tively-Le., they were issued t o E.G.O. as “unknowns.” The precision of the procedure, not including the weighing step or preliminary separation of interfering ions from aluminum, was determined by analyzing ten aliquots from the same sample solution after separation of interfering ions. Results for sample 62 (Table 11) are shown in Table 111. These values show that the change from aqueous medium to absolute methanol, with subsequent development of the color,
can be accomplished nith good precision. LITERATURE CITED
W.,Neuman, W. F., hlulvran, B. J., ANAL. CHEM.21. 1358 (1949). . ( 2 ) Fletcher, hl. H., White, C. E., Pheftel. hl. S.. IND. ENQ. CHEM.. AKAL.ED. 18. 179 119461. (3) Harvey, A: E.,’ Manning, D. L., J. Am. Chem. SOC.72, 4488 (1950). (4)Jcb, P., Ann. chim.9, 113 (1928). ( a ) hul’berg, L. M., Mustafin, I. S.,
(1) Cucci, hl.
Dokladg Akad. h-auk S.S.S.R. 77, 285
(1951).
(6) Marshall, P. G., J . Chem. SOC.1931,
3206. ( 7 ) Sandell, E. B., “Colorimetric Determination of Traces of Metals,” 2nd ed., Interscience, New York, 1950. (8) Schemer, J. A., Mogerman, W. D., J. Research Natl. Bur. Standards 21. 103 (1938). (9) Yoe, J. H., Jones, A. L., IKD.ENG. CHEM.,AZAL. ED.16, 111 (1944). RECEIVED for review August 5, 1958. Accepted October 29, 1958.
Spectrographic Determination of Aluminum, Iron, Potassium, and Silicon in Tungstic Oxide THOMAS
J.
VELEKER and RUDOLPH DYCK
Chemical & Mefallurgical Division, Sylvania Elecfric Producfs Inc., Towanda, Pa.
b A method for the quantitative spectrographic determination of aluminum, iron, potassium, and silicon in tungstic oxide uses a powder-pellet spark technique for the aluminum, iron, and silicon determinations, and a high voltage alternating current arc method for potassium, The spark excitation of tungsten in the presence of graphite powder causes an erratic formation o f tungsten carbide, making it very difficult to obtain reproducible intensity ratios using secondary tungsten lines for internal standardization. In contrast, the addition of anhydrous sodium tungstate as a buffer in the preparation of the pellets gave a very reproducible spark which stopped the formation of tungsten carbide during the sparking cycle. This procedure i s used as a routine control for aluminum, iron, potassium, and silicon, in one step of the manufacture of tungsten wire and rod from the ore. It covers the following elements and ranges, aluminum 0.005 to 0.1 5%, iron 0.008 to O.l%, potassium 0.025 to 0.35%, and silicon 0.02 to 0.50%.
T
HE manufacture of tungsten wire and rod for the electrical industry requires stringent control of the physical and chemical properties throughout the stages of manufacture from the ore to the finished product. Tungsten cannot be smelted like many of the common nonferrous metals because of its high melting point. Consequently, the process involves a combination of hydro- and powder metallurgy. The control of impurities is very important in the chemical processing steps from the digehon of the ore in sodium
rial to tungstic oxide. After firing, the material is blended on a Wig-L-Bug t o ensure homogeneity. One part of sample, 1 part of SP-1 Rational graphite powder, and 1/4 part of pure anhydrous sodium tungstate by volume are mixed thoroughly for 30 seconds in a Wig-LBug using a 1-inch polystyrene vial and 3/8-in~hPlexiglas ball. Enough sample is mixed to conveniently prepare two pellets ( 3 / l e x inch) on an Applied Research Laboratories’ briquetting press. The pellets are prepared by applying 1000 pounds’ pressure (28,000 p.s.i.) for 5 seconds, the pressure is released, and then applied for 10 more seconds. A firm pellet weighing about 180 mg. is formed. Samples are sparked in duplicate. DETERMINATION OF POTASSIUM. The blue oxide is oxidized as above. The resulting tungstic oxide is mixed with dry lithium sulfate. Six hundred milligrams of tungstic oxide are mixed with 100 mg. of lithium sulfate in the KigL-Bug for 30 seconds. The sample is then loaded in triplicate into brass electrodes by tamping. ELECTRODES. High purity graphite electrodes are used for the aluminum, iron, and silicon determinations. The sample-bearing electrode is inch in diameter by 3/4 inch in length, with crater 0.194 inch in diameter and inch in depth The counter is a 15” cone-shaped electrode inch in diameter and 3/4 inch long. For the potassium determination, the electrodes are made from free machining yellow brass rod inch in diameter. The sample-bearing electrode is ‘/4 inch in diameter by inch long, with a PROCEDURE crater inch in diameter by l / s inch in Sample Preparation. DETERMI- depth. The sides have a 15” slope, so that a t the crater top the wall thicknesb NATION OF ALUMINUM, IRON, AND SILIis 0.010 f 0.002 inch. The brass counter CON. The blue oxide (WaOll) is oxidized electrode is the same design as the in fused silica combustion boats a t graphite counter electrode above. 750” C. for 1 hour to convert the matehydroxide to the final gray metal powder. The metallurgical and chemical makeup of the tungsten metal powder greatly influences the quality of the wire and rod made from it ( 7 ) . This spectrographic method offers a rapid, precise control of aluminum, iron, potassium, and silicon in one of the processing steps. Pure tungstic oxide is produced and then partially reduced to the tungsten blue oxide (W4011). At this stage several of the impurities which have been removedalumina, potassium, and silica-are carefully added in trace amounts to the blue oxide. These additives act as cleansing agents during the sintering of the ingots pressed from the powder and also as grain growth inhibitors. The iron is a residual impurity and must be controlled, because it lowers the recrystallization temperatures in tungsten Tire (6, ‘7). The blue oxide is then reduced in hydrogen to the gray metal powder. Little has been published on the spectrographic determination of trace impurities in tungsten. Pheline (6) lists a method for determining silicon in tungstic oxide using a 220-volt, 2ampere arc. Gentry and Mitchell (1) and Lounamaa (3) describe the analysis of tungstic oxide using direct current arc excitation with graphite as a buffer, and formation of tungsten carbideduring the arcing cycle.
YQl. 31, NO, 3, MARCH 1959
387
I .o
I
I
I
I
I
SI 2 5 0 6 8 9 9 K 7664907
7
0
-
IO
20
w
30
z
2
t
40
v1
4
t" c z V W
50
60
W K
a
j
70
1
I
I
1
1
80
90
100 0-20
20-40
40.60
60-80
80-IOC
100-120
120-140
TIME-SECONDS
The following equipment is used: a Bausch & Lomb Optical Co. large littrow quartz spectrograph; National Spectrographic Laboratories Inc. spark source, Model J4200; high voltage alternating current arc, 2200 volts, varying from 2.5 to 15 amperes; Eational Spectrographic Laboratories Inc. projection microphotometer; Applied Research Laboratories developing machine; Applied Research Laboratories briquetting press; and a Spex Industries, Inc., Crescent Wig-L-Bug. Excitation conditions are listed in Table I. Plate Development. The spectrographic plates are developed in an Applied Research Laboratories developing machine a t 21' C. for 2.5 minutes using D-19 developer. Chrome alum solution is used for 1.25 minutes to condition the emulsion for fast drying; then the plates are fixed in
Table 1.
Figure 2.
Curves show rates of volatilization of elements during excitation in relation to lines chosen for internal standardization
hypo solution until clear. The plates are washed for 3 minutes and dried rapidly with heat. Emulsion Calibration and Photometry. The calibration curve of the emulsion is determined with a rotating eight-step sector (ratio 1 t o 1.585) placed directly before the slit. For the aluminum, iron, and silicon determinations, an EK33 plate is used with an iron arc. The iron lines used are 2817.505 and 2838.120 A. (4). From the stepped spectrogram, the selected iron lines are read on the densitometer. For the potassium determination, an I L plate is used with a tungsten lamp to calibrate the visible region.
Excitation Conditions
-41, Fe, and Si Determinations Excitation Spectrograph N.S.L. spark source Wave length region, 2400-3200 A. 20-micron slit width Secondary voltage, 10,000 Inductance, 200 ph. added 1.8-mm. slit height Capacitance, 0.01 pf. Rotating sector, lOO7? opens Resistance, residual 62.5 cm., source to slit distance R.F. amperage, 7.2 Primary voltage, 160 Sparkpower, set at 7 Discharges/half cycle, 5 Auxiliary air gap, 3 mm. Air pressure, 2.25 p.s.i. Analytical gap, 2 mm. Exposure time, 30 seconds, n o preburn Conditions for K Determination High voltage alternating current arc Wave length region, 4000-8500 A. Volts, 2200 alternating current 20-micron slit width Amperes, 6 1.8-mm. slit height Analytical gap, 1 mm. Rotating sector, 7570 open4 Exposure time, 60 seconds 62.5 cm., source to slit distance No preburn a Rotating sector at 100yGopen allows 50% of incident radiation to reach slit. ~~
388
ANALYTICAL CHEMISTRY
Moving plate study
The transmittance readings are plotted on semilogarithmic paper against the assigned relative logarithmic intensity values. From the emulsion calibration curves, the working curves are plotted by relating the relative logarithmic intensity ratios, obtained from the transmittance of the selected line pairs, with respect to concentration (Figure 1). Two standards are placed on each plate to correct the logarithmic intensity ratios of the unknown samples to the working curves. EXPERIMENTAL
It was advantageous during preliminary investigations on sample preparation to convert the tungsten blue oxide (WaOIl) to tungstic oxide (WOa) for analysis, because the blue oxide powder is inconvenient to handle; also, the trioxide form is stable and forms firm pellets for sparking. Pellets were made by mixing the tungstic oxide with pure anhydrous sodium tungstate and graphite. In determining optimum excitation conditions, it was difficult to obtain reproducible spectra and logarithmic intensity ratios. Experimentation, confirmed by x-ray diffraction studies, showed that the tungsten combines erratically with the graphite during sparking to form tungsten carbide. Because its formation is not uniform, the spark does not consume the sample in a reproducible manner. This carbide formation in the spark d e presses the tungsten spectrum and enhances the impurity lines which result
I
0
I
I
I
4 +I F
Figure 3. Moving plate study on logarithmic intensity ratios independence of logarithmic intensity ratios with time is shown during first port of ex. citation period
I W 2504.527
I AI 3092 285 I W 3092 713
1 I
I
7 '
i 1
run under these conditions (Figures 2 and 3). Sample preparation conditions are given under procedure. N o background correction was needed to construct the working curve. PREPARATION OF STANDARDS
1 Fe 2599.396
J
Standards were prepared on a labI K E ~ ~ S Goratory scale similar to the manner in 1L'812652, which these elements are added to the k i l blue oxide in production batches. A K 1 I I I I 1 master standard was made by taking 0-20 20-40 40-60 60-80 80-100 100-120 Dure blue oxide and doping it with TI ME - SECONDS standard solutions of the -elements. The solutions were added in quantities so as just to moisten the powder. Care in variable intensity ratios. The addiobtained in the concentration range was taken to avoid contact between the tion of anhydrous sodium tungstate desired. More than half the sample solutions and the vessel containing the depresses this effect-that is, it causes remained after 2 minutes of sparking. powder. the spark to consume the sample uniFor the determination of potassium, The resultant material was dried, formly and prevents the formation of high voltage alternating current arc ground in a mortar, then oxidized in carbide. The sodium tungstate must excitation was used. The volatilizafused silica boats a t 750" C. for 1 hour. be anhydrous or the resulting spectra tion of tungstic oxide in graphite There was no silica pickup during oxidawill be depressed and reproduce badly. electrodes gave considerable background tion. After firing to tungstic oxide, Tungsten emits a complex spectrum and continuous radiation in the visible. the powder was screened and blended in the ultraviolet coupled with intense The use of brass electrodes gave spectra to produce a homogeneous mixture. continuous background radiation. The almost entirely free of background. The prepared master standard contained choice of analysis lines is limited because Potassium-free lithium sulfate was added 0.15% aluminum, 0.10% iron, 0.47% of interferences by secondary tungsten to the tungstic oxide to serve as an silicon, and 0.32% potassium. From lines. Tungsten was used as an internal internal standard. this standard, a series of lower standards standard for the spark method and Volatilization of the potassium ocwas prepared by the dry powder techlithium was added as internal standard curred during the first part of the arc nique of mixing varying amounts of for the potassium determination. The burn. A 60-second exposure at 6 pure tungstic oxide with the master spectral lines chosen are tabulated in amperes, 2200 volts alternating current, standard. The base material was sufTable 11. gave reproducible logarithmic intensity ficiently pure and a determination of Various spark parameters were inratios. A moving plate study was these elements was not required. vestigated by running moving plate studies under various excitation conditions. Using 0.005-pf. capacitance, the inductance values ivere varied from Table 11. Spectral Lines, Range, and Precision 300 to 800 ph., and the discharges per Element Line, A. Internal Std. Line, A. Range, % PrecisionIuyo half cycle from four to seven. At 7 A1 3092.713 0.005-0.15 W 3092.285 0.01-pf. capacitance, the inductance 4 Fe 2599.396 0.008-0.10 W 2601.433 9 K 7664.907 0.025-0.35 Li 8126.52 was varied from 100 to 500 ph. and the Si 2506.899 JT 2504,527 0.02 -0.50 5 discharges per half cycle from three to a Precision is expressed as the coefficientof variation, v , and is calculated as follows: six. KOadded resistance was put into the circuit because this mould cause a decrease in the intensity of the spectrum and some loss in sensitivitv. where c = av. concn., per cent A capacitance of 0.1 pf. with 200-ph. d = differenceof detn. from the mean inductance and five discharges per half 7t = No. of detns. cycle gave the best sparklike condensed discharge with an approximate Table 111. Data on Precision linear decrease in emission of radiation. Per Cent Moving plate studies under these conElement A1 Fe Si K ditions are shown in Figure 2 and 3. The discharge was stable and relatively Standard I (doped) 0.042 0.018 0.075 0.040 Spec. results on 0.040 0.018 0.080 0.036 independent of time a t five discharges. std. 1 0.044 0.017 0.078 0.040 Above six discharges per half cycle, 0.038 0,019 0.070 0.036 the spark became somewhat violent 0.040 0.020 0.070 0.044 and hard to control. 0.040 0,018 0.065 0.041 0.040 0,018 0,080 0.034 The 0.01-pf. capacitance in the oscillatory circuit was needed t o give Chem. Spec. Chem. Spec. Chem. Spec. a sufficient flow of current through the Sample electrodes, resulting in increased emis541 0.015 0.017 0.12 0.14 sion. The 200-ph. inductance damped 543 0.0025
