The Frobenius-Schur indicator of a compact group #
For a finite-dimensional continuous representation π of a compact group G, the
Frobenius-Schur indicator is the Haar average of the character over squares,
ν₂(π) = ∫_G χ_π(g²) dμ,
with μ the normalized Haar measure of TauCeti/RepresentationTheory/Compact/Haar.lean. It is the
compact-group form of the finite average |G|⁻¹ ∑_g χ(g²) built in
TauCeti/RepresentationTheory/CharacterTable/FrobeniusSchur/Basic.lean; the integral replaces the
average and nothing else changes in the statement.
What the indicator computes is the signed difference of the two square characters. Over any field
χ_{Sym²}(g) - χ_{Λ²}(g) = χ(g²) (Representation.char_symmetricSquare_sub_char_exteriorSquare),
so integrating that identity gives
ν₂(π) = ∫_G χ_{Sym²π} - ∫_G χ_{Λ²π},
and the companion identity χ(g)² = χ_{Sym²}(g) + χ_{Λ²}(g) gives the two integrals separately in
terms of ν₂(π) and ∫_G χ_π². Those two integrals are the (complexified) dimensions of the
invariants of the two squares once the projection onto invariants built from the Haar averaging
operator is available; here only the character-level identities are proved, which is what makes them
independent of that development.
The two square characters are not, a priori, characters of continuous representations: the
symmetric and exterior square of a continuous representation carry a continuous action, but that
is not what is used. The closed formulas χ_{Sym²}(g) = ½(χ(g)² + χ(g²)) and
χ_{Λ²}(g) = ½(χ(g)² - χ(g²)), valid because 2 ≠ 0, exhibit both as continuous functions of g
directly, and that is all integration needs; that continuity is
TauCeti/RepresentationTheory/Continuous/Square/Character.lean, which needs neither a group nor a
measure, and only the integrability it yields is proved here.
For a unitary representation the indicator is real
(ContRepresentation.conj_frobeniusSchurIndicator): conjugating the character inverts its
argument, (g²)⁻¹ = (g⁻¹)², and Haar measure on a compact group is inversion invariant. Its
modulus is at most dim V. That it takes only the values 1, 0, -1 on an irreducible is the
reality trichotomy, which is not proved here but in
TauCeti/RepresentationTheory/Compact/FrobeniusSchur/Trichotomy.lean; the reading of those three
values as invariant bilinear forms is
TauCeti/RepresentationTheory/Compact/FrobeniusSchur/InvariantForm.lean.
Main definitions #
ContRepresentation.frobeniusSchurIndicator:ν₂(π) = ∫_G χ_π(g²) dμ.
Main statements #
ContRepresentation.frobeniusSchurIndicator_eq_of_equiv: the indicator depends only on the equivalence class of the representation.ContRepresentation.frobeniusSchurIndicator_eq_sub_integral: the indicator is the difference of the two square-character integrals,ν₂(π) = ∫ χ_{Sym²π} - ∫ χ_{Λ²π}.ContRepresentation.integral_character_sq: the two square-character integrals add up to∫ χ_π², so together with the previous statement they determine both:ContRepresentation.integral_character_symmetricPower_twoandContRepresentation.integral_character_exteriorPower_twoare the resulting closed forms½(∫ χ_π² ± ν₂(π)).ContRepresentation.conj_frobeniusSchurIndicatorandContRepresentation.im_frobeniusSchurIndicator: the indicator of a unitary representation is real.ContRepresentation.norm_frobeniusSchurIndicator_le: it has modulus at mostdim V.ContRepresentation.frobeniusSchurIndicator_trivial: the indicator of the trivial representation is its dimension, in particular1on the line.
Implementation notes #
The continuity hypothesis hπ : Continuous π is the one ContRepresentation.character
already carries: ContRepresentation bundles continuous action operators without asking that the
map to them be continuous. Since it is a Prop argument, two indicators built from different
continuity proofs are equal by proof irrelevance.
Everything below is declared in the root ContRepresentation namespace rather than in
TauCeti.ContRepresentation, so that π.frobeniusSchurIndicator hπ elaborates:
ContRepresentation is Mathlib's type, and a namespace for it nested inside TauCeti does not
support dot notation.
The ambient TauCeti names this file consumes, such as TauCeti.haarProb and
TauCeti.integrable_continuousMap, are brought in by open. The character and its formulas are
methods in ContRepresentation.
The scalars are ℂ: the symmetric- and exterior-power representations of
TauCeti/RepresentationTheory/SymmetricPower.lean and
TauCeti/RepresentationTheory/ExteriorPower.lean, whose characters the square identities below
speak of, are built over a base ring in Type, so a general RCLike field in an arbitrary universe
would not even let the statements be formed.
The mathematical development follows Daniel Bump, Lie Groups, second edition, Chapter 2, and Bröcker-tom Dieck, Representations of Compact Lie Groups, Chapter II.
The character read along the squaring map is integrable against normalized Haar measure, so the Frobenius-Schur indicator below is a genuine average and not the Bochner integral's junk value.
The Frobenius-Schur indicator of a finite-dimensional continuous representation of a
compact group: the Haar average ν₂(π) = ∫_G χ_π(g²) dμ of the character over squares.
Equations
- π.frobeniusSchurIndicator hπ = ∫ (g : G), (π.character hπ) (g * g) ∂TauCeti.haarProb G
Instances For
The defining integral of the Frobenius-Schur indicator.
Equivalent representations have the same Frobenius-Schur indicator, since they have the
same character (Representation.char_iso). The two continuity witnesses are unrelated: the
indicator reads only the underlying representation, so nothing links them.
The Frobenius-Schur indicator of the trivial representation is its dimension: every character
value is dim V, and Haar measure is normalized. On the line this is ν₂ = 1.
The two square characters #
Over any field the difference of the symmetric-square and exterior-square characters is the
character at the square, and their sum is the square of the character. Integrating both identities
expresses the two square-character integrals through ν₂(π) and ∫ χ_π².
The symmetric-square character is integrable against normalized Haar measure.
The exterior-square character is integrable against normalized Haar measure.
The Frobenius-Schur indicator is the signed difference of the two square-character
integrals, ν₂(π) = ∫ χ_{Sym²π} - ∫ χ_{Λ²π}.
This is the Haar average of Representation.char_symmetricSquare_sub_char_exteriorSquare, which
holds over every field, characteristic two included. Once the averaging projection identifies
∫ χ_ρ with the dimension of the invariants of ρ, the right-hand side becomes the signed count
dim (Sym²V)ᴳ - dim (Λ²V)ᴳ of the finite-group statement.
The two square-character integrals add up to the integral of the squared character,
∫ χ_π² = ∫ χ_{Sym²π} + ∫ χ_{Λ²π}. Together with
ContRepresentation.frobeniusSchurIndicator_eq_sub_integral this determines both.
The symmetric-square character integrates to ½(∫ χ_π² + ν₂(π)).
The exterior-square character integrates to ½(∫ χ_π² - ν₂(π)).
The Frobenius-Schur indicator of a unitary representation is real.
Conjugating the character inverts its argument
(ContRepresentation.character_apply_inv), and (g * g)⁻¹ = g⁻¹ * g⁻¹, so the conjugate
indicator is the Haar average of g ↦ χ(g⁻¹ * g⁻¹). Haar measure on a compact group is inversion
invariant, which returns that average to the original one.
The Frobenius-Schur indicator of a unitary representation has vanishing imaginary part: it is
a real number, the form in which the reality trichotomy ν₂ ∈ {1, 0, -1} is stated.
The Frobenius-Schur indicator of a unitary representation has modulus at most the dimension: every character value of a unitary representation does, and Haar measure is normalized.