Fermionic QBP block channels with filtered Majorana jumps, and the sampler
DefinitionQDSGibbs_Supplement_fermiBlockChanneldobrushin-shlosmangibbs-samplinglindbladianmathematical-physicsp2o-gran-per-chapterp2o-plan-paperp2o-v1quantum-information
For the Fermi–Hubbard Hamiltonian and a block :
- The QBP operator (7.15): , , with the QBP transform (3.2).
- The jumps (7.17) and , for , , .
- (7.18) , , ,
- The sampler over all nonempty labelled cubes of side .
Formalization Note , , the Bohr-frequency calculus and the exponential series are the base mission's qbpF, qbpG, bohrCalc, sexp; the ODE solution is chosen by Classical.epsilon as in the base qbpPathOp. uses the full Hamiltonian , and the filter uses the block Hamiltonian .
Definition code
import Mathlib
import Definitions.Def_QDSGibbs_blockChannel
import Definitions.Def_QDSGibbs_Supplement_fhLattice
open scoped Matrix
namespace QDSGibbs.Supplement
/-- The **fermionic QBP operator** `O_B = O_B(1)`, where
`O′_B(v) = −(β/4) F_{H − (1−v) W_∂B}(W_∂B) O_B(v)`, `O_B(0) = I` (Rouzé–Stilck França,
arXiv:2610.03578, held file sha256 ede55815…, (7.15), p.65), with `F` the QBP transform (3.2).
**Formalization Note** `F` is the base `qbpF`. The path is chosen by `Classical.epsilon` among the
solutions on all of `ℝ` (derivatives entrywise), exactly as the base `qbpPathOp`; the coefficient is
continuous, so the linear ODE has a unique global solution whose value at `1` is the paper's. -/
noncomputable def fqbpOp {Λ : Type} [Fintype Λ] [DecidableEq Λ] {M : ℕ} (β : ℝ)
(H : FermiHubbardMix.HamData Λ M) (B : Finset Λ) : FermiHubbardMix.Op M :=
(Classical.epsilon fun O : ℝ → FermiHubbardMix.Op M => O 0 = 1 ∧
∀ (u : ℝ) (x y : FermiCap.Gaussian.Qubits M), HasDerivAt (fun s => O s x y)
(((-(β / 4)) • qbpF β (H.ham - (1 - u) • crossF H B) (crossF H B) * O u) x y) u) 1
/-- The **dressed filtered Majorana jumps** (arXiv:2610.03578, (7.17), p.66):
`J_{B;i,s,r} = 2^{−1/2} (1 + e^{β ad_{H_B}})^{−1/2}(γ_{i,s,r})`, `L_{B;i,s,r} = O_B J_{B;i,s,r} O_B^{−1}`.
**Formalization Note** The filter is the base Bohr-frequency calculus `bohrCalc` of the block
Hamiltonian `H_B` (2.7) with `q(ω) = (1 + e^{βω})^{−1/2}`; `⁻¹` is Mathlib's matrix inverse (`O_B` is
invertible). -/
noncomputable def fJump {Λ : Type} [Fintype Λ] [DecidableEq Λ] {M : ℕ} (β : ℝ)
(ord : Λ × Fin 2 ≃ Fin M) (H : FermiHubbardMix.HamData Λ M) (B : Finset Λ) (i : Λ)
(s r : Fin 2) : FermiHubbardMix.Op M :=
fqbpOp β H B *
(((Real.sqrt 2)⁻¹ : ℝ) : ℂ) • bohrCalc (hamOnF H B)
(fun ω => (((Real.sqrt (1 + Real.exp (β * ω)))⁻¹ : ℝ) : ℂ)) (majF ord i s r) *
(fqbpOp β H B)⁻¹
/-- The jump map `T_B(X) = Σ_{i ∈ B} Σ_{s,r} L_{B;i,s,r} X L*_{B;i,s,r}` (arXiv:2610.03578, (7.18),
p.66). -/
noncomputable def fT {Λ : Type} [Fintype Λ] [DecidableEq Λ] {M : ℕ} (β : ℝ)
(ord : Λ × Fin 2 ≃ Fin M) (H : FermiHubbardMix.HamData Λ M) (B : Finset Λ)
(X : FermiHubbardMix.Op M) : FermiHubbardMix.Op M :=
∑ i ∈ B, ∑ s : Fin 2, ∑ r : Fin 2, fJump β ord H B i s r * X * (fJump β ord H B i s r)ᴴ
/-- The loss operator `S_B = Σ_{i ∈ B} Σ_{s,r} L*_{B;i,s,r} L_{B;i,s,r}` (arXiv:2610.03578, (7.18),
p.66). -/
noncomputable def fS {Λ : Type} [Fintype Λ] [DecidableEq Λ] {M : ℕ} (β : ℝ)
(ord : Λ × Fin 2 ≃ Fin M) (H : FermiHubbardMix.HamData Λ M) (B : Finset Λ) :
FermiHubbardMix.Op M :=
∑ i ∈ B, ∑ s : Fin 2, ∑ r : Fin 2, (fJump β ord H B i s r)ᴴ * fJump β ord H B i s r
/-- The **fermionic block generator** `K_B(X) = T_B(X) − ½{S_B, X} − i[G_B, X]`, `G_B = G_H(S_B)`
with the full Hamiltonian `H` (arXiv:2610.03578, (7.18), p.66).
**Formalization Note** `G_H` is the base coherent map `qbpG β H` (`(i/2) tanh((β/4) ad_H)`, (3.12)). -/
noncomputable def fK {Λ : Type} [Fintype Λ] [DecidableEq Λ] {M : ℕ} (β : ℝ)
(ord : Λ × Fin 2 ≃ Fin M) (H : FermiHubbardMix.HamData Λ M) (B : Finset Λ)
(X : FermiHubbardMix.Op M) : FermiHubbardMix.Op M :=
fT β ord H B X - (1 / 2 : ℂ) • (fS β ord H B * X + X * fS β ord H B) -
Complex.I • (qbpG β H.ham (fS β ord H B) * X - X * qbpG β H.ham (fS β ord H B))
/-- The **fermionic block channel** `Φ_B(H, T) = e^{T K_B}` (arXiv:2610.03578, (7.18), p.66), with the
base exponential series `sexp`. -/
noncomputable def fermiBlockChannel {Λ : Type} [Fintype Λ] [DecidableEq Λ] {M : ℕ} (β : ℝ)
(ord : Λ × Fin 2 ≃ Fin M) (H : FermiHubbardMix.HamData Λ M) (B : Finset Λ) (T : ℝ) :
FermiHubbardMix.Op M → FermiHubbardMix.Op M :=
sexp (fK β ord H B) T
/-- The **fermionic block sampler** `L = ℓ^{−D} Σ_B (Φ_B(H, T) − id)` over all nonempty labelled cubes
of side `ℓ` (arXiv:2610.03578, Theorem 1.1, p.6; p.66). -/
noncomputable def fermiSampler {D M : ℕ} (Λ : Finset (Fin D → ℤ)) (β : ℝ)
(ord : Λ × Fin 2 ≃ Fin M) (H : FermiHubbardMix.HamData Λ M) (ℓ : ℕ) (T : ℝ)
(X : FermiHubbardMix.Op M) : FermiHubbardMix.Op M :=
(((ℓ : ℂ) ^ D)⁻¹) • ∑ z ∈ cubeLabels Λ ℓ, (fermiBlockChannel β ord H (cube Λ ℓ z) T X - X)
end QDSGibbs.Supplement
Source
Rouzé–Stilck França, Rapid mixing of Gibbs samplers via quantum Dobrushin–Shlosman conditions, arXiv:2610.03578 (held file sha256 ede55815…), (7.15) p.65, (7.17) and (7.18) p.66, Theorem 1.1 p.6