Trace element analysis in water by proton activation - Analytical

Determination of arsenic, cadmium and zinc in river water by neutron activation analysis. S. Sava , L. Zikovsky , J. Boisvert. Journal of Radioanalyti...
1 downloads 0 Views 614KB Size
Trace Element Analysis in Water by Proton Activation S. F. Bankert Department of Applied Science, University of California, Davis, Calif.

S. D. Bloom and G. D. Sauter Lawrence Livermore Laboratory, Livermore, and Department of Applied Science, University of California, pavis, Calif.

Charged particle activation analysis (CPAA) has been applied to the measurement of trace amounts of pollutants in calibrated water samples. The activating particles are 14.7 MeV protons in a microampere (or less) beam from the Lawrence. Livermore Laboratory 30-inch cyclotron. The short half-live gamma spectrum ( T I I P S 2 min) of the activated sample is assayed in a Nal-Ge(Li) anti: coincidence detector system. A fast transit system carries the sample from the bombardment site to the counting position in about 1.8 seconds. As of now, the technique is capable of simultaneously detecting ppm quantities of boron, nitrogen, bromine, selenium, sodium, cadmium, and chromium in concentrated water samples evaporated to dryness on tantalum foils.

For many years both public and private agencies have been involved in the analysis of water from many sources -lakes, rivers, oceans, and water reprocessing plants. These analyses may include not only tests for such gross sample properties as conductivity, turbidity, and acidity, but also measurements of the concentration of elements such as Na, C1, S, N, Mg, As, B, U, Se, and Cd ( I ) . Various techniques for the determination of elemental concentrations in water samples are currently available. These include chemical tests, neutron activation analyses, and X-ray fluorescence measurements. Although the ideal situation would be to analyze the sample just as it comes from the source, all of these methods require a certain amount of sample preparation. Unfortunately, no single technique is comprehensive; in fact, there are some elements of interest ( e . g . , K) that presently cannot be adequately measured in the trace concentrations in which they normally occur in water samples. Chemical testing is undoubtedly the most widely used method, the many available tests covering a broad range of elements (2). However, it is usually fairly slow, and generally a separate test must be made for each element of interest. Neutron activation and X-ray fluorescence, although more rapid on an element per hour basis, are not so widely used and are incapable of measuring low atomic number elements. There is an obvious need for improving our current ability to measure the concentrations of various elements in water samples, both by refining existing methods and by developing new techniques. Charged particle activation analysis (CPAA) by multiple gamma spectroscopy could well be a useful supplement to existing techniques. It allows rapid, simultaneous measurement of many elements, including those with low atomic numbers. The experimental method is very similar to most existing methods of (1) "Report of the Committee on Water Quality Criteria," FWPCA, April 1, 1968. H. A. Swonson and H. L. Baldwin. "A Primer on Water Quality," USDI, 1965. "Manual on Industrial Wastewater," 2nd ed., 1966 printing, ASTM Pub. No. 148-1. (2) "Standard Methods for the Examination of Water and Wastewater," Amer. Public Heaith Assoc. inc., Twelfth Edition, 1966.

692

ANALYTICAL CHEMISTRY, VOL. 45,

NO. 4,

APRIL 1973

neutron activation analysis (NAA). In either case, a pre.pared sample is activated by irradiating it with nuclear particles, and the resulting gamma ray spectrum of the activated sample is then measured with an appropriate gamma detector. The number of gamma rays detected at an energy characteristic of' a given element can then be related, using a previously obtained calibration curve, to the amount of that element that was present in the original sample. In this paper we emphasize that when we speak of CPAA or NAA we shall mean specifically this approach (3). Like neutron activation, proton activation is potentially capable of detecting many elements with each activation. In fact, it is possible to activate all elements with an atomic number greater than 4. although detection of the resulting radioactivities is another problem, as we shall see. In addition, proton activation is particularly suited to rapid sample analysis since the activated isotopes, in general, have shorter half-lives (seconds and niinut,es in lieu of minutes and hours) than their neutron-produced coun; terparts. As might be expected, the half-life is the determining factor for the optimum irradiation and the counting times, and hence the total analysis time. In this paper, we briefly describe our effort to develop a rapid, sensitive technique (based on proton activation analysis) for determining the elemental content of water samples. While our attempt to analyze unprepared samples failed because of the intense positron annihilation background, due primarily to the activation of oxygen, we found that evaporating the samples considerably alleviated the background problem, although "real" sample analysis is still complicated by the oxygen activation. Nonetheless, we feel that proton activation analysis is a promising approach to the problem of trace element analysis not only in water but also in other media of environmental interest. The work we report here was essentially a feasibility study; our aim was to examine the potential of the technique. We have made no attempt to introduce all the refinements that are possible or even practical. The optimization of the technique for any particular element or group of elements will involve, a t a minimum, determination of the most appropriate proton energies and the best activation-counting cycle. These in turn include studies of proton-induced reaction cross-sections and the interferences between various elements. All of these investigations remain to be done.

EXPERIMENTAL Liquid Target Experiments. T h e ideal approach in trace element analysis in the case of water would be t o take a sample from a stream. lake. or river a n d analyze it w i t h o u t having t o prepare or alter it i n ' a n y way. In t h e liquid target experiments described here, a n a t t e m p t was made t u do exactly this. (3) An excellent review of CPAA using short-lived nuclides is given in J. L. Debrun. D. C. Riddle. and E A. Schweikert, Ana/. Chern.. 4 4 , 1386 (1972)

-Pressure

-1

Inlet

l i t e r Reservoir

1 cm. d i a . x 1/2 em. deep t a r g e t r e s e r v o i r w i t h 1/2 m i l Nickel f r o n t cover f o i l .

t e r Cooled 1 . 0 0 ~

250 ml. Reservoir

Transit time from target t o counting reservoir less t h a n 1 sec. a t 2 p . r . i . g . a i r p r e s s u r e .

Figure 1. Schematic plan of the liquid target experiment

Figure 1 is a schematic diagram of the configuration of the flowing water system which was used. The salt solutions used in the tests of the system varied from 10 ppm to 1000 ppm and were produced by dissolving known amounts of salt in one or two liters of distilled water. These prepared samples were forced to flow from a storage reservoir through the proton beam to a second reservoir in front of the Ge(Li) detector. As a prelude to a run, the system was flushed for a few seconds. Then the test solution was allowed to flow, and the proton beam shutter was opened for 20 seconds. Immediately after the irradiation, the flow was stopped and the gamma activity of the irradiated solution was counted for 100 seconds. Generally the accumulated charge was less than 0.1 F C , because the beam current had to be held below 25 nA to avoid saturating the detector-amplifier system. The vast majority of this activity was background, due almost entirely to the 511 keV gamma ray resulting from the 1 6 0 ( p , a ) 1 3 X reaction. 13X is a 10.0 minute p+ emitter, thus the 511 keV annihilation radiation. At high .concentrations, it was possible t o see very small peaks due to N a and C1. Similar background problems exist for activation with alphas [160((u,n)19Ne, 18 sec. fit] and deuterons [160(d,n)17F,66 sec, B + ) . Thus charged particle activation analysis of liquid water samples is limited to cases where the element concentrations are high, such as in sea water and other brines; in its present form, CPAA is not a practical technique for trace element analysis of unprepared water samples. Evaporated Sample Experiments. Because of the very low sensitivity of the liquid sample technique due to the severe background problems caused by leO, it was vital to remove as much of the 511 keV activity as possible. This was done by evaporating the sample and by slightly increasing the complexity of the electronics. The preparation of calibration samples was much the same. Basically, a known amount of salt was dissolved in distilled water to create concentrated standard solutions. A total of 40 11 of these solutions was then evaporated to dryness on a T a foil after a concentration step of 100 to 1. Care was taken to reject samples with visible wall losses, and the procedure was always carried out in the same manner to keep vaporization losses the same in all samples. The T a foils were 3/s inch in diameter and had a 5-mm diameter, 1-mm deep indentation pressed into them. Tantalum was chosen because it does not activate appreciably for the proton energies used here. The sample foils were carried to the cyclotron beam by a pneumatic "rabbit" with a transit time, from irradiation position t o counting position, of 1.8 seconds. The amount of beam hitting the foil was controlled by a pneumatic shutter and a series of collimators. The collimators were chosen to be 6 mm in diameter in order to provide a beam that entirely covered the evaporated drop in the 5-mm diameter indentation. This limited the maximum beam current on the target to about 1.5 PA. Collimators before and after the nickel exit foil reduced the activation of the target holder, a tantalum ring. Control of each irradiation cycle was done electronically by the monitor timing and cycle control system (see Figure 2 ) . At the

end of each irradiation cycle, the data stored in the multichannel analyzer were transferred onto paper tape for later processing. Spectra were taken using a KaI detector in anti-coincidence (at 180") with the Ce(Li) detector (see Figure 2). This was done to reduce the unwanted 511 keV activity in the spectra produced by residual oxygen in the sample and other positron emitters produced by the irradiation. The effect of the XaI anti-coincidence was to reduce the unwanted 511 keV peak by a factor of2 to 3. In order to keep the dead time of the analyzer a t a reasonable level (less than 30%). the beam currents used in the irradiations varied from 0.1 to 1.0 FA, depending on the sample characteristics. Forty-second irradiation and 60-second count times were chosen because they correspond to 2 to 3 times the half-lives of most of the elements studied. Limits on the activity of the sample were set by the maximum count rate (about 30 to 35 X lo3 counts/second) that could be tolerated by the analyzer and the detector-amplifier systems. Improvements in sensitivity could be made if these systems were capable of handling higher count rates.

DATA ANALYSIS As indicated in the previous section, the data from each experimental run (1 foil per run) were transferred onto paper tape. These tapes were then converted to computer cards for analysis on a CDC-3400 computer. The detector system was calibrated for energy response by placing standard calibration sources in front of the detector and taking a 1000-second spectrum. A third-order polynomial was least squares fitted to the calibration peak energies and centroids. From this polynomial. the energies of peaks in later spectra could be determined. Spectra from the irradiated sample foils were analyzed in the simplest possible manner. These spectra were of two classes. The first class consisted of spectra from foils on which a known amount of a known salt was deposited ( L e , , calibration foils), and the second of unknown amounts of unknown salts. The former were used to calibrate the system in terms of its response (counts in a given peak per unit integrated beam current) as a function of the concentration (ppm) of a given element in a sample; the latter were used to test the proton activation method after the system had been calibrated. From the calibration foils and the energy us. channel calibration, the locations (energies) of the peaks from each element of interest were known. Each peak was taken as five analyzer channels wide-the channel containing the maximum number of counts plus the adjacent two channels on each side. The sum of the counts in these five channels was identified as the total area under the peak, and the square root of the total area was taken as the statistical error in ANALYTICAL CHEMISTRY, VOL. 45, NO. 4, APRIL 1973

693

I

R a b b i t Control Apparatus

1-1 ~

I

\ \ \

R a b b i t in Bombard Position

:

f

Accelerator

,

Target Material

i

I

I I

I

i

\ E x i t Foil

Beam Shutter-/ Rabbit Tube

1

Cycle Control

1

C E (ti) Detector and p r e - a m p

I7

i \ \

4

Amplifier

f

7Count

Coincidence Counter

1

Position J

IDiscriminator]

~

Figure 2. Schematic plan of the evaporated target experiment

Pprn Br

Figure 3. Average yield as a function of bromine concentration. Reaction: 7 9 B r ( p , n )79 Kr, T 1 / 2 = 55.0sec, EY = 126.5 KeV

the peak. The background count per channel was determined by fitting a straight line to channels on either side of the five-channel peak. The background was subtracted from the total peak area to get the net peak area. The net peak area and statistical error were then normalized by dividing by the integrated current. For each element of interest, the final calibration for sensitivity was achieved by averaging all the normalized peak areas (counts/pC) for all the calibration foils of the same concentration (ppm). A straight line was leastsquares fitted to the counts/FC us. ppm data. The least squares fit was a weighted fit using the inverse of the total error in the counts/pC a t each concentration as the weighting factor. The total error was taken as the square root of the sum of the squares of the statistical error, the 694

ANALYTICAL CHEMISTRY, VOL. 45, NO. 4 , APRIL 1973

0 0

I

I

I

I

I

4

0

I2

16

20

24

Ppm €3 Figure 4. Average yield as a function of boron concentration. Reaction: l o B ( p , n ) l O C T, 1 / 2 = 19.4sec, E-, = 715.4 keV sample preparation error (if it could be evaluated), and the variance in the counts/& for foils of the same concentration.

RESULTS AND DISCUSSION Although we tried to measure concentrations of a large number of elements, we were successful in seeing peaks only from the following seven: IOB, I4N, 23Na, W r , soSe, and 111Cd. At high concentrations (100 ppm), under the irradiation conditions described above, no peaks were observable in the following cases: Si, Ag, C1, K, P, S,As, Al, Mg, Hg, Pb, and Ca. Typical calibration curves are shown in Figures 3 through 6. They are plots of yield, or normalized area (in counts/&) under the peak for the element being consid-

I

I

I

I

,

10,000 z

1,000 7

2 K

-

l0-

-

-

-

w

-

-

I)

v)

c

0

u00

-z 0 (D

100

‘--------jJ

-

--v-.z

;

-

Table I . Detection Limits, Half-Lives, and Characteristic Radiations of Elemental Water-Pollutants Detected in the Present Work Radiations“ Element

T,/,

Type

Ey

M inimum kg

detected

Minimum ppm detected

Boron 19.4sec p+ 0.72 2.4 0.6 p+ 2.31 2.2 0.6 Nitrogen 71.0sec 1.8 0.4 12.1 sec fit 0.44 Sodium Chromium 21 min pt 1.43 4.5 1.1 4.8 sec IT 0.21 5.8 1.4 Selenium Bromine 55 sec IT 0.13 3.8 0.9 IT 0.54 6.1 1.5 Cadmium 8 min .p* refer to positron or electron emissions: I T means isomeric transition. E7 is the energy of the gamma ray (in MeV) used for identification purposes. The bombard/count cycle and details pertaining to it are given in the text. Most of the bombardments used a total integrated current of =20 @C.The entries in the last two columns pertain to calibration, samples only (see Results and Discussion). Ppm

N

Figure 6. Average yield as a function of nitrogen concentration. Reaction: 14N(p,n)140 = 71.0sec, fY = 2309.5 k e V The slight positive zero-intercept of the ordinate in this case is due to the fact that a certain amount of (atmospheric) air adhered to the foil through the bombard part of the cycle

ered, as a function of the concentration (in ppm) of the element in the sample. The solid curve is a linear least squares fit of the data. The dashed curves are the maximum and minimum straight lines due to statistical uncertainties in the fitted straight line coefficients. Table I lists the minimum amounts and concentrations detected (in the calibration samples) of each element that we were able to detect with the present system. This estimate, therefore, does not include the effects of interferences in typical “real” gamma spectra, which will be discussed later. It is important to note that the energy of the detected radiation grossly affects it detectability in “dirty” samples. For instance, the detection of Se depends on the ability to extract a 210-keV peak from a background which is an order of magnitude higher than the background in the vicinity of the Na peak a t 440 keV. A “mock” sample of many elements is shown in Figure 7 , illustrating the appearance of a spectrum free of (mainly oxygen) interference.

The main sources of error in our work are counting statistics, sample preparation uncertainties, and irradiation uncertainties due to inconsistency in rabbit alignment, all of which are independent of the beam characteristics. In addition to these three, there is one more cause of error which is completely a function of the beam intensity. This error would not be important if it did not change significantly with different beam intensities. The end result is that the total measured integrated current is not the actual integrated current falling on the target. Thus, the experimental points given in Figures 3 to 6 are averages for several different beam intensities, which is, of course, a source of error. To adjust for this error, we use a modification of a technique used in neutron activation analysis. In neutron activation analysis (NAA), the most common method for trace element measurement is to run a standard reference source along with the unknown. The intensity associated with the standard peak (or peaks) is then measured against the unknown peak intensities. This method is independent of neutron flux variations and requires only that a sensitivity calibration has been made for all trace elements against the standard element (or elements). In proton activation analysis, exactly the same method can be employed, using the 23r\;a(p,n)23Mg reaction as the reference source with the slight modification that the apparent variation of the Na sensitivity with the ANALYTICAL CHEMISTRY. VOL. 45,

NO. 4,

APRIL 1973

0

695

7

1

Table II. Standard Ratios for Elements of Interest oo(x),a

standardized sensitivity cts/&/ppm

Element

- oO

10

20

30

40

F Machine calibration factor, B(q), as a function of integrated current (see text) Figure 8.

Sodium Bromine Selenium Cadmium Boron Chromium Nitrogen

set at 9.0 @ 17.0 pC 93.3 f 7.8 30.4 f 8.5 1.7 f 0.1 2.0 i 0.4 2.8 f 0.6 0.8 f 0.2

Rlxi,'

standard ratio

1.o 10.4 3.4

0.19

0.22 0.31 0.089

= o o ( x )and R f x l are defined in the text.

by the method of least squares to the data points in Figure 8. This line is given by

B ( q ) = (0.115 f 0.0020) X q

Energy, keV

Gamma ray spectrum of a typical "real" sample, ih Evnironmental Engineering Laboratory water. Integrated current was 2.0 pC. Bombard-count cycle is given in text

Figure 9. this case

beam current be taken into account. To parametrize this method precisely, the following quantities can be defined:

= counts/pC/ppm for element x for measured integrated current q u,,(x) u ( x , q o ) 3 standardized sensitivity-ie., the sensitivity at some arbitrary qo B ( 9 ) = u(Na,q)/u,(Na) = calibration factor of the machine R ( x ) = uo(x)/uo(Na) E standard ratio = ratio of a0(x) for element x to uo(Na) We emphasize further that the timing sequence is fixed for all measurements (see below), while the beam current could be (and was) varied. It will be noted that, except for the standard ratio, R ( x ) , all quantities are a function of the measured integrated current, q. Thus R ( x ) is the only quantity which is machine independent. It depends strictly on the cross-sections for producing the radioactivities in question at the proton energy of 14.7 MeV. However, all quantities, including R ( x ) , will vary, depending on the particular bombard-count cycle involved. In the present work, a standard 40-second bombard-60-second count cycle was employed, as it turned out to be about optimum for the half-lives and cross-sections pertaining to the seven elements investigated. Other timing cycles would require new calibrations, of course. An important point regarding the use of the reference method in this case is that the machine calibration factor B ( q ) should not vary too much from unity over the range of beam current utilized. This not only has the obvious advantage of making the whole method more reliable, but it also makes it possible to rely on the secondary method of directly using Figures 3 to 6 for confirmation of results. That in fact B ( q ) did stay close to unity is shown in Figure 8, which is a plot of the calibration factor, B ( q ) , for various integrated current levels. A straight line was fitted u ( x , q ) =sensitivity

696

ANALYTICAL CHEMISTRY, VOL. 45, NO. 4, APRIL 1973

+ (0.863 f 0.107)

where q is the measured integrated current (pC). Table I1 gives the measured results for the standardized sensitivities, u,(n), and the standard ratios,, R ( x ) . Using the data in Table I1 combined with a B ( q ) curve for sodium, a set of calibration curves can be created for any other comparable system from the data presented here, assuming the same bombard-count cycle and the same proton energy, 14.7 MeV. The analysis was used on several "real" samples. The results of these particular runs are given in Table 111. The results are within experimental error and the range of applicability of the curves involved. The UCD(1) and (2) samples, while from different taps, have the same well source. Determinations of the contents of these samples by other methods are also given in Table 111. The gamma spectrum representing UCD(1) and UCD(2) samples is shown in Figure 9. The presence of Na (63 ppm) is clear. Due to the type of interference shown in Figure 9, only upper limits could be placed on the following elements: Bromine Selenium Cadmium Boron Chromium

5 2.2 ppm 5 3.3 ppm 5 3.6ppm 5 1.4ppm 5 2.7 ppm.

Somewhat lower limits would characterize the Lake Tahoe samples, which are significantly freer of oxide deposits.

CONCLUSION In the preceding we have shown the feasibility of using proton activation for rapid, sensitive analysis of water samples. This technique allows measurement in the ppm range for seveal of the elements considered (bromine, selenium, sodium, cadmium, boron, and chromium) and in the tenth of a ppm range for nitrogen. These sensitivities are comparable to X-ray fluorescence sensitivities and lower than those of neutron activation and chemical methods (including atomic absorption). The sensitivity for total N is, however, superior to any other technique, and the rapidity and simplicity of the boron measurement is superior to existing techniques for boron, depending on the background-interference problem (see above) due mainly to oxygenicontaining deposits. (Chemical measurements of boron concentrations require about 10 minutes per sample.) In the work presented here, the emphasis has been on water samples. However, the method is applicable to any

Table Ill. Proton Activation Results for Unknown Samples. All Values Are Given in ppm. Total Nitrogen:

Samplea

Present method

UCD(2) Colorado River Lake Tahoe So. Lake Tahoe

50.38 f 0.20 50.50 i 0.28 50.3 f 0.2 10.3 f 0.2

EHS~

KjeldahlC

0.0f 5.0 0.0 5.0

*