EVALUATION OF STATISTICAL TESTS FOR SUBSTRUCTURE IN CLUSTERS OF GALAXIES

We use N-body simulations of galaxy cluster mergers to evaluate new an d published statistical tests of substructure. These tests include 22 one-dimensional (normality) tests, four two-dimensional (spatial) test s, and five three-dimensional (velocity-spatial) tests. These tests ar e statistical in that they provide a significance level for the presen ce of substructure. All the tests are applied to the same data files s o that their relative sensitivity can be compared. The data files cont ain positions and velocities of dark matter particles drawn randomly f rom the simulations. Three noncosmological simulations are run in whic h the subclusters begin as King spheres: a single isothermal cluster, a merger of a 1/6 mass subcluster, and a merger of a 1/3 mass subclust er. In this way, we examine the dependence of the tests on the subclus ter's relative mass. We examine also the dependence on the total sampl e size, the epoch of merger (pre- and post-core-crossing), and the pro jection angle of the merger axis. The results allow a quantitative com parison of the effectiveness of each estimator under different observa tional scenarios. In general, the higher the dimensionality of the tes t, the more sensitive it is to substructure. The sensitivity of indivi dual diagnostics depends on the line of sight relative to the merger a xis. The three-dimensional tests are least sensitive to lines of sight perpendicular to the merger axis and are most sensitive to lines of s ight 45 degrees-60 degrees from perpendicular. The two-dimensional tes ts are most sensitive to lines of sight perpendicular to the merger ax is. The one-dimensional tests are the most sensitive to lines of sight parallel to the merger axis. No single substructure test is the most sensitive in all situations. Therefore, we recommend that a battery of tests be applied to each cluster. We provide a score for each test re flecting its relative sensitivity. We find that clusters with no mergi ng components but a velocity dispersion gradient have an increased lik elihood for a ''false-positive'' response from some three-dimensional tests.We examine also the signatures of merger and the detectability o f mergers in redshift surveys. We find that cluster masses are overest imated by up to a factor of 2 for clusters undergoing mergers. We plot the dependence of mass estimators on projection angle and epoch of me rger. We find that the detectability of postmerger states is hampered by small search radii (e.g., <2.0 h(75)(-1) Mpc) in redshift surveys.