Introduction to Lie groups

Kirill Zaynullin
Winter 2012
Contents
1 Topological spaces 2
2 Manifolds 3
3 Examples of manifolds 5
4 Dierentiable functions on manifolds 6
5 Vector bundles over manifolds 8
6 Lie groups 9
7 Types of Lie groups 11
1
Preliminaries
1 Topological spaces
Let X be a topological space. Given a base for a topology on X and a point
x X, the set of all basis sets containing x forms a local base at x.
1.1 Denition. A topological space X is called rst countable if each point
has a countable local base, i.e. for each point x X there exists a sequence
U
1
, U
2
, . . . of open neighbourhoods of x such that for any open neighbourhood
V of x there exists an integer i with U
i
contained in V . A topological space X
is called second countable if its topology has a countable base, i.e. there exists
some countable collection U
i
of open subsets of X such that any open subset
of X can be written as a union of elements of some subfamily of U
i
.
Observe that if one has a countable base for a topology on X then one clearly
has a countable local base at every point x X. Therefore,
second countable = first countable.
1.2 Examples. Euclidean space R
n
with its usual topology is second countable.
The base is the set of all open balls with rational radii and whose centers have
rational coordinates.
1.3 Exercise. Provide an example of a rst countable topological space which
is not second countable.
Properties of second countable spaces:
Every second countable space contains a countable dense subset, i.e. is
separable
Every open cover of a second countable space has a countable subcover
A continuous, open image of a second-countable space is second-countable
Every subspace of a second-countable space is second-countable
Quotients of second-countable spaces need not be second-countable
Any countable product of a second-countable space is second-countable,
although uncountable products need not be
Any base for a second-countable space has a countable subfamily which is
still a base
Every collection of disjoint open sets in a second-countable space is count-
able
1.4 Exercise. Prove one of the above properties.
2
1.5 Denition. A Hausdor space, separated space or T
2
-space is a topological
space in which any two distinct points have disjoint neighbourhoods.
1.6 Examples. The real numbers R (under the standard metric topology on
real numbers) is a Hausdor space. More generally, all metric spaces are Haus-
dor.
Properties of Hausdor spaces:
Subspaces and products of Hausdor spaces are Hausdor
Quotient spaces of Hausdor spaces need not be Hausdor. In fact, every
topological space can be realized as the quotient of some Hausdor space
1.7 Exercise. Provide an example of a Hausdor space which is not second
countable.
From now on by a topological space we mean always a second countable
Hausdor space.
1.8 Denition. A topological space X is called compact if each of its open
covers has a nite subcover.
Observe that a compact subset of a topological space is always closed.
2 Manifolds
2.1 Denition. A continuous (real) atlas on a topological space X is a collec-
tion of pairs (U
i
,
i
) called coordinate charts, where the U
i
are open sets which
cover X, and
i
: U
i
R
n
is a homeomorphism of U
i
onto an open subset of
R
n
. The maps

ij
=
j

1
i
[
i(UiUj)
:
i
(U
i
U
j
)

j
(U
i
U
j
)
are called the transition maps. Observe that each
ij
is a homeomorphism
between
i
(U
i
U
j
) and
j
(U
i
U
j
).
2.2 Denition. A C
k
-atlas, k 0, on X is an atlas whose all functions
i
and
all transition functions
ij
are in C
k
, i.e. are k-times continuously dierentiable.
Note that for k = 0 we obtain the usual continuous atlas. Similarly, a C

-atlas
is an atlas whose transition functions are smooth, i.e. have derivatives of all
orders.
2.3 Denition. Two C
k
-atlases are called equivalent if their union is again a
C
k
-atlas. This denes an equivalence relation on the set of all C
k
-atlases. An
equivalence class of C
k
-atlases on X is called a C
k
-structure on X.
Properties of C
k
-structures:
3
By the very dention the union of any two continuous (for k = 0) atlases is
automatically a continuous atlas, i.e. all continuous atlases are equivalent.
So there is only one continuous (C
0
-)structure on X.
There exist a topological space with continuous structure which admit no
C
1
-structure [Kervaires theorem].
For k > 0 it is known [Whitneys theorem] that every C
k
-atlas is uniquely
smoothable, i.e. for every C
k
-atlas there exists a unique C
k
-equivalent
C

-atlas. Moreover, two C

k
-atlases that are equivalent to a single C

-
atlas are equivalent as C
k
-atlases. In other words, there is a one-to-one
correspondence between C
k
-structures and smooth (C

-)structures.
On S
7
there exist 28 dierent smooth structures [Milnor].
It is conjectured that S
4
has at least two dierent smooth structures
[smooth Poincare conjecture]
2.4 Denition. A topological space X together with a C
k
-structure is called
a C
k
-manifold. In particular, a C
0
-manifold is called a topological manifold; a
C
k
-manifold for k > 0 is called a dierentiable manifold and a C

-manifold is
called a smooth manifold.
The properties of C
k
-structures show that there is no distinct notion of a
C
k
-manifold other than a topological manifold (k = 0) and a smooth manifold
(k = ).
2.5 Denition. The dimension n of R
n
appearing in the denition of a chart-
atlas-structure is called the dimension of a manifold. A manifold of dimension
one is called a (real) curve. A manifold of dimension two is called a (real)
surface.
2.6 Remark. The denitions of this section are based on the following three
concepts: the eld R, the topological space R
n
and the property C
k
, k > 0.
Replacing each of them we obtain the following generalizations of the notion of
a manifold:
If R is replaced by some other eld F (F = C or F = Z
p
) keeping the
Euclidean topology on F
n
and the transition maps are just continuous
functions, then we obtain the notion of a (topological) F-manifold, e.g. a
topological complex manifold or a topological p-adic manifold.
If the property C
k
is replaced by the property of being analytic, i.e. the
transition maps are analytic (they can be expressed as power series), then
we obtain the notion of an analytic manifold. For example, the sphere can
be given an analytic structure.
If R is replaced by C and the property C
k
, k > 0 is replaced by the prop-
erty of being holomorphic, then we obtain the notion of a holomorphic
complex manifold.
4
3 Examples of manifolds
1. The Euclidean space R
n
and its open subspace provide examples of non-
compact real smooth manifolds.
2. The circle
S
1
= (x
1
, x
2
) R
2
[ x
2
1
+x
2
2
= 1
is a compact smooth curve. Indeed, consider the cover S
1
= U
n
U
s
, where U
n
is obtained from S
1
by removing the north pole and U
s
the south. Consider
the stereographic projections
n
: U
n
R and
s
: U
s
R to the coordinate
line from the north and the south pole, respectively. In this way we obtain two
charts (U
n
,
n
) and (U
s
,
s
) with smooth transition functions.
3. The sphere
S
2
= (x
1
, x
2
, x
3
) R
3
[ x
2
1
+x
2
2
+x
2
3
= 1
is a compact smooth surface. The n-dimensional sphere S
n
provides an example
of a compact smooth manifold.
4. The projective line
RP
1
:= linear subspaces of dimension one in R
2

is a compact smooth curve.

5. The projective plane
RP
2
:= linear subspaces of dimension one in R
3

is a compact smooth surface.

6. The n-dimensional projective space
RP
n
:= linear subspaces of dimension one in R
n+1
.
provides an example of a compact smooth manifold.
3.1 Exercise. Show that the set of points of RP
n
is the quotient set of S
n
modulo the following equivalence relation: two points of S
n
are said to be
equivalent if they are opposite.
3.2 Exercise. Show that RP
1
is a smooth curve (provide charts with smooth
transition functions).
5
7. The set
Gr(k; n) := linear subspaces of dimension k in R
n
, where 0 < k < n,
is a compact smooth manifold of dimension k(n k) called the Grassmannian
of k-dimensional planes in R
n
.
By the very denition we have Gr(1; n) = RP
n1
.
3.3 Exercise. Show that there is a bijection between the sets of points of
Gr(k; n) and Gr(n k; n).
3.4 Exercise. Consider the set of n n-matrices M
n
(R). Consider an equiv-
alence relation on M
n
(R) given by: two matrices A and B are equivalent if A
and B are similar. Let Q be the respective quotient set and let Q
k
denote the
subset consisting of classes of matrices of rank k. Show that there is a bijection
between the set of points of Gr(k, n) and the set Q
k
.
8. Let 0 < k
1
< . . . < k
r
< n. We dene a set
Fl(k
1
, . . . , k
r
; n) := V
1
V
2
. . . V
r
R
n
[ dimV
i
= k
i
.
A point of this set is an increasing sequence of linear subspaces (called a ag)
V
1
V
2
. . . V
r
, where increasing means each is a proper subspace of the
next. The set Fl(k
1
, . . . , k
r
; n) is a compact smooth manifold called the ag
manifold.
The ag manifold Fl(1, 2, . . . , n 1; n) is called the complete ag manifold.
By the very denition we have Gr(k, n) = Fl(k; n).
4 Dierentiable functions on manifolds
4.1 Denition. A real valued function f on an n-dimensional dierentiable
manifold X is called dierentiable at a point x X if it is dierentiable in any
coordinate chart dened around x. More precisely, if (U
i
,
i
) is a chart where
U
i
is an open set in X containing x and
i
: U
i
R
n
is the map dening the
chart, then f is dierentiable if and only if
f
1
i
: (U
i
) R
n
R
is dierentiable at
i
(x). One can show that this notion doesnt depend on the
choice of an equivalent atlas.
In a similar way, replacing the word dierentiable by k-times dierentiable
we dene the notion of a C
k
-function on X (denoted f C
k
(X)) and a smooth
function on X (denoted f C

(X)).
From now on we assume that X is a C
k
-manifold and f C
k
(X).
4.2 Denition. Suppose that : R X is a function such that
i
is a
dierentiable curve for any chart (U
i
,
i
). Then is called a dierentiable curve
in X.
6
4.3 Denition. Let f be a real valued function on an n-dimensional dieren-
tiable manifold X. Let : R X be a dierentiable curve in X with (0) = x.
We dene the directional derivative of f at x along to be

f(x) := lim
t0
f((t)) f(x)
t
.
4.4 Lemma. If
1
and
2
are two dierentiable curves in X such that
1
(0) =

2
(0) = x, and in any coordinate chart (U
i
,
i
),

i
(x) = lim
t0

i
(
1
(t))
i
(x)
t
= lim
t0

i
(
2
(t))
i
(x)
t
=
2

i
(x).
then the directional derivatives of f at x along
1
and along
2
are the same.
4.5 Denition. A tangent vector at x X is an equivalence class of dieren-
tiable curves with (0) = x, modulo the equivalence relation

1

2

1

i
(x) =
2

i
(x)
in all coordinate charts (U
i
,
i
). Hence, the equivalence classes are curves
through x with a prescribed velocity vector at x.
The collection of all tangent vectors at x forms a vector space called the
tangent space to X at x, denoted T
x
X.
4.6 Corollary. The directional derivative depends only on the tangent vector
of the curve at x. So we can dene
v
f(x) :=

f(x), where v is a tangent

vector for a curve , i.e. the equivalence class of .
4.7 Denition. Given a function f : X R we dene a linear map
Df(x): T
x
X R by v
v
f(x).
called the dierential of f at x.
Observe that replacing a real-valued function f in the denition 4.3 by a
function f : X R
m
, we obtain the usual gradient

f(x) of f along and

the denition 4.7 extends into
4.8 Denition. Given a function f : X R
m
we dene a linear map
Df(x): T
x
X R
m
by v
v
f(x).
called the dierential of f at x.
What does it mean to have a dierentiable map between two manifolds ?
4.9 Denition. Let X and Y be two C
k
-manifolds. Let f : X Y be a
function (between sets). We say that f is continuously k-dierentiable, i.e.
f C
k
(X, Y ), if
j
f[
Ui

1
i
are continuously k-dierentiable for all charts
(U
i
,
i
) and (V
j
,
j
) of X and Y respectively.
7
What about the dierential ?
Let f : U V be a function from an open subset U R
n
to an open subset
V of R
m
. For any point x U, the Jacobian of f at x (with respect to the
standard coordinates) is the matrix representation of the total derivative of f
at x, which is a linear map
Df(x): R
n
R
m
.
We want to generalize it to the case of a function f : X Y between manifolds.
4.10 Denition. Let f : X Y be a C
k
-function between C
k
-manifolds. We
dene it dierential at the point x X to be the linear map
Df(x): T
x
X T
f(x)
Y
by sending an equivalence class of a curve : R X to the equivalence class
of the curve f : R Y . One can check that this denition is compatible
with coordinate charts, i.e. restricted to U
i
containing x the formula for the
dierential D
i
(x) is given in 4.8.
5 Vector bundles over manifolds
5.1 Denition. By a vector bundle of rank m 0 over a C
k
-manifold X we
mean a C
k
-manifold E together with a C
k
-function p: E X such that for
every point x X there exists a chart U
i
containing x such that
Locally E looks as R
m
U
i
,
i.e. p restricted to p
1
(U
i
) factors as
p: p
1
(U
i
)

i
R
m
U
i
pr2
U
i
,
where the last map is the projection on the second coordinate.
In particular, p
1
(x) R
m
for every point x X.
Compatible with other such charts,
i.e. for another such chart U
j
R
m

i
p
1
(x)

j
R
m
is a linear automorphism (invertible linear transformation) of R
m
.
Note that as a set E can be viewed as a disjoint union of R
m
s, i.e.
E H
xX
(R
m
x)
Observe also that dimE = dimX +m.
5.2 Examples.
8
1 The bundle E = R
m
X is called a trivial bundle of rank m.
2 The cylinder is an example of a trivial bundle of rank 1 over S
1
.
3 Mobius strip is an example of a non-trivial bundle over S
1
.
4 Let E = H
xX
T
x
X x be the disjoint union of all tangent spaces of X.
One can endow E with a structure of a C
k
-manifold called the tangent bundle
to X and denoted TX. Observe that dimTX = 2 dimX.
5 Given a manifold X in R
N
we can dene its normal bundle NX as NX =
H
xX
N
x
Xx, where N
x
X is a vector space orthogonal to the tangent space
T
x
X.
6 Tautological bundle
k
over a Grassmannian Gr(k, n) is dened by

k
= H
xGr(k,n)
V
x
x,
where V
x
is the k-dimensional linear subspace representing the point x. This is
a non-tirvial bundle over Gr(k, n).
In particular, if k = 1, i.e. Gr(k, n) = RT
n1
, then its tautological bundle

1
is the Mobius strip over the projective space and is also denoted by O(1).
7 If E is a bundle over X, then replacing the vector spaces appearing in the
denition of E by dual spaces, by direct sum of spaces, by tensor products of
spaces we obtain the dual bundle, the direct sum of bundles and the tensor
product of bundles. For instance, the dual to the tangent bundle is called the
cotangent bundle; the dual to the normal bundle is called the conormal bundle.
6 Lie groups
6.1 Denition. A (real) Lie group is a nite-dimensional real (C

-) smooth
manifold together with two (C

-) smooth functions between the manifolds

the multiplication : GG G, (x, y) = xy
the inverse inv : G G, inv(x) = x
1
which satisfy the usual group axioms (associativity, existence of an inverse and
the neutral element e), i.e. (x, (y, z)) = ((x, y), z), (x, e) = (e, x) = x,
(x, inv(x)) = (inv(x), x) = e, where e G denotes the neutral element (as
we use the multiplicative notation we will also denote e by 1).
The latter means that G has a structure of a group at the same time, i.e.
Lie group = Smooth Manifold + Group
9
6.2 Examples. Euclidean space R
n
with ordinary vector addition as the
group operation is an n-dimensional noncompact abelian Lie group.
The circle group S
1
consisting of angles mod 2 under addition or, alter-
nately, the complex numbers with absolute value 1 under multiplication.
This is a one-dimensional compact connected abelian Lie group.
The 3-sphere S
3
forms a Lie group by identication with the set of quater-
nions of unit norm.
The only other spheres that admit the structure of a Lie group is the 0-
sphere S
0
(real numbers with absolute value 1). By the way, the sphere S
7
carries the structure of a Lie quasigroup (a nonassociative group), which
can be identied with the set of unit octonions.
The 2 2 real invertible matrices form a group under multiplication, de-
noted by GL(2):
GL(2) =

A =

a b
c d

[ det A = ad bc ,= 0

.
This is a 4-dimensional noncompact Lie group. This group is discon-
nected; it has two connected components corresponding to the positive
and negative values of the determinant.
The rotation matrices form a subgroup of GL(2), denoted by SO(2). Using
the rotation angle as a parameter, this group can be parametrized as
follows:
SO(2) =

cos sin
sin cos

[ R/2Z

.
Addition of the angles corresponds to multiplication of the elements of
SO(2), and taking the opposite angle corresponds to inversion. Thus
both multiplication and inversion are dierentiable maps.
Hence, SO(2) S
1
(the example above).
The special unitary group of degree n, denoted SU(n), is the group of
n n unitary matrices with determinant 1. The group operation is that
of matrix multiplication.
Consider the group
SU(2) =

[ , C, [[
2
+[[
2
= 1

Now consider the following map:

: C
2
M
2
(C)
(, ) =

.
10
Since C
2
R
4
and M
2
(C) R
8
we can see that is an injective real
linear map and, hence, an embedding. Now considering the restriction
of to the 3-sphere S
3
, we can see that this is an embedding of the 3-
sphere onto a compact submanifold of M
2
(C). However it is also clear
that (S
3
) = SU(2). Therefore S
3
SU(2) and so SU(2) is a compact,
connected Lie group.
Another way to say this is that the group SU(2) is isomorphic to the group
of quaternions of norm 1, i.e. SU(2) S
3
.
Since unit quaternions can be used to represent rotations in 3-dimensional
space (up to a sign), There is a surjective homomorphism from SU(2) to
the rotation group SO(3) whose kernel is 1. Hence, SO(3) RP
3
.
6.3 Remark. Hilberts 5th problem asked whether replacing dierentiable
manifolds with (C
0
-) topological ones can yield new examples. The answer
to this question is NO: [1952, Gleason, Montgomery and Zippin] showed that
if a group G is a topological manifold with continuous group operations, then
there exists exactly one smooth structure on G which turns it into a Lie group.
7 Types of Lie groups
11