GS03 — Topological Insulators and Chern Bands

Field Value
Domain Grassmannian Systems
System Occupied band subspace over Brillouin zone
Group U(k) acting on Gr(k, 2k)
H^k tier
ISA Forge (β ≈ β*)
Status Validated
Opcodes ORBIT · TWIST · BIND
Papers Paper 574

Physical system

A topological insulator has an energy gap separating k occupied bands from unoccupied bands at every crystal momentum k ∈ 𝕋^d (the Brillouin zone torus). The occupied-band subspace V(k) ∈ Gr(k, n) varies smoothly over the Brillouin zone, defining a vector bundle E → 𝕋^d. Whether the insulator is topologically trivial or non-trivial depends on the topology of this bundle — specifically its Chern classes c₁(E), c₂(E), … ∈ H^{2k}(𝕋^d, ℤ).

The Chern number c₁ ∈ ℤ = ∫_{𝕋²} F/2π (where F = dA is the Berry curvature 2-form) is the BIND invariant: the integral of the Berry curvature over the Brillouin zone 2-torus. The quantised Hall conductance is σ_xy = c₁ e²/h — one conductance quantum per Chern number unit. This is the TKNN formula (Thouless, Kohmoto, Nightingale, den Nijs 1982), the first topological invariant in condensed matter physics.


Target category

Vect(𝕋^d) — the category of complex vector bundles over the Brillouin zone torus 𝕋^d = (S¹)^d, with smooth unitary bundle maps as morphisms. Objects: occupied-band bundles E → 𝕋^d of rank k. Classification: stable isomorphism classes of rank-k bundles over 𝕋^d are given by K-theory K̃(𝕋^d); the Chern character maps K̃(𝕋^d) → H^{even}(𝕋^d, ℚ), and the integer Chern classes are in H^{2j}(𝕋^d, ℤ).

Interpretation functor

F: C → Vect(𝕋^d) defined by:

Opcode F(opcode)
ORBIT Band evolution: V(k) → V(k+dk) as crystal momentum k moves around the Brillouin zone; adiabatic parallel transport of the occupied subspace
TWIST Berry connection A_μ(k) = i⟨u_n(k)|∂_{k_μ}|u_n(k)⟩: the H¹ connection 1-form on the bundle E; Berry phase γ = ∮_C A accumulated around a loop C in the BZ
BIND Berry curvature F_μν = ∂μ A_ν − ∂_ν A_μ + [A_μ, A_ν]: the H² curvature 2-form; Chern number c₁ = (1/2π)∫{𝕋²} tr(F) ∈ ℤ

ISA programme

BLOCH:   LABEL[H(k) | Bloch Hamiltonian at each k]   -- parametric family over BZ
DIAG:    ORBIT[H(k)|u_n(k)> = E_n(k)|u_n(k)>]       -- eigenvalue problem at each k
SUBSP:   LABEL[V(k) = span{|u_1(k)>,...,|u_k(k)>}]  -- occupied subspace in Gr(k,n)
CONNECT: TWIST[A_mu(k) = i<u|d_k u>]                -- Berry connection (H1)
CURV:    BIND[F = dA + A^A]                          -- Berry curvature (H2)
CHERN:   BIND[c_1 = (1/2pi) int_T2 tr(F)]           -- Chern number (integer)
CONDUCT: LABEL[sigma_xy = c_1 * e^2/h]              -- Hall conductance (LABEL output)
EDGE:    LABEL[n_edge = |c_1| chiral edge modes]     -- bulk-edge correspondence

Computable output

  • Chern number c₁ ∈ ℤ: the H² BIND invariant. Computable numerically via the Fukui-Hatsugai-Suzuki discretisation: divide the BZ into a mesh, compute Berry phases around each plaquette, sum the winding numbers. The result is always an exact integer (modular arithmetic over ℤ). For the Haldane model on honeycomb lattice: c₁ = ±1 in the topological phase, 0 in the trivial phase.
  • Quantised Hall conductance σ_xy = c₁ e²/h: measured experimentally to nine significant figures in quantum Hall systems. The TKNN formula is one of the most precisely tested results in condensed matter physics.
  • Chiral edge modes: by the bulk-edge correspondence (a topological version of the BIND closure condition), a system with |c₁| = n has exactly n chiral conducting modes at each edge, immune to backscattering. This is the H² output that cannot be obtained from H⁰ (band gap) or H¹ (Berry phase along a line) alone — it requires the integral over the full 2D Brillouin zone.
  • Z₂ topological invariants (time-reversal symmetric insulators, Kane-Mele): when Kramers degeneracy is enforced, the Chern number vanishes but a ℤ₂ invariant ν ∈ {0,1} persists. This is the BIND obstruction in the symplectic symmetry class — the H² class in real K-theory (KO-theory) rather than complex K-theory.

H^k structure of band topology

H^k tier Invariant Physical signature ISA
H⁰ Band gap E_gap at every k Insulating bulk ORBIT (eigenvalue)
Berry phase γ = ∮ A along loop Polarisation; Zak phase; WCC TWIST
Chern number c₁ = ∫ F Hall conductance; edge modes BIND
Chern-Simons invariant θ Axion electrodynamics (3D TI) BIND²

The H³ level (3D topological insulators, magnetoelectric θ-term) requires a BIND² opcode — the Chern-Simons form CS = tr(A∧dA + 2/3 A∧A∧A) is the secondary characteristic class whose boundary is tr(F∧F). This connects to the Yang-Mills instanton (G01): the topological insulator’s θ-term is the condensed-matter analogue of the QCD θ-vacuum.

Connection to the Grassmannian framework (Paper 574)

The occupied-band subspace V(k) ∈ Gr(k, n) and the molecular orbital subspace V ∈ Gr(k, n) in chemistry (C06, P01) are the same mathematical object — a k-plane varying over a parameter space. The parameter space is the Brillouin zone 𝕋^d in condensed matter and the nuclear configuration space ℝ^{3N} in chemistry. The Berry phase (TWIST) and Chern number (BIND) appear in both:

Chemistry Condensed matter ISA
Molecular Berry phase around conical intersection Berry phase around Dirac point in BZ TWIST
Non-adiabatic coupling (NAMD) Inter-band matrix elements TWIST failure
θ_G angle (alchemi) Principal angle between V(k) and V(k+dk) ORBIT metric
NOON spectrum Occupation numbers of Bloch bands LABEL
CASSCF convergence = Schubert crossing Band gap closure = topological transition β* snap

The topological phase transition — where c₁ changes from 0 to 1 — is a Schubert variety crossing in Gr(k, n) as a function of Hamiltonian parameters. The gap closes (σₖ² = σₖ₊₁²) at the transition: this is the same β* snap as the CASSCF convergence threshold in x596c. Both are instances of the same categorical event: a k-plane hitting a codimension-1 Schubert variety in Gr(k, n).

Validation

  • TKNN formula: Thouless, Kohmoto, Nightingale & den Nijs (1982). The original proof; Chern number = Hall conductance. Confirmed experimentally in GaAs/AlGaAs heterostructures (von Klitzing 1980 Nobel Prize in Physics 1985).
  • Haldane model: Haldane (1988). First theoretical model with c₁ = ±1 without Landau levels; realised experimentally in cold atoms (Jotzu et al. 2014).
  • Kane-Mele ℤ₂ invariant: Kane & Mele (2005). Validated in HgTe quantum wells (König et al. 2007); Nobel Prize in Physics 2016 (Haldane, Kosterlitz, Thouless).
  • Bulk-edge correspondence: Hatsugai (1993). Exact for Chern insulators; proved via K-theory for general symmetry classes (Kitaev 2009 periodic table).

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