Theorem 17 — Quantum Gibbs preparation at the full spectral threshold
OpenTFIMPhaseDiagram.Classical.Supplement.theorem_17Let , let be an arbitrary symmetric real interaction matrix, and let be arbitrary signed, site-dependent real fields. Write and , with . Suppose , including equality. For every a uniform quantum algorithm prepares a density operator satisfying
One polynomial bounds the classical generation, elementary quantum gates and workspace by . A published finite Real RAM generates the actual finite circuit. Its state is averaged over its internal tape and reduced by partial trace; the error is the full Schatten-one norm, without a postselection or success-probability assumption.
The statement also retains the fixed-gate realization promised in footnote 16 and Appendix B: on rational coefficients of bit length at most , an actual finite-bit compiler prepares the same accuracy with polynomial cost in the displayed parameters and . Both programs and their polynomial constants precede every varying input.
Differences from the source:
- encoding: a uniform quantum algorithm and its output state → charged Real RAM and bit-machine compilers, H/T/CNOT execution and discarded workspace; reason: operational uniform circuit semantics with actual matrix execution, including internal randomness
- encoding: polynomial runtime and rational-input finite-precision overhead → one uniform integer C,d and a ceiling polynomial, with the bit length added for the fixed-gate implementation; reason: integer budgets express the displayed polynomial and Appendix B overhead without changing any instance
- encoding: any real target accuracy e in (0,1) → the same domain, with precisionBits(e) for finite compilation; reason: rounding the requested precision upwards in bits attains every real target tolerance
import Definitions.Def_TFIMPhaseDiagram_Classical_Supplement_computation set_option autoImplicit false noncomputable section
namespace TFIMPhaseDiagram.Classical.Supplement
/-- Theorem 17, p.11, and its fixed finite gate implementation (footnote 16,
Appendix B). Both classical compilers and all executed gates/workspace are charged.
The real-input clause is not replaced by the rational implementation clause. -/
theorem theorem_17 :
∃ (P : RAMProgram) (f : Word → Word) (M : BitMachine f) (C d : ℕ), 0 < C ∧
(∀ (n : ℕ), 0 < n → ∀ (J : Matrix (Fin n) (Fin n) ℝ) (h η : Fin n → ℝ)
(e : ℝ), J.IsSymm → 0 < e → e < 1 → spectralWidth J * alpha (minField η) ≤ 1 →
PreparesReal P (quantumInput J h η e) (prepBudget C d J h η e) (density J h η) e) ∧
(∀ (n b : ℕ), 0 < n → ∀ (J : Matrix (Fin n) (Fin n) ℚ) (h η : Fin n → ℚ)
(e : ℝ), J.IsSymm → inputBitsBound b J h η → 0 < e → e < 1 →
let Jr : Matrix (Fin n) (Fin n) ℝ := fun i j => J i j
let hr : Fin n → ℝ := fun i => h i
let ηr : Fin n → ℝ := fun i => η i
spectralWidth Jr * alpha (minField ηr) ≤ 1 →
let input := rationalInput J h η (precisionBits e)
let c := circuitDecode (f input)
let B := C * ⌈(1+n+b+maxEntry Jr+maxField hr+maxField ηr+e⁻¹)^d⌉₊
M.time input.length ≤ B ∧ circuitCost n c ≤ B ∧
schattenOne (circuitDensity n c - density Jr hr ηr) ≤ e) := by sorry
end TFIMPhaseDiagram.Classical.Supplement