Details of Some Representation Spaces For (3) : I A A A A
Details of Some Representation Spaces For (3) : I A A A A
Details of Some Representation Spaces For (3) : I A A A A
The defining representation is of course just the set of matrices that one uses to identify
what is meant by the group in the first place. As a group, given a set of basis vectors in C3 ,
these are then the set of all 3 × 3 unitary matrices with determinant +1. As an algebra, these
are the set of all 3 × 3 skew-Hermitean, traceless matrices, Xa , usually created by writing
them as i times Hermitean, traceless matrices. A standard (complete) set of those Hermitean
traceless matrices, {λa |a = 1, . . . , 8}, was originally given by Gell-Mann, being a plausible
generalization of the Pauli matrices used for studies of the group SU(2). It is customary, then,
to use a set of 8 real parameters, aa , lying in an appropriate (finite) range of values, such that
any group element may be written in the form
i a a
g = g(a) ≡ e− 2 a λa
≡ e−ia Ta
. (1.1)
(Note that some people put +i there instead of −i; used consistently this simply changes the
signs of the parameters, and nothing more.)
In the abstract algebra we often use the Chevalley basis, {h1 , h2 }, along with the associated
simple roots, {α1 , α2 }, and their eigen-elements, {e1 , e2 } and {f1 , f2 }. The simple roots have
values on the Cartan subalgebra given by the Cartan matrix, A, with elements Aij ≡ αi (hj ),
all of which are integers:
This defining representation is variously denoted in many distinct ways, all equivalent:
a. [1 0] describes it in terms of its highest weight, the (two) integers indicating multiples of
the lattice basis vectors, {µa }r1 , which are vectors reciprocal to the basis of root vectors,
{αa }r1 . (Each of these sets of vectors constitute a basis for the root space, which is the
space of vectors dual to the Cartan subalgebra of the set of generators, i.e., the Lie algebra,
of the group in question. The Cartan subalgebra has dimension r, usually referred to as
the rank of the Lie algebra.) As this particular representation is described simply by one
a vector space of three dimensions. This is not necessarily completely descriptive because
there could be more than one inequivalent three-dimensional representation, which would
then require bars, primes, etc., in order to distinguish them, along with a table giving the
meaning of those bars and primes. In fact, for SU(3) there is one other, which, as we will
c. the tensor components v a also form a common description for this representation. (Since
this is the defining representation, this is truly just the statement that this is the most
natural sort of a vector, in analogy with the set of coordinate differences, xa .) As is usual
with tensor components, they really describe a tensor relative to some chosen, fixed set of
basis elements. In this case, using the language of “ket-vectors,” we suppose that |vi is
some vector in the representation space, that { |a i | a = 1, 2, 3} form a basis for the space,
and, then, that the vector can be written in terms of its components with respect to that
basis, i.e., |vi = v a |a i, where there is a sum on all the allowed values of a.
d. the Young tableaux symbol, ¤ , is also used to describe the type of tensor representation
it is. This simple symbol simply means that the tensor in question has one (contravariant)
index, so that there are not really any symmetry questions to ask. For more complicated
Knowing that this representation has highest weight [1 0], we would then want to find
the rest of the allowed weights, to describe a basis for the entire vector space. The standard
calculation, using lowering operators, tells us that this representation has a total of 3 weights,
2
each one with a single associated weight vector, since the fundamental representation weights
where the other columns indicate various equivalent sets of nomenclature for the same things,
i.e., the three weight vectors. Each of the statements that these are equivalent nomenclatures
We begin that process with the first column, which simply lists the three allowed weights.
These are the sets of eigenvalues for the r = 2 members of the representation of the basis
elements that have been already chosen for the Cartan subalgebra. Denoting an arbitrary such
weight by [m n], the (Dirac notion) ket-vector that is the eigenvector for that eigenvalue by
|[m, n]i, and the representatives of the hi ’s by Hi , this means that H1 |[m, n]i = m|[m n]i and
H2 |[m, n]i = n|[m n]i. The second column only gives simpler names to those 3 vectors so that,
for example H1 |2 i = −1|2 i and H2 |2 i = +1|2 i. The third column, however, now begins to treat
our vectors, and operators, in terms of a specific choice of basis, so that they are represented
as row matrices (row vectors) and square matrices, respectively. In particular, it picks out the
three eigenvectors themselves as the basis set, so that their representations, as given in this
third column, are just Kronecker delta’s, i.e., 0’s and 1’s. One must then check that it is in fact
just these eigenvectors that were chosen as a basis when Gell-Mann wrote down his λ-matrices,
although surely one would suspect this. We do this by creating the representations for the hi ’s
in terms of the actual 3 × 3 matrices, and checking that they have the right eigenvalues on
3
these vectors. For the representations of the hi ’s, we have
+1 0 0 √ 0 0 0
H1 = λ3 = 2Tz = 0 −1 0 , H2 = 3 1
2 λ8 − 2 λ3 = 32 Y −Tz = 0 +1 0 . (1.4)
0 0 0 0 0 −1
As they are diagonal, we know that the basis is indeed their joint eigenvectors. We can read the
three weights off by just reading jointly down their diagonals. This gives the expected answers,
indicating, also, that we have the eigenvectors in the correct order. However, it is worth
continuing to check the rest of the properties that weight vectors should have, namely that
they should be transformed one into another via the proper lowering and/or raising operators.
In this case, knowing the weight diagrams, and their relationships to the roots, we can write
E1 |[1, 0]i = 0 = E2 |[1, 0]i , F1 |[1, 0]i = |[−1, 1]i , F2 |[1, 0]i = 0
E1 |[−1, 1]i = |[1, 0]i , E2 |[−1, 1]i = 0 , F1 |[−1, 1]i = 0 , F2 |[−1, 1]i = |[0, −1]i , (1.5)
E1 |[0, −1]i = 0 , E2 |[0, −1]i = |[−1, 1]i , F1 |[0, −1]i = 0 = F2 |[0, −1]i .
On the other hand, via the 3 × 3 λ-matrices, we have their specific representatives:
0 1 0 0 0 0
1
E1 = 2 (λ1 +iλ2 ) = 0 0 0 = (F1 )T , 1
E2 = 2 (λ6 +iλ7 ) = 0 0 1 = (F2 )T . (1.6)
0 0 0 0 0 0
It is then straightforward matrix arithmetic to verify that each of the requirements listed in
Eqs. (1.5) are indeed satisfied by the matrices in Eqs. (1.6), acting on the vectors as listed
in Eqs. (1.3). One should do so! The rest of the description of the matrices defining the Lie
algebra are then given by the remaining non-zero commutators, which are simply
0 0 1
1
[E1 , E2 ] = 2 (λ4 + iλ5 ) = 0 0 0 = −([F1 , F2 ])T . (1.7)
0 0 0
The set of these eight matrices do close upon themselves upon multiplication using the com-
mutator product.
4
At this point it is worth commenting on the particular labelling given for the “simpler
names” for the vectors given in the second column, or, if you prefer, on the ordering that
was given to the basis vectors |a i. This was made so that their tensor components, v a , would
be, respectively, δ1a , δ2a , and δ3a . Keep track of this as we go through higher-level tensor
representations.
The further columns of Eq. (1.3) concern the standard physical interpretation of these
weight vectors in the “Standard Model” in particle physics. (Before describing them, I should
surely note that there are many other physical systems in which the group SU(3) is an im-
portant symmetry group; an example is gravity-induced ocean waves far out at sea, described
by the Boussinesq equation.) The first of the second set of columns gives the names of the
three quarks, simply as symbols, under the standard identifications, while the next one, the
fifth column, gives those same names, as labels for Dirac ket-vectors. They are the up quark,
u, the down quark, d, and the strange quark, s, respectively. The last column lists the appro-
priate physical properties of these quarks, which allows the respective identifications vis-a-vis
Gell-Mann’s names for the choices of diagonal λ-matrices, i.e., their values of Tz and Y . Given
the relationships to H1 and H2 , as in Eqs. (1.4), one can use the individual weight vectors
to calculate the values for the z-component of isospin, i.e., the eigenvalue for Tz , and the
hypercharge, the eigenvalue for Y . It is interesting to plot these various objects on a lattice
diagram for su(3), where we see that while the two simple roots meet at an angle of 120◦ and
the two reciprocal lattice vectors meet at an angle of 60◦ , the physical quantities are linear
combinations chosen so as to be orthogonal, i.e., they meet at 90◦ . Such a plot is on the next
page.
As a “summary” of the content of this discussion, and especially Eqs. (1.3), we can de-
scribe the physical content of this representation via a single column vector, which contains as
elements the vectors for the three particles it describes. They are put there so that the tensor
components that describe a state with just one of those particles is written by simply replacing
5
its symbol with a one, and the other symbols with a zero:
u
d . (1.8)
s
that D(g1 )D(g2 ) = D(g1 g2 ). Denoting D(gi ) simply by Di , and D(g1 g2 ) by D12 , we can try
to use various standard operations on linear operators to get other representations, which only
sometimes works:
D1 D2 = D12 =⇒ (D2 )−1 (D1 )−1 = (D12 )−1 , (D2 )† (D1 )† = (D12 )† ,
(D1 )∗ (D2 )∗ = (D12 )∗ , {(D1 )† }−1 {(D2 )† }−1 = {(D12 )† }−1 , {(D1 )T }−1 {(D2 )T }−1 , (2.1)
which gives us three additional representations for any given one: the conjugate one, the
may well be that some of these are equivalent to the original one, so that not all of these are
any compact, simple Lie group can be chosen so that the matrices are unitary; therefore for
our case, for example, the last of the three entries in Eqs. (2.1) is identical to the original
representation. As well, the other two entries there are identical, so that we have acquired at
most one other by this means. In the case of SU(3), it turns out that this one other one, the
complex conjugate one, is indeed inequivalent, in general, and in particular for the defining
one. Therefore, there is one other 3-dimensional representation, which we label by the symbol
a a a
Ta∗ a
(−Ta∗ ) a
(−TaT )
e−ia Ta
≡ D(ga ) = (e−ia Ta ∗
) = e+ia = e−ia = e−ia (2.2)
6
where the last equality is caused by the fact that the Ta are all Hermitean matrices, so that
conjugate representation for the generators. In particular, one should note that it leaves
invariant the form of the Chevalley commutation relations, Eqs. (1.2), so that these would be
acceptable names for these matrices. However, as it puts minus signs into the representations
of the simple raising and lowering operators it violates the standard versions of the choices of
phase for such things, due originally to Condon and Shortley, for su(2), and more recently to
de Swart for su(3)—see Rev. Mod. Phys. 35 916-939 (1963). Luckily one can notice that
the Chevalley relations are unchanged if one agrees to change the signs of just the raising and
lowering operators, doing nothing else. Therefore, following de Swart, we will agree to label
H i = − Hi ,
E i = +Fi , F i = +Ei ,
(2.3)
0 0 0
E 12 ≡ [E 1 , E 2 ] = [F1 , F2 ] = 0 0 0 = −([F 1 , F 2 ])T .
−1 0 0
Since in this representation, the matrices that describe the diagonal elements have opposite
signs to those that were used in the 3-representation, the various weights for the 3 have opposite
signs from the 3-representation; therefore the weights for this representation are [−1 0], [1 − 1],
and [0 1]. As well, since the raising and lowering operators have been switched, it follows that
(the negative of) the old highest weight will be the lowest weight, and vice versa. Therefore
we recognize this representation as that one with highest weight [0 1]; i.e., the other basic (or
7
Eqs. (1.3) will then be
Most of the columns are, by now, either self-explanatory or already explained, but not quite
all. The three weights have their weight vectors numbered in a way that might be unexpected;
this comes about from our desire to ensure that the particles and their associated antiparticles
have the same number, only with the location being switched from contravariant to covari-
ant. Therefore we may think of the mapping between the two sorts of levels as the particle
conjugation operator. Therefore we will treat an arbitrary vector in this vector space, for the
them from the other ones. Also notice that by taking the matrix presentations for these vectors
as row vectors instead of column vectors, it automatically forces us to use the transpose of the
original representation matrices when they are being acted upon, only insisting that we recall
correctly the various minus signs involved. This is intentional and one should not now go
ahead and use an extra transpose on the matrices, unless it is the intent to write them in the
One column, however, in the table is quite new; it is the one with 3×3 matrices in it. While
there is not any operator in these tensor spaces that raises or lowers indices, in the mode of the
“metric” in a physical coordinate space, there are two universal, invariant tensors within the
purview of SU(n), namely the Levi-Civita symbol, ²a1 a2 ...an , used to create the determinants
8
the Kronecker delta symbol, δba , which are the components of the identity matrix—invariant
because everything commutes with the identity:
We may therefore use the Levi-Civita symbol to trade any covariant tensor index for two
upper, skew-symmetric indices. In 3 dimensions, the two approaches have exactly the same
number of degrees of freedom, namely 3. The Levi-Civita symbol trades a covariant index
of 1, for example, with the contravariant pair {2, 3}. Either the tensor components va , or
the equivalent tensor components, W bc = W [bc] may be used to describe the vectors in this
representation space. The matrices shown are then the presentations of the contravariant,
skew-symmetric tensor components for the vectors.
Considering the indices of both the fundamental in terms of contravariant indices only
allows them to be considered equivalently from the point of view of Young tableaux:
3 ⇐⇒ ¤ , 3 ⇐⇒ ¤
¤. (2.6)
The single vector summary statement about the particle content of the representation is then
the following:
0 s −d
(u d s) ⇐⇒ −s 0 u . (2.7)
d −u 0
I first want to perform the simplest of the various products of representations, namely the
product of the defining representation and its conjugate one, the result for which is already
known to be
3 ⊗ 3 = [1 0] ⊗ [0 1] = [1 1] ⊕ [0 0] = 8 ⊕ 1 , (3.1)
¤⊗¤ ¤¤ ¤
¤=¤ ⊕ ¤.
¤
9
When the tensors for these two representations are treated simultaneously, each with 3 compo-
nents, a total of 9 components are generated, described by the generic tensor v a wb ; however,
this representation is no longer irreducible, i.e., there are sub-vector spaces that transform
only among themselves. In this particular case it is the multiples of the identity that trans-
form among themselves alone, and the other, traceless parts of the tensor that also tranform
among themselves alone; this is simply the statement made earlier that the components of
the identity tensor—the Kronecker delta symbol—are left invariant by every group element.
Therefore, we may use the Kronecker delta to reduce this representation into its irreducible
parts:
v a wb = (v a wb − 13 δba v m wm ) + 13 δba v m wm ≡ W a b + 13 δba W . (3.2)
This is simply the mathematical justification for the decomposition already stated in Eq. (3.1),
since the (traceless) tensor components W a b have eight independent degrees of freedom, and
the single invariant W corresponds to a single degree of freedom, namely an SU(3)-scalar.
Now, what happens to the weights in this product, and the particle identifications? This is
the interesting part that we want to understand better.
To begin with we need to see what the representation of the generators of the algebra
look like, as they act on the vectors in this 8-dimensional representation, as labelled in tensor
terms. Writing out the previous equation, Eq. (3.2), in terms of the symbolic matrices that
present the tensor components, we have
u uu ud us
d ( u d s ) = du dd ds
s su sd ss
(2uu − dd − ss)/3 ud us 1 0 0
W
= du (2dd − uu − ss)/3 ds + 0 1 0 ,
3
su sd (2ss − uu − dd)/3 0 0 1
u
W ≡ ( u d s ) d = uu + dd + ss .
s
(3.3)
Notice that the matrix presentation for W a b is in fact traceless!
10
Since this is a direct product representation, that means that for every group element g,
there is a (3×3) matrix D(g) that gives the action of that group element on the space of vectors
that span the representation space for the 3 = [1 0] representation, and the associated matrix,
D(g) for the 3 = [0 1] representation, as generated by the matrices defined in Eqs. (2.3). When
both act together, of course the one representation simply ignores the vectors for the other
representation, and then we have the reverse operation for the other one; this is the meaning of
“the direct product”; therefore, the action of the generators alone is given by T ⊗ I3 + I3 ⊗ T ,
for any abstract Lie algebra elements t ∈ G:
¡ ¢
T (W a b ) = T a c W c b − ηT T c b W a c = T a c δbd − ηT δca T d b W c d ≡ {T [1 1] }ad cb W c d , (3.4)
We begin by calculating the weights of the various states in the matrix presentation of the
tensor for the representation, using the two Cartan subalgebra elements. As a simple example,
we calculate the weight vector for the upper right-hand corner of the matrix presentation:
[1 1]
Hi W 1 3 = H [1 1] us = {Hi ⊗ I3 + I3 ⊗ H i }us = {Hi ⊗ I3 − I3 ⊗ (H)i }us
(3.5)
= (Hi u)s − u(Hi s) = [1 0] us + u{[0 1] s} = ([1 0] + [0 1])us = [1 1]us .
11
Repeating this sort of calculation for the entire matrix of elements W a b , we present the matrix
with just the weights that are obtained for each element:
[0 0] [2 − 1] [1 1]
weights for elements: [−2 1] [0 0] [−1 2] . (3.6)
[−1 − 1] [1 − 2] [0 0]
As expected, every location corresponds to only one weight; however, there are two, or perhaps
three, locations with the same weight, namely [0 0]. This was expected since we knew that
[0 0] was degenerate, i.e., there are two different vectors in the representation space that have
this weight. (There are not three; the appearance of three is simply because there are only 8
degrees of freedom in this matrix, since it is traceless, so one of the 3 appearances of [0 0] is
just the negative of the sum of the other two.)
To break the degeneracy on the weights, we need to recall that one of them corresponds
to an isospin singlet, with I = 0 as well as Iz = 0, while the other one is simply the Iz = 0
member of an isospin triplet, with I = 1. Therefore, we expect the triplet member to have
contributions only in the W 1 1 and W 2 2 locations, since these are the places that the matrix
Tz can reach. We therefore choose A as the triplet state, and B as the singlet. Insuring that
A occurs only in the two upper locations, and that the matrix is traceless gives us
W 11 = A + B , W 2 2 = −A + B , W 3 3 = −2B . (3.7)
The result is
uu − dd uu + dd − 2ss
A= √ , B= √ . (3.8)
2 6
allowing us to re-write explicitly our matrix one last time, describing in detail the elements
irreducibly, in terms of the usual pseudoscalar meson octet members, where A becomes the
neutral π 0 and B the neutral η 0 meson:
(uu − dd)/2 + (uu + dd − 2ss)/6 ud us
du (dd − uu)/2 + (uu + dd − 2ss)/6 ds
su sd (2ss − uu − dd)/3
0 √ 0
√ + +
π / 2+η / 6 √π √ K
− 0 0
= π −π / 2 + η / 6 K0 .
− 0 0
√
K K −2η / 6
(3.9)
12
IV. The Baryon Representations
We now want to consider the quark triple product, and its tensor version:
3 ⊗ 3 ⊗ 3 = [1 0]3 = [3 0] ⊕ [1 1] ⊕ [1 1] ⊕ [0 0] = 10 ⊕ 8 ⊕ 8 ⊕ 1 , (4.1)
¤ ⊗ ¤ ⊗ ¤ = ¤¤¤ ⊕ ¤¤ ⊕ ¤¤ ⊕ ¤
¤.
¤ ¤ ¤
The following equality splits the simple triple product into its constituent irreducible parts,
following the usual rules for creating appropriate Young Tableaux:
ta ub v c = 16 {ta ub v c + ta uc v b + tb uc v a + tb ua v c + tc ua v b + tc ub v a }
+ 13 {ta ub v c + tb ua v c − tc ub v a − tb uc v a }
+ 31 {ta ub v c + tc ub v a − tb ua v c − tc ua v b }
+ 61 {ta ub v c − ta uc v b + tb uc v a − tb ua v c + tc ua v b − tc ub v a } ,
≡Dabc + √2 ²acd X b d
6
+ √2 ²abd Y c d
6
+ √1 ²abc Z
6
, (4.2)
X bd ≡ √1 ²def (te ub v f
6
+ tb ue v f ) ,
Y bd ≡ √1 ²def (te uf v b
6
+ tb uf v e ) ,
Z≡ √1 ²def td ue v f .
6
a. The tensor components Dabc are totally symmetric in their indices, and thereby constitute
the 10-dimensional representation, [3 0].
b. The next two braces of components are skew-symmetric in a, c and in a, b, respectively,
and therefore allow us to pick out the equivalent tensors X b d and Y c d , which one sees by
calculation are traceless; they constitute two copies of the [1 1] representation, which are
each 8 dimensional.
c. The last brace of components are skew-symmetric in every pair of indices, and are therefore
simply proportional to the Levi-Civita symbol itself, since we have only 3 values for these
indices to take. This is then just a scalar representation, [0 0], and is 1 dimensional.
13
Picking one of the eight-dimensional representations, say X a b , we may calculate weights
by using the individual 3 × 3 representations of the Cartan subalgebra for each part. While
each representative will be composed of three quarks, since it is skew in two of those three
indices, it is not allowed to have all three the same. Therefore the highest weight will be
the one composed of two u-quarks and one d-quark, composed so as to create the element
X 1 3 ∝ ²312 t1 u1 v 2 + ²321 t1 u2 v 3 = 2uud − duu − udu. This has Iz = 1/2 and Y = 1, for
physical particles, we identify this one as the proton, p, and normalize it appropriately for a
physical state:
√
[1 , 1] = p+ = (2uud − duu − udu)/ 6 = X 1 3 . (4.3)
Then we can easily outline the other, non-degenerate members of the octet:
√
[−1 , 2] = n0 = X 23 = (−udd − dud + 2ddu)/ 6 ,
√
[2 , −1] = Σ+ = X 12 = (2uus − suu − usu)/ 6 ,
√
[−2 , 1] = Σ− = X 21 = (2dds − sdd − dsd)/ 6 , (4.4)
√
[1 , −2] = Ξ0 = X 32 = (2ssu − uss − sus)/ 6 ,
√
[−1 , −1] = Ξ− = X 31 = ((2ssd − dss − sds)/ 6 .
The two independent but degenerate states that occur in the diagonal elements of the matrix
both have weight [0 , 0]. Using the method espoused at Eqs. (3.7), and the tensor products in
√
X 1 1 = (dus + uds − sud − usd)/ 6 = A + B ,
√
X 2 2 = (sdu + dsu − uds − dus)/ 6 = −A + B ,
√
X 3 3 = (usd + sud − dsu − sdu)/ 6 = −2B ,
0 √ √
Σ = 2A = (2uds + 2dus − usd − sud − dsu − sdu)/ 12 ,
=⇒ √ (4.5)
0
Λ = 6B = (dsu + sdu − usd − sud)/2 ,
14
Inserting all these states into the matrix form for X a b , we then have
Σ0 Λ0
√
2
+ √
6
Σ+ p+
0
Λ0
((X a b )) =
Σ− Σ
−√ 2
+ √
6
n0 . (4.6)
0
Λ
Ξ− Ξ0 −2 √ 6
15