- Email: [email protected]
Hydrothermal stability of microporous silica and niobia–silica membranes
Journal of Membrane Science 319 (2008) 256–263
Contents lists available at ScienceDirect
Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci
Hydrothermal stability of microporous silica and niobia–silica membranes V. Boffa 1 , D.H.A. Blank, J.E. ten Elshof ∗ MESA Institute for Nanotechnology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
a r t i c l e
i n f o
Article history: Received 5 January 2008 Received in revised form 14 March 2008 Accepted 23 March 2008 Available online 30 March 2008 Keywords: Hydrothermal stability Niobia Silica Microporous membrane Permporometry
a b s t r a c t The hydrothermal stability of microporous niobia–silica membranes was investigated and compared with silica membranes. The membranes were exposed to hydrothermal conditions at 150 and 200 ◦ C for 70 h. The change of pore structure before and after exposure to steam was probed by single-gas permeation measurements and nanopermporometry. The hydrothermal stability of the niobia–silica membrane was found to be higher than that of silica. After hydrothermal treatment at 200 ◦ C, the hydrogen permeance of the niobia–silica top layer had declined by 32%, while the H2 permeance of the silica top layer was reduced by 73%. The apparent activation energies of the H2 permeance were 12.2 ± 0.2 and 15.3 ± 0.7 kJ mol−1 for silica and niobia–silica, respectively. Nanopermporometry experiments on the silica membrane were in semi-quantitative agreement with the gas permeation data. The data suggest that densification of the top layer occurred predominantly in those areas with the highest convex curvatures, thereby increasing the effective transport path of helium and hydrogen across the membrane. © 2008 Elsevier B.V. All rights reserved.
1. Introduction We recently developed a niobia–silica microporous membrane [1] that shows a very low permeability of CO2 . It may find application in enabling technologies for CO2 sequestration, or in membrane reactors to recover H2 from hydrogen synthesis processes such as steam reforming [2] and water–gas shift reaction [3]. Industrial application of ceramic membranes requires stability under working conditions for at least several years. Hydrothermally stable supports consisting of La-doped ␥-alumina mesoporous layers coated on ␣-alumina macroporous supports have already been developed [4]. Unfortunately, the long-term durability of gas-selective microporous top layers, which are all based on silica, is limited in the presence of steam [5]. Crystallization of microporous silica hardly occurs below 1100 ◦ C [6], but it undergoes drastic structural changes in the presence of steam below 200 ◦ C [7–11]. Hydrothermal exposure leads to collapse of the porous structure, yielding dense impermeable materials. Microcracks may form as a result of the stresses that develop during these structural rearrangements. Densification occurs mostly in the early stages of exposure to steam, but microcrack forma-
∗ Corresponding author. Tel.: +31 53 489 2695; fax: +31 53 489 3595. E-mail address: [email protected] (J.E. ten Elshof). 1 Present address: Dipartimento di Chimica Generale ed Organica Applicata, Universita` di Torino, Corso M.D’Azeglio 48, 10125 Torino, Italy. 0376-7388/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2008.03.042
tion is less predictable. Loss of selectivity due to this type of deterioration often occurs after extended periods of exposure to steam. Himai et al. [7] proposed that Si O bonds are broken upon interaction with water at high temperature, creating vicinal hydroxyl pairs that are subject to recondensation. This mechanism, consisting of the cyclic destruction and reconstruction of Si O Si network bonds, gives some flexibility to the material, so that it can reorganize itself eventually in a more stable and denser state. The hydroxylation of silica finally results in depolymerization [12], because silicic acid dissolves in the thin film of adsorbed water that is covering the material during exposure to steam. The silicic acid can freely migrate through the sorbed water layer until it precipitates and recombines with the silica matrix due to saturation. Dissolution (the depolymerization of silica) occurs fastest in those areas where the convex curvature is largest. This process is dominant in microporous silica, where pores have larger degrees of curvature than in mesoporous silica and are filled by water even at low water partial pressures. Microporous silica can be stabilized by introducing transition metals in the silica network [13–17]. Oxygen forms more polar and stable bonds with transition metals than with silicon. Furthermore, most transition metals have coordination numbers larger than that of silicon [18]. Thus, they tend to yield more closely packed and crystalline structures than pure silica. To determine the changes in pore size distributions upon hydrothermal treatment, different monitoring techniques can be
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
employed. One way is to carry out single-gas permeation experiments with a series of gases with different kinetic diameters. Assuming that separation is based on size differences only, a low permeance corresponds to a small number of pores that is accessible for a species with a certain size. The advantage of this approach is that it is not dependent on any theoretical model. The results may nevertheless be affected by differences in interactions between different molecular probes and the membrane pore walls [1]. Another established technique to determine pore size distributions in supported thin films is permporometry. Its advantage is that all pore sizes are accessed with the same molecular probe, so that pore size distributions are not influenced by probe–wall interactions. On the other hand, estimates of pore size distributions can only obtained by applying theoretical models and making assumptions regarding the shape of the pores. The permporometry technique was originally developed in the 1990s for mesoporous membranes [19]. Pore size distributions can be calculated by applying the Knudsen [20] and Kelvin [21] equation. More recently, Tsuru developed the nanopermporometry technique, which is able to access micropores [22,23]. However, the interpretation of nanopermporometry data is more complex than conventional permporometry data. The Knudsen and Kelvin equations are not valid in the microporous regime [22], and simple alternative models are not available in literature. Tsuru et al. defined the average pore size as the Kelvin diameter at which the permeance of the vapor-containing gas was half of the permeance of the dry gas [22]. Here the Kelvin diameter is two times the Kelvin radius (rK ), which is defined by the Kelvin equation: ln Pr = −
s Vmol cos 1 RT rK
(1)
where Pr is the relative vapor pressure, s (J m−2 ) the gas–solid interfacial tension, Vmol (m3 mol−1 ) the molar volume, and is the contact angle of the condensed phase on the pore wall. The latter value is normally assumed to be zero. In this work the hydrothermal stability of a niobia–silica microporous membrane was investigated and compared with that of pure silica. Nanopermporometry and single-gas permeation measurements were employed to probe the change of pore structure of the membrane after exposure to steam at elevated temperatures. 2. Experimental 2.1. Sol synthesis The niobia-doped silica sol (denoted as NS sol) was prepared by adding 11 ml of tetraethyl orthosilicate (Aldrich, 99.999% pure; denoted as TEOS) to 10.5 ml of ethanol. An aqueous solution of
257
nitric acid was dropped in the mixture until the final molar ratio (Si)–OC2 H5 :H2 O:HNO3 was 1:0.5:0.01. We have chosen to express the chemical composition of the sol in this way because alkoxy groups, water and acid are the moieties that participate directly in the reaction. This mixture was heated under reflux at 60 ◦ C for 2 h. A 1-M solution of niobium(V) penta(n-butoxide) (Gelest) in n-butanol (Aldrich, anhydrous) was added slowly to the mixture. Aqueous nitric acid was dropped into the mixture to restore the initial composition (M)–O–R:H2 O:HNO3 to 1:0.5:0.01, with M = Si or Nb and R = C2 H5 or C4 H9 . The sol was refluxed at 60 ◦ C for 5 h. Then it was cooled down to 20 ◦ C and kept at this temperature for several days until it was used for characterization and membrane preparation. Prior to coating membranes, the sol was diluted 11.5 times with ethanol. The silica sol was prepared by adding 11 ml of 0.73N aqueous nitric acid to 21 ml of tetraethylorthosilicate (TEOS, 98%, Aldrich) dissolved in 21 ml ethanol (≥99.5%, Aldrich). The addition was carried out drop by drop under vigorous stirring, while the reaction flask was cooled in an ice bath. The mixture was then refluxed at 60 ◦ C for 3 h. The resulting silica sol was diluted 18 times in ethanol and kept at −5 ◦ C for a few days until it was used for preparation of membranes. 2.2. Preparation of silica membranes Flat supports were prepared by coating a 6% La-doped boehmite sol on a 2.0-mm thick ␣-alumina disk (∅ 39 mm) as described elsewhere [4]. After calcination at 600 ◦ C a mesoporous 6% La-doped ␥-alumina layer with average pore size of 5 nm was formed, as verified by conventional permporometry [10]. Silica and niobia–silica membranes were prepared by dipping the supports in the sols at an angular speed of 0.06 rad s−1 . The membranes were calcined for 3 h in air at 500 ◦ C, with heating and cooling rates of 0.5 ◦ C/min. Since the silica sol yielded an extremely thin layer, the coating and calcination cycle was repeated to increase film thickness and cover any remaining defects or uncoated areas of the membrane. The niobia–silica membrane is further referred to as NS membrane. 2.3. Single-gas permeation experiments Permeation measurements were carried out using the setup shown in Fig. 1. In the single-gas permeation experiments all valves were closed with the exception of valve number 2, and the setup was used as a dead-end mode permeation setup [24]. In a typical experiment, the membrane was first degassed at 200 ◦ C for 2 days in a flowing helium atmosphere. Then the flows of He, H2 , CO2 , N2 , CH4 and SF6 were measured sequentially, employing a pressure of 5 bar on the side of the feed, and atmospheric pressure on the per-
Fig. 1. Schematic diagram of the experimental setup.
258
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
meate side. Steady state was ensured by waiting at least 3 h prior to sampling data. In the case of hydrogen and helium the measurements were carried out between 90 and 200 ◦ C to determine the apparent activation energies of permeance of these gases. The volumetric gas flow on the permeate side was measured with a soap film flow meter. 2.4. Nanopermporometry experiments Nanopermporometry experiments were performed using steam (kinetic diameter ∼0.21 nm) as condensable vapor and hydrogen (0.29 nm) as incondensable gas. The setup shown in Fig. 1 was used, since the relative pressure of water on the feed side can be controlled accurately by the mass flow controllers MFC1 and MFC2. The pressure controller was excluded from the active part of the gas permeation setup by closing valve number 2, and opening the other valves. Valve 7 was regulated manually to keep the H2 pressure gradient across the membrane at 1 kPa. While flushing the membrane with dry hydrogen the sample temperature was decreased from 200 to 90 ◦ C. The temperature of the water bath was kept at 32 ◦ C. The steam partial pressure in the feed was gradually increased by varying the gas flows through the two mass flow controllers MFC1 and MFC2. Pore size distributions were calculated from the change of hydrogen permeance with changing steam partial pressure, as discussed in more detail below. 2.5. Hydrothermal treatments The hydrothermal stability of the membranes was tested by exposure to steam. The membranes were first exposed to a steam partial pressure of 0.56 bar at 150 ◦ C for 70 h, using hydrogen as sweep gas. The total overpressure at the feed side was 4 bar. Then the sample was dried at 200 ◦ C for 2 days. Afterwards, a series of single-gas measurements was carried out at 200 ◦ C using gases with increasing kinetic diameter. The pore size distribution of the silica top layer was measured by nanopermporometry under the same experimental conditions as used for the fresh sample. After exposure to 0.56 bar of steam at 200 ◦ C for 70 h, the gas permeation and nanopermporometry experiments were repeated a third time. 2.6. Scanning electron microscopy Scanning electron microscopy (SEM) images were taken on a LEO 1550 FEG.
3. Results and discussion 3.1. Membrane morphology The SEM images in Fig. 2 shows cross-sections of the silica and NS membrane. The gas-selective layers have thicknesses of about 30 and 150 nm for silica and NS, respectively. Silica and NS-derived powders were X-ray amorphous after calcination at 500 ◦ C. 3.2. Single-gas permeation experiments: pore size distribution In the Knudsen regime, which covers the pore size range between 2 and 50 nm, the permeability of a membrane F is inversely proportional to the square root of the molecular mass of the permeating species [20]:
F =K
1 MRT
(2)
where K is a constant that depends on pore shape, membrane area and thickness of the top layer, R the gas constant, T the absolute temperature, and M the molar mass of the gas. Single-gas permeation measurements on mesoporous ␥alumina supports reported earlier [20] showed that the gas flow rate has a Knudsen-type relation with temperature and molecular mass. The results of gas permeation experiments on membranes that include microporous silica-based top layers are listed in Tables 1 and 2. The kinetic diameters of the corresponding probe molecules are also given. No correlation with the trend predicted by Eq. (1) was observed. In general, the permeance decreased gradually with increasing kinetic diameter. The permeance of SF6 in silica was immeasurably low. This indicates that the majority of pores in the silica layer has a diameter between 0.3 and 0.4 nm, and a negligible number of pores larger than 0.55 nm. The helium permeance of the NS membrane was about 12 times higher than that of nitrogen, so a large fraction of pores has dimensions smaller than the size of nitrogen (0.36 nm). But in contrast to silica, the NS membrane was also permeable to SF6 . This implies the presence of a small fraction of pores larger than ∼0.55 nm. As we reported elsewhere [1], the NS membrane has an exceptionally low permeability for CO2 . This strongly suggests that the larger pores through which SF6 diffused must be in the microporous range (<2 nm). Otherwise the presence of mesopores would have resulted in a much higher permeability of CO2 [1].
Fig. 2. SEM images of cross sections of (a) microporous silica top-layer and (b) a NS gas-selective top layer.
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
259
Table 1 Single-gas permeance of supported silica membrane before and after the first (HT150) and second (HT200) hydrothermal treatment Molecule
He H2 CO2 N2 CH4 SF6
dk (nm)a
0.26 0.29 0.33 0.36 0.38 0.55
Fresh sample
After HT150
Fb
FH2 /FX
Fb
FH2 /FX
Fb
FH2 /FX
1.0 1 46 1.5 × 102 5.6 × 102 ≥103
3.7 × 10−7
0.85 1 33 1.4 × 102 2.0 × 102 ≥10 3
3.7 × 10−7 3.0 × 10−7 6.3 × 10−9 2.1 × 10−9 1.1 × 10−9 ≤1 × 10−10
0.75 1 47 1.4 × 102 2.7 × 102 ≥103
6.1 × 10−7
6.3 × 10−7 1.3 × 10−8 4.2 × 10−9 1.1 × 10−9 ≤1 × 10−10
3.1 × 10−7 9.5 × 10−9 2.2 × 10−9 1.5 × 10−9 ≤1 × 10−10
After HT200
The ideal hydrogen to X permselectivity is reported for every set of measurements. a Kinetic diameters d taken from Ref. [25]. k b Permeance (F) is expressed in mol s−1 m−2 Pa−1 .
Table 2 Single-gas permeances of supported NS membrane before and after the first (HT150) and second (HT200) hydrothermal treatment Molecule
dk (nm)a
Fresh sample Fb
He H2 CO2 N2 CH4 SF6
0.26 0.29 0.33 0.36 0.38 0.55
7.2 × 10−8
3.9 × 10−8 8.5 × 10−10 6.2 × 10−9 5.4 × 10−9 2.9 × 10−9
After HT150 FH2 /FX
Fb
0.54 1 46 6.3 7.2 13
5.2 × 10−8
3.7 × 10−8 8.2 × 10−10 6.2 × 10−9 5.8 × 10−9 2.9 × 10−9
After HT200 FH2 /FX
Fb
FH2 /FX
0.71 1 45 6.0 6.4 13
4.0 × 10−8
0.68 1 33 4.0 5.0 9.0
2.7 × 10−8 8.3 × 10−10 6.7 × 10−9 5.4 × 10−9 3.0 × 10−9
The ideal hydrogen to X permselectivity is reported for every set of measurements. a Kinetic diameters d taken from Ref. [25]. k b Permeance (F) is expressed in mol s−1 m−2 Pa−1.
3.3. Single-gas permeation experiments: hydrothermal stability The overall permeance after both hydrothermal treatments is listed in Tables 1 and 2. The abbreviations HT150 and HT200 are used to indicate the data acquired after the first and second hydrothermal treatment, respectively. Exposure to steam at high temperature led to densification of the microporous silica phase. This reduced the permeance of hydrogen and helium. The densification of silica was significant after HT150, but less pronounced after HT200. The permeance of larger molecules did not increase after either HT150 or HT200. This indicates the absence of cracks in the film. The overall He permeance of the niobia–silica membrane decreased from 7.2 × 10−8 to 5.2 × 10−8 mol Pa−1 m−2 s−1 after HT150 and to 4.0 × 10−8 mol Pa−1 m−2 s−1 after HT200. The decrease of H2 permeance after HT200 was less pronounced than that of helium. The permeance of molecules with a larger kinetic diameter remained almost unaffected by the treatment. The decrease of membrane permselectivity towards mixtures of hydrogen and other gases is therefore due to the decline of hydrogen permeance. The permeance of the gas-selective top layer of both membranes was calculated by considering that the membrane consists of a series of layers, each with a certain resistance to gas flow. The resistance of each layer is proportional to the reciprocal permeance. The permeance of the gas-selective layer Ftop can thus be determined from the total permeance Ftot and the permeance of the support Fsup via Eq. (3) [24]: 1 1 1 = + Ftot Ftop Fsup
The hydrogen permeance of the silica top layer at 200 ◦ C was 1.6 × 10−6 mol Pa−1 m−2 s−1 , and 3.8 × 10−8 mol Pa−1 m−2 s−1 for NS. Defining permeability as permeance per unit membrane thickness, this implies that the H2 permeabilities at this temperature were 4.5 × 10−14 mol Pa−1 m−1 s−1 for silica and 5.7 × 10−15 mol Pa−1 m−1 s−1 for NS. The relatively low permeability of NS is probably related to the lower hydrolysis and acid ratios at which the niobia–silica sol was synthesized. The incorporation of unreacted precursors and small clusters into the NS layer may be the cause of the relatively dense layer. Alternatively, the lower permeability of NS might also be due to the incorporation of Nb ions in a microporous silica network. The effects of hydrothermal treatment of silica and NS are compared in Fig. 3. Hydrothermal densification is more prominent for silica than for NS. The helium permeance of silica was reduced to 21% after HT150, and the hydrogen permeance to 27%. HT200
(3)
The permeance of the support Fsup to different probes at various temperatures was measured directly, or extrapolated by fitting Eq. (3) to a series of experimental data on different probe molecules, in a range of temperatures between 90 and 200 ◦ C.
Fig. 3. Influence of HT150 and HT200 on permeance of hydrogen and helium in silica top layer and NS top layer.
260
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
Fig. 4. Arrhenius plot of the hydrogen and helium permeance in silica top layer, both fresh and after HT200.
did not have any further effect. The decline of permeance was less pronounced in niobia–silica. After HT150, the hydrogen permeance was reduced to about 95%, and the helium permeance to 68%. This suggests that only a small fraction of micropores of NS underwent substantial densification upon hydrothermal treatment, namely those pores that were small enough to be accessible to He but not to H2 . HT200 caused an overall decline of permeance to 68% for hydrogen and 47% for helium. Apparently, the introduction of niobium in the silica framework yielded a more stable material than pure silica. The effect of Nb on hydrothermal stability is evident after HT150. However, microporous structures are thermodynamically unstable by definition and therefore remain prone to densification. An alternative explanation for the higher stability of NS could also be that it has slightly larger pores than silica. Smaller pores are thermodynamically more sensitive to hydrothermal degradation than larger ones, but it is very unlikely that this is the sole effect. The data presented here do not allow estimation
to which extent these two effects contribute each to the stability of the NS layer. 3.4. Effect of hydrothermal modification on activation energy of permeance The Arrhenius plots of the hydrogen and helium permeances of the silica and NS top layers are shown in Figs. 4 and 5. The data were calculated using Eq. (3). The transport of gases in the microporous regime occurs via a thermally activated surface diffusion mechanism. The transport rate can be expressed in terms of a modified Fick law [24]: F = F0 exp
E a −
RT
(4)
where F is the membrane permeance (flux per unit of driving force), Ea is the apparent activation energy, and F0 is a temperature inde-
Fig. 5. Arrhenius plot of the hydrogen and helium permeance in NS top layer, both fresh and after HT200.
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
pendent parameter that has the dimensions of permeance. The value of F0 depends on the interaction between pore wall and gas molecule, thickness of the gas-selective layer, and on pore shape and tortuosity. The straight lines in the figures show linear fits to the data with slopes equal to −Ea [kJ mol−1 ]. The apparent activation energy of permeance of silica was 12.8 ± 0.3 kJ mol−1 for He and 12.2 ± 0.2 kJ mol−1 for H2 . After HT200, the activation energies were 12.9 ± 0.1 kJ mol−1 for He and 11.9 ± 0.4 kJ mol−1 for H2 . These data do not indicate any change of the predominant transport mechanism upon hydrothermal treatment. The experimental data of both hydrogen and helium for NS at lower temperatures deviated from the exponential behavior predicted by Eq. (4). This can be attributed to the existence of two parallel transport paths through the top layer. One path is via the selective micropores, the other path via larger pores. The latter one is probably associated with a Knudsen-type diffusion mechanism [20]. Knudsen diffusion does not depend strongly on temperature; its rate is proportional to T−0.5 . The relative contribution of Knudsen diffusion will therefore increase with decreasing temperature and may even become dominant, as seems to be the case for NS at low temperatures. This is consistent with the fact that deviation from Eq. (4) was not observed for the silica membrane, where gas permeation experiments had already provided sufficient proof for the absence of large micropores or mesopores. The temperature below which the permeance of the NS membrane deviated significantly from the prediction of Eq. (4) were 70 and 90 ◦ C for He and H2 , respectively. After hydrothermal exposure, these values shifted to 130 ◦ C for hydrogen and 115 ◦ C for helium. Activation energies of 19.1 ± 0.3 kJ mol−1 for helium and 15.3 ± 0.7 kJ mol−1 for hydrogen were calculated from the permeance data of the fresh NS membrane in Fig. 5. After HT2, these values were similar, i.e., 19.3 ± 0.6 and 15.7 ± 0.7 kJ mol−1 , respectively. Qualitatively similar observations were made by De Vos et al., who reported an apparent activation energy of hydrogen permeance of ∼8 kJ mol−1 for silica membranes calcined at either 400 or 600 ◦ C [24]. Silica membranes calcined at 600 ◦ C were denser and had smaller pores than those calcined at 400 ◦ C [24,26], but no difference in activation energy was observed. 3.5. Nanopermporometry measurements: pore size distribution The Kelvin and Knudsen equations were used to calculate the corresponding Kelvin diameter from the experimental data. Eq. (5) [27] was used to analyze the permporometry curves shown in Fig. 6: n(rK ) = −
K rK3
RTM 2
F
acc
r
261
Fig. 6. Analysis of microporous silica membrane by nanopermporometry at 90 ◦ C. H2 permeance versus relative pressure of steam is indicated by the thin drawn line. Pore size distribution (thick line) was calculated from this curve using Eq. (5).
less pronounced. At a relative pressure of about 20%, the hydrogen permeance was less than 10% of its initial value. The estimated pore size distribution of the silica membrane as calculated from Eq. (5) is also shown in Fig. 6. The curve indicates a monomodal distribution with an average Kelvin diameter of about 0.5 nm. Fig. 7 shows the permeance as function of the kinetic diameter of the gas probe molecules, after correction for the support resistance. In contrast to the interpretation of nanopermporometry data, the permeance versus kinetic diameter data are empirical and provide a direct indication of the pore size distribution. The effective pore size distribution obtained with nanopermporometry is shown for the sake of comparison. Application of Eq. (5) seems to lead to an overestimation of the pore diameter by about 0.2 nm. However, it is also possible that an appreciable permeance occurs only in pores that are at least 1–2 A˚ larger than the kinetic diameter of the permeating molecule. This would imply that nanopermporometry cannot measure pores smaller than 0.4–0.5 nm, since the water, hydrogen and helium molecules used in that experiment have a kinetic diameter of about 0.2–0.3 nm.
(5)
where n represents the number of pores and the parameter rK expresses the effective radius corresponding to a certain relative pressure of steam as calculated by the Kelvin equation. The final term at the right-hand side of the equation expresses the derivative of permeance over rK . Although the assumptions underlying these equations are not valid in the microporous regime, they allow qualitative and semi-quantitative comparison of pore size distributions of membranes, and are therefore suitable parameters to monitor the degradation of membranes. The permeability of the NS layer was too low to acquire sufficiently accurate data. Fig. 6 shows the variation of hydrogen permeance of silica as function of the partial pressure of steam. A significant reduction of H2 flux was observed when the relative pressure of steam was increased from 1.1 to 7.0%, so most of the micropores were closed by capillary condensation of water at the latter vapor pressure. At higher relative pressures, the reduction of hydrogen permeance with relative pressure of water was much
Fig. 7. Permeance of fresh silica top layer as function of the kinetic diameter of the probe molecule (solid line) and pore size distribution calculated from nanopermporemetry data (dashed line).
262
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
Fig. 8. Effect of hydrothermal treatment on hydrogen permeance in silica membrane as measured by nanopermporometry: (a) normalized permeance and (b) calculated pore size distribution.
Fig. 9. Scheme of the effect of hydrothermal densification on microporous structure. The effective path length of a He molecule is increased after steam exposure, while the permeation of SF6 in the larger pores of the NS membrane remains unaffected.
3.6. Nanopermporometry measurements: hydrothermal stability
4. Conclusions
We also applied nanopermporometry to characterize degradation phenomena in microporous silica membranes by investigating changes in the pore size distributions of the gas-selective top layers after hydrothermal treatment. Results are depicted in Fig. 8. The data support the trends of the single-gas permeation measurements. As can be seen in Fig. 8b, the fraction of pores larger than 0.6 nm had decreased after hydrothermal treatment. Apparently, densification narrowed all pores indiscriminately, reducing both the number of large pores, and the permeance of the membrane to all gases. The change of pore size is more pronounced during the first hydrothermal treatment at 150 ◦ C. This is in agreement with the findings of the single-gas permeation experiments. A possible mechanism for the change of pore structure upon hydrothermal treatment is schematically illustrated in Fig. 9. The pore structure of microporous silica [28], and probably of NS too, roughly resembles that of a sponge. The pores with larger curvature, i.e., the smaller micropores, are more vulnerable to hydrothermal attack than pores with smaller curvature, such as larger micropores and mesopores. According to the densification mechanism proposed here, only a small fraction of the pore structure was densified during the limited time of exposure to hydrothermal conditions. The smallest pores in the microporous thin layer became blocked, thereby increasing the real length of the transport path of small molecules through the membrane, and lowering the membrane selectivity. This explains why the reduction of hydrogen and helium permeance was substantial, while no variation of activation energy was recorded.
The NS membrane was found to be more stable under hydrothermal conditions than the silica membrane. This is probably the result of the incorporation of Nb ions into the silica matrix, which yields hydrothermally stable Nb O Si bonds, although differences in pore size, density and layer thickness may also play a secondary role in stabilizing the NS membrane network. Hydrothermal treatment led to a decrease of the permeability of all gases, but the decrease was more prominent for small gases like He and H2 . No apparent changes occurred in the mechanism of gas transport through the membranes. Longer stability tests will reveal to which extent the presence of Nb ions stabilizes the microporous network by slowing down the kinetics of densification. It was shown that the nanopermporometry technique can be applied to estimate pore size distributions of microporous layers in a semi-quantitative way. This fast and non-destructive technique may be applied to other microporous systems as well. The nanopermporometry and single-gas permeation data yielded qualitatively similar results. References [1] V. Boffa, J.E. ten Elshof, A.V. Petukhov, D.H.A. Blank, Chem. Sus. Chem. 1 (2008) 437–443. [2] D.W. Lee, B. Sea, K.Y. Lee, K.H. Lee, Ind. Eng. Chem. Res. 41 (2002) 3594. [3] A. Brunetti, G. Barbieri, E. Drioli, T. Granato, K.H. Lee, Chem. Eng. Sci. 62 (2007) 5621. [4] A. Nijmeijer, H. Kruidhof, R. Bredesen, H. Verweij, J. Am. Ceram. Soc. 84 (2001) 136. [5] S. Giessler, J.C.D. da Costa, G.Q. Lu, J. Nanosci. Nanotechnol. 1 (2001) 331.
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263 [6] R.K. Iler, The Chemistry of Silica, Wiley, New York, 1979. [7] H. Himai, H. Morimoto, A. Tominaga, H. Hiraschima, J. Sol–Gel Sci. Technol. 10 (1997) 45. [8] G.P. Fotou, Y.S. Lin, S.E. Pratsinis, J. Mater. Sci. 30 (1995) 2803. [9] S.W. Nam, G.R. Gavalas, AIChE Symp. Ser. 85 (1989) 68. [10] R.S.A. de Lange, K. Keizer, A.J. Burggraaf, Ind. Eng. Chem. Res. 34 (1995) 3838. [11] S. Giessler, J.C. Diniz da Costa, G.Q. Lu, J. Nanosci. Nanotechnol. 1 (2001) 331. [12] R. Leboda, E. Mendyk, Mater. Chem. Phys. 27 (1991) 189. [13] M. Asaeda, Y. Sakou, J.H. Yang, K. Shimasaki, J. Membr. Sci. 209 (2002) 163. [14] W. Liu, B.Q. Zhang, X.F. Liu, L.M. Xu, Chin. J. Chem. Eng. 14 (2006) 31. [15] J. Sekulic, M.W.J. Luiten, J.E. ten Elshof, N.E. Benes, K. Keizer, Desalination 148 (2002) 19. [16] M. Kanezashi, M. Asaeda, J. Chem. Eng. Jpn. 38 (2005) 908. [17] M. Kanezashi, M. Asaeda, J. Membr. Sci. 271 (2006) 86. [18] F.A. Kettle, Physical Inorganic Chemistry: A Coordination Chemistry Approach, Spectrum Academic Publisher, Oxford, 1996, pp. 31–46.
263
[19] G.Z. Cao, J. Meijerink, H.W. Brinkman, A.J. Burggraaf, J. Membr. Sci. 83 (1993) 221. [20] R.R. Bhave, Inorganic membranes: synthesis, in: Characteristics and Applications, Van Nostrand Reinhold, New York, 1991, p. 338. [21] D.D. Do, Adsorption Analysis: Equilibria and Kinetics, vol. 1, Imperial College Press, London, 1998, p. 115. [22] T. Tsuru, T. Hino, T. Yoshioka, M. Asaeda, J. Membr. Sci. 186 (2001) 257. [23] T. Tsuru, Y. Takata, H. Kondo, F. Hirano, T. Yoshioka, M. Asaeda, Sep. Purif. Technol. 32 (2003) 23. [24] R.M. de Vos, H. Verweij, J. Membr. Sci. 143 (1998) 37. [25] D.W. Breck, Zeolite Molecular Sieves—Structure, Chemistry, and Use, John Wiley & Sons, New York, 1974, p. 636. [26] R.M. de Vos, H. Verweij, Science 279 (1998) 1710. [27] V. Boffa, J.E. ten Elshof, D.H.A. Blank, Micropor. Mesopor. Mater. 100 (2007) 173. [28] T. Yoshioka, T. Tsuru, M. Asaeda, Mol. Phys. 102 (2004) 191.
Contents lists available at ScienceDirect
Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci
Hydrothermal stability of microporous silica and niobia–silica membranes V. Boffa 1 , D.H.A. Blank, J.E. ten Elshof ∗ MESA Institute for Nanotechnology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
a r t i c l e
i n f o
Article history: Received 5 January 2008 Received in revised form 14 March 2008 Accepted 23 March 2008 Available online 30 March 2008 Keywords: Hydrothermal stability Niobia Silica Microporous membrane Permporometry
a b s t r a c t The hydrothermal stability of microporous niobia–silica membranes was investigated and compared with silica membranes. The membranes were exposed to hydrothermal conditions at 150 and 200 ◦ C for 70 h. The change of pore structure before and after exposure to steam was probed by single-gas permeation measurements and nanopermporometry. The hydrothermal stability of the niobia–silica membrane was found to be higher than that of silica. After hydrothermal treatment at 200 ◦ C, the hydrogen permeance of the niobia–silica top layer had declined by 32%, while the H2 permeance of the silica top layer was reduced by 73%. The apparent activation energies of the H2 permeance were 12.2 ± 0.2 and 15.3 ± 0.7 kJ mol−1 for silica and niobia–silica, respectively. Nanopermporometry experiments on the silica membrane were in semi-quantitative agreement with the gas permeation data. The data suggest that densification of the top layer occurred predominantly in those areas with the highest convex curvatures, thereby increasing the effective transport path of helium and hydrogen across the membrane. © 2008 Elsevier B.V. All rights reserved.
1. Introduction We recently developed a niobia–silica microporous membrane [1] that shows a very low permeability of CO2 . It may find application in enabling technologies for CO2 sequestration, or in membrane reactors to recover H2 from hydrogen synthesis processes such as steam reforming [2] and water–gas shift reaction [3]. Industrial application of ceramic membranes requires stability under working conditions for at least several years. Hydrothermally stable supports consisting of La-doped ␥-alumina mesoporous layers coated on ␣-alumina macroporous supports have already been developed [4]. Unfortunately, the long-term durability of gas-selective microporous top layers, which are all based on silica, is limited in the presence of steam [5]. Crystallization of microporous silica hardly occurs below 1100 ◦ C [6], but it undergoes drastic structural changes in the presence of steam below 200 ◦ C [7–11]. Hydrothermal exposure leads to collapse of the porous structure, yielding dense impermeable materials. Microcracks may form as a result of the stresses that develop during these structural rearrangements. Densification occurs mostly in the early stages of exposure to steam, but microcrack forma-
∗ Corresponding author. Tel.: +31 53 489 2695; fax: +31 53 489 3595. E-mail address: [email protected] (J.E. ten Elshof). 1 Present address: Dipartimento di Chimica Generale ed Organica Applicata, Universita` di Torino, Corso M.D’Azeglio 48, 10125 Torino, Italy. 0376-7388/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2008.03.042
tion is less predictable. Loss of selectivity due to this type of deterioration often occurs after extended periods of exposure to steam. Himai et al. [7] proposed that Si O bonds are broken upon interaction with water at high temperature, creating vicinal hydroxyl pairs that are subject to recondensation. This mechanism, consisting of the cyclic destruction and reconstruction of Si O Si network bonds, gives some flexibility to the material, so that it can reorganize itself eventually in a more stable and denser state. The hydroxylation of silica finally results in depolymerization [12], because silicic acid dissolves in the thin film of adsorbed water that is covering the material during exposure to steam. The silicic acid can freely migrate through the sorbed water layer until it precipitates and recombines with the silica matrix due to saturation. Dissolution (the depolymerization of silica) occurs fastest in those areas where the convex curvature is largest. This process is dominant in microporous silica, where pores have larger degrees of curvature than in mesoporous silica and are filled by water even at low water partial pressures. Microporous silica can be stabilized by introducing transition metals in the silica network [13–17]. Oxygen forms more polar and stable bonds with transition metals than with silicon. Furthermore, most transition metals have coordination numbers larger than that of silicon [18]. Thus, they tend to yield more closely packed and crystalline structures than pure silica. To determine the changes in pore size distributions upon hydrothermal treatment, different monitoring techniques can be
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
employed. One way is to carry out single-gas permeation experiments with a series of gases with different kinetic diameters. Assuming that separation is based on size differences only, a low permeance corresponds to a small number of pores that is accessible for a species with a certain size. The advantage of this approach is that it is not dependent on any theoretical model. The results may nevertheless be affected by differences in interactions between different molecular probes and the membrane pore walls [1]. Another established technique to determine pore size distributions in supported thin films is permporometry. Its advantage is that all pore sizes are accessed with the same molecular probe, so that pore size distributions are not influenced by probe–wall interactions. On the other hand, estimates of pore size distributions can only obtained by applying theoretical models and making assumptions regarding the shape of the pores. The permporometry technique was originally developed in the 1990s for mesoporous membranes [19]. Pore size distributions can be calculated by applying the Knudsen [20] and Kelvin [21] equation. More recently, Tsuru developed the nanopermporometry technique, which is able to access micropores [22,23]. However, the interpretation of nanopermporometry data is more complex than conventional permporometry data. The Knudsen and Kelvin equations are not valid in the microporous regime [22], and simple alternative models are not available in literature. Tsuru et al. defined the average pore size as the Kelvin diameter at which the permeance of the vapor-containing gas was half of the permeance of the dry gas [22]. Here the Kelvin diameter is two times the Kelvin radius (rK ), which is defined by the Kelvin equation: ln Pr = −
s Vmol cos 1 RT rK
(1)
where Pr is the relative vapor pressure, s (J m−2 ) the gas–solid interfacial tension, Vmol (m3 mol−1 ) the molar volume, and is the contact angle of the condensed phase on the pore wall. The latter value is normally assumed to be zero. In this work the hydrothermal stability of a niobia–silica microporous membrane was investigated and compared with that of pure silica. Nanopermporometry and single-gas permeation measurements were employed to probe the change of pore structure of the membrane after exposure to steam at elevated temperatures. 2. Experimental 2.1. Sol synthesis The niobia-doped silica sol (denoted as NS sol) was prepared by adding 11 ml of tetraethyl orthosilicate (Aldrich, 99.999% pure; denoted as TEOS) to 10.5 ml of ethanol. An aqueous solution of
257
nitric acid was dropped in the mixture until the final molar ratio (Si)–OC2 H5 :H2 O:HNO3 was 1:0.5:0.01. We have chosen to express the chemical composition of the sol in this way because alkoxy groups, water and acid are the moieties that participate directly in the reaction. This mixture was heated under reflux at 60 ◦ C for 2 h. A 1-M solution of niobium(V) penta(n-butoxide) (Gelest) in n-butanol (Aldrich, anhydrous) was added slowly to the mixture. Aqueous nitric acid was dropped into the mixture to restore the initial composition (M)–O–R:H2 O:HNO3 to 1:0.5:0.01, with M = Si or Nb and R = C2 H5 or C4 H9 . The sol was refluxed at 60 ◦ C for 5 h. Then it was cooled down to 20 ◦ C and kept at this temperature for several days until it was used for characterization and membrane preparation. Prior to coating membranes, the sol was diluted 11.5 times with ethanol. The silica sol was prepared by adding 11 ml of 0.73N aqueous nitric acid to 21 ml of tetraethylorthosilicate (TEOS, 98%, Aldrich) dissolved in 21 ml ethanol (≥99.5%, Aldrich). The addition was carried out drop by drop under vigorous stirring, while the reaction flask was cooled in an ice bath. The mixture was then refluxed at 60 ◦ C for 3 h. The resulting silica sol was diluted 18 times in ethanol and kept at −5 ◦ C for a few days until it was used for preparation of membranes. 2.2. Preparation of silica membranes Flat supports were prepared by coating a 6% La-doped boehmite sol on a 2.0-mm thick ␣-alumina disk (∅ 39 mm) as described elsewhere [4]. After calcination at 600 ◦ C a mesoporous 6% La-doped ␥-alumina layer with average pore size of 5 nm was formed, as verified by conventional permporometry [10]. Silica and niobia–silica membranes were prepared by dipping the supports in the sols at an angular speed of 0.06 rad s−1 . The membranes were calcined for 3 h in air at 500 ◦ C, with heating and cooling rates of 0.5 ◦ C/min. Since the silica sol yielded an extremely thin layer, the coating and calcination cycle was repeated to increase film thickness and cover any remaining defects or uncoated areas of the membrane. The niobia–silica membrane is further referred to as NS membrane. 2.3. Single-gas permeation experiments Permeation measurements were carried out using the setup shown in Fig. 1. In the single-gas permeation experiments all valves were closed with the exception of valve number 2, and the setup was used as a dead-end mode permeation setup [24]. In a typical experiment, the membrane was first degassed at 200 ◦ C for 2 days in a flowing helium atmosphere. Then the flows of He, H2 , CO2 , N2 , CH4 and SF6 were measured sequentially, employing a pressure of 5 bar on the side of the feed, and atmospheric pressure on the per-
Fig. 1. Schematic diagram of the experimental setup.
258
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
meate side. Steady state was ensured by waiting at least 3 h prior to sampling data. In the case of hydrogen and helium the measurements were carried out between 90 and 200 ◦ C to determine the apparent activation energies of permeance of these gases. The volumetric gas flow on the permeate side was measured with a soap film flow meter. 2.4. Nanopermporometry experiments Nanopermporometry experiments were performed using steam (kinetic diameter ∼0.21 nm) as condensable vapor and hydrogen (0.29 nm) as incondensable gas. The setup shown in Fig. 1 was used, since the relative pressure of water on the feed side can be controlled accurately by the mass flow controllers MFC1 and MFC2. The pressure controller was excluded from the active part of the gas permeation setup by closing valve number 2, and opening the other valves. Valve 7 was regulated manually to keep the H2 pressure gradient across the membrane at 1 kPa. While flushing the membrane with dry hydrogen the sample temperature was decreased from 200 to 90 ◦ C. The temperature of the water bath was kept at 32 ◦ C. The steam partial pressure in the feed was gradually increased by varying the gas flows through the two mass flow controllers MFC1 and MFC2. Pore size distributions were calculated from the change of hydrogen permeance with changing steam partial pressure, as discussed in more detail below. 2.5. Hydrothermal treatments The hydrothermal stability of the membranes was tested by exposure to steam. The membranes were first exposed to a steam partial pressure of 0.56 bar at 150 ◦ C for 70 h, using hydrogen as sweep gas. The total overpressure at the feed side was 4 bar. Then the sample was dried at 200 ◦ C for 2 days. Afterwards, a series of single-gas measurements was carried out at 200 ◦ C using gases with increasing kinetic diameter. The pore size distribution of the silica top layer was measured by nanopermporometry under the same experimental conditions as used for the fresh sample. After exposure to 0.56 bar of steam at 200 ◦ C for 70 h, the gas permeation and nanopermporometry experiments were repeated a third time. 2.6. Scanning electron microscopy Scanning electron microscopy (SEM) images were taken on a LEO 1550 FEG.
3. Results and discussion 3.1. Membrane morphology The SEM images in Fig. 2 shows cross-sections of the silica and NS membrane. The gas-selective layers have thicknesses of about 30 and 150 nm for silica and NS, respectively. Silica and NS-derived powders were X-ray amorphous after calcination at 500 ◦ C. 3.2. Single-gas permeation experiments: pore size distribution In the Knudsen regime, which covers the pore size range between 2 and 50 nm, the permeability of a membrane F is inversely proportional to the square root of the molecular mass of the permeating species [20]:
F =K
1 MRT
(2)
where K is a constant that depends on pore shape, membrane area and thickness of the top layer, R the gas constant, T the absolute temperature, and M the molar mass of the gas. Single-gas permeation measurements on mesoporous ␥alumina supports reported earlier [20] showed that the gas flow rate has a Knudsen-type relation with temperature and molecular mass. The results of gas permeation experiments on membranes that include microporous silica-based top layers are listed in Tables 1 and 2. The kinetic diameters of the corresponding probe molecules are also given. No correlation with the trend predicted by Eq. (1) was observed. In general, the permeance decreased gradually with increasing kinetic diameter. The permeance of SF6 in silica was immeasurably low. This indicates that the majority of pores in the silica layer has a diameter between 0.3 and 0.4 nm, and a negligible number of pores larger than 0.55 nm. The helium permeance of the NS membrane was about 12 times higher than that of nitrogen, so a large fraction of pores has dimensions smaller than the size of nitrogen (0.36 nm). But in contrast to silica, the NS membrane was also permeable to SF6 . This implies the presence of a small fraction of pores larger than ∼0.55 nm. As we reported elsewhere [1], the NS membrane has an exceptionally low permeability for CO2 . This strongly suggests that the larger pores through which SF6 diffused must be in the microporous range (<2 nm). Otherwise the presence of mesopores would have resulted in a much higher permeability of CO2 [1].
Fig. 2. SEM images of cross sections of (a) microporous silica top-layer and (b) a NS gas-selective top layer.
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
259
Table 1 Single-gas permeance of supported silica membrane before and after the first (HT150) and second (HT200) hydrothermal treatment Molecule
He H2 CO2 N2 CH4 SF6
dk (nm)a
0.26 0.29 0.33 0.36 0.38 0.55
Fresh sample
After HT150
Fb
FH2 /FX
Fb
FH2 /FX
Fb
FH2 /FX
1.0 1 46 1.5 × 102 5.6 × 102 ≥103
3.7 × 10−7
0.85 1 33 1.4 × 102 2.0 × 102 ≥10 3
3.7 × 10−7 3.0 × 10−7 6.3 × 10−9 2.1 × 10−9 1.1 × 10−9 ≤1 × 10−10
0.75 1 47 1.4 × 102 2.7 × 102 ≥103
6.1 × 10−7
6.3 × 10−7 1.3 × 10−8 4.2 × 10−9 1.1 × 10−9 ≤1 × 10−10
3.1 × 10−7 9.5 × 10−9 2.2 × 10−9 1.5 × 10−9 ≤1 × 10−10
After HT200
The ideal hydrogen to X permselectivity is reported for every set of measurements. a Kinetic diameters d taken from Ref. [25]. k b Permeance (F) is expressed in mol s−1 m−2 Pa−1 .
Table 2 Single-gas permeances of supported NS membrane before and after the first (HT150) and second (HT200) hydrothermal treatment Molecule
dk (nm)a
Fresh sample Fb
He H2 CO2 N2 CH4 SF6
0.26 0.29 0.33 0.36 0.38 0.55
7.2 × 10−8
3.9 × 10−8 8.5 × 10−10 6.2 × 10−9 5.4 × 10−9 2.9 × 10−9
After HT150 FH2 /FX
Fb
0.54 1 46 6.3 7.2 13
5.2 × 10−8
3.7 × 10−8 8.2 × 10−10 6.2 × 10−9 5.8 × 10−9 2.9 × 10−9
After HT200 FH2 /FX
Fb
FH2 /FX
0.71 1 45 6.0 6.4 13
4.0 × 10−8
0.68 1 33 4.0 5.0 9.0
2.7 × 10−8 8.3 × 10−10 6.7 × 10−9 5.4 × 10−9 3.0 × 10−9
The ideal hydrogen to X permselectivity is reported for every set of measurements. a Kinetic diameters d taken from Ref. [25]. k b Permeance (F) is expressed in mol s−1 m−2 Pa−1.
3.3. Single-gas permeation experiments: hydrothermal stability The overall permeance after both hydrothermal treatments is listed in Tables 1 and 2. The abbreviations HT150 and HT200 are used to indicate the data acquired after the first and second hydrothermal treatment, respectively. Exposure to steam at high temperature led to densification of the microporous silica phase. This reduced the permeance of hydrogen and helium. The densification of silica was significant after HT150, but less pronounced after HT200. The permeance of larger molecules did not increase after either HT150 or HT200. This indicates the absence of cracks in the film. The overall He permeance of the niobia–silica membrane decreased from 7.2 × 10−8 to 5.2 × 10−8 mol Pa−1 m−2 s−1 after HT150 and to 4.0 × 10−8 mol Pa−1 m−2 s−1 after HT200. The decrease of H2 permeance after HT200 was less pronounced than that of helium. The permeance of molecules with a larger kinetic diameter remained almost unaffected by the treatment. The decrease of membrane permselectivity towards mixtures of hydrogen and other gases is therefore due to the decline of hydrogen permeance. The permeance of the gas-selective top layer of both membranes was calculated by considering that the membrane consists of a series of layers, each with a certain resistance to gas flow. The resistance of each layer is proportional to the reciprocal permeance. The permeance of the gas-selective layer Ftop can thus be determined from the total permeance Ftot and the permeance of the support Fsup via Eq. (3) [24]: 1 1 1 = + Ftot Ftop Fsup
The hydrogen permeance of the silica top layer at 200 ◦ C was 1.6 × 10−6 mol Pa−1 m−2 s−1 , and 3.8 × 10−8 mol Pa−1 m−2 s−1 for NS. Defining permeability as permeance per unit membrane thickness, this implies that the H2 permeabilities at this temperature were 4.5 × 10−14 mol Pa−1 m−1 s−1 for silica and 5.7 × 10−15 mol Pa−1 m−1 s−1 for NS. The relatively low permeability of NS is probably related to the lower hydrolysis and acid ratios at which the niobia–silica sol was synthesized. The incorporation of unreacted precursors and small clusters into the NS layer may be the cause of the relatively dense layer. Alternatively, the lower permeability of NS might also be due to the incorporation of Nb ions in a microporous silica network. The effects of hydrothermal treatment of silica and NS are compared in Fig. 3. Hydrothermal densification is more prominent for silica than for NS. The helium permeance of silica was reduced to 21% after HT150, and the hydrogen permeance to 27%. HT200
(3)
The permeance of the support Fsup to different probes at various temperatures was measured directly, or extrapolated by fitting Eq. (3) to a series of experimental data on different probe molecules, in a range of temperatures between 90 and 200 ◦ C.
Fig. 3. Influence of HT150 and HT200 on permeance of hydrogen and helium in silica top layer and NS top layer.
260
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
Fig. 4. Arrhenius plot of the hydrogen and helium permeance in silica top layer, both fresh and after HT200.
did not have any further effect. The decline of permeance was less pronounced in niobia–silica. After HT150, the hydrogen permeance was reduced to about 95%, and the helium permeance to 68%. This suggests that only a small fraction of micropores of NS underwent substantial densification upon hydrothermal treatment, namely those pores that were small enough to be accessible to He but not to H2 . HT200 caused an overall decline of permeance to 68% for hydrogen and 47% for helium. Apparently, the introduction of niobium in the silica framework yielded a more stable material than pure silica. The effect of Nb on hydrothermal stability is evident after HT150. However, microporous structures are thermodynamically unstable by definition and therefore remain prone to densification. An alternative explanation for the higher stability of NS could also be that it has slightly larger pores than silica. Smaller pores are thermodynamically more sensitive to hydrothermal degradation than larger ones, but it is very unlikely that this is the sole effect. The data presented here do not allow estimation
to which extent these two effects contribute each to the stability of the NS layer. 3.4. Effect of hydrothermal modification on activation energy of permeance The Arrhenius plots of the hydrogen and helium permeances of the silica and NS top layers are shown in Figs. 4 and 5. The data were calculated using Eq. (3). The transport of gases in the microporous regime occurs via a thermally activated surface diffusion mechanism. The transport rate can be expressed in terms of a modified Fick law [24]: F = F0 exp
E a −
RT
(4)
where F is the membrane permeance (flux per unit of driving force), Ea is the apparent activation energy, and F0 is a temperature inde-
Fig. 5. Arrhenius plot of the hydrogen and helium permeance in NS top layer, both fresh and after HT200.
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
pendent parameter that has the dimensions of permeance. The value of F0 depends on the interaction between pore wall and gas molecule, thickness of the gas-selective layer, and on pore shape and tortuosity. The straight lines in the figures show linear fits to the data with slopes equal to −Ea [kJ mol−1 ]. The apparent activation energy of permeance of silica was 12.8 ± 0.3 kJ mol−1 for He and 12.2 ± 0.2 kJ mol−1 for H2 . After HT200, the activation energies were 12.9 ± 0.1 kJ mol−1 for He and 11.9 ± 0.4 kJ mol−1 for H2 . These data do not indicate any change of the predominant transport mechanism upon hydrothermal treatment. The experimental data of both hydrogen and helium for NS at lower temperatures deviated from the exponential behavior predicted by Eq. (4). This can be attributed to the existence of two parallel transport paths through the top layer. One path is via the selective micropores, the other path via larger pores. The latter one is probably associated with a Knudsen-type diffusion mechanism [20]. Knudsen diffusion does not depend strongly on temperature; its rate is proportional to T−0.5 . The relative contribution of Knudsen diffusion will therefore increase with decreasing temperature and may even become dominant, as seems to be the case for NS at low temperatures. This is consistent with the fact that deviation from Eq. (4) was not observed for the silica membrane, where gas permeation experiments had already provided sufficient proof for the absence of large micropores or mesopores. The temperature below which the permeance of the NS membrane deviated significantly from the prediction of Eq. (4) were 70 and 90 ◦ C for He and H2 , respectively. After hydrothermal exposure, these values shifted to 130 ◦ C for hydrogen and 115 ◦ C for helium. Activation energies of 19.1 ± 0.3 kJ mol−1 for helium and 15.3 ± 0.7 kJ mol−1 for hydrogen were calculated from the permeance data of the fresh NS membrane in Fig. 5. After HT2, these values were similar, i.e., 19.3 ± 0.6 and 15.7 ± 0.7 kJ mol−1 , respectively. Qualitatively similar observations were made by De Vos et al., who reported an apparent activation energy of hydrogen permeance of ∼8 kJ mol−1 for silica membranes calcined at either 400 or 600 ◦ C [24]. Silica membranes calcined at 600 ◦ C were denser and had smaller pores than those calcined at 400 ◦ C [24,26], but no difference in activation energy was observed. 3.5. Nanopermporometry measurements: pore size distribution The Kelvin and Knudsen equations were used to calculate the corresponding Kelvin diameter from the experimental data. Eq. (5) [27] was used to analyze the permporometry curves shown in Fig. 6: n(rK ) = −
K rK3
RTM 2
F
acc
r
261
Fig. 6. Analysis of microporous silica membrane by nanopermporometry at 90 ◦ C. H2 permeance versus relative pressure of steam is indicated by the thin drawn line. Pore size distribution (thick line) was calculated from this curve using Eq. (5).
less pronounced. At a relative pressure of about 20%, the hydrogen permeance was less than 10% of its initial value. The estimated pore size distribution of the silica membrane as calculated from Eq. (5) is also shown in Fig. 6. The curve indicates a monomodal distribution with an average Kelvin diameter of about 0.5 nm. Fig. 7 shows the permeance as function of the kinetic diameter of the gas probe molecules, after correction for the support resistance. In contrast to the interpretation of nanopermporometry data, the permeance versus kinetic diameter data are empirical and provide a direct indication of the pore size distribution. The effective pore size distribution obtained with nanopermporometry is shown for the sake of comparison. Application of Eq. (5) seems to lead to an overestimation of the pore diameter by about 0.2 nm. However, it is also possible that an appreciable permeance occurs only in pores that are at least 1–2 A˚ larger than the kinetic diameter of the permeating molecule. This would imply that nanopermporometry cannot measure pores smaller than 0.4–0.5 nm, since the water, hydrogen and helium molecules used in that experiment have a kinetic diameter of about 0.2–0.3 nm.
(5)
where n represents the number of pores and the parameter rK expresses the effective radius corresponding to a certain relative pressure of steam as calculated by the Kelvin equation. The final term at the right-hand side of the equation expresses the derivative of permeance over rK . Although the assumptions underlying these equations are not valid in the microporous regime, they allow qualitative and semi-quantitative comparison of pore size distributions of membranes, and are therefore suitable parameters to monitor the degradation of membranes. The permeability of the NS layer was too low to acquire sufficiently accurate data. Fig. 6 shows the variation of hydrogen permeance of silica as function of the partial pressure of steam. A significant reduction of H2 flux was observed when the relative pressure of steam was increased from 1.1 to 7.0%, so most of the micropores were closed by capillary condensation of water at the latter vapor pressure. At higher relative pressures, the reduction of hydrogen permeance with relative pressure of water was much
Fig. 7. Permeance of fresh silica top layer as function of the kinetic diameter of the probe molecule (solid line) and pore size distribution calculated from nanopermporemetry data (dashed line).
262
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263
Fig. 8. Effect of hydrothermal treatment on hydrogen permeance in silica membrane as measured by nanopermporometry: (a) normalized permeance and (b) calculated pore size distribution.
Fig. 9. Scheme of the effect of hydrothermal densification on microporous structure. The effective path length of a He molecule is increased after steam exposure, while the permeation of SF6 in the larger pores of the NS membrane remains unaffected.
3.6. Nanopermporometry measurements: hydrothermal stability
4. Conclusions
We also applied nanopermporometry to characterize degradation phenomena in microporous silica membranes by investigating changes in the pore size distributions of the gas-selective top layers after hydrothermal treatment. Results are depicted in Fig. 8. The data support the trends of the single-gas permeation measurements. As can be seen in Fig. 8b, the fraction of pores larger than 0.6 nm had decreased after hydrothermal treatment. Apparently, densification narrowed all pores indiscriminately, reducing both the number of large pores, and the permeance of the membrane to all gases. The change of pore size is more pronounced during the first hydrothermal treatment at 150 ◦ C. This is in agreement with the findings of the single-gas permeation experiments. A possible mechanism for the change of pore structure upon hydrothermal treatment is schematically illustrated in Fig. 9. The pore structure of microporous silica [28], and probably of NS too, roughly resembles that of a sponge. The pores with larger curvature, i.e., the smaller micropores, are more vulnerable to hydrothermal attack than pores with smaller curvature, such as larger micropores and mesopores. According to the densification mechanism proposed here, only a small fraction of the pore structure was densified during the limited time of exposure to hydrothermal conditions. The smallest pores in the microporous thin layer became blocked, thereby increasing the real length of the transport path of small molecules through the membrane, and lowering the membrane selectivity. This explains why the reduction of hydrogen and helium permeance was substantial, while no variation of activation energy was recorded.
The NS membrane was found to be more stable under hydrothermal conditions than the silica membrane. This is probably the result of the incorporation of Nb ions into the silica matrix, which yields hydrothermally stable Nb O Si bonds, although differences in pore size, density and layer thickness may also play a secondary role in stabilizing the NS membrane network. Hydrothermal treatment led to a decrease of the permeability of all gases, but the decrease was more prominent for small gases like He and H2 . No apparent changes occurred in the mechanism of gas transport through the membranes. Longer stability tests will reveal to which extent the presence of Nb ions stabilizes the microporous network by slowing down the kinetics of densification. It was shown that the nanopermporometry technique can be applied to estimate pore size distributions of microporous layers in a semi-quantitative way. This fast and non-destructive technique may be applied to other microporous systems as well. The nanopermporometry and single-gas permeation data yielded qualitatively similar results. References [1] V. Boffa, J.E. ten Elshof, A.V. Petukhov, D.H.A. Blank, Chem. Sus. Chem. 1 (2008) 437–443. [2] D.W. Lee, B. Sea, K.Y. Lee, K.H. Lee, Ind. Eng. Chem. Res. 41 (2002) 3594. [3] A. Brunetti, G. Barbieri, E. Drioli, T. Granato, K.H. Lee, Chem. Eng. Sci. 62 (2007) 5621. [4] A. Nijmeijer, H. Kruidhof, R. Bredesen, H. Verweij, J. Am. Ceram. Soc. 84 (2001) 136. [5] S. Giessler, J.C.D. da Costa, G.Q. Lu, J. Nanosci. Nanotechnol. 1 (2001) 331.
V. Boffa et al. / Journal of Membrane Science 319 (2008) 256–263 [6] R.K. Iler, The Chemistry of Silica, Wiley, New York, 1979. [7] H. Himai, H. Morimoto, A. Tominaga, H. Hiraschima, J. Sol–Gel Sci. Technol. 10 (1997) 45. [8] G.P. Fotou, Y.S. Lin, S.E. Pratsinis, J. Mater. Sci. 30 (1995) 2803. [9] S.W. Nam, G.R. Gavalas, AIChE Symp. Ser. 85 (1989) 68. [10] R.S.A. de Lange, K. Keizer, A.J. Burggraaf, Ind. Eng. Chem. Res. 34 (1995) 3838. [11] S. Giessler, J.C. Diniz da Costa, G.Q. Lu, J. Nanosci. Nanotechnol. 1 (2001) 331. [12] R. Leboda, E. Mendyk, Mater. Chem. Phys. 27 (1991) 189. [13] M. Asaeda, Y. Sakou, J.H. Yang, K. Shimasaki, J. Membr. Sci. 209 (2002) 163. [14] W. Liu, B.Q. Zhang, X.F. Liu, L.M. Xu, Chin. J. Chem. Eng. 14 (2006) 31. [15] J. Sekulic, M.W.J. Luiten, J.E. ten Elshof, N.E. Benes, K. Keizer, Desalination 148 (2002) 19. [16] M. Kanezashi, M. Asaeda, J. Chem. Eng. Jpn. 38 (2005) 908. [17] M. Kanezashi, M. Asaeda, J. Membr. Sci. 271 (2006) 86. [18] F.A. Kettle, Physical Inorganic Chemistry: A Coordination Chemistry Approach, Spectrum Academic Publisher, Oxford, 1996, pp. 31–46.
263
[19] G.Z. Cao, J. Meijerink, H.W. Brinkman, A.J. Burggraaf, J. Membr. Sci. 83 (1993) 221. [20] R.R. Bhave, Inorganic membranes: synthesis, in: Characteristics and Applications, Van Nostrand Reinhold, New York, 1991, p. 338. [21] D.D. Do, Adsorption Analysis: Equilibria and Kinetics, vol. 1, Imperial College Press, London, 1998, p. 115. [22] T. Tsuru, T. Hino, T. Yoshioka, M. Asaeda, J. Membr. Sci. 186 (2001) 257. [23] T. Tsuru, Y. Takata, H. Kondo, F. Hirano, T. Yoshioka, M. Asaeda, Sep. Purif. Technol. 32 (2003) 23. [24] R.M. de Vos, H. Verweij, J. Membr. Sci. 143 (1998) 37. [25] D.W. Breck, Zeolite Molecular Sieves—Structure, Chemistry, and Use, John Wiley & Sons, New York, 1974, p. 636. [26] R.M. de Vos, H. Verweij, Science 279 (1998) 1710. [27] V. Boffa, J.E. ten Elshof, D.H.A. Blank, Micropor. Mesopor. Mater. 100 (2007) 173. [28] T. Yoshioka, T. Tsuru, M. Asaeda, Mol. Phys. 102 (2004) 191.