Lecture Notes Part 2
Lecture Notes Part 2
25 maj 2007
Continuous groups
We begin with reminding us of the denition of a group.
Denition 1 (Group). A set of elements a(
) ( form a group if
1. There exists an composition law
a b = ab
that is associative, i.e.
(ab)c = a(bc).
2. The set is closed under this law, i.e.
ab = c (; a, b (.
3. There exists an element a(
The main dierence from the previous groups we have so far studied is that the ele-
ments are functions of n real parameters
= (
1
, . . . ,
n
).
Example 1. SO(2) The two dimesional rotational group SO(2), some times also denoted
R
2
, which transforms a vector to a rotated vector
_
x
1
x
2
_
= R
2
_
x
1
x
2
_
; R
2
=
_
cos sin
sin cos
_
has one parameter .
Strictly speaking, this is the dening representation of the SO(2) group. We will later
come to that denition.
)a( ) = a(
) (, can be reformulat-
ed to that there must exist a mapping
f(
1
,
2
) : , that takes two points from
the parameter space and gives a new point in the parameter space.
We can reformulate the other rules of the group in this mapping as
1
1. Associativity
f(
,
f( ,
)) =
f(
f(
, ),
)
2. Identity
f(
0,
) =
f(
0) =
3. Inverse
f(
) =
0; a(
) = a(
)
1
Example 2. Rotation In the case of rotations we see that this mapping is just that of
adding angles. More specically + = , + ( + ) = ( + ) + , + 0 = 0 + = ,
= 0 .
We have already seen the most of the classications of groups, simple, semi-simple and
Abelian. We here introduce another classication.
Denition 3 (Compact). A group is compact if the parameter space is bounded.
Example 3. Rotations In the case of rotations we can choose the parameter space to be
the interval I = [, ], so it is a compact group.
Representations
Denition 4 (Representation). A representation is a mapping of the group elements
onto matrices
a(
) D(a(
))
such that
D(a(
))D(a( )) = D(a(
)a( ))
2
Fundamental representation
This representation is a faithful representation that denes the matrices.
Example 4. SO(2) The special orthogonal group (SO(2)) are 2 2, real and orthogonal
matrices A, i.e.
A
1
= A
T
and det A = 1
It has one parameter, is compact and Abelian.
The classication special of a group implies that the determinant of the matrices are
one. In all groups, it is demanded that the determinant is non-zero to ensure the existence
of an inverse for every element, so that is a property we always have.
Example 5. GL(2) The general linear group (GL(2)) are 2 2, real matrices with four
parameters.
Example 6. SL(2) The special linear group (SL(2)) are 2 2, real matrices with three
parameters, since we have the condition that det A = 1.
Example 7. U(2) The unitary group (U(2)) are 2 2, complex and unitary matrices, i.e.
A
1
= A
Example 8. SU(2) The special unitary group (SU(2)) are 2 2, complex and unitary
matrices with det A = 1 and therefor three parameters.
Example 10. SL(2, C) The special linear complex group SL(2, C) are 2 2 complex
matrices with det A = 1 and so six parameters. This group is homomorphic to the proper
Lorentz group consisting of boosts and rotations (but not inversions).
3
Lie Algebras
To nd the important properties of a Lie group it is sucient to study the correspond-
ing Lie Algebra. This is obtained from elements which dier only innitesimally from the
identity.
Let us Taylor expand an element of the dening representation
D(a(
)) = 1 +i
k
X
k
+. . .
where
X
k
= i
k
D(a(
))
0
.
The matrices X
k
are called the generators of the group. Observe that we cannot
just multiply the generators to obtain all the other elements of the group, the method will
be some what more complicated.
Note that if D is unitary, then X is Hermitian, i.e. X
k
= X
k
.
A arbitrary group element in the dening representation is given by
D(a(
)) = e
i
k
X
k
=
n
i(
X)
n
n!
.
Example 11. SO(2)
e
i
= exp
_
i
_
3
1
i
2
1
+i
2
3
__
The generators X
k
form a vector space with dimension n, the same as the number of pa-
rameters. So we can add them and multiply with real scalars and still be in the same space.
Example 12. SO(2) For SO(2) we have a generator
X =
2
=
_
0 i
i 0
_
))D(a( )) = D(a(
))
4
becomes
e
i
X
e
i
X
= e
i
X
,= e
i(
+ )
X
since we deal with matrices they need not to commute. So in general
[X
i
, X
k
] ,= 0
unless we deal with an Abelian group.
We need to study the commutator of two dening representations close to the identity,
dened as
D(a(
))D(a( ))D(a(
))
1
D(a( ))
1
= D(a(
))
and if it is not equal to identity we have a non-zero commutator.
Taylor expansion gives
1 +
j
k
[X
j
, X
k
] +. . . = 1 +i
l
X
l
where we can write
l
= C
l
jk
k
, which follows from the boundary conditions of the
mapping
= g(
, ). Namely g(
0) =
0 = g(
0, ). So we nd that
[X
j
, X
k
] = iC
l
jk
X
l
This relation is what we call a Lie algebra and the constants C
l
jk
are called the structure
constants. The structure constants dene the group completely and have the following
usefull properties
Real: C
l
jk
R if the generators are unitary.
Antisymmetric: C
l
jk
= C
l
kj
, which follows from that the commutator is antisym-
metric.
Example 13. Generators of angular momentum The generators of angular momentum
satises
[J
j
, J
k
] = i
jkl
J
l
SU(2)
where
jkl
is the Levi-Civita tensor which is antisymmetric in all its indexes.
The structure constants are basis dependent, but as we will see one can always choose
a basis such that the structure constant is antisymmetric in all indexes.
For each representation D(a(
)) = e
i
T
; [T
j
, T
k
] = iC
l
jk
T
l
5
So the structure constants are the same for all representations of generators. In the fol-
lowing we will mainly focus on the generators themselves. Properties such as equivalence,
reducibility can be transfered from the group (representaion of elements) to the algebra
(generators of the representation).
Adjoint representation
This is a representation of the generators dened by the structure constants
(T
j
)
l
k
= iC
l
jk
where the lower index represents columns and the upper index represents rows. The di-
mension of this representation is the same as the number of parameters.
Example 14. SU(2) SU(2) has three parameters so in the adjoint representation we have
3 3 matrices.
T
1
=
_
_
111
121
131
112
122
132
113
123
133
_
_
=
_
_
0 0 0
0 0 i
0 i 0
_
_
T
2
= . . . =
_
_
0 0 i
0 0 0
i 0 0
_
_
T
3
= . . . =
_
_
0 i 0
i 0 0
0 0 0
_
_
We can diagonalize the metric since it is symmetric, and we may choose a basis where
we can write the metric as
g
ij
=
ij
; > 0
since the metric is real. This is however only true for compact Lie algebras.
Example 15. SU(2)
g
ij
=
ikl
jlk
= 2
ij
ij
and k
i
,= 0 for semi-simple groups, in other cases at least one is zero.
Theorem 1. Cartans criterion The determinant of the metric is non-zero if and only if (
is semi-simple, i.e.
det g ,= 0 ( is semi-simple
As it has been said earlier we can for compact groups choose a basis where the structure
constants are fully antisymmetric.
C
l
jk
= C
m
jk
ml
=
1
C
m
jk
g
ml
=
1
C
m
jk
Tr
_
T
A
m
T
A
l
_
=
1
Tr
_
C
m
jk
T
A
m
T
A
l
_
=
i
Tr
__
T
A
j
, T
A
k
T
A
l
_
= C
jkl
since the trace is cyclic. From now on we will only use this basis.
Generators acts on states in the same way as representations of group elements, i.e.
T
k
[ i ) = [ j ) j [ T
k
[ i ) = [ j ) [T
k
]
ji
.
Theorem 2. Inverse metric For semi-simple groups the Cartan criterion provides the
existence of an inverse metric, dened such that
g
jk
g
kl
=
j
l
7
Denition 6 (C
1
). We dene the quadratic Casimir operator as
C
1
= g
jk
T
j
T
k
which commutes with every generator.
Theorem 3. Racahs theorem The number of Casimir operators is equal to the rank m
of the group.
Note that the value of the Casimir operator depends on the representation. In addition
it depends on the normalisation used ().
Example 16. SU(2) This group has rank (SU(2)) = 1, so we have one Casimir operator
C
1
=
1
2
J
2
(g
ij
=
1
ij
; = 2).
1
2
J
2
[ j, m ) =
1
2
j(j + 1)[ j, m )
where j is the spin or angular momentum.
j
.
([T
j
, T
k
])
= (ic
ijk
T
m
)
= ic
jkm
T
m
= ic
jkm
(T
m
)
= [T
J
, T
k
]
Denition 7 (Real). It there exists a similarity transformation such that
ST
j
S
1
= T
j
then D is said to be real.
8
SU(2)
Rotational group SO(3)
The Hamiltonian is invariant under rotation
H
(r) = U(
)H(r)U(
)
1
= H(r)
and the states are transformed as
[ ) [
) = U(
)[ )
these rotations form the group SO(3). Arbitrary rotation can be written as
U(
) = e
i
J
; = 1.
The generators obey
[J
j
, J
k
] = i
jkl
J
l
ans spherical symmetry gives
[J
i
, H] = 0; i
and at most one of the J
j
s can be diagonalized simultaneously.
In wave-mechanics
n [ lm ) = Y
m
l
()
forms a 2l + 1 dimensional basis for an irreducible representation with a given l.
We have one quadratic Casimir operator
C
1
=
1
2
J
2
.
We will now see that it is possible to construct the irreducible representations without
using C
1
, which is the standard procedure in books on quantum mechanics. The reson for
doing this is that we want a procedure that can be generalized to any compact group. We
begin with forming the ladder operators
J
=
1
2
(J
1
iJ
2
).
Diagonalize J
3
: J
3
[ j, m, ) = m[ j, m, ) where is all other degrees of freedom describ-
ing the state. We can take
j, m, [ j, m, ) =
] = J
[J
+
, J
] = J
3
which gives
J
3
(J
[ j, m, )) = . . . = (m1)J
[ j, m, ) .
Since this is a nite dimensional representation there is a state with maximal m = j
(the highest weight state) for which
J
+
[ j, j, ) = 0;
we dene
J
[ j, j, ) N
j
()[ j, j 1, )
where [ j, j 1, ) also can be chosen orthonormal. So we have
j, j 1, [ j, j 1, ) =
1
N
j
()N
j
()
j, j, [ J
+
J
[ j, j, ) = . . . =
j
N
j
()N
j
()
.
Here we see that N
j
() =
j =: N
j
is independent of . By applying ladder operators we
also get
J
[ j, j k, ) = N
jk
[ j, j k 1, )
J
+
[ j, j k 1, ) = N
jk
[ j, j k, )
So remains unchanged when applying the ladder operators. This gives
[J
+
, J
][ j, j k, ) = J
3
[ j, j k, ) N
2
jk
= N
2
jk+1
+j k
The recursion relation can be solved from
N
2
j
= j
N
2
j1
N
2
j
= j 1
. . .
N
2
jk
N
2
jk1
= j k
Taking the sum on both sides gives
N
2
jk
= (k + 1)j
k(j + 1)
2
=
1
2
(k + 1)(2j k)
By letting k = j m we get
N
m
=
1
2
_
(j +m)(j m+ 1) (1)
10
which we recognize. And we have obtained this without the use of the Casimir. What is
the minimal value of m? Lets apply J
[ j, j l, ) = 0; for some l.
We must have that N
jl
= 0 so from (1) we obtain l = 2j. So since j =
l
2
and l Z we
have that j is either integer of half integer.
For each we get a 2l +1 dimensional subspace, and so the representation is reducible,
but we have assumed that the representation is irreducible. This means that must be
unique. So we can instead label our states
[ j, m )
This is the spinj representation of SU(2).
Example 17. Let j =
1
2
and m =
1
2
, so
J
3
[
1
2
,
1
2
) =
1
2
[
1
2
,
1
2
)
apply a rotation by 2 about the z-axis,
= (0, 0, 2).
(x) = x [
1
2
,
1
2
) x [ exp(i
J) [
1
2
,
1
2
)
= x [ i(0, 0, 2) (0, 0,
1
2
) [
1
2
,
1
2
) = x [
1
2
,
1
2
)
The wave function changes sign.