roseking's picture
Sync from openai/math@faee2cf485a8 (part 23)
607ee45 verified
Raw History Blame Contribute Delete
30.1 kB
import Mathlib
import OAI.RepresentationTheory.Unitary.Basic
namespace OAI
noncomputable section
open scoped BigOperators Matrix.Norms.L2Operator ComplexOrder
attribute [local instance] Classical.propDecidable
namespace CoordinateSweeps
namespace SignedTensor
/- p even basis symbols followed by p odd ones, exactly E⊕O in05. -/
abbrev Letter (p : ℕ) := Fin (2*p)
abbrev Word (p l : ℕ) := Fin l → Letter p
def odd {p : ℕ} (a : Letter p) : Prop := p ≤ a.val
/- The sign from reordering the odd factors, with positions ordered as in
ordinary tensor words. There is no choice of orientation hidden here. -/
def koszulExponent {p l : ℕ} (σ : Equiv.Perm (Fin l)) (w : Word p l) : ℕ :=
((Equiv.Perm.finPairsLT l).filter (fun ij =>
odd (w ij.1) ∧ odd (w ij.2) ∧ σ ij.1 < σ ij.2)).card
def phase {p l : ℕ} (σ : Equiv.Perm (Fin l)) (w : Word p l) : ℂ :=
(-1)^koszulExponent σ w
/- The underlying ordinary slot permutation; gh applies h first. -/
def wordPerm {p l : ℕ} (σ : Equiv.Perm (Fin l)) : Equiv.Perm (Word p l) where
toFun w i := w (σ⁻¹ i)
invFun w i := w (σ i)
left_inv w := by funext i; simp
right_inv w := by funext i; simp
/- Matrix of the exact parity-signed action in the tensor word basis. -/
def matrix {p l : ℕ} (σ : Equiv.Perm (Fin l)) : Matrix (Word p l) (Word p l) ℂ :=
fun u w => if u=wordPerm σ w then phase σ w else 0
lemma phase_star_mul {p l : ℕ} (σ : Equiv.Perm (Fin l)) (w : Word p l) :
star (phase σ w)*phase σ w=1 := by
simp only [phase,star_pow,star_neg,star_one]
rw [← mul_pow]
simp
lemma phase_prod {p l : ℕ} (σ : Equiv.Perm (Fin l)) (w : Word p l) :
phase σ w = ∏ ij ∈ Equiv.Perm.finPairsLT l,
(if odd (w ij.1) ∧ odd (w ij.2) ∧ σ ij.1 < σ ij.2 then (-1 : ℂ) else 1) := by
rw [Finset.prod_ite]
simp [phase,koszulExponent]
lemma wordPerm_mul {p l : ℕ} (σ τ : Equiv.Perm (Fin l)) :
wordPerm (p := p) (σ*τ) = wordPerm σ * wordPerm τ := by
ext w i
simp [wordPerm]
/- The exact Koszul cocycle, proved from pair inversions, not an action axiom. -/
theorem phase_mul {p l : ℕ} (σ τ : Equiv.Perm (Fin l)) (w : Word p l) :
phase (σ*τ) w = phase σ (wordPerm τ w) * phase τ w := by
rw [phase_prod,phase_prod,phase_prod]
have hreindex :
(∏ ij ∈ Equiv.Perm.finPairsLT l,
(if odd (wordPerm τ w ij.1) ∧ odd (wordPerm τ w ij.2) ∧ σ ij.1 < σ ij.2
then (-1 : ℂ) else 1)) =
∏ ij ∈ Equiv.Perm.finPairsLT l,
(if odd (wordPerm τ w (Equiv.Perm.signBijAux τ ij).1) ∧
odd (wordPerm τ w (Equiv.Perm.signBijAux τ ij).2) ∧
σ (Equiv.Perm.signBijAux τ ij).1 < σ (Equiv.Perm.signBijAux τ ij).2
then (-1 : ℂ) else 1) := by
symm
exact Finset.prod_nbij (Equiv.Perm.signBijAux τ)
Equiv.Perm.signBijAux_mem Equiv.Perm.signBijAux_injOn
Equiv.Perm.signBijAux_surj (fun _ _ => rfl)
rw [hreindex,← Finset.prod_mul_distrib]
apply Finset.prod_congr rfl
rintro ⟨i,j⟩ hij
have hji : j < i := Equiv.Perm.mem_finPairsLT.mp hij
have hne : τ j ≠ τ i := τ.injective.ne hji.ne
have hneσ : σ (τ j) ≠ σ (τ i) := σ.injective.ne hne
by_cases hi : odd (w i)
· by_cases hj : odd (w j)
· by_cases ht : τ j < τ i
· simp [Equiv.Perm.signBijAux,wordPerm,Equiv.Perm.mul_apply,ht,hi,hj,ht.not_gt]
· have ht' : τ i < τ j := lt_of_le_of_ne (le_of_not_gt ht) hne.symm
by_cases hs : σ (τ i) < σ (τ j)
· simp [Equiv.Perm.signBijAux,wordPerm,Equiv.Perm.mul_apply,ht,ht',hi,hj,hs,hs.not_gt]
· have hs' : σ (τ j) < σ (τ i) := lt_of_le_of_ne (le_of_not_gt hs) hneσ
simp [Equiv.Perm.signBijAux,wordPerm,Equiv.Perm.mul_apply,ht,ht',hi,hj,hs,hs']
· by_cases ht : τ j < τ i <;>
simp [Equiv.Perm.signBijAux,wordPerm,Equiv.Perm.mul_apply,ht,hi,hj]
· by_cases ht : τ j < τ i <;>
simp [Equiv.Perm.signBijAux,wordPerm,Equiv.Perm.mul_apply,ht,hi]
/- Unitarity follows directly, without assuming the Schur-Weyl
or Pieri decomposition. -/
theorem matrix_unitary {p l : ℕ} (σ : Equiv.Perm (Fin l)) :
(matrix (p := p) σ).conjTranspose * matrix σ=1 := by
ext u w
simp only [Matrix.mul_apply,Matrix.conjTranspose_apply,matrix]
rw [Finset.sum_eq_single (wordPerm σ u)]
· by_cases hu : u=w
· subst w
simp only [ite_true,Matrix.one_apply_eq,phase_star_mul]
· have hne : wordPerm σ u ≠ wordPerm σ w :=
fun hh => hu ((wordPerm σ).injective hh)
simp [hne,hu]
· intro v hv hne
simp [hne]
· simp
lemma phase_one {p l : ℕ} (w : Word p l) : phase (1 : Equiv.Perm (Fin l)) w=1 := by
rw [phase_prod]
apply Finset.prod_eq_one
intro ij hij
have hji := Equiv.Perm.mem_finPairsLT.mp hij
simp [Equiv.Perm.one_apply,hji.not_gt]
theorem matrix_one (p l : ℕ) : matrix (p := p) (1 : Equiv.Perm (Fin l))=1 := by
ext u w
simp [matrix,wordPerm,phase_one,Matrix.one_apply]
theorem matrix_mul {p l : ℕ} (σ τ : Equiv.Perm (Fin l)) :
matrix (p := p) (σ*τ)=matrix σ * matrix τ := by
ext u w
simp only [Matrix.mul_apply,matrix]
rw [Finset.sum_eq_single (wordPerm τ w)]
· simp only [ite_true,wordPerm_mul,Equiv.Perm.mul_apply,phase_mul]
split_ifs <;> simp
· intro v hv hne
simp [hne]
· simp
/- Exact signed tensor representation of the symmetric group. -/
def action (p l : ℕ) : Equiv.Perm (Fin l) →* Matrix (Word p l) (Word p l) ℂ where
toFun := matrix
map_one' := matrix_one p l
map_mul' := matrix_mul
/- Ordinary (not graded) tensor power of a one-site density, as in05. -/
def densityPower {p : ℕ} (l : ℕ) (r : Matrix (Letter p) (Letter p) ℂ) :
Matrix (Word p l) (Word p l) ℂ := fun u w => ∏ i, r (u i) (w i)
lemma densityPower_mul {p : ℕ} (l : ℕ) (a b : Matrix (Letter p) (Letter p) ℂ) :
densityPower l (a*b)=densityPower l a * densityPower l b := by
ext u w
simp only [densityPower,Matrix.mul_apply]
rw [Fintype.prod_sum]
apply Finset.sum_congr rfl
intro v hv
exact Finset.prod_mul_distrib
lemma densityPower_star {p : ℕ} (l : ℕ) (r : Matrix (Letter p) (Letter p) ℂ) :
densityPower l r.conjTranspose=(densityPower l r).conjTranspose := by
ext u w
simp [densityPower,Matrix.conjTranspose_apply]
lemma densityPower_trace {p : ℕ} (l : ℕ) (r : Matrix (Letter p) (Letter p) ℂ) :
Matrix.trace (densityPower l r)=(Matrix.trace r)^l := by
change (∑ w : Word p l, ∏ i, r (w i) (w i))=(∑ i, r i i)^l
exact (Fintype.sum_pow (fun i : Letter p => r i i) l).symm
open scoped MatrixOrder in
lemma densityPower_posSemidef {p : ℕ} (l : ℕ) (r : Matrix (Letter p) (Letter p) ℂ)
(hr : r.PosSemidef) : (densityPower l r).PosSemidef := by
obtain ⟨b,rfl⟩ := CStarAlgebra.nonneg_iff_eq_star_mul_self.mp hr.nonneg
rw [densityPower_mul,Matrix.star_eq_conjTranspose,densityPower_star]
exact Matrix.posSemidef_conjTranspose_mul_self _
/- Even means positive trace-one and block diagonal for E,O, exactly as in05. -/
def EvenDensity (p : ℕ) := {r : Matrix (Letter p) (Letter p) ℂ //
r.PosSemidef ∧ Matrix.trace r=1 ∧
∀ i j, (odd i ↔ ¬ odd j) → r i j=0}
instance (p : ℕ) : MeasurableSpace (EvenDensity p) :=
MeasurableSpace.comap Subtype.val (borel (Matrix (Letter p) (Letter p) ℂ))
lemma phase_eq_of_parity {p l : ℕ} (σ : Equiv.Perm (Fin l))
(u w : Word p l) (h : ∀ i, odd (u i) ↔ odd (w i)) : phase σ u=phase σ w := by
simp only [phase,koszulExponent]
congr 2
apply Finset.filter_congr
intro ij hij
rw [h ij.1,h ij.2]
lemma densityPower_parity_of_ne_zero {p l : ℕ} (r : EvenDensity p)
{u w : Word p l} (h : densityPower l r.val u w ≠ 0) :
∀ i, odd (u i) ↔ odd (w i) := by
intro i
have he : r.val (u i) (w i) ≠ 0 :=
(Finset.prod_ne_zero_iff.mp h) i (Finset.mem_univ _)
by_contra hn
apply he
apply r.property.2.2
tauto
lemma densityPower_permute {p l : ℕ} (σ : Equiv.Perm (Fin l))
(r : Matrix (Letter p) (Letter p) ℂ) (u w : Word p l) :
densityPower l r (wordPerm σ u) (wordPerm σ w)=densityPower l r u w := by
simp only [densityPower,wordPerm,Equiv.coe_fn_mk]
exact Equiv.prod_comp σ⁻¹ (fun i => r (u i) (w i))
/- Even densities commute with the signed action; omission of evenness here
would be false. This is also the block-reordering claim at the end of05. -/
theorem densityPower_commute {p l : ℕ} (r : EvenDensity p)
(σ : Equiv.Perm (Fin l)) :
matrix σ * densityPower l r.val=densityPower l r.val * matrix σ := by
ext u w
simp only [Matrix.mul_apply,matrix]
rw [Finset.sum_eq_single ((wordPerm σ).symm u),
Finset.sum_eq_single (wordPerm σ w)]
· simp only [(wordPerm σ).apply_symm_apply,ite_true]
have hd : densityPower l r.val u (wordPerm σ w)=
densityPower l r.val ((wordPerm σ).symm u) w := by
simpa using densityPower_permute σ r.val ((wordPerm σ).symm u) w
rw [hd]
by_cases hn : densityPower l r.val ((wordPerm σ).symm u) w=0
· simp [hn]
· rw [phase_eq_of_parity σ _ _ (densityPower_parity_of_ne_zero r hn),mul_comm]
· intro v hv hne
simp [hne]
· simp
· intro v hv hne
have hh : u ≠ wordPerm σ v := by
intro he
apply hne
exact (Equiv.eq_symm_apply _).mpr he.symm
simp [hh]
· simp
lemma evenDensityPower_trace {p : ℕ} (l : ℕ) (r : EvenDensity p) :
Matrix.trace (densityPower l r.val)=1 := by
rw [densityPower_trace,r.property.2.1,one_pow]
/- The standard central character formula for the isotypic projection. Its
projection properties require the signed action laws (not assumed here). -/
def isotypic {p l : ℕ} (ρ : UnitaryIrrep (Equiv.Perm (Fin l))) :
Matrix (Word p l) (Word p l) ℂ :=
((ρ.dimension : ℂ)/(Fintype.card (Equiv.Perm (Fin l)) : ℂ)) •
∑ σ, Matrix.trace (ρ.matrix σ⁻¹) • matrix σ
/- Exact signed-density domination05 eq13, expressed for every irreducible
(including absent ones whose projection is zero). The mixture is a genuine
probability measure on the source's even density matrices, not an unproved
positive operator supplied as a hypothesis. -/
def Domination : Prop :=
∃ C : ℝ, 0 < C ∧ ∀ s p : ℕ, 1 ≤ p → p ≤ s → ∀ l : ℕ, l ≤ s →
∀ ρ : UnitaryIrrep (Equiv.Perm (Fin l)),
∃ μ : MeasureTheory.Measure (EvenDensity p), MeasureTheory.IsProbabilityMeasure μ ∧
(Complex.ofReal (Real.exp (Real.log ρ.dimension+C*p^2*Real.log (s+1))) •
(∫ r : EvenDensity p, densityPower l r.val ∂μ) - isotypic (p := p) ρ).PosSemidef
end SignedTensor
end CoordinateSweeps
/- A direct counting approach to the polynomial multiplicity bound in05,
without replacing Schur-Weyl or the density inequality by an axiom.
The signed tensor commutant is determined by pair-word histograms. -/
namespace CoordinateSweeps.SignedTensor
lemma phase_ne_zero {p l : ℕ} (σ : Equiv.Perm (Fin l)) (w : Word p l) : phase σ w ≠ 0 := by
intro h
have hh := phase_star_mul σ w
simp [h] at hh
lemma mul_signed_entry {p l : ℕ} (A : Matrix (Word p l) (Word p l) ℂ)
(σ : Equiv.Perm (Fin l)) (u w : Word p l) :
(A*matrix (p := p) σ) u w=A u (wordPerm σ w)*phase σ w := by
simp only [Matrix.mul_apply,matrix]
rw [Finset.sum_eq_single (wordPerm σ w)]
· simp
· intro v hv hne; simp [hne]
· simp
lemma signed_mul_entry {p l : ℕ} (A : Matrix (Word p l) (Word p l) ℂ)
(σ : Equiv.Perm (Fin l)) (u w : Word p l) :
(matrix (p := p) σ*A) (wordPerm σ u) w=phase σ u*A u w := by
simp only [Matrix.mul_apply,matrix]
rw [Finset.sum_eq_single u]
· simp
· intro v hv hne
have hne' : wordPerm σ u ≠ wordPerm σ v :=
fun hh => hne (((wordPerm σ).injective hh).symm)
simp [hne']
· simp
/- The actual signed-action commutant as a linear subspace of matrices. -/
def commutant (p l : ℕ) : Submodule ℂ (Matrix (Word p l) (Word p l) ℂ) where
carrier := {A | ∀ σ : Equiv.Perm (Fin l), A*matrix σ=matrix σ*A}
zero_mem' := by simp
add_mem' hA hB := by intro σ; simp only [add_mul,mul_add,hA σ,hB σ]
smul_mem' c A hA := by intro σ; simp only [smul_mul_assoc,mul_smul_comm,hA σ]
lemma commutant_zero_transport {p l : ℕ} (A : commutant p l)
(σ : Equiv.Perm (Fin l)) {u w : Word p l} (h : A.val u w=0) :
A.val (wordPerm σ u) (wordPerm σ w)=0 := by
have hh := congrArg (fun B : Matrix (Word p l) (Word p l) ℂ => B (wordPerm σ u) w)
(A.property σ)
rw [mul_signed_entry,signed_mul_entry,h,mul_zero] at hh
exact (mul_eq_zero.mp hh).resolve_right (phase_ne_zero σ w)
abbrev Profile (p l : ℕ) := (Letter p × Letter p) → Fin (l+1)
def profile {p l : ℕ} (u w : Word p l) : Profile p l := fun a =>
⟨(Finset.univ.filter (fun i => (u i,w i)=a)).card,
Nat.lt_succ_iff.mpr (by
simpa using Finset.card_filter_le (Finset.univ : Finset (Fin l))
(fun i => (u i,w i)=a))⟩
lemma profile_fiber_card {p l : ℕ} (u w : Word p l) (a : Letter p × Letter p) :
Fintype.card {i : Fin l // (u i,w i)=a}=(profile u w a).val := by
simp [profile,Fintype.card_subtype]
/- Two word pairs have the same histogram exactly when a position permutation
carries one to the other. This supplies the finite orbit-counting parameter. -/
lemma profile_eq_exists_perm {p l : ℕ} {u w u' w' : Word p l}
(hh : profile u w=profile u' w') :
∃ σ : Equiv.Perm (Fin l), wordPerm σ u=u' ∧ wordPerm σ w=w' := by
have hc (a : Letter p × Letter p) : Fintype.card {i : Fin l // (u i,w i)=a}=
Fintype.card {i : Fin l // (u' i,w' i)=a} := by
rw [profile_fiber_card,profile_fiber_card,hh]
let e (a : Letter p × Letter p) : {i : Fin l // (u i,w i)=a} ≃
{i : Fin l // (u' i,w' i)=a} := Fintype.equivOfCardEq (hc a)
let σ : Equiv.Perm (Fin l) := Equiv.ofFiberEquiv e
have he (i : Fin l) : (u' (σ i),w' (σ i))=(u i,w i) := Equiv.ofFiberEquiv_map e i
refine ⟨σ,?_,?_⟩
· funext i
have hi := congrArg Prod.fst (he (σ⁻¹ i))
simpa [wordPerm] using hi.symm
· funext i
have hi := congrArg Prod.snd (he (σ⁻¹ i))
simpa [wordPerm] using hi.symm
abbrev OccurringProfile (p l : ℕ) :=
{q : Profile p l // q ∈ Set.range (fun uw : Word p l × Word p l => profile uw.1 uw.2)}
def profileRep {p l : ℕ} (q : OccurringProfile p l) : Word p l × Word p l :=
Classical.choose q.property
lemma profileRep_spec {p l : ℕ} (q : OccurringProfile p l) :
profile (profileRep q).1 (profileRep q).2=q.val := Classical.choose_spec q.property
def commutantEval (p l : ℕ) : commutant p l →ₗ[ℂ] (OccurringProfile p l → ℂ) where
toFun A q := A.val (profileRep q).1 (profileRep q).2
map_add' _ _ := rfl
map_smul' _ _ := rfl
lemma commutantEval_injective (p l : ℕ) : Function.Injective (commutantEval p l) := by
apply (LinearMap.ker_eq_bot).mp
apply LinearMap.ker_eq_bot'.mpr
intro A hA
apply Subtype.ext
ext u w
let q : OccurringProfile p l := ⟨profile u w,Set.mem_range_self (u,w)⟩
have he : A.val (profileRep q).1 (profileRep q).2=0 := congrFun hA q
have hp : profile (profileRep q).1 (profileRep q).2=profile u w := profileRep_spec q
obtain ⟨σ,hσu,hσw⟩ := profile_eq_exists_perm hp
have hh := commutant_zero_transport A σ he
simpa [hσu,hσw] using hh
/- Polynomial signed-commutant bound of exactly the O(p² log(s+1)) scale
required by the main induction. No classification of partitions is used. -/
theorem commutant_finrank_le (p l : ℕ) :
Module.finrank ℂ (commutant p l) ≤ (l+1)^((2*p)*(2*p)) := by
have h := LinearMap.finrank_le_finrank_of_injective (commutantEval_injective p l)
have hc : Fintype.card (OccurringProfile p l) ≤ Fintype.card (Profile p l) :=
Fintype.card_subtype_le _
have he : Fintype.card (Profile p l)=(l+1)^((2*p)*(2*p)) := by
simp [Profile,Letter]
simpa [he] using h.trans (by simpa using hc)
end CoordinateSweeps.SignedTensor
namespace CoordinateSweeps.SignedTensor
lemma matrix_inv {p l : ℕ} (σ : Equiv.Perm (Fin l)) :
matrix (p := p) σ⁻¹=(matrix (p := p) σ).conjTranspose := by
calc
matrix σ⁻¹=1*matrix (p := p) σ⁻¹ := (one_mul _).symm
_ = ((matrix σ).conjTranspose*matrix σ)*matrix (p := p) σ⁻¹ := by rw [matrix_unitary]
_ = (matrix (p := p) σ).conjTranspose := by
rw [mul_assoc,← matrix_mul,mul_inv_cancel,matrix_one,mul_one]
abbrev tensorRepresentation (p l : ℕ) := UnitaryIrrep.matrixRepresentation (action p l)
/- An actual intertwining endomorphism has its coefficient matrix in the
signed commutant. This is not an assumed decomposition of the tensor action. -/
def endomorphismMatrix (p l : ℕ) :
(tensorRepresentation p l).IntertwiningMap (tensorRepresentation p l) →ₗ[ℂ]
commutant p l where
toFun F := ⟨LinearMap.toMatrixAlgEquiv' F.toLinearMap,by
intro σ
apply Matrix.toLinAlgEquiv'.injective
simp only [map_mul,Matrix.toLinAlgEquiv'_toMatrixAlgEquiv']
exact F.isIntertwining' σ⟩
map_add' F H := by
apply Subtype.ext
exact map_add LinearMap.toMatrixAlgEquiv' _ _
map_smul' c F := by
apply Subtype.ext
exact map_smul LinearMap.toMatrixAlgEquiv'.toLinearMap c _
lemma endomorphismMatrix_injective (p l : ℕ) :
Function.Injective (endomorphismMatrix p l) := by
intro F H hh
apply Representation.IntertwiningMap.ext
exact LinearMap.toMatrixAlgEquiv'.injective (congrArg Subtype.val hh)
theorem tensor_endomorphisms_finrank_le (p l : ℕ) :
Module.finrank ℂ ((tensorRepresentation p l).IntertwiningMap (tensorRepresentation p l)) ≤
(l+1)^((2*p)*(2*p)) :=
(LinearMap.finrank_le_finrank_of_injective (endomorphismMatrix_injective p l)).trans
(commutant_finrank_le p l)
/- The polynomial multiplicity clause of05 signed density lemma, proved
without Young diagrams. Multiplicity is the actual space of equivariant embeddings. -/
theorem multiplicity_le (p l : ℕ) (ρ : UnitaryIrrep (Equiv.Perm (Fin l))) :
Module.finrank ℂ (ρ.asRepresentation.IntertwiningMap (tensorRepresentation p l)) ≤
(l+1)^((2*p)*(2*p)) := by
have hh := ρ.multiplicity_sq_le_commutant (action p l)
(fun σ => matrix_inv σ |>.symm)
have hm : Module.finrank ℂ (ρ.asRepresentation.IntertwiningMap (tensorRepresentation p l)) ≤
(Module.finrank ℂ (ρ.asRepresentation.IntertwiningMap (tensorRepresentation p l)))^2 := by
exact Nat.le_self_pow (by omega) _
exact hm.trans (hh.trans (tensor_endomorphisms_finrank_le p l))
lemma isotypic_eq_projector {p l : ℕ} (ρ : UnitaryIrrep (Equiv.Perm (Fin l))) :
isotypic (p := p) ρ=ρ.projector (action p l) := by
simp only [isotypic,UnitaryIrrep.projector,UnitaryIrrep.coefficient,Finset.smul_sum,mul_smul]
rfl
theorem isotypic_posSemidef {p l : ℕ} (ρ : UnitaryIrrep (Equiv.Perm (Fin l))) :
(isotypic (p := p) ρ).PosSemidef := by
rw [isotypic_eq_projector]
exact ρ.projector_posSemidef (action p l) (fun σ => (matrix_inv σ).symm)
theorem isotypic_trace {p l : ℕ} (ρ : UnitaryIrrep (Equiv.Perm (Fin l))) :
Matrix.trace (isotypic (p := p) ρ)=(ρ.dimension : ℂ)*
Module.finrank ℂ (ρ.asRepresentation.IntertwiningMap (tensorRepresentation p l)) := by
rw [isotypic_eq_projector]
simp only [UnitaryIrrep.projector,Matrix.trace_sum,Matrix.trace_smul,
UnitaryIrrep.coefficient,smul_eq_mul]
calc
_ = ((ρ.dimension : ℂ)/(Fintype.card (Equiv.Perm (Fin l)) : ℂ)) *
∑ σ, Matrix.trace (action p l σ)*Matrix.trace (ρ.matrix σ⁻¹) := by
rw [Finset.mul_sum]
apply Finset.sum_congr rfl
intro σ hσ
ring
_ = _ := by
rw [ρ.character_sum_multiplicity (action p l)]
have hN : (Fintype.card (Equiv.Perm (Fin l)) : ℂ) ≠ 0 :=
Nat.cast_ne_zero.mpr Fintype.card_ne_zero
field_simp
end CoordinateSweeps.SignedTensor
/- Mismatched symbols in a purification
obey a Pauli exclusion rule. This is derived from the literal Koszul action. -/
namespace CoordinateSweeps.SignedTensor
lemma swap01_inversion_iff (n : ℕ) (i j : Fin (n+2)) :
(j < i ∧ Equiv.swap (0 : Fin (n+2)) 1 i < Equiv.swap 0 1 j) ↔
i = 1 ∧ j = 0 := by
simp only [Equiv.swap_apply_def]
split_ifs <;> simp_all
omega
lemma phase_swap01 {p n : ℕ} (w : Word p (n+2)) :
phase (Equiv.swap (0 : Fin (n+2)) 1) w =
if odd (w 1) ∧ odd (w 0) then (-1 : ℂ) else 1 := by
have he : ((Equiv.Perm.finPairsLT (n+2)).filter (fun ij =>
odd (w ij.1) ∧ odd (w ij.2) ∧
Equiv.swap (0 : Fin (n+2)) 1 ij.1 < Equiv.swap 0 1 ij.2)) =
if odd (w 1) ∧ odd (w 0) then {⟨1,0⟩} else ∅ := by
ext ⟨i,j⟩
simp only [Finset.mem_filter, Equiv.Perm.mem_finPairsLT]
have hi := swap01_inversion_iff n i j
split_ifs with hw
· simp only [Finset.mem_singleton, Sigma.mk.inj_iff, heq_eq_eq]
constructor
· rintro ⟨hji,hiw,hjw,hs⟩
exact hi.mp ⟨hji,hs⟩
· rintro ⟨rfl,rfl⟩
exact ⟨by simp, hw.1, hw.2, by simp⟩
· simp only [Finset.notMem_empty, iff_false]
rintro ⟨hji,hiw,hjw,hs⟩
obtain ⟨rfl,rfl⟩ := hi.mp ⟨hji,hs⟩
exact hw ⟨hiw,hjw⟩
unfold phase koszulExponent
rw [he]
split_ifs <;> simp
lemma wordPerm_swap01_of_equal {p n : ℕ} (w : Word p (n+2)) (hw : w 0 = w 1) :
wordPerm (Equiv.swap (0 : Fin (n+2)) 1) w = w := by
ext i
simp only [wordPerm, Equiv.coe_fn_mk, Equiv.swap_inv, Equiv.swap_apply_def]
split_ifs <;> simp_all
/- Two repeated, parity-mismatched pair symbols force a zero entry in every
commuting matrix, hence in the canonical purification of an isotypic projector. -/
theorem commutant_entry_zero_adjacent_mismatch {p n : ℕ} (A : commutant p (n+2))
(u w : Word p (n+2)) (hu : u 0 = u 1) (hw : w 0 = w 1)
(hm : odd (u 0) ↔ ¬ odd (w 0)) : A.val u w = 0 := by
let σ := Equiv.swap (0 : Fin (n+2)) 1
have h := congrArg (fun B : Matrix (Word p (n+2)) (Word p (n+2)) ℂ =>
B (wordPerm σ u) w) (A.property σ)
rw [mul_signed_entry, signed_mul_entry,
wordPerm_swap01_of_equal u hu, wordPerm_swap01_of_equal w hw] at h
dsimp [σ] at h
rw [phase_swap01, phase_swap01, ← hu, ← hw] at h
by_cases ho : odd (u 0)
· have hw' : ¬ odd (w 0) := hm.mp ho
simp only [ho, hw', and_self, ↓reduceIte, mul_one, neg_one_mul] at h
linear_combination (1/2 : ℂ)*h
· have hw' : odd (w 0) := Classical.byContradiction (fun hh => ho (hm.mpr hh))
simp only [ho, hw', and_self, ↓reduceIte, one_mul, mul_neg_one] at h
linear_combination (-1/2 : ℂ)*h
/- An explicit two-point transporter, avoiding any unproved transitivity
principle or choice of ordering of the other positions. -/
lemma exists_perm_pair {X : Type*} [DecidableEq X] (a b c d : X)
(hab : a ≠ b) (hcd : c ≠ d) :
∃ g : Equiv.Perm X, g a = c ∧ g b = d := by
let s := Equiv.swap a c
have hs : s b ≠ c := by
intro h
have he : s a = c := by simp [s]
exact hab (s.injective (he.trans h.symm))
refine ⟨Equiv.swap (s b) d * s, ?_, ?_⟩
· simp only [Equiv.Perm.mul_apply]
have he : s a = c := by simp [s]
rw [he, Equiv.swap_apply_of_ne_of_ne hs.symm hcd]
· simp
/- All repeated parity-mismatched pair symbols are impossible, at arbitrary
positions. Distinct mismatched symbols will therefore account for every
exceptional slot in the postselection argument. -/
theorem commutant_entry_zero_repeated_mismatch {p l : ℕ} (A : commutant p l)
(u w : Word p l) (i j : Fin l) (hij : i ≠ j) (hu : u i = u j) (hw : w i = w j)
(hm : odd (u i) ↔ ¬ odd (w i)) : A.val u w = 0 := by
have hl : 2 ≤ l := by
have hn : i.val ≠ j.val := fun h => hij (Fin.ext h)
have hi := i.isLt
have hj := j.isLt
omega
obtain ⟨n, rfl⟩ : ∃ n, l = n+2 := ⟨l-2, by omega⟩
obtain ⟨g,hi,hj⟩ := exists_perm_pair i j (0 : Fin (n+2)) 1 hij (by simp)
let u' := wordPerm g u
let w' := wordPerm g w
have hue (x : Fin (n+2)) : u' (g x) = u x := by simp [u',wordPerm]
have hwe (x : Fin (n+2)) : w' (g x) = w x := by simp [w',wordPerm]
have hui : u' 0 = u i := hi ▸ hue i
have huj : u' 1 = u j := hj ▸ hue j
have hwi : w' 0 = w i := hi ▸ hwe i
have hwj : w' 1 = w j := hj ▸ hwe j
have hz : A.val u' w' = 0 := by
apply commutant_entry_zero_adjacent_mismatch A u' w'
· exact hui.trans (hu.trans huj.symm)
· exact hwi.trans (hw.trans hwj.symm)
· simpa [hui,hwi] using hm
have hz' := commutant_zero_transport A g⁻¹ hz
simpa [u',w',wordPerm] using hz'
end CoordinateSweeps.SignedTensor
namespace CoordinateSweeps.SignedTensor
/- Exceptional slots of a vectorized commuting matrix. -/
def mismatchSlots {p l : ℕ} (u w : Word p l) : Finset (Fin l) :=
Finset.univ.filter (fun i => odd (u i) ↔ ¬ odd (w i))
/- The exclusion bound needed for even-density postselection. The coarse
constant four rather than two is harmless in the source's absolute Cp² error. -/
theorem mismatchSlots_card_le {p l : ℕ} (A : commutant p l)
(u w : Word p l) (hne : A.val u w ≠ 0) :
(mismatchSlots u w).card ≤ (2*p)*(2*p) := by
classical
let f : {i // i ∈ mismatchSlots u w} → Letter p × Letter p := fun i => (u i.val,w i.val)
have hf : Function.Injective f := by
intro i j he
apply Subtype.ext
by_contra hij
apply hne
exact commutant_entry_zero_repeated_mismatch A u w i.val j.val hij
(congrArg Prod.fst he) (congrArg Prod.snd he) (Finset.mem_filter.mp i.property).2
have hc := Fintype.card_le_of_injective f hf
simpa only [Fintype.card_coe, Fintype.card_prod, Fintype.card_fin] using hc
end CoordinateSweeps.SignedTensor
namespace CoordinateSweeps.SignedTensor
/- The actual Koszul regrouping phase from (V⊗V̄)^⊗l to
V^⊗l⊗V̄^⊗l. It will cancel, not be dropped, under partial trace. -/
def regroupPhase {p l : ℕ} (u w : Word p l) : ℂ :=
∏ ij ∈ Equiv.Perm.finPairsLT l,
if odd (u ij.1) ∧ odd (w ij.2) then (-1 : ℂ) else 1
lemma phase_square {p l : ℕ} (σ : Equiv.Perm (Fin l)) (u : Word p l) :
phase σ u * phase σ u = 1 := by
simpa only [phase,star_pow,star_neg,star_one] using phase_star_mul σ u
lemma regroupPhase_square {p l : ℕ} (u w : Word p l) :
regroupPhase u w * regroupPhase u w = 1 := by
unfold regroupPhase
rw [← Finset.prod_mul_distrib]
apply Finset.prod_eq_one
intro ij _
split_ifs <;> norm_num
lemma regroupPhase_star {p l : ℕ} (u w : Word p l) :
star (regroupPhase u w) = regroupPhase u w := by
unfold regroupPhase
rw [star_prod]
apply Finset.prod_congr rfl
intro ij _
split_ifs <;> simp
/- Permutations of the matched slots have no residual super-sign after
literal Koszul regrouping; the positions of mismatched symbols are fixed. -/
theorem regroupPhase_permute {p l : ℕ} (σ : Equiv.Perm (Fin l)) (u w : Word p l)
(hfix : ∀ i, (odd (u i) ↔ ¬ odd (w i)) → σ i = i) :
regroupPhase (wordPerm σ u) (wordPerm σ w) * phase σ u * phase σ w =
regroupPhase u w := by
unfold regroupPhase
rw [phase_prod,phase_prod]
have hreindex :
(∏ ij ∈ Equiv.Perm.finPairsLT l,
(if odd (wordPerm σ u ij.1) ∧ odd (wordPerm σ w ij.2) then (-1 : ℂ) else 1)) =
∏ ij ∈ Equiv.Perm.finPairsLT l,
(if odd (wordPerm σ u (Equiv.Perm.signBijAux σ ij).1) ∧
odd (wordPerm σ w (Equiv.Perm.signBijAux σ ij).2) then (-1 : ℂ) else 1) := by
symm
exact Finset.prod_nbij (Equiv.Perm.signBijAux σ)
Equiv.Perm.signBijAux_mem Equiv.Perm.signBijAux_injOn
Equiv.Perm.signBijAux_surj (fun _ _ => rfl)
rw [hreindex,← Finset.prod_mul_distrib,← Finset.prod_mul_distrib]
apply Finset.prod_congr rfl
rintro ⟨i,j⟩ hij
have hji : j < i := Equiv.Perm.mem_finPairsLT.mp hij
have hne : σ j ≠ σ i := σ.injective.ne hji.ne
by_cases ht : σ j < σ i
· simp [Equiv.Perm.signBijAux,wordPerm,ht,ht.not_gt]
· have ht' : σ i < σ j := lt_of_le_of_ne (le_of_not_gt ht) hne.symm
have hm : ¬ ((odd (u i) ↔ ¬ odd (w i)) ∧ (odd (u j) ↔ ¬ odd (w j))) := by
rintro ⟨hi,hj⟩
have hi' := hfix i hi
have hj' := hfix j hj
rw [hi',hj'] at ht'
exact hji.not_gt ht'
by_cases hui : odd (u i) <;> by_cases huj : odd (u j) <;>
by_cases hwi : odd (w i) <;> by_cases hwj : odd (w j) <;>
simp [Equiv.Perm.signBijAux,wordPerm,ht,ht',hui,huj,hwi,hwj] at hm ⊢
lemma commutant_transport {p l : ℕ} (A : commutant p l) (σ : Equiv.Perm (Fin l))
(u w : Word p l) :
A.val (wordPerm σ u) (wordPerm σ w) = phase σ u * phase σ w * A.val u w := by
have hh := congrArg (fun B : Matrix (Word p l) (Word p l) ℂ => B (wordPerm σ u) w)
(A.property σ)
rw [mul_signed_entry,signed_mul_entry] at hh
calc
_ = (A.val (wordPerm σ u) (wordPerm σ w) * phase σ w) * phase σ w := by
rw [mul_assoc,phase_square,mul_one]
_ = (phase σ u * A.val u w) * phase σ w := by rw [hh]
_ = _ := by ring
/- Rephased purification components are ordinary symmetric on the free,
matched-pair slots, a proved conclusion from the signed commutant relation. -/
theorem regrouped_commutant_invariant {p l : ℕ} (A : commutant p l)
(σ : Equiv.Perm (Fin l)) (u w : Word p l)
(hfix : ∀ i, (odd (u i) ↔ ¬ odd (w i)) → σ i = i) :
regroupPhase (wordPerm σ u) (wordPerm σ w) *
A.val (wordPerm σ u) (wordPerm σ w) = regroupPhase u w * A.val u w := by
rw [commutant_transport]
calc
_ = (regroupPhase (wordPerm σ u) (wordPerm σ w) * phase σ u * phase σ w) * A.val u w := by ring
_ = _ := by rw [regroupPhase_permute σ u w hfix]
/- Equal local parities in the system copies give equal regrouping signs.
This is what justifies cancellation under the actual auxiliary partial trace. -/
lemma regroupPhase_of_same_parity {p l : ℕ} (u u' w : Word p l)
(h : ∀ i, odd (u i) ↔ odd (u' i)) : regroupPhase u w = regroupPhase u' w := by
unfold regroupPhase
apply Finset.prod_congr rfl
intro ij _
rw [h ij.1]
end CoordinateSweeps.SignedTensor
end
open scoped Matrix.Norms.L2Operator
end OAI