Documentation

CsdLean4.Mathlib.Analysis.Matrix.AdiabaticPhase

The phase: Berry's geometric phase survives the adiabatic limit #

Category: 1-Mathlib (CSD-free; staged for upstream). BACKLOG #164 (out of #88; closes the loop #56 opened — the general theorem behind Berry's model).

Along a gapped C² path Γ with a nondegenerate branch (mult = 1, so rank P = 1), the orbit s ↦ A s ψ₀ of the adiabatic intertwiner A (#160, A' = K A, K = [Ṗ, P]) is a unit section of the eigenline whose connection form vanishes — it is BP-1's horizontal lift (GeometricPhase.horizontalLift). Consequently, for a closed eigenvector loop ψ 1 = e^{iφ} ψ 0, A 1 ψ₀ = e^{iβ} ψ₀ with β = geometricPhase ψ 1 φ the geometric phase of the loop, and the physical propagator at speed T satisfies

U_T 1 ψ₀ = e^{−iT∫₀¹ lam} e^{iβ} ψ₀ + O(1/T):

dynamical phase times Berry's phase, with the error of #162.

The identification is a uniqueness-of-ODE step. For a C¹ unit section ψ of the eigenline (P ψ = ψ, ‖ψ‖ = 1), K ψ = (1 − P) ψ̇ (differentiate P ψ = ψ), and for a rank-one P, P ψ̇ = ⟪ψ, ψ̇⟫ ψ = i Im⟪ψ, ψ̇⟫ ψ (Re⟪ψ, ψ̇⟫ = 0 by unit norm), so K ψ = ψ̇ − i A ψ with A = Im⟪ψ, ψ̇⟫ the connection form. The horizontal lift e^{−i∫A} ψ therefore solves χ' = K χ, χ 0 = ψ 0, and so does A s ψ 0; the vector-level uniqueness for propagators (Heisenberg constancy of A† χ) identifies them.

Honest scope #

References: M. V. Berry, Proc. R. Soc. A 392 (1984) 45; B. Simon, PRL 51 (1983) 2167; T. Kato, J. Phys. Soc. Japan 5 (1950) §3; GeometricPhase.lean (BP-1, #10), AdiabaticBornFock.lean (#163), AdiabaticPropagators.lean (#161), AdiabaticIntertwiner.lean (#160); specs/berry-phase-scoping.md BP-1, BP-5; specs/BACKLOG.md #164, #88, #56.

The derivative of s ↦ M s (v s) #

toEuclideanCLM, with scalars restricted to ℝ, as a real continuous linear map on matrices.

Equations
  • One or more equations did not get rendered due to their size.
Instances For
    theorem HasDerivWithinAt.toEuclideanCLM_apply {N : ℕ} {M : ℝ → Matrix (Fin N) (Fin N) ℂ} {M' : Matrix (Fin N) (Fin N) ℂ} {v : ℝ → EuclideanSpace ℂ (Fin N)} {v' : EuclideanSpace ℂ (Fin N)} {s : Set ℝ} {t : ℝ} (hM : HasDerivWithinAt M M' s t) (hv : HasDerivWithinAt v v' s t) :
    HasDerivWithinAt (fun (u : ℝ) => (Matrix.toEuclideanCLM (M u)) (v u)) ((Matrix.toEuclideanCLM M') (v t) + (Matrix.toEuclideanCLM (M t)) v') s t
    theorem HasDerivAt.toEuclideanCLM_apply {N : ℕ} {M : ℝ → Matrix (Fin N) (Fin N) ℂ} {M' : Matrix (Fin N) (Fin N) ℂ} {v : ℝ → EuclideanSpace ℂ (Fin N)} {v' : EuclideanSpace ℂ (Fin N)} {t : ℝ} (hM : HasDerivAt M M' t) (hv : HasDerivAt v v' t) :
    HasDerivAt (fun (u : ℝ) => (Matrix.toEuclideanCLM (M u)) (v u)) ((Matrix.toEuclideanCLM M') (v t) + (Matrix.toEuclideanCLM (M t)) v') t

    Vector-level uniqueness for propagators #

    theorem Matrix.IsPropagator.conjTranspose_apply_eq {N : ℕ} {G U : ℝ → Matrix (Fin N) (Fin N) ℂ} {T : ℝ} (hU : IsPropagator G T U) (hG : ∀ s ∈ Set.Icc 0 T, (G s).conjTranspose = -G s) {v : ℝ → EuclideanSpace ℂ (Fin N)} (hv : ∀ s ∈ Set.Icc 0 T, HasDerivWithinAt v ((toEuclideanCLM (G s)) (v s)) (Set.Icc 0 T) s) (s : ℝ) :
    s ∈ Set.Icc 0 T → (toEuclideanCLM (U s).conjTranspose) (v s) = v 0

    ★ Heisenberg constancy on vectors: for an anti-Hermitian generator, U† v ≡ v 0 along every solution v' = G v.

    theorem Matrix.IsPropagator.eq_apply_init {N : ℕ} {G U : ℝ → Matrix (Fin N) (Fin N) ℂ} {T : ℝ} (hU : IsPropagator G T U) (hG : ∀ s ∈ Set.Icc 0 T, (G s).conjTranspose = -G s) {v : ℝ → EuclideanSpace ℂ (Fin N)} (hv : ∀ s ∈ Set.Icc 0 T, HasDerivWithinAt v ((toEuclideanCLM (G s)) (v s)) (Set.Icc 0 T) s) (s : ℝ) :
    s ∈ Set.Icc 0 T → v s = (toEuclideanCLM (U s)) (v 0)

    ★ Uniqueness on vectors: every solution of v' = G v is U v 0.

    Rank-one projectors #

    theorem Matrix.toEuclideanCLM_apply_eq_inner_smul {N : ℕ} {P : Matrix (Fin N) (Fin N) ℂ} (hP : P.IsHermitian) (hrank : P.rank = 1) {ψ : EuclideanSpace ℂ (Fin N)} (hψ : (toEuclideanCLM P) ψ = ψ) (hunit : ‖ψ‖ = 1) (v : EuclideanSpace ℂ (Fin N)) :
    (toEuclideanCLM P) v = inner ℂ ψ v • ψ

    A rank-one Hermitian projector with a unit vector ψ in its range is v ↦ ⟪ψ, v⟫ ψ.

    theorem Matrix.eq_inner_smul_of_toEuclideanCLM_apply_eq {N : ℕ} {P : Matrix (Fin N) (Fin N) ℂ} (hP : P.IsHermitian) (hrank : P.rank = 1) {ψ y : EuclideanSpace ℂ (Fin N)} (hψ : (toEuclideanCLM P) ψ = ψ) (hunit : ‖ψ‖ = 1) (hy : (toEuclideanCLM P) y = y) :
    y = inner ℂ ψ y • ψ

    Every vector of the range of a rank-one Hermitian projector is a multiple of a unit vector in it: P y = y → y = ⟪ψ, y⟫ ψ.

    Two small general facts #

    theorem inner_deriv_eq_im_mul_I {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℂ E] {ψ : ℝ → E} {ψ' : E} {t : ℝ} (hψ : HasDerivAt ψ ψ' t) (hunit : ∀ (t : ℝ), ‖ψ t‖ = 1) :
    inner ℂ (ψ t) ψ' = ↑(inner ℂ (ψ t) ψ').im * Complex.I

    Along a unit curve, ⟪ψ, ψ̇⟫ is purely imaginary: ⟪ψ, ψ̇⟫ = i Im⟪ψ, ψ̇⟫.

    theorem exists_eq_exp_smul_of_norm_eq {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℂ E] {x y : E} {c : ℂ} (hx : ‖x‖ = 1) (hy : ‖y‖ = 1) (h : y = c • x) :
    ∃ (β : ℝ), y = Complex.exp (↑β * Complex.I) • x

    Two unit vectors, one a multiple of the other, differ by a phase.

    The horizontal lift starts where the curve does.

    Rank one and closed paths #

    theorem Matrix.GappedPath.rank_proj {N : ℕ} (Γ : GappedPath N) (s : ℝ) :
    (Γ.proj s).rank = Γ.mult s

    rank (P s) = mult s.

    theorem Matrix.GappedPath.rank_proj_eq_one {N : ℕ} (Γ : GappedPath N) {n : ℕ∞} (hH : ContDiff ℝ (↑n) Γ.H) (hmult : Γ.mult 0 = 1) (s : ℝ) :
    (Γ.proj s).rank = 1

    A nondegenerate branch (mult 0 = 1) has rank-one projectors everywhere.

    theorem Matrix.GappedPath.proj_one_eq_of_closed {N : ℕ} (Γ : GappedPath N) (hH : Γ.H 1 = Γ.H 0) (hlam : Γ.lam 1 = Γ.lam 0) :
    Γ.proj 1 = Γ.proj 0

    A closed path (H 1 = H 0, lam 1 = lam 0) has P 1 = P 0.

    theorem Matrix.GappedPath.exists_cyclic_of_proj_one_eq {N : ℕ} (Γ : GappedPath N) {n : ℕ∞} (hH : ContDiff ℝ (↑n) Γ.H) (hmult : Γ.mult 0 = 1) (hP : Γ.proj 1 = Γ.proj 0) {ψ : ℝ → EuclideanSpace ℂ (Fin N)} (hunit : ∀ (t : ℝ), ‖ψ t‖ = 1) (hsec : ∀ (t : ℝ), (toEuclideanCLM (Γ.proj t)) (ψ t) = ψ t) :
    ∃ (φ : ℝ), ψ 1 = Complex.exp (↑φ * Complex.I) • ψ 0

    A unit section of a rank-one eigenline with P 1 = P 0 closes up to a phase.

    K on a unit section of the eigenline #

    theorem Matrix.GappedPath.K_apply_section {N : ℕ} (Γ : GappedPath N) {n : ℕ∞} (hn : 1 ≤ n) (hH : ContDiff ℝ (↑n) Γ.H) {ψ : ℝ → EuclideanSpace ℂ (Fin N)} {ψ' : EuclideanSpace ℂ (Fin N)} {s : ℝ} (hψ : HasDerivAt ψ ψ' s) (hsec : ∀ (t : ℝ), (toEuclideanCLM (Γ.proj t)) (ψ t) = ψ t) :
    (toEuclideanCLM (Γ.K s)) (ψ s) = ψ' - (toEuclideanCLM (Γ.proj s)) ψ'

    ★ K ψ = (1 − P) ψ̇ for a differentiable section ψ of the eigenline.

    theorem Matrix.GappedPath.K_apply_unit_section {N : ℕ} (Γ : GappedPath N) {n : ℕ∞} (hn : 1 ≤ n) (hH : ContDiff ℝ (↑n) Γ.H) (hmult : Γ.mult 0 = 1) {ψ : ℝ → EuclideanSpace ℂ (Fin N)} (hψ : ContDiff ℝ 1 ψ) (hunit : ∀ (t : ℝ), ‖ψ t‖ = 1) (hsec : ∀ (t : ℝ), (toEuclideanCLM (Γ.proj t)) (ψ t) = ψ t) (s : ℝ) :
    (toEuclideanCLM (Γ.K s)) (ψ s) = deriv ψ s - (↑(GeometricPhase.connectionForm ψ s) * Complex.I) • ψ s

    ★★ K ψ = ψ̇ − i A ψ for a C¹ unit section ψ of a nondegenerate eigenline, A the connection form Im⟪ψ, ψ̇⟫: Kato's generator acts on the section as covariant differentiation.

    theorem Matrix.GappedPath.hasDerivAt_horizontalLift_K {N : ℕ} (Γ : GappedPath N) {n : ℕ∞} (hn : 1 ≤ n) (hH : ContDiff ℝ (↑n) Γ.H) (hmult : Γ.mult 0 = 1) {ψ : ℝ → EuclideanSpace ℂ (Fin N)} (hψ : ContDiff ℝ 1 ψ) (hunit : ∀ (t : ℝ), ‖ψ t‖ = 1) (hsec : ∀ (t : ℝ), (toEuclideanCLM (Γ.proj t)) (ψ t) = ψ t) (s : ℝ) :

    ★★ The horizontal lift of a unit section solves Kato's equation χ' = K χ.

    The intertwiner's orbit is the horizontal lift #

    theorem Matrix.GappedPath.IsIntertwiner.toEuclideanCLM_apply_eq_horizontalLift {N : ℕ} {Γ : GappedPath N} {A : ℝ → Matrix (Fin N) (Fin N) ℂ} (hA : Γ.IsIntertwiner 1 A) {n : ℕ∞} (hn : 1 ≤ n) (hH : ContDiff ℝ (↑n) Γ.H) (hmult : Γ.mult 0 = 1) {ψ : ℝ → EuclideanSpace ℂ (Fin N)} (hψ : ContDiff ℝ 1 ψ) (hunit : ∀ (t : ℝ), ‖ψ t‖ = 1) (hsec : ∀ (t : ℝ), (toEuclideanCLM (Γ.proj t)) (ψ t) = ψ t) (s : ℝ) :

    ★★★ The adiabatic intertwiner parallel-transports the eigenvector: for a nondegenerate branch and any C¹ unit section ψ of the eigenline, A s (ψ 0) = horizontalLift ψ s on [0, 1] — the orbit of the intertwiner is BP-1's horizontal lift.

    theorem Matrix.GappedPath.IsIntertwiner.toEuclideanCLM_apply_one_eq {N : ℕ} {Γ : GappedPath N} {A : ℝ → Matrix (Fin N) (Fin N) ℂ} (hA : Γ.IsIntertwiner 1 A) {n : ℕ∞} (hn : 1 ≤ n) (hH : ContDiff ℝ (↑n) Γ.H) (hmult : Γ.mult 0 = 1) {ψ : ℝ → EuclideanSpace ℂ (Fin N)} (hψ : ContDiff ℝ 1 ψ) (hunit : ∀ (t : ℝ), ‖ψ t‖ = 1) (hsec : ∀ (t : ℝ), (toEuclideanCLM (Γ.proj t)) (ψ t) = ψ t) {φ : ℝ} (hcyc : ψ 1 = Complex.exp (↑φ * Complex.I) • ψ 0) :

    ★★ The holonomy is Berry's phase: for a closed eigenvector loop ψ 1 = e^{iφ} ψ 0, A 1 (ψ 0) = e^{iβ} ψ 0 with β = geometricPhase ψ 1 φ.

    theorem Matrix.GappedPath.IsIntertwiner.exists_apply_one_eq {N : ℕ} {Γ : GappedPath N} {A : ℝ → Matrix (Fin N) (Fin N) ℂ} (hA : Γ.IsIntertwiner 1 A) {n : ℕ∞} (hn : 1 ≤ n) (hH : ContDiff ℝ (↑n) Γ.H) (hmult : Γ.mult 0 = 1) (hP : Γ.proj 1 = Γ.proj 0) {x : EuclideanSpace ℂ (Fin N)} (hx : (toEuclideanCLM (Γ.proj 0)) x = x) (hunit : ‖x‖ = 1) :
    ∃ (β : ℝ), (toEuclideanCLM (A 1)) x = Complex.exp (↑β * Complex.I) • x

    The holonomy exists without choosing a section: for a nondegenerate branch with P 1 = P 0 and a unit ψ₀ ∈ range (P 0), A 1 ψ₀ = e^{iβ} ψ₀ for some β.

    The phase theorem #

    theorem Matrix.GappedPath.IsAdPropagator.toEuclideanCLM_apply_eq {N : ℕ} {Γ : GappedPath N} {T : ℝ} {U A : ℝ → Matrix (Fin N) (Fin N) ℂ} (hU : Γ.IsAdPropagator T U) (hA : Γ.IsIntertwiner 1 A) {n : ℕ∞} (hn : 1 ≤ n) (hH : ContDiff ℝ (↑n) Γ.H) {s : ℝ} (hs : s ∈ Set.Icc 0 1) {x : EuclideanSpace ℂ (Fin N)} (hx : (toEuclideanCLM (Γ.proj 0)) x = x) :
    (toEuclideanCLM (U s)) x = Γ.dynPhase T s • (toEuclideanCLM (A s)) x

    ★★ U_A s ψ₀ = e^{−iT∫₀ˢ lam} · A s ψ₀ for P 0 ψ₀ = ψ₀ (#161 on vectors).

    theorem Matrix.GappedPath.exists_norm_sub_dynPhase_smul_le {N : ℕ} (Γ : GappedPath N) {n : ℕ∞} (hn : 2 ≤ n) (hH : ContDiff ℝ (↑n) Γ.H) (hex : ∀ (s : ℝ), ∃ (j : Fin N), ⋯.eigenvalues j = Γ.lam s) :
    ∃ (C : ℝ), 0 ≤ C ∧ ∀ (T : ℝ), 0 < T → ∀ (UT A : ℝ → Matrix (Fin N) (Fin N) ℂ), Γ.IsPhysPropagator T UT → Γ.IsIntertwiner 1 A → ∀ s ∈ Set.Icc 0 1, ∀ (x : EuclideanSpace ℂ (Fin N)), (toEuclideanCLM (Γ.proj 0)) x = x → ‖(toEuclideanCLM (UT s)) x - Γ.dynPhase T s • (toEuclideanCLM (A s)) x‖ ≤ C / T * ‖x‖

    ★★ The adiabatic approximation with the parallel-transported eigenvector: ‖U_T s ψ₀ − e^{−iT∫₀ˢ lam} A s ψ₀‖ ≤ (C / T) ‖ψ₀‖ for P 0 ψ₀ = ψ₀, uniformly on [0, 1] (any multiplicity).

    theorem Matrix.GappedPath.exists_norm_sub_phase_smul_le {N : ℕ} (Γ : GappedPath N) {n : ℕ∞} (hn : 2 ≤ n) (hH : ContDiff ℝ (↑n) Γ.H) (hex : ∀ (s : ℝ), ∃ (j : Fin N), ⋯.eigenvalues j = Γ.lam s) (hmult : Γ.mult 0 = 1) {ψ : ℝ → EuclideanSpace ℂ (Fin N)} (hψ : ContDiff ℝ 1 ψ) (hunit : ∀ (t : ℝ), ‖ψ t‖ = 1) (hsec : ∀ (t : ℝ), (toEuclideanCLM (Γ.proj t)) (ψ t) = ψ t) {φ : ℝ} (hcyc : ψ 1 = Complex.exp (↑φ * Complex.I) • ψ 0) :
    ∃ (C : ℝ), 0 ≤ C ∧ ∀ (T : ℝ), 0 < T → ∀ (UT : ℝ → Matrix (Fin N) (Fin N) ℂ), Γ.IsPhysPropagator T UT → ‖(toEuclideanCLM (UT 1)) (ψ 0) - (Γ.dynPhase T 1 * Complex.exp (↑(GeometricPhase.geometricPhase ψ 1 φ) * Complex.I)) • ψ 0‖ ≤ C / T

    ★★★ The phase theorem (Berry's phase from Kato's theorem). Along a gapped C² path with a nondegenerate branch and a closed C¹ unit eigenvector loop ψ (ψ 1 = e^{iφ} ψ 0), there is a constant C such that for every speed T > 0 and physical propagator U_T,

    ‖U_T 1 (ψ 0) − e^{−iT∫₀¹ lam} e^{iβ} (ψ 0)‖ ≤ C / T, β = geometricPhase ψ 1 φ:

    the state returns with the dynamical phase times Berry's geometric phase, up to O(1/T).