Effects of Injection Schemes and Column Geometry on the

Christa M. Snyder , William R. Alley , Jr. , Margit I. Campos , Martin Svoboda , John A. Goetz , Jaqueline A. Vasseur , Stephen C. Jacobson , and Milo...
0 downloads 0 Views 2MB Size
Anal. Chem. 1994,66,1107-1113

Effects of Injection Schemes and Column Geometry on the Performance of Microchip Electrophoresis Devices Stephen C. JacOb80n,t Roland HergenrWer,t Lance B. Koutny,t R. J. Warmack,* and J. Mlchael Ramsey'gt Chemical and Analyticai Sciences Division and Health Sciences Research Division, Oak Ridge National Laboratoty, Oak Rme, Tennessee 3783 1-6 142

A glass microchip column was fabricated for free-solution electrophoresis. The channels were wet chemically etched on a substrate using standard photolithographic techniques and were sealed using a direct bonding technique. Two methods of sample introduction are evaluated employing a cross-type channel geometry. With the preferred sample loading method, the volume of the sample plug can be accurately controlled and is time independent, enabling a constant volume to be injected. The injection can be controlled electronically and does not induce any electrophoretic mobility based bias. Separations were performed on a compact chip with a serpentine column geometry that has a 165-mm separation channel in an area of less than 10 mm X 10 mm. Band-broadening effects in a serpentine column pattern were studied. The microfabricationof analyticalinstrumentation provides an interesting alternative for chemical sensing. A common approach to chemical sensor design is to rely on some general physical property observable for quantification, such as an electrode current or acoustic wave propagation, and a chemically selective coating for chemical specificity.' Such approaches have had limited success with complex mixtures of molecular species. The design of high-performance chemically selective coatings is presently a difficult task. Sensor concepts based on conventional laboratory approaches to chemical measurement problems benefit from a large knowledge base and thus may be more amendable to a priori designs. In addition, microfabricated chemical instruments are likely to be more versatile,allowing a single deviceto address differing sensing problems. Microinstrumentation could well derive benefits similar to microelectronics, Le., low cost, compact size, high speed,2 and parallel analyses. One avenue for the miniaturization of chemical instrumentation is to micromachine a monolithicdevice using standard lithographic,etching, and deposition techniques, Le., a mi~rochip.~ Chemical separation techniques appear to be a promising area for microinstrumentation. Micromachined columns have been fabricated for gas chromatography,4liquid chromatogr a p h ~and , ~ capillary electrophoresisk9 and evaluated in the cases of gas chromatography and capillary electrophoresis. 7 Chemical and Analytical Sciences Division. 3 Health Sciences Research Division.

(1) Murray, R. W.; Darsy, R. E.; Heineman, W. R.; Janata, J.;Seitz, W. R., Eds. Chemical Sensors and Microinstrumentation; ACS Sympcsium Serics 403;

American Chemical Society, Washington, DC, 1989. (2) Jacobson, S. C.; Hergenrbder, R.; Koutny, L. B.;Ramsey, J. M.A w l . Chem., following paper in this issue. (3) Manz, A.; Fettinger, J. C.; Verpoorte, E.; LUdi, H.; Widmer, H. M.; Harrison, D. J. Trends Anal. Chem. 1991, 10, 144. (4) Terry,S.C.; Jerman, J. H.; Angell, J. B. IEEE Trans. Electron Dcuices 1979, 26, 1880.

0003-2700/94/0366- 1107$04.50/0 0 1994 American Chemical Society

Capillary electroph~resis~O-'~ as an analytical technique appears the most promising because of the experimental simplicity and the dependence of the resolution on the electric field strength, not the length of the column. Three aspects of microchip electrophoresis are addressed in this paper: substrate/cover slip bonding, sampleintroduction,and column geometry. Previ0usly,6-~in order to join the substrate with the etched channels and a cover plate, the two pieces had to be melted together. Another avenue is direct bonding, which allows the surfaces of the substrate and the cover slip to be bonded below their melting points. Although not necessary for glass, higher melting point substrates, e.g., quartz, can be bonded much more easily with such a technique. Two drawbackswith conventionalcapillary electrophoresis is that the sample introduction method of exchanging sample and buffer reservoirs is time consuming and lacks precision, and with an electrokinetic injection scheme, ions with greater mobilities are disproportionately introduced in larger quantities. The mechanical exchange of capillaries among various reservoirs is of course not possible with a monolithic micromachined device and alternative injection schemes must be used. The concept of using electroosmotic flow to perform valveless injections on a microchip electrophoresisdevice was previously described$*' This injector was essentially a threeport device that performed electrokinetic injections without mechanical exchange of capillaries but still with mobilitybased injection bias. An alternative approach for injection on a microchip electrophoresis device was investigated that eliminates mobility-based injection bias. Two different modes of injector operation were studied. The preferred method provides an injection volume that becomes time independent, leading to greater injection precision. We have also studied the impact of a serpentine separation channel geometry on separation efficiency. Band-broadening phenomena must be understood to optimize the trade-offs between compact device designs and separation performance. A straight channel design would minimize contributions from (5) Manz, A.; Miyahara, Y.; Miura, J.; Watanabe, Y.; Miyagi, H.; Sato, K. Sew. Actuators 1990, B l , 249. (6) Manz. A.; Harrison, J.; Verpoorte, E. M. J.; Fcttinger, J. C.; Paulus, A.; LUdi, H.; Widmer, H. M. J . Chromatogr. 1992, 593, 253.

(7) Harrison, D. J.; Manz, A.; Fan, Z.; LUdi, H.; Widmer, H. M. A w l . Chem. 1992, 64, 1926. (8) Seiler, K.; Harrison, D. J.; Manz, A. A w l . Chem. 1993, 65, 1481. (9) Harrison, D. J.; Fluri, K.; Seiler, K.; Fan, Z.; Effenhauser, C. S.; Manz, A. Science 1993, 261, 895. (IO) Hjerten, S. Chromatogr. Reo.1967, 9, 122. (1 1) Virtanen, R. Acta Polytech. S c a d . , Appl. Phys. Ser. 1974, No.123, 1. (12) Mikkers, F.; Evcraerts, F.; Vcrheggcn, T.J . Chromatogr. 1979, 169, 11. ( 1 3 ) Jorgenson, J . W.; Lukacs, K. D. Anal. Chem. 1981, 53, 1298.

Analyflcai Chemktry. Voi. 66, No. 7, April 1, 1994 I107

cover slip

Figure 2. Scanning electron microscope image of a cross section of the channel. The bar at the bottom represents a length of 11.1 pm. substrate

Figure 1. Schematicof serpentine channelgeometryfor the microchip with the large circle representingthe cover slip and the smaller circles the reservoirs.

electric field and geometriceffects whereas alternative layouts such as the serpentine pattern used here will minimize device dimensions. EXPERIMENTAL SECTION The column on the microchip was constructed on a 50 mm X 25 mm glass microscope slide (Corning, Inc. No. 2947) and covered by a 22-mm circular coverslip. The generation of the channels involves standard photolithographic procedures followed by chemical wet etching.14 The column image was transferred onto the slide with a positive photoresist (Shipley 181 1) and an e-beam written chrome mask (Institute of Advanced Manufacturing Sciences, Inc.), and the channel was chemically wet etched using a HF/NH4F solution. The cover slip was then bonded to the slide using a direct bonding technique whereby the slide and cover slip surfaces were first hydrolyzed in a dilute NH40H/H202 solution and then joined. The chip was annealed at 500 OC in order to ensure proper adhesion of the cover slip to the slide. Following the bonding, cylindrical plastic reservoirs were affixed to the chip using epoxy. Figure 1 shows a schematic of the column design. The channels are designated by the name assigned in Figure 1 from the terminus up to the intersection of the four channels, and each channel has an accompanying reservoir mounted above the channel at the edge of the cover slip. The enclosed column length is 9.4 mm from the analyte reservoir to the injection cross, 9.4 mm from the analyte waste reservoir to the injection cross, 6.4 mm from the buffer reservoir to the injectionkross, and 171 mm from the injection cross to waste reservoir. The radius of all of the turns is 0.16 mm. The cross section of the channel is seen in the scanning electron (14) KO,W. H.; Suminto, J. T. In Sensors;Gopel, W., Hasse, J., Zennel, J. N. Eds.; VCH: Weinhein, Germany, 1989; Vol. 1, pp 107-168.

1108

Analytical Chemistry, Vol. 66, No. 7, April I, 1994

microscope image in Figure 2. The dimensions of the channel are 10 pm deep, 90 pm wide at the top, and 70 pm wide at the bottom. Two experimental apparatuses were used to analyze chip dynamics via analyte fluorescence. A charge coupled device (CCD) camera monitored designated areas of the chip and a photomultiplier tube (PMT) tracked single-point events. The CCD (Princeton Instruments, Inc. TE/CCD-5 12TKM) camera was mounted on a stereomicroscope (Nikon SMZU), and the chip was illuminated using an argon ion laser (514.5 nm, Coherent Innova 90) operating at 3 W with the beam expanded to -20 mm. The point detection scheme employed a helium/neon laser (543 nm, PMS Electro-Optics LHGP-005 1) with an electrometer (Keithley 617) to monitor response of the PMT (Oriel 77340). The power supplies (Spellman CZEl000R) for the electrophoresis were operated between +0.3 and +4.4 kV relative to ground. The chip was operated under a “sample loading” mode and a “separation” mode. In the sample loading mode, two types of sample introduction were investigated. For a “floating” sampleloading, a potential was applied to the analyte reservoir with the analyte waste reservoir grounded and with the buffer and the waste reservoirs floating. For a “pinched” sample loading, potentials were applied to the analyte, the buffer, and the waste reservoirs with the analyte waste reservoir grounded in order to control the injection plug shape. In the separation mode, the potential was applied to the buffer reservoir with the waste reservoir grounded and with analyte and analyte waste reservoirs at approximately half of the potential of the buffer reservoir. The analytes used for the diagnostic experiments were rhodamine B and sulforhodamine101(Exciton Chemical Co., Inc.) at 60 pM for the CCD images and 6 pM for the point detection. A sodium tetraborate buffer (50 mM, pH 9.2) was the mobile phase in all experiments. The laser dyes studied are easily separated, but for the sake of column diagnostics, the information is invaluable.

1.0

-."

/

'

'

0.8 -

Y

E

1

0.6 0.4 -

0.0

b

Figure 3. (a) CCD image of injection cross with no fluorescent analyte present; CCD images of same region for (b) a pinched sample loading and (c) a floating sample loading using rhodamine B.

RESULTS AND DISCUSSION The two types of sample loading were tested for sample introduction into the separation column. Rhodamine B was used as the test analyte for the sample loading studies. The analyte is placed in the analyte reservoir and in both injection schemes is pumped in the direction of the analyte waste reservoir. CCD images of the two types of injections are depicted in Figure 3. Figure 3a is a CCD image of the injection cross with no fluorescent sample present. Figure 3b shows the pinched sample loading prior to being switched to the separation mode where the analyte is pumped electrophoretically and electroosmoticallyfrom the analyte reservoir to the analyte waste reservoir (left to right) with the buffer from the buffer reservoir (top) and the waste reservoir (bottom) traveling toward the analyte waste reservoir (right). The brighter (lighter) portions of the image are due to sample fluorescence. The voltages applied to the analyte, buffer, analyte waste, and waste reservoirs were 90%,90%,0%, and 100%of the power supply output and correspond to electric field strengths in the channels of 270,400,690, and 20 V/cm respectively, for an applied potential of 1 .O kV. Consequently, the analyte in the injection cross has a trapezoidal shape and is constricted and diluted in analyte waste channel (right) by

1 0

25

50

time

75

100

[SI

Figure 4. Variation of amount of analyte in Injection area with time for a pinched sample loading (circle) and a floating sample loading (square) using rhodamine B.

this flow pattern. Figure 3c shows a floating sample loading. The analyte is pumped from the analyte reservoir to the analyte waste reservoir as in the pinched sample loading except no potential is applied to the buffer and waste reservoirs. By not controlling the flow of mobile phase in the buffer and separation channels, the analyte is free to move into these channels through eddy flow, resulting in a more diffuse injection plug. When the pinched and floating sample loadings are compared, the pinched sample loading is superior in two areas: temporal stability and plug length. When two or more analytes with vastly different mobilities are to be analyzed, a sample loading with temporal stability ensures that equal volumes of the faster and slower moving analytes are introduced into the separation column. A smaller plug length leads to a higher efficiency and, consequently, to a greater component capacity for a given instrument. To determine the temporal stability of each sample loading method, a series of CCD fluorescence images was collected at 1 5 s intervals starting just prior to the analyte reaching the injection cross. An estimate of the amount of analyte that is to be injected was determined by integrating the fluorescence in the intersection and the buffer and separation channels. This integrated fluorescenceis plotted versus time in Figure 4. For the pinched sample loading (circle), a stability of 1% rsd is seen, which is comparable to the stability of the illuminating laser. For the floating sample loading (square), the amount of analyte to be injected into the column increases with time because of the flow anisotropy. For a 30-s loading period, the volume of the injection plug in the injection cross is -90 pL and stable for the pinched sample loading versus -300 pL and continuously increasing with time for a floating sample loading. When the switch is made to the separation mode, some of the analyte in the analyte channel and the analyte waste channel is injected into the separation column. For the pinched sample loadingin Figure 3, the volume of the injection plug becomes 144 f 17 pL with a plug length of 158 f 19 pm following the switch to the separation mode. These dimensions are estimated from a series of CCD images taken immediately after the switch is made to the separation mode. The reproducibility for the two injection modes was tested by integrating the area of the band profile following introduction into the separation channel by monitoring the Analytical Chemistry, Vol. 66,No. 7, April 1, 1994

1109

10

12

14 time

10

16

18

[SI

12

14 16 18 time [SI Flgure 5. Reproducibilityof (a, top) pinched sample loadings and (b, bottom) floating sample loadingsfor five replicateinjectionsof rhodamine B using a separation length of 9 mm and a separation field strength of 120 V/cm.

separation channel at a point 9 mm from the intersection. For five injections with a duration of 40 s, the reproducibility for the pinched sample loading (Figure sa) is 1.7% rsd and for the floating sample loading (Figure 5b) is 2.7% rsd. The pinched sample loading, not surprisingly, has a higher reproducibility because of the temporal stability. With electronically controlled voltageswitching,the rsd is expected to improve for both schemes. The injection plug length and, ultimately, the resolution between analytes depend largely on both the flow pattern of the analyte and the dimensions of the injection cross. For this column, the channel width is -80 pm, but a channel width of 10 pm is feasible which would decrease the volume of the injection plug to -2 pL with a pinched sample loading. After the sample has been pumped into the cross region of the chip, the voltages are manuallyswitched from the sample loading to the separation mode of operation. In Figure 6 the CCD images record the separation process at 1-s intervals with Figure 6a showing a schematic of the section of the chip imaged and with Figure 6b-e showing the separation unfold in time. Figure 6b again shows the pinched sample loading with the applied voltages at the analyte, buffer, and waste reservoirs equal. Parts c-e of Figure 6 show the plug moving away from the intersection at 1,2, and 3 s, respectively, after switching to run mode. In Figure 6c the injection plug is migrating around a 90' turn, and band distortion is obvious dus to the inner portion of the plug traveling less distance in 1110

Analytical Chemistry, Vol. 66, No. 7, April I , 1994

Figure 6. (a) Schematicof region imaged (injectioncross); CCD images of (b) a pinched sample loading and (c-e) a separation of rhodamine B (less retained) and sulforhodamine (more retained) at 1, 2, and 3 s, respectively,after switchingto the separationmode usinga separation field strength of 150 V/cm.

the turn than the outer portion. By Figure 6d the analytes have separated into distinct bands, which are distorted in the shape of a parallelogram. In Figure 6e the bands are well separated and have attained a more rectangular shape i.e., collapse of the parallelogram due to radial diffusion, an additional contribution to efficiency loss. When the switch is made from the sample loading mode to the separation mode, a clean break of the injection plug from the analyte stream is mandatory to avoid tailing. This is achieved by pumping the mobile phase from the buffer channel into the analyte, analyte waste, and separation channels simultaneously by maintaining the potential at the intersection below the potential of the buffer reservoir and above the potentials of the analyte, analyte waste, and waste reservoirs. For these experiments, the intersection was maintained at 66% of the potential of the buffer reservoir during the run mode. This provided sufficient flow of the analyte back away from the injection cross down the analyte and analyte waste channels without decreasing the field strength in the separation channel significantly. This threeway flow is demonstrated in Figure 6c-e as the analytes in the analyte and analyte waste channels (left and right,

t. 0

300

200

100 time [a]

Flgure 7. Electropherograms at 33 (top), 99 (middle), and 165 mm (bottom) from the point of injection for rhodamine B (less retained)and sulfomodamine (more retained) using a separation field strength of 170 Vlcm and pinched sample loading. The electropherogramshave been offset vertically to facilltate viewing. 40000

op

(3)

1

where linj and ldet are the channel lengths of the injection plug and detection path, respectively, and L is the length of the separation column between the point of injection and detection. Often, these two contributions are negligible compared to the other contributions. The contribution of the axial diffusion to the band broadening follows from the Einstein equation:"

30000

ta

I

(I

i

where Hinj, Hdct, Hdiff, and Hgcocorrespond to the contributions to the plate height from the injection plug length, detector path length, molecular diffusion in the axial direction, and geometry of the column, respectively. The contributions of the injection plug length and detector path length are constant, the contribution of the axial diffusion depends on the linear velocities of the analytes, and the contribution of the column geometry depends on the number of turns per unit length in the column. The contribution to the plate height from Joule heating is neglected because heat dissipation was below 1 W/m for all experiments.15 The contributionsof the injection plug length and detection path length can be calculated byI6

2o000 10000

1

0

0

5

10 length [cml

1s

20

Flgure 8. Varlatlon of the plate number wlth separation channel length for modemine 6 (circle) and sulforhodamine (square) with best linear fits (line) using a separation field strength of 170 V/cm and pinched sample loading.

respectively) move farther away from the intersection with time. Three-way flow permits well-defined, reproducible injections (Figure 5) with minimal bleeding of the analyte into the separation channel. To obtain electropherograms in the conventional manner, single-point detection with the helium/" laser was used at different locations down the axis of the separation column. The efficiency at 10 evenly spaced positions was monitored, each constituting a separate experiment. Figure 7 depicts selected electropherograms at 33,99, and 165 mm from the point of injection. 'The efficiency data are plotted in Figure 8. At 165 mm from the point of injection, the efficiencies of rhodamine B and sulforhodamine are 38 100 and 29 000 plates, respectively. Efficiencies of this magnitude are sufficient for many separation applications. The linearity of the data provides information about the uniformity and quality of the channel down the length of the column. If a defect in the channel, e.g., a large etch pit, was present, a sharp decrease in the efficiency would result. This is not the case. The primary contributions to band broadening for these experiments are shown in the following equation for the total plate height:

Hdiff= 2DJu

(4)

where Dm and u are the diffusion coefficient of the analyte and the linear velocity of the analyte, respectively. The key to minimizing this contribution is performing the separation as quickly as possible. Less straightforward is the contribution of the geometry of the column to the plate height because of the serpentine pattern of the column. A geometry such as this serpentine enables a significantly greater column length to be employed for the separation in a much smaller area (a 165-mm column length for the serpentine geometry versus 10-mm column length for a straight geometry in a 10 mm X 10 mm chip area). However, by bending the column around 180' turns, sacrificing column efficiency for column length becomes a concern. We will only consider simple geometrical broadening effects here although other contributions such as electric field effects are quite feasible, as discussed below. For each turn the difference in the length of the column between the inside and the outside is N

dl = e dr

(5)

where 8 is the angle of the turn and dr is the difference in the radius of the turn between the inside and the outside of the channel, i.e., the channel width. With simple geometrical broadening, we assume that the analyte is traveling at the (15) Monnig, C. A.; Jorgenson, J. W. Anal. Chem. 1991, 63, 802. (16) Sternberg, J. C. Adu. Chromatogr. 1966, 2, 205. (17) Giddings,J. C. Dynamicsof Chromatography,Part I: Principles and Theory; Marcel Dekker: New York, 1965; Chapter 2.

Analytical Chemistty, Vol. 66, No. 7, April 1, 1994

1111

same velocity on both the inside of the column and the outside of the column through the turn. Consequently, the shift or distortion in the band profile is equivalent to dl in eq 5. For the 909 turn immediately following the injection cross in the separation channel (Figure 6a), 8 and dr are equal to 7r/2 and 90 pm, respectively, and consequently, dl is 140 pm, which should be the contribution to band broadening for this one turn. From the experimental data in Figure 6d, the shift in the band shape is calculated from the center of the bands for each row of pixels from this CCD image, and the band distortion is estimated to be 150 pm. The slight discrepancy (7%) can be attributed to the nonrectangular injection plug. The geometry of the turns could also produce an additional distortion due to spatially varying electric field strengths within the turn. Thus, an ion on the inside of the turn would experience a higher electric field than an ion on the outside of a turn, resulting in a greater linear velocity for the ion on the inside. Coupled with thedifference in the lengths of travel, an ion on the inside of a turn and an ion on the outside would be separated by a distance 2dl following a turn, e.g. 280 pm for a 90° turn. This distortion is too large compared to the experimentally measured band distortion of 150 pm observed in Figure 6d. Consequently, the contribution to the plate height equation due to the turns in the serpentine pattern of the separation column is written as follows:

where n is the number of identical turns, w is the width at the top of the channel, and 8 is the angle of a single turn. The transit time between turns is assumed to be long compared to the time todiffuseacross thechannel. Therefore the position of an individual molecule in the cross section of the channel, from turn to turn, is random, and Hgcois a function of n rather than n2.18 Assuming the number of turns per unit distance is constant, then Hgaowill be constant. Importantly, this contribution to the plate height decreases as the square of the channel width. The contributions from injection plug length, detector path length, and column geometry are constant for a given experiment while the contribution from the axial diffusion decreases with increasing linear velocity. In Figure 9, the plate height versus linear velocity data demonstrate an increasing efficiencywith increasing linear velocity. The three constant terms are combined, added to an axial diffusion term, and fitted to the experimental data. The sum of the three constant contributions for rhodamine B and sulforhodamine are 4.02 and 5.39 pm, respectively, and the diffusion coefficients, D,, are 3.27 X 10-6 and 3.33 X 10-6 cm2/s, respectively. The individual constant plate height contributions for this experiment are estimated. The injection plug length is equal to 300 pm for the the injection in Figure 6, and Hinj from eq 2 is 0.23 pm. Since the detection is on-column, the detector path length is equal to the laser spot size, which is 100 pm. Therefore, (eq 3) is equal to 25 nm. Because the separation was monitored at a distance 33 mm from the point

-

(18) Giddings,

J. C. J. Chem. Educ. 1958, 35, 588.

1112 Analytical Chemistry, Vol. 66, No. 7, April 1, 1994

X 8 -

. 0.0

0.2

0.4

0.6

0.8

1.0

u ~ d s l

figure 0. Variation of the plate height (H) wlth linear veioclty (u) for rhodamine B (circle) and suiforhodamine (square)with best fits of eq 1 (lines) using a separation length of 33 mm and pinched sample loading.

of injection, the separation column includes one 90' turn (described above) and four 180' turns. Using eq 6, Hgcois equal to 0.85 pm. The fitted constant values from the data in Figure 9,4.02 for rhodamine B and 5.39 for sulforhodamine, are not equal and are both greater than the sum of the three estimated constant contributions of 0.90 pm. Limitations of the present apparatus prevented the use of field strengths greater than 170 V/cm, and therefore, more comprehensive data could not be generated. CONCLUSIONS The use of microchip devices for liquid-phase separations and analysis appears to be quite promising. Several important aspects for improving the performance of these devices were discussed in this work. Cold bonding of cover plates to enclose micromachined channels will allow the use of a broad range of substrate materials and could lead to greater device yields. An injection scheme was presented that provides high performance with respect to accuracy, precision, and small volume while remaining simplistic with respect to fabrication. The device used in these studies has a smaller arealcolumn length ratio than previously reported due to the use of a serpentine column g e ~ m e t r y .The ~ ~ use of such geometries will be necessary to move microchip liquid-phase analysis devices into the sensor realm. Band-broadening phenomena associated with the serpentine structure, while measurable, do not appear to be a severely limiting problem for capillary electrophoresisimplementations employing field strengths less than -200 V/cm. The plate height associated with the serpentine geometry was experimentally measured and found to be approximately what would be expected for path length differences associated with the turns. This additional plate height was calculated to be 1 pm for the 165-mm column length. Reduction in channel widths should allow reduction of this effect to an acceptable level.

-

ACKNOWLEDGMENT This research was sponsored by the US.Department of Energy (DOE), Office of Research and Development. Oak Ridge National Laboratory is managed by Martin Marietta

Energy Systems, Inc., for the U.S.Department of Energy under Contract DE-AC05-840R21400. Also, this research was sponsored in part by an appointment for S.C.J. to the Alexander Hollaender Distinguished Postdoctoral Fellowship Program by the ‘sa for L’B*K*to the Laboratory Cooperative Postgraduate Research Training Program, and for R.H. to the ORNL postdoctoral research associate program. These programs are administered by the

Oak Ridge Institute for Science and Education and ORNL. The authors acknowledge many useful discussions with Drs. Roswitha S.Ramsey and Georges Guiochon. for review October 5, 1993. Accept& January 10, 1994.a Abstract published in Advance ACS Abstracts, February 15, 1994.

Anal)rtical Chemistty, Vol. 66, No. 7, April 1, 1994

1I 1 5