DISCUSSION
The results of the present work demonstrate that the proposed principle is valid. Even with the sample, inexpensive apparatus described in the present paper, surprisingly accurate and useful results can be obtained. The accuracy of this technique could be significantly improved by improving the accuracy of each gauge measurement. The transmission gauge accuracy could be improved by using a temperaturecontrolled sample and a slightly more elaborate calibration curve. The transmission calibration should take scattering and Bremsstrahlung losses into account as well as simple ionization losses. This could be done by fitting the calibration to the form
a=1
kl
p
'2to,Zt2/Ai)
(11)
2 = 1
where kl is a constant. Smith and Otvos (6)report a standard deviation of 0.170 for hydrogen in their work on the transmission gauge principle. They used a calibration form similar to
Equation 11 and their samples were controlled to a constant temperature. The backscatter gauge accuracy could be improved by using a sample cell with a thin covering on top so that the source-to-sample-to-detector distance could be reproduced more exactly. Gray, Clarey, and Beamer (2) used such a sample cell with a top covering of a thin sheet of mica. They obtained a standard deviation of 0.03% for hydrogen in their work. With discussed improvements to the two individual gauges, it should be possible to obtain measurement standard deviations of less than 0.1% for hydrogen, and less than o.5yO for carbon and oxygen. This precision would be possible for a total analysis time of less than 10 minutes. Although the present work concentrated on the three-component system hydrogen, carbon, and oxygen, other three-component systems containing hydrogen as one component could also be analyzed by this technique. Any other three-component system of this type, with the exception of the system hydrogen, carbon, and nitrogen, should be easier to analyze than the present one because of the greater
difference in the backscatter response to combinations of elements other than carbon and oxygen. One simply has to calibrate for the desired threecomponent system. ACKNOWLEDGMENT
The authors are grateful for the assistance of K. L. Fletcher in taking the experimental data for this paper and to B. R. Roberts for constructing the transmission and backscatter gauges and sample holders. LITERATURE CITED
(1) Berthold, R., 'fcontinuous Measuring
of Hydrogen in Hxdrocarbons by Beta-ray Absorption, Proc., Second U. N . International Conf. o n Peaceful
Uses of Atomic Energy 19, 288 (1958). (2) Gray, P. R., Clarey, D. H., Beamer, W. H., ANAL.CHEM.31, 2065 (1959). (3) Jacobs, R. B., Lewis, L. G., Piehl, F. J., Ibid., 28, 324 (1956). (4) Muller, D. C., Ibid., 29, 975 (1957). (5) Muller, R. H., Ibid., p. 969. (6) Smith, V. N., Otvos, J. W., Ibid., 26, 359 (1954). RECEIVED for review December 7, 1964. Accepted January 21, 1965. This work
was sponsored by the Division of Nuclear Education and Training of the Cnited States Atomic Energy Commission.
Trace Oxygen Determination in Cesium Metal and the Problem of Recoils from the Atmosphere during Fast-Neutron Activations OSWALD
U.
ANDERS and DAVID W. BRIDEN
Radiochemistry Research laboratory, The Dow Chemical Co., Midland, Mich.
b A method to determine oxygen traces of 5 to 500 p.p.m. in highly reactive materials (cesium metal) by fast neutron activation analysis has been developed. The preparation of the low-cost, low-oxygen-content polyethylene vials used is described. The effect of recoiling nuclei from the octivation of the atmosphere surrounding the sample vial has been studied and found to be proportional to the surface area in contact with air during the irradiation.
S
recent articles discuss the determination of oxygen by fastneutron activation analysis using the 0l6 ( n , p ) W reaction to induce radioactive W6for counting (1, 3-5). It appears possible to carry out relatively interference-free determinations for chemically bound and free oxygen if present in levels above 700 pg. (1). EVERAL
530
ANALYTICAL CHEMISTRY
Authors using more sensitive counting techniques, such as well-type scintillation detectors and low discriminator level settings on their gamma spectrometers-e. g., 1.2 m.e.v. (3)-are able to report detection limits of 200 pg. and less, but such elements as All B, C1, F, Mg, N , Na, P, S,Si, V (3, 5 ) , Th, and U interfere. Even if these interfering elements pve absent from the samples to be analyzed, oxygen present in the sample container itself is a problem ( 3 ) and actually emerges as the limiting factor to the sensitivity of the methods. Several laboratories circumvent the container problem by transferring the activated samples to a nonactive container before counting and are thus able to detect even smaller amounts of oxygen ( 2 ) . For the case of cesium metal analysis, this latter approach is prohibited by the extreme chemical reactivity of the
sample and the short half life of the radioactivity to be measured. I t is also necessary for cesium metal samples with oxygen contents in the range of 50 to 100 pg. to employ a well-type NaI(T1) counter for counting the induced 6" and suffer possible interferences, in order to obtain significant counting statistics. INSTRUMENTAL
For routine use in this laboratory samples are hermetically sealed into inexpensive commercially available polyethylene vials. They are then activated for 30 seconds in the 14m.e.v. neutron flux of a Texas Nuclear Corp. neutron generator, employing a 1-curie per sq. inch tritium target to produce neutrons by the reaction T(d,n)He. After activation the sample is transferred by a pneumatic system to a funnel above the steel castle of a 5-inch-diameterJ 5-inch-high NaI (Tl) scintillation counter. The pulses from
this counter are amplified and those equivalent to gamma energies between 4.5 and 7 m.e.v. are selected by a single channel analyzer and counted with a scaler. Four seconds are allowed for the transfer to the counter before the scaler starts counting the sample for an 18-second interval. The timing is done automatically. The activation flux of -2 X IOs n’sq. cm./sec. is measured by a fast neutron counter positioned behind the sample during activation. Its output is recorded on a strip chart recorder and used for the normalization of the data ( 1 ) . The sample is removed from the counter by the operator. If deemed necessary, he can also intercept the incoming activated samples, check them for damages, and insert them into the counter by hand. EXPERIMENTAL
I t was decided to use polyethylene vials, if possible, also for the cesiummetal samples, in order to keep the container cost lorn and avoid interferences from other easily activated container materials. Even if free from oxygen, containers of copper, aluminum, or stainless steel provide a sufficient number of high-energy pulses due to pile-up of loiier eneigy pulses to simulate the 6.1-m.e.v. gamma-rays of N16 and cause severe interferences. The commercial polyethylene vials used for packaging our samples seemed, however, to contain 3 mg. or more of oxygen. This amounted in the blanks to more than 1500 counts and a standard deviation of 0.1 mg. due to statistics alone. .liter several unsuccessful attempts to procure commercial vials sufficiently loa in osygen content, it appeared necessary to manufacture our own vials in the laboratory. Various commeroial grade polyethylene pellets a e r e surveyed to find a suitable low-oxygen resin. The assay wab carried out by packaging the pellets in our regular vials and determining the increase in apparent oxygen content over and above empty vials activated at the same time as blanks. For several polymer grades no increase, but rather reductions in count rates per flux n ere encountered. One of these, a resin made by the Phillips process, was chosen for molding the new vials. Molding of Low-Oxygen Polyethyle n e Vials. T h e only way the oxygen content of the polyethylene in a molded product can increase over t h a t of the molding material is by ovidation of the hot molten polymer during the molding process and the use of a mold-release agent containing oxygen. .Lroiding the need for a release agent, keeping the molding temperature low, and excluding air from contacting the melt should reduce this pickup of oxj gen. L o n molding temperatures, however, also require higher molding
pressures and tend to induce brittleness in the products. Optimum conditions thus had to be selected experimentally. A . suitable mold was made in the shop to be operated b y a small injection molding machine, equipped to permit blanketing the molten material with nitrogen and thus exclude air from contacting the polymer. Vials were then molded from the selected resin a t different temperatures and their oxygen content was checked. S o difference in oxygen content could be detected for extrusion temperatures between 320’ and 450’F. Analysis of Molded Vials. T h e empty vials were first analyzed using air as driving gas in the pneumatic transfer system. T h e air present in the vials during activation was purged out with a jet of air immediately before insertion into the counter. Comparators were made from dilute solutions of isoamyl alcohol and n-butyl alcohol in pure paraffin oil. Vials with only paraffin oil were simultaneously run as blanks for the comparators. When the apparent oxygen content per gram polymer of the empty vials was compared with the corresponding values for the same vials after they were filled with pure paraffin oil as in the blanks, it was observed that the oxygen content decreased. This was puzzling since the activated air was flushed out of the empty vials and paraffin oil must have contained some trace of oxygen impurity. It was also observed that vials filled with paraffin oil showed the same count rates per flux as vials filled with polymer pellets in mineral oil. When a further series of experiments showed the same oxygen content per gram of polymer for the above-mentioned empty vials and polymer pellets irradiated inside air-containing polyethylene vials and placed in the counter after activation without container, further study of the apparent paradox was indicated. Six sets of vials were weighed and analyzed for oxygen in the following way:
A. A set of five vials, average weight 3.85 grams, was filled with nitrogen, kept in nitrogen for one week, and run
Table I.
Av. w t . of polymer,
Expt. A
B C D E F
grama 3.85 3.85 3.85 3.85 3.85 7.86
in the pneumatic system using nitrogen as driving gas. The atmosphere inside the counter was air. €3. A set of 4 vials containing air, average weight 3.85 grams, was kept in air. The vials were introduced into the nitrogen atmosphere of the pneumatic system less than minute before activation. The air contained in the vials during the activation was purged out with a jet of air prior to insertion into the counter. C. -1set of two vials from set B was run about 10 minutes after set B, but this time the air was purged out with nitrogen inside a nitrogen glove box about minute before the activation. D. A set of five vials, average weight 3.85 grams, was filled with nitrogen and kept in nitrogen for several days until ‘/2 minute before activation in air. Air was used for the driving gas of the pneumatic system. E. A set of five vials containing air and kept in air also was activated in air, but the air content was purged out with a jet of air immediately before counting. F. One vial containing twelve 4.5cm.-lone: and 0.3-cm.-diameter ~ o l v ethylene rods of the same materi’al ‘ts the vials contained air and was kept and activated in air several times. The air in the voids between the rods was purged out with a jet of air before insertion into the counter. The results of the six experiments are given in Table I. It is seen from Experiments B and C that the presence of air inside the vial during the activation greatly influences the apparent oxygen content. If it is assumed that vials A were essentially free of dissolved oxygen gas, the concentration of dissolved oxygen in the vials is given by C - A = 0.008 mg. per gram of polymer. Since conceivably some dissolved oxygen might have diffused out of the vial in the 1 minute during which samples C were inside the nitrogen atmosphere, the oxygen content value of 0.008 mg. per gram represents a lower limit. We define as “oxygen excess” that amount of apparent oxygen content which is not accounted for by the bound oxygen content, of the polymer, as determined by experiment 4,or the dissolved oxygen, given by the difference of the results of experiments C and A.
Results of Analyses of Polyethylene Vials
Apparent oxygen content per gram polymer, mg. /g. 0.051 0.133 0.059 0.133 0,203 0.145
Oxygen excess, per gram polymer, mg./g.
Surface area Oxygen in contact excess/area, with air, sq. cm. mg./sq. cm.
0.074
30
0.0096
0.074 0.144 0.086
33 63 125
0.0086 0,0088 0 0054
VOL. 37, NO. 4, APRIL 1965
531
Subtracting the value of C from B, D, E, and F, in column 3 of Table I, we obtain the oxygen excess values per gram of polymer which are given in column 4 of Table I. There is no correlation between these results unless we consider the surface area in contact with air during the activations. The 2-dram vials were 5.6 em. high and had an outside diameter of 1.65 em. The inside surface area was 30 sq. cm.; the outside surface area 33 sq. em. The polyethylene rods in vial F had an area of 5.2 sq. em. each. If we calculate the oxygen excess per surface area in contact a i t h air during the activation-i.e., column 2 times column 3 divided by column 5-we get column 6 of Table I. The error of the data in Table I is about 7%, because of the statistics of the counts. If n e compare B, D, and E we see that the oxygen excess per square centimeter surface exposed to air during activation is a constant. The areas for our experiments varied over a factor of 2. The data also indicate that the oxygen
Table II.
Sample 1
2 3 4 6 6
7 8
9
Individual Determinations of Oxygen Content of Cesium Metal Oxygen content SIg. 0-vial P.p.m. P.p.m. (0.150 Weight, added found Median hIg 0 grams mg. es.) 0 0 250 8 7 0.100 12.359 4 0.050 0 200 7 0,090 0 240 0 0.127 11 11 0 277 11 864 11 0.127 0 277 11 0.136 0 286 45 31 0 740 0,590 45 13.066 43 31 43 0 700 0.550 12.858 42 33 48 0.595 0 745 12.260 38 0 612 0.462 42 0 670 0.520 33 0.581 46 43 0 731 12.503 0.492 39 0 642 0.540 43 0 690 44 45 33 0 699 0.549 12,183 38 0,459 0 609 44 0.540 0 690 63 0,770 62 79 0 920 12.403 1,125 90 1 275 71 1 028 0.878 79 1 137 0,987 124 124 119 1.508 1 658 12,096
10
11,738
11 12 13
1 2 , 705
14
12,497 12 505
15
532
excess of the samples activated in contact with air is almost an order of magnitude larger than the oxygen found due to dissolved gas in the polymer of the vial. If this oxygen excess is proportional to the exposed surface, this surface effect must be due to the activation of the surrounding air. I n particular, the phenomenon can be explained by assuming activations taking place in the oxygen of the air close to the polymer surface, and the surface of the vial acting as a catcher for the recoiling N16 nuclei produced by these reactions. The rate of this collection of recoils should be a constant for a unit area of polymer surface, if in contact with a layer of air thicker than the path of the recoiling X16nuclei. I n cases like F, where the air layer over part of the surface is less thick than the range of recoiling