Real-time analysis of radiation spectra by iterative bayesian estimation. Application to elemental analysis by X-ray spectra

In conventional spectrum analysis in a PHA system, system build-in program usually applys to the accumulated histogram data for a certain time after the analogue-to-digital conversion. Hence, this type of analysis is inherently made in the batch data processing mode, not real time mode. A noble and quite simple method has been recently developed for the radiation spectrum unfolding problems by one of the present authors. This method is only on the basis of the Bayes' theorem, and has the following features: the spectrum inference is made in the iterative fashion as getting the pulse height datum for each count by a detector; hence it enables us to estimate the energy spectrum in true 'real time' mode in principle; it provides intrinsically non-negative estimates because the estimated values are treated as the probability, and almost non-oscillating estimates because the estimation implicitly satisfies the principle of the maximum information entropy; the estimate is free from the numerical ill-condition which frequently appears in the matrix inversion type methods, e.g., the least squares fitting method. In this paper, we apply the Bayesian method to the problem of the elemental analysis using spectra of the characteristic X-rays induced by bombarding the energetic ions from accelerators, called PIXE (particle induced X-ray emission). Benefit of the real-time analysis in PIXE is very great because it provides approximate values of the elemental concentration during the measurement as the time dependent data; this means that the analysis method enables us to correct the effect of gradual decreasing of yields of a few volatile elements in the irradiated sample if exist; it can also notify us a brake of backing/sample foil due to the ion beam bombardment by sudden decrease of the rate of all X-ray yields. We are also able to adjust the data collection parameters or various experimental parameters during the measurement, such as collection time, beam intensity, and suitable X-ray filter selection (see 3.1), etc. in accordance with the sample condition. This adaptability is quite beneficial to realize the automatization of the analysis which should be adopted in the 'industrial PIXE'. (author)