8.

Continuous groups 1

8. Continuous groups
The discussion so far has focused on finite groups, containing a number of
discrete elements. There are also groups with infinitely many elements. As
a very simple example, the set of all integers forms a group with respect to
addition. The elements of the group are denumerable, but there are infinitely
many of them.

Another simple example is the set of rotations through some angle about
a fixed axis, which form a group with respect to consecutive action. The
elements of the group are labelled by the continuous variable θ, the angle of
rotation. There are infinitely many of them and they are not denumerable.
This is an example of a continuous group.

In general, the elements of a continuous group may be labelled by a

number of continuous variables. For instance, the group of rotations in three
dimensions with a fixed point is labelled by three angles — two to define the
direction in space of the axis of rotation through the fixed point, the third
to specify the angle of rotation.

For rotation groups, the range of the continuous variables is generally

finite. An example of a continuous group with a variable of infinite range is
the group of translations along a line by a distance x, where −∞ < x < ∞.

The continuous variables labelling the group elements are generally taken
to be real. The composition rules relating the labels of a product to the
labels of its factors may be either simple or complicated.
[For rotations Rθ through an angle θ about a fixed axis, Rθ Rφ =
Rθ+φ . For translations Tx by a distance x along a fixed line,
Tx Ty = Tx+y . Non-zero complex numbers form a group under
the usual rule of multiplication, the two real labels of an element
being its real and imaginary parts. Denoting x + iy by Cx,y , the
composition rule is Cx,y Cv,w = Cxv−yw,xw+yv .]

Square matrices over a field are closed under matrix multiplication. A

unit matrix exists, and non-singular matrices may be inverted, so matrix
groups can be defined, labelled by their entries. Different matrix groups may
be defined by imposing various restrictions on the entries in the matrices.

The set of non-singular n × n matrices with complex entries constitutes

the general linear group GL(n, C) and has 2n2 real parameters. Restricting it
to real entries produces the real general linear group GL(n, R), with n2 real

Introductory Algebra for Physicists Michael W. Kirson

8. Continuous groups 2

parameters. The subset of unitary matrices [U † U = 1] with complex entries

is closed under multiplication and defines the unitary group U(n) with n2
independent real parameters.

[n2 complex entries means 2n2 real parameters. The unitarity

condition imposes n real conditions from the diagonal entries
and n(n − 1)/2 complex conditions from the off-diagonal entries.
There remain 2n2 − n − 2.n(n − 1)/2 = n2 independent parame-
ters.]

Unitary matrices, interpreted

Pn ∗ as transformations of complex vectors, preserve
the quadratic form i=1 zi zi . From unitarity, | det U| = 1 =⇒ det U = eiφ ,
2

where φ is a real function of the matrix parameters. The subset of U(n) for
which det U = 1 comprises the special unitary group (or unimodular unitary
group) SU(n), with n2 − 1 independent real parameters (since φ = 0 imposes
an additional condition on the n2 independent parameters). It is an invariant
subgroup of U(n).

Related to the unitary matrices

Pm ∗are thosePm+n matrices which preserve the
more general quadratic form i=1 zi zi − i=m+1 zi∗ zi . They,  too, forma
1m 0
group U(m, n). In terms of the modified unit matrix Jm,n = ,
0 −1n
†
where 1m is an m×m unit matrix, each such matrix satisfies Um,n Jm,n Um,n =
2
Jm,n , so | det Um,n | = 1. The group SU(m, n) satisfies, in addition, the
condition det Um,n = 1. From the definition it follows that U(m, n) = U(n, m)
and U(n, 0) = U(0, n) = U(n).

Analogous to the unitary matrices is the subset of orthogonal matrices

[O tr O = 1, where Atr is the transpose of the matrix A] with real entries,
which is also closed under multiplication and defines the orthogonal group
O(n) with n(n − 1)/2 real parameters.

[n2 real entries, n conditions from the diagonal entries of the or-
thogonality condition, n(n−1)/2 conditions from the off-diagonal
entries.]
Pn 2
Orthogonal matrices preserve the real quadratic form i=1 xi . From or-
2
thogonality, (det O) = 1 =⇒ det O = ±1. The subset of O(n) for which
det O = 1 comprises the special orthogonal group (or unimodular orthogo-
nal group) SO(n), an invariant subgroup of O(n). Choosing det O = 1 is a
process of selection, which does not impose an additional condition on the
parameters. There remain n(n − 1)/2 independent real parameters.

Introductory Algebra for Physicists Michael W. Kirson

8. Continuous groups 3

PmThere2
also
Pm+noccur the groups O(m, n), which preserve the quadratic form
2
i=1 xi − i=m+1 xi and for which O(m, n) = O(n, m) and O(n, 0) =
O(0, n) = O(n). Similarly, there are unimodular groups SO(m, n) in which
every matrix has unit determinant.

Also of interest are the 2n-dimensional

P2n matrices
Pn which preserve the skew-
symmetric quadratic form i,j=1 xi Jij yj = i=1 (xi yi+n − xi+n yi ), where
 
0 1
J = is a 2n × 2n matrix made of n × n unit and zero matrices.
−1 0
These 2n-dimensional symplectic matrices constitute the symplectic group
Sp(2n) with n(2n + 1) real parameters. It can be shown that the symplectic
matrices automatically have unit determinant, i.e. they are unimodular.

[It should be noted that alternative notations are sometimes used.

For instance, the group denoted here by Sp(2n) is sometimes
denoted Sp(n, R).]

Continuous groups are often parametrized in such a way that the identity
element is characterized by vanishing parameters. Infinitesimal values of the
parameters then characterize elements which are in some sense close to the
identity. Other elements can then be obtained by frequent application of
elements close to the identity.

As a concrete example, consider again rotations about some axis. Parametris-

ing them by the angle of rotation θ, the identity has θ = 0 and infinitesi-
mally small values of θ correspond to infinitesimally small rotations. A large
enough number of infinitesimal rotations will produce any desired finite ro-
tation. There is therefore considerable interest in the nature and properties
of group elements close to the identity.
[Note that not all group elements need be accessible by repeated
application of elements close to the identity. In the case of the
orthogonal group O(n), for instance, those elements having a de-
terminant of −1 cannot be obtained by continuous variation from
the identity, which has determinant +1.]
Consider an element of a continuous group which is infinitesimally close
to the identity, 1 + ǫM, where ǫ is an infinitesimal. The defining proper-
ties of the group (unitary, orthogonal, symplectic, . . . ) will constrain M.
For example, the condition of unitarity, to lowest order in infinitesimals, is
(1 + ǫM)† (1 + ǫM) = 1 =⇒ M † = −M , i.e. M must be anti-Hermitian. If,
in addition, the group element is required to have unit determinant, M must
have a vanishing trace.

Introductory Algebra for Physicists Michael W. Kirson

8. Continuous groups 4

[The n × n Hermitian matrix iM has n real eigenvalues. In the

basis of its eigenvectors, it has only diagonal Q
elements, say αk .
In thisPbasis, the determinant ofP(1 + ǫM) is nk=1 (1 − iǫαk ) =
1 − iǫ nk=1 αk + O(ǫ2 ) = 1 =⇒ nk=1 αk = 0. The result is true
in general, because the determinant and trace of a matrix are
independent of the basis.]
The group SU(n) of unimodular unitary n × n matrices can be generated
by the traceless anti-Hermitian n × n matrices. There are n2 − 1 independent
such matrices, so SU(n) has n2 − 1 infinitesimal generators. In the same
way, the other matrix groups described above can be assigned finite sets of
infinitesimal generators. In terms of its infinitesimal generators
P {Gi }, the
elements of a continuous group can be represented as exp( i αi Gi ), where
the numbers {αi } could supply a convenient parametrization of the group.

For such a form of the group elements to be acceptable, the product of

two such elements must also take the same form, or else the group will not
be closed under multiplication. But the Baker-Campbell-Hausdorff formula
asserts that eX eY = eZ , where X, Y, Z are matrices and Z is given by the
expansion Z = X + Y + [X, Y ]/2 + ([[X, Y ], Y ] + [X, [X, Y ]])/12 + · · ·. All the
terms in the expansion, from the third one on, are expressed
P P
purely inPterms
αi M i
of multiple commutators of X and Y . So requiring e i e βiM i = e i γi M i
i

for some appropriate γi ({α}, {β}) means requiring that [Mi , Mj ] is a linear
combination of the {Mi }.

The continuous groups of interest in physics are Lie groups, whose el-
ements are analytic functions of the continuous parameters. They can be
expressed in terms of infinitesimal generators defined by derivatives of group
elements, with respect to the parameters, close to the identity. The require-
ment that the commutator of any two infinitesimal generators is a linear
combination of the infinitesimal generators confers on the set of generators
a mathematical structure called a Lie algebra. These algebras will be the
subject of the rest of the course. The above discussion indicates that any Lie
group can be associated with a Lie algebra. Lie proved that any Lie algebra
can be associated with a Lie group.

Examples
1. The group of unimodular, unitary 2 × 2 matrices is SU(2). Each matrix
has four complex entries, so eight real parameters are required to specify it.
The condition of unitarity imposes four real restrictions (two real diagonal
restrictions and one complex off-diagonal restriction) and the condition that

Introductory Algebra for Physicists Michael W. Kirson

8. Continuous groups 5

the determinant be +1 imposes one more. There remain three independent

real parameters. This is a three-parameter continuous group. The infinitesi-
mal generators of the group are traceless, anti-Hermitian 2 × 2 matrices. The
most general such matrix takes the form
 
ia b + ic
−b + ic −ia
     
i 0 0 1 0 i
=a +b +c
0 −i −1 0 i 0

=aX1 + bX2 + cX3 ,

where X1 , X2 , X3 generate the associated Lie algebra su(2). (It is a widely-
used convention to denote a group and its associated algebra by the same
name, with the group name in upper-case letters and the algebra name in
lower-case letters. This convention will be adopted here.) The commutators
of the generators may be evaluated straightforwardly:
[X1 , X2 ] = 2X3 , [X2 , X3 ] = 2X1 , [X3 , X1 ] = 2X2 .
2. The group of unimodular, orthogonal 3 × 3 matrices is SO(3). Each
matrix has nine real entries, but the condition of orthogonality imposes three
diagonal and three (symmetric) off-diagonal conditions, so there are three
independent real parameters. The set of orthogonal 3×3 matrices divides into
two subsets — one of matrices with determinant −1, the other of matrices
with determinant +1. The latter subset is relevant here, and constitutes
a three-parameter continuous group. Near the identity, the condition of
orthogonality is (1 + ǫM)tr (1 + ǫM) = 1, where Atr is the transpose of the
matrix A and ǫ is an infinitesimal. To lowest order in infinitesimals, this
becomes M tr = −M, so the infinitesimal generators of the group are real,
antisymmetric 3 × 3 matrices. The most general such matrix takes the form
 
0 a b
 −a 0 c 
−b −c 0
     
0 1 0 0 0 1 0 0 0
= a  −1 0 0  + b  0 0 0  + c  0 0 1 
0 0 0 −1 0 0 0 −1 0

= aY1 + bY2 + cY3 ,

where Y1 , Y2 , Y3 generate the associated Lie algebra so(3). By straightforward
evaluation, the commutators of the generators are:

Introductory Algebra for Physicists Michael W. Kirson

8. Continuous groups 6

[Y1 , Y2] = −Y3 , [Y2 , Y3] = −Y1 , [Y3 , Y1 ] = −Y2 .

3. The group of unimodular, unitary 3×3 matrices is SU(3). Each matrix has
nine complex entries. Unitarity imposes three real diagonal conditions and
three complex off-diagonal conditions, while the requirement of unit deter-
minant imposes one further real condition. There remain eight independent
real parameters. The infinitesimal generators are traceless, anti-Hermitian
3 × 3 matrices. The most general such matrix takes the form
 
ia b + ic d + ie
 −b + ic if g + ih 
−d + ie −g + ih −ia − if
       
i 0 0 0 1 0 0 i 0 0 0 1
= a  0 0 0  + b  −1 0 0  + c  i 0 0  + d  0 0 0 
0 0 −i 0 0 0 0 0 0 −1 0 0
       
0 0 i 0 0 0 0 0 0 0 0 0
+e  0 0 0  + f  0 i 0  + g  0 0 1  + h  0 0 i 
i 0 0 0 0 −i 0 −1 0 0 i 0

= aZ1 + bZ2 + cZ3 + dZ4 + eZ5 + f Z6 + gZ7 + hZ8 ,

where Z1 , Z2, . . . , Z8 generate the associated Lie algebra su(3). The commu-
tators of the generators are found to be:
[Z1 , Z2 ] = Z3 [Z1 , Z3 ] = −Z2 [Z1 , Z4 ] = 2Z5 [Z1 , Z5 ] = −2Z4
[Z1 , Z6 ] = 0 [Z1 , Z7 ] = Z8 [Z1 , Z8 ] = −Z7 [Z2 , Z3 ] = 2Z1 − 2Z6
[Z2 , Z4 ] = −Z7 [Z2 , Z5 ] = −Z8 [Z2 , Z6 ] = Z3 [Z2 , Z7 ] = Z4
[Z2 , Z8 ] = Z5 [Z3 , Z4 ] = Z8 [Z3 , Z5 ] = −Z7 [Z3 , Z6 ] = −Z2
[Z3 , Z7 ] = Z5 [Z3 , Z8 ] = −Z4 [Z4 , Z5 ] = 2Z1 [Z4 , Z6 ] = −Z5
[Z4 , Z7 ] = −Z2 [Z4 , Z8 ] = Z3 [Z5 , Z6 ] = Z4 [Z5 , Z7 ] = −Z3
[Z5 , Z8 ] = −Z2 [Z6 , Z7 ] = 2Z8 [Z6 , Z8 ] = −2Z7 [Z7 , Z8 ] = 2Z6

Introductory Algebra for Physicists Michael W. Kirson