The fundamental group: Minutes of Discussion

- 4
Scribe: Pavithran S Iyer

Helpful insights from: Prof. Arvind, Prof. Sanjeev, Prof. Chanchal Kumar
Ritabrata, Debmalaya, Amol and Shruti
Presented & Discussed on: 9th Sept, 2011
1 Recap
Brief recap of the presentation on 26th Aug, 2011:
1. Covering spaces:
p : E B is called a covering map when for every x B, a neighborhood U B
such that p
1
(U) = H

where V

E and p[
V
is a homeomorphism. Pictorially,
there are a set of every patches in the space E, each of which are homeomorphic to U.
Example: p : R S
1
dened by p(x) = (cos(2x), sin(2x)), is a covering map. There
were four open sets containing the neighborhood of each point, the inverses of each of
which were disjoint neighborhoods in R
1
: p
1
(a > 0) =
_
n
1
4
, n +
1
4
_
, p
1
(b >
0) =
_
n, n +
1
2
_
, p
1
(a < 0) =
_
n
1
4
, n +
1
4
_
, p
1
(b < 0) =
_
n, n
1
2
_
;
where (a, b) S
1
and n Z.
Theorem: If p : E B is a covering map, then the induced homomorphism p

:
(E) (B) is an injection. Consequently, (E) is a subgroup of (B). To Prove,
rst we construct a mapping between equivalence classes of E and B, then prove its a
homomorphism and nally using path lifting properties, we show its kernel is a singleton
set implying p

is injective.
(a) p : E B is surjective but not injective
(b) p

: (E) (B) is injective not surjective

(c) Question: How to show that explicitly that (B) has more elements that (E).
(d) Consequences: p : S
1
R
1
, p : S
2
S
1
S
1
cannot be covering maps.

Note that the formalisms developed and discussed in this particular presentation is not required for study-
ing particle statistics, but having familiarized ourselves with a reasonable number of topological concepts, it is
worthwhile to look at the reason behind some important physical considerations in the theory of spin-half particles.

email: [email protected] , [email protected]

1
(e) The converse does not necessarily hold. Ex: p : RP
2
S
1
is not a covering map
whereas (RP
2
) Z
2
and (S
1
) = Z.
2 Topological Groups
We dened a topological space as a pair (X, ) where 2
X
satisfying certain properties. Given
a set, one can also think of imposing
I
a group
II
structure on it (or maybe there naturally exists
a group structure) with respect to some binary operation, say . Hence the triple (X, , ) would
be such that (X, ) form a topological space and (X, ) form a group. Moreover, we demand that
whenever three points in the topological space are related by a group operation: x
1
x
2
= x
3
, all
points in the neighborhood of x
1
and x
2
must also be related by the same group operation to any
point in the neighborhood of x
3
.
Figure 1: Suppose V X is a neighborhood of the point x
3
X and x
1
x
2
= x
3
, then there
must exist U
1
, U
2
X which are neighborhoods of points x
1
, x
2
X respectively, such that the
set U
1
U
2
= x y[x U
1
, y U
2
lies in V , i.e, U
1
U
2
V . In otherwords, f : X X X
where f(x
1
, x
2
) = x
3
, must be a continuous map.
In otherwords, the group multiplication operation must be a continuous function on the topo-
logical space.
Denition 2.1 A set of elements S, along with a binary operation

dened on the set and a

collection of subsets 2
S
is said to form a topological group when:
I
Imposing a group structure can be done by having a bijection from each element of the set to an element of
another set which forms a group under some operation. A natural group structure is found when elements of the
set X themselves show group properties under some binary operation.
II
A set X forms a group under some binary operation when the following properties are satised:
1. x
1
, x
2
X, x
1
x
2
X.
2. x
1
, x
2
, x
3
X, x
1
(x
2
x
3
) = (x
1
x
2
) x
3
.
3. e X such that x X, e x = x e = x.
4. x X, y X such that x y = y x = e.
2
1. (S, ) forms a topological space
2. (S, ) forms a group
3. The group operations f : SS S and g : S S describing multiplication: s
3
= f(s
1
, s
2
) =
s
1
s
2
and inverse: t = g(s) = s
1
, where s
1
, s
2
, s, t S, must be continuous maps.
The last condition is important because it makes the algebraic and the topological conditions
consistent. If the inverse
III
operation was not continuous, then the topological properties of the
space would be lost under this operation, as the result of which the elements would no longer form
the same topological group. In otherwords, if the set of elements X form a group, then the set of
elements obtained by taking inverse of every element in X must form the same group. However,
if X as a topological space was connected
IV
and the inverse map was not continuous, then the
image of X under this map need not be connected, thereby implying that the same topological
group cannot be obtained under the inverse map. Similarly, if the multiplication operation was
not continuous and the elements of X forms a group then the set formed by multiplying each pair
of elements from X would not form the same topological group (since the topological property of
the space would be lost).
Remark Since a standard topology is always dened given any set, we can omit and refer to
the set X as a topological group, separately specifying the operation under which the elements of
the set show the group properties.
Remark When a group G (under some binary operation) contains a set which can be given a
standard topology, then we can talk of the topological properties of G. For instance, phrases like
G is path connected . . . can be used. Implicitly, it means that the topological space formed by
the underlying set in G (along with the dened topology) is path connected.
We have already seen that the set set of all equivalence classes of loops forms a group under
the product operation

dened on the equivalence classes. In a topological group, the points of

the topological space also form a group under some operation

. We would now relate these two

products by giving an analogous denition for

on the set of equivalence classes. Note that any

denition we give for this product must be well dened on equivalence classes, i.e; the (new) product
of two equivalence classes of any two loops must be the same as the product of equivalence classes
of two other corresponding homotopic loops. We dene the group operation

on equivalence
classes as: [] [] = [ ] which is a point-wise product. Moreover, ( )(t) = (t) (t),
as a consequence of which the (new) multiplication is well dened on the equivalence classes.
As a consequence of the denition of this new product of equivalence classes, which is just the
equivalence class of the point-wise product, we see that the fundamental group turns out to be
abelian.
III
Moreover, the inverse map is a homeomorphism from X onto itself. If g : X X describes the inverse operation
g(x) = x
1
, notice that the inverse function g
1
: X X is nothing but the identity map g
1
(x
1
) = x which is
continuous and bijective. Hence we see that the inverse map is continuous, bijective and has a continuous inverse,
thereby being a homeomorphism.
IV
Connectedness is just one example of a topological property. In general it can be any property like Path
connectedness, Simply connectedness, Compactness, etc.
3
Lemma 2.1 The fundamental group of a topological group is abelian
V
.
Proof Let , eta : I X be two loops in the topological group T, under the operation . Notice
that each loop is homotopic to its product with the constant loop c
e
and since the multiplication is
well dened on the equivalence classes, we see that [ c
e
] = [c
e
] = [] and [ c
e
] = [c
e
] = [].
Hence the product of the equivalence classes [] and [] can be expressed as [ ] = [ c
e
] [e
e
]
as well as: [ ] = [c
e
] [ c
e
]. Using the usual method of expressing the concatenated path,
we see that there are two expressions for the same point-vise product [ ]:
Using
_

_
c
e
when 0 t
1
2
(2t 1) when
1
2
t 1
and
_

_
(2t) when 0 t
1
2
c
e
when
1
2
t 1
( )(t) (t) (t)
_

_
(2t) when 0 t
1
2
(2t 1) when
1
2
t 1
= ( )(t) [ ] = [ ]
and using
_

_
(2t) when 0 t
1
2
c
e
when
1
2
t 1
and
_

_
c
e
when 0 t
1
2
(2t 1) when
1
2
t 1
( )(t) (t) (t)
_

_
(2t) when 0 t
1
2
(2t 1) when
1
2
t 1
= ( )(t) [ ] = [ ]
Hence it can be seen that: [ ] = [ ] = [ ] implying: [] [] = [] [] or, in otherwords,
the fundamental group is abelian.
Example R, Z form topological groups under the addition dened for real numbers and integers
respectively. More generally R
n
forms a group under the addition operation dened for ntuples
as adding each component of the two ntuples to obtain a third one.
Example R 0, C form topological groups under the multiplication operation dened for
real and complex numbers respectively. Similar to the previous case, R
n
0 and C
n
also
form topological groups under multiplication dened for ntuples of real and complex numbers
respectively (as multiplying the individual coordinates).
Example S
1
= z C[[z[ = 1 forms a topological group under the multiplication operation
dened for complex numbers. The set satises the group axioms:
Closure: z
1
, z
2
S
1
, we have [z
1
z
2
[ = [z
1
[ [z
2
[ = 1. Hence z
1
z
2
S
1
.
Identity: z S
1
, we have 1 z = z 1 = z. Hence 1 is the identity element of S
1
.
Inverse: zS
1
, we have [z
1
[ = [ z[ = [z[ = 1. Hence z
1
S
1
.
V
A group (G, ) is abelian when g
1
g
2
= g
2
g
1
g
1
, g
2
G
4
Associativity follows from the multiplication rule dened for complex numbers and the prop-
erty: [z
1
(z
2
z
3
)[ = [(z
1
z
2
z
3
[, z
1
, z
2
, z
3
S
1
.
Hence the circle has a natural group structure.
Example GL
n
(R), consisting of all nn real non-singular matrices ([M[ , = 0, M G), forms a
topological group under the matrix multiplication operation. Before calling GL
n
(R) a topological
group, we must dene the topology on the set of matrices, which is not very obvious. This is
done by mapping each n n matrix to a point in the topological space R
n
2
(which would have n
2
coordinates). Each element of the n n matrix is mapped to a unique coordinate of the point in
R
n
2
. Product of two matrices in GL
n
(R), given by M
3
= f(M
1
, M
2
) = M
1
M
2
is mapped set of
coordinates which are expressible as a polynomial in the coordinates of M
1
and M
2
, thereby making
the function continuous. Similarly, the inverse also would be a continuous function R
n
2
R
n
2
where the coordinates of the inverse point can be expressed as a polynomial in the coordinates
of the the point. Hence one can talk about topological properties of GL
n
(R), implicitly meaning
to refer to the topological properties of the subspace of R
n
2
to which each matrix in GL
n
(R) is
mapped to.
Lemma 2.2 GL
n
(R) has two connected components.
Proof Recall that a continuous function between two spaces f : X Y preserves the connected-
ness of the space, i.e, the number of connected components in X and Y would be the same. The
determinant a continuous function
VI
det : GL
n
(R) R

. This is because the determinant of a

matrix can be expressed as a polynomial in the matrix entries. Hence, the number of connected
components in R

must be equal to that in GL

n
(R), implying that GL
n
(R) has two connected
components. Moreover, these connected components are nothing but two sets of n n matrices,
one of which contains all matrices with positive determinant and the other contains all matrices
with negative determinant.
3 Groups structure on special topological spaces
Lie groups are topological groups dened on a particular class of spaces, called a Manifold.
3.1 Topological Manifold
Loosely speaking, a manifold is a topological space where a coordinate system (or coordinate
systems) can be dened to express all points in the space. This means that dierent sections of
this space resemble some other space where a coordinate system can be dened and an obvious
(as well as a general) type of topological space with a coordinate system is R
n
. In otherwords, a
manifold is a topological space which is locally homeomorphic to R
n
for some n, which is referred
to as the dimension of the manifold. In R
n
and hence in a manifold, one can dene not only
continuous but also dierentiable functions.
Denition 3.1 An n-dimensional manifold /
n
consists of
VI
Notation: R

R 0 is the punctured real line.

5
A topological space T
A collection of maps

: T R
n
, such that for U

T, the map (U

) = V

R
n
describes
a homeomorphism.
The homeomorphism (

) : R
n
R
n
whose action is given by (

)(

(U

)) =

(U

) are bijective, continuous, invertible and dierentiable, U

T.
The last condition says that in addition to the existence of local coordinate system, the transition
from one coordinate system to another as we move across the space, must be continuous. In
otherwords, if open neighborhoods U

and U

have dierent local coordinate systems (which are

essentially mappings from these neighborhoods to R
n
)

and

respectively, then the open set

common to both these neighborhoods U

should be expressible by both coordinate systems

as well as

. Moreover, the function (

) which describes the change in the coordinate

system from

to

must be continuous, invertible, bijective as well as dierentiable.

Having looked at the denition of manifolds, here are some examples:
Example Needless to say, R
n
is an ndimensional manifold.
Example S
1
is 1-dimensional manifold, since its locally homeomorphic to R
1
. Moreover, we have
already seen that R
1
is the covering space for S
1
. In this case, a single coordinate system suces
to describe all points on the circle.
Example The n-sphere is a n-dimensional manifold, since it is locally homeomorphic to R
n
and
globally homeomorphic to R
n
0. The n-torus is also an n-dimensional manifold since R
n
is the
covering space
VII
for the n-torus.
3.2 Lie groups
As stated initially, Lie groups are topological groups where the underlying topological space is a
manifold.
Denition 3.2 A Lie group is an n-dimensional topological manifold /
n
whose points param-
eterize the elements of a group, i.e, the set g(x)[x /
n
forms a group with respect to some
binary operation . /
n
must satisfy two conditions:
The function : /
n
/
n
/
n
, given by (x, y) = z must be dierentiable whenever
g(x) g(y) = g(z).
The function : /
n
/
n
, given by (x) = y must be dierentiable whenever g(x)
1
=
g(y).
where x, y, z /
n
.
VII
Note that the n-torus is formed by the quotient of R
n
with the lattice group Z
n
.
6
Note that the above condition is similar
VIII
to that for the topological group except that we have
demanded the group multiplication and group inverse operations to correspond to not only con-
tinuous but dierentiable functions on the manifold. Once again, we remark that the requirement
of multiplication and inverse maps being dierentiable makes the both the conditions of /
n
being
a manifold and parametrizing the elements of a group, consistent.
Example GL
n
(R), the set of all nn real non-singular matrices is a Lie group since the underlying
manifold is a subspace of 1
n
2
where each matrix is mapped to a point in this subspace. The
mapping is given by mapping every entry of the n n matrix to a unique coordinate of a point
in R
n
2
. Similarly, GL
n
(C), the set of all n n complex non-singular matrices is a Lie group since
every n n complex matrix can be mapped to a n 2n real matrix, which in turn is mapped
uniquely to a subspace of R
2n
2
.
4 Rotations and Rotation paths
Since the physical space under consideration is R
3
, we would now want to look at all automorphisms
on this space. These are nothing but the rotation operations which form a group under composition.
Moreover, this group is isomorphic to SO(3). Eulers rotation theorem states that any rotation
in R
3
would preserve an element of the 3dimensional vector space, which is called the axis of
rotation. Hence any rotation can be viewed as rotation by some angle about an axis n and
consequently any element of SO(3) can be parameterized using an axis n and an angle . We
would revisit the use of this parameterization shortly in (prop. 5.1).
4.1 Rotation in physics
A physical rotation is not the same as an automorphism (on the real 3dimensional vector space)
because a vector does not instantly take on a new conguration on rotation, rather it passes
continuously through a set of intermediate congurations. However the rotation operation or
the element in SO(3) does not seem to capture this point. The notion of a continuous rotation
by SO(3) is emphasized by saying that any intermediate conguration of the vector during the
process of this rotation (in R
3
) can be reached by the action of some rotation operator (in SO(3))
on the initial conguration of the vector. In otherwords, corresponding to every conguration
from a continuous set of congurations the vector takes in R
3
as it is rotated, there is a rotation
operator in SO(3) which can instantly rotate the vector (from its initial conguration), to this
conguration. This continuous set of such rotation operators in SO(3) is called a rotation path.
The initial conguration of the vector corresponds to the identity element of SO(3), which would
be starting point of the rotation path.
Denition 4.1 A rotation path is a continuous map : I O(n) where (0) = I
O(n)
.
VIII
Instead of considering the points in the manifold to form a group, we have considered a general case where
each point can be mapped to some element from another set, which forms a group under some binary operation.
This can be said for the topological group as well and hence this is not the dierence between the way Topological
groups and Lie groups are dened.
7
Remark As the rotation path is dened to be continuous, its starting point lies in the identity
component of O(3) (which is SO(3)) and I is a connected set, we see that the entire rotation path
(for any general rotation) lies in SO(3). Subsequently, SO(3) is called the rotation group.
We will now formally (by means of a lemma) relate the actual path
v
traced by a vector v in the
n-dimensional vector space as it is continuously rotated, to the corresponding path in SO(n)
consisting of all the rotation operators required to reach each of the intermediate congurations
(from the starting conguration). This must hold for every vector in the n-dimensional vector
space, but it suces to show the above property just for the basis vectors as every other vector
can be expressed as linear combinations of the basis. Notice that each point on the path denoted
by (t) is a rotation matrix in SO(n).
Theorem 4.1 Let B R
n
be a basis. A map : I SO(n) is continuous if and only if x B,
the map
x
: I R
n
,
x
(t) = ((t))x.
Proof Let x
j
B. The path
(x
j
)
: I R
n
dened as
(x
j
)
(t) = ((t))x
j
is continuous if
and only if each component of ((t))x
j
is continuous. The action of the matrix (t) on x
j
gives
the intermediate conguration of the vector x
j
at time t: [((t))x
j
]
i
=

k
((t))
ik
x
k
. If every
component must be continuous then ((t))
ij
must be continuous in t. But every entry of the
matrix being continuous implies that the matrix is continuous in t. Hence we see that must be
a continuous map.
5 Topologies of SO(3) and SU(2)
5.1 Topology of SO(3)
We now want to look at the homotopy classes of paths in SO(3), the precise reason for which
would soon become evident (sec. 6). Moreover, as we have decided to look at rotation paths which
is more physical rather than rotations (as automorphisms on the real vector space), dierent
types of rotations must be characterized by dierent types or in otherwords, dierent homotopy
classes of rotation paths. To nd the homotopy classes of paths in SO(3) we must rst have a
space whose points parametrize all elements in SO(3), which we saw was a 3-disk of radius , with
its antipodal points identied. Since we have the antipodal points on the boundary identied, any
path which reaches a point on the boundary, and instantly reappears at the antipodal point
is still a continuous path. Hence we may have loops which start from a point, pass through a
boundary point, reappear at the antipodal point and go back to the initial point. Notice that
such loops are not homotopic to the constant loop because two of its points must always remain
antipodal, thereby implying that one cannot have a continuous family of loops between this and
the constant loop.
8
Figure 2: Two non-homotopic loops in SO(3). The path connecting A and A

is also a loop since

these two points are identied, thereby implying that any path between them is homotopic to the
constant loop based at either of the points.
Since any loop based at a point g denotes a 2 rotation starting from a particular conguration
(depending on the point g, we can take g to be the identity), this immediately implies that there
are 2 rotations that do not have the same rotation path, or in otherwords do not correspond to
the same action on the vector being rotated.
Figure 3: In the second gure, the path is a loop since points A and A

are identied. This loop

cannot be homotopic to the constant loop or to the loop in the rst case because neither of them
have antipodal points. Loosely speaking, if we try to shrink the loop to the loop in the rst case,
then once the points A and A

are no more antipodal (which must happen at some stage of the

shrinking process), then the loop would snap. However, in the third gure, we have the case of
a 4 rotation where the rotation path starts from 0 (denoting the I), to A, to the antipodal point
A

, then to B (thereby completing one 2 rotation), then to the antipodal point B

and nally
back to 0, thereby completing another 2 rotation.
Notice that since points A

and B (in g. 3) are independent (they are not antipodal) and

belong to the same loop, they can be moved close to each other, resulting in a similar motion of
the corresponding antipodal points A and B

too. When A

and B coincide, the loop in the third

9
gure would resemble the one in the rst gure, which is clearly null-homotopic. Hence we see
that there are only two homotopy classes of loops in SO(3). Moreover, we see that if we denote the
loops homotopic to the one represented in the rst case as [rst] and those homotopic to the loop
in the second case as [second], we have [second] [second] = [rst]. The multiplication algebra can
be summarized as:
[rst]] [second]
[rst]] [rst]] [second]
[second] [second] [rst]]
(1)
We will make use of this result shortly in (sec. 6). Moreover, the fundamental group of SO(3) as
isomorphic to Z
2
, whose lowest dimensional representations are +1 for [rst] and 1 for [second].
5.2 Relating their topologies
Firstly, we look at two results, each of which relates a physically important Lie group to a topo-
logical surface. The rst result concerns the SO(3) group formed by the all 3 3 real orthogonal
matrices with unit determinant
IX
and the second contains the SU(2) group formed by all 2 2
complex unitary matrices with unit determinant. The physical signicance of this group is de-
scribed in detail in a dierent section, here our motivation to look at the result is to arrive at some
topological space that is easy to visualize geometrically that SO(3). In otherwords, we would like
to look for topological spaces homeomorphic to SO(3) so that we can look at their relation with
spaces homeomorphic to SU(2) instead of directly looking for any topological relations between
SU(2) and SO(3).
Proposition 5.1 SO(3) is homeomorphic to RP
3
.
Proof Proposition 5.2 SO(3) is homeomorphic to the 3-disk of radius with antipodal points
identied.
Proof We already saw that each element of SO(3) can be mapped to a point on the 3disk (radius
), the direction of whose position vector denoted the axis of rotation and the magnitude of the
position vector denoted the angle of rotation. As a consequence of rotation by about either n
or n being the same, we had identied the antipodal points on this disk. This mapping from
SO(3) to the 3disk is clearly bijective and continuous. The inverse mapping maps every vector
v to an element of SO(3) whose eigenvector is v and which rotates every vector perpendicular to
v by [v[. This map too is continuous. Hence we see that SO(3) (as a Lie Group) is homeomorphic
to 3disk of radius with antipodal points identied.
Proposition 5.3 D
3
with antipodal points identied is homeomorphic to RP
3
.
Proof Firstly, note that RP
3
is the Real projective space, a set of all lines in R
4
passing through
the origin. It can also be interpreted as the space obtained by identifying the antipodal points
of the three sphere, i.e, if we dene an equivalence relation on points in S
3
as x y i y = -x,
IX
Note that SO(3) stands for Special Orthogonal group of 3-dimensions, where Special is used to denote that
elements of the group have unit determinant.
10
then: RP
3
=
S
3

. Consider the projection of every point on S

3
on to the equatorial plane (which
would be D
3
), p : S
3
D
3
, given by (x
1
, x
2
, x
3
, x
4
) (x
1
, x
2
, x
3
, 0). This is certainly a two-to-
one (continuous) mapping from S
3
to D
3
since antipodal points on S
3
would be mapped to the
same point on the equatorial plane D
3
. But if we have identied the antipodal points on S
3
, then
the image of this map is
_
(x
1
, x
2
, x
3
, 0)[x
2
1
+ x
2
2
+ x
2
3
1
_
with (x
1
, x
2
, x
3
, 0) = (x
1
, x
2
, x
3
, 0)
whenever x
2
1
+x
2
2
+x
2
3
= 1. In otherwords, the image of this map is just D
3
with antipodal points
identied. Moreover, this mapping becomes continuous as well as bijective. Its inverse would be a
map p
1
: D
3
S
3
, given by (x
1
, x
2
, x
3
, 0)
_
x
1
, x
2
, x
3
,

_
(1 x
2
1
x
2
2
x
2
3
)

_
, which is again
continuous and moreover bijective since antipodal points have been identied on S
3
. Hence the
mapping p : S
3
D
3
is a homeomorphism.
Since both SO(3) are well as RP
3
are each homeomorphic to D
3
with opposite points identi-
ed and homeomorphism is an equivalence relation (by transitivity), we conclude that SO(3) is
homeomorphic to RP
3
.
The second result relates the special unitary group SU(2) to another easily visualizable topo-
logical space than SU(2).
Proposition 5.4 SU(2) is homeomorphic to S
3
.
Proof S
3
is referred to as the 3-sphere and as a set, S
3
=
_
(x
1
, x
2
, x
3
, x
4
) R
4
[x
2
1
+ x
2
2
+ x
2
3
+ x
2
4
= 1
_
.
It is a three dimensional manifold in the four dimensional space R
4
. To see this homeomorphism,
we will take a general 22 complex matrix, impose the conditions for it to lie in the SU(2) group.
We would notice that under these conditions, each matrix maps to a point in a subspace of R
4
with a unit length position vector, as a consequence of which the subspace is identied as S
3
.
Let M =
_
z
1
z
2
z
3
z
4
_
GL
n
C. For M to be an element of SU(2), we must have:
M

M = I which implies that M

is the inverse of M, i.e, M

= M
1
. Using the explicit
form of M in terms of the matrix entries, we have
X
:
M

=
_
z
1
z
3
z
2
z
4
_
and M
1
=
_
z
4
z
2
z
3
z
1
_
Equating the above two matrices gives: z
1
= z
4
, z
3
= z
2
, z
2
= z
3
and

4
1
= z
1
. M can
now be rewritten as M =
_
z
1
z
2
z
2
z
1
_
. Hence we see M can be mapped to a an element in
R
4
, or in general SU(2) can be mapped to a subspace of R
4
.
X
The inverse of a matrix is dened by M
1
=
adj(M)
det(M)
where adj(M) denoted the adjoint matrix of M which is
obtained by taking a transpose of the cofactor matrix (the cofactor matrix C of M is dened as C
ij
= det(M
ij
)
(1)
i+j
where M
ij
is the minor matrix obtained by deleting the i
th
and j
th
rows and columns (respectively) of M)
and det(M) = 1 for matrices in SU(2).
11
det(M) = 1: Using the above matrix form for M, we have det(M) = z
1
z
1
+ z
2
z
2
= 1.
Expressing the complex entries z
1
, z
2
in terms of their real and imaginary parts as z
1
= x
1
+ix
2
and z
2
= x
3
+ ix
4
, the unit determinant condition on M translates to:
x
2
1
+ x
2
2
+ x
2
3
+ x
2
4
= 1
Hence we see that all points in the subspace of R
4
to which SU(2) matrices are mapped
satisfy the above relation. In otherwords, we see that the corresponding subspace of R
4
is
nothing but the 3-sphere.
Hence we see that manifold on which SU(2) is dened, is homeomorphic to the 3-sphere.
Having given these two homeomorphisms, we are now justied in considering topological rela-
tions between RP
3
and S
3
to be the same between SO(3) and SU(2) (respectively).
Lemma 5.5 The map p : S
3
RP
3
is a double cover.
Proof It suces to give an explicit map from S
3
to RP
3
which satises the properties of a covering
map (review point: 1) and the bre over a point in RP
3
has two points. Intuitively, we need a
map such that the inverse image of each line (passing through the origin) is a discrete set of two
points on the sphere and the easiest method to do this is to map every point on the sphere to the
line passing through it (and the origin). Notice that the bre over a line would be two antipodal
points on the sphere and similarly, for any open neighborhood of lines their inverse image would
be two patches on the sphere on the antipodal ends, and moreover, restricting the mapping to
one patch, it is a continuous bijection with a continuous inverse.
Proposition 5.6 The map p : S
3
RP
3
given by p(x
1
, x
2
, x
3
, x
4
) =
XI
[x
1
: x
2
: x
3
: x
4
] is a
covering where the bre over each line has two points.
Proof Firstly, note that p : S
3
RP
3
is surjective since every line passing through the origin
also passes through some point on the sphere and hence lies in the image of p. Let U RP
3
be
a open neighborhood of a line l RP
3
. U is nothing but two solid cones, sharing a vertex at
the origin with (curved) bases on the antipodal ends of the sphere. The inverse map of this open
neighborhood is precisely the two open sets V
1
, V
2
in S
3
(corresponding to the bases of the solid
cones). Since there are two of them, the bre over every line has two points (conversely every line
passing through the origin intersects the sphere at two points). Restricting to any V
i
, the map p
is a homeomorphism since it is a bijection (every point is mapped to only one line, which is the
one passing through that point), continuous and has a continuous inverse.
Hence we see that S
3
is a double cover of RP
3
.
Corollary 5.7 Since S
3
is homeomorphic to SU(2) and RP
3
is homeomorphic to SO(3), we say
that SU(2) is a double cover of SO(3).
XI
The notation [x : y] means to refer to all pairs (x
i
, y
i
) that are related by a scalar multiple, i.e, [x : y] = [u : v]
i x = u and y = v.
12
6 Quantum Evolution and topology of group of operations
We would now consider operations on quantum states, which implies that the rotation operators
must be unitary. Hence given a group G of operations, we must construct its unitary representation
consisting of a unitary matrix U(g) for every g G, which act on the state-space of the system.
Moreover, if G is a Lie-group, we can demand the unitary operations U(g) to vary smoothly with
g. The unitary action on a quantum state is only specied upto a phase, i.e, any two states
related by a phase factor (a complex number in general) is essentially the same with respect to
measurement. As a consequence of the same ambiguity, each eigenvalue of U(g) is specied upto
an arbitrary phase, say (g). Let us specialize to the case where the state-space of the system
is N dimensional implying that the matrix U(g) is N N. Since each eigenvalue has a complex
scalar multiple of (g), the condition det(U(g)) = 1 implies that
N
= 1.
As a consequence of the smooth variation of U(g) with g and the fact that its action on a state
vector can only be specied upto a phase factor, we see that as we go along a closed loop l based at
g, in general, the corresponding unitary operator undergoes the transformation U(g)
l
(g)U(g).
Since rotations are characterized by rotation paths, we neednt assume that loops which are not
equivalent to each other produce the same (arbitrary) complex scalar multiple. However, if l and
l

are homotopic, then the physical rotations are the same, implying
l
=
l
. For convenience,
we may make
l
0
(g) = 1 where l
0
is the constant loop based at g. Similarly, if the rotation path
comprises of n loops, the corresponding unitary operator varies as: U(g)
l
1

l
2

ln
U(g)
where
l
i
=
l
j
when l
i
is homotopic to l
j
. Moreover if g is the identity element, one can say that
there exists some 2 rotation which do not leave the state-vector invariant. We would from now
on consider only loops based at I which corresponds to the initial conguration of the state-vector
being rotated.
Let us now take G = SO(3), which we saw has two homotopy classes of paths (eq. 1) that
form a group under the multiplication operation (dened on the equivalence classes). Hence we see
a 2 rotations along the loops l, l

which belong (respectively) to [first] and [second], results in

the transformation, in the corresponding element of the unitary representation, by scalar multiples
of
l
and
l
respectively. Moreover, since the unitary representation of SO(3) also satisfy the
same multiplication table (eq. 1), we see that
l
= 1 and
l
= 1. Note that the state-vectors
on which the group SO(3) acts are nothing but the angular momentum states labelled as [j, m
where j m j and j =
n
2
for n Z
+
, where the dimension of the state-space is given by
N = 2j +1. Combining the conditions
N
= 1 (arising from det(U) = 1), = 1 and N = 2j +1,
we see that whenever j is integral, then N is odd implying that the only solution for
N
= 1 is
= 1. Hence for states with integral spin (states with integral j)
l
= 1 for all loops, whether
contractible or not. In otherwords, a 2 rotation brings the state-vector back to its original
conguration. On the contrary, if j is half-integral, then N is even which implies that
N
= 1
has two solutions, = 1. Hence the multiplicative complex factor is dierent for dierent loops.
As there are only two types of loops (2 rotations), by convention we take this scalar multiple
to be 1 for null-homotopic loops and 1 for the loops in the other class. Hence we see that the
unitary representations for SO(3) do not result in the same conguration after a 2 in the case
of half-integral spin states. In otherwords, the half-integral spin representations of SO(3) are
multivalued. One way of getting rid of the multivaluedness is to keep tract of whether the number
of 2 rotations being performed is even or odd, just like how we deal with the multivalued nature
13
of the complex square-root function or the complex logarithm function.
But we would want to do better, by having a group of unitary matrices and moreover one
which is simply connected. A group which is the unitary representation of SO(3) and is simply
connected is SU(2), which we saw is the double cover of SO(3). In this regard, SU(2) is the group
used to examine rotation of angular momentum state-vectors with half-integral spins or spin half
particles, the basis for which is the set of three Pauli matrices
_
0 1
1 0
_
,
_
0 i
i 0
_
and
_
1 0
0 1
_
.
References
[1] [BOOK]: Topology, by Munkres, J.R., 2nd edition, Prentice Hall, 2000, http://books.google.
co.in/books?id=XjoZAQAAIAAJ , http://www.amazon.com/Topology-2nd-James-Munkres/
dp/0131816292
[2] [Book, Chapter 2] Algebraic Topology, by Allen Hatcher, http://www.math.cornell.edu/
~
hatcher/AT/ATpage.html
[3] [Book, Chapter 2]: Lie groups, physics, and geometry: an introduction for physicists, en-
gineers and chemists, by Robert Gilmore, (published:) Cambridge University Press, ISBN:
9780521884006, ICCN: 2008272822, http://books.google.com/books?id=MHExMkMFzvcC,
http://einstein.drexel.edu/
~
bob/LieGroups.html
[4] [research-article]: What is a Topological Group?, by Deane Montgomery, The American Math-
ematical Monthly, Vol. 52, Number 7, pp. 302-307, July 1945, ISSN: 00029890, (published:)
Mathematical Association of America, http://www.jstor.org/stable/2305290
[5] Supplementary Notes: Spin, topology, SU(2)SO(3) etc, by Ben Allanach, http://www.damtp.
cam.ac.uk/user/examples/D18S.pdf
[6] Rotations, Rotation Paths, And Quantum Spin, by Michael Thvedt, http://www.math.
uchicago.edu/
~
may/VIGRE/VIGRE2006/PAPERS/Thvedt.pdf
[7] [Wikipedia Resources]: Wikipedia pages for
Topological Group: http://en.wikipedia.org/w/index.php?title=Topological_
group&oldid=443233546, last modied: 5th Aug, 2011; at 1947 Hrs.
Manifold: http://en.wikipedia.org/w/index.php?title=Manifold&oldid=
446910218, last modied: 27th Aug, 2011; at 0142 Hrs.
Charts on SO(3): http://en.wikipedia.org/w/index.php?title=Charts_on_SO(3)
&oldid=447144297, last modied: 28th Aug, 2011; at 1435 Hrs.
Lie Group: http://en.wikipedia.org/w/index.php?title=Lie_group&oldid=
446751257, last modied: 26th Aug, 2011; at 0131 Hrs.
14