LIDT definitions: LDT / MakingMeasurementsProjective / QXPLayerIdentities / ProjectorApprox
DefinitionLIDT_MIPStarRE_LDT_MakingMeasurementsProjective_QXPLayerIdentities_ProjectorApproxlidt
This bundle proves the projector approximation properties of the QXP layer. It establishes positivity, Hermiticity, idempotence, and mass bounds for the P operators, proves the cross-sum identity, and derives the P-versus-Q approximation in general and rank-range settings.
Principal definitions:
pa_nonneg: positivity of the P operator.pa_hermitian: Hermiticity of the P operator.pa_idempotent: idempotence of the P operator.pa_mass_le_one: the total P mass bound.xHat_cross_sum_eq_sqrt: the XHat cross-sum square-root identity.q_p_cross_close: the cross-term closeness estimate.pQApprox: the general P-versus-Q approximation statement.pQApprox_ofRankReductionSigmaRangePositiveGram: the positive-Gram rank-range specialization.
Formalization credit: Sirui Lu, Ruixuan Deng, Yanqiao Zhu, and Zhengfeng Ji (arXiv:2609.19814). Mathematical source: Zhengfeng Ji, Anand Natarajan, Thomas Vidick, John Wright, and Henry Yuen (arXiv:2009.12982).
Definition code
-- Generated from MIPStarRE/LDT/MakingMeasurementsProjective/QXPLayerIdentities/ProjectorApprox.lean by skeleton subtraction.
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_AxisParallelLine
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_DiagonalLine
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_Distribution
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_DistributionAvg
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_DistributionUniformSums
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_LinePolynomials
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_LowDegreePolynomial
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_MeasurementLift
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_OpFamily
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_OperatorExpectations
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_PMFAverages
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_PMFUniformAverages
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_ParametersBase
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_ParametersFiniteAnswers
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_QuantumState
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_SubMeasurementCore
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_SubMeasurementFamilies
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_TensorPlacement
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_Projectivization
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_QXPLayerIdentities_LayerAlgebra
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_QXPLayerIdentities_PositiveGram_Completion
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_QXPLayerIdentities_PositiveGram_Rows
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_QXPLayerIdentities_PositiveGram_Sigma
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_QXPLayer_AlmostProjective
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_QXPLayer_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_QXPLayer_QCompleteness
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_QXPLayer_RankReduction_LowRank
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_QXPLayer_RankReduction_Sigma
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_QXPLayer_TruncationCombinatorics
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_Statements
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_BipartiteSelfConsistency_Completion
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_BipartiteSelfConsistency_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_BipartiteSelfConsistency_Local
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_CauchySchwarz
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_ComparisonCore
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_CompletionTransfer
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_DistanceBounds
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichMain_Completeness
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichPrep_ApproxDelta
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichPrep_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichPrep_InnerProduct
import Definitions.Def_LIDT_MIPStarRE_LDT_Tactic_QuantumNonneg
import Definitions.Def_LIDT_MIPStarRE_LDT_Test_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_Test_StrategyCore
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteHilbert
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_Basic
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_BlockDiagonal
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_NormalizedTrace
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_Order
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_TracePairing
import Definitions.Def_LIDT_MIPStarRE_Quantum_Measurement
import Definitions.Def_LIDT_MIPStarRE_Quantum_ProjectorONB
import Mathlib
import Mathlib.Algebra.BigOperators.Group.Finset.Sigma
import Mathlib.Algebra.BigOperators.Ring.Finset
import Mathlib.Algebra.Order.BigOperators.Group.Finset
import Mathlib.Analysis.InnerProductSpace.Adjoint
import Mathlib.Analysis.InnerProductSpace.GramMatrix
import Mathlib.Analysis.InnerProductSpace.PiL2
import Mathlib.Analysis.Matrix.Spectrum
import Mathlib.Analysis.MeanInequalitiesPow
import Mathlib.Analysis.Real.Sqrt
import Mathlib.Data.Finset.Max
import Mathlib.Data.Finset.Powerset
import Mathlib.Data.Fintype.EquivFin
import Mathlib.Probability.ProbabilityMassFunction.Integrals
import Mathlib.Probability.ProbabilityMassFunction.Monad
open scoped _root_.Matrix
namespace LIDT
namespace MIPStarRE.LDT.MakingMeasurementsProjective
end MIPStarRE.LDT.MakingMeasurementsProjective
namespace MIPStarRE.LDT.Preliminaries
end MIPStarRE.LDT.Preliminaries
/-!
# Section 5 — P-Q approximation for QXP layers
The final comparison estimates between the projective family `P` produced
from `XHat` and the original projective `Q` layer.
-/
@[expose] public section
open scoped BigOperators MatrixOrder _root_.Matrix ComplexOrder
namespace MIPStarRE.LDT.MakingMeasurementsProjective
open MIPStarRE.LDT
universe uOutcome uι
noncomputable section
private lemma pa_nonneg {Outcome : Type*}
{ι : Type*} [Fintype ι] [DecidableEq ι]
[Fintype Outcome]
(data : QXPLayerData Outcome ι) (a : Outcome) :
0 ≤ Pa data a := by
rcases pProjectivity data with ⟨P, hP⟩
simpa [hP a] using P.outcome_pos a
private lemma pa_hermitian {Outcome : Type*}
{ι : Type*} [Fintype ι] [DecidableEq ι]
[Fintype Outcome]
(data : QXPLayerData Outcome ι) (a : Outcome) :
(Pa data a)ᴴ = Pa data a :=
(Matrix.nonneg_iff_posSemidef.mp (pa_nonneg data a)).isHermitian.eq
private lemma pa_idempotent {Outcome : Type*}
{ι : Type*} [Fintype ι] [DecidableEq ι]
[Fintype Outcome]
(data : QXPLayerData Outcome ι) (a : Outcome) :
Pa data a * Pa data a = Pa data a := by
rcases pProjectivity data with ⟨P, hP⟩
simpa [hP a] using P.proj a
private lemma pa_mass_le_one {Outcome : Type*}
{ι : Type*} [Fintype ι] [DecidableEq ι]
[Fintype Outcome]
(data : QXPLayerData Outcome ι) (ψ : QuantumState ι)
(hψ : ψ.IsNormalized) :
(∑ a : Outcome, ev ψ (Pa data a)) ≤ 1 := by
rcases pProjectivity data with ⟨P, hP⟩
calc
(∑ a : Outcome, ev ψ (Pa data a))
= ∑ a : Outcome, ev ψ (P.outcome a) := by
refine Finset.sum_congr rfl ?_
intro a _
rw [hP a]
_ = ev ψ (∑ a : Outcome, P.outcome a) := by
exact (ev_sum ψ P.outcome).symm
_ = ev ψ P.total := by rw [P.sum_eq_total]
_ ≤ ev ψ (1 : MIPStarRE.Quantum.Op ι) := ev_mono ψ _ _ P.total_le_one
_ = 1 := ev_one_of_isNormalized ψ hψ
private lemma xHat_cross_sum_eq_sqrt {Outcome : Type*}
{ι : Type*} [Fintype ι] [DecidableEq ι]
[Fintype Outcome]
(data : QXPLayerData Outcome ι) :
(∑ a : Outcome, (Xa data a)ᴴ * data.xHat * Pa data a) =
CFC.sqrt (QTotal data.qLayer) := by
have hterm : ∀ a : Outcome,
(Xa data a)ᴴ * data.xHat * Pa data a =
data.xᴴ * Ta data.qLayer a * data.xHat := by
intro a
have hTa : (Ta data.qLayer a)ᴴ = Ta data.qLayer a := by
simpa [Ta] using ProjMeas.outcome_hermitian data.qLayer.t a
calc
(Xa data a)ᴴ * data.xHat * Pa data a
= (data.xᴴ * Ta data.qLayer a) * data.xHat *
(data.xHatᴴ * Ta data.qLayer a * data.xHat) := by
simp [Xa, Pa, Matrix.conjTranspose_mul, hTa, Matrix.mul_assoc]
_ = data.xᴴ * Ta data.qLayer a * (data.xHat * data.xHatᴴ) *
Ta data.qLayer a * data.xHat := by
simp [Matrix.mul_assoc]
_ = data.xᴴ * Ta data.qLayer a * Ta data.qLayer a * data.xHat := by
simp [data.xHat_coisometry, Matrix.mul_assoc]
_ = data.xᴴ * Ta data.qLayer a * data.xHat := by
simp [Ta, data.qLayer.t.proj a, Matrix.mul_assoc]
calc
(∑ a : Outcome, (Xa data a)ᴴ * data.xHat * Pa data a)
= ∑ a : Outcome, data.xᴴ * Ta data.qLayer a * data.xHat := by
refine Finset.sum_congr rfl ?_
intro a _
exact hterm a
_ = (∑ a : Outcome, data.xᴴ * Ta data.qLayer a) * data.xHat := by
simpa using
(Matrix.sum_mul (s := Finset.univ)
(f := fun a : Outcome => data.xᴴ * Ta data.qLayer a)
(M := data.xHat)).symm
_ = data.xᴴ * (∑ a : Outcome, Ta data.qLayer a) * data.xHat := by
have hsum : (∑ a : Outcome, data.xᴴ * Ta data.qLayer a) =
data.xᴴ * (∑ a : Outcome, Ta data.qLayer a) := by
simpa using
(Matrix.mul_sum (s := Finset.univ)
(f := fun a : Outcome => Ta data.qLayer a)
(M := data.xᴴ)).symm
rw [hsum]
_ = data.xᴴ * data.xHat := by
simpa [Ta] using
congrArg (fun M => data.xᴴ * M * data.xHat) data.qLayer.t.sum_eq
_ = CFC.sqrt (QTotal data.qLayer) := data.xHat_mixed
private lemma q_p_cross_close {Outcome : Type*}
{ι : Type*} [Fintype ι] [DecidableEq ι]
[Fintype Outcome]
(ψ : QuantumState ι)
(A : Measurement Outcome ι) (ζ : Error)
(data : QXPLayerData Outcome ι)
(hψ : ψ.IsNormalized)
(hζ : 0 ≤ ζ) (hζ_small : ζ ≤ 1 / (4 : Error))
(hRank : RankReductionWitness ψ A ζ data.qLayer) :
|(∑ a : Outcome, ev ψ (Qa data.qLayer a * Pa data a)) -
ev ψ (CFC.sqrt (QTotal data.qLayer))| ≤
2 * zetaQuarterRoot ζ := by
let D : Matrix data.qLayer.auxSpace.carrier ι ℂ := data.x - data.xHat
let first : Outcome → Error := fun a =>
ev ψ ((Xa data a)ᴴ * (D * Dᴴ) * Xa data a)
let second : Outcome → Error := fun a =>
ev ψ ((Pa data a)ᴴ * Pa data a)
have hfirst_nonneg : ∀ a : Outcome, 0 ≤ first a := by
intro a
dsimp [first, D]
simpa [Matrix.conjTranspose_mul, Matrix.mul_assoc] using
ev_adjoint_self_nonneg ψ ((data.x - data.xHat)ᴴ * Xa data a)
have hsecond_nonneg : ∀ a : Outcome, 0 ≤ second a := by
intro a
dsimp [second]
exact ev_adjoint_self_nonneg ψ (Pa data a)
have hfirst_le : (∑ a : Outcome, first a) ≤
4 * spectralTruncationError ζ := by
have hpoint : ∀ a : Outcome,
first a ≤ ev ψ (Qa data.qLayer a * QTotal data.qLayer * Qa data.qLayer a -
Qa data.qLayer a) := by
intro a
have hrect :
(Xa data a)ᴴ * (D * Dᴴ) * Xa data a ≤
(Xa data a)ᴴ *
((data.x * data.xᴴ - 1) * (data.x * data.xᴴ - 1)) *
Xa data a := by
apply sub_nonneg.mp
have hpsd :
0 ≤ (Xa data a)ᴴ *
(((data.x * data.xᴴ - 1) * (data.x * data.xᴴ - 1)) - D * Dᴴ) *
Xa data a := by
exact
(Matrix.PosSemidef.conjTranspose_mul_mul_same
(Matrix.nonneg_iff_posSemidef.mp <|
sub_nonneg.mpr (by simpa [D] using squaredDifference data))
(Xa data a)).nonneg
simpa [Matrix.mul_sub, Matrix.sub_mul, Matrix.mul_assoc] using hpsd
calc
first a ≤ ev ψ ((Xa data a)ᴴ *
((data.x * data.xᴴ - 1) * (data.x * data.xᴴ - 1)) *
Xa data a) := by
dsimp [first]
exact ev_mono ψ _ _ hrect
_ = ev ψ (Qa data.qLayer a * QTotal data.qLayer * Qa data.qLayer a -
Qa data.qLayer a) := by
rw [xExpressionToQExpression data a]
calc
(∑ a : Outcome, first a)
≤ ∑ a : Outcome,
ev ψ (Qa data.qLayer a * QTotal data.qLayer * Qa data.qLayer a -
Qa data.qLayer a) := by
exact Finset.sum_le_sum fun a _ => hpoint a
_ = ev ψ (∑ a : Outcome,
(Qa data.qLayer a * QTotal data.qLayer * Qa data.qLayer a -
Qa data.qLayer a)) := by
exact (ev_sum ψ (fun a : Outcome =>
Qa data.qLayer a * QTotal data.qLayer * Qa data.qLayer a -
Qa data.qLayer a)).symm
_ ≤ ev ψ ((((4 : Error) * spectralTruncationError ζ) : ℂ) •
(1 : MIPStarRE.Quantum.Op ι)) := by
exact ev_mono ψ _ _ (qAlmostProjective ψ A ζ data.qLayer hζ hζ_small hRank)
_ = 4 * spectralTruncationError ζ := by
simpa [Complex.ofReal_mul, ev_one_of_isNormalized ψ hψ] using
ev_scale ψ (4 * spectralTruncationError ζ)
(1 : MIPStarRE.Quantum.Op ι)
have hsecond_le : (∑ a : Outcome, second a) ≤ 1 := by
calc
(∑ a : Outcome, second a)
= ∑ a : Outcome, ev ψ (Pa data a) := by
refine Finset.sum_congr rfl ?_
intro a _
simp [second, pa_hermitian data a, pa_idempotent data a]
_ ≤ 1 := pa_mass_le_one data ψ hψ
have hhat_ev :
(∑ a : Outcome, ev ψ ((Xa data a)ᴴ * data.xHat * Pa data a)) =
ev ψ (CFC.sqrt (QTotal data.qLayer)) := by
calc
(∑ a : Outcome, ev ψ ((Xa data a)ᴴ * data.xHat * Pa data a))
= ev ψ (∑ a : Outcome, (Xa data a)ᴴ * data.xHat * Pa data a) := by
exact (ev_sum ψ (fun a : Outcome =>
(Xa data a)ᴴ * data.xHat * Pa data a)).symm
_ = ev ψ (CFC.sqrt (QTotal data.qLayer)) := by
rw [xHat_cross_sum_eq_sqrt data]
have hdiff_sum :
((∑ a : Outcome, ev ψ (Qa data.qLayer a * Pa data a)) -
∑ a : Outcome, ev ψ ((Xa data a)ᴴ * data.xHat * Pa data a)) =
∑ a : Outcome, ev ψ (((Xa data a)ᴴ * D) * Pa data a) := by
rw [← Finset.sum_sub_distrib]
refine Finset.sum_congr rfl ?_
intro a _
calc
ev ψ (Qa data.qLayer a * Pa data a) -
ev ψ ((Xa data a)ᴴ * data.xHat * Pa data a)
= ev ψ (Qa data.qLayer a * Pa data a -
(Xa data a)ᴴ * data.xHat * Pa data a) := by
rw [← ev_sub]
_ = ev ψ (((Xa data a)ᴴ * D) * Pa data a) := by
congr 1
rw [(qaRestated data a).2.2]
dsimp [D]
rw [← Matrix.sub_mul, ← Matrix.mul_sub]
have hdiff_abs :
|(∑ a : Outcome, ev ψ (Qa data.qLayer a * Pa data a)) -
∑ a : Outcome, ev ψ ((Xa data a)ᴴ * data.xHat * Pa data a)| ≤
Real.sqrt (∑ a : Outcome, first a) * Real.sqrt (∑ a : Outcome, second a) := by
calc
|(∑ a : Outcome, ev ψ (Qa data.qLayer a * Pa data a)) -
∑ a : Outcome, ev ψ ((Xa data a)ᴴ * data.xHat * Pa data a)|
= |∑ a : Outcome, ev ψ (((Xa data a)ᴴ * D) * Pa data a)| := by
rw [hdiff_sum]
_ ≤ ∑ a : Outcome, |ev ψ (((Xa data a)ᴴ * D) * Pa data a)| :=
Finset.abs_sum_le_sum_abs _ _
_ ≤ ∑ a : Outcome, Real.sqrt (first a) * Real.sqrt (second a) := by
refine Finset.sum_le_sum ?_
intro a _
dsimp [first, second]
simpa [D, Matrix.conjTranspose_mul, Matrix.mul_assoc] using
ev_abs_mul_le_sqrt ψ ((Xa data a)ᴴ * D) (Pa data a)
_ ≤ Real.sqrt (∑ a : Outcome, first a) *
Real.sqrt (∑ a : Outcome, second a) := by
exact Real.sum_sqrt_mul_sqrt_le (s := Finset.univ)
(f := fun a : Outcome => first a)
(g := fun a : Outcome => second a)
(fun a => hfirst_nonneg a) (fun a => hsecond_nonneg a)
have hsqrt_first : Real.sqrt (∑ a : Outcome, first a) ≤
Real.sqrt (4 * spectralTruncationError ζ) := by
exact Real.sqrt_le_sqrt hfirst_le
have hsqrt_second : Real.sqrt (∑ a : Outcome, second a) ≤ 1 := by
have h := Real.sqrt_le_sqrt hsecond_le
simpa using h
calc
|(∑ a : Outcome, ev ψ (Qa data.qLayer a * Pa data a)) -
ev ψ (CFC.sqrt (QTotal data.qLayer))|
= |(∑ a : Outcome, ev ψ (Qa data.qLayer a * Pa data a)) -
∑ a : Outcome, ev ψ ((Xa data a)ᴴ * data.xHat * Pa data a)| := by
rw [hhat_ev]
_ ≤ Real.sqrt (∑ a : Outcome, first a) *
Real.sqrt (∑ a : Outcome, second a) := hdiff_abs
_ ≤ Real.sqrt (4 * spectralTruncationError ζ) * 1 := by
exact mul_le_mul hsqrt_first hsqrt_second
(Real.sqrt_nonneg _) (Real.sqrt_nonneg _)
_ = 2 * zetaQuarterRoot ζ := by
have hsqrt_rpow :
Real.sqrt (ζ ^ (1 / (2 : Error))) = zetaQuarterRoot ζ := by
rw [Real.sqrt_eq_rpow, zetaQuarterRoot, ← Real.rpow_mul hζ]
congr 1
ring
dsimp [spectralTruncationError]
rw [Real.sqrt_mul (by positivity : 0 ≤ (4 : Error)), hsqrt_rpow]
ring
/-- **`P` is close to `Q`** (`lem:P-Q-approx`).
The final internal comparison in the paper's repair step is derived from the
primitive `X/XHat/P` identities in `QXPLayerData`, the rank-reduction witness,
and the standard small-error hypotheses. No closeness bound is stored inside
`QXPLayerData`; the proof below follows the paper's expansion through
`squaredDifference`, `qAlmostProjective`, and `sqrtQCompleteness`. -/
lemma pQApprox {Outcome : Type*}
{ι : Type*} [Fintype ι] [DecidableEq ι]
[Fintype Outcome]
(ψ : QuantumState ι)
(A : Measurement Outcome ι) (ζ : Error)
(data : QXPLayerData Outcome ι)
(hψ : ψ.IsNormalized)
(hζ : 0 ≤ ζ) (hζ_small : ζ ≤ 1 / (4 : Error)) :
RankReductionWitness ψ A ζ data.qLayer →
SDDOpRel ψ (uniformDistribution Unit)
(constOpFamily data.qLayer.q)
(constOpFamily (PFamily data))
(30 * zetaQuarterRoot ζ) := by
intro hRank
let S : Error := ∑ a : Outcome, ev ψ (Qa data.qLayer a * Pa data a)
have hcross_close : |S - ev ψ (CFC.sqrt (QTotal data.qLayer))| ≤
2 * zetaQuarterRoot ζ := by
simpa [S] using q_p_cross_close ψ A ζ data hψ hζ hζ_small hRank
have hsqrt_complete : ev ψ (CFC.sqrt (QTotal data.qLayer)) ≥
1 - 12 * zetaQuarterRoot ζ :=
sqrtQCompleteness ψ A ζ data.qLayer hψ hζ hζ_small hRank
have hS_lower : S ≥ 1 - 14 * zetaQuarterRoot ζ := by
have hleft := (abs_le.mp hcross_close).1
nlinarith
have hq_mass : ev ψ (QTotal data.qLayer) ≤
1 + 2 * spectralTruncationError ζ := by
calc
ev ψ (QTotal data.qLayer)
≤ ev ψ ((((1 : Error) + 2 * spectralTruncationError ζ) : ℂ) •
(1 : MIPStarRE.Quantum.Op ι)) := ev_mono ψ _ _ hRank.total_le
_ = 1 + 2 * spectralTruncationError ζ := by
simpa [ev_one_of_isNormalized ψ hψ] using
ev_scale ψ (1 + 2 * spectralTruncationError ζ)
(1 : MIPStarRE.Quantum.Op ι)
have hp_mass : (∑ a : Outcome, ev ψ (Pa data a)) ≤ 1 :=
pa_mass_le_one data ψ hψ
have hcross_symm :
(∑ a : Outcome, ev ψ (Pa data a * Qa data.qLayer a)) = S := by
dsimp [S]
refine Finset.sum_congr rfl ?_
intro a _
exact ev_mul_comm_of_psd ψ (Pa data a) (Qa data.qLayer a)
(pa_nonneg data a) (hRank.outcome_nonneg a)
have hq_sum :
(∑ a : Outcome, ev ψ (Qa data.qLayer a)) = ev ψ (QTotal data.qLayer) := by
calc
(∑ a : Outcome, ev ψ (Qa data.qLayer a))
= ev ψ (∑ a : Outcome, Qa data.qLayer a) := by
exact (ev_sum ψ (Qa data.qLayer)).symm
_ = ev ψ (QTotal data.qLayer) := by rw [hRank.sum_eq_total]
have hqsddeq :
qSDDOp ψ data.qLayer.q (PFamily data) =
ev ψ (QTotal data.qLayer) + (∑ a : Outcome, ev ψ (Pa data a)) -
S - S := by
unfold qSDDOp qSDDCore
calc
(∑ a : Outcome,
ev ψ (((data.qLayer.q.outcome a - (PFamily data).outcome a)ᴴ) *
(data.qLayer.q.outcome a - (PFamily data).outcome a)))
= ∑ a : Outcome,
(ev ψ (Qa data.qLayer a) + ev ψ (Pa data a) -
ev ψ (Qa data.qLayer a * Pa data a) -
ev ψ (Pa data a * Qa data.qLayer a)) := by
refine Finset.sum_congr rfl ?_
intro a _
have hQaH : (Qa data.qLayer a)ᴴ = Qa data.qLayer a :=
(hRank.projective a).isSelfAdjoint.isHermitian.eq
have hPaH : (Pa data a)ᴴ = Pa data a := pa_hermitian data a
have hQaSq : Qa data.qLayer a * Qa data.qLayer a = Qa data.qLayer a :=
(hRank.projective a).isIdempotentElem.eq
have hPaSq : Pa data a * Pa data a = Pa data a := pa_idempotent data a
calc
ev ψ (((data.qLayer.q.outcome a - (PFamily data).outcome a)ᴴ) *
(data.qLayer.q.outcome a - (PFamily data).outcome a))
= ev ψ (((Qa data.qLayer a - Pa data a)ᴴ) *
(Qa data.qLayer a - Pa data a)) := by
rfl
_ = ev ψ ((Qa data.qLayer a + Pa data a) -
Qa data.qLayer a * Pa data a -
Pa data a * Qa data.qLayer a) := by
congr 1
rw [Matrix.conjTranspose_sub, hQaH, hPaH]
calc
(Qa data.qLayer a - Pa data a) *
(Qa data.qLayer a - Pa data a)
= Qa data.qLayer a * Qa data.qLayer a -
Qa data.qLayer a * Pa data a -
Pa data a * Qa data.qLayer a +
Pa data a * Pa data a := by
noncomm_ring
_ = (Qa data.qLayer a + Pa data a) -
Qa data.qLayer a * Pa data a -
Pa data a * Qa data.qLayer a := by
rw [hQaSq, hPaSq]
noncomm_ring
_ = ev ψ (Qa data.qLayer a) + ev ψ (Pa data a) -
ev ψ (Qa data.qLayer a * Pa data a) -
ev ψ (Pa data a * Qa data.qLayer a) := by
rw [ev_sub, ev_sub, ev_add]
_ = (∑ a : Outcome, ev ψ (Qa data.qLayer a)) +
(∑ a : Outcome, ev ψ (Pa data a)) -
(∑ a : Outcome, ev ψ (Qa data.qLayer a * Pa data a)) -
(∑ a : Outcome, ev ψ (Pa data a * Qa data.qLayer a)) := by
simp [Finset.sum_add_distrib, Finset.sum_sub_distrib]
_ = ev ψ (QTotal data.qLayer) + (∑ a : Outcome, ev ψ (Pa data a)) -
S - S := by
rw [hq_sum, hcross_symm]
have hqSDD_bound : qSDDOp ψ data.qLayer.q (PFamily data) ≤
30 * zetaQuarterRoot ζ := by
calc
qSDDOp ψ data.qLayer.q (PFamily data)
= ev ψ (QTotal data.qLayer) + (∑ a : Outcome, ev ψ (Pa data a)) -
S - S := hqsddeq
_ ≤ (1 + 2 * spectralTruncationError ζ) + 1 -
(1 - 14 * zetaQuarterRoot ζ) -
(1 - 14 * zetaQuarterRoot ζ) := by
nlinarith
_ = 2 * spectralTruncationError ζ + 28 * zetaQuarterRoot ζ := by ring
_ ≤ 30 * zetaQuarterRoot ζ := by
have hε_le := spectralTruncationError_le_zetaQuarterRoot ζ hζ hζ_small
nlinarith
constructor
simpa [sddErrorOp, avgOver, uniformDistribution, constOpFamily] using hqSDD_bound
/-- Apply `lem:P-Q-approx` to the positive-Gram sigma-space QXP layer.
This is the constructor-facing form of the local `Q -> X -> Xhat -> P` stage.
The rank-reduction witness supplies the sigma-space `X`; the positive-Gram
polar construction supplies `Xhat`; and the theorem concludes the paper's
`P`-versus-`Q` approximation without any external SVD data. -/
lemma pQApprox_ofRankReductionSigmaRangePositiveGram
{Outcome : Type uOutcome} [Fintype Outcome] [DecidableEq Outcome]
{ι : Type uι} [Fintype ι] [DecidableEq ι]
(ψ : QuantumState ι)
(A : Measurement Outcome ι) (ζ : Error)
{qLayer : QLayerData Outcome ι}
(hRank : RankReductionWitness ψ A ζ qLayer)
[Nonempty (FiniteHilbertSpace.sigmaFinCarrier
(fun a : Outcome => (qLayer.q.outcome a).rank))]
(hψ : ψ.IsNormalized)
(hζ : 0 ≤ ζ) (hζ_small : ζ ≤ 1 / (4 : Error)) :
∃ xHat : Matrix (sigmaRangeCarrier qLayer.q) ι ℂ,
xHat * xHatᴴ =
(1 : MIPStarRE.Quantum.Op (sigmaRangeCarrier qLayer.q)) ∧
(sigmaFinRangeEmbedding qLayer.q.outcome hRank.projective)ᴴ * xHat =
CFC.sqrt (QTotal qLayer) ∧
∃ data : QXPLayerData Outcome ι,
∃ hq : data.qLayer = sigmaRangeQLayer qLayer.q,
hq ▸ data.x =
(show Matrix (sigmaRangeQLayer qLayer.q).auxSpace.carrier ι ℂ from
sigmaFinRangeEmbedding qLayer.q.outcome hRank.projective) ∧
hq ▸ data.xHat =
(show Matrix (sigmaRangeQLayer qLayer.q).auxSpace.carrier ι ℂ from
xHat) ∧
SDDOpRel ψ (uniformDistribution Unit)
(constOpFamily data.qLayer.q)
(constOpFamily (PFamily data))
(30 * zetaQuarterRoot ζ) := by
obtain ⟨xHat, hxHat_coisometry, hxHat_mixed, data, hq, hx, hxHat⟩ :=
exists_qxpLayerData_ofRankReductionSigmaRangePositiveGram hRank
refine ⟨xHat, hxHat_coisometry, hxHat_mixed, data, hq, hx, hxHat, ?_⟩
exact pQApprox ψ A ζ data hψ hζ hζ_small
(hq.symm ▸ hRank.toSigmaRangeQLayer)
end
end MIPStarRE.LDT.MakingMeasurementsProjective
end
end LIDT
Source