Invariant sesquilinear forms on a representation #
A sesquilinear form H : V →ₗ⋆[R] V →ₗ[R] R on the space of a representation ρ is invariant
when every ρ g preserves it, H (ρ g x) (ρ g y) = H x y. This is the Hermitian companion of the
bilinear TauCeti.Representation.IsInvariantForm, and the two together are what produce a real
structure on a complex representation with Frobenius-Schur indicator 1.
The vocabulary and the construction that produces invariant forms use only the star-semiring
algebra of the scalars, and are stated for a module over an arbitrary commutative star semiring:
the invariant forms are a submodule of all sesquilinear forms, and summing an arbitrary form over
the group makes it invariant, exactly as
TauCeti/RepresentationTheory/InvariantForm/SumOfConjugates.lean does for a bilinear form.
Positivity is what asks for more, and the second half of the file is over ℂ. Here the sum does
more work than it does over ℝ: there the sum of the coordinate dot products is merely nonzero,
while over ℂ the sum of the coordinate Hermitian forms is still positive definite, because
each summand is, and a sum of nonnegative complex numbers vanishes only when every summand does.
So the construction gives an invariant inner product outright, not just a nonzero invariant form.
Positive semidefiniteness is Mathlib's LinearMap.IsPosSemidef for the complex order (activate it
with open scoped ComplexOrder), which packages Hermitian symmetry conj (H x y) = H y x
(LinearMap.IsSymm for the conjugation starRingEnd ℂ) together with 0 ≤ H x x. Definiteness
is carried separately as H x x ≠ 0 off the origin, which is the form the Schur-lemma argument
downstream consumes.
The positivity half is stated over ℂ rather than over an RCLike field. The order that
LinearMap.IsPosSemidef needs is the one Complex.partialOrder supplies in the ComplexOrder
scope, and RCLike supplies a second, lower-priority partial order in the same scope; stating the
positivity results over RCLike would fix the wrong one of the two for the ℂ-only consumer,
TauCeti/RepresentationTheory/InvariantForm/StructureMap.lean.
Main definitions #
Representation.IsInvariantSesqForm: a sesquilinear form invariant for a representation, with its unfoldingRepresentation.isInvariantSesqForm_iffand the accessorRepresentation.IsInvariantSesqForm.apply.Representation.invariantSesqForms: the invariant sesquilinear forms, as a submodule of all sesquilinear forms onV.Representation.sumOfConjugatesSesqForm: the sum of a sesquilinear form over the conjugates of a representation of a finite monoid.
Main results #
Representation.isInvariantSesqForm_sumOfConjugatesSesqForm: that sum is invariant.Representation.isPosSemidef_sumOfConjugatesSesqForm: that sum is positive semidefinite if the form is.Representation.exists_isInvariantSesqForm_isPosSemidef_apply_self_ne_zero: a finite-dimensional complex representation of a finite group carries a positive definite invariant Hermitian form. This is the unitarian trick for a finite group, with the undivided sum in place of the average.Representation.IsInvariantSesqForm.surjective_of_apply_self_ne_zero: a definite invariant sesquilinear form on a finite-dimensional space makes every action map surjective, so a representation carrying one needs no group inverses.
References #
- Character theory roadmap,
Layer 7: the "compatible Hermitian form" that the realizability target
frobeniusSchurIndicatorRep_eq_one_realizableis built from. - J.-P. Serre, Linear Representations of Finite Groups, GTM 42 (1977), §13.2.
Invariant sesquilinear forms #
A sesquilinear form H is invariant for a representation ρ when every ρ g preserves
it: H (ρ g x) (ρ g y) = H x y. This is the Hermitian companion of
TauCeti.Representation.IsInvariantForm, stated pointwise because a sesquilinear form is out of
reach of LinearMap.compl₁₂, which asks for a form linear in its first argument.
Instances For
A sesquilinear form is invariant for ρ exactly when it satisfies the pointwise equation
H (ρ g x) (ρ g y) = H x y.
An invariant sesquilinear form takes the same value on ρ g x and ρ g y as it does on x
and y.
The invariant sesquilinear forms of ρ, as a submodule of all sesquilinear forms on V.
Instances For
Membership in Representation.invariantSesqForms is invariance.
The zero form is invariant.
A sum of invariant sesquilinear forms is invariant.
A scalar multiple of an invariant sesquilinear form is invariant.
The sum of a sesquilinear form over the conjugates of a representation of a finite monoid,
∑ g, H (ρ g ·) (ρ g ·). No division by the order of the monoid is performed, exactly as in
TauCeti.Representation.sumOfConjugatesForm for a bilinear form; when G is a group the plain sum
is already invariant.
Instances For
The sum over the conjugates of a Hermitian form is Hermitian.
Surjectivity of the action #
A definite invariant sesquilinear form makes the action surjective. Invariance carries the
self-pairing of ρ g x back to that of x, so definiteness makes ρ g injective, and on a
finite-dimensional space an injective endomorphism is surjective. A representation carrying such a
form therefore needs no group inverses: an argument that would use ρ g⁻¹ to move a vector into
the image of ρ g can use this instead, and so applies to a representation of a mere monoid.
Invariance of the conjugate sum #
The sum of a sesquilinear form over the conjugates of a representation of a finite group is
invariant: translating the summation index by h moves the sum for ρ h x and ρ h y back to the
sum for x and y.
Positivity of the conjugate sum #
The sum over the conjugates of a positive semidefinite form is positive semidefinite.
A vector with vanishing self-pairing for the conjugate sum of a nonnegative form already has
vanishing self-pairing for the form itself: the summand at g = 1 is one of the nonnegative
summands of a vanishing sum.
The coordinate Hermitian form and the unitarian trick #
A finite-dimensional complex representation of a finite group carries a positive definite
invariant Hermitian form. Take the coordinate Hermitian form of a basis -- the sesquilinear form
of the identity matrix, ∑ i, conj (b.repr x i) * b.repr y i -- and sum it over the conjugates.
The sum stays Hermitian and nonnegative, and it stays definite because a vanishing sum of
nonnegative complex numbers has vanishing summands, in particular the one at g = 1.
This is the unitarian trick for a finite group, with the undivided sum in place of the average, so
no invertibility of the group order is needed. Over ℝ the same recipe gives only a nonzero
invariant form (TauCeti.Representation.exists_isInvariantForm_isSymm_ne_zero); it is the
definiteness of the Hermitian form that makes this statement stronger.