CM01 — Hubbard Model Mott Transition

Field Value
Domain Condensed Matter
System Hubbard model at half-filling
Group SU(2) (spin rotation)
H^k tier
ISA Forge (β ≈ β*)
Status Validated
Opcodes ORBIT · TWIST · BIND · LABEL
Papers Paper 563

Physical system

The Hubbard model is the minimal model of interacting electrons on a lattice:

H = −t Σ_{⟨ij⟩,σ} c†{iσ} c{jσ} + U Σ_i n_{i↑} n_{i↓}

The hopping term t allows electrons to delocalise (ORBIT across the lattice); the on-site repulsion U penalises double occupancy (two electrons on the same site). At half-filling (one electron per site on average), the competition between t and U drives the Mott metal-insulator transition at a critical ratio U_c/t:

  • U/t ≪ 1 (metallic): electrons delocalise freely; Fermi liquid; ORBIT dominates; band theory applies.
  • U/t ≈ U_c/t ≈ 1.8 (critical, β* snap): double occupancy collapses; spectral weight transfers from quasiparticle peak to Hubbard bands; the Fermi surface topology changes.
  • U/t ≫ 1 (Mott insulator): electrons localise; one per site; charge gap 2U opens; spin physics described by Heisenberg antiferromagnet with J = 4t²/U; BIND dominates (superexchange = ring exchange = H² holonomy).

x563c experiment (Paper 563, SHA 174927e): Galerkin solver confirmed the Mott β* snap at U/t ≈ 1.8 via double-occupancy collapse — directly validating the ISA snap condition for the Hubbard model.


Target category

FermiLat(N) — the category of fermionic lattice models on N sites with SU(2) spin symmetry. Objects: N-site Fock spaces ⊗ᵢ (|0⟩, |↑⟩, |↓⟩, |↑↓⟩). Morphisms: particle-hole symmetric unitaries preserving the half-filling constraint ⟨n⟩ = 1. The Mott insulator is the terminal object in the U/t → ∞ limit: a product state of localised spins with no charge fluctuations.

Interpretation functor

F: C → FermiLat(N) defined by:

Opcode F(opcode)
ORBIT Hopping: c†{jσ} c{iσ} moves an electron from site i to site j; the kinetic ORBIT on the lattice; generates the Fermi sea in the non-interacting limit
TWIST Exchange: virtual hopping c†{iσ} c{jσ} c†{jσ’} c{iσ’} creates a Berry phase in the spin sector; generates the Heisenberg exchange J = 4t²/U in the Mott limit; the H¹ spin correlation
BIND Superexchange ring: four-site ring exchange t⁴/U³ term; generates the four-spin BIND interaction; becomes important near the Mott transition where ring exchange corrections to Heisenberg model are O(t/U)²
LABEL Double occupancy D = ⟨n_{i↑} n_{i↓}⟩: the order parameter; D ≈ 0.25 (free) → D ≈ 0 (Mott); the β* snap is D collapsing from ~0.15 to ~0.02 at U_c/t ≈ 1.8 (x563c)

ISA programme

INIT:     LABEL[half-filling: <n> = 1 per site]         -- constraint
FREE:     ORBIT[t * c†_j c_i | hopping, U=0 limit]      -- Fermi sea (H0)
INTERACT: LABEL[U * n_up * n_dn | on-site repulsion]    -- H0 energy penalty
DOUBLOCC: LABEL[D = <n_up n_dn> | double occupancy]     -- order parameter
SNAP?:    LABEL[D collapsing? | U/t near 1.8]           -- beta* test (x563c)
EXCHANGE: TWIST[J = 4t^2/U | spin exchange in Mott limit]  -- H1 spin physics
RING:     BIND[K = 4t^4/U^3 | ring exchange]            -- H2 correction
OUTPUT:   LABEL[metal if U<Uc, Mott insulator if U>Uc]  -- phase label

Computable output

  • Double occupancy D(U/t): computed by x563c (Galerkin solver) for the 1D Hubbard chain. D decreases monotonically from D = 0.25 (U=0, free fermions) through D ≈ 0.15 (U/t = 1) to D ≈ 0.02 (U/t = 4). The β* snap at U/t ≈ 1.8 is where dD/d(U/t) is maximum — the sharpest change in double occupancy, the ISA snap condition. Validated against exact Bethe ansatz results (Lieb-Wu 1968) for 1D; DMFT for infinite dimensions.
  • Mott gap Δ = U − W (U large): charge gap that opens in the insulating phase. For U ≫ t: Δ ≈ U − 2zt where z is the coordination number. The two Hubbard bands (lower: singly-occupied; upper: doubly-occupied) are LABEL eigenvalues separated by Δ. In 1D (Bethe ansatz): Δ = 0 for all U > 0 (no true Mott transition in 1D at T=0), but D still shows the crossover.
  • Heisenberg exchange J = 4t²/U in the Mott limit: the TWIST output. Néel temperature T_N ∝ J; spin-wave velocity v_s = J√2 (square lattice). Confirmed in cuprate parent compounds (La₂CuO₄: J ≈ 130 meV, U/t ≈ 8).
  • Spectral weight transfer: in ARPES, the quasiparticle peak at the Fermi energy loses spectral weight Z as U increases; Z → 0 at the Mott transition (Brinkman-Rice theory). Z is the ORBIT amplitude — it measures how much of the electron propagation is coherent (ORBIT-like) vs incoherent (blocked by U).

The Mott transition as β* snap

The Mott transition is the archetype of a correlation-driven β* snap: it is not driven by symmetry breaking (no order parameter in the Landau sense for the charge sector at half-filling in the paramagnetic Mott state) but by the change in the ORBIT topology of the Fermi surface.

In the ISA framework:

  • Metallic phase (U < U_c): the Fermi surface is a large ORBIT enclosing half the Brillouin zone (Luttinger theorem). ORBIT is the dominant opcode; TWIST gives spin fluctuations; BIND is perturbative.
  • At U_c (β* snap): the quasiparticle residue Z → 0; the ORBIT coherence collapses; spectral weight transfers from quasiparticle peak to incoherent Hubbard bands. The Fermi surface topology changes from a large electron Fermi surface to (in the Mott state) no Fermi surface at all.
  • Mott insulating phase (U > U_c): no ORBIT (charge is localised); TWIST dominates (spin exchange J = 4t²/U via virtual hopping); BIND appears as ring exchange corrections K = 4t⁴/U³.

The snap condition from x563c: double occupancy D collapses at U/t ≈ 1.8 for the Hubbard chain. This is the ISA snap: D is the ORBIT overlap (probability of two electrons sharing a site), and it measures the coherence of the hopping ORBIT. When D → 0, ORBIT is suppressed and TWIST takes over.

Connection to cuprate superconductivity

The cuprate high-temperature superconductors (La₂CuO₄, YBa₂Cu₃O₇, …) are doped Mott insulators — they sit at U/t just above the Mott transition, doped with holes that can move through the localised spin background. The ISA story:

Doping level Phase ISA
0 (undoped) Mott antiferromagnet TWIST (Heisenberg J)
Under-doped Pseudogap TWIST + partial ORBIT (doped holes)
Optimal doped d-wave superconductor BIND (Cooper pairs via spin fluctuation exchange)
Over-doped Conventional metal ORBIT (Fermi liquid recovers)

The superconducting dome is the region where BIND (Cooper pairing from TWIST spin fluctuations) wins over ORBIT (Fermi liquid). The pseudogap is the TWIST- dominated regime where spin correlations persist but charge order is absent. The Mott transition at U_c is the β* snap that makes all of this possible — without it, cuprates would be ordinary metals.

Cold-atom validation

The Hubbard model has been realised exactly in ultracold fermionic atoms (⁴⁰K or ⁶Li) in optical lattices, where U/t is tunable via Feshbach resonances and lattice depth:

  • Jördens et al. (2008), Nature 455, 204: Mott insulating state observed at half-filling via double-occupancy suppression — directly measuring D ≈ 0 in the Mott phase. The ISA LABEL output (D → 0) confirmed.
  • Greif et al. (2013), Science 340, 1307: short-range spin correlations (TWIST, Heisenberg exchange J) measured via spin-sensitive imaging.
  • Mazurenko et al. (2017), Nature 545, 462: long-range antiferromagnetic order (TWIST ordering) imaged directly at U/t ≈ 7.

Validation

  • Lieb & Wu (1968): exact Bethe ansatz solution for 1D Hubbard chain; no true Mott gap in 1D at T=0, but D(U/t) exactly computed.
  • x563c (Paper 563, SHA 174927e): Galerkin solver; β* snap at U/t ≈ 1.8 confirmed numerically.
  • DMFT (Georges et al. 1996, Rev. Mod. Phys.): exact solution in d=∞; Mott transition at U_c/t = 2.9√2; Z → 0 confirmed analytically.
  • Cuprates: J ≈ 130 meV in La₂CuO₄ from neutron scattering (Coldea et al. 2001); consistent with J = 4t²/U at U/t ≈ 8.

Part of the ISA Zoo. Categorical foundations: Paper 591.