How Accurate Are Our Processed ENDF Cross Sections?

Let me start by reassuring you that currently our nuclear data processing codes are very accurate in the calculations that they perform INSIDE COMNPUTERS. However, most of them drop the ball in what should be a trivial final step to output their results into the ENDF format. This is obviously a very important step, because without accurately outputting their results we would not be able to confidently use their results in our applications. This is indeed a very important step, but unfortunately it is one that is not given the attention it deserves; hence we come to the purpose of this paper. Here I document first the state of a number of nuclear data processing codes as of February 2012, when this comparison began, and then the current state, November 2013, of the same codes. I have delayed publishing results until now to give participants time to distribute updated codes and data. The codes compared include, in alphabetical order: AMPX, NJOY, PREPRO, and SAMMY/SAMRML. During this time we have seen considerable improvement in output results, and as a direct result of this study we now have four codes that produce high precision results, but this is still a long way from ensuring that all codes that handle nuclear data are maintaining the accuracy that we require today. In the first part of this report I consider the precision of our tabulated energies; here we see obvious flaws when less-precise output is used. In the second part I consider the precision of our cross sections; here we see more subtle flaws. The important point to stress is that once these flaws are recognized it is relatively easy to eliminate them and produce high precision energies and cross sections.