In these exercises, G always denotes a finite group. Unlessindicated otherwise, all vector spaces are assumed to be finitedimensional over the field F=C. The representationspace V of a representation of G is always assumed to be finitedimensional over C.
Let ϕ:G→F× be a group homomorphism; since F×=GL1(F), we can think of ϕ as a1-dimensional representation (ϕ,F) of G.
If V is any representation of G, we can form a newrepresentation ϕ⊗V. The underlying vector space forthis representation is just V, and the "new" action of an elementg∈G on a vector v is given by the rule g⋆v=ϕ(g)gv.
a. Prove that if V is irreducible, then ϕ⊗V is also irreducible.
::: {.solution}
We prove the following statement: (∗) if W⊂V is a subspace, then Wis invariant for the original action of G if and only if itis invariant for the ⋆ action of G.
First note that (∗) immediately implies the assertion of (a).
To test invariance, let w∈W and let g∈G. Since Wis a linear subspace and since ϕ(g) is a non-zero scalar,it is immediate that gw∈W⟺g⋆w=ϕ(g)gw∈W Since this holds for all w and all g, (∗) follows.
:::
b. Prove that if χ denotes the character of V, then thecharacter of ϕ⊗V is given by ϕ⋅χ; inother words, the trace of the action of g∈G on ϕ⊗V is given by χϕ⊗V(g)=tr(v↦g⋆v)=ϕ(g)χ(g).
::: {.solution}We just need to compute the trace of the linear mapping V→V given byv↦g⋆v.
If the action of g on V is given by the linear mappingρ(g), then χV(g)=tr(ρ(g)).
Now, the ⋆-action of g is given by the linear mappingv↦g⋆v=ϕ(g)ρ(g)v.
So χϕ⊗V(g)=tr(ϕ(g)ρ(g)).For any scalar s∈k, trace of the linear mapping sρ(g) isgiven by tr(sρ(g))=str(ρ(g))=sχV(g)("linearity of the trace").
Thus χϕ⊗V(g)=tr(ϕ(g)ρ(g))=ϕ(g)χV(g).
:::
c. Recall that in class we saw that S3 has an irreducible representation V2 of dimension 2whose character ψ2 is given by
gψ212(12)0(123)−1
Observe that sgnψ=ψ and conclude thatV2≃sgn⊗V2, wheresgn:Sn→{±1}⊂C×is the sign homomorphism.
On the other hand, S4 has an irreducible representation V3 of dimension 3whose character ψ3 is given by
gψ313(12)1(123)0(1234)−1(12)(34)−1
(I'm not asking you to confirm that ψ3 is irreducible, though it would bestraightforward to check that ⟨ψ3,ψ3⟩=1).
Prove that V3≃sgn⊗V3 asS4-representations.
(In particular, S4 has at least twoirreducible representations of dimension 3.)
::: {.solution}
We first consider the representation V2 of S3. Writeχ2 of the character of this irreducible representation.The character of sgnχ2 is then given by the product sgnχ2.
gψ2sgnsgnψ21212(12)0−10(123)−11−1
Inspecting the table we see that ψ2=sgnψ2. This shows that V2 is isomorphicto sgn⊗V2 as representations forS3.
:::
Let V be a representation of G.
For an irreducible representation L, consider the set S={S⊆V∣S≃L} of all invariantsubspaces that are isomorphic to L as G-representations.
Put V(L)=S∈S∑S.
a. Prove that V(L) is an invariant subspace, and show that V(L) is isomorphic to a direct sumV(L)≃L⊕⋯⊕Las G-representations.
::: {.solution}
First of all, we note more generally that if I is an index setand if Wi⊂V is a G-invariant subspace for each i∈I, then i∈I∑Wi is again aninvariant subspace. (The proof is straightforward from thedefinitions). This confirms that V(L) is an invariantsubspace.
To prove the remaining assertion, we proceed as follows.
Let us say that a G-representation W is L-isotypic ifevery irreducible invariant subspace of W is isomorphic toL.
It is immediate that V(L) is L-isotypic. We are going toprove:
If W is an L-isotypicG-representation, then W is isomorphicto a direct sum W≃L⊕⋯⊕L.
Proceed by induction on dimW. If dimW=0 then W={0} and theresult is immediate (W is the direct sum of zero copies of L).
Now observe that if dimW>0 then W contains aninvariant subspace isomorphic to L, so that dimW≥dimL.
Now if dimW=dimL, then W≃L and theresult holds in this case.
Finally, suppose that dimW>dimL and let S⊂W be an invariant subspace with S≃L.
By complete reducibility we may find an invariant subspace U⊂W such that W is the internal direct sum W=S⊕U.
Since dimW=dimS+dimU, we have dimU<dimW. Moreover, U is also L-isotypic. So by induction ondimension, we know that U≃L⊕⋯⊕L,(say, a direct sum of d copies of L).
But thenW=S⊕U≃L⊕(L⊕⋯⊕L)=L⊕L⊕⋯⊕Lis isomorphic to a direct sum of d+1 copies of L.
:::
b. Prove that the quotient representationV/V(L) has noinvariant subspaces isomorphic to L as G-representations.
::: {.solution}
Write π:V→V/V(L) for the quotient map v↦v+V(L); thus π is a surjective homomorphism ofG-representations.
Suppose by way of contradiction that S⊂V/V(L) is aninvariant subspace isomorphic to L, and let S′⊂Vbe the inverse image under π of S:S′=π−1(S).
Then S′ is an invariant subspace of V containing V(L), andthe restriction of π to S′ defines a surjective mappingπ∣S′:S′→S≃L.
If K denotes the kernel of π∣S′, then complete reducibilityimplies that there is an invariant subspace M of V such thatS′ is the internal direct sum(∗)S′=K⊕M.
In particular, the invariant subspace M is isomorphic to Las G-representations. But then by definition we have M⊂V(L) contradicting the condition M∩K={0} which must hold by (∗). This contradiction proves the result.
:::
c. If L1,L2,⋯,Lm is a complete set of non-isomorphicirreducible representations for G, prove that V is theinternal direct sum V=i=1⨁mV(Li).
::: {.solution}
We first note that V is equal to the sum i=1∑mV(Li); indeed, if W=∑i=1mV(Li), then bycomplete reducibility V=W⊕W′ for an invariantsubspace W′. But if W′=0 then W′ contains anirreducible invariant subspace, so that W′∩V(Li)=0 for some i and hence W′∩W=0; this is impossiblesince the internal sum V=W⊕W′ is direct. This argument showsthat W′=0 and hence that V=W.
Finally, we show thatthe sum i=1∑mV(Li)is direct, i.e. that for each j we have(♣)V(Lj)∩i=j∑V(Li)=0.
Wrote I for the intersection appearing in (♣); thus,I is an invariant subspace of V. If I is non-zero, it hasan irreducible invariant subspace S. Since I⊂V(Lj)and since V(Lj) is Lj-isotypic, we conclude thatS≃Lj.
But then S∩V(Li)=0 for every i=j so thatS∩i=j∑V(Li)=0.Since I⊂i=j∑V(Li), we conclude that I=0.
This completes the proof that V is the direct sum of the V(Li), as required.:::
Let χ be the character of a representation V of G. For g∈G provethat χ(g)=χ(g−1).
Is it true for any arbitrary class function f:G→Cthat f(g)=f(g−1) for every g? (Give a proof ora counterexample...)
::: {.solution}
Let ρ(g):V→V denote the linear automorphism of Vdetermined by the action of g∈G.Then χ(g)=tr(ρ(g)).
Now, since ρ(g) has finite order, say n, its minimal polynomial dividesTn−1∈C[T], and hence every eigenvalue of ρ(g)is an n-th root of unity.
For any n-th root of unity ζ, note that ζ=ζ−1.
Write α1,⋯,αd for the eigenvalues of ρ(g),with multiplicity (so that d=dimV).Notice that ρ(g−1) has eigenvalues α1−1,⋯,αd−1.
Thus
χ(g)=i=1∑dαiandχ(g−1)=i=1∑dαi−1.
Now, we see thatχ(g)=i=1∑dαi=i=1∑dαi−1=χ(g−1)as required.
It is not true that f(g)=f(g−1) for every classfunction f and every g∈G. Indeed, let f=αδ1 bea multiple of the characterisitic function δ1 of the trivial conjugacy class{1}.
Then f(1)=α while f(1−1)=f(1)=α, so that if α∈R, we havef(1)=f(1−1).
:::
For a prime number p, let k=Fp=Z/pZ be the field with p elements. Let V bean n-dimensional vector space over Fp for somenatural number n, and let ⟨⋅,⋅⟩:V×V→k be a non-degenerate bilinear form on V.
(A common example would be to take V=Fpn thefield of order pn, and ⟨α,β⟩=trFpn/Fp(αβ) the trace pairing).
Let us fix a non-trivial group homomorphism ψ:k→C× (recall that k=Z/pZ is anadditive group, while C× is multiplicative). Thusψ(α+β)=ψ(α)ψ(β)for allα,β∈k. If you want an explicit choice, set ψ(j+pZ)=exp(j⋅2πi/p)=exp(2πi/p)j.
For a vector v∈V, consider the mappingΨv:V→C× given by theruleΨv(w)=ψ(⟨w,v⟩).
a. Show that Ψv is a group homomorphism V→C×.
::: {.solution}
For w,w′∈V notice thatΨv(w+w′)=ψ(⟨w+w′,v⟩)=ψ(⟨w,v⟩+⟨w′,v⟩)=ψ(⟨w,v⟩)⋅ψ(⟨w′,v⟩)=Ψv(w)⋅Ψv(w′)since the form is bilinearsince ψ is a group homomby definition.This confirms that Ψv is a group homomorphism.
:::
b. Show that the assignment v↦Ψv is injective(one-to-one).
(This assignment is a function V→Hom(V,C×). In fact, it is a grouphomomorphism. Do you see why? How do you makeHom(V,C×) into a group?)
::: {.solution}
One checks that Hom(V,C×) is amultiplicitive group (this is the dual groupV of V, mentioned in thelectures); the product of ϕ,ψ∈V is given by the ruleg↦ϕ(g)⋅ψ(g).
We note that the assignment v↦Ψv is a group homomorphism. For v,v′∈V we mustargue thatΨv+v′=ΨvΨv′.
For w∈W we haveΨv+v′(w)====ψ(⟨v+v′,w⟩)ψ(⟨v,w⟩+⟨v′,w⟩)ψ(⟨v,w⟩)⋅ψ(⟨v′,w⟩)Ψv(w)⋅Ψv′(w)since the form is bilinearsince ψ is a group homomby defn
Now to show that v↦Ψv is injective, it is enough to argue thatthe kernel of this mapping is {0}.
So, suppose that Ψv is the identity element ofV. In other words, suppose that Ψv(w)=1 forevery w∈V. This shows that ψ(⟨v,w⟩)=1 for every w∈V. Since ψ is a non-trivialhomomorphism Fp→C×, we know thatkerψ={0} (remember that k has prime order...) andwe conclude that ⟨v,w⟩=0 for every w∈W.
(Note that ⟨v,w⟩=0 is an equality in k=Fp=Z/pZ).
Since the form ⟨⋅,⋅⟩ is non-degenerate, sowe may now conclude that v=0.
This proves that the kernel of the mapping v↦Ψv is{0}, hence the mapping is injective.
:::
b. Show that any group homomorphism Ψ:V→C×has the form Ψ=Ψv for some v∈V.
Conclude that there are exactly ∣V∣=qn group homomorphismsV→C×.
::: {.solution}
We observed in class that for any finite abelian group A, there is an isomorphismA≃A.
In particular, ∣A∣=∣A∣.
Applying this in the case A=V, we conclude that∣V∣=∣V∣=∣Hom(V,C×)∣.
Now, we have define an injective mapping v↦Ψv:V→(V).Since the domain and co-domain of this mapping are finite of the same order,the mapping v↦Ψv is also surjective.
Thus the pigeonhole principal shows that every homomorphismΨ:V→C× has the form Ψ=Ψv forsome v∈V, as required.