Smooth eigenvector frame and reduced-resolvent vector for a gapped linear interpolation
ProvedQAlgorithms.Nannicini.adiabatic_linInterp_frameThere is a universal constant such that the following holds. Let be Hermitian, , , and let be continuous with a simple eigenvalue of separated by at least from the rest of the spectrum for every . Let be a unit eigenvector of for . Then there are functions with such that for all : and have derivatives , within ; is a unit eigenvector of with eigenvalue ; ; and (operator norm). Proof idea: parallel transport with the reduced resolvent (), ; bound via .
Preamble
import Definitions.Def_QAlgorithms_Defs_Adiabatic set_option autoImplicit false
Formal statement
universe u
namespace QAlgorithms.Nannicini
/-- Spectral input for the adiabatic theorem along `H(s) = (1 − s)H_I + sH_F`. If `λ(s)` is a
continuous simple eigenvalue of `H(s)` with gap `γ > 0` on `[0, 1]`, the eigenvector `φ₀` of
`H(0)` can be transported to a differentiable unit eigenvector family `u(s)` (eigenvalue `λ(s)`),
and `u′(s)` lies in the range of `H(s) − λ(s)`: `u′ = (H − λ)v` with `v` differentiable,
`‖v‖ ≤ K‖H_F − H_I‖/γ²`, `‖v′‖ ≤ K‖H_F − H_I‖²/γ³`, for a universal constant `K`. (One can take
`u′ = −R D u` (parallel transport) and `v = R u′`, `R` the reduced resolvent, `D = H_F − H_I`.) -/
theorem adiabatic_linInterp_frame : ∃ K : ℝ, ∀ {ι : Type u} [Fintype ι] [DecidableEq ι]
(HI HF : Matrix ι ι ℂ), HI.IsHermitian → HF.IsHermitian →
∀ (lam : ℝ → ℝ), ContinuousOn lam (Set.Icc 0 1) →
∀ (φ0 : EuclideanSpace ℂ ι), IsState φ0 → act (linInterp HI HF 0) φ0 = (lam 0 : ℂ) • φ0 →
∀ (γ : ℝ), 0 < γ →
(∀ s ∈ Set.Icc (0 : ℝ) 1, HasGapAround (linInterp HI HF s) (lam s) γ) →
∃ u u' v v' : ℝ → EuclideanSpace ℂ ι, u 0 = φ0 ∧
∀ s ∈ Set.Icc (0 : ℝ) 1,
HasDerivWithinAt u (u' s) (Set.Icc 0 1) s ∧ HasDerivWithinAt v (v' s) (Set.Icc 0 1) s ∧
IsState (u s) ∧ act (linInterp HI HF s) (u s) = (lam s : ℂ) • u s ∧
act (linInterp HI HF s) (v s) - (lam s : ℂ) • v s = u' s ∧
‖v s‖ ≤ K * specNorm (HF - HI) / γ ^ 2 ∧
‖v' s‖ ≤ K * specNorm (HF - HI) ^ 2 / γ ^ 3 := by
sorry
end QAlgorithms.Nannicini
Source
Giacomo Nannicini, Quantum algorithms for optimizers, 2025, v5, §9.2 (proof of Corollary 9.24); cf. Jansen–Ruskai–Seiler, J. Math. Phys. 48 (2007) 102111, and Kato, J. Phys. Soc. Japan 5 (1950). https://arxiv.org/pdf/2408.07086v5