Counting Properties of Toluene Triton Water Mixtures. Because the greatest difficulties are found in the counting of weak /3 emitters, tritium has been used to evaluate the various mixtures. As shown in Table I, addition of the several Triton surfactants t o toluene solutions of scintillator in the absence of water has relatively small effects o n the counting efficiency of tritium. Mixtures of Triton and scintillation solution are useful for solubilizing small amounts of biological material for counting ( 5 ) . F o r example, we have used these mixtures to count E. coli cells precipitated with trichloroacetic acid. Tables I1 and I11 show the counting efficiencies of tritiated water in suitable mixtures of Triton, scintillation solution, and water in the cold and at room temperatures, respectively. The differences in counting efficiencies between the two sets of conditions may be caused by a variety of factors such as differences in instrumentation and differences in light yield from the phosphors at the different temperatures. Within each group, however, the counting efficiencies of the solubilized samples are similar to those of the emulsions even though many of the emulsions are quite murky. The temperature range over which the mixtures form homogeneous solutions can be adjusted to any desired value by the use of appropriate mixtures of Triton surfactants as illustrated by the mixtures of Triton X-45 and X-114 in Table I1 and Triton X-100 and X-114 in (5) R. C. Meade and R. Stiglitz. I t i t . J . Appl. Radial. Isotopes, 13, 11 (1962).
Table 111. Counting Efficiency of Tritium in Triton Toluene Water Mixtures a t Ambient Temperature" Counting Physical state Counting efficiency, Triton of mixture rate, cpm % X-114 Clear mobile 7670 10 solution X-114 :X- 100 (2 :1) Clear mobile 7133 9.4 solution x-102 Viscous 7934 10.4 translucent emulsion a Each vial contains 5 ml of water with 7.6 X 104 dpm of tritiated water and 17 ml of a mixture of 2 parts ofscintillation solution to 1 part of Triton. Samples were counted at approximately 22 "C with a Packard Tri-Carb Model 574 liquid scintillation spectrometer; gain, 65 %; lower discriminator, 35; upper discriminator, 1000. The mixture containing Triton X-114 was clear between 17 and 21 "C and the mixture containing both Triton X-100 and Triton X-114 was clear between 22 and 28 "C. Table 111. For emulsion counting, Triton X-102 has the greatest utility for it can be used over a much wider temperature range than the others. RECEIVED for review June 24, 1968. Accepted August 14, 1968. Work supported in part by Grant No. G M 10317 from the National Institutes of Health.
High Precision Activation Analysis of Sodium with an Internal Standard Technique R . H. M a r s h a n d W. Allie, Jr. Ford Motor Co., Scientific Research Staff, Dearborn, Mich. 48121 MANYWORKERS have used activation analysis t o determine sodium in a variety of materials employing radiochemical separations. Reuland and Voight ( I ) analyzed tungsten bronzes for sodium nondestrictively, by coincident y counting. Lightowlers ( 2 ) determined sodium in a diamond with a pulse height analyzer. Schroeder and Winchester ( 3 ) determined sodium in silicate rocks nondestructively by discriminating against y rays having energies less than 2.6 meV, and using sodium carbonate as a standard. Kawashima ( 4 ) determined dysprosium in yttrium oxide using the yttrium as an internal standard. Standard deviation for a single determination was about 10%. Recently, Lutz and DeSoete (5)employed sodium as a n internal standard for the determination of carbon in sodium by photon activation analysis. The problem under consideration here was that of the determination of macro amounts of sodium in a material which could only be readily solubilized by a basic fusion. Because of the possibility of contamination, loss of sample, dilution error, etc., to which the usual methods used for the deter(1) R. J. Reuland and A. F.Voigt, ANAL.CHEM., 35, 1263 (1963). (2) E. C. Lightowlers, ibid., 35, 1285 (1963). (3) G. L. Schroeder and J. W. Winchester, ibid., 34, 96 (1962). (4) Toshi Kawashima, Deriki Tsushin Kendyusho Keiikya Jitsuyoka Hokokir, 13, 1835 (1964). (5) G. J. Lutz and D. A. DeSoete, Anal. Chem., 40, 818(1968).
mination of sodium can be subjected, a n attempt was made to develop a neutron activation technique which would be highly precise and accurate. Although neutron activation is usually thought of as a trace method, proper care should allow the technique to be used for macro quantities. One of the severest limitations t o the precision of activation analysis is nonuniform neutron fluxes which are found in reactors and which require flux mapping in order t o apply corrections. Also, the maintenance of strictly uniform irradiation and counting geometry is difficult. The use of a n internal standard obviates the necessity of mapping neutron fluxes in reactors and maintaining strictly uniform counting geometry. EXPERIMENTAL
Apparatus and Reagents. The following special equipment and reagents were used: a cooled, lithium-drifted germanium detector and 400-channel analyzer with a resolution of 5-6 keV F W H M for cobalt-60 radiation, and a lithium-drifted germanium detector and 2048-channel analyzer with a digital spectrum stabilizer having a resolution of 3.5 keV F W H M . Potassium carbonate containing less than 2 ppm of sodium, obtained from United Mineral and Chemical Corporation. All other materials used are common t o a n analytical laboratory. Procedures. The first procedure employed was as follows: the standards were prepared by weighing 10-mg portions of AS203 and 20-mg portions of N a 2 C 0 3 into 1-cm long polyVOL. 40, NO. 13, NOVEMBER 1968
* 2037
Table I. Analyses of Several Batches of Ceramic Average Std dev Av dev, CPG As/CPG Na 3.46 3.45 0.36 0.88 3.47 3.41 3.48 3.40 Per cent Na10 9.91 9.40 9.38 CPG As/CPG Na 3.28 3.20 3.27
9.56
3.26
0.30
2.4
0.041
0.96
Per cent Nap0 9.01 9.53
9.27
0.37
2.8
CPG As/CPG Na 2.71 2.65 2.65
2.61
0.035
1 .o
Per cent Na20 9.09 9.33 9.81
9.41
0.31
2.9
3.20
0.032
0.12
CPG As/CPG Na 3.18 3.18 3.20 3.25 Per cent Nap0 10.62 10.08
a
Naz 0
=
Y . Nz . W , GF
Ni
.
100
W2
where Y is the average ratio of corrected counts per gram of arsenic to corrected counts per gram of sodium in the standards; Nl is the corrected counts of arsenic; N , is the corrected counts of sodium; Wl is the weight of arsenic; W z is the weight of sample; and GF is the combined gravimetric factors. RESULTS AND DISCUSSION
10.35
0.38
2.6
ethylene vials. The vials were then sealed, the materials mixed o n a vibrator (Wig-L-Bug), and irradiated in a flux of approximately 1.5 X 1013N/cm2/sec for 3 minutes. After the samples were allowed to “cool” for 20 hours to permit the trapped argon-41 t o decay, they were counted for approximately 30 minutes at a n analyzer dead time of 20% in order to obtain good counting statistics for the 559-keV arsenic peak and 1.73-meV sodium double escape peak which were the most sensitive peaks in the spectrum. In addition, plain arsenic and sodium samples were irradiated for use as background blanks. Each spectrum was then plotted out, the background blanks subtracted out by visual comparison, and the arsenic and sodium peak areas determined by integrating over each peak. These values were corrected for decay, both back to the beginning of each count and back to the beginning of the first count. The ratios of arsenic counts per gram to sodium counts per gram gave results with a precision of *3% relative average deviation. The precision of these results was considered unacceptable, but indicative of that which could be obtained with a minimum of effort on the analyst’s part. The procedure which proved to be superior is as follows: Approximately 10 mg of As203and 20 mg of N a 2 C 0 3were accurately weighed into a series of 1-cm long polyethylene vials. The powders were mixed thoroughly, and the stirring rod and walls of the vial were washed down with anhydrous diethyl ether which had been saturated with polystyrene. The polystyrene pins the powders down and prevents losses 2038
in subsequent operations. The ether was evaporated and the vial filled with molten polyethylene. The sealing operation was best done by melting low-melt index polyethylene (available from Union Carbide Corp.) in a beaker which was heated carefully over a burner or on a hot plate. The polyethylene was poured into the vial while the vial was held in a dish of water and allowed to remain in the water until the polyethylene had set. The vials were then cut down to size for convenience and used as standards. The samples were prepared in the same manner, weighing out enough sample to obtain approximately 20 mg of sodium, Pure arsenic and sodium background blanks were prepared similarly. The standards and ceramic samples were then irradiated for four minutes in the pneumatic tube of the reactor and cooled overnight to allow any short lived activities and trapped argon-41 to decay. Each vial was then counted for a sufficient time t o provide good counting statistics--i.e., u = 0.7% or better. Clock times must be recorded accurately for decay corrections. Results were obtained by plotting out the spectra, subtracting out the background blanks, correcting for decay, and calculating the per cent sodium oxide. The calculation was as follows:
ANALYTICAL CHEMISTRY
The results obtained with the first procedure were inadequate. Two major sources of error were apparent. In the first place, the ability of the technique to be independent of flux variations depends on the materials being very thoroughly mixed. Merely shaking them in a vibrator for a while is not enough. With the powder spread over the inside walls of the vial, various portions would experience widely varying fluxes. Any inhomogeneities would readily be detected. If the samples were only fairly well mixed but confined to one small location, they would experience a more uniform flux and inhomogeneities would be less conspicuous. This was accomplished by mixing the material in the open vial with a quartz stirring rod. The rod and sides of the vial were then washed down with ether, the ether evaporated, and the vial partially filled with molten polyethylene. The vial was cut down, producing a fairly small, sealed standard sample. This provided two other advantageous results. Practically all the air (and argon) is excluded so the 20-hr “cooling” period is eliminated, and a fairly large number of specimens can be irradiated in the pneumatic tubes at the reactor at the same time, gaining in convenience (no sodium-24 from the pool water to wash off), as well as subjecting the specimens to a more thermalized flux. A well thermalized flux is important because of the epithermal neutron absorption resonance peaks of arsenic. When samples were irradiated in the core where the flux contains a noticeably larger portion of epithermal neutrons than in the pneumatic tubes, the results were considerably less precise. The second major source of error was in subtracting out the pure sodium and arsenic background blanks. Attempts to reproduce the data obtained from the visual com-
parison gave deviations of the same order as previously mentioned. A better method is to analyze the data by computer. The method is as follows: Ala
=
Aa
- Rs Ba
A's
=
As
- Ra Bs
a
Where: Aa = area of the arsenic sample peak As = area of the sodium sample peak Ra = ratio of the area of the arsenic sample peak to the area of the arsenic background blank peak Rs = ratio of the area of the sodium sample peak to the area of the sodium background blank peak Ba = area under the sodium background blank spectrum corresponding to the position of the arsenic peak Bs = area under the arsenic background blank spectrum corresponding to the position of the sodium peak Thus, two values, corrected for contributions from the other nuclide are obtained for the sample arsenic and sodium peak areas. The values are only a n improved approximation, however, because the original values used to obtain the ratio of sample peak area to background peak were not corrected. The accuracy of the correction is improved to the limit of the accuracy of the data by repeating the process with the new values obtained until there is no observable change in value. The final values are then corrected for decay during the individual count ;
X=
Ylit 1
- exp ( - a t )
(3)
and for decay back to the start of the first count; 2
=
X exp(Af)
(4)
The per cent sodium oxide is then calculated as in Equation 1, Values in Table I for the ratios of counts per gram arsenic over counts per gram sodium show the improvement obtained. Because the errors in analyzing ceramic samples were excessive, the method of sample preparation was investigated. Previous attempts had been made to determine sodium in reagent grade potassium carbonate with potassium as a n internal standard. The first set of results are presented in Table 11-a. I t was suspected that the scatter in the values was caused by nonhomogeneity. The bottle of reagent grade potassium carbonate was placed in the same horizontal shaker mixing device used in preparing the ceramic samples and shaken overnight. The analysis was repeated and the results showed little improvement (Table 11b). Shaking the bottle for an entire weekend showed no improvement (Table 11-c). Shaking a sample in the vibrator for three hours, however, gave results within the error expected from the data. See Table 11-d. Better data in this latter case could not be obtained because the counting times were approaching the half-life time of the sodium and potassium. The horizontal shaker mixing device used in preparing the ceramic samples was wholly inadequate and obviously incapable of reducing the error below the 2 to 3 found. The fact that the samples were as homogeneous as they were must have been caused by some other preparative step. To further test this conclusion, a large portion of the ceramic sample was thoroughly mixed in the vibrator and samples taken from this and run. The results are shown in Table 111. These results demonstrate that errors previously encountered were caused by sample nonhomogeneity.
Table 11. Repeated Analyses of Na2C03 in Reagent Grade KzC03 Average, Std dev, x 10~3. x 10-3 Av dev, CPG KiCPG Na, X 10-3 4.78 4.9 0.24 3.5 5.12
Per cent Na2C03 x 10-3 7.05 5.28 CPG KjCPG Na, X 10-3 4.92 5.09 4.98
6.17
1.2
5.00
0.084
1.2
3.61
0.77
15.4
5.01
0.12
1.8
7.66
1.5
14.0
5.01
0.12
1.8
14.3
Per cent Na2C03,
x
10-3
4.44 2.92 3.46 e
CPG KjCPG Na, X 5.14 4.91 4.97
Per cent Na2C03, x 10-3 9.26 7.39 6.22 CPG K/CPG Na, X 5.14 4.91 4.97
Per cent Na2CO3, x 10-2 1.22 I . 14 x 10-2 0.08 X 5.2e 1.06 1.12 a Samples from original bottle submitted without adequate blending. * Samples taken after having been mixed overnight in a horizontal shaker. Samples taken after having been mixed 3 days in a horizontal shaker. Samples taken after having been mixed 3 hours in a vibrator. e Value is within the range expected from the counting statistics available.
To test the accuracy of the method, samples which had been analyzed by atomic absorption spectrometry were run by this method. The results are shown in Table IV. I n order to see if the method could be used for trace analysis, an attempt was made to analyze similar material which had been deposited on quartz slides. The sample weights were of the order of 10 Mg and were supposed to be identical, as they were made a t the same time. Standards were prepared, and internal standards were added by the evaporation of solutions. Krylon clear plastic spray was used as a protective cover. The samples were irradiated for one hour a t a flux of 1.5 X l O I 3 N/cm2/sec, and counted. The data were handled as before and the results are given in Table V. The precision is VOL. 40, NO. 13, NOVEMBER 1968
2039
~~
Table 111. Analysis of Well Mixed Ceramic Average Std dev Av dev, %, CPG As/CPG Na 2.89 2.89 0.039 0.97 2.85 2.92
Table V. NanO in Alumina on Thin Films Average Std dev Av dev, CPG As/CPG Na 4.14 4.20 0.06 1.03 4.19 4.26
Per cent NaeO 9.31 9.34 9.64
Per cent NatO 8.49 7.37 7.72
CPGAs/CPG Na 2.78 2.12 2.79 2.76 Per cent NaaO 8.10 7.99 8.16 8.03
9.44
0.18
1.4
2.77
0.032
0.87
8.07
0.075
0.74
Table IV. Comparison of Neutron Activation and Atomic Absorption Results Results obtained by neutron activation analysis: Per cent NanO Average Std dev Av dev, %, 8.10 8.07 0.075 0.74 7.99 8.16 8.03 8.06 8.01 0.05 0.48 7.99 7.97 7.98 8.00 0.14 1.19 7.89 7.97 8.08 Results obtained by atomic absorption spectrometry: 8.12 8.01 0.089 0.87 7.93 8.03 7.94
thought to be as good as the ability to produce identical samples. All the results presented so far were from data taken with the 400-channel analyzer. Peaks consisted of three or four channels. The computer program employed required data from the three highest channels in each peak. Because of this limitation, the precision of the method is reduced by shifts in the spectrum caused by gain and zero drift during and between measurements. This difficulty can be overcome with a pulse height analyzer which has more channels and a digital spectrum stabilizer. By integrating the entire peak area, the counting statistics can be improved without resorting t o long counting times. The stabilizer ensures reproducible peak shapes and the integration can be carried out with a digital computer. An example of the precision which can be obtained is shown in Table VI. The
2040
0
ANALYTICAL CHEMISTRY
7.86
0.57
5.3
Table VI. Analysis of Ceramic Samples with a 2048-Channel Analyzer with a Digital Spectrum Stabilizer Average Std dev Av dev, CPG As/CPG Na 5.200 5.205 0.0069 0.100 5.202 5.213 Per cent Nan0 8.166 8.154 8.152
8.157
0.012
0.109
standards were prepared by the evaporation of solutions t o eliminate weighing errors. The primary sources of error in the method are: weighing errors which, when weighing into polyethylene vials, can be significant unless care is taken to avoid static charge build u p ; gain and zero drift in the electronics which can be eliminated or compensated for by a digital spectrum stabilizer and total peak integration; the presence of epithermal neutrons in the flux; improper mixing of sample and standard; and the ability to collect sufficient data to obtain good counting statistics. The major contribution to the error of the results presented in Table VI is probably random counting errors and small weighing errors. An additional source of error can be introduced by counting at different rates when the dead time of the analyzer is high. At dead times greater than 50%, errors as large as 6 % have been found; at dead times below 20 %, small changes in dead time become insignificant. Ceramic samples have been analyzed on a routine basis with this method but without a digital spectrum stabilizer or total peak integration. The standard deviations for a single determination have been consistently within 1.5 %. More recent work indicates that 10 mg of copper oxide can be substituted for arsenic oxide, greatly reducing errors caused by the epithermal neutron flux. ACKNOWLEDGMENT
The authors thank W. R. Pierson for his helpful discussions and suggestions, and the use of the germanium detector which he fabricated, L. C . Westwood and M. Teklinski for performing the sodium analyses by atomic absorption spectrometry, and J. Jones and T. Arber of the Phoenix Memorial Laboratory for testing the method on a routine basis. RECEIVED for review March 21, 1968. Accepted August 5, 1968.
