Enhancement of Extraction Efficiency and Reduction of Boundary

Anal. Chem. 1999, 71, 4407-4412

Enhancement of Extraction Efficiency and Reduction of Boundary Layer Effects in Pulse Introduction Membrane Extraction Xuemei Guo and Somenath Mitra*

Department of Chemical Engineering, Chemistry and Environmental Science, New Jersey Institute of Technology, Newark, New Jersey, 07102

Membrane separation has emerged as an attractive alternative for interfacing an extraction step directly to a gas chromatograph or to a mass spectrometer. In pulse introduction (or flow injection type) membrane extraction, a sample pulse is introduced onto an eluent stream that transports it onto the membrane. Since a fixed sample volume is injected, the detector response is directly proportional to the extraction efficiency. This in turn depends on membrane module design, flow conditions, etc. Also, when water contacts a membrane, a static boundary layer is formed at the membrane surface that serves as an additional diffusional barrier to the permeation process. Consequently, permeation slows down, which lowers the speed of analysis. In this paper, methods of increasing the extraction efficiency and decreasing boundary layer effects are presented. The goal is to have higher sensitivity at a shorter analysis time. A stream of nitrogen is introduced into the membrane after sample elution to eliminate the aqueous boundary layer. This technique is found to be effective not only for faster analysis, but also for increasing extraction efficiency. Isolation of analytes from an aqueous matrix is usually the first step in the analysis of organic compounds in water. Membrane separation has emerged as an attractive alternative for interfacing an extraction step directly to a gas chromatograph1-5 or to a mass spectrometer.6-10 The on-line extraction can be carried out continuously, thus facilitating real-time monitoring. Owning to * Corresponding author. E-mail: [email protected]. (1) Xu, Y.; Mitra, S. J. Chromatogr. A 1994, 688, 171-180. (2) Mitra, S.; Zhu, N.; Zhang, X.; Kebbekus, B. J. Chromatogr. A 1996, 736, 165-173. (3) Mitra, S.; Zhong, L.; Zhu, N.; Guo, X. J. Microcolumn Sep. 1996, 8 (1), 21-27. (4) Mitra, S.; Guo, X. Anal. Lett. 1998, 31 (2), 367-379. (5) Guo, X.; Mitra, S. J. Chromatogr. A, 1998, 826, 39. (6) Slivon, L. E.; Bauer, M. R.; Ho, J. S.; Budde, W. L. Anal. Chem. 1991, 63, (13), 1335-1340. (7) Kotiaho, T.; Lauritsen, F. R.; Choudhury, T. K.; Cooks, R. G. Anal. Chem. 1991, 63 (18), 875A-883A. (8) Savickas, P. J.; Lapack, M. A.; Tou, J. C. Anal. Chem. 1989, 61, 23322336. (9) Tsai, G.-J.; Austin, G. D.; Syu, M. J.; Tsao, G. Anal. Chem. 1991, 63, 24602465. (10) Virkki, V. T.; Ketola, R. A.; Ojala, M.; Kotiaho, T.; Kompa, V.; Grove, A.; Facchetti, S. Anal. Chem. 1995, 67, 1421-1425. 10.1021/ac990352s CCC: $18.00 Published on Web 08/20/1999

© 1999 American Chemical Society

their large active surface area per unit volume, hollow fibers have been found to be the membrane geometry of choice. The sample is either introduced continuously1-3 or as a pulse (flow injection type)4,5,9,10 during membrane extraction. In the former approach, the water sample continuously contacts the feed side of the membrane and the organics are removed continuously on the permeate side. Here, the measurement is made after the permeation has reached steady state, which can take a fairly long time.11 No measurements can be made during the transition period when the system is approaching equilibration. Another limitation of continuous extraction is that the analysis of small-volume individual samples is not possible: some sample has to pass through the membrane to reach equilibrium. The pulsed, or flow injection, approach has been used widely in membrane introduction mass spectrometry,9,10 where the organics are directly introduced into the vacuum in the mass spectrometer. A similar approach has been reported recently for interfacing membrane extraction with GC as well.4,5,11 Referred to as the pulse introduction membrane extraction, here a sample pulse is introduced onto an eluent stream that transports it onto the membrane. A stream of nitrogen is introduced into the membrane after sample elution3,4 to eliminate the aqueous boundary layer. Consequently, this is somewhat different from conventional flow injection. The permeated organics are pneumatically transported by a flow of inert gas, concentrated, and injected into a GC. This approach is also applicable for GC/MS and MS interfaces. When water contacts a membrane, a static boundary layer is formed at the membrane surface due to the poor mixing of the aqueous phase with the membrane. The boundary layer serves as an additional diffusional barrier to the permeation process. The degree of mixing depends on the Reynolds number (Re), which is a function of the flow conditions:

Re ) Fνd/µ

(1)

where F is density, ν is velocity, µ is viscosity and d is the diameter of the tubing. The higher the Re, the better is the mixing and the less are the boundary layer effects. A Re of 20 000-30 000 is required to reach turbulent conditions, which eliminate the (11) Guo, X.; Mitra, S. Theoretical Analysis of Non-steady State, Pulse Introduction Membrane Extraction with a Sorbent Trap Interface for Gas Chromatographic Detection. Anal. Chem., submitted.

Analytical Chemistry, Vol. 71, No. 19, October 1, 1999 4407

boundary layer. Such high Re requires a high flow rate, where the pressure drop is enough to rupture the membrane. High flow rates also reduce extraction efficiency, thus lowering sensitivity. In a typical analytical-type membrane extraction Re is less than 300, and a thick boundary layer is encountered. Since the organics have high solubility in the membrane material, a concentration depletion zone is formed in the boundary layer that impedes the mass transfer from water to membrane. The overall mass transfer resistance is comprised of the individual resistance in series such that

1/K ) 1/kwb + 1/km + 1/kgb

(2)

where the K is the overall mass transfer coefficient and kwb, km, and kgb are the mass transfer coefficients for the aqueous boundary layer, the membrane, and the gaseous boundary layer, respectively. An overall diffusivity can be calculated11 considering the three resistances in series. The mechanism of membrane extraction from gaseous and aqueous samples is quite similar.3 Extraction rates of organic molecules from a gas sample have been reported to be much faster than that from an aqueous sample.3 This demonstrates that the aqueous boundary layer is the major impediment to permeation in the extraction of aqueous samples. The relative importance of the boundary layer resistance depends on the hydrodynamic conditions, the membrane thickness as well as the nature of analytes. For analytes that have high diffusivity and high partition coefficient in the membrane, the boundary layer is the major resistance to permeation when thin membranes are used.9,12,13 For example, it has been reported12 that the liquid phase accounted for 90% of the total mass transfer resistance for toluene permeation using polyester-block-polyamide membrane when the membrane thickness was less than 0.4 mm. It has also been reported12 that the permeation of compounds having high Henry’s constant through a hydrophobic membrane is limited by the boundary layer resistance. Another issue that is of considerable interest is the extraction efficiency. Since a fixed sample volume is introduced in the pulse introduction approach, a higher extraction efficiency results in more sample going into the detector and consequently higher detector response. Injection volumes cannot be increased indefinitely to enhance sensitivity, because the permeation will take a longer time and reduce the speed of analysis. Thus, it is important to optimize conditions that will allow high sensitivity and fast instrument response. Methods of increasing the extraction efficiency and decreasing the boundary layer effects are presented in this paper. The goal is to have higher sensitivity at a shorter analysis time. Preliminary published results4,5 have shown that the boundary layer can be eliminated by introducing a gas stream to cleanse (or purge) the membrane. The effects of gas purge on the extraction process were studied in detail. EXPERIMENTAL SECTION The schematic diagram of the experimental system is shown in Figure 1. The aqueous sample is to be introduced as a pulse (12) Raghunath, B.; Hwang, S. T. J. Membr. Sci. 1992, 65, 147-161. (13) Psaume, R.; Aptel, Ph.; Aurelle, Y.; Mora, J. C.; Bersillon, J. L. J. Membr. Sci. 1988, 36, 373-384.

4408 Analytical Chemistry, Vol. 71, No. 19, October 1, 1999

Figure 1. Schematic diagram of the pulse introduction membrane extraction system.

into the membrane. An eluent stream is used for transporting the sample to the membrane module. A countercurrent gas stream strips the analytes that permeate through. The GC interface is accomplished using a microsorbent trap (referred to as microtrap) which concentrates the organics before injecting into the GC. The sample size used was anywhere from 100 µL to 10 mL depending upon analysis requisite. The injection was accomplished by a pneumatically controlled six-port valve (Valco Instruments Co. Inc., Houston, TX). The eluent was HPLC-grade high-purity water. An HPLC pump was used as the eluent pump. A threeway valve was used for introducing a nitrogen stream into the membrane module. Typical operation consisted of injecting a sample onto the eluent steam. A few minutes were allowed for the permeation to complete. The permeated analytes were pneumatically transported by a stripping gas to the microtrap. The analytes were concentrated and then desorbed into the GC for analysis. For each injection, nitrogen could be introduced into the system prior to and/or after sample elution to clean the membrane. During continuous on-line monitoring,5 the aqueous stream continuously flowed through the sample loop of the valve and periodically injections were made into the membrane module, corresponding to each injection, a chromatogram was obtained. The membrane module was made of composite hollow fiber membranes of dimensions 0.260 mm o.d. × 0.206 mm i.d. (Applied Membrane Technology, Minnetonka, MN). The length was anywhere from 10 to 70 cm. The composite structure comprised of 1 µm thick homogeneous siloxane film as the active layer supported on a layer of microporous polypropylene. The membrane module was constructed by inserting the fibers into a 1/8 in. o.d. stainless tubing. At each end of the tubing, a “T” unit (Components & Controls Inc., Carlstadt, NJ) was used to connect the inlet and outlet for both the nitrogen and the aqueous stream. The connection points of membrane and “T” units were sealed with epoxy to separate nitrogen and aqueous phase. The membrane module was either straight or spiraled. Spiraling the straight module into a circular form of 6-10 cm diameter made the latter. A segment of one spiral is shown in Figure 1. The microtrap was a small-diameter silica-lined tube packed with a small amount of adsorbent. It had low thermal mass and could be heated and cooled rapidly. When nitrogen-carrying organics flowed through the microtrap, the organics were trapped

and concentrated. An electrical current resistively heated the microtrap, and the desorption pulse served as an injection for GC analysis. The details of the microtrap and its working principle have been presented else where.14-16 A 15-cm-long, 0.53 mm i.d. silica-lined tubing (Restek Corp., Bellefonte, PA) packed with Carbotrap C (Supelco, Supelco Park, PA) served as the microtrap. A 7-10 A current was supplied from a 40 V ac power source to heat the microtrap. The duration and the interval between the heat pulses were controlled using a microprocessor-based controller fabricated in-house. A HP 5890 Series II GC (Hewlett-Packard Co., Avondale, PA) equipped with a flame ion detector and a 30-m-long, 0.53 mm o.d. × 0.21 mm i.d. SE-54 megabore column with 2.4 µm thick stationary phase was used for GC separation. HP Chemstation 3365 software was used for data acquisition and analysis. RESULTS AND DISCUSSION By studying the influence of flow rate on membrane permeation, Psaume et al.13 observed that the pervaporation rate of organics is limited by the solute-depleted boundary layer at the membrane-liquid interface. Thereafter, based on mass balance across the membrane, the solute flux through the membrane wall was expressed as

Ji ) ((QeF)/A)(w′i,e - w′i,s)

(3)

where Qe is the water flow rate, F is density, A is the internal surface area, and w′i,e and w′i,s are weight fraction of analyte at inlet and outlet, respectively. Assuming steady state, stagnant polarization layer and negligible convective flow, the flux was

Ji,z ) kiF(w′i,z - w′i,z,m )

(4)

where Di is diffusivity of solute i in water, the mass transfer coefficient ki ) Di /σ (σ is the boundary layer thickness) and the w′i,z and w′i,z,m are the concentration on either side of the boundary layer. The mass balance for a hollow fiber is

Q′eF dw′i,z ) -2πRJi,z dz

(5)

where R is the radius of the fiber. Assuming w′i,z,m to be zero, and the integration of eq 5 with boundary conditions

z ) 0,

w′i,0 ) w′i,e

z ) L,

w′i,L ) w′i,s

where L is the fiber length, results in

ln(w′i,e/w′i,s) ) kiA/Q′e

(6)

The ki was estimated using the Leveque correlation:16 (14) Mitra, S.; Chen, Y. J. Chromatogr. 1993, 648, 415-421. (15) Mitra, S.; Lai, A. J. Chromatogr. Sci. 1995, 33, 285. (16) In Handbook of Industrial Membrane Technology; Porter, M. C., Ed.; Noyes Publications: Park Ridge, NJ, 1990; pp 174-175.

(ki2R/Di) ) 1.62 (d2ν/LDi)

(7)

where d is fiber diameter and ν is velocity. The extraction efficiency, EE, was thus derived as

EE ) (w′i,e - w′i,s)/w′i,e ) 1 - exp(-6.48 (Di2/3L2/3/d4/3ν2/3)) (8) A somewhat similar expression has been reported elsewhere17 for a porous membrane and assuming that the mass transfer is limited by the liquid phase. The mass transfer mechanism is the same whether steady state is reached or not. Thus, the above equation may be adopted for non-steady-state permeation in the pulse introduction approach by introducing a response factor that accounts for the departure from steady state. In fact, our experiments showed that extraction efficiency in the pulse introduction approach could be within 2% of the equilibrium value when the eluent flow rates were very low. Therefore the system response in pulse introduction can be given as

M ) KCV(EE) ) KCV[1 - exp(-6.48 (Di2/3L2/3/d4/3ν2/3))] (9) where M is the system response, K is the analyte-dependent response factor, C is the analyte concentration in water sample, and V is the injection volume. To see if the above equation was applicable in pulse introduction membrane extraction, the system response was plotted as a function of {1 - exp[-6.48(Da2/3L2/3/d4/3ν2/3)]} as shown in Figure 2. Diffusivities of benzene and toluene in water were calculated from the Wilke-Chang equation,18 and the molar volumes of benzene and toluene were obtained from literature.19 The experiments were done with an 8-mL sample containing 67.5 ppb benzene and 75 ppb toluene. Water velocity was varied while the fiber diameter and its length were fixed. A linear fit with r2 of 0.993 and 0.996 was obtained for benzene and toluene, respectively. This demonstrated the applicability of the above equation to pulse introduction membrane extraction. Since eq 9 was obtained by assuming that the permeation across the boundary layer is the rate-determining step, this fit further confirmed the importance of the boundary layer. Flow rate is an important parameter according to the above equation. As flow rate increased, the residence time decreased and there was less time for permeation.11 Consequently a large fraction of the analytes went through unextracted. In continuous membrane extraction, it has been reported1 that the system response increased with flow rate increase because more sample was brought into the membrane. Eventually, the response stabilized because extraction efficiency decreased, resulting in no additional permeation flux. During pulse introduction, the sample amount is fixed, so a reduction in extraction efficiency reduces (17) Pratt, K. F.; Pawliszyn, J. Anal. Chem. 1992, 64, 2101-2106. (18) Wilke, C. R. In The properties of gases and liquids; 3rd ed.; Reid, R. C., Prausnitz, J. M., Sherwood, T. K., Eds.; McGraw-Hill: New York, 1977. (19) Lee, B. I. In The properties of gases and liquids, Reid, R. C., Prausnitz, J. M., Sherwood, T. K., Eds.; McGraw-Hill: New York, 1977.

Analytical Chemistry, Vol. 71, No. 19, October 1, 1999

4409

Figure 2. System response as a function of [1 - exp(k/ν2/3)]. The sample contained 67.5 ppb benzene and 75 ppb toluene, and injection volume was 8 mL.

Figure 3. Response as a function of velocity. The sample contained 67.5 ppb benzene and 75 ppb toluene, and injection volume was 8 mL.

Figure 4. Permeation profiles as a function of injection volume; the sample was 48.4 ppb toluene and the eluent flow rate was 0.8 mL/ min.

system response. Figure 3 is a plot of system response vs flow rate, and it is seen that the response decreased as the eluent flow rate increased. Sample volume also affects the system response. At the same flow rate, system response is proportional to sample size.5 This has been reported before11 and can also be seen from Figure 4, where the area under the permeation profile (which is proportional to detector response) increases with injection volume. A large sample contains a larger mass of analytes resulting in a larger detector response. However, a large sample also stays longer in the membrane increasing the lag time, and this is seen from Figure 4. The lag time in pulse introduction membrane extraction 4410 Analytical Chemistry, Vol. 71, No. 19, October 1, 1999

Figure 5. System response as a function of membrane length. A 0.7 mL sample containing 30 ppb benzene and 40 ppb chlorobenzene at an eluent flow rate of 1 mL/min was used.

is defined as the time interval between 10% of the maximum response in the ascending part and the descending part in the response profile as shown in Figure 4. It is a measure of the duration over which the permeation takes place. Lag time is an important parameter, because it is directly related to the speed of analysis. The next sample can be injected only after the permeation from previous sample has reached completion. To enhance sensitivity, one may consider increasing the sample size or decreasing the flow rate. A larger volume clearly increases lag time. A slow flow rate also reduces Re and results in slower mass transfer due to the formation of a thick boundary layer. The combination of longer residence time and slower mass transfer results in longer lag time.11 Thus flow rate also provides a tradeoff between sensitivity and lag time. Membrane Module Design. From a practical point of view, the lag time is the limiting factor that prevents the use of very large sample volume or very low flow rates. To enhance the performance of pulse introduction membrane extraction, alternate approaches need to be studied to obtain high extraction efficiency and short lag time. Increasing the membrane active surface area can increase extraction efficiency. Multiple hollow fibers packed into a module have been used to increase active surface area and consequently system response.3 According to eq 9, extraction efficiency increases with the membrane length as long as it is shorter than that required for exhaustive, quantitative extraction. System response as a function of membrane length (other conditions remaining constant) is presented in Figure 5 for a 0.7 mL sample containing 30 ppb benzene and 40 ppb chlorobenzene. It is also seen that increasing membrane length is an effective way of increasing extraction efficiency and thus sensitivity. When small sample volume is used, short lag time is obtained. On a 10-cm membrane module, and for a 0.7-mL sample containing 40 ppb each benzene and toluene, lag times were 2.6 and 3.3 min for the former and the latter. The lag times increased to 2.8 min for benzene and to 4.1 min for toluene on a 40-cm fiber module. Therefore, the fiber length did not significantly affect lag time. A comparison of sensitivity for 10- and 40-cm-long membrane fiber is presented in Figure 6. The figure shows that longer the membrane fiber, the higher is the sensitivity due to its higher extraction efficiency. The ratio of the slopes of calibration curves from the 40-cm module to that from the 10 cm was 2.02. In general, a long membrane module with multiple fibers appears to be more

Figure 6. Comparison of sensitivity for membrane modules of different lengths. Injection volume was a 0.7 mL sample at an eluent flow rate of 1 mL/min.

Figure 7. Effect of nitrogen purge on system response and lag time. The nitrogen was turned on after a predetermined period of water elution. The experiments were done with 37.5 ppb toluene, and injection volume was 3 mL at eluent flow rate of 1 mL/min.

practical than reducing eluent flow rate and/or increasing sample size. For example, according to eq 8, the extraction efficiency of benzene would be 90% at an eluent flow rate of 1 mL/min and using a 70-cm-long module comprising of seven hollow fibers. The same could be achieved with a 19-cm-long module with 12 fibers and at an eluent flow rate of 0.5 mL/min. However, the latter would have significantly longer lag time due to the lower flow rates. Previous studies have shown that the boundary layer resistance can be reduced by increasing Re.12,13 However, as mentioned before, increased flow rate reduces sample residence time and consequently the system response. Having a flow pattern that changes direction can bring about more turbulence. This would reduce the thickness of the boundary layer and introduce more mixing of the sample on the membrane surface. A spiral membrane module was fabricated. The spiral configuration allowed the sample to flow in a circular path inside the membrane, thus disrupting the boundary layer. This geometry also provided a practical way of making compact, yet long membrane modules. For example, the 70-cm-long module mentioned above was made into a spiral of 8 cm diameter. A comparison of sensitivity between the straight and the spiral module of the same length is also presented in Figure 6. The spiral configuration consistently generated higher response than the straight module. Elimination of Boundary Layer by Inert Gas Purge. In pulse introduction membrane extraction, either a gas or a liquid can be used as the eluent. The main difference between gases and liquids is the transport properties, i.e., diffusivity, viscosity, and density. Membrane extraction from gaseous and aqueous mediums follow similar mechanisms, and Re can be used to normalize the difference between them.3 As diffusion coefficients are significantly higher in gases, the boundary layer resistance is practically negligible compared to the aqueous layer on the membrane. Consequently, the lag time in gas extraction is significantly shorter compared to that in water. However, the problem with gas elution of the water sample is the high pressure required for sample elution resulting in a high flow rate which reduces the residence time and consequently the extraction efficiency. Short lag time is required to eliminate memory effect so that analysis can be completed quickly. It is also an important issue in continuous on-line analysis. Here, a sample pulse is introduced into the membrane and an eluent serves as a carrier fluid. The axial mixing of the sample with the eluent broadens the input

pulse, and so does slow permeation through the boundary layer and the membrane.11 The tailing part of the sample input profile makes little contribution to the system response but significantly increases the lag time. The response profile for toluene is presented in Figure 7, where water was used for eluting the sample for a predetermined period of time. Then a flow of inert gas, such as nitrogen, was used to purge the membrane to eliminate the tailing part of sample input and the boundary layer. A three-way valve shown in Figure 1 was used to introduce the nitrogen purge. The purge was turned on depending upon the performance requisite. The effect of purge interval (interval between sample input and nitrogen purge) is seen in Figure 7. Lag times as a function of purge interval are also presented for benzene and chlorobenzene in Table 1. If no nitrogen purge was used for toluene, the response continued for nearly 25 min. If the purge was turned on after 1 min, then the lag time was only 5 min. However, some sensitivity was lost because much of the sample passes unextracted. In general, longer the purge interval, the closer is the response to the maximum possible sensitivity, but the longer is the lag time. If the purge is turned on where the profile tailing is seen, relatively less sensitivity is lost but lag time is significantly reduced. For example, it can be seen from Table 1 that when the purge was turned on 6 min after sample elution, the lag time for toluene reduced by nearly 50%, with only 16.8% loss in sensitivity. In general, the tradeoff between lag time and sensitivity depends on what point in time the nitrogen purge is turned on. For a highconcentration sample, early nitrogen purge is recommended, whereas for a low-concentration sample, the purge should be initiated later. Membrane Precleaning with Inert Gas Purge. An advantage of the nitrogen purge is that it cleans the membrane by eliminating the aqueous boundary layer that is formed on its surface. If the membrane is precleaned, and the liquid eluent is used only to elute the sample, then no boundary layer is encountered initially and higher permeation rate is obtained. Eventually, as the sample elutes, the boundary layer is reestablished. Consequently, higher overall extraction efficiency is expected when precleaned membrane is used. Once again the N2 purge is turned on at a predetermined interval to eliminate the tailing part of the permeation profile. This also serves as the preclean for the next extraction-analysis. A comparison between precleaned and water-eluted membrane is presented in Figure 8. Analytical Chemistry, Vol. 71, No. 19, October 1, 1999

4411

Table 1. Reduction in Lag Time and Loss of Sensitivity during Nitrogen-Purged Elution

water elution 9 min water elution followed by nitrogen purging 6 min water elution followed by nitrogen purging 5 min water elution followed by nitrogen purging 4 min water elution followed by nitrogen purging 3.6 min water elution followed by nitrogen purging 2 min water elution followed by nitrogen purging 1 min water elution followed by nitrogen purging nitrogen elution only

lag time for benzene (min)

lag time for toluene (min)

lag time for chlorobenzene (min)

reduction in system response for toluene (%)

23 14 12.5 11 9.5 9.5 8 6.5 3.5

24.5 14 12.5 9.5 9.5 8 6.5 5 2

24.5 15.5 12.5 9.5 8 8 8 6.5 5

0 19.8 16.8 32.9 46.2 56.3 72.1 84.8 96.7

probably never formed. However, when gas purging is not used, a fully developed boundary layer is always encountered. CONCLUSIONS

Figure 8. Permeation profiles for water-eluted sample and when membrane precleaning was done. Postelution N2 purge was turned on 6 min after injection for the later. A 2 mL sample containing 27.3 ppb toluene at an eluent flow rate of 0.8 mL/min was used.

It is clear that, along with a shorter lag time, the response was significantly higher in the precleaned membrane. In this experiment with toluene, the response was 33% higher with the precleaned membrane. Since the contact period with the membrane was only 6 min, a fully developed boundary layer was

4412 Analytical Chemistry, Vol. 71, No. 19, October 1, 1999

Pulse introduction membrane extraction was studied in terms of lag time and sensitivity. Sensitivity increased with extraction efficiency, which could be increased by increasing the length of the membrane or by increasing residence time. Injecting a larger volume is another way of enhancing sensitivity, but it also results in a longer lag time. A spiral membrane module showed only minor enhancement in sensitivity. Purging the membrane with an inert gas prior to membrane extraction cleansed the membrane surface, resulting in higher extraction efficiency. Purging the membrane after sample elution was an effective way of reducing the lag time without sacrificing too much sensitivity.

Received April 6, 1999. Accepted June 29, 1999. AC990352S