Sensitivity analysis of random two-body interactions

The input to the configuration-interaction shell model includes many dozens or hundreds of independent two-body matrix elements. Previous studies have shown that when fitting to experimental low-lying spectra, the greatest sensitivity is to only a few linear combinations of matrix elements. Here we consider interactions drawn from the two-body random ensemble and find that the low-lying spectra are also most sensitive to only a few linear combinations of two-body matrix elements, in a fashion nearly indistinguishable from an interaction empirically fit to data. We find in particular the spectra for both random and empirical interactions are sensitive to similar matrix elements, which we analyze using monopole and contact interactions.