Professional Documents
Culture Documents
Curtis
Matrix Groups
I
Springer-Verlag
New York Heidelberg Berlin
Dr. Morton L. Curtis
Department of Mathematics
Rice University
Houston, Texas 77001
USA
(Universitext)
Includes index.
I. Matrix groups. I. Title.
QAI71.C87 512'.2 79-23523
987654321
the spring of 1977. It is assumed that the students know some linear
be an invariant.
We define quotient groups and then note that we get new groups
G whenever a matrix group has nontrivial center.
center
of the text. All of our matrix groups are in some euclidean space
We prove that any two maximal tori are conjugate and that, if G is
know the dimension, center and rank of all of our matrix groups and
groups and covering groups. The only new groups which arise are the
show that
izer is the semidirect product of the torus and the Weyl group, we
Spin(2n+l) ~ ~
center - center
Spin(2n+l)
The normalizer in SO(2n+l) ) splits
center
for n = 1,2,3, ...
Lie groups.
CONTENTS
Chapter 3 Homomorphisms 35
A. Curves in a Vector Space 35
B. Smooth Homomorphisms 41
C. Exercises 43
Chapter 6 Topology 73
A. Introduction 73
B. Continuity of Functions, Open Sets, Closed Sets 74
C. Connected Sets, Compact Sets 79
D. Subspace Topology, Countable Bases 82
E. Manifolds 86
F. Exercises 89
Chapter 7 Maximal Tori 92
A. Cartesian Products of Groups 92
B. Maximal Tori in Groups 95
C. Centers Again 100
D. Exercises 104
xii
A. Groups
X x Y (x,y)lx EX and y E Y} ,
the curly brackets being read as "the set of all" and the vertical
p SxS~S
operation
~:GxG-+G
a,b,c E G we have
(ab)c a(bc) .
~(~(a,b),c) = ~(a,~(b,c))
Note that properties (ii) and (iii) leave open the possibilities
that there may be more than one identity element and that an element
may have more than one inverse. But neither of these can happen.
b eb (ca)b c(ab) ce c •
Exaroples
verse for 2
(4) The set ~ - (0) (i.e., all nonzero rationals) is a group under
multiplication.
x + Y
The identity is
4
fO g : S -? S
Then this is the identity element for G for this operation. Then
have ab = ba •
In the examples above, (1), (3), (4), (5), and (6) are abelian
(Exercise. )
homomorphisms.
What this means is that we can first multiply a and b (using the
H. So
For a € G we have
-1 -1
er(a)er(a ) er(aa ) rr (e) e' ,
-1 -1
showing that er(a ) (er(a))
example, the map er: R ...,. R (er(x) = (x,x)) is injective, and the map
6
2
p R ~ R (o(x,y) = x) is not injective.
by
x
cr (x) a
Then cr is a homomorphism
cr(x + y) a(x)a(y) .
Thus these t~o groups are isomorphic--not only that, but there are lots
of isomorphisms.
cr is injective ~ cr-l(e / ) e .
7
-1 -1
cr(a) cr(b) .. cr(a) cr(b) e' .. o(ab ) e'
and
ab- 1 e .. a = b .
B. Fie1ds, Quaternions
a(b + c) ab + ac
p1ication we try
(1,0)(0,1) = (0,0) .
8
Now (0,0) is the additive identity or "zero" and we would have two
a field because:
ab '" 0 •
-1 -1
but a (ab) (a a)b lb o so b o.
(ac - bd,ad + bf) + (ad - bf,af + be) and we easily check that these
are equal. Next we need to see that if (a,b) '" (0,0) then it has a
9
a -b
(a, b) (2 2' 2 2) (1,0) ,
a +b a +b
as you ean readily verify. Thus we have made R2 into a field whieh
we denote by C and eall the eomplex numbers.
You may know that there is a simple mnemonie deviee for remem-
Thus
X € ~ be x + iO .
x + y = x + iO + Y + iO (x + y) + iO
2
S~we have taken the field R as all (x,O) in ~ and extended
10
3
to a field on R Now for the bad news.
R3 into a field.
~i
Proof: Take basis vectors l,i,j so
3
that any element of ~ can be written
-j ia - b + ijc
-j ia - b + (a + ib + jc)c
-j (a - b) + i(a + b) + jC 2 .
2
This implies c -1 , contradicting CER.
and define 1 i j k
1 1 i j k
i i -1 k -j
j j -k -1 i
k k j -i -1
+ k(aw + bx + bz - ey)
4
R with this mu1tip1ieation is ea11ed the quaternions. It is easy
axioms (i), (ii), (iii)' are readi1y verified exeept for showing
q = a + ib + je + bd
we set
q
-1 a- ib -
2 2 2
kd ;e -
a +b +e +d
-1 -1
and readi1y verify that qq = 1 = q q
k E {R,r:, ll.} •
This makes kn into an abelian group with identity ~ (0, ... ,0) .
For c E k we define
n
and this makes k into a vector space over k (for k = ll. we must
0(CX)
so is 1\10 $ .
Proof: (I\iO$)(cx+dy)
elements from k
map $(M) by
(~ .~) (CX ~ l Yl
d )
q cx + dy
n n
c (T)
q n
+ d (q~l) qYn
unique). The first row of M is the n-tuple ~(1,0, ... ,0), the
Note that if the matrix A gives the linear map ~ and the
way:
A + B (a .. + b .. )
~J ~J
lS
n x n array.
A(B + C) AB + AC
(B + C)A BA+CA •
plicative inverse.
group.
(iii)
GL(l,R) = R - (0)
GL(l,c) = C - (0)
GL(l,'H) = 1H - (0)
17
because all nonzero elements are units. GL(2,p) is the set of units
a,b,c,d E 1'l.}
GL(2,p) = t(~ ~)
ad - bc " 0
a.ö-Sy.
Then det(i
i
1)
.
= k - (-k) = 2k " 0 , but this matrix cannot be a unit
or the corresponding linear map would be an isomorphism, whereas
dJ(b) = y. Then
18
(/)(ab) E (/)(G) .
-1 -1
Finally, suppose x E (/)(G) . Then x = (/)(a) and so x (/)(a )
injective homomorphism
We begin with
defined by
I\i(x+iy+jz+kd) = (
X+i Y
-Z-iW)
z-iw x-iy
Lemma 9:
Lemma 10:
By LeUlUla 9
E. Exercises
rJ> is defined to be
20
we have
C = (y E G I xy = yx for all x E G)
centralizer C(S) of S by
N(S) = (x E G I xSx- l = S1
and call N(S) the normalizer of S. Show that C(S) c N(S) and
A B
HK {hk I h E H, k E K}
A rBs .)
H, then KH is a subgroup of G
iq
as follows: represent ~ E C as ~ = pe with p ~ 0 and set
cos e sin e)
JP (
-sin e cos e
A. Inner products
Namely,
for x ER, x =x .
For a x + iy E C, a x - iy .
~ + B
aB = Ba.
a.B = a.B
proof: Exercise.
Ixl=~.
the symbol
t-
(xA,y) - (x,y A) •
B. Orthogonal groups
Definition:
(t}(n,k) = [A E M (k)
n
I (xA,yA) = (x,y) for all x,y E kn } .
proposition 3: (t}(n,k) is ~~
so that
AB E (t}(n, k)
If A E (t}(n,k) we have
(eiA,ejA) (ei,e j ) ö ij
- {: if
if
i j
i ,f j
t- t ÄA
Thus AA = I But then is also the identity since
t(tÄA ) t(ÄtA) AtÄ t Ä A- 1
= = I Thus a 1eft hand and right
Finally,
-1 -1 -1 -1
(xA ,yA ) (xA ~,yA A) (x,y) ,
-1
showing that A E tt)(n,k) q.e.d.
27
(i) A € ~(n,k)
t- -1
(vi) A=A
Proof: Exercise.
This gives (x,y) + (y,X) = (xA,yA) + (yA,xA) and since (,) over
F is symmetric, this proves
28
(e.A,e.A) + (e.A,e.A) = 0 .
~ J J ~
<eiA,e.A)(x.X.
J ~ J
o
U(l) is just the set of all comp1ex numbers of length one. This is
I}
It is an interesting fact that these are the on1y spheres which can
be groups.
(det A)(det A) = 1 .
29
define
and eall this the special orthogonal group (also ealled the
just the refleetion in the first axis, and has determinant equal to
-1 .
!!l(n,k) for n = 1,2, ... and k E (R,r,E} ) and our major goal is to
find out whieh of these are isomorphie. The basic idea will be to
groups whieh are isomorphie must have the same invariants. This will
make it possible to say that eertain groups are not isomorphie. But
must look for isomorphisms. In this seetion we will give one iso-
morphism.
Suppose you suspeet that Sp(l) and SU(2) are isomorphie.
How wou1d you try to find an isomorphism? Sp(l) is the set of all
mu1tipl1cation.
~ : Sp(l) ~ SU(2) .
-e-id) (a-ib
a-ib e+id
ö a. and y e
D. Reflect~ons in pn
uJ. 01
thus
r = (v,U) .
v - 2ru v - 2(v,u)u .
(0\ .~).
using this basis, the ref1ection ~ is given by the matrix
to By Proposition 4 A is orthogonal. So
32
given by
so
COS 0. sin ~)
A (
-sin ~ COS 0.
angle r1.
E. Exercises
1. Prove Proposition 1.
2. Prove Proposition 4.
that
(cos 8 sin 8)
-sin 8 cos 8
)..A E U(n)
2
6. Let LI and L2 be lines through the origin in P Show
What can you say about the determinant of a idempotent matrix? What
whenever i > j
AB is not nilpotent.
whenever i > j
define these.
is an open interval in R.
y' (e)
a
•
e b
Now ,
Mn (e) , Mn (H)
Mn (") can all be considered to be real
2 2
vector spaees (of dimensions n , 2n 2 and 4n ). I f G is a matrix
zero 0 I rI r E R
matrix
(va)(u) = v(u)a(u)
ferentiable at c and
(va) (u)
so that
37
(vo) , (u)
the set of all tangent vectors v' (0) to curves v:(a,b) .... G,
o(u) = V(ru) .
v
-l.......---+-----<~ 1
U(l)
-i
38
Let
y(t) =
V 1 - sin 2 \,1t - sin 2 vt - sin 2 ).t + i sin ut + j sin vt + k sin kt
2
Example 3: dim GL(n, R) n
A E GL(n, R) If v is any
curve cr in Mn(R) by
39
o(t) tv + I .
2
A simi1ar argument shows that dim GL(n,C) = 2n .
We will now get upper bounds for the dimensions of O(n), U(n)
A + tA = 0 ;
so(n) , then
(A + B) + t(A + B) o,
denote the matrix whose entries are all zero except the ij entry,
E ij only for i < j, it is easy to see that they form a basis for
n(n-l)
(n-l) + (n-2) + ... + 1 = 2 of them .
over C .
n(n-l) 2
n + 2
=
2 n.
proof: Exercise.
C € Mn (E) skew-symplectic i f
C + t c o.
In the exercises one shows that the set sp(n) of such matrices is a
3n + 4 n(n-l) n(2n+l) .
2
t-
B (u) e (u) I.
Thus its derivative is zero, and the result follows from Proposition 1.
Corollary:
2
Dim U(n) ~ n
B. ~ horooroorphisms
p : (a,b) ~ G
d~(ap'(O) + ba'(O»
is $o~ and
Proof: The first part 1s obv1ous. For the second, let y'(O)
-1
Proof: I/J 0 I/J is the identity, so is the
-1
identity. Thus dI/J is injective and is surjective. I/J I/J is
-1
the identity, so is the identity.. Thus dI/J is
c. Exercises
CO"
sin t
y(t) -sion t cos t
0 :)
Show that y is a curve in SO(3) and find y' (0) Show that
2
(y )'(0) = 2y' (0)
o
cos t
o(t) - (:
-sin t
Ca1cu1ate 0' (0) Write the matrix for y(t)o(t) and verify that
iTTt
e 0 0
i TTt
2
P (t) 0 e 0
i nt
2
0 0 e
A. Exponential of ~ matrix
groups.)
where A2 means the matrix product AA, etc. We say that this
2
sequence converges if each of the n real-number sequences
converges.
converges.
(k-l)! nm
n
k-2 k-l
m
T
x
This exponential behaves somewhat like the familiar e (x € R) •
e
o I.
Also:
A+B AB
e e e
few terms.
47
A2 2 3 2 2 3
A+B
e I+A+B+-2 +AB+~+~+AB+AB
2 6 2 2
+~+
6
2 3 2 3
AB
e e (I+A+ A2 +~ + ... )(I+B+ B2 +B6 + ... )
2 2 3 2 2 3
I+A+B+~+AB+~+~+AB+AB +~+
2 2 6 2 2 6
e
0 A-A
proof: A and -A eOUlUlute, so I e
Proof: We have I = e
o proving
A
that e is orthogonal.
we see that
say: (i) it does not say that every orthogonal matrix is some eA
in some detail.
o x
0. - (-x 0)' x e: R •
2
To ca1cu1ate e"' , we ca1cu1ate the powers of 0. • 0.
2
- (-: _:2)
0.
3 -(:3 _:3) CI.
4
- (': X:) CI.
, (0 5 X')
-
-x 0
etc. Then
eCl. _
(10)+(0
o 1 -x
0) 1 ( 0
2 + 3T 3 X)
0
1
+2T (-:
2
_:3 )
-x x
1 x4 0 )
~
+ . ( 0 x 4 +sr
-x 0
S
1 (0 5 x ) + ...
x2 x 4 x6
1 - 2T + 4T - GY + .•• - cos x , etc.
We find that
and B ~ nonsingular , ~
49
BAB- l _ Be~-l
e
B. Logarithm
x
Just as e is defined for all x € R and log x is defined
only for x > O. the logarithm of a matrix will be defined only for
matrices near to the identity matrix I.
Let X be areal n xn matrix and set
n k k+l k k
~ T-'I'""K = k+l ne ~ ne .
n €
exp(V) c U Then
A A 2 1 A2 2 1 A2 3
log e (A + 2T + ... ) - '2(A + 2T + ... ) + '3(A + 2T + ... ) +
A2 A2 A3 A3 A3
= A + [2T - T] + [6 - T + 3] + ... = A
2 3
(i) is similar. log X = (X-I) - ~
2
+ ~ -
3
1 X (X_I)2 1 (X_I)2 2
e og =[I+(X-I)- 2 + ... )+2T(X-I)- 2 + ... )
1 (X_I)2 3
+ TI [(X-I) - 2 + ... ) +
2 2 333
= X - ~
2 + ~
2 + lr~
3 - ~
2 + Q.:l.L.)
6 + ...
= X •
is one-to-one near 0
I = xSc
so 0 = log xSc = log X + log Sc = log X + t(log X) showing log X
is skew-symmetric.
so(n)
t!l(n)-SO(n)
log
o
-
exp
log defined
here
here
C. One-parameter subgroups
is a smooth homomorphism
uA 2 A2
y(u) = e I+uA+u 2T+
such that
uA
y(u) e
with
y(u) = ecr(u) •
Let cr' (0) =A We just need to show that cr(u) is a line through
So
some one-parameter subgroup. We will see now that this is also true
A y' (0) •
Thus
t --
p(u) p(u) = I
so that
p'(O)+to'(O) 0, Le.,
o.
y(u) t y(u)
--
I .
subgroups.
matrices. Thus
2
dim U(n) dim su(n) = n
diagonal terms;
We c1ear1y have
To prove (ii) we just write it out. The sum of the diagonal terms in
AB is
55
Since ~ and C are commutative one easily checks that these are
equal.
Clearly
(iii) Tr(I) n.
Also
Tr(A)
-1
Proof: By (ii) , Tr(B(AB- l )) Tr( (AB )B) Tr(AI) TrA .
Tr(A) A
e = det (e ) .
We will prove this later, but a few comments are in order here.
First off, (f) looks wrong because the left hand side depends only
this is true for the right-hand side. The point is that det and
B is non singular
-1
A -1
det (e BAB ) det (Be-~ )
We will prove (t) once we have found how to put matrices in simpler
Tr : u(n) -> C
all of iR. From the rank theorem in linear algebra we know that
Thus the dimension of Tr-l(O) is just one less than dim u(n) i.e.
since
Tr (C) o .. 1 = eTr(C)
C
.. e E SU(n) .
D. Lie algebras
2 _x 2 O )
then o. ( 2
o -x
[A,B] AB - BA .
57
o .
Thus so(n) , su(n) and sp(n) become algebras (over R) with this
Property (iv) is called the Jacobi identity and its proof is a routine
verification.
(i) ... (iv) is called a Lie algebra. (One could clearly consider com-
So we have the trivial product (which obviously satisfies (i) ... (iv)).
We must have
so
we get the trivial Lie algebra. For any other choice we get a non-
trivial Lie algebra. In the exercises one shows that these non trivial
algebras, but will simply look at two which arise quite naturally.
a
o
-c
[i,j] k
and defining [j ,k] i
[k, i] j
E. Exereises
nor skew-symmetrie.
2
4. Prove that any two nontrivial Lie algebras on Rare iso-
3
A. The homomorphism p : 5 ~ 50(3)
50(3) , and, for all we know now, they might be isomorphie. In this
4
Proof: As veetor spaees over R , JH and R are the same.
4
So L is surely a linear map of F- for i f a,b E R and
q
o.,a E E we have
L (ao.
q
+ ba) aqo. + bqa = aL (0.) + bL (a) .
q q
4
calculate (L (i) ,L (j)
q q
(using ( , ) for ~ ) . We get
p(q)(a.) qa.q
x is a real quaternion
Note also that p(q) is the inverse of p(q) in the group tt;(4)
Together these two observations imply that p(q) maps the 3-space
3
Proposition 2· p : S -> Stt;( 3) is a surjective homomorphism and
Ker(p) [1,-1) c S3 .
3
Proof: If Ql,q2 E Sand a. E Span(i,j,k) , then
62
Tnus p is a nomomorpnism.
a E (l,-l) •
Finally we need to snow tnat p is surjeetive. This will be
Let q =a + ib + je + kd We want
(a + ib + je + kd)(k)(a - ib - je - kd) =k , or
(a + ib + je + kd)(ka - jb + ie + d) = k , so
ad - be + be - ad 0 (autom~ieally) ,
ae + bd + ae + bd =0 or 2(ae + bd) 0,
1 .
Now
2 2 2 2
a+d+b+e 1
so
63
o or b o e .
q = a + dk (with a2 + d2 1)
(a + kd)(i)(a - kd) j
(a + kd)(ia + jd) j
o, a ±d
2
ad + ad 1 if a = d , 2a 1 ,
(a + ka)(j)(a - ka) -i
(a + ka)(ja - ia) -1
2 2
a a o.
1 1 1
kl:-o
80 q = - + k - or q = Both will give the desired
.fi /2 J'I .fi
element of 80(3) (This shou1d be enough to eonviee us that we
Note that this does ~ prove that 83 and 80(3) are not iso-
B. Centers
C [x E G I xy yx for all y E G} ,
3
Proposition 3: The center of S = Sp(l) is [l,-l} , whereas
ai - b - ck + dj ai - b + ck - dj
2 3
and this implies b o. So q = a and a 1 . Thus Center S
= P,-l}
T = ('0'"-s~ 9
sin 8
cos 8
° ~)
since T c: 50(3) Consider the standard basis e1 (1,0,0) ,
e 2 = (0,1,0) , e 3 = (0,0,1) for R3 ".
this imp1ies a = 0= b
it is a rotation because:
is itse1f a rotation.
2 2
Let I/J R ~ R denote such an element of ~(2) . For any
rotation
cos 8 sin 8)
t = (
-sin A cos 8
8) and de t I/J
2 2
( 0. a. + B
-B a.
66
A € T •
(co,
~)
e sin e
A = -Si: e cos A
0
and we let
R = (~ ~ ~)
-1 0 0
€ SO ( 3) .
sin e
co,,) ( 0
C
0
AR cos e -sin e = -sin e cos e
:)- M
0 0 -cos e sin 8
Proposition 3 is proved.
tori. We conclude this chapter with a bit more abstract theory which
c. Quotient groups
C(x) Hx - (hx I h € H} .
-1
Also Hx -= Hy .. xy € H .. Y € C(x) .. X € C(y) • These equivalence
classes are called right ~ of H.
xH {xh I h EH} .
Let G/H denote the set whose elements are the rightcosets of
H in G
is 80(3) .
There is a natural map Tl: G -> G/H given by Tl(x) = Hx. In the
nel H.
-1 -1
xyx y
-1 -1
is called the commutator of x and y (because (xyx y )(yx) = xy).
Now the product of two commutators is not necessarily a commutator,
g ~ abelian group.
-1 -1 -1 -1
(xyx y )(yxy x ) e,
-1 -1
showing [G,G] is a subgroup. Let Z E G and xyx y E [G,G] .
Then
-1 -1 -1 -1 -1 -1 -1 -1 -1
z(xyx y )z = zxy(z (xy) (xy)z)x «yz) (yz))y z
-1 -1 -1 -1 -1 -1
= [z(xy)z (xy) }[x(yz)x (yz) }(yzy z } E [G,G]
normal subgroup.
-1 -1 -1
xy(yx) xyx y E [G,G] .
q.e.d.
C is its center, then G/C will have trivial center. But this need
w(x) : G ..,. G
-1 -1
Ez w(x) (y) xyx y Then G/C has nontrivial center ~ ~x E G-C
such that
w(x)(G) c: C .
Proof: ~
Cx E center G/C
-1 -1
CxCy Cxy Cyx CyCx so that xyx y E C for a11 Y E G •
D. Exercises
-1
R -1 0 Lx(g) = xgx
x
(L (i),L (k») = 0 .
q q
G ~ K
I "
G/H
-1
Show that rf, 0 "
is weIl defined and gives an isomorphism of G/H
onto K.
their centers.
Chapter 6
Topology
A. Introduetion
eompaetness, whieh some of our groups have and others do not. These
deeide whieh of our groups have them. This will be done in seetions
Band C.
for open sets, a eoneept we will need in our study of maximal tori in
,,~U,
~'-i-,1>A.(:,Z)
d(x,y)
of the vector x - y .)
n
proposition 1: This is a metric on R
Schwarz inequality.
(x+ty,x+ty) ~ 0 .
2
(x,X) + 2(x,y)t + (y,y)t ~ 0 .
polynomial can have only one minimum.) Thus the discriminant cannot
be positive; i.e.,
2
(2(x,y») - 4(y,y)(x,X\ ~ 0 .
So
2
(*) (x,y) ~ (x,X)(y,y\ .
and call this the open ball with center x and radius r. Open balls
Rm
f(x) E B(f(a),e) .
Another way of saying this is: Given € > 0 there ex1sts 0 > 0
d(f(a),f(x)) < e .
by
f(xl " ~: if
if
x < 0
x:::>: 0
f:A~".
is continuous at each a E A .
continuous.
points of A Let
i, ... , k
and let ö be the smallest of these. Then for any e > 0 any ele-
such element of A) .
Proposition 2: If A C Rn and
and thus
showing z E B(y,s) .
but no B(l,r) lies in (0,1] since every such ball contains num-
comp1ement Rn C is open.
x to points of K. Then
B(x,ö) c Rn - K ,
proving K is c10sed.
y:[O,l]-+D
y(t) = (x + t(y-x))
is a path in Rn from x to y
y det
[0,11 ~ ~(n) ~ R
tB
y(t) e
det e B = +1 so eB E SO(n)
81
closed.
However, when we put closed and bounded together, they are then
both preserved.
U V n W •
n
Note that i f W is an open set in R , then UcW is open in
n
W if and on1y i f it is open in R
n
For Wc R the co11ection of all open sets of W is the
subspace topo1o~l of W.
Of course, not every open set is an open ball, but open balls suffice
to give all open sets by taking unions (the "empty" union being
allowed) .
n
Definition: A co11ection of open sets in R is a
n
basis for open sets if every open set in R is a union of sorne of
the V 's
n
83
The set of all open balls in Rn iS, of course, a basis for open
sets. But so is
and we can give the same definition as above for the notion of a
We want to get bases for open sets which are "minimial" in the
sense that they have no more sets than needed to do the job. The
13 1245781
1, 2, 2' 2' 3, 3' 3' 3' 3' 3' 3' 4, 4'
1
1, -1, 2, -2, 2' -2' 2'
1 1. -2"3 3, -3, •.. contains all of 0).
ri .xi/i2xi3·· .
r = 'Y1Y2 Y3··· Yj = 5
"
Let where if x .. 5 and Yj 1 if
JJ
x .. = 5 Then
JJ
"
rl because Yl xn
r "
r " r2
because , etc.
Y2 " x 22
But r EI.
Proof: Let A
we can write
because following the path shown below will include all ofAx B .
85
n
Proposition 7: (and hence any WeR ) has a countable
set of balls.
B(y,s) c V .
s
d(x,y)-<:"3
s s
and let r be a rational number satisfying "3 -<: r -<: 2
86
E. Manifolds
and f- l is continuous.
open. )
n
A manifold is aspace which "locally" looks like some R
87
Lx 0 exp: B (0, r) -7 G
is not closed. Because 0 E Mn(k) - GL(n,k) but every ball with cen-
(xB,yB) # (x,y) .
Mn(k) such that for B' E B(B,s) we have (xB',yB') I (x,y) Thus
N c M. If M is connected, then N =M .
p [0,1] -->M
-1
with p (0) = x and p (1) = y Then o (M - N) is an open set in
jective.
is an orthonormal
SO(3) .
q.e.d.
for all n
F. Exercises
n
3. Show that for Ac R and f: A ~ Rm then f is continuous
Rn
4. Show that i f A,B are connected sets in and A n B '" r/l ,
then A UB is connected.
that
-1
S U xHx
XEG
is connected.
is continuous) .
n
8. Let A C Rn and XE" We say that x is a limit point
of A (x 1p A) if every
B(x,r) n A
( x 1p C ~ X E C ) .
91
(n + m) x (n + m) matrices by
(g,h) I 0)
n[ ( g
~ 1m
n m
this.
Let Then
93
= SI x SI is called a 2-torus.
groups.
k x k matrices.
y(s)v(t)o(s)o(t)
v(s)o(s) y(t)o(t)
(yo)(s)(yo)(t) .
(TG)e to G at e.
H = G •
H .
Thus 2 3
W UUU uU U ... cH.
for some
so H =G .
torus.
Proposition 3:
e
I
fI(
~)J
e
eos sin
T = 1-Si~ e
cos e
o
T:,t:T'cSO(3) .
prove that:
back to our proof that Center 50(3) {I) in Chapter V, and you will
Proposition 4:
J}
81 sin 81 0 0
{( co, \)1 cos 81 0 0
elements of T then rb E T
pairs
(rotation of V , rotation of W) •
Claim:
tity on W. Then
~(el) = ae l + be 2 + ce 3 + de 4
~rb(el) = ae l + be 2 + c / e 3 + d / e 4
97
above proofs.
Proposition 5:
n of the form
Cl. = E T •
such Cl.'S can move any vector which is not a multiple of el . Hence
98
1, ... ,n .
Thus
and T is maximal.
proposition 6:
T o
Proof:
i el
A matrix
ie
n
0- of the form
use
99
Now for n > 2 the same proof as used for U(n) will work, but
for n = 2
T - (,:, e~i')
and we do not have matrices er _
,-
(1 0)0 e- 19
to use. But for n =2
(a b (i 0)
rb d) E SU(2) and er. E T
c 0 -i
then
Thus b 0 = c and ~ E T .
Proof: Just as for U(n) we can show that any element of Sp(n)
100
iS10
which conunutes with all
(
o
e . . i
e Sn
) must be diagonal with the
C. Centers again
Now that we know maximal tori in our matrix groups, we are able
Proof: We have seen that if any element conunutes with all ele-
Center c T .
conunute with the matrix j1, it follows that the diagonal elements
must be real (and since they are of unit length) they are ±l 1t
is an exercise to show that they all have the same sign. Now I and
(: 0... 0)
1
0 0 ... 0
A = 0 1. 0
o '.1
For SU(n) we note that the same argument will show that an
ie ine
det(e I) e
q.e.d.
a diagonal matrix.
Co,
)
91 sin 91 0
A -Si~ 9 1 cos 9 1 0
0 *
(J 0 ~ 0)
Let
0
1
0 01 SO(k)
P E
. 0
o.
1
0 ') 0)
°
-1_1 -1
F or examp 1e , ( is in the center of SO(4) , but ( -1
O -1
-1 -1
is not even in SO(3) •
103
groups.
2
U(n) n
SU(n)
2n(2n-l)
SO(2n)
2 (I,-I}
Sp(n) (I,-I}
D. Exercises
q is a complex number.
explicitely, let
we form
n
K is a subgroup of II
105
ments of H. ete.)
A. General remarks
-1
G xTx
-1
Y E xTx
be a linear map.
107
3
Examples: Let V Rand W be the 2-plane spanned by el
and If
~)
sin
( cos 8
o= -Si; 8 cos
0
then W is o- stable.
2
Let V = R and 0 = (0
1
1)
0
Then W [(x,x)} is o- stable
0(rv) A(rv)
(We will have to be a little more careful when using our "skew" field
VO.) (v c V I tb(v)
Thus V(A) includes the zero vector 0 and all eigenvectors having
V(O) = (v c V I mev) = O}
is just the kernel (or null space) of the linear map tb.
Then linear maps tb: V -> V correspond 1-1 with elements of Mn(k) ,
Proof: =>
is unitary.
(xA,yA) (X,y)
orthogonal complement
is also A - stable.
t- -1
(xA,y) (x,y A) = (x,yA )
-1
Now A is an isomorphism of W onto W so yA E W Thus
cover U(n) .
e. to v.
J J
BAB- 1
{~'..~)
Since BAB- 1 is unitary, each L e is a unit vector; 1. e. ,
J j
I Aj I = 1 Thus BAB -1 E T
q.e.d.
T = {( ~1 ~8nl Z8 j o}
cover SU(n) .
111
det(B/A(B/)-l) = det A = 1
1
If we choose U E C such that det B' and set
n
u
BuB',
-1
U xTx
XEG
-1
(!j(n)" U xTx
XE (!) (n)
But we will prove (t) for SO(n) and this will, incidentally,
prove that
SO(n) is connected
112
n n
Definition: A linear map 5 ~ ~ ~ is symmetrie if
n
(x5,Y) (x,y5) for a11 x,y € R
have
t t
(x5,y) = (xA,y) + (x A,y) = (x,y A) + (x,yA) (x,5y)
t(x) = ~
(x,X)
We have
(v+ty,vSHyS) 0. (t)
f (t)
(v+ty,v+ty) a (t)
f '(0)
~/(O)a(O) - a'(O)~(O)
(a (0)) 2
«vS - (V,vS)v),y) 0
vS (v,vS)v q.e.d.
for SO(n) .
114
subspaee W of Rn with
dim W E (1,2)
S. Consider wand wA
-1
wA + wA or ~wA .
/lW + BwA E W .
Then
2
(/lW + BwA)A /lwA + BwA
n.wA + B üwA - w)
(-B)w + (~+B~)wA E W
A E SO(n) as fo110ws.
v l ,v 2 '··· ,vk',vk+l.:.,vk+2:
Wn _k
-2-
CAC- l E T .
D =(~:,0)
o '1
D. ( t) for Sp (n)
2
Proof: Given q E JH we must find h E JH such that h = q
2
h =q gives
2 2
8 - y
y 2~y = z 2".~ w
n
proposition 8: For A E Sp(n) there exists a nonzero v E JH
such that
W Span(v,vA) is A stable.
t (x) ~
(x,x)
117
V E
f(t) ~(v+ty),tER,
or
-1
o «vA - (vA,v»)v,y) + (y,(vA - (vA,v)v) .
-1
(vA + vA 2(vA,v)v,y) o
-1
vA + vA = 2(vA,v)v
or
(*) 2(vA,v)vA .
A - stable.
We have
2
a.vA + S(vA )
2
a.()., - 2(vA,v) + 1) O.
Since ]H has square roots, ",e can find )., such that
2
)., - 2(vA,v) + 1 = 0 q.e.d.
BAB- 1
119
E Sp(n) , 1 ,
we get
-1
QBA(QB)
E. Re1fections in Rn (again)
through 0 perpendicu1ar to x.
the formula
~(y) y - 2~x
(x,x')
Proof: ,,
,,
,
,
t-~-- p(y)
.k------'~ Y
x
120
r = ß.n So
(X,x)
p(y) ~ x
y - (x ,X)
and ~(y) - 2 ~ x
(X,X)
w' ,
to W and CL2' •.. '~n a basis fpr W. Then BAB- 1 is the identity
-1 -1
on CL 2B , •.. ,CLnB and
sin 8 1 sin 8
cos 8 ) , ... , B
sin 8
m) arranged a10ng
1 m
m
~1'··· '~m
Let ~i be ref1ection in the hyperplane formed by all coordinates
Let be ref1ection in
1
BAB- 1 E T
q.e.d.
F. Exercises
'::'*i)
cos !!.
4
1. Show that A (
,50(2) h., no A - ,"b1.
-S1n
•
'4'1i
2
subspace in ~ other than (o} Find all A - stab1e subspaces in
2 1)
R for A = (0
1
°
2. Find all eigenvectors and eigenva1ues of
(0 1) , (1 0) , (1 ~) , (0 1)
1
° °° ° ° 1
3. In llin sho~ that i f (x,y) has zero real part for every
y , then x = °
4. Let D Show that ~(n) - SQ(n) SQ(n)D .
2
6. Let Ll ,L 2 be two lines through 0 in R with angle e
between them. Let $1 be reflection in Ll and $2 be reflection
A = BCB- l
BCB- l C -1
Be B
0)
A
e e
o .0) (~
and with I Aj I 1 so eC
C = (" •
·"n An
e
Al An A +· .. +>.
Thus det e A = det (BeCB- l ) det e
C
e e e l n
= e TrC = e TrA .)
Chapter 9
Conjugacy of Maxi mal Tori
A. Monogenie groups
is dense in ,,2
wi th Y c ethen C = X .
2 3
f x ,x,x,···1
125
Note that the additive group F is not monogenie, nor are any of
x,2x,3x,4x, ...
will not form a dense set in SI because they will all lie in the
set
1 2 q-l q
[q,q' ... '-q-,q O}
proposition 1: SI is monogenie
such that
We proceed as folIows.
Then we can
q.e.d.
we get
-1
Y E xTx
-1
Now xTx is a group, so
y 2 3 -1
(y,y ,y , ... ) c xTx ,
have
T' n xTx -1 -1
T' or xTx ::> T'
-1
But xTx is a maximal torus in G so xTx -1 T'
of a maximal torus in G.
U(l) 1
SU(2) 3
SO(2) 1
SO(3) 3 {T}
Sp(l) 3 {T,-I}=71/2
Rank 2:
U(2) 4
SU(3) 8 'll/3
SO(4) 6 (I,-I}=z/2
SO(5) 10 (I}
Sp(2) 10 {I,-Il
Rank 3:
U(3) 9 Sl
SU(4) 15 71./4
SO(6) 15 (I,-Il=Z/2
SO(7) 21 {Tl
Sp(3) 21 (I,-Il=71./2
dirn U < dirn SU < dirn SO(even) < dirn SO(odd) = dirn Sp .
For rank 2 on1y 50(5) and Sp(2) have the same dimension
For rank 3 SU(4) and SO(6) have the same dimension and
and again these are not isomorphie with any of our other groups except
we might have
of our groups are simple, and, when they are not, what normal sub-
groups they have. We state the result, but a proof would take us
Theorem: ~ ~ SU(n)
center - center
SO(2n+l)
SO(2n)
center
~
center
(wa)(t) w(t)a(t) .
p(<U) w(l)
is a surjective homomorphism.
131
so p is a homomorphism.
in G.
connected.
groups are
r--!
SO(n)
r--J
We see that p SO(n) -'> SO(n) is a 2-1 homomorphism. We construct
E. Exercises
Let GO Jx G I :[ pa th w : [0, 1] G
l
E -'>
w(O) = 0 w(l) = x
group of G.
A. Clifford Algebras
exarople, we have used the algebra Mn(R) in which the group of units
algebra Ck .
2
and set el = -1 , = -1 and This specifies
1 f') 1
e l f') i
e 2 f') j
e l e 2 f')k
gives an isoroorphism
2
e. = -1 and e.e. -e e if i.f j
~ J ~ i j
(e. . .. e.
~l ~r
basis).
as a subalgebra of Ck '
divisors of zero.
with operations
r(x,y) (rx,ry)
as follows: ~ t C is represented by
(i ?) --> el
° ~
(j ?) --> e2
° J
(k 0) --> e3
°-k
(0 k) --> e4
k 0
gives isomorphisms
c ! Cl
E! C2
1HEllE 1 C3
M2 (lH) 1 C4
1
:
( ~ -~ -~) (~-~ -; -:)
0 0, 0 -~ 0 0 0
k
Let R denote the k-space in Ck spanned by e 1 , ... ,e k and
then
k-1 *
SeC k .!: .~. , each x E Sk-1 is a unit.
Proof: Let 1 .
Then
For example,
1* -e.
~
-eie j
Clearly **
(x) x. Also, we have (xy) * y *x * because
«e·l···e.
~ ~r
)(e·l···e.
J Js
)) * (-e. ) ... (-e·l)(-e. ) ... (-e. l )
Js J ~r ~
J Js ~
*
(e. l ·· .e. ) (e. l ... e. )
~r
*
= -e. e.) .)
~ J
k-l
Proposition 4: If u E S C Pin(k) , then 0 (u) is
k
reflection in R in the hyperplane perpendicular to u.
139
Consider
the p;,oposition.
k
It now fo110ws that for each u E Pin(k) , p(u) does map R into
k
Rand it is orthogonal since it is a product of ref1ections. In
that p is surjective.
-1
Definition: Spin(k) = p (SQ(k)),
by S
o= tel,-el} This is just tel,e l
2
= -1, e l
3
= -e l , e l
4
",.1)
Z
"''4'
Now p(e l ) = p(-e l ) is a reflection and
1 2 2
S = tael + be 2 I a + b = 1)
S l} ,
1
Pin(2) = S \J Sand Spin(2) S
Spin(3) l}
spin groups,
k ~ 3 ,
r-1
(e. 1 .,.e. )(e. e.) (-1) e. (e. 1 " oe. )e.
1 1r 1r J 1r 1 1r J
r-1 r
(-1) (-1) (e. e i .) (e 1" ,ei)
1r J i r
(-1) (2n-1)r 1
C. The isomorphisms
1, ... , k-l .
and
part of Ck
!/J(x* )
matrices
143
The others work the same way, and the verification for
MtM = I so that
1 = ~(M
L-_
M) = ~(M)$(M)
* and
Thus
SpineS) c ~~(Sp(2»
Spin(6)'c ~w(SU(4» ,
D. Exercises
1. Show that
2. Show that
3. Show that
A. Normalizers
N(S) = (x E G I XSx- l = S}
subgroup of N(S)
A : G -;. G
x
.
Our standard maximal torus in Sp(l) is T = (a+~b I a 2+b 2 I) .
Let N be the normalizer of T in Sp(l) If q E N then
qiq -i
2 2 2 2
x- + Y z w -1 ,
and since
2 2 2 2
x + Y + z + w 1
2 2
we get x = 0 y Furthermore, i f q jz +kw with z + w 1
147
-i Hence
r"
torus T' in SO(3) Reca11 that
m
sin
T' -s~n 8 cos e
0
V (cos
J)( c~s
cos 8
(c~s
Sl.n
r/J
r/J
sin r/J
-cos r/J -sin 8
sin 8
cos
0)(
o Sl.n
r/J
r/J
sin r/J
-cos r/J o sin 88 ~in
= cos 88 0)
0
0 0 0 0 1 0 0 - 0 0 1
I t is 1ess routine (but also instructive) to show that these are the
of T'
So we have:
148
proposition 1: N' ~ N .
and isomorphie quotient groups (namely Z/2), but N and N' are
~ being the inelusion and B being the quotient map. We say say
cross section.)
Then T' ~ N' ~ Z/2 splits (send identity to identity and send
a homomorphism).
groups.
proof: ~
-1 -1
x = Y'r for some 'r E T So Ax(t) = xtx -1 Y'rt(y'r)
-1
Y'rt'r Y
-1
and sinee T is abelian this is just yty A (t)
Y
-1 -1 -1 -1-1
If xtx yty for all t E T , then (y x)t(y x) = t so
-1 -1
that y x eommutes with all t E T aud T is maximal so y x E T .
150
Corollary: W acts on T .
-1
Proof: For w E W choose x E 8 (w) and define w(t)
GG T
G XT~
~yT
81 G G
W
.
id
151
Spin(2n+l) ~ ~
center - center
Spin(2n+l)
in splits, but the normalizer of a maximal torus in
center
Sp(n)
center does ~ split. So for n = 3,4,5, ...
Spin(2n+l) ~ Sp(n)
~ = (jl.~) ~
Let
o ·1
E Sp(n) . We claim that E N. Take any
152
t €
(1o 0) eien
in T and we ca1cu1ate
'I1 t 'l1
-1
o ("1
(lI.?) 0 )("
~""ei9n ""I
'1
0) ;"
iA 1
je (-j)
0
i8 2
e
0 i9 n
e
i8 1
. -i8 -1
and je (-j) j(cos 81 + i sin 81) (- J) = e 1 so that 'I1 t 'l1
.0)
is aga in in T
2 (-11 •
Let w = B (T'\) Then, since '11
o '1
€ T we have that
2 2
w = B(T'\ ) = 1 E W. On the other hand, since y is assumed to be a
'1')' y(w)
2 2
('rI' ) y(w ) y(l) I
l
T1 '
. ) ( i8 )
( J" " , ." .iB
n
o
,,'=
i8
n
e
Thus
. i81. i 8l
2
Je Je
2i8 2
o
(" ,) e
o e
2i8
n
-1
o
" I
o e
2i8
n
the general argument will be clear. The idea is like for the proof
to a relation in N
o
~) (~ ~)
1
Let 1 and cr o and one checks easily
o o
that ~,cr are in N Let
2 2
Also since ~ and cr are in T we have
2 2
wl 1 and W z 1
Also
(~ ~)
j
~cr 0
0
in T So we have
(y(w l ))
zE [I,-I} (= center)
Z
(y(w Z)) E (I,-I}
4
(y(wl)y(w Z)) E [I,-I}
So we will assume that we have ~' in the coset of '1"\, cr' in the
ISS
(" '",
Proof: We have
Je 0
)A,)
i 82
~ , ~t 0 e
0 0
-I and Tl'
~:'
(_l)m/dJ
a' 0
0
i~)
Proof: We have
a' at
a' o
o
156
2
Then (0' )
we can take °
= 1/)1 and e
i0 2 m il/)
(-1) e ,and Lemma b is proved.
We finish (t) by
4
Lemma c: (Ti'a' )
Tl'a' o
o
D. SO(n) sp1its
We showed that the normalizer in SO(3) sp1it, but that was easy
because we knew the Wey1 group was just Z/2, so we simp1y had to
Let
cos 8 1 sin 9 1 0 0 0
91 cos 9 1 0 0 0
0 0 -sin 8 2 cos 8 2 0
0 0 0 0 0
PA P and
so that Then
-1
p'TA R or P'T PA p' ,
have
158
for P2 .
(~ j) ~)
0 0 0 0 0 1
1 0 0 0 1 0
A -
0
0
0
0
1
0
1
0
0
B =
(1 1
0
0
0
0
0
0
0
0
Also B B- l and
"
"""
"
""
"
", BA
",
----------~--------------~~----_7 91
""
""
""- "-
" SB = w2
X
159
We11,
(~
0 0 1
J)
0 1 0
AB 0 0 0
1 0 0
0 0 0
Then
~)
1 0 0
(l
Z 0 0 0
(AB) = 0 0 1
0 1 0
0 0 0
and so
~)
0 0 0
(i
1 0 0
(AB) 4 = 0 1 0 I
0 0 1 ,
0 0 0
,(l ~)
1 0
0 0 c- 1
so (5) . Let C
0 0
Then = C and
0 1
~)
0 0
0 1
(1
Let B Then B E SO(4) , B- 1 = B and
1 0
0 0
or of the form
~~0
~
one changing one sign but not permuting. Such a matrix cannot have
as required.
E. Exercises
sin
21/J)
-cos 21/J
r
( cos e
-sin e
sin
cos :) r
( c~s
s~n
e
e
-sin
cos :)
S. Show that the on1y isomorphism of s1 = [eie} which
A. Differentiable manifolds
!Wo charts dJ: U -> M and 1iJ: V -> Mare said to overlap smoothly
164
-1
i f (either ~(U) n ~(V) ~, or) 1\1 0 t1J is a diffeomorphism.
charts which over1ap smooth1y with the given ones in the atlas. Since
If two new charts over1ap smooth1y with the given charts in the atlas,
a differentiab1e structure.
is not a diffeomorphism.
n
If we have a smooth curve or surface in some ~ then the concept
165
own right - not necessarily sitting in sorne euclidean space. How then
We used A(p) to denote this set of pairs (U,f) since we are going
that these operations make A(p) into areal vector space. Then we
define
two functions f,g such that f(p) = 0 g(p) , then for any tangent
Any chart tJ;: U ..,. M (with P E tJ; (U)) in the differentiable struc-
~i (p) (f)
-1
be the i th partial derivative of f 0 rb at tJ; (p) Thus
~(f) + 'lief)
al(p) , ..• , !n(p) in TM We will show that these are a basis for
P
TM
p
We may assume ~(O) = P Then for any f E A(p) we can write
n
f(x) f(p) + ~
i=l
o in R. We note that
f(x)-f(O) _ f'(O)
x
Hx)
x
2
f(x) f(O) + f'(O)x + $(x)x
q.e.d.
Xf : W -7 R
we have Xf smooth.
then so is X + Y ,
(rX(f) = r(X(f») . Thus the smooth veetor fields form a veetor spaee.
have
smooth function Xi
If we have two smooth vector fields X and Y we can get a real-
f ~ X (Yf)
P
f ~ X (Yf) - Y (Xf)
P P
We assert that
since X satisfies (ii), and this is just the right hand side
p
evaluated at p Of course, we have a similar formula for Y(fg)
Then
and similarly
Thus
(i) are true since they are true for both terms. So
XY - YX is a tangent vector at p
p p
[X,Y] = XY - YX
P P P
171
[ ,
~(f 0 1\1)
A(q) to Rand assigns the same real number to two functions which
= f(q)d1\l~(g) + d.JJ~(f)g(q)
c. Lie groups
the operations
G ~ G
-1
(a,b) I-i> ab a I-i> a
dL gXe
173
(dL : T G -'> T G). Such vector fields are called left-invariant; Le.
g e g
a vector field X on G is left-invariant if it satisfies (t) .
so is [X,Y] .
= X (Y (f 0 L )) - Y (X (f 0 L ))
e g e g
Xg (Yf) - Yg (Xf)
Now one can see easily that X,Y left invariant implies X+Y
Examples:
morphisms - functions
morphisms - homomorphisms
ordinary composition
ordinary composition
1\ 0 y
diffeomorphism.
F : Cl -> C2
we have either
A ----=. F (A)
Cl1 1 F(rJ)
s1 1 F(S)
C -F(C)
or
8i r F(S)
C ------,. F ( C)
it is a contravariant functor.
algebra homomorphisms.
~(G) and t(H) we have a linear map of the Lie algebras. We must
we calculate
algebra is the Lie algebra of some Lie group. Indeed there is the
following theorem (which we are not prepared to prove here - see page
algebra t(G). The catch is that H may not be a Lie group at all.
If, on the other hand, ~TT is rational then H will be a circle sub-
D. Connected groups
be a subset of X
Y = X •
178
x E Y
x E Y •
then H = G .
Proof: Let
U2 {XY!XEU,YEU}
u
m
but then
e in G, then f is surject1ve.
connected subgroup of G
-1
,. (t) (p (t»
-1
is a path from e to x
that U c GO .
see that x E GO
We are going to need an exponential ~ from T G to G and
e
we cannot use the one from Chapter IV because the elements of T G
e
and G cannot be assumed to be matrices. If y: (-€,€) ~ G is a
at e to be
one shows that for matrix groups this agrees with our definition in
Chapter II) .
exp(,,) y(l)
o (t) exp(t,,)
To see this hold t fixed and consider the curve cr(u) = y(tu)
have proved
inverse). This is still true in our more general setting and is based
on the
of p .
books).
E. Abelian groups
n
Lemma: If CeR is bounded and infinite then C has a limit
point.
infinite set.
each edge into two equal parts and cut Kl into zn equal cubes of
side length ~ Call those cubes Kll,KlZ, ... ,KlZn At least one
Then
that Y = tY l 'Y2' ... } has no limit point and that X = tXl'x2, ... }
Yl = f(x l ) E B(y,l) .
Abe1ian group 4
Algebra 15
Algebra of matrices 15
Atlas 163
Binary operation 1
Bounded set 81
Category 173
Center of a group 20
C10sed manifo1d 87
C10sed set 79
Commutator 69
Commutator subgroup 69
Compact set 81
Comp1ex numbers 9
Conjugation in ~, C, E 23
187
Conjugate of a matrix 24
Connected set 79
Continuity of a function 76
Cosets of a subgroup 67
Countability 83
Diffeoroorphism 163
Discrete subgroup 94
Exponentia1 of a matrix 45
Fie1d 7
Funetor 174
Group 2
Indempotent matrix 33
Injeetive homomorphism 5
Inner produet 23
Isomorphism (linea r) 14
Isomorphism of groups 6
Jaeobi identity 57
Kerne1 of a homomorphism 19
Length of a veetor 24
Linear map 12
Manifold 87
Maximal torus 95
Maximal torus in SO(n) 97
Maximal tori in U(n) and SU(n) 97, 98
Maximal torus in Sp(n) 99
Metrie 74
Normal subgroup 20
Normalizer ( of a set in a group ) 20
Normalizers (of max. tori) in Sp(l) and SO(3) 147
One-parameter subgroup 51
Open ball 75
Open set 78
Orthogonal groups 27
Path 80
Path in a group 130
Pin(k) 137
Primitive root of unity 22
Projeetion 33
190
Quaternions 1H 11
Quotient group 68
Rotation group 27
Schwarz inequality 75
Skew-Hermetian matrix 40
Skew-Symmetric matrix 39
Skew-symplectic matrix 40
Smooth homomorphism 41
Split group extension 148
Spin(k) 140
Surjective homomorphism 5
Symmetric group 4
Torus 93
Trace of a matrix 54
Transpose of a matrix 24
Triangle inequality 74
Unipotent matrix 34
Unitary group 27