Comparing plasma population kinetics codes: Review of the NLTE-3 Kinetics Workshop

Abstract

Comparison of plasma population kinetics codes, which has become an important tool for their testing and verification, was the subject of the third Non-LTE Code Comparison Workshop held at the National Institute of Standards and Technology in December 2003. The motivations for the workshop are presented here, together with its technical organization. The cases studied range from carbon to gold, and include both steady-state and time-dependent calculations. The set of results represents the current capabilities in kinetics modeling, and illustrates the effectiveness of this workshop approach to code comparisons. It was found that good agreement between codes and with experiments is observed near closed shells, but significant differences exist otherwise, or when density effects are important. On a more general level, we discuss how to meaningfully compare kinetics data. We conclude with plans for the next workshop and for a kinetics database.

Introduction

During the last decade, the advances in hardware manufacture and scientific software resulted in the development of numerical simulation codes which are becoming increasingly complex. As a result, verification and validation (VV) of such codes are now considered mandatory for establishing credibility of computational results [1]. This approach is especially important for the modeling of plasma population kinetics which includes diverse physical processes and numerous atomic states, sometimes reaching tens and hundreds of thousands. In turn, lack of non-trivial analytic results and the paucity of reliable experimental data emphasize the need for well-designed computational benchmark experiments providing the necessary testbeds for comparison of different codes.

The VV procedures for plasma population kinetic codes have been developed since the mid-1990s when the first non-LTE kinetics workshop, NLTE-1 [2], was organized in 1996. NLTE-1 adopted the format used with success in the LTE Opacity Workshops [3], namely prior selection of test cases and subsequent submission of data in well-defined formats, followed by comparison and analysis during the workshop. The substantial disagreement between codes, for the mean ion charge $\overline{Z}$—a coarse, averaged parameter—clearly demonstrated the need for such meetings. The workshop brought to light the role of autoionizing states in kinetics, and the importance of high-$n$ states in recombination. NLTE-2 [4], held in 2001 as a Virtual Workshop over the Internet, concentrated on emissivity comparisons. The large spread in the results, much greater than observed in the Opacity Workshops, motivated further investigation into the level populations. As a result, NLTE-3 was held during December 1–5, 2003, at the National Institute of Standards and Technology (NIST). As with NLTE-1, NIST provided a unique venue for fruitful discussions between about 20 researchers from 5 different countries.

In this review, we present the organization of the meeting, the set of cases studied, and the results. Some of the material has been presented in a more condensed form [5]. For conciseness the data appear here mostly in the form of graphs for $\overline{Z}$, although additional results are shown as appropriate. One of the goals of the meeting was to develop tools for meaningful code comparisons. The final section of the paper is therefore devoted to a discussion of useful code diagnostics.