Monte Carlo Markov chain method for analyzing nuclear reaction spectra

The precision with which reaction yields can be determined is often limited by the difficulty of extracting the relevant peaks from large backgrounds which may arise either from competing reaction processes or unknown and physically uninteresting phenomena. This difficulty is compounded when, as is often the case, the statistical distribution of counts attributable to the background or to the reaction peaks cannot be assigned a particular analytic or parametric form. It may also be the case that only weak physical arguments constrain the peak shapes, so that the use of a rigidly parametric analysis method would be grossly misleading. In an attempt to confront these challenging issues in data reduction, a sampling-based method for resolving complex reaction spectra into its various constituents has been developed. Employing Monte Carlo Markov Chain Methods within the context of the Bayesian statistical paradigm, the system can be run either in parametric or non-parametric mode. In the non-parametric mode, minimal constraints on the peak shapes are derived from general considerations of statistical processes which contribute to the topology of the peak shape. Probabilistic constraints, which serve to bias the solution toward a particular shape or distribution can also be imposed. Several versions of this algorithm have been coded in Fortran and used to analyze the spectra arising from protons scattered through time deuterated carbon targets at 315 keV. All explanation of the algorithm, discussion of problems and future research directions, and samples of analyzed spectra will be presented