PH7009: Group Theory and Symmetry 7009:
Classroom Exercise 2 Supplement
TA: Jason Payne
Here we will briefly introduced the notion of induced representations in order to clarify what
is going on in Problem 1 of Classroom Exercise 2. We will closely follow the discussion (and
notation) given in 3.11 of [1], just so that you have something else to reference, if needed.
The setup-up is as follows: consider a group G, a subgroup H, and a d-dimensional repre-
sentation r of H, i.e. a homomorphism r : H GL(d). The goal of the method of induced
representations is then to take this representation r of the subgroup H, and construct a repre-
sentation of the entire group G.
As you all should be familiar with by now, having r means that for each element h H,
we have an operator M(h)[r] acting on a d-dimensional vector space1 H. Given some basis |ii
(where i = 1, 2, , d) for H, this means that we can write each M(h)[r] as a d d matrix:
M(h)[r] : |ii 7 M(h)[r] ij |ji
(where we are using the Einstein summation convention on the js). Now, in order to move
towards all of G, rather than just H, we consider a new space:
G H/,
which is just some fancy notation for the space obtained by taking all possible pairs (g, |ii) with
g G and |ii H (thats the G H part), and then identifying (i.e. treating as if they are the
same) all pairs of the form
(gh, |ii) g, M(h)[r] ij |ji .
In other words, we absorb any pieces of h showing up in a given element g 0 = gh G into |ii
using the representation r.
Now, suppose that the coset space G/H consists of N cosets, and write them as
H, g1 H, g2 H, , gN 1 H,
where g1 , , gN 1 G \ H. Recall that these form a partition of G, i.e.
G = H g1 H g2 H gN 1 H, 2
i.e. that every element of G lies in on (and only one) coset. This means that the aforementioned
space is simply given by
G H/ = (G/H) H,
since any two pairs (g, |ii) and (g 0 , |ji) with g, g 0 gk H are considered as the same.
So.... who cares? What can we do with this (G/H) H gadget? Well, we can easily turn it
into a dN -dimensional vector space VdN by treating the pairs (gk , |ii) as a basis vectors |gk , ii,
and then taking (formal) linear combinations (similar to what one does when forming the group
algebra, so that you can define the regular representation):
VdN := {ki |gk , ii}.
1
In [1] they talk about having a Hilbert space, but this isnt needed for what were talking about here, so lets
just forget about that...
2
... and gi H gj H = for all i 6= j...
Furtheremore, and more importantly, we can easily define a dN -dimensional representation
IndGH [r] : G GL(dN ) of all of G on VdN as follows:
IndGH [r](g)|gk , ii = |ggk , ii.
One can check that this makes sense mathematically by looking through the details on page 62
of [1], but we wont dwell on that here: suffice to say, it works.
Note
This representation IndGH [r] definitely does not have to be irreducible, in general, as you
will see.
So, we were able to successfully construct a representation of the entire group G starting
from just a representation of one of its subgroups. The relevance of this to Problem 1(a)3 lies in
considering the case when G = T := G4 o Z3 , H = G4 , and taking r as a trivial representation
G4 . Since (whether G4 is D2 or Z4 ) there are only 1-dimensional (irreducible) representations of
G4 , we will have d = 1; moreover, T /G4 = Z3 , so N = 3. In other words, we will ultimately end
up with a dN = 3-dimensional representation at the end. This means that we have a basis |10 i
(the prime is just to distinguish this vector from another vector later) for the 1D vector space
C that G4 is being represented on, which we combine with the three elements (e, b, and b2 ) of
T /G4 in order to obtain a basis for the space (T /G4 ) C:
|e, 10 i, |b, 10 i, and |b2 , 10 i.
Since the 10 in these vectors is superfluous in this case, we can throw them out and simply write
our basis as
|e, 10 i |ei, |b, 10 i |bi, and |b2 , 10 i |b2 i.
Now, how does (as an illustrative example) IndTG4 [r](b), which we will write as (b) for simplicity
(and to be rather suggestive) behave? Well, based on what we said earlier, we have
(b)|ei = |bei = |bi, (b)|bi = |bbi = |b2 i, (b)|b2 i = |bb2 i = |ei.
In other words, the vectors |ei, |bi, |b2 i are nothing more than the vectors |1i, |2i, |3i that were
given in the problem!
This is what is going on throughout Problem 1: you are given some representation of G4 ,
from which you can construct a representation of the entire group T that acts on the space
spanned by |1i, |2i, and |3i. This representation is defined in the way described above, i.e. it
basically boils down to just the group multiplication!
References
[1] P. Ramond, Group Theory A Physicists Survey, CUP (2010).
[2] W.-K. Tung, Group Theory in Physics, World Scientific, 2003 reprint.
3
... and similar things could be said about the rest of Problem 1 in particular, one can adapt what is said
here to complete Problem 1(b) without using the Theorem 3.6 and the Corollary to Theorem 3.7 in [2]...
