Professional Documents
Culture Documents
Contents
I.1. A group G is a set of elements together with a multiplication law which asso-
ciates a third element to any pair of elements of G: (a, b ) -+ (ab) = c in such a
way that the following conditions hold:
i) associativity (ab)c= a(bc) = abc; a , b, c E G ; l )
ii) t,here exists an unit element e E G such that for any a E G : a e = ea = a ;
iii) for any a E G there exists an element a-l, called the inverse of a, such that:
aa-1 = a-1 a =e.
The subgroup G' of G is an invariant or normal subgroup if hgh-1 E Gf' for any
h E G and any 9 E 0'.
1.3. A mapping v of a group G, into another G, is called a homomorphism if :
a -+ p ( a ) , b +-~ ( b ) implies
a b - 9 (ab) = P'(a)v(b).
is called an onto honiomorphism if for any a' E G, there is an a E G, such that
v ( a ) = a'.
Let c' be the unit element of G,. The set K , of the elements of G, which are mapped
into e', is called the kernel of the homomorphism p. It is easily shown that K ,
is a normal subgroup of G,.
A one-to-one homomorphism of GI onto G, is called an isomorphism. I n this case
K , = {e,}2).
1.4.Let G be a subgroup of G, and a any fixed element of G. The set of all products
ah when h runs over the whole G', is called a right G' coset, indicated in what
follows as aG'. If two cosets aG', bG' have one element in common they are in
fact coincident. Therefore the whole group G' decomposes into disjoint C' right
cosets. I n the same way one can define left G' cosets.
When G' is a normal subgroup of G, then for any a E G: aG' = G'a. I n fact
the general element of aG' has the form:
ah h c G'
But aha-, = h', h' E G ' ; hence:
a h = h'a
that is any element of aG' is in G'a and conversely. Let G' be a normal subgroup
of G, and let us indicate with G I G the set of all distinct G' cosets in G. The set
GIG' equipped with the following multiplication law :
turns out to be a group, called the factor group of G with respect t o G'. Due t o
the fact that G' is a normal subgroup, ii is easy to see that the above introduced
multiplication law satJisfies1.1 i-iii. We observe that the unit element of G/c' is
e G'.
Given a homomorphism of a group G, onto a group G, we can form the factor
group G J K , because K , is a normal subgroup of GI. Observe that:
if and only if a and b belong to the same K , coset. I n fact let e2 be the unit element.
of G,: the previous condition is equivalent to:
e2 = p(a)p(b)-l = ~ ( a b - l )
1.5. Examples
RR'= 1 (1)
(RTis the transpose matrix, RZ = Rki).
For any R E 0, the correspondence R -+{Rik} is one-to-one; in terms of the
matrices, the product in 0, reduces t o the usual matrix product so that 0, is
isomorphic t o this matrix group. In what follows we will identify them.
It follows from (1) that det R = 41. The subset of the R's with determinant
+ 1, is by itself a group, called R,, and its elements are the proper rotations of E,.
R, is a normal subgroup of 0,, because for any element R E 0,, and for any
R' E R,:
det (RR'R-I) = -+I.
The factor group O,lR, has only two elements: the cosets el = R, and e2 =
= -I R,, where I is the unit matrix. Product rules are:
elel = e2e2 = e,
e1e2 = e2e1 = e2.
21"
282 G. DE FRANCESCHI
and L. MAIANI
b) X U,.
The set X U , of two diniensional unitary unimodular (determinant = 1) complex
matrices forms a group with respect t o the usual multiplication law for matrices.
The general form of a SU , matrix is :
GI = ( )
0 1
1 0 6,= (; -i) CT, 0
= (l -10)
and the unit rnatrix c0 as follows:
where a = (al,
a2,a,) and the real numbers ai(i = 0 , .. ,,3) satisfy:
a; + + +; = 1 .
01; 01; (41
U
Putting a = -in, n2 = 1, 1 2 0, n = - -
la 1
b y (4)we have:
01; + 12 = 1
so that we can set,:
e o 5 e 5 2n
O1o =
i= sin -
. e
2
so that :
U
e
= cos - - isin - n
e . CT .
2 2
n"=X.0
and observe that: *
det X = - /XI2
For any U E S U,, we define the transformation3) :
2, = U X U + .
3) If U = (Uik}
is a square complex matrix, we define:
- - -
u : ( U ) i , = (U<,<)
-
u+: ( u+)i,= ( U S J .
Group Theory and Unitary Symmetry Models 283
--
matrix U . Then from ( 7 ) we have:
-I--
c) s u3.
The set of 3;: 3 complex, unitary unimodular mat,rices also forms a group wit)h
respect to t,he usual matrix multiplication law, and this group is c,alled S U,.
d) The set of proper Lorentz transformations forms a group indicated as L.;”
which is isomorphic t’oa matrix group : the se.t of 4 y 4 real matrices ,1 such that :
=G
i) Alz’CJAl
being G t,he ma.t’rix:
ii) det ,l = 1 . 21
G= 6-;-; :) 0 0
0 --1
0
.Xi = 2 .likXh. + ai
k
(where A tE L<, and a is a four-vector) or simply:
S’= AX + a,.
Applying a transformation determined by t’he pa.ir (a,,L l ) , and then t,hc t r a w
formation (a,’?.I.’), we obt,ain a ne,w element, of P;”
(a‘,.l’)(a,.I)
= (a’ + L4’a,Ll’Lt).
With this multiplication rule P: is a group.
It is easy t,o see that the unit element of P,: is
(0, I ) ( I = unit element of L < )
and that the inverse of ( a , -1)is
(@, A)-1 = (- A,l-la,k 1 ) .
The subset
C(0,/I)}
is a subgroup of P;” isomorphic to L%,and the subset
{ ( a ,I ) }
is an a,beliari subgroup, isomorphic to the translations group in four-space. More-
over { (a,I ) } is an invariant subgroup.
T ( X , 1) = g@).
I n particular a closed curve g (x)will be reducible to the point f if it is reconciliable
with the constant function :
*) It must be observed that the notion of compactness of a topological group is intrinsic, and
can be given without referring t o a particular parametrization of the group, just as the in-
troduction of a topology in a group (also for this topic see [ I ] ) .
286 G. DE BRANCESCHI
and L. MAIANI
1.7. E x a m p l e s
We give here some examples t o illustrate the relevant concepts introduced in the
previous paragraph, as well as t o establish some useful results concerning the
groups which are of interest t o us.
Let us begin with R,. I n sect 1.5 we have identified any rotation by a unit vector
n and an angle 8 (0 5 8 2n).The drawback of this is that: R(8,n) = R(" dZ -
- 8 , - n) .We can instead obtain a one-to-one correspondence between rotations
and t.hree-dimensional vectors stemming from the origin of E,, with lenght less
than or equal to 52: to any rotation of an angle 8 (0 5 8 (= n)in the counter clock-
wise sense around a unit vector n, we associate the vector a = 8 . n(1a 1 5 n);
conversely given OL, 8 and n can be obtained as:
The end points of these vectors fill a sphere of radius z,and we note that the same
rotation corresponds t o points on the surface diametrically opposite. Hence it is
necessary t o identify those points in order t o preserve the one-to-one correspon-
dence property.
Since the set of parameters is a bounded closed connected subset of E,, R, is a
compact connected group. It is instead not simply connected, as can be seen if one
considers a curve connecting two diametrically opposed points on the surface of
the sphere. These two points correspond t o the same element of R,, so that the
curve is effectively closed, but is cannot be reduced t o a point.
Group Theory and Unitary Symmetry Models 287
Let us consider a rotation R (0, n).From 1.5 (9)it follows that the matrix R which
corresponds t o the rotation R (1.5a) has the form
1,3
Rii = (1 - cos8) ninj + 2k sin 66ikjnk + cos 0 (10)
where nk are the components of n, < i l k is the Levi-Civita tensor, and aiiis the Kro-
necker tensor.
This expression is equivalent to
0.m 1
R = R ( u ) = e o 7 l . z = 2 If j k ( n . Z ) k ='e"'z (11)
6 I%.
where the 3 x 3 matrices L'i are defined as follows:
0 0 0 / 0 0 1 0 -1 0
0 1 0 -1 0 0
This can be seen using in (11) the relations:
a, = J , ( R ,?).
Now consider the product R' . R" = R of two elements of R,5), and call respccti-
vcly a,p, S the parameters of R', R", R. The coefficients R,, of R are analytic
functions of a a n d p . Hence S, being an analytic function of three of such coef-
ficients, is an analytic function of tx and p.
This demonstrates, together with the fact that the parameters of R-I are obviously
analytic functions of those of R, that R, is a 3-dimensional Lie group.
I n an exactly analogous way it is possible, using ( 5 ) and the properties of Pauli
matrices, to write any S U , matrix as :
U = exp (.:
-a -11. )
cs = exp (-ia cs) (14)
0
0 ~ 0 ~ - - . 2n
a= ~
2
5, such that R',R', R E N .
388 and L. MAIAKI
G . DE FRANCESCHI
where PI, F , . ..P, are eight hermitian traceless independent matrices that are
listed, with their commutation and anticommutation rules in [3].
Formula (15) shows that S U , is a connected, %dimensional group and in fact it
may be shown that XU, is also simply-connected.
An example of a non compact Lie group is the proper Lorentz group (1.5d). If we
choose as parameters of an element of its matrix coefficients, the subset of El,
so obtained is not bounded; in fact, in the coefficients of matrices belonging to the
subgroup of special Lorentz transformations there appear expressions like :
2. Linear Spaces
Oik = 0 k = 1 . . .n
otherwise they will be said linearly dependent.
We say that L is n-dimensional if there exist n linearly independent elements,
+
whereas any n k vectors are always linearly dependent ( k 2 1). Any set of n
linearly independent vectors ei (i = 1, . . . n ) is called a basis for L, and we can
express in a unique way any vector x as a linear combination of them:
rL
x=2xiei.
i=l
A transformation T
x + X' = T(x) x E L, X' E L'
of the n-dimensional linear space L into the m-dimensional linear space L', is
called a linear operator if
T(x + y) = T(x) + T(y) T ( o t ~=) a T ( z ) .
so that
Hence the coordinates of the transformed vector I' are obtained from those of x,
by means of a ( m w n ) matrix T k z ,that uniquely represents the given t,ransfor-
mation in the bases {ei} and { e : } .
If in a linear space L, for any fixed n, there exist n linearly independent vectors,
then L is said t o be infinite dimensional.
2.3. A subset 1 of a linear space L, such that any linear combination of elements of
1 belongs to it, is called a subspace (or linear manifold) of L. L is said to be the
direct sum of the subspaces l,, I,, . . . if it happens that any vector x of L can be
expressed uniquely as a linear combination of vectors contained in I,, I,. . . We
will write:
L = l 1 @ 1 , @ 1,. . .
2.4. A correspondence of pairs of vectors of a complex linear space L into the com-
plex numbers :
x>Y + (X, Y)
290 G. DE FRANCESCHI
and L. MAIANI
is a scalar product in L.
We observe that in any finite-dimensional space it is always possible to define a
scalar product (this is not true in general for infinite dimensional spaces: the exj-
stence of a scalar product must be assumed as an additional hypothesis) : in fact if
xi are the coordinates of x in a fixed basis and yi are those of y, we define
and it is easy to see that all the previous conditions are satisfied.
Any linear space in which a scalar product can be defined is called a Hilbert space.
I n the infinite dimensiona1case the additional hypothesis of completeness is required
(see ~41).
With the aid of the scalar product it is possible to introduce the concept of lenght
of a vector x :
llxll = i(c4.
If now T is a linear operator which maps L into itself, then we will say that T is
bounded if there exists a positive number C such that for any vector x =+ 0 :
for any z, y, E L
when T = T+, T is called self adjoint or hermitian; when T f T = T Ti- = 1 , T
is said to be unitary.
3. Representation of Groups
3.2. E x a m p l e s
Consider again R,. We have seen that there is a one-to-one correspondence be-
tween rotation and 3 x 3 real orthogonal matrices with determinant equal to 1.
Those matrices are operators in a 3-dimensional real linear space, and obviously
this correspondence fullfills conditions ( i ) ,so that it is a representation of R,. I n
addition this is a one-to-one representation, i. e. a faithful one.
Let now consider the set L2 of the complex valued functions y ( x ) defined in E,,
such that
J IY ( X V d3X
exists. This is a vector space, and also a Hilbert space, with the scalar product
defined as
( Y , 9)= Y (2) s
91 (2)d 3 2 .
The Schrodinger equation for a particle of mass m in a given potential V ( x )is
[g v2 +(E -
I
V(2)) y ( z ) = 0
and for certain classes of potentials (for example Coulomb potential) there exist in
L2 solutions of ( 2 ) corresponding to bound states. I n this case let US call L E the
subset (c L2) of the solutions of ( 2 ) corresponding to the same eigenvalue E . LEip
obviously a linear space.
Define for any rotation R E R, the operator T ( R )in L2 as
( T ( R ) y ) ( x=
) y ' ( ~=
) y(R-'z).
If V ( x )is a central potential, i.e.
V(Z) = V(lsl)
then T ( R )maps L E into itself. I n fact let y ( z ) E LE then
[- V2
1
+ (E - V ( 2 ) ) y(R-la) = [ 2 l - V'2 + ( E - V(2'))
1 y(z') = 0
892 and L. & i A 4 1 ~ N I
G. DB FRANCESCIXI
and obviously the identity of R, is mapped into the unit operator. Hence we have
a representation of R, in LE,which is furthermore unitary.
3.3. a) E q u i v a l e n c e of r e p r e s e n t a t i o n s
Let T , and T , be two representations of a given group in the spaces L, and L,.
They are equivalent if there exists a one-to-one linear mapping A of L, onto L,
such that
TI(9)A = ATz(9)
for any g E G. In this case we will write T , T,. -
The set of all representations of G decomposes into classes of equivalent represen-
tations, and the fact that two equivalent representations are essentially the
same thing, permits us t o limit our study t o inequivalent representations.
b) Reducible representations.
A subspace 1 of L is said t o be invariant for a representation of G in L, if
T ( g )x E 1 when x E 1
for any g E G. (0and L are always invariant (trivial) subspaces). If a represen-
tation has no invariant subspaces other than 0 and L, it is said t o be irreducible.
We observe that if T ( g ) is a finite dimensional reducible representation of G
in L, and I is an invariant subspace, we can choose a basis in L such that I is
spanned by the first elements of the basis, while the remaining ones span a sub-
space 1'(L = 1 @ Z').
The matrices corresponding t o the operators T ( 9 ) in this basis are of the block
form
When Q(g) = 0, 1 and 1' are both invariant. I n this case we will say that T (g)
decomposes into the direct sum of T1( 9 ) and T,,( 9 ):
T = Ti @ Ti,.
Group Theory and Unitary Symmetry Models 893
I n general we will say that a representation T of G in L is decomposable if it is
possible to write L as the direct sum of invariant subspaces I,, I,, . . . SO that
implies
n=l, zw=o
which are impossible to be satisfied for any z if w +
0. Hence T ( z )is not de-
composable.
Furthermore a decomposable representation is not always completely reducible :
for example the representation
/ l z 0 o\
9 -+T ( 9 )
9-1 -+ T ( 9 9 = (T(s))-l= (T(9))f
so that if x E I , y E 1L
T(g-1)x E I
294 and L. MAIANI
G. D E FRANCESCEII
hence :
i.e.
If both T 1and T 1 i are irreducible, the theorem is proved. Otherwise there will
be invariant subspaces contained in I and (or) 1I. I n this case we repeat the above
arguments, decomposing T I and (or) T1I . Being L finite dimensional the process
must end after a finite number of steps leading t o a full reduction of T . A weaker
result holds in the infinite dimensional case, namely the unitarity of the represen-
tation guarantees only its decomposability.
4.1. I n this section we want t o show that in the case of Lie groups, the problem
of finding out the irreducible representations is essentially equivalent t o that
of finding finite sets of operators obeying certain commutation rules, or, more
Group Theory and Unitary Symmetry Models 395
T ( g )= T(ocl,.. .,a,) g E N .
It can be shown [5] that the operator functions T = T (al,. . ., a,) are analytic')
so that tJhere exist n operators Ik (they will be called infinitesimal generators),
defined as
I n the same way, from eq. (14), for the 2-dimensional representation of X U , we
find the generators
T ( f g )Y k - l ) = T ( j d T(g-l)x= T ( f )x = Y ( f )
i.e.
Y ( f )= T ( f g )y(g-'). (3)
') ..,mnj.
For this we mean that any matrix element T i j(a,,...,a,)is a n analytic function of (al,.
y(0,. ..O) =x
T ( 0 , .. . O ) = 1.
(8)
where the real numbers Cg, (structure constants) depend upon the derivatives
of Sbt evaluated in ocl = ocz = . . . = 0, i.e. they are independent from the
particular representation chosen. By virtue of (9), C;, satisfy
Consider now a real vector space is' of a dimension n equal to bhe dimension of
the Lie group B, and let {Ak} (k = 1 . . . n) be a basis in S. With the aid of C;, a
composition rule in S can be defined as follows:
[x,y1 = - bf,XI y, x: E
[x, [Y,zl] + [ z , [x,YI] + [Y, [ z , XI] = 0 (Jacobi identity)
which are analogous t o the usual properties of the commutator of two operators.
The vector space S equipped with the composition law just defined is called the
Lie algebra A ( G ) associated t o the group G. It seems that this definition depends
(through Cj,) upon the particular parametrization of N . However if we make an
analytic change of variables in N (i.e. if g E N and g = (a:, . . . a;), then
a; = &;(a,, . . . a,)
are invertible analytic functions in all the arguments), we obtain a set of new
structure constants CfFk,which are related to Cil through a non singular matrix
ail in the following way :
= 2 akqail ci](a-'),h
9,L s
298 G. DE FRANCESCHI
and L. MAIANI
x -+ A (x)
such that
i) A ( a x + Py) = & A ( % )+ /3A(y) (linearity)
ii) A ([x,Y1)= [ A (4,
A (Y)l
where now [ A , B ] means the commutator of A and B.
Wc then see that starting from a representation of G
9 + T(d
its infinitesimal generators can be thought as a representation of a basis in A ( G ) ,
which extends by linearity to a representation of A (G). The usefulness of intro-
ducing A ( G ) is that the converse is also essentially true, in a way to be explained
below.
Let then {Ak} be a basis in A ( G ) and let
l k +Ak
LAi, A k l = 2
h
C&Ah (12)
(13)
T ( 0 ,. . . O ) =3
The integrability conditions of (13), can be expressed in terms of the SAk(al, . . . oln)
and it can be shown ([l]cap. IX) that they are satisfied in a neighbourhood of
the point (0, . . . 0) due to the definition of Shk (see eq. (6))and to eq. (12),which
is the form that the integrability condition assumes in the point (0, . . . 0). Hence
(13) is solvable in a suitable neighbourhood N' of the point (0, . . . O), giving us a
correspondence between the elements g E G, contained in a neighbourhood of
the unit element, and linear operators T (9) in L.
It may be verified, in a rather cumbersome way, that if g,g',gg' E N', then
T ( g 9 ' ) = T ( g )T ( g ' ) ,
and for this we refer the reader t o [ 6 ] .
Group Theory and Unitary Symmetry Models 299
Let now g be an arbitrary element of G ; by a result quoted previously, if G is
connected, we can express g as a product of elements of N ’ :
g = gi - * * gk 91,92, * * .) gk E ”.
It would be tempting a t this point to define
obtaining in this way a representation of the full group in L (it is obvious that
this definition satisfies 3.1 (1)).However, if
g = 91 . . . g k = 9; . . . gf,, gi, 91 E N‘
we are not sure that
This gives t o G a structure of Lie group, and in addition G is compact and connec-
ted. This group is one dimensional and so is A ( G ) ; hence any operator A on a
lincar space L, determines a representation of A ( G ) .
The simplest example a t hand is the representation of A ( G ) over the one dimen-
sional complex linear space. Consider the linear operator ik (i.e. the operator
that multiplies by i k , ii real number). We take i k as the representative of the
generator of A ( G ) , and equation (13) reads
T ( e )= 1
T(B)= eih.8
where B belongs t o a suitable neighbourhood N of 8 = 0. Let now g ( 0 ) be an
element of G. There exists an integer n such that B/n belongs to this neighbour-
hood, and we can define:
T(g(6))= T(g(B/n))’l= e f k o .
However the element which corresponds to 0 = & n can be written (with a
suitable n) as
):(
g and g ()-:
300 G. DE FRANCESCHI
and L. MAIANI
The failure of this method is due to the fact that G is not simply connected. For
simply connected groups instead eq. (14)is fulfilled (see [7]).
I n any case given a connected Lie groups G, with a Lie algebra ,I (G), it is possible
in an essentially unique way, to construct a connected Lie group G, having the
same Lie algebra -l(G), which is in addition simply connected [ I ] .
From this it follows the existence in d and G of two neighbourhoods fi and N
of the unit elements which are in a one-to-one bicontinuous correspondence, in
such a way that if
s, L s . i E N
stf E
then
TW,) =T(i)
in a way that preserves the associativity of the multiplication law in 6/Kp.This
correspondence is then a representation of 8/Kp in L and also, being 6/KTiso-
morphic to G, a representation of G in L.
Summarizing we can say that :
i) if G is simply connected, a representation of A ( G ) in L determines uniquely
a representation of G ;
ii) if G is not simply connected, a representation of A ( G )in L determines a repre-
sentation of Q which reduces t o a representation of G if and only if the kernel
K , of the homomorphism 6 4 G is mapped into the identity operator. To
exemplify let us consider again the case of R, and SU,. We have seen that the
Group Theory and Unitary Symmetry Models 301
infinitesimal generators of X U , verify the same commutation rules as the genera-
tors of R,, i.e. they have the same Lie algebra.
I n addition XU, is simply connected, so that XU, is the covering group of R,.
I n the section 1.5b, the homomorphism of SU , onto R, has been proved, together
with the fact that the kernel K , is the set of the two matrices f 1. We will see
later that any irreducible representation of A (X U ) , is uniquely determined by an
integer or half integer non negative number j , in such a way that its dimension
is 2 j+ 1. When j is an integer the elements 1 are mapped into unity, SO
that these representations are in fact representations of R,.
4.4.Quite independently from Lie groups, we can define a real, n-dimensional
Lie algebra as a set 2’of elements such that:
i) Y is a real n-dimensional linear space ;
ii) there exists in dp a composition law indicated as x, y --f [x,y] linear in x and y,
antisymmetric, satisfying Jacobi’s identity (see sect. 4.3.).
As before we define a representation of Y in a linear space L to be a mapping
x + A (x) of the elements of Y into linear operators of L satisfying conditions:
a) A (ax + +
BY) = & A(4 PA (Y);
b) A ([x,y]) = commutator of A ( x ) and A (y) = [ A (x),A (y)].
where ad(x) is the linear mapping of Y into itself (qua vector space) defined as
ad (X)Y = [x, Y1.
Furthermore
ad ( a x + Bz) = oi ad (2) + /3 ad(z)
[ad (x),ad @)I = ad ([x,21)
by virtue of Jacobi identity. Hence the correspondence
z --f ad (2)
is in fact a representation of Y.
An invariant subspace for the regular representation is called an ideal 9:
YCJ
if
ad (x)y = [z, y] E 9 for any x E 2’
I n particular 9 is abelian when z c 9 y E 9 iniplies [x,y] = 0.
Presence or absence of ideals has extremely important consequences for the
structure of the Lie algebra it,self and Lie algebras are divided in three classes
accordingly :
302 G. DE FRANCESCHI
and L. MAIANI
Hence given 6, one has to identify all its central discrete subgroups, which is a
relatively simple task. Then by making the corresponding quotient groups, one
find all the required groups.
For example see the case of R, and XU,. & 1 is the only discrete central sub-
group of XU,, so that R, is the only non simply connected group having the
same Lie algebra as S U,.
5 . I. Let L and L' be two linear spaces, respectively of dimensions n and n'.
Consider the set L @ L' of all formal sumss)
If now g + T ( g )and (g --f T'(g) are two representations of the same group G
in L and L', a new representation of G in L @ L' (indicated as T @ T ' ) can
be defined as follows :
9 + T ( d 0T ' W
where
(T ( 9 ) O T' ( 9 ) ) I m > Im') =.S
Tsm (Y)
SS'
T's'm, ( 9 ) IS) IS')
and for an arbitrary vector I x) :
8) Where it is convenient we use for vectors the Dirac notation: lm) for em (sect. 2.2).
304 G. DE FRANCESCHI
and L. MAIANI
It is easy t o see that T @ T' is unitary when T and TI are; but even in the
case that T and T' are irreducible, T @ Tf is not so. However, according to the
general statements in section (3.4b-c) being T @ T' unitary, it will be fully
reducible: the space L @ L' can be written as a direct sum of invariant sub-
spaces lisa which transform according to irreducible representations Ti.a( 9 )g, :
L @ L' = @ Zi,=
%,a
T @ T' = @ T i , . i = I . . .K.
i,a
where the index r labels the basis vectors of li,&. Collecting all these vectors we
obtain an orthonormal basis in L @ L; hence there exists a unitary matrix
G(m,m ' ; r , i, a)
6.1. S c h u r ' s l e m m a
i)Let T , ( 9 )and 17, ( 9 )be two irreducible, inequivalent, finite dimensional represen-
tations of a group G in the linear spaces L, and L,.
*) I n general there will be several irreducible subspaces transforming according the same irre-
ducible representation. These subspaces are distinguished by the additional label a, whereas i
distinguishes between group of subspaces transforming according inequivalent representations.
Group Theory and Unitary Symmetry Models 305
Any linear operator A mapping L, into L,, such that
T,(g)A = AT,(g) for any y E G
is the null operator.
Proof: Let N , be the set of vectors in L, such that
A X= 0 x E NA.
Y RA
i.e. AX
T , ( g ) y= T , ( g ) A x = A T , ( g ) x
i.e. T,(g)y E RA when y E R,.
so that the theorem is proved if we show that any operator A' which commute=
with all the operators of an irreducible representation of G is a multiple of the unit
element (in fact if this is true we have A U-1 = 21 i.e. A = A U).
Any operator A' has a t least an eigenvector x +
0 belonging t o some eigen-
value A:
A'x = Ax.
306 and L. MAIANI
G. DE FRANCESCM
Let V be the linear manifold spanned by the vectors belonging to this eigenvnlue
( V $. 0). V is invariant under T , ( 9 )in fact :
X E v
implies
A’T,(g)x = T , ( g ) A ’ x = LT,(g)x
i.e.
T,(g)x E V .
But V + 0, and T , is irreducible. Hence it follows V = L,, i.e.
A‘ = LI.
As before, the theorem is true for any bounded operator A‘ commuting with the
operators T (9) of a n unitary irreducible representation of G in any Hilbert space.
6.2. We use the results of sect. 6.1 to determine the structure of an operator T,
which is invariant under an arbitrary unitary representation of a group G.
I n particular we will determine the form of its matrix elements, with respect
t o a fixed basis.
To be definite let g --f U ( g ) be an unitary representation of G in a Hilbert
space X.We require it t o be completely reducible, which is always the case
when G is compact or finite. Let T be an operator mapping X into itself such that :
T U ( g ) = U ( g ) T forany g E G. (1)
3” can be decomposed (sect. 5.1) into a direct sum of invariant irreducible sub-
spaces :
311 = 0(0
lid = 0 4,. (2).
a i i a
( W e )(9) - U(iB)( 9 ) ) .
@ia
i.e.
(q,
F@P)= % ( ia/?)
, dijd,,. (7)
6.3 W i g n e r - E c k a r t t h e o r e m
This theorem, which is valid for any compact group, can be seen as a generali-
zation of the preceeding statements on matrix elements of invariant operators.
Let g --f U ( 9 ) be an unitary representation of G into the Hilbert space X . Suppose
we have a finite number of operators T k such that
(Ti,@
: @{@)= 2 C ( i r , l k ; j s y ) ( i a ~ ~ T ~ ~ ~ j ~ ) y (9)
Y
“Lj =
c + 1 , 0 , -1
llz, -‘/z
j=i
i=
Consider a set of two operators TYz(i = 5’1,) transforming as the j = 1/2
representation, i.e.
U’IZ @ U’I. 1 u1 @ UQ
~1 0 i j i l a= TJ~I/. 0 ~ J ~ L
Group Theory and Unitary Symmetry Models 309
Hence when j = 1 the matrix elements vanish unless j' = ' I 2 and viceversa.
Non vanishing matrix elements are :
(1 m' 1 T;z/ 1 l z r n > = (
72;:;I(1, '1%)
('I2 m' I TP I 1m) = Czf;$ I 1).
I
0 0 0 0 0 0 - 0
O O F F
0 0 0 0 0 -1 O S -1 0 0 0 0 I
0-- P
1/3
0 0 0 'I2 / 3l p0 0 0 0
ll21 l12
/; P
--
0 0 - -p 0 0
I
rn: 1 0-1 Y 2 -'I2
j: 1 '12
I n this case, using formulas of sect. 6.2, we find the following structure:
-
1 0 -1
1
7.1. In the development of Quantum Physics increasing attention has been given,
at various stages, t o invariance principles. Particularly, in the physics of elemen-
tary particles the systematic inquiry and exploitation of the 40 called internal
310 G. DE FRANCESCHI
and L. M~IANI
symmetries led, in the last few years, t o remarkable successes in contrast to the
unsatisfactory status of dynamical calculations. It appears that, due t o peculiar
features of the quantum mechanical description, the theory of group represen-
tations provides the natural device to handle such invariance principles.
Physical laws express correlations among observable events. The latter of course
take place in space and time so that t o specify any of them we need certain
coordinates with respect to a fixed space-time frame of reference. I n addition we
need some other “internal parameters” (such as charge, baryonic number etc.)
which uniquely determine the nature of the objects under consideration.
Physical laws are then functional relations among sets of such “coordinates”.
Loosely speaking a symmetry is a transformation on the coordinates which
leaves invariant these relations.
There are invariance principles which we believe to be valid for any kind of
phenomena, and these are: invariance under translations in space and time and
spatial rotations. AS stressed by WIGNER[ 8 ] , the mere possibility of comparing
results of experiments made in different places and a t different times (i.e. the
reproducibility of phenomena) is based on this assumption.
At the same level of universality we accept the assumption that the physical
laws are the same in all frames of reference differing for an uniform rectilinear
motion. All this is summarized in the statement that Physics is invariant under
inhomogeneous proper Lorentz transformations.
There are many other symmetries in elementary particle physics which are shared
only by certain kinds of processes (as for example is the case for the SU, or iso-
topic spin symmetry which is valid only for strong interactions). We postpone
the study of these topics to a brief sketch of TVigner’s analysis of relativistic
invariance in quantum mechanics.
7 . 2 . I n the formalism of quantum mechanics, there is a normalized vector y in
a Hilbert space X , corresponding t o any physical situation we can set up in
laboratory. The normalization condition determines y only up t o a phase factor,
so that what is really relevant is a set Y of vectors different from one another
by phase factors. Y is called a unit i n X . There is in addition a self-adjoint opera-
tor A corresponding to any observable quantity a and the connection between
theory and experiments is contained in the statement that the average value
of u in the situation Y is
The existence of !Pfor any Y and g is equivalent t o what is called the homogeneity
of Minkowski space-time, and combined with i) makes us sure that all physical
operations possible in a given frame, are possible in any other frame connected
to it by a Lorentz transformation. Condition ii) tells us that the connections
between any two states depend only upon their relative motion or position.
Finally iii) merely expressed the fact that P< is a group and that two slightly
different transformations must produce nearly the same effect. From i) and iii)
it follows that Y and Y g belong to the same coherent sector.
From i)-ii) it can be shown [11,121 that there is a unitary operator U ( g ) corres-
ponding to each g E P<, such that
U ( g ) y e 1Vs when Y E Y.
U (9)is defined up to a factor of modulus one, in that the substitution U (9) --f U'(g)
= w(g) U ( g )( l w ( g ) 1 = 1) gives us another set of admissible operators.
However there exist a neighborhood N of the unit element in P< and a particular
choice of these phase factors such that
a) g E N , g +- U ( g ) is a continuous mapping;
b) g,g', g - g' E N implies g . g '+ U ( g g ' ) = U ( g ) U ( g ' ) ,
are the total angular momentum operators of the system. P, and M,, are a
representation of the Lie algebra of P< . Their commutation relations are listed
in [9].I n particular
[P,,, Po] = [Ji,Po] = 0
23 Zeitschrift ,,Fortschritte der Physik". Heft 7
312 G . DE FRANCESCHI
and L. MAIANI
which leave Y invariant (here the labels a,/3refer only to internal degrees of
freedom such as charge, baryonic number, hypercharge etc. whereas space-time
labels are neglected).
10) Due to the fact that P, P, commutes with all the infinitesimal generators, in any irre-
fi
ducible representation is must be a multiple of the unit operator.
Group Theory and Unitary Symmetry Models 3 13
Such transformations are induced on the fields by unitary operators U (9) (g E G)
which are a representation of (2 in the Hilbert space of the strongly interacting
particles. The infinitesimal generators multiplied by -i are n selfadjoint opera-
tors Q k . Their expression in terms of the fields can be found by requiring that the
infinitesimal transformations leave unchanged the Lagrangian. I n fact one can
construct in terms of the fields n divergence-free currents
Ji(x) Ic = 1,. . .n
a q ( X ) =0
such that the operators
J^ Ji (z)d3z
satisfy the cornmutatmionrelations characterizing the Lie algebra of G, and are just
the infinitesimal generators of the transformation (1). From ( 2 ) it follows that
such operators are constant of motion (this result is just the quantum counter-
part of the classical Noether’s theorem [ l a ] ) .
Consider now one particle states. They are obtained by applying t o the vacuum
state a creation operator a: (again space-time labels are omitted) which obviously
satisfies
u a ; u-1 = C t a B a ;
B
so that
[J (a: I 0)) = 2 t a p @ i I 0))
B
i.e. one particle states transform like a representation of G . Moreover, since the
& i s are constant operators, the Hamiltonian which is the time displacements
generator, commutes with them, being therefore an invariant operator under G ;
the same applies to the mass operator. If G is compact, then the representation
of G over one particle states is completely reducible and the irreducible compo-
nents correspond to states of particles with the same mass.
We can simultaneously diagonalize a number of Qk equal t o the rank of G, then
we find multiplets of particles with equal masses, distinguished (apart from
possible degeneracies11) by the eigenvalues of the diagonal Qk’S, i.e. by certain
“internal” quantum numbers.
Passing t o multiparticle states we observe that they transform under G as follows
(1,2 . . . take into account space time as well as other degrees of freedom whicb
are uneffected by G ) i.e. as tensor product of one particle representations.
Suppose
I q:, 1); I z,; 2 ) ; . . .
to be one particle states which are eigenstates of &k with eigenvalues q!, qt, . . .
Then, under the unitary transformation u = (1 +
ieQk), the multiparticle
23*
314 G. DE BRANCESCHI
and L. MAIANI
state
I q!, 1) I YE, 2 ) * - * (4)
transforms as
l2) This group called U , has as elements the complex number eia (m mod 2 n), with the multi-
plication law:
eta e@ = ei ( a + B) .
As for any Abelian group, its irreducible representations are one dimensional: in fact let
eia -+ U ( a ) be a n irreducible representation over a linear space L ; then for any @
Group Theory and Unitary Symmetry Models 315
The corresponding conserved current is the charge-density current, and point a)
expresses charge conservation. However being this group Abelian, its irreducible
representations are 1-dimensional, so that each resulting multiplet contains only
one particle. I n this case there are no consequences other than mere charge con-
servation.
Conservation of any additive charge-like quantity (e.g. baryonic number or
hypercharge) can be obtained in this way.
7.4. On the basis of previous considerations it should be clear how things actually
go. Strong interactions display additive conservation laws, and moreover the
great variety of strongly interacting particles (hadrons) seems t o divide naturally
into sets of particles with very analoguos properties (for example x+,x-,no;K+,Ko;
p, n and so on). One takes these experimental facts as an indication that the
underlying dynamics possesses a non abelian symmetry group. Then one tries
with some Lie group and compares with experiments the relations which can be
found in the way outlined before.
Imagine a world in which only proton, neutron, x+, x-, no,are present, as was
the situation a t the time when the isotopic spin was introduced.
The mass spectrum of these particles provides a very clear evidence for the
existence of a non Abelian symmetry group, whose representation on one particle
states splits up into two components, the pion and the nucleon. Inside each
multiplet, particles are distinguished by only one quantum number (charge),
so that we are led t o a non Abelian Lie group of rank one. As we will see later
there is only one compact simple group of this kind, i.e. XU,. I t s Lie algebra
is spanned by three elements T,, T 2 ,T,, satisfying the product rules of angular
momentum, and its irreducible representations are labeled by a number T Z 0
such that:
i) 3 T is an integer :
ii) the dimension of the representation is 2 T + 1;
iii) the spectrum of T , consists of numbers T , T - 1, ..., - T .
We have then to assign the nucleon t o the T = and the pion to the T = 1
representations. T is called isotopic spin (I-spin) [15].
If we assuae, as a convention, that particles correspond to eigenstates of T,.
then T , has t o be connected to the charge operator. I n fact the following relation
holds for pions and nucleons :
1
Q - -
2
N = T, (N = baryonic number). (5)
I n this context we have two additive conservation laws: charge and baryonic
number. The latter is derived as invariance under a gauge group, whereas charge
conservation is included in isotopic spin conservation. Note that TIand T , do
not commute neither with T, nor with charge, so that they do not correspond
to observable quantities.
where A3 and A' depend only upon space-time labels. All the other matrix ele-
ments vanish. I n conclusion we can express the amplitudes of all processes like
(6) in terms of only two amplitudes which are function of space-time variables,
but do not depend anymore on the charge variables.
It is a well known fact, that experimentally a t an energy near 190 MeV for the
incident, pion the amplitude A 3 greatly dominates: neglecting A' we find a t
that energy a well determined ratio for the following processes:
a) x+ + p 4 x + + p
b) x- + p +x- + p
c) x - + p + x o + n .
Rate a : Rate b : Rate c = 9 : 1:2 which is well verified experimentally.
Since isotopic spin has been introduced many other hadrons have been found,
together with another additive conservation law : hypercharge ( Y ) conservation.
However all hadrons still fit well into isomultiplets when relation ( 5 ) is modi-
fied as
and all the experimental findings are consistent with the assumption that strong
interactions are invariant under XU, [ 1 6 ] .I n this context, as for baryonic number,
hypercharge conpervation is considered to derive from invariance under a gauge
group, quite independently from isotopic spin conservation.
Group Theory and Unitary Symmetry Models 317
This situation can be summarized stating that the symmetry group for strong
interactions is the direct product13) of two gauge groups ( N and Y-conservation)
times XU, (I-spin conservation). As a consequence N , (T, T,)and Y quantum
numbers appear in a completely uncorrelated way.
7.5. U n i t a r y s y m m e t r y m o d e l s
I n unitary symmetry models one tries t o derive T, and Y conservation from in-
variance under a group which does not break into the direct product XU, x U , ( Y ) ,
introducing in this way relations between particles belonging t o isomultiplets with
different hypercharge. It is a fact that as yet nobody has succeeded in extending
such procedure to include N-conservation which, as before, is derived from a
separated gauge group [.271.
The feeling for such a higher symmetry is not strongly substantiated a t first sight
by experimental evidence. I n fact according t o results derived in sect. 7.3) par-
ticles would be organized into supermultiplets, (i.e. irreducible representations)
behaving as elementary objects under strong interactions ; but now inside same
supermultiplets there would appear particles differing by Y as well as by T,.
This is in conflict with the experimental evidence in that Arnlm between particles
differing by Y are quite large and not imputable t o non strong interactions (for
example mA - mN z 175 MeV, m E - m, 380 MeV). Hence we must conclude
that the idea of a higher symmetry in the sense above specified, cannot be literally
applied.
Nevertheless the following interpretation has been proposed : there is a symmetrical
component in strong interactions which is responsible of the gross structure of
particles world; in addition there is a weaker component to be treated as a per-
turbation responsible of the departures from the exact symmetry. It is understood
that both components are charge independent as well as strangeness conserving.
13) Given two groups G, and G,, their direct product G, x G, is defined as the set of ordered
pairs (g, 9,) (9, E Gi) with the multiplication law
and it can be shown that all the representations of G = G, x G, can be put into this form.
U(g,,g,) is irreducible if and only if U(gl) and U ( g z )are. I n our particular case we have the
group
xu,X U,(N) X Ul(Y)
It should be noted that the weaker component has not been till now satisfactory
identified. However the idea of a simmetry breaking interaction treated as first
order perturbation, has provided us with corrections to the predictions derived
from a pure symmetrical model, which are consistent with experimental findings.
7.6. Previous reasoning obviously do not indicate us what the symmetry group
tor strong interactions actually is. Following general requirements however seem
to be quite reasonable, and are usually imposed on possible candidates:
i) this group must be a Lie group. I n fact we want t o identify additive conserved
quantities such as T,and Y with its infinitesimal generators;
ii) is must be compact: this assures that its irreducible unitary representations
are finite-dimensional, so that wen can fill up resulting supermultiplets with a
finite number of particles (see sect. 2.2);
iii) it must be semi-simple (see sect. 4.4).This restriction is mainly due t o practical
reasons: for semisimple Lie groups in fact there is a complete mathematical theory,
which is not the case non for semisimple groups;
iv) the rank of the group, i.e. the rank of its Lie algebra, must be two, because we
require two conserved commuting quantities i.e. T,and Y ;
v) is must contain a subgroup isomorphic t o XU, in order to recover the isotopic
spin symmetry. Actually this does not bear any restriction in that any semisimple
Lie group has this property (see later sect. 9.7 b).
To construct a concrete theory we need a t this point a characterization of Lie
groups as well as a classification of their irreducible representa,tions.
I n next sections we will study these topics with some detail.
I n sect. 4 we studied the relations between Lie groups and Lie algebras, and the
conclusion achieved was that there is a one-to-one correspondence between Lie
algebras and Lie groups, so that instead of studying Lie groups one can study the
corresponding Lie algebras and their representations, which is more convenient
by a mathematical point of view.
8.1. We have given in sect. 4.4 the definition of n-dimensional Lie algebra of
rank r. At the same time we noted that the mapping
x-tadx xcY
8.2. C o m p l e x i f i c a t i o n
then
where Cih are the structure constants relative to the basis choosen in P.This pro-
duct satisfies condition 4.4ii), and obviously this definition of Yc does not depend
upon the particular basis choosen. I n I,the form (x,y) is simply
=+
This form is not degenerate if and only if det(x,, xk) 0 ; this condition is the
same whether we consider Y or Y,,so that if I is semisimple so is Y c .I n addition
the rank of Yc is the same of that of Y .
8.3. S t r u c t u r e t h e o r y
We will be, content here t o state without proof all results relevant for applications.
A complete derivation can be found in the excellent book by JACOBSON
[18]14).
An easier treatment can be found in [19]. See in addition the very readable paper by
14)
RACAH[ZO].
320 and L. MAIANI
G. DE FRANCESCHI
Y=Y,Of,
n
where the direct sum runs over all non vanishing roots and Y ois the eigenspace
belonging to the root (0, . . ., 0). Obviously Yo contains V, but a stronger
result holds, namely
111. Po = 9 , (3)
and furthermore each Y , (01 +
0) is one dimensional.
It follows the there are n - r non zero roots.
We can extend the basis {hi}to a basis in P,adding t o these elements all the vec-
tors e, where 01 is a non zero root and e, is a vector spanning Y",.
IV. Consider the restriction of the trace form (1) to V. It is determined by the matrix :
Hence
ai =git = 2 (ay2 0 .
0
Suppose that for some value of i , Ai = 0. Then a{ = 0 for any a,so that
[hi, e,] = 0 .
Moreover
[hi, h;] = 0 ,
SOthat adhi is represented in the basis (hi, ea) by the null matrix.
This has the consequence
(h:,s)= 0 for any x E Y
i.e. (by Cartan's criterion) 4 = 0 which is excluded.
I n conclusion we see that gii is a non singular, positive definite, real matrix. We will
indicate with g'j its inverse
gikgki = dij.
k
With the aid of this metric tensor we define a scalar product between roots:
The difference between ai= 2 gijai and atcan be removed by performing a. real
linear transformation on hi's, which reduces gijinto the form Bij. I n this basis the
scalar product between roots is written as
B) = cmipi.
i
In the basis (hi,e,) part of the multiplication rules are defined in terms of roots
[hi, hi] = 0
[hi,e,] = aie, .
We will see that the same applies to all multiplication rules between hi and e,,
in the sense that roots determine all the structure constant relative t o the basis
{hi,e,}. This depends on peculiar properties of roots, which are the very heart of
structure theory.
332 G. DE FRANCESCHI
and L. MAIANI
Properties of roots.
I n the following k n and N + 1 are defined as
k a = (ka,,. . ., ka,)
01 +1 = (a1 + P I . . . ., + 14.
&T
Being in addition
-r 5 q - r 5q (-r' 5 q' - r' 5 q')
we obtain that if 01 and 1 are non zero root's, then
are non zero roots. The first (second) 0s obtained by reflecting ( L Y )with respect to
the plane orthogonal t o LY (p).
The general feature of the multiplication table can be understood in terms of the
following relation :
We note that this normalization determines e, and e-, up to factors d,, d-, such
that dad-, = 1.
c) E + + +
j3 0, 01 j3 is a root. By eq. (9) and by unidimensionality of Ye+@it
follows :
[ea, epl = Nage,+,*
One can show that
N,,p = N,,-,-p = N-a-p,a = -Npa (11)
and in addition, by disposing of factors d,, d-, in e,, e-a, one can assume
- - Xa,+.
N a, -
I n this case
VIII .
where q, and r are the integers determining the g-string containing /3. We see that
+
when a @ is a root, N , , 0. =+
This relation determines Nabup to a sign which must be chosen so t o satisfy ( l l ) ,
(12).
Collecting all these results, we can write the complete multiplication table relative
.to the basis {hi,e,, e-,}:
[hi,hi] = 0
[hi,e+al = fa i e i a
[e,, e-,I = 2 &hi = 2 aih&
j ii
e,I =
(a+ -,!I,a + j3 is not a root) = 0
(a$. -is, a + j3
Lea,
is a root) = Nape,,,.
A l l structure constants relative to this basis are determined by roots and are all
real. I n addition a linear non singular transformation on the his does not change
the form of these products. I n fact if we have
where
= 2jAi7aj; OC" == 2 a'(A-')il,
a
so that the scalar product between roots is dependent upon the particular ba,sis
choosen in %?.From this it follows that although there are in %? bases in which N ~ S
are complex numbers, in any case (a,a) is a positive number.
By (14) we see that the roots of a Lie algebra, determine uniquely its structure.
Hence a classification of semisimple Lie algebras of rank r is equivalente t o find all
sets of r-dimensional real vectors which satisfy V, VI, (7). This is the argument of
next section.
8.4. To begin with we introduce now an ordering between roots in the following
way.
Given two roots a and ,B, a is said to be greater than ,8 if the first non vanishing
component of a - ,8 _= (ai - Pi) is a positive number. I n particular a root is
positive if it is greater than zero.
Of course this ordering depends on the basis choosen in %? and it is the same that
the ordering of words in a dictionary.
We introduce another useful concept. A root 01 is simple if:
a ) a is a positive root,
b) a cannot be written as sum of two positive roots.
Two important properties of simple roots are nearly immediate.
If a and ,8 are simple roots, then:
i) a - 0 is not a root. If a - ,8 were a positive root, then a would be equal to
(a - 8) + 8, i.e. would not be simple. Conversely if a - 8 were a negative root,
+
then ,8 - a would be positive, and ,8 would be equal to (,8 - a) a , i.e. 0 would
not be simple;
ii) (a,8) 5 0. By i) if we consider the a-string containing ,!?,we see that r = 0 , SO
that
-1 -1 120° 1
-1 -2 135" 2
-1 -3 150" 3
-2 -1 135" '/z
-3 -1 150" 'I3
0 0 goo arbitrary
I n addition in the latter case ((rob = 90") neither a - 8, nor (x i3 are roots (see +
properties 8.4i, ji).
8.5. On the basis of stated properties of simple roots, we illustrate the classi-
fication of (complex) semisimple Lie algebras.
Let us first examine the case in which simple roots split up into groups of simple
roots, such that any root of each group is orthogonal to all roots belonging to
different groups, whereas inside each group there is no root orthogonal to all the
other ones:
am,. . . a ( % ) ; p,. . . p.'; y w , , , . y'rm'
r1 + r2 + --.+ r, =r.
The whole root diagram splits up into mutually orthogonal parts CL, . . . , 8. . . : . . . :
y , . . . and in addition 0: & j3 . . . ; . . . 01 y , . . .; p y . . . are not roots, so that
for the corresponding e , , . . .; e p , . . . ; ey . . . we have
Summarizing we see that the basis {hie,,e-,} can be decomposed into the direct
sum of bases {hL1),ear e-a}, {hi2),ep, e+), . . . { h r ,e,, e-y} such that all products
between elements of different groups vanish.
Each linear manifold spanned by such bases is evidently a subalgebra in 2' and it
is even an ideal, so that the existence of roots orthogonal t o all the others implies Y
to be not simple. The converse is also true, i.e.
326 G. DE FRANGESCHI
and L. MAIANI
C l a s s i f i c a t i o n of s i m p l e L i e a l g e b r a s
AS it should be clear from last section, the problem of determining all simple Lie
algebras of a fixed rank r , is equivalent to finding all sets of r simple roots satis-
fying V, VI, (7), and the condition, that no one of them is orthogonal to all the
others.
The essential results of the structure theory can be formulated as following.
XII. Lenght of simple roots can assume a t most two values.
Keeping this in mind, the set of simple roots of a simple Lie algebra can be con-
veniently described in a graphical way introduced by E. B. DYNXIN:
( I ) t o any simple root we associate a circle :
(2) two circles are connected by one, two, or three lines when the angle between
corresponding roots is respectively 120°, 135",or 150".
If the roots are orthogonal, circles are not connected.
(3) Circles corresponding to shorter roots are blackened.
The only simple algebras are then defined by following diagrams (for any fixed r ) :
Nome of the
algebra
Dynkin diagram Dimensionality
1 Remarks
I n particular there are only three distinct Lie algebras of rank two, i.e. A,,
C2 = B,, G,.
8.6. C l a s s i f i c a t i o n of s i m p l e , c o m p a c t Lie g r o u p s
As said in sect. 7.5 these groups are of the main concern in unitary symmetry
models. To any of them we can uniquely associate a real Lie algebra Tr, and it is
remarkable that, as showed by H. WEYL,the compactness of the group reflects in
T, in that its trace form (1) is negative definite. I n view of this circumstance, Y,.itself
is called a compact (real) Lie algebra.
To carry out structure theory it has been convenient to consider complex Lie al-
gebras. Now, whereas any real algebra Y , uniquely define its complex extension
Y c ,the converse is not true, in that the same -Y, can be obtained starting from dif-
ferent real Lie algebras i.e. from different Lie groups (this is for example the case
of R, and of the 3-dimensional Lorentz group, which have the same complex Lie
algebra Al).
However, as again has been showed by H. WEYL,for any semisimple complex Lie
algebra -YC.there is essentially one real semisimple compact Lie algebra whose com-
plex extension is Yc. This has the meaning that in Ye there exists a basis such that:
i) all its structure constants are real;
ii) the real Lie algebra spanned by this basis is compact.
I n particular starting from the canonical basis {hi,ear e-a} it can be easily shown
that the basis
f.-
- - .$ h i ; fa = -. + e-J; ga = - (e, - e-J (16)
( a runs over positive roots)
is c o ~ n p a c t l ~ ) .
8.7. E x a m p l e s
a ) xu,
We have seen in sect. 4.1 that the real Lie algebra associated t'o SU2(R,) is span-
ned by three elements I,: I,, I , with the product rules:
[Il,1 2 1 = I , > [I2,131 = I,, [I3,I,l =I,.
15) To verify that (5, z) < O ( z =+ 0)
whenever z is a real linear combination of the elements
(16), one has simply to use the orthogonality relations (ki,e,) = 0,(ern,ep) = 6-, , 0 which
follow from (1') and from the normalization condition (e,x, e - a ) = 1.
This is a 3-dimensional simple Lie algebra of rank 1. For the general element
x = C ei I t (ci real) it is easy to see that Tr (adx adz) < 0 as it must be, being
i
SLT2compact. Let us choose I , as the element spanning V. Finding roots is equi-
valent to solve the following eigenvalue equation (with a + 0 ):
adI,(z,) = [I,, XJ = ax,, x, = clIl + c 2 1 , + c31,.
which is equivalent to :
so that we have
1 - I I 3
[e;, el_] = - h, = -- h,
2 Le,, e - ] 2
being (e!+, e l ) = 1 as one can easily verify.
Root space is one-dimensional, and we have two non zero roots: -& 1, and one
simple root. The corresponding Dynkin diagram is made of a single circle, so that
this Lie algebra is just A,.
Usually as basis are taken the elements
Hence h:, e l . span a subalgebra in Y which is identical with A,, so that by the
considerations made in sect. 4.4we see that the compact group associated to any
Y contains subgroups isomorphic to XU,.
c) A ,
We start from the Dynkin diagram:
From this we see that there are two simple roots dl)and d2)of equal lenghts,
making, an angle of 120 degrees:
Hence the a(,) string containing d 2 ) consists of the two elements @), a(,) d 2 ) +
+
and the reversed string contains dl)and dl) a(,). A , is 8-dimensional and has
rank 2 , so that we expect six non zero roots a t all: in fact they are
t t
I
a) bi
Fig. 1. a) Root diagramin of S U s b) Root diagram of C ,
I n order to construct explicitly our Lie algebra, we evaluate now covariant com-
ponents of a's ( N ~in
) a fixed frame of reference in %'. Each choice of the frame will
lead US to a well defined set of structure constants in a certain basis {hi,e,, e-J.
The most convenient choice is to refer LY'S to orthogonal axes, i.e. to axes such
that :
(hi,hi) = g i j = 2 cyi cy, = 6ij.
d
24*
330 G. DE FRANCESCHI
and L. MAIANI
I n the notation of the fig. 2 (were we have relabeled the roots and for simplicity
we use lower indices):
-- -a, -a\
-02
/ ja 3 ( - ,1- - )
2v3
1
2
Fig. 2 .
being q = 1, r = 0. We choose
i
Group Theory and Unitary Symmetry Models 33 1
SO that the remaining multiplication rules are
1 1
[e,, e-,] = - e [e-,, e-,] = -e .
1/6 -2 is -l
We note that this basis differs from that given in (16) only by real factors which do
not affect its compactness and have been introduced in order to have product
rules of the form
13
[A,,LEI = c
m
flkmam 1, = ...8
where flkmis the completely antisymmetric tensor given in [3]. f l k , n defines the struc-
ture constants of XU,, (see sect. 1.7), which is then the compact group associated
to A,.
d) Calculation of roots
We outline here a method of calculating the roots of a simple Lie algebra based on
properties IX, X of simple roots.
If N is a positive root we will say that oc lies in the nth level when:
01 = 2 kin(<), n = 2 k i .
i i
Property X) makes us sure that any root of the nth level is obtained by adding a
simple root to some positive root belonging t o the (n - level. I n particular if
the nth level is empty, all the successive levels are also empty.
Suppose we know all roots up to the nth level, and let N = 2kia(!) belong to such
i
level. Then we can ascertain whether oc - lack),for any non negative integer I ,
( d k ) is a simple root) is a root or not, so that we know the number r , relative to the
dk))-stringcontaining a.Furthermore :
332 and L. MAIAXI
0. DE FRANCESCHI
dk)))
and the right hand side of this equation is a known number ( Z ( d i ) , dk))/(a@),
is known from D y n h diagram). I n this way we obtain p, and if p > 0, a dk) is +
a root of the (n + 1)thlevel. With this procedure, by ( X ) , varying and a(k)we
+
obtain all the ( n 1)th level roots. Since we already know the roots of the Ith
level from Dynkin's diagram (i.e. the simple roots) and in addition a(?)- d k )is
-
never a root when d i ) and d k ) are simple, this method can be used as a recurrence
procedure to find all positive roots of the given algebra.
Let us try with G,. The Dynkin diagram is:
El a.?
and from (15) we obt,ain:
but r = 1, so that q = 0.
a1 + 201, is a root.
and r = 0, so that q = 1
but r = 3 so that q = 0.
Cartan’s
Dimension of G =
denomination Compact group G associated to L dimension of $p
of 9
9.1. We recall here that by representation of a Lie algebra into a complex linear
space L we mean a linear mapping x -+ T ( x ) where x E Y , T (2)is a linear
operator in L, satisfying the condition:
We will treat here only finite dimensional representations, for which the follo-
wing Weyl’s theorem applies :
I. Any finite-dimensional representation of a semisimple Lie algebra is completely
reducible. Hence we can limit ourselves to irreducible representations.
Chosen a basis {hi,e,, e-a} in Y,we will indicate with { H i , E,, E-,} the correspon-
,ding operators in any given representation.
11. It is possible t o choose among equivalent representation, a particular one in a
Hilbert space, in which: H I = Hiand EL= E-,16).
la)A representation of .f gives us a representation of the associated compact real Lie algebra,
which in turn generat,es a representation of the corresponding compact group. Call it W ( g ) .
From what we said in sect. 3.4 we can always change W ( g )by an equivalence transformation
( W ( g )-+ w(g)= A W(g)A-l) so to obtain a n unitary representation.
Under the same equivalence transformation the operators Fi, P,, G, representing the com-
pact basis (16) go into the operators:
F: = AFiA-l etc.
which are antihermitian, so that Iii,E,, E-, transform into operators satysfying:
(H$+ = H { , (E;)+= EL,.
This result is not essential from a mathematical point of view, in that what really matters is
the possibility of diagonalizing the operators H i ’ s which is assured by the fact that Hi’s re-
present a Cartan subalgebra. (Continued on page 334.)
334 G . DE FRANCESCHI
and L. MAIAEI
Then the Hi’ s, being commuting Hermitian operators, are simultaneously diago-
nalizable, and have real eigenvalues. If M = (illl,.. .M,) is a set of eigenvalues
on a simultaneous eigenvector
Hi IM) = Mi [ M )
9.3. P r o p e r t i e s of w e i g h t s
Let IM) be a vector belonging t o LM,then by the commutat,ion relations between
H i and E,, we obtain:
HdE, IN) = (.i + Mi)E, IM). (2)
Let us suppose E , IM) =+ 0. Then (2) tells us that M $- n = ( M , al. . . M , + +
+ a,) is a weight, and E, IM) belongs t o LM+D. If E,E, IM) =/= 0 we can repeat
the reasoning concluding that M +
2 n is a weight and that E, E , I M ) belongs
to By recurrence if ( E J k 1M) =/= 0 , then M +
( k n ) is a weight and
(E,)”M) belongs to LM+!<. . Being L finite dimensional this procedure must end,
so that there exists an integer q such that (E,)qIM) 0 , i.e. M + -+
qa is a weight,
whereas (EJg+llM)= 0. By analogy we can work with E-&, obtaining an in-
teger r such that
(WrIM) =I=0
(E-,)‘+l I M ) = 0.
Hence weight vectors dispose intcj strings generated by roots and the E,’s behave as
usual raising and lowering operators.
With the aid of the tensor qij we introduce a scalar products between weights and
roots, as well as between weights :
We note that the close resemblance between the stated properties of weights, and
the properties of roots listed in sect. 8.3 is not surprising in that roots are simply
the weights of a particular representation of Y,i.e. the regular representation.
We introduce now an ordering between the weights of an arbitrary representation.
We recall that the r-simple roots constitute a basis in the space of the r-dimensional
real vectors, so that for any weight M we can write:
M =2
z
Midi)(,(it = ith simple root). (8)
We will say that M > M’ if the first non zero component of the vector M - M’
is greater than zero. Since there is only a finite number of distinct weights for
any representation, among them there is a maximal weight, i.e. a weight which
is greater than all the others. This definition has the consequence that if a is a
positive root and \A>an eigenvector belonging t o the maximal weight A, then
E, IA) = 0. (Otherwise E , IA) would be a vector belonging to the weight A + a
which, since a > 0 , is greater than A ) .
Let R be a representation of Y in the linear space L, and \ A , l), / A , 2 ) . . . \ A , k ) ,
be independent eigenvectors belonging to the maximal weight A. Consider the
subspace P spanned by vectors
Hi’s is a linear combination of vectors (9) differing only by the order in whicl
E,’s appear. The corresponding weight is
and in addition
/ A ,1) = In,v1) + IA, w,), IA, w,) € u,
i.e.
H, IiL 1) = A, 1) = H , In,Vl> + H , IA,t’J
+
(Wz - 4)IA, ~ 1 ) ( H t - At) ] A ,~ 2 ==) 0
( H , - A,) I A, vl> E vl; (W,- A,) I A, v2> E w2 (by invariance of vl, vz)
so that
(Hz - A) IA, ~ 1 = ) 0
and the same for I A, w2). Hence in Z1 there exist two independent vectors 1 A, v,),
I A , v2) belonging t o the weight A , which is impossible.
Hence IA, wl) or ] A ,v2) must vanish. Suppose / A ,w,) 0, then +
14 1) = In, v1)
which implies P c w1 whereas by hypothesis l1 =3 v,. We conclude that
I1 = wl, v2 3 0, and Z1 is irreducible.
The great relevance of the concept of maximal weight suggested in part by pre-
vious considerations can be appreciated from the following theorems due t o
Cartan.
VI. Two irreducible representations having the same maximal weight are equi-
valent.
TIT. An r-component vector A is the maximal weight for some irreducible repre-
sentation of Y if and only if
2 ( 4 ,(I))
=--
O1& (&, &))
is a non negative integer for any simple root a(*)of 2’.
Hence, once we have chosen a set of simple roots d L )any , collection of non
negative integers (AulA,1,.. .A,,) uniquely defines an irreducible representation
of 2’ and all representations are obtained in this way.
VIII. If A is a maximal weight of a givcn irrcducible representation of Y,then
any other weight M has the form
Now q is a known number (because we know all the weights up to the n t h level)
and so is the right, hand side of this equation, so that we can ascertain whether
M - dk)is a weight or not. By varying M and we obtain all the weights
+
belonging to the ( n l)thlevel.
Hence starting from the zero level, i.e. from the maximal weight (there are no
negative levels because A + dk) is not a weight for any dk)) by this recurrence
method we can construct all the weights of the representation. This method does
not provide for each weight M the corresponding multiplicity, i.e. the dimen-
sionality of the manifold Ln1appearing in (1).Being the representation determined,
up to an equivalence, by its maximal weight 11, these multiplicities must be
derivable from A , ill and from the roots, but no simple formula can be given
for them. Instead we will give later for XU, a simple rule which allows one to
read directly these multiplicities from the weight diagram.
9.S.a) Till now we have analyzed properties which are the same for equivalent
representations (in particular weight diagmms).
I n practical calculations it is necessary to pick-up from each equivalence class
a particular representative, i. e. standard matrix representation of the elements
hi,e,, e-#. It is convenient to choose a representation in which a basis is constituted
by normalized eigenvectors of Hi's as well as of the other commuting operators
which are necessary to remove degeneracies (sect. 9.1). Of course with this choice
the Hi's are represented by diagonal matrices, with coefficients determined by
the weight diagram.
Using the commut'ation relations and the string property of weights, one ca i t
det,ermine the matrix elements of E,, E , up to certain phase factors. The proce-
dure is quite analogous, although considerably more complicated, to that employed
in usual angular momentum theory [8].It is then necessary to make a definite
phase convention. At the same time this convention fixes phases in the Clebsch-
Gordan coefficients.
9.3. a) Weight diagram of XU,
The Lie algebra of 8 U , has rank one, so that its irreducible representations are
characterized by s single non negative integer Aal . For any maximal weight 11,
338 G . D E FRANCESCHI
and L. MAIANI
we have :
Since a1 = 1 (sect. 8.7a) we see that 11, i.e. the covariant component of the
maximal weight along a ( l ) is an integer or half integer number which we call
j. There is only one string i.e. the &)-string containing A , so that by applying
E-,(I,to the eigenvector belonging to il, we generate all the vector space of the
representation. There are 2 j + 1 independent vectors a t all, i.e.
9.3.b) W e i g h t d i a g r a m s of S U ,
Being X U , of rank two, its irreducible representations can be labeled by two non
negative integers (m,n) where, with the notat’ions of fig.2 , we have
11 = nza2 + n3a3
and, by taking the scalar product of bot,h members with a2and with as, we obtain
vb = 2n, - n3
i.e.
n=-n 2 + 2,n,
Hence with respect t o the basis chosen for roots in sect. 8 . 7 ~we
) have
Group Theory and Unitary Symmetry Models 339
These components are respectively the eigenvalues of H , and H , on the manifold
L,,. We will construct now the weight diagram for ar, arbitrary irreducible repre-
sentation (m, n ) with a graphical method which lead to the resultmore quickly
than the general one, outlined in sect.
9.2.
i) We draw with respect t o two ortho-
gonal axes the vector rl of components
(m, + 1212 (3, m - n / 6 ) as well as the two
simple roots nz, n3 of components
Mi = -1 - in,; 0 5 i 5 m.
iii) From (7) we see that if LX-is a root, by reflecting a weight through an axis
orthogonal to n we obtain another weight, so that the weight diagram goes into
itself by such operation.
Let us indicate with r,, r,, r3 the axes through the origin orthogonal to nl, a,, n,.
Reflection through r3 must carry a into itself (so that r3 intersects a in its mid-
point) and send b into 6,. Hence in the region B there are no weights (by reflec-
ting such weights through r, we would obtain weights ending in A ) . Furthermore
b, contains the end points of new weights which are the reflected of those ending
in b.
Segments a,, a l , b, are obtained by reflections through rz and r,. The figure SO
obtained has the properties that no weight ends outside it and there are weights
ending on its vertices and on its sides, distances between two consecutive end
points being equal to 1 a,1 = I n, 1.
Let us see now how we can find all the weights of the diagram.
Consider an arbitrary weight M = il-- in, - hn3 (i, h non negative integers).
When i 5 rn, A - ia, is a weight ending on b, so that all weights M with
i 5 m can be obtained from strings starting from weights ending on 6 . When
i 2 m, we write
t r I
(l.'O)
3
(3.0) (2!2)
10 27
Wig 4. Weight diagrams of some S O s representations
and it is easy to see that the end points of the weights so obtained fill up the whole
hexagon with a triangular pattern as shown in fig. 4.In addition this figure shows
that the points are disposed in layers. Now the following rule applies:
Weights of the external layer are all simple, i.e. not degenerate.
Starting from the external layer, multiplicity increases by one a t each layer,
until a triangular path is reached: when this occurs multiplicity does not increase
any more. When n = 0 or m = 0 the hexagon degenerates in a triangle and
all the weights are simple.
When n = m the diagram is a regular hexagon and multiplicity increases until
one reach the last layer, which is constituted by a single point. I n fig. 4 are given
the weight diagrams of some between the most used representations of S U , ,
with the respective multiplicities.
We said before that in any irreducible representation of S V 2 ,there is one operator,
commuting with H , and H , , which removes all the degeneracies.
It can be seen that the operator
1 ., T = 'i2 2H, = I
1 ,, ,, T = '1, 2H, = -1
1 ,, T = O 2H2=0.
We shall see later that in the eightfold way model the operators H i , T2, HI,
are identified with T i T2,Y , so that this rule gives us a decomposition of each
XU, supermultiplet into isospin multiplets.
9.4. T e n s o r p r o d u c t of r e p r e s e n t a t i o n s
Let el and e2 be two irreducible representations of Y into the linear spaces L,
and L,:
5 E Y x + el (x); el (5) = linear operator in L,
x -+e, (x); e2(x)= linear operator in L,
and let fW, MZ ,...; N 1 , N 2,...; be the weights of the two representations,
1_W),IM,),. . . INl), I N2),.., the bases formed with the corresponding eigenvectors
(for sake of simplicity we do not write explicitly the eigenvalues of the additional
operators needed to remove all the degeneracies : they are however understood) :
i.e. the weights of p are obtained by adding together the weights of el and e2
in all possible ways. I n particular for the greatest weight we have
A = A1 + 112
At= A:* + %>
where A1 and A2 are the maximal weights of el and e2.
(for any simple root d2))
Moreover the eigenspace corresponding to A is always one-dimensional, and it is
spanned by the vector
I A1>I A2)*
I n general e splits up in a direct sum of irreducible components
p = 0 On'. (16)
n.
Each of them, according to VI, is characterized by its maximal weight A', which
we have chosen as a label in (16). I n general in this formula will appear many
irreducible equivalent components, i. e. terms with the same A'. Although general
formulae can be given, characterizing which are the irreducible components and
how many times they appear in (16) [I81 yet these formulae are extremely com-
plicatedls) and are not used in practicel9).
la) This is not the case of A,(SU,) for which the decomposition (16) is explicity given by
the well know Clebsch-Gordan formula :
19) I n practical calculations we need not only the decomposition (16), but also explicitly the
matrix connecting the basis lMk) INz) t o the basis spanning the irreducible components,
i.e. we need all the Clebsch-Gordan coefficients involved. Such coefficients in general have not
been calculated; however in the case of S U , they are tabulated [29,33] for all tensor product of
interest in physics.
Group Theory and Unitary Symmetry Models 343
Anyway it is a very simple task t o isolate a particular term in (16)) i.e. the
component pA where A is the greatest weight of e. In fact if we construct the
manifold spanned by the vectors
E-, E.-pE-, . . . 1.1)
E+, E-,,. . . lowering operators of e ) we obtain
(a,/3, y , . . . = positive roots: EM&,
an irreducible invariant subspace (see sect. 9.2) and obviously if we consider
the restriction of e to this manifold, we obtain a representation having the maxi-
mal weight equal t o A . Being A simple, en occurs only once in (16).
Consider now the r-irreducible inequivalent representations ei with maximal
weights A(i),such that
2 (A({),
a@))
= Bik.
(&)) La@))
and isolating the irreducible component of greatest weight. I n fact this component
belongs to the weight
A = A,, n ( 1 ) + A,&I(Z) + + A,* A@)*.a
so that the eigenvalue of 5(hi)are just minus one times the eigenvalues of e (hi),
e
i.e. the weight diagram of is obtained by reflecting through the origin the dia-
e
gram of e. I n particular is equivalent t o e if and only if its weight diagram is
invariant under this reflection.
Consider a vector x E L with components (xi) and a vector y E L* with com-
ponents (yi): by definition applying transformations e and we have e,
x'i = C & X k
k
Y; = - @?Yk-
Having L and L* the same dimension, there exists always a one-to-one corre-
spondence between their elements. Let
x = A y x E L , Y EL*
be such correspondence. If we transform x with e and y with e, the resulting
vectors will not be connected by A , unless e is equivalent t o
P P
Fig. 5 .
9 --f T(d
is the representation of this group generated by 0, then
is that generated by 5.
Group Theory and Unitary Symmetry Models 345
I n the case of S U 3 we note that all weights diagrams are symmetrical for reflec-
tions around y-axis (this axis is in fact orthogonal to the root a,) so that, to
obtain the diagram of the adjoint representation, we have only to reverse it with
respect to the x-axis.
A glance to figure 5 is sufficient to conclude that if e is the representation (m, n ) ,
e
then is (n,m). I n particular Q is equivalent to if and only if m = rL (in this case
in fact they have the same maximal weight).
9.6. E x p l i c i t c o n s t r u c t i o n of X U , r e p r e s e n t a t i o n s
The 3 x 3 antihermitian traceless matrices
A”,=-- i (O1 0 0)
2
0 i (;-g :) 6 A,=-- f 0 -1 O 0)
0
A”*=-qo .) ..=-:;q0
0 0 0 0 0 0 0,
0 0 1 o o i 0 0 0
2 0 0 0 l) (19)
1 0 0 i o o 0 1 0
2 0 g
d(; 0 -p) : :)
0 0-2
satisfy the commutation relations given in sect. 8 . 7 so
~ that they are a 3-dimen-
sional representation of the compact basis of A,, and their real combinations span
a representation (in fact irreducible) of the Lie algebra of SU,. The operators
representing h, and h,, are (see sect. 8 . 7 ~ )
By considering the simple roots a, and a3 we easily see that Al is the greatesi
weight and that
so that this representation is the (1,O) one, whose diagram is reported in fig. 4.
25*
346 and L. MAIANI
G. DE FRANCEXHI
The adjoint representation is obtained by taking the matrices -AT, and is the
(0,1) representation, inequivalent to the (1,O) one. These representations which
are the fundamental ones, being the only 3-dimensional representations of A ,
will be called 3 and 3.
By taking the matrices
1.8
zakak
T(a,. . .ag)= e k (ak real numbers),
This is the representation which to each matrix of XU, associates the complex
conjugate matrix.
The linear spaces of representations 3 and 3 are
L; : ,, )I Y 9) 93 (Yl, Yz,Y3).
I f Uikis a matrix of S U,, then for the representation 3 we have
U + T ( U ) : x ' = T(U)x
xli I= c k
UikXk,
whereas for 5
U -+ T(U), y' = ? ( U ) y
Y; = 3g i k Y k .
k
The tensor product
is just the vector space of the 3n$+ncomponents objects (tensors) (sect. 5.1)
and the product representation of XU, in this space is constituted by the opera-
tors Ti%)( U ) defined as (sum over repeated indices is understood)
Group Theory and Unitary Symmetry Models 347
From (15) sect. 9.4 we see that the compact basis of the Lie algebra of XU, is
represented by the operators A(k)acting as
where ,I:are
$) the matrices listed in (19).
9.7. Let us consider the tensor product representation (reducible when k > 1):
(3)k = 3 @3 @3 0 . . . @ 3.
L
.
where i', 2'. .,k' are again the numbers 1,2 . . k: rearranged in some way speci-
fied by p . We associate to p a linear operator p acting on L k , defined as:
i.e. p commutes with T ( k ) ( U ) .Let us call & the set of all the operators p and
of all their linear combinations: clearly all the elements of Z;, commute with
348 and L. MAIANI
G . DE FRANCESCHI
T ( k ) ( U ) ’ sbut
, the importance of Z;,lies in the fact that the converse is also true,
i.e. [ 2 2 ] :
all the operators commuting with T(”( U)’s belong t o &.
Hence Y,’s (which reduce the representation) are in &.
Their characterization can be achieved using the so called Young tableaux.
Consider an arbitrary partition of k objects into groups of A,, Az . . . A, elements
(Al 2 A, 2 . . . 2 I,). This partition is indicated by a tableau made of r rows
containing A,, A,, . . . , A, boxes :
Fill now this tableau with the numbers 1, 2 , . . . k in all possible ways consistent
with the rule: numbers must increase, in each row from left and in cach column
from above.
Consider for example k = 4. Then we have five tableaux:
1, := 4
E
P
a, =3 A, = A,
A,= i
A, = 1, := 1, = 0 A, =1 A3 = I.,
A, = A, = 0 A, = 2
A, = A, =1
1, =0
which, filled in all possible ways., give:
U U
(1 2 4) (3) or simply (1 2 4)
stands for:
1+2
2+4
4-tl
3+3
and :
stands for:
- x i a i z i r i l-
- xi,i2i4ia yiiiai&- Xi&ipi,
+ Xiri1i4ia
+ Xizi3irii + Xi&iais + Xi&isil.
It can be shown that:
i) all Y's are proportional to projections:
Y
Y2 =- cY so that - is a projection,
c
iv j Y's corresponding to the same partition A,, .. . A,, differing by the arrangement
of 1, 2 . . .k project over equivalent representations.
It can be easily verified that any Young symmetrizer projects over tensors anti-
symmetrical in indices appearing in the same column of the corresponding tableau.
Now our indices run over 1 , 2 , 3 so that each tableau with more than three rows
gives an identically vanishing projection (this is the case of the tableau (21)l) and
need not to be considered.
We want to see now how each irreducible (m, n) representation of SU, can be
obtained isolating a particular irreducible component in a suitable product
3 @ 3 . . . 03 [21].
Consider the tensor product T'w2%'= (3)m+2n.We claim that the irreducible
manifold L characterized by the Young tableau
1 2... n mi-%
(33)
m+n+l ... m+2n
then
-
TiE; ( U )F = cn ) E X = E TmiZn
T'm' (WX,
where T:?;( U ) is any element of the representation (3)mx @)" = 3 @ 3 . . . 3 @
@
-
5.. .@ 3
n
and with E we have indicated the mapping (24). This shows that the
(s)",
whose restriction t o L"is equivalent to the restriction of T(m+zn) to L.
Proof will be
-
complete if we show that the vector belonging to the maximal weight
of (3)" x (3)"is contained in LE(see sect. 5.4).
Now, according t o the matrices of sect. 9.6, the vector belonging to the maximal
weight of the representation 3 is
(X)i = 8il = (1, 090) 9
(,-)tf+':"i"+"
3 ...3% = &+ll 8in+z 1 , . . (yn+rn1 . dj12 . . . djn2
$,+,+I .-
352 G. DE FRANCESCHI
and L. MAIANI
For example previous theorem allows one to conclude that linear manifolds on
which tableaux (21b, c, d) project, transform as the (2,l) representation, whereas
those corresponding t o (21e, f ) transform as (0,2).
I n the composition of the tensor product ( 3 ) k also
, tableaux with three rows appear.
To which irreducible representation do they correspond? We will see this referring
t o an example.
Consider the manifold of tensors
i.e.
so that the irreducible component (21i )is equivalent t o the representation3, which
corresponds t o the tableau . This reasoning can be repeated for tableaux with an
arbitrary number s of columns with three boxes, leading to the conclusion that :
+-
is equivalent t o en+
tS+
Group Theory and Unitary Symmetry Models 353
(3)* (4,O)
a
L--y.--
b cd
--
+ (2,1) + (2,1) 4-(2,1) -t(0,2) + (0,2) + (1,O) + (130) + (1,O)
ef ghi
(312 = (0,1) + (290)
(31, = (3,O) + (1,1)+ (14) + (0,O).
= ui,, . . .
P(U ) ' s are of course symmetrical in upper and lower indices and, due t o unitarity
of the U's, they are traceless. The linear manifold spanned by F ( U ) ' s is contained
in Le and is obvously invariant for the representation (3)" x (sp. Since LEis
irreducible (as we saw before) and this manifold is surely not zero, we coiiclude that
it is identical with LE;i.e. LE contains tensors symmetrical and traceless.
ii) Call LF the manifold of all tensors symmetrical in upper and lower indices and
traceless. From i) Lz 3 Le.We show now that the dimension of LT equals that of
LE,so that LT = LE.
354 G. DE FRANCESCHI
and L. MAIANI
. .
A tensor Fii;;:;: symmetrical in upper and lower indices, has
independent components (remember that indices i , j go over 1 , 2 , 3), and the num-
ber of independent conditions (26) equals the number of independent components
of a tensor symmetrical in (m - 1)upper an ( n - 1 )lower indices, i.e. it is equal to:
(m - 1 ) 3 - 1 + m(m + 1 ) n ( n + 1)
( m-1 4 (28)
, JT'
Group Theory and Unitary Symmetry Models 355
is associated t o the representation (k1,..kn-J, and tableaux differing by an arbi-
trary number of columns with n rows characterize equivalent representations. The
dimensionality of the representation (lcl,..lc,,-l) is given by the formula :
We list here for example some SU, representations, together with their dimensiona-
lity (the contragradient of ( k l . . . kn+) is (/en+ k,,-2 :..,k,))
r
m
R e d u c t i o n of t e n s o r p r o d u c t of a r b i t r a r y X U , r e p r e s e n t a t i o n s .
I n finding the irreducible components of the product (m, n) @ (m', n') again the
technique of Young tableaux can be used. We give here without proof a simple rule
for making this reduction [22]. We illustrate this rule referring to the product
0 (IJ).
Write down the corresponding tableaux, having filled with symbols a and b first
and second row of one of the two, arbitrarily selected :
Then add to the empty tableau the boxes of the first row of the second one in all
possible ways consistent with the rules :
i) there must never appear two a's in the same column;
ii) for each resulting tableau, containing A,, A2, A, boxes in first, second, third row,
it must be 31, 2 A2 2 A,.
I n our case we have six possible tableaux:
(4)
356 and L. MAIANI
G. DE FRANCESCHI
To all these tableaux we add now, in all possible ways, all boxes containing b,
again consistently with ii) and with the additional rules ;
iii) there must never appear two b’s in the same column;
iv) for each tableau so obtained write the sequence of a’s and b’s obtained reading
first row from right to left, then the second row from right to left and so on: only
tableaux are allowed such t,hat a t any stage of this sequence the number of a’s
already read is greater than or equal t o the corresponding number of b’s.
From tableau (1) we obtain the following allowed tableaux
8 @8=1 + 8 + 10 + E+27
8 @ 10 = 35 + 2 7 + 10 + 8
8 @ 10 = 8 +lo + 27 + 35.
Group Theory and Unitary Symmetry Models 357
The same procedure can be applied t o the reduction of the tensor product of arbi-
trary S U , representations, labeling with c , d, ... boxes in third, fourt'h, .. rows. At
the end of course we will drop tableaux containing more than n rows.
Consider the case of XU, and the product:
35@35=(1,0,0,0,1) @(1,0,0,0,1)
+
+
4
e
+ +
e
e a a e
0
in fact one can imagine situations in which the symmetry breaking interactions,
which are certainly present (see sect. 7.5), make some members of a supermultiplet
so unstable, that they are not practically ob-
servable. On the contrary the discovery of a
predicted particle provides strong evidence in
favour of the implied model;
ii) from the symmetry scheme one can deduce
relations between amplitudes of different pro-
-x-
I+ - cesses; (see the example of iso-spin, sect. 7.4).
iii) by assuming certain transformation pro-
perties of weak and electromagnetic interac-
tion Lagrangians of hadrons under the sym-
--
L. 70 metry group, with the aid of Wigner-Eckart
theorem (sect. 6.3), one can derive relations be-
Fig. 6. tween the amplitudes of weakand electromagne-
tic processes involving such particles.
I n what follows, we will focus our attention on the “eightfold way” model pro-
posed by NE’EMAN[24] and GELL-MANN [3], which has proved t o be the most
successful one 20).
10.2. a) I n the eightfold way model one chooses XU, as the underlying symmetry
group, and associates the eight “stable” baryons N , Z, A’, E: t o the basis vectors of
its (1, I ) eight dimensional representation, as indicated in fig. 6.
We can easily deduce the relations between T,,Y , and the diagonal elements H I ,
H,, by remembering (sect. 9.3b) that in the representation (m,n ) of S U , the eigen-
values of H , and H , corresponding t o the maximal weight A are
m+n m-n
A= ___ -).
( 2 f 3 ’ 6
Hence, with m = n = 1 we have
T,= fCC H ,
20) For a concise discussion concerning the other symmetry models see [25, 261.
Group Theory and Unitary Symmetry Models 359
and in addition, by requiring p t o correspond to / A - &,I),
Y = 2H,.
The operator T 2 defined in sect 9.3b, which removes all degeneracies inside a
supermultiplet, is just the square of the isospin. We will choose normalized simul-
taneous eigenvectors of T 2 ,T,,Y as a basis in any irreducible representation.
The pseudoscalar mesons also fit very nicely into an octet according to the scheme
reported in fig. 7.
1 t
--n - -P'
I I
Fig. 7. Fig. 8.
It should be noted that the y1 meson, which quite naturally completes the octet,
has been discovered after the introduction of the eighfold way.
When we try to arrange the vector mesons into the scheme, we get in trouble.
I n fact we would like to assign the nine mesons
p(T = 1, I'= 0 ) , K*(T = Y = I ) , Z*(T=1/,,
Y = - l),w = (T = 0, Y = 0), v ( T = 0, I' = 0)
to the same irreducible representation. However,
since T,, T 2 , Y are a complete set of commuting __
operators inside each irreducible representation, it
is not possible to fit in the same supermultiplet two
distinct isosinglets (a,rp) with the same hypercharge.
The usual assignment is t o put eight mesons into
an octet (fig. 8) and the remaining one into a singlet
$7:
(i.e., in the (0,O)representation), and the question
arises whether the w or the 'p particle is to be put Fig. 9.
in the singlet.
We shall see in the following that this question can be consistently resolved.
I n considering the baryon-meson resonances, we have t o find a representation
containing an isospin 3/2 multiplet with hypercharge equal t o one, corresponding
t o the well known NZp2,x - N resonance. The lowest representations contai-
ning this isomultiplet are I0 and 27. The first one accomodates wery well the
N = 1, J p = 3/2+ resonances N&, Yf, Z*, Q- (fig. 9).
Whereas N& Yy, S* were a t hand when it was proposed to assign them t o
the decuplet, the Q- was not yet known: the model predicted its quantum numbers
( N = I , T = 0, Y = - 2 ) as well as its mass M o = 1676 MeV.
20 Zeitschrift ,,Fortschritte der Physik", Heft 7
360 G. DE FRANCESCHI
and L. MAIANI
On this indications extensive searches for the Q- have been carried out, until a
particle with right mass and hypercharge has been found studying K-p reactions
a t 5.0 GeV/c [27].
The positive result of the search has been considered as one of the most brilliant
successes of the eightfold way model.
There is another particle, namely the Y,* (mass = 1405 MeV, T = 0, Y = 0,
J p = l/;) which usually is sssigned to a (0,O)representation.
(p) 938.2
N 1 (n) 939.6 1 'IS
A" 1 2
!'+ 1115.4 0 0
__
+ 1189.4
c 1 'i2+ -1197.1 0 1
1192.4
1 -1 'I2
1 *In-? 1 1405 0 0
__
N* 1 "/2+ 1236 2 1 3iZ
__
1 "8' 1382.1 & 0.9 0 1
__ ___ _ _ _ _ _ _ ~ __
1 "2+ 1529.1 f 1.0 -1 'I2
__
R- 1 3 /z +? 1675 f 3 -2 0
Tc 0 0-
__-
i 3 139.6
135.0 0 I
K 0 0-
+ 493.8 1
498.0
-_
P 0 1- 763 f 4 0 1
__ .- -.
w 0 1- 782.8 0.5 0 0
---
1 when i, . . . i, is a permutation of
I
1 1 . . . 1 2 2 . . . 2 3 3 . . . 3 a n d j l . . . jR is
. .
T;;::;:
=
---
kl k2 h3
a permutation of 11 . . 1 22
0 otherwise
hl
. . . 2 33 . . . 3
h2 h3
1 2
y = -
3
(k1 - h,) + 31 - ( k , - h2)- 7 ( k , - h,).
They are a basis in the manifold of t,he tensors which are symmetrical in upper
and lower indices, so that to obtain a basis in the (WL,n) representation, which
diagonalizes T 3 and Y , it is necessary to take those linear combinations of (1)
which are traceless. Among them we will select those linear combinations, which
correspond to eigenstates of T2.
In what follows we will be concerned only with the eight and ten dimensional
representations.
I n the case of the (8) representation we have to consider tensors with two indices.
We have not to impose any particular symmetry property, but only the trace
condition :
Ti = 0 .
2 6*
362 and L. MAIANI
G. D E FRANCESCHI
T(,), T ( 5 )T, ( 9 )are not traceless. There are only two independet traceless lines
combinations
TO) - T(5)
+
T(1) T'(5) - 2T(9) (2)
which span, together with T ( z ) ,T(3),T(41, T ( 6 )T(,),
, T ( s ) the (1.1) represen-
tation. It is worth noticing that the general tensor of ( 1 , l ) can be thought as
a three by three traceless matrix, and the transformation law 9.6 (20) can be
written as a matrix product
T + TI = UTU-1 U E RU,.
Then the operators representing the S U 3 Lie algebra act as
( A T )= [A,T I .
where ii is any linear combination of the matrices given in sect 9.6 (19). It is
then very easy t o see that the tensors (2) are eigenstates of T z with eigenvalues,
respectively, 2 and 0.
From this point of view, recalling what we said in the previous subsection, we
see that :
i) in the case of stable baryons, we have the following assignments :
ic
I O 0
1
0 - 0 et,c.
1s
-2
0 0 -
ljc
Which are symbolically summarized in the matrix :
Co Ro
P
Group Theory and Unitary Symmetry Nodels 363
Factors i/@,I/@ have been introduced so that our basis tensors are normalized
with respect to the scalar product
T ' ) = 2 Tj (2"';) = Trace [T(T')+]
(T, .
ij
For what concerns the 10-dimensional (3,O) representation, we notice that it is.
spanned by the symmetrical tensors :
1 when i j k is a permutation of (a./I,y = 1,2,3)
(T,&# =
0 otherwise
Y T2
111 3i2 1
112 '/a 1
122 -Y z 1
222 - 3i2 1
113 1 0
123 0 0
223 -1 0
133
233
333
'I2
--%
0
1 -1
-1
-2
10.3. A n t i p a r t i c l e m u l t i p l e t s
So far we have not explicity considered the assignments of antiparticles.
If la) is a particle state, then the corresponding antiparticle state IZ) has t h e
same space-time properties, but the values of all charges (Q, N , Y , etc.) a r e
reversed. 1 a} and) ; 1 are connected by the charge conjugation operator C
la> = C l 4 ,
which can be assumed t o be a unitary operator satisfying :
P = 1.
It is a well known fact that strong interactions are symmetrical under C , so that
if particles exhibit, the SU, symmetry, relative t o strong interactions the same
behaviour must be displayed by the corresponding antiparticles.
364 G. D E FRANCESCHI
and L. MAIANI
CQ + QC = {C,Q)= 0
{C,N}= 0
{ C , Y ) = 0,
the linear manifold spanned by the basis vectors of the (m, n) supermultiplet
of particles with baryonic number N particles and with
n -
(the bar indicates that these particles have baryonic number - I).
Group Theory and Unitary Symmetry Models 365
The pseudoscalar meson multiplet is mapped into itself by C since mesons have
baryonic number equal t o zero :
M + ( M -) T = (S+rl
-‘
J12 ‘J16
-
-
................
..........................
x- .-*I
I n the case of vector mesons a minus sign appears due t o the fact that neutral
vector mesons have negative charge conjugation.
10.4. Mass f o r m u l a e
H=H 0 + HI,
where H , is invariant under SU,. For what concerns H I , it must be charge
independent and hypercharge conserving
[HI, Ti] = [ H I , Y ] = 0, i = 1 , 2 , 3
so that H , is constant inside each isomultiplet (we do not consider a t this point
electromagnetic and weak contributions t o the mass).
Let us label each strongly interacting particle in its rest frame with the quantum
numbers N , J p , I,, T,, T 2 , Y , where 3, is the label of the S U , representation
t o which it is assigned. Physical masses are then the eigenvalues of the matrix:
( N ‘ , J t P ‘ ,A‘, T 2 ,TI2, Y‘ IHI N , J p , I., T,, T 2 , Y ) .
We have written N and J p for completeness but obviously H commutes with
them, so that the relevant matrix elements are :
T i , TI2, Y‘ IHI N , J p , i,
( N J p ,if, T , , T 2 , Y ) = (1’TiT’2Y’ IH( i T 3 T 2 Y )=
= (3L’TiT‘2Y‘
lHol IT3T2Y) + (i’TjT‘2Y’IH,l I.T,T2Y). (9)
T o simplify notations we assume that particles with the particular values of
N a n d J p considered, group together in only two supermultiplets il and 1’. This
case easily extends to the general one, and in addition this is the most complicated
situation that has been found till now.
Using the assumptions made on H , and H I , the matrix element (9) writes as :
&.W BY^,
mo(i) 6~11 + mi(T2, Y ) A A4 *~ p
~ T ~ T ~ , z
~ Y YQT,T
, S. (10)
366 and L. MAIANI
G. DE FRANCESCHI
m,(T2, P ) H
ml(T2,Y ) P ~
for each value of T 2and Y occurring both in d and 1’.
I n the case of stable baryons, pseudoscalar mesons, and decuplet resonances,
foreach value of N and Jp,only one representation is present, so that H , has
the form
Using perturbation theory in YMs one would write mass correctioils as the expec-
tation value of a power serie in 2 ’ ~ s .If we assume Y5f.l~
t o transform like the
T 2 = T, = Y = 0 member of an octet, we are led in first order approximation
to our assumption on H I .
Group Theory and Unitary Symmetry Models 367
With this proviso we can obtain the general form of matrix elements (13), using
Wigner-Eckart theorem (sect. 6.3.).
We write H , as T$b,oand introduce the notation
C(1, T 2 ,T,, Y ; 8 , 0 , 0 , 0 ; pyT2T,Y)
for the Clebsch-Gordan coefficient connecting the vectors
12, T2T,Y) 18, 0, 0, 0); I&T2T,y)3
where the latter vectors span the standard basis which decomposes the tensor
product I @ 8. The suffix y distinguishes between equivalent representations
appearing in the decomposition.
Then, according to Wigner-Eckart theorem, we have :
m = m, + m,,(T2,y) = m, + 2 C(1, T2,T,, I';8 , 0, 0, 0; pyT2T,Y). (1IIT(8)IIp,,).
Y
(14)
Consider stable baryons. In this case I = 8, and 8 @ 8 = 1 8 8 10 + + + +
+ +l ! 27, so that the regular representation appears twice in this reduction,
and we have two reduced matrix elements.
Using the Clebsch-Gordan coefficients reported in [29] one finds:
mN = m,
1/5 ( 8 11 T @jj) 8,)
--
I0
+ -2I ( 8 (1 T@)I[ 8,)
We have four masses and three unknown parameters. Their elimination leads
t o the relation (firstly given by G-ELL-MANN)
:
is
2
MQ- = N o + ~ (10 11 T(*)I[ 10).
is
368 and L.MAIANI
G. DE FRANCESCHI
Since there are two unknown parameters and four masses, one can obtain:
M p - M y + = ME* - MQ-
(146 3 MeV) (146 & 4 MeV)
2ME* M = y,* + MQ-
(2058 & 2 MeV) (2057 & 3 MeV).
These formulae simply state that masses in tjhe decuplet are equally spaced.
The mass of Q- can be obtained from the known masses of N* and Y:. This
predicted value, as we said, is amazingly close to the experimental one.
For the octet of the pseudoscalar mesons the argument is the same as that made
for the stable baryons. Thus we get a formula analogous t o (16):
3 1
mg -+ mE = -m,
2
-+ -mn
2
(992 MeV) = (891 MeV).
10.5. o - cp m i x i n g
We have previously assigned eight vector mesons to an octet and the ninth to
a n S U , singlet; but we had no way to decide whether the w or the cp had t o be
placed in the singlet. With the aid of the mass formula for the octet, which has
of course the same form as (19), we find:
(ms = mass of the ieosinglet i.e. of the T2= Y = 0 member of the octet)
Inserting the known values for mIc*,mp, one obtains m8 = 930 & 3 MeV. This
value is intermediate between 7np = 1019, 5 & 3 MeV and ma= 782. & 0.5
MeV; however both differences rn; - mg = 18 . lo4MeV2, mi - rn; = 2 5 .
- lo4MeV2 are very large.
If we insist that the mass formula must hold even in this case t o an accuracy
comparable to that obtained for the three other cases considered, we must con-
Group Theory and Unitary Symmetry Models 369
clude that neither o nor 'p can be identified sic e t simpliciter with the isosinglet
of the octet.
However physical particles are eigenstates of:
H = H, +HI.
0 M ; + ( P IHlIP) 0 0 0
H E 0 0 2M; +(K*IH,JK*) 0 0
0 0 0 J!f; + (8 IHlI8) (8 lHlI1)
\ 0 0 0 (1 IHI 18) Jq + (1 IHI I),
where
811 = (8 /HI I 8), E~~ = (8 I H , I 1) etc.
Changing, if necessary, phases of the states 1 S), 11) we can take E~~ real so that
from the hermiticity of H,, =E ~ ~ .
We write:
('p)=cos8j8) +sinoil).
I w) = - sin 8 18) + cos 8 I I).
The angle B must be determined requiring I 'p) and 1 w) t o be eigenstates of (21)
with eigenvalues equal to their physical masses squared [30].
From this one finds :
m$ - mz
(tan 8)%= = 0,69; 8 = 39'50'
mi - m2,
+
where mi = M $ E~~ is the value given by the mass formula for the octect.
According to these results 'p and w appear to be superpositions of pure S U , states,
in contrast to all the other particles we have considered [31].
It is important to note that, in contrast to the other cases, the introduction of
the mixing angle 8 rises to four the number of parameters needed to describe
the vector mesons mass spectrum (8, M , , Ml, (8 I H , I 8)). Having four masses
a t our disposal m p , mp,m,, mp it is not possible in this context to have any test
of the theory. The possibility of determining 0 in an independent way will be
discussed later.
370 G. DE FRANCESCHI
and L. MAIAXI
Our derivation of the sum rules (16), (17), (19). 20), was based on the knowledge
of the Clebsch-Gordan coefficients involved.
Alternatively one can find the general structure inside each irreducible represen-
tation of an operator, such as EI,, commuting with TI, T,,T,, Y and transforming
according to the regular representation of S U,.
Such an operator must have the form:
11, = f ( T 2 , Y ) = a 1 + 6Y + C T 2 +dY2 + * * *.
M=a.l+bY+c 4 (25)
j26)I
For the decuplet resonance one has the relation T = 1/2 Y + 1, which reduces
(26) to
m = a' bY , + (27)
giving the equal spacing rule.
The general formula (24) has been given by S. OKUBO[32]
10.6. B a r y o n - m e s o n Y u k a w a c o u p l i n g s
It is interesting to find out all the Yukawa-type couplings between baryons and
pseudoscalar mesons which are invariant under SU,, I n fact we will find that
these couplings involve only two constants so that writing them in the usual
isotopic spin form, one can derive relations between coupling constants such
as gNNm gZAm gNZK etc.
I n a field-theoretical model one assumes baryons and mesons fields t o behave
under X U , as tensor operators (see sect. 6.3) belonging t o the eight dimensional
representation.
The Yukawa-type interaction Lagrangian has the form
Group Theory and Unitary Symmetry Models 371
Trace ( B y 5M B ) (29)
is invariant.
Instead of (28) and (29) we will use the so called F and D combinations defined as
a = - - -9,
i y 9'
then all coupling constants can be expressed in terms of oc and g. We refer the
reader to [33] for a complete list of these relations.
372 G. DE FRANCESCHI
and L. MAIANI
x + Pr
K+K;
3 + (%)Y
Consider now the pseudoscahr-vector meson couplings : also in this case we may
have two invariant combinations of terms like
a,v, =0,
Group Theory and Unitary Symmetry Models 373
is equivalent to a divergence, i.e. it adds a divergence to the Lagrangian, so that
it can be assumed to vanish. I n fact :
Hence also this coupling is p n x P.I n contrast with the previous case the vector
meson singlet cannot be coupled to pseudoscalar mesons : in fact the only possible
coupling would be
TO-fKtK
is forbidden, whereas
w,+K+K
is allowed, i,e. only the componend of the 'p particle on the octet can decay in
in K K. This fact can be used in principle to determine the w - rp mixing indepen-
dently from mass formulae [37].
The decays of the baryon decuplet resonances allowed by energy, T,, T2 and Y
conservation are
(Q- is stable against electromagnetic and strong decays because of its mass,
which is less than the threshold of the E K channel which is the only open for
these interactions). These decays, in the limit of exact SU,, are described by it
single amplitude. In fact in this case the matrix elements involved are of the type
(all the other amplitudes can be obtained from these by isotopic spin symmetry)
where factors multiplying a are the proper Clebsch-Gordan coefficients.
If we want t o compare (32) with experiments we have t o take into account mass
differences. The most simple thing to do is t o introduce mass differences into the
phase space factors which multiply 1 M l2 in the expression for rates, leaving un-
touched relations (32).
Predictions so obtained are in an unpleasant disagreement with experiments.
For example one has :
4 P$ 2
Rate (yl*'-+cx) = (phasespace ratio) x -- = - . - N 12%
Rate (Yr+ -+A n ) 6 Pi 3
(pr,* = momentum of Z or A particle) ,
11.1. We know that the electromagnetic field interacts with hadrons in such a
way t o conserve T, and Y ; moreover, due t o the smallness of the coupling
constant, we can describe such interaction with a pertubative method, starting
from an interaction Lagrangian of the form :
Group Theory and Unitary Symniatry Models 375
Here j, (x)is the electromagnetic current of hadrons. The structure of this operator
depends upon the dynamics of strong interactions themselves, which we a t
present do not know in detail.
Hence the Lagrangian (1) has to be considered as a phenomenological device
which allows us t o explicitly consider the dependence on electromagnetic field
of these interactions, whereas unknown effects of strong interactions are lumped
into the local operator j, (x).We note that the matrix elements of j, (x)between
physical states are connected t o measurable quantities (form factors).
By definition we have
Jj4(z)d32 =Q (2)
and by charge, Y , T , conservation :
a+{, (c) 0
Jj4(2)d3x = Q = T , + -21 Y = 3 H 1 + H , .
This suggests j, (x)t o be composed of two parts: the current of the third com-
ponent of total isospin plus one half the hypercharge current
and now j , ( T 3 ) ( ~ ) and j,( y, (x)have the same transformation properties (under
X U,) as T,and Y . From this it is easy t o see that j, (x)commutes with E,, and
the same for Pint.If we put (see sect. 8.7.c)
we see that U,, U , have the same commutation relations of the isotopic spin
generators, and are called the U-spin generators. Hence electromagnetic inter-
actions a t all perturbative orders conserve charge and U-spin, which in this
case play the same role as hypercharge and I-spin for medium strong interactions.
From the previous analysis we see that everything we said for medium strong
interactions can be applied to electromagnetic interactions, substituting the
partition into I-spin and Y-multiplets of irreducible S U , representations with
a partition into U-spin and Q-multiplets.
We give for reference the decomposition of the pseudoscalar mesons octet into
U-spin multiplets (Pig. 10).
27 Zeitschrift ,,Fortschritte der Physik", Heft 7
376 and L. MAIANI
(i.DE FRANCESCHL
One sees that n+,K+ and K-, n- constitute two U-spin doublets whereas K, and
KOare members of an U-spin triplet. no and q are eigenstates of U,, with eigen-
value zero, but they are not eigenstates of U2. Instead the combinations
KO I K*
Relations (5) are of course riot satisfied by the actual masses; this is natural
because we have neclected the important contribution of medium strong inter-
actions. We can however detlucc from ( 5 )the relation:
1
- -,u(XoA) = p ( n ) - p(h)
1/3
- fi p ( R X o )= p ( n ) - p ( Z o ) ,
where , u ( X o ~ 2is) called the transition magnetic moment between Zo and A, and
appears for example in the amplitude of the decay X 0--t A -k y.
The experimental information available up to now does not allow to test any-one
of the relations (7).
For pseudoscalar mesons we obtain easily the results [ a d ] that the form factors
of K+ and n+ arc equal, whereas the form factors of KO and are zero.
I n fact :
form factor (KO) = form factor (KO) by U-spin
form factor ( K O ) = - form factor (KO) by Charge-conjugation
c) q and no - two photons decay - We said previously that the U , = Q = 0
eigenstates of U 2 are (in the pseudoscalar meson octet)
27"
378 and L. MAIANI
G . DE FRANCESCHI
Observe now that 1 +) cannot decay, by U-conservation, into two photons (which
have U = 0) so that the amplitudes for (no, q) -+ 2 y are equal t o
I n the amplitude for the procesb (8) the matrix element of the electromagnetic
current between the w(or rp) state and the vacuum is involved. Now if we write
(sect. 10.5 (22))
1 'p) = cos 0 18) sin 8 I 1) +
lW)=sinep) +coseji),
we see that we have t o evaluate the matrix elements :
0! jca (XI 10)
(81 i P ( 4 10).
Group Theory and Unitary Symmetry Models 379
j p ( x ) transforms as the S representation and, whereas the product 8 @ S con-
tains the singlet representation (to which the vacuum is assigned) this represen-
tation is not contained in the product 8 @ 1. Hence, by Wigner-Eckart theorem,
the first matrix element vanishes. We conclude tha,t an SU, singlet cannot decay
through the one photon channel, so that only the components of w and 'p over
the octet can go (at first order) into e+ +
e- (p+ p-) . +
The ratio
r(rp-+e+e-)
= (phase space corrections) tan2 8
r ( w -+e+e-)
12.1. Very exciting results have been obtained by the application of S U , sym-
metry to the field of weak interactions of hadrons. We will not give here an
extensive discussion of all the topics involved, limiting ourselves t o sketch the
theory for leptonic decays21). These processes have the general form
A -+ I + -+ B + B' +
V]
A --+l +v1,
where A , B, B' . . . are strongly interacting particles, 1 is a lepton (e, p), V I the
corresponding neutrino. A few significant examples are :
AT AS A&
+ e+ + v,
I
n+ -+ no -1 0 -1
z++ p+ + vw -1 0 -1
AS=0
n+p+e-+G6 1 0 1
Z+-th+e++v, -1 0 -1
I
(1)
+
K+-tnO e+ + v , -1 -1
AX=O
K+ -+ p+ v,+ -1 -1
A+p+e-+ie '12 1 1
Z - + n f e - +V, '12 1 1
J , and j, are the weak currents associated to hadrons and leptons, and G is the
weak coupling constant determined from p-decay.
21) For a more detailed treatment of weak interactions see [as, 43, 441.
380 6. DE FRANCESOHI
and L. MAIANI
For what concerns j,, expermental findings agree with the form (in terms of
lepton fields)
jCL =6 , L V + Y 5 ) ve +P Y P ( 1 + Y d V p , (3)
whereas the structure of J,, which is thought to be determined by stronginter-
actions, is not known in detail. The amplitudes of the processes we are consider-
ing are expressed as matrix elements of ( 2 )
the matrix elements of j,A can be calculated, so that the actual difficultyis consti-
tuted by the other terms.
For what concerns space-time properties, experiments indicate that Jll can be
splitted up in two terms: one transforming as a vector and the other as a pseudo-
vector : we will refer to them as t o the vector and axial-vector currents :
J p = JF + 4; .
Thc first one is responsible e.g. for the p-decay of x + :
x+ --f xo + e+ + vo ,
and the other for the usual x decay:
x+ + p+ + v,.
Let us define A S , A T,, A Q respectively as the changes of strangeness, isospin,
charge suffered by hadrons22) (see (1)).Then we can divide leptonic decays into
two classes: A S = 0, and AS =+
0 decays. Experiments indicate [43] that the
following selection rules are satisfied within errors (which are however rather
large) :
i) for A X = 0 decays, IAT,I = 1 (hence O Q = & I )
ii) for AS =+ 0 decays A S = A Q , ~AAS'I= 1 (hence lAT,j = 112).
This selection rule forbids for example the process :
C + + n -t-c++v,
which actually has not been seen [as].
We arc then led t o write J Mas:
+ J F i l ) + J f C O ) + p"1'
J
,-
- JFiO)
/* ' (4)
where J,Y(O)is the strangeness conserving and tJ;h-il) is the A S = 1 part of the
vector current, and the same for axial currents.
22) These changes are not independent. From the Gell-Mann Nisliijima formula:
1
A Q = A T , 3- - - - A S .
2
Group Theory and Unitary Syrnmct,ry Moclcls 381
In t'he context of isotopic spin symmetry one could attempt t o explain selection
rule i), assuming the .1 S = 0 current to tranforni under S U - as a tensor operator,
belonging to the isospin one representation.
The CVC (conserved vector Current) theory of PEYNMANN and GELL-MANX[46]
embodies this assumption in a much stronger statement; they identify the 3 S = 0
vector currents JL'OJ,(JzioJ)+ with the T,= & 1 components of the isospin
current, i.e. of the current which arises from S U , invariance (see sect. 7 . 3 . ) .
The T , = 0 component is the isovector part of the electromagnetic current. As
a consequence, insofar we neglect electromagnetic effects, JFoiand (JjJ7'O')-+ are
conserved :
a,,Jp =. 0 i, (,J"<"J)-r-
' / A ii 10.
We quote three significant examples of predictions made by CVC theory [42]:
a ) t,he agreement between the Fermi constant in p-decay of nuclei and the Ferrni
constant in p-decay ;
b) the rate of the pion p-decay can be calculated from the neutron p-decay obta-
ing a result in agreement with experiments :
c) using Wigner-Eckart theorem one can express the matrix elements of JiL7[Oi,
(J;'O')+, for example between nucleon states, in terms of the electromagnetic
current matrix elements, i.e. of nucleon e. m. form factors which are known from
e - N scattering experiments. Preliminary data on neutrino experiments seem
to support this prediction.
The 3s = 0 axial current is also assumed to transform as an isovector, but in
contrast to JJ(OJit is not conserved.
The simplest generalization t o include strange particles decays is t o assume J:
and J;1' to possess well-defined transformation properties under SU ,. I n parti-
cular if we assume JJ and J f to belong to octets of tensor operators, it,is obvious
that selection rules i), ii) are fulfilled (of course the converse is not true: i)ii) do
not imply octet currents).
Now XU, symmetry provides us an octet of vector currents (see sect. 7.3) which,
in absence of symmetry breaking interactions, are conserved. We could then iden-
tify various pieces of Jzwith such currents. This however would imply each current
to be coupled to leptons with the same strenght, i.e. with the same coupling
constant as the A S = 0 part. On the contrary experiments give coupling con-
stants for strange particle decays which are smaller of an order of magnitude than
the 3 S = 0 couplings.
N. CABIBBOhas assumed [17-48] that the vector current coupled to leptons
has the form
J L = oos 0 J:;(OJ$- sin 0 J;,-'I), (5)
where JL(O), (J;(OJ)+,J:L*'l',(JF[li).4, are the T , = & I , 1' = 0, 1', = & 1/2,
We give here a brief account of predictions which can be obtained from this
theory in the case of baryons decays. For a detailed discussion see [43, 491.
Matrix elements of JZ between baryon states can be expressed in terms of 0
and of two reduced matrices (8 @ 8 contains the eight representation twice).
By analogy with sect. 10.6 let us call them F,-, D,. I n the limit of zero momentum
transfer, they can be written as :
F , (0) = 1, D, (0) = 0 ,
so that the matrix elements of J L are determined by 0.
For what concerns axial currents, again we have two reduced matrices, P, D
which in the same limit as (7) can be written as
( p i+o)
P~ = G ( P J ) F ( k 2 ) y ~ ~ y 5 U(Ic2 )
0, = i ( P J D(k2)Y,Y5 U(Pi), (8)
but now the lack of conservation of J f does not allow t o obtain additional con-
ditions on F ( 0 ) ,D (0).
Concluding we see t h a t the baryons decays (in the limit li2 + O , absence of
symmetry breaking interactions 23) are described in terms of three numbers :
6( O ) , D (01, 6.
H. COURANTe t al. [45] find two sets of parameters consistent with data, both
with near the same value of 0, but differing for the ratio FID.
It is remarkable that near the same value of 0(0 ~ l0.25) i has been given by
CABIBBOin the paper previously quoted, comparing the decays :
K+ -> p+ + v,
iT+ + p+ f vp.
(I(+I (JA'I')+10)
(K+ I (JA'O')+10)
i.e. to tan2 0.
23) When the symmetry breaking interaction is taken into account at the first order, it can be
shown that our conclusions on tho vector current remain correct [SO].
Group Theory and Unitary Symmetry Models 3 83
13.1. The idea of a higher symmetry, in the concrete formulation of the eightfold
way model, undoubtely greatly improves the phenomenological description of
the behaviour of strongly interacting particles. However the very fact that it
works rises the need of understanding a t a deeper, dynamical level, how the sym-
metry is brought about (as well as its partial violation).
For an up-to-date discussion of the various attempts made in this direction, using
bootstrap technique a.s well as field theoretical methods, see [23]. To the latter
class belongs SCHWINCER’S W , model [51] as well as the popular “quarks” or
“aces” model (proposed by ZWEICand GELL-MA”].
Another interesting problem is that of the connection between internal and
space-time symmetries.
I n this context very promising is the S U , model proposed by GURSEY,RADICATI
and PAIS[52]; see also [53]) who treat on the same footing spin, isospin and
hypercharge.
References
[?3] A. SALAM, Symmetry of Strong Interactions, Intern. Conf. on High Ener. Phys.. Dubna
(1964).
[24] Y. NE'EYAN,Nuclear Phys. 26, 222 (1961).
[HI R. E. BEHRENDS, J. DREITLEIN,C. FRONSDAL, W. LEE, Rev. mod. Phys. 34, 1 (1962).
[26] R. GATTO, Theoretical Physics - Seminar IAEA (1963).
[27] V. E. B-4RNER et al., Phys. Rev. Letters 12, 134 (1964).
[28] A. H. ROSENFELD et al., Rev. mod. Phys. 36,977 (1964).
[29] P. MCNAMEE, S. J. and FRANK CHILTON,Rev. mod. Phys. 36, 1005 (1964).
[30] S. OKUBO,Phys. Letters 5, 165 (1963).
[31] J. J. SAKURAI, Phys. Rev. Letters 8, 472 (1962).
[32] S. OKUBO,Progr. theor. Phys. (Kyoto), 27, 949 (1962).
[33] J. DE SWART,Rev. mod. Phys. 35, 916 (1963).
[34] A. W. MARTIN,K. C. WALI, Phys. Rev. 130,2455 (1963).
[35] S. HATSUEADE et al., Photoproduction of Strange Particles and Unitarity Symmetry,
Rochester. U-R-875-52 (1964).
[36] J. J. SAKURAI, Proc. of the Intern. School of Phys. ,,Enrico Fermi" Course XXVI.
[37] R. H. DALITZ,Proc. of the Sienna Conf. on Elem. Part. (1963), Plenary Sessions, vol.
11.
[35] V. GUPTA,V. SINQH,Phys. Rev. 135, B 1442 (1964).
[39] S. COLEMAN, S. L. GLASHOW, Phys. Rev. Letters 6, 423 (1961).
[40] N. CABIBBO, R. GATTO, Nuovo Cimento 21, 872 (1961).
[ U ] R. GATTO,Proc. of the Intern. Sienna Conf. on Elem. Part. (1963), Plenary Sessions,
vol. 11.
[42] J. D. JACKSON, Weak Interactions, Brandeis Summer Institute 1962, Benjamin, New
York (1963).
[43] L. B. OKUN,Weak Interactions and Unitary Symmetry, Institute of Theor. and Exper.
Phys., Moscow (1964) (in russian).
[44] Weak Interactions and Higher Symmetries, Acta Physica Austriaca. Suppl. I Springer
Verlag Wien - New York (1964).
[45] H. COURANT et al., Phys. Rev. 136, B 1791 (1964).
[46] R. P. FEYXMAX, M. GELL-MANN, Phys. Rev. 109, 193 (1958).
[17] N. CABIBBO, Phys. Rev. Letters 10, 531 (1963).
[48] Proc. of the Intern. School of Phys. ,,Ettore Majorana", Benjamin, New York (1964).
[29] L. WOLFENSTEIN, Phys. Rev. 135, B 1436 (1964).
[SO] M. ADEMOLLO,R. GATTO,Phys. Rev. Letters 13, 264 (1964).
[51] J. SCHWINGER, Phys. Rev. Letters 12, 237 (1964).
[52] F. GURSEY,L. A. RADICATI, Phys. Rev. Letters 13, 173 (1964).
A. PAIS,Phys. Rev. Letters 13, 175 (1964);
F. GURSEY,A. PAIS,L. A. RADICATI, Phys. Rev. Letters 13, 299 (1964).
[ 5 3 ] B. SAKITA, Phys. Rev. 135, B 1756 (1964).
Herausgeber: Prof. Dr. Frank Easchluhn, Prof. Dr. Artur Liische, Prof. Dr. Rudolf Ritschl und Prof. Dr. Robert
Rompe; Manuskripte sind zu richten an die Schriftleitung: DipLPhys. Lutx Rothkirch, 11. Physikalisches Institut
der Humboldt-Universitat Berlin, 104 Berlin, Hessische Str. 2. Verlag : Akademie-Verlag GmbH, 108 Berlin.
Leipziger Str. 3-4, Fernruf: 220441, Telex-Nr. 011773, Postscheckkonto: Berlin 35021. Die Zeitschrift ,,Fort-
schritte der Physik" erscheint monatlich; Berugspreis eines Heftes MDN 6, -. Bestellnummer dieses Heftes:
1007/13/7. - Satz und Druck: VEB Druckhaus ,,Maxim Gorki", 74 Altenburg, Bez. Leiprig, CarI-von-Ossietzky-
St,r. 30- 31. - Veroffentlicht unter der Lizenznummer 1324 des Presseamtes beimvorsitzenden des Ministerratea
dor Deutschen Demokratischen Republik.