Documentation

TauCeti.RepresentationTheory.Induction.DoubleCosetPairing

The pairing of two permutation characters counts double cosets #

The character of a permutation representation k[X] counts fixed points, so the normalized pairing of two permutation characters is the Burnside average |G|⁻¹ ∑ g, |X^g| · |Y^g|, which counts the orbits of G on X × Y. Taking X = G ⧸ H and Y = G ⧸ K and identifying those orbits with double cosets gives

⟨Ind_H^G 1, Ind_K^G 1⟩_G = #(H \ G / K),

the pairing of two induced trivial representations.

Main statements #

Implementation notes #

All the identities are equalities in the coefficient field k, with the counts cast into k, and the normalized ones assume IsUnit (Nat.card G : k) so that the |G|⁻¹ in TauCeti.ClassFunction.characterPairing is meaningful. No algebraic closure and no orthogonality of irreducible characters is used: the proof is Burnside's lemma, not a decomposition of the permutation representation into irreducibles. In particular the statements are valid in characteristic p for p ∤ |G| and without assuming k algebraically closed; Maschke's theorem does make the permutation representation semisimple under exactly this hypothesis, but nothing below uses that, only the fact that its character counts fixed points.

The pairing ⟨χ, ψ⟩ = |G|⁻¹ ∑ g, χ g * ψ g⁻¹ is bilinear rather than Hermitian, and the inversion in its second argument is exactly what MulAction.fixedBy_inv absorbs.

References #

This is the "⟨Ind_H^G 1, Ind_H^G 1⟩_G = #(H \ G / H)" clause of the permutation-character item of Layer 2 in TauCetiRoadmap/RepresentationTheory/InductionRestriction/README.md, proved here for two possibly different subgroups.

Burnside's lemma for two permutation characters. The unnormalized sum ∑ g, χ_X(g) · χ_Y(g⁻¹) is the number of orbits of G on X × Y times the order of G.

This form carries no division, so it holds over any field.

The pairing of two permutation characters counts orbits on the product. For a finite group G whose order is invertible in k, the normalized pairing of the characters of k[X] and k[Y] is the number of orbits of G on X × Y.

The pairing of two coset permutation characters counts double cosets. For a finite group G whose order is invertible in k, the normalized pairing of the characters of k[G ⧸ H] and k[G ⧸ K] is the number of double cosets H \ G / K.

The pairing of two coset permutation characters with the trivial character removed. For a finite group G whose order is invertible in k, subtracting the character of the one-point G-set from each of the characters of k[G ⧸ H] and k[G ⧸ K] drops their pairing from the number of double cosets H \ G / K to one less than it: the three extra Burnside terms are 1 each, because G is transitive on each coset space and on the point.

For K = H this is the norm of the augmentation, or Steinberg, character of k[G ⧸ H], whose value is therefore #(H \ G / H) - 1.

The pairing of two induced trivial representations counts double cosets. For a finite group G whose order is invertible in k, the pairing of the characters of Ind_H^G 1 and Ind_K^G 1 is the number of double cosets H \ G / K.

For K = H this is the classical statement that ⟨Ind_H^G 1, Ind_H^G 1⟩_G counts H \ G / H.