sharp peak of ferrocene. By stopping the spinning of the sample tube and slightly detuning the field homogeneity controls most of the ringing was eliminated. The audio modulator phase balance of the V-3521 Integrator was critical in measuring the area integrals, because the solvent peak and sample and reference peaks could not be adjusted to exact absorption modes siniultaneously with an improperly balanced 2KC audio modulation phase.
The residual intensity from the large solvent peak introduced error. In quantitative S M R , rf. saturation is always a problem. This is especially true of the present work in which the peak intensities of diluted polymer solutions are being measured. The possibility of error due to rf. saturation a t the rf. power level of 70-db. units a t the magnetic field sweep rate corresponding to about 15 c.p.s. per second was checked and found to be negligible.
LITERATURE CITED
(1) Porter, R. S., Nicksic, S. W.> Johnson, J. F., AKAL.CHEM.35, 1948 (1963). HVSGYrr CHEN' MARSHALL E. LEWIS
Research Division U. S.Industrial Chemicals Co. Division of Sational Distillers and Chemical Corp. Cincinnati 37, Ohio Present address, Research Division, Goodyear Tire& Rubber Co., Akron, Ohio
Com bustion-Conductometric Determination of Less Than 10 p.p.m. Carbon in Tungsten SIR: Tungsten metal which has been processed by floating-zone refining commonly contains less than 10 p.p.m. of carbon by weight. The quantitative determination of carbon in this material is of interest because it affords a method for measuring the effectiveness of the purification process and also because carbon apparently has an appreciable effect on the mechanical properties of the metal (6). However, quantitative determination of the carbon in this low range has proved to be a difficult task with the conventional combustion conductometric method. h limitation of the technique a t lorn concentrations is the magnitude and variation of the blank, a major part of which arises from materials such as iron and tin added to the crucible ostensibly to promote heating and to aid in the formation of a less viscous mixture. Haymes and Ollar (3) reduce contamination from this source by prefusing an electrolytic iron bath in a helium atmosphere. Although the presence of the additive materials in the crucible is apparently essential for the quantitative recovery of carbon from some metals and alloys (2, 4),it is questionable that it is necessary or desirable in the evolution of carbon from tungsten or other refractory metals. For example, it has been observed that when iron chips were fused in the presence of tungsten granules the unreacted tungsten metal was formed into an agglomerate apparently by the stirring effect of the high-frequency field. This has been observed by other investigators such as Huber and Chase (4)>mho recommend the crushing and optical examination of the fused sample for unreacted metal. In the preient work, it was found that rarely would a crushing procedure alone disclose the presence of unreacted tungsten. Careful sectioning of the crucibles, however, showed that in about SOYo of the fusions the tungsten metal could be found intact. Thus it appeared that the elimination of all additives to the crucible by heating the 1396
ANALYTICAL CHEMISTRY
sample directly would not only simplify the operation, but would be a major step toward achieving a reproducible blank and, therefore, a lower limit of determination for carbon. The purposes of this investigation were to achieve improved precision a t the lower carbon concentrations and, in addition, to prove the accuracy of the method when applied to refractory metals, particularly tungsten. The oxidation step was investigated, and a new combustion technique is recommended. Precision and accuracy are reported, and the resultant analytical procedure is outlined in detail. EXPERIMENTAL
Apparatus and Reagents. Induction furnace, Model 521; conductivity cell, illodel 515; aluminum oxide crucible, No. 528-35; crucible covers, S o . 528-42; zirconium oxide crucibles. N o . 501-45; iron chips, S o . 501-77; tin granules, S o . 501-76; tin capsules, Yo. 501-59; and quartzenclosed carbon crucible, No. 550-182 (Laboratory Equipment Corp.). Ba(OH)&olution, 1 gram of Ba(OH)2.8 H z 0 per liter. WC and W2C, 99.9% purity, (hdamas Carbide Corp.). Potassium acid phthalate, primary standard, aqueous solution containing 100 gg. of carbon per liter. Platinum disk susceptors, 1-inch diameter, fabricated from platinum foil and formed to fit the bottom of the sample crucible. Preparation of Samples. Samples obtained as solid metal are crushed in a hardened steel mortar until the bulk of the material passes a 16-mesh screen and is retained on a n 80-mesh screen. T h e small amount of material passing the fine screen is recombined with the sample granules because carbon segregation may occur. A considerable amount of iron may be introduced into the sample during this operation and may result in high values for carbon if not completely removed. This is accomplished by digestion in 1 : 1 HCI followed by thorough rinsing with water and drying with ether. Samples in powdered form do not require
special treatment, although the combustion conditions described later for pulverized metal are modified for powders. Procedure. Because instrumental reliability may vary from day to day, it is desirable to make some basic checks on the instrument prior to running samples. The variation in conductivity of the Ba(OH)2 solution subjected to a continuous flow of oxygen for a period of hour should not exceed 0.1 ohm. After completion of this test for instrument stability the furnace is conditioned by heating until readings for an 8-minute period do not exceed 0.3 ohm for multiple determinations, as determined with quartz-enclosed carbon or platinum susceptors of the type to be described. These readings were obtained with solutions that contained approximately 0.7 gram of Ba(OH)2.8Hs0 per liter. A sample weighing between 2 and 8 grams, depending on carbon content, is placed in an aluminum oxide crucible which has been previously heated in oxygen a t 900" C. for 15 minutes. hfter introducing the crucible into the furnace, pure oxygen is flushed through the furnace and conductivity cell for 2 minutes at a rate of 250 to 300 ml. per minute to divest the system of COn which may be picked up during the loading operation. The furnace is then energized for an 8-minute combustion cycle, although the formation of tungstic oxide is generally complete in 2 minutes as indicated by the rapid fall of the plate current. Dial readings are taken a t a selected period in the temperature cycle of the thermostatically controlled bath near the end of the collection period, The instrument used in this work gave the most reproducible readings 60 seconds after completion of a heat-on cycle. The instrument readings are converted to micrograms of carbon by u4ng a ])reviously prepared calibration curve. For powdered samples, place a plntinum disk, 1 inch in diameter and l'az inch thick under and in good thermal contact with the crucible containing the sample. An alternate procedure for igniting powders entails placing a
~
Table I.
Comparison of Blank Levels with and without Additive hdditive material and lot designation
Iron chips screened,a 1 gram 1
Total blank, pg. of carbon 26 0 Standarddeviation,bpgof 4 0 carbon Limits of detection per gram sample p p m range of 3sb Laboratory Equipment Corp
2 39 4 1 4
3 39 5 2 5
4 40 1 4 0
4 2 to 12 0
Tin granules," 1 gram
Tin capsules"
5 12 3 3 7
12 0
5 10 8 2 4
15 7 6 0
3 0
7 2to8 1
9 0to18 0
ElectrolYtic
Series of apparatus blanks, direct
19 2
11 3
1 1
2 7
4 9, 3 7, 4 3, 2 8 13,20,18,15
3 3to8 1
3 8to6 ( 1 7 to 3 O ) c
~
.I. - 1 Based on 2-gram sarrple square of platinum foil 0.003 inch X inch X inch, or larger, in the bottom of the crucible and covering it with mctal po\\der ( 1 ) . Both forms of susceptors should be previously conditioned by heating in the induction furnace under the conditions described for tungsten metal for a period of 2 minutes. Calibration Curve. Prepare an aqueous solution of primary standard potassium acid phthalate containing 100 pg. of carbon per nil. ( 5 ) and, with a microburet, add appropriate aliquots of solution to porous crucible covers, not exceeding an addition of 0.7 ml. a t one time. Dry the covers a t 110" C. for approximately 1 hour. Place the covers on top of a quartzenclosed carbon crucible and heat inductively for 2 minutss. Collect the evolved CO, for an addiional 6 minutes. In an alternate proi:edure, weighed increments of Kational Bureau of Standards sample 55e are contained in a zirconium oxide crucible and oxidized using the quartz-enclosed carbon crucible according to the previously outlined procedure. Since the blanks for the calibration and tungsten procedures are the same, it is convenient to plot all values as uncorrected. DISCUSSION AND RESULTS
Experiments were ccnducted to obtain definitive data on the limitations imposed b y a variety of typical additives when oxidized alone. The data from these experiments, shown in Table I, represent ten replicate determinations for each set of conditi0.w. Included in the table is a series of apparatus blanks obtained by heating a quartz-enclosed carbon crucible. These blanks are indicative of the ultimate detection limit for carbon by this procedure. ,4 conclusion drawn from these data is that the carbon content of the additives is a significant source of error in determination a t low carbon concentrations. Although this error can be minimized by careful selection of additive materials, a variation between lots of material is still likely to occur. Data are shown for the lower limit based o i the apparatus when using a 2-gram ?,ample, because this is the minimum recommended
weight for the direct combustion of tungsten under the conditions to be described. The data show that, to minimize the effect of carbon blank on the analytical result, it is necessary to oxidize the tungiten sample without resorting to the addition of other substances to the crucible. Investigations in this laboratory show that a number of metals including tungsten can be completely oxidized by direct induction heating-providing that the bulk of the sample granule.. are between 16 to 80 mesh. Tungsten chips larger than 16 mesh do not couple effectively; their surface/volume ratio ib too small to permit efficient induction heating, because in the megacycle frequency range the transfer of energy is
Table II.
primarily a surface phenomenon. Fine powders, while advantageous from the standpoint of surface area, provide a discontinuous path for high frequency energy and t,hus are also heated ineffectively. Tungsten metal granules of the proper size not only couple effectively, but, once the oxidation of tungsten commences, the metal is converted into a fused oxide mass by the exothermic reaction Fyhich raises the crucible temperature to greater than 1450' C. Samples received as powders finer t'han 80 mesh require adding wfficient heat to init,iate the oxidation, which is then selfsust,aining. The platinum susceptors described for this purpose are simple, inexpensive, and do not produce a measurable blank. If difficulty is encoun-
Recovery of Carbon Based on Potassium Acid Phthalate Calibration
(Per cent deviation of 0.1 ohm in blank correction: 7 at 30 pg. of carbon, 2 at 90 pg. of carbon) Recovered Recovered Added %; dev. Added % dev. from rg of from erg. of carbon carbon KHCsHdOr Sample KHCsH404 Sample WC + 2 grams of W 25 24 28 24 W2C t 26 25 0 2 grams of FV 10
+
28
26 -.
33 31
28
30 30 Av. 46 47 57 67 74 74 80
Av.
28 31 35 42 46 53 56
+7
C8 I
-
- 12 38 $12 +3.3 -13 2 +8 0
+
64
Av .
Fe (XBS 55e)
$11 +I1 -11 +lo +5
30 30 30 30
24
+I5 -9 +4.6
50 90 90 Av.
96 90 95 96 93 90
Av .
+4 +7 4-7 +2
++ 0
-3 0 12 +2 13 +11 +5 +7 4
+ +
1-3 $3 +l5 +6 0 -6 -6 +O +2
3 3 0 7 0 1 78 1
2 2
4 1
+4 4
+2 7 +4.3
VOL. 36, NO. 7, JUNE 1964
1397
tered in igniting powdered samples in this way, it may be necessary to increase the furnace power with the grid current t a p switch. I n the procedure described, the calibration of the conductivity cell is accomplished by combustion of measured amounts of either potassium acid phthalate or NBS 55e. This calibrating procedure, however, does not disclxe the degree of carbon recovery frJm the tungsten sample, and this is necessary to place the determination on a quantitative basis. The validity of reporting carbon in tungsten, based on recoveries from iron, steel, or organic compiunds, appeared to warrant further testing because these procedures are based on the assumption that the carbon contents of the metal sample and the calibrating material are totally evolved. Recovery experiments were designed to assess effectiveness of the evolution of trace amounts of carbon from tungsten metal, with the direct-combustion procedure, by doping samples with near stoichiometric WC and W2C. The carbides of tungsten were chosen because it appeared they would most nearly simulate the chemical nature of carbon in tungsten. The combustiongravimetric value for carbon did not significantly deviate from the stoichiometric value and the latter was therefore used in calculating quantities of carbon. T o facilitate weighing microgram amounts, the powdered carbides were blended with pure tungsten powder containing less than 10 p.p.m. of carbon to form a mix containing approximately 1% carbon by weight. The additional tungsten added to the crucible with the carbides was negligible and did not require a correction.
The experimental procedure consisted of adding a few milligrams of the WCtungsten or R2C-tungsten mixtures directly to a refractory crucible. The carbides were then covered with 2 grams of the tungsten-base material, which was prepared from electron-beam melted tungsten that produced a signal of 0.4 ohm for a 2-gram sample. Table I1 is a summary of recovery experiments for WC and W2C and includes comparative data for potassium acid phthalate and the iron standard, S13S 55e. The recoverieh from the carbides were coniistently greater than those obtained for the organic compound or the standard iron sample and may have been caused by slight differences in the blank corrections for the different procedures. The deviations were not considered significant for the purposes of this experiment, which adequately demonstrates the accuracy of the direct combustion procedure. Precision studies were made on 4gram samples of tungsten metal containing about 4.0 p.p.m. carbon. The relative standard deviation calculated for 10 determinations made over a period of one week was 12%. This improved precision results from the use of increased sample weights and a more constant blank. Since this method is successful in determining the carbon content in tungsten, the extension of the directcombustion method to metals such as molybdenum, tantalum, niobium, zirconium, titanium, chromium, and copper is also of interest. These metals will couple directly if the particle size is similar to that described for tungsten. For these more ductile metals it is often necessary to provide the proper size
by cutting or shearing. In sheet form up to l / , 6 inch in thickness, these metals couple and oxidize especially rapidly by placing a section in the bottom of the crucible with the sheet oriented perpendicular to the axis of the induction coil. Complete oxidation is observed for molybdenum only if the sublimation of molybdic oxide is allowed to go to completion. I t appears, therefore, that a special trapping device is desirable to prevent contamination of the conductometric system by the volatile oxide. Tantalum, niobium, zirconium, titanium, chromium, and copper undergo fubion of the oxides similar to tungsten. Preliminary recovery experiments using TaC and S b C in the metals indicate quantitative recover-y of carbon under the conditions described for tungsten. LITERATURE CITED
(1) Edwards, G. W., Linde Co., Division of Union Carbide, Indianapolis, Ind.,
private communication, 1962.
(2) Elwell, JV. -,T., Wood, D. F., “The
Analysis of Titanium, Zirconium and Their Alloys,” p. 23, Wiley, 1961. ( 3 ) Haymes, J. E., Ollar, A., Bureau of Mines RI-6005, p. 17 (1962). ( 4 ) Huber, F. E., Chase, D. L., Chemist Analyst 50, 71 (1961). ( 5 ) McKinley, T. D., Advisory Group for Aeronautical Research & Development Working Paper M-33, Analysis of Refractory Metals. (6) Stephens, J. R., AIME Meeting, October 28-November 2, 1962. WILLIAMA. GORDON JUDSON W. GRAAB ZITAT. TUMNEY Lewis Research Center National Aeronautics and Space Administration Cleveland, Ohio 44135
Separation and Determination of Uranium in High Zirconium Alloys SIR: The need for the quantitative estimation of uranium in zirconiumaluminum alloys arose from pyrometallurgical studies on the purification of uranium in zirconium-uranium alloys. The determination of microgram quantities of uranium in the zirconium alloys (Zr.%ls) posed a problem because of the high mole ratio of zirconium to uranium of approximately 6 X lo4. Several spectrophotometric procedures have been reported f6r the determination of uranium in the presence of large amounts of zirconium ( 2 , 6 ) . The procedure reporting the highest tolerance for zirconium states that zirconium interferes when the zirconium-touranium mole ratio exceeds 2.4 X l o 3 (5). A fluorometric procedure is reported for the determination of uranium in zirconium metal (’7). This procedure 1398
ANALYTICAL CHEMISTRY
uses ethyl acet,ate as an extractant and appreciable zirconium is extracted with the uranium. The quenching of uranium fluorescence due t,o the extracted zirconium is compensated for by making a standard addition of uranium to an aliquot of the sample being analyzed. Thus a technique which would quantitatively separate uranium from zirconium and eliminate quenching would be desirable. h comprehensive discussion of separation procedures for uranium is given in the treatise (4). The desired species, when present as a trace constituent, is usually separated from the ‘matrix. In some instances, large quantities of impurities are removed from the desired constituenti.e., solvent extraction of impurities from uranium(ITl) by cupferron or diethyldithiocarbamate. However, removal of matrix quantities by precipitation is not
common because of loss of the desired species by coprecipitation. K e have used the bromomandelic procedure to separate fission-product zirconium plus carrier zirconium from an aluminum-magnesium matrix. The zirconium bromomandelate precipitate carries approximately 11% of the niobium-95 and less than 1% of fissionproduct strontium, yttrium, ruthenium, or cerium. h study of the coprecipitation of uranium by the bromomandelate precipitate was made to evaluate the possibility of a direct determination of uranium in the filtrate. EXPERIMENTAL
Apparatus. An Atomic 1095 Scalcr equipped with a 27r flow counter was used for alpha counting. The fluorometer has been described ( 1 ) .
