The study presented in this report was initiated at the request of the NEA Nuclear Science Committee as part of the programme of work of the Agency in the area of Partitioning and Transmutation of radioactive nuclear waste. The aim of the exercise was to determine the predictive power of current nuclear reaction models and codes when calculating activation yields in the intermediate energy range (up to 5000 MeV). The results of the study help assess the needs for improvements in the nuclear reaction codes which would be used in calculating nuclear reaction processes for the design of various accelerator-based transmutation concepts. This exercise is the third one in a series. It was preceded by intercomparisons of calculations of double differential spectra in thin targets and of particle production and transport in thick targets.

Emphasis has been placed on evaluating the quality of calculated activation yields for a wide range of target elements (O, Al, Fe, Co, Zr, and Au) by comparison with high-quality experimental data. Calculated results from 29 contributions by 18 participants or participating collaborations have been compared with a total of nearly 6000 experimental cross sections for 202 target/product combinations. 22 different models or codes have been applied giving a representative survey on today's modeling capabilities. Most major codes participated in the exercise.

This report gives detailed information about the different models and codes used and surveys extensively the target element and product nuclide coverage of the different contributions. The largest part of the report is dedicated to the graphical representation of the results. Comparative plots demonstrate significant differences between reaction cross sections calculated by the different models and codes. However, these differences cannot account for the partially extreme differences among the calculations for individual target/product combinations. Such reaction-wise comparisons of calculated and experimental activation yields as function of proton energy make up the main body of this report. They give detailed information about the predictive power of models and codes when calculating cross sections for the production of residual nuclides from thresholds up to 5000 MeV.

The agreement or disagreement between experimental and calculated data is quantitatively described by factors of deviation which were calculated for each contribution reaction-wise and globally. From these deviation factors and from the entire exercise we may conclude that modeling calculations of intermediate energy activation yields on a predictive basis may at best have uncertainties of the order of a factor of two. Frequently, average deviations are much lager and individual reaction-wise deviations may go up to two or three orders of magnitude. There are no general over- or underestimates by individual models or codes, but rather a broad scatter of calculated data. Occasionally, the calculations are contradictory among each other by up to 3 orders of magnitude for a given reaction.

Problems are encountered which are connected with the calculation of nuclear masses, binding energies and consequently Q-values, with the consideration of shell effects and the various level density formulas used, with the neglect of competition between g- and particle deexcitation of excited intermediate nuclei, and , last but not least, with the basic modeling of medium energy fission and Fermi break-up. The causes of the individual deviations are multi-factorial and can - for a given model, code or contribution - only be evaluated by model and parameter exercises for a wide range of reactions. The present intercomparison gives a first survey over these problems. It should be understood by the modelers and code developers as a starting point for the improvement of models and codes.