Introduction To Quantum Groups

a
r
X
i
v
:
q
-
a
l
g
/
9
7
0
4
0
0
2
v
2

1
7

J
u
n

1
9
9
7
Introduction to Quantum Groups
P. Podles
and E. M uller
Department of Mathematical Methods in Physics, Faculty of Physics, University of War-
saw, Ho za 74, 00682 Warszawa, Poland. E-mail: podles@fuw.edu.pl
Graduiertenkolleg Mathematik im Bereich ihrer Wechselwirkung mit der Physik, De-
partment of Mathematics, Munich University, Theresienstrae 39, 80333 M unchen, Ger-
many. E-mail: emueller@rz.mathematik.uni-muenchen.de
Abstract
We give an elementary introduction to the theory of algebraic and topological quantum
groups (in the spirit of S. L. Woronowicz). In particular, we recall the basic facts from Hopf
(-) algebra theory, theory of compact (matrix) quantum groups and the theory of their
actions on compact quantum spaces. We also provide the most important examples, in-
cluding the classication of quantum SL(2)-groups, their real forms and quantum spheres.
We also consider quantum SL
q
(N)-groups and quantum Lorentz groups.
Contents
1. Introduction and physical motivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Polynomials on classical groups of matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3. Examples of quantum groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4. -Structures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5. Compact Hopf -algebras. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6. Actions on Quantum Spaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
7. Quantum Lorentz groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
8. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Supported by Graduiertenkolleg Mathematik im Bereich ihrer Wechselwirkung mit der

Physik, Dept. of Mathematics, Munich University and by Polish KBN grant No. 2 P301
020 07
1. Introduction and physical motivations
What are quantum groups?
Let G be a group in the usual sense, i. e. a set satisfying the group axioms, and k be a
eld. With this group one can associate a commutative, associative k-algebra of functions
from G to k with pointwise algebra structure, i. e. for any two elements f and f
, for any
scalar k, and g G we have
(f +f
)(g) := f(g) +f
(g), (f)(g) := f(g), (ff
)(g) := f(g)f
(g)
If G is a topological group, usually only continuous functions are considered, and for an
algebraic group the functions are normally polynomial functions. These algebras are called
algebras of functions on G. These algebras inherit some extra structures and axioms for
those structures from the group structure and its axioms on G. Locally compact groups
can be reconstructed from this algebra.
Now the algebra is deformed or quantized, i. e. the algebra structure is changed so that
the algebra is not commutative any more, but the extra structures and axioms for them
remain the same. This algebra is called algebra of functions on a quantum group, where
quantum group is just an abstract object described by the deformed algebra. This
process can be summarized as follows:
classical group G
axioms of a group
quantum group
(abstract object)

commutative algebra of
functions on G with
corresponding extra axioms
forget about
group
non-commutative algebra with

same extra axioms; algebra of
functions on a quantum group
There is a similar concept of quantum spaces: If G acts on a set X (e. g. a vector space),
there is a corresponding so-called coaction of the commutative algebra of functions on G
on the commutative algebra of functions on X satisfying certain axioms. The latter algebra
can often be deformed/quantized into a non-commutative algebra, called the algebra of
functions on a quantum space with a similar coaction. There are three ways of considering
algebras of functions on a group and their deformations:
(a) polynomial functions Poly(G) (developed by Woronowicz and Drinfeld),
(b) continuous functions C(G), if G is a topological group (developed by Woronowicz),
(c) formal power series (developed by Drinfeld).
Only the rst two approaches will be dealt with in the sequel. They include representation
theory, Peter-Weyl theory, Tannaka-Krein theory, and actions on quantum spaces.
Chapter 1. Introduction and physical motivations 3
There is a second approach to quantum groups. If G is a connected, simply connected
Lie group, G can be reconstructed from the universal enveloping algebra U(g) of the
corresponding Lie algebra g. The algebra U(g) again inherits some extra structures and
axioms and can be deformed. The deformed universal enveloping algebra can be regarded
as universal enveloping algebra corresponding to a quantum group. One can consider
(d) the quantized universal enveloping algebra U
q
(g) (developed by Jimbo),
(e) formal power series (to be more precise, the ring of formal power series in h over
a free algebra, subject to certain relations which are the same as for U(g) in the
case h = 0. From this ring the algebra U
q
(g) can be extracted. This approach has
been developed by Drinfeld).
This approach will not be used in the sequel.
Physical motivations
There are some physical motivations for quantum groups including
1. integrable modelshandled with approach (e),
2. conformal eld theoryhandled with approach (e),
3. physical models based on quantized space-timehandled with approaches (a), (b),
and (e).
The last motivation shall be explained in more detail. One of the main problems in
Quantum Field Theory (QFT) is to join QFT and General Relativity Theory in a consistent
way. It seems that in such a new theory it would be impossible to study the geometry of the
space when very small volumes are considered. If you consider a cube in space, each vertex
of it having Plancks length or less, and measure simultaneously the three coordinates x,
y, and z of a particle in it, then the uncertainty of the measurement, i. e. the errors x,
y, and z are very small, whence by Heisenbergs uncertainty relation the errors of the
coordinates of the momentum are big and therefore the uncertainty of the energy E is
big, too. Since the energy is positive, the expected value E) of the energy is big, and the
smaller the cube the bigger the energy, which at a certain stage generates a black hole.
Therefore the observation of the geometry of the space gives it a dierent geometry, which
makes this observation useless (We have used here the arguments by Professor W. Nahm).
Quantum mechanics says that physical quantities such as momentum and position, which
can be measured, correspond to self-adjoint operators on a Hilbert space. Its elements de-
scribe possible states of a physical system. When a quantity is measured, the state is
projected onto an eigenvector of the operator, and the result of the measurement is the
corresponding eigenvalue. Two quantities can be measured simultaneously if and only if
the corresponding operators commute. In usual quantum mechanics the operators corre-
sponding to the three coordinates of space commute and can be measured simultaneously,
which leads to the problem with the black hole. Thus it is reasonable to assume that the
operators corresponding to the coordinates x, y, and z do not commute (whence they
cannot be measured simultaneously). Hence the commutative algebra generated by the
operators corresponding to x, y, and z, which is isomorphic to the algebra of polynomials
on R
3
, is replaced by a non-commutative algebra on a quantum space. In order to give
sense to self-adjoint operators, this algebra should be a -algebra.
1.1. Denition: (a) A -algebra is a C-algebra A equipped with an antilinear, antimulti-
plicative involution : A A, i. e. for all a, b A and C the following holds:
(a +b)
= a
+b
, (a)
=

a
, (ab)
= b
, (a
= a.
(b) Let A, B be -algebras. An algebra homomorphism : A B is called -homo-
morphism, if (a
) = (a)
for all a A.
Physical experiments should be comparable and reproducible, i. e. the same experiment
performed at dierent places and times ought to give the same result. Therefore the theory
should be invariant with respect to certain symmetry groups (containing translations in
time and space). But the classical (symmetry) groups do not t well to quantum spaces,
so they have to be changed to quantum groups, too. (Example: The group SO
3
(R) of
rotations in three-dimensional space acts on the sphere S
2
. When the algebra of functions
on S
2
is properly deformed such that the algebra becomes non-commutative, then there
is no reasonable coaction of the usual algebra of functions on SO
3
(R) any more. [P1,
Remark 2])
There is another motivationdeformation of an existing physical theory may help to
understand the theory in a better way. It can reveal why the theory works, what is a
consequence, and what is just a coincidence.
Example [P4]: After looking at deformations of standard Dirac theory, the covariance
of the Dirac equation can be seen more directlyon the level of groups rather than Lie
algebras. For the wave vector there is the equation

=
0
, where
0
also appears in
the Dirac equation. In the deformed theory there is

=
A with A ,=
0
in general, so
that A =
0
is just a coincidence, and the condition A =
0
is not really important for the
theory.
In physics all symmetry groups are groups of matrices or can be described with groups
of matrices, therefore the case of matrix groups is considered.
Acknowledgment
These lecture notes were written down by E. M. after the lectures by P. P. given at
the Department of Mathematics, Munich University. The rst author is very grateful to
Professor Dr. Hans-J urgen Schneider for his warm hospitality at Munich University. We
thank him for his useful remarks. An earlier version of these lectures was given by P. P.
at Kyoto University in 199091. The rst author would like to express his gratitude to
Professor Huzihiro Araki for his kind hospitality and encouragement.
2. Polynomials on classical groups of matrices
Notations
In the sequel the base eld of all vector spaces and algebras is the eld C of complex
numbers. A unital algebra is an (associative) algebra with a unit element, and a unital
mapping is a mapping between unital algebras which sends the unit element to the unit
element.
Let N, N
0
and R denote the sets of positive integers, non-negative integers and real
numbers respectively and x M, N N. Let A be a unital algebra and let M
MN
(A) be
the vector space of M N-matrices with entries in A. If M = N, M
N
(A) := M
NN
(A) is
a unital algebra. For each matrix M M
MN
(A) let M
ij
be the entry at the i-th row and
j-th column of M. Let B be another algebra, : A B a map and M M
MN
(A). Then
(M) is shorthand for the matrix in M
MN
(B) with entries (M
ij
). The group GL(N, C)
of invertible N N-matrices with complex entries is equipped with a topology inherited
from the norm topology of the vector space M
N
(C)

= C
N
2
. The neutral element of a group
is denoted by e.
Let C
N
denote the space of row vectors and
N
C the space of column vectors. Using
matrix multiplication, C
N
can be regarded as dual space of
N
C. If e
1
, . . . , e
N
is a basis
of
N
C, then there is a dual basis e
1
, . . . , e
N
of C
N
such that e
i
e
j
=
ij
for all i, j N.
In a similar way there are dual bases of the k-fold tensor products (
N
C)
k
and (C
N
)
k
.
In the sequel the indices i, j, i
, j
, k denote positive integers less or equal to N.

Let 1
N
denote the identity matrix with N rows and columns or the identity endomor-
phism of C
N
or
N
C.
Functions on groups
Let G be an arbitrary subgroup of the group GL(N, C). Let Fun(G) be the algebra of
complex valued functions on G. This algebra is unital with unit element 1: G C, g 1
and is a -algebra, where for all f Fun(G) the function f
is dened by f
(g) := f(g)
for all g G.
For all i and j, the coecient functions
u
ij
: G C, g g
ij
and u
1
ij
: G C, g (g
1
)
ij
belong to Fun(G). Then the matrices u := (u
ij
)
1i,jN
and u
1
:= (u
1
ij
)
1i,jN
belong to
M
N
(Fun(G)) and are inverses of each other in M
N
(Fun(G)). This justies the notation u
1
.
2.1. Denition: Let Pol(G) be the subalgebra of Fun(G) generated by the elements u
ij
and u
1
ij
for all i and j.
Chapter 2. Polynomials on classical groups of matrices 7
Remark: This algebra is automatically unital because of the relation 1 =
n
k=1
u
1k
u
1
k1
.
The algebra is called algebra of holomorphic polynomials on G, too.
2.2. Lemma: If G SL(N, C) then Pol(G) is already generated by the elements u
ij
.
Proof: By the usual formula for the inverse of a matrix, (g
1
)
ij
= (1)
i+j
det g
j,i
/ det(g)
for all g G, where the (N 1) (N 1)-matrix g
j,i
is obtained from g by deleting the
j-th row and the i-th column. But det(g) = 1, whence also u
1
ij
is a polynomial in the
functions u
i
j
.
2.3. Denition: Let Poly(G) be the -subalgebra of Fun(G) generated by the elements
u
ij
and u
1
ij
.
Usually the algebra Poly(G) is considerably bigger than Pol(G).
2.4. Lemma: If G is a compact subgroup of GL(N, C), then Poly(G) is generated by the
elements u
ij
as -subalgebra.
Proof: The map : G R
+
, g [ det(g)[ is a group homomorphism from G into the
multiplicative group of positive real numbers. Since is continuous and G is compact, the
image of is a compact subgroup of R
+
. But 1 is the only compact subgroup of R
+
,
whence (g) = 1 for all g G. Therefore
1 = det(g)det(g) = det(g) det(( g
ij
)
1i,jN
).
Thus det(u) is invertible in Poly(G) with inverse det((u
ij
)
1i,jN
), whence the elements
u
1
ij
can be expressed by the u
i
j
and u
j
.
2.5. Remark: Let I be an index set and let G be a subgroup of
I
GL(N
, C). Each
element g of this group can be written as g = (g
)
I
with g
GL(N
, C) for all I
and dene
u
ij
, (u
)
1
ij
: G C, u
ij
(g) := (g
)
ij
, (u
)
1
ij
:= (g
1
)
ij
for all g G. The algebras Pol(G) and Poly(G) are generated by the elements u
ij
and
(u
)
1
ij
as algebras or -algebras, respectively. This generalization covers all compact
groups G, because the group homomorphism
G
G
GL(dim(), C), g ((g))
G
where

G is the set of nite dimensional irreducible representations of G, is injective if G
is compact (cf. Tannaka-Krein duality).
The multiplication, unit, and the inverse on G lead to the following extra structures on
Fun(G):
: Fun(G) Fun(GG), (f)(g, h) := f(gh) for all g, h G (Comultiplication),
: Fun(G) C, (f) := f(e) (Counit),
S: Fun(G) Fun(G), (Sf)(g) := f(g
1
) for all g G (Antipode).
These maps are unital -homomorphisms. The (algebraic) tensor product Fun(G)Fun(G)
is the vector subspace of Fun(G G) generated by elements u v, where u, v Fun(G),
by dening (u v)(g, h) := u(g)v(h) for all g, h G. Equality only holds if G is nite.
The axioms for the group structure on G are reected by certain axioms for the extra
structures on Fun(G). Let f be an element of Fun(G) such that (f) Fun(G) Fun(G).
Since the multiplication in G is associative, we have
(id)(f) = (id )(f). (1)
The property of the neutral element, namely ge = eg = g for all g G, leads to the
equation
( id)(f) = (id )(f) = f. (2)
(Here the usual identication CV

= V C
= V for all C-vector spaces is used). Let the

linear map : Fun(G) Fun(G) Fun(G), f f
ff
be induced by the multiplication

in Fun(G). Then the properties gg
1
= g
1
g = e of the inverse can be expressed as
(S id)(f) = (id S)(f) = (f)1. (3)
2.6. Denition: A unital algebra H is called Hopf algebra, if there are unital algebra
homomorphisms : H H H and : H C and a linear map S: H H satisfying
axioms (1)(3) for all f H.
The following lemma gives examples of Hopf algebras and shows why Pol(G) and the
elements u
1
ij
are interesting.
2.7. Lemma: Pol(G) is a Hopf algebra satisfying
u
ij
=
N
k=1
u
ik
u
kj
, u
1
ij
=
N
k=1
u
1
kj
u
1
ik
,
(u
ij
) = (u
1
ij
) =
i,j
, S(u
ij
) = u
1
ij
, S(u
1
ij
) = u
ij
.
If G is nite, also Fun(G) is a Hopf algebra.
Proof: For all g, h G,
u
ij
(g, h) = u
ij
(gh) = (gh)
ij
=
N
k=1
g
ik
h
kj
=
N
k=1
u
ik
(g)u
kj
(h) =
N
k=1
(u
ik
u
kj
)(g, h).
A similar computation yields the formula for (u
1
ij
). Therefore the image of Pol(G) under
is contained in Pol(G) Pol(G). The values of the counit can be computed: (u
ij
) =
(u
1
ij
) = e
ij
=
i,j
. The equations for the antipode follow from (S(u
ij
))(g) = u
ij
(g
1
) =
u
1
ij
(g) for all g G. The Hopf algebra axioms are clearly satised, because Pol(G) is a
subalgebra of Fun(G).
If G is nite, then Fun(G) is a Hopf algebra because Fun(G) Fun(G) = Fun(GG).
The following general theorem for Hopf algebras can be inferred from [A].
2.8. Theorem: Let H be a Hopf algebra with unit element 1.
(a) The maps and S are unique if is xed.
(b) S is a unital antihomomorphism
(c) If : H H H H, x y y x denotes the ip automorphism, then
S = (S S), S = .
(d) Let S
: H H be a C-linear map. Then the following are equivalent:

(i) (id S
)(f) = (S
id)(f) = (f)1 for all f H,

(ii) S S
= S
S = id.
2.9. Remark: (a) In general, the antipode of a Hopf algebra is not invertible.
(b) A map S
such as in part (d) of Theorem 2.8 is called skew antipode, and there is an-
other Hopf algebra structure on H with comultiplication , counit and antipode S
.
(c) A motivation for the fact, that the counit, but not the antipode is an algebra homo-
morphism, if H is not commutative: Since and the identity are algebra homomor-
phisms, there is no reason following from axiom (2) that should not be an algebra
homomorphism. But the map in axiom (3) is an algebra homomorphism if and
only if H is commutative. Therefore it should not be expected that S is an algebra
homomorphism.
For all f Fun(G) satisfying (f) Fun(G) Fun(G), the following equation holds:
(f
)(x, y) = f
(xy) = f(xy) = f(x, y) =
f
1
(x)f
2
(y) =
=
1
(x)f
2
(y) = ( )(f)(x, y)
for all x, y G. This motivates the following denition.
2.10. Denition: A unital algebra H is called a Hopf -algebra, if H is both a Hopf
algebra and a -algebra such that (f
) = ( )f for all f H.
From the denitions and Lemma 2.7 follows immediately
2.11. Lemma: Poly(G) is a Hopf -algebra, and if G is nite, also Fun(G) is a Hopf
-algebra.
2.12. Proposition: Let H be a Hopf -algebra. Then
(a) For all x H, (x
) = (x), i. e. is a -homomorphism.
(b) S S = id, in particular, S is bijective.
Proof: (a) Since the map H C, x (x
) satises the properties of the counit, both

are equal by Theorem 2.8, part (a), whence the assertion follows.
(b) The map S satises all properties of the skew antipode. By Theorem 2.8, part (d)
it is equal to it. This implies the two equivalent equalities SS = id
H
= SS.
Elements of representation theory
Let H be a Hopf algebra.
2.13. Denition: Let k be a positive integer. A matrix v M
k
(H) is called corepresen-
tation, if the entries satisfy the following relations for all indices a and b.
(a) v
ab
=
k
c=1
v
ac
v
cb
,
(b) (v
ab
) =
a,b
The number dimv := k is called the degree of the corepresentation, and the elements v
ab
are called the matrix elements of the corepresentation.
2.14. Remark: (a) Let v be a corepresentation of a Hopf algebra H. Then S(v
ab
) = (v
1
)
ab
for all indices a, b. Thus Condition (b) of Denition 2.13 can be equivalently replaced
by invertibility of v (note that Condition (a) implies (v)v = v).
(b) Let G be a classical group of matrices and H one of the Hopf algebras Pol(G) or
Poly(G). Let
v: G M
k
(C), g (v
ab
(g))
1a,bk
be a map such that all functions v
ab
are contained in H. Then (v
ab
)
1a,bk
is a
corepresentation if and only if v is a representation of G.
Proof: (a) This follows from the axioms for the antipode of a Hopf algebra.
(b) For all x, y G the following equations hold.
(v
ab
)(x, y) = v
ab
(xy) = (v(xy))
ab
,
(
k
c=1
v
ac
v
cb
)(x, y) =
k
c=1
v
ac
(x)v
cb
(y) = (v(x)v(y))
ab
.
Therefore condition (a) in Denition 2.13 is equivalent to v(xy) = v(x)v(y). A com-
putation of (v
ab
) shows that condition (b) is equivalent to v(e) = 1
k
.
Now x a Hopf algebra H.
2.15. Denition: Let v and w be two corepresentations of H.
(a) Then v w and v w are corepresentations of H, where v w is a matrix with
dim(v) + dim(w) rows and columns given by
_
v 0
0 w
_
,
and the matrix of v w has dim(v) dim(w) rows and columns and entries given by
(v w)
ij,kl
:= v
ik
w
jl
, where the indices i, k take values between 1 and dim(v) and
the indices j, l between 1 and dim(w).
(b) A dim(w)dim(v) matrix A over C intertwines v with w, if Av = wA. Dene Mor(v, w)
as vector space of intertwining matrices between v and w. The elements of Mor(v, w)
can be regarded as C-linear maps from C
dimv
to C
dimw
. The corepresentations v and
w are said to be equivalent (v

= w) if dim(v) = dim(w) and there is an invertible
element in Mor(v, w).
2.16. Denition-Lemma: Let w be a corepresentation of dimension N and V
N
C a
subspace of dimension l. Then the following are equivalent:
(a) For each Hom(H, C) the statement (w)V V holds.
(b) There is a corepresentation v and a basis a
1
, . . . , a
l
of V such that for the Nl-matrix
A
l
:= (a
1
a
l
) the equation wA
l
= A
l
v holds. This is equivalent to the condition
that A
l
is an injective intertwiner of v with w.
(c) There is a corepresentation v of dimension l and an invertible matrix A, the rst l
columns of which are a basis of V and such that
wA = A
_
v
0
_
.
If one of the equivalent conditions holds, then V is called w-invariant subspace, and the
corepresentation v in part (b) and (c) is called subcorepresentation of w and we write
v = w[
V
(Note that v depends on the chosen basis of V ).
Proof: (a) ((b) (c)). Let a
1
, . . . , a
l
be a basis of V and extend it to a basis a
1
, . . . , a
N
of
N
C. Then let A
l
be the N l-matrix (a
1
a
l
) and A be the N N-matrix
(a
1
a
N
). Then A is invertible and let B := A
1
wA. Let Hom(H, C). Then
A
1
(w)A = (B). Now condition (a) means that there is a matrix C
M
l
(C) such
that (w)A
l
= A
l
C
, whence (B) looks like

_
(v)
0
_
,
where v is the submatrix of B consisting of the rst l rows and columns. Since this
holds for all linear forms, there is the matrix equation
A
1
wA =
_
v
0
_
wA = A
_
v
0
_
or, equivalently, by restriction wA
l
= A
l
v.
(b) (a). From (b) it follows for all Hom(H, C) that (w)A
l
= (v)A
l
, which gives
(w)V V .
2.17. Denition: Let w a corepresentation.
(a) w is said to be irreducible if w ,= 0 and there is no subcorepresentation v such that
0 < dim(v) < dim(w).
(b) w is called completely reducible if w is equivalent to a direct sum of irreducible sub-
corepresentations.
2.18. Lemma: The intersection of invariant subspaces is an invariant subspace.
Proof: This follows directly from Denition-Lemma 2.16, part (a).
2.19. Lemma: Let A Mor(v, w). Then Ker(A) is v-invariant and Im(A) is w-invariant.
Proof: Use Denition-Lemma 2.16, part (a). For each Hom(H, C) the equation A(v) =
(w)A follows. If x Ker(A) then A(v)x = (w)Ax = 0, whence (v)x Ker(A) and the
kernel is v-invariant. If y Im(A), say y = Az, then (w)y = (w)Az = A(v)z is in the
image of A, too.
2.20. Lemma (Schur): Let v, w be irreducible corepresentations. If v and w are not
equivalent, then Mor(v, w) = 0. If v is irreducible, then Mor(v, v) = C1, where 1 is the
identity.
Proof: Let A Mor(v, w) 0. Since v and w are irreducible, by Lemma 2.19, A must
be injective and surjective, whence v and w are equivalent. Now let w = v and be
an eigenvalue of A Mor(v, v). Then A1 Mor(v, v) is not injective and therefore
vanishes.
2.21. Remark: There is a relationship between nite dimensional right comodules of H
and corepresentations.
2.22. Theorem: Let H be a Hopf algebra.
(a) The matrix elements of corepresentations span H.
1
1
This result is related to the fact that each element of a Hopf algebra is contained in a nite
dimensional subcoalgebra.
(b) The matrix elements of a set of non-equivalent irreducible corepresentations are lin-
early independent.
(c) The following are equivalent:
(i) There is a set T of non-equivalent irreducible corepresentations such that the matrix
elements of them form a basis of H.
(ii) Each corepresentation is completely reducible.
2
Moreover if (i) holds then T contains all non-equivalent irreducible corepresentations.
Proof: (a) Let x H. Then there is a number N N, linearly independent elements
x
1
, . . . , x
N
and y
1
, . . . , y
N
in H such that (x) =
N
j=1
x
j
y
j
. By coassociativity,
N
j=1
(x
j
) y
j
=
N
j=1
x
j
(y
j
), whence there are elements v
ij
of H such that
(x
j
) =
N
i=1
x
i
v
ij
for all j. Using coassociativity and the properties for the counit, from these equations
it follows that the elements v
ij
are matrix elements of a corepresentation and x
j
=
i
(x
i
)v
ij
for all j. But then
x =
N
j=1
x
j
(y
j
) =
N
i,j=1
(y
j
)(x
i
)v
ij
is a linear combination of matrix elements.
(b) Use the arguments in the proof of [W2, Proposition 4.7].
(c) The conclusion (ii) (i) is now obvious, because by (a), the Hopf algebra is spanned
by matrix elements of irreducible corepresentations, which are linearly independent
by (b). The conclusion (i) (ii) is proved in [P2, Appendix]. The last remark follows
from (b).
2.23. Proposition: Let v
[ I and v
[ J be sets of irreducible corepresen-

tations such that
I
v
J
v
.
Then the multiplicities of equivalence classes of irreducible corepresentations are the same
on both sides.
3
Proof: The set Mor(
) can be computed using Schurs lemma (Lemma 2.20).

But this set must contain an invertible element, since both direct sums are equivalent.
2
In the language of Hopf algebras this means that H is cosemisimple.
3
cf. Krull-Remak-Schmidt theorem
2.24. Denition-Lemma: (a) Let w be a corepresentation of a Hopf algebra H. Then
also the matrix w
c
with matrix elements w
c
ij
:= S(w
ji
) is a corepresentation, the
contragradient corepresentation to w.
(b) Let w be a corepresentation of a Hopf -algebra H. Then also the matrix w with
matrix elements w
ij
:= w
ij
is a corepresentation. Dene w
to be the transpose of w.
(c) A corepresentation w of a Hopf -algebra is called unitary if w = w
c
or equivalently
ww
= w
w = 1
dimw
.
Proof: (a) and (b) follow from the identities S = (S S), S = , = ( ).
3. Examples of quantum groups
Quantum SL(2)-groups
The simplest Lie group over the complex numbers, which is interesting and important
in physics, is SL(2, C). We want to nd quantum analogues of Pol(SL(2, C)). The corep-
resentations of this Hopf algebra have the following properties:
(1) The irreducible corepresentations are w
, where 2 N
0
.
(2) dim(w
) = 2 + 1 for all ,
(3) w

= w
||
w
||+1
w
+
(Clebsch Gordan),
(4) Each corepresentation is completely reducible, or equivalently, the matrix elements
w
ij
span the Hopf algebra.
Remark: The fundamental corepresentation is w := w
1/2
given by
g (g
ij
)
1i,j2
for g SL(2, C), and w
0
is the identity.
3.1. Denition: A quantum SL(2)-group is a Hopf algebra satisfying the properties
(1)(4).
3.2. Theorem: Up to isomorphism there are the following quantum SL(2)-groups H. The
Hopf algebra H is generated by the matrix elements w
ij
(1 i, j 2) of the fundamental
corepresentation w := w
1/2
and relations
(w w)E = E, E
(w w) = E
,
where the base eld C is canonically embedded into H and there is the following extra
relation between the row vector E
C
2
C
2
and the column vector E
2
C
2
C:
Let e
1
, e
2
be a basis of
2
C and e
1
, e
2
be a dual basis of C
2
. There is the following
presentation:
E =
2
i,j=1
E
ij
e
i
e
j
, E
=
2
i,j=1
E
ij
e
i
e
j
.
Then the 2 2 matrices with entries E
ij
and E
ij
are inverses. There is a basis e
1
, e
2
of
2
C such that
E = e
1
e
2
qe
2
e
1
=
_
0 1
q 0
_
or E = e
1
e
2
e
2
e
1
+e
1
e
1
=
_
1 1
1 0
_
,
where q C0 must not be a non-real root of unity. In the rst case the quantum group
is called the standard deformation SL
q
(2), in the second case it is called the non-standard
deformation SL
t=1
(2). The non-standard deformation SL
t=1
(2) is not isomorphic to any
of the standard deformations, and two standard deformations SL
q
(2) and SL
q
(2) are
isomorphic if and only if q = q
or qq
= 1.
Chapter 3. Examples of quantum groups 16
3.3. Remark: (a) There is a set of non-standard deformations SL
t
(2) indexed by a pa-
rameter t C 0 corresponding to the vector E
t
= e
1
e
2
e
2
e
1
+te
1
e
1
, but
they are all equivalent to the deformation for t = 1, because if the basis vector e
1
is
replaced by e
1
= e
1
t then
tE
t
= e
1
e
2
e
2
e
1
+e
1
e
1
=E
1
.
Since the relations remain the same when E is multiplied by a non-zero scalar, the
Hopf algebras are isomorphic.
(b) For t 0, the vector E
t
tends to the vector for q = 1.
(c) Parts of the proof of Theorem 3.2 can be found e. g. in [DV], [W4], [KP].
To prepare the proof of Theorem 3.2, some extra denitions and lemmas are useful.
3.4. Denition: Let q be a complex number. Then a Hecke algebra of degree n is a unital
algebra generated by elements
1
, . . . ,
n1
subject to the relations
k+1
k
=
k+1
k+1
for 1 k n 2,
(
k
1)(
k
+q
2
1) = 0,
l
=
l
k
for [k l[ 2.
From these relations follows an important property of Hecke algebras and quotients of
them:
3.5. Denition-Lemma: Let / be a Hecke algebra as in Denition 3.4. Let be an ele-
ment of the symmetric group
n
of degree n, i. e. a permutation of the set I := 1, 2, . . . , n.
Then can be written as the composition of transpositions t
j
(where t
j
interchanges the
elements j and j + 1 of I). The minimal number of such transpositions is called length
of and is denoted by l(). Let = t
k
1
t
k
l
be a decomposition of into a minimal
number l = l() of transpositions. Then
k
1

k
l
does not depend on the actual choice of
transpositions as far as their number is minimal. Therefore
:=
k
1

k
l
is well-dened.
3.6. Denition-Lemma: Let / be a Hecke algebra as in Denition 3.4. Then dene the
element
S
n
:=
n
q
2l()
/.
This element satises the property (
k
1)S
n
= 0 for 1 k n 1.
Proof: Let k be an integer between 1 and n 1. Let
n
be a permutation such that
(k) > (k + 1) and let
:= t
k
. If t
k
1
t
k
l
is a decomposition of
into a minimal
number of transpositions then t
k
t
k
1
t
k
l
= t
k
is a decomposition of into a minimal

number of transpositions and l() = l(
) + 1. Therefore for all k

S
n
=
_

n
(k)>(k+1)
q
2l()
n
(k)<(k+1)
q
2l()
_
=
=
_
k
q
2
n
(k)<(k+1)
q
2l()
n
(k)<(k+1)
q
2l()
_
= (q
2
k
+ 1)
n
(k)<(k+1)
q
2l()
and hence
(
k
1)S
n
= (
k
1)(1 +q
2
k
)
. .
=0 (Hecke algebra)
n
(k)<(k+1)
q
2l()
= 0.
3.7. Remark: The Hecke algebra is a generalization of the symmetric group, and for
q = 1 the Hecke algebra relations are just the relations between the transpositions of
the symmetric group. Let V be a vector space. The symmetric group acts on V
n
by
permutations of the tensor factors. The operator
k
corresponding to a transposition t
k
has the eigenvalues 1 and -1. The intersection of the kernels of all
k
1 or of the kernels
of all
k
+ 1 are called totally symmetric vectors or totally antisymmetric vectors,
respectively. When a Hecke algebra (or a quotient of it) acts on V
n
, then the eigenvalues
are 1 and q
2
due to the second Hecke algebra relation. The intersection of the kernels of
all
k
1 or of the kernels of all
k
+q
2
is called the space of totally q-symmetric vectors
or totally q-antisymmetric vectors. The element S
n
is called symmetrization operator,
which is justied by Denition-Lemma 3.6, which also explains the factor q
2l()
in the
denition of S
n
.
Proof of Theorem 3.2: Let K =
2
C be the space of column vectors and K
= C
2
the dual
space of row vectors. w
0
has always the matrix element 1 (because (1) = 1 1). Let w
be the fundamental corepresentation w
1/2
. Then w w

= w
0
w
1
, which is equivalent to
the matrix _
1 0
0 w
_
,
where w
M
3
(C). This matrix has the column eigenvector E = (1 0 0 0)
T
and the row
eigenvector E
= (1 0 0 0). Therefore there are the relations

E
(w w) = E
= w
0
E
, (w w)E = E = Ew
0
.
Thus the vectors E and E
, considered as 4 1 oder 1 4 matrices, intertwine w w and

w
0
. Moreover
E
E ,= 0. (4)
Now (E
1
2
)(1
2
E) can be regarded as an intertwiner of w with w, because w

=
w w
0

= w
0
w and
(E
1
2
)(1
2
E)(ww
0
) = (E
1
2
)(www)(1
2
E) = (w
0
w)(E
1
2
)(1
2
E).
Since w is irreducible, by Schurs Lemma (2.20) (E
1
2
)(1
2
E) is a multiple of the
identity, say times the identity. Using the coordinate representation of E and E
with
respect to a basis e
i
e
j
[ 1 i, j 2 of KK and the dual basis e
i
e
j
[ 1 i, j 2
of K
,
E =
i,j
E
ij
e
i
e
j
, E
i,j
E
ij
e
i
e
j
,
this condition becomes
2
k=1
E
ik
E
kj
=
ij
.
Therefore the matrices

E with entries E
ij
and

E
with entries E
ij
satisfy
E = 1
2
.
If = 0 then

E must have rank 1, because if it has rank 2, then E
= 0 and if it has rank 0

then E = 0 in contradiction to (4). Hence

E
ij
= x
i
y
j
for some x
1
, x
2
, y
1
, y
2
C and E
has the form E = x y, where x = x
1
e
1
+x
2
e
2
, y = y
1
e
1
+ y
2
e
2
. From (w w)E = E it
follows that
wx wy = x y, x wy = w
1
x y.
Both sides are in KKA. Applying id
K
id
A
, where is a linear form on K such
that (x) = 1, we get
wy = y ( id
A
)(w
1
x).
Therefore Cy is an w-invariant subspace in contradiction to the fact that w is irreducible.
Thus ,= 0, and by scaling of E
which does not change the relations, one gets

E
=

E
1
.
The vector E in KK can be written as sum of a symmetric tensor E
sym
, i. e. an element
of K K which is invariant with respect to the ip automorphism of K K, mapping
x y to y x, and an antisymmetric tensor E
asym
satisfying (E
asym
) = E
asym
, dened
by E
sym
=
1
2
(E + (E)) and E
asym
=
1
2
(E (E)). Symmetric tensors
i,j
a
ij
e
i
e
j
in K K, where a
ij
= a
ji
for all i, j, can be identied with quadratic forms Q on K
,
namely Q(
i
v
i
e
i
) =
i,j
a
ij
v
i
v
j
. In particular there are bases such that E
sym
has one of
the following presentations:
(a) E
sym
= e
1
e
2
+e
2
e
1
if Q has rank 2,
(b) E
sym
= e
1
e
1
if Q has rank 1,
(c) E
sym
= 0 if Q has rank 0.
With respect to any basis e
1
, e
2
of K an antisymmetric tensor E
asym
is a scalar multiple
of e
1
e
2
e
2
e
1
. Therefore E has one of the following presentations:
(a) E = (1 +c)e
1
e
2
+ (1 c)e
2
e
1
with c C. Since

E has rank 2, both coecients
must not vanish. Therefore E is a scalar multiple of e
1
e
2
qe
2
e
1
, where q =
c1
c+1
and q C 0, 1.
(b) E = e
1
e
1
+c(e
1
e
2
e
2
e
1
), where c C0, because

E has rank 2. Therefore E
is a scalar multiple of e
1
e
2
e
2
e
1
+te
1
e
1
, where t =
1
c
. According to Remark 3.3
this is equivalent to the vector for SL
t=1
(2). In this case let q := 1.
(c) E = c(e
1
e
2
e
2
e
1
), where c C0. This is the case q = 1 which is not included
in (a).
Now let the associative, unitary algebra H
0
be generated by the elements , , , subject
to the relations (vv)E = E, E
(vv) = E
for v = (
). There are uniquely determined

comultiplication, counit and antipode such that this algebra becomes a Hopf algebra and v
is a corepresentation (see Proposition 3.8 and Proposition 3.9 below).
Since the relations between the generators of H
0
are satised in H, there is a Hopf algebra
map
: H
0
H, v
ij
w
ij
.
We shall study the corepresentation theory of H
0
.
Consider the 4 4 matrix := 1
4
+ qE E
(where E and E
are again 4 1 and 1 4

matrices, respectively). Then is an element of the vector space Mor(v v, v v). It
satises the relations
( 1
4
)( +q
2
1
4
) = 0, ( 1
2
)(1
2
)( 1
2
) = (1
2
)( 1
2
)(1
2
).
Fix an integer n 2 and dene for integers k satisfying 0 < k < n:
k
= 1
2
1
2
. .
k1
1
2
1
2
. .
nk1
.
These are operators on the n-fold tensor product K
n
and intertwine v
n
with v
n
. They
satisfy the Hecke algebra relations (cf. Denition 3.4).
Now dene the operators
as in Denition-Lemma 3.5 and the symmetrization operator

as in Denition-Lemma 3.6:
S
n
:=
n
q
2l()
.
Due to Denition-Lemma 3.6 it takes values in
K
n/2
:= x K
n
[ k:
k
(x) = x =
k
Ker(
k
1).
The dimension of the space K
n/2
is n + 1. (Proof: analyze relations on coordinates of
elements of K
n/2
or see [W4]). The space K
n/2
is v
n
-invariant as intersection of the
kernels of the intertwiners
k
1 by Lemma 2.18 and Lemma 2.19 and a right comodule.
Let v
n/2
denote the corresponding subcorepresentation of v
n
as in Denition-Lemma 2.16.
Then v
n/2
is a corepresentation of dimension n+1. By denition, v = v
1/2
, v
0
is the trivial
one-dimensional corepresentation. At this moment we assume that q is not a non-real root
of unity. For all s
1
2
N
0
, one has
(A) v
k
is irreducible for all k s +
1
2
,
(B) v
k
v

= v
k+
1
2
v
k
1
2
, where by denition v
1/2
:= 0, for all k s.
These statements will be proved by induction: The case s = 0 follows from the result on
monomials below.
Suppose the statements are true for s replaced by s
1
2
.
We want to decompose v
s
v and consider the map
: K
s
1
2
K
s
K, x (S
2s
1
2
)(x E)
(antisymmetrization-symmetrization procedure). Note that E K K. The map is well
dened due to the property of the symmetrization operator and intertwines v
(2s1)

=
v
(2s1)
v
0
with v
2s
v. By inspection
(
2s
1)(e
1
e
1
)
. .
2s1 factors
,= 0
if q is not a non-real root of unity. Therefore (e
1
e
1
) / K
s+
1
2
and Ker() ,= K
s
1
2
.
By induction hypothesis, there is no proper non-trivial v
2s1
-invariant subspace of K
s
1
2
and the kernel of is invariant by Lemma 2.19. Consequently is injective and Im()
corresponds to v
s
1
2
. Moreover
Im() K
s+
1
2
,= Im().
Since v
s
1
2
is irreducible, there is no proper non-trivial v
s
1
2
-invariant subspace of Im()
and Im() K
s+
1
2
= 0. Thus
K
s
1
2
K
s+
1
2

= Im() K
s+
1
2
K
s
K.
Equality follows by dimension arguments (dimK
t
= 2t + 1) and yields (B). By denition
of the tensor product of representations, the monomials in , , , of degree smaller or
equal to 2s+1 are linear combinations of the matrix elements of v
(2s+1)
. Using result (B)
yields that they are linear combinations of matrix elements of v
0
, v
1
2
, . . . , v
s+
1
2
. The space
of these monomials has dimension
2s+2
k=1
k
2
, as in the classical case for Pol(SL(2, C)). This
has been shown in [W4], [WZ2]. Since v
t
has (2t +1)
2
matrix elements for all t
1
2
N
0
, the
space spanned by the matrix elements of v
0
, v
1
2
, . . . , v
s+1/2
has this dimension if and only
if all matrix elements are linearly independent. Hence (A) follows. Now it is easy to prove
that H
0
is a quantum SL(2)-group.
The matrix (v
s
) is a corepresentation of H because respects and .
Then (v
s
)

= w
s
for all s
1
2
N
0
. Proof by induction: The assertion is trivial for s = 0
and s =
1
2
. Suppose the statement is true for all non-negative half integers smaller than s.
Then by part (B), v
s
1
2
v

= v
s1
v
s
. Since is an algebra homomorphism, by the
denitions of direct sum and tensor product of corepresentations the following holds
w
s1
(v
s
)

= (v
s1
) (v
s
)

= (v
s1
v
s
)

= (v
s
1
2
v) =
= (v
s
1
2
) (v)

= w
s
1
2
w

= w
s1
w
s
.
Due to condition (4) for the quantum SL(2)-group, the corepresentation (v
s
) is com-
pletely reducible, whence by Proposition 2.23 (v
s
)

= w
s
. Thus is an isomorphism and
H can be identied with H
0
.
Now we consider the case when q is a non-real root of unity (see e. g. [KP]). Let q be a
non-real root of unity of order N. Dene
N
0
:=
_
N if N is odd,
N/2 if N is even.
Then H
0
has a corepresentation
4
z =
_
N
0
N
0
N
0
N
0
_
.
Then v
k
v

= v
k+
1
2
v
k
1
2
for k < (N
0
1)/2, v
k
is irreducible for k
1
2
(N
0
1), and
v
(N
0
1)/2
v

=
_
_
v
1
2
N
0
1

0 z
0 0 v
(N
0
1)/2
_
_
.
It is possible to show (v
k
)

= w
k
for k
1
2
(N
0
1) as before, but on the other hand
w
1
2
N
0
1
w
1
2
N
0
= w
1
2
(N
0
1)
w

= (v
1
2
(N
0
1)
v)

=
=
_
_
w
1
2
N
0
1

0 (z)
0 0 w
1
2
N
0
1
_
_
= w
1
2
N
0
1
(z) w
1
2
N
0
1
,
4
cf. [T2, part 5.2]
because corepresentations in H are completely reducible. But this is a contradiction to
Proposition 2.23.
Let q
1
and q
2
be two values such that SL
q
1
(2)

= SL
q
2
(2) (q
1
, q
2
C Y t = 1,
where the subset Y contains 0 and all non-real roots of unity), i. e. that the Hopf algebras
are isomorphic. Then the fundamental representation w
1
is mapped to w
2
, i. e. they are
equivalent: w
1
= Qw
2
Q
1
. Let E
1
, E
2
be the corresponding eigenvectors. Then
(w
1
w
1
)E
1
= E
1
, (Qw
2
Q
1
Qw
2
Q
1
)E
1
= E
1

(w
2
w
2
)((Q
1
Q
1
)E
1
) = (Q
1
Q
1
)E
1
= E
2
with C 0, because the space of eigenvectors of w
2
w
2
for the eigenvalue 1 is one-
dimensional. There are symmetric tensors E
sym
1
, E
sym
2
and antisymmetric tensors E
asym
1
,
E
asym
2
such that E
1
= E
sym
1
+E
asym
1
and E
2
= E
sym
2
+E
asym
2
. Therefore
E
1
= (QQ)E
2
E
sym
1
= (QQ)E
sym
2
, E
asym
1
= (QQ)E
asym
2
(5)
and E
sym
1
and E
sym
2
have the same rank. If the rank is 0 or 1, it is the same deformation,
and if the rank is 2, one can use (5) and the fact that the rank of (Qe
1
Qe
2
) is one, to
get q
1
= q
2
or q
1
q
2
= 1 (in the last case the isomorphism is given by e
1
e
2
).
Quantum SL(N)-groups
Let N be a positive integer greater than 1. The Hopf algebra H of the group SL(N, C)
corresponds to the commutative unital algebra generated by the matrix elements w
ij
for
1 i, j N of a fundamental corepresentation w subject to the relations
w
N
E = E, E
w
N
= E
(6)
where E and E
are classical completely antisymmetric elements of (

N
C)
N
and (C
N
)
N
respectively, i. e. with respect to a basis e
1
, . . . , e
N
of
N
C and a dual basis e
1
, . . . , e
of C
N
, they can be presented as
E =
n
(1)
l()
e
(1)
e
(N)
, E
n
(1)
l()
e
(1)
e
(N)
. (7)
Then the relations just mean (assuming commutativity) that the determinant of the ma-
trix w is one. For SL(2) this is just e
1
e
2
e
2
e
1
and e
1
e
2
e
2
e
1
, which is changed
to e
1
e
2
qe
2
e
1
and up to a non-zero factor to e
1
e
2
qe
2
e
1
in the standard
deformation SL
q
(2). Therefore it is natural to dene
E
q
=
n
(q)
l()
e
(1)
e
(N)
, E
q
=
n
(q)
l()
e
(1)
e
(N)
(8)
and to consider the relations
w
N
E
q
= E
q
, E
q
w
N
= E
q
. (9)
For q not being a non-real root of unity they imply (cf. [W3])
w
2
= w
2
(10)
where
(e
i
e
j
) :=
_
qe
j
e
i
if i < j,
qe
j
e
i
+ (1 q
2
)e
i
e
j
if i > j,
e
i
e
i
if i = j,
for i, j = 1, . . . , N. Now SL
q
(N) is introduced as the unital algebra generated by w
ij
for
1 i, j N subject to the relations (9), (10) (cf. [P2]). One can check that this denition
coincides with the standard one (cf. [Dr], [R]).
The following proposition shows that all unital algebras with relations dened by inter-
twiners are bialgebras. If the intertwiners are chosen badly, the bialgebras can be small
and uninteresting. For each matrix w and each n N dene the matrix w
n
as for corep-
resentations in Denition 2.15 and let w
0
:= 1
1
.
3.8. Proposition: Let H be the universal unital algebra generated by elements w
ij
for
1 i, j N, which are the entries of a matrix w subject to relations
E
m
w
s
m
= w
t
m
E
m
(11)
for m in an index set I, s
m
, t
m
N
0
and E
m
M
N
t
m
N
s
m(C). Then there exist a unique
comultiplication and counit such that H is a bialgebra and w is a corepresentation of H.
Proof: (a) Uniqueness: We must have w
ij
=
N
k=1
w
ik
w
kj
and (w
ij
) =
ij
for all i
and j. Since and are unital algebra homomorphisms, they are uniquely determined
if they exist.
(b) Existence: Dene w
ij
:=
N
k=1
w
ik
C
w
kj
HH for all i and j. The matrix w with
entries w
ij
also satises the relations (11), because w
n
= w
n
C
w
n
follows from
the rule (a b)(c d) = (ac bd) and
E
m
w
s
m
= E
m
(w
s
m
C
w
s
m
) = w
t
m
E
m

C
w
s
m
=
= w
t
m
C
E
m
w
s
m
= w
t
m
C
w
t
m
E
m
= w
t
m
E
m
,
because the entries of E
m
are just complex numbers. Dene w
ij
:=
ij
for all i, j.
Then the matrix w with entries w
ij
satises the properties
E
m
w
s
m
= E
m
, w
t
m
E
m
= E
m
,
whence it satises relations (11). Now the universality of H gives the existence of
unital homomorphisms , such that (w
ij
) = w
ij
and (w
ij
) = w
ij
. It is enough to
check Conditions (1) and (2) for bialgebras (cf. Denition 2.6) for elements f = w
ij
when they are obvious.
3.9. Proposition: Let the conditions of Proposition 3.8 be satised. Let e
1
, . . . , e
N
be
a basis of
N
C and e
1
, . . . , e
N
be a dual basis of C
N
. Moreover assume that there exist
positive integers s and t and elements E Mor(1
1
, w
t
) and E
Mor(w
s
, 1
1
) such that
E =
N
k=1
e
k
f
k
, E
=
N
k=1
f
k
e
k
such that the elements f
k
(
N
C)
t1
and f
k
(C
N
)
s1
are linearly independent.
Then the matrix w
1
exists and there is a uniquely determined antipode S such that the
bialgebra H is a Hopf algebra.
Proof: From the relation w
t
E = E it follows that
(w w
(t1)
)E = E
N
k=1
we
k
w
(t1)
f
k
=
N
k=1
e
k
f
k
. (12)
Since the elements f
k
of (
N
C)
(t1)
are linearly independent, there are elements g
k
of the
dual space (C
N
)
(t1)
such that g
i
f
j
=
ij
. Apply e
i
g
j
to Equation (12):
N
k=1
e
i
we
k
g
j
w
(t1)
f
k
=
N
k=1
e
i
e
k
g
j
f
k
= 1 g
j
f
i
=
ij
.
Therefore the matrix G with entries G
kj
:= g
j
w
(t1)
f
k
is a right inverse to w. From the
second condition it follows in a similar way that there is a left inverse of w. Thus w
1
exists. Finally, when to the relation
E
m
w
s
m
= w
t
m
E
m
,
(w
s
m
)
1
= (w
1
)
op
s
m
is applied to the right and (w
t
m
)
1
= (w
1
)
op
t
m
to the left (the
tensor product
op
is with respect to the algebra H
op
with opposite multiplication),
then
(w
1
)
op
t
m
E
m
= E
m
(w
1
)
op
s
m
.
Therefore there is a unital algebra homomorphism S: H H
op
such that S(w) = w
1
.
Equivalently, S: H H is a unital antihomomorphism. It is enough to check Condition (3)
for the antipode (cf. Denition 2.6) for f = w
ij
when it is obvious. Uniqueness of S follows
from Theorem 2.8.
3.10. Remark: For the quantum SL(N) group take I = 1, 2, 3, E
1
= E
q
, t
1
= N,
s
1
= 0, E
2
= E
q
, t
2
= 0, s
2
= N, E
3
= , t
3
= s
3
= 2. Then the algebras SL
q
(N) are
Hopf algebras.
3.11. Remark: For 0 < q 1 the corepresentation theory of SL
q
(N) is the same as for
the classical SL(N) (cf. [W3], [P2]). If q is transcendental, see [R], [H]. If q C 0
is not a non-real root of unity, see [PW]. There are deformations of the orthogonal and
symplectic groups [RTF], [T1] (cf. [P2]).
4. -Structures
In the classical theory there exist -structures on Pol(SL(2)) which give the Hopf
-algebras Poly(SU(2)), Poly(SU(1, 1)) and Poly(SL(2, R)). We will classify the Hopf
-algebra structures on the quantum SL(2)-groups H described in Theorem 3.2. Firstly
recall that H is generated as an algebra by the matrix elements of a 22 matrix w subject
to the relations
(w w)E = E, E
(w w) = E
or equivalently
j,l
w
ij
w
kl
E
jl
= E
ik
,
i,k
E
ik
w
ij
w
kl
= E
jl
. (13)
4.1. Lemma: Let be an (anti-)linear comultiplicative algebra (anti-)automorphism of a
quantum SL(2)-group H. Then
(a) there exists a matrix Q GL(2, C) such that (w) = QwQ
1
.
(b) If and only if the matrix Q GL(2, C) satises the conditions
(Q
1
Q
1
)E = cE, E
(QQ) = c
(14)
for some numbers c, c
C 0, there is a Hopf algebra automorphism of H

such that (w) = QwQ
1
. Moreover, all Hopf algebra automorphisms of H can be
described in this way.
(c) Let denote the linear twist (interchanging factors) and let

E and

E
denote the
elements of
2
C
2
C and C
2
C
2
with conjugate complex coecients with respect to
the bases e
i
e
j
, e
i
e
j
. Then if and only if the matrix Q GL(2, C) satises the
conditions
(Q
1
Q
1
)

E = cE,

E
(QQ) = c
(15)
for some c, c
C 0, there is an antilinear, comultiplicative, algebra antiautomor-

phism of H such that (w) = QwQ
1
.
(d) Let the antilinear involutive comultiplicative algebra antiautomorphisms ,

and
the corresponding matrices Q and

Q be dened as in (a). Then the Hopf algebra H
equipped with -structures and

gives isomorphic Hopf -algebras if and only if

is equivalent to up to a Hopf algebra automorphism (i. e.

=
1
) if and only
if

Q = c

A
1
QA where c C 0 and A GL(2, C) corresponds to via (b).
Proof: (a) Since is comultiplicative, the matrix (w) is a corepresentation. The following
conclusions follow from the fact that is bijective: w is irreducible if and only if the
matrix elements w
ij
are linearly independent if and only if the matrix elements (w
ij
)
are linearly independent if and only if (w) is irreducible. But there is only one
Chapter 4. -Structures 26
irreducible corepresentation of dimension 2 up to isomorphism, therefore there is a
matrix Q GL(2, C) such that
(w) = QwQ
1
. (16)
(b) Since the trivial corepresentation appears in the direct sum decomposition w w

=
w
1
w
0
only once, by Lemma 2.20 the space of intertwiners in Mor(ww, w
0
) is one-
dimensional. Thus Condition (14) is equivalent to the condition that (Q
1
Q
1
)E
intertwines w w with w
0
and E
(QQ) intertwines w
0
with w w:
(w w)(Q
1
Q
1
)E = (Q
1
Q
1
)E (QwQ
1
QwQ
1
)E = E
and E
(QQ)(w w) = E
(QQ) E
(QwQ
1
QwQ
1
) = E
.
_
(17)
Let be a Hopf algebra automorphism of H. Then by part (a), there is a matrix
Q GL(2, C) such that (w) = QwQ
1
. The automorphism must map the rela-
tions between the generators of H to relations in H, therefore Equation (17) holds.
Conversely, let Equation (17) be satised. Let T be the free associative unital algebra
generated by the matrix elements of w and let 1 be the two-sided ideal generated by
the relations (13). Then the map can be dened as unital algebra homomorphism
on T such that (w) = QwQ
1
. Equation (17) shows that maps 1 to 1, therefore
it induces a unital algebra homomorphism on H = T/1. Such preserves the Hopf
algebra structure of H. Moreover, replacing Q by Q
1
(Equation (14) still holds for
c
1
and (c
)
1
) we get
1
.
(c) The proof is similar as in part (b). The only changes arise from the fact that should
be an antilinear algebra antiautomorphism instead of a linear algebra automorphism.
Therefore, applied to relations (13) yields
j,l
(w
kl
)(w
ij
)

E
jl
=

E
ik
,
i,k
ik
(w
kl
)(w
ij
) =

E
jl
or shortly
((w) (w))

E =

E,

E
((w) (w)) =

E
.
Using
2
= id
H
, we get the desired results.
(d)
1
(w) = (A
1
wA) = (

A
1
QwQ
1

A) =

A
1
QAwA
1
Q
1

A, while

(w) =
Qw

Q
1
. The left hand sides are equal if and only if

Q
1

A
1
QA Mor(w, w) = C1
2
.
Remark. It is easy to check that the second condition in (14) (and also the second
condition in (15)) is redundant.
4.2. Theorem: All non-equivalent Hopf -algebra structures on the quantum SL
q
(2)-
groups H are dened by w = QwQ
1
, where
(a) Q = (
1
0
0
1
), [q[ = 1. Then w = w. This algebra is called Poly(SL
q
(2, R)).
(b) Q = (
0
1
q
0
), q R 0. Then w
Bw = wBw
= B, for B := (
1
0
0
1
). This algebra is
called Poly(SU
q
(1, 1)).
(c) Q = (
0
1
q
0
), q R 0. Then w is unitary. This algebra is called Poly(SU
q
(2)).
The only equivalence among them is Poly(SL
1
(2, R))
= Poly(SU
1
(1, 1)).
For the non-standard deformation SL
t=1
(2) there is only one Hopf -algebra structure (up
to equivalence), namely for Q = (
1
0
0
1
).
Except for (c), the above corepresentations w are not equivalent to unitary ones (The
above examples were given in [W1], [RTF], [W4]).
Ideas of the proof: Since the map is an antilinear comultiplicative algebra antiauto-
morphism, by Lemma 4.1, part (a) there is a matrix Q GL(2, C) such that w = QwQ
1
.
By part (c) of Lemma 4.1, the map can be an algebra antiautomorphism if and only
if Q satises the condition
(Q
1
Q
1
)

E = cE
for some c C 0. The equation
2
= id
H
is equivalent to
QQ = d1
2
with d C0. Q is determined up to the equivalence relation as in Lemma 4.1, part (d).
Consider the standard quantum deformations SL
q
(2), q ,= 1, rst. From the relations (13)
it follows that there are only the following characters (algebra homomorphisms) : H C:
a
(w) =
_
a 0
0 a
1
_
and in addition to that for q = 1:
a
(w) =
_
0 a
a
1
0
_
,
where a C 0 (Relations (13) are equivalent to
w
11
w
12
= qw
12
w
11
, w
11
w
21
= qw
21
w
11
, w
12
w
22
= qw
22
w
12
,
w
21
w
22
= qw
22
w
21
, w
12
w
21
= w
21
w
12
,
w
11
w
22
qw
12
w
21
= w
22
w
11
q
1
w
12
w
21
= 1,
and the numbers (w
ij
) should satisfy the same relations).
Now the following trick can be used in order to compute all possible -structures: If is a
character, then also the map
#
: x (x
) is a character, because C is commutative.

Then for any a C 0 there exists b C 0 such that
#
a
=
b
or (for q = 1)
#
a
=
b
. Applying both sides to w, we get that Q is a diagonal or antidiagonal matrix.
Similarly (use ), isomophisms of Hopf algebras are given by diagonal or (q = 1)
antidiagonal matrices. Then we use the other conditions for Q and part (d) of Lemma 4.1.
For the non-standard deformation SL
t=1
(2) split E into E
sym
and E
asym
as in the proof
of Theorem 3.2. Then consider Q with respect to both. For q = 1, equivalent Qs can be
regarded as matrices of the same antilinear mapping j such that j
2
= d id (j is equivalent
to kj for some k C 0). Then d = 1 corresponds to (a), (b) while d = 1 to (c).
4.3. Remark [RTF], [P2].
(a) There exist the following -structures on SL
q
(N):
(i) For [q[ = 1 you can choose w = w. The corresponding quantum group is called
SL
q
(N, R).
(ii) If q is real then for
1
, . . . ,
N
1 there are -structures such that w
Bw =
wBw
= B, where B is a diagonal matrix with diagonal elements

1
, . . . ,
N
. The
corresponding quantum group is called SU
q
(N;
1
, . . . ,
N
). For
1
= =
N
= 1
we get the quantum group SU
q
(N), in which w is a unitary corepresentation.
(b) There are also -structures on the orthogonal and symplectic quantum groups.
5. Compact Hopf -algebras
In this chapter we follow [W2], [W3], [Ko]. Let / be a Hopf -algebra.
5.1. Denition: / is called compact if there are unitary corepresentations such that their
matrix elements generate / as algebra.
Example: The fundamental corepresentation of Poly(SU
q
(N)) is unitary and generates
Pol(SL
q
(N)) as algebra.
5.2. Lemma: Let / be a compact Hopf -algebra.
(a) The matrix elements of unitary corepresentations span /.
(b) Let v be a unitary corepresentation. Then v is equivalent to a direct sum of irreducible
unitary corepresentations.
(c) The matrix elements of non-equivalent irreducible unitary corepresentations form a
linear basis of /.
(d) Each irreducible corepresentation is equivalent to a unitary one.
(e) Each corepresentation is completely reducible (into irreducible ones). Since the irre-
ducible corepresentations are equivalent to unitary corepresentations, each corepre-
sentation is equivalent to a unitary corepresentation.
Proof: (a) By denition, / is spanned by matrix elements of tensor products of unitary
corepresentations, but tensor products of unitary corepresentations are unitary.
(b) Proof by induction with respect to the dimension d of corepresentations. If d = 1
or the corepresentation is irreducible, then there is nothing to do. Now assume that
the corepresentation v is not irreducible. Then choose an orthonormal basis of an
invariant proper subspace L and add some more orthonormal elements in order to
get an orthonormal basis B of C
dimv
. The transition from the standard basis to B is
unitary and intertwines v with a unitary corepresentation
_
A B
0 C
_
=: w,
where A, B, C are matrices of suitable size and with at least one entry. Since w is
unitary, w = w
c
or S(w) = w
or equivalently
_
S(A) S(B)
0 S(C)
_
=
_
A
0
B
_
.
Therefore B = 0, moreover A and C are unitary and w a direct sum of them (notice
that L
is also invariant and C = w[

L
). By induction hypothesis, the corepresen-
tations A and C of dimensions less than d are direct sums of irreducible unitary
corepresentations, whence w is a direct sum of irreducible unitary corepresentations.
(c) This follows from (a), (b), and Theorem 2.22, part (b).
(d) and (e) follow from Theorem 2.22, part (c).
Chapter 5. Compact Hopf -algebras 30
Remark: All irreducible corepresentations can be obtained by decomposition of tensor
products of those unitary corepresentations which generate / as algebra (cf. Lemma 5.2,
part (a)).
Peter-Weyl Theory and Haar measure
Let / be a compact Hopf -algebra. Let 1 be an index set and let u
[ 1 be a
complete set of non-equivalent irreducible unitary corepresentations. Let I := u
0
be the
one dimensional corepresentation. Then the elements u
mn
form a basis of / (Lemma 5.2,
part (c)).
5.3. Denition: The Haar measure is a linear functional on / dened by
h(u
mn
) =
,0
.
Since the u
mn
are matrix elements of corepresentations, for all x / the Haar measure
satises the equations
(h id
A
)(x) = (id
A
h)(x) = h(x)1, h(1) = 1. (18)
(By denition, also h(S(x)) = h(x) holds for all x /.)
In order to compute h on products, some preparation is necessary.
5.4. Lemma: For each 1 there is a strictly positive denite matrix F
such that
(u
)
cc
= F
F
1
.
Proof: For each 1, the matrix u
is also a corepresentation and equivalent to a unitary

one, say u
: Q
Q
1
= u
. Then u
= (u
)
c
and
(u
)
cc
= (u
)
c
= (Q
1
)
c
= Q
T
(u
)
c
(Q
1
)
T
= Q
T
(Q
1
)
T
= Q
T
Q
1
(Q
1
)
T
and therefore (u
)
cc
= F
F
1
where F
= Q
T
(Q
T
is a strictly positive denite

matrix.
Fix an irreducible corepresentation v and let n := dim(v). Since S(v) is the inverse of v,
there are intertwiners
AI = (v v
c
)A, B(v
c
v) = IB,
where A =
n
k=1
e
k
e
k
and B =
n
k=1
e
k
e
k
.
5.5. Lemma: Let v and w be irreducible representations of dimensions n and m respec-
tively. Then
(a) Mor(v
c
w, I)
= Mor(w, v), Mor(v

c
v, I) = CB.
(b) Mor(I, w v
c
)

= Mor(v, w), Mor(I, v v
c
) = CA.
Proof: (a) If X intertwines v
c
w with I then X(v
c
w) = IX and
(1
n
X)(A1
m
)(I w) = (1
n
X)(v v
c
w)(A1
m
) = (v I)(1
n
X)(A1
m
).
Since I w

= w and v I

= v, (1
n
X)(A1
n
) can be regarded as intertwiner of w
and v. Conversely, let Y Mor(w, v). Then Y w = vY and
B(1
n
Y )(v
c
w) = B(v
c
v)(1
n
Y ) = IB(1
n
Y ).
Therefore B(1
n
Y ) intertwines v
c
w with I. The maps between Mor(v
c
w, I) and
Mor(w, v) are inverses of each other because (1
n
B)(A1
n
) = (B1
n
)(1
n
A) = 1
n
.
The second statement follows from the rst with Schurs Lemma 2.20.
(b) is proved in a similar way.
Now the Haar measure is computed on certain products of basis elements:
5.6. Theorem: The Haar measure satises the Peter-Weyl-Woronowicz relations:
h(u
mn
u
jl
) =
,
(F
)
ln
mj
Tr(F
)
(19)
and
h(u
jl
mn
) =
,
(F
1
)
mj
ln
Tr(F
1
)
(20)
for all , 1, 1 m, n dim(u
), 1 j, l dim(u
).
Proof: Let w be any corepresentation (of dimension N). Application of h id and id h
to w
ij
yields together with Equation (18)
h(w)w = wh(w) = h(w)1.
This matrix equation means
N
k=1
h(w)
ik
w
kj
= h(w)
ij
1 =
N
k=1
w
ik
h(w)
kj
or equivalently that for each i the row vector with coordinates h(w)
ij
for j = 1, . . . , N
intertwines w with I and for each j the column vector with coordinates h(w)
ij
for i =
1, . . . , N intertwines I with w. These facts will be applied to the sets Mor(I, u
c
) and
Mor(u
cc
u
c
, I) for , 1.
Therefore for w = u
c
and for xed indices k, l, the element (h(u
ik
u
jl
c
))
1i,jN
is
in Mor(I, u
c
). By Lemma 5.5 it vanishes for ,= and is a multiple of A for = .
Thus there are numbers
kl
C such that
h(u
ik
u
jl
c
) =
,
kl
ij
(21)
for all i, j, k, l. Similarly, for w = u
cc
u
c
and for xed indices i, j, the element
(h(u
ik
cc
u
jl
c
))
1k,lN
is in Mor(u
cc
u
c
, I) = CB. Therefore there are numbers
ij
C
such that
h(u
ik
cc
u
jl
c
) =
ij
kl
. (22)
But from Lemma 5.4, u
= F
1
(u
)
cc
F
, which yields by linearity and Equation (22) the

equation
h(u
mn
u
jl
c
) =
i,k
(F
1
)
mi
h(u
ik
cc
u
jl
c
)(F
)
kn
= (F
)
ln
i
(F
1
)
mi
ij
.
Comparison with Equation (21) and u
jl
c
= (u
jl
)
yields
h(u
mn
u
jl
) = c
,
(F
)
ln
mj
for some c
C. These constants can be evaluated using the unitarity of u
:
1 = h(1) =
n
h(u
mn
u
mn
) = c
n
(F
)
nn
= c
Tr(F
).
The trace of F
is positive because F
is positive denite. This proves Equation (19). The

other equation is proved in a similar way.
5.7. Remark: (a) Since the matrices F
can be scaled by a positive number, we normalize

them by the condition Tr(F
) = Tr(F
1
). After normalization they are uniquely

determined.
(b) Example: In the standard deformation SU
q
(2) for q R 0,
F
0
= (1), F
1/2
=
_
[q[
1
0
0 [q[
_
.
Proof: w
1/2
= w = (
), and S(w) = (

q
q
1
). Then
w
cc
= S
2
(w) =
_
q
2
q
2

_
= F
1/2
wF
1
1/2
where F
1/2
is as desired. Note that the absolute value of q must be used, because the
eigenvalues of a positive denite matrix must be positive.
5.8. Theorem (Positivity of the Haar measure)
For all x /, h(x
x) 0, and equality only holds for x = 0.

Proof: Since / has a basis u
mn
[ 1 m, n dim(u
), I, a general element a of /
can be written as
a =
m,n,
a
mn
u
mn
.
By the second Peter-Weyl-Woronowicz relation (20)
h(a
a) =
,m,n,p
(a
mp
(F
1
)
mn
a
np
)
Tr(F
)
,
in which the sums
m,n
a
mp
(F
1
)
mn
a
np
are strictly positive unless all coecients a
mp
for xed , p vanish, because the matrices F
1
are strictly positive denite for all .

5.9. Corollary (Scalar product)
There is a scalar product on / dened by (a [ b) := h(a
b) for all a, b /.
Proof: This inner product is antilinear in the rst argument and linear in the second
argument by denition and positive denite by Theorem 5.8.
5.10. Corollary (Modular Homomorphism)
There is a uniquely determined algebra automorphism of / such that h(ab) = h(b(a))
for all a, b /. It is dened on elements of the basis as
(u
mn
) = (F
)
mn
.
Proof: Uniqueness: Let a be an element of / and let a
, a
/ such that for all b /

the equation
h(ab) = h(ba
) = h(ba
)
holds. Then h(b(a
)) = 0 for all b /, whence a
= a
by Corollary 5.9.
Existence: From the second Peter-Weyl-Woronowicz relation it follows that
h(u
jl
(u
mn
)) =

,
(F
)
ln
mj
Tr(F
)
= h(u
mn
u
jl
).
Therefore by linearity h(ab) = h(b(a)) for all a, b /. Moreover F
0
= (1) implies
(1) = 1, and for all a, b, c /,
h(a(bc)) = h(bca) = h(ca(b)) = h(a(b)(c)).
Therefore is a unital algebra homomorphism. Since F
is invertible, also is invert-

ible.
C
-structure
For any Hilbert space H let (.[.)
H
denote the inner product and B(H) the set of bounded
linear operators on H. Then B(H) is a -algebra. Let / be a compact Hopf -algebra and
consider the set
:= : / B(H) [ H Hilbert space, unital -homomorphism
(it is enough to consider some xed H with dim(H) dim(/) as cardinal numbers, thus
is actually a set).
Fix and let H be the corresponding Hilbert space. Let u
be a unitary corepre-
sentation of /. Then (u
) is a unitary matrix in M
dimu
(B(H)) and
m
(u
mn
)
(u
mn
) = 1
for all n dimu
. Therefore for all x H and k dimu
(x [ x)
H
=
m
((u
mn
)
(u
mn
)x [ x)
H
=
=
m
((u
mn
)x [ (u
mn
)x)
H
((u
kn
)x [ (u
kn
)x),
whence the operator norm |(u
kn
)| is at most 1, and for each a =
,m,n
a
mn
u
mn
/
there is the inequality
|(a)|
,m,n
[a
mn
[ < .
Therefore the following denition is possible:
5.11. Denition-Lemma: There is a norm |.|
C
on / such that for all a /,
|a|
C
= sup
|(a)|.
Moreover this norm satises the equations |ab|
C
|a|
C
|b|
C
, |a
|
C
= |a|
C
,
|a
a|
C
= |a|
2
C
for all a, b /.
Proof: The main problem is to show |a|
C
= 0 a = 0 for a /. The inner product
(. [ .) on / induces a norm |.|
(.|.)
(cf. Corollary 5.9). For each a / let
0
(a) denote the
operator of left multiplication by x on /. Then for all x /
m
|
0
(u
mn
)(x)|
2
(.|.)
= h(x
m
(u
mn
)
mn
)
. .
=1
x) = h(x
x) = |x|
2
(.|.)
,
whence the operator norm |.|
(.|.)
of
0
(u
mn
) is at most 1. For all a =
,m,n
a
mn
u
mn
/
|
0
(a)|
(.|.)

,m,n
[a
mn
[.
Therefore for each a / the operator
0
(a) is bounded on / and can be extended to the
completion H of / with respect to the norm |.|
(.|.)
as a bounded linear operator
0
(a)
with same operator norm |
0
(a)|
(.|.)
:= |
0
(a)|
(.|.)
. Therefore
0
, and
|a|
C
= 0 |
0
(a)|
(.|.)
= 0 |
0
(a)1|
(.|.)
= 0 |a|
(.|.)
= 0 a = 0.
The other properties of this norm follow from the corresponding properties of the operator
norms of the representations in .
5.12. Denition: Let A be the closure of / with respect to the norm |.|
C
. Then A is a
C
-algebra by Denition-Lemma 5.11.

The following properties of C
-algebras are useful:

5.13. Proposition: Let A be a C
-algebra. Then
(a) There is a Hilbert space H such that A can be embedded as closed -subalgebra
into B(H) [D, 2.6.1].
(b) Let B be another C
-algebra. Then each -homomorphism from A to B is continuous

[D, 1.3.7].
The comultiplication of / can be extended to a -homomorphism from A to A

A,
where A

A denotes the (topological) tensor product of C
-algebras, dened as follows:

Let H be a Hilbert space and let : A B(H) be an embedding of C
-algebras. Then
A

A is identied with the closure of ( )(A A) in B(H

H), where H

H is the
(topological) tensor product of Hilbert spaces. The C
-algebra A

A does not depend
(up to isomorphisms) on the embedding [D, 2.12.15]. The map
/

// B(H

H)
is a -homomorphism called
1
. Since H

H is a Hilbert space,
1
belongs to and can
be extended to a -homomorphism on A. It is again called .
5.14. Denition: A compact matrix quantum group is a pair (A, ) or shortly A where
(a) A is a unital C
-algebra generated by some elements u

ij
A for 1 i, j N and
some positive integer N,
(b) : A A

A is a unital -homomorphism such that (u
ij
) =
N
k=1
u
ik
u
kj
for
all i, j,
(c) the matrices u and u are invertible.
5.15. Remark: (a) Let / be a Hopf -algebra generated as unital algebra by matrix ele-
ments of one unitary corepresentation u or (equivalently) generated as unital -algebra
by matrix elements of a corepresentation v such that v and v are equivalent to uni-
tary corepresentations. Then the C
-algebra constructed as above is a compact matrix

quantum group.
(b) For all positive integers N the compact Hopf -algebra of SU
q
(N) gives rise to a
compact matrix quantum group.
(c) The general example of a compact matrix quantum group comes from C
-algebras A
as in (a) after dividing by closed two-sided ideals I x A: h(x
x) = 0 such that
induces a -homomorphism A/I A/I

A/I.
5.16. Theorem: Let A be a compact matrix quantum group constructed as in Re-
mark 5.15, part (c).
(a) Then [h(x)[ |x|
C
for all x /, therefore h can be extended to a (positive)
continuous functional on A, which will be denoted by h again.
(b) The algebra / is embedded into A (because for all x / 0 the inequality
h(x
x) > 0 holds).
(c) Any corepresentation of A (in the sense v
ab
=
c
v
ac
v
cb
, v
1
exists) has matrix
elements in / and thus / can be recovered from A as the span of matrix elements of
corepresentations.
5.17. Remark: For I
1
:= x A: h(x
x) = 0 (it is a closed two-sided ideal due to [W2,

p. 656]), h is faithful on A/I
1
(i. e. h(x
x) = 0 x = 0), while for I

2
:= 0, is
continuous on A/I
2

= A. In the case of SU
q
(2), I
1
and I
2
coincide, cf. [P3, Remark 6].
The notion of compact matrix quantum groups generalizes that of algebras of continuous
functions on compact groups of matrices. To be more precise: Let G be a compact group
of matrices. Then there is a Haar measure on G. There is an inner product on C(G)
given by
(, ) :=
_
G
d.
for , C(G). The algebra Poly(G) as in Denition 2.3 is a compact Hopf -algebra
(cf. proof of Lemma 2.4). The inner product as above can also be expressed as h(
).
Therefore the completion of Poly(G) with respect to the norm |.|
(.|.)
is the same as L
2
(G),
and the completion of Poly(G) with respect to the norm |.|
C
is the same as C(G). Here
the comultiplication : C(G) C(G)

C(G)

= C(GG) is given by ()(g, h) = (gh)
for all g, h G and C(G) (cf. Chapter 2). In the following, each compact topological
space is by denition a Hausdor space. There are one-to-one correspondences induced by
Gelfands theorem:
compact topological spaces X unital commutative C
-algebras C(X)
continuous mappings : X Y unital -homomorphisms
: C(Y ) C(X)
cartesian product X Y topological tensor product C(X)

C(Y )
compact group of matrices G compact matrix quantum group C(G)
for commutative / = Poly(G)
6. Actions on Quantum Spaces
Denition and spectral decomposition [P3, Section 1]
This chapter deals with a topological counterpart of right comodule algebras. Let V be
a topological vector space and Z V a subset. Then Z) denotes the closure of the linear
span of the elements of Z in V .
6.1. Denition: Let (A, ) be a compact matrix quantum group and B a unital C
-
algebra. The unital -homomorphism : B B

A is called a coaction for A on B
if
(a) ( id
A
) = (id
B
),
(b) B A = (id
B
y)(x) [ x B, y A).
6.2. Remark: (a) Let G be a compact group of matrices, X a compact topological space
and XG X, (x, g) xg for x X and g G, an action. Then there is a coaction
: C(X) C(XG) given by ()(x, g) = (xg) for all C(X), g G, x X. The
properties x(gh) = (xg)h and xe = x for all g, h G, x X correspond to Conditions
(a) and (b) in Denition 6.1 respectively. Given a coaction as in Denition 6.1 for
commutative A and B, the group action can be recovered by Gelfands theorem.
(b) Quantum analogues of left actions are considered in [P3, Remark 7].
6.3. Theorem: Let A be a compact matrix quantum group, B a unital C
-algebra and
a coaction. Then there exists a maximal -subalgebra B of B such that B is dense in B
and an / right comodule algebra, i. e. for := [
B
:
(B) B /, ( id) = (id ), (id ) = id.
For each 1 there is a set I
such that the algebra B has a basis e

rk
for 1, r I
,
1 k dim(u
) such that
(e
rk
) =
s
e
rs
u
sk
.
Idea of proof (cf. [P3, Theorem 1.5]): From the Peter-Weyl-Woronowicz relation (20) it
follows that there are elements x
sm
/ which span / such that the continuous linear
functionals
sm
: A C, x h(x
sm
x)
satisfy
sm
(u
kr
) =
,
sk
mr
. Then the operators
E
sm
= (id
B

sm
): B B
Chapter 6. Actions on Quantum Spaces 39
have properties of matrix units. The traces
s
E
ss
are projections onto subspaces W
B
which contain all elements x B such that
(x) B (
ik
Cu
ik
).
Construction of the basis: For each 1 let e
r1
[ r I
be a basis of the vector space

Im(E
11
) and e
rs
:= E
s1
(e
r1
). Let B denote the linear span of all elements e
rs
. Then
the closure of B is
E
sm
(x) [ x B, , s, m) = E
sm
(x) [ x B, , s, m) =
= (id h)(id x
sm
)(x) [ x B, , s, m) =
= (id h)(id y)(x) [ y A, x B)) =
= (id h)(B

A)) = B.
6.4. Denition: Let a compact matrix quantum group A coact by on a quantum
space B.
(a) For each 1, the number c
denotes the cardinality of I
as in Theorem 6.3 and is

called multiplicity of u
in the spectrum of .
(b) For each 1 let W
be the linear span of the elements e

rs
as in Theorem 6.3.
Quantum spheres [P1]
Since the quantum groups SU
q
(2) and SU
1/q
(2) are isomorphic by Theorem 3.2, we can
restrict ourselves to the case q [1, 1] 0. For the quantum SU(2) groups, 1 is the set
of non-negative half integers and u
k
= w
k
for k 1. We want to classify coactions of
SU
q
(2) such that
(1) c
k
=
_
1 if k N
0
0 if k N
0
+
1
2
,
(2) the subspaces W
0
and W
1
generate B as a C
-algebra.
The pairs (B, ) are called quantum spheres (cf. the case q = 1 in Theorem 6.5 below).
For convenience, the matrix elements of the unitary irreducible corepresentations of SU
q
(2)
will be indexed by numbers in the index set
N
:= , + 1, . . . ,
instead of the index set 1, . . . , 2 + 1 for each
1
2
N
0
.
6.5. Theorem [P1]: In the case q = 1 there is only one object B = C(S
2
) and the coaction
is induced by the standard right action of SU(2) on the sphere S
2
. Here W
0
= C1 and
W
1
= Cx +Cy +Cz. Then Condition (2) means that the coordinates x, y, z separate the
points of S
2
by the Stone-Weierstrass theorem.
In the case q = 1 there is only one object B
1,0
with coaction
1,0
.
In the case 1 < q < 1 and q ,= 0 there areup to isomorphismsthe following quantum
spaces B
qc
for c R
+
0
. The C
-algebra B
qc
is generated by the elements e
1
, e
0
, e
1
of W
1
subject to the relations
e
i
= e
i
for i 1, 0, 1,
(1 +q
2
)(e
1
e
1
+q
2
e
1
e
1
) +e
2
0
= 1,
e
0
e
1
q
2
e
1
e
0
= e
1
(1 +q
2
)(e
1
e
1
e
1
e
1
) + (1 q
2
)e
2
0
= e
0
,
e
1
e
0
q
2
e
0
e
1
= e
1
,
_
_
(23)
where
=
_
1 q
2
if c R
0 if c =
and =
_
(1 +q
2
)
2
q
2
c + 1 if c R
(1 +q
2
)
2
q
2
if c = .
The coaction
qc
is given by
(e
i
) =
1
j=1
e
j
u
1
ji
for i 1, 0, 1. Here we choose a non-unitary form
u
1
=
_
_
2
(q
2
+ 1) q
2
q
1
1 + (q +q
1
)
q
1
2
(q +q
1
)
2
_
_
.
Ideas of proof: Due to Theorem 6.3 and Condition (1), the algebra B has the linear basis
e
k
[ N
0
, k N
such that
(e
k
) =
sN
e
s
u
sk
for N
0
, k N
.
Therefore the e
k
s are analogues of spherical harmonics. One has (u
1
lk
)
= u
1
l,k
. Then
(e
k
) =
l
e
l
(u
1
l,k
)
l
e
l
u
1
lk
.
From the irreducibility of u
1
it follows that there is a constant c such that e
k
= ce
k
for
all k. Moreover the modulus of c is one because of e
k
= (e
k
)
= (ce
k
)
= c ce
k
. Thus it
is possible to achieve c = 1 by scaling the elements e
k
with a suitable complex number of
modulus one.
Now consider products of the generators: Because of the Clebsch-Gordan relation u
1
u
1

=
u
0
u
1
u
2
there are injective intertwiners G
Mor(u
, u
1
u
1
) for 0, 1, 2. From
the equation
(e
k
e
l
) =
m,r
e
m
e
r
u
1
mk
u
1
rl
it follows for the elements e
,t
:=
k,l
e
k
e
l
G
kl,t
:
( e
,t
) =
k,l,m,r
e
m
e
r
u
1
mk
u
1
rl
G
kl,t
=
n
_
m,r
e
m
e
r
G
mr,n
. .
= e
,n
_
u
nt
.
Therefore the elements e
,t
satisfy the same relations for the coaction as the elements e
k
.
Since the corepresentations u
are irreducible, there are constants
C such that
e
,t
=
e
,t
. For 0, 1 this gives relations for the generators:
k,l
e
k
e
l
G
1
rl,t
= e
t
(here =
1
),
k,l
e
k
e
l
G
0
rl,0
= 1 (here =
0
).
These are the relations (23) for the quantum spheres. Applying to both sides, we
obtain that and are real. There is still the freedom of scaling the e
k
s by a non-zero
real number. Consider the case 0 < [q[ < 1. If does not vanish, it can be scaled to the
value = 1 q
2
. Then dene c by
= (1 +q
2
)
2
q
2
c + 1.
The existence of a faithful C
-norm on B implies that c is a non-negative number. It

remains = 0, positive (B is a C
-algebra). Then can be scaled to the value (1+q

2
)
2
q
2
.
These (B, )s are indeed quantum spheres. No extra relation can be imposed, because
then we would get a coaction for a quantum subspace. But c
0
= 1 means that the space is
homogeneous (cf. [P3, Denition 1.8]), and from the facts that h is faithful (i. e. h(x
x) =
0 x = 0) and the counit is continuous (cf. Remark 5.17) it follows here that the
homogeneous space corresponding to B has no non-trivial homogeneous subspaces (this
idea stands behind the proof in the paper [P1]).
The case q = 1 can be handled similarly, and the case q = 1 reduces to q = 1.
6.6. Remark: (a) If the rst condition for the quantum spheres is weakened to c
0
= c
1
= 1,
there are some more homogeneous spaces for c c(2), c(3), . . ., 0 < [q[ < 1, where
c(n) = q
2n
/(1 +q
2n
)
2
for all n N.
These objects satisfy the conditions
c
k
=
_
1 if k = 0, 1, . . . , n 1
0 otherwise.
There exist analogues of these objects in the case q = 1. They correspond (cf. [P1])
to the adjoint action of SU(2) on U(su(2)) taken in its n-dimensional irreducible
-representation (X
= X for X su(2)).
(b) For 0 < [q[ < 1, c R
+
0
c(2), c(3), . . . the quantum sphere S
2
qc
= (B
qc
,
qc
)
is a quotient space if and only if c = 0, embeddable (i. e. can be regarded as a non-
zero C
-subalgebra of A, where is induced by the comultiplication) if and only if

c [0, ], and homogeneous for all considered c (for the compact groups of matrices
these three notions coincide).
(c) An algebraic version of Theorem 6.5 can be found in [S].
7. Quantum Lorentz groups (cf. [WZ2])
The algebra / = Poly(SL(2, C)) is called the algebra of polynomials on the Lorentz
group. Its corepresentations have the following properties (cf. Chapter 3):
(1) There are irreducible corepresentations w
for
1
2
N
0
such that all non-equivalent
irreducible corepresentations are w
for ,
1
2
N
0
.
(2) dim(w
) = 2 + 1 for all ,
(3) w

= w
||
w
||+1
w
+
(Clebsch Gordan),
(4) Each corepresentation is completely reducible, or equivalently, the matrix elements
w
ij
(w
kl
)
give a basis of /.
(5) For all ,
1
2
N
0
the corepresentations w
and w
are equivalent.
7.1. Denition: A quantum Lorentz group is a Hopf -algebra / satisfying properties
(1)(5).
7.2. Theorem: Up to isomorphisms, all quantum Lorentz groups / are given as follows:
The Hopf -algebra / is generated by the matrix elements w
ij
(1 i, j 2) of the
fundamental corepresentation w := w
1/2
and relations
(i) (w w)E = E,
(ii) E
(w w) = E
,
(iii) X(w w) = ( w w)X,
where the base eld C is canonically embedded into /, the vectors E
C
2
C
2
and
E
2
C
2
C are the same as in Theorem 3.2 and X M
4
(C) satises the properties:
(iv) X is invertible,
(v) there is a scalar factor c C 0 such that

X = cX,
(vi) the intertwiners 1
2
E and (X 1
2
)(1
2
X)(E 1
2
) in Mor( w, w w w) are
proportional (note that w

= w
0
w

= w w
0
).
Idea of proof: Necessity of relations: Restrict attention to the corepresentations w
rst.
Their matrix elements give a basis of a quantum SL(2)-group H as in Theorem 3.2. This
shows conditions (i) and (ii) and gives E and E
. From assertions (1) and (4) it follows

that there is a linear isomorphism
/

= H H

= HH
, w
kl
(w
mn
)
kl
(w
mn
)
kl

C
(w
mn
)
,
where denotes multiplication. Assertion (5) for = =
1
2
shows that there is a bijective
intertwiner X Mor(w w, w w), which gives conditions (iii) and (iv). Apply the map
to (iii) and use the formula v w = ( w v) as in the proof of Lemma 4.1, part (c):
X(w w) = ( w w)

X

X(w w) = ( w w)

X

X(w w) = ( w w)

X.
Chapter 7. Quantum Lorentz groups (cf. [WZ2]) 44
Since ww and w w are irreducible, the intertwiners X and

X must be proportional,
which gives Condition (v). The last condition follows, because both 1
2
1
2
and
X(1
2
1
2
E
)(1
2
X 1
2
)(E 1
2
1
2
)
are elements of Mor( w w).
Existence: We set / := H H
with H as in Theorem 3.2 and laborously introduce the

Hopf -algebra stucture on / by means of (iii)(vi).
Suciency of relations: More relations would make the elements w
ij
(w
kl
)
linearly depen-
dent.
7.3. Remark: (a) Possible matrices X have been found (up to isomorphisms of the cor-
responding Hopf -algebras) in [WZ2].
(b) There is also a topological structure for two examples of / ([PW1], [WZ1]) which uses
the notion of aliated elements [W5].
(c) Quantum Poincare groups arise by adding translations [PW2].
(d) Quantum analogues of Poly(SL(N, C)) were considered in [P2] (cf. [Z]).
8. References
[A] E. Abe, Hopf algebras, Cambridge University Press 1980
[D] J. Dixmier, C
-Algebras, North Holland 1977

[Dr] V.G. Drinfeld, Quantum groups, Proceedings ICM-1986, Berkeley, AMS, 1987,
pp. 798820
[DV] M. Dubois-Violette, G. Launer, The quantum group of a non-degenerate bilinear
form, Phys. Lett. B 245 (1990), 175-177
[H] T. Hayashi, Quantum deformation of classical groups, Publ. RIMS, Kyoto 28
(1992), 5781
[KP] P. Kondratowicz, P. Podles, On representation theory of quantum SL
q
(2) groups
at roots of unity, hep-th/9405079, Quantum groups and quantum spaces, R.
Budzy nski et al. (eds.), Banach Center Publications 40 (1997), 223248, Inst.
of Math., Polish Acad. Sci.
[Ko] T.H. Koornwinder, General compact quantum groups, a tutorial, hep-th/9401114,
in: Representations of Lie groups and quantum groups, V. Baldoni & M. A. Pi-
cardello (eds.), Pitman Res. Notes Math. Ser. 311, Longman Scientic & Techni-
cal, 1994, pp. 46128
[P1] P. Podles, Quantum Spheres, Lett. Math. Phys. 14 (1987), 193202
[P2] P. Podles, Complex quantum groups and their real representations, Publ. RIMS,
Kyoto 28 (1992), 709745
[P3] P. Podles, Symmetries of Quantum Spaces. Subgroups and Quotient Spaces of
Quantum SU(2) and SO(3) Groups, Commun. Math. Phys., 170 (1995), 120
[P4] P. Podles, The Dirac operator and gamma matrices for quantum Minkowski
spaces, q-alg/9703014, will appear in J. Math. Phys.
[PW] B. Parshall, J. P. Wang, Quantum linear groups, Mem. AMS 439, 1991, AMS
[PW1] P. Podles, S. L. Woronowicz, Quantum Deformation of Lorentz group, Commun.
Math. Phys. 130 (1990), 381431
[PW2] P. Podles, S. L. Woronowicz, On the classication of Quantum Poincare Groups,
Commun. Math. Phys. 178 (1996), 6182 and references therein
[R] M. Rosso, Alg`ebres enveloppantes quantiees, groupes quantiques compacts de
matrices et calcul dierentiel non commutatif, Duke Math. J. 61,1 (1990), 1140
[RTF] N. Yu. Reshetikhin, L. A. Takhtadzyan, L. D. Faddeev, Quantization of Lie
groups and Lie algebras, Leningrad Math. J. 1(1) (1990), 193225
[S] J. Apel, K. Schm udgen, Classication of three-dimensional covariant dierential
calculi on Podles quantum spheres and on related spaces, Lett. Math. Phys. 32
(1994), 2536
[T1] M. Takeuchi, Quantum orthogonal and symplectic groups and their embedding
Chapter 8. References 46
into quantum GL, Proc. Japan Acad., 65 (1989), Series A, No. 2, 5558
[T2] M. Takeuchi, Some Topics on GL
q
(n), J. Algebra 147 (1992), 379410
[W1] S. L. Woronowicz, Twisted SU(2) group. An example of a non-commutative
dierential calculus, Publ. RIMS, Kyoto 23 (1987), 117181
[W2] S. L. Woronowicz, Compact matrix pseudogroups, Commun. Math. Phys. 111
(1987), 613665
[W3] S. L. Woronowicz, Tannaka-Krein duality for compact matrix quantum groups.
Twisted SU(N) groups, Invent. Math. 93 (1988), 3576
[W4] S. L. Woronowicz, New quantum deformation of SL(2, C). Hopf algebra level,
Rep. Math. Phys. 30 (1991), 259269; cf. also Yu. I. Manin, Le cons College de
France (1989); E. E. Demidov, Yu. I. Manin, E. E. Mukhin, D. V. Zhdanovich,
Non-standard quantum deformations of GL(n) and constant solutions of the Yang-
Baxter equation, Prog. Theor. Phys. Suppl. 102 (1990), 203218; S. Zakrzewski,
A Hopf star-algebra of polynomials on the quantum SL(2, R) for a unitary R-
matrix, Lett. Math. Phys. 22 (1991), 287289; H. Ewen, O. Ogievetsky, J. Wess,
Quantum matrices in two dimensions, Lett. Math. Phys. 22 (1991), 297305
[W5] S. L. Woronowicz, Unbounded elements aliated with C
-algebras and non-

compact quantum groups, Commun. Math. Phys. 136 (1991), 399432; S. L.
Woronowicz, C
-algebras generated by unbounded elements, Rev. Math. Phys.,

Vol. 7, No. 3 (1995), 481521
[WZ1] S. L. Woronowicz, S. Zakrzewski, Quantum Lorentz group having Gauss decom-
position property, Publ. RIMS, Kyoto, 28 (1992), 809824
[WZ2] S. L. Woronowicz, S. Zakrzewski, Quantum deformations of the Lorentz group.
The Hopf -algebra level, Comp. Math. 90 (1994), 211243
[Z] S. Zakrzewski, Realications of complex quantum groups, in Groups and related
topics, R. Gielerak et al. (eds.), Kluwer Academic Publishers, 1992, pp. 83100

Introduction To Quantum Groups

Uploaded by

Introduction To Quantum Groups

Uploaded by

a

Supported by Graduiertenkolleg Mathematik im Bereich ihrer Wechselwirkung mit der

(g), (f)(g) := f(g), (ff

non-commutative algebra with

, k denote positive integers less or equal to N.

= V for all C-vector spaces is used). Let the

be induced by the multiplication

: H H be a C-linear map. Then the following are equivalent:

id)(f) = (f)1 for all f H,

(xy) = f(xy) = f(x, y) =

) satises the properties of the counit, both

, whence (B) looks like

[ J be sets of irreducible corepresen-

) can be computed using Schurs lemma (Lemma 2.20).

is a decomposition of into a minimal

) + 1. Therefore for all k

= (1 0 0 0). Therefore there are the relations

, considered as 4 1 oder 1 4 matrices, intertwine w w and

= 0 and if it has rank 0

which does not change the relations, one gets

). There are uniquely determined

are again 4 1 and 1 4

as in Denition-Lemma 3.5 and the symmetrization operator

are classical completely antisymmetric elements of (

C 0, there is a Hopf algebra automorphism of H

C 0, there is an antilinear, comultiplicative, algebra antiautomor-

is equivalent to up to a Hopf algebra automorphism (i. e.

) is a character, because C is commutative.

= B, where B is a diagonal matrix with diagonal elements

is also invariant and C = w[

is also a corepresentation and equivalent to a unitary

is a strictly positive denite

= Mor(w, v), Mor(v

, which yields by linearity and Equation (22) the

C. These constants can be evaluated using the unitarity of u

is positive denite. This proves Equation (19). The

can be scaled by a positive number, we normalize

). After normalization they are uniquely

x) 0, and equality only holds for x = 0.

are strictly positive denite for all .

/ such that for all b /

)) = 0 for all b /, whence a

is invertible, also is invert-

. Therefore for all x H and k dimu

-algebra by Denition-Lemma 5.11.

-algebras are useful:

-algebra. Then each -homomorphism from A to B is continuous

-algebras, dened as follows:

-algebra generated by some elements u

-algebra constructed as above is a compact matrix

x) = 0 (it is a closed two-sided ideal due to [W2,

x) = 0 x = 0), while for I

such that the algebra B has a basis e

be a basis of the vector space

denotes the cardinality of I

as in Theorem 6.3 and is

be the linear span of the elements e

are irreducible, there are constants

-norm on B implies that c is a non-negative number. It

-algebra). Then can be scaled to the value (1+q

-subalgebra of A, where is induced by the comultiplication) if and only if

. From assertions (1) and (4) it follows

with H as in Theorem 3.2 and laborously introduce the

-Algebras, North Holland 1977

-algebras and non-

-algebras generated by unbounded elements, Rev. Math. Phys.,

You might also like