Instrumental Activation Analysis Using High-Resolution 7-Spectrometry and Automated Data Analysis Paul Biloen, Jan Dorrepaal, and Herman B. van der Heijde KoninklijkelShell-Laboratorium,Amsterdam, The Netherlands A nondestructive method of instrumental activation analysis using a Ge(Li) detector is described, whose procedure is independent of sample type. It produces an estimated precision for any one element detected and calculates maximum concentrations of the undetected elements. I n an average sample, six to nine concentrations and approximately thirty maximum concentrations are determined simultaneously. Only a few criteria are sufficient for a fully automated data analysis, which includes screening for systematic errors. The performance of this method of analysis can be accurately predicted from detector characteristics and isotope constants.
SEVERAL AUTHORS (1-3) have recently reported on promising systems for nondestructive instrumental activation analysis (IAA). With the aid of high-resolution germanium detectors they have succeeded in making multielement analyses without the use of radiochemical separations. However, the purely instrumental method has its limitations (3-9, and it is commonly recognized that the precisions and sensitivities that can be attained vary with the overall sample composition. A central, multifunctional, research laboratory like ours requires element analyses of a wide variety of materials. These range from, say, novel catalyst compositions tried out in our exploratory research groups to corrosion products sent in for identification by our operating companies in many parts of the world. Availability of general techniques of survey analysis is therefore important, even if they give information which is incomplete or less precise than more traditional methods. When such methods are applied to new materials, however, the precision and sensitivity should be manifest. Therefore, in the IAA system described, the emphasis has been laid on the evaluation of detection limits and precisions. In addition, we present an analysis of the performance of the method in terms of fundamental parameters, such as detector characteristics and isotope constants. INSTRUMENTATION
Samples are irradiated in the pool of a light-water-moderated tank-in-pool type nuclear reactor : the H F R reactor (6) of The Netherlands Reactor Center. The neutron flux at the irradiation site is approximately 4 x 1012 neutrons. cm-2. sec-1. The minimum delay time between irradiation and measurement a t our laboratory is about 90 minutes. The components of the spectrometer system-listed in Table I-have been selected on the following criteria: high count rate capability, sufficiently large number of data channels, and availability of a n interface between the analyzer and ( 1 ) 0. U. Anders, ANAL. CHEM., 41,428 (1969). ( 2 ) R . Dams. J. A. Robbins, K. A. Rahn, and J. W. Winchester, ibid., 42, 861 (1970). (3) J. Turkstra. P. J. Pretorius. and W. J. de Wet, ibid., P 835. (4j R. f . Coleman, Proceedings, 1968 International Conference on Modern Trends In Activation Analysis, NBS Spec. Pirbl., 312, Vol. 11, p 1262. (5) F. Girardi, Proceedings, 1968 International Conference on Modern Trends in Activation Analysis, . , N E S SDec. Publ., 312, VOl. I, p 577. (6) IAEA-Directory of Nuclear Reuctiotw, 3, 173 (1960). 288
Table I. Detector
Components of the Gamma Spectrometer : True coaxial; 30 cm3 Energy resolution at 1.33 MeV: 3.6 keV Peak-to-Compton ratio: 11: 1 Relative efficiency at 1.33 MeV: 4% Capacitance: 30 pF Manufactured by Ortec in 1968 Preamplifier : Ortec 118 A : Ortec 450 Amplier Pile-up rejection: Ortec 442 Analyser : Nuclear Chicago 25603 4096 channels 100 Mc-ADC Magnetic tape : Ampex T7., continuous drive IBM/System 360 compatible Control characters : generated in interface Interface : Manufactured by Nuclear Chicago
the computer-compatible magnetic tape. Input pulse rates up to 150,000 p/sec, corresponding to count rates of approximately 50,000 c/sec, can be processed without any serious spectrum degradation. The detector/preamplifier combination is now outdated, especially with regard to its peak-to-Compton ratio. The data tapes are processed batchwise on a SBS-Sigma 7 computer. Software information is entered on the tape via the first few channels of the multichannel analyzer. The processing of a single spectrum requires approximately 35 seconds computer time. We use a core memory of approximately 30 K, and 20 K disk locations. GENERAL OUTLINE OF PROCEDURE
The irradiation is performed batchwise. Each batch comprises sixteen samples and five flux monitors, which are packed in a 500-ml polyethylene bottle, separated by polyethylene spacers. The samples are small: approximately 15 mg. The induced gamma activity is integrally recorded in the 100-2000 keV energy region. The individual samples are routinely measured after two or three different cooling intervals. In the computerized data analysis, the gamma spectrum is inspected for individual peaks, Isotopes are identified by peak position. From peak areas isotopic, and from them elemental, abundances are calculated. Upper limits for concentrations of undetected elements are calculated on the basis of inspection of the background at positions where, according to the reference data, peaks should be located. For reference we use a permanent library of data organized isotope-wise. Reference peak areas have, for a number of isotopes. been obtained from measurements on standard materials, for the remaining ones they have been calculatedfor our particular detector and detection geometry-from data compilations of Hollander (7) and Dams and Adams (8). ( 7 ) C. M. Lederer, J. M. Hollander, and I. Perlman, “Tables of
Isotopes,” Wiley, New York, N.Y., 1968. (8) R. Darns and F. Adams, Radiochim. Acta, 10, 1 (1968).
ANALYTICAL CHEMISTRY, VOL. 45, NO. 2, FEBRUARY 1973
The neutron flux is monitored with lanthanum; we use the 1596-keV emission of l40La. The entire batch is subjected to a uniaxial rotation of approximately one cycle per second throughout the irradiation. The times of irradiation and cooling-both of the order of several hours-are measured externally. Following a procedure described by Anders ( I ) , we monitor the “live” counting time by feeding pulses from a mercury-relayed pulser to the input of the preamplifier. The pulses are counted as a separate peak on the highenergy side of the spectrum. We use an “end-on” sample detector geometry; the sample can be placed at three different, well-defined, distances from the detector front end. GENERAL FEATURES OF RESULTS
To our experience, the limitations are mainly due to peakbackground interference. The Compton backgrounds of isotopes with strong activity hamper or impair the measurement of those of weak activity having lower gamma energy and half-life. Consequently, the limit of detection or precision that can be attained for a given element varies greatly with the overall sample composition. This is in line with the findings of other authors (3-5). From analyses of widely different materials, such as ceramics, catalysts, and steels, we found as average results: (1) Number of elements detected: 6 to 9. Precisions varying from 2 to 20% relative. ( 2 ) Number of undetected elements with a calculated maximum concentration W,,, < 1 wt: approximately 30, with W,,, < 0.01 % wt: approximately 15. These results have been obtained with fixed irradiation times of 2 hours and measurements after three fixed cooling times (90 minutes, 30 hours, 100 hours) with counting intervals of approximately 5, 15, and 45 minutes, respectively. The procedure is independent of sample type. The computer output of two such routine analyses is shown in Tables 11, A and B. The data analysis is fully automated, but sample handling is not yet mechanized. At the present stage of development, a complete analysis takes approximately one and a half manhours. PRECISION
From our first results, we gained the impression that there were three main causes of imprecision. Counting statistics dominated in approximately 60% of all results. For strong peaks in “hot” samples, the uncertainty in live-time measuring was observed to be the important parameter. In the absence of these two causes, the uncertainty in neutron flux for the individual samples in the batch was dominant. We currently calculate the peak areas and their standard deviation due to counting statistics by the Covell method (9). Similarly, the standard deviation in live-time measuring is calculated from the pulser peak. The uncertainty in neutron flux is calculated from the dispersion of the activity of the five flux monitors irradiated in one batch. To this end, three of the five flux monitors are used for monitoring of the residual flux inhomogeneity in a plane perpendicular to the rotation axis. The other two are located at fixed vertical distances from this plane. Their activity is corrected for a flux gradient (characteristic of the irradiation site) of 673crn. Any remaining dispersion in activity is then regarded as an uncertainty in the axial neutron flux gradient. Current flux uncertainties move around 1 % relative.
Table IIA. Computer Output of the Routine Analysis of a Co-Mo-A120a Catalyst Element Conc % wt Element Conc % wt 17 C1