Speaker
Description
Evaluated nuclear data files are expected to provide consistent data sets which represent the best knowledge of reaction observables of a specific group of nuclei. They are essential prerequisites for the development of novel technologies, such as fusion research, but also yield important input for several fields of science such as nuclear medicine, astrophysics, radiation safety, space science etc. Triggered by current developments an extension of the evaluations in incident particles ($\gamma$, p, d, $\alpha$) and energy range (up to collision energies of about 200 MeV) as well as the provision of reliable uncertainty information was requested. Especially, the extension of the energy range has an important impact on the evaluation procedure because the scarcity of experimental data at higher energies has required an increased weight of nuclear models in the evaluation process. In order to consistently combine experimental data with theoretical information from nuclear models, Bayesian statistics is well suited and is the basis of most evaluation procedures at present. Despite the worldwide efforts to develop consistent evaluation procedures in the last decade, there are still some open problems. One major difficulty is the diversity of nuclear models used in a complete evaluation as a-priori information. Especially, the different concept of R-matrix theory, suitable for the resonance region, and the statistical model and its variants applied at higher collision energies generates problems. Consequently the evaluations of observables in the two energy regions are treated independently and mutual correlations are ignored because of the different underlying concepts.
In order to perform a unique evaluation over the combined energy region we suggested the Unified Bayesian Evaluation Technique. In this statistically consistent procedure the models are considered simultaneously on the level of observables. An essential ingredient is the definition of a proper transition function which guarantees a continuous transition between the models. Thus a-priori mean values of all observables and the associated unified a-priori covariance matrix can be generated. This a-priori information is combined with available experimental data and experimental uncertainties as well as model defects via Bayesian statistics. Thus unified a-posteriori mean values and an associated covariance matrix can be generated.
In this contribution we present recent advances to make the Unified Bayesian Evaluation Technique a general working tool for the generation of evaluated data files, especially for light nuclear systems. Most important steps are the proper definition of transition functions, the determination of an independent set of observables, the proper choice of covariance functions [1,2] for the description of model defects on the basis of Gaussian processes and the realisation of a large scale evaluation procedure on the basis of the modified GLS method [3]. The advances in each of these steps will be presented and its feasibility demonstrated by application to realistic examples. Finally an outlook on further developments of the Unified Bayesian Evaluation Technique will be given.
Acknowledgement: The work has received funding from the Euratom Project SANDA under grant agreement No 847552. The views and opinions expressed herein do not necessarily reflect those of the European Commission.
References:
[1] G. Schnabel, H. Leeb, EPJ Web of Conferences 111, 09001 (2016).
[2] H. Leeb, Th. Srdinko, EPJ Web of Conferences 211, 07006 (2019).
[3] G. Schnabel, H. Leeb, Nucl. Inst. Meth. A 841, 87 (2017).