Anal. Chem. 1999, 71, 1016-1020
Time-Dependent Permeance of Gas Mixtures through Zeolite Membranes Kevin H. Bennett and Kelsey D. Cook*
Department of Chemistry, University of Tennessee, Knoxville, Tennessee 37996-1600 John L. Falconer and Richard D. Noble
Department of Chemical Engineering, University of Colorado, Boulder, Colorado 80309-0424
The time-dependent permeation behavior of binary gas mixtures through a ZSM-5 zeolite membrane was studied. Although steady-state permeation rates were indistinguishable for CO2 and N2 or for cis- and trans-2-butene in binary mixtures, differences in the rate of approach to steady state allowed component distinction. In “normal” systems, one component is initially enriched in the permeate following application of a pulse of analyte gas to the membrane, and then disappears more quickly upon termination of the pulse. Mixtures of cis- and trans-2butene exhibit qualitatively different behavior; the permeate is enriched in cis-2-butene during both the leading and trailing edges of a sample pulse (though not at steady state). These differences in permeation behavior reflect different balances among multiple transport mechanisms through the zeolite membrane, thought to reflect a combination of selective component sorption and intracrystalline diffusion; in the case of cis- and trans-2butene, these two factors oppose one another. It is known that this mechanistic complexity can engender synergistic effects, wherein the presence of one component can affect the permeation of another. These may limit applicability to true “unknowns”, but resulting complications should be less problematic in well-defined process applications. Membrane inlet mass spectrometry (MIMS)1 has been used to enhance selectivity and sensitivity of on-line process mass spectrometry.2,3 Response time (e.g., for following changes in stream composition) is important for monitoring and control applications; in general, MIMS response times are limited by analyte permeation times through the membrane. Permeation times, in turn, are related inter alia to membrane thickness; thinner membranes give faster response (within the limits of membrane robustness). Silicone rubber membranes are used most often, providing selectivity for analysis of volatile organics in aqueous streams; applications include fermentation monitoring, waste stream analysis, and ambient air monitoring.2 Discrimination (1) Srinivasan, N.; Johnson, R. C.; Kasthurikrishnan, N.; Wong, P.; Cooks, R. G. Anal. Chim. Acta 1997, 350, 257-271. (2) Cook, K. D.; Bennett, K. H.; Haddix M. L. Ind. Eng. Chem. Res., in press. (3) Blaser, W. W.; Bredeweg, R. A.; Harner, R. S.; LaPack, M. A.; Leugers, A.; Martin, D. P.; Pell, R. J.; Workman, J.; Wright, L. G. Anal. Chem. 1995, 67, 47R-70R.
1016 Analytical Chemistry, Vol. 71, No. 5, March 1, 1999
against water and hydrophilic solutes (including salts) is effective, providing very low backgrounds and excellent limits of detection for dissolved organic compounds.4 Imbedding powdered zeolite crystals within a silicone membrane can increase the total flux of organics from organic/water mixtures.5 Additional adsorptive selectivity can also result. Several other membrane types have been used to provide different selectivities.6 For example, microporous Teflon membranes derive selectivity primarily from steric (size exclusion) effects.7 Even in cases where mixture components have identical steadystate membrane permeances, differences in the rates at which different components approach steady state can sometimes be exploited to help resolve mixtures.8 In such “dynamic MIMS” applications, a sample is pulsed to the inlet side of a membrane and the species-dependent permeation delay in the mass spectrometric ion signal derived from gas sampled from the other side of the membrane induces a phase shift in signals for different components that can be related to sample composition.8 Recently,9,10 we assessed the potential MIMS utility of a new and unusual class of zeolite membranes.11 In contrast to the silicone membrane-doping experiments,5 the zeolite crystals in these experiments are grown directly as a thin film on a porous ceramic support, resulting in greatly enhanced physical robustness.12 The thin-film geometry can also allow for fast response to changes in stream composition, making these membranes good candidates for process-monitoring applications. Permeation of molecules through these membranes has been described as a five-step process involving molecular adsorption, transport to the pores, intracrystalline transport, transport out of the pores, and desorp(4) Soni, M.; Sauer, S.; Amy, J. W.; Wong, P.; Cooks, R. G. Anal. Chem. 1995, 67, 1409-1412. (5) Hennepe, H. J. C.; Boswerger, W. B. F.; Bargeman, D.; Mulder, M. H. V., Smolders, C. A. J. Membr. Sci. 1994, 89, 185-196. (6) Maden, A. J.; Hayward, M. J. Anal. Chem. 1996, 68, 1805-1811. (7) Kasthurikrishnan, N.; Cooks, R. G. Talanta 1995, 42, 1325-1334. (8) Overney, F. L.; Enke, C. G. J. Am. Soc. Mass Spectrom. 1996, 7, 93-100. (9) Cook, K. D.; Bennett, K. H.; Haddix, M. L.; Keator, E. A.; Seebach, G. L.; Falconer J. L. J. Process Anal. Chem. 1998, 3 (3-4), 115-124. (10) Bennett, K. H.; Cook, K. D. 46th ASMS Conference on Mass Spectrometry and Allied Topics; Orlando, FL, 1998; p 1465. (11) Coronas, J.; Noble, R. D.; Falconer, J. L. Ind. Eng. Chem. Res. 1998, 37, 166-176. (12) Funke, H. H.; Argo, A. M.; Falconer, J. L.; Noble, R. D. Ind. Eng. Chem. Res. 1997, 36, 137-143. 10.1021/ac980991n CCC: $18.00
© 1999 American Chemical Society Published on Web 01/21/1999
tion.13 All five steps can be influenced by factors such as the temperature, pressure, and concentration of the gas(es) passing through the membrane.14 The complexity of the situation is evidenced by the fact that relative permeances of pure gases are sometimes inverted when the gases are sampled together in mixtures. For example, when hydrogen and n-butane were sampled separately through a zeolite membrane, hydrogen permeated faster by a factor of ∼20, consistent with its high diffusivity.15 However, when the gases were sampled as a binary mixture, n-butane was enriched significantly (by a factor of 125 relative to the feed mixture) in the permeate once steady state was achieved.15 This is thought to be due to preferential adsorption of n-butane; in essence, the large (0.43-nm kinetic diameter15) adsorbed butane molecules “block” the pore access of the small (0.29-nm15) hydrogen molecules.16 (The kinetic diameter is a measure of the minimum effective diameter of a molecule, based on the Lennard-Jones relationship.17 It is used to predict permeation behavior when steric factors are limiting.) Such synergistic behavior is strongly dependent on concentration; there must be sufficient n-butane to saturate the membrane pores. Intercomponent interactions such as these constitute potentially complicating matrix effects that might severely limit applicability to true “unknowns” but are of less concern for relatively well-defined process monitoring applications. One feature that makes zeolites attractive for MIMS applications is their extraordinary selectivity (S, eq 1), where CA is the
S ) (CA/CB)p/(CA/CB)f
(1)
concentration of the compound of interest and CB is the concentration of the other compound in a binary mixture (or the sum of other species’ concentrations in more complex mixtures); p and f correspond to permeate and feed, respectively.15 Illustrative of the selectivity achievable with these membranes, Ssteady-state,n-hexane ) 2000 in an isomeric n-hexane/2,2-dimethylbutane mixture.11,18 This high selectivity is again due primarily to preferential adsorption; here, n-hexane adsorbs strongly due to a better “fit” to the membrane pores, effectively inhibiting permeation of 2,2-dimethylbutane. Selectivities among organics are rarely if ever reported for “conventional” membranes; as noted above, these are generally used to separate organics from water, rather than distinguishing among organics. For mixtures where data are available, zeolite selectivities are comparable to those found for polymeric membranes (e.g., S ) 650 for acetone/water for a silicone membrane and S ) 250 for acetone/water for a zeolite membrane).19 The multiple and potentially competitive factors affecting permeance through a zeolite membrane20 can provide a source (13) Barrer, R. M. J. Chem. Soc., Faraday Trans. 1990, 86, 1123-1130. (14) Graaf, J. M.; Kapteijn, F.; Moulijn, J. A. Struct. Catal. Reactors; Chem. Ind. 1998, 71, 543-573. (15) Bakker, W. J. W.; Kapteijn, F.; Poppe, J.; Moulijn, J. A. J. Membr. Sci. 1996, 117, 57-78. (16) Krishna, R.; Smit, B.; Vlugt, T. J. H. J. Phys. Chem. A 1998, 102, 77277730. (17) Breck, D. W. Zeolite Molecular Sieves: Structure, Chemistry, and Use; Wiley: New York, 1974; p 636. (18) Vroon, Z. A. E.P.; Keizer, K.; Gilde, M. J.; Verweij, H.; Burggraaf, A. J. J. Membr. Sci. 1996, 113, 293-300. (19) Liu, Q.; Noble, R. D.; Falconer, J. L.; Funke, H. H. J. Membr. Sci. 1996, 117, 163-174. (20) Goldon, C.; Sircar, S. J. Colloid and Interface Sci. 1994, 162, 182-188.
Table 1. Kinetic Diameters, Polarizabilities, and Polarities of the Compounds Tested
compound
kinetic diam (nm)
polarizability (10-24 cm3)24
dipole moment (D)24
diffusivity (10-8 m2 s-1)
cis-2-butene trans-2-butene carbon dioxide nitrogen ethane propane
0.4823 0.4523 0.3315 0.3615 0.3815 0.4315
8.49 2.91 1.74 4.47 6.29
0.33 0 0 0 0 0.084
1.0623 0.8023 0.722 1.322 1.722 0.7522
of the time dependence required for dynamic MIMS. This report tests for such time dependence for three binary gaseous mixtures passing through a ZSM-5 zeolite membrane. The systems tested have small or negligible steady-state selectivities (due inter alia to small differences in polarity, polarizability, kinetic diameter, and/or diffusivity; see Table 1); there is no need to resort to dynamic MIMS when selectivities are large. The work aims to improve our understanding of permeation through these membranes, while providing a preliminary assessment of the feasibility of exploiting them in sensor applications using dynamic MIMS. Specific application to resolving a pair of stereoisomers is included. EXPERIMENTAL SECTION A ZSM-5 zeolite membrane was synthesized on the inside surface of a cylindrical porous alumina support (US-Filters, 1.0cm o.d., 0.65-cm i.d., 4.8 cm long) at the University of Colorado. The membrane synthesis procedure has been described in detail.21 The membrane was used at ambient temperature (∼23 °C). All samples were prepared using CP grade gases (Matheson Gas Products, Marrow, GA). Mixtures were prepared on-line using mass flow controllers with 2-µm filters (Brooks Instruments, Hatfield, PA) to inject pure components into a 1/8-in. stainless steel mixing “T” (Swagelok, Solon, OH). The total sample flow was 10 standard cm3/min (sccm). The sample mixture subsequently flowed through 1/8-in. stainless steel tubing to a second 1/8-in. mixing “T”, where it was mixed with 10 sccm He (total gas flow 20 sccm). The gas mixture then flowed to a switching valve (model 6UW, Valco, Houston, TX) which was manually activated to generate 2-min sample pulses into the membrane inlet (Figure 1). Between replicate pulses, sample was directed to an exhaust vent and 20 sccm He was directed to the membrane so that the sample flow controllers were not disturbed by flow interruptions. Following the switching valve, the mixture or He carrier flowed through the interior of the membrane tube, which was secured by two silicone O-rings, allowing a separate flow of He (20 sccm) to sweep the permeate gas from the outside of the tube to the “quick inlet” of an ABB-Extrel (Pittsburgh, PA) Questor IV quadrupole process mass spectrometer. This inlet is a “T” splitter fitting which directs a small portion of the gas (∼0.04%) through a 25-µm-i.d. silica capillary (Polymicro, Phoenix, AZ) directly into the electron ionization (100 eV electrons) ion source. The balance of the gas mixture is directed to a waste vent. Ion signals were generally acquired in the selected ion monitoring mode. All ions comprising >1% relative abundance in full-scan reference spectra (21) Jiea, M. D.; Chen, B.; Noble, R. D.; Falconer, J. L. J. Membr. Sci. 1994, 90, 1-10.
Analytical Chemistry, Vol. 71, No. 5, March 1, 1999
1017
Figure 1. Modified inlet for use with zeolite membranes.
of the compounds of interest were recorded using a Faraday cup detector for optimum precision. In all cases, triplicate pulse measurements were made to confirm the reproducibility of permeation behavior. Data shown in the figures represent one pulse but are representative of the three. Reference spectra of pure compounds (10 sccm) mixed with He carrier gas (10 sccm) were collected without the membrane. Concentration estimates were obtained by treating mixture spectra as linear combinations of reference mass spectra. Relative sensitivities were assessed using a mass spectrum from a reference mixture (also acquired without the membrane to avoid fractionation). Data workup employed least-squares routines incorporated into the Questor IV data system or the MATLAB software package (The Mathworks Inc., Natick, MA). RESULTS AND DISCUSSION A CO2/N2 binary mixture was chosen for preliminary tests because each component includes a “unique” peak in its mass spectrum (no overlap with peaks for the other component; i.e., mass-to-charge ratio (m/z) 14 for N2 and 12, 16, or 44 for CO2). Figure 2a shows two representative ion signals and the calculated permeate concentrations obtained when a single 2-min pulse of a 50/50 CO2/N2 (v/v) mixture was passed through a zeolite membrane. Gas was applied to the membrane at time t ) 0, but due to the dead volume and time required for analyte to permeate the membrane, ion intensities remained quite low for several seconds. The signal at m/z 44 exceeded 3 times the standard deviation of triplicate background measurements (σbkg) in ∼10 s and became the most intense analyte peak in the spectrum in ∼90 s. During the period of low intensity, estimated “concentrations” (based on the background signals) are not thought to be meaningful and oscillate widely. This situation recurred several seconds after withdrawal of the sample; the sample pulse ended at 2.0 min, and the signal at m/z 44 fell below 3 σbkg ∼90 s later. Figure 2b shows the selectivities calculated at various times during the pulse using eq 1. It can be seen that operation at steady state (indicated by the plateau at the middle of the pulse) would not enable resolution of the CO2/N2 mixture (Ssteady-state ≈ 1 for both components). However, at the onset of the sample pulse, N2 penetrated the membrane more quickly than CO2, resulting in significant enrichment (Smax, N2 > 3; Figure 2b). During the trailing edge of the pulse, N2 also cleared the membrane first, resulting in enrichment in CO2 (Smax, CO2 ≈ 4; Figure 2b). It should be noted that the enrichment in CO2 extended for almost 1 min before the 1018 Analytical Chemistry, Vol. 71, No. 5, March 1, 1999
Figure 2. (a) Time dependence of typical analyte ion intensities (m/z 28 and 44; AU, arbitrary units) and calculated concentrations in the permeate when a pulse of a 50/50 (v/v) carbon dioxide/nitrogen mixture is sampled through a zeolite membrane. (b) Time dependence of selectivities (eq 1) from the experiment of (a).
ion intensities became so small that concentrations became uncertain. The higher diffusivity22 of N2 evidently allows it to penetrate first, while the greater polarizability of CO2 promotes slightly stronger adsorption20 and thus enrichment in the trailing edge of the pulse. Size exclusion factors are unlikely to play a key role here because CO2 and N2 have similar kinetic diameters (Table 1), both smaller than the zeolite pore size (0.55 nm 15). As a result, the more strongly adsorbed CO2 does not “block” N2, so there is no steady-state partitioning once the pores are “saturated” (unlike the case in the H2/butane system). Enhanced diffusivity and reduced adsorption15 (in this case attributable to lower polarity; Table 1) can also account for the relatively rapid permeation of ethane in a 25/75 (v/v) binary mixture with propane (Figure 3). In this case, there is some selectivity even at steady state (Figure 3b, Ssteady-state,ethane ≈ 1.1), possibly due to the (relatively) large difference in diffusivities (Table 1). Enrichments away from steady state were significantly higher, with Smax, ethane ≈ 16 during the leading edge of the pulse and Smax, propane ≈ 7 during the trailing edge of the pulse. The “symmetry” of Figures 2 and 3 may be considered indicative of “normal” behavior; i.e., one component both penetrates and clears the membrane more quickly than the other. In these cases, higher diffusivity and lower adsorption of one component may both promote its faster permeation. Resulting enrichments of different components early and late in the pulse provide time windows when each component could be isolated or monitored with reduced interference from the other. In repetitive pulse experiments, there would be no need or advantage to waiting for attainment of steady state; operation with a faster (22) Bakker, W. J. W.; Broeke, L. J. P.; Kapteijn, F. Moulijn, J. A. AIChE J. 1997, 43, 2203-2214.
Figure 4. “Validation plot” of cis-2-butene concentrations determined from mass spectral deconvolution (monitoring m/z 27 and 39) versus actual concentration for binary mixtures of cis- and trans-2butene.
Figure 3. (a) Time dependence of typical analyte ion intensities (m/z 29 and 30; AU, arbitrary units) and calculated concentrations in the permeate when a pulse of a 25/75 (v/v) ethane/propane mixture is sampled through a zeolite membrane. (b) Time dependence of selectivities (eq 1) from the experiment of (a).
(automated) switching valve could shorten or eliminate the lowselectivity “plateau” in the middle of Figures 2b and 3b. Enrichments during the leading and trailing edges of fast, repetitive pulses will be reduced relative to those in single pulse experiments (due to partial saturation of the membrane), but significant and useful enrichments can persist. The extent of saturation will depend on the relation between the pulse and membrane response times, as well as flow rates, temperature, etc. Of course, selectivity was not a critical issue in the analysis of the two simple model systems described above, since the spectra of the components are sufficiently distinct to allow simultaneous analysis with no temporal resolution.9 A more challenging and interesting prospective application can be drawn from the analysis of the stereoisomers cis- and trans-2-butene. Earlier work9 found that simultaneous mass spectrometric analysis of these isomers was feasible. For example, Figure 4 is a “validation plot” of calculated (via spectral deconvolution; see Experimental section) versus actual concentration for butene mixtures sampled (without the membrane) over a fairly wide concentration range. The largest error is 9.8% relative, and the correlation coefficient (r2) is 0.9976. This represents a slight improvement in accuracy (r2 ) 0.9886 in ref 9) attributable to improvements in the algorithm used to select the ion signals monitored (the revised algorithm will be described in a later publication). While this performance may be adequate for many process applications, it is significantly less than what can be achieved with some other mixtures (e.g., r2 ) 0.999 993 for a 1-butene/isobutylene validation plot9). Thus there is room for improvement on quantitation by use of a membrane sensor, if the membrane can provide the needed extra dimension of differentiation for this system. A possible complication in applying the membrane to stereoisomers stems from the ability of zeolites
Figure 5. (a) Time dependence of a typical analyte ion intensity (m/z 39; AU, arbitrary units) and calculated concentrations in the permeate when a pulse of a 60/40 cis-/trans-2-butene mixture is sampled through a zeolite membrane. (b) Time dependence of selectivities (eq 1) from the experiment of (a).
to catalyze isomer interconversion. However, after acquiring single-component reference spectra without the membrane, each isomer was correctly identified when passed separately through the membrane (i.e., calculated concentration, 100%; data not shown), confirming that there was no interconversion by this membrane at the temperature employed. Figure 5a shows the response to a single pulse of a 60/40 (v/ v) mixture of cis- and trans-2-butene through the membrane. The data again present a typical ion signal (m/z 39) and the analyte concentrations calculated during the pulse. (In this case, complete overlap between analyte spectra precludes plotting separate ion intensities representative of each component.) Steady-state permeances for the two components were identical (Ssteady state ) 1, Figure 5b). However, in marked contrast to the behavior in Figures 2 and 3, the permeate was significantly enriched in cis2-butene both during pulse initiation and in the trailing edge of the pulse (Smax,cis > 10). In other words, cis-2-butene penetrated the membrane more quickly than trans-2-butene upon application of the pulse but was depleted more slowly when the analyte flow Analytical Chemistry, Vol. 71, No. 5, March 1, 1999
1019
was interrupted. This “abnormal” permeation behavior must result from offsetting effects. The more polar cis-2-butene (Table 1) should be more strongly adsorbed. Slow release would then promote the observed enrichment in this isomer late in the pulse. In the cases described earlier, low diffusivity and/or preferential adsorption delayed initial component breakthrough, accounting for the depletion of CO2 and propane early in Figures 2 and 3, respectively. This is inconsistent with the observed enrichment in the cis isomer during the leading edge of the pulse in Figure 5. Unlike CO2 and propane, cis-2-butene is known to have higher diffusivity23 than its more weakly adsorbed cosolute (compare the diffusivities of CO2/N2, propane/ethane, and cis- and trans-2-butene in Table 1). Evidently, the difference in adsorption of the stereoisomers is so small that the adsorptive depletion of cis-2butene during pulse initiation is more than compensated by its enhanced diffusivity, accounting for the initial enrichment in the cis isomer. Apparently, the difference in diffusivity is the more important factor is determining the order of breakthrough, except in cases where the polarity difference is so large that one component severely impedes the permeation of the other (the H2/butane case described above). This explanation is tentative; the mechanism(s) involved need further study. CONCLUSIONS These initial studies of time-dependent permeation behavior prove that MIMS with zeolite membranes can provide significant time-dependent component enrichment, even in cases where there (23) Shah, D. B.; Chokchai-acha, S.; Hayhurst, D. T. J. Chem. Soc., Faraday Trans. 1993, 89, 3161-3167. (24) CRC, Handbook of Chemistry and Physics, 51st ed.; Cleveland, OH, 1971.
1020 Analytical Chemistry, Vol. 71, No. 5, March 1, 1999
is no detectable difference in the steady-state permeances. The mixtures tested show a range of permeation behavior depending on the balance between (or among) selective sorption and intracrystalline transport mechanisms. Future studies will assess the potential utility of these time-dependent enrichments for dynamic MIMS, with an aim of improving the accuracy, precision, and/or sensitivity of quantitation of isomers. Assessment of the effect of temperature will be important; although little temperature dependence was noted in preliminary studies analogous to Figure 5, with membrane temperatures varying between 20 and 80 °C (data not shown), the selectivity of these membranes often does show strong temperature dependence.15 Synergistic effects will also have to be addressed for continuously pulsed systems, as they may differ from those observed for single pulses in the present study. ACKNOWLEDGMENT Support for this work was provided in part by NSF “Tie” grants (EEC-9528067 and EEC-9528068) to the UT Measurement and Control Engineering Center and the CU Center for Separations Using Thin Films (NSF-supported Industry/University Cooperative Research Centers). Instrumentation gifts and loans from ABBExtrel, Lubrizol, Rosemount/Brooks, and Union Carbide are gratefully acknowledged.
Received for review December 2, 1998. AC980991N
September
3,
1998.
Accepted
