The atomic limit of the Hubbard model is a simple single-site problem which can be solved exactly, and all one- and two-particle Green’s functions can be obtained analytically. These solutions can thus serve as a means of critiquing the success of various approximate theories which might be applied to the full Hubbard model. In particular, we have examined the $T$-matrix approximation for the attractive Hubbard model in the atomic limit, which should give reasonable results at low electronic densities, if one can avoid the spurious phase transition that results when a fully non-self-consistent $T$-matrix approximation is employed—previously we have shown that any level of self-consistency guarantees that this phase transition is correctly suppressed to zero temperature in two dimensions or less. Here, a minimally self-consistent $T$-matrix approximation is shown to be successful in reproducing the exact results for the atomic limit, while fully self-consistent $T$-matrix results do not agree with the known solutions. Of particular note is that the minimally self-consistent $T$-matrix approximation reproduces not only one- and two-particle (static) thermodynamic quantities, but it also exactly reproduces the one-particle spectral function at low but nonzero temperatures. We also make a comparison to the two-particle self-consistent approach of Vilk and Tremblay, and find that the minimally self-consistent $T$-matrix theory can give better results over a broader temperature range.

• Received 28 January 2005

DOI:https://doi.org/10.1103/PhysRevB.71.155111