The Ricci curvature tensor of a bi-invariant metric for semi-simple group

is proportional to the metric

1
Ric = g , (3.211)
4
and, therefore, the scalar curvature is
n
R= . (3.212)
4

3.12 SL(2, C) and Its Subgroups: SL(2, R), SU(2), SU(1, 1)

and SO(1, 2)
The group SL(2, C) is the group of non-degenerate complex matrices with
unit determinant. Of course, it has the subgroup SL(2, R) of real non-
degenerate matrices with unit determinant, as well as the subgroup SU(2) of
complex unitary matrices with unit determinant. The Lie algebra of SL(2, C)
consists of traceless complex matrices, while the Lie algebra of SL(2, R) con-
sists of traceless real matrices and the Lie algebra of SU(2) consists of trace-
less complex anti-Hermitian matrices. Obviously, the dimensions of these
groups are

dim SL(2, C) = 6 , dim SL(2, R) = dim SU(2) = 3 . (3.213)

In the Lie algebra of SL(2, C) we can choose the following basis

     
1 0 i 1 0 1 1 −i 0
σ1 =  , σ2 =  , σ3 =  ,
2 i 0 2 −1 0 2 0 i
     (3.214)

1 0 1  1  0 −i  1  −1 0 
Σ1 = , Σ2 = , Σ3 = ,
2 1 0 2 i 0 2 0 1
(3.215)
so that the matrices σj are anti-Hermitian and the matrices Σk are Hermitian,
that is,
σj† = −σj Σ†k = Σk . (3.216)

riemgeom.tex; April 9, 2010; 13:23; p. 61

The Lie algebra formed by these matrices has the form

[σi , σj ] = εijk σk , (3.217)

[σi , Σj ] = εijk Σk , (3.218)
[Σi , Σj ] = −εijk σk , (3.219)

where εijk is the Levi-Civita symbol. More explicitly,

[σ1 , σ2 ] = σ3 , [σ2 , σ3 ] = σ1 , [σ3 , σ1 ] = σ2 , (3.220)

[σ1 , Σ2 ] = Σ3 , [σ2 , Σ3 ] = Σ1 , [σ3 , Σ1 ] = Σ2 , (3.221)
[Σ1 , σ2 ] = Σ3 , [Σ2 , σ3 ] = Σ1 , [Σ3 , σ1 ] = Σ2 , (3.222)
[Σ1 , Σ2 ] = −σ3 , [Σ2 , Σ3 ] = −σ1 , [Σ3 , Σ1 ] = −σ2 . (3.223)

Now, we immediately see that this algebra has a subalgebra formed by

the matrices σ1 , σs and σ3 . This is the Lie algebra of SU(2).
Another subalgebra is formed by the real traceless matrices

B1 = Σ 1 , B2 = σ2 , B3 = Σ3 . (3.224)

It has the form

[B1 , B2 ] = B3 , [B2 , B3 ] = B1 , [B3 , B1 ] = −B2 . (3.225)

This is the Lie algebra of SL(2, R). We see that it is very similar to the Lie
algebra of SU(2). The only difference is the sign of the last commutator.
The third subalgebra is formed by the matrices

A1 = σ 3 , A2 = Σ 1 , A3 = Σ 2 , (3.226)

it has the form

[A1 , A2 ] = A3 , [A2 , A3 ] = −A1 , [A3 , A1 ] = A2 . (3.227)

This is the algebra of complex traceless matrices that satisfy the identity

A†i = −ηAi η −1 , (3.228)

where η is the matrix  

1 0
η= . (3.229)
0 −1

riemgeom.tex; April 9, 2010; 13:23; p. 62

The corresponding group SU(1, 1) is the group of complex matrices with unit
determinant that preserve the bilinear form η in the sense
U † = ηU −1 η −1 . (3.230)
This algebra is isomorphic to the Lie algebra of the group SO(1, 2).
In the following we will consider these algebras simultaneously. We in-
troduce parameters (λi ) = (λ1 , λ2 , λ3 ) such that the algebra is given by
[G1 , G2 ] = λ3 G3 , [G2 , G3 ] = λ1 G1 , [G3 , G1 ] = λ2 G2 . (3.231)
Then for (λi ) = (1, 1, 1) this is the algebra of SU(2), while for (λ i ) = (1, −1, 1)
this is the algebra of SL(2, R), and for (λi ) = (−1, 1, 1) this is the algebra of
SU(1, 1).
Notice also that if one of the parameters is equal to zero, this describes
the solvable Lie algebra of the group E(2) of motions of the R 2 , for example, if
(λi ) = (1, 0, 1) this is the Lie algebra of the group of motions of the xz-plane
in R3 . In this case, G2 generates rotations around the y-axis and G1 and G3
generate translations along the x-axis and the z-axis. If two parameters are
equal to zero, then this is the nilpotent Heisenberg algebra H3 (R). Finally,
if all three parameters are equal to zero then this is just an Abelian algebra
of R3 . We will only consider the case when all parameters are not equal to
zero.
The structure constants of the group define the generators of the adjoint
representation
     
0 0 0 0 0 λ1 0 −λ1 0
     
     
C1 =  0 0 −λ2  , C2 =  0 0 0  , C3 =  λ2 0 0 ,
     
0 λ3 0 −λ3 0 0 0 0 0
(3.232)
We see that the matrices Ca are traceless, which follows, of course, from the
fact that the group is uni-modular. Moreover, for (λ i ) all matrices Ca are
anti-symmetric.
Let xa be the canonical coordinates on the group. Then
 
0 −λ1 x3 λ1 x2
 
 
C(x) = Ca xa =  λ2 x3 0 −λ2 x1  . (3.233)
 
−λ3 x2 λ3 x1 0

riemgeom.tex; April 9, 2010; 13:23; p. 63

 This enables us to compute the Cartan metric γ = (γab ). We obtain
 
λ2 λ3 0 0
1  
 
γ = − (tr Ca Cb ) =  0 λ 1 λ3 0 . (3.234)
2  
0 0 λ 1 λ2

We see that if all parameters λi are not equal to zero, then this metric is
non-degenerate, and, therefore, the groups SL(2, R), SU(2) and SU(1, 1) are
semisimple.
We parametrize the parameters λi by complex parameters µi such that

λi = µ2i , (3.235)

and introduce the parameters

ω 1 = µ 2 µ3 , ω 2 = µ 1 µ3 , ω 3 = µ 1 µ2 . (3.236)

Then
ω2 ω3 ω1 ω3 ω1 ω2
λ1 = , λ2 = , λ3 = , (3.237)
ω1 ω2 ω3
and the Cartan metric takes the form
 
ω2 0 0
 1 
 
γ =  0 ω22 0 . (3.238)
 
0 0 ω32

This suggest rescaling the coordinates by

yi = ω i xi , (no summation!) (3.239)

Notice that these coordinates are complex, in general. Then

 
0 − ωω12 y3 ωω31 y2
 
 
C(x) =  ωω1 y3 0 − ω3
y
ω2 1 
. (3.240)
 2
− ωω31 y2 ωω23 y1 0

riemgeom.tex; April 9, 2010; 13:23; p. 64

 Now, let Ω be a square root of the Cartan metric defined by
 
ω1 0 0
 
 
Ω =  0 ω2 0  , (3.241)
 
0 0 ω3

so that
Ω2 = γ . (3.242)
Then it is easy to check that

C(x) = ΩS(x)Ω−1 , (3.243)

where S(x) is anti-symmetric matrix

 
0 −y3 y2
 
 
S(x) =  y3 0 −y1  . (3.244)
 
−y2 y1 0

To rewrite it in a more compact and more familiar form, we introduce

the generators Ta of the group SU(2) in adjoint representation defined by

(Ta )bc = εbac , (3.245)

or, more explicitly,

     
0 0 0 0 0 1 0 −1 0
     
     
T1 =  0 0 −1 , T2 =  0 0 0  , T3 =  1 0 0  .
     
0 1 0 −1 0 0 0 0 0
(3.246)
Then the matrix S(x) takes the form

S(x) = Ta ya , (3.247)

or
Sab (x) = −εabc yc . (3.248)

riemgeom.tex; April 9, 2010; 13:23; p. 65

Therefore, the matrix C is now

Cab (x) = −Ωad εdf c Ω−1 c

fb y . (3.249)

One can show that the products of the matrices Ta form the Lie algebra

[Ta , Tb ] = εabc Tc , (3.250)

and, moreover, their products are

(Ta Tb )cd = −δab δcd + δad δbc . (3.251)

This form of the matrix S(x) and the matrix C(x) greatly simplifies the
calculations. Indeed, one can easily compute the square of the matrix S

S 2 = −y 2 P , (3.252)

where

y 2 = y12 + y22 + y32 = γab xa xb = ω12 x21 + ω22 x22 + ω32 x23 , (3.253)

and P is a projection on the plane orthogonal to vector y a defined by

ya yb
Pab (x) = δab − , (3.254)
y2
and satisfying the equations

P2 = P , P S = SP = S , tr P = 2 . (3.255)

Therefore,
S 2n = (−y 2 )n P , S 2n+1 = (−y 2 )n S . (3.256)
Thus, for any analytic function of S one can compute
� �
1 1
f (S) = f (0)I+ [f (r) + f (−r)] − f (0) P + [f (r) − f (−r)] S , (3.257)
2 2r
�
where r = −y 2 . By using this equation one can compute now any analytic
function of the matrix C(x); we get
� �
1 1
f (C) = f (0)I + [f (r) + f (−r)] − f (0) Π + [f (r) − f (−r)] C ,
2 2r
(3.258)

riemgeom.tex; April 9, 2010; 13:23; p. 66

where Π is another projection defined by
Π = ΩP Ω−1 . (3.259)
It satisfies the identities
Π2 = Π , CΠ = ΠC = C , tr Π = 2 . (3.260)
Notice that the projection Π could be determined by the square of the
matrix C,
C 2 = −y 2 Π . (3.261)
Notice that the matrix C(x) and the invariant y 2 are real; therefore, the
projection Π is also real.
Let us introduce rescale coordinates according to
x̃a = ωa2 xa = λa xa , (no summation!) , (3.262)
so that
y 2 = x̃a xa . (3.263)
Then the projection Π has the form
1
Πab = δab − x̃a xb . (3.264)
y2
Now, we can compute everything in canonical coordinates. The matrix
X determining the left-invariant one-forms has the form
� �
exp C − I sinh r cosh r − 1
X= =I+ −1 Π+ C, (3.265)
C r r2
and its inverse, L = X −1 , determining the left-invariant vector fields is
C �r �r� � 1
L= =I+ coth −1 Π− C. (3.266)
exp C − I 2 2 2
The matrix Y determining the one-forms and the matrix R = Y −1 determin-
ing the right-invariant vector fields are obtained by just changing the sign of
x, which is equivalent to changing the sign of C, that is,
� �
I − exp(−C) sinh r cosh r − 1
Y = =I+ −1 Π− C, (3.267)
C r r2
C �r �r� � 1
R= =I+ coth −1 Π+ C. (3.268)
I − exp(−C) 2 2 2
Now, one can obtain the metric, the connection, and the curvature.

riemgeom.tex; April 9, 2010; 13:23; p. 67