METHODS FOR SAMPLING AND
INORGANIC ANALYSIS OF COAL
U.S. Geological Survey
Bulletin 1823
Edited by D.W. Golightly and F.O. Simon
THE DETERMINATION OF 41 ELEMENTS IN WHOLE COAL BY INSTRUMENTAL NEUTRON ACTIVATION ANALYSIS
By C.A. Palmer and P.A. Baedecker
Forty-one elements have been determined in coal by a combination of short and long irradiations using instrumental neutron activation analysis (INAA). The factors that lead to errors in analysis, such as spectral overlaps, low sensitivity, and multiple sources of the indicator radionuclide, are discussed. Detection limits for the elements determinable by INAA are given and data for National Bureau of Standards coal and fly ash standards are presented.
Instrumental neutron activation analysis (INAA) is a versatile technique for elemental analysis because it has very low detection limits for many elements, lends itself to automation, and provides precise data for many major, minor, and trace elements. The application of INAA to the analysis of coal has been described in a number of publications (e.g., Block and Dams, 1973; Ondov and others, 1975; Rowe and Steinnes, 1977a, b; Swaine, 1985). For analyses based on the measurement of long-lived nuclides (t1/2 > 1 day), 30 elements can be determined routinely in most coal samples. Two additional elements can be determined in certain coals. Extension of the technique to the measurement of short-lived activities, with rapid sample transfer and short irradiation and counting times, makes possible the determination of an additional nine elements.
Activation analysis is based on the measurement of the activity from radionuclides that are produced by nuclear reactions on naturally occurring isotopes of the element to be determined in the sample. Reactor neutrons are most commonly used for inducing the nuclear transformations because of their availability, their relatively high probability for neutron-induced reactions, and their relative freedom from problems due to matrix effects (self-shielding). For example, the determination of arsenic is carried out by the following reaction:
where the amount of arsenic in the sample is determined by counting the induced 76As activity. The activity of the indicator radionuclide produced during the irradiation is directly proportional to the amount of element of interest in the sample, and the analytical determination is generally made by comparing the activity induced in the sample against the activity measured for well-characterized standard samples. The activities of the samples and standards are most commonly measured by gamma-ray spectroscopy because potentially interfering activities can generally be discriminated against by looking at gamma-rays having unique energies for the indicator radionuclide and because gamma counting is relatively free from matrix effects (self-absorption). Semiconductor detectors, such as high-purity germanium and lithium-drifted germanium (Ge(Li)) diodes, are generally used for gamma-ray spectroscopy because of their excellent resolution, which permits the separation of closely spaced lines. These devices convert the gamma-ray signal for the irradiated samples to electrical signals that can be sorted according to amplitude by an analog-to-digital converter; the pulses within each amplitude interval are counted by a multichannel scaler, with each channel corresponding to a given interval of gamma energy. A typical gamma-ray spectrum of an irradiated coal sample, measured 5 days after irradiation, is shown in figure 11. The radioactivity in the sample is thus measured by determining the area of a specific gamma photopeak of interest above an underlying background continuum.
If only relatively long-lived (t1/2 > 12 h) indicator radionuclides are employed in the analysis, up to 32 elements can be determined using relatively long irradiations and two sample counts following irradiation. The elements, their indicator radionuclides, half-lives, gamma-ray lines, detection limits, and potential spectral interferences are listed in table 7. The estimation of detection limits for INAA is subject to considerable uncertainty. This uncertainty exists because the estimate is based on the signal-to-background ratio for each photopeak of each sample being counted and the Ge(Li) detector employed, where the Compton continuum from higher-energy gamma-rays contributes to the background. The detection limits are therefore dependent on sample composition. These detection limits, presented in table 7, are for coal samples having concentrations of Na, Sc, Fe, Co, and La similar to National Bureau of Standards (NBS) standard coal 1632, because the gamma-rays from the activation products of these elements dominate the spectrum of most activated coal samples and limit the sensitivities for the determinations of other elements.
The pertinent information for the short-lived species that can be measured for the determination of nine additional elements is listed in table 8.
Those species with half-lives of less than 1 h are generally measured after irradiation for 5 min at a flux of 3 ´ 1012 units in the pneumatic tube system of the U.S. Geological Survey 1 MW TRIGA reactor in Denver, Colo., and counted after approximately 10 min of decay. Manganese and dysprosium are measured after a 2 h irradiation and approximately 3 h of decay.
EXPERIMENTAL METHODS AND PROCEDURES
The TRIGA research reactor facility of the U.S. Geological Survey is utilized for most of the INAA work conducted in our laboratories. Powdered samples, weighing 500 mg, are heat sealed in 1.5 cm3 polyethylene vials. Eighty samples, including standards, can be irradiated in the "lazy susan" facility of the reactor for 8 h at a flux of 6 ´ 1012 n/cm2s. The samples are packaged in 7.4 cm3 snap-top polyethylene vials and inserted into a single TRIGA tube in the "lazy susan." Because of a modest flux variation along the tube, the 40 samples and standards on each level are treated as separate sample sets during the subsequent counting and data processing procedures.
Samples, as small as 2 mg, can be analyzed after irradiation in the U.S. Geological Survey TRIGA reactor by increasing the irradiation time to 16 or 24 h. The sensitivity is reduced for the determination of several elements in samples weighing less than 200 mg, and less than 20 elements can be detected for most samples weighing less than 10 mg.
Three multiple-element standards, or flux monitors, are currently used for the routine analysis of coal: NBS 1632 (coal) and 1633 (fly ash) and Eastman-Kodak gelatin multicomponent reference material (TEG-50-B). The concentration values used for NBS 1632 and 1633 are those of Ondov and others (1975). Standard TEG-50-B has been calibrated against single-element standards for As, Co, Cu, Hg, Sb, Se, and Zn. In addition, NBS standard reference materials 1632a and 1633a have been characterized in the U.S. Geological Survey laboratories for use as standards. Data for all five standard samples are listed in table 9. Antimony has been observed to be inhomogeneous in NBS 1632 (Ondov and others, 1975) and has a low concentration in NBS 1633. Thus, TEG-50-B is used as a standard for antimony. For short irradiations, these standards are supplemented with analytical-grade sulfur and spectroscopic grade CaC0Recently, 3 and MgO. work has begun on a new U.S. Geological Survey standard that will consist of a spiked coal from the Lower Bakerstown coal bed. After testing of this standard is complete, the NBS standards will probably be used only for quality-control purposes.
Following an 8-h irradiation in the U.S. Geological Survey reactor in Denver, Colo., and the subsequent four days of transcontinental shipping of the samples, all samples are placed in 7.4 cm3 polyethylene vials. These vials serve as transfer containers for the automatic sample changers used with the high-resolution coaxial Ge(Li) detectors for gamma-ray spectroscopy. The detectors are coupled to multichannel pulse-height analyzers, which are capable of dividing the spectrum into 4,096 energy increments or channels. The analyzers automatically repeat a cycle of data acquisition, sample changing, and read-out of the spectral data to disk storage. The automatic sample changer used with each detector is mechanically identical to that described by Massoni and others (1973), but the electronic interface has been completely redesigned to employ solid-state circuitry. Eight sample changers are controlled by a single interface. At the end of each preset counting period, the multichannel analyzer signals the controller to change samples. The sample changer is a gravity feed device that uses compressed air to eject a sample from the counting station after counting is completed.
All spectra are processed using the computer program SPECTRA (Baedecker, 1976, 1980; Grossman and Baedecker, 1986). The program contains algorithms for smoothing the spectral data; searching out all peaks in the spectrum; determining the areas of both well-resolved, single peaks and unresolved (or overlapping) multiple peaks; and determining the energies of the photopeaks (including corrections for zero and gain drift of the spectrometer). The calculations of elemental concentrations include corrections for spectral interferences, decay, and neutron-flux variations, as well as pulse pileup effects. When a gamma-ray line has been specified for use in the analysis and has been observed in a standard material but not in a sample, the program computes an upper limit to the concentration of the element in question. The peak area used for the calculation is 10 times the standard deviation of the back-ground within the normal integration limits at the expected peak location in the spectrum. The program corrects for overlapping spectral lines that are too poorly resolved to be recognized as multiplets when an interference-free line of the interfering radionuclide can be observed in the spectrum and used for correction. For example, the area of the 264.6-keV line of 75Se can be corrected for interference from the 264.1-keV line of 182Ta. This correction is accomplished by using the peak area of the 1221.3-keV line of 182Ta and the known ratio of the intensity of the 264.1-keV line to the intensity of the 1221.3-keV line.
Thus, elemental concentration data, based on each specified peak, are stored in a disk file that is generated during execution of the SPECTRA program. A second program, SUMMARY1, is executed after the successful execution of SPECTRA, to average the results from multiple lines for a given element and to generate a report of analysis for a single counting of a sample set. Peaks that are averaged are specified to the program in order of decreasing priority. Results from lower-priority peaks are included in the reported result only if they agree within 2 standard deviations of the mean based on the peak(s) given higher priority. Each result included in the mean is weighted by its estimated standard deviation based on counting statistics. The weighted average results and the estimated standard deviations for each element are stored in a disk file for long-term storage.
Following the completion of all counts on each set of samples during a 2-month decay period, a third program, SUMMARY2, is executed to average the results from multiple counts and to generate a report of analysis. Again, the mean value for each element is a weighted mean based on the estimated counting error. If uranium has been determined via the 2.35-day 239Np, SUMMARY2 corrects for fission product interferences in the determination of Zr, Mo, La, Ce, Nd, and Sm using the correction factors listed in table 10. Where the magnitude of the correction exceeds 20 percent of the value reported, the value is flagged in the report of analysis. The calculation of the fission product interference for lanthanum is complicated by the fact that the fission product 140Ba has a longer half-life (12.7 days) than its daughter product, the indicator radionuclide 140La. We have not measured the fission product interference for lanthanum at the normal decay time of 6 days for coal samples. We use the calculated value of 0.02 part per million of lanthanum per part per million of uranium for computing the fission product interference, which is generally less than 1 percent for most coal samples. The experimental values for other elements listed in table 10 are not decay dependent and were measured at a decay time of 10 days after irradiation. Because the experimental values agree with the calculated values to within 25 percent and the fission product interferences for most coals are below 2 percent, small errors in the interference-correction factor result in negligible errors in the data.
An estimate of the precision for each sample set is made by using replicate samples to determine the reproducibility of the measurements. Several replicate samples are generally determined within each sample set. In addition, many elements are determined by two or more counting cycles.
Agreement in both cases is usually within errors determined by counting statistics. When this is not the case, the results are flagged in the data reduction procedures using programs described by Grossman and Baedecker (1986). An estimate of the accuracy of the data is obtained by running control standards that can then be compared to "best values" determined from the literature or from previously determined averages. Reference materials 1632a and 1633a are most commonly used as control standards. "Best values" for these are given in table 9. This procedure also allows for an estimate of the precision of the data between different sample sets. Computerized procedures (including graphic presentation of the quality-control data) for comparing measured values with "best values" were also descrlbed by Grossman and Baedecker (1986).
Up to 41 elements can be determined in coal and fly ash by a combination of short and long irradiations for INAA. Except for dysprosium, all the elements determined using short-lived indicator radionuclides are light or first-row transition elements, which can be better determined by other multiple-element instrumental techniques. The use of short irradiations and counting times is primarily useful for the analysis of small samples where it is desirable to get as complete an analysis as is possible on a single sample (e.g., mineral, size, or density separates). Of the elements with half-lives of less than 1 h, the lines from the activation products of Mg, Al, and V are most readily observed in the spectra obtained within 10 min of irradiation. The remaining four elements listed in table 8 (S, Cl, Ca, and Ti) are determinable with a detectability poorer than that for Mg, Al, and V.
The following discussions provide more detailed evaluations of the problems associated with the determination of specific elements.
Sulfur is determined in the short counting cycle, 5 min after irradiation. It is determined from the reaction
The determination of sulfur is subject to error due to natural variation in the isotopic abundance of 36S. A maximum error of ±29 percent has been calculated, but errors from +4.4 to -6.5 percent have been observed for results on natural material when compared with a standard with d 34S = 0. Because of the low natural abundance of 36S (0.01 percent) and its low cross section (0.14 b), the determination of sulfur has poor sensitivity, and errors from counting statistics are likely to be much higher than errors in isotopic composition for most moderate- to low-sulfur coals.
Although the concentration of magnesium is moderately high in most coal samples, it is relatively difficult to measure, because 27Mg is produced in the reactor from two sources: from the (h ,g ) reaction or 26Mg and from the (h ,r ) reaction or 27Al. Because most coal samples having high magnesium concentrations commonly have even higher aluminum concentrations, large corrections must be made. The correction, which depends on the neutron spectrum for a given reactor, is about 35 percent for NBS 1632 irradiated in the pneunmatic tube of the U.S. Geological Survey Denver reactor.
The accuracy for the determination of gallium is greatly dependent on concentrations of iron and the speed with which the sample is counted. The major photopeak of 31Ga (834 keV) has an interference from the 835-keV photopeak of 54Mn that is formed from the (h ,r ) reaction of 54Fe. Although the 54Fe peak is relatively small for most coal samples, the half-life of gallium is only 14 h. For samples having high concentrations of iron or a large counting delay, this interference can be a major problem. For NBS 1632, the relative error is 1.2 percent after 3 days, but the error increases to 70 percent after 6 days.
The determination of chromium suffers from potential spectral interference from 177Lu and 147Nd. For NBS 1632, the corrections for these interferences amount to £ 1 percent, but they can be higher in coals with lower concentrations of chromium or higher contents of the lanthanide elements.
The 1,115-keV photopeak of 65Zn falls on the low-energy tail from the 1,120-keV photopeak from 46Sc, which complicates the evaluation of the base area for the photopeak. This 65Zn photopeak is normally treated by the computer algorithm as part of a triplet along with the 1,112-keV photopeak of 152Eu. Also, a small interference exists from 160Tb, which amounts to 1 percent zinc in NBS 1632.
The 264.6-keV photopeak from selenium is best measured in coal samples in a count taken from 3 weeks to 2 months after irradiation. A correction must generally be made for an interference from 182Ta. The correction is 8 percent of the measured selenium in NBS 1632.
Zirconium is difficult to determine by INAA due to spectral interferences from 152Eu, 154Eu, and 160Tb, which interfere with the lines from 92Zr and its daughter product, 35.1-day 95Nb. For NBS 1632, the correction on the 95Nb 765.8-keV line is 10 percent. Also, a correction is necessary for the fission product interference based on the measured uranium content, which for fly ash NBS 1632 amounts to 45 percent. For these reasons, the determination of zirconium is semiquantitative at best.
Molybdenum, like zirconium, falls near the low-mass maximum of the fission yield curve of 235U, and a large fission-interference correction is generally required. The interference correction would be roughly 46 percent of the value for NBS 1632 and 1633.
The determination of barium is based on the measurement of the most intense 131Ba line at 496.3 keV in the Ge(Li) spectrum taken within a week of irradiation. There is a potential spectral interference from fission product 103Ru, which cannot be corrected because of the absence of any other interference-free line from 103Ru of sufficient intensity. An estimate of the interference calculated from fundamental parameters and confirmed by experiment is 2.9 ´ e0.0402 t (apparent parts per million of barium per part per million of uranium), where t is the time after irradiation in days. For NBS 1632, 103Ru contributes 1.5 percent to the intensity of the barium photopeak seven days after irradiation, and for NBS 1633, it contributes 1.7 percent.
The determination of lanthanum also can be affected by fission product interferences. The measurement is complicated by the relatively long lived, 12.8-day 140Ba, which is a fission product precursor to the indicator radionuclide 140La. Four days after irradiation, the interference from 140Ba is 0.0084 (apparent parts per million of lanthanum per part per million of uranium), or a 0.1-percent correction for NBS 1632. For later counts, the correction will increase; at a counting time of 10 days, the interference is 0.092, or a 1.2-percent correction for NBS 1632.
The determination of terbium is made difficult by spectral interferences. The most intense line in the Ge(Li) spectrum, at 298.6 keV, occasionally suffers from interferences from the 233Pa 299.9-keV line. The computer program must first check to see if both lines have been detected and resolved by the multiplet analysis algorithm; if not, the program then applies a correction for the interference, which for NBS 1632 will be 50 percent, at a decay time of two months. The 879.4-keV line is also observed in most spectra as a weak line on the low-energy tail of the generally intense 146Sc 889.3-keV line, which renders the base area difficult to evaluate. The 965.8-keV line may suffer from interference from 152Eu, and while the 1,177.9-keV line is free from interferences, it has much poorer sensitivity.
Photopeaks from 101-h 175Yb and 32-day 169Yb are normally detected in Ge(Li) spectra of activated coal. Most photopeaks from both nuclides incur spectral interferences, although ytterbium is generally well determined using the 396.1-keV line of the shorter-lived isotope. There are minor interferences from 147Nd and 233Pa; the correction is 3 percent for NBS 1632.
Mercury can be determined in many coal samples by measuring the intensity of the 279.2-keV line of the 46.6-day 203Hg. A substantial spectral-interference correction is generally required due to 75Se (60 percent for NBS 1632). Because the 175Se 264.6-keV line, which is used to make the correction, suffers from a generally small interference from 182Ta, the algorithm must treat the 264.6-keV line before computing a mercury content from the 279.2-keV photopeak.
Rowe and Steinnes (1977a, b) have described the application of epithermal neutron irradiations to the analysis of coal. Many of the spectral and fission product interferences discussed in this section can be reduced and greater sensitivity realized by this technique, particularly for the determination of Ni, Mo, Rb, Sr, Cs, Tb, Ta, W, and U. However, the technique is limited to small (~50 mg) samples and requires special packaging in aluminum foil and irradiation inside cadmium capsules. Because of the added cost and labor involved in the special handling of samples for epithermal neutron activation analysis (and increased radiological hazard of handling radioactive coal powders), the technique is not suitable for routine application but is useful for special problems.
Baedecker, P.A., 1976, SPECTRA: computer reduction of gamma-ray spectroscopic data, in Taylor, R.E., ed., Advances in obsidian glass studies: Park Ridge, New Jersey, Noyes Press, p. 334-349.
_______ 1980, Comparisons of peak-search and photopeak-integration methods in the computer analysis of gamma-ray spectra, in Vogt, J.R., ed., International Conference on Nuclear Methods in Environmental and Energy Research, 4th, 1980, Proceedings: U.S. Department of Energy CONF-800433, p. 15-24.
Block, C., and Dams, R., 1973, Determination of trace elements in coal by instrumental neutron activation analysis: Analytica Chimica Acta, v. 68, no. 1, p. 11-24.
Grossman, J.N., and Baedecker, P.A., 1986, Computer graphics for quality control in the INAA of geological samples, Proceedings: Modern Trends in Activation Analysis, Copenhagen, Denmark, June 23-27, 1986, p. 571 -578 .
Massoni, C.J., Fones, R.V., and Simon, F.O., 1973, A pneumatic sample changer for gamma-ray spectroscopy: Review of Scientific Instruments, v. 44, no. 9, p. 1350-1352.
Ondov, J.M., Zoller, W.H., Olmez, llhan, Aras, N.K., Gordon, G.E., Rancitelli, L.A., Abel, K.H., Filby, R.H., Shah, K.R., and Ragaini, R.C., 1975, Elemental concentrations in the National Bureau of Standards' environmental coal and fly ash standard reference materials: Analytical Chemistry, v. 47, no. 7, p. 1102-1109.
Rowe, J.J., and Steinnes, E., 1977a, Determination of 30 elements in coal and fly ash by thermal and epithermal neutron-activation analysis: Talanta, v. 24, p. 433-439.
_______ 1977b, Instrumental activation analysis of coal and fly ash with thermal and epithermal neutrons: Journal of Radioanalytical Chemistry, v. 37, no. 2, p. 849-856.
Swaine, D.J., 1985, Modern methods in bituminous coal analysis: Trace elements: CRC Critical Reviews in Analytical Chemistry, v. 15, no. 4, p. 315-346.
[ Contents | Coal Quality CD ]
Created by the EERT WWW Staff.
[an error occurred while processing this directive]