I-
z
W
a
1.0
50
VOLTAGE
I
‘Ail
L/
0
i
Figure 5. chloride
Cadmium in 1-butanol, 0.1M with lithium
J
Cell realstance, 1.3 megohms
1.0
VOLTAGE
Figure 4. chloride
Cadmium in 0.1M aqueous potassium Cell resistance:
Curve a, 500 ohms Curve b, 1.2 megohms
erratic electrode behavior. As was mentioned in the earlier article (d), maximum suppressors help greatly. I n addition, the use of rubber connection in the dropping-mercury-electrode assembly eliminated certain similar troubles which had been encountered when Tygon tubing was employed. WAVE CHARACTERISTICS
In Figure 4, two polarograms of cadmium in 0.1M aqueous potassium chloride are shown. Curve a was obtained with a cell having a resistance of 500 ohms, employing a Sargent Model XXI Polnrograph without the cornpenaator, while for curve b, the same solution was used in a cell of 1.2-megohm resistance, the polarograms being made on the aame.Polarograph but with the compen-
sator. In Figure 5 is shown the wave for cadmium in 1-butanol, 0.1M with lithium chloride, which was obtained on the same Polarograph with the compensator. Cell resistance in this case was approximately 1.3 megohms. Numerous runs with this apparatus have shown the half-wave potentials to be independent of resistance. Testa of diffusion currents have shown them to be independent of resistance and have also shown the usual linear relationship between diffusion current and concentration. Numerous tests of these principles have been made and the apparatus also has been employed, with excellent results, for amperometric titrations in such solvents as 1:l mixtures of iso-octane and %propanol. These amperometrio titrations will be the subject of future papers.
ACKNOWLEDGMENT
The authors express their appreciation for the financial assistance by the Research Foundation of Oklahoma State University and for valuable assistance by Paul Sherrick and John Lulay of E. H. Sargent & Co. in constructing the cell and in designing and constructing the compensator circuit which was used in most of this research. LITERATURE CITED
(1) Arthur, Paul, Lewis, P. A., Lloyd, N. A. ANAL.CHEM.26, 1853 (1954). (2) Arthur, Paul Lewis, P. A., Lloyd, N. A., Vanderkarn, R. H., ZW., 33, 488 (1961). (3) Arthur, Paul, Lyons, Harold, Ibid., 24,1422 (1952). (4) Jackson, W., Jr., Elving, P. J., Zbid., 28,378 (1956). (5) Kelley, M. T., Fisher, D. J., Jonea, H. C. Zbid., 32,1262 (1960). (6) Keliey, M. T.,Jonea, H. C., Fisher, D. J., Zbid., 31, 1475 (1959). (7) Nicholson, M. M.,Ibul., 27, 1364 ( 1955). (8) Oka, S.,Zbid., 30, 1635 (1958). RECEIVEDfor review M a y 9, 1960. Accepted March 6, 1961.
Spectrog ra phic Ana lysis of Silicon Tetrachloride for Trace Amounts of Boron THOMAS J. VELEKER and EM11 J. MEHALCHICK Chemical and Metallurgical Division, Sylvania Electric Products, Inc., Towanda, Pa.
b An emission spectrographic procedure is described for determining boron in silicon tetrachloride in the range 0.8 to 50 p.p.b. Samples are preconcentrated by partial hydrolysis with a dilute aqueous methyl cyanide solution. A direct current arc technique is used with a Stallwood jet and argon gas. The coefficient of variation for the procedure is 14%.
S
is a basic raw material used in the preparation of high purity silicon for semiconductor uses. The control of impurities in this material is very important. Boron is particularly troublesome since it is difficult to remove by methods of zone purification once the bar of silicon has been produced. It has a distribution coefficient of 0.9 and ILICON TETRACHLORIDE
does not move much during zone refining (1).
Traces of boron have a pronounced effect on the electrical properties of silicon. Boron can be determined in the silicon by calculation from electrical measurements; however, compensating impurities must be removed or the results are ambiguous. There were no methods found in the literatire for deVOL 33, NO. 6, MAY I961
* 767
P
z 0 L
LINE
P
I Z 4 V 7 133
e,
Z4*?6 I 24S6 778 C u Z 6 3 4 931
OIL'
'
'
-*
' -2
'
I
'
'
'
i z
0
'
'
14
'
LOG I N T E H S I T V RdTlO
Figure curves
1.
Spectrographic
working
termining traces of boron in silic6n tetrachloride. There are several references for the determination of this element in silicon, but only one method was able to detect boron down to 1 p.p.b. Morrison ( 4 ) describes a spectrographic procedure for determining boron in silicon. It is based upon a preconcentration technique using dialysis through a cation-permeable membrane. EXPERIMENTAL
Apparatus. A Bausch and Lomb dual grating spectrograph is used with a 110-2s source unit (National Spectrographic Laboratories, Inc.). The direct current is rated at 300 volts with current control from 5 t o 30 amperes. A N.S.L. projection-type microphotometer is employed. A Stallwood jet, purchased from Spex Industries, Inc., is used with argon gas during excitation. PRECONCEN"R4TION EQUIPMENT. A 50-ml. transparent quartz crucible is used to hold the silicon tetrachloride. An argon chamber about 3 inches in diameter and 5 inches high was made out of transparent quartz. The top of the chamber is closed with the exception of a '/,-inch port to introduce
Table 1. Excitation Conditions 300-volt d.c. arc SA-1 spectrographic date
13 ampercs
30;060 grating, 4500 to 5500 A., 2nd order 45-second exposure, 0.0 biprism setting no preburn Zmm. arc gap, 1.0-mm.primary apphotograph off erture cathode Zstep lens before 51-cm. source to slit slit (100 and 33% distance TI
Stdwood jet with 20-micron slit width argon gas 2O-c.f.h. argon gae 3-mm. slit height flow ~~~~~
768
~~
~
ANALYTICAL CHEMISTRY
the argon gas a t tho flow rate of 10 c.f.h. The bottom of the chamber is open and rests directly on the hot plate over the sample crucible. Two escape ports about l/, inch in size are located on the bottom side to allow the argon gas and evaporated silicon tetrachloride to escape. Other miscellaneous pieces of apparatus used included ti magnetic stirrer, a Teflon-covered magnet, and a Wig-GBug from Spex Industries, Inc. Reagents. The reagent used in the preconrentration step is prepared by adding 6 ml. of distilled water to methyl cyanide t o give 50 ml. of solution. The amount of water used determines the weight of the residue obtained, which in turn determines the concentration factor. This may be varied from 3.0 to 12 ml. of water per 50 ml. of solution if other concentration factors are desired. The buffer is prepared by mixing cupric hydroxyfluoride (CuOHF) in National SP-2 graphite powder to give a 25y0 mixture. The spec pure copper salt is purchased from Spex Industries, Inc. Procedure. Samples are preconcentrated for analysis. Sixty grams (40.5 ml.) of silicon tetrachloride are placed in a 4-ounce polyethylene bottle which contains a Teflon-covered magnet. With agitation on a magnetic stirrer, 15 to 20 drops of a n aqueous methyl cyanide solution are added. The bottle is then capped and agitated for 15 minutes, after which the suspension is allowed to stand for 16 hours. The suspension is transferred to a 50ml. weighed transparent quartz crucible and gently evaporated to dryness in an argon atmosphere on a constant temperature hot plate. The remaining silica and coprecipitated boric acid mixture are dried for 1/2 hour a t 150' C. and then fired for another hour a t 500' C. in a muffle furnace to remove the last traces of hydrochloric acid and water. The crucible and contents are then cooled to room temperature in a desiccator, reweighed, and the residue is transferred to a vial for analysis. The final weight before and after transfer is recorded. The residue will weigh from 60 to 80 mg. Maximum transfer is desirable and 90 to 95% is usually attainable. The transferred silica residue is blended with equal parts by weight of buffer. Samples are mixed for 30 seconds in 1-inch vials on the Wig-L Rug. The samples are then loaded into United high purity graphite electrodes (Type L1988) by tamping the crater full, and are then arced in duplicate. The counter electrode is a 15' coneshaped electrode l/, inch in diameter. Anode excitation is used with the direct current arc. Excitation. The excitation conditions a r e listed in Table I. The sample-bearing electrode is placed into the Stallwood jet, which is attached to the lower clamp of the arcspark s t m d . The electrode tip is adjusted to protrude 1.5 mm. above the jet. T h e argon gas is set to flow
? !
?O
i t 00
so
?a
TRUE 8 0 R E A D I N G B E N E A T H
*o
.O
100
8 2 4 9 7 . 7 3 3 4 . LINL
Figure 2. Curve used for background correction
a t the rate of 20 c.f.h. during the exposures. The radiation from the analytical gap is focused on the primary aperture of the spectrograph to photograph off the cathode side of the discharge. Plate Development. The spectrographic plates are developed in a Jarrell-Ash developing machine at 21' C. for 2.5 minutes using D-19 developer. An acetic acid stop solution is used for 30 seconds followed by fixing in Eastman Kodak rapid fixer. Plates are rinsed in running water for 3 minutes, then dried rapidly with hot forced air. Photometry. T h e SA-1 spectrographic plate is calibrated by the two-step method (ratio of steps 1 to 1.585). An iron spectrum is excited by a low amperage direct current arc on iron powder in a crater electrode. A molten bead forms which gives a steady arc. From the emulsion calibration curve, the working curves are drawn by relating the relative logarithmic intensity ratios of the selected line pairs to concentration (Figure I). An 8-micron slit is used on the N.S.L. microphotometer. Each spectrum is photographed through a two-step lens giving 100 and 33% transmission. The boron line at 2497.733 A. is corrected for background to improve the linearity of the curve at low concentrations. The main background in the spectrum is not due to continuous radiation, but to the presence of weak S i 0 bands. The argon gas flow through the Stallwood jet does not completely depress these bands. One light band lies directly beneath the boron line. Its transmittance is determined by plotting the transmittance of another S i 0 band line a t 24973 A. against the band line when boron is absent a t various exposure levels (Figure 2). From this curve the true transmittance reading beneath the boron line can be determined so a background correction can then be made. Calculation of Results. The working curves are constructed on the basis of a concentration factor of 1000. Sixty grams of silicon tetrachloride are treated to give a rcsidue of about 60 mg. of silica. This residue can wcigh more or less than
60 mg. since the amount of the water in the methyl cyanide solution can be varied. Also, some differences are due to sample manipulation. Consequently, if the final weight before transfer deviates from 60 mg., the result obtained from the working curves must be corrected as follows.
-
boron (p.p.b. from ourve) x final wt. in mz. before tranefer
BorOn(,.,.b.)
60
Preparation of Standard Samples. Standard samples were prepared by two techniques t o establish the working curves and check the preconcentration procedure. BORICACID METHOD. The first series was made by doping pure silica with known quantities of boric acid. The silica was prepared from purified silicon tetrachloride by the direct hydrolysis in distilled ice water. A tenfold excess of water was used. The resultant silica was fired at 500" C. to remove the last traces of water. A standard boron solution was made by dissolving powdered boron metal in nitric acid. A graduated series of boric acid standard solutions was added in sufficient quantities to just moisten the silica and give a range of 0.4 to 50 p.p.m. of boron. Care was taken to avoid cont a c t between the solution and the vessel containing the powder. The standard silica samples were then fired again at 500' C. for hour. BORONTRICHLORIDE METHOD.The second series of standards was made by doping silicon tetrachloride with known quantities of boron trichloride. A known quantity of purified silicon tetrachloride wae weighed in a dry volumetric flask. The boron trichloride was liquefied in a liquid nitrogen trap, and a small amount added to the volumetric flask which was then closed and allowed to come to room temperature. The contents were reweighed and the amount of boron added waa calculated. From this master standard, dilutions were made with pure silicon tetrachloride to give a series of standards in the part per billion range. Since the components are volatile and are hydrolyzed easily, care must be taken to maintain a dry system throughout the preparation of the standard solutions. RESULTS
I n order to obtain a sensitivity in the part per billion range for boron, a preconcentration technique is needed. Using a dual spectrograph with standard optics before the slit and a 30,000-lineper-inch grating, the boron was detectable at 0.5 to 1.0 p.p.m. when photographed in the second order. Consequently, the samples needed to be concentrated by a factor of 1OOO. A Stallwood jet with argon gas was used to improve the reproducibility and depress the Si0 bands which cause interference with the boron lines used.
TYL,SECONOS
Figure 3.
Moving plate study
Table II. EfFect of Reaction Time on Amount of Boron Recovered
Reaction Time, Hoursa 0 2 6 16 16
Boron, P.P.B. Added Found 2.1 3.1 2.3 3.1 2.9 3.1 2.9 3.1 Residual
