Section 2, p. 4 — the Haar average is the selfadjoint idempotent onto the invariant tensors
ProvedWeingartenCalc.Fundamental.projection_onto_invariantshaar-measureinvariant-theoryorthogonal-projectionp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1weingarten-calculus
Let be a compact Hausdorff topological group with its Haar probability measure , let be a continuous unitary representation on , fix , and let
be the Haar average of the tensor-power matrices on . Then:
- is selfadjoint: ;
- is idempotent: ;
- the image of is the space of -invariant tensors,
Together, is the orthogonal projection of onto . This is what reduces the computation of Weingarten integrals to the invariant theory of : once a basis of the invariant tensors is known, is determined by linear algebra.
Formalization Note is the conjugate transpose Pᴴ; the image is the range of the linear map on (Fin d → Fin N) → ℂ; the invariant subspace is defined from the representation itself.
Preamble
import Mathlib import Definitions.Def_WeingartenCalc_Fundamental_TensorPower open MeasureTheory Matrix
Formal statement
namespace WeingartenCalc.Fundamental
theorem projection_onto_invariants
{G : Type*} [Group G] [TopologicalSpace G] [IsTopologicalGroup G] [CompactSpace G]
[T2Space G] [MeasurableSpace G] [BorelSpace G] (μ : Measure G) [μ.IsHaarMeasure]
[IsProbabilityMeasure μ] {N : ℕ} (U : G →* Matrix.unitaryGroup (Fin N) ℂ)
(hU : Continuous U) (d : ℕ) :
(averagingOperator μ U d)ᴴ = averagingOperator μ U d ∧
averagingOperator μ U d * averagingOperator μ U d = averagingOperator μ U d ∧
LinearMap.range (averagingOperator μ U d).mulVecLin = invariants U d := by sorry
end WeingartenCalc.Fundamental
Source
Collins, Matsumoto & Novak, The Weingarten Calculus, arXiv:2109.14890v1, p. 4, Section 2 ("Thus P is a selfadjoint idempotent, and as such P orthogonally projects H⊗d onto its image, which is the space of G-invariant tensors")
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.