1.1. Lecture 1

1.1.1. Definition of a representation

Let kk be a field (we will almost always take k=k=\mathbb{C}), and let GG be a group.

Definition 1.1.1.

A representation of GG over kk is a pair (π,V)(\pi,V), where

  1. (i)

    VV is a vector space over kk, and

  2. (ii)

    π:GGL(V)\pi:G\rightarrow\operatorname{GL}(V) is a group homomorphism.

We also say that π\pi is a representation of GG on VV, or simply that π\pi is a representation of GG; this makes sense since the vector space VV is part of the definition of π\pi.

The dimension of a representation (π,V)(\pi,V) is the dimension of VV. We will often denote this by dimπ\operatorname{dim}\pi.

There is another way to think of this. Suppose that (ρ,V)(\rho,V) is a representation of GG. Then we may define an action of GG on VV by letting

g𝐯:=ρ(g)𝐯g\cdot{\bf v}:=\rho(g){\bf v}

for all gGg\in G and 𝐯V{\bf v}\in V. This is an action because ρ\rho is a homomorphism, and it is linear, meaning that for every gGg\in G the map taking 𝐯ρ(g)𝐯{\bf v}\mapsto\rho(g){\bf v} is a linear map on VV. Conversely, given a linear action of GG on VV, we can define a representation (π,V)(\pi,V) of GG by π(g)𝐯:=g𝐯\pi(g){\bf v}:=g\cdot{\bf v}. In other words:

A representation of GG is a linear action on a vector space.

Suppose that (π,V)(\pi,V) is a representation of a group GG and dimπ=n\operatorname{dim}\pi=n. By choosing a basis for VV, we may identify VV with knk^{n}, and under this identification any (invertible) linear map VVV\rightarrow V is just the same thing as an (invertible) n×nn\times n matrix.

Thus, once you choose a basis, a representation is just the same as a homomorphism GGLn(k)G\rightarrow\operatorname{GL}_{n}(k). In particular:

A 1-dimensional representation of GG is a homomorphism Gk×G\rightarrow k^{\times}.

1.1.2. Examples

Example 1.1.2.

If VV is any vector space, then we can always take ρ:GGL(V)\rho:G\rightarrow\operatorname{GL}(V) to be the homomorphism sending every element to the identity. We call this the trivial representation of GG on VV.

Example 1.1.3.

Let G=SnG=S_{n}. Recall that there is a homomorphism

ϵ:Sn{±1}\epsilon:S_{n}\rightarrow\{\pm 1\}

taking a permutation to its sign. Since {±1}×\{\pm 1\}\subseteq\mathbb{C}^{\times}, this gives a 1-dimensional representation of SnS_{n}, called the sign representation.

Example 1.1.4.

Suppose that G=(,+)G=({\mathbb{Z}},+). Then, if ρ\rho is a representation of GG, it is completely determined by VV and the invertible linear map ρ(1):VV\rho(1):V\rightarrow V (which may be any element of GL(V)\operatorname{GL}(V)). This is because we then have

ρ(n)=ρ(1++1)=ρ(1)n.\rho(n)=\rho(1+\ldots+1)=\rho(1)^{n}.

Thus, a representation of {\mathbb{Z}} is just a vector space VV together with an invertible linear map from VV to itself.

We can push this a bit further. Suppose that GG is a cyclic group of order nn with generator aa, hence

G=a|an=e.G=\left\langle a\,|\,a^{n}=e\right\rangle.

A representation (π,V)(\pi,V) of GG is once again determined by VV and ρ(a)\rho(a), which can be any linear map T:VVT:V\rightarrow V such that Tn=IdT^{n}=\mathrm{Id}.

Many interesting examples arise from geometry.

Example 1.1.5.

Let G=DnG=D_{n} be the dihedral group of order 2n2n, the group of symmetries (rotations and reflections) of a regular nn-gon. Since each rotation/reflection is an invertible linear map from 22\mathbb{R}^{2}\rightarrow\mathbb{R}^{2}, we get a representation ρ\rho of GG on 2\mathbb{R}^{2}. Letting rr be counter-clockwise rotation by 2π/n2\pi/n radians and ss be reflection in the horizontal axis, recall that DnD_{n} has the presentation

r,s|rn=s2=e,sr=r1s.\left\langle r,s\,|\,r^{n}=s^{2}=e,sr=r^{-1}s\right\rangle.

As an explicit homomorphism ρ:DnGL2()\rho:D_{n}\rightarrow\operatorname{GL}_{2}(\mathbb{R}), we have

ρ(r)=(cos(θn)sin(θn)sin(θn)cos(θn)),ρ(s)=(1001),\rho(r)=\left(\begin{matrix}\cos(\theta_{n})&-\sin(\theta_{n})\\ \sin(\theta_{n})&\cos(\theta_{n})\end{matrix}\right),\qquad\rho(s)=\left(% \begin{matrix}1&0\\ 0&-1\end{matrix}\right),

where θn=2π/n\theta_{n}=2\pi/n. Since GL2()GL2()\operatorname{GL}_{2}({\mathbb{R}})\subseteq\operatorname{GL}_{2}({\mathbb{C}}), we may also view this as a representation of DnD_{n} on 2{\mathbb{C}}^{2}; we call (ρ,2)(\rho,{\mathbb{C}}^{2}) the defining representation of DnD_{n}.

Example 1.1.6.

Let G=S4G=S_{4}. You might remember that this is isomorphic to the group of symmetries (rotations/reflections) of the regular tetrahedron in 3\mathbb{R}^{3}. We therefore get a representation

ρ:S4GL3().\rho:S_{4}\rightarrow\operatorname{GL}_{3}(\mathbb{R}).

It would be a slightly unpleasant exercise to work the matrices out explicitly.

Note that S4S_{4} is also isomorphic to the group of rotations of the cube, giving another (different!) 3-dimensional representation.

Another source of representations comes from actions of groups on (usually finite) sets.

Example 1.1.7.

Define a representation (π,kn)(\pi,k^{n}) of SnS_{n} via

π(σ)(x1𝐞1++xn𝐞n)=x1𝐞σ(1)++xn𝐞σ(n),\pi(\sigma)\left(x_{1}{\bf e}_{1}+\ldots+x_{n}{\bf e}_{n}\right)=x_{1}{\bf e}_% {\sigma(1)}+\ldots+x_{n}{\bf e}_{\sigma(n)},

where 𝐞1,,𝐞n{\bf e}_{1},\ldots,{\bf e}_{n} is the standard basis. This is called the permutation representation of SnS_{n} on knk^{n}.

Important: If we write elements of knk^{n} as (x1,,xn)t(x_{1},\ldots,x_{n})\,^{\mathrm{t}}\! (as we commonly do), then it is not the case that

π(σ)(x1xn)=(xσ(1)xσ(n)).\pi(\sigma)\left(\begin{matrix}x_{1}\\ \vdots\\ x_{n}\end{matrix}\right)=\left(\begin{matrix}x_{\sigma(1)}\\ \vdots\\ x_{\sigma(n)}\end{matrix}\right).

This actually would define a right action, not a left action. The correct formula is

π(σ)(x1xn)=(xσ1(1)xσ1(n)).\pi(\sigma)\left(\begin{matrix}x_{1}\\ \vdots\\ x_{n}\end{matrix}\right)=\left(\begin{matrix}x_{\sigma^{-1}(1)}\\ \vdots\\ x_{\sigma^{-1}(n)}\end{matrix}\right).

This may be generalised whenever we have a group action on a set. Firstly, we associate to any set XX an kk-vector space:

Definition 1.1.8.

Let XX be a set. The free vector space k(X)k(X) over kk generated by XX is the vector space consisting of all formal sums

xXzxx,\sum_{x\in X}z_{x}x,

where zxkz_{x}\in k and zx=0z_{x}=0 for all but finitely many xXx\in X. The vector space operations are as follows: vector addition is given by

(xXzxx)+(xXyxx):=(xX(zx+yx)x),\left(\sum_{x\in X}z_{x}x\right)+\left(\sum_{x\in X}y_{x}x\right):=\left(\sum_% {x\in X}(z_{x}+y_{x})x\right),

and scalar multiplication

λ(xXzxx):=(xXλzxx)\lambda\left(\sum_{x\in X}z_{x}x\right):=\left(\sum_{x\in X}\lambda z_{x}x\right)

for all λk\lambda\in k.

Observe that a basis for k(X)k(X) is {x}xX\{x\}_{x\in X} (here x=1xx=1x, i.e. the formal sum yXzyy\sum_{y\in X}z_{y}y, with zy=1z_{y}=1 if y=xy=x and zy=0z_{y}=0 otherwise)

Definition 1.1.9.

Given a group GG acting on a set XX, we define the permutation representation (π,k(X))\big{(}\pi,k(X)\big{)} for this action by

π(g)(xXzxx):=(xXzx(gx))=(xXzg1xx).\pi(g)\left(\sum_{x\in X}z_{x}x\right):=\left(\sum_{x\in X}z_{x}(g\cdot x)% \right)=\left(\sum_{x\in X}z_{g^{-1}\cdot x}x\right).

1.1.3. Exercises

.

Remark 1.1.10.

We elaborate on two ways of checking whether a map π:GGL(V)\pi:G\rightarrow\mathrm{GL}(V) is actually a representation. This point was discussed a bit during the lecture, but since it will be used frequently throughout the course, I would like to emphasise this a bit further:

(i) Assume that a rule or formula is given for π(g)\pi(g) for every gGg\in G. In order to verify that (π,V)(\pi,V) really is a representation, one needs to either check or show that π(gh)=π(g)π(h)\pi(gh)=\pi(g)\pi(h) for all g,hGg,h\in G.

(ii) Often, we will define π\pi by where it sends the generators g1,,gkg_{1},\ldots,g_{k} of a group GG. Since every element of GG can be written as a word in g1,,gkg_{1},\ldots,g_{k}, if π\pi is required to be a homomorphism, then there is no choice in the values of π(g)\pi(g) for the other group elements gg: π(g)\pi(g) must be the product of the π(gi)\pi(g_{i})s corresponding to the decomposition of gg into a product of gig_{i}s. In order for this procedure to give a well-defined homomorphism π\pi on all of GG, it is both necessary and sufficient that the π(gi)\pi(g_{i})s satisfy the same relations as the gig_{i}s. (This is a general fact about any group homomorphism, not just ones into GL(V)\mathrm{GL}(V)).

Compare the phrasing of problems 1.1.10 and 1.1.10 below!

Problem 12. Verify that (π,3)(\pi,\mathbb{C}^{3}) is a representation of S3S_{3}, where π(σ)GL3()\pi(\sigma)\in\mathrm{GL}_{3}(\mathbb{C}) is defined via

π(σ)(z1z2z3)=(zσ1(1)zσ1(2)zσ1(3))(z1z2z3)3.\pi(\sigma)\left(\begin{matrix}z_{1}\\ z_{2}\\ z_{3}\end{matrix}\right)=\left(\begin{matrix}z_{\sigma^{-1}(1)}\\ z_{\sigma^{-1}(2)}\\ z_{\sigma^{-1}(3)}\end{matrix}\right)\qquad\forall\left(\begin{matrix}z_{1}\\ z_{2}\\ z_{3}\end{matrix}\right)\in\mathbb{C}^{3}.

Problem 13. Writing D3=r,s|r3=s2=rsrs=eD_{3}=\langle r,s\,|\,r^{3}=s^{2}=rsrs=e\rangle, let ρ(r),ρ(s)GL2()\rho(r),\rho(s)\in\mathrm{GL}_{2}(\mathbb{C}) be given by

ρ(r)=(cos(2π/3)sin(2π/3)sin(2π/3)cos(2π/3)),ρ(s)=(1001).\rho(r)=\left(\begin{matrix}\cos(2\pi/3)&-\sin(2\pi/3)\\ \sin(2\pi/3)&\cos(2\pi/3)\end{matrix}\right),\qquad\rho(s)=\left(\begin{matrix% }1&0\\ 0&-1\end{matrix}\right).

Show that this defines a representation (ρ,2)(\rho,\mathbb{C}^{2}) of D3D_{3}.

Problem 14.

  1. (a)

    Find the matrices of all the elements of S3S_{3} for the permutation representation (π,3)(\pi,{\mathbb{C}}^{3}), with respect to the basis 𝐞1,𝐞2,𝐞3{\bf e}_{1},{\bf e}_{2},{\bf e}_{3}.

  2. (b)

    Find another basis such that the matrices all take the form

    (1000??0??),\begin{pmatrix}1&0&0\\ 0&?&?\\ 0&?&?\end{pmatrix},

    and determine the unknown entries for your basis.

Problem 15. Let (π,V)(\pi,V) and (ρ,W)(\rho,W) be two representations of a group GG. Show that (πρ,VW)(\pi\oplus\rho,V\oplus W) is a representation of GG, where

πρ(g)(v,w):=(π(g)v,ρ(g)w)gG,vV,wW.\pi\oplus\rho(g)(v,w):=(\pi(g)v,\rho(g)w)\qquad\forall g\in G,\,v\in V,\,w\in W.

Remark: this is called the (direct) sum of two representations.

Problem 16. Let (π,V)(\pi,V) be a representation of a group GG. Given a subgroup HGH\leq G, show that (π|H,V)(\pi|_{H},V) is a representation of HH.

Problem 17. Given a group action GXG\circlearrowright X, let V={f:X}V=\{f:X\rightarrow\mathbb{C}\}.

a) Show that (π,V)(\pi,V) is a representation of GG, where for every gGg\in G and fVf\in V, π(g)fV\pi(g)f\in V is defined via the formula

(π(g)f)(x)=f(g1x)xX.\big{(}\pi(g)f\big{)}(x)=f(g^{-1}\cdot x)\qquad\forall x\in X.

b) Show that (π~,V)(\widetilde{\pi},V) is not a representation of GG, where for every gGg\in G and fVf\in V, π~(g)fV\widetilde{\pi}(g)f\in V is defined via the formula

(π~(g)f)(x)=f(gx)xX.\big{(}\widetilde{\pi}(g)f\big{)}(x)=f(g\cdot x)\qquad\forall x\in X.

Problem 18. Prove the following:

Proposition 1.1.11.

Let GG be a finite group and (π,)(\pi,{\mathbb{C}}) a representation of GG. Show that for every gGg\in G, there exists ng{0,1,2,|G|1}n_{g}\in\{0,1,2,\ldots|G|-1\} such that

π(g)=e2πing/|G|.\pi(g)=e^{2\pi in_{g}/|G|}.

Problem 19. Find all 1-dimensional representations of D4D_{4}.