Hubbard Lattice: Electron Localization at the Mott Crossover

A 2x2 plaquette of sites: electrons hop (sampled from the exact thermal state) and a red halo marks a costly double occupancy. Drag to orbit.

Switch to 2D
Controls

Presets

Parameters

Quality

Camera

Model checks

The science

Model. The Fermi-Hubbard model has two competing terms: electrons hop between neighbouring sites with amplitude t (kinetic energy, favours delocalisation) and two electrons on the same site cost U: H = -t Σ (c†iσ cjσ + h.c.) + U Σ ni↑ ni↓. At half filling and large U/t each site holds exactly one electron: a Mott insulator, whose spins interact through the superexchange J = 4t²/U.

What is computed. A 4-site ring (a 2×2 plaquette) with 2 up and 2 down electrons has only 36 basis states, so the Hamiltonian is diagonalised exactly (cyclic Jacobi). Everything shown is a canonical thermal average over all 36 eigenstates: double occupancy D = ⟨n↑n↓⟩, kinetic energy, local moment 1 - 2D, free energy and entropy. The animated electrons are configurations sampled from those exact probabilities.

Checks. U = 0 gives D = 1/4 exactly; the same code on a 2-site dimer reproduces the closed form E0 = (U - √(U² + 16t²))/2; the Hellmann-Feynman relation ∂F/∂U = Σ⟨n↑n↓⟩ ties the free energy to D; at U = 200t the ground energy approaches the Heisenberg-ring value -12t²/U.

Honest simplification. A 4-site cluster cannot show a true Mott phase transition (that needs the thermodynamic limit, e.g. dynamical mean-field theory in infinite dimensions). Here the metal-to-insulator behaviour appears as a smooth crossover of D from 1/4 toward 0 as U/t grows; the exact scale (U of order the bandwidth, about 8t for a square lattice) is only indicative for a cluster.

Loading model...
Loading Blender model...
-
-
Made with 🟠 Blender · Three.js