Fys3120: On the LORENTZ GROUP

DEFINITION OF A GROUP:
A group G = {gi } is a set (collection of elements) with the following
properties:
* ∃ multiplication such that gi gj ∈ G ;
* ∃ a unity element g0 = 1 such that gi g0 = g0 gi = gi ;
* (g1 g2 ) · g3 = g1 · (g2 g3 ) (Associative law)
* ∃ an inverse (gi )−1 for all gi ∈ G
LIE GROUPS
In theoretical physics one most often need to study continuous groups,
called Lie groups. A Lie group is a group where the elements are functions
of continuous parameters:

g(α1 , α2 , ...., αn ) , (1)

where all αi might be chosen real. Important groups are for instance
n o
U(1) = eiα ; α ∈ R , (2)

i.e. the collection of all phase factors, and

n o
SU(n) = U = n × n matrix (n = 2, 3, ..) ; Ukl ∈ C , U † U = 1 , detU = 1
(3)
Here R is the set of all real numbers, C is the set of all complex numbers,
and k and l run from 1 to n.
One may show that the group elements of SU(n) have (n2 − 1) inde-
pendent real parameters. later we will use the group SU(2) to study three-
dimensional rotations.

THE LORENTZ GROUP

Most often, we think of a Lorentz transformation as a boost (special
Lorentz transformation) along an axis, for instance the x-axis, with the fol-
lowing relations between the following position and time coordinates:

x′ = γ (x − v t) , t′ = γ (t − v x/c2 ) ; γ = (1 − v 2 /c2 ) , (4)

between the inertial frames S and S ′ . The easiest way to obtain the trans-
formations in (4) might be the following: First we write down the relations

x′ = γ (x − v t) ; x = γ (x′ + v t′ ) , (5)

where γ is an unknown constant. Here we have assumed that the relations

between the position and time coordinates are linear (A linear movement
in one inertial frame is also linear in another frame.) Further, x′ must be

1
proportional to (x − vt) because the origo x′ = 0 of S ′ is moving with
constant velocity v in the frame S,- i.e. x′ = 0 must imply (x − v t) = 0.
Then this plus linearity implies the first equation in (5). The second equation
in (5) follows from the symmetry of the problem: The frame S moves with
velocity (−v) with respect to S ′ . Next we consider a light signal travelling
to the right, starting at time t = t′ = 0 in both frames (the clocks in S and
S ′ are assumed to be synchronized when they pass each other at x = x′ = 0
and t = t′ = 0). Then we use the postulate that the velocity of light is the
same in both frames, such that

x = ct ; x′ = c t′ . (6)

From the equations in (5) and (6) one easily finds the value for γ and t′ in
equation(4).
But Lorentz transformations are defined much wider than in (5). Lorentz
transformations are all coordinate transformations which conserve the quadrate
of a four vector xµ , i.e. conserve x · x ≡ xµ xµ = (~x)2 − c2 t2 , where
x0 = − x0 = c t. Among such are also rotations (which conserve (~x)2 sepa-
rately) a subgroup. We will first discuss the Rotation group,- and afterwards
study the boosts.
The rotation group
A 3-dimensional rotation may be expressed as:

xi → x′i = Rij xj . (7)

Here and hereafter sum over repeated indices is understood. (Einstein’s sum
convention). The matrix R has the property that the length of the position
vector is conserved under the transformation:

(x~′ )2 = (~x)2 . (8)

The rotation group may be considered to be the set of all 3 × 3 matrices R

where the inverse R−1 is the same as the transpose RT :
n o
O(3) = R ; Rkl ∈ R ; RT · R = 1 . (9)

Simple spesial cases are rotations R~n (φ) an angle φ around (the same)
axis ~n (where |~n| = 1), for example around the x-, y-, or the z-xis:
     
1 0 0 cφ 0 −sφ cφ s φ 0
     
Rx =  0 cφ sφ  , Ry =  0 1 0  , Rz =  −sφ cφ 0  (10)
0 −sφ cφ sφ 0 cφ 0 0 1

where cφ ≡ cosφ og sφ ≡ sinφ. Rotations which are continous with the unit
element (as the transformations in (10)) are in the group SO(3), different
from O(3) which also contains space invertion, ~x → −~x.

2
 Rotations around an axis ~n constitute a subgroup

R~n (φ1 ) · R~n (φ2 ) = R~n (φ1 + φ2 ) , (11)

which follows from

cosφ1 cosφ2 − sinφ1 sinφ2 = cos(φ1 + φ2 ) ,

cosφ1 sinφ2 + sinφ1 cosφ2 = sin(φ1 + φ2 ) . (12)

Such rotations about the same axis commute with each other and we say that
they constitute an Abelian sub-group. In general, rotations around different
axes will not commute.

Lorentz transformations
A Lorentz transformation of a four vector xµ = (x0 , ~x) is in analogy with
(7) given by
xµ → x′µ = Lµ ν xν , (13)
where
ν
(x′ )2 ≡ x′µ x′µ = xµ xµ ≡ x2 ⇒ Lα µ Lα ν = g µν ⇒ (L−1 )µ = Lν µ . (14)

The Lorentz-group may be intepreted as the set of all 4×4 matrices where
the inverse is equal to its transpose:
n o
L ≡ O(3, 1) = L ; Lµ ν ∈ R ; LT · L = 1 , (15)

Special cases are the boosts L~e (χ) an hyperbolic angle χ along an axis, for
instance the x-, y-, or z -axis:
     
cχ −sχ 0 0 cχ 0 −sχ 0 cχ 0 0 −sχ
 −sχ cχ 0 0   0 1 0 0   0 1 0 0 
     
Lx =   , Ly =   , Lz =  (16)
 0 0 1 0   −sχ 0 cχ 0   0 0 1 0 
0 0 0 1 0 0 0 1 −sχ 0 0 cχ

where cχ ≡ coshχ og sχ ≡ sinhχ, and further

coshχ = γ = (1 − β 2 )−1/2 ; sinhχ = β γ , β ≡ v/c , (17)

where v is the speed for the boost and c is the speed of light. Transformations
which exclude space and time inversion are within the more restictive group
SO(3, 1).
For boosts along the same axis along ~e is

L~e (χ1 ) · L~e (χ2 ) = L~e (χ1 + χ2 ) , (18)

Which follows from addition of hyperbloic angles.

coshφ1 coshχ2 + sinhχ1 sinhχ2 = cosh(χ1 + χ2 ) ,

coshχ1 sinhχ2 + sinhχ1 coshχ2 = sinh(χ1 + χ2 ) . (19)

3
The speed v for the total boost corresponding to the hyperbolic angle χ =
χ1 + χ2 will be
β1 + β2 v1 + v2
v = c · β = c · tanh(χ1 + χ2 ) = c = (20)
1 + β1 · β2 1 + v1 v2 /c2

Boosts along the same axis commute (They constitute an Abelian sub
group). In general, boosts along different axes will not commute. The set of
all boosts is not a group in the mathematical sense, that is, the product of
two boosts is not a pure boost.
Rotations constitute (in the matematical sense) a sub-set of the Lorenzt-
group. As 4 × 4 matrices, they are given by

L0 0 = 1 , L0 k = Lk 0 = 0 , Lk l = Rk l = Rkl (21)

Lie Algebra
The group elements in a Lie group may be written
( )
X
X = exp i αk Mk , (22)
k

where αj ; j = 1, 2, ..., N is a set real parameters. and the generators

!
∂X
Mk = −i , (23)
∂αk {αj =0}

satisfiy a Lie algebra (sum over m is understood):

[Mk , Ml ] = ifklm Mm , (24)

where fklm is a set coefficients (structure constants of the group) completely

antisymmetric in the indices klm.
If {Mk } is a set of generators for a group, we may obtain a new set of
generators {Mk′ } (-which also satisfies the Lie algebra in (24)) by means of a
similarity transformation:

Mk → Mk′ = S Mk S −1 , (25)

where S is an n × n matrix (with existing inverse). Because

X = eY ⇒ detX = eT rY , (26)

we obtain from (when) detX = 1 that

T rMk = 0 , (27)

for all k = 1, 2, ..., N.

4
 Lie-algebra for the rotation group
We find the generators for the rotation group from the expression for an
infinitesimal rotation around an axis along the unit vektor ~n:
R~n (δφ) = 1 + δφ i ~n · J~ + O((δφ)2 ) (28)
From this we obtain the generators for rotations around the x, y- and the
z-axis:
     
0 0 0 0 0 i 0 −i 0
J1 = 
 0 0 −i 
 , J 2 = 
 0 0 0 
 , J 3 = 
 i 0 0 

(29)
0 i 0 −i 0 0 0 0 0
which can also be written: .
(Jk )ln = −iεkln , (30)
which gives the well known algebra which is the same as for angular momenta
~
h̄J:
[Jk , Jl ] = iεkln Jn . (31)
A rotation around the unit vector ~n may be written:
~ = 1 + i sinφ ~n · J~ + (cosφ − 1)(~n · J)
R = exp(iφ ~n · J) ~2 (32)
where we have expanded the exponential function , used
~ 3 = ~n · J~ ,
(~n · J) (33)
and then summed the series.
The above expressions for Ji are not those usually used in quantum me-
chanics, where one most often uses the spherical notation with a diagonal
J3 = J3s , which can be obtained by means of a transformations as in (25):

     
0 1 0 0 −i 0 1 0 0
1   1    
J1s = √  1 0 1  , J2s = √  i 0 −i  , J3s =  0 0 0  ,(34)
2 0 1 0 2 0 i 0 0 0 −1
which will also satisfy the Lie-algebra.

Lie-algebra for the Lorentz group

We find the generators for boosts from the expressions for an infinitesimal
boost along the unit vector ~e:
~ + O((δχ)2 ) .
R~e (δχ) = 1 + δχ i ~e · K (35)
which gives the generators for boosts along the x, y- and z-xis:
     
0 i 0 0 0 0 i 0 0 0 0 i
 i 0 0 0   0 0 0 0   0 0 0 0 
     
Kx =   , Ky =   , Kz =   (36)
 0 0 0 0   i 0 0 0   0 0 0 0 
0 0 0 0 0 0 0 0 i 0 0 0

5
 Multiplying these we find the Lie-algebra for the complete Lorentz group:

[Jk , Jl ] = iǫklm Jm , [Jk , Kl ] = iǫklm Km , [Kk , Kl ] = −iǫklm Jm , (37)

where Jk are the rotation generators and Kl are the boost generators (k, l, m =
1, 2, 3). In the Minkowski representation, the rotasjon generator are in prin-
ciple 4x4 matrices, but with only zeros for all time components.
A pure boost along the unit vector ~e may be written
~ = 1 + i sinhχ ~e · K
L = B~e = exp(iχ ~e · K) ~ − (coshχ − 1)(~e · K)
~ 2 (38)

where we have used

~ 3 = −~e · K
(~e · K) ~ ; coshχ = (1 − v 2 /c2 )1/2 (39)

From and (38) we recover the transformation matrices in (16).

Connection between SU(2) and spatial rotations

Given an arbitrary three vector ~x. We define the 2x2 matrix

S ≡ ~x · ~σ ⇒ detS = −~x2 . (40)

(The determinant is easily found by means of the standard expressions for

the Pauli matrices). Now we let S transform after Uǫ SU(2) :

S → S ′ = USU † . (41)

Then
S ′ = x~′ · ~σ ⇒ detS ′ = −(x~′ )2 . (42)
(One might think that S ′ also might contain a term proportional to the
unit matrix; but because the Pauli matrices are traceless and T r(S ′ ) =
T r(U † U S) = T r(S) this is excluded). On the other hand,

detS ′ = detU · detS · detU † = |detU|2 · detS = detS , (43)

such that the transformation (41) generates a rotation of ~x, conserving its
lenght:
2
x~′ = ~x2 → x′i = Rij xj . (44)
From (7), (40) and (42) and T r(σi σj ) = 2δij we find the coefficients
1  
Rij = T r σi Uσj U † . (45)
2
Note that +U and −U give the same rotation, showing that SU(2) and
SO(3) are not completely equivalent. The connection between SU(2) and
the rotation group means that gruppa betyr at

U1 · U2 = U3 ⇒ R(U1 ) · R(U2 ) = R(U3 ) , (46)

6
which shows that SO(3) is a representation of SU(2).
If we write
i
U = exp( θ ~n · ~σ ) = cos(θ/2) + i~n · ~σ sin(θ/2) , (47)
2
this represents a rotation an angle θ around the direction ~n (we have |~n| = 1).
In this case we find

Rij = cosθ δij + (1 − cosθ) ni nj + sinθ ǫijk nk , (48)

which is the rotation matrix in three spatial dimensions. Further, if χ is a

two component spinor, ~v ≡ χ†~σ χ will transform as a vector, i.e.

vi → vi′ = Rij vj when χ → χ′ = Uχ , (49)

In the fundamental representation of SU(2) the generators are (k=1,2,3):

1 (1/2)
Lk = σk ≡ Sk , (50)
2
which corresponds to spin 1/2. Higher representations (s = 1, 3/2, ...) have
the same Lie algebra h i
(s) (s) (s)
Sk , Sl = i ǫklm Sm , (51)
where ǫklm is the Levi-Civita tensor in three dimensions. But other matrix
relations are different for different spin. For example, σj 2 = 1 is only valid
(1) (1) (1)
for spin 1/2. For spin 1, (Sj )2 6= 1, but (Sj ) = Sj . The spin 1 rep-
resentation of SU(2) corresponds to the fundamental representation of the
(1)
rotation group, often given in the cartesian representation. (Sk )ln = −iǫkln .

Connection between SL(2C) and Lorentz transformations

The group SL(2C) is defined by

SL(2C) = {A = 2 × 2 matrix ; Akl ∈ C , detA = 1} . (52)

The group SL(2C) is related to the Lorentz group in the same way as SU(2)
is related to the rotation group. Let

T ≡ xµ σ µ ⇒ detT = xµ xµ ≡ x2 , (53)

where we have defined

σ µ ≡ (1, ~σ) , and similarly one defines σe µ ≡ (−1, ~σ) . (54)

Now, let T transform as A ∈ SL(2C) :

T → T ′ = AT A† . (55)

7
Then we have
T ′ = x′µ σ µ ⇒ detT ′ = (x′ )2 . (56)
On the other hand,

detT ′ = detA · detT · detA† = |detA|2 · detT = detT , (57)

such that the norm (“length”) of the four vector is conserved:

(x′ )2 = x2 . (58)

Then we find in analogy with (45) :

1  µ 
Lµ ν = T r σe Aσν A† . (59)
2
The connection between SL(2C) and the Lorentz group means that

A1 · A2 = A3 ⇒ L(A1 ) · L(A2 ) = L(A3 ) , (60)

i.e the Lorentz group is a representation of SL(2C).

Further, the quantity v µ ≡ χ† σ µ χ will transform as a four vector, i.e..

vµ → vµ′ = Lµ ν vν when χ → χ′ = Aχ , (61)

and χ is a two component spinor. In the SL(2C) representation, the gener-

ators are given by
1 ~ = i ~σ
J~ = ~σ and K (62)
2 2
One can show that an arbitrary element of SL(2C) can be written as

A = H ·U where H † = H and U † = U −1 . (63)

which means that any Lorentz transformation might be written as a boost

times a rotation. Furthermore, H can be written in analogy with U in (47):

~ = exp( 1 ψ ~e · ~σ ) = cosh(ψ/2) + ~e · ~σ sinh(ψ/2) . (64)

H = exp(−iψ ~e · K)
2

One may show that all elements in SL(2C) may be factorized in a boost
and a rotation. From the connection between SL(2C) and the Lorentz group,
we obtain the Lorentz transformations L:

L(A) = L(H) · L(U) , (65)

where L(H) = B(H) is a pure boost and L(U) = R(U) is a pure rotation.

8
 Proof:
From an arbitrary element A ∈ SL(2C) we define a new matrix F :

F ≡ A A† ⇒ F † = (A† )† A† = A A† = F . (66)

Because F is hermittean and positive definite it must have real and positive
eigen values f1 og f2 ,and there must exist a matrix V which transform F to
diagonal form:
!
′ † f1 0
F → F = VFV = . (67)
0 f2

Thus the square root of this matrix must also be hermittean:

√ √ !
f1 0
H′ ≡ F ′ = √ , (68)
0 f2

Transforming back we obtain:

√
H = F = V † H′ V . (69)

We define

U ≡ H −1 A ⇒ U U † = H −1 A A† H −1 = H −1 (H H) H −1 = 1. (70)

From the practical side, the point is that 2×2 matrices in SL(2C) are
simpler to handle than the 4×4 matrices L in the Minkowski representation,
such that Lorentz transformations may be factorized by means of v SL(2C).
To do this factorization explicite, we must of course find H −1 from F .