376
Anal. Chem. 1987,59, 376-379
(IO) Hasebe, K.; Kakizaki, T.; Yoshida, H. Fresenius’ Z . Anal. Chem. 1985, 322,486.
(11) Hasebe, K.; Kakizaki, T.; Tanaka, S.; Yoshlda. H. Nippon Kagaku Kaishi 1984, 557. (12) Bardina, S. M.; Chikryzova, E. G. Zh. Anal. Khim. 1978, 33,358. (13) Zaitsev, P. M.; Zaitseva. 2. V.; Zhdanov, S. I.; Nikolaeva, T. D. Zh. Anal. Khim. 1980, 35, 1951. (14) Hasebe, K.: Kakizaki, T.; Yoshida, H. Anal. Sci. 1985, 7 , 85. (15) Lahr, S.K.; Finklea, H. 0.; Schultz, F. A. J. Nectroanal. Chem. 1984,
763,237. (16) Progress in Inorganic Chemistry; Vicek, A. A,, Cotton, F. A,. Eds.; Interscience: New York, 1963; Vol. 5, p 228. (17) Parry, E. P.; Osteryoung, R. A. Anal. Chem. 1985, 37, 1634. (18) Hemmi, H.; Hasebe, K.; Ohzeki, K.; Karnbara, T. Talanta 1984, 37, 319. (19) Hasebe, K.; Yamamoto, Y.; Ohzeki, K.; Karnbara. T. Fresenius’ Z . Anal. Chem. 1986, 323,464.
(20) Dryhurst, G.; Kadish, K. M.; Schelier, F.; Renneberg, R. Biological Electrochemistry; Academic Press: New York, 1982; Vol. I, pp 116-179. (21) Qwttro, V. D.;Wybenga, D.; Studnitz, W. V.; Brunjes, S. J. Lab. Clin. Med. 1964, 63,864. (22) Dutrieu. J.; Delrnotte. Y. A. Fresenius’ Z . Anal. Chem. 1984, 377, 124. (23) Krstulovic, A. M.; Brown, P. R. Reversed-Phase High-Performance Liquid Chromatography; Wiley: New York, 1982; p 215. (24) Soldin, S.J.; Gilbert Hill, J. Clin. Chem. (Winston-Salem, N.C.)1980, 26,291. (25) Saito, M.; Kitarnura, M.; Niwa, M. Rinsho Kagaku Bunseki; Tokyo Kagaku Dojin: Tokyo, 1971: p 78.
RECEIVED for review April 25, 1986. Resubmitted August 7 , 1986. Accepted September 5 , 1986.
Open-Tubular Mixer for Gradient Preparatlon in Microbore High-Performance Liquid Chromatography Karel Slais’ and Roland W. Frei* Department of Analytical Chemistry, Free University, De Boelelaan 1083, 1081 HV Amsterdam, T h e Netherlands Microbore high-performance liquid chromatography (micro-HPLC) has become a very important separation technique because of its well-documented advantages (1-4). Similar to standard HPLC, gradient elution in micro-HPLC is a powerful technique for eluting solutes with a wide range of polarity, for improving solute detectability, and for increasing the speed of analysis ( 5 ) . However, the miniaturization of devices that can produce precise and accurate gradients is not a simple matter. A number of approaches to this problem have been suggested. One of the most straightforward is to use two high-pressure syringe pumps modified for supplying low flow rates (6-8). These gradient devices are capable of retention time reproducibility of 0.1-0.5 % relative standard deviation (% RSD) (6-8). The solvents delivered by two pumps should be mixed in a static or dynamic mixer if sensitive solute detection is needed (7-9). However, these gradient instruments are not readily available, and they are expensive ( 5 ) . Therefore, a considerable effort was made toward producing gradients for micro-HPLC in an indirect way by using a single high-pressure pump. Ishii et al., in one of the first papers dealing with microHPLC, used an open tube for storing a gradient prepared before the separation (10). In a more recent application of this technique, a retention time reproducibility of 0.9-1.3 % RSD was reported (11). Several authors have used microexponential diluters (12-15). In such a system, the gradient is formed in a stirred mixing vessel. After the initial solvent is introduced via a switching valve, the final solvent is allowed to dilute the initial solvent exponentially. Retention time reproducibilities are reported to be 1.8 (13) or 1-1.5% (14) RSD. A similar approach to gradient elution made use of a series of linked micro-exponential diluters (16, 17). Recently a mixing device consisting of two coaxial tubes was described (18). While the outer tube serves as a mixer and initial solvent container, the inner one has a number of small holes along its length. Thus the solvent of final eluent strength flowing through the inner tube enters the mixer at several places. This leads to a continuously increasing final solvent percentage in the liquid leaving the mixer. The retention time reproducibility for phenols using this device was in range 0.2-0.3% RSD. On leave from the Institute of Analytical Chemistry, Czechoslovak Academy of Sciences, 611 42 Brno, Czechoslovakia.
All the above-mentioned mixing devices suffer from low flexibility in obtaining varying gradient profiles since the geometrical configuration of the mixers must be altered to change the gradient shape. To achieve increased flexibility, a system generating “breakthrough gradients” was introduced (19). Such gradients are produced by eluting the interface between weak and strong eluent from packed “gradient generator” columns. The shape of the gradient can then be changed either by changing the column dimensions and/or the diameter of the packing particles or by altering the refill flow during preparation of the gradient. It was demonstrated that by decreasing the flow rate of the flushing cycle 8-fold, the gradient volume could be diminished to about 50% of its original value, when a 1% solution of acetone in water was mixed with pure water (19). In this paper we describe a device for the inexpensive preparation of continuous mobile phase gradients for micro-HPLC. The device is based on storing and mixing two solvents in an open tubular loop of suitable dimensions and shape. The loop is fitted to a six-port valve in the back-flush mode, so that it can be partially or completely filled by the initial solvent. The final solvent is delivered by a high-pressure syringe pump. With a change of the volume of initial solvent introduced, the length and steepness of the gradient can conveniently be varied. Technical parameters together with performance evaluation under micro-HPLC conditions are given.
THEORY Let us consider a tube of circular cross section, which is fitted in the back-flush mode to a six-port valve. While the valve is in the position shown in Figure 1,solvent A, delivered by the pump, moves through the tube from a to b. When the valve is switched, solvent B, which is miscible with solvent A, enters the tube at position b. The boundary between both solvents is thus moving from b to a. When the valve is switched to the original position, the boundary moves in the opposite direction (reversed flush) and leaves the tube at position b for the separation column. While traveling in both directions, the boundary width is continuously increasing. When the procedure described is used to generate boundaries of variable width, the large dependence of boundary width on solvent volume passing through the tube is advantageous. The volumetric variance, av2,for a sigmoidal profile of two very similar miscible solvents can be calculated in the same
C 1987 American Chemical Society 0003-2700/87/0359-0376$01.50/0
ANALYTICAL CHEMISTRY, VOL. 59, NO. 2, JANUARY 1987
possible to use other changes in tube geometry for bandwidth control, such as squeezing, twisting, waving (28),or zigzag shaping (27). A similar approach can be used for bandwidth control when packed beds, rather than open tubes, are used for mixing. Increasing the packing particle diameter from position b t o a will lead to a more drastic increase of a v / V than in the case of a packing with uniform particle diameter.
1 2
Figure 1. Diagram of mixing device: 1, spiral shaped tube: 2, coiled capillary; 3, six-port valve: 4, high-pressure syringe pump delivering final solvent; 5, initial solvent delivery; 6, output to detector or sampling device and microcolumn; 7, waste; a, b, ends of mixing loop.
way as that for a narrow band moving under the same conditions (20). In a straight tube the Taylor (21) equation holds with UAZ
=-
377
->
E( ro4LF
where ro is the inner radius of the tube, L is the length traveled by the boundary, F is the solvent flow rate, and D is the diffusion coefficient in the solvent. It follows from eq 1that the boundary volume standard deviation normalized per volume unit, uV/ V , depends on the square root of the length traveled by the boundary. An excessive tube length and a large isocratic delay would result in flat and broad boundary profiles. In short tubes (plate number less than 30, L < 0.4FID) the dependence of uV/ V on the volume passed is less sharp, as was demonstrated earlier (22,23).When the tube is coiled, uV/ V is decreased both for short (23)and long (24,25)tubes. The influence of coiling on uV/ V is shown in eq 2 (25),where
rl is the coil radius, 11 is the coefficient of dynamic viscosity, and p is the density. When assuming values typical for micm2/s, p = l g/mL, F = l hL/s, 7 = l cro-HPLC (D = CP = P a s) and tube parameters ro = 0.5 mm and rl = 30 mm it appears that coiling reduces the volumetric variance to about 76% of its value for straight tubes. At higher flow rates and coil curvatures eq 1 does not hold any more. However, it was shown (25-27) that the bandwidth reaches some maximum value and then decreases reciprocally with the flow. From the above it can be deduced that a v / V can be controlled effectively by varying both ro and r l , while keeping other variables constant. In the case of a straight tube, increasing ro along the tube will lead to a steeper dependence of uV/ V on the solvent volume passed through the tube than is the case for a tube of constant Po. Similarly, when considering a coiled tube under the conditions of high radial mixing, an increase of rl along the tube will lead to an even more pronounced dependence of av/ V on the solvent volume pumped through the tube than in the case of a coil of constant r1.
Thus, one can conclude that the tube should have a small ro and rl a t position b while at position a a larger ro and rl would be desired. Since a continuous change of ro is not easy to achieve, it is possible t o couple several tubes of different ro in series and to coil each piece with a different rl. It is also
EXPERIMENTAL SECTION Equipment and Materials. The device for gradient preparation consisted of two stainless steel capillaries 1and 2 (see Figure 1) coupled in one loop and mounted to a six-point valve (3) (Rheodyne, Cotati, CA) in the back-flush mode. Capillary 1 of 0.1 mm i.d. and 80 cm length was spirally coiled so that the coiling radius uniformly increased from 2 to 30 mm. The reproducible preparation of the mixing tubes was accomplished by winding the capillaries on a suitable cone and then shaping them into the final flat spiral form. Capillary 2 of 0.25 mm i.d. and length 200 cm was coiled with a constant diameter of 2.5 cm. Capillary 1 was used for solvent mixing and capillary 2 for adjusting the length of the initial isocratic portion of the gradient profile. Both capillaries were coupled by a zero dead volume Valco (Houston, TX) union. The homemade high-pressure syringe pump of 50-mL volume (4)was used to deliver solvent A (final solvent; highest elution power). Solvent B (initial solvent) ( 5 ) was delivered by a Gilson reciprocating pump (Villiersle Bel, France, Model 302). The flow rate of both solvents was checked by using a calibrated capillary and a stopwatch. The delivery time of liquid B was measured with the stopwatch while manually operating the Gilson pump. The homemade six-port sampling valve containing a 5-WLinner sampling loop was used to introduce the aqueous test mixture of chlorophenols. The microbore column used was a 100 X 1mm glass-lined stainless steel tube packed with Spherisorb ODS-2,5 wm (Phase Separation,Waddinxveen,The Netherlands). A Kratos Spectroflow 757 (Ramsey,NJ) equipped with 0.5-pL flow cell was used as the detector. It was adjusted to 220 nm for monitoring the gradient profile and to 280 nm for monitoring the chromatogram of the chlorophenols. Chromatograms were recorded on a BD 8 registration millivoltmeter (Kipp and Zoonen, Delft, The Netherlands). The solvents used were prepared from deionized water, HPLC quality methanol (Baker, Deventer, The Netherlands), sodium dihydrogen phosphate, phosphoric acid, and sodium nitrate (Baker Analyzed reagents). Buffer pH values were measured with a PW 9409 pH meter (Philips, Eindhoven, The Netherlands) before mixing with methanol. The chlorophenols used for preparing the test mixture were obtained from Aldrich (Beerse, Belgium). Operating the Mixing Device. The preparation of gradients and gradient chromatographic separations were carried out as follows: In the standby mode, the six-port valve (3) (see Figure 1)was switched to configuration “G”,so that solvent A delivered from the high-pressure syringe pump (4)was allowed to enter the loop (1and 2) position a. When the loop was completely filled by solvent A, valve 3 was switched to configuration “P”. In this configuration, liquid A bypassed the loop and flowed into the sampling device and column or directly to the detector. Simultaneously, the desired volume of solvent B was introduced to the loop at position b. Liquid B was delivered by the pump or by a syringe if the reproducibility was not critical. After the valve was switched again to the configuration G, solvent B soon appeared at output 6 of the valve. The percentage of solvent A in the liquid leaving the mixing device was then continuously decreased depending on the volume of solvent B introduced. The time gap between the appearance of solvent B and the beginning of the continuously increasing solvent A concentration was used for reconditioning the microcolumn and to inject the sample. When the gradient run was complete, valve 3 was switched to configuration P and the mixing loop refilled with the solvent B again. RESULTSAND DISCUSSION The influence of the flow rate of solvent B delivery on the gradient profile obtained was studied. The flow rate during the gradient run (valve 3 in position G) was kept constant at 1.4pL/s. The amount of solvent B introduced into the mixer was 0.3 mL.
378
ANALYTICAL CHEMISTRY, VOL. 59, NO. 2, JANUARY 1987
G
4
G
T
I *
i
'B
G
-I ~
2
0
6
4
8
t, [min] Figure 2. Gradient profiles obtained at different flow rates of solvent B delivery: A (final solvent), 0.02 M NaH,PO,, pH 3,in 80% methanol (v/v); B (initial solvent), as A, spiked with lO-'M NaNO,; Gradient flow rate, F = 1.4 pLls; Amount of B delivered, 0.3 mL; G, switch of gradient run; numbers on curves correspond to flow rates of solvent B delivery in mL/min. - -
c
/A
iii
'B
0,3 .-
0
2
4
6
8
10
12
14
16
18
20
tR\mlnJ
Figure 3. Dependence of gradient profile on amount of B introduced: delivery of B, 1 mL/min; gradient run, 1.0 pL/s; A, in Figure 3,ii and B in Figure 3i, see Figure 2; B, in Figure 3ii, 0.02 M NaH,PO,, pH 3, in 30% methanol (v/v) spiked with lo-' M NaO,; numbers on curves correspond to amount of solvent B introduced in mL.
The composition of both solvents A and B was the same except that solvent B was spiked with lo4 M NaN03to follow the profie by UV detection at 220 nm. The gradient profile was monitored by the detector directly connected to output 6 of the mixing device (see Figure 1). It can be seen from Figure 2 that even with a 50-fold change in flow rate of solvent B, the gradient profile obtained altered only slightly. This can be explained by the fact that the broadening of the boundary a t the conditions of the gradient run was close to its maximum value; see discussion of eq 2. Hence, the refilling step does not contribute significantly to the total broadening of the boundary profile. This situation permits the use of short times for the solvent B delivery and, thus, enables rapid preparation of the next gradient run. The dependence of the gradient shape on the volume of solvent B introduced into the mixer is shown in Figure 3. The flow rate of solvent B delivery, although not critical, was kept at 1 mL/min. The flow rate during the gradient run was 1.0 pL/s. The detector was connected to the device output 6. From Figure 3, it follows that gradient steepness and length vary considerably with the amount of solvent B delivered. When the volume of solvent B introduced is increased 5 times, the tangent in the central region of the gradient profile decreases by about 5 times (see Figure 3i). For other modifications in the steepness of the gradient, changes in the geometry of the mixing device can be used; see discussion of eq 2. Substantial differences in the composition and density of liquids A and B do not have a great influence on the gradient profiles obtained, as follows from a comparison of parts i and ii of Figure 3. This means, that the device suggested can be used for generating gradient profiles under the conditions generally used for micro-HPLC.
*
I
w
?I
1
V
0,5
---L'L-----t, [ min] Figure 4. Dependence of chromatographic profile on amount of B delivered: for condBions for gradient preparation and monitoring see Figure 3ii; chromatogram monitoring, microcolumn, 100 X 1 mm, packed with Spherisorb ODS-2,5 pm; UV detector, Kratos 757, adjusted to 280 nm; sample, 5-pL solution of chlorophenok in water; peak identification and concentration in sample (mg/L), (1) phenol (62), (2) Qchlorophenol(62), (3)2,4dichlorophenol(64), (4) 2,4,6-trlchlorophenoI (54),(5)2,3,4,5-tetrachlorophenol (184), (6) pentachlorophenol (370); G, gradient switch; I, sample injection. Numbers on chromatograms correspond to amount of B introduced in mL.
The performance of the proposed mixing device under typical micro-HPLC conditions was demonstrated with the gradient separation of chlorophenols. The apparatus used for micro-HPLC analysis was identical with that used for obtaining gradient profiles shown in Figure 3ii, except for inserting the sampling valve and microcolumn between the mixing device output and the detector.
Anal. Chem. 1987, 59,379-382
Table I. Retention Time ( t R )Reproducibility for Chlorophenols Separated in the Gradient Produced by an Open Tubular Mixera
tR (s) at the following peak no. 1
2
3
4
5
6
274.6 309.5 350.1 406.1 434 mean, n = 5 192 1.00 1.14 2.04 1.77 1.08 U 1.50 0.39 0.32 0.33 0.50 0.41 brei, % 0.78 "Gradient run flow rate 1.4 pL/s. Other conditions and peak identification as given for Figure 4. The influence of the gradient shape on the chromatograms is shown in Figure 4. It can be seen that the length of the chromatogram and the peak separation increases with increasing volume of solvent B introduced into the mixing device. Thus, the possibility for optimization of both resolution and analysis time is given. The reproducibility of a gradient profile is most conveniently evaluated by observing the retention time reproducibility of the components of some test mixture (5, 7, 8, 18). The gradient generator proposed in this paper was evaluated by using the conditions for the separation of chlorophenols by reverse-phase micro-HPLC presented in Figure 4iii. The flow rate during the gradient run was adjusted to 1.4 gL/s. The time gap between gradient switch (G) and sample injection (I) was always 90 s. The retention data obtained from five consecutive separations are summarized in Table I. It can be seen, that with the exception of the first peak, the RSD of the retention times varies between 0.3 and 0.5% RSD. The slightly higher variability of the first peak retention (0.78% RSD) can probably be explained by its higher sensitivity to flow rate changes caused by differences in the viscosities of solvents A and B. The reproducibility of the retention times can probably be further improved by electronically timing the delivery of solvent B. On the other hand, when the retention time reproducibility is not highly critical, solvent B can also be introduced by hand with a syringe. The small dependence of gradient profile on the flow rate of solvent B delivery (see Figure 2) can be an important advantage in such a case. Considering the solvent properties taken in the Theory, geometrical parameters presented in the Exerimental Section and the Poiseuille equation, it appears that for delivering 0.3 mL of solvent B within 0.3 min, a pressure of -0.35 atm is needed. Hence, the use of syringe or simple low-pressure pumps is also feasible from this point of view.
CONCLUSION The open tubular mixer proposed in this paper can effectively generate gradients for micro-HPLC. The duration and steepness of the gradient can be readily changed in a wide range by simply changing the volume of the initial solvent introduced in the mixing device. The design also has the potential for further minimization to fit packed or open ca-
379
pillary LC. Good reproducibility and simple control of gradient profiles render the proposed device suitable for automated chromatographic analysis. On the other hand, the simplicity of the design and its low cost are favorable for rapid optimization of chromatograms, even when operated manually and in a less well equipped laboratory.
ACKNOWLEDGMENT J. C. Gluckman is acknowledged for carefully reading the manuscript.
LITERATURE CITED Small Bore Liquid Chromatography Columns : Their Properties and Uses. Chemical Analysis, Scott, R. P. W., Ed.; Wiley: New York, 1984; Vol. 72. Novotny, M.; Ishii, D. "Microcolumn Separations"; J. Chromatogr. Libr., Vol. 30; Elsevier: Amsterdam, 1985. Sagallno, N., Jr.; Shih-Hsien, H.; Floyd, T. R.; Raglione, T. V.; Hartwick, R. A. J. Chromatogr. Sci. 1985, 2 3 , 238-246. Verzele, M.; Dewaele, C. I n "The Science of Chromatography", Brunner, F., Ed.; Chromatogr. Libr., Vol. 32; Elsevier: Amsterdam, 1985; pp 435-447. Schwartz, H. E.; Berry, V. V. LC Mag. 1985, 3, 1024 Scott, R. P. W.; Kucera, P. J. Chromatogr. 1979, 785, 27-41. Schwartz, H. E.; Karger, B. L.; Kucera, P. Anal. Chem. 1983, 5 5 , 152-1760. Schwartz, H. E.; Brownlee, R. Am. Lab. (Fairfield, Conn.) 1984, 161101. 43-58. Powley, C. R.; Howard, W. A.; Rogers, L. B. J. Chromatogr. 1984, 299, 4355. Ishii, D.; Asai, K.; Hibi, K.; Jonokuchi, T.; Nagaya, M. J. Chromatogr. 1977, 144, 157-168. Simpson, R. A.; Schachterle, S. D. "Design and Performance of Microbore System", presented at the Pittsburgh Conference and Exposition on Analytical Chemistry and Applied Spectroscopy, Atlantic City, NJ, 1984; abstract 650. Takeuchi, T.; Ishii, D. J. Chromatogr. 1982, 253, 41-47. Hirata, Y.; Nakata, F. J. Chromatogr. 1984. 294, 357-360. Karlsson, K. E.; Novotny, M. HRC CC,J. High Resolut. Chromatogr. Chromatogr. Commun. 1984, 7 , 411-413. Saito, M.; Wada, A.; Hibi, K.; Takahashi, M. Ind. Res. Dev. 1983, 2 5 ( 4 ) 102-106. Berry, V. V.; Takeuchi, T.; Ishii, D. 21st International Symposium "Advances in Chromatography", Oslo, Norway, 1985. Berry, V. V. Am. Lab. Fairfield, 1885, 77(10), 33-37, 39-41. Slais, K.; Preussler, V. HRC CC,J . High Resolut. Chromatogr. Chromatogr . Commun, in press. Berry, V. V.; Ishil, D.; Takeuchi, T. HRC CC,J. High Resolut. Chromatogr. Chromatogr . Commun . 1985, 8 , 659-664. Glddlngs, J. C. Dynamics of Chromatography,Part I ; Marcel Dekker: New York, 1965; p 87-88. Taylor, G. I.Proc. R . SOC.London A 1953, 219, 186-203. Golay, M. J. E.; Atwood, J. G. J. Chromatogr. 1979, 186, 353-370. Atwood, J. G.; Golay, M. J. E. J . Chromatogr. 1981, 278, 97-122. Golay, M. J. E. J. Chromatogr. 1979, 186, 341-351. Tijssen, R. Sep. Sci. Technol. 1978, 73, 681-722. Hofmann, K.; Halisz, I. J. Chromatogr. 1979, 773, 211-228. Katz, E. D.; Scott, R. P. W. J. Chromatogr. 1983, 268, 169-175. Hofmann, K.; Halisz, I. J. Chromatogr. 1980, 799, 3-22.
RECEIVEDfor review June 23,1986. Accepted September 8, 1986.
Determination of Trace Levels of Trimethylamine in Air by Gas Chromatography/Surface Ionization Organic Mass Spectrometry Toshihiro Fujii* and Toshihumi Kitai National Institute for Environmental Studies, Tsukuba, Zbaraki 305, J a p a n The Occurrence and determination of aliphatic amines have received a great deal of attention in recent years (I,2). These foul-smelling compounds have been found in a number of ambient environments (3-5) and become a source of serious social and psychological problems. They are also involved in nitrosamine synthesis in air (6),because methylamines react with NO, and 02.
The necessity of determining these compounds a t low levels in complex matrices has resulted in a number of analytical schemes (7-10).Usually concentration prior to identification is necessary. Solvent absorption (7)and adsorption on porous are well-known concentration techniques. polymers (3,8,9) The final analysis has been done with a flame ionization detector or a nitrogen-selective detector combined with gas
0003-2700/87/0359-0379$01.50/00 1987 American Chemical Society
