It should be noted that considerably larger sample sizes were required for the colorimetric analysis than were necessary for comparable emission measurements. For example, duplicate emission analysis of water sample E required 0.2 ml to obtain a measurement 14 times the blank, whereas the colorimetric procedure required 10 ml to obtain a measurement 8 times the blank. The water sample data included in Table I offer a point of interest. Sample A is from the runoff of Morning Glory Pool (a high arsenic source) in Yellowstone National Park while the remaining samples were taken at succeeding points down the drainage system from the same pool. Consequently, the
decreasing As concentrations reflect dilution and/or removal effects in the drainage field. In view of the results summarized herein, one may conclude that the microwave emission technique offers a convenient, sensitive, and reliable method for the determination of arsenic in a wide variety of sample types. Because the method is based on the separation of arsenic as arsine, it may be generally anticipated that the method will not be subject to deleterious chemical or physical interference problems. RECEIVED for review January 3, 1972. Accepted March 7, 1972. Research supported by NSF Grant No. GP-21306.
Determination of MetalIic-Element Impurities in Uranium Hexafluoride by Spark Source Mass Spectrometry Olin H. Howard Oak Ridge Gaseous Diffusion Plant, Union Carbide Corp., Nuclear Dioision, Oak Ridge, Tenn. 37830
URANIUM HEXAFLUORIDE (UFs) delivered to the Atomic Energy Commission (A.E.C.) for enrichment in zr5U must be of a specified purity (I). Specified maximum concentrations of some metallic-element impurities are shown in Table I. These elements are usually determined by several different optical spectrometric procedures (2), some of which necessitate preliminary separation of the elements from the U. The capability of spark source mass spectrometry (SSMS) for determining many elements in a single analysis, with high sensitivity and without separation from the matrix, prompted an investigation of its suitability for determining impurities in UFs. A procedure was developed by which nearly all of approximately 35 elements of interest can be determined to 1 ppm or less, with an average relative standard deviation of =t33 %, and an analysis time of 5 hours. Several papers on the determination of elements in nonconductive powders by SSMS have been written by other authors (3-7). Usually the sample is mixed with a conductive powder, e.g., graphite or silver, and pressed into solid form for sparking in the mass spectrometer. The procedure described herein is similar since the UF6 is converted to uranium oxide (U3Os) and compressed with silver powder for the analysis. EXPERIMENTAL
Procedure. The UF6 is sampled as a liquid (2),hydrolyzed, and diluted to a known uranium concentration of approximately 0.1 gram per gram of solution. To a weight of the (1) “Uranium Hexafluoride: Base Charges, Use Charges, Special Charges, Table of Enriching Services, Specifications, and Packaging,” Fed. Register, No. 32, 16289 (1967). (2) J. C. Barton, C. W. Weber, L. A. Smith, and W. D. Hedge, At. Energy Camm. Rept. K-L-6140,Oak Ridge, Tenn., April 1967. (3) G. D. Nicholls, A. L. Graham, Elizabeth Williams, and Margaret Wood, ANAL.CHEM.,39,584 (1967). (4) R. M. Jones, W. F. Kuhn, and Charles Varsel, ibid., 40, 10 (1968). (5) Riley D. Carver and Paul G. Johnson, Appl. Specrrosc., 22, 431 (1968). (6) G. H. Morrison and A. T. Kashuba, ANAL.CHEM.,41, 1842 (1969). (7) R. Brown and P. G. T. Vossen, ibid., 42,1820 (1970). 1482
ANALYTICAL CHEMISTRY, VOL. 44, NO. 8, JULY 1972
Table I. A.E.C. Specifications for Metallic-Element Impurities in UF6 Elements Ti Nb Ru Sb Ta Elements forming nonvolatile fluorides V Mo W 233U
Maximum concn, ppm 1 1 1 1 1
300(Total) 200 200 200 500
Basis U U U U U
U
solution containing 0.5 gram of uranium is added 5 pg of yttrium (10 ppm, U basis) as internal standard. (Yttrium was chosen because it was not expected to be in the UF6, was readily available, and did not interfere with the analysis.) Sulfuric acid, 0.25 ml, is added. The solution is evaporated, and the residue, U02S04,is ignited at 850 “C for 10 min to U308. (The HzS04enables removal of the fluorine by evaporation of HF, preventing loss of volatile metal fluorides.) The U308is mixed with 0.5 gram of high-purity (Cominco American) silver powder (weight ratio AgiU = 1). Two electrodes, 1/~6-in. diameter X in. long, are pressed with a molding die using a polyethylene mold (AEI Scientific Apparatus). Preliminary pressing of partial electrodes cleans the mold, particularly of fragments of polyethylene which contribute hydrocarbon background to the spectrum. A thin strip of tetrafluoroethylene placed between the sample and die acts as a lubricant, preventing distortion or breakage of electrodes. The electrodes are mounted in the mass spectrometer (AEI MS702) with Norbeck type, silver electrode holders (Vacumetrics, Inc.) This spring-clamp holder is preferred over the standard screw-clamp, tantalum holder for this application because it is more convenient, causes less electrode breakage, and is not a potential source of tantalum contamination. TO
further reduce tantalum contamination (from 10 ppm to 0.3 ppm), the No. 2 tantalum ion-source plate (first grounded plate) was replaced with a silver plate, thus enabling determination of tantalum to less than 1 ppm. The electrodes are sparked with 30 kV. The width of the ion-beam defining slit is 0.005 in. A series of exposures to 200 nanocoulombs (nC) is made ranging from 1 x on a photoplate, in the mje range of 7-250. Spark-pulse rate and length are increased gradually from 10 to 300 per sec, and from 25 to 200 psec, respectively, between the shortest and longest exposures. Changing the pulse rate and length possibly contributes to the observed analytical variability but is necessary, in the absence of a beam chopper, to the completion of the exposures within a practical time, e.g., 1 hr. The effect would be minimized if the standard and impurity lines could be measured on the same exposure. This is not always possible, however, because of the extremes of line intensities between the 10-ppm, monoisotopic Y and, e.g., a 1-ppm, multiisotopic impurity element. The exposed photoplate is developed, the optical densities of the 89Y+line are measured for several exposures, and a calibration curve of optical density DS. logarithm of exposure is plotted. The curve is practically linear in the optical density range of 0.03-0.3. Its use is confined to this range which, in this case, corresponds to an exposure range of 3-30 nC. Among the exposures obtained are 3, 5.5, 10, 17, and 30 nC; hence, 5 points are plotted on the straight-line portion of the curve. Although, theoretically, only 2 points would define the useful portion of the curve, 5 reduce the effect of scatter. The optical densities of selected isotopic lines of the impurity elements are measured. Generally, the single-charge line of the most abundant isotope of each element is used, for maximum sensitivity. Exposures are selected which give optical densities of less than 0.3 so as to be within the linear portion of the calibration curve. Impurity concentration is calculated in the conventional manner (8) using the concentration of the Y , the ratio of the standard exposure (obtained from the calibration curve) to the exposure of the measured impurity line, relative sensitivity factor, and published isotopic abundance and atomic weight. Results are corrected for reagent blanks which are determined by analyzing a pure solution of hydrolyzed UF6. Relative sensitivity factors (RSFs, ratios of uncorrected to known concentrations) are determined by analyzing a solution of hydrolyzed UF6 contain] ng known concentrations of the elements of interest. RESULTS AND DISCUSSION
RSFs, precisions, and detection limits for 31 elements and the isotope *3aU are shown in Table 11. The results are based on S separate analyses of a solution of hydrolyzed UFGcontaining 10 wt ppm of the impurities on a uranium basis. The RSFs (relative to 1 for Y ) range from 0.8 (Ta and Th) to 14 (Na) and average 1.6, excluding Na. The precisions, as relative standard deviations, range from +12% (Nb) to + 5 S % (Pb) and average 1 3 3 x . Most of the variability is probably due to nonhomogeneous electrodes (3). Two sets of detection limits are shown. Instrument limits can be attained with an exposure of 200 nC if the reagent blanks are zero. Lower limits could probably be attained for some elements with a larger exposure. Method limits include reagent blanks, most of which are 1 ppm or less. To ascertain the precision and accuracy with which the elements Ti, Nb, Sb, and Ta can be determined at their 1 ppm-specification level (Table I), each was added to a solu(8) R. D. Craig, G. A. Errock, and J. D. Waldron, Advan. Muss Spectrom., 1,145 (1959).
Table 11. Relative Sensitivity Factors, Precisions, and Detection Limits Elementa added Li Na Mg A1 P K
Ca Ti V
Cr Mn
Fe co Ni
cu Zn Sr Zr
Nb Mo
Sn Sb Ba Sm
Gd DY Ta W Pb Bi Th P33U
Average a
Re1 std RSFb 1 .o 14 1.5 4.5 1.5 3 .O 1.8 2.3 1.4 1.6 1.5 1.8 1.7 1.8 1.6 1.2 1.5 0.9 1.1 1.1 1.2 1.3 2.2 1.1 1.1 1.1 0.8 1 .o 2.7 1.8 0.8 1.7 1 .@
dev,
z
=t50 1 47 f27 127 138
i40 120
i.14
i25 i20 122 143 125 140 f24 i38 128 116 i12 135 i48 135 f33 +24 i24 i40 1 22 rt 36 i58 1 57 i50 -Ir 50 zt33
Detection limit, wt PPm Instrument Method 0.004 0.001 0.01 0.003 0.02 0.01 0.01 0.03 0.02 0.02 0.02 0.02 0.02 0.03 0.03 0.03 0.04 0.1 0.05 0.3 0.2 0.1 0.05 0.3 0.5 0.3 0.1 0.4 0.1 0.1 1
1 0.18
0.4 1 1 0.5 0.7 2 3 0.4 0.02 0.1 0.2 2 0.02 0.3 0.8 0.6 0.04 0.1 0.05 0.3 0.2 0.1 0.05 0.3 0.5 0.3 0.3 0.4 0.2 0.1 1 1 0.56
Concentration = 10 pprn, U basis. Relative to 1 for yttrium. Na excluded.
tion of hydrolyzed UF6 to a concentration of 1 pprn on a U basis, and 8 determinations were made. The RSFs in Table I1 were used in calculating the results shown in Table 111. The relative standard deviations for Ti, Sb, and Ta are considerably larger than at 10 ppm (see Table 11), possibly due to variations in small background interferences. The relative errors indicate that the RSFs determined at 10 ppm are applicable at 1 ppm, with the possible exception of Ti, which might be biased by a small background interference not properly corrected for by the blank. Metallic elements observed to be lost during the conversion of UF6 to U308are Ru (no spectral lines) and, to some extent, Cd, (weak and highly variable spectral lines), thus precluding their determination. The fact that the RSFs (which, in this case, are empirical correction factors) for the elements in Table I1 are approximately 1 or larger, relative to Y whose oxide is refractory, indicates that little, if any, of the elements is lost during sample preparation. (Some loss is acceptable, so long as it is reproducible, since the RSFs are overall, empirical correction factors.) N o other elements, besides those listed in Table 11, were studied. Typical impurity levels in the UF6 received for enrichment are usually lower than specifications, with a few of the more common, nonvolatile-fluoride elements being in the range of 1-10 ppm. ANALYTICAL CHEMISTRY, VOL. 44, NO. 8, JULY 1972
1483
Element Ti Nb Sb Ta
Added
Table 111. Precision and Accuracy at 1 pprn Concentration, ppm Found Found Blank (corrected)
1
2.0
1
1.1
1 1
1.o 1.4
1.6 1.1 0.9 1.1
0.4 0.0 0.1
0.3
Analysis time, beginning with the solution of hydrolyzed UF6, is approximately 5 hours. In summary, spark source mass spectrometry has been used to determine 31 metallic elements and the isotope 233U in UF6 with a detection limit of 1 ppm or less, an average relative standard deviation of *33 %, and an analysis time of 5 hours. Although the method might be categorized as semi-quantitative, its precision is considered adequate for the purpose, and its relative inclusiveness, simplicity, speed,
Re1 std dev, 7Z
Re1 error,
+56
60 10
+13 ir 56 i64
- 10 10
and sensitivity make it attractive for determining metallicelement impurities in UFb. RECEIVED for review December 3, 1971. Accepted March 16, 1972. Presented at Fourteenth Conference on Analytical Chemistry in Nuclear Technology, Gatlinburg, Tenn., October 1970. Work performed at the Oak Ridge Gaseous Diffusion plant operated by Union Carbide Corporation for the U. S. Atomic Energy Commission.
Exact Mass Measurement Accuracy from CEC 21-llOB Mass Spectrometer and Commercial Data System DS-30 R. S. Gohlke G . P. Happ,] D. P. Maier, and D. W. Stewart Research Laboratories, Eastman Kodak Company, Rochester, N . Y . 14650 A DOUBLE FOCUSING, high-resolution mass spectrometer (Consolidated Electrodynamics Corp., Model 21-1 lOB), designed for electrical and photoplate recording, has been interfaced with a commercially available data acquisition and processing system (DS-30, AEI Scientific Apparatus, Elmsford, N.Y. 10523). An in-depth discussion of the parameters that must be considered in the use of small computers to process high-resolution mass spectral data has been presented (I). Our initial work was done in 1970 using AEI’s DS-20, a less versatile program which gave entirely comparable mass measuring accuracy. In our case, the DS-30 system utilizes a PDP-8/1 computer (Digital Equipment Corp., Maynard, Mass. 01754) (4096word memory) with high-speed paper-tape punch and reader, two 32 K disks, an ADC-1 analog-to-digital converter, and an ASR33 Teletype unit. If additional storage capacity is required, an RF-08 disk (256 K words) may be used. The mass spectrometer-computer interface is supplied by AEI and consists of an adjustable RC-clock (1-12.5 kHz), signal biasing and threshold controls, and requires an input signal in the range of 0 to approximately - 10 V. Operating programs and data are stored on the disks and called into core at the appropriate time by a specialized disk monitor system, SERF (SERial File Disk Monitor). SERF and DS-30, of which it is a part, were written by Applied Data Research, Princeton, N.J. The ASR33 keyboard provides the operator with program control. Data output can be
obtained on the Teletype or high-speed paper-tape punch. (Data output can also be obtained on a line printer or on a magnetic-tape unit, but these facilities are not a part of our system.) Basically, DS-30 samples and digitizes the analog signal from the mass spectrometer at a rate determined by the adjustable clock (typically 7 kHz); rejects values less than the preset threshold; and, in real time, computes the area and time centroid of each ion signal comprising the mass spectrum. In practice, we scan the mass range of 900 to 50 in approximately 90 seconds. Subsequent to data acquisition, the program calculates exact ionic masses by comparison of the stored time centroids with those derived from the known masses of a perfluorocarbon reference material run in admixture with the unknown sample. Although any reference material, fluorinated or not, can be used, we normally choose high-boiling perfluorokerosene, supplied by Pierce Chemical Co., Rockford, Ill. In DS20, only PFK could be used. The conversion of time centroids to exact mass (which requires 5-10 seconds) can be delayed indefinitely to permit the uninterrupted acquisition of multiple scans on the same sample. Approximately 20 spectra can be stored on 64 K disks and over 100 can be accommodated if the RF-08 (256 K) disk is used. A description of the numerous operating subroutines and data output formats available within DS-30 is available elsewhere (2). In addition to a 0- to - 10-V signal, DS-30 requires that the mass spectrometer be capable of providing an exponential
Present address, P. 0. Box 28033, San Diego, Calif. 92128. (1) R. T. Klimowski, R. Venkataraghavan, F. W. McLafferty, and E. B. Delany, Org. Mass. Spectrom., 4,17 (1971). 1484
ANALYTICAL CHEMISTRY, VOL. 44, NO. 8 , JULY 1972
(2) P. Powers and M. J. Wallington, paper F3 presented at the
Nineteenth Annual Conference on Mass Spectrometry and Allied Topics, Atlanta, Ga., May 2-7, 1971.
