The system varies somewhat with pH, one calibration curve being needed for p H 5 to 7 and another for p H 7 to 9; and (2) the tolerance limits are less for the cations of Al, Bi, Cd, Hg, Th, and V (see Tables 111 and IV). RECOMMENDED METHOD
Using the apparatus and reagents already described, the following method is recommended for determining iron with either 6-hydroxy-1,’l-phenanthroline or 8-quinolinol. The Sample. Procure a representative sample, and subject it to any necessary preparative treatment to obtain a usable solution. If necessary, remove any interfering ions to bring them within the tolerance limits listed in Tables I t o IV. The solution should be sufficiently acidic to prevent precipitation of the hydrous oxide. Calibration Curve. (A)
&HYDROXY-
With a pipet transfer to 25-ml. volumetric flasks 0.30, 0.60, 1.00, 1.30, 1.60, and 2.00 ml. of the stock Fefs ion solution. To each flask add 6.25 ml. of 0.20M KH(C8H4Or),10 ml. of 1-propanol, and 2.0 ml. of 0.20M ligand. Dilute each to the mark with distilled water and mix. Adjust to p H 5.5 i 0.5 with 5M NaOH (the solution turns purple). Measure the transmittancy (absorbancy) in a 1-cm. cell a t 570 mp and plot the calibration curve. 1,7-PHENANTHROLINE.
If p H 7 to 9 is preferable, follow the previous directions but use 6.25 ml. of the 0.20M buffer containing KCl-H3BOs, and adjust the p H to 8 f 1. (B) ~-QUINOLIKOL. With a pipet transfer to 25-ml. volumetric flasks 0.50, 1.00, 1.50, 2.00, and 2.50 ml. of the stock Fe+3 ion solution. Add 2.0 ml. of 0.2094 ligand to each flask. Otherwise follow the same procedure as described for 6-hydroxy-1,7-phenanthroline, but measure the transmittancy (absorbancy) a t 585 mp. Use the phthalate buffer for pH 5 to 6 and the borate buffer for pH 7 to 9. Either range gives the same calibration curve. Procedure. The preferable chromogenic reagent and the p H for measurement depend upon the amounts and nature of any interfering ions present in the solution of the sample. Consult Tables I to IV for tolerance limits. Use an aliquot of the sample solution containing an amount of iron within the limits of the calibration curve. Filter or centrifuge out any precipitate which may form after pH adjustment. Measure the transmittancy (absorbancy) and determine the amount of iron from the calibration curve. ACKNOWLEDGMENT
The authors gratefully acknowledge the fellowship grant of Allied Chemical and Dye Corp., which financially supported this work.
LITERATURE CITED
(1) Druey, J., Schmidt, P., Helv. Chim. Acta 33,1083 (1950). (2) Duswalt, J. M., “Analytical Applications of Some &Hydroxyquinoline
Derivatives t o the Spectrophotometric Determination of Iron and Vanadium,” Ph.D. thesis, Purdue Eniversity, 1961. (3) Dusyalt, J. M., “Preparation and Analytical Application of 5-Fluoro-& hydroxyquinoline-7-sulfonic acid,” M.S. thesis, p. 6, Purdue Cniversity, 1957.
(4) Ghosh, T. N., Roy, A. C., Banerjee, S., J . Indian Chem. SOC.2 2 , 219 (1945). (.5.) Haworth, R. D., Sykes, W. O., J . Chem. Soc.’1944,311. (6) Hollingshead, R. G. W., Chcm. & Ind. (London) 1954, 344. (7) Hollingshead, R. G. T., “Oxine and Its Derivatives,” Vol. I, p. 280, Butterworths, London, 1954. (.8,) Matsumura. K.. J . Am. Chem. SOC. 52,3974 (1930). ’ (9) Sandell, E. B., Spinder, D. C., Ibid., 71,3806 (1949). (IO) Snell, F. D., Snell, C. T., “Colorimetric Methods of Analysis,” Vol. 11, D. 324. Van Nostrand. New York, i949. ‘ (11) Swank, H. W., “Spectrophotometric Study of Colorimetric Methods of Determining Iron,” Ph. D. thesis, Part 11,p. 8, Purdue University, 1937. (12) Swank, H. W., Mellon, M. G., IND. E N Q . CHEM.,ANAL.ED.9,406 (1937). (13) Tomkinson, J. C., Williams, R. J. P., J . Chem. Soc. 1958, 1153, 2010. (14) Welcher, F. J., “Organic Analytical Reagents,” Vol. I, p. 308, Van Nostrand, New York, 1947. RECEIVED for review April 21, 1961. Accepted September 5, 1961.
Analysis of Multicomponent Methyl- and Phenylchlorosilane Solutions CARL A. HlRT Silicone Products Department, General Electric Co., Waterford, N. Y.
b The properties of silicone resins are largely dependent on relative amounts of various chlorosilane monomers in solution before polymerization. Analysis of chlorosilane mixtures is satisfactorily performed by mass spectrometric techniques based on pressure sensitivities. The standard deviation is less than 0.5 mole for all compounds present. Calibrations based on known mixtures rather than on pure compounds give more accurate results.
70
T
monomers from which siloxane polymers are made are based on various substituted chlorosilanes (9). By blending together various monomers before hydrolysis it is possible to obtain a wide range of molecular weights and degrees of cross linking in the polymers. Since the properties of the final polymer depend to a large degree on the composiHE
1786
ANALYTICAL CHEMISTRY
tion of the monomer blend, it is necessary to have an accurate analysis before the blend is hydrolyzed. Mixtures of chlorosilanes have been previously analyzed in our laboratories by analytical distillation, but this is time-consuming. Data in a recent article (7) show that mixtures of the type described in this article might be analyzed by infrared techniques. The application of the mass spectrometer to analysis of mixtures of methylchlorosilanes has been demonstrated by Norton (10) and Brewer (4). Taylor et al. (11) report the results on a mixture of trimethylchlorosilane and hexamethyldisilane. Other applications of mass spectrometers have been described in the literature, as is evident from excellent review articles (6). The mass spectrometer procedure described in this article for mixtures of
methyl- and phenylchlorosilanes is rapid and minimizes any hydrolysis errors due to sample manipulation. EXPERIMENTAL
The General Electric analytical mass spectrometer was used for this work. All spectra were obtained with a tungsten filament using an acceleration potential of 2000 volts and an ionization potential of 70 volts. The instrument was equipped with a Consolidated Electrodynamics micromanometer and stainless steel inlet system. The original oven thermostat was replaced with a Sargent Thermonitor in order to reduce temperature variations in the oven. The improvement in temperature control eliminated the zero drift on the micromanometer, and increased the precision in measuring pressure by 50%. The oven temperature was controlled a t 135” C.
-+
:c
METHYLTRICHLOROSILANE Me Si CI 3
100
c
a
'5 - S i C I t 50-r"l 25-
IO0
c
* CH3 Si CI2*
I
CH3SiCI3'
+
I DIMETHYLDICHLOROSILANE Mc2SiC12
CH3!iC12
'I
I
M/e
Figure 1.
Samples are introduced using the indium capillary technique (8). Two capillaries approximately 1.5 inches long and 0.010 inch in internal diameter are filled with sample by capillary action. The filled capillaries are sealed off and placed in a tube equipped with an Asco-Seal 12/30 standardtaper joint (available from Arthur F. Smith Co.). After the air has been pumped off, samples are vaporized by melting the capillaries (m.p. 117" C.). Pressures for analysis are approximately 135 microns on the high pressure side of the leak. -411 chlorosilanes used in this work were chromatographically pure. They were obtained by laboratory distillation of material manufactured in our plant. The mass spectra showed no evidence of impurities. Fragmentation of Chlorosilanes in M a s s Spectrometer. A typical chlorosilane blend used in the manufacture of silicone resins is made u p of dimethyldichlorosilane, diphenyldichlorosilane, phenyltrichlorosilane, and methyltrichlorosilane (9). The principal peaks in the mass spectra of these compounds are shown in Figure 1. Only peaks of 5% or more relative intensity in the mass range above 63 mass units are shown. ~IETHYLCHLOROSILAI~E SPECTRA.In the spectra of the methylchlorosilanes the most intense peak is due to the loss of a methyl group from the parent molecule. Thus for dimethyldichlorosilane the peak a t mass number 113 due to the CH3SiC12ion is the strongest, while for methyltrichlorosilane the peaks a t mass
Mass spectra of phenyl- and methylchlorosilanes
numbers 133 and 135 due to SiClsion are the strongest. As shown in Figure 1, more than one peak is assigned to the CH3SiCls and SiC13 ions. The additional peaks are due mainly to the C13' isotope, with smaller contributions from the Siz9and C13 isotopes. In the spectrum of dimethyldichlorosilane the ratios of the 113,115, and 117peak heights to the 113 peak height have the expected values of 1, 0.64, and 0.10 for an ion containing two chlorine atoms ( 2 , 5 ) . Likewise the expected values of 1,0.96,and 0.31 for an ion containing three chlorine atoms (2, 5 ) are observed for the peaks a t 133, 135, and 137 for the SiCla ion in the spectrum of methyltrichlorosilane. The preceding relative abundance ratios for ions containing two and three chlorine atoms differ slightly from those given by Beynon (3), because he used slightly different isotopic chlorine values than Bernstein (2). The peak a t 93 in the dimethyldichlorosilane spectrum is due t o the (CHJzSiCl ion. The peaks due to the parent molecule are a t 128 and 130 for the dimethyldichlorosilane and a t 148 and 150 for the methyltrichlorosilane. The peaks a t 63 and 65 are due to the Sic1 ion and are found in all the spectra. The ratio of the 65 to 63 peak height shows the 1 to 3 ratio expected for the chlorine 35 and 37 isotopes ( 2 , 5 ) . PHEXYLCHLOROSILANE SPECTRA. The phenylchlorosilanes do not behave as simply as the methylchlorosilane in the mass spectrometer. Phenyltrichloro-
silane has three equally intense peaks. Those a t mass 210 and 212 are due to the parent molecule, while the peak a t 175 is due to the loss of a chlorine atom from the parent molecule. The other peaka in the groups assigned to these two ione are due to isotopes. The ratios of the 210,212, and 214 peak heights to the 210 peak height have the expected values of 1, 0.96, and 0.31, respectively, for an ion containing three chlorine atoms (2, 6 ) . Likewise, the ratios of the 175,177, and 179 peak heights have the expected values of 1, 0.64, and 0.10 for an ion containing two chlorine atoms ( 2 , s ) . The SiCla ion is formed by loss of a phenyl group from the parent molecule. The phenyl ion is present a t mass number 77. The peak a t mass number 112 is due to the CeHsCl ion formed by rearrangement. In the mass spectrum of diphenyldichlorosilane the strongest peak occurs a t mass number 154. The parent molecule ion is present a t mass number 252. The peaks a t 175, 177, and 179 are due to the CBH6SiC12ion formed by loss of a phenyl group from the parent molecule. The ratios of the peaks assigned to the parent molecule and the CBH6SiCla ion have the values expected for an ion containing two chlorine atoms (2, 5 ) . The phenyl ion is present a t mass number 77. The phenyl group in the phenylchlorosilanes shows a marked tendency toward rearrangement reactions. In diphenyldichlorosilane the phenyl groups rearrange to form biphenyl, which results in the most intense peak in the spectrum a t mass number 154. In VOL 33, NO. 12, NOVEMBER 1961
1787
Table 1.
Compound MezSiClz MeSiCL CsHsSiCla ( CeH6)zSiClnb
Calibration Data
Analysis Peak 128 148 210 217 148 128
Ion Parent Parent Parent (C6H6)ZSiCl
Sensitivity,a Divisions/Micron From From pure solution compounds 3.25 4.94 23.7 5.00 0.05 0.24
2.75 3.44 20.5 5.03 0.039 0.26
Average of two determinations. Corrections for changes in instrument conditions are based on sensitivity of (C&)*Sic&. a b
Table II. Results on Known Mixtures Mean ErroP
Sam-
Compd. Solution True calibra- calibraCompound Mole % tion tion lb +2SiC12 14.87 - 1 . 5 - 0 . 2 +SICl, 9.64 0 . 4 -0.05 MeSiCla 14.43 - 1 . 5 -0.07 MezSiClt 61.06 0.9 0.4 2b +zSiC12 29.45 - 1 . 3 0.5 0.6 0.08 +Sicla 16.60 9.34 1 . 2 -0.05 MeSiClr MezSiClz 44.61 - 0 . 5 - 0 . 6 * 0.7* 3b dzSiCl2 30.07 - 0 . 7 1.5 -0.2 +Sicla 17.76 0 . 5 -0.2* MeSiCla 9 . 6 9 MelSiC12 42.49 - 1 . 3 0.3 0.6 40 +zSiClz 30.17 16.60 0.6 +Sicla 9.55 -0.2 MeSiCla -1.0 Me2SiC12 43.69 Average difference with regard to sign of test results from true values. * Mean error calculated from 3 and 4 detns., respectively, for pure compd. and solution calibration. 0 Mean error calculated from 2 detns.
&
5
Table 111.
Standard Deviationsa
99 % Confidence Limits for Av. of 3 Detns., Mole $70
SI Quantity Measured Mole $70 0.58 +zSiCIz 0.312 0.130 0.24 c$Slc& 0.073 0.14 MeSiCla 0.46 MezSiClz 0.242 5 Calculated from 10 re licate analyses of known mixture by 2 anafysts.
phenyltrichlorosilane the rearrangement of the phenyl group results in the formation of chlorobenzene, which is detected a t mass number 112. Analytical Peaks. The peaks chosen for analysis are mass numbers 128, 148, 210, and 217. They are shown as open bars in Figure 1 . Since there is no need t o determine low concentrations in the present system, i t was not necessary t o use the most sensitive peaks. Peaks with the 1788
ANALYTICAL CHEMISTRY
least amount of interference from other compounds in the system were chosen for the analysis. Except for diphenyldichlorosilane, all the analysis peaks are due to the molecule ion. The (C6Hs)ZSiCl ion at mass number 217 for diphenyldichlorosilane was chosen because it is closer to the 210 peak than the 252 parent peak and thus reduces instrument scanning time. The peak at mass number 154 was not used because of possible errors due to biphenyl contamination. The analytical procedure is based on the following equations:
+ St&P+D
Hiis = ST4,PT
(2)
Hzio = SF,P+T
(3)
Hz17 = LS'$,D,P+~
(4)
where H = peak height S = sensitivity P = pressure D = MezSiClz $D = (CeH&SiC12 T = MeSiCla $T = CeH&3iC1, The additional terms in Equations 1 and 2 used for the determination of dimethyldichlorosilane and methyltrichlorosilane account for small contributions from diphenyldichlorosilane. The sensitivity constants in Equations l t o 4 were determined by two different procedures. One set of sensitivity values was obtained from the mass spectrum of the pure compounds run a t known pressure. The other set of sensitivity values was obtained by substituting in Equations 1 to 4 calculated partial pressures and peak heights from the mass spectrum of a known solution of the four compounds. Since there were six unknown sensitivities and only four equations, it was necessary to determine the s?$and S?$ by multiplying the StP, determined from solution, by the peak height ratios and found from the mass spectrum of pure diphenyldichlorosilane. The calibration data are summarized in Table I.
I n actual practice Equations 1 to 4 are solved for P in terms of S and H and results calculated using a desk calculator. Observed peak heights are corrected for changes in instrument sensitivity as determined from the known pressure of a diphenyldichlorosilane sample. Results on Known Solutions. T o determine the accuracy of the method, mixtures of known composition were prepared with the following concentration ranges: (C&)zSiClt CsHeSlC13 MeSiCla MezSiClz
15-45 10-30 5-15 20-60
mole % mole yo mole 7c mole %
The solution calibration v, as obtained from a mixture of the four chlorosilanes having individual concentrations approximately a t the mid-point of the above ranges. The mixture used for calibration was not used to check accuracy. The known mixtures were analyzed a number of times and the results calculated using the calibration data obtained from pure compounds and a known mixture. Results were normalized to 100%. The errors as determined by the difference between the test results and the true values were calculated for each analysis. These errors were averaged with regard to sign and are shown in Table 11. Comparison of these mean errors shows that the solution calibration gives more accurate results. That this is a significant conclusion is shown by comparing the mean errors with the 99% confidence limits tabulated in Table 111. All the mean errors tabulated for the pure compound calibration are greater than the 99% confidence limits listed in Table 111, while only three (marked with asterisk) of the mean errors tabulated for the solution calibration exceed the 99% confidence limits. Thus all the results obtained from the pure compound calibration differ from the true values by more than esperimental error, while only three of the results obtained by solution calibration differ significantly from true values. The standard deviations and 99% confidence limits listed in Table I11 were calculated (12) from 10 replicate analyses of a known mixture using the solution calibration. These standard deviations and confidence limits also apply to results obtained from the pure compound calibration, since the same calibration was used throughout this investigation and thus has no effect on the precision of the results obtained. The 99% confidence limits for the average of three determinations were calculated by the usual procedure (fa) of multiplying the standard deviation by Student's t value and dividing by the
square root of the number of determinations. I t is realized that the 99% confidence limits would be somewhat lower for the results obtained from the solution calibration, since the average of four instead of three determinations was used. However, the slight difference was not significant in this investigation. In order to determine long-term accuracy, a freshly prepared standard was analyzed (sample 4, Table 11) using the solution calibration. The mean errors do not differ significantly from experimental error. The calibration was 5 months old a t the time of analysis. Also, in the time interval between calibration and analysis, a new filament had been installed in the source. I t is not unreasonable that the solution calibration gives results closer to the true values. This is so because the calibration data are obtained under conditions closer to those for actual analysis of unknowns. For example, the partial pressures obtained for the various compounds in the known mixture used for calibration are closer t o the partial pressures obtained from the known solutions used to check the method than is the approximately 100-micron pressure used for each compound during the pure compound calibration. Also, any interference errors ( 2 ) tend to cancel, since they are present during calibration as well as during analysis of unknowns. These effects could be investigated, but
they were beyond the scope of the present work. Errors Due to Hydrolysis. Because of the reactive nature of chlorosilanes with moisture, i t was necessary to determine the errors due to hydrolysis of samples submitted for analysis. It was found t h a t 1% by weight of water could cause errors of more than 10 relative 7,. However, leaving samples in uncapped bottles for 30 minutes caused no significant errors. This is much longer than the time needed to remove a sample for analysis. Any significant amount of hydrolysis can be detected by observing the polysiloxane peaks a t mass numbers 187, 207, and 281. The 281 peak, which is due to a fragment ion from the polysiloxane, is the most sensitive. Detecting the hydrogen chloride given off during hydrolysis is not satisfactory, since it might be lost from the sample because of its volatility. The HC1+ ion is frequently observed as background in the mass spectrometer. Pumpout Time after Analysis. The diphenyldichlorosilane has the lowest vapor pressure in the present system (b.p. 304’ C. a t 760 mm.). Two minutes after opening the reservoir to the pump the strongest peak a t mass number 154 was reduced to 0.2% of its original height. This is equivalent to 0.2 mole Yo diphenyldichlorosilane.
ACKNOWLEDGMENT
The author is grateful to A. S.Crouse, who obtained most of the experimental data for this n-ork. LITERATURE CITED
(1) Barnard, G. P., “Modern Mass Spectrometry,” pp. 65-7, Institute of Physics, London, 1953. (2) Bernstein, R. B., Semelup, G. P., Arends, C. B., ANAL.CHEM.25, 139 (1953). (3) Beynon, J. H., “Mass Spectrometry and Its Applications to Organic Chemistry,” p. 298, Elsevier, New York, 1960. (4) Brewer, S.D., unpublished work. (5) Dibeler, V. H., “Organic Analysis,” p. 418, Interscience, New York, 1956. (6) Dibeler, V. H., Reese, R. M., ANAL. CHEM.32,211 R (1960). (7) Grenoble, M . E., Launer, P. J., A p p l . Spectroscopy 14,85 (1960). (8) Grubb, H. M., Ehrhardt, C. H., VanderHarr, R. W.,Moeller, W. H., ASTM Committee E-14 meeting, Los Angeles, Calif., 1959. (9) Mea!, R. K.,Lewis, F. M., “Silicones, pp. 107-10, Reinhold, New York, 1959. (10) Norton, F. J., unpublished work. (11) Taylor, R. C., Brown, R. A., Young, 14’. S.,Headington, C. E., ANAL.CHEM. 20,396 (1948). (12) Youden, W. J., “Statistical Methods for Chemists,” Wley, New York, 1957.
RECEIVEDfor review April 27, 1961. Accepted September 7, 1961. 8th Annual Meeting of ASTM Committee E-14 on Mass Spectrometry at Atlantic City, N. J., June 26 t o July 1, 1960.
Report on Recommended Specifications for Microchemical Apparatus Oxygen Combustion Flask Committee on Microchemical Apparatus, Division of Analytical Chemistry, American Chemical Society AL STEYERMARK, Chairman, Hoffmann-La Roche Inc., Nutley, N. J. H. K. ALBER, Arthur H. Thomas Co., Philadelphia, Pa.
V. A. ALUISE, Hercules Powder Co., Wilmington, Del.
E. W . D. HUFFMAN, Huffman Microanalytical Laboratories, Wheafridge, Colo. E. L. JOLLEY, Corning Glass Works, Corning, N. Y. J. A. KUCK, College of the City of New York, New York, N. Y., and American Cyanamid Co., Stamford, Conn. J. J. MORAN, Kimble Glass Co., Vineland, N. J. C. L. OGG, Eastern Utilization Research Branch, Agricultural Research Service, U. S. Department of Agriculture, Philadelphia, Pa. C. E. PIETRI, U. S. Atomic Energy Commission, New Brunswick, N. J.
I
ACCORDANCE with the practice followed in previous reports of the Committee on Microchemical Apparatus (1, 9, 4,these specifications are for pieces of apparatus that are either the most widely used in their respective fields of application or are an improvement over such apparatus according N
to tests made by members of this committee and cooperating chemists. In this report, specifications are recommended for the conventional apparatus used in connection with oxygen flask combustion procedures (7, 8). Types of apparatus employing electric ignition (2, 6, 6) are not included be-
cause of lack of experience with these forms. Figure 1 shows two sizes of boro300- and 5Wml. silicate flasks-the sizes-and a glass stopper with platinum sample holder which is used with a flask of either size. Although oxygen flask combustion VOL. 33,
NO. 12, NOVEMBER 1961
1789