s
variance of any quantity x S, = standard deviation of any quantity x S,/X = fractional error in any quantity 1: SI, Sz,S,,S4 = gross counts from first, second, third, and fourth countings of a standard ts , h$2, ts,, ts, = counting times for standards SI,8 2 , S3, and S4 Itlo = counting time for backgrosnd Bo tnl = counting time for background B1 =
ZZ
tu, too
T
uo U1
UQ, Uli
counting time after internal standard addition = counting time for sample = numerator of Equation 1 = initial gross count of sampie = gross count of sample after internal standard addition = gross initial count of sample i = gross count of sample i after internal standard addition
=
ACKNOWLEDGMENT
C. S.Rice and R. E. Shultz of this laboratory were the originators of an IBM program by which many of the calculations presented here were made. LITERATURE CITED
(1) Herberg, R. J., ANAL. CHEM. 33, 1308 (1961). ( 2 ) Whisman, M. L., Eccleston, B. H., Armstrong, F. E., Ibid., 32, 484 (1960).
RECEIVEDfor review July 13, 1962. Accepted March 21, 1063.
Determination of Oxygen in Gallium Arsenide by Neutron Activation Analysis R. F. BAILEY and D. A . ROSS RCA Laboratories, Pririceton, N. J.
b Neutron activation of oxygen in gallium arsenide has been carried out using the 0 l 6 ( T , nIF'8 reaction with the tritons being produced by the Li6(n, T)He4 reaction. The maximum sensitivity of the method is about 7 X gram of oxygen in gallium arsenide, representins1 5 atomic parts per million in the samples used. The actual bulk concentraiion in the GaAs 18 atomic samples used was 72 parts per million.
S
METHODS for the determination of 0xyge.n are satisfactory when there is a large amount of material with a high oxygen Concentration. With small samples and an oxygen concentration in the submicrogram range, the standard methods arc no longer adequate. For these low concentrations where most instrumental methods fail, analysts have been turiiing to activation analysis using both accelerators and nuclear reactors (1-3, 6, 18). Direct activation of the oxygen by neutrons gives rise tc short-lived isotopes which can be used to measure the total concentratior in the bulk and physically or chemically sorbed on the surface and in the surrounding atmosphere (12, 17, 19). To study penetmtion profiles or bulk cor.centrations only, it is necessary for the rdioactive isotope to have :t long enough life for radioc1icmir:d procedures ;such as etching arid chernical separations. Activation hy charged particles froni nn rlccelerator 1t:i.s I ~ c t ~Icported n (IC). One rnetliod is to use the 0l6(T,n)FiSrcaction (4, 7 , IS, 16). The F18 appears to be the TANDARD
most suitable isotope resulting from oxygen bombardment. It has a half life of 1.87 hours, which is long enough to carry out radiochemical separations and to etch the sample to eliminate the surface oxygen which might mask that in the bulk. It decays by positron emission and the resulting annihilation radiation is easy to count with a gammaray spectrometer (11). To carry out this type of activation with a nuclear reactor, it is necessary to produce the tritons through a secondary reaction. This can be done by bombarding lithium, resulting in the reaction Li6(n, T)He4. The reaction is exoergic with a Q value of 4.78 m.e.v. (8). The tritons are emitted with an energy of slightly over 2 m.e.v., which is sufficient to penetrate about 0.002 inch of density-5 material. By placing a layer of lithium on each side of a wafer of the material under investigation, a relatively-uniform flux of tritons is produced throughout the sample, and the resulting F18 can be used to obtain the bulk oxygen concentration. EXPERIMENTAL
The technique to be described was developed to determine the oxygen concentration in materials. GaAs n-as used in this experiment. Electrical measurements showed the presence of c,arrier traps at concentrations of the order of 1017 cm.-3 (9, 14). Analysis of rhc.iuica1impurities other than oxygen with conventional techniques showed tha>t, none n w c ~ ~ r c s c ni nt these coricciitratioiis. Of tlic methods for thc: tletcrmination of oxygen, activation appeared to be the only one that would
eliminate surface oxides and permit measurement of the bulk oxygen concentration. The technique used was to irradiate thin wafers of GaAs wrapped in lithium metal. The tritons produced by the neutrons in the lithium had enough energy to penetrate the GaAs wafer, producing FIE. If the neutron flux and the effective cross section for the reaction is known, it is possible to calculate the oxygen concentration. While the cross section for both the LiG(n, T)He4 and the 016(T, n)F@ reactions are known, the triton flux in the sample is difficult to calculate because of neutron-flux depression in the lithium and the variation in lithium thickness. The effective activation cross section was therefore determined experimentally. To do this, intimately mixed powder samples of lithium-oxygen compounds were irradiated. The effective activation cross section was determined from the amount of Fl8 present. In addition, lithium-wrapped powdered oxides were irradiated. The values obtained were internally consistent and agreed with a value obtained by Osmond and Smales (15) using powdered B e 0 mixed with LiF. The measured activation cross section was 5 X 10-28 The only possible competing reaction which might also produce F18 was F19(n, 2n)F18. The yields of the GaGg((n, 2n)Ga68 and the AS^^(^, 2 n ) A ~ 7 were ~ too low to determine in the present experiment. This n-as checked by GaAs blanks put t>liroughthe separation procedure. To tleterniiiie whether the fluorine conceiitlratioii could be a serious sourcc of error, it was necessary to measure the cross section for the reaction in a modified fission sliectrum which wits not list,cd iri thc literature. This was doiie by irradiating various fluoride samples, such as Teflon and ammonium fluoride, VOL. 35, NO. 7, JUNE 1963
791
lo(
I
I
IO'
I
IC?
102
NEUTRON ENERGY
Figure 1.
I
IO
100
(m.0.v.)
Neutron spectrum
in the pneumatic facility in the Industrial Reactor Laboratories, Inc. 5-Mw. swimming-pool reactor, jointly owned by RCA and nine other corporations. The 0.51-m.e.v. annihilation peak was then counted and followed for several half lives to establish identity. The neutron flux and energy spectrum is shown in Figure 1. The cross section for the F19(n, 2n)F18 m s determined to be 1.45 f 0.23 X 10-6 barns. All cross-section values given for this reaction have been made with monoenergetic neutrons rather than in a modified fission spectrum (6). If the fluorine concentration were the same as the suspected oxygen concentration a t about 100 atomic p.p.m., the error mould be less than 0.3%. As fluorine had been a possible chemical impurity and could be measured below this value with conventional techniques, it was known that its concentration was well below that of the suspect oxygen. Its interference was therefore considered negligible. F18 produced by reaction of the lithium metal, and oxygen or water vapor in the air vias also considered negligible because of the cleaning and etching of the sample after irradiation. Sensitivity. The ultimate sensitivity of the method is limited by the ability to handle large, thin wafers of GaAs wrapped in lithium metal, and by the competing gamma rays resulting from activation of the matrix material which interfered with the counting of the F18 activity. The radioactivity per gram of oxygen resulting from triton bombardment of 0 1 6 a t a thermal neutron flux of 2 x l O I 3 n/cm.2/second is given by hi A = -
J
I
I
I
'01
Substitution gives A = 1.1 X 10 d./second/gram of oxygen. The counting efficiency using the 3inch NaI(T1) scintillation crystal is 37% at 0.511 m.e.v., the energy of the annihilation gamma ray resulting from the decay of the positron emitted from the FL8. The efficiency was determined by calibration using a standard of Ka22 and comparing the number of photons under the 0.511-m.e.v. peak vs the total number expected. The chemical separation required to reduce the concentration of the radioactive Ga and As reduced the F18 concentration by 20Yob. The resulting over-all efficiency i5 about 30%. This gives a counting rate of N = 3.3 x lo7 c./second/gram oxygen. Assuming that a sufficient number of counts can be accumulated to permit an accurate subtraction of the background counts, the ultimate sensitivity of detection is of the order of 7 X 10-9 gram of oxygen. This would be equivalent to 10 c.p.m. for the sample over a background of 6 c.p.m. a t 0.511 m.e.v. The maximum drift of the counting system was =t0.5Y0per 24 hours. The short term drift was the same. The chemical separation procedure required to reduce the Ga and As
(1 - e - M )
1Tf
disintegrations/second/gram O2 (1)
where N M
Avogadro's number Atomic weight of oxygen = 16 Activation cross section = 5 X 10-28cm.~ 0 = Isotopic abundance of O L E 9 = Neutron flux = 2 x 1013 n/cm. a/second X = Decay constant of F1* = 1.03 X 10-4second t = Irradiation time = 300 seconds = = c =
792
ANALYTICAL CHEMISTRY
400
GAMMA RAY ENERG'l-(kJe6;)
Figure 2. Spectrum of GaAs sample taken a t 60-minute intervals
activity gave decontamination factors of between 10,000 and 40,000, depending on method used. Enough activity still remained, however, to produce interference with the FIB gamma-ray peaks. Arsenic was a particular problem since it had a strong peak a t 0.560 m.e.v. which almost completely masked the 0.511 F'8 peak. A differential method of analysis was used to separate the activities in these two peaks. In this method, spectral curves are taken a t frequent intervals after the sample preparation on a multi-channel, gammaray analyzer. Since the half lives of the two isotopes varied widely, the As76 being 26.6 hours compared to the 1.87hour FIB,a marked variation could be seen in the front end of the main arsenic peak. By calculating this difference in activity, it was possible to establish the counting rate for the FIE when it left the reactor. This interference reduced the sensitivity of the method to about 10-8 gram of oxygen. Sample spectral curves are shown in Figure 2. Sample Preparation. The wafers of GaAs were taken from crystal ES53. They were sliced perpendicular to the direction of crystal growth and lapped to a final thickness of 0.002 inch. The measured surface area was between 20 and 50 mm.2 To check the effect of the cold-worked surface due to the lapping, some of the rrafers were rough lapped to about 0.004 inch thickness and chemically etched to their final thickness of 0.002 inch. The samples were nrapped in lithium metal, fresh cut and rolled to l/32 inch thick under nitrogen, and placed in polyethylene containers for insertion into the pneumatic irradiation facility. The irradiation time was 5 minutes in a flux of about 2 x 1013 n/cm.2/second, limited by the heat developed in the sample n-hich softened the plastic container. Other types of containers became too radioactive and would have required the use of a hot cell. While the sensitivity n-as limited by the short irradiation time, it was more convenient to handle the samples a' soon as possible after irradiation. Radiochemical Separation. ilfter irradiation, the sample was etched for varying lengths of time t o remove the surface FIS, and then dissolved in a small amount of HX03 -I- HCl (4:1), and 30 mg. of fluorine carrier was added. The solution was extracted tnice with ethyl ether to remove the bulk of the Ga. Excess AgS03 was added to precipitate the C1- present. After filtration, the pH was adjusted to 7 and the Ag&OI formed was filtered off. The resulting solution was made t o volume and an aliquot (5 ml.) was removed and counted. The recovery of F18 was 80% with this procedure, and the decontamination factor was about 10,000. -4second separation method was used for most of the later samples. It was the same as the first one through the cther extraction. At this point, La(K'03)3was added in excess and the T,aF3 collected, n ashed, and counted. The recovery of F18 was again about SO%, but the decontamination factor
Table 1.
Etch depth, inches
OI6 vs.
Depth
Oxygen,
p.p.m. (atomic)
Std. dev.
75,000 7,000 775 295
i71
TI,^^ half life (minutes)
Ground samples: 0 2 2.7 4 2
I
\
\
-
".\I
100
200
309 400 TIME MIN I
Figure 3. Half life o f trum stripping
F18
500
10-4 10-4
10-4 10-3
108 f 3 110 f 10 106 1 3 109 1 4 109 f 5
1 7 l t 3 1 1
so
Etched samples:
4
0
x x x x
0 1 2
38,500 155 10-4 65 10-3 All samples run in duplicate or triplirate.
x x
109 f 2 108 f 3 111 f 3
1 2
f l
600
from spec-
increased to about 40,000. Undoubtedly, a second precipitation with additional carrier at this point would have eased data interpretation considerably; however, because of laboratory problems this \vas not done at this time. All chemicals used with the exception of the lithium metal were Baker Analyzed Reagents. The lithium was purified grade obtained from Fisher Chemical. A number of runs u5ing radioactive tracers wore made to determine the yield and the decontamination factors. To ensure that thi: etching did not introduce artifacts, a curve of weight us. etch time was plotted and was found to be linear with time. I n addition] one sample was heavily etched before irradiation and compared with another which had not had an initial etch. These two were merely washed after irradiation. The resultant F18 concentration agreed within a factor of two. One heavily-etched sample was examined microscopic ally and an estimate made of the increased surface area due to etch pits. It was about 200 mm.2 21s. the 20 t 3 50 mm.* for the unetched samples. This did not appear t o affect the results, so it is concluded that the surface rerioval is approximately linear with etch time. RESULl'S
The separated aliquots containing the F'8 were counted with a 3-inch diameter by 3 inches long NaI(T1) well-type scintillatio,i crystal. The output from the scintillator was fed through a nonoverloading amplifier and to a 200-channel gamma-ray analyzer. Counting was started approximately 1 hour after removal of the sample from the reactor, and spectra were taken at appro:uimately one-half hour intervals. Soms typical spectra are shown in Figure 2. By measuring the decrease in counting rate at the low-energy end of the As76peak, which corresponded to the 0.511-m.e.v. an-
1 1
70
60
I-
39 60 61 6 2 63 6 4 6 5 66 67 68 69 70 71 CHANNEL NUMBER
Figure 4.
F'*
72
73
spectrum after stripping
nihilation radiation, it was possible to measure the half life of the F18. This 5 minutes per sample, averaged 109 and an example is shown in Figure 3 for one of the samples. By measuring the number of counts due to the F18activity and extrapolating t h b back to zero time (time of removal from the reactor), the peak due to the fluorine was reconstructed. A typical peak is shown in Figure 4. The sum of this activity, corrected for counter efficiency, chemical-separation yield, and effective cross section, represents the concentration of oxygen in the unetched part of the GaAs wafer. Table I shows some typical results of a number of runs. Figure 5 shows these points plotted us. depth in the 0.002inch thick wafers. The decrease in oxygen concentration with surface etch shown in Figure 5 is probably due to increased diffusion from the surface, which had a high concentration of defects resulting from cold working the surface during the
lapping. The cold-worked region appeared to reach a depth of the order of 5 microns (IO). The leveling-off of the oxygen concentration beyond this depth presumably represents the bulk concentration (72 I 18 atomic p.p.m.). The error is, in this case, an estimation. To investigate this surface diffusion, wafers uere prepared using a chemical etch for the final finish which should eliminate the cold work surface. These results show lovc er surface concentrations but the same final, bulk concentration. Shortly after the completion of the above work, a hIetro-Vickers AIS7 solids mass spectrometer became available and a sample from the end of the same crystal was analyzed for oxygen. After repeated sparking and extreme care in removing surface and ambient oxygen, the concentration was found to be 30 i 15 atomic p.p.m. Considering the low levels of concentration, it was felt that this was satisfactory agreement.
d.1
d.0
d.2
2.3
d.4
d.5
&--
SAMPLE THICKNESS (INCHES X 1 0 9
Figure
5.
0lG Concentration
vs. depth
VOL. 35, NO. 7, JUNE 1963
793
(4) Bernstein, R. B., Iiatz, J. L., Ibid., 11,46 (1953). 15) Cohen, B. L.,. Phvs. Rev. 81, 184 (1951). ‘ (6) Coleman, R. F., Analyst 86, 39 (1961). (7) Coleman, It. F., Perkin, J. L., Ibid., 84, 233 (1959). (8) Everling, F., Koenig, L. ;?., Mattauch, J. H. E., Wapztra, A4. H., 1960 Kuclear Data Tables, Part 1, Xational Acad-
ACKNOWLEDGMENT
The authors thank J. R. Woolston for the mass spectrometric analysis and L. R. Weisberg for supplying the ingot. LITERATURE CITED
emy of Sciences, National Research Council, February 1961. ( 9 ) Fuller, E. S., Kaiser, W., Thurmond, C. D., Phys. Chem. Solids 17, 301
(1) Anders, 0. V., ANAL.CHEM.33, 1706 (1961). ( 2 ) Bate, 1,. C., Leddicotte, G. W.j Pittsburgh Conference on Analytical
(1961). (10) Gatos, H. C., Am. Inst. Chem. Engrs. Conf., New York, N. Y., December 1961. (11) Heath, R. L., AEC Research and Development Rept. IDO-16408, July 1, 1957.
Chemistry and Applied Spectroscopy, March 1958. (3) Beard, D. B., Johnson, R. G., Bradshaw, W. G., Aucleonics 17, 90 (1959).
(12) Koch, R. C., “Activation AnalJsis Handbook,” ASTIA DOC f.214941, December 15, 1958. (13) bhrkowitz, 5. S., Mahony, J. D., ANAL.CHEM.34,329’(1962). (14) Morrison, G. H., Cosgrove, J. F., Ibid., 27,810 (1955). (15) Osmond, R. G., Smales, A. Anal. Chim. Acta 10, 117 (1964). (16) Smales, A. A Pate, B. P., ANAL CHEW24,717 (1952). (17) Steele, E. L., Rfeinke, W. W., Ibid., 34, 185 (1962). 118) Thommon. B. A., Ibid., 33, 583 . (1961). ’ (19) Veal, D. J., Cook, C. F., Ibid., 34, 178 (1962).
RECEIVEDfor review July 24, 1962. Accepted March 22, 1963.
Quantitative Interpretation of Color Quenching in Liquid Scintillator Systems HARLEY H. ROSS’ and ROGER E. YERICK2 Oak Ridge Institute of Nuclear Studies, Oak Ridge, Tenn.
b A general approach to the quantitative interpretation of color quenching in liquid scintillator systems i s developed and applied to the specific case of carbon-1 4 in dioxane scintillator. The observed linear relationship between per cent quenching and concentration of color-quenching species permits the evaluation of quenching coefficients. In the perfectly general system, a complete series solution requires the use of a computer program. For systems in which the absorption spectrum of the colored material i s simple, good agreement between predicted and observed quenching can b e obtained b y determining only a few coefficients. In the system investigated, agreement i s achieved with the evaluation of only three coefficients. The linearity of color quenching makes it possible to estimate easily the amount o f quenching in systems in which only one quenching agent of known identity i s present. In any system, a comparison o f the predicted color quenching with the observed total quenching permits calculation of the amount o f chemical quenching within the system.
T
methods of sample preparation for liquid scintillation counting often rcsult in systems that are considerahly less than ideal from the standpoint of counting efficiency. Several techniques have been investigated for correcting for total quenching by R ~ internal standard method (6, 11, 12) o r by an approximation tcchiiique (‘7). \1/ hile work has been done on cheiiiical
794
quenching (2, 8),little has been published on the general phenomenon of color quenching. Some specific instances have been noted. Leffingwell, Riess, and Melville (IO) observed the effect of color on iron-59 systems containing the colored tris-l,lO-phenanthroline iron(I1) chelate but noted an enhancement of photon yield, relative to their colorless standards, that was chelate-concentration dependent. Baille (1) found a difference in over-all magnitude between chemical quenching and color quenching at higher concentrations, and Herberg (9) investigated the effect of color in obtaining suitable background solutions for certain types of biological materials. Helf and White (6) studied the quenching effects of organic nitrocompounds as a function of their ultraviolet absorption spectra but did not suggest a way to correct for this other than to change the chemistry of the counting system. Ilalvorsen (4) used an extrapolation technique to correct for color quenching in tissue samples. However, this technique can be used only for the same or similar samples. In the present study, a n attempt was made to correlate color quenching with concentration and wavelength absorption maximum of the quenching material for a particular liquid scintillator sy st em .
HE USUAL
ANALYTICAL CHEMISTRY
counts recorded were the sum of the red and green registers. The photomultiplier high voltage mas selected t o give balance point operation in the red channel for (2-14in a n unquenched sample. Screwcap 20-ml. glass vials were used as sample bottles. Spectral measurements were made with a Rausch & Lomb Spectronic-20 colorimeter with matched I/*-inch test tubes. The scintillating medium consisted of a dioxane solution containing 200 grams per liter of naphthalene, 7 grams per liter of 2,Bdiphenyloxazole (PPO), and 0.3 gram per liter of 1,4bis - ( 5 - phenyl - 2 - oxazolyl) - benzene (POPOP). Twenty milliliters of scintillating solution were used in each sample. Coloring agents used mere watersoluble FD&C coal-tar dyes. Aqueous solutions of these mere prepared of such dilution that from 10 p1. to 700 pl. were required to give the desired absorbances in the liquid scintillator medium. The weight of dye in these solutions varied from ahout 1 pg. to 100 pg. Activity was introduced as toluene-l-CI4 dissolved in toluene (0.1 /Jc. to 0.01 pc.). Absorbances of the colored solutions were measured against a dioxane blank, because the liquid scintillator blank proved to have too high an absorbance fcr proper spectrophctometer operation. The indirect method of measuring absorbances of liquid scintillator alone us. dioxane blank, and liquid scintillator plus dye against dioxane blank, and
EXPERIMENTAL I
Apparatus and Reagents. C‘ounting w a s done wit,h :? F:irkaril Tri-CIarh liqiiicl sc:iiitill:tticrii spcctriomcltr,
Modo1 314. 1~isciiriiiii:ttor scttiiiys were 10, 50, and 100 yolts, and the
1 Present address, Analytlral C‘hmnstrvThvision, Oak Ridge National I,:thoratiirj,
P.O. Box X, Oak Ridgc, Tenn.
2 Prcsent address, Lamar State Collq!c~ of Technology, Beaumont, Texas
