Completely positive order and environmental filters
DefinitionCRCD_QuantumChannelContinuity_Filterchannel-renyi-continuity-jwang226quantum-information
For superoperators , write when is completely positive at every matrix amplification. A rectangular dilation determines
An environmental map acts on the dilation as . These interfaces retain the output system and trace out only the environment. Their supporting operator identities relate environmental contractions and defects to completely positive domination and dilation error bounds; the contraction and positivity conditions are explicit in each applicable result.
Definition code
import Mathlib.Analysis.CStarAlgebra.CompletelyPositiveMap
import Mathlib.Analysis.CStarAlgebra.ContinuousLinearMap
import Mathlib.Analysis.InnerProductSpace.PiL2
import Mathlib.Analysis.InnerProductSpace.StarOrder
import Mathlib.Analysis.InnerProductSpace.TensorProduct
import Mathlib.Analysis.InnerProductSpace.Trace
import Mathlib.Analysis.SpecialFunctions.ContinuousFunctionalCalculus.Rpow.Basic
import Mathlib.LinearAlgebra.Matrix.ToLin
import Mathlib.LinearAlgebra.Trace
import Mathlib.Tactic.Abel
import Mathlib.Tactic.Linarith
import Mathlib.Topology.Algebra.Module.LinearMapPiProd
import Definitions.Def_CRCD_Quantum_QuantumMechanics_QuantumChannel_part_4
import Definitions.Def_CRCD_Quantum_QuantumMechanics_QuantumState
/-
Copyright (c) 2026 Jinzhao Wang. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Jinzhao Wang (AI-assisted formalization)
-/
/-!
# Concrete dilation algebra used by the fixed-environment filter
`dilationChannel` is an actual completely positive map obtained by partial
trace, using Lean-Quantum's finite-dimensional Hilbert-space definitions.
The results below verify the channel-level parallelogram inequality and the
environment-contraction step of the manuscript. The SDP attainment theorem
in Gour (2026), equation (44), is not asserted here.
-/
open QuantumState QuantumChannel TensorProduct
open scoped ComplexOrder
namespace QuantumChannelContinuity
universe u
set_option maxHeartbeats 800000
variable {A B E F : Type u} [Qudit A] [Qudit B] [Qudit E] [Qudit F]
/-- Complete-positive order, including all matrix amplifications. -/
def CPLe (Φ Ψ : QuantumChannel.T A B) : Prop :=
IsCompletelyPositive (Ψ - Φ)
theorem cp_add {Φ Ψ : QuantumChannel.T A B}
(hΦ : IsCompletelyPositive Φ) (hΨ : IsCompletelyPositive Ψ) :
IsCompletelyPositive (Φ + Ψ) := by
apply (isCompletelyPositive_iff_cstarMatrix_nonneg _).mpr
intro k M hM
have hmap : M.map (Φ + Ψ) = M.map Φ + M.map Ψ := by
ext i j x
simp [CStarMatrix.map_apply]
rw [hmap]
exact add_nonneg
((isCompletelyPositive_iff_cstarMatrix_nonneg _).mp hΦ k M hM)
((isCompletelyPositive_iff_cstarMatrix_nonneg _).mp hΨ k M hM)
theorem CPLe.trans {Φ Ψ Ω : QuantumChannel.T A B}
(hΦΨ : CPLe Φ Ψ) (hΨΩ : CPLe Ψ Ω) : CPLe Φ Ω := by
have h := cp_add hΦΨ hΨΩ
have hid : (Ψ - Φ) + (Ω - Ψ) = Ω - Φ := by abel
simpa only [CPLe, hid] using h
/-- The channel associated with a rectangular dilation operator. -/
noncomputable def dilationChannel (V : A →ₗ[ℂ] B ⊗[ℂ] E) : QuantumChannel.T A B :=
(TrRight (ℋ₂ := B) (ℋ₃ := E)).comp (krausTerm V)
theorem dilationChannel_cp (V : A →ₗ[ℂ] B ⊗[ℂ] E) :
IsCompletelyPositive (dilationChannel V) :=
comp_isCompletelyPositive _ _ (krausTerm_isCompletelyPositive V)
(TrRight_isCompletelyPositive (stdOrthonormalBasis ℂ E))
/-- The parallelogram identity holds before and after the partial trace. -/
theorem dilationChannel_parallelogram (U W : A →ₗ[ℂ] B ⊗[ℂ] E) :
(2 : ℂ) • dilationChannel U + (2 : ℂ) • dilationChannel W -
dilationChannel (U - W) = dilationChannel (U + W) := by
have h : (2 : ℂ) • krausTerm U + (2 : ℂ) • krausTerm W -
krausTerm (U - W) = krausTerm (U + W) := by
ext1 X
simp only [LinearMap.add_apply, LinearMap.sub_apply,
krausTerm, LinearMap.coe_mk, AddHom.coe_mk, map_add, map_sub, two_smul,
LinearMap.add_comp, LinearMap.sub_comp, LinearMap.comp_add, LinearMap.comp_sub]
abel
ext1 X
have hx := congrArg (fun T : QuantumChannel.T A (B ⊗[ℂ] E) => TrRight (T X)) h
simpa [dilationChannel, LinearMap.comp_apply] using hx
/-- Equation `Φ_(U-W) ≤cp 2 Φ_U + 2 Φ_W`, with complete positivity proved. -/
theorem dilationChannel_sub_le (U W : A →ₗ[ℂ] B ⊗[ℂ] E) :
CPLe (dilationChannel (U - W))
((2 : ℂ) • dilationChannel U + (2 : ℂ) • dilationChannel W) := by
unfold CPLe
rw [dilationChannel_parallelogram]
exact dilationChannel_cp (U + W)
/-- Taking a partial trace of an elementary tensor operator. -/
theorem trRight_map (X : L B) (Y : L E) :
TrRight (TensorProduct.map X Y) = (Tr Y) • X := by
have hcomm : conjugateEnd (TensorProduct.comm ℂ B E) (TensorProduct.map X Y) =
TensorProduct.map Y X := by
ext b e
simp [conjugateEnd]
change Tr₂ (conjugateEnd (TensorProduct.comm ℂ B E) (TensorProduct.map X Y)) = _
rw [hcomm, ← l_tensor_equiv_symm_tmul]
exact Tr₂_l_tensor_equiv_symm_tmul Y X
/-- Environmental action on a dilation, leaving its output system fixed. -/
noncomputable def environmentMap (U : E →ₗ[ℂ] F) : B ⊗[ℂ] E →ₗ[ℂ] B ⊗[ℂ] F :=
TensorProduct.map LinearMap.id U
theorem environment_kraus_map (U : E →ₗ[ℂ] F) (X : L B) (Y : L E) :
krausTerm (environmentMap (B := B) U) (TensorProduct.map X Y) =
TensorProduct.map X (krausTerm U Y) := by
simp only [krausTerm, environmentMap, LinearMap.coe_mk, AddHom.coe_mk,
TensorProduct.adjoint_map, LinearMap.adjoint_id]
rw [← TensorProduct.map_comp, ← TensorProduct.map_comp]
simp
/-- The discarded environmental part is itself a completely positive map.
The relation on `U,R` is the exact isometric-completion identity. -/
theorem environment_defect_identity
(U : E →ₗ[ℂ] F) (R : L E)
(hUR : (LinearMap.adjoint U).comp U + (LinearMap.adjoint R).comp R = 1) :
TrRight - (TrRight (ℋ₂ := B) (ℋ₃ := F)).comp
(krausTerm (environmentMap (B := B) U)) =
(TrRight (ℋ₂ := B) (ℋ₃ := E)).comp
(krausTerm (environmentMap (B := B) R)) := by
have helem (X : L B) (Y : L E) :
TrRight (TensorProduct.map X Y) -
TrRight (krausTerm (environmentMap (B := B) U) (TensorProduct.map X Y)) =
TrRight (krausTerm (environmentMap (B := B) R) (TensorProduct.map X Y)) := by
rw [environment_kraus_map, environment_kraus_map,
trRight_map, trRight_map, trRight_map]
have hU : Tr (krausTerm U Y) = Tr (((LinearMap.adjoint U).comp U).comp Y) := by
exact (LinearMap.trace_comp_comm' (U.comp Y) (LinearMap.adjoint U)).symm
have hR : Tr (krausTerm R Y) = Tr (((LinearMap.adjoint R).comp R).comp Y) := by
exact (LinearMap.trace_comp_comm' (R.comp Y) (LinearMap.adjoint R)).symm
have htrace : Tr (krausTerm U Y) + Tr (krausTerm R Y) = Tr Y := by
rw [hU, hR, ← map_add, ← LinearMap.add_comp, hUR]
rfl
rw [← sub_smul]
congr 1
exact sub_eq_iff_eq_add.mpr (by simpa [add_comm] using htrace.symm)
ext1 Z
obtain ⟨z, rfl⟩ := (l_tensor_equiv (ℋ₁ := B) (ℋ₂ := E)).symm.surjective Z
induction z using TensorProduct.induction_on with
| zero => simp
| tmul X Y => simpa [l_tensor_equiv_symm_tmul] using helem X Y
| add z w hz hw => simpa only [map_add] using congrArg₂ (· + ·) hz hw
/-- Finite-dimensional Douglas factorization, proved by extending the map on
the range of `S` by zero on its orthogonal complement. This is the linear
algebra needed when deriving dilation alignment from a Choi/Gram inequality. -/
theorem exists_contraction_factor
(T : A →ₗ[ℂ] F) (S : A →ₗ[ℂ] E)
(hTS : ∀ x, ‖T x‖ ≤ ‖S x‖) :
∃ U : E →ₗ[ℂ] F, (LinearMap.adjoint U).comp U ≤ 1 ∧ T = U.comp S := by
have hker : LinearMap.ker S ≤ LinearMap.ker T := by
intro x hx
rw [LinearMap.mem_ker] at hx ⊢
apply norm_eq_zero.mp
exact le_antisymm (by simpa [hx] using hTS x) (norm_nonneg _)
let K := LinearMap.range S
let D : K →ₗ[ℂ] F :=
((LinearMap.ker S).liftQ T hker).comp S.quotKerEquivRange.symm.toLinearMap
have hD (x : A) : D ⟨S x, LinearMap.mem_range_self S x⟩ = T x := by
change ((LinearMap.ker S).liftQ T hker)
(S.quotKerEquivRange.symm ⟨S x, LinearMap.mem_range_self S x⟩) = T x
rw [S.quotKerEquivRange_symm_apply_image]
exact Submodule.liftQ_apply _ _ _
have hDnorm (z : K) : ‖D z‖ ≤ ‖z‖ := by
obtain ⟨x, hx⟩ := z.property
have hz : z = ⟨S x, LinearMap.mem_range_self S x⟩ := Subtype.ext hx.symm
rw [hz, hD]
exact hTS x
let U : E →ₗ[ℂ] F := D.comp K.orthogonalProjection.toLinearMap
have hUnorm (x : E) : ‖U x‖ ≤ ‖x‖ :=
(hDnorm (K.orthogonalProjection x)).trans (K.norm_orthogonalProjection_apply_le x)
refine ⟨U, ?_, ?_⟩
· apply sub_nonneg.mp
apply (LinearMap.nonneg_iff_isPositive _).mpr
refine ⟨LinearMap.IsSymmetric.id.sub (LinearMap.isSymmetric_adjoint_comp_self U), ?_⟩
intro x
have hn := hUnorm x
simp only [LinearMap.sub_apply, Module.End.one_apply, LinearMap.comp_apply,
inner_sub_left, LinearMap.adjoint_inner_left, map_sub,
inner_self_eq_norm_sq (𝕜 := ℂ)]
nlinarith [norm_nonneg (U x), norm_nonneg x]
· ext x
have hp : K.orthogonalProjection (S x) = ⟨S x, LinearMap.mem_range_self S x⟩ := by
apply Subtype.ext
exact K.starProjection_eq_self_iff.mpr (LinearMap.mem_range_self S x)
change T x = D (K.orthogonalProjection (S x))
rw [hp, hD]
/-- Douglas factorization stated directly in terms of Gram-operator order. -/
theorem exists_contraction_factor_of_gram_le
(T : A →ₗ[ℂ] F) (S : A →ₗ[ℂ] E)
(hTS : (LinearMap.adjoint T).comp T ≤ (LinearMap.adjoint S).comp S) :
∃ U : E →ₗ[ℂ] F, (LinearMap.adjoint U).comp U ≤ 1 ∧ T = U.comp S := by
apply exists_contraction_factor T S
intro x
have hp := (LinearMap.nonneg_iff_isPositive _).mp (sub_nonneg.mpr hTS)
have hx := hp.re_inner_nonneg_right x
simp only [LinearMap.sub_apply, LinearMap.comp_apply, inner_sub_right,
LinearMap.adjoint_inner_right, map_sub, inner_self_eq_norm_sq (𝕜 := ℂ)] at hx
nlinarith [norm_nonneg (T x), norm_nonneg (S x)]
/-- Every environmental contraction has a positive square-root defect. -/
theorem exists_environment_defect (U : E →ₗ[ℂ] F)
(hU : (LinearMap.adjoint U).comp U ≤ 1) :
∃ R : L E, (LinearMap.adjoint U).comp U + (LinearMap.adjoint R).comp R = 1 := by
let D : L E := 1 - (LinearMap.adjoint U).comp U
have hD : 0 ≤ D := sub_nonneg.mpr hU
let R : L E := CFC.sqrt D
have hR : star R = R := (IsSelfAdjoint.of_nonneg (CFC.sqrt_nonneg D)).star_eq
have hRR : (LinearMap.adjoint R).comp R = D := by
change star R * R = D
rw [hR]
exact CFC.sqrt_mul_sqrt_self D hD
exact ⟨R, by rw [hRR]; simp [D]⟩
/-- A contraction acting only on the environment decreases partial trace in
complete-positive order. No positivity or invertibility of its input is
assumed: complete positivity includes arbitrary reference systems. -/
theorem environment_partialTrace_cp (U : E →ₗ[ℂ] F)
(hU : (LinearMap.adjoint U).comp U ≤ 1) :
IsCompletelyPositive
(TrRight - (TrRight (ℋ₂ := B) (ℋ₃ := F)).comp
(krausTerm (environmentMap (B := B) U))) := by
obtain ⟨R, hR⟩ := exists_environment_defect U hU
rw [environment_defect_identity U R hR]
exact dilationChannel_cp (environmentMap (B := B) R)
theorem krausTerm_comp (U : E →ₗ[ℂ] F) (V : A →ₗ[ℂ] E) :
krausTerm (U.comp V) = (krausTerm U).comp (krausTerm V) := by
ext X x
simp [krausTerm, LinearMap.adjoint_comp]
/-- The environment-transfer domination required in the fixed-dilation
lemma, proved for arbitrary supplied dilation and contraction. -/
theorem dilationChannel_environment_le
(U : E →ₗ[ℂ] F) (V : A →ₗ[ℂ] B ⊗[ℂ] E)
(hU : (LinearMap.adjoint U).comp U ≤ 1) :
CPLe (dilationChannel ((environmentMap (B := B) U).comp V))
(dilationChannel V) := by
unfold CPLe dilationChannel
rw [krausTerm_comp, ← LinearMap.comp_assoc, ← LinearMap.sub_comp]
exact comp_isCompletelyPositive _ _ (krausTerm_isCompletelyPositive V)
(environment_partialTrace_cp U hU)
/-- Tensoring an environmental contraction with the identity remains a
contraction in the Hilbert-space operator norm. -/
theorem environmentMap_norm_le_one (U : E →ₗ[ℂ] F)
(hU : (LinearMap.adjoint U).comp U ≤ 1) :
‖(environmentMap (B := B) U).toContinuousLinearMap‖ ≤ 1 := by
obtain ⟨R, hR⟩ := exists_environment_defect U hU
let T := environmentMap (B := B) U
let S := environmentMap (B := B) R
have hTS : (LinearMap.adjoint T).comp T + (LinearMap.adjoint S).comp S = 1 := by
simp only [T, S, environmentMap, TensorProduct.adjoint_map,
LinearMap.adjoint_id, ← TensorProduct.map_comp, LinearMap.id_comp,
← TensorProduct.map_add_right, hR]
exact TensorProduct.map_id
apply ContinuousLinearMap.opNorm_le_bound _ zero_le_one
intro x
have hx := congrArg (fun K : L (B ⊗[ℂ] E) => RCLike.re (inner ℂ x (K x))) hTS
simp only [LinearMap.add_apply, LinearMap.comp_apply, Module.End.one_apply,
inner_add_right, LinearMap.adjoint_inner_right,
map_add, inner_self_eq_norm_sq (𝕜 := ℂ)] at hx
change ‖T x‖ ≤ 1 * ‖x‖
nlinarith [sq_nonneg ‖S x‖, norm_nonneg (T x), norm_nonneg x]
/-- Transferring a two-piece decomposition to a prescribed environment
preserves its operator-norm error. The alignment equality is explicit;
its existence is the separate dilation-uniqueness theorem. -/
theorem fixed_environment_error
(U : E →ₗ[ℂ] F) (V : A →ₗ[ℂ] B ⊗[ℂ] F)
(X Y : A →ₗ[ℂ] B ⊗[ℂ] E)
(hU : (LinearMap.adjoint U).comp U ≤ 1)
(hV : (environmentMap (B := B) U).comp (X + Y) = V) :
‖(V - (environmentMap (B := B) U).comp X).toContinuousLinearMap‖ ≤
‖Y.toContinuousLinearMap‖ := by
have hid : V - (environmentMap (B := B) U).comp X =
(environmentMap (B := B) U).comp Y := by
rw [← hV, LinearMap.comp_add, add_sub_cancel_left]
rw [hid]
change ‖(environmentMap (B := B) U).toContinuousLinearMap.comp
Y.toContinuousLinearMap‖ ≤ ‖Y.toContinuousLinearMap‖
calc
_ ≤ ‖(environmentMap (B := B) U).toContinuousLinearMap‖ *
‖Y.toContinuousLinearMap‖ := ContinuousLinearMap.opNorm_comp_le _ _
_ ≤ 1 * ‖Y.toContinuousLinearMap‖ := mul_le_mul_of_nonneg_right
(environmentMap_norm_le_one (B := B) U hU) (norm_nonneg Y.toContinuousLinearMap)
_ = _ := one_mul _
end QuantumChannelContinuity
Source