Anal. Chem. 1990, 62, 2453-2457
2453
Determination of (Methylcyclopentadieny1)manganese Tricarbonyl in Gasolines by Gas Chromatography with Flame Photometric Detection Walter A. Aue,* Brian Millier, and Xun-Yun Sun Department of Chemistry, Dalhousie University, Halifax, Nova Scotia B3H 4J3, Canada
The common gasoline additive (methyicyclopentadieny1)mangarwe Mcarbonyl (YMT) can be determined down to 0.6 ppm (w/w) levels In gasoline, by uslng atomlc emlsslon from a dmple, Inexpensive, and commercially available flame photometric detector. Other manganese compounds respond shdlar to MMT. auenddng caused by the carbon background is Insignllcant. The chemllumlnescence from manganese compares favorably wlth signals from compounds of sulfur, phosphows, and lead in gasoline matrlces under a varlety of spectral condltions. Selectlvlty ratios of three to six orders of magnltude can be easily obtalned for Mn vis-a-vis Pb, P, S, and C In duakhannel dmerentlai modes. Any element can be suppressed and such spectral tests can be used to ascertain the identity and purlty of the MMT peak.
INTRODUCTION (Methylcyclopentadieny1)manganesetricarbonyl (MMT), CH3C5H4Mn(C0)3,is frequently used as an antiknock agent in gasolines. Considerable literature exists about the determination of this (I) and other gasoline additives. The methodologies range from simple analyses for total manganese (e.g. ref 2) to sophisticated techniques of combining gas chromatography with emission or absorption plasma spectrophotometry. The most sensitive approach makes use of the hydrogen-atmosphere flame ionization detector, which can g/s of manganese (and 7.2 X detect 1.7 X g/s of lead), at selectivities against carbon compounds of 4 to 5 orders of magnitude ( 3 ) . Excellent performance has also been achieved with an atmospheric pressure helium microwave plasma emission system coupled to a gas chromatograph: the minimum detectable flow of Mn was 0.25 X g/s (that of P b 0.49 X lo-’* g/s), with selectivities larger than 6 orders of magnitude (4). Gas chromatographic separation may also precede atomic emission from a dc argon plasma (minimum detectable amount (MDA) = 3 ng of Mn) (5) or atomic absorption in a slotted quartz tube atomizer (MDA = 0.2 ng of Mn) (6). Note that minimum detectable amounts (or flows) are cited here solely to allow a limited comparison of different approaches. Detection limits are only one-and frequently not even the most important-of analytical performance criteria. Beyond analysis, technical and environmental health aspecta of manganese in general, and MMT in particular, have been covered well in the literature (e.g. refs 7 and 8). We were interested in selectively determining MMT in gasoline by using simple, inexpensive, and commercially available instrumentation. We were also interested in extending the analytical methodology to trace amounts of MMT and to other heteroorganics associated with gasolines. Since the flame photometric detector (FPD) is known to produce sensitive responses to a few transition metals-the latter in the form of volatile compounds amenable to GC-we wanted
* To whom correspondence should be directed. 0003-2700/90/0362-2453$02.50/0
to find out whether it could also detect compounds of manganese. In particular, we wanted to obtain the spectrum of manganese luminescence (if any) in the FPD, establish the best analytical conditions and calibration curves for the title compound in a gasoline matrix via single and dual channel modes, assess possible quenching effects by the carbon background, estimate the potential interference from other relevant species such as compounds of S, Pb, and P, increase the selectivity of Mn vs such elements, and develop spectral tests to probe the identity and purity of the MMT peak.
EXPERIMENTAL SECTION This study used a basic gas chromatographwith a dual-channel flame photometric detector (Shimadzu GC-4BMPF)and a short packed column (100 X 0.3 cm i.d. glass, 5% OV-101 on Chromosorb W, 100/120 mesh). The typical column temperature employed for MMT was 130 “C. Injection port and detector base temperatures were set at least 30’ higher (between 160 and 200 O C ) than that of the column. The nitrogen flow through the column was usually 27 mL/min, with an additional 20 mL/min of nitrogen introduced to the detector via the air supply l i e . The detector was run without the conventional quartz chimney but with an efficient exhaust duct situated above the flame gas outlet. The luminescence was monitored by two Hamamatsu R-268 photomultiplier tubes (bialkali, 300-650 nm, maximum response at 420 nm). Under “regular conditions” (vide infra), the hydrogen supply to the detector was 300 mL/min; the air supply was 80 mL/min. To obtain the spectrum of the manganese luminescence, one of the photomultiplier channels of the FPD was replaced by a Jarrel-Ash Model 82-415quarter-meter monochromator with a 1180 grooves/“ grating blazed for 500 nm, a pair of 2.1-mm slits, and the same Hamamatsu R-268 photomultiplier. An acetone solution of MMT was repeatedly injected and the wavelength setting manually advanced (thereby eliminating possible background emissions). Before the manganese spectrum was measured, response to MMT had been optimized by varying hydrogen and air flows in an “open” mode, i.e. without wavelength selection save for the response profile of the photomultiplier tube. After the manganese. spectrum became available, response was again optimized in a two-channel configuration with 405- and 540-nm three-cavity interference filters (Ditric), respectively. (Note that a 405-nm filter was used to monitor the 403-nm lines.) The spectral characteristics of these and other interference filterswere checked in the lab, and central wavelengths listed refer to these measurements. Unless otherwise noted, band-passes were in the 9-12 nm range. For differential flame photometry, the signals from the two channels were combined after amplification by the respective electrometers. The interface circuit shown in Figure 1 was used for matching signal intensities (of gasoline matrix components) and for reduction of high-frequencynoise before feeding the signal to a Linear dual-channel 1 mV chart recorder. To simulate the potential quenching effect of a hydrocarbon matrix on manganese response, methane as a quencher was introduced via an exponential dilution flask into the hydrogen supply line (whichjoins the column effluent at the detector base, long before the latter reaches the flame). In accord with photometric custom, the quencher concentration was calculated on 0 1990 American Chemlcal Society
2454
ANALYTICAL CHEMISTRY, VOL. 62, NO. 22, NOVEMBER 15, 1990
electrometer #1 R
e
500 ohms
elecirometer # 2
C 0
--
R D
R
100
180‘1:
10t/min
Figure 3. Temprature-programmedchromatography of manganese compounds: (1) 1 ng of cyclopentadienylmangnesetricarbonyl (cymantrene);(2) 1 ng of (methylcycbpentadhyl)maenyl)mangenesetrlcarbonyl (MMT);(3a,3b) 1.5 ng of “(pentamethylcyclopentadienyl)manganese
A
tricarbonyl” as received.
min.
Figure 2. Peak from 0.20 ng of CH&5H,Mn(CO)3: 600 mL/min; single channel at 405 nm.
I
hydrogen flow ca.
w
the total (hydrogen plus air plus nitrogen) flow through the detector. The manganese compounds were obtained from Strem Chemical, Newburyport, MA. They were kept in the freezer and used without further purification. Other test compounds were of standard laboratory quality.
RESULTS AND DISCUSSION Even the fist experiments clearly established manganese as belonging to the roster of elements that produce analytically useful responses in the FPD. After some optimization, the minimum detectable amount of (methylcyclopentadieny1)manganese tricarbonyl was approximately 5 X mol/s at SIN = 2 (or about 2 X lo-“ mol/s according to the more lenient S l a = 3 IUPAC definition) at the ”regular conditions” given in the Experimental Section. If the hydrogen flow is doubled, this MDA improves by about a factor of 2; if a differential mode is used, it worsens slightly. Figure 2 shows a chromatogram of 0.20 ng of MMT at a high hydrogen flow rate to demonstrate sensitivity; Figure 3 shows a separation of (nominally) three manganese compounds under “regular” conditions and a t somewhat higher concentrations to demonstrate the applicability to other Mn-containing analytes. (Note that one of the compounds, cymantrene, is commonly used as an internal standard for the gas chromatographic determination of MMT. However, in our case MMT was analytically so well behaved that an internal standard was considered unnecessary.) The spectral distribution of the luminescence originating from MMT, which was obtained by repeatedly injecting the analyte at different wavelengths, is shown in Figure 4. The emission at 403 nm corresponds to the two strongest arc lines of atomic manganese, found a t 4030.76 and 4033.07 A (also to the strong line at 4034.49 A). The emission at 540 nm may represent one or both of the 5394.67- and 5432.55-A lines. While the latter are hardly visible in an arc, they share with the 403-nm lines the distinction of being transitions to the ground state (9). Transitions to the ground state are favored in the FPD as compared to the high-temperature emission sources, presumably because of the lower and closer defined energy of a chemiluminescent process (cf. ref 10). Our presumption of a chemiluminescent (as opposed to a thermal) process rests on the fact that, while some transition elements
,
.
I
300
,
,
I
I
/
*
I
400
I
.
I
I
I
I
/
!
?
,
500
,
I
I
I
I
>
*
/
C
b
/
I
I
I ,
600
WAMLMGM, nm Figure 4. Spectrum obtained by repeatedly injecting 2 pg of CH&H,Mn(CO), under “regular” conditions (see Experimental Section): band-pass ca. 6.7 nm.
like
manganese respond with remarkable-and unpredictable-sensitivity in the cool flame of the FPD, others do not respond a t all. That is consistent with chemiluminescence but inconsistent with thermal excitation. The assumption that the 540-nm emission is atomic in origin is also supported by the fact that the 405- and 540-nm channels kept showing similar behavior under the wide variation of hydrogen and air flows employed in optimization runs. For analytical purposes-such as improving the signal-to-noise ratio and discriminatingagainst other FPD-active elements-the combined 403-nm lines offer the best performance (as they also do in conventional, high-temperature atomic emission from oxyhydrogen and oxyacetylene flames). Figure 5 shows calibration curves measured at 405 nm for MMT and dodecane in acetone. Also included is a calibration curve measured at 540 nm (the second strong Mn emission) and one that describes the differential response from a gwline sample. (”Differential” means in this case that the response measured at 448 nm-after adjustment of amplitude for best overall suppression of gasoline constituents-is electronically deducted from the response at 405 nm, thereby yielding a single recorder trace.) The differential calibration curves from acetone and gasoline solutions turn out to be virtually identical. The response selectivity for the manganese compound against a carbon standard (dodecane) on the single 405-nm channel is about 5 X lo3 (w/w) (or 7 X lo4 on a (mol/s C)/ (mol/s Mn) basis) as determined from the minimum detectable amounts. This selectivity ratio improves slightly if
ANALYTICAL CHEMISTRY, VOL. 62, NO. 22, NOVEMBER 15, 1990
2455
20
E Cp
1.5
1.0 0
1000
2000 3000 PPm Flgwe 6. Stern-Volmer plot, showing the quenching of luminescence from CH,C,H,Mn(CO), and dcidecane, by methane present In the detector gases: R,, response in the absence of quencher; R , response in the presence of quencher; Q, concentration of quencher in ppm (vlv) of total detector gases. QJ
AMOUNT INJECTED, gram Flgure 5. Calibration curves: 0, CH,CSH,Mn(C0), (MMT) in acetone, single channel, 405 nm; 0,MMT in gasoline, single channel, 405 nm; A, MMT in gasoline, dual differential channel (405-448 nm, carbon suppressed); 0, MMT in acetone, single channel, 540 nm; 0,dodecane in acetone, single channel, 405 nm; A, dodecane in acetone, dual differential channel (405-448 nm, carbon suppressed).
the hydrogen flow is increased and drastically if two channels are operated in the differential, carbon-suppressing mode. Nominally higher selectivities than the one shown in Figure 5-Le. higher than 1.5 X lo5 (w/w) (or 2 X lo6 (mol/s C)/ (mol/s Mn))-can be easily achieved by exactly matching the response of the two channels for the very carbon standard used in the selectivity measurement. However, since different carbon compounds can differ in spectra and response (e.g. aromatics > aliphatics), we maintained the general setting chosen earlier for the best overall suppression of gasoline matrix peaks. The linear range for the manganese compound is 2 X lo4 (4.3 orders of magnitude). This value does not vary significantly among single and double channel modes or among samples dissolved in gasolines and in acetone. The minimum detectable concentration-at SIN = 2, with a 0.5-wL injection and without analyte enrichment-is about 0.6 ppm (weight of MMT/weight of gasoline). This lower limit could likely be improved by adding a preconcentration step or by using a capillary column. However, such seems hardly necessary for routine determination of the much higher MMT levels found in typical gasoline samples. It is well-known-for instance from the determination of sulfur-containing compounds in fossil fuels-that the lowtemperature chemiluminescence characteristic of the FPD can be severely quenched by coeluting peaks. We therefore ran quenching curves for MMT and a standard carbon compound, by introducing varying levels of methane (as the quencher) through an exponential dilution flask. The result is presented in the form of Stern-Volmer plots in Figure 6. Manganese response is less affected by the background presence of carbon (in methane) than is the response of the carbon standard itself. More importantly, it takes about 3600 ppm (v/v) of methane in the detector atmosphere (which corresponds to 65000 ppm if calculated on the nitrogen flowing through the column) to reduce the manganese peak by half. Thus, unless the manganese peak is subject to severe overlap by a very large carbon peak, interference originating from luminescence quenching can be considered negligible. There are, however, other types of spectral interference that could assume importance in the flame photometric detection of manganese in gasoline (if not circumvented by chromatographic separation). Most importantly, the atomic emission of Mn at 403 nm is located very close to one of the major bands of Sp The emission from lead stretches over that region as well. The HPO bands are far away but, for the analysis of
Table I. Molar Selectivity Ratio for Manganese' filter none 405 405/448* 405/36P
MnvsC
MnvsP
MnvsPb
9.7 x 7.1 x 2.3 X 1.7 X
0.6 2.2 x io2 1.7 X lo3 1.0 X lo2
7.4 2.0 x io2 3.0 X lo2 1.6 X lo2
103 104 lo6 lo'
Formally calculated as the minimum detectable molar amount [signal/(peak-to-peak noise) = 21 of interfering element (C, P, or Pb) per second, divided by the minimum detectable moles of Mn per second. bTuned for suppression of dodecane. cTuned for suppression of di-n-butyl disulfide. Compounds used Mn, (methylcyclopentadieny1)manganese tricarbonyl; C, dodecane; P, triethyl phosphate; Pb, tetraethyllead.
an automotive product, it still would seem prudent to characterize the selectivity against phosphorus. Besides, the initial work on the FPD (11)involved S and P compounds and these are often best determined in this manner. Finally, the luminescence from carbon compounds (as opposed to their quenching effect on Mn emission) is relatively strong at 403 nm. Table I lists the molar selectivity ratios for manganese versus carbon, phosphorus, and lead, under three increasingly selective spectral conditions (plus one under sulfur suppression). The manganese determination never reaches specificity (infinite selectivity) vis-a-vis a hydrocarbon, but this is mainly so because very large amounts of any compound do disturb the weak flame and hence the baseline. Sulfur is even more likely than C, P, and P b to interfere in the determination of Mn, but its selectivity is more difficult to characterize because of the quadratic nature of its S2-based response and its susceptibility to detector contamination (sulfurresidues in the detedor increase analyte sulfur response and linearize the calibration curve in the lower concentration range). The advantage of switching from single to dual differential channel operation for sulfur suppression is therefore illustrated here by chromatography of the same mixture. Figure 7 shows the approximately 50-fold gain in selectivity of Mn vs S, which could be further improved by fine-tuning. How far a particular selectivity is to be increased depends on the type of sample to be analyzed and the amount of MMT expected. That can vary all the way from a high-octane fuel to an extract of contaminated soil. Samples of premium unleaded gasolines bought domestically from different companies in Halifax could, in fact, be analyzed for MMT (after a 1 0 1 dilution with hexanes) without having to resort to spectral discrimination at all, i.e. without the aid of an interference filter. The left chromatogram of Figure 8 shows such an analysis. For comparison, the middle and right-side
2456
ANALYTICAL CHEMISTRY, VOL. 62, NO. 22, NOVEMBER 15, 1990
n l
Mn
i:l:
JJ
3'0 mtmn dot 310"Clminot Figure 7. Chromatography of 10-fold diluted, unleaded gasdine doped with a sulfur compound: (left) 405-nm filter, attn. 16, Mn peak height 126 PA; (right) 4051365-nm dlfferential filters, attn. 2 (S-suppressed); Mn, 1.5 ng of MMT; S, 250 ng of di-feri-butyl disulfide.
MMT
MMT I
S
,
33
4
'9th"
1
150% 30
1GWn 7
1501:
iotimin
15&
Figwe B. Tempwature-progran?med ctwmtowaphy of (old)rnleaded gasoline containing 3 ppm (w1w) of MMT: upper left, open (no interference filter); upper right, 405-nm filter (manganese); lower left, 4051365 nm differential (sulfur suppressed); lower right, 405/448 nm differential(carbon suppressed). Mn peak helghts in same sequence from upper left to lower right: 78.7, 47.2, 55.1, and 50.4 PA.
Figwe 8. Temperature-programmed chromatography of premium unleaded gasoline: (left) open (no M e r e n c e W, higher attenuation), MMT peak height 227 PA; (mlddle) 405-nm fitter (manganese); (right)
MMT
i
4051448 nm differential (carbon suppressed), MMT peak 189 PA.
chromatogram in Figure 8 show the same sample (at lower attenuation) as viewed through a 405-nm single channel and 405/448dual differential channels, respectively. The resulting increase in selectivity is large but, while important for trace work, is not really needed for the routine determination of MMT in gasolines. The amount of manganese as MMT in this sample is, incidentally, 7.0 mg of Mn/L and thus at about half the legal upper limit. The permissible concentration of Mn is 18 mg/L (approximately 80 ppm MMT by weight). To demonstrate the determination of a much lower concentration, 3 ppm of MMT was added to an old (practically MMT-free) unleaded gasoline. The resulting chromatograms, shown in Figure 9, provide a graphic demonstration how the 3 ppm peak of MMT, barely credible in open mode (upper left), becomes clearer in the Mn single-channel mode (upper right) and, save for a sulfurcontaining matrix component, dominates the chromatogram in carbon-suppressed dual-channel differential mode (lower right). The sulfur-suppressed differential mode (lower left) establishes that the large peak marked "S" did indeed arise from a sulfur compound. A variety of different filters could be used to suppress sulfur efficiently; however, it should be noted that the extent to which carbon response is simultaneously changed does vary in accordance with the chosen wavelength. Although a positive peak identification is hardly de rigeur under routine circumstances, cases may arise where the identity and/or homogeneity of the MMT peak is truly in
-
30
100Clmin
l5O&
1
I
30
10OClmin
150&
Flgure 10. Temperature-programmed chromatography of premium unleaded gasoline contalning CB. 80 ppm (w/w) of MMT: (left) 405-nm fllter, MMT peak helght 1.26 nA; (right) 4051540 nm differential (MMT suppressed).
doubt. One way of narrowing down the alternatives has been described above: it relies on diminishing or canceling potentially interfering heteroelements while keeping the Mn response intact. The opposite approach is also possible: it depends on canceling the Mn response itself: The disappearance of the suspected MMT peak then establishes its nature as a manganese compound. This is demonstrated in Figure 10 with a regular chromatogram on the left and a Mn-suppressed one on the right (in the latter case, carbon
Anal. Chem. 1990, 62, 2457-2460
response is strong and inverted). Note that any element that should happen to have the same response ratio as manganese at the two wavelengths could theoretically produce the same effect; however, the probability of this occurring is remote. Beyond MMT in gasolines, such dual-channel differential methodologies can be used to analytical advantage in a wide variety of samples containing FPD-active elements.
2457
(7) Hob, M. C. B. ARernathres to Lead in Gasoline; The Royal Society of Canada, Commission on Lead in the Environment, Minister of Supply and Services Canada: Ottawa, Ontario, 1988. (8) kngenese; EnvkonmentalHealth Crlterla 17; World Health Organhation: Geneva, Switzerland, 1981. (9) Meggers, W. F.; Corliss, C. H.; Scribner, B. F. Tabks of Spectfal-Line Intensities, 2nd ed.; NBS Monograph 145; U.S. Government Printing Office: Washington, DC, 1975. (IO) Sun, X.-Y.; Aue, W. A. J . Chromtogr. 1989, 467, 75. (11) Brcdy, S.S.;Chaney, J. E. J . Gas Chromtogr. 1988, 4 , 42.
LITERATURE CITED (1) Crompton, T. R. comprehensive Crgmometallic Analysis; Plenum Press: New York, 1907; pp 418-423, 763-785. (2) Smith, G. W.: Palmby, A. K. And. Chem. 1959, 31, 1798. (3) DuPuls, M. D.; Hill, H. H., Jr. Anal. Chem. 1979, 57, 292. (4) Qulmby, B. D.; Uden. P. C.; Barnes. R. M. Anal. Chem. 1978, 50, 2112. (5) M e n , P. C.; Barnes. R. M.; DiSanro. F. P. Anal. Chem. 1978, 50, 852. (6) Coe, M.; Cruz, R.; vanloon, J. C. Anal. Chim. Acta 1980, 120, 171.
RECEIVED for review May 9,1990. Accepted August 13,1990. This study was supported by NSERC Operating Grant A9604. Excerpts from this study were presented at the 72nd CIC Conference (Victoria, BC, June 1989) and the 20th Ohio Valley Chromatography Symposium (Hueston Woods, OH, June 1989). This material forms part of the doctoral thesis requirements of X.Y.S.
Indirect Inductively Coupled Plasma Atomic Emission Determination of Fluoride in Water Samples by Flow Injection Solvent Extraction Jamshid L. Manzoori' and Akira Miyazaki* National Research Institute for Pollution and Resources, 16-3 Onogawa, Tsukuba, Zbaraki 305,Japan
An indirect determination of fluorlde in water by Inductively
coupled plasma atomic emissbn spectrometry combined with flow InJectlon coupled with solvent extraction Is reported in this paper. A manlfokl for rapkl determination of fluoride has been designed that uses a single coil for complex formation and extraction. The method involves the formatlon of ianthanum/akarln compkxone/tluorlde complex and its extraction Into h e x a d contahrlng N,N-dlethyianHlne. The concentration of fluorlde is determined indirectly by Introduction of the organic layer Into the plasma and measurement of the emlsrlon intensity of the La I1 333.75-nm line. The optknwn experimental conditions for the determination are described. A coiled groove phase separator fitted with a grid and PTFE porous membrane was used In this work. The sampling rate was 36 samples per hour and the caiibratlon graphs were bear from 0.09 to 1.3 WmL. The relatlve standard devlatkn found was 2.16% for 290 pL of 1 pg/mL of fluorlde. The method Is selective and has been applied satisfactorily to the determination of fluoride In water samples.
INTRODUCTION In recent years there has been a growing trend toward the application of flow injection analysis (FIA) to the determination of fluoride. FIA systems for fluoride determination by ion-selective electrodes (1-31,spectrophotometry with cerium or lanthanum alizarin complexone (La-AC) (4,5), and microwave induced plasma (6)have been developed. Among these methods ion-selective electrodes in FIA offer remarkable detection limits, but their applicationto real samples is limited because of interference effects. Other methods lack sensitivity Present address: Department of Chemistry, Faculty of Science, University of Tabriz, Tabriz, Iran. 0003-2700/90/0362-2457$02.50/0
and simplicity; thus further efforts to develop simple and sensitive methods appear to be worthwhile. Since the introduction of liquid-liquid extraction based on the flow injection principle by Karlberg and Thelander (7) and Bergamin et al. (8)several papers have been published on this topic (9-12),but its application to inductively coupled plasma atomic emission spectroscopy (ICP-AES) has been limited and few papers concerning solvent extraction in FIA with ICP-AES have appeared in the literature (13,14). The direct determination of fluoride by ICP-AES is difficult since its resonance line appears at 95.5 nm; consequentlywork with fluoride has utilized nonresonance lines, which leads to poor detection limits. However, the application of indirect methods seems to be promising. Miyazaki and Bansho (15) have proposed an indirect method for the determination of fluoride based on ICP-AES measurement of La-AC/fluoride complex (La-AC-F). We recently described a useful extension of FIA combined with ICP-AES to the determination of total phosphorus and phosphate (16). The aim of this work is to develop a flow injection solvent extraction combined with ICP-AES for the determination of fluoride as an alternative to the manual method reported by Miyazaki and Bansho (15).The method involves the formation of La-AC-F and extraction into hexanol containing N,N-diethylaniline and determination of La I1 333.75-nm emission in the organic layer by ICP-AES.
EXPERIMENTAL SECTION Apparatus. A FIA Star 5020 Analyzer (Tecator,Sweden) with two pumps and a FIA Star 5102-001 variable volume injector was used. A schematic diagram of the flow injection system is shown in Figure 1. Aqueous solutions were introduced into the system by peristaltic pumps fitted with Tygon pump tubes. A double plunger micropump (NP DX-2, Nihon Seimitsu Kagaku, Japan) was used for the organic solution. A T-segmentor (Sanuki Kogyo Ltd., Japan) in which the aqueous phase flows straight through and organic phase at right angles was used for mixing organic and 0 1990 American Chemical Society
