On-site Coulomb repulsion ($U>0$)

The energy cost $U=\iint |a({\mathbf{r}}_1){|}^2|a({\mathbf{r}}_2){|}^2{e}^2/[4\pi {ϵ}_0|{\mathbf{r}}_1-{\mathbf{r}}_2|]{\mathrm{d}}^3{r}_1{\mathrm{d}}^3{r}_2$ for double occupancy of a localized orbital. It is positive for an ordinary nonzero orbital. A model retaining only this integral also neglects intersite and mixed Coulomb terms; orbital orthogonality alone does not make those terms vanish.

An isolated spin-degenerate orbital coupled to a particle reservoir has Hamiltonian $H=ϵ(n_{\uparrow }+n_{\downarrow })+U{n}_{\uparrow }{n}_{\downarrow }$. The four occupation microstates have weights $1,z,z,q{z}^2$, where $z=e^{\beta (\mu -ϵ)}$ and $q=e^{-\beta U}$. Thus $\Xi =1+2z+q{z}^2$ and $\bar{n}=(2z+2q{z}^2)/\Xi$. The factor two counts distinct spin states. At $U=0$ the grand partition function factorizes into two independent fermionic orbitals; positive U separates the two electron-addition thresholds.

For a repulsive single orbital at temperatures small compared with U, chemical potentials strictly between the first and second addition energies select a singly occupied defect. Its mean occupation is nearly one while its charge-number variance is exponentially small away from the thermally broadened step regions. The remaining spin degeneracy does not itself produce charge-number fluctuations.

With $z=e^{\beta (\mu -ϵ)}$, $q=e^{-\beta U}$ and $\Xi =1+2z+q{z}^2$, direct four-state averaging givesFor $\beta U\gg 1$ it has two separated peaks of height $1/4$ at the spin-degeneracy-shifted charge steps. At the lower bare crossing $\mu =ϵ$, three microstates share equal limiting weight, giving $\bar{n}=2/3$ and variance $2/9$. The upper bare crossing similarly gives mean $4/3$ and variance $2/9$. Between crossings occupation is pinned near one and fluctuations are suppressed. The identity is grand-canonical; fixing total particle number would change the fluctuation question.

Near the first addition threshold of a repulsive single-orbital defect, the empty and singly occupied grand weights are 1 and $2e^{\beta (\mu -ϵ)}$. Equal weights occur at $\mu =ϵ-k_{B}T\ln 2$. Near the second threshold, the single/double aggregate weights are in ratio $2:e^{\beta (\mu -ϵ-U)}$, giving half-step position $ϵ+U+k_{B}T\ln 2$. These are entropic finite-temperature shifts, which vanish in energy units as $T\to 0$. They must still be retained when evaluating statistics exactly at an unshifted degeneracy point.