Superconductivity in the Attractive Hubbard Model: The Double Hubbard--I Approximation

TL;DR

This paper applies a Hubbard--I decoupling scheme to the attractive Hubbard model, deriving Green functions with four symmetric poles, and explores superconducting order parameters with self-consistent spectra calculations, extending mean-field BCS results.

Contribution

It introduces a Hubbard--I based approximation for the attractive Hubbard model, analyzing Green functions with four poles and self-consistent spectra, and generalizes BCS theory to include local attractive interactions.

Findings

Green functions have four symmetric poles.

Self-consistent calculation of spectra and weights is necessary.

For s-wave pairing, results reduce to BCS with a renormalized chemical potential.

Abstract

Using the Dyson equation of motion for both the diagonal one-particle Green function, $G(\vec{k},\omega)$ and off--diagonal Green function, $F(\vec{k},\omega)$, at the level of the Hubbard--I decoupling scheme, we have found that they have four poles symmetric in pairs, justifying a more elaborated calculation done by the Z\"urich group by means of the $T$-Matrix approach (Pedersen et al, Z. Physik B {\bf 103}, 21 (1997)) and the moment approach of Nolting (Z. Physik {\bf 255}, 25 (1972)). We find that the energy spectra and the weights of $G(\vec{k},\omega)$ and $F(\vec{k},\omega)$ have to be calculated self-consistently. $G(\vec{k},\omega)$ satisfies the first two moments while $F(\vec{k},\omega)$ the first sum rule. Our {\it order parameter} $\alpha(T)$ is given by $1/N_s \sum_{{\vec{k}}} \epsilon({\vec{k}})\Delta({\vec{k}})$. Due to the fact that we have a purely local attractive interaction $\Delta(\vec{k})$ can be of {\it any} s--type wave. However, for a {\it pure s--wave}, for which $\alpha(T) = 0$, we go back to the mean--field $BCS$ results, with a renormalized chemical potential. In this case, the off--diagonal Green function, $F(\vec{k},\omega)$, satisfies the first two off--diagonal sum rules. We explicitly state the range of validity of our approximation.