gram was intentionally started a t -70" or -85' C. instead of the usual -75' C., the peak areas did not change by more than 3%. The small peak ( A ) at 0 seconds occurred in all samples because of pressure changes during sampling. The relatively small hydrogen peak is a consequence of poor sensitivity. Although the size of this peak can be increased somewhat by increasing the hydrogen concentration, this approach is limited because of the rapid change in the slope of the hydrogen calibration curve a t high concentration. Peaks ( B )and (C) occurred only when the hydrogen fluoride was approaching dryness and are believed to arise from fluorine reactions. They were seldom apparent when the water concentration was kept above 0.2 mole yo (4). The small displacement from the base line before the ozone peak is an indication of ozone decomposition. Such decomposition also occurred during the calibration, so it is not considered a source of serious error. An application of this analysis for monitoring an electrolysis experiment is shown in Figure 3. The rapid Sampling permitted changes in product yields to be followed closely-for example, the rapid increase in ozone during the first hour and the drop in oxygen difluoride
after about 3 hours. Moreover, the displacement of hydrogen from 100% indicates the extent of depolarization, and the discrepancy between anode and cathode (Hz) totals may indicate unidentified, probably nonvolatile, anode products. Such results are giving a more complete picture (4) of the reaction than was previously possible (I,fJ,Q).
is detectable only above Presumably, the use of detectors and/or larger permit detection down less.
ACKNOWLEDGMENT
We thank J. Marlcovich for assisting with the experimental work. LITERATURE CITED
CONCLUSfONS
In our experiments, the formation of fluorine was a condition we wished to and could avoid (4). However, the fact that fluorine shows discrete peaks in the chromatograms indicates that the gas chromatographic technique might be extended to include this gas. Our attempts to passivate the equipment with fluorine were not successful; but we did not try extensive daily passivation ( 7 ) ,which may be effective for such studies as the reactions of fluorine with water or base (8). Alternatively, such reactions could be run with excess water or base to give complete fluorine conversion and thus avoid passivation problems. Another possible extension is to such important areas as the determination of ozone and other gases for air pollution studies. In our present setup, ozone
200-300 p.p.m. more sensitive samples would to 1 p.p.m. or
(1) Briner, E., Tolun, R., Helv. Chim. Acta. 3 1 , 172 (1948). (2) Cady, G. H., J . A m . Chem. SOC.5 7 , 246 (1935). (3) Cook, G. A,, et al., Advan. Chem. Ser. 21,44 (1959). (4) Donohue, J, A., Nevitt, T. D., Zletz, A., Ibid., 5 4 , 192 (1966). ( 5 ) Donohue, J. A,, Wilson, W. A. (to Standard Oil Co., Ind.) U. S. Patent 3 , 1 3 4 , 6 5 6 (May 26, 1964). (6) Engelbrecht, A, Nachbaur, E., Monatsh. Chem. 90, 367 (1959). (7) Hamlin, A. G., et aZ., AXAL. CHEY. 3 5 , 2037 (1963). (8) Hensley, A. L., Barney, J. E., Ibid., 3 2 , 828 (1960). (9) Rodgers, H. H., Evans, S.,Johnson, J. H., J. Electrochem. Soc. 111, 701 (1964).
RECEIVEDfor review July 26, 1966. Accepted September 26, 1966. Work
~Research ~ ~ Office-Durham ~ e " c ' tunder " sE " eAsA ~ r~:$~ Contract DA-31-124-ARO(D)-78.
Quantitative Gas Chromatographic Analysis of Metals, Alloys, and Metal Oxides, Carbides, Sulfides, and Salts RICHARD S. JUVET, Jr. and RICHARD L. FISHER Deparfment o f Chemistry and Chemical Engineering, University o f Illinois, Urbana, 111.
6 7 80 7
b This paper describes a
with CC1, in a sealed capillary, no studies of the quantitative gas chromatographic analysis of nonvolatile inorganic compounds have yet appeared. In this paper it is demonstrated that certain pure metals, alloys, carbides, oxides, sulfides, and metal salts may be fluorinated vith elemental fluorine and quantitatively determined in a rapid gas chromatographic procedure.
rapid, direct method for the quantitative analysis of alloys and certain metal oxides, carbides, sulfides, acetates, and nitrates by in sifu reaction with Fz in a specially designed reactorinjection system of a gas chromatograph followed by separation and analysis of the volatile metal fluorides on a chemically conditioned column. The reaction and elution properties of the elements: U, S, Se, Te, W, Mo, Re, Si, 8, Os, V, Ir, and Pt in various chemical forms have been evaluated. Quantitative determination of the first seven elements is reported.
c
INTEREST has developed recently in the quantitative determination of metal halides. Hamlin, Iveson, and Phillips (4) determined UFe in mixtures of UFa, Clz, ClK, ClF, and HF. Phillips and Timms (9) employed the quantitative ONSIDERABLE
1860
ANALYTICAL CHEMISTRY
gas chromatographic determination of silicon and germanium tetrachlorides produced by reaction of silanes and germanes with gold(II1) chloride as a measure of silicon and germanium in mixed hydrides. Dennison and Freund ( I ) described the analysis of SnCLAsCls mixtures, and Sie, Bleumer, and Rijnders (11), recently determined a number of metal chlorides in mixtures quantitatively. Juvet and coworkers (6, 14) described a sensitive and selective, corrosion-free, flame photometric detection system giving linear response for metal halides over four orders of magnitude. The present authors in a previous publication (6) showed that several volatile metal fluorides can be chromatographed on a chemically conditioned poly(tetrafluoroethy1ene) column coated with poly(trifluoromonochloroethylene). With the exception of Sievers, Wheeler, and ROSS' (1.9) interesting work on the determination of Tic14 following reaction of TiOz
EXPERIMENTAL
Chemicals. The chemicals used in
this study were obtained from commercial suppliers as follows: W, Mo, Re, Os, UC2 (Research Inorganic Chemical Co.); UOZ, Mo&, WC, VC (Alpha Inorganics, Inc.); UOz(CzHs0z)z * 2Hz0, UOz(T'JOs)2 * 6Hz0, Mooa, VzOs, S (J. T. Baker Chemical Co.); Se (B. F. Drackenfelt and Co., N. Y.); Te (C. A. F. Kahlbaum, Berlin); Mo-W alloys (courtesy P. R. Mallory and Co., Inc., and Climax Molybdenum Coo of Michigan); MOSZ (native molybdenite crystals, A. D.
-TO
COLUMN
VACUUM
kF L U O R I N E Figure 1 .
Reactor-injection system for in situ fluorinations
Mackay, Industrial and Rare Minerals and hIetals) ; Fz(Air Products Corp.). These materials were used without further purification except for treatment of the Mo, W, Re, IIo2C, and R C with Hzat 1500’ C. in a quartz tube to remove traces of oxides or absorbed oxygen gas present in these finely powdered materials. Apparatus. A modified Micro-Tek Model 2000-R research gas chromatograph was used throughout this study. Modifications consisted of the use of a Gow-Afac stainless-steel thermal conductivity cell equipped with nickel filaments resistant to corrosion by metal fluorides and the addition of the stainless steel reactor-injection system illustrated in Figure 1. The reaction chamber and gas sampling valve were constructed of Type 316 stainless steel. Tubing connecting the sampling valve t o the reactor and leading to the column was heavy-walled Type 316 stainless steel, l/&nch i.d. The miniature needle valves (Matheson Co.) shown have Type 303 stainless bodies and Type 316 stainless stems. Vacuum and fluorine inlet lines were 1/4-inch copper. Total volume of the 3/8-in~h diameter reaction cavity was variable over the range 1 to 3 ml., depending upon the position of the PTFE insert holding the nickel-ribbon, sample holder-heater assembly. The helium inlet and outlet tubes were connected t o provide the minimum dead volume. All connections directly in contact with the contents of the cell were helium arc welded. Clearances between the PTFE insert and the body of the cavity as well as the spacing between the plunger rod and !Tall of the gas sampling valve TTere made as small as practical but are exaggerated in Figure 1 for clarity. Stainless steel was free from visible attack by fluorine up t o 250’ C. Above this temperature a non-adherent layer of fluorides slowly formed. The heating element of the sample holder-heater assembly was made from a l/Ie-inch wide, 6-inch long strip of 0.005-inch nickel ribbon (Matheson, Coleman and Bell). The ribbon was wound into a 1/4-inch diameter helix, flattened, and silver soldered to two l/g-inch brass electrical leads. A 3/g-
inch diameter P T F E rod drilled to accommodate the brass leads served as an electrical insulator. A brass compression fitting was used to seal the P T F E around the rods. The entire assembly was sealed inside the reaction cavity, as shown in the diagram, by means of a neoprene “O”-ring. The four-port gas sampling valve was equipped with neoprene “O”-rings which were found t o give satisfactory service with fluorine under pressure when lubricated lightly with Kel-F oil. In the absence of lubrication, the valve would stick after remaining static for some time in the presence of Fg. The “0”rings were replaced twice a week to avoid leakage by gradual hardening of the neoprene. Metal Fluoride Preparation. Finely divided samples of Mo, W, AlozC, and WC were weighed in a nickel boat, placed in a quartz tube, and heated t o 1500’ C. in a H2 atmosphere for approximately 15 minutes to reduce oxides or adsorbed oxygen gas which interfered with the analysis owing t o the formation of higher boiling oxyfluorides (WOF4, b.p. = 187’ C.; hl0OF4, b.p. = 185’ C.). Samples of MoOa were quantitatively reduced with Hz t o the metal with subsequent accurate analysis following fluorination. MoOs and W03 could not be converted directly to the hexafluorides by heating in Fz in the reactor system described. Other samples were weighed directly to the nearest microgram on a x 1/2-inch piece of nickel or aluminum foil using a Model M-10 Cahn Electrobalance. The foil was then placed directly on the nickel heating element. Short circuiting did not occur because the contact resistance of the sample container with the heater was very large with respect to the 0.5ohm resistance of the nickel heating element. After inserting the sample in the reaction cavity, the cavity was evacuated to a pressure of less than 1 mm. Hg, Fz gas was admitted under a pressure of 10 p.s.i., and a potential of 3 volts 8.0. was applied t o the heating element for a 1-minute period. This potential was sufficient to heat the nickel resistance heater to a dull red heat.
Xo visible attack on the nickel foil occurred at this temperature. After the 1-minute heating period, power was turned off and the reaction products were immediately swept into the column through the gas sampling valve. Provided the temperature of the reaction cavity was not in excess of 150’ C., this procedure gave quantitative reactions with all the samples studied with the exception of VC, P t , and Ir. In normal operation the cavity was maintained at 75’ C., well above the boiling point of the least volatile material studied (UFO,b.p., 56.5’ C.). Column. The column used throughout this study was a 22-foot, l/*-inch diameter P T F E column packed with 15% w./w. Kel-F oil No. 10 (Minnesota Mining and Manufacturing Co.) on 40- to 60-mesh Chromosorb T (JohnsManville Co.). Packing was performed a t liquid nitrogen temperatures as previously described ( 6 ) , and the column was conditioned with ClF3 or FP prior to use to remove moisture and traces of reactive organic matter. RESULTS AND DISCUSSION
Column
Conditioning
Require-
ments, Siliceous diatomaceous earth supports generally react with metal fluorides and cannot be used in these studies. Chromosorb T (Teflon-6) is nonreactive with most metal fluorides and is the only solid support found useful to date for the quantitative deiermination of these reactive materials. Teflon-6 support has an unusual property vorthy of mention. There is evidence that some solution or adsorption of the hexafluorides occurs in the PTFE solid support. With all the hexafluorides studied, it was necessary t o saturate the column with the fluoride being analyzed before consistent and reproducible results could be obtained. Saturating the column with one fluoride does not saturate it for other fluorides. Before an alloy or mixture can be analyzed quantitatively, the column must be saturated with respect to all components being determined. Xoreover, after conditioning the column with a hexafluoride, the recorder base line was shifted slightly up-scale and would not return to the original position for several hours after the initial injection. T h e n the column was operated in m unconditioned state, small samples gave larger relative errors than larger samples. Saturation is easily accomplished by injecting several large samples of the hexafluoride onto the column before quantitative measurements are made. Once the column is saturated, it remains in this condition approximately one-half hour. If longer times than this elapse between injections, somewhat lower results are obtained than expected. Other authors (4,?) have also noted solution or adsorption effects on PTFE. VOL. 38, NO. 13,
DECEMBER 1966
0
1861
Analysis of Mo and W Compounds. Calibration curves were prepared for It7 and M o (Figure 2) using both hydrogen-purified powdered metals and pure M o and W mires from a 100watt hniplex light bulb. Pure 1%‘ wa3 obtained from the filament and >do from the filament support or stabilizer wire, commonly called the “pig tail.” Calibration curves were identical using metals from these two murces. Linear response was noted with both ill0 and W over the range shown in Figure 2, and the curves pass through the origin. Although the linear range extends to at least 5 mg. for both metals, reproducibility decreased above 1.5 mg. Using these calibration curves, samples of blo-W alloys, hlo,C, WC, and M o S n were analyzed, and the results are presented in Table I. Alloy samples in the form of fine turnings or small chips were analyzed one week apart. After preparing a W calibration curve, three samples of alloy were analyzed for 1V content only. After preparing a 110calibration curve six days later, three more determinations were run and &Io and \I7 were simultaneously determined. No significant deviations were noticed for the values found for W on the different dates, thus indisating no serious change in colunin
Table 1.
Gas Chromatographic Analysis of Tungsten and Molybdenum Alloys and Compounds
Sample wt., mg.
w
1.980 1.278
0.624 0.395 0.314 0,315 0.493 0.080
1,026 1,002
1.556 0,258 1.098 0.778 0.960 1.522 1.074 1.062 0.488
0.352 0.236 0.310 0,490 0.351 0.321
0.506 0.942 0.822
0.481 0.876 0.778
0.250 0.714 0.874
... ...
0.152
...
Analysis 1.435 0.829 0.956 0.686
1862
mg. Element
Figure 2. Calibration curves for fluorinated samples of SI Mol Se, W, Te, UOZ,and Re showing relative response
...
...
...
...
ANALYTICAL CHEMISTRY
W, 70 % ‘Found Error Found Climax Mo-W Alloy (Reported Analysis, 30.9% W 0.001% C ) +0.6 ... ... 31.5 ... 30.9 0.0 *.. -0.3 .,. 30.6 31.4 $0.5 66.0 0:66l 31.7 $0.8 68.7 1.069 31.0 +o. 1 68.2 0.176 Mean values 31.2 f 0 . 4 +0.3 67.6 f 1 . 4 Mallory &Io-W Alloys (Reported Value, ca. 30% W ) ... 32.1 ... ... ... 30.4 . ~ . .,. 32.3 ... ... ... 32.2 ... ... 67.2 32.7 ... 0.722 65.4 30.2 ... 0.694 68.8 0.336 31.6 ... 67.1 f 1.7 Mean values 31.6 f 1.0 Analysis of W C (Reported Purity 98%; Formula 70 W,93.9%) ... 95.1 I .2 *.. 93.0 0.9 ... 94.8 0.9 1.0 Mean values 94.3 A 1.1 Analysis of MOZC[Reported Purity 98yo; % Mo (Calcd.), 927,] 0.233 ... ... 93.1 *.. ... 92.5 0.659 *.. ~ . . 93.5 0.817 Mean values 93.0 A 0.5 of Molybdenite Crystals (MoSz) (Purity, Unknown; Formula % hlo, 59.9%)
Fonnd, mg.
?VI0
+
.
0.791 0.455 0.525 0.382
.,. . I .
.
.
I
...
... ... ... ...
I
.
55.1 54.9 54.8 55.7
Mean values 55.1 rt 0 . 4
Error
...
... ...
... .*.
..,
.,.
... ... .,. * . *
... ,..
~ . . I
.
.
...
$1.1 +0.5
+1.5 +1.0 Purity, calcd. 92.3% 91.8 91.6 93.0 92 2% I
Re FI
F*tREAGTIOR
PRODUCTS
O*F,
WF6
P
IMPURITY
I N RHENIUM
s
8L?
ROOF8
PTFE column, 22-feet column temp., 75' C.;
X %-inch 0.d. containing 15% w./w. Kel-F No. 10 oil on 40- to 60-mesh Chromosorb T.
Carrier gas flow rate, 2 8 ml./min. He;
sample size ca. 1 mg. of W , Re, and Os metals
characteristics or detector response. The results obtained in all deterniinations showed precision and accuracy coniparable t o quantitative gas chromatographic methods for organic materials. Based on the quantitative behavior of the carbides and ;\loSz, the conclusion may be drawn that this method should be directly adaptable to many other W and M o compounds including the nitrides, borides, selenides, tellurides, and silicides. Samples of compounds that contain oxygen such as molybdates or tungstates may also be applicable to this method of analysis following reduction with H2 prior to fluorination. Analysis of Uranium Compounds. Uranium hexafluoride is one of the easiest of metal fluorides to determine gas chromatographically. Owing to the high stability of UFe, i t may be prepared directly from uranium oxides and carbides. The synthesis of U F O
Table
II.
Analysis of Uranium Compounds
Error, %
U, %
U found, mg.
Sample wt., mg.
U O Z ( N O ~ ) Z - ~(Reported H~O purity, 99.87,; calcd. % U, 47.37$) 3.156 1.52 48.3 f1.0 2.732 1.31 48.0 1-0.7 1.869 47.6 0.89 +0.3 Mean values 48.0 f 0 . 3 +0.7 UO2(CzH80z)2 -2H20 (Reported purity, 99.67,; calcd. % U, 55.9%) 2.985 1.70 56.9 +1.0 2.507 1.44 57.4 +1.5 1 653 0.95 57.5 $1.6 1.028 0.59 57.4 +1.5 Mean values 57.6 A 0 . 4 +1.4 I
0.412 1.109 1.873
UCZ(Reported purity, 99.9%; calcd. % U, 90.7%) 0.358 87.0 0.974 87.8 1.655 88.5 Mean values 87.8 f 0 . 7
-3.7 -2.9 -2.2 -2.9
7
I
4
,
I
6
I
I
8
I
I
10 Figure 4.
Column temp., dry ice-acetone; as in Figure 3
I
I
I
I
I
I
12 14 16 18 TIME (MI N . ) Elution of SiF4, BF3, SFS, Seh, and TeF6
sample sizer: 0.2 mg. of S, 0.5 mg. of Se, 0.9 mg. of Te plus trace amount of
BzOa
I
I
I
20
and quartz powder.
22
i F
Other parameters
VOL. 38, NO. 13, DECEMBER 1 9 6 6
1863
from COZ, U308, L-02, UOZF,, UC2, U303P207,and UF4 by reaction with Fzhas been reported to be quantitative (IO,I S ) . Samples of T-Oa(NOa)2.fiH20, U0,(C2H302)2.2Hz0, UC2, and KO2 were determined quantitatively gas chromatographically, and the results are given in Table 11. The calibration curve shown in Figure 2 was prepared using samples of COz of 99.5% purity. High results for the uranyl salts are attributed to loss of water of hydration during weighing of the finely powdered crystals. Crystals freshly obtained from the container were transparent until exposed t o the atmosphere whereupon they shortly became translucent during the loss of water of hydration. In order to reduce the number of elements consuniing fluorine in the fluorination step, weighed samples of the nitrate and acetate were first heated slowly in air with the sample heater outside the reaction cavity until the water of crystallization was removed and the sample had thermally decomposed into oxides of uranium. UC2 ignites spontaneously in air under friction to form the oxide, and this leads to low results in the analysis owing to a decrease in weight per cent of uranium upon oxidation. Attempts to analyze carnotite ores were unsuccessful. From high grade carnotite ore only a trace of cF6 was formed. Interference may be caused by the presence of CaO, MgO, and K20, which, on forming the corresponding fluorides, adsorb and mechanically trap CF6. Analysis of Re and Os Metals. Figure 3 shows a Chromatogram of the separation of JvF6, ReF6, OSF6, and ReOFS. The peak assignments were made with the aid of standard spot tests (2, 8). The peak labeled ReOFj mas not identified as such, but was shown to contain Re. A literature search revealed the only compounds of Re volatile at the temperature of the column are ReF6 (b.p., 33.8' 6.) and ReOFS (b.p., 73.0' C.). Treatment of the finely-divided Re powder with H2at 1500' C. eliminated the ReOFb peak. An unidentified impurity, not completely separated from the fluorine peak, was always observed in the analysis of the powdered rhenium metal even after reduction with hydrogen. The simultaneous quantitative determination of W and Re was demonstrated by the analysis of a tungsten301, rhenium theimal conductivity filament (Gom-Mac Instrument Go.). The W-Re filament burns vigorously in a F2 atmosphere. Results were as follows: Wt. of sample, 1.584 mg.; mg. W found, 1.519; % W, 96.0%; mg. Re found, 0.043; 70Re, 2.7%. Although Re gave a linear calibration curve (Figure 2 ) when the column was 1864
4
ANALYTICAL CHEMISTRY
6 t REAGTlON
SAMPLE: V 6
PRODUCTS
VOF3
\ 2
4
Figure 5. a.
b. e.
SAMPLE: VP Os
SAMPLE: V C f AIR
8 IO 12 14 T I M E (MIN.) Fluorination products of vanadium compounds
6
5 mg. of VC in absence of air 1 mg. of VaOa 5 mg. of V C in presence of air.
Column temp., 55' C.
conditioned as previously described, osmium could not be eluted quantitatively. After several samples of OsFB were eluted, the column packing became noticeably gray, an indication of some decomposition or reaction on the column. OsF6, b.p., 47.5' C., is one of the more reactive hexafluorides and could conceivably react with the solid support, the liquid phase, or with decomposition products on t8he column from previously eluted materials. At the time of the osmium studies, the column had been used for more than 300 separations of various metal fluorides.
16
Other parameters as in Figure 3
Attempts to extend the analysis of the platinum group metals to include IrF6 and PtF6,were unsuccessful owing to reaction or decomposition of the hexafluorides on the column. The intensely black deposits which formed on the column could only be removed by passing pure F2 or ClFa through the column at 50' C. for several minutes. , Analysis of S, §e, and Te. Figure 4 shows a chroniatogram of the elution and separation of SFs (b.p., -63.7' C.), SeFe (b.p., -45.9' C.), and TeF6 (b.p., -38.4' c.) a t dry ice-acetone temperatures. These compounds were prepared in situ by
reaction of the elements. Peak assignment was made by comparing retention times with chromatograms of separate components. These hexafluorides are readily prepared quantitatively and are particularly easy to handle gas chromatographically. Extensive column conditioning is not required for the quaptitative elution of these compounds. Peaks of SiFl and BF3, prepared from Bz03 and powdered quartz which were added to the mixture, were unresolved a t this column temperature. Since Kel-F oil No. 10 begins to solidify at -6' C., separation of SFe, SeF8, and TeFo a t dry ice-acetone temperatures is apparently due to gas-solid adsorption rather than to partitioning. Elution of VF6. Attempts to determine vanadium compounds quantitatively failed owing to oxyfluoride formation and reaction of the fluoride with the packing material. The elution behavior of vanadium is shown in Figure 5 , The peaks labeled V F j and VOFI mere shown t o contain vanadium by standard spot tests ( 3 ) . The peak labeled VOFI was formed only when oxygen was present. This niaterial condensed as needle-like colorless crystals in the exit tube of the chromatograph when eluted in large quantities. VOF3 is the only known oxyfluoride which would give this behavior. The
yield of VFs (b.p., 47.9" C.) is greatly dependent upon the reaction conditions. When the heating element is maintained a t red heat during the one-minute reaction period and the reactor block is maintained a t 50" C., small amounts of VF6 are produced by reaction of Fz with VeOa. If the reactor block is increased in temperature to 125" C. or if the heating element is lowered in temperature, no VF6 is produced. In the reaction of F2 with VC, VF6 is formed in small yields when the reaction block is maintained below 50" C.; at higher temperatures, no VFs is formed. The formation of a dark gray coloration in the column suggests reaction between JTF5and the packing material. Conclusions. This work demonstrates that gas chromatography is readily adaptable to the quantitative analysis of many inorganic compounds and alloys. The procedure is rapid, often requiring no more than 20 minutes per sample. Relative errors approaching 1% are common with well - behaved systems. -4 thermodynamic study of the solutesolvent interactions of these systems will be published elsewhere. LITERATURE CITED
(1) Dennison, J. E., Freund, H., AUL.
CHEM.37, 1766 (1965).
Quantitative Analysis of Aqueous by Gas Chrornatogruphy
(2) Feigl, F., "Spot Tests," Vol. I, p. 137, Elsevier, New York, 1954. (3) Ibid., p. 118. (4) Hamlin, A. G., Iveson, G., Phillips, T. R., ANAL.CHEM.38,306 (1963). (5) Juvet, R. S., Durbin, R. P., Ibid., p. 565. (6) Juvet, R. S.,Fisher, R. L., Ibid., 37, 1752 (1965). (7) Mattraw, H. C., Hawkins, N. J., Carpenter, D. R., Sahal, W. W., J . Chem. Phys. 23, 985, (1955). (8) Meites, L., ed., 'Handbook of Snalytical Chemistry," p. 2-17, &ICGraw-Hill. New York., ~1863. - - (9) Philli s,' C. S. G., Timms, P. L., Ax.4~.&EM. 35, 505 (1963);( (10) Rodden, C. J., ed., Analytical Chemistry of the Manhattan Project," Div. TIII, Vol. 1, p. 40, RIcGrawHill, New Yo&*-,..-"". lain (11) Sie, 8.T., EUeumer. J. P. 9.. R.iinders. G. R. A., Separation Sci. 1, 41 (1966): (12)- Sievers, R. E., Wheeler, G., Ross, m.D., ANAL.CHEM. 38, 306, (1966). (13) Simons, J. H., ed., "Fluorine Chemistry," 1'01. VI p. 106, Academic Press 1964. (14) Zado, F. &I., Juvet, R. S., ANAL. CHEM.38, 569 (1966).
RECEIVEDfor review July 29, 1966. Accepted October 4, 1966, Research supported by the National Science Foundation (Grant NSF-GP-2616 and continuation grant NSF-GP-5151). Presented in part at the 6th International Symposium on Gas Chromatography and Associated Techniques, Rome, Italy, September 20-23, 1966.
Alcohol
Mixtures
CLAIRE BLUESTEIN and HOWARD N. POSMANTER Cenfral Research, Witco Chemical Co., Inc., Oakland, N . J.
b An improved gas chromatographic procedure for determination of water and C1 to Cg alcohols in mixtures was developed. The mixtures were separated on a 12-foot X I/d-inch column packed with 5% Triton X-305 on Teflon using helium carrier gas and a thermal conductivity detector. The lower limit of detection of any component by this procedure was of the mixture. The about upper limit of water which this column will effectively handle is about 8 mg. Even at high water concentrations, the peaks show negligible tailing. The calculation of unknown samples was made b y u new variation of the internal normalization method which is applicable to any three or mare component systems. A table of area fractions vs. weight fractions based on a known standard was constructed. The sample composition range for
o.3Y0
such a table must be limited to obtain accurate results with unknowns. This calculation procedure has advantages over usual internal normalization methods in that sample size and method of area computation can be variable. Also, it is readily utilized for routine control of plant process mixtures. This general procedure was adapted also for aqueous alcohol extracts of nonvolatile compounds.
N
gas chromatographic procedures for the analysis of aqueous alcohol mixtures have been published. I n the quantitative analysis by gas chromatography of polar mixtures containing water, a major problem had existed because of tailing, particularly in the water peak (8,Q).When the direct determination of water in a sample was not necessary, many turned to the use UXEROUS
of the flame ionization detector which is purportedly unaffected by water in samples, though this is not strictly true (4). Where a direct determination of water 13-as required and the usual thermal conductivity detector was available, it has been generally accepted that a Teflon column packing has been the best solution to the problem of tailing ('J
R1
6)'
For the substrate, we have preferred to use Triton X-305 because this improved the separation of the ternary mixtures with which we were working, decreased the relative retention time of the water peak, and allowed for more rapid elution of the higher boiling components a t a relatively lorn column temperature. The thermal conductivity detector response was used to calculate quantitative amounts of water and alcohols in the mixtures. Present quantitative calculation VOL. 38, NO. 13, DECEMBER 1966
e
1865
