are considered. A typical experiment was carried out as follows: A sample of silica powder (Grace Davison, Davicat) was mortared and suspended in water. Table 1 lists the structural properties of the silica materials, as determined by nitrogen adsorption experiments. The pore size distributions estimated by application of the Barrett, Joyner, and Halenda (BJH) method to the desorption branches are depicted in Figure 2.39 The IRE was placed in a custom-built jig that permits coating the crystal very concisely. Note that high precision is necessary, because if the seal presses on the solid layer, either the IRE breaks or the cell is not leakproof. A few drops of the suspension were given on the crystal, which was subsequently dried at 50 °C and 25 mbar. Preliminary thermogravimetric analysis (TGA) results with SI1402 samples showed that the physisorbed water desorbs at 80 °C and atmospheric pressure; therefore the water should be removed almost completely from the surface. Figure 3 shows a scanning electron microscopy (SEM) picture of the coated silica layer on the IRE. The two white bars and the corresponding numbers indicate the size of the IRE and the silica layer, being 2.18 mm and 244 µm, respectively. The silica layer is much thicker than the ATR penetration depth. The picture inset shows a 5-foldmagnified part of the silica layer. It can be seen that the interparticle void space (filled with CO2 during the experiments) (39) Barrett, E. P.; Joyner, L. G.; Halenda, P. P. J. Am. Chem. Soc. 1951, 73, 373.
is not negligible and will therefore contribute significantly to the IR spectra. The IRE was mounted into the view-cell, and after flushing the cell with nitrogen reference spectra for both sides of the IRE were taken. A certain amount of carbon dioxide (e.g., 10 g; CarbaGas; purity, 99.995%) was dosed into the cell. After the cell content equilibrated at the desired conditions (pressure or density was controlled by adjusting the volume), spectra of both sides of the IRE were recorded and the difference of the two spectra was calculated. For comparison, a sample with 100 nm SiO2 on the ZnSe crystal was prepared by electron beam evaporation, using a BalTech BAE 370 vacuum coating system, similarly as in ref 40. X-ray photoelectron spectroscopy (XPS; Leybold Heraeus LHS11 MCD, see ref 40) showed that the ZnSe crystal was mostly covered by SiO2 (signal for Zn < 0.2%). The XPS peaks were shifted using the C 1s signal (248.7 eV) resulting in the O 1s peak at 532.9 eV. Safety note: The experiments described in this paper involve the use of relatively high pressure and require equipment with the appropriate pressure rating.
Results and Discussion Density Dependence of the ν2 Infrared Band of CO2. For the ATR-IR investigations, the ν2 infrared band of CO2 was used to study the density versus pressure behavior of CO2 in the bulk fluid and in the pores. Figure 4 shows the ν2 and ν3 rotation-vibration bands of CO2 recorded in transmission mode with a resolution of 0.3 cm-1 at 21 °C and 1.6 bar. As the state of aggregation is a low-density gas phase, the single rotational bands are clearly visible and the P-, Q- (only for ν2), and R-branches can easily be discriminated.41 With increasing density, the rotational structure of the P- and R-branches is lost due to the shorter mean free path length, the P- and R-branches move together, and the average absorbance wavenumber is red-shifted in accordance with observations of other authors42 and in a similar way as was reported for hydrogen chloride.43 Above the saturation pressure, the IRE is covered with the liquid phase and the branches merge to one single signal. The very strong Q-branch of the ν2 band renders this the signal of choice (40) Grunwaldt, J. D.; Go¨bel, U.; Baiker, A. Fresenius’ J. Anal. Chem. 1997, 358, 96. (41) Schrader, B. Infrared and Raman Spectroscopy: Methods and Applications; VCH: Weinheim, 1995. (42) Kazarian, S. G.; Briscoe, B. J.; Coombs, D.; Poulter, G. Spectrosc. Eur. 1999, 11, 10. (43) Buback, M.; Franck, E. U. Ber. Bunsen-Ges. Phys. Chem 1971, 75, 33.
Near-Critical CO2 in Porous Silica
Figure 4. High-resolution spectrum of CO2 at 21 °C and 1.6 bar showing the P- and R-branches of the ν3 band and the P-, Q-, and R-branches of the ν2 band. The spectrum was measured in transmission mode through ZnSe windows accumulating 50 scans at a resolution of 0.3 cm-1. The background spectrum was measured with nitrogen.
Langmuir, Vol. 20, No. 7, 2004 2893
Figure 5. Shift of the ν2 band maximum of carbon dioxide with increasing density. To be able to compare the band maxima, all spectra were smoothed and normalized in the shown frequency region. The corresponding densities are (in kg m-3) (A) 37, (B) 142, (C) 208, (D) 526, (E) 588, and (F) 775.
to detect the density dependence, because the ν3 and the overtone bands have no Q-branch,41 and at low densities the average wavenumber has to be calculated from the band maxima of the P- and R-branches. The strongly absorbing ν2 and ν3 bands of CO2 are hardly accessible at higher pressures by conventional transmission experiments, because very short path lengths are required ( 50 nm). As no pores in the region between 20 and 100 nm could be measured in the nitrogen adsorption measurements, confinement-driven effects of the (macroporous) interparticular void space can be excluded. By integration of the difference bands, it is possible to calculate a quantity that can be compared to the excess adsorption (also known as Gibbs adsorption) of CO2 on silica. According to the ordinary definition of Gibbs adsorption, the amount adsorbed is the excess material present in the pores and on the surface of the adsorbent over and above that corresponding to the density of the gas in the bulk phase at that temperature and pressure.57
ne ) ntot - VvoidFgas
(1)
Vvoid stands for the total minus the solid volume and is usually determined by volume difference experiments (pycnometry) with helium. (It is assumed that the noble gas does not adsorb on the surface.14) Equation 1 can be rearranged to
ne ) Vads(Fads - Fgas)
(2)
where Vads stands for the volume and Fads for the density of the adsorbed phase, respectively.58 Equation 2 illustrates the physical interpretation of the excess adsorption. If the adsorbed-phase volume were filled with bulk gas, then ne is the amount adsorbed in excess of that volume. With increasing pressure, this excess soon reaches a maximum. As the pressure is increased further, at supercritical conditions, the density of the gas phase gradually approaches that of the adsorbed phase, and at subcritical conditions condensation occurs. In the former case, the amount adsorbed measured experimentally and calculated according to the above definition must become zero; in the latter case it even gets negative. Hence, because the excess adsorption calculation neglects the volume occupied by the adsorbed phase in calculating the amount of unadsorbed gas (the entire void volume is viewed as being available to the unadsorbed gas), the high-pressure adsorption isotherms must exhibit a maximum even by elementary considerations. The results of the integration of the difference spectra for the three silica materials are shown in Figure 11 and Figure 12. The origin was added to all figures as the starting point of the isotherm. However, the lines con(57) Menon, P. G. Chem. Rev. 1968, 68, 277. (58) Sudibandriyo, M.; Pan, Z. J.; Fitzgerald, J. E.; Robinson, R. L.; Gasem, E. A. M. Langmuir 2003, 19, 5323.
Figure 11. Excess adsorption type isotherms of carbon dioxide on silicas with different pore diameters ((A) 2 nm; (B) 14.5 nm) calculated from the integration of the difference bands at 660 cm-1. Measured at 21 and 22 °C, respectively. The lines are guides to the eye.
necting the data points are just guides to the eye and do not represent any adsorption characteristics at low pressures. With the SI1402 material, measurements were carried out at two different temperatures. The general shape of the curve is the same for all four plots. The amount of excess adsorbed molecules increases very slowly. Only shortly before the saturation pressure is reached, a very strong increase can be observed. As seen in Figure 10, when condensation occurs the signal gets negative, and so does the integral. The strong increase in excess adsorption when approaching the saturation pressure can be understood as formation of thick adsorption layers on the external surface and in the pores. This interpretation is supported by other authors who measured similar behavior with different techniques.27,28 Chen et al. even reported desorption hysteresis up to the bulk critical temperature,28 but this was probably due to condensation in the tubings of the experimental setup.55 The characteristics of the three isotherms at 21 and 22 °C seem to be nearly identical in that there is a steep increase of adsorbed molecules close to the saturation pressure. With increasing pore size, the radius of the curve bending at the beginning of the jump increases. When the relative distribution is the same, the absolute difference of the pore sizes increases with increasing pore diameter (e.g., for SI1402 the pore diameter varies between 4 and 12 nm, compared to 6-20 nm for the SI1301, cf. Figure 2), and pore condensation is observed rather as a blurred than a sharp signal. For the curve of SI1401 at 25 °C, the jump in adsorption gives way to a smooth curve, even with a data point between the maximum and the negative value. The latter progression indicates different adsorption and phase behavior, namely, absence of capillary con-
Near-Critical CO2 in Porous Silica
Langmuir, Vol. 20, No. 7, 2004 2897
Figure 13. Difference spectra of carbon dioxide adsorbing on SI1401 (pore size, 2 nm). Isochoric temperature ramp of subcritical to supercritical conditions. The curves correspond to the following temperatures and pressures: (A) 22.5 °C, 61.7 bar; (B) 27 °C, 68.1 bar; (C) 30 °C, 74.3 bar; (D) 31 °C, 74.7 bar (critical point); (E) 35 °C, 83.7 bar.
Figure 12. Excess adsorption type isotherms of carbon dioxide on silicas with a pore diameter of 10 nm measured at 21 °C (A) and 25 °C (B) calculated from the integration of the difference bands at 660 cm-1. The lines are guides to the eye.
densation. Hence, the pore critical temperature for carbon dioxide adsorption on SI1402 can be confined to lie between 21 and 25 °C. This value fits into the data set collected by Morishige et al.,59 who plotted (Tcp - Tc)/Tc versus d/rp according to the relation between capillary critical temperatures and pore radius developed by Evans and Marconi.16 The Tcp reported here seems to follow the observed trend that Tch and Tcp for CO2 are considerably higher than for other adsorbates.21 The question remains of why pore condensation can be observed for the SI1401 silica, as because of the much smaller pore radius, Tcp should be at a considerably lower temperature than 21 °C. Adsorption of Supercritical Carbon Dioxide. To investigate the phase behavior in the pores along an isochoric temperature ramp, 15 g of CO2 was dosed to the cell, at a volume of 30 mL, resulting in an average density of 500 kg m-3. The changes in difference spectra for the SI1401 are shown in Figure 13. The first curve corresponds to liquid CO2 at 22.5 °C. There is a small positive band around 668 cm-1 and a strong negative signal with its maximum at 661 cm-1. If there are more subtle bands around 661 cm-1, they are covered by the latter. The second spectrum at 27 °C shows a slight shift of the broad band toward higher wavenumbers, and shoulders appear at 665 and 556 cm-1. The absorbance of the negative signal decreases continuously with increasing temperature. But at the critical point, the signal at 662 cm-1 gives way to a broad but weak positive signal at lower wavenumbers. Interestingly, the small band at 668 cm-1 almost did not change up to 31 °C. It is only in the supercritical region where it suddenly vanishes. At 35 °C, the signals over the (59) Morishige, K.; Ito, M. J. Chem. Phys. 2002, 117, 8036.
Figure 14. Difference spectra of supercritical carbon dioxide (35 °C) adsorbing on SI1401 (pore size, 2 nm). The corresponding density for each spectrum is given in the order of the arrow crossing the curve at 670 cm-1.
whole spectral region covered by the figure are flattened; in other words, the bulk and the silica spectra nearly look the same. This series of measurements shows that the differences of fluid phase behavior in the bulk and in pores almost vanish at temperatures above the bulk critical temperature. The broad but weak band still visible at supercritical conditions seems to be similar to the bands discussed for Figure 8. The difference from the subcritical (but gaseous in bulk) state is that the 660 cm-1 signal (standing for the species adsorbed in the multilayer) almost disappeared. Hence, with increasing temperature the multilayer gets thinner. The small signal at 667 cm-1 denotes CO2 species at gaslike densities. It only vanishes at supercritical temperatures. An isothermal adsorption experiment with SI1401 was also carried out above the bulk critical temperature. Figure 14 shows the difference spectra varying with density (see the arrow for the right sequence according to the indicated density values). The red-shift of the strong negative signal increases with increasing pressure. The maximum of the positive band almost does not shift, whereas at 526 kg m-3 a sudden displacement can be observed, and the absorbance difference gets lower. Note that the absorbance values in this figure, and therefore also the differences
2898
Langmuir, Vol. 20, No. 7, 2004
Schneider et al.
Table 2. Influence of Temperature and Pressure on Bulk and Pore Phase Behavior temperaturea pressureb Ttp < T < Tch
P , Psat P e Psat P > Psat
Tch < T < Tcp P , Psat P e Psat P > Psat Tcp < T < Tc
P , Psat P e Psat P > Psat
T g Tc
P , Pc P ≈ Pc P > Pc
T > Tc
F < Fc F > Fc
T . Tc
F < Fc F > Fc
phase behavior bulk: gas pore: monolayer bulk: gas pore: condensation, hysteresis bulk: liquid pore: liquid bulk: gas pore: monolayer bulk: gas pore: strong increase in adsorption, condensation bulk: liquid pore: liquid bulk: gas pore: monolayer bulk: gas pore: continuous pore filling bulk: liquid pore: “supercritical” bulk: gaslike pore: low adsorption bulk: clustering pore: bulk: pore: bulk: pore: bulk: pore: bulk:
depletion effect liquidlike liquidlike gaslike low adsorption liquidlike liquidlike gaslike
pore: very low adsorption bulk: dense gaslike pore: low adsorption
remarks gas-phase reaction conditions unfavorable conditions for reaction usual conditions for trickle bed or slurry reactor gas-phase reaction conditions no hysteresis f easy desorption possible low diffusivity gas-phase reaction conditions no condensation, density only in pores tunable high density in pore, but not tunable; good solubility, but low diffusivity good diffusivity, low solubility good tunable bulk solubility, enhanced pore diffusivity, tunable pore solubility good solubility, low diffusivity good diffusivity, low solubility fair solubility, low to fair diffusivity very low bulk solubility, good diffusivity in pores, gas-phase reaction conditions fair to good solubility, fair diffusivity, very high pressures needed
a Temperatures: T , triple point temperature; T , critical hysteresis temperature; T , critical capillary temperature; T , bulk critical tp ch cp c temperature. b Pressures: Psat, saturation pressure of the solvent; Pc, bulk critical pressure. At higher temperatures, densities are indicated instead of pressures. Fc, bulk critical density.
between bulk fluid and fluid in the solid layer, are generally much lower compared to the measurements at subcritical conditions. The spectra at supercritical temperatures look quite different from the subcritical ones. The negative signals at higher wavenumbers give hint that in the bulk phase the density is continuously shifted toward higher densities. The positive bands at lower wavenumbers indicate that the density of the adsorbate layers is higher than the bulk fluid phase density. For the same reasons as at subcritical conditions, there must be a maximum for the excess adsorption, which seems to lie here between 450 and 550 kg m-3. Steriotis et al. also found evidence by neutron diffraction that CO2 on a microporous carbon sample slightly above the critical temperature but below the critical pressure is in a densified state, comparable to dense supercritical fluid or even bulk liquid at pressures well below the critical pressure.60 Also Su¨er et al. concluded from self-diffusion experiments of n-heptane (Tc, 267 °C; pc, 27.4 bar) in zeolite L at 13.8 bar and 350 °C that the density inside the pores is much higher than the density of the bulk SCF at the same temperature and pressure.61 Hence, a fluid above the critical temperature does not exhibit the same density in and outside the pores but can be distinguished using ATR-IR spectroscopy. Only at high pressures is the density (60) Steriotis, T. A.; Stefanopoulos, K. L.; Mitropoulos, A. C.; Kanellopoulos, N. K.; Hoser, A.; Hofmann, M. Appl. Phys. A 2002, 74, S1333. (61) Su¨er, M. G.; Dardas, Z.; Lu, Y. P.; Moser, W. R.; Ma, Y. H. AIChE J. 1997, 43, 1717.
inside and outside the pores similar, which seems to approach in a continuous manner. Consequences for Adsorption Processes and Heterogeneously Catalyzed Reactions in Supercritical Fluids. The measurements showed that at all conditions, there is a high-density adsorbate layer on the (inner and outer) solid surface. At temperatures between Tch and Tcp, the layer thickness increases rapidly as the pressure approaches the saturation pressure, but usually no desorption hysteresis can be observed. Between Tcp and Tc, there is no sudden jump in layer thickness but a continuous increase. At supercritical conditions, the continuous curve is flattened, but the shape with the maximum stays. This means that the density of the fluid in the pores is higher, but with increasing pressure and temperature, the bulk density approaches the density inside the pores. These phenomena have an impact on adsorption processes and heterogeneously catalyzed reactions in SCFs, for the latter especially if they are mass transport limited. The solubility of a reactant in a SCF strongly depends on the density.62 Hence, the increased density within the pores is advantageous for systems where the solubility of one reactant is determining the overall reaction rate. In the case of catalyst deactivation by blocking of active sites or pores (e.g., coking in alkene isomerization), the enhanced solvent power of the SCF can lead to faster desorption and hence to improved catalyst lifetimes.63 On the other hand, at conditions where a liquidlike phase is (62) Chrastil, J. J. Phys. Chem. 1982, 86, 3016. (63) Subramaniam, B. Appl. Catal., A 2001, 212, 199.
Near-Critical CO2 in Porous Silica
formed in the pores while the bulk phase behavior still is gaslike, the emerging phase boundary may lead to a slowdown of the overall reaction rate due to the additional mass transport resistance. However, solubility is not the only parameter that should be considered. Zhou and Wang calculated diffusivities of supercritical carbon dioxide in slit pores with a combination of grand canonical Monte Carlo simulations and molecular dynamics.64 As could be expected, the diffusion coefficients of CO2 molecules confined to slit pores at supercritical conditions are strongly dependent on the fluid density in the pore. Diffusivity generally decreases with increasing density at constant temperature. Su¨er et al. reported significantly reduced diffusivities under supercritical conditions due to increased fluid density in the pores for hydrocarbons in Zeolite L.61 Lai and Tan observed for CO2 and toluene in activated carbon that the diffusivity depends much more strongly on the fluid density than on temperature or the amount of solute adsorbed on the surface.65 These observations show that the determination of ideal conditions for adsorption processes and heterogeneously catalyzed reactions in SCFs is more difficult than often anticipated. However, by careful optimization it should be possible to find temperatures and pressures where the density is in an intermediate range, so the solvent power of the SCF is strong enough to dissolve reasonable amounts of reactants but the diffusivity is not reduced too much. An attempt to summarize the discussed changes of density, solubility, and diffusivity in the bulk and the pore phase for the whole temperature and pressure range is made in Table 2. Optimization of reaction conditions can be accelerated by careful inspection of the bulk phase behavior and in situ adsorption measurements with appropriate tools. As the capillary critical temperatures depend on the pore size, it is also important to choose the right support material for the catalytically active species. Consequently, by combined consideration of solubility, adsorption and diffusion characteristics, and their tunability, it emerges that for adsorption processes and (64) Zhou, J.; Wang, W. C. Langmuir 2000, 16, 8063. (65) Lai, C. C.; Tan, C. S. Ind. Eng. Chem. Res. 1993, 32, 1717.
Langmuir, Vol. 20, No. 7, 2004 2899
heterogeneously catalyzed reactions in supercritical fluids, solids of intermediate pore size and SCFs at intermediate density are best suited. Conclusions A new approach to investigate adsorption and phase behavior of supercritical fluids confined to porous solids has been presented, using carbon dioxide on mesoporous silica with different pore diameters as a model system. In situ ATR-FTIR measurements using the SBSR technique were performed to subtract interfering signals of the bulk phase. Thus the observed signals could be attributed either to the bulk fluid or to the pore fluid phase. Measurements without silica provided a relation between the bulk fluid density and the peak shift of the ν2 vibrational band of CO2, which in turn permitted an estimate of the fluid density inside the pores. Integration of the IR difference bands resulted in an excess adsorption type quantity that allowed interpretation of the adsorption behavior. It was concluded that at all temperatures and pressures investigated, and even under supercritical conditions, the fluid density in the pores was higher than in the bulk. The influence of the fluid density inside the pores on solubility and diffusivity and its impact on catalytic reactions have been discussed. Considering all the mentioned points, it can be concluded that it is a necessary prerequisite to investigate the so far almost neglected field of phase behavior of SCFs within porous materials. Proper understanding of the mass transfer and adsorption processes in pores under SCF conditions may pave the way to optimally tailored pore structures for catalysts and adsorbents. Acknowledgment. Financial support of this work by the Swiss Federal Office of Energy is kindly acknowledged. We thank F. Krumeich for the SEM investigation, M. Maciejewski for the TGA experiments, A. Urakawa for performing the model calculations, and T. Bu¨rgi for his help and advice. Grace Division is acknowledged for providing the silica samples. LA035778N
