2390
Anal. Chem. 1982,5 4 , 2390-2392 212
BEFORE
7.46
235
AFTER
~,
’
of the problem, and we concluded that the GC/MS isolation valve was. In fact, in an early attempt to solve this problem we dismantled the instrument and cleaned the valve. Afterward, we were able to detect acridine, which had previously been undetectable. Following the modification of the interface region, as described above, the same four-component standard produced the total ion chromatogram shown in Figure 2 “after”. All four compounds were successfully eluted into the mass spectrometer ion source with increased sensitivity and with no chemical alteration (see Figure 3 “after”).
282
237
I65
I = I
R
L
I
m/e Flgure 3. Mass spectra DDT observed before and affer modification of the HP 5995. Only the after spectrum is correct.
2 “before” and “after”). Even real time monitoring of the major chlorine cluster ion of m / e = 264 failed to produce a peak. Even more alarming, the mass spectrum of the last peak (see Figure 3 “before”)was not that of DDT as it should have been. It was, in fact, the combined mass spectra of 1,l’-(dichloroethenylidene)bis[4-chlorobenzene] (mol wt = 316) (mol (DDE) and l,l’-(chloroethylidene)bis[4-chlorobenzene] wt = 282). Since we could obtain a correct mass spectrum of DDT when it was introduced through the direct probe, it appeared that the GC/MS interface device was dehydrochlorinating this compound. Thus, the chemical activity of the interface presented real concerns about the use of this instrument. We suspected that the GC/MS isolation valve was responsible for these losses and compound alteration for three reasons: (a) When the four test compounds were introduced into the mass spectrometer via the direct probe the correct spectra with good sensitivity were obtained. (b) The same incorrect DDT spectrum and total ion chromatogram were obtained using the jet separator and an open-split interface (3). (c) All four of these compounds were correctly detected with the H P 5985B. Both the H P 5985B and H P 5995 had the same type of injector and column. Therefore, the injector, column, and mass spectrometer could not have been the source
CONCLUSIONS We have described our successful modification of H P 5985B, H P 5982, and H P 5995 GC/MS units for direct introduction of fused silica columns in close proximity to the ionizing beam in the ion source of the instrument. For the H P 5985B and H P 5982, the modification increased transmission of compounds with long GC retention times. Not only was sensitivity improved for the H P 5995 but, more importantly, the chemical reactivity and selectivity of the isolation valve were eliminated by its removal. In all three cases, we are confident that our approach was the best available solution. Though the manufacturer’s designed flexibility of use (packed or capillary columns) and ion source vacuum isolation have been sacrificed, these losses have been inconsequential in view of substantial gains in data quality and sensitivity. (Sketches of the hardware modification are available on request to R.A.H.) ACKNOWLEDGMENT We thank the Hewlett-Packard Gorp. for the gift of their 5995 GC/MS system, John Dorsett and the staff of the Chemistry Department Machine Shop for skilled craftsmanship, and Walter E. Reed (of UCLA) for helpful discussions. LITERATURE CITED (1) Watson, J. T.; Blemann, K. Anal. Chem. 1965, 37, 844. (2) Ryhage, R. Anal. Chem. 1964, 36, 759. (3) Kenyon, C. N.; Goodley, P. C. Presented at 29th Conference on Mass Spectrometry and Allied Topics; Minneapolis, MN, May 24-29, 1981.
RECEIVED for review May 21, 1982. Accepted July 23, 1982. The U.S. Department of Energy (Grant No. 80 EV-10449) and the U S . Environmental Protection Agency (Grant No. R808865) supported this work.
Automated System for Solvent Extraction Kinetic Studies Hltoshl Wataral,’ Larry Cunnlngham, and Henry Frelser” Department of Chemistry, University of Arizona, Tucson, Arizona 8572 1
Growing attention to the study of solvent extraction kinetics requires increasingly more convenient apparatus for extraction rate measurements. The apparatus reported previously (1) by our laboratory has given less equivocal data than has been obtained by other methods (2) used for extraction kinetic studies such as Lewis cell (3),falling drop method (4),in-liquid ejection method (5), and the AKUFVE apparatus (6),as a consequence of the highly efficient two-phase mixing necessary for rapid mass transfer between two phases. The apparatus described here is a further significant improvement incorporating continuous monitoring of the rate of extraction and essentially instantaneous data analysis, which are accomPresent address: Department of Chemistry, Faculty of Education, Akita University, Akita, Japan.
plished by introduction of a Teflon phase separator and an on-line minicomputer. A chemical system used to illustrate the efficacy of the automated system is the extraction of Ni(I1) with dithizone (dithiophenylcarbazone), which was studied earlier in this laboratory (7).
EXPERIMENTAL SECTION The schematic diagram of the extraction kinetic apparatus is shown in Figure 1. The extraction vessel is a 200-mL Morton flask fitted with a high speed stirrer (0-20 000 rpm) (Cole-Palmer Instrument Co.) and a Teflon phase separator. Stirrer blades are made of Teflon and the stainless steel stirrer shaft is covered with irradiated polyolefin tubing t o prevent contamination from the underlying stainless steel. The phase separator consists of a bored Teflon cylinder (9 mm 0.d. and 24 mm in length) wrapped with
0003-2700/82/0354-2390$01.25/00 1982 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 54, NO. 13, NOVEMBER 1982
.Ol
2391
.
.oo
? 0.0
100
200
TJME
300
400
ISECONL151
r
, -
Figure 1. Diagram of the computer-asslsted extractlon kinetics measurement system: (A) high-speed stirrer, (B) stirrer shaft, (C) sample Inlet, (D) Teflon stirring bar, (E) Teflon phase separator, (F) water bath, (G) flow cell, (H) spectrophotometer, (I) peristaltlc pump, (J) chart recorder, (K) 10 be IUD converter, (L) clock, (M) minicomputer, (N) dual-floppy disk drive, (0)printer, (P) digital plotter.
a specially made Teflon cloth (Kokuboseiki Co., Tokyo, Japan), capable of selectively filtering water-immisciblesolvents from an intimate admixture with an aqueous phase. The organic phase, once removed by the separator, is introduced to the flow cell (50 pL) attached to the Gilford Model 2400 spectrophotometer and returned to the vessel by the aid of a Gilson Minipuls I1 peristaltic pump. All tubing is made of Teflon and Acidiflex. The total liquid volume of the tubing system is 1.26 mL; the volume between the phase separator and the flow cell center is 0.65 mL. Since the flow rate used is 5.0 mL/min, the time delay alter phase separation was estimated to be of the order of 8 s. This is no problem since pseudo-first-order conditions can be maintained. The spectrophotometer is interfaced to a Data General Corp. Nova 2/10 minicomputer equipped with 32K core memory and a Xebec Systems, Inc., dual-floppy disk drive. Other electronics required, as well as the clock which allows wide selection of the data-acquisition rate, are constructed with an ADD 8000 ana.. log/digital design module (E & L Instruments, Inc., Derby, CT). A Date1 EK10-B integrating analog-to-digital converter has sufficient speed and resolution for all measurements (maximum conversion time of 6 ms with 10 bit resolution). Data so obtained are then stored on disk and later displayed graphically (Figure 2). The software provideei corrections for base line drift and for grossly high absorbance readings which occur when occasional air bubbles pass through the flow cell. The rate constant calculation is based on several hundred experimental points and is made over a t least two half-lives to allow use in a variety of' chemical systems. Solvent absorbance values are either requested by the program or measured by the system upon completion of the experiment. All programming was done in Fortran IV and assembly language (copieo will be provided upon request). Since absorbance readings are made on the organic phase, the course of the reaction can be followed either by a drop in the extractant concentration or a rise in the metal chelate concentration in this phase. The ease with which a kinetics study could be performed is illustrated in the following example. Fifty milliliters of dithizone in chloroform was added to the Morton flask and pumped through the system until a stable absorbance reading was observed. Then, 50 mL of an aqueous nicker solution was carefully added to avoid any premature mixing of the two phases. The extraction was started with the high-speed stirrer, and after a delay time of 8 s the A/D was read by the computer, usually at a frequency of 1 Hz. The progress of the experiment was observed on a chart recorder and could be terminated manually when the system reached equilibrium.
RESULTS AND DISCUSSION Absorbance data enabled calculation of the pseudo-firstorder extraction rate coneitant, kobsd (&I, using the following equation with least-squares fitting: (1) In ((AF - A & / ( A F- At) = h o b s d t
U
2.5
LU L \
2.0
.
1.5
.
U 0 I
LU L
-
I
-=
.50
~
0.0
50
TIME
100
150
15ECOND51
Figure 2. Typical result of kinetics determination. Upper figure is absorbance vs. tlme. Lower figure is first-order rate law plot which gave a kobsdvalue of (1.419 f 0.008) X lo-* s, pH 7.46.
Table I. Kinetic Data for the Extraction Dithizone (at 25 "C)' half-life, run pH S log kobsd 60.0 -1.937 1 7.54 81.3 -2.069 2 7.54 -2.046 77.1 3 7.55 57.9 -1.922 4 7.40 42.2 -1.785 5 7.51 73.6 -2.026 6 7.35 55.4 -1.903 7 7.46 67.9 -1.991 8 7.46 45.9 -1.821 9 7.61
of Nickel by
1ogklb 3.73 3.60 3.61 3.89 3.91 3.83 3.85 3.76 3.78
mean lit.
3.77 j: 0.11 3.78 ( 5 ) M in CHCl,, [Ni2+]= 1.0 X a [Dz]= (4.3-8.5) X M. log k , = log kobsd - log [Ni2+]- pH - log ( K . / K ~ 1.R
where Ao, At, and AF correspond to absorbance values at t = 0, at time t , and a t equilibrium, respectively. A typical result is shown in Figure 2, where a simple exponential increase in absorbance a t 740 nm, Le., increase in nickel dithizonate concentration in the organic phase, was continuously monitored. A first-order rate law as applied to these data resulted in excellent linearity over two half-lives (correlation coefficient of 0.998). The rate-determining step in the nickel extraction with dithizone has been established to be the formation of 1:l chelate in the aqueous phase, k l . As seen in Table I, the average value for log k, obtained in the present study agreed very well with that reported previously by our laboratory and demonstrates the consistency of the two different methods (5). The real lag time in this apparatus is determined by the time required for phase separation. The volume of 0.27 mL inside the phase separator and the flow rate of 5 mL/rnin suggests that the phase separation time should be less than 3 s. Thus, extraction rates with half-lives as short as 10 s are expected to be easily measured. Indeed, systems having 20-s half-lives with a zinc di(n-butylpheny1)thiocarbazone system
Anal. Chem. 1982, 5 4 , 2392-2393
2392
were measured successfully with this apparatus. A dramatic saving in time as well as significant upgrading in the quality of the kinetic data is made possible using this apparatus. In addition, the system permits the examination of two liquid phase systems under conditions where a larger interfacial area is generated.
ACKNOWLEDGMENT The authors wish to thank N. Suzuki for helpful scientific suggestions and S. Tamura, who invented the Teflon phase separator.
LITERATURE CITED (1) Carter, S. P.; Frelser, H. Anal. Chem. 1979, 5 7 , 1100. (2) Tavlarldes, L. L. Cbem. Eng. Commun. 1981, 8, 133. (3) Fleming, C. A. Rep.-Nafl. Inst. Metall. ( S . Afr.) 1976, No. 7793. (4) Hanson, C.; Whewell, R. J.; Hughes, M. A. J. Inorg. Nucl. Cbem. 1975, 37, 2303.
(5) Watarai, H.; Suzuki, N. Inorg. Nucl. Chem. Lett. 1974, 70,431. (6) Rydberg, J.; et al. Acta Chem. Scand. 1989, 23,2773,2781, 2797. (7) Ohashi, K.; Freiser, H. Anal. Cbem. 1980, 52, 767.
RECEIVED for review June 1, 1982. Accepted July 27, 1982. The project was supported by a grant from the National Science Foundation.
Determination of Phosphoryl Chloride and Detection of Phosphorus Trichloride in Electronic Grade Trichlorosilane by Gas Chromatography with Thermionic Detection Brenda K. Schulte and Larry W. Shlve” Monsanto Company, St. Peters, Missouri 63376
The assay for phosphorus in trichlorosilane (TCS) is an important part of the evaluation of this raw material which is used in the production of electronic grade silicon. Producers and users of electronic grade TCS try to minimize its phosphorus content because, as the TCS is decomposed, the phosphorus in it is deposited along with the silicon and affects its electronic properties. Accurate methods for phosphorus analysis are therefore needed to aid in the control of the TCS purification processes. Presently this need is met by either a classical spectrophotometric method (1,2) or a silicon quality test method (3, 4 ) which involves making silicon from the TCS. The weaknesses of both methods are (1)large sample sizes are needed (0.1-1 L), (2) several hours of sample preparation time are required, and (3) samples are easily contaminated during the preparation step. Gas-liquid chromatographic (GLC) methods have routinely provided the solution to these types of problems. We describe and evaluate in this report a GLC method for direct analysis of the phosphorus content of TCS. A nitrogen-phosphorus detector (NPD) provides the high sensitivity required for detection of the phosphorus compounds in the TCS ( 5 ) .
EXPERIMENTAL SECTION Apparatus. All analyses were done with a Perkin-Elmer Sigma 1B gas chromatograph equipped with a nitrogen-phosphorus detector. A four-port Carle switching valve was situated in the line between the column and the detector. Connected to the other two ports were a helium-carrying line (pressure controlled flow) and a vent line. The chromatographic separation was done with a 12 ft X 2 mm i.d. glass column packed with 20% OV-101 on Supelcoport. Procedure. Standards. Four external standards of POCl, (Ultrapure grade, Atfa Prcducta, Danvers, MA) in electronic grade trichlorosilane (Union Carbide, Sistersville, WV), with concentrations in the range from 11.0 ppba to 0.33 ppba, were prepared and stored in all-Teflon bottles. The TCS used for the standards contained no detectable POC13. The abbreviation “ppba” refers to the ratio of phosphorus atoms to silicon atoms as expressed in parts per billion. The mixed etchant (Ashland Chemical Co., Columbus, OH) which was to be used for syringe cleaning was diluted with an equal volume of water. G L C I N P D Operating Conditions. The following GLC operating conditions were used: oven temperature, 40 “C; injector temperature, 150 “C; helium flow rate, 35 mL/min (pressure controlled). The nitrogen-phosphorus detector was operated in the N-mode under these conditions: detector temperature, 100 “C; hydrogen pressure, 15-18 psig; air pressure, 30 psig; bead
Table I. Retention Times and Event Timing for the Detection of Phosphoryl Chloride in Trichlorosilane abs
retention compd time, min event time, min switch “in” 15.4 0.89 CH, detector zero 15.55 0.97 PH, 2.47 data collection start 15.6 HSiC1, 20.0 4.21 data collection end SiC1, PCI 3 7.4 switch “out’’ 20.1 POC1, 16.4 method end 35.0 current adjust, 3.5. The detector’s P-mode is recommended for qualitative analysis only. Sample Analysis. An aliquot (0.1 pL-3 wL) of the sample was injected onto the GLC column. The syringes were kept free of silica by flushing the barrels with diluted mixed etchant-a mixture of hydrofluoric,acetic, and nitric acids-immediately after use and then quickly rinsing them with deionized water. A t the start of the analysis, the switching valve was positioned so that the eluate was vented and only helium flowed through the NPD. The switching valve was turned “in”just in time to send the eluted PC13 and POCIBthrough the detector and then turned back “out” again. Absolute retention times and suggested switching and detector zeroing times are listed in Table I. A typical chromatogram of electronic grade TCS is shown in Figure 1. Calculations. The concentration of POCl, in samples of an unknown was determined according to the external standard method of analysis (6). Calibration curves of peak height vs. picograms of POC13 were plotted and then used to relate a peak height to the amount of POC13 in a sample with unknown concentration.
RESULTS AND DISCUSSION The trichlorosilane samples which were analyzed contained CO.1 to 1 ppba POCl, and a far lesser amount of PC13. A typical chromatogram is shown in Figure 1. The linear range of the detector was found to be greater than lo4 and included the concentration range of interest to this work. The precision and accuracy for this external standard method were found to be &IO% and f8% (one relative standard deviation), respectively, by repetitive analysis of a 16.7-pg sample of POC1,. The minimum detectable limit was 350 fg of POCl,. Usable lifetime of the alkali bead exceeded 300 h (2 months of 40-h work weeks). Two precautions were required to ensure the desired detector sensitivity and to obtain reproducible results. First,
0 1982 American Chemical Society 0003-2700/82/0354-2392$01.25/0