Invariant bilinear forms on a representation #
A bilinear form B on the space of a representation ρ of G is invariant when every ρ g
preserves it, B (ρ g x) (ρ g y) = B x y. The invariant forms are a submodule of all bilinear
forms, and read as maps V → V* they are exactly the intertwiners from ρ to its dual
representation -- the concrete form of the statement that they are the G-invariants of a space
of bilinear forms. Cutting the invariant forms down by symmetry, respectively by alternation,
gives two further submodules, and away from characteristic two they meet only in 0.
On an irreducible representation the invariant forms are tightly constrained, and the three
constraints proved here are the ones the Frobenius-Schur trichotomy is read off from. First, the
left radical of an invariant form is the kernel of an intertwiner out of an irreducible
representation, so a nonzero invariant form on one is nondegenerate. Second, over an
algebraically closed field and in finite dimensions, Schur's lemma makes a nonzero invariant form
unique up to a scalar: the invariant forms are the line it spans. Third -- and this is the
point -- the flip of an invariant form is invariant, so a nonzero invariant form is a scalar
multiple of its own flip; flipping twice squares the scalar to 1, and the form is therefore
either symmetric or the negative of its flip.
Away from characteristic two the second alternative is exactly alternation, which gives the
orthogonal/symplectic dichotomy: an irreducible representation carries at most a line of invariant
forms, and -- in characteristic other than two -- any nonzero one on it is either symmetric or
alternating.
Nothing here needs a finite group, and almost nothing needs finite dimensions: invariance, and the
symmetric and alternating invariant forms with it, are defined for a representation of a monoid on
a module over a commutative semiring, and only the results that invoke Schur's lemma ask for an
algebraically closed field and a finite-dimensional space. The g⁻¹ in
TauCeti.Representation.IsInvariantForm.apply_left, and with it the comparison with the dual
representation, is what makes G a group.
Main definitions #
TauCeti.Representation.IsInvariantForm:B (ρ g x) (ρ g y) = B x yfor allg,x,y.TauCeti.Representation.invariantForms: the invariant forms, as a submodule ofBilinForm k V.TauCeti.Representation.symmetricInvariantFormsandTauCeti.Representation.alternatingInvariantForms: the invariant forms that are symmetric, respectively alternating.TauCeti.Representation.invariantFormsEquivIntertwiningMapDual: the invariant forms ofρas the intertwiners fromρto its dual.
Main results #
TauCeti.Representation.isInvariantForm_iff_isIntertwiningMap: a form is invariant exactly when it intertwinesρwithρ.dual.TauCeti.Representation.invariantForms_eq_invariants_linHom_dual: the invariant forms are the invariants oflinHom ρ ρ.dual.TauCeti.Representation.symmetricInvariantForms_inf_alternatingInvariantForms: away from characteristic two the symmetric and the alternating invariant forms meet only in0.TauCeti.Representation.IsInvariantForm.nondegenerateandTauCeti.Representation.IsInvariantForm.nondegenerate_iff_ne_zero: an invariant form on an irreducible representation is nondegenerate exactly when it is nonzero.TauCeti.Representation.IsInvariantForm.exists_eq_smul: over an algebraically closed field and in finite dimensions, an invariant form on an irreducible representation is a scalar multiple of any nonzero one.TauCeti.Representation.IsInvariantForm.invariantForms_eq_span: over an algebraically closed field and in finite dimensions, a nonzero invariant form on an irreducible representation spans all of them.TauCeti.Representation.IsInvariantForm.flip_eq_or_flip_eq_neg: in every characteristic, the flip of such a form is the form itself or its negative.TauCeti.Representation.IsInvariantForm.isSymm_or_isAlt: away from characteristic two -- the hypothesis(2 : k) ≠ 0-- such a form is symmetric or alternating.TauCeti.Representation.exists_isSymm_or_exists_isAlt_or_invariantForms_eq_bot: the case split that dichotomy gives, that an irreducible representation carries a nonzero invariant symmetric form, or a nonzero invariant alternating one, or no nonzero invariant form at all.
Implementation notes #
TauCeti.Representation.IsInvariantForm is the statement that every ρ g is an isometry of the
form, TauCeti.BilinForm.IsIsometry B (ρ g), so that the invariant forms of a representation and
the isometry group of a form are the same notion read two ways. Its body is not exposed:
TauCeti.Representation.isInvariantForm_iff_isIsometry and
TauCeti.Representation.isInvariantForm_iff introduce its isometry and pointwise forms, while
TauCeti.Representation.IsInvariantForm.isIsometry and
TauCeti.Representation.IsInvariantForm.apply eliminate them, so nothing outside this file has to
unfold the definition. The pointwise iff is deliberately not a simp lemma: unfolding
invariance into its quantified equation would take
TauCeti.Representation.isInvariantForm_zero and
TauCeti.Representation.mem_invariantForms out of simp-normal form, which the simpNF linter
rejects. What the file adds around that pair is the two rewritings that are not
immediate -- moving a single ρ g across the form at the cost of an inverse
(TauCeti.Representation.IsInvariantForm.apply_left), and the identification with intertwiners
into the dual. That identification is recorded twice: once unbundled as
TauCeti.Representation.isInvariantForm_iff_isIntertwiningMap, which is the shape the proofs
below use, and once as the linear equivalence
TauCeti.Representation.invariantFormsEquivIntertwiningMapDual, which is the shape a dimension
count of the invariant forms needs. The equivalence is not built by hand: the invariant forms are
literally the invariants of linHom ρ ρ.dual
(TauCeti.Representation.invariantForms_eq_invariants_linHom_dual), so it is Mathlib's
Representation.invariantsEquivIntertwiningMap transported along that equality. Mathlib's
Representation.invariants asks for a commutative ring and an additive group, which is why the two
declarations sit in a narrower section than the rest of the file; the additive group on the space
of maps V → V* is supplied explicitly, because a representation records only the additive monoid
and the elaborator does not solve for a group structure inducing a given one. The equivalence is
not exposed,
so a downstream module uses it opaquely -- a dimension count goes through LinearEquiv.finrank_eq
and never looks at a value. When a value is needed, the two simp lemmas
TauCeti.Representation.invariantFormsEquivIntertwiningMapDual_apply_toLinearMap and
TauCeti.Representation.invariantFormsEquivIntertwiningMapDual_symm_apply_coe are what identify
it: they say the equivalence moves a form and an intertwiner to each other unchanged.
That identification is also what controls the left radical: the form is an intertwiner out of an
irreducible representation, so TauCeti.Representation.IsInvariantForm.ker_eq_bot is Mathlib's
Representation.IsIrreducible.injective_or_eq_zero rather than a hand-built subrepresentation
argument. Schur's lemma enters through Mathlib's
Representation.IsIrreducible.algebraMap_intertwiningMap_bijective_of_isAlgClosed, the same
theorem that ContRepresentation.exists_eq_smul_one_of_isIrreducible rests on for a
continuous representation.
References #
- Character theory roadmap,
Layer 7,
IsInvariantFormand the invariant symmetric and alternating forms that the values+1and-1of the Frobenius-Schur indicator are characterized by. The indicator itself isTauCeti.Representation.frobeniusSchurIndicator; this file supplies the forms, not the count. - J.-P. Serre, Linear Representations of Finite Groups, GTM 42 (1977), §13.2.
- I. M. Isaacs, Character Theory of Finite Groups (1976), Chapter 4.
The invariant forms of a representation #
A bilinear form B is invariant for a representation ρ when every ρ g preserves it:
B (ρ g x) (ρ g y) = B x y; that is, when every ρ g is an isometry of B.
Equations
- TauCeti.Representation.IsInvariantForm ρ B = ∀ (g : G), TauCeti.BilinForm.IsIsometry B (ρ g)
Instances For
A form is invariant for ρ exactly when every ρ g is an isometry of the form.
Every element of a representation is an isometry of an invariant form.
A form is invariant for ρ exactly when it satisfies the pointwise equation
B (ρ g x) (ρ g y) = B x y.
An invariant form takes the same value on ρ g x and ρ g y as it does on x and y.
The invariant bilinear forms of ρ, as a submodule of all bilinear forms on V.
Equations
- TauCeti.Representation.invariantForms ρ = { carrier := {B : LinearMap.BilinForm k V | TauCeti.Representation.IsInvariantForm ρ B}, add_mem' := ⋯, zero_mem' := ⋯, smul_mem' := ⋯ }
Instances For
Membership in TauCeti.Representation.invariantForms is invariance.
The zero form is invariant.
A sum of invariant forms is invariant.
A scalar multiple of an invariant form is invariant.
Exchanging the two arguments of an invariant form leaves it invariant.
Every bilinear form is invariant for the trivial representation, which acts by the identity.
The symmetric and the alternating invariant forms #
The invariant symmetric bilinear forms of ρ, as a submodule of all bilinear forms.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The invariant alternating bilinear forms of ρ, as a submodule of all bilinear forms.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Membership in TauCeti.Representation.symmetricInvariantForms is invariance together with
symmetry.
Membership in TauCeti.Representation.alternatingInvariantForms is invariance together with
alternation.
An invariant symmetric form is in particular invariant.
An invariant alternating form is in particular invariant.
Away from characteristic two, a form cannot be both symmetric and alternating, so the two
invariant subspaces meet only in 0. All that is asked of k is that 2 be regular, which over
a field is IsRegular.of_ne_zero.
Invariance as intertwining with the dual representation #
Moving a single ρ g across an invariant form replaces it by its inverse on the other side.
This is the shape invariance is used in whenever only one of the two arguments carries the action,
as in the radical of the form or in a comparison with the dual representation.
A bilinear form is invariant exactly when it intertwines ρ with its dual. Read as a
linear map V → V*, an invariant form is a map of representations from ρ to ρ.dual, and
conversely. This is what makes the invariant forms a Hom space rather than just a submodule,
and it is how Schur's lemma reaches them.
The invariant forms of ρ are the invariants of linHom ρ ρ.dual. Invariance of a form
is intertwining of the map V → V* it is, which is Mathlib's membership criterion for the
invariants of a Hom representation.
The invariant forms of ρ are the intertwiners from ρ to its dual. This is the bundled
form of TauCeti.Representation.isInvariantForm_iff_isIntertwiningMap, and it is what puts the
invariant forms in reach of machinery that counts intertwiners: it is Mathlib's
Representation.invariantsEquivIntertwiningMap, transported along
TauCeti.Representation.invariantForms_eq_invariants_linHom_dual.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The intertwiner attached to an invariant form is that form, read as a map V → V*.
The invariant form attached to an intertwiner ρ → ρ.dual is that intertwiner, read as a
bilinear form.
Invariant forms on an irreducible representation #
A nonzero invariant form on an irreducible representation has trivial left radical.
A nonzero invariant form on an irreducible representation is nondegenerate.
On an irreducible representation, an invariant form is nondegenerate exactly when it is
nonzero. The interesting direction is TauCeti.Representation.IsInvariantForm.nondegenerate; the
other is Mathlib's LinearMap.BilinForm.Nondegenerate.ne_zero, once the space an irreducible
representation acts on is known to be nonzero.
An invariant form on an irreducible representation is unique up to a scalar: over an
algebraically closed field and in finite dimensions, every invariant form C is a scalar multiple
of a nonzero invariant form B.
A nonzero invariant form on an irreducible representation spans all of them: the invariant forms are a line.
A nonzero invariant form on an irreducible representation is its own flip up to sign: its flip is either the form itself or its negative. This holds in every characteristic.
The orthogonal/symplectic dichotomy. Away from characteristic two, a nonzero invariant form on a finite-dimensional irreducible representation over an algebraically closed field is either symmetric or alternating.
An irreducible representation carries a symmetric, an alternating, or no invariant form.
Away from characteristic two, over an algebraically closed field it carries a nonzero invariant
symmetric form, or a nonzero invariant alternating form, or no nonzero invariant form at all. This
is the case split the three values of the Frobenius-Schur indicator are read off from, for a finite
group in TauCeti/RepresentationTheory/CharacterTable/FrobeniusSchur/Trichotomy.lean and for a
compact group in TauCeti/RepresentationTheory/Compact/FrobeniusSchur/InvariantForm.lean.