Physics 251 Solution Set 4 Spring 2013

1. This problem concerns the Lie group SO(4) and its Lie algebra so(4).
(a) Work out the Lie algebra so(4) and verify that so(4) ∼
= so(3) ⊕ so(3).

The defining representation of the Lie algebra so(n) is

so(n) = {A | A ∈ gl(n, R) such that AT = −A} ,

where gl(n, R) is the set of all real n × n matrices. Recall that a suitable basis for the
defining representation of so(3), which consists of all 3 × 3 real antisymmetric matrices, is
(Ai )jk = −ǫijk , where i, j and k can take on the values 1, 2 and 3. To find a suitable basis
for the defining representation of so(4), one can generalize the Ai of so(3) by choosing
 
0
 −ǫ
ijk 0 
(Ai )jk =  , where i, j, k = 1, 2, 3 . (1)
 0 
0 0 0 0

Since a 4×4 real antisymmetric matrix has six independent parameters, we need to choose
three additional linearly-independent antisymmetric matrices to complete the basis for
so(4). We therefore introduce three antisymmetric matrices Bi by placing a 1 in one of
the non-diagonal elements of the fourth row (and a corresponding −1 required by the
antisymmetry property of the matrix), with all other elements zero. That is, a suitable
basis for so(4) is given by:
     
0 0 0 0 0 0 1 0 0 −1 0 0
0 0 −1 0  0 0 0 0 1 0 0 0
A1 =  0 1 0 0
, A 2 = 
−1 0 0 0
, A 3 = 0 0 0 0 ,
 

0 0 0 0 0 0 0 0 0 0 0 0
     
0 0 0 −1 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 −1 0 0 0 0
B1 = 
0
, B2 =  , B3 =  .
0 0 0 0 0 0 0 0 0 0 −1
1 0 0 0 0 1 0 0 0 0 1 0

One can easily verify that the six generators of so(4) satisfy the following commutation
relations:

[Ai , Aj ] = ǫijk Ak , [Bi , Bj ] = ǫijk Ak , [Ai , Bj ] = ǫijk Bk . (2)

Note that the commutation relations satisfied by the Ai are precisely those of so(3), which
is not surprising in light of eq. (1).

1
 The form of the commutators given in eq. (2) is not completely transparent. To un-
derstand the implications of eq. (2), it is convenient to define a new set of so(4) generators
that are real linear combinations of the Ai and Bi . Thus, we define,

Xi ≡ 21 (Ai + Bi ) , Yi ≡ 12 (Ai − Bi ) , where i = 1, 2, 3 . (3)

Using eq. (2), it is a simple matter to work out the commutation relations among the Xi
and Yi :
[Xi , Xj ] = ǫijk Xk , [Yi , Yj ] = ǫijk Yk , [Xi , Yj ] = 0 . (4)
Thus, we have succeeding in writing the so(4) commutation relations in such a way that
the generators {Xi } and {Yi } are decoupled. In particular, the {Xi } and {Yi } each satisfy
so(3) commutation relations. Hence, so(4) is a direct sum of two independent so(3) Lie
algebras. That is,1
so(4) ∼
= so(3) ⊕ so(3) . (5)

(b) What is the universal covering group of SO(4)? What is the center of SO(4)?
What is the adjoint group corresponding to the so(4) Lie algebra?

Since the universal covering group of SO(3) is SU(2), we can use eq. (5) to conclude that
the universal covering group of SO(4) is SU(2 )⊗ SU(2).2 In particular,

SO(4) ∼
= SU(2) ⊗ SU(2)/Z2 . (6)

To justify eq. (6), consider the centers of SO(4) and SU(2) ⊗ SU(2). The center of SO(4)
consists of all orthogonal matrices of unit determinant that are multiples of the identity.
There are only two such matrices, 14×4 and −14×4 , where 14×4 is the 4×4 identity matrix.
Hence,
Z(SO(4)) = Z2 .
The center of SU(2) is {12×2 , −12×2 } ∼= Z2 so that

Z(SU(2) ⊗ SU(2)) = Z2 ⊗ Z2 .

Thus, only one Z2 factor can appear in eq. (6).

Finally, the adjoint group by definition has a trivial center. Thus, the adjoint group
of SO(4) can be expressed in a number of equivalent forms,

SO(4)/Z2 ∼
= SO(3) ⊗ SO(3) ∼
= SU(2) ⊗ SU(2)/Z2 ⊗ Z2 ,

where we have made use of the well-known isomorphism, SO(3)∼ =SU(2)/Z2 . Indeed,
SO(3) ⊗ SO(3) has a trivial center since SO(3) has a trivial center.
1
Since su(2) ∼
= so(3) as Lie algebras, we can equally well write so(4) ∼
= su(2) ⊕ su(2).
2
Since the Lie group is obtained by exponentiation of the Lie algebra, a direct sum of Lie algebras
correspond to a direct product of Lie groups.

2
 (c) Calculate the Killing form of so(4) and verify that this Lie algebra is semisimple
and compact.

The Cartan-Killing form can be expressed in terms of the Lie algebra structure constants,
d c
gab = fac fbd . (7)

In this expression, the indices a, b, c and d range over 1, 2, . . . , 6, corresponding to the

six generators of so(4). It is easiest to evaluate gab in the basis {Xi , Yj } [cf, eqs. (3) and
(4)]. In this basis,

ǫijk ,
 for a = i , b = j , and d = k ,
d
fac = ǫijk , for a = i + 3 , b = j + 3 , and d = k + 3 ,

0, otherwise ,


where i, j and k range over 1, 2 and 3. Plugging into eq. (7) yields

gab = −2δab ,

which indicates that so(4) is a semi-simple and compact Lie algebra.

2. A Lie algebra g is defined by the commutation relations of the generators,

c
[ea , eb ] = fab ec .

Consider the finite-dimensional matrix representations of the ea . We shall denote the cor-
responding generators in the adjoint representation by Fa and in an arbitrary irreducible
representation R by Ra . The dimension of the adjoint representation, d, is equal to the
dimension of the Lie algebra g, while the dimension of R will be denoted by dR .
(a) Show that the Cartan-Killing metric gab can be written as gab = Tr(Fa Fb ).

The Cartan-Killing metric can be expressed in terms of the structure constants as follows,

gij = fiℓk fjk

ℓ
.

On the other hand, matrix elements of the adjoint representation are given by:

(Fi )j k = fikj .

Therefore,
Tr(Fi Fj ) = (Fi )k ℓ (Fj )ℓ k = fiℓk fjk
ℓ
= gij .

3
 (b) If g is a simple real compact Lie algebra, prove that for any irreducible represen-
tation R,
Tr(Ra Rb ) = cR gab ,
where cR is called the index of the irreducible representation R.

Consider a d-dimensional Lie algebra g, whose generators are represented by the matrices
Ra . These matrices satisfy the Lie algebra commutation relations,
c
[Ra , Rb ] = fab Rc , where a, b, c = 1, 2, . . . , d . (8)

We first note the following identity:

 
Tr [Ra , Rb ]Rc = Tr Ra [Rb , Rc ] . (9)

The proof of eq. (9) is straightforward:

 
Tr [Ra , Rb ]Rc = Tr (Ra Rb − Rb Ra )Rc = Tr(Ra Rb Rc ) − Tr(Rb Ra Rc )
 
= Tr(Ra Rb Rc ) − Tr(Ra Rc Rb ) = Tr Ra (Rb Rc − Rc Rb ) = Tr Ra [Rb , Rc ] ,

after using the cyclic properties of the trace. Making use of eq. (8) in eq. (9) yields:
d
fab Tr(Rd Rc ) = fbcd Tr(Ra Rd ) . (10)

To make further progress, recall that fabc ≡ gad fbcd is totally antisymmetric under the
interchange of any pair of indices a, b and c. It follows that

fbcd = g ad fabc , (11)

where g ad is the inverse Cartan metric tensor. It is convenient to multiply both sides of
eq. (10) by g ea to obtain:

g eafab
d
Tr(Rd Rc ) = g eafbcd Tr(Ra Rd ) . (12)

Using eq. (11) and the antisymmetry properties of fabh ,

g ea fab
d
= g ea g hdfhab = g eag hd fabh = g hdfbh
e
.

Inserting this result into eq. (12) yields

g hdfbh
e
Tr(Rd Rc ) = g eafbcd Tr(Ra Rd ) . (13)

Consider the d × d matrix whose matrix elements are

Ah c ≡ g hd Tr(Rd Rc ) . (14)

We can then rewrite eq. (13) in the following form:

e
fbh Ah c = fbcd Ae d . (15)

4
 e
We recognize fbh = (Fb )e h and fbcd = (Fb )d c . Hence, eq. (15) is equivalent to the ec
component of the matrix equation,
Fb A = AFb ,
for all b = 1, 2, . . . , d.
We proved in class that the adjoint representation of a simple Lie algebra (whose
generators are represented by the matrices Fb ) is irreducible. Applying Schur’s second
lemma to representations of Lie algebras,3 any matrix that commutes with all the Fb must
be a multiple of the identity. Hence, A = cI or equivalently.
g ed Tr(Rd Rc ) = cR δce ,
where cR is some complex constant. Using g ed geh = δhd , it immediately follows that
Tr(Rh Rc ) = cR ghc , (16)
which is the desired result.

(c) The quadratic Casimir operator is defined as C2 ≡ g ab ea eb where g ab is the inverse

of gab . Recall that C2 commutes with all elements of the Lie algebra. Hence, by Schur’s
lemma, C2 must be a multiple of the identity operator. Let us write C2 = C2 (R)I, where I
is the dR × dR identity matrix and C2 (R) is the eigenvalue of the Casimir operator in the
irreducible representation R. As noted above, d is the dimension of the Lie algebra g.
Show that C2 (R) is related to the index cR by
dcR
C2 (R) = .
dR
Check this formula in the case that R is the adjoint representation.

By definition,
C2 (R)I = g ab Ra Rb , (17)
where I is the dR × dR identity matrix, dR is the dimension of the representation R, and
a, b = 1, 2, . . . , d. Taking the trace of eq. (17) and using eq. (16), it follows that:
dR C2 (R) = g ab Tr(Ra Rb ) = cR g ab gab = cR d ,
since g ab gab = δaa = d. Hence, solving for C2 (R), one obtains:
dcR
C2 (R) = . (18)
dR
3
A review of the proof given in class of Schur’s lemmas (which were applied to group representations)
reveals that it also applies to representations of Lie algebras. Indeed, for any algebraic structure A ,
Schur’s second lemma states that if there exists a matrix M such that D(A)M = M D(A) for all A ∈ A ,
where D(A) is an n-dimensional irreducible matrix representation of A (over a complex representation
space Cn ), then it follows that M must be a multiple of the identity matrix. In particular, any element
of a Lie algebra A can be expressed as some linear combination of the the generators Aa (which serve as
a basis for the Lie algebra). Consequently, if D(Aa )M = M D(Aa ) for all a = 1, 2, . . . , d, then it follows
that D(A)M = M D(A) for all A ∈ A , and Schur’s second lemma applies.

5
 For the adjoint representation (usually denoted by R = A), we have dA = d. Moreover,
the adjoint representation generators are (Ra )b c = fac
b
, as shown in class. Hence,
Tr(Ra Rd ) = (Ra )b c (Rd )c b = fac
b c
fdb = gad ,
where we used the definition of the Cartan metric tensor at the last step. Comparing
this result with that of eq. (16) yields cA = 1. Hence, eq. (18) implies that C2 (A) = 1 in
agreement with the theorem proven in class.

(d) Compute the index of an arbitrary irreducible representation of su(2).

For su(2), the irreducible representations are labeled by j = 0, 21 , 1, 32 . . .. The quadratic

Casimir operator is proportional to J12 + J22 + J32 , where [Ji , Jj ] = iǫijk Jk in the physicist
convention. Since the eigenvalue of of J12 + J22 + J32 is j(j + 1), we shall adjust the
overall normalization of the Casimir operator so that C2 (A) = 1. Given that the adjoint
representation of su(2) corresponds to j = 1, it follows that:

C2 (j) = 21 j(j + 1) .
We can now use eq. (18) to obtain the index of an irreducible representation of su(2).
Using dR = 2j + 1 for the irreducible representation labeled by j, it follow that the index
cR is given by
c(j) = 16 j(j + 1)(2j + 1) .
In the defining representation, j = 12 , and we find cF ≡ c( 12 ) = 41 . In the adjoint
representation, j = 1 and we find that cA ≡ c(1) = 1 as expected from part (b).

(e) Compute the index of the defining representation of su(3) and generalize this result
to su(n).

First, consider the Lie algebra su(3). We choose the generators in the defining represen-
tation to be the Gell-Mann matrices, 21 λa . Following the mathematician’s conventions,
we define Ta ≡ − 21 iλa so that
[Ta , Tb ] = fabc Tc ,
where the fabc are the totally antisymmetric structure constants in the convention where
the Ta satisfy
Tr(Ta Tb ) = − 14 Tr(λa λb ) = − 21 δab , (19)
using the explicit form for the Gell-Mann matrices displayed in class. In this basis choice,
c c
gab = fad fbd = −3δab ,
using the explicit form for the su(3) structure constants listed in the class handout on
SU(3). The index of the defining representation, usually denoted by cF (since physicists
also refer to this representation as the fundamental representation), can be obtained from
eq. (16),
Tr(Ta Tb ) = cF (−3δab ) .

6
Using eq. (19) to compute the trace, we end up with

cF = 16 .

To generalize these results to su(n), we shall make use of the construction of the su(n)
Lie algebra given in class. There, we defined traceless n × n Hermitian matrices,
1
(F a b )cd = δbc δda − δba δcd ,
n
which satisfy the commutation relations,

[F a b , F c d ] = δda F c b − δbc F a d . (20)

The generalized Gell-Mann matrices are:

   
1 2 σ1 0 1 2 σ2 0
λ1 = F 2 + F 1 = , λ2 = i(F 2 − F 1) = ,
0 0 0 0
 
1 2 σ3 0
λ3 = F 1 −F 2 = , etc. (21)
0 0

where the Pauli matrices occupy the upper left 2 × 2 block of the n × n matrix generators
(with all other elements zero). In the mathematician’s convention, we define Ta = − 12 iλa
and [Ta , Tb ] = fabc Tc , where the fabc are totally antisymmetric and Tr(Ta Tb ) ∝ δab . To
compute the constant of proportionality, one can check for example that

Tr(T3 T3 ) = − 41 Tr(λ3 λ3 ) = − 12 ,

using eq. (21). Clearly, the constant of proportionality does not depend on the choice of a
and b. Hence, it follows that the generators of su(n) in the defining representation satisfy

Tr(Ta Tb ) = − 21 δab . (22)

Next, we evaluate the Cartan metric tensor, which is given by:

c d
gab = fad fbc . (23)

In the convention where the generators satisfy Tr(Ta Tb ) ∝ δab , the Cartan metric tensor
also satisfies gab ∝ δab , in light of eq. (16). To determine the proportionality constant,
consider
[T3 , Tc ] = f3cd Td .
We can evaluate g33 = f3dc f3cd by examining eq. (20). In particular,

[T3 , F 2 1 ] = F 2 1 , [T3 , F 1 2 ] = −F 1 2 , [T3 , F a 1 ] = 21 F a 1 , [T3 , F 1 a ] = − 21 F 1 a ,

[T3 , F a 2 ] = − 12 F a 1 , [T3 , F 2 a ] = 12 F 1 a , [T3 , F a b ] = [T3 , F b a ] = 0 , (24)

7
for a 6= b and a, b = 3, 4, . . . , n. Note that the non-diagonal generators Tc of the form
F a b + F b a and i(F a b − F b a ) for a < b with a = 1 or a = 2 are the only generators that do
not commute with T3 . Eq. (24) provides the necessary information to evaluate g33 ,


g33 = (+1)(−1) + (n − 1) 12 )(− 21 = −n .
where the first term on the right-hand side derives from f312 f321 , whereas the remaining
terms derive from the remaining combination of non-zero structure constants. That is,
c d
gab = fad fbc = −nδab .
The index of the defining representation can be obtained from eq. (16),
Tr(Ta Tb ) = cF (−nδab ) .
Using eq. (22) to compute the trace, we end up with
1
cF = . (25)
2n
One sees that this general result is consistent with the corresponding results of su(2) and
su(3) previously obtained.

Remarks:
Using eqs. (18) and (25), one can compute the eigenvalue of the quadratic Casimir
operator in the defining representation of su(n). In particular, since d = n2 − 1, dF = n
and cF = 1/(2n), it follows that:
n2 − 1
C2 (F ) = .
2n2
Moreover, the Casimir operator in the defining representation of su(n) is given by
C2 (A) = 1 ,
according to the theorem proved in class. However, note that the Casimir operator of
su(n) is defined in an arbitrary irreducible representation R by
n −1 2

ab 1 X
C2 = g Ra Rb = − Ra Ra , (26)
n a=1

where we have used eq. (23) [recall that g ab is the inverse of gab ]. In the physics literature,
in the case of su(n) one typically defines C2 by omitting the overall factor of 1/n in
eq. (26). Consequently, C2 (R) is a factor of n larger than indicated above, in which case
n2 − 1
C2 (F ) = , Ca (A) = n .
2n
Additional details on the Casimir operator and index of an irreducible representation
of a simple Lie algebra can be found in the class handout entitled, The eigenvalues of the
quadratic Casimir operator and second-order indices of a simple Lie algebra.

8
3. Consider the simple Lie algebra generated by the ten 4 × 4 matrices: σa ⊗ I, σa ⊗ τ1 ,
σa ⊗ τ3 and I ⊗ τ2 , where (I, σa ) and (I, τa ) are the

2 × 2 identity and Pauli matrices in
1 0
orthogonal spaces. For example, since τ3 = 0 −1 , we obtain in block matrix form:
 
σa 0
σa ⊗ τ3 = , (a = 1, 2, 3) ,
0 −σa
where 0 is the 2 × 2 zero matrix. The remaining seven matrices can be likewise obtained.
Take H1 = σ3 ⊗ I and H2 = σ3 ⊗ τ3 as the Cartan subalgebra. Note that if A, B, C, and
D are 2 × 2 matrices, then (A ⊗ B)(C ⊗ D) = AC ⊗ BD.
(a) Find the roots. Normalize the roots such that the shortest root vector has length 1.

First, we write out the ten generators explicitly in block matrix form:
 
0 σa
Aa ≡ σa ⊗ τ1 = , (a = 1, 2, 3) ,
σa 0
 
σa 0
Ba ≡ σa ⊗ τ3 = , (a = 1, 2, 3) ,
0 −σa
 
σa 0
Ca ≡ σa ⊗ I = , (a = 1, 2, 3) ,
0 σa
 
0 −iI
D ≡ I ⊗ τ2 = . (27)
iI 0
To check that these generators actually generate a Lie algebra, we work out all the com-
mutation relations:

[Aa , Ab ] = 2iǫabc Cc , [Ba , Bb ] = 2iǫabc Cc , [Ca , Cb ] = 2iǫabc Cc ,

[Aa , Bb ] = −2iδab D , [Aa , Cb ] = 2iǫabc Ac , [Ba , Cb ] = 2iǫabc Bc ,
[Aa , D] = 2iBa , [Ba , D] = −2iAa , [Ca , D] = 0 , (28)

where we have used σa σb = Iδab + iǫabc σc . For example,

     
0 σa σb 0 σb 0 0 σa
[Aa , Bb ] = Aa Bb − Bb Aa = −
σa 0 0 −σb 0 −σb σa 0
 
0 −(σa σb + σb σa )
=
σa σb + σb σa 0
 
0 −2Iδab
= = −2iδab D . (29)
2Iδab 0
Alternatively, one can derive the commutation relations displayed in eq. (28) by em-
ploying the direct product representation of the Lie algebra generators given in eq. (27)

9
and using (A ⊗ B)(C ⊗ D) = AC ⊗ BD. For example, eq. (29) can also be obtained as
follows.

[Aa , Bb ] = (σa ⊗ τ1 )(σb ⊗ τ3 ) − (σb ⊗ τ3 )(σa ⊗ τ1 )

= (σa σb ) ⊗ (τ1 τ3 ) − (σb σa ) ⊗ (τ3 τ1 )
= (σa σb ) ⊗ (−iτ2 ) − (σb σa ) ⊗ (iτ2 )
= (σa σb + σb σa ) ⊗ (−iτ2 )
= (2Iδab ) ⊗ (−iτ2 ) = −2iδab I ⊗ τ2 = −2iδab D .

All other commutation relations are easily derived using either of the methods shown
above. Thus, the ten generators {Aa , Ba , Ca , D} generate a Lie algebra, since the com-
mutation relations close.
To determine the roots, we take H1 = σ3 ⊗ I = C3 and H2 = σ3 ⊗ τ3 = B3 to generate
the Cartan subalgebra. Indeed, these two generators are diagonal in the representation
given in eq. (27). We now rewrite the commutation relations given in eq. (28) in the
Cartan-Weyl form. Starting from the commutation relations,

[B3 , A1 ] = [B3 , A2 ] = 0 , [C3 , A1 ] = 2iA2 , [C3 , A2 ] = −2iA1 ,

it is clear that we should define A± ≡ A1 ± iA2 , in which case,

[B2 , A± ] = 0 , [C3 , A± ] = ±2A± . (30)

Next, we focus on the commutation relations,

[B3 , A3 ] = 2iD , [B3 , D] = −2iA3 , [C3 , A3 ] = [C3 , D] = 0 .

These results motivate the definition D± ≡ A3 ± iD, in which case,

[B3 , D± ] = ±2D± , [C3 , D± ] = 0 . (31)

The remaining commutation relations are:

[B3 , B1 ] = 2iC2 , [B3 , B2 ] = −2iC1 , [B3 , C1 ] = 2iB2 , [B3 , C2 ] = −2iB1 ,

(32)
[C3 , B1 ] = 2iB2 , [C3 , B2 ] = −2iB1 , [C3 , C1 ] = 2iC2 , [C3 , C2 ] = −2iC1 .
(33)

Defining B± ≡ B1 ± iB2 and C± ≡ C1 ± iC2 , eqs. (32) and (33) can be rewritten as:

[B3 , B± ] = ±2C± , [B3 , C± ] = ±2B± , [C3 , B± ] = ±2B± ,

[C3 , C± ] = ±2C± .
(34)
Thus, if we define F± ≡ B± + C± and G± ≡ B± − C± , the eq. (34) will be in Cartan-Weyl
form,

[B3 , F± ] = ±2F± , [B3 , F± ] = ±2F± , [C3 , G± ] = ∓2G± , [C3 , G± ] = ±2G± .

(35)

10
To summarize, eqs. (30), (31) and (35) provide the Cartan-Weyl form for the commutation
relations among the generators of the Cartan subalgebra and the off-diagonal generators
Eα ≡ {A± , D± , E± , F± }.
The root vectors are defined by the Cartan-Weyl form for the Lie algebra commutation
relations, [Hi , Eα ] = αi Eα , for i = 1, 2, . . . , r, where r is the number of diagonal generators
(and is equal to the rank of the Lie algebra). In the present example, r = 2, H1 = C3 ,
H2 = B3 and the off diagonal generators are Eα ≡ {A± , D± , E± , F± }. Hence, we identify
the root vectors derived from the non-diagonal generators:

A± : ±(0, 2) , D± : ±(2, 0) , (36)

F± : ±(2, 2) , G± : ±(−2, 2) , (37)

where the first entry of the root vector is the eigenvalue of adC3 and the second entry of
the root vector is the eigenvalue of adB3 The Cartan metric can be computed from the
formula derived in class, X
gij = αi αj .
α

From the four root vectors obtained in eqs. (36) and (37), we immediately obtain

gij = 24δij . (38)

1
The inverse Cartan metric is g ij = δ .
24 ij
One can now define the inner product on the
root space,
(α, β) = g ij αi βj . (39)
The squared-length of a root vector α is given by
2
X
ij
(α, α) = g αi αj = αi αi .
i=1

It is convenient to redefine the inner product given in eq. (39) by introducing an overall
multiplicative positive constant such that the new inner product is Euclidean,
X
(α, β) = αi βi .
i

Moreover, we are always free to rescale the generators of the Cartan subalgebra (which
rescales the root vectors) in such a way that the shortest root vector has length 1. In
these conventions, the rescaled roots are given by [cf. eqs. (36) and (37)]:

±(0, 1) , ±(1, 0) , ±(1, 1) , ±(−1, 1) .

and the corresponding root diagram is shown in Fig. 1, which we recognize as the root
diagram for the rank-2 Lie algebra sp(2) ∼
= so(5).4
4
In the notation used here, sp(n) is a Lie algebra of rank n. However, many books denote this Lie
algebra by sp(2n). Both conventions are common in the mathematics and physics literature.

11
 (0, 1)
(−1, 1) (1, 1)

(−1, 0) (1, 0)

(−1, −1) (1, −1)

(0, −1)

Figure 1: The root diagram for sp(2) ∼

= so(5).

(b) Determine the simple roots and evaluate the corresponding Cartan matrix. Deduce
the Dynkin diagram for this Lie algebra and identify it by name.

The simple roots correspond to the two smallest positive roots. These are

α1 ≡ (0, 1) , and α2 ≡ (1, −1) . (40)

It is a simple matter to check that the other two positive roots can be expressed as sums
of simple roots,
(1, 0) = α1 + α2 , (1, 1) = 2α1 + α2 .
The Cartan matrix is defined by:5
2(αi , αj )
Aij = , (41)
(αi , αi )
P
where the inner product (α, β) ≡ i αi βi in the convention where gij = δij . Using
eq. (40), we obtain A11 = A22 = 2, A12 = −2 and A21 = −1. That is,
 
2 −2
A= . (42)
−1 2

The structure of the Dynkin diagram depends on the angle between the two simple roots:
(α1 , α2 ) −1
cos ϕα1 α2 = p =√ .
(α1 , α1 )(α2 , α2 ) 2

Hence ϕα1 α2 = 135◦ , which corresponds to a double line connecting the two balls of the
Dynkin diagram. Hence, the Dynkin diagram corresponding to the Lie algebra, whose
simple roots are given by eq. (40), is exhibited in Fig. 2, where the shaded ball corresponds

12
 α1 α2
Figure 2: The Dynkin diagram for sp(2) ∼
= so(5).

to the simple root of the smallest length. In Cartan’s notation, this Lie algebra is B2 ∼
= C2 ,
∼
which corresponds to sp(2) = so(5) as noted at the end of part (a).

(c) The fundamental weights mi are defined such that

2(mi , αj )
= δij , for i, j = 1, 2, . . . , r , (43)
(αj , αj )
where αj ∈ Π (the set of simple roots) and r is the rank of the Lie algebra. Using the
results of part (b), determine the fundamental weights.

We can solve for the mi by expanding the fundamental weight vectors in terms of the
simple roots:
Xr
mi = cki αk .
k=1

Inserting this expression into eq. (43) yields,

r
X 2(αk , αj )
cki = δij ,
k=1
(αj , αj )

which can be expressed in terms of the Cartan matrix A,

r
X
cki Ajk = δij .
k=1

This implies that c = A−1 , and we conclude that

r
X
mi = (A−1 )ki αk . (44)
k=1

Using the Cartan matrix given in eq. (42), the inverse is easily obtained:
 
−1 1 2 2
A =2 .
1 2
Thus, eq. (44) yields the two fundamental weights of sp(2) ∼
= so(5),
m1 = α1 + 12 α2 = ( 12 , 21 ) , (45)
m2 = α1 + α2 = (1 , 0) , (46)
where we have used eq. (40) for the simple roots.
5
Warning: in the mathematics literature, eq. (41) is often employed as the definition of the transposed
Cartan matrix. You should check carefully when using results from books on Lie algebras.

13
 Each of the two fundamental weights is the highest weight for a particular irreducible
representation of sp(2) ∼
= so(5), sometimes called a fundamental representation. In gen-
eral, a simple Lie algebra of rank ℓ possesses precisely ℓ inequivalent fundamental ir-
reducible representations. Given a highest weight, it is possible to construct all the
weights of the corresponding irreducible representation by a technique that is exhibited
in Appendix A. Thus, one can construct the weight diagrams corresponding to the two
fundamental representations corresponding to m1 and m2 , respectively (cf. Fig. 4 in
Appendix A).

Appendix A: Obtaining a complete set of weights

The complete set of weights for the irreducible representations of sp(2) ∼
= so(5) corre-
sponding to the highest weights m1 and m2 , respectively, can be obtained by the method
of block weight diagrams described in Robert N. Cahn, Semi-Simple Lie Algebras and
Their Representations (Dover Publications, Inc., Mineola, NY, 2006).6 Given a high-
est weight M , the corresponding Dynkin labels ki (which are non-negative integers) are
defined by7
2(M , αi )
ki ≡ , where αi ∈ Π . (47)
(αi , αi )
The irreducible representations of the Lie algebra g are often denoted by placing the ith
Dynkin label ki above the ith ball of the Dynkin diagram (which corresponds to the ith
simple root αi ). Thus, the block weight diagrams corresponding to the two fundamental
representations of sp(2) ∼
= so(5) are exhibited below.
1 1
and

1 0 0 1

−1 1 2 −1

1 −1 0 0

−1 0 −2 1

0 −1

Figure 3: The block weight diagrams of the fundamental irreducible representations of sp(2) ∼
= so(5).

The block weight diagrams are obtained using the theorem quoted in class that estab-
lishes strings of weights of the form
2(m, αi ) 2(m, αi )
− kAji , for values of k = 0, 1, 2, , . . . , .
(αi , αi ) (αi , αi )
6
However, note that Cahn defines the Cartan matrix that is the transpose of our definition.
7
Do not confuse the Dynkin labels of a weight with its coordinates in weight space given in eqs. (45)
and (46). For example, the fundamental weight m1 = α1 + 21 α2 = ( 12 , 21 ), whereas its Dynkin labels are
(k1 , k2 ) = (1 , 0).

14
Thus, starting with any weight m, the Dynkin labels for the weights appearing below it
in the block weight diagram are obtained by subtracting off the jth column of the Cartan
matrix n times, where n is the jth positive Dynkin label of the weight.8 Applying the
above algorithm has produced the Dynkin labels of the four weights corresponding to
the representation specified by m1 and the five weights corresponding to representation
specified by m2 . In this method, the computation of the multiplicity of a given weight
requires additional analysis. But, for the simple cases treated above, all weights appear
with multiplicity equal to one, in which case the dimension of the representation is simply
equal to the number of weights in the block weight diagram.
Hence, the representations depicted by the block weight diagrams of Fig. 3 are four-
dimensional and five-dimensional, respectively, The four-dimensional representation, cor-
responding to the highest weight m1 , is precisely the matrix representation given in
eq. (27). This is either the defining representation of sp(2) or the spinor representation of
so(5).9 In contrast, m2 is the highest weight of a five-dimensional representation, which
corresponds to the defining representation of so(5).
It is instructive to re-express the weights in terms of its coordinates in the vector
space spanned by the simple roots. The weights can then be depicted as vectors in
a two-dimensional plane. Given a weight specified by its Dynkin labels (k1 , k2 ), the
corresponding weight m is obtained by solving the equations [cf. eq. (47)]:

2(m, α1 ) 2(m, α2 )
k1 ≡ , k2 ≡ . (48)
(α1 , α1 ) (α2 , α2 )

To solve for m in terms of k1 and k2 , we expand m as a linear combination of simple

roots [which are given explicitly in eq. (40)],

m = c1 α1 + c2 α2 . (49)

Inserting this expression for m into eq. (48), it follows that:

2(c1 α1 + c2 α2 , α1 )
k1 = = 2c1 − 2c2 ,
(α1 , α1 )
2(c1 α1 + c2 α2 , α2 )
k2 = = −c1 + 2c2 ,
(α2 , α2 )

where we have used (α1 , α1 ) = 1, (α1 , α2 ) = −1 and (α2 , α2 ) = 2. Solving for c1 and c2
then yields:
c1 = k 1 + k 2 , c2 = 21 k1 + k2 . (50)
Hence, using eqs. (40) and (50), the weight m specified by eq. (49) is given by:


m = 21 k1 + k2 , 12 k1 . (51)
8
If there is more than one positive Dynkin label, then the block weight diagram branches. This does
not occur in the present example.
9
Since sp(2) ∼= so(5), one is free to regard the representations obtained above as representations of
either Lie algebra. However, the interpretation of the representation depends on which choice of Lie
algebra is made.

15
As a check, if m = m1 then k1 = 1 and k2 = 0, in which case c1 = 1, c2 = 21 and
m1 = ( 21 , 21 ) in agreement with eq. (45). Likewise, if m = m2 then k1 = 0 and k2 = 1,
in which case c1 = c2 = 1 and m1 = (1 , 0) in agreement with eq. (46).
One can use eq. (51) to obtain the coordinates of all the weights exhibited in Fig. 3.
For the four-dimensional representation specified by the Dynkin labels (1, 0) and the five-
dimensional representation specified by the Dynkin labels (0, 1), the corresponding weight
space diagrams are given in Fig. 4.10

T2 T2
(0, 1)

− 21 , 12 1 1



2, 2

(0, 0) (1, 0)
T1 (−1, 0) T1

− 21 , − 21 1 1



2, −2

(0, −1)

Figure 4: The weight diagrams of the irreducible representations of sp(2) ∼

= so(5), with dimensions four
[left] and five [right], respectively.

In particular, T1 ≡ 21 H1 = 12 C3 and T2 ≡ 12 H2 = 12 B3 are the diagonal generators normal-

ized such that the shortest root vector has length 1. Given the explicit four-dimensional
representation
 1 1 in eq. (27), one can check that the weight vectors exhibited in Fig. 4,
( 2 , 2 ) , ( 2 , − 2 ) , (− 12 , 12 ), (− 12 , − 12 ) ,11 satisfy the eigenvalue equations,12
1 1

Ti |mi = mi |mi , for i = 1, 2 , (52)

where m = (m1 , m2 ) are the coordinates in the  T1 –T2 plane. The weights of the 13five-
dimensional representation shown in Fig. 4, (1, 0), (0, 1), (0, 0), (0, −1), (−1, 0) , in-
clude a zero weight (indicated by the filled circle at the origin of the weight diagram). To
check that eq. (52) is satisfied in this latter case, it is straightforward to construct explicit
five-dimensional matrix representations of T1 and T2 , which are the diagonal generators
of the defining representation of so(5).
10
As previously noted, all weights shown in the two weight space diagrams above have multiplicity one,
which means that the corresponding simultaneous eigenvector |mi defined in eq. (52) is unique.
11
The corresponding Dynkin indices, obtained in Fig. 3, are (k1 , k2 ) = (1, 0) , (−1, 1) , (1, −1) and
(−1, 0), respectively.
12
Sometimes, the eigenvalues m1 and m2 are called weights and the corresponding eigenvector |mi is
called the weight vector. However, it is more common to refer to the weight vector m of a weight space
diagram as the vector whose coordinates (m1 , m2 ) are given by the eigenvalues of T1 and T2 .
13
The corresponding Dynkin indices, obtained in Fig. 3, are (k1 , k2 ) = (0, 1) , (2, −1) , (0, 0) , (−2, 1)
and (0, −1), respectively.

16