, ,

Let

K

be a field of characteristic 0 and let

C_{n d}

be the pure trace algebra generated by all traces of products

tr (X_{i_{1}} \dots X_{i_{k}})

of the generic

n \times n

matrices

X_{i} = (x_{p q}^{(i)})

p, q = 1, \dots, n

i = 1, \dots, d

. The algebra

C_{n d}

coincides with the algebra of invariants of the general linear group

G L_{n} = G L_{n} (K)

acting by simultaneous conjugation on

d

matrices of size

n \times n

For a background on the algebras of matrix invariants see, e.g. [F, DF] . Traditionally, a result giving the explicit generators of the algebra of invariants of a linear group

G

is called a first fundamental theorem of the invariant theory of

G

and a result describing the relations between the generators a second fundamental theorem.

Classical invariant theory implies that

C_{n d}

is finitely generated. By the Noether normalization theorem

C_{n d}

contains a homogeneous set of algebraically independent elements

{f_{1}, \dots, f_{k}}

called a homogeneous system of parameters, where

k = (d - 1) n^{2} + 1

is the transcendence degree of the quotient field of

C_{n d}

, such that

C_{n d}

is integral over the polynomial algebra

S = K [f_{1}, \dots, f_{k}]

. A more precise result, see Van den Bergh [V] , gives that

C_{n d}

is a finitely generated free module over a suitable chosen

S

. An upper bound for the generating set of the algebra

C_{n d}

is given in terms of PI-algebras. By the Nagata-Higman theorem the polynomial identity

x^{n} = 0

implies the identity

x_{1} \dots x_{N} = 0

for some

N = N (n)

. If

N

is minimal with this property, then

C_{n d}

is generated by traces of products

tr (X_{i_{1}} \dots X_{i_{k}})

of degree

k \leq N

. This estimate is sharp if

d

is sufficiently large. The values of

N (n)

are between

n (n + 1) / 2

and

n^{2}

, bounds obtained respectively by Kuzmin [Ku] and Razmyslov [R] (see also [DF] for an exposition of the results in [Ku] ). The only explicitly known values of

N (n)

are

N (2) = 3

and

N (3) = 6

(an old result of Dubnov [Du] ), and

N (4) = 10

(a result of Vaughan-Lee [VL] ). Recently, Shestakov and Zhukavets [SZ] obtained that the class of nilpotency of 2-generated nil algebras of nil index 5 is 15. All these results agree with the conjecture of Kuzmin [Ku] that

N (n) = n (n + 1) / 2

Explicit minimal sets of generators of

C_{n d}

are found in few cases only. By a theorem of Sibirskii [S] ,

C_{2 d}

is generated by

tr (X_{i})

1 \leq i \leq d

tr (X_{i} X_{j})

1 \leq i \leq j \leq d

tr (X_{i} X_{j} X_{k})

1 \leq i < j < k \leq d

. Teranishi [T1] found the following system of generators of

C_{32}

\begin{matrix} tr (X), tr (Y), tr (X^{2}), tr (X Y), tr (Y^{2}), \end{matrix}

\begin{matrix}  \end{matrix}

\begin{matrix} tr (X^{3}), tr (X^{2} Y), tr (X Y^{2}), tr (Y^{3}), tr (X^{2} Y^{2}), tr (X^{2} Y^{2} X Y), \end{matrix}

where

X

Y

are generic

3 \times 3

matrices. He showed that the first ten of these generators form a homogeneous system of parameters of

C_{32}

and

C_{32}

is a free module with generators 1 and

tr (X^{2} Y^{2} X Y)

over the polynomial algebra on these ten elements. Abeasis and Pittaluga [AP] found a system of generators of

C_{3 d}

, for any

d \geq 2

, in terms of representation theory of the symmetric and general linear groups, in the spirit of its use in theory of PI-algebras. Teranishi [T1, T2] found also a minimal set of generators and a homogeneous system of parameters of

C_{42}

, all of them being traces of products of the generators.

It has turned out that the systems of generators of Sibirskii [S] for

C_{2 d}

and of Teranishi [T1] for

C_{32}

are not very convenient if we want to find the defining relations of the algebra. (Compare the relation for

C_{23}

found by Sibirskii [S] with the defining relations for

C_{2 d}

of one of the authors [D2] with respect to another natural system of generators.) Also, it follows from the description of the generators of

C_{32}

, that

tr (X^{2} Y^{2} X Y)

satisfies a quadratic equation with coefficients depending on the other ten generators. The explicit (but very complicated) form of the equation was found by Nakamoto [N] , over

Z

, with respect to a slightly different system of generators.

A much simple relation with respect to the system of generators of

C_{32}

in [AP] was found by Aslaksen, Drensky and Sadikova in [ADS] .

The purpose of the present paper is to find a minimal system of generators of

C_{42}

in the spirit of the work of Abeasis and Pittaluga [AP] used in their description of the generators of

C_{n d}

. Applied to

C_{42}

, the idea in [AP] is the following. The general linear group

G L_{2}

acts canonically on the free associative algebra

K 〈 x, y 〉

and this induces an action on the pure trace algebra

P (x, y) = K [tr (z_{1} \dots z_{k}) | z_{i} = x, y, i = 1, \dots, k, k = 1, 2, \dots],

and on the algebra

C_{42}

(because

C_{42}

is a homomorphic image of

P (x, y)

under the natural homomorphism extending the map

x \to X

y \to Y

, where

X

and

Y

are two generic

4 \times 4

matrices). The algebra

C_{42}

has a system of generators of degree

\leq 10

. Without loss of generality we may assume that this system consists of traces of products

tr (Z_{1} \dots Z_{k})

Z_{i} = X, Y

. Let

T_{k}

be the subalgebra of

C_{42}

generated by all traces

tr (Z_{1} \dots Z_{l})

of degree

l \leq k

Z_{i} = X, Y

. Clearly,

T_{k}

is also a

G L_{2}

-submodule of

C_{42}

. Let

C_{42}^{(k + 1)}

be the homogeneous component of degree

k + 1

C_{42}

. Then the intersection

T_{k} \cap C_{42}^{(k + 1)}

is a

G L_{2}

-module and has a complement

G_{k + 1}

C_{42}^{(k + 1)}

, which is the

G L_{2}

-module of the “new” generators of degree

k + 1

We may assume that

G_{k + 1}

is a submodule of the

G L_{2}

-module spanned by traces of products

tr (Z_{1} \dots Z_{k + 1})

of degree

k + 1

. The

G L_{2}

-module of the generators of

C_{42}

G = G_{1} \oplus G_{2} \oplus \dots \oplus G_{10} .

The authors of [AP] work with the multilinear elements in the free algebra which are consequences of the polynomial identity

x^{n} = 0

. Instead, we prefer to work directly in the algebras

P (x, y)

and

C_{42}

with more essential use of representation theory of the general linear group, in the spirit of [D1] . Our main result is that the minimal generating

G L_{2}

-module

G

C_{42}

decomposes as

\begin{matrix} G = W (1, 0) \oplus W (2, 0) \oplus W (3, 0) \oplus W (4, 0) \oplus W (2, 2) \oplus W (3, 2) \end{matrix}

\begin{matrix} \oplus W (4, 2) \oplus W (3, 3) \oplus W (4, 3) \oplus W (5, 3) \oplus W (4, 4) \oplus W (6, 3) \oplus W (5, 5), \end{matrix}

\begin{matrix}  \end{matrix}

where

W (λ_{1}, λ_{2})

is the irreducible

G L_{2}

-module corresponding to the partition

(λ_{1}, λ_{2})

. For

λ = (λ_{1}, λ_{2}) \neq (5, 5)

, one may choose as a generator of

W (λ_{1}, λ_{2})

the element

w_{λ} (X, Y) = tr ((X Y - Y X)^{λ_{2}} X^{λ_{1} - λ_{2}}) .

For

λ = (5, 5)

a generator of

W (5, 5)

w_{(5, 5)} (X, Y) = tr ((X Y - Y X)^{3} (X^{2} Y^{2} - X Y Y X - Y X X Y + Y^{2} X^{2})) .

It seems to us that our generators are more convenient than the usual ones, if we want to find a system of defining relations of

T_{42}

. Most of the computations are performed with Maple, using standard functions only. Some of the obtained relations and technical results may be used for concrete calculations with traces of generic

4 \times 4

matrices.

1 Preliminaries

The group

G L_{2}

acts in a canonical way on the vector space with basis

{x, y}

and this action induces a diagonal action on the free algebra

K 〈 x, y 〉

g (z_{1} \dots z_{k}) = g (z_{1}) \dots g (z_{k}), z_{i} = x, y, g \in G L_{2} .

The action of

G L_{2}

K 〈 x, y 〉

induces an action on the pure trace algebra

P (x, y)

g (tr (z_{1} \dots z_{k})) = tr (g (z_{1}) \dots g (z_{k})), z_{i} = x, y, g \in G L_{2} .

If we fix the generic

n \times n

matrices

X, Y

, the algebra of invariants

C_{n 2}

becomes a homomorphic image of

P (x, y)

under the natural homomorphism extending the map

x \to X

y \to Y

and the

G L_{2}

action is transferred on

C_{n 2}

. The

G L_{2}

-module

K 〈 x, y 〉

is a direct sum of irreducible polynomial modules, described in terms of partitions

λ = (λ_{1}, λ_{2})

. We denote by

W (λ)

the corresponding

G L_{2}

-module.

The

G L_{2}

-submodules and factor modules

W

K 〈 x, y 〉

inherit its natural bigrading which counts the entries of

x

and

y

in each monomial. We denote by

W^{(p, q)}

the corresponding homogeneous component. The formal power series

H (W, t, u) = \sum_{p, q \geq 0} dim (W^{(p, q)}) t^{p} u^{q}

is called the Hilbert series of

W

. The Hilbert series of

W (λ)

is the Schur function

S_{λ} (t, u)

which, in the case of two variables, has the very simple form

S_{λ} (t, u) = (t u)^{λ_{2}} (t^{λ_{1} - λ_{2}} + t^{λ_{1} - λ_{2} - 1} u + \dots + t u^{λ_{1} - λ_{2} - 1} + u^{λ_{1} - λ_{2}}) .

The Hilbert series of

W

plays the role of its character, and

W = \oplus W^{\oplus m (λ)} (λ), m (λ) \in N \cup {0},

i.e.

W (λ)

participates in

W

with multiplicity

m (λ)

, if and only if

H (W, t, u) = \sum m (λ) S_{λ} (t, u) .

The Hilbert series of

C_{42}

was calculated by Teranishi [T2] , with some misprints in the expression, and, corrected, by Berele and Stembridge [BS] :

\begin{matrix} H (C_{42}, t, u) = \frac{P_{C} (t, u)}{(1 - t) (1 - u) Q_{C} (t, u)}, \end{matrix}

(1)

P_{C} (t, u) = (1 - e_{2} + e_{2}^{2}) (1 - e_{1} e_{2} + e_{1} e_{2}^{2} + e_{1}^{2} e_{2}^{2} + e_{1} e_{2}^{3} - e_{1} e_{2}^{4} + e_{2}^{6}),

e_{1} = t + u, e_{2} = t u,

Q_{C} (t, u) = (1 - t^{2}) (1 - t^{3}) (1 - t^{4}) (1 - u^{2}) (1 - u^{3}) (1 - u^{4})

(1 - t u)^{2} (1 - t^{2} u)^{2} (1 - t u^{2})^{2} (1 - t^{3} u) (1 - t u^{3}) (1 - t^{2} u^{2}) .

We shall need the decomposition as a sum of Schur functions of the first 11 homogeneous components

h_{n} (t, u)

of the symmetric formal power series

\begin{matrix} h (t, u) = \frac{P_{C} (t, u)}{Q_{C} (t, u)} = (1 - t) (1 - u) H (C_{42}, t, u) = \sum_{n \geq 0} h_{n} (t, u) . \end{matrix}

(2)

Using the formula

\begin{matrix} \frac{1}{1 - z} = 1 + z + z^{2} + \dots, \end{matrix}

(3)

direct calculations show that

\begin{matrix} \begin{matrix} h_{0} = 1, h_{1} = 0, h_{2} = t^{2} + t u + u^{2} = S_{(2, 0)} (t, u), \\ h_{3} = S_{(3, 0)} (t, u), h_{4} = 2 S_{(4, 0)} (t, u) + 2 S_{(2, 2)} (t, u), \\ h_{5} = S_{(5, 0)} (t, u) + S_{(4, 1)} (t, u) + 2 S_{(3, 2)} (t, u), \\ h_{6} = 3 S_{(6, 0)} (t, u) + S_{(5, 1)} (t, u) + 5 S_{(4, 2)} (t, u) + S_{(3, 3)} (t, u), \\ h_{7} = 2 S_{(7, 0)} (t, u) + 2 S_{(6, 1)} (t, u) + 5 S_{(5, 2)} (t, u) + 4 S_{(4, 3)} (t, u), \\ h_{8} = 4 S_{(8, 0)} (t, u) + 2 S_{(7, 1)} (t, u) + 10 S_{(6, 2)} (t, u) + 6 S_{(5, 3)} (t, u) + 8 S_{(4, 4)} (t, u), \\ h_{9} = 3 S_{(9, 0)} (t, u) + 3 S_{(8, 1)} (t, u) + 10 S_{(7, 2)} (t, u) + 13 S_{(6, 3)} (t, u) + 8 S_{(5, 4)} (t, u), \\ h_{10} = 5 S_{(10, 0)} (t, u) + 4 S_{(9, 1)} (t, u) + 16 S_{(8, 2)} (t, u) \\ + 16 S_{(7, 3)} (t, u) + 24 S_{(6, 4)} (t, u) + 5 S_{(5, 5)} (t, u) . \end{matrix} \end{matrix}

(4)

Any submodule

W (λ) = W (λ_{1}, λ_{2})

K 〈 x, y 〉

is generated by a unique, up to a multiplicative constant, homogeneous element

w_{λ} (x, y)

of degree

λ_{1}

and

λ_{2}

with respect to

x

and

y

, respectively, called the highest weight vector of

W (λ)

It is characterized by the following property, see [DEP, ADF] and [K] for the version which we need. We state it for two variables only. The homogeneous polynomial

w_{λ} (x, y) \in K 〈 x, y 〉

of degree

(λ_{1}, λ_{2})

is a highest weight vector for some

W (λ_{1}, λ_{2})

if and only if

δ (w_{λ} (x, y)) = 0

, where

δ

is the derivation of

K 〈 x, y 〉

defined by

δ (x) = 0

δ (y) = x

. If

W_{i}

i = 1, \dots, k

, are

k

isomorphic copies of

W (λ)

and

w_{i} \in W_{i}

are highest weight vectors, then the highest weight vector of any submodule

W (λ)

of the direct sum

W_{1} \oplus \dots \oplus W_{k}

has the form

ξ_{1} w_{1} + \dots + ξ_{k} w_{k}

for some

ξ_{i} \in K

Any

k

linearly independent highest weight vectors can serve as generators of the

G L_{2}

-module

W_{1} \oplus \dots \oplus W_{k}

. Of course, similar arguments work for submodules and homomorphic images of the

G L_{2}

-module

K 〈 x, y 〉

The multiplicity of

W (λ)

K 〈 x, y 〉

is equal to the degree

d (λ)

of the corresponding irreducible representation of the symmetric group

S_{k}

k = λ_{1} + λ_{2}

, or, equivalently, to the number of standard

λ

-tableaux. For the standard

λ

-tableau

T_{σ}

=

σ (1)

\dots

σ (2 λ_{2} - 1)

σ (2 λ_{2} + 1)

\dots

σ (k)

σ (2)

\dots

σ (2 λ_{2})

corresponding to

σ \in S_{k}

k = λ_{1} + λ_{2}

, we associate a highest weight vector

w (T_{σ})

K 〈 x, y 〉

. When

σ = ɛ

is the identity of

S_{k}

we fix

w (T_{ɛ}) = (x y - y x)^{λ_{2}} x^{λ_{1} - λ_{2}}

= \sum_{ρ_{1}, \dots, ρ_{s} \in S_{2}} sign (ρ_{1} \dots ρ_{s}) z_{ρ_{1} (1)} z_{ρ_{1} (2)} \dots z_{ρ_{s} (1)} z_{ρ_{s} (2)} x^{λ_{1} - λ_{2}},

where

z_{1} = x

z_{2} = y

s = λ_{2}

. For

σ

arbitrary, we define

w (T_{σ})

in a similar way, but the skew-symmetries are in positions

(σ (1), σ (2)), \dots, (σ (2 s - 1), σ (2 s))

instead of the positions

(1, 2), \dots, (2 s - 1, 2 s)

(and the positions with fixed

x

are

σ (2 s + 1), \dots, σ (k)

instead of

2 s + 1, \dots, k

). For the standard

(p, q)

-tableau

T

=

a_{1}

\dots

a_{q}

a_{q + 1}

\dots

a_{p}

b_{1}

\dots

b_{q}

we denote the corresponding highest weight vector

tr (w (T))

w (T) = w (\begin{matrix} a_{1} & \cdot \cdot \cdot & a_{q} & a_{q + 1} & \cdot \cdot \cdot & a_{p} \\ b_{1} & \cdot \cdot \cdot & b_{q} \end{matrix}) .

2 The $G L_{2}$ -modules of formal traces

We denote by

U_{n}

the vector subspace of all traces of formal products of length

n

in the variables

x

and

y

in the pure trace algebra

P (x, y)

Proposition 2.1. The

G L_{2}

-modules

U_{n}

spanned by the traces of the products of length

n \leq 10

are decomposed as follows:

U_{1} = W (1), U_{2} = W (2), U_{3} = W (3), U_{4} = W (4) \oplus W (2, 2),

U_{5} = W (5) \oplus W (3, 2), U_{6} = W (6) \oplus 2 W (4, 2) \oplus W (3, 3),

U_{7} = W (7) \oplus 2 W (5, 2) \oplus 2 W (4, 3),

U_{8} = W (8) \oplus 3 W (6, 2) \oplus 3 W (5, 3) \oplus 3 W (4, 4),

U_{9} = W (9) \oplus 3 W (7, 2) \oplus 6 W (6, 3) \oplus 4 W (5, 4),

U_{10} = W (10) \oplus 4 W (8, 2) \oplus 7 W (7, 3) \oplus 10 W (6, 4) \oplus 4 W (5, 5) .

Proof. We calculate the Hilbert series $H (U_{n}, t, u)$ of $U_{n}$ . Since $H (U_{n}, t, u)$ is a symmetric polynomial in $t, u$ , it is sufficient to count the number $h_{n k}$ of traces of degree $k$ with respect to $x$ and $n - k$ with respect to $y$ , $k \geq n - k$ . There is one trace depending on $x$ only, $tr (x^{n})$ , and one trace, $tr (x^{n - 1} y)$ , of degree $(n - 1, 1)$ .
Hence $h_{n 0} = h_{n 1} = 1$ and we may consider the case $k \leq n - 2$ only. Since the traces are invariant under cyclic permutation, we may assume that they are of the form $tr (x^{a_{1}} y^{b_{1}} \dots x^{a_{p}} y^{b_{p}}),$ where $a_{i}, b_{i} > 0$ and $a_{1} + \dots + a_{p} = k$ , $b_{1} + \dots + b_{p} = n - k$ . We order the words $x^{a_{1}} y^{b_{1}} \dots x^{a_{p}} y^{b_{p}}$ comparing lexicographically the $2 p$ -tuples $(a_{1}, \dots, a_{p}, b_{1}, \dots, b_{p})$ and fix the representative of $tr (x^{a_{1}} y^{b_{1}} \dots x^{a_{p}} y^{b_{p}})$ to be as high as possible.
Cases $n = 1, 2, 3$ . These cases are obvious because the coefficients $h_{n k}, h_{n, n - k}$ of $H (U_{n}, t, v)$ are equal to 1 for $k = 0, 1$ and $H (U_{1}, t, v) = t + u = S_{(1)} (t, u),$ $H (U_{2}, t, u) = t^{2} + t u + u^{2} = S_{(2)} (t, u),$ $H (U_{3}, t, u) = t^{3} + t^{2} u + t u^{2} + u^{3} = S_{(3)} (t, u) .$ This means that $U_{1} = W (1)$ , $U_{2} = W (2)$ and $U_{3} = W (3)$ .
Case $n = 4$ . The only elements of degree $(2, 2)$ are $tr (x^{2} y^{2}), tr (x y x y)$ and this implies $H (U_{4}, t, u) = t^{4} + t^{3} u + 2 t^{2} u^{2} + t u^{3} + u^{4} = S_{(4)} (t, u) + S_{(2^{2})} (t, u),$ and hence $U_{4} = W (4) \oplus W (2, 2)$ .
Case $n = 5$ . The traces of degree $(3, 2)$ are $tr (x^{3} y^{2})$ , $tr (x^{2} y x y)$ and $H (U_{5}, t, u) = t^{5} + t^{4} u + 2 t^{3} u^{2} + 2 t^{2} u^{3} + t u^{4} + u^{5} = S_{(5)} (t, u) + S_{(3, 2)} (t, u),$ which gives $U_{5} = W (5) \oplus W (3, 2)$ . Case $n = 6$ . The traces of degree $(4, 2)$ are $tr (x^{4} y^{2})$ , $tr (x^{3} y x y)$ , $tr (x^{2} y x^{2} y)$ , these of degree $(3, 3)$ are $tr (x^{3} y^{3})$ , $tr (x^{2} y^{2} x y)$ , $tr (x^{2} y x y^{2})$ , $tr (x y x y x y)$ , and $H (U_{6}, t, u) = t^{6} + t^{5} u + 3 t^{4} u^{2} + 4 t^{3} u^{3} + 3 t^{2} u^{4} + t u^{5} + u^{6}$ $= S_{(6)} (t, u) + 2 S_{(4, 2)} (t, u) + S_{(3^{2})} (t, u),$ $U_{6} = W (6) \oplus 2 W (4, 2) \oplus W (3, 3) .$ Case $n = 7$ . The traces of degree $(5, 2)$ are $tr (x^{5} y^{2})$ , $tr (x^{4} y x y)$ , $tr (x^{3} y x^{2} y)$ , these of degree $(4, 3)$ are $tr (x^{4} y^{3})$ , $tr (x^{3} y^{2} x y)$ , $tr (x^{3} y x y^{2})$ , $tr (x^{2} y^{2} x^{2} y)$ , $tr (x^{2} y x y x y)$ and $H (U_{7}, t, u) = t^{7} + t^{6} u + 3 t^{5} u^{2} + 5 t^{4} u^{3} + 5 t^{3} u^{4} + 3 t^{2} u^{5} + t u^{6} + u^{7}$ $= S_{(7)} (t, u) + 2 S_{(5, 2)} (t, u) + 2 S_{(4, 3)} (t, u),$ $U_{7} = W (7) \oplus 2 W (5, 2) \oplus 2 W (4, 3) .$ Case $n = 8$ . The traces of degree $(6, 2)$ are $tr (x^{6} y^{2}), tr (x^{a_{1}} y x^{a_{2}} y), (a_{1}, a_{2}) = (5, 1), (4, 2), (3, 3),$ and $h_{82} = 4$ , the traces of degree $(5, 3)$ are $tr (x^{5} y^{3}),$ $tr (x^{a_{1}} y^{b_{1}} x^{a_{2}} y^{b_{2}}), (a_{1}, a_{2}) = (4, 1), (3, 2), (b_{1}, b_{2}) = (2, 1), (1, 2),$ $tr (x^{a_{1}} y x^{a_{2}} y x^{a_{3}} y), (a_{1}, a_{2}, a_{3}) = (3, 1, 1), (2, 2, 1),$ $h_{83} = 7$ . Finally, the traces of degree $(4, 4)$ are $tr (x^{4} y^{4}), tr (x^{a_{1}} y^{b_{1}} x^{a_{2}} y^{b_{2}}),$ $(a_{1}, a_{2}) = (3, 1), (b_{1}, b_{2}) = (3, 1), (2, 2), (1, 3),$ $(a_{1}, a_{2}) = (2, 2), (b_{1}, b_{2}) = (3, 1), (2, 2),$ $tr (x^{2} y^{b_{1}} x y^{b_{2}} x y^{b_{3}}), (b_{1}, b_{2}, b_{3}) = (2, 1, 1), (1, 2, 1), (1, 1, 2),$ $tr (x y x y x y x y) .$ Hence $h_{84} = 10$ , $H (U_{8}, t, u) = t^{8} + t^{7} u + 4 t^{6} u^{2} + 7 t^{5} u^{3} + 10 t^{4} u^{4} + 7 t^{3} u^{5} + 4 t^{2} u^{6} + t u^{7} + u^{8}$ $= S_{(8)} (t, u) + 3 S_{(6, 2)} (t, u) + 3 S_{(5, 3)} (t, u) + 3 S_{(4, 4)} (t, u),$ $U_{8} = W (8) \oplus 3 W (6, 2) \oplus 3 W (5, 3) \oplus 3 W (4, 4) .$ Case $n = 9$ . The traces of degree $(7, 2)$ are $tr (x^{7} y^{2}), tr (x^{a_{1}} y x^{a_{2}} y), (a_{1}, a_{2}) = (6, 1), (5, 2), (4, 3),$ and $h_{92} = 4$ , these of degree $(6, 3)$ are $tr (x^{6} y^{3}), tr (x^{a_{1}} y^{b_{1}} x^{a_{2}} y^{b_{2}}),$ $(a_{1}, a_{2}) = (5, 1), (4, 2), (b_{1}, b_{2}) = (2, 1), (1, 2), (a_{1}, a_{2}) = (3, 3), (b_{1}, b_{2}) = (2, 1),$ $tr (x^{a_{1}} y x^{a_{2}} y x^{a_{3}} y), (a_{1}, a_{2}, a_{3}) = (4, 1, 1), (3, 2, 1), (3, 1, 2), (2, 2, 2),$ $h_{93} = 10$ . The traces of degree $(5, 4)$ are $tr (x^{5} y^{4}), tr (x^{a_{1}} y^{b_{1}} x^{a_{2}} y^{b_{2}}), (a_{1}, a_{2}) = (4, 1), (3, 2), (b_{1}, b_{2}) = (3, 1), (2, 2), (1, 3),$ $tr (x^{a_{1}} y^{b_{1}} x^{a_{2}} y^{b_{2}} x^{a_{3}} y^{b_{3}}),$ $(a_{1}, a_{2}, a_{3}) = (3, 1, 1), (2, 2, 1), (b_{1}, b_{2}, b_{3}) = (2, 1, 1), (1, 2, 1), (1, 1, 2),$ $tr (x^{2} y x y x y x y),$ $h_{94} = 14$ . The Hilbert series and the decomposition of $U_{9}$ are, respectively, $H (U_{9}, t, u) = t^{9} + t^{8} u + 4 t^{7} u^{2} + 10 t^{6} u^{3} + 14 t^{5} u^{4}$ $+ 14 t^{4} u^{5} + 10 t^{3} u^{6} + 4 t^{2} u^{7} + t u^{8} + u^{9}$ $= S_{(9)} (t, u) + 3 S_{(7, 2)} (t, u) + 6 S_{(6, 3)} (t, u) + 4 S_{(5, 4)} (t, u),$ $U_{9} = W (9) \oplus 3 W (7, 2) \oplus 6 W (6, 3) \oplus 4 W (5, 4) .$ Case $n = 10$ . The traces of degree $(8, 2)$ are $tr (x^{8} y^{2}), tr (x^{a_{1}} y x^{a_{2}} y), (a_{1}, a_{2}) = (7, 1), (6, 2), (5, 3), (4, 4),$ and $h_{10, 2} = 5$ , these of degree $(7, 3)$ are $tr (x^{7} y^{3}), tr (x^{a_{1}} y^{b_{1}} x^{a_{2}} y^{b_{2}}),$ $(a_{1}, a_{2}) = (6, 1), (5, 2), (4, 3), (b_{1}, b_{2}) = (2, 1), (1, 2),$ $tr (x^{a_{1}} y x^{a_{2}} y x^{a_{3}} y), (a_{1}, a_{2}, a_{3}) = (5, 1, 1), (4, 2, 1), (4, 1, 2), (3, 3, 1), (3, 2, 2),$ $h_{10, 3} = 12$ . The traces of degree $(6, 4)$ are $tr (x^{6} y^{4}), tr (x^{a_{1}} y^{b_{1}} x^{a_{2}} y^{b_{2}}), (a_{1}, a_{2}) = (5, 1), (4, 2), (b_{1}, b_{2}) = (3, 1), (2, 2), (1, 3),$ $(a_{1}, a_{2}) = (3, 3), (b_{1}, b_{2}) = (3, 1), (2, 2),$ $tr (x^{a_{1}} y^{b_{1}} x^{a_{2}} y^{b_{2}} x^{a_{3}} y^{b_{3}}),$ $(a_{1}, a_{2}, a_{3}) = (4, 1, 1), (3, 2, 1), (3, 1, 2), (b_{1}, b_{2}, b_{3}) = (2, 1, 1), (1, 2, 1), (1, 1, 2),$ $(a_{1}, a_{2}, a_{3}) = (2, 2, 2), (b_{1}, b_{2}, b_{3}) = (2, 1, 1),$ $tr (x^{a_{1}} y x^{a_{2}} y x^{a_{3}} y x^{a_{4}} y), (a_{1}, a_{2}, a_{3}, a_{4}) = (3, 1, 1, 1), (2, 2, 1, 1), (2, 1, 2, 1),$ $h_{10, 4} = 22$ . Finally, the traces of degree $(5, 5)$ are $tr (x^{5} y^{5}), tr (x^{a_{1}} y^{b_{1}} x^{a_{2}} y^{b_{2}}),$ $(a_{1}, a_{2}) = (4, 1), (3, 2), (b_{1}, b_{2}) = (4, 1), (3, 2), (2, 3), (1, 4),$ $tr (x^{a_{1}} y^{b_{1}} x^{a_{2}} y^{b_{2}} x^{a_{3}} y^{b_{3}}),$ $(a_{1}, a_{2}, a_{3}) = (3, 1, 1), (2, 2, 1),$ $(b_{1}, b_{2}, b_{3}) = (3, 1, 1), (1, 3, 1), (1, 1, 3), (2, 2, 1), (2, 1, 2), (1, 2, 2),$ $tr (x^{2} y^{b_{1}} x y^{b_{2}} x y^{b_{3}} x y^{b_{4}}),$ $(b_{1}, b_{2}, b_{3}, b_{4}) = (2, 1, 1, 1), (1, 2, 1, 1), (1, 1, 2, 1), (1, 1, 1, 2),$ $tr (x y x y x y x y x y),$ hence $h_{10, 5} = 26$ . The Hilbert series and the decomposition of $U_{10}$ are, respectively, $H (U_{10}, t, u) = t^{10} + t^{9} u + 5 t^{8} u^{2} + 12 t^{7} u^{3} + 22 t^{6} u^{4}$ $+ 26 t^{5} u^{5} + 22 t^{4} u^{6} + 12 t^{3} u^{7} + 5 t^{2} u^{8} + t u^{9} + u^{10},$ $= S_{(10)} (t, u) + 4 S_{(8, 2)} (t, u) + 7 S_{(7, 3)} (t, u) + 10 S_{(6, 4)} (t, u) + 4 S_{(5, 5)} (t, u),$ $U_{10} = W (10) \oplus 4 W (8, 2) \oplus 7 W (7, 3) \oplus 10 W (6, 4) \oplus 4 W (5, 5) .$ □

Lemma 2.2. For

λ = (λ_{1}, λ_{2}) ⊢ n

n \leq 10

, the following trace polynomials form a basis of the subspace of

U_{n}

consisting of all

λ

-highest weight vectors:

λ = (n)

n \geq 1

w = w (\begin{matrix} 1 & \cdot \cdot \cdot & n \end{matrix}) = tr (x^{n});

λ = (2, 2)

w = \frac{1}{2} w (\begin{matrix} 1 & 3 \\ 2 & 4 \end{matrix}) = \frac{1}{2} tr ([x, y]^{2}) = - tr (x^{2} y^{2}) + tr (x y x y);

λ = (3, 2)

w = w (\begin{matrix} 1 & 3 & 5 \\ 2 & 4 \end{matrix}) = tr ([x, y]^{2} x) = - tr (x^{3} y^{2}) + tr (x^{2} y x y);

λ = (4, 2)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 6 \\ 2 & 4 \end{matrix}) = tr ([x, y]^{2} x^{2}) = - tr (x^{4} y^{2}) + 2 tr (x^{3} y x y) - tr (x^{2} y x^{2} y),

w_{2} = w (\begin{matrix} 1 & 2 & 5 & 6 \\ 3 & 4 \end{matrix}) = tr (x^{4} y^{2}) - tr (x^{2} y x^{2} y);

λ = (3, 3)

w = \frac{1}{3} w (\begin{matrix} 1 & 3 & 5 \\ 2 & 4 & 6 \end{matrix}) = tr (x^{2} y^{2} x y) - tr (x^{2} y x y^{2});

λ = (5, 2)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 6 & 7 \\ 2 & 4 \end{matrix}) = tr ([x, y]^{2} x^{3}) = - tr (x^{5} y^{2}) + 2 tr (x^{4} y x y) - tr (x^{3} y x^{2} y)),

w_{2} = w (\begin{matrix} 1 & 2 & 5 & 6 & 7 \\ 3 & 4 \end{matrix}) = tr (x^{5} y^{2}) - tr (x^{3} y x^{2} y);

λ = (4, 3)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 7 \\ 2 & 4 & 6 \end{matrix}) = tr (x^{3} y^{2} x y) - tr (x^{3} y x y^{2}),

w_{2} = w (\begin{matrix} 1 & 2 & 5 & 7 \\ 3 & 4 & 6 \end{matrix}) = - tr (x^{4} y^{3}) + 2 tr (x^{3} y^{2} x y)

+ tr (x^{3} y x y^{2}) - tr (x^{2} y^{2} x^{2} y) - tr (x^{2} y x y x y);

λ = (6, 2)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 6 & 7 & 8 \\ 2 & 4 \end{matrix})

= tr ([x, y]^{2} x^{4}) = - tr (x^{6} y^{2}) + 2 tr (x^{5} y x y) - tr (x^{4} y x^{2} y),

w_{2} = w (\begin{matrix} 1 & 2 & 5 & 6 & 7 & 8 \\ 3 & 4 \end{matrix}) = tr (x^{6} y^{2}) - tr (x^{4} y x^{2} y),

w_{3} = w (\begin{matrix} 1 & 3 & 4 & 6 & 7 & 8 \\ 2 & 5 \end{matrix}) = - tr (x^{6} y^{2}) + tr (x^{5} y x y) + tr (x^{4} y x^{2} y) - tr (x^{3} y x^{3} y);

λ = (5, 3)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 7 & 8 \\ 2 & 4 & 6 \end{matrix})

= tr (x^{4} y^{2} x y) - tr (x^{4} y x y^{2}) - tr (x^{3} y^{2} x^{2} y) + tr (x^{3} y x^{2} y^{2}),

w_{2} = w (\begin{matrix} 1 & 2 & 5 & 7 & 8 \\ 3 & 4 & 6 \end{matrix}) = - tr (x^{5} y^{3}) + 2 tr (x^{4} y^{2} x y)

- 2 tr (x^{3} y^{2} x^{2} y) + 2 tr (x^{3} y x^{2} y^{2}) - tr (x^{2} y x^{2} y x y),

w_{3} = w (\begin{matrix} 1 & 3 & 5 & 6 & 8 \\ 2 & 4 & 7 \end{matrix}) = tr (x^{4} y^{2} x y) - tr (x^{4} y x y^{2});

λ = (4, 4)

w_{1} = \frac{1}{2} w (\begin{matrix} 1 & 3 & 5 & 7 \\ 2 & 4 & 6 & 8 \end{matrix}) = \frac{1}{2} tr ([x, y]^{4}) = - 2 tr (x^{2} y^{2} x y x y) + 2 tr (x^{2} y x y^{2} x y)

- 2 tr (x^{2} y x y x y^{2}) + tr (x^{2} y^{2} x^{2} y^{2}) + tr (x y x y x y x y),

w_{2} = w (\begin{matrix} 1 & 2 & 5 & 7 \\ 3 & 4 & 6 & 8 \end{matrix}) = tr (x^{3} y^{3} x y) - 2 tr (x^{3} y^{2} x y^{2}) + tr (x^{3} y x y^{3})

- tr (x^{2} y^{2} x y x y) + 2 tr (x^{2} y x y^{2} x y) - tr (x^{2} y x y x y^{2}) - 2 tr (x^{2} y^{3} x^{2} y) + 2 tr (x^{2} y^{2} x^{2} y^{2}),

w_{3} = \frac{1}{2} w (\begin{matrix} 1 & 2 & 5 & 6 \\ 3 & 4 & 7 & 8 \end{matrix}) = tr (x^{4} y^{4}) - 2 tr (x^{3} y^{2} x y^{2}) + tr (x^{2} y x y^{2} x y)

- 2 tr (x^{2} y^{3} x^{2} y) + 2 tr (x^{2} y^{2} x^{2} y^{2});

λ = (7, 2)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 6 & 7 & 8 & 9 \\ 2 & 4 \end{matrix})

= tr ([x, y]^{2} x^{5}) = - tr (x^{7} y^{2}) + 2 tr (x^{6} y x y) - tr (x^{5} y x^{2} y),

w_{2} = w (\begin{matrix} 1 & 2 & 5 & 6 & 7 & 8 & 9 \\ 3 & 4 \end{matrix}) = tr (x^{7} y^{2}) - tr (x^{5} y x^{2} y),

w_{3} = w (\begin{matrix} 1 & 2 & 4 & 6 & 7 & 8 & 9 \\ 3 & 5 \end{matrix}) = tr (x^{6} y x y) - tr (x^{4} y x^{3} y);

λ = (6, 3)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 7 & 8 & 9 \\ 2 & 4 & 6 \end{matrix}) = tr ([x, y]^{3} x^{3}) = tr (x^{5} y^{2} x y) - tr (x^{5} y x y^{2})

- tr (x^{4} y^{2} x^{2} y) + tr (x^{4} y x^{2} y^{2}) - tr (x^{3} y x^{2} y x y) + tr (x^{3} y x y x^{2} y),

w_{2} = w (\begin{matrix} 1 & 2 & 5 & 7 & 8 & 9 \\ 3 & 4 & 6 \end{matrix}) = - tr (x^{6} y^{3}) + 2 tr (x^{5} y^{2} x y)

- 2 tr (x^{4} y^{2} x^{2} y) + tr (x^{4} y x^{2} y^{2}) + tr (x^{3} y^{2} x^{3} y) - tr (x^{3} y x^{2} y x y),

w_{3} = w (\begin{matrix} 1 & 2 & 4 & 7 & 8 & 9 \\ 3 & 5 & 6 \end{matrix})

= tr (x^{5} y x y^{2}) - tr (x^{5} y^{2} x y) + tr (x^{4} y x^{2} y^{2}) - tr (x^{4} y^{2} x^{2} y),

w_{4} = w (\begin{matrix} 1 & 3 & 5 & 6 & 8 & 9 \\ 2 & 4 & 7 \end{matrix}) = tr (x^{5} y^{2} x y) - tr (x^{5} y x y^{2}) + tr (x^{4} y x^{2} y^{2})

- tr (x^{4} y x y x y) - tr (x^{3} y^{2} x^{3} y) + tr (x^{3} y x^{2} y x y) + tr (x^{3} y x y x^{2} y) - tr (x^{2} y x^{2} y x^{2} y),

w_{5} = w (\begin{matrix} 1 & 3 & 5 & 6 & 7 & 9 \\ 2 & 4 & 8 \end{matrix})

= tr (x^{5} y^{2} x y) - tr (x^{5} y x y^{2}) + tr (x^{3} y x^{2} y x y) - tr (x^{3} y x y x^{2} y),

w_{6} = w (\begin{matrix} 1 & 2 & 4 & 6 & 8 & 9 \\ 3 & 5 & 7 \end{matrix}) = - tr (x^{5} y^{2} x y) + 2 tr (x^{4} y x^{2} y^{2})

+ tr (x^{4} y x y x y) - tr (x^{3} y^{2} x^{3} y) - tr (x^{3} y x y x^{2} y);

λ = (5, 4)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 7 & 9 \\ 2 & 4 & 6 & 8 \end{matrix}) = tr (x^{3} y^{2} x^{2} y^{2}) - tr (x^{3} y^{2} x y x y)

+ tr (x^{3} y x y^{2} x y) - tr (x^{3} y x y x y^{2}) - tr (x^{2} y^{2} x^{2} y x y) + tr (x^{2} y x y x y x y),

w_{2} = w (\begin{matrix} 1 & 2 & 5 & 7 & 9 \\ 3 & 4 & 6 & 8 \end{matrix}) = tr (x^{4} y^{3} x y) - tr (x^{4} y^{2} x y^{2})

- tr (x^{3} y^{3} x^{2} y) + tr (x^{3} y x y x y^{2}) + tr (x^{2} y^{2} x^{2} y x y) - tr (x^{2} y x y x y x y),

w_{3} = w (\begin{matrix} 1 & 2 & 4 & 7 & 9 \\ 3 & 5 & 6 & 8 \end{matrix}) = tr (x^{4} y^{3} x y) - tr (x^{4} y x y^{3})

- tr (x^{3} y^{3} x^{2} y) + tr (x^{3} y x^{2} y^{3}) - tr (x^{3} y^{2} x y x y) + tr (x^{3} y x y x y^{2}),

w_{4} = w (\begin{matrix} 1 & 2 & 3 & 4 & 9 \\ 5 & 6 & 7 & 8 \end{matrix}) = tr (x^{5} y^{4}) - tr (x^{4} y^{3} x y)

+ tr (x^{4} y^{2} x y^{2}) - tr (x^{4} y x y^{3}) - tr (x^{3} y x y^{2} x y) + tr (x^{2} y x y x y x y);

λ = (8, 2)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 6 & 7 & 8 & 9 & 10 \\ 2 & 4 \end{matrix}) = - tr (x^{8} y^{2}) + 2 tr (x^{7} y x y) - tr (x^{6} y x^{2} y),

w_{2} = w (\begin{matrix} 1 & 3 & 4 & 6 & 7 & 8 & 9 & 10 \\ 2 & 5 \end{matrix})

= - tr (x^{8} y^{2}) + tr (x^{7} y x y) + tr (x^{6} y x^{2} y) - tr (x^{5} y x^{3} y),

w_{3} = w (\begin{matrix} 1 & 3 & 4 & 5 & 7 & 8 & 9 & 10 \\ 2 & 6 \end{matrix})

= - tr (x^{8} y^{2}) + tr (x^{7} y x y) + tr (x^{5} y x^{3} y) - tr (x^{4} y x^{4} y),

w_{4} = w (\begin{matrix} 1 & 3 & 4 & 5 & 6 & 8 & 9 & 10 \\ 2 & 7 \end{matrix})

= - tr (x^{8} y^{2}) + tr (x^{7} y x y) - tr (x^{5} y x^{3} y) + tr (x^{4} y x^{4} y);

λ = (7, 3)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 7 & 8 & 9 & 10 \\ 2 & 4 & 6 \end{matrix}),

w_{2} = w (\begin{matrix} 1 & 3 & 5 & 6 & 8 & 9 & 10 \\ 2 & 4 & 7 \end{matrix}),

w_{3} = w (\begin{matrix} 1 & 3 & 5 & 6 & 7 & 9 & 10 \\ 2 & 4 & 8 \end{matrix}),

w_{4} = w (\begin{matrix} 1 & 3 & 5 & 6 & 7 & 8 & 10 \\ 2 & 4 & 9 \end{matrix}),

w_{5} = w (\begin{matrix} 1 & 2 & 5 & 6 & 7 & 9 & 10 \\ 3 & 4 & 8 \end{matrix}),

w_{6} = w (\begin{matrix} 1 & 3 & 4 & 7 & 8 & 9 & 10 \\ 2 & 5 & 6 \end{matrix}),

w_{7} = w (\begin{matrix} 1 & 2 & 3 & 7 & 8 & 9 & 10 \\ 4 & 5 & 6 \end{matrix});

λ = (6, 4)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 7 & 9 & 10 \\ 2 & 4 & 6 & 8 \end{matrix}),

w_{2} = w (\begin{matrix} 1 & 2 & 3 & 4 & 5 & 6 \\ 7 & 8 & 9 & 10 \end{matrix}),

w_{3} = w (\begin{matrix} 1 & 2 & 3 & 7 & 9 & 10 \\ 4 & 5 & 6 & 8 \end{matrix}),

w_{4} = w (\begin{matrix} 1 & 2 & 5 & 7 & 9 & 10 \\ 3 & 4 & 6 & 8 \end{matrix}),

w_{5} = w (\begin{matrix} 1 & 2 & 5 & 6 & 9 & 10 \\ 3 & 4 & 7 & 8 \end{matrix}),

w_{6} = w (\begin{matrix} 1 & 2 & 5 & 6 & 7 & 10 \\ 3 & 4 & 8 & 9 \end{matrix}),

w_{7} = w (\begin{matrix} 1 & 3 & 4 & 7 & 8 & 10 \\ 2 & 5 & 6 & 9 \end{matrix}),

w_{8} = w (\begin{matrix} 1 & 2 & 4 & 6 & 8 & 10 \\ 3 & 5 & 7 & 9 \end{matrix}),

w_{9} = w (\begin{matrix} 1 & 3 & 4 & 5 & 8 & 9 \\ 2 & 6 & 7 & 10 \end{matrix}),

w_{10} = w (\begin{matrix} 1 & 3 & 4 & 7 & 9 & 10 \\ 2 & 5 & 6 & 8 \end{matrix});

λ = (5, 5)

w_{1} = w (\begin{matrix} 1 & 3 & 5 & 7 & 9 \\ 2 & 4 & 6 & 8 & 10 \end{matrix}),

w_{2} = w (\begin{matrix} 1 & 3 & 5 & 7 & 8 \\ 2 & 4 & 6 & 9 & 10 \end{matrix}),

w_{3} = w (\begin{matrix} 1 & 3 & 5 & 6 & 7 \\ 2 & 4 & 8 & 9 & 10 \end{matrix}),

w_{4} = w (\begin{matrix} 1 & 2 & 3 & 4 & 9 \\ 5 & 6 & 7 & 8 & 10 \end{matrix}) .

Proof. For each $λ ⊢ n$ , the number of highest weight vectors $w_{i}$ given above, coincides with the multiplicity of $W (λ)$ in $U_{n}$ . The explicit form of $w_{i}$ is obtained by direct calculations. It is sufficient to prove that the $w_{i}$ s are linearly independent in $U_{n}$ . For most of the cases it is almost obvious. In the more complicated cases the linear independence is obtained by calculating the rank of the matrix with entries equal to the coefficients of some of $tr (x^{a_{1}} y^{b_{1}} \dots x^{a_{p}} y^{b_{p}})$ in the expression of $w_{i}$ . For example, for $λ = (6, 3)$ we compare the coefficients of $tr (x^{6} y^{3}), tr (x^{5} y^{2} x y), tr (x^{5} y x y^{2}), tr (x^{4} y^{2} x^{2} y), tr (x^{4} y x^{2} y^{2}), tr (x^{3} y x^{2} y x y)$ in the presentation of $w_{1}, \dots, w_{6}$ and obtain the matrix $(\begin{matrix} 0 & 0 & 1 & 0 & 1 & 0 \\ 1 & 1 & - 1 & - 1 & - 1 & 2 \\ - 1 & - 1 & - 1 & 0 & - 1 & - 2 \\ - 1 & 0 & 1 & 0 & 1 & 0 \\ 1 & 1 & - 1 & 2 & 0 & 0 \\ - 1 & 1 & 0 & 0 & 0 & - 1 \end{matrix}) .$ Its rank is equal to 6, and this gives that the trace polynomials $w_{1}, \dots, w_{6}$ are linearly independent in $U_{9}$ . □

3 The new generators

We fix the generic

4 \times 4

matrices

X = (x_{i j})

and

Y = (y_{i j})

. Changing the variables on the diagonals of

X

and

Y

, we can present them in the form

\begin{matrix} X = \frac{1}{4} tr (X) e + x, Y = \frac{1}{4} tr (Y) e + y, \end{matrix}

(5)

where

e

is the identity

4 \times 4

matrix and

x

y

are generic traceless matrices. As in the case of generic matrices without restrictions on the trace, we may assume that

x

is diagonal. We fix the notation

x = (\begin{matrix} x_{1} & 0 & 0 & 0 \\ 0 & x_{2} & 0 & 0 \\ 0 & 0 & x_{3} & 0 \\ 0 & 0 & 0 & - (x_{1} + x_{2} + x_{3}) \end{matrix}),

y = (\begin{matrix} y_{11} & y_{12} & y_{13} & y_{14} \\ y_{21} & y_{22} & y_{23} & y_{24} \\ y_{31} & y_{32} & y_{33} & y_{34} \\ y_{41} & y_{42} & y_{43} & - (y_{11} + y_{22} + y_{33}) \end{matrix}),

where

x_{1}, x_{2}, x_{3}

y_{i j}

are algebraically independent commuting variables. The algebra

C_{42}

is generated by

tr (X), tr (Y)

and the traces of products of

x

and

y

. We denote by

C_{0}

the algebra generated by

tr (z_{1} \dots z_{k})

z_{i} = x, y

, and

k \geq 2

. Clearly,

\begin{matrix} C_{42} \sim = K [tr (X), tr (Y)] \otimes_{K} C_{0}, \end{matrix}

(6)

and it is sufficient to find a minimal system of generators of

C_{0}

. Our idea is the following. We want to build inductively the

G L_{2}

-module

G

of the a minimal system of generators. We start with the

G L_{2}

-module

G_{1} = W (1)

generated by

tr (X)

, and further, we assume that the generators of higher degree are in

C_{0}

. Assume that we have already found the module

G_{2} \oplus \dots \oplus G_{n - 1}

of generators of degree

< n

C_{0}

Then we consider the symmetric algebra

K [G_{2} \oplus \dots \oplus G_{n - 1}]

and all trace identities for

M_{4} (K)

of the form

f (x, y) - g (x, y) = 0

, where

f (x, y) \in K [G_{2} \oplus \dots \oplus G_{n - 1}]

and

g (x, y) \in U_{n}

. The elements

g (x, y)

generate a

G L_{2}

-submodule

R_{n}

U_{n}

. Then the

G L_{2}

-submodule

G_{n}

U_{n}

is a complement of

R_{n}

U_{n}

. We may identify

G_{n}

with a

G L_{2}

-submodule of

C_{0}

. We add

G_{n}

to the module of generators and call it the

G L_{2}

-module of new generators of degree

n

. Then

G_{2} \oplus \dots \oplus G_{10}

is a

G L_{2}

-module of generators of the algebra of invariants of two traceless

4 \times 4

matrices and

G = G_{1} \oplus (G_{2} \oplus \dots \oplus G_{10})

is a minimal

G L_{2}

-module of generators of

C_{42}

. Let the homogeneous component of degree

n

K [G_{2} \oplus \dots \oplus G_{n - 1}]

K [G_{2} \oplus \dots \oplus G_{n - 1}]^{(n)} = \sum_{(λ_{1}, λ_{2}) ⊢ n} p (λ) W (λ) .

Similarly, we have the decomposition

U_{n} = \sum_{(λ_{1}, λ_{2}) ⊢ n} q (λ) W (λ)

from Proposition 2.1 and the decomposition

R_{n} = \sum_{(λ_{1}, λ_{2}) ⊢ n} r (λ) W (λ) \subset U_{n},

with unknown multiplicities

r (λ)

. If

v_{1}, \dots, v_{p (λ)} \in K [G_{2} \oplus \dots \oplus G_{n - 1}]^{(n)}

are linearly independent highest weight vectors of the direct sum of

p = p (λ)

isomorphic copies of

W (λ)

and

w_{1}, \dots, w_{q} \in U_{n}

q = q (λ)

, are linearly independent highest weight vectors in

q W (λ) \in U_{n}

, then the highest weight vectors of the

G L_{2}

-submodule

r W (λ) \subset R_{n}

r = r (λ)

, are all trace identities of the form

w (x, y) = (ξ_{1} v_{1} + \dots + ξ_{p} v_{p}) - (η_{1} w_{1} + \dots + η_{q} w_{q}) = 0,

where

ξ_{1}, \dots, ξ_{p}, η_{1}, \dots, η_{q}

are constants from the field

K

. We consider the elements

ξ_{1}, \dots, ξ_{p}

η_{1}, \dots, η_{q}

as unknowns. The entries of the evaluation of

w (x, y)

on the generic traceless matrices

x, y

are equal to 0. In this way we obtain a linear homogeneous system with respect to the unknowns

ξ_{1}, \dots, ξ_{p}

η_{1}, \dots, η_{q}

, and with coefficients which are polynomials in

x_{1}, x_{2}, x_{3}

and

y_{i j}

. The module

R_{n}

is generated by all

η_{1} w_{1} + \dots + η_{q} w_{q}

which correspond to solutions

ξ_{1}, \dots, ξ_{p}

η_{1}, \dots, η_{q}

The highest weight vectors

w_{1}, \dots, w_{q}

are given in the previous section. In order to find

v_{1}, \dots, v_{p}

, we need first to know the multiplicity

p (λ)

W (λ)

K [G_{2} \oplus \dots \oplus G_{n - 1}]^{(n)}

. If

G_{2} \oplus \dots \oplus G_{n - 1} = W_{1} \oplus \dots \oplus W_{m}

, where the

G L_{2}

-modules

W_{1}, \dots, W_{m}

are irreducible, then

K [G_{2} \oplus \dots \oplus G_{n - 1}] = K [W_{1}] \otimes \dots \otimes K [W_{m}] .

W_{i}

is homogeneous of degree

d_{i}

, and

W_{i}^{\otimes_{s} b_{i}}

denotes the

b_{i}

symmetric power of

W_{i}

, then

K [G_{2} \oplus \dots \oplus G_{n - 1}]^{(n)} = \oplus (W_{1}^{\otimes_{s} b_{1}} \otimes \dots \otimes W_{m}^{\otimes_{s} b_{m}}),

where the direct sum is on all

(b_{1}, \dots, b_{m})

with

b_{1} d_{1} + \dots + b_{m} d_{m} = n

. The decomposition of

W_{i}^{\otimes_{s} b_{i}}

can be obtained from the Hilbert series of

K [W_{i}]

and this of the usual tensor product by the Littlewood-Richardson rule. Form computational point of view it is convenient to calculate the homogeneous component of degree

n

of the Hilbert series of

K [G_{2} \oplus \dots \oplus G_{n - 1}]

. If

H (W_{i}, t, u) = a_{i 0} t^{d_{i}} + a_{i 1} t^{d_{i} - 1} u + \dots + a_{i, d_{i} - 1} t u^{d_{i} - 1} + a_{i d_{i}} u^{d_{i}}, a_{i j} \geq 0,

is the Hilbert series of

W_{i}

, then

H (K [G_{2} \oplus \dots \oplus G_{n - 1}], t, u) =^{m} {\prod_{i = 1}}^{d_{i}} \prod_{j = 0} \frac{1}{(1 - t^{j} u^{d_{i} - j})^{a_{i j}}} .

Then the homogeneous component of degree

n

of the Hilbert series can be obtained using ( 3 ). The highest weight vectors can be found using the constructive approach with the derivations. We start the realization of this scheme step by step.

Lemma 3.1. For

n \leq 5

, the

G L_{2}

-modules

G_{n}

of the new generators of

C_{0}

are

G_{2} = W (2, 0), G_{3} = W (3, 0), G_{4} = W (4, 0) \oplus W (2, 2), G_{5} = W (3, 2) .

Proof. By Proposition 2.1

U_{2} \oplus U_{3} \oplus U_{4} \oplus U_{5} = W (2) \oplus W (3) \oplus W (4) \oplus W (2, 2) \oplus W (5, 0) \oplus W (3, 2) .

Hence

G_{2} \oplus G_{3} \oplus G_{4} \oplus G_{5}

is a homomorphic image of this module. We shall show that

W (5, 0)

belongs to the subalgebra generated by

G_{2} \oplus G_{3} \oplus G_{4}

. Since

W (5, 0)

is generated by

tr (x^{5})

, it is sufficient to show that

tr (x^{5})

can be expressed in

C_{0}

in terms of

tr (x^{2})

tr (x^{3})

, and

tr (x^{4})

. By the Cayley-Hamilton theorem,

x

satisfies an equation of degree 4 with coefficients depending on

tr (x^{2})

tr (x^{3})

, and

tr (x^{4})

. Direct verification gives that

x^{4} - \frac{1}{2} tr (x^{2}) x^{2} - \frac{1}{3} tr (x^{3}) x + (- \frac{1}{4} {tr}^{2} (x^{2}) + \frac{1}{8} tr (x^{4})) e = 0 .

Multiplying by

x

, taking the trace, and using that

tr (x) = 0

, we obtain

\begin{matrix} tr (x^{5}) - \frac{5}{6} tr (x^{2}) tr (x^{3}) = 0 . \end{matrix}

(7)

(In the same way, by easy induction, we obtain that

tr (x^{n}) \in K [tr (x^{2}), tr (x^{3}), tr (x^{4})]

C_{0}

for any

n \geq 5

.) Hence, the

G L_{2}

-module of the new generators of

C_{0}

, of degree

\leq 5

, is a homomorphic image of the module

F = W (2) \oplus W (3) \oplus W (4) \oplus W (2, 2) \oplus W (3, 2) .

The Hilbert series of

F

H (F, t, u) = (t^{2} + t u + u^{2}) + (t^{3} + t^{2} u + t u^{2} + u^{3})

+ (t^{4} + t^{3} u + 2 t^{2} u^{2} + t u^{3} + u^{4}) + (t^{3} u^{2} + t^{2} u^{3})

and the Hilbert series of the symmetric algebra of

F

H (K [F], t, u) = \frac{1}{q_{2} (t, u) q_{3} (t, u) q_{4} (t, u) q_{5} (t, u)},

\begin{matrix} \begin{matrix} q_{2} (t, u) = (1 - t^{2}) (1 - t u) (1 - u^{2}), \\ q_{3} (t, u) = (1 - t^{3}) (1 - t^{2} u) (1 - t u^{2}) (1 - u^{3}), \\ q_{4} (t, u) = (1 - t^{4}) (1 - t^{3} u) (1 - t^{2} u^{2})^{2} (1 - t u^{3}) (1 - u^{4}), \\ q_{5} (t, u) = (1 - t^{3} u^{2}) (1 - t^{2} u^{3}) . \end{matrix} \end{matrix}

(8)

The algebra

K [F]

maps naturally on the subalgebra of

C_{0}

generated by

G_{2} \oplus G_{3} \oplus G_{4} \oplus G_{5}

, and the coefficients of the Hilbert series of

K [F]

are bigger or equal to the corresponding coefficients of the Hilbert series of this subalgebra. By direct calculation, we compare the coefficients of the homogeneous components of degree

\leq 5

of the Hilbert series of

H (K [F], t, u)

and

H (C_{0}, t, u)

. We find that they coincide and this implies that the homogeneous components of degree

\leq 5

K [F]

and

C_{0}

are isomorphic as

G L_{2}

-modules, completing the proof of the lemma. □

In the following lemmas the elements

w_{i}

are taken from Lemma 2.2. The relations have been found using the scheme described in the beginning of the section. The proofs are direct verifications, performed by Maple.

Lemma 3.2. For

λ = (4, 2)

, the following is a relation in

C_{0}

6 w_{1} - 12 w_{2} + 6 v_{1} + 2 v_{2} - 3 v_{3} - 5 v_{4} = 0,

where

v_{1} = tr (x^{3}) tr (x y^{2}) - {tr}^{2} (x^{2} y),

v_{2} = 3 tr (x^{4}) tr (y^{2}) - 6 tr (x^{3} y) tr (x y) + (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{2}),

v_{3} = tr (x^{2}) (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y)),

v_{4} = \frac{1}{2} tr ([x, y]^{2}) tr (x^{2}) .

Lemma 3.3. The following relations hold in

C_{0}

For

λ = (5, 2)

- 24 w_{1} + 18 v_{1} + 4 v_{2} + 16 v_{3} - 3 v_{4} - 4 v_{5} = 0,

24 w_{2} + 6 v_{1} - 4 v_{2} + 3 v_{4} + 4 v_{5} = 0,

where

v_{1} = tr ([x, y]^{2} x) tr (x^{2}),

v_{2} = 3 tr (x^{4}) tr (x y^{2}) - 6 tr (x^{3} y) tr (x^{2} y) + (tr (x y x y) + 2 tr (x^{2} y^{2})) tr (x^{3}),

v_{3} = \frac{1}{2} (tr ([x, y]^{2}) tr (x^{3}),

v_{4} = tr (x^{2}) (tr (x^{2}) tr (x y^{2}) - 2 tr (x y) tr (x^{2} y) + tr (y^{2}) tr (x^{3}),

v_{5} = (- tr (x^{2})) tr (y^{2}) + {tr}^{2} (x y)) tr (x^{3}) .

For

λ = (4, 3)

4 w_{1} - 8 w_{2} - 10 v_{1} - 4 v_{2} + 3 v_{3} = 0,

where

v_{1} = tr ([x, y]^{2} x) tr (x y) - tr ([x, y]^{2} y) tr (x^{2}),

v_{2} = tr (x^{4}) tr (y^{3}) - 3 tr (x^{3} y) tr (x y^{2})

+ (tr (x y x y) + 2 tr (x^{2} y^{2})) tr (x^{2} y) - tr (x y^{3}) tr (x^{3}),

v_{3} = {tr}^{2} (x^{2}) tr (y^{3}) - 3 tr (x^{2}) tr (x y) tr (x y^{2})

+ tr (x^{2}) tr (y^{2}) tr (x^{2} y) + 2 {tr}^{2} (x y) tr (x^{2} y) - tr (x y) tr (y^{2}) tr (x^{3}) .

Lemma 3.4. The following relations hold in

C_{0}

For

λ = (6, 2)

12 w_{1} - 6 v_{1} - 4 v_{2} - 6 v_{4} + 3 v_{7} = 0,

24 w_{2} - 6 v_{1} + 8 v_{2} - 6 v_{5} - 2 v_{6} + 5 v_{7} - 6 v_{8} + 6 v_{9} + 3 v_{10} = 0,

48 w_{3} + 18 v_{1} - 40 v_{2} - 16 v_{3} - 8 v_{4} + 18 v_{5} + 10 v_{6} - 13 v_{7} + 2 v_{8} - 6 v_{9} - 9 v_{10} = 0,

where

v_{1} = tr ([x, y]^{2} x^{2}) tr (x^{2}),

v_{2} = tr ([x, y]^{2} x) tr (x^{3}),

v_{3} = tr (x^{4}) (2 tr (x^{2} y^{2}) + tr (x y x y)) - 3 {tr}^{2} (x^{3} y),

v_{4} = \frac{1}{2} tr ([x, y]^{2}) tr (x^{4}),

v_{5} = tr (x^{4}) (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y),

v_{6} = 3 tr (x^{4}) {tr}^{2} (x y) - 6 tr (x^{3} y) tr (x^{2}) tr (x y) + (2 tr (x^{2} y^{2}) + tr (x y x y)) {tr}^{2} (x^{2}),

v_{7} = \frac{1}{2} tr ([x, y]^{2}) {tr}^{2} (x^{2}),

v_{8} = tr (x^{3}) (tr (x^{3}) tr (y^{2}) - 2 tr (x^{2} y) tr (x y) + tr (x y^{2}) tr (x^{2})),

v_{9} = {tr}^{2} (x^{3}) tr (y^{2}) - 2 tr (x^{3}) tr (x^{2} y) tr (x y) + {tr}^{2} (x^{2} y) tr (x^{2}),

v_{10} = (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y)) {tr}^{2} (x^{2}) .

For

λ = (5, 3)

16 w_{1} - 8 w_{2} - v_{1} - 4 v_{2} - 4 v_{3} + 2 v_{4} - v_{5} = 0,

2 w_{3} - v_{2} = 0,

where

v_{1} = - 2 tr ([x, y]^{2} x^{2}) tr (x y) + (tr ([x, y]^{2} x y) + tr ([x, y]^{2} y x)) tr (x^{2}),

v_{2} = (tr (x^{2} y^{2} x y) - tr (y^{2} x^{2} y x)) tr (x^{2}),

v_{3} = tr ([x, y]^{2} x) tr (x^{2} y) - tr ([x, y]^{2} y) tr (x^{3}),

v_{4} = tr (x^{4}) tr (x y) tr (y^{2}) - tr (x^{3} y) tr (x^{2}) tr (y^{2})

- 2 tr (x^{3} y) {tr}^{2} (x y) + (2 tr (x^{2} y^{2}) + tr (x y x y) tr (x^{2}) tr (x y) - tr (x y^{3}) {tr}^{2} (x^{2}),

v_{5} = - 2 tr (x^{3}) tr (x y^{2}) tr (x y) + 2 tr x^{2} y) tr (x^{2} y) tr (x y)

+ tr (x^{3}) tr (y^{3}) tr (x^{2}) - tr (x^{2} y) tr (x y^{2}) tr (x^{2}) .

For

λ = (4, 4)

- 24 w_{1} + 12 w_{2} - 12 v_{1} - 4 v_{2} + 16 v_{3} + v_{4} + 16 v_{5} = 0,

- 72 w_{1} + 24 w_{3} - 42 v_{1} - 20 v_{2} + 56 v_{3} + 3 v_{4} + 66 v_{5} + 6 v_{6} + 6 v_{7} = 0,

where

v_{1} = tr ([x, y]^{2} x^{2}) tr (y^{2}) - (tr ([x, y]^{2} x y) + tr ([x, y]^{2} y x)) tr (x y) + tr ([x, y]^{2} y^{2}) tr (x^{2}),

v_{2} = 3 tr (x^{4}) tr (y^{4}) - 12 tr (x^{3} y) tr (x y^{3}) + (2 tr (x^{2} y^{2}) + tr (x y x y))^{2},

v_{3} = \frac{1}{4} {tr}^{2} ([x, y]^{2}),

v_{4} = 3 tr (x^{4}) {tr}^{2} (y^{2}) - 12 tr (x^{3} y) tr (x y) tr (y^{2})

+ 2 (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{2}) tr (y^{2}) + 4 (2 tr (x^{2} y^{2}) + tr (x y x y)) {tr}^{2} (x y)

- 12 tr (x y^{3}) tr (x^{2}) tr (x y) + 3 tr (y^{4}) {tr}^{2} (x^{2}),

v_{5} = \frac{1}{2} tr ([x, y]^{2}) (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y)),

v_{6} = - tr (x^{3}) tr (x y^{2}) tr (y^{2}) + {tr}^{2} (x^{2} y) tr (y^{2}) + tr (x^{3}) tr (y^{3}) tr (x y)

- tr (x^{2} y) tr (x y^{2}) tr (x y) - tr (x^{2} y) tr (y^{3}) tr (x^{2}) + {tr}^{2} (x y^{2}) tr (x^{2}),

v_{7} = (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y))^{2} .

Lemma 3.5. The following relations hold in

C_{0}

between the elements

w_{i}

and the elements

v_{j}

given below:

For

λ = (7, 2)

48 w_{1} - 16 v_{1} - 12 v_{2} - 12 v_{3} - 4 v_{5} - 16 v_{7} - v_{9} + 4 v_{10} = 0,

144 w_{2} - 24 v_{1} + 36 v_{2} - 24 v_{4} - 20 v_{5} + 48 v_{6} + 20 v_{7} - 24 v_{8} + 9 v_{9} = 0,

48 w_{3} - 12 v_{2} - 4 v_{5} - 8 v_{7} - v_{9} + 4 v_{10} = 0,

v_{1} = tr ([x, y]^{2} x^{2}) tr (x^{3}),

v_{2} = tr ([x, y]^{2} x) tr (x^{4}),

v_{3} = tr ([x, y]^{2} x) {tr}^{2} (x^{2}),

v_{4} = tr (x^{4}) (tr (x^{3}) tr (y^{2}) + tr (x y^{2}) tr (x^{2}) - 2 tr (x^{2} y) tr (x y)),

v_{5} = 3 tr (x^{4}) tr (x y^{2}) tr (x^{2})

+ (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{3}) tr (x^{2}) - 6 tr (x^{3} y) tr (x^{2} y) tr (x^{2}),

v_{6} = tr (x^{4}) tr (x y^{2}) tr (x^{2}) - tr (x^{4}) tr (x^{2} y) tr (x y)

+ tr (x^{3} y) tr (x^{3}) tr (x y) - tr (x^{3} y) tr (x^{2} y) tr (x^{2}),

v_{7} = \frac{1}{2} tr ([x, y]^{2}) tr (x^{3}) tr (x^{2}),

v_{8} = tr (x^{3}) (tr (x^{3}) tr (x y^{2}) - {tr}^{2} (x^{2} y)),

v_{9} = (tr (x^{3}) tr (y^{2}) - 2 tr (x^{2} y) tr (x y) + tr (x y^{2}) tr (x^{2})) {tr}^{2} (x^{2}),

v_{10} = (tr (x^{3}) {tr}^{2} (x y) - 2 tr (x^{2} y) tr (x^{2}) tr (x y) + tr (x y^{2}) {tr}^{2} (x^{2})) tr (x^{2}) .

For

λ = (6, 3)

- 24 w_{1} + 48 w_{2} - 12 v_{1} + 3 v_{2} - 24 v_{3} + 36 v_{4} + 12 v_{5}

- 6 v_{6} - 2 v_{7} + 8 v_{8} + 7 v_{9} + 3 v_{10} - 15 v_{11} - 7 v_{12} = 0,

2 w_{3} + v_{1} = 0,

24 w_{4} - 12 v_{1} - 12 v_{2} - 12 v_{3} - 24 v_{4} + 6 v_{5} - 4 v_{7} + 4 v_{8} - 16 v_{9} - 6 v_{11} - v_{12} = 0,

w_{1} + w_{5} - v_{1} - v_{3} = 0,

24 w_{1} + 48 w_{6} + 3 v_{2} - 24 v_{3} - 12 v_{4} - 18 v_{5}

- 6 v_{6} - 10 v_{7} - 4 v_{8} - 9 v_{9} + 3 v_{10} + 3 v_{11} + 6 v_{12} = 0,

v_{1} = tr ([x, y]^{3} x) tr (x^{2}),

v_{2} = 2 tr ([x, y] x^{2}) tr (x^{2} y) - (tr ([x, y]^{2} x y) + tr ([x, y]^{2} y x)) tr (x^{3}),

v_{3} = (tr (x^{2} y^{2} x y) - tr (y^{2} x^{2} y x)) (tr (x^{3}),

v_{4} = tr ([x, y]^{2} x) tr (x^{3} y) - tr ([x, y]^{2} y) tr (x^{4}),

v_{5} = (tr ([x, y]^{2} x) tr (x y) - tr ([x, y]^{2} y) tr (x^{2})) tr (x^{2}),

v_{6} = (tr (x^{4}) tr (x^{2} y) - tr (x^{3} y) tr (x^{3})) tr (y^{2})

+ 2 (- tr (x^{4}) tr (x y^{2}) + tr (x^{3} y) tr (x^{2} y)) tr (x y)

+ (2 tr (x^{3} y) tr (x y^{2}) - (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{2} y) + tr (x y^{3}) tr (x^{3})) tr (x^{2}),

v_{7} = (3 tr (x^{4}) tr (x y^{2}) - 6 tr (x^{3} y) tr (x^{2} y) + (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{3})) tr (x y)

+ (- 3 tr (x^{3} y) tr (x y^{2}) + 2 (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{2} y) - 3 tr (x y^{3}) tr (x^{3})) tr (x^{2}),

v_{8} = (tr (x^{4}) tr (y^{3}) - 3 tr (x^{3} y) tr (x y^{2})

+ (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{2} y) - tr (x y^{3}) tr (x^{3})) tr (x^{2}),

v_{9} = \frac{1}{2} tr ([x, y]^{2}) (tr (x^{3}) tr (x y) - tr (x^{2} y) tr (x^{2})),

v_{10} = {tr}^{2} (x^{3}) tr (y^{3}) - 3 tr (x^{3}) tr (x^{2} y) tr (x y^{2}) + 2 {tr}^{3} (x^{2} y),

v_{11} = tr (x^{3}) tr (x^{2}) tr (x y) tr (y^{2}) - tr (x^{3}) {tr}^{3} (x y)

- tr (x^{2} y) {tr}^{2} (x^{2}) tr (y^{2}) + tr (x^{2} y) tr (x^{2}) {tr}^{2} (x y),

v_{12} = - 3 tr (x^{3}) tr (x^{2}) tr (x y) tr (y^{2}) + 2 tr (x^{3}) {tr}^{3} (x y)

+ 3 tr (x^{2} y) {tr}^{2} (x^{2}) tr (y^{2}) - 3 tr (x y^{2}) {tr}^{2} (x^{2}) tr (x y) + tr (y^{3}) {tr}^{3} (x^{2}) .

For

λ = (5, 4)

w_{1} - v_{3} = 0,

24 w_{2} + 12 v_{1} - 6 v_{2} + 12 v_{3} - 6 v_{4} - 2 v_{5} + 2 v_{6} + v_{7} = 0,

w_{3} + v_{1} = 0,

24 w_{4} + 6 v_{2} + 12 v_{4} - 4 v_{6} + v_{7} + 3 v_{8} = 0,

v_{1} = tr ([x, y]^{3} x) tr (x y) - tr ([x, y]^{3} y) tr (x^{2}),

v_{2} = tr ([x, y]^{2} x^{2}) tr (x y^{2})

- (tr ([x, y]^{2} x y) + tr ([x, y]^{2} y x)) tr (x^{2} y) + tr ([x, y]^{2} y^{2}) tr (x^{3}),

v_{3} = \frac{1}{2} tr ([x, y]^{2}) tr ([x, y]^{2} x),

v_{4} = tr ([x, y]^{2} x) (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y)),

v_{5} = (3 tr (x^{4}) tr (x y^{2}) - 6 tr (x^{3} y) tr (x^{2} y) + (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{3})) tr (y^{2})

+ (- tr (x^{4}) tr (y^{3}) - 3 tr (x^{3} y) tr (x y^{2})

+ 3 (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{2} y) - 5 tr (x y^{3}) tr (x^{3})) tr (x y)

+ (tr (x^{3} y) tr (y^{3}) - 3 tr (x y^{3}) tr (x^{2} y) + 2 tr (y^{4}) tr (x^{3})) tr (x^{2}),

v_{6} = (tr (x^{4}) tr (y^{3}) - 3 tr (x^{3} y) tr (x y^{2})

+ (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{2} y) - tr (x y^{3}) tr (x^{3})) tr (x y)

+ (- tr (x^{3} y) tr (y^{3}) + (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x y^{2})

- 3 tr (x y^{3}) tr (x^{2} y) + tr (y^{4}) tr (x^{3})) tr (x^{2}),

v_{7} = \frac{1}{2} tr ([x, y]^{2}) (tr (x^{3}) tr (y^{2}) - 2 tr (x^{2} y) tr (x y) + tr (x y^{2}) tr (x^{2})),

v_{8} = (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y)) (tr (x^{3}) tr (y^{2}) - 2 tr (x^{2} y) tr (x y) + tr (x y^{2}) tr (x^{2})) .

Lemma 3.6. For

λ = (8, 2)

the elements

w_{i}

and

v_{j}

satisfy the following relations in

C_{0}

360 w_{1} - 90 v_{1} - 45 v_{2} - 150 v_{3} - 90 v_{6} - 4 v_{7} - 36 v_{8}

- 80 v_{9} - 45 v_{12} + v_{13} - 10 v_{14} + 9 v_{15} = 0,

720 w_{2} + 90 v_{1} + 90 v_{2} - 420 v_{3} + 90 v_{4} - 90 v_{5} - 135 v_{6} + 90 v_{8}

- 80 v_{9} + 12 v_{10} + 9 v_{11} + 60 v_{12} - 2 v_{13} - 20 v_{14} - 18 v_{15} - 45 v_{16} = 0,

720 w_{3} - 90 v_{1} - 180 v_{2} + 90 v_{4} + 150 v_{5} - 195 v_{6} + 8 v_{7}

+ 162 v_{8} - 12 v_{10} - 99 v_{11} - 150 v_{12} + 40 v_{14} + 45 v_{16} = 0,

1440 w_{4} + 270 v_{2} - 960 v_{3} - 240 v_{5} - 120 v_{6} - 16 v_{7} - 144 v_{8}

- 320 v_{9} + 30 v_{10} + 180 v_{11} + 195 v_{12} - 80 v_{14} + 90 v_{15} - 135 v_{16} = 0,

where

v_{1} = tr ([x, y]^{2} x^{2}) tr (x^{4}),

v_{2} = tr ([x, y]^{2} x^{2}) {tr}^{2} (x^{2}),

v_{3} = tr ([x, y]^{2} x) tr (x^{3}) tr (x^{2}),

v_{4} = {tr}^{2} (x^{3} y) tr (x^{2}) - 2 tr (x^{4}) tr (x^{3} y) tr (x y) + {tr}^{2} (x^{4}) (y^{2}),

v_{5} = (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{4}) tr (x^{2}) - 3 {tr}^{2} (x^{3} y) tr (x^{2}),

v_{6} = \frac{1}{2} tr ([x, y]^{2}) tr (x^{4}) tr (x^{2}),

v_{7} = 5 (2 tr (x^{2} y^{2}) + tr (x y x y)) {tr}^{2} (x^{3}) - 30 tr (x^{3} y) tr (x^{2} y) tr (x^{3})

+ 3 tr (x^{4}) (2 tr (x y^{2}) tr (x^{3}) + 3 {tr}^{2} (x^{2} y)),

v_{8} = tr (x^{4}) (tr (x y^{2}) tr (x^{3}) - {tr}^{2} (x^{2} y)),

v_{9} = \frac{1}{2} tr ([x, y]^{2}) {tr}^{2} (x^{3}),

v_{10} = (5 (2 tr (x^{2} y^{2}) + tr (x y x y)) {tr}^{2} (x^{2}) - 30 tr (x^{3} y) tr (x y) tr (x^{2})

+ 3 tr (x^{4}) (4 {tr}^{2} (x y) + tr (y^{2}) tr (x^{2}))) tr (x^{2}),

v_{11} = tr (x^{4}) (tr (y^{2}) tr (x^{2}) - {tr}^{2} (x y)) tr (x^{2}),

v_{12} = (tr (x^{2} y^{2}) - tr (x y x y)) {tr}^{3} (x^{2}),

v_{13} = 5 {tr}^{2} (x^{3}) tr (x^{3}) (tr (y^{2}) tr (x^{2}) + 2 {tr}^{2} (x y)) - 30 tr (x^{3}) tr (x^{2} y) tr (x y) tr (x^{2})

+ 3 (2 tr (x^{3}) tr (x y^{2}) + 3 {tr}^{2} (x^{2} y)) {tr}^{2} (x^{2}),

v_{14} = {tr}^{2} (x^{3}) (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y)),

v_{15} = (tr (x^{3}) tr (x y^{2}) - {tr}^{2} (x^{2} y)) {tr}^{2} (x^{2}),

v_{16} = (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y)) {tr}^{3} (x^{2}) .

Lemma 3.7. In

C_{0}

and for

λ = (7, 3)

, the elements

w_{1}, \dots, w_{7}

and

v_{1} = tr ([x, y]^{3} x^{2}) tr (x^{2}),

v_{2} = tr ([x, y]^{3} x) tr (x^{3}),

v_{3} = (tr ([x, y]^{2} x y) + tr ([x, y]^{2} y x)) tr (x^{4}) - 2 tr ([x, y]^{2} x^{2}) tr (x^{3} y),

v_{4} = (tr (x^{2} y^{2} x y) - tr (y^{2} x^{2} y x)) tr (x^{4}),

v_{5} = ((tr ([x, y]^{2} x y) + tr ([x, y]^{2} y x)) tr (x^{2}) - 2 tr ([x, y]^{2} x^{2}) tr (x y)) tr (x^{2}),

v_{6} = (tr (x^{2} y^{2} x y) - tr y^{2} x^{2} y x)) {tr}^{2} (x^{2}),

v_{7} = (tr ([x, y]^{2} x) tr (x^{2} y) - tr ([x, y]^{2} y) tr (x^{3})) tr (x^{2}),

v_{8} = tr ([x, y]^{2} x) (tr (x^{3}) tr (x y) - tr (x^{2} y) tr (x^{2})),

v_{9} = (- 6 {tr}^{2} (x^{3} y) + 2 tr (x^{4}) (2 tr (x^{2} y^{2}) + tr (x y x y))) tr (x y)

+ (tr (x^{3} y) (2 tr (x^{2} y^{2}) + tr (x y x y)) - 3 tr (x^{4}) tr (x y^{3})) tr (x^{2}),

v_{10} = \frac{1}{2} tr ([x, y]^{2}) (tr (x^{4}) tr (x y) - tr (x^{3} y) tr (x^{2})),

v_{11} = tr (x^{4}) tr (x^{3}) tr (y^{3}) - 3 tr (x^{4}) tr (x^{2} y) tr (x y^{2}) + 6 tr (x^{3} y) {tr}^{2} (x^{2} y)

- 2 (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{3}) tr (x^{2} y) + 2 tr (x y^{3}) {tr}^{2} (x^{3}),

v_{12} = - tr (x^{4}) tr (x^{2} y) tr (x y^{2}) + tr (x^{3} y) tr (x^{3}) tr (x y^{2}) + 2 tr (x^{3} y) {tr}^{2} (x^{2} y)

- (2 tr (x^{2} y^{2}) + tr (x y x y)) tr (x^{3}) tr (x^{2} y) + tr (x y^{3}) {tr}^{2} (x^{3}),

v_{13} = - tr (x^{4}) tr (x^{2}) tr (x y) tr (y^{2}) + tr (x^{4}) {tr}^{3} (x y)

+ tr (x^{3} y) {tr}^{2} (x^{2}) tr (y^{2}) - tr (x^{3} y) tr (x^{2}) {tr}^{2} (x y),

v_{14} = tr (x^{4}) {tr}^{3} (x y - 3 tr (x^{3} y) tr (x^{2}) {tr}^{2} (x y)

+ (2 tr (x^{2} y^{2}) + tr (x y x y)) {tr}^{2} (x^{2}) tr (x y) - tr (x y^{3}) {tr}^{3} (x^{2}),

v_{15} = {tr}^{2} (x^{3}) tr (x y) tr (y^{2}) - tr (x^{3}) tr (x^{2} y) tr (x^{2}) tr (y^{2}) - 2 tr (x^{3}) tr (x^{2} y) {tr}^{2} (x y)

+ tr (x^{3}) tr (x y^{2}) tr (x^{2}) tr (x y) + 2 {tr}^{2} (x^{2} y) tr (x^{2}) tr (x y) - tr (x^{2} y) tr (x y^{2}) {tr}^{2} (x^{2}),

v_{16} = - 2 tr (x^{3}) tr (x y^{2}) tr (x^{2}) tr (x y) + 2 {tr}^{2} (x^{2} y) tr (x^{2}) tr (x y)

+ tr (x^{3}) tr (y^{3}) {tr}^{2} (x^{2} - tr (x^{2} y) tr (x y^{2}) {tr}^{2} (x^{2}),

satisfy the relations

24 w_{1} - 12 v_{1} - 8 v_{2} - 18 v_{4} + 9 v_{6} = 0,

48 w_{2} - 24 v_{2} + 12 v_{3} - 24 v_{4} - 3 v_{5} - 8 v_{7} + 4 v_{8} - 12 v_{10} + 4 v_{11} - 2 v_{15} - v_{16} = 0,

48 w_{3} - 24 v_{1} - 8 v_{2} + 12 v_{3} - 12 v_{4} - 3 v_{5} + 6 v_{6} - 8 v_{7}

+ 4 v_{8} - 12 v_{10} + 4 v_{11} - 2 v_{15} - v_{16} = 0,

12 w_{4} - 4 v_{2} - 3 v_{6} = 0,

48 w_{5} - 60 v_{1} + 8 v_{2} + 6 v_{3} - 24 v_{4} + 36 v_{6} + 10 v_{7}

+ 4 v_{8} - 4 v_{9} - 10 v_{10} + 10 v_{11} - 12 v_{12} + 6 v_{13} - 2 v_{15} - v_{16} = 0,

48 w_{6} - 48 v_{1} - 8 v_{2} - 36 v_{4} - 3 v_{5} + 30 v_{6} - 32 v_{7}

- 20 v_{8} - 8 v_{11} + 24 v_{12} - 6 v_{13} + 6 v_{14} - 6 v_{15} + 3 v_{16} = 0,

12 w_{7} - 6 v_{1} + 4 v_{2} + 3 v_{6} = 0 .

Lemma 3.8. The relation

\sum_{i = 1}^{10} α_{i} w_{i} + \sum_{j = 1}^{24} β_{j} v_{j} = 0,

holds in

C_{0}

between

w_{i}

and

v_{j}

for

λ = (6, 4)

, where

v_{1} = (tr ([x, y]^{3} x y) + tr ([x, y]^{3} y x)) tr (x^{2}) - 2 tr ([x, y]^{3} x^{2}) tr (x y),

v_{2} = tr ([x, y]^{4}) tr (x^{2}),

v_{3} = tr ([x, y]^{3} x) tr (x^{2} y) - tr ([x, y]^{3} y) tr (x^{3}),

v_{4} = tr ([x, y]^{2} x^{2}) (2 tr (x^{2} y^{2}) + tr (x y x y))

- 3 (tr ([x, y]^{2} x y) + tr ([x, y]^{2} y x)) tr (x^{3} y) + 3 tr ([x, y]^{2} y^{2}) tr (x^{4}),

v_{5} = \frac{1}{2} tr ([x, y]^{2} x^{2}) tr ([x, y]^{2}),

v_{6} = tr ([x, y]^{2} x^{2}) {tr}^{2} (x y)

- (tr ([x, y]^{2} y x) + tr ([x, y]^{2} x y)) tr (x^{2}) tr (x y) + tr ([x, y]^{2} y^{2}) {tr}^{2} (x^{2}),

v_{7} = tr ([x, y]^{2} x^{2}) (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y)),

v_{8} = {tr}^{2} t ([x, y]^{2} x),

v_{9} = tr ([x, y]^{2} x) (tr (x^{3}) tr (y^{2}) - 2 tr (x^{2} y) tr (x y) + tr (x y^{2}) tr (x^{2})),

v_{10} = tr ([x, y]^{2} x) (- tr (x^{3}) tr (y^{2}) + tr (x^{2} y) tr (x y))

+ tr ([x, y]^{2} y) (tr (x^{3}) tr (x y) - tr (x^{2} y) tr (x^{2})),

v_{11} = (tr (x^{4}) tr (x^{2} y^{2}) + tr (x^{4}) tr (x y x y) - 2 {tr}^{2} (x^{3} y)) tr (y^{2})

+ (- 2 tr (x^{4}) tr (x y^{3}) + 2 tr (x^{3} y) tr (x^{2} y^{2})) tr (x y)

+ (tr (x^{4}) tr (y^{4}) - 2 tr (x^{3} y) tr (x y^{3}) + tr (x^{2} y^{2}) tr (x y x y)) tr (x^{2}),

v_{12} = (tr (x^{4}) tr (x^{2} y^{2}) - {tr}^{2} (x^{3} y)) tr (y^{2})

+ (- tr (x^{4}) tr (x y^{3}) + tr (x^{3} y) tr (x y x y)) tr (x y)

+ (tr (x^{3} y) tr (x y^{3}) - tr (x^{2} y^{2}) tr (x y x y)) tr (x^{2}),

v_{13} = (tr (x^{4}) tr (y^{4}) - 4 tr (x^{3} y) tr (x y^{3}) + {tr}^{2} (x^{2} y^{2}) + 2 tr (x^{2} y^{2}) tr (x y x y)) tr (x^{2}),

v_{14} = (- tr (x^{4}) tr (y^{4}) + 4 tr (x^{3} y) tr (x y^{3}) - 4 tr (x^{2} y^{2}) tr (x y x y) + {tr}^{2} (x y x y)) tr (x^{2}),

v_{15} = tr (x^{4} tr (x^{2} y) tr (y^{3}) + tr (x^{3} y) (- tr (x^{3}) tr (y^{3}) - 3 tr (x^{2} y) tr (x y^{2}))

+ tr (x^{2} y^{2}) (3 tr (x^{3}) tr (x y^{2}) + {tr}^{2} (x^{2} y))

+ 2 tr (x y x y) {tr}^{2} (x^{2} y) - 4 tr (x y^{3}) tr (x^{3}) tr (x^{2} y) + tr (y^{4}) {tr}^{2} (x^{3}),

v_{16} = tr (x^{4}) {tr}^{2} (x y^{2}) - 4 tr (x^{3} y) tr (x^{2} y) tr (x y^{2}) + tr (x^{2} y^{2}) (2 tr (x^{3}) tr (x y^{2})

+ 2 {tr}^{2} (x^{2} y)) + 2 tr (x y x y) {tr}^{2} (x^{2} y) - 4 tr (x y^{3}) tr (x^{3}) tr (x^{2} y) + tr (y^{4}) {tr}^{2} (x^{3}),

v_{17} = tr (x^{2} y^{2}) (- tr (x^{3}) tr (x y^{2}) + {tr}^{2} (x^{2} y)) + tr (x y x y) (tr (x^{3}) tr (x y^{2}) - {tr}^{2} (x^{2} y)),

v_{18} = tr (x^{4}) (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y)) tr (y^{2})

+ tr (x^{3} y) (- 2 tr (x^{2}) tr (y^{2}) + 2 {tr}^{2} (x y)) tr (x y)

+ tr (x^{2} y^{2}) tr (x^{2}) (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y)),

v_{19} = tr (x^{4}) (- tr (x^{2}) tr (y^{2}) + {tr}^{2} (x y)) tr (y^{2})

+ tr (x^{3} y) (2 tr (x^{2}) tr (y^{2}) - 2 {tr}^{2} (x y)) tr (x y)

+ tr (x y x y) tr (x^{2}) (- tr (x^{2}) tr (y^{2}) + {tr}^{2} (x y)),

v_{20} = tr (x^{4}) tr (x^{2} {tr}^{2} (y^{2}) - 4 tr (x^{3} y) tr (x^{2}) tr (x y) tr (y^{2}) + 4 tr (x^{2} y^{2}) tr (x^{2}) {tr}^{2} (x y)

+ 2 tr (x y x y) {tr}^{2} (x^{2}) tr (y^{2}) - 4 tr (x y^{3}) {tr}^{2} (x^{2}) tr (x y) + tr (y^{4}) {tr}^{3} (x^{2}),

v_{21} = {tr}^{2} (x^{3}) {tr}^{2} (y^{2}) - 4 tr (x^{3}) tr (x^{2} y) tr (x y) tr (y^{2})

+ 2 {tr}^{2} (x^{2} y) tr (x^{2}) tr (y^{2}) + 4 tr (x^{3}) tr (x y^{2}) {tr}^{2} (x y) + (- 2 tr (x^{3}) tr (y^{3})

- 2 tr (x^{2} y) tr (x y^{2})) tr (x^{2}) tr (x y) + (2 tr (x^{2} y) tr (y^{3}) - {tr}^{2} (x y^{2})) {tr}^{2} (x^{2}),

v_{22} = ((tr (x^{3}) tr (x y^{2}) - {tr}^{2} (x^{2} y)) tr (y^{2})

+ (- tr (x^{3}) tr (y^{3}) + tr (x^{2} y) tr (x y^{2})) tr (x y)

+ (tr (x^{2} y) tr (y^{3}) - {tr}^{2} (x y^{2}))) tr (x^{2}),

v_{23} = (- tr (x^{3}) tr (x y^{2}) + {tr}^{2} (x^{2} y)) {tr}^{2} (x y)

+ (tr (x^{3}) tr (y^{3}) - tr (x^{2} y) tr (x y^{2})) tr (x^{2}) tr (x y) + (- tr (x^{2} y) tr (y^{3}) + {tr}^{2} (x y^{2})) {tr}^{2} (x^{2}),

v_{24} = (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y))^{2} tr (x^{2}),

and

β_{1} = \frac{1}{4} (α_{3} + α_{6} - α_{7} - α_{9}),

β_{2} = \frac{1}{24} (- 6 α_{1} - 3 α_{2} - 8 α_{3} - 8 α_{4} - 17 α_{5} + 2 α_{6} - 5 α_{7} - α_{8} - 2 α_{9} - 6 α_{10}),

β_{3} = \frac{1}{2} (α_{3} + α_{4} - α_{6} + α_{7} - 2 α_{9}),

β_{4} = \frac{1}{12} (- α_{2} + α_{3} + α_{7} - α_{9}),

β_{5} = \frac{1}{12} (- 12 α_{1} - 2 α_{2} + 5 α_{3} - 3 α_{4} + 3 α_{6} + 2 α_{7} + 4 α_{9} - 3 α_{10}),

β_{6} = \frac{1}{48} (- 3 α_{2} - 28 α_{3} - 16 α_{4} - 37 α_{5} + 4 α_{6} - 13 α_{7} - 5 α_{8} - 10 α_{9} - 12 α_{10}),

β_{7} = \frac{1}{48} (- 15 α_{2} - 40 α_{3} - 16 α_{4} - 37 α_{5} + 4 α_{6} - 13 α_{7} - 5 α_{8} - 22 α_{9} - 24 α_{10}),

β_{8} = \frac{1}{4} (- α_{2} + α_{3} + 3 α_{4} + 3 α_{5} - α_{6} - α_{7} + α_{8} - 3 α_{10}),

β_{9} = \frac{1}{12} (+ 3 α_{2} + 5 α_{3} + 2 α_{4} + 5 α_{5} + α_{6} + 2 α_{7} + α_{8} - α_{9}),

β_{10} = \frac{1}{12} (- 2 α_{3} + α_{4} - 2 α_{5} + 2 α_{6} - 2 α_{7} - 4 α_{8} + α_{9} - 6 α_{10}),

β_{11} = \frac{1}{8} (+ α_{2} - α_{3} - 2 α_{4} + α_{7} - α_{9} + 2 α_{10}),

β_{12} = \frac{1}{4} (- α_{2} + 7 α_{3} + 2 α_{4} - 2 α_{6} + α_{7} + 3 α_{9} + 2 α_{10}),

β_{13} = \frac{1}{24} (+ 12 α_{1} - 6 α_{2} + 13 α_{3} + 7 α_{4} - 14 α_{5} - 7 α_{6} - 2 α_{7} - 10 α_{8} + 10 α_{9} + 3 α_{10}),

β_{14} = \frac{1}{24} (+ 12 α_{1} + 6 α_{2} + 8 α_{3} + 11 α_{4} + 17 α_{5} - 5 α_{6} + 2 α_{7} + α_{8} + 5 α_{9} + 9 α_{10}),

β_{15} = \frac{1}{12} (- α_{2} - α_{3} - 2 α_{4} - 8 α_{5} + 2 α_{6} - α_{7} - 4 α_{8} + 3 α_{9}),

β_{16} = \frac{1}{4} (+ α_{4} + 3 α_{5} - α_{6} + α_{8} - α_{9} - α_{10}),

β_{17} = \frac{1}{12} (+ α_{2} - 2 α_{3} - α_{4} + 2 α_{5} + α_{6} + α_{7} - 2 α_{8} - 3 α_{9} - 3 α_{10}),

β_{18} = \frac{1}{48} (- 15 α_{2} - 64 α_{3} - 22 α_{4} - 43 α_{5} + 10 α_{6} - 19 α_{7} - 11 α_{8} - 28 α_{9} - 36 α_{10}),

β_{19} = \frac{1}{48} (- 9 α_{2} - 46 α_{3} - 22 α_{4} - 43 α_{5} + 10 α_{6} - 13 α_{7} - 11 α_{8} - 22 α_{9} - 24 α_{10}),

β_{20} = \frac{1}{96} (+ 9 α_{2} + 4 α_{3} + 10 α_{4} + 25 α_{5} + 2 α_{6} + α_{7} + 5 α_{8} + 4 α_{9}),

β_{21} = \frac{1}{48} (+ 3 α_{2} + 9 α_{3} - α_{4} + 3 α_{5} - α_{6} + 5 α_{7} + 5 α_{8} + 11 α_{10}),

β_{22} = \frac{1}{48} (- 3 α_{2} + 18 α_{3} + 4 α_{4} + 15 α_{5} - 14 α_{6} + 7 α_{7} + 7 α_{8} + 10 α_{10}),

β_{23} = \frac{1}{4} (+ 2 α_{3} + α_{5} - α_{6} + α_{7} + α_{8} + 2 α_{10}),

β_{24} = \frac{1}{16} (+ α_{2} + 2 α_{3} + α_{5} + α_{7} + α_{8} + 2 α_{10}) .

Lemma 3.9. For

λ = (5, 5)

C_{0}

w_{1} - 5 v_{2} = 0,

w_{2} - w_{3} + v_{1} - 4 v_{2} - v_{3} = 0,

w_{3} + w_{4} - \frac{5}{2} v_{1} - v_{2} + 2 v_{3} = 0,

where

v_{1} = tr ([x, y]^{3} x^{2}) tr (y^{2}) - (tr ([x, y]^{3} y x) + tr ([x, y]^{3} x y) tr (x y) + tr ([x, y]^{3} y^{2}) tr (x^{2}),

v_{2} = \frac{1}{2} (tr (x^{2} y^{2} x y) - tr (y^{2} x^{2} y x)) tr ([x, y]^{2}),

v_{3} = (tr (x^{2} y^{2} x y) - tr (y^{2} x^{2} y x)) (tr (x^{2}) tr (y^{2}) - {tr}^{2} (x y)),

v_{4} = (tr (x^{3}) tr (y^{2}) - 2 tr (x^{2} y) tr (x y) + tr (x y^{2}) tr (x^{2})) tr ([x, y]^{2} y)

- (tr (y^{3}) tr (x^{2}) - 2 tr (x y^{2}) tr (x y) + tr (x^{2} y) tr (y^{2})) tr ([x, y]^{2} x) .

The following theorem is the main result of the paper.

Theorem 3.10. The minimal generating

G L_{2}

-module

G

of the pure trace algebra

C_{42}

of two generic

4 \times 4

matrices decomposes as

\begin{matrix} G = W (1, 0) \oplus W (2, 0) \oplus W (3, 0) \oplus W (4, 0) \oplus W (2, 2) \oplus W (3, 2) \end{matrix}

\begin{matrix} \oplus W (4, 2) \oplus W (3, 3) \oplus W (4, 3) \oplus W (5, 3) \oplus W (4, 4) \oplus W (6, 3) \oplus W (5, 5) . \end{matrix}

\begin{matrix}  \end{matrix}

For all

λ = (λ_{1}, λ_{2}) \neq (5, 5)

, the module

W (λ_{1}, λ_{2})

is generated by the element

w_{λ} (X, Y) = tr ((X Y - Y X)^{λ_{2}} X^{λ_{1} - λ_{2}}) .

For

λ = (5, 5)

a generator of

W (5, 5)

w_{(5, 5)} (X, Y) = tr ((X Y - Y X)^{3} (X^{2} Y^{2} - X Y Y X - Y X X Y + Y^{2} X^{2})) .

Proof. Using ( 5 ), we may replace the elements

tr (Z_{1} \dots Z_{k})

Z_{i} = X, Y

k \geq 2

, modulo the subalgebra generated by products of traces of degree

< k

, with

tr (z_{1} \dots z_{k})

, where

z_{i} = x, y

are generic traceless matrices. Hence, it is sufficient to show that the algebra

C_{0}

has a minimal

G L_{2}

-module of generators

\begin{matrix} \begin{matrix} M = W (2, 0) \oplus W (3, 0) \oplus W (4, 0) \oplus W (2, 2) \oplus W (3, 2) \\ \oplus W (4, 2) \oplus W (3, 3) \oplus W (4, 3) \oplus W (5, 3) \oplus W (4, 4) \oplus W (6, 3) \oplus W (5, 5), \end{matrix} \end{matrix}

(9)

each

W (λ_{1}, λ_{2})

(λ_{1}, λ_{2}) \neq (5, 5)

, is generated by

w_{λ} (x, y) = tr ((x y - y x)^{λ_{2}} x^{λ_{1} - λ_{2}}),

and for

λ = (5, 5)

a generator of

W (5, 5)

\begin{matrix} w_{(5, 5)} (x, y) = tr ((x y - y x)^{3} (x^{2} y^{2} - x y y x - y x x y + y^{2} x^{2})) . \end{matrix}

(10)

There are three typical cases. We consider the image of the

G L_{2}

-modules

U_{n}

C_{0}

under the natural homomorphism

π : P (x, y) \to C_{0}

. We may assume that the generators of

C_{0}

are in

π (U_{2} \oplus \dots \oplus U_{10})

. Lemma 3.1 allows to consider the cases

n \geq 6

only.

Let

λ = (n)

n = 6, \dots, 10

. This case was handled in Lemma 3.1. The only

G L_{2}

-module

W (n)

π (U_{n})

is generated by

tr (x^{n})

. By ( 7 ),

tr (x^{5})

belongs to the subalgebra of

C_{0}

generated by traces of lower degree. Similar equations hold for the other

n \geq 6

. Hence,

W (n)

does not participate in any minimal

G L_{2}

-module of generators of

C_{0}

for

n = 6, \dots, 10

Let

λ = (6, 2)

. By Proposition 2.1,

W (6, 2)

participates with multiplicity 3 in

U_{8}

. The highest weight vectors of the three copies of

W (6, 2)

are

w_{1}, w_{2}, w_{3}

from Lemma 2.2. By the relations in Lemma 3.4,

π (w_{1}), π (w_{2}), π (w_{3})

are linear combinations of

v_{1}, \dots, v_{10}

, and hence belong to the subalgebra of

C_{0}

generated by the elements of degree

\leq 7

(and even of degree

\leq 6

). Therefore, as in the case

λ = (n)

, the minimal

G L_{2}

-modules of generators of

C_{0}

do not contain

W (6, 2)

The cases

λ = (5, 2), (7, 2), (5, 4), (8, 2), (7, 3), (6, 4)

are similar.

Let

λ = (5, 3)

. By Proposition 2.1 again,

W (5, 3)

participates with multiplicity 3 in

U_{8}

. The corresponding highest weight vectors are

w_{1}, w_{2}, w_{3}

from Lemma 2.2, and the element

w_{1}

is equal to

tr ((x y - y x)^{3} x^{2})

. By the relations in Lemma 3.4,

π (w_{2}), π (w_{3})

are linear combinations of

π (w_{1})

and

v_{1}, \dots, v_{5}

. Hence they belong to the subalgebra of

C_{0}

generated by the elements of degree

\leq 7

and by the

G L_{2}

-module generated by

w_{1}

. In this way, the multiplicity of

W (5, 3)

in the minimal

G L_{2}

-module of generators is bounded from above by 1. The cases

λ = (4, 2), (4, 3), (4, 4), (6, 3)

are similar. The case

λ = (5, 5)

is analogous but one has to express the elements

w_{1}, w_{3}, w_{4}

in terms of

w_{2}

and

v_{1}, v_{2}, v_{3}, v_{4}

. It is easy to see that

w_{2}

is equal to ( 10 ). Finally, the multiplicity of

W (3, 3)

U_{6}

is already 1. In this way we obtain that the minimal

G L_{2}

-module of generators of

C_{0}

is a homomorphism image of the module

M

from ( 9 ), and the algebra

C_{0}

is a homomorphic image of the symmetric algebra

K [M]

M

Clearly, the coefficients of the Hilbert series of

K [M]

are bigger or equal to the coefficients of the Hilbert series of

C_{0}

. The Hilbert series of

C_{42}

is given in ( 1 ). The factor

(1 - t) (1 - u)

in its denominator corresponds to the polynomial algebra

K [tr (X), tr (Y)]

in ( 6 ). Hence, the Hilbert series of

C_{0}

is equal to

h (t, u) = P_{C} (t, u) / Q_{C} (t, u)

from ( 2 ). It follows from ( 9 ) that the Hilbert series of

K [M]

H (K [M], t, u) = \frac{1}{q_{2} (t, u) \dots q_{10} (t, u)},

where

q_{2} (t, u), \dots, q_{5} (t, u)

are given in ( 8 ) and

q_{6} (t, u) = (1 - t^{4} u^{2}) (1 - t^{3} u^{3})^{2} (1 - t^{2} u^{4}),

q_{7} (t, u) = (1 - t^{4} u^{3}) (1 - t^{3} u^{4}),

q_{8} (t, u) = (1 - t^{5} u^{3}) (1 - t^{4} u^{4})^{2} (1 - t^{3} u^{5}),

q_{9} (t, u) = (1 - t^{6} u^{3}) (1 - t^{5} u^{4}) (1 - t^{4} u^{5}) (1 - t^{3} u^{6}), q_{10} (t, u) = (1 - t^{5} u^{5}) .

By direct calculations we see that the coefficients of the homogeneous components of degree

\leq 10

H (K [M], t, u)

are the same as the coefficients

h_{0}, h_{1}, \dots, h_{10}

from ( 4 ). This implies that the generators ( 9 ) satisfy no relations of degree

\leq 10

and competes the proof of the theorem. □

Remark 3.11. Since

tr ([X, Y]) = 0

, the equation ( 7 ) gives that the element

tr ([X, Y]^{5})

C_{42}

belongs to the subalgebra generated by traces of lower degree.

Remark 3.12. Calculating the first homogeneous components of the difference

H (C_{0}, t, u) - H (M, t, u)

, we see that

H (C_{0}, t, u) - H (M, t, u) = (S_{(7, 5)} (t, u) + 2 S_{(6, 6)} (t, u)) + (S_{(8, 5)} (t, u) + 2 S_{(7, 6)} (t, u))

+ (2 S_{(9, 5)} (t, u) + 6 S_{(8, 6)} (t, u) + 2 S_{(7, 7)} (t, u))

+ (2 S_{(10, 5)} (t, u) + 9 S_{(9, 6)} (t, u) + 7 S_{(8, 7)} (t, u)) .

Hence, the

G L_{2}

-modules of the defining relations of degree 12 and 13 are, respectively,

W (7, 5) \oplus 2 W (6, 6)

and

W (8, 5) \oplus 2 W (7, 6)

. In principle, the defining relations can be found as those in Lemmas 3.2 – 3.9, but the calculations should be more complicated.

Remark 3.13. The homogeneous system of parameters of

C_{42}

found by Teranishi [T1, T2] contains all traces of degree

\leq 4

and two elements of degree

(4, 2)

and

(2, 4)

, respectively. Translated in our language, we can choose for a homogeneous system of parameters of

C_{42}

the elements

tr (X), tr (Y), tr (X^{2}), tr (X Y), tr (Y^{2}),

tr (X^{3}), tr (X^{2} Y), tr (X Y^{2}), tr (Y^{3}),

tr (X^{4}), tr (X^{3} Y), tr (X^{2} Y^{2}), tr (X Y^{3}), tr (Y^{4}),

and the polynomials

w_{(4, 2)} (X, Y) = tr ((X Y - Y X)^{2} X^{2}), w_{(4, 2)} (Y, X) = tr ((X Y - Y X)^{2} Y^{2})

from Theorem 3.10.

References

S. Abeasis, M. Pittaluga, On a minimal set of generators for the invariants of $3 \times 3$ matrices, Commun. Algebra 17 (1989), 487-499.
G. Almkvist, W. Dicks, E. Formanek, Hilbert series of fixed free algebras and noncommutative classical invariant theory, J. Algebra 93 (1985), 189-214.
H. Aslaksen, V. Drensky, L. Sadikova, Defining relations of invariants of two $3 \times 3$ matrices, preprint.
A. Berele, J.R. Stembridge, Denominators for the Poincaré series of invariants of small matrices, Israel J. Math. 114 (1999), 157-175.
C. De Concini, D. Eisenbud, C. Procesi, Young diagrams and determinantal varieties, Invent. Math. 56 (1980), 129-165.
V. Drensky, Free Algebras and PI-Algebras, Springer-Verlag, Singapore, 1999.
V. Drensky, Defining relations for the algebra of invariants of $2 \times 2$ matrices, Algebras and Representation Theory 6 (2003), 193-214.
V. Drensky, E. Formanek, Polynomial Identity Rings, Advanced Courses in Mathematics, CRM Barcelona, Birkhäuser (to appear).
J. Dubnov, Sur une généralisation de l'équation de Hamilton-Cayley et sur les invariants simultanés de plusieurs affineurs, Proc. Seminar on Vector and Tensor Analysis, Mechanics Research Inst., Moscow State Univ. 2/3 (1935), 351-367 (see also Zbl. für Math. 12 (1935), p. 176).
E. Formanek, The Polynomial Identities and Invariants of $n \times n$ Matrices, CBMS Regional Conf. Series in Math. 78, Published for the Confer. Board of the Math. Sci. Washington DC, AMS, Providence RI, 1991.
P.E. Koshlukov, Polynomial identities for a family of simple Jordan algebras, Commun. Algebra 16 (1988), 1325-1371.
E.N. Kuzmin, On the Nagata-Higman theorem (Russian), in “Mathematical Structures, Computational Mathematics, Mathematical Modelling. Proc. Dedicated to the 60th Birthday of Acad. L. Iliev”, Sofia, 1975, 101-107.
K. Nakamoto, The structure of the invariant ring of two matrices of degree $3$ , J. Pure Appl. Algebra 166 (2002), No. 1-2, 125-148.
Yu.P. Razmyslov, Trace identities of full matrix algebras over a field of characteristic zero (Russian), Izv. Akad. Nauk SSSR, Ser. Mat. 38 (1974), 723-756. Translation: Math. USSR, Izv. 8 (1974), 727-760.
I. Shestakov, N. Zhukavets, On associative algebras satisfying the identity $x^{5} = 0$ , Algebra Discrete Math. (Lugansk) (2004), No. 1, 112-120.
K.S. Sibirskii, Algebraic invariants for a set of matrices (Russian), Sib. Mat. Zh. 9 (1968), No. 1, 152-164. Translation: Siber. Math. J. 9 (1968), 115-124.
Y. Teranishi, The ring of invariants of matrices, Nagoya Math. J. 104 (1986), 149-161.
Y. Teranishi, Linear Diophantine equations and invariant theory of matrices, Commutative algebra and combinatorics (Kyoto, 1985), 259-275, Adv. Stud. Pure Math., 11, North-Holland, Amsterdam, 1987.
M. Van den Bergh, Explicit rational forms for the Poincaré series of the trace rings of generic matrices, Isr. J. Math. 73 (1991), 17-31.
M. R. Vaughan-Lee, An algorithm for computing graded algebras, J. Symbolic Comput. 16 (1993), 345-354.

Institute of Mathematics and Informatics, Bulgarian Academy of Sciences, 1113 Sofia, Bulgaria E-mail address : drensky@math.bas.bg Fachgruppe Informatik, RWTH Aachen, 52056 Aachen, Germany E-mail address : sadikova@stce.rwth-aachen.de

Vesselin Drensky and Liliya Sadikova