1372 I
I
I
I
I
I
m(d) S to
t, v)
particle size distribution function particle diameter selectivity (=S& void time retention time retention time for particles of unit diameter flow rate through channel retention volume channel thickness
Y
4 1
5,
l(i
4w
Greek Alphabet density difference between particle and carrier AP particle density P
V
P w
vr
b
&
LITERATURE CITED (1) Gddings. J. C. Anal. Chem. 1981, 53, 1170A-1175A. (2) Giddings, J. C. S e p . Sci. Technol. 1984, 19, 831-847.
DIAMETER (pm) Figure 11. Comparisonof mass-bad size distributioncurves for the 5-50-pm sample obtained from channels I and 11.
using polystyrene latex beads and then by changing the rotation rate of the centrifuge to compensate for the density difference between the sample material and the latex calibrant. This procedure is shown to yield good results when tested by (a) microscopy, (b) self-consistency between two different FFF channel structures, and (c) a comparison with an external (NIST) cumulative size distribution curve. Further refinements in this approach should lead to increased speed and accuracy.
b C
c(t,) d
G
L m
GLOSSARY channel breadth mass concentration in effluent corrected fractogram signal particle diameter acceleration channel length particle mass
(4) Karaiskakis, G.; Myers, M. N.; Caldweli. K. D.; Giddings, J. C. Anal. Chem. W81, 5 3 , 1314-1317. (5) Peterson, R . E., 11; Myers, M. N.; Gddings, J. C. S e p . Sci. Technol. 1984, 19, 307-319. (6) Segr!, G.; Silberberg. A. Nature 1861, 789, 209-210. (7) Segre, G.; Silberberg, A. J. FluidMech. 1962, 74, 115-135. (8) Segrd, G.; Silberberg, A. J. FluidMech. 1982, 74, 136-157. (9) Saffman, P. G. J. Fluid Mech. 1965,22, 385-400. (10) Cox, R. G.; Brenner. H. Chem. Eng. Sci. 1988,23, 147-173. (11) Ho,8. P.; Leal, L. G. J. fluid Mech. 1974,6 5 , 365-400. (12) Vasseur, P.; Cox, R. G. J. Fluid Mech. 1976, 78, 385-413. (13) Cox. R. G.; Hsu, S. K. Int. J. Multiphase Flow 1977, 3,201-222. (14) Leal, L. G. Annu. Rev. FluidMech. 1980, 12,435-476. (15) Ratanathanawongs, S.K.; Giddings. J. C. J. Chromatogr. 1989, 467, 341-356. (16) CaMwell, K. D.;Nguyen, T. T.; Myers, M. N.; Giddings, J. C. Sep , Sci. Techno/. 1970, 14, 935-946. (17) G i i n g s , J. C.; Myers, M. N.; Yang, F. J. F.; Smith, L. K. I n ColloMand
Interface Science; Kerker, M., Ed.; Academic Press: New York, 1976;Vol. IV, pp. 381-398. (18) Koch, T.; Giddings, J. C. Anal. Chem. 1986,58,994-997. (19) Giddings, J. C. S e p . Sci. Technol. 1989,24, 755-768. (20) Moon. M. H.; Myers, M. N.; Giddings, J. C. J. Chromatogr. ISSO. 517,
423-433.
RECEIVED for review December 17, 1990. Accepted April 16, 1991. This work was supported by Public Health Service Grant GM10851-33 from the National Institutes of Health.
Nanoliter-Scale Multireflection Cell for Absorption Detection in Capillary Electrophoresis Tiansong Wang, Joseph H. Aiken, Carmen W. Huie, and Richard A. Hartwick* Department of Chemistry, State University of New York at Binghamton, Binghamton, New York 13902 A multlreflective absorption cell for CZE is fabricated and examined by both static and dynamlc measurements. The new cell Is characterlred by Improved sensittvlty as compared to conventlonal single-pass cells, wlth no increase In cell volume. A 40-fdd Improvement in sensltlvtty is obtalned when compared to a single-pass cell, wlth similar noise levels. A concentratton detectton ilmlt of 6.5 X lod M for Millant green Is estimated for the new cell design from static measurements. A theoretical analysis of cell performance using ray tracing for both axial and radlai reflections in the new cell shows good agreement with experlmentai results.
INTRODUCTION Detection is one of most important and rapidly developing areas in capillary zone electrophoresis (CZE). Various detection principles, such as spectrophotometric, mass spectrometric, electrochemical, and radiometric detection, have
been applied to CZE (1). Although the sensitivity of UVvisible absorption detection is perhaps the lowest among these detection methods, UV detectors are the most popular detector currently utilized in CZE because of their simplicity and versatility. Most UV detectors in CZE employ single-pass detection, which was first introduced by Yang for capillary liquid chromatography (2). In this design, the light beam travels normal to the capillary axis,crossing the capillary only a single time. According to Beer’s law,
A = tbc (1) where A is the absorbance, t is the molar extinction coefficient of the sample, b is the light path length, and c is the sample concentration. The low sensitivity in single-pass detection is predictable because the path length is limited by the inner diameter of the capillary, typically 50-75 Fm. Therefore, even with state of the art detectors, concentration detection limits are rarely lower than ca. 10% M.
0003-2700/91/0363-1372$02.50/0 62 1991 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 83, NO. 14, JULY 15, 1991
For a given molecule in a particular separation environment, absorption detection limits can be improved by decreasing the noise and/or increasing the path length. Some UV detectorsfor CZE have noise levels as low as 2 X lab au, resulting in detection limits on the order of ca. lo* M for a good chromophore. Further reductions in noise levels are possible but remain technically difficult. The other way to improve the detection limit is to increase the effect path length. Rectangular capillaries have been investigated for CZE (3). With a 20-fold increase in path length, a 15-fold increase in sensitivity was obtained. A z-shaped absorption cell has been developed for capillary liquid chromatography (4). The path length was about 2 cm and sensitivity enhancement was 100-500 times. However, the volume of the z-shaped cell was too large for CZE. Another technology called axial-beam absorption detection has recently been introduced to open tubular capillary liquid chromatography (5). The path length in axial-beam detection was estimated to be lo00 times compared to that in single-pass detection, and a 192-fold improvement in detection limit was obtained. However, axial-beam detection requires precise alignment between the capillary and the light source and/or restricts the choice of mobile phase to solutions with refractive indices higher than that of fused silica. Multireflection of the interrogative light beam with mirrors is a well-known technique to increase effective path length and thus to increase the sensitivity of absorption detection. A multireflection cell was first introduced by White (6) in order to analyze low-level gaseous samples by infrared spectroscopy, and multireflection cells (path length 1to over 100 m) have been a regular accessory for infrared analysis of gas (7,B). To date, however, this technique has not been applied to nanoliter-scale separations such as CZE and micro-LC, despite the obvious needs in both technologies for improved detection sensitivities. In this paper, a successful nanoliterscale absorption cell based on multireflection from externally mirrored capillaries is r e p o d . The performance of the new cell was evaluated statically and was applied to CZE separations. The basic theory of the new cell is presented by using ray-tracing simulations.
THEORY The light ray tracing in a capillary can be established by a geometrical optical approach, and some patterns for single-pass detection have been reported (9-11). For a multireflection cell, the ray tracing is more complex than that in single-pass cell, since the light path must be traced both radially and axially. According to geometrical optics (12), when a light ray impinges on a reflective surface, reflection will take place and the reflection angle is equal to the incident angle. When a light ray crosses from medium 1to medium 2, refraction will take place and the refraction angle follows Snell's law: nl sin 8, = n2 sin O2 (2) where dl and f12 are the incident and the refraction angle and nland n2 are the refractive index of medium 1 and medium 2, respectively. There are two situations in the multireflection cell: reflection along the axis of a capillary and reflection along the circumference of a capillary. Axial Reflection. The ray tracing can be seen from Figure 1. According to eq 2, O2 = arcsin [(nlsin Ol)/n21 (3) From trigonometry, ll = d, tan d2 (4) and
e3 = e2
(5)
1373
T dl
i
d2
P
l-----L------l
Figure 1. Ray tracing of axial reflection.
-Y
n'
w
Figure 2. Ray tracing of radial reflection.
Similarly, l2 and l3 can be obtained. Then, the total length that the ray traveled is
L = I,
+ l2 + l3
(6)
In the current cell design, the exterior of the capillary is silver mirrored. When the ray strikes the silver coating, it will be reflected with Be = 0, and the tracing will follow the same pattern established above. After traveling a distance D along the axis of a capillary, the number of reflections of the ray will be
N = D/L - 1
(7)
N also indicates the increased time in path length. Assume a capillary with 75-pm i.d. X 364-pm o.d., nl = 1.000, n2 = 1.458, n3 = 1.333(water),and el = 5; then L will be 0.0222 mm. For a traveling distance of 1mm, N is equal to 44,that is, 44 time increase in the path length. Radial Reflection. The ray tracing is shown in Figure 2. The incident ray is assumed to be parallel with the X axis and its position is y,. According to trigonometry, coordinate x1 will be
Coordinates x2,y2, x3,y3,and x4,y4can be established by the method in ref 10. When the ray reflects at point 4, it is obvious that
e.,
=
e3
(9)
Therefore, the ray tracings before and after reflection are symmetric along the X',-X'axis, which is rotated by an angle of cy with regard to the X,-X axis. The angle a depends on both the incident position (xl,yl) of the ray and the refractive indices of each medium.
1374
ANALYTICAL CHEMISTRY, VOL. 63, NO.
14,
JULY
15, 1991
X
X
Flgure 3. Typical ray tracing after four reflections in a multireflection
cell (75-pm 1.d. X 364-pm 0.d.). (a) Incident position y , = 0.10; (b) Incldent posltion y , = 0.20.
5
9b
Flgure 5. Schematic diagram of apparatus used for multlreflectlon measurements. 1, He-Ne laser: 2, conventional cell; 3, capillary; 4, filter; 5, photomultiplier tube (PMT): 6, PMT high-voltage supply; 7,
detection system: 8, recorder; 9a and 9b, buffer reservoir: high-voltage supply; 11, rotary stage.
I
L>L D:
3
‘Lb
3ete:t;r
Figure 4. Multireflection cell.
Since the ray tracing is symmetric along the X’,-X’axis, further coordinates such as x5,y5can easily be established by mathematic operation. Firsly, convert x3a3 to x3)jY3/ as follows: x3’ = x 3 cos a y 3 sin a (10)
+
y3’ = -x3 sin a
+ y 3 cos cy
(11)
The coordinates xgl,ygl will be = x3’
(12)
Ys/ = -Y3’
(13)
xg’
Secondly, convert xgl,ygl to x5,y5 as follows: x 5 = x i cos a
- y5‘ sin a
(14)
y5 = xs/ sin a
+ y5’ cos a
(15)
Figure 3 presents two typical ray tracings after four reflections in a capillary with 75-pm i.d. and 364-pm o.d., nl = 1.000, n2 = 1.458, and n3 = 1.333(water) from which the distribution of light along the circumference of the capillary can be seen.
EXPERIMENTAL SECTION Fabrication of the Cell. A schematic diagram of the multireflection cell is shown in Figure 4. The capillary was made of fused silica with 75-pm i.d. X 364-pm 0.d. and 51 cm in length (Polymicro Technologies, Inc., Phoenix, AZ). The capillary also served as the separation column in CZE. An opening of about 1 cm was made on the capillary simply by burning off the polyimide coating of the capillary. Then the silver layer was deposited on the opening by redox reaction of Ag(NH&+ and glucose (see Reagents subsection). Black paint (Illinois Bronze Paint Co., Lake Zurich, IL) was applied on the silver layer to protect it from physical damage. The light windows made on the cell were separated by distance D , and D2 (0.8 and 1.5 mm, respectively). The cell volume calculated from D2 was 6.6 nL. Apparatus. A schematic diagram of the apparatus is shown in Figure 5. A 5-mW He-Ne laser (Model 1105P, Uniphase, Sunnyvale, CA) was used as the light source. The capillary with the multireflection cell was mounted on a conventional cell, which has been described elsewhere (13). The conventional cell and a 632.8-nm interference filter (Corion Corp., Holliston, MA) were taped on the window of a photomultiplier tube (R928, Hamamatsu, Somerset, NJ). The entire assembly was then mounted
10, CZE
on a rotary stage, which allowed f i e adjustment of incident angle. The photomultiplier tube was operated by a high-voltage supply (2K10, Power Designs Inc., Westbury, NY).The laser beam first passed through the aperture (60 pm X 1.4 mm) of the conventional cell and then struck one of the windows of the multireflection cell. The light intensity exiting the other window was detected by a photomultiplier tube that either was connected to a Keithley 177 digital multimeter (Keithley Instruments, Inc., Cleveland, OH) in static measurement or was recorded by an Oriel 7072 detection system (Oriel Corp., Stratford, CT) and a OmniScribe A5111-5 chart recorder (Houston Instrument, Austin, TX) in CZE running. A PS/MK3OPO2.5 high-voltage power supply (Glassman High Voltage, Whitehouse Station, NJ) was used for CZE analysis. Single-pass detection also was conducted by using a conventional cell (13) in order to compare the results with the multireflection detection. A similar capillary (75-pm i.d. X 364-pm 0.d. and 47 cm in length) was used. The aperture was 60 pm X 0.22 mm and was placed behind the capillary, i.e., a laser beam passed the capillary first and then the aperture. Evaluation Procedure. The performance of both the multireflection and the conventional cell was examined and compared by static measurement and CZE running. The test compound was brilliant green. In static measurement, the capillary was filled with distilled water and brilliant green-water solution altemately and signal currents were recorded. In CZE running, 0.01 M citrate buffer of pH 4.5 was used and brilliant green was dissolved in the buffer. Bare capillaries were used without special treatment except to wash with 1% NaOH for 20 min. The sample was injected by the siphon method (level difference, 5 cm, 30 s). The running voltage was 15 kV, and the current was 0.03 mA. When necessary, data were converted into absorbance: A = log Uo/O (16) where Z, and Z are currents measured from the reference and the sample solution, respectively. Reagents. Siluer Deposit. Solution A . Dissolve 1g of silver nitrate in 20 mL of distilled water, add concentrated ammonium hydroxide until the solution is clear, then add 20 mL of 4% sodium hydroxide, and use ammonium hydroxide to redissolve all precipitate (WARNING: Solution A can generates explosive compounds; do not store solution A over 12 h). Solution B. Dissolve 0.5 g of glucose in 20 mL of distilled water. The silver deposit reaction will take place when equal amounts of solution A and B are mixed. Brilliant green was obtained from Aldrich (Milwaukee, WI), ammonium hydroxide from Corco (Fairless Hills, PA), and all other chemicals from Fisher (Fair Lawn, NJ).
RESULTS AND DISCUSSION Incident Angle. The incident angle 0 (see Figure 4) is a critical parameter because it affects the light intensity arriving a t the photodetector and controls the number of reflections in the capillary, thus controlling the detection sensitivity. Figure 6 indicates the relationship of output light intensity to incident angle 0. When 6’ ;= 0, the light intensity is very low (only 10% of the maximal intensity). The output light
ANALYTICAL CHEMISTRY, VOL. 03, NO. 14, JULY 15, 1991 3.0
1375
1
8.6
-
e.0
-
1.6
-
1.0
-
lo-'
1 o-?
0.01
'
'
'
'
'
'
'
o
e
4
(I
e
io
iz
14
' io
'
1
18
80
I
Flgure 8. Relationship of incident angle to output light intensky.
0.016
-
0.090
-
0.016
-
0.010
-
0.006
n
lo-'
I
io-'
Incident Angle(degree)
10-0
1 r 5
10-4
Sample C o n c e n t r a t i o n ( V )
Figure 8. Linear dynamic range of detection with multireflection cell.
a
4
a
a
io
io
14
in
_-
Incident Angle(degree)
Flguro 7. Relationships of incident angle to number of reflection and to absorbance.
intensity quickly increases with increasing B from 0 to 7O, has a maximum a t 7O,and then slowly decreases with increasing 8. The relationships of 8 to the number of reflections and to the sensitivity are shown in Figure 7. In theory, the sensitivity of a multireflection cell is proportional to the number of reflections, and reducing the incident angle will increase the number of reflections. Therefore, decreasing 6 will enhance the sensitivity. This appears to be true for B in the range &loo. However, such a relationship does not hold at smaller incident anglw. When B is reduced h m 5 to 3O, the calculated number of reflections increases from 44 to 74, which means that the path length will be increased by about 1.7 times. However, the absorbance only increased by 2.8% experimentally, and the loss of light intensity was about 60%. Therefore, an incident angle of 5 O was chosen in the following experiments. Linear Dynamic Range and Detection Limit. The linear dynamic range of the multireflection cell was examined by static measurement with B = 5 O , and the results are shown in Figure 8. The linearity established from Figure 8 is 2 orders of magnitude. In static measurement, noise is about 4.3 x lo-' au, and the absorbance of 1.1 X lo-' M brilliant green is 0.0022 au. The calculated detection limit (signal to noise ratio = 3) is 6.5 X 10" M. Although brilliant green has a high molar extinction coefficient (c = 8.2 X l@ at 633 nm and pH 5.61, such a detection limit still is very impressive. When applied to actual CZE separations, the calculated concentration limit of detection was increased to 3.0 X lo-' M (injected concentration)because of band dispersion and increased noise levels. Comparison w i t h Single-Pass Cell. The performance of the multireflection cell was compared with that of the single-pass cell by both static and dynamic measurements. The results are summarized in Table I, and two electropherograms are presented in Figure 9. It can be seen that the sensitivity of the multireflection cell is over 40 times higher
L. H 1 min
Flgure 9. Electrophoregrams of 1.1 X M brilliant green. (a) Obtained from single-pass cell; (b) obtained from multirefiection cell.
See text for conditions.
Table I. Performance Comparison of Multireflection Cell (mc) to Single-Pass Cell (sc) condition
absorbancea mc (0 = 5 O ) sc
static measurement 0.230 static noise 4 x 10"' CZE measurement 0.0899 8.3 X lo-' CZE noise a 1.1 x M brilliant green.
A,,IA,
0.0055
41.8
0.0022 6.6 X lo4
40.9
4 x 10-4
than that of the single-pass cell, which is very close to the calculated increase of path length (44times). Noise levels were similar for both cells. Cell Design. Literally, the sensitivity of the multireflection cell can be increased further by increasing the distance between the two windows. In practice, however, the distance D2will be restricted by loss in separation efficiency. According to previous research (13),the distance should not be over la of the peak width, and for a typical peak in CZE, la is about 0.9mm. Besides the loss in efficiency, longer distance coupled with more reflections also reduces the output light intensity
Anal. Chem. 1991, 63, 1376-1379
1376
from the exit window because of the loss in light power during reflection, and at the extreme, the output light intensity will be lower than the detector noise, resulting in no detectable signal. Recently, Werle and Slemr (14) have found there is an optimal number of reflections for multireflection cell, and the number depends on the incident light power, the mirror reflectivity, and the modulation method of laser. Therefore, optimization of cell length will be important for different reflective coatings and light sources. Another important aspect of cell design is the shape and dimensions of the windows. The critical parameters are the axial length and radial width of incident window and the shape of exit window. Although no experiments were conducted, theoretical analysis is very helpful: (a) In Figure 4,there are two typical incident rays. Ray 1 enters the capillary and ray 2 exits the window after the first reflection. Ideally, all rays between ray 1 and ray 2 should enter the capillary. In order to obtain higher output light intensity from the exit window, the axial length of the incident window should not be smaller than 2L (about 0.05-0.1 mm), which allows the maximal amount of light to enter the capillary. (b) According to the theory of radial reflection, the rays that do not pass the inner diameter of the capillary during the first segment of travel in the capillary (because their incident positions are too far away from the center of the capillary) will never pass the inner diameter in their entire travel and thus have no contribution to sample absorption. Therefore, the radial width of incident window should be equal to or slightly narrower than the inner diameter of a capillary in order to cut off the useless light. (c) Also from the theory of radial reflection, a ray will rotate an angle after each reflection (Figure 3) with the angle depending on both incident position of the ray and the refractive index of each medium. It is expected that, after many times of reflection, rays will be distributed around the entire cir-
cumference of the capillary. Therefore, the exit window should ideally be a ring-shaped window, and a ring-shaped photodetector should be used in order to collect the entire light output.
ACKNOWLEDGMENT We acknowledge Robin Robinson from the Department of Chemistry, State University of New York at Binghamton, for his assistance in the computer simulations of the ray-tracing studies and Scott Weinberger from Spectra-Physics Analytical Division for his helpful discussions. LITERATURE CITED Ewing, A. G.; Wallingford, R. A.; Oleflrowicz, T. M. Anal. Chem. 1989, 6 1 , 292A-303A. Yang, F. HRC&CC, J . High Resolof. Chromafogr. Chromatogr. 1981, 4 , 83-85. Tsuda, T.; Sweedler, J. V.; Zare, R. N. Anal. Chem. 1990, 62, 2149-2152. Chervet, J. P.; Ursem, M.; Salzmann, J. P.; Vannoort, R. W. HRC&CC J . Hgh Reson. Chromafogr. Chromafogr. 1989, 12, 278-281. Xi, X.; Yeung, E. S. Anal. Cbem. 1990, 62, 1580-1585. White, J. U. J . Opt. SOC. Am. 1942, 3 2 , 285-288. Willard, H. H.; Merritt, L. L., Jr.; Dean, J. A,; Settle, F. A., Jr. Insfrumental Methods of Analysis, 7th ed.; Wadsworth: Belmont. 1988; p 306. Kiss-Eross, K. I n Comprehensive Analyfical Chemistry; Svehla, G., Ed.; Elsevier: Amsterdam, 1976; Vol. V I , pp 165-168. Vindevogel, J.; Schuddinck, G.; Dewaele, C.; Verzele, M. J . High Resoluf. Chromafogr. Chromafogr. Commun. 1988, 1 1 , 317-321. Synovec, R. E. Anal. Chem. 1987, 59, 2877-2884. Bruno, A. E.; Gassmann, E.; Pericles. N.; Anton, K. Anal. Chem. 198g7 61 876-883. Finchman. W. H. A.: Freeman, M. H. Opfics, 8th ed.;Butterworths: London, 1974; pp 15-21. Wang, T.; Hartwick, R. A,; Champlin. P. B. J . Chromafogr. 1989, 462, 147-1 54. Werle, P.; Slemr, F. Appl. Optics W91, 3 0 , 430-434. I
RECEIVED for review January 28, 1991. Accepted April 23, 1991. This research was supported by grants from the Spectra-Physics Analytical Division and from Unilever Research.
Gas Chromatographic Determination of Trace Impurities in High-Purity Nitrogen Using Hydrogen Storage Alloys as a Nitrogen Adsorbent Hiroshi Ogino,* Yoko Aomura, and Tetsuya Seki
Technical Research Laboratory, Toyo Sanso Company, Ltd., 3-3, Mizue-cho, Kawasaki-ku, Kawasaki-shi, Kanagawa 210, J a p a n
A comblnatlon of two dlfferent hydrogen storage alloys was used to assist in the gas chromatographic determlnatlon of trace impurities, such as neon, argon, krypton, methane, and xenon, in nitrogen. Thls system conslsts of a gas chromatograph combined, in series, with two precoiumns filled with WMn-type and Ca-type hydrogen storage alloys. The former alloy efficiently retalns nitrogen at 250 O C and under the pressure of the carrier gas. The latter alloy is used in order to more efficiently adsorb hydrogen, whlch is released from the former alloy, at ambient temperature. By using a photoionlzatlon detector, the detection limits achieved were as follows: 3.7, 0.08, 0.03, 0.04, and 0.09 ppm for Ne, Ar, Kr, CH,, and Xe, respectively.
* To whom
correspondence
should b e addressed.
INTRODUCTION High-purity nitrogen used as an inert gas plays an important role in the field of semiconductor manufacture and other high-technology industries. Therefore, impurity analysis is becoming important for the quality control of gases. In general, for the determination of inherent gas impurities in industrial gases, gas chromatography (GC)has been widely used as the preferred method ( I ) . The photoionization detector (PID), which is based on the emission from a direct current discharge in helium gas ( 2 ) ,is the universal sensitive detector ( I , 2 ) and has especially high sensitivity for permanent gases. We reported in previous papers (3-5) that the PID is suitable for the determination of trace amounts of such gases. However, it has still been hard to separate and accurately determine trace amounts of impurities at less than parts per million levels in nitrogen: the use of a high-sensitivity
0003-2700/91/0363-1376$02.50/00 1991 American Chemical Society
