Tensor-product norm bounds and contractive Gram operators
DefinitionCRCD_QuantumChannelContinuity_FilterTensorchannel-renyi-continuity-jwang226quantum-information
For linear maps and between finite-dimensional complex Hilbert spaces, their tensor-product operator satisfies
In particular, tensoring either side with an identity does not increase the norm: and . Here all norms are continuous-linear-map operator norms. If , then in the positive-operator order. These inequalities also cover zero-dimensional spaces and zero operators; nontriviality is not a premise.
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_QuantumChannelContinuity_Filter
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)
-/
/-! # Hilbert operator-norm bounds for tensor products -/
open QuantumState QuantumChannel TensorProduct
open scoped ComplexOrder
namespace QuantumChannelContinuity
universe u
variable {A B E F : Type u} [Qudit A] [Qudit B] [Qudit E] [Qudit F]
set_option maxHeartbeats 800000
theorem adjoint_comp_le_one_of_norm_le_one (U : E →ₗ[ℂ] F)
(hU : ‖U.toContinuousLinearMap‖ ≤ 1) : (LinearMap.adjoint U).comp U ≤ 1 := by
apply sub_nonneg.mp
apply (LinearMap.nonneg_iff_isPositive _).mpr
refine ⟨LinearMap.IsSymmetric.id.sub (LinearMap.isSymmetric_adjoint_comp_self U), ?_⟩
intro x
have h := U.toContinuousLinearMap.le_opNorm x
have hn : ‖U x‖ ≤ ‖x‖ := by
exact h.trans (by simpa using mul_le_mul_of_nonneg_right hU (norm_nonneg 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]
/-- Tensoring an arbitrary operator with the identity does not increase its
operator norm; this includes zero-dimensional spaces and the zero operator. -/
theorem environmentMap_norm_le (U : E →ₗ[ℂ] F) :
‖(environmentMap (B := B) U).toContinuousLinearMap‖ ≤ ‖U.toContinuousLinearMap‖ := by
let c := ‖U.toContinuousLinearMap‖
have hc : 0 ≤ c := norm_nonneg U.toContinuousLinearMap
apply ContinuousLinearMap.opNorm_le_bound _ hc
intro x
by_cases hc0 : c = 0
· have hzero : U = 0 := by
apply LinearMap.toContinuousLinearMap.injective
exact norm_eq_zero.mp hc0
simp [hzero, environmentMap, hc0]
· have hnorm : ‖(((c : ℂ)⁻¹ • U).toContinuousLinearMap)‖ ≤ 1 := by
change ‖(c : ℂ)⁻¹ • U.toContinuousLinearMap‖ ≤ 1
rw [norm_smul, norm_inv, Complex.norm_real, Real.norm_of_nonneg hc]
change c⁻¹ * c ≤ 1
rw [inv_mul_cancel₀ hc0]
have h := environmentMap_norm_le_one (B := B) ((c : ℂ)⁻¹ • U)
(adjoint_comp_le_one_of_norm_le_one _ hnorm)
have hp : ‖environmentMap (B := B) ((c : ℂ)⁻¹ • U) x‖ ≤ ‖x‖ := by
exact ((environmentMap (B := B) ((c : ℂ)⁻¹ • U)).toContinuousLinearMap.le_opNorm x).trans
(by simpa using mul_le_mul_of_nonneg_right h (norm_nonneg x))
have heq : environmentMap (B := B) ((c : ℂ)⁻¹ • U) =
(c : ℂ)⁻¹ • environmentMap (B := B) U := by
simp [environmentMap, TensorProduct.map_smul_right]
rw [heq, LinearMap.smul_apply, norm_smul, norm_inv,
Complex.norm_real, Real.norm_of_nonneg hc] at hp
have hm := mul_le_mul_of_nonneg_left hp hc
simpa only [← mul_assoc, mul_inv_cancel₀ hc0, one_mul] using hm
theorem tensorMap_left_norm_le (U : A →ₗ[ℂ] B) :
‖(TensorProduct.map U (LinearMap.id : L E)).toContinuousLinearMap‖ ≤
‖U.toContinuousLinearMap‖ := by
apply ContinuousLinearMap.opNorm_le_bound _ (norm_nonneg U.toContinuousLinearMap)
intro x
have hx : TensorProduct.comm ℂ B E (TensorProduct.map U (LinearMap.id : L E) x) =
environmentMap (B := E) U (TensorProduct.comm ℂ A E x) := by
induction x using TensorProduct.induction_on with
| zero => simp
| tmul a e => simp [environmentMap]
| add x y hx hy => simp only [map_add, hx, hy]
change ‖TensorProduct.map U (LinearMap.id : L E) x‖ ≤ ‖U.toContinuousLinearMap‖ * ‖x‖
calc
_ = ‖TensorProduct.comm ℂ B E (TensorProduct.map U (LinearMap.id : L E) x)‖ :=
(TensorProduct.norm_comm _).symm
_ = ‖environmentMap (B := E) U (TensorProduct.comm ℂ A E x)‖ := congrArg norm hx
_ ≤ ‖(environmentMap (B := E) U).toContinuousLinearMap‖ *
‖TensorProduct.comm ℂ A E x‖ :=
(environmentMap (B := E) U).toContinuousLinearMap.le_opNorm _
_ ≤ ‖U.toContinuousLinearMap‖ * ‖x‖ := by
rw [TensorProduct.norm_comm]
exact mul_le_mul_of_nonneg_right (environmentMap_norm_le (B := E) U) (norm_nonneg x)
/-- Operator norms are submultiplicative under tensor products, on the
actual Hilbert tensor product used by the quantum-channel construction. -/
theorem tensorMap_norm_le (U : A →ₗ[ℂ] B) (V : E →ₗ[ℂ] F) :
‖(TensorProduct.map U V).toContinuousLinearMap‖ ≤
‖U.toContinuousLinearMap‖ * ‖V.toContinuousLinearMap‖ := by
have heq : TensorProduct.map U V = (environmentMap (B := B) V).comp
(TensorProduct.map U (LinearMap.id : L E)) := by
simp only [environmentMap, ← TensorProduct.map_comp, LinearMap.id_comp,
LinearMap.comp_id]
rw [heq]
change ‖(environmentMap (B := B) V).toContinuousLinearMap.comp
(TensorProduct.map U (LinearMap.id : L E)).toContinuousLinearMap‖ ≤ _
calc
_ ≤ ‖(environmentMap (B := B) V).toContinuousLinearMap‖ *
‖(TensorProduct.map U (LinearMap.id : L E)).toContinuousLinearMap‖ :=
ContinuousLinearMap.opNorm_comp_le _ _
_ ≤ ‖V.toContinuousLinearMap‖ * ‖U.toContinuousLinearMap‖ :=
mul_le_mul (environmentMap_norm_le (B := B) V) (tensorMap_left_norm_le (E := E) U)
(norm_nonneg (TensorProduct.map U (LinearMap.id : L E)).toContinuousLinearMap)
(norm_nonneg V.toContinuousLinearMap)
_ = _ := mul_comm _ _
end QuantumChannelContinuity
Source