Isotopic enrichment and pulse shape discrimination for measurement

---

Isotopic Enrichment and Pulse Shape Discrimination for Measurement of Atmospheric Argon-37 W. M. Rutherford* Monsanto Research Corporation, Mound Laboratory, Miamisburg, Ohio 45342

John Evans Brookhaven National Laboratory, Upton, N. Y. 1 175 1

L. A. Currie Analytical Chemistry Division, National Bureau of Standards, Washington. D.C. 20234

High sensitivlty measurements of cosmic-ray-produced 37Ar from the Southern Hemisphere have been completed with the aid of thermal diffusion isotopic enrichment and lowlevel proportlonal countlng with rlse-tlme circuitry. Thermal diffusion proved to be an excellent means for increasing the very low natural specific activity of 37Ar. Good enrichment factors (-70) were achieved in a period (-6 days) significantly less than the half-llfe (35.1 days) of 37Ar. Results for Southern Hemisphere samples confirmed that the measurement process was sufficiently sensitive to quantltatively determine the natural, troposphere levels of 37Ar, but they suggested the possibility of some contamination of the Southern Hemisphere 37Ar from artificial sources.

Radionuclides which are produced by the nuclear interactions of cosmic radiation with the Earth’s atmosphere have been of considerable interest since the discovery of radiocarbon dating ( 1 ) . Investigations of the production and subsequent behavior of such radionuclides can provide very useful information for a number of scientific disciplinesranging, for example, from cosmic ray physics to hydrology ( 2 ) .Rare gas nuclides are a particularly interesting class to study, for they are not influenced by chemical processes that modify the behavior of more reactive species. Because of this, and because of its relatively short half-life (35.1 days), cosmic ray production 37Aris quite promising for the study of atmospheric mixing processes. The aim of the experiments described below was to evaluate the feasibility of measuring natural 37Ar with sufficient precision to permit meaningful deductions concerning atmospheric mixing. The eventual objective, the determination of vertical diffusion coefficients, may be accomplished by using a) the 37Ar vertical production profile and b) the observed 37Ar atmospheric concentration profile as input for a meterological mixing model (3,4 ) . Measurement of the natural levels of the atmospheric radionuclides is quite difficult, especially when there is considerable isotopic dilution, as with the argon radioisotopes. Thus, for cosmic ray production 37Ar, the mean tropospheric production rate corresponds t o a concentration of only -0.003 dpm/l.-Ar ( 5 ) .In order to obtain the requisite sensitivity for these measurements, it was necessary t o evaluate means for background reduction (rise-time discrimination) and specific activity increase (isotopic enrichment). As will be seen below, both of these routes were fruitful. An additional problem arose, however. This was the (temporary) contamination of the Northern Hemisphere with noncosmic 37Ar, presumably due primarily to nuclear testing, with a possible contribution from reactor

operation. For this reason, we sought and obtained samples of argon from the Southern Hemisphere, as the nuclear lifetime was short enough to preclude serious contamination from interhemispheric mixing (6, 7). Sensitivity Requirements. The minimum 37Ar concentration ( ZQ) that can be quantitatively measured (relative standard deviation = 10%) is given by Equation 1.

where ZQ = radioactive concentration (dpm/l.-Ar); LQ = number of counts for quantitative determination = 50(1 [I + !~B/25]’/~); Teff = Effective counting time = ~ ( -1 7 = mean life for 37Ar (50.6 days); !JB = expected number of background counts collected in counting time, t ; V = volume of argon counted, a t STP; Z = isotopic enrichment factor; and E = 37Ar detection efficiency. Derivation and discussion of the expression for LQ may be found in References 6, 7, and 8, which develop also the basic statistical criteria for establishing limits of detection and quantitation. As noted in References 6 and 7, the counting time, which affects the magnitudes of pg and Teff,has an optimum value a t about two half-lives. (Note that we consider in Equation 1 only the imprecision arising from counting statistics, as this represents the limiting value for the overall precision.) In order to construct a system which is sufficiently sensitive to measure the average concentration of cosmic ray produced 37Ar in the troposphere, we must first determine the factors in Equation 1 for which ZQ = 0.003 dpm/l. Such an evaluation is conveniently carried out by examining the product ( VZ)Q for different measurement times and counter backgrounds. In Table I, (VZ)Q is presented for three measurement times and four background rates, assuming E = 0.70. The first two measurement times are reasonable

+

Table I. ( V I ) Qvs. Background Rate and Counting Time0 Counting Time 2rl/2(l.01X

Background r a t e

0 cpm

103min

104min

IO5 min)

48

5.1

0.87

0.005 cpmb 50 7 .O 2.4 0.08 cpmc 73 17.0 8.3 0.5 cpm 130 39.0 20 .o a ( V 1 ) is ~ the product of argon volume (STP) and 37Ar enrichment factor corresponding to the determination (RSD = 1 0 % )of ’?Ar at the level 0.003 dpm/l-Ar. b Brookhaven Counter ( V = 0.15 l., STP) used in this work, incorating rise-time discrimination. CBern Counter ( V = 5.0 I . , STP) used at the Physikalisches Institut ( 5 ) .

ANALYTICAL CHEMISTRY, VOL. 48, NO. 3, MARCH 1976

607

TITANIUM 900

C

Table 11. Characteristics of Hot-wire Thermal Diffusion Columns in the 11-Column Argon Cascade Cold wall radius 0.953 cm Hot wall radius 0.08 cm Length 7.32 m Hot wall temperature 800 “C Cold wall temperature 30 “C Operating pressure 550 mm Hg Initial transport coefficient 1.994 x g/s (0.9 x theoretical value; one mass unit basis) Remixing coefficient 0.01375 g cm/s

SAMPLE OUT* -TOP

RESERVOIR

THERMAL DIFFUSION COLUMN -CIRCULATING

PUMP

T

- B O T T O M RESERVOIR

Figure 1. Schematic diagram of 11-column thermal diffusion cas-

cade for argon-37 enrichment limits for low-level counting, lo4 min being approximately 1 week, and the final measurement time--2t1/2 = IO5 minyields the best sensitivity for 37Ar. (Note that Teff, which can never exceed T , is but 5.46 X l o 4 min for the last measurement time.) Of the four background rates listed in Table I, two (0 and 0.5 cpm) represent limiting values for typical high-sensitivity gas counters, while the remaining two correspond to the rates of the counters used in Bern and a t Brookhaven (this work). These two counters are certainly among the best currently available for the measurement of small quantities of electron capturing or low energy &emitting nuclides. The numbers appearing in Table I, ( VZ)Q, represent simply the argon volumes required for counting unenriched ( I = 1) samples. I t is thus evident that the two “real” counters, as currently used, are not sufficiently sensitive in the absence of isotopic enrichment. Other points which may be noted in Table I include a) the diminishing gain in sensitivity for counting times exceeding -1 week, and b) the relatively slight gain when the background rate is reduced below 0.005 cpm. In fact, it can be readily shown ( 8 ) , that WB 5 11 counts may be considered “effectively zero” in terms of its effect on ZQ. Finally, the product, ( VZ)Q, appearing in Table I allows one to derive the enrichment factor required for a given counter and counting time. Thus, for a 1-week counting period, the Bern counter would require I Q = 17.515.0 = 3.4 and the Brookhaven counter, ZQ = 7.010.15 = 47. Additional considerations which enter into the selection of counters include the facts that a) for a given (VZ)Q, the counter having the smaller capacity ( V ) may be expected to have the smaller background rate and therefore the greater sensitivity, and b) the currently available enrichment facilities are suitable for producing relatively small quantities ( V 300 ml) of highly enriched ( I = 100) argon isotopes.

ISOTOPIC ENRICHMENT Isotopic enrichment has been successfully used to obtain increased sensitivity for the detection of low levels of carbon-14 and tritium, and several authors have described equipment and procedures for quantitative sample enrichment by gas phase thermal diffusion (10-14). The enrichment of 37Ar can also be accomplished by thermal diffusion, provided that a reasonable separation can be accomplished in a time which is short relative to the half-life. The combined requirements of a high enrichment factor and a short time of operation cannot be readily fulfilled by use of a single thermal diffusion column. A multiple column cascade is needed. The cascade, or series-parallel arrangement of columns, must have the following characteristics: 1) High rate of initial transport of 608

ANALYTICAL CHEMISTRY, VOL. 48, NO.

3,

MARCH 1976

37Ar. 2) Sufficient length to reach an enrichment factor, a t the steady-state, approximately equal to the square of the desired enrichment factor. (This assures an adequate transport rate throughout the separation process. The transport rate, of course, decreases to zero a t the steady state.) 3) Low material holdup in the high enrichment end of the system. An existing argon isotope separation cascade a t Mound Laboratory has characteristics similar to those required. The configuration of the cascade is depicted schematically in Figure 1, and the dimensions and operating parameters of the columns are given in Table 11. The cascade has a large reservoir of 85 STP liters a t the bottom end and a small enriched sample reservoir a t the top. The seven coiumns in parallel a t the bottom provide a high rate of transport of the rare light isotopes from the large reservoir to the cascade. The 5-column series length of the cascade, nearly 37 m, provides the potential for large separation factors between the enriched sample reservoir and the bottom reservoir. The cascade is designed for steady-state, continuous-flow separation of the stable isotopes argon-36 and argon-38. Although it is suitable for the transient batch enrichment of 3iAr, it is not necessarily an optimum configuration for this purpose. Southern Hemisphere argon samples were taken from the air separation plants of Commonwealth Industrial Gases, Ltd., in Sydney, Australia, and La Oxiegena S.A.I.C. in Buenos Aires, Argentina. The dates of sampling and other pertinent information are given in Table 111. Following purification over titanium sponge a t 900 “C, a 100-1. portion of each sample was placed in the previously evacuated separation system. The hot wires of the columns and the circulating pumps were turned on, and the system pressure was adjusted to the operating value of 550 mm Hg.by addition or removal of small amounts of gas as required. Thereafter, mass spectrometric samples of 2 STP ml each were taken daily. The runs were terminated when the analytical results indicated that an argon-37 enrichment factor of 70 or greater had been reached. The enrichment sample loop was isolated and the final sample was removed through the bed of titanium sponge a t 900 “C. The composition in the enriched sample loop as a function of time is depicted for the second run in Figure 2. There were some difficulties with the equipment a t the start of the first run; hence, the effective starting time is uncertain by as much as 0.1 day. The holdup in the product sample loop was larger for the second run than it was for the first (0.55 g as opposed to 0.45 g); thus, the isotopic concentrations changed a t a somewhat slower rate in the second run. When the experimental work was completed, calculations were undertaken to match the theoretically calculated performance of the cascade to the observed performance. I t was originally planned that the column transport parameters, which are not precisely known, would be var-

--

Table 111. Enrichment of Argon-37 in Southern Hemisphere Samples Run 1 Sample location

Sample No. Sample volume, l . - h Date of sampling Received at Mound Enrichment run started

Run 2

Port Kembla, Buenos Aires, NSW, Australia Argentina

NBS-26 100 I. 1 2 Jan 72 26 Jan 72 7 Feb 72 27 Jan 72 28 Jan 72 8 Feb 72 -2 p.m. 9:30 a.m. Enrichment run ended 3 Feb 72 14 Feb 72 1 : i 5 p.m. 2.45 p.m. 0.55 g Holdup in enriched sample loop 0.45 g Final composition (mol % ) Argon-36 41.8 39.0 Argon - 38 1.72 175 Argon-40 56.5 59.3 Calculated argo n-37 72 3 7oi 1 enrichment factor

END OF RUN

*

NBS-23 100 1.

+

ied in the calculations and that a trial and error process would be required to effect a match. The argon-37 enrichment factor would be taken from the set of calculations which most closely represented the enrichment of argon-36 and argon-38 as a function of time. Accordingly, initial approximations to the column transport ( 1 5 ) coefficients were evaluated from theory based on the use of the Dymond-Alder (DA) intermolecular potential model to predict transport properties of argon (16). (Potential parameters for the DA model were taken t o be: elk = 142.8 K and u = 3.258 8, on the basis of recent viscosity data; the Kihara second approximations (17) were used to calculate properties.) On the basis of previous experience with similar systems, the initial transport coefficient was taken to be 90% of the theoretical value calculated on the basis of the DA potential. The transient behavior of the cascade was subsequently calculated by numerical integration of the system of nonlinear differential equations which describes the change of the concentrations with time (18).Thus, for a multicomponent mixture, the rate of change of the mass fraction of component i is given by

where wi is the mass fraction of component i, and the d, are the atomic mass differences relative to an arbitrarily chosen key component, e.g., di = mi - mk, where mi is the atomic mass of component i, and mk is the atomic mass of the key component. The quantity is the mass holdup per unit length, H, is the reduced initial transport coefficient, K is the remixing coefficient, n is the number of components, and z is the vertical coordinate. The equations given above were solved by an implicit forward difference method, subject to material balance constraints. Provisions were made in the computer program to take into account the material held up in the various reservoirs and interstage connections. An arbitrary, low value of 1 X mol % was assumed for the starting concentration of argon-37, and argon-37 enrichments were calculated relative to this value. I t can be shown from Equation 2 that the calculated enrichment factor is independent of the choice of the starting composition if the concentration of argon-37 remains negligible a t all times and all positions in the cascade.

0

2 TIME D A Y S

Flgure 2. Isotopic enrichment of atmospheric argon sample from Buenos Aires, Argentina (NBS-26) The solid lines are calculated

With minor adjustments to the estimated amount of material held up in the enriched sample loop, it was possible to duplicate for each run the experimental concentrations of argon-36 and argon-38 within the estimated error limits of the analyses. T h e adjustments required were less than 10% of the first estimation of the holdup. Adjustments to the column coefficients were unnecessary. (It should be noted that the reported difference in the enriched sample holdup between runs 1 and 2 was the result of a change in the sample loop pressure between the two runs and was not created by the adjustment procedure.) Final enrichment factors for argon-37 were 72 & 3 for sample 1, from Port Kembla, NSW, Australia, and 70 & 1 for sample 2 from Buenos Aires, Argentina. The uncertainty estimates are based on combined effects of analytical error, time scale uncertainty, and uncertainty in volume of the enriched sample loop. As noted earlier, there were minor operational difficulties during the first run; these account for the larger spread. The uncertainty estimates were very crude and should not be taken as standard deviations in the statistical sense.

ARGON-37 MEASUREMENTS F i n a l Purification a n d Counting. Following isotopic enrichment, the samples were shipped to the laboratory of Raymond Davis, Jr., a t Brookhaven National Laboratory for final purification and radiochemical assay. The volume of each sample was a t this point about 300 cm3. Only about half of this was eventually used for the measurement owing to restrictions on the sizes of the existing counters. Purification of the samples consisted primarily of passage over hot titanium (900 "C) to remove any residual 0 2 or Nz,followed by passage over room temperature titanium t o eliminate any potential tritium contamination. No effort was made to separate s5Kr from the sample since it was felt that the original fractional distillation of the sample a t the air reduction plant, followed by isotopic enrichment of a low atomic weight fraction, should have totally removed any krypton. The sample was collected in a large volume automatic Toepler pump, and its volume was accurately determined. The sample was finally transferred to the counter through a small charcoal trap (-78 'C) for radon removal. An aliquot of pure methane, sufficient to make a 95% Ar, 5% CH4 mixture, was also added to the counter. Two simple cylindrical proportional counters were used for the radiochemical assay of the two samples. With the ANALYTICAL CHEMISTRY, VOL. 48, NO. 3, MARCH 1976

609

ADP

1

Table IV. Results for NBS-23 and -26 NBS-23 (Australia)

272.0 t 19.0 0.55, 11025

374.0 i 22.0 0.56, 14500 t , min 72 70 I 150 143 V (STP) (ml) 12 26 t,a 1972 44 57 Tb days 0.0078, C o (dpm/l.-Ar) 0.0083, 8.2% Uncertainty-random 6.1% ( u p e s t i m a t e d systematic 4.7% 4.7% a Time of atmospheric sampling. b Mean time of counting. S t u;J (counts)

E

E

NBS-26 (Argentina)

Figure 3. Representation of the two-dimensional spectra of ADP vs. E Counts within the 37Arwindow ( Yi)are delimited by the x-ray band (X) and the 37Ar energy band (dotted lines). Low ionization density events (y) generally lie below the diagonal x-ray band

exception of the cathode materials, the counters were essentially identical. The counter envelopes were constructed of high purity quartz and the anode material was 1-mil tungsten wire. The cathode materials were zone-refined aluminum and high purity iron foil for NBS 23 and NBS 26, respectively. Both counters had a sensitive volume of 115 cm3 and a cathode area of 200 cm2, Within the 37Ar window, the backgrounds of the two counters were not significantly different. In order to enhance the sensitivity and specificity of the 37Ar measurement, a pulse rise-time system developed for the Brookhaven Solar Neutrino Experiment (19),by V. Radeka and L. Rogers, was adopted for use with these counters. The rise-time system measures two parameters. The E-signal, which represents the energy deposited in the counter, is presented in the conventional way via a standard spectroscopy amplifier. In addition, the charge collected by the pre-amplifier in the first 10 nanoseconds is integrated, differentiated, and stretched for pulse height analysis. The resulting ADP-signal (ADP = amplitude of the differentiated pulse), together with the E -signal, were fed into the two ADC inputs of a 4096 channel dual parameter multichannel analyzer. Events due to true 37Ar electron capture decays have, for a given total energy deposition, a much higher specific ionization (Le., shorter ionization tracks) in the counter than do minimum ionizing Compton electrons from y-ray interactions in the counter wall. (The latter type of event is responsible for most of the normal background in x-ray proportional counters.) As a consequence of the difference in track length, the Auger event gives a larger ADP pulse than the Compton event, on the average. On a two-dimensional plot of ADP vs. E , xrays or Auger electron events define a line of slope greater than that of Compton events-thus providing a means of discrimination against background events while simultaneously recording these background events. This information, together with appropriate calibration data can be utilized to calculate the background spillover into the x-ray region, thus reducing the required counting time by eliminating the necessity of independent background measurements. Calculations and Results. Analysis of the two-dimensional array of counts was carried out using a two-dimensional window situated along the diagonal corresponding to x rays, and somewhat less than two fwhm in each (ADP, E) direction. (The array is shown schematically in Figure 3.) Calibration of the window was accomplished using “Co as being representative of background pulse shapes, and 55Fe, 610

ANALYTICAL CHEMISTRY, VOL. 48, NO. 3, MARCH 1976

as an x-ray emitter. The total counts inside ( Y i ) and outside ( Y o )the ADP window and within the E window, may be represented as,

Yi = Si

+ B; = Sai + Bpi + ei

(34

Yo = So

+ Bo = Sa, + Bpo + e ,

(3b)

Equations 3 simply partition the total signal (S) and background ( B ) counts into compartments within (i) and without (0)the ADP window, by means of the normalized calibration vectors a and j3 (e represents the random error vector) a and j3 were determined with 55Fe and “Co, as noted above. For the window selected: a = (0.985, 0.015); j3 = (0.27, 0.73). Thus, 98.5% of the 37Ar peak lay within the ADP window, and the rejection ratio was 0.73. (The rejection ratio, Po, represents the fraction of background pulses rejected.) Within the E window selected, the total 37Ar detection efficiency (integrated over ADP) was about 56% for the two samples. As will be later shown, the window-background (Bi)was equivalent to about 8 countslday. Solution of Equations 3 simply involves simultaneous equations, and Poisson counting errors may be propagated to give, = (@,Yi- @iY,)/Dand &4 = (p;Yi

+ p,?Yo)1/2/D

where D = P0aL- pia,. For NBS-23 and -26, the following observations were made:

NBS-23 NBS-26

Yi

YO

S

OS

327 445

165 214

271.6 373.5

19.1 22.2

Counting time, t , min

11 025

14500

The remaining pertinent parameters and the calculated results for these two samples are given in Table IV. Note that the total random error is slightly larger than that due to Poisson counting statistics, due principally to random error components of V and I . The advantage of pulse shape analysis in reducing the background (as shown by the rejection ratio) and a t the same time giving a “real time” measure of the background may be seen by solving Equations 3 for B . This gives for both samples total background rates (within the energy window) of 0.020 cpm, and net background rates ( B i l t ) of only 0.005 cpm. An attempt was made to confirm the radiochemical purity of the final argon sample by decay analysis, but the availability of counters and the low activity levels did not make it possible to make a very precise check. Independent counts of the two samples were, however, consistent with the 35.1 day half-life. Because of the lack of precision in

0 5

these comparisons, we could only demonstrate that the decay rate corresponded to a half-life greater than -17 days. Finally, upon examination of the results in Table IV, one sees that the results for the two samples were not statistically significantly different, and that the random counting errors were well below 10%. Although the sampling sites (Australia, Argentina) differed by 150 degrees of longitude, the latitudes were nearly the same (35' S),and the sampling times differed by only two weeks.

I

I

I

I

I

I

I

I

I

I

r-

01

DISCUSSION I t is of great interest to compare our Southern Hemisphere results with those of samples that we had shipped to the University of Bern, taken from both hemispheres. Samples of the unenriched argon from both Southern Hemisphere sites were sent to H. Loosli of the Physikalisches Institut. These samples were measured for about 1 week each, and gave results (20,211:

0 01

\

9

0 001

Australia -0.0085 f 0.0029 Argentina -0.0061 f 0.0022 These results are consistent with those quoted earlier, but, of course, the power of the comparison is somewhat limited because of the imprecision of the measurements made without enrichment. Comparison with Northern Hemisphere 37Ar levels is possible via measurements carried out a t Bern on a number of samples, including several shipped by NBS and one from Bombay. Results for the Bern measurements are indicated by the solid line as shown in Figure 4 (20, 21). Examination of the specific activity of atmospheric 37Ar vs. time in the two hemispheres makes it quite clear that the level in the southern Hemisphere was lower in January 1972 by a factor of 4 t o 6. On the other hand, it is equally clear that the level that we observed for the Southern Hemisphere is significantly higher than the aforementioned mean tropospheric level, due to cosmic ray interactions, of about 0.003 dpm/l.-Ar. Although this difference could be due in part to North-South mixing, it seems unlikely that this could account for most of it, because such mixing times are of the order of 2 years (6, 7). The most likely source of artificial 37Ar in the Southern Hemisphere a t the time our samples were collected is the series of French nuclear tests carried out during the summer of 1971 in the South Pacific. As has been observed by Loosli e t al. (20, 21), high yield nuclear tests in the Northern Hemisphere have resulted in initial 37Ar concentrations which exceed the natural levels by a factor of one hundred or more. Taking into account the 35-day half-life of 37Ar, it thus becomes clear that our January 1972 samples may have been significantly contaminated from the high yield nuclear test which took place in mid-August 1971 in the South Pacific. Contributions of 37Arfrom Southern Hemisphere reactor operation cannot be entirely excluded, although they are not known to have been significant for the samples collected in January 1972, particularly in comparison t o the residue from the foregoing nuclear tests. On the other hand, continued increases in the production of nuclear energy, especially in the Northern Hemisphere, are likely to distort and possibly even mask the natural tropospheric production of 37Ar. Matuszek, Paperiello, and Kunz have shown, for example, that air collected within 100 km of a nuclear reactor may contain excess 37Arranging from about 2% to 3 X 10% of the natural (cosmic-ray produced) level, depending upon the nature of the reactor (22). Clearly, the higher levels of contamination are not yet with us, as the observed

0 WOl

Figure 4. Ground-level 37Ar concentrations vs. time The upper solid line is taken from Northern Hemisphere measurements by the Bern group (20,21).using a 35-day half-life. The two points shown represent our Southern Hemisphere measurements of enriched samples. The upper dotted line represents the estimated average cosmic ray tropospheric production rate, and the lower dotted line, the approximate production-inflight

tropospheric levels decay approximately to expected natural levels following a large nuclear test. The possibility of significant reactor contributions (210% natural) cannot, however, be ignored in the future with increasing density and power levels of nuclear reactors. An additional question that may be raised with respect to the Southern Hemisphere samples, is whether there might be significant additional amounts of cosmic-ray induced 37Ar during the long, high-altitude air flight (from Australia). The question is not a trivial one to treat, because one faces slight nucleonic attenuation by the aircraft and cylinder, and yet the multiplicity of secondary neutrons is clearly higher than in the atmosphere. (Surface transport was, of course, out of the question because of the half-life of 37Ar.) A very crude estimate of this excess production was made using the approach of La1 and Peters (2) and assuming a total flight time of 20 h a t 35 000 ft ( ~ 2 4 0 g/cm2). The resulting estimate which is shown as the bottom dashed line in Figure 4 proves not to be a very significant contribution. In conclusion, i t has been demonstrated t h a t thermal diffusion enrichment operates successfully on a time scale which is acceptable, in terms of the 37Ar half-life. The degree of enrichment achieved makes feasible the measurement of natural tropospheric levels of 37Arwith the requisite precision (