Bound for the cocharacters of the identities of irreducible representations of $\mathfrak{sl}_2(\mathbb{C})$
Extracted source text ยท arXiv 2112.06870
%% September 14, 2021
\documentclass{amsart}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{question}[theorem]{Question}
\newtheorem{problem}[theorem]{Problem}
\newtheorem{conjecture}[theorem]{Conjecture}
\theoremstyle{definition}
\newtheorem{definition}[theorem]{Definition}
\newtheorem{example}[theorem]{Example}
\newtheorem{examples}[theorem]{Examples}
\newtheorem{remark}[theorem]{Remark}
\newtheorem{remarks}[theorem]{Remarks}
\newtheorem{algorithm}[theorem]{Algorithm}
\def\cocoa{{\hbox{\rm C\kern-.13em o\kern-.07em C\kern-.13em o\kern-.15em A}}}
%========================================================
\usepackage{comment}
\usepackage{amsmath, amssymb,url}
\usepackage{amscd,amssymb}
\usepackage{youngtab}
\newenvironment{proofof}[1]{\noindent{\it Proof of
#1.}}{\hfill$\square$\\\mbox{}}
%=====================================================
\begin{document}
\title[Cocharacters of representations of $\mathfrak{sl}_2(\mathbb{C})$]
{Bound for the cocharacters of the identities of irreducible representations of $\mathfrak{sl}_2(\mathbb{C})$}
\author[M\'aty\'as Domokos]
{M\'aty\'as Domokos}
\address{Alfr\'ed R\'enyi Institute of Mathematics,
Re\'altanoda utca 13-15, 1053 Budapest, Hungary}
\email{domokos.matyas@renyi.hu}
\thanks{Partially supported by the Hungarian National Research, Development and Innovation Office, NKFIH K 138828, K 132002.}
\subjclass[2010]{16R30; 16R10; 17B01; 17B20; 20C30.}
\keywords{weak polynomial identities, simple Lie algebra, irreducible representation, cocharacter sequence}
\maketitle
\newenvironment{dedication}
{\vspace{6ex}\begin{quotation}\begin{center}\begin{em}}
{\par\end{em}\end{center}\end{quotation}}
\begin{dedication}
{Dedicated to Vesselin Drensky on his $70$th birthday}
\end{dedication}
\begin{abstract}For each irreducible finite dimensional representation of the Lie algebra $\mathfrak{sl}_2(\mathbb{C})$ of $2\times 2$ traceless matrices, an explicit uniform upper bound is given for the multiplicities in the cocharacter sequence of the polynomial identities satisfied by the given representation.
\end{abstract}
\section{Introduction} \label{sec:intro}
Let $\rho:\mathfrak{g}\to \mathfrak{gl}(V)$ be a finite dimensional representation of the Lie algebra $\mathfrak{g}$ over a field $K$ of characteristic zero; that is, $\mathfrak{gl}(V)=\mathrm{End}_K(V)$, the space of all $K$-linear transformations of the finite dimensional $K$-vector space $V$, viewed as a Lie algebra with Lie product $[A,B]:=A\circ B-B\circ A$ for $A,B\in \mathrm{End}_K(V)$, and $\rho$ is a
homomorphism of Lie algebras. Denote by $F_m:=K\langle x_1,\dots,x_m\rangle$ the free associative $K$-algebra with $m$ generators. Consider $F_m$ a subalgebra of $F_{m+1}$ in the obvious way, and write
$F:=\bigcup_{m=1}^\infty F_m$ for the free associative algebra of countable rank.
We say that \emph{$f=0$ is an identity of the representation $\rho$ of $\mathfrak{g}$} (or briefly, of the pair $(\mathfrak{g},\rho)$) for some $f\in F_m$ if for any elements
$A_1,\dots,A_m\in \mathfrak{g}$ we have the following equality in the associative $K$-algebra $\mathrm{End}_K(V)$:
\[f(\rho(A_1),\dots,\rho(A_n))=0\in \mathrm{End}_K(V).\]
Note that an identity of the representation $\rho$ of the Lie algebra $\mathfrak{g}$ is also called in the literature a \emph{weak polynomial identity} for the pair $(\mathrm{End}_K(V),\rho(\mathfrak{g}))$.
This notion was introduced and powerfully applied first by Razmyslov \cite{Razmyslov:1, Razmyslov:2, Razmyslov:3, Razmyslov:4}
(see Drensky \cite{drensky} for a recent survey on weak polynomial identities).
Set
\[I(\mathfrak{g},\rho):=\{f\in F\mid f=0\text{ is an identity of }(\mathfrak{g},\rho)\}.\]
Clearly $I(\mathfrak{g},\rho)$ is an ideal in $F$ stable with respect to all $K$-algebra endomorphisms of $F$ of the form $x_i\mapsto u_i$, where $u_i$ for $i=1,2,\dots$ is an element of the Lie subalgebra of $F$ generated by $x_1,x_2,\dots$.
In particular, the general linear group $\mathrm{GL}_m(K)$ acts on $F_m$ via $K$-algebra automorphisms: for $g=(g_{ij})_{i,j=1}^m$ we have $g\cdot x_j=\sum_{i=1}^mg_{ij}x_i$, and $I(\mathfrak{g},\rho)\cap F_m$ is a $\mathrm{GL}_m(K)$-invariant subspace of $F_m$.
The \emph{multilinear component} of $F_m$ is
\[P_m:=\mathrm{Span}_K\{x_{\pi(1)}\cdots x_{\pi(m)}\mid \pi\in S_m\},\]
where $S_m$ is the symmetric group of degree $m$.
It is well known that when $\mathrm{char}(K)=0$, the ideal
$I(\mathfrak{g},\rho)$ is determined by the multilinear components $I(\mathfrak{g},\rho)\cap P_m$, $m=1,2,\dots$.
Identifying $S_m$ with the subgroup of permutation matrices in $\mathrm{GL}_m(K)$ we get its action on $F_m$ via $K$-algebra automorphisms (more explicitly, $\pi\in S_m$ is the automorphism of $F_m$ given by $x_i\mapsto x_{\pi(i)}$), and
the subspaces $P_m$ and $I(\mathfrak{g},\rho)\cap P_m$ are $S_m$-invariant.
Define the \emph{$m$th cocharacter of $(\mathfrak{g},\rho)$} as
\[\chi_m(\mathfrak{g},\rho):=\text{ the character of the }S_m\text{-module }P_m/(I(\mathfrak{g},\rho)\cap P_m).\]
We call
\[\chi(\mathfrak{g},\rho):=(\chi_m(\mathfrak{g},\rho)\mid m=1,2,\dots)\]
the \emph{cocharacter sequence of} $(\mathfrak{g},\rho)$.
The irreducible $S_m$-modules are labeled by partitions of $m$; let $\chi^{\lambda}$ denote the character of the irreducible
$S_m$-module associated to the partition $\lambda=(\lambda_1,\dots,\lambda_m)\vdash m$. We have
\[\chi_m(\mathfrak{g},\rho)=\sum_{\lambda\vdash m}\mathrm{mult}_{\lambda}(\mathfrak{g},\rho)\chi^{\lambda},\]
and we are interested in the multiplicities $\mathrm{mult}_{\lambda}(\mathfrak{g},\rho)$ of the irreducible $S_m$-characters in the
cocharacter sequence.
Note that the value of $\chi_m(\mathfrak{g},\rho)$ on the identity element of $S_m$ is
\[c_m(\mathfrak{g},\rho):=\dim_K(P_m/(I(\mathfrak{g},\rho)\cap P_m)),\]
and
\[(c_m(\mathfrak{g},\rho)\mid m=1,2,\dots)\]
is called the \emph{codimension sequence of}
$(\mathfrak{g},\rho)$. It was proved by Gordienko \cite{gordienko}
that $\lim_{m\to\infty}\root m\of{c_m(\mathfrak{g},\rho)}$ exists and is an integer.
As is observed in \cite[Example 3]{gordienko}, an obvious upper bound for
$c_m(\mathfrak{g},\rho)$ can be obtained from the fact that there is a natural $K$-linear embedding
\begin{equation}\label{eq:embedding}
P_m/(I(\mathfrak{g},\rho)\cap P_m)\hookrightarrow
\mathrm{Hom}_K(\rho(\mathfrak{g})^{\otimes m}, \mathrm{End}_K(V)).
\end{equation}
Our starting observation is that the adjoint representation of $\mathfrak{g}$ on itself induces a natural representation of $\mathfrak{g}$ on $\rho(\mathfrak{g})^{\otimes m}$
(the $m$th tensor power of $\rho(\mathfrak{g})$) and on $ \mathrm{End}_K(V)$, such that
the image of the embedding \eqref{eq:embedding} is contained in the subspace of $\mathfrak{g}$-module homomorphisms from $\rho(\mathfrak{g})^{\otimes m}$ to $\mathrm{End}_K(V)$. So
\eqref{eq:embedding} can be refined as
\begin{equation}\label{eq:2embedding}
P_m/(I(\mathfrak{g},\rho)\cap P_m)\hookrightarrow
\mathrm{Hom}_{\mathfrak{g}}(\rho(\mathfrak{g})^{\otimes m}, \mathrm{End}_K(V)).
\end{equation}
This will be used to give an upper bound for the multiplicities in the cocharacter sequence
$\chi(\frak{sl}_2(\mathbb{C}),\rho^{(d)})$ of the $d$-dimensional irreducible representation
\[\rho^{(d)}:\mathfrak{sl}_2(\mathbb{C})\to \mathfrak{gl}(\mathbb{C}^{d})=\mathbb{C}^{d\times d}\]
of $\mathfrak{sl}_2(\mathbb{C})$
for $d=1,2,\dots$. Note that throughout the paper we shall identify $\mathfrak{gl}(\mathbb{C}^d)$ with
the associative algebra $\mathbb{C}^{d\times d}$ of $d\times d$ complex matrices, viewed as a Lie algebra with Lie bracket $[A,B]=AB-BA$.
\begin{theorem}\label{thm:main}
The multiplicity $\mathrm{mult}_{\lambda}(\frak{sl}_2(\mathbb{C}),\rho^{(d)})$ in
$\chi(\frak{sl}_2(\mathbb{C}),\rho^{(d)})$ is non-zero only if $\lambda=(\lambda_1,\lambda_2,\lambda_3)$ (i.e. $\lambda$ has at most $3$ non-zero parts),
and in this case we have the inequality
\[\mathrm{mult}_{\lambda}(\frak{sl}_2(\mathbb{C}),\rho^{(d)})\le 3^{d-2}.\]
\end{theorem}
\begin{remark} \label{remark:1} (i) The exact values of $\mathrm{mult}_{\lambda}(\frak{sl}_2(\mathbb{C}),\rho^{(d)})$ are known for $d\le 3$. For $d=1$ all the multiplicities are obviously zero.
It was proved in \cite{Procesi:2} (see also \cite[Exercise 12.6.12]{Drensky:2}) that
\[\mathrm{mult}_{\lambda}(\frak{sl}_2(\mathbb{C}),\rho^{(2)})= 1 \text{ for all }\lambda=(\lambda_1,\lambda_2,\lambda_3). \]
The multiplicities $\mathrm{mult}_{\lambda}(\frak{sl}_2(\mathbb{C}),\rho^{(3)})$ are computed in
\cite[Theorem 3.7, Proposition 3.8]{DD}. It turns out that
$\mathrm{mult}_{\lambda}(\frak{sl}_2(\mathbb{C}),\rho^{(3)})\in \{1,2,3\}$ for each $\lambda=(\lambda_1,\lambda_2,\lambda_3)$.
(ii) Theorem~\ref{thm:main} shows in particular that for each dimension $d$, there is a uniform bound (depending on $d$ only) for the multiplicities $\mathrm{mult}_{\lambda}(\frak{sl}_2(\mathbb{C}),\rho^{(d)})$. For comparison we mention that the
multiplicities in the cocharacter sequence of the ordinary polynomial identities of $2\times 2$ matrices are unbounded:
see \cite{Drensky:84} and \cite{Formanek}. For example, for any partition $\lambda=(\lambda_1,\lambda_2)$ with $\lambda_2>0$, the multiplicity is $(\lambda_1-\lambda_2+1)\lambda_2$. On the other hand, the cocharacter multiplicities of any PI algebra are polynomially bounded by \cite{berele-regev}.
(iii) There is no uniform upper bound independent of $d$ for the multiplicities $\mathrm{mult}_{\lambda}(\frak{sl}_2(\mathbb{C}),\rho^{(d)})$, because by Proposition~\ref{prop:lowerbound}, $\max\{\mathrm{mult}_{\lambda}(\frak{sl}_2(\mathbb{C}),\rho^{(d)})\mid m=1,2,\dots,\ \lambda\vdash m\}\ge d-1$ for $d\ge 2$.
(iv) The irreducible representations of $\mathfrak{sl}_2(\mathbb{C})$ are defined over $\mathbb{Q}$.
For any field $K$ of characteristic zero and any positive integer $d$, the Lie algebra $\mathfrak{sl}_2(K)$ has a unique
(up to isomorphism) $d$-dimensional irreducible representation $\rho^{(d)}_K$ over $K$.
By well-known general arguments, the multiplicities $\mathrm{mult}_{\lambda}(\frak{sl}_2(K),\rho^{(d)}_K)$ do not depend on $K$.
Therefore Theorem~\ref{thm:main} implies that
$\mathrm{mult}_{\lambda}(\frak{sl}_2(K),\rho^{(d)}_K)\le 3^{d-2}$ for any field $K$ of characteristic zero.
(v) A different interpretation and approach to the study of
$\mathrm{Hom}_{\mathfrak{g}}(\rho(\mathfrak{g})^{\otimes m}, \mathrm{End}_K(V))$
for $\mathfrak{g}=\mathfrak{sl}_2(\mathbb{C})$ and $\rho=\rho^{(d)}$ is given in our parallel
preprint \cite{Domokos}, using classical invariant theory.
\end{remark}
We close the introduction by mentioning the recent paper of da Silva Macedo and Koshlukov \cite[Theorem 3.7]{dSM-K}, where the codimension growth of polynomial identities of representations of Lie algebras is studied.
In particular, in \cite[Theorem 3.7]{dSM-K} the identities of representations of $\frak{sl}_2(\mathbb{C})$ play a decisive role.
\section{Matrix computations}
Denote by $\widetilde\rho^{(d)}:\mathfrak{sl}_2(\mathbb{C})\to
\mathfrak{gl}(\mathbb{C}^{d\times d})$
the representation given by
\begin{equation}\label{eq:rhotilde}
\widetilde\rho^{(d)}(A)(L)= \rho^{(d)}(A) L-L \rho^{(d)}(A) \text{ for }A\in \mathfrak{sl}_2(\mathbb{C}), \
L\in \mathbb{C}^{d\times d}.
\end{equation}
We have $\widetilde\rho^{(d)}\cong \rho^{(d)}\otimes {\rho^{(d)}}^*$.
The representations of $\mathfrak{sl}_2(\mathbb{C})$ are self-dual, and so by the Clebsch-Gordan rules we have
\begin{equation}\label{eq:clebsch-gordan}
\widetilde\rho^{(d)}\cong \rho^{(d)}\otimes {\rho^{(d)}}\cong\bigoplus_{n=1}^d\rho^{(2n-1)}.
\end{equation}
We shall need an explicit decomposition of $\mathbb{C}^{d\times d}$ as a direct sum of minimal
$\widetilde\rho^{(d)}$-invariant subspaces.
Set
\[e:=\left(\begin{array}{cc}0 & 1 \\0 & 0\end{array}\right), \quad f:=\left(\begin{array}{cc}0 & 0 \\1 & 0\end{array}\right),
\quad h:=\left(\begin{array}{cc}1 & 0 \\0 & -1\end{array}\right),\]
so $e,f,h$ is a $\mathbb{C}$-vector space basis of $\mathfrak{sl}_2(\mathbb{C})$, with $[h,e]=2e$, $[h,f]=-2f$, and $[e,f]=h$.
Recall that given a representation $\psi:\mathfrak{sl}_2(\mathbb{C})\to \mathfrak{gl}(V)$, by a \emph{highest weight vector} we mean a non-zero element $w\in V$ such that $\psi(e)(w)=0\in V$ and $\psi(h)(w)=n w$ for some
non-negative integer $n$ (the non-negative integer $\lambda$ is called the \emph{weight} of $w$); in this case
$w$ generates a minimal $\mathfrak{sl}_2(\mathbb{C})$-invariant subspace in $V$, on which the representation of $\mathfrak{sl}_2(\mathbb{C})$ is isomorphic to $\rho^{(n+1)}$. Moreover, any finite dimensional irreducible $\mathfrak{sl}_2(\mathbb{C})$-module contains a unique (up to non-zero scalar multiples) highest weight vector.
\begin{lemma}\label{lemma:sl2decomposition of end}
Consider the $\mathfrak{sl}_2(\mathbb{C})$-module $\mathbb{C}^{d\times d}$ via the representation $\widetilde\rho^{(d)}$. To simplify notation set $\rho:=\rho^{(d)}$ and
$\widetilde\rho:=\widetilde\rho^{(d)}$.
\begin{itemize}
\item[(i)] $\rho(e)^n$ is a highest weight vector in $\mathbb{C}^{d\times d}$ of weight $2n$
for $n=0,1,\dots,d-1$.
\item[(ii)] $\rho(e)^{n-1}$ generates a minimal $\widetilde\rho$-invariant subspace $V_n$ on which $\mathfrak{sl}_2(\mathbb{C})$ acts via $\rho^{(2n-1)}$ for $n=1,\dots,d$.
\item[(iii)] $\mathbb{C}^{d\times d}=\bigoplus_{n=1}^dV_n$.
\item[(iv)] For $L_1\in V_{n_1}$ and $L_2\in V_{n_2}$ with $1\le n_1\neq n_2\le d$ we have
$\mathrm{Tr}(L_1 L_2)=0$.
\end{itemize}
\end{lemma}
\begin{proof} (i) We have $\widetilde\rho(e)(\rho(e)^n)=\rho(e) \rho(e)^n-\rho(e)^n \rho(e)=0$
and
\[\widetilde\rho(h)(\rho(e)^n)=\rho([h,e]) \rho(e)^{n-1}+\rho(e) \rho([h,e]) \rho(e)^{n-2}+\cdots+
\rho(e)^{n-1} \rho([h,e])=2n\rho(e)^n.\]
This shows that $\rho(e)^n$ is a highest weight vector of weight $2n$ for the representation $\widetilde\rho$.
(ii) Statement (i) implies that $\rho(e)^{n-1}$ generates an irreducible $\mathfrak{sl}_2(\mathbb{C})$-submodule of $\widetilde\rho$ isomorphic to $\rho^{(2n-1)}$ for $n=1,\dots,d$.
(iii) follows from (ii) and \eqref{eq:clebsch-gordan}.
(iv) Consider the symmetric non-degenerate bilinear form
\[\beta:\mathbb{C}^{d\times d}\times \mathbb{C}^{d\times d}\to \mathbb{C}, \quad
(L,M)\mapsto \mathrm{Tr}(LM).\]
Note that $\beta$ is $\widetilde\rho$-invariant:
\begin{align*}\beta([\rho(A),L],M)+\beta(L,[\rho(A),M])=\mathrm{Tr}([\rho(A),L]M)+
\mathrm{Tr}(L[\rho(A),M])\\
=\mathrm{Tr}([\rho(A),LM])=0 \quad \text{ for any }A\in \mathfrak{sl}_2(\mathbb{C}).
\end{align*}
The radical of the bilinear form $\beta_{V_n}:V_n\times V_n\mapsto \mathbb{C}$ (the restriction of $\beta$ to $V_n\times V_n$) is a $\widetilde\rho$-invariant subspace in $V_n$, so it is either $V_n$ or $\{0\}$. We claim that it is not $V_n$. Indeed, $V_n$ contains a non-zero diagonal matrix $D$ with real entries, since the zero weight subspace in $\mathbb{C}^{d\times d}$ (with respect to $\widetilde\rho(h)$) is the subspace of diagonal matrices, and $V_n$ intersects the zero-weight space in a $1$-dimensional subspace (defined over the reals). Now being a sum of squares of non-zero real numbers, $0\neq \mathrm{Tr}(D^2)=\beta(D,D)$.
Thus $\beta_{V_n}$ is non-degenerate.
The representation $\widetilde\rho$ is multiplicity free by \eqref{eq:clebsch-gordan}, and by (ii) and (iii),
every $\widetilde\rho$-invariant subspace is of the form $\sum_{j\in J}V_j$ for some subset $J\subseteq \{1,2,\dots,d\}$.
As we showed above, the orthogonal complement of $V_n$ (with respect to $\beta$) is disjoint from $V_n$, so it is the sum of the other minimal invariant subspaces $V_j$, $j\in \{1,\dots,d\}\setminus\{n\}$.
\end{proof}
The representation $\rho^{(2)}$ is the defining representation of $\mathfrak{sl}_2(\mathbb{C})$ on $\mathbb{C}^2$,
and $\rho^{(d)}$ is the $(d-1)$th symmetric tensor power of $\rho^{(2)}$. Denote by $x,y$ the standard basis vectors in $\mathbb{C}^2$, and take the basis $x^{d-1},x^{d-1}y,\dots,y^{d-1}$ in the $(d-1)$th symmetric tensor power of $\mathbb{C}^{2}$. Then
denoting by
$E_{i,j}$ the matrix unit with entry $1$ in the $(i,j)$ position and zeros in all other positions, the representation
$\rho^{(d)}$ as a matrix representation $\mathfrak{sl}_2(\mathbb{C})\to \mathbb{C}^{d\times d}$ is given as follows:
\[\rho^{(d)}(e)=\sum_{i=1}^{d-1}iE_{i,i+1},\quad \rho^{(d)}(f)=\sum_{i=1}^{d-1}(d-i)E_{i+1,i}, \quad
\rho^{(d)}(h)=\sum_{i=1}^d(d+1-2i)E_{i,i}\]
\begin{lemma} \label{lemma:products of length d}
For $d\ge 3$ the $\mathbb{C}$-vector space $\mathbb{C}^{d\times d}$ is spanned by
\[\{\rho^{(d)}(A_1) \cdots \rho^{(d)}(A_{d-1})\mid A_1,\dots,A_{d-1}\in \mathfrak{sl}_2(\mathbb{C})\}.\]
\end{lemma}
\begin{proof} To simplify the notation write $\rho:=\rho^{(d)}$ and $\widetilde\rho:=\widetilde\rho^{(d)}$.
Let $\mathcal{L}$ be the subspace of $\mathbb{C}^{d\times d}$ spanned by
the products $\rho(A_1) \cdots\rho(A_{d-1})$, where $A_1,\dots,A_{d-1}\in \mathfrak{sl}_2(\mathbb{C})$.
Clearly $\mathcal{L}$ is a $\widetilde\rho$-invariant subspace of $\mathbb{C}^{d\times d}$.
Since the representation $\widetilde\rho$ is multiplicity free by \eqref{eq:clebsch-gordan}, we have
$\mathcal{L}=\sum_{j\in J}V_j$ for some subset $J\subseteq \{1,2,\dots,d\}$
by Lemma~\ref{lemma:sl2decomposition of end} (ii) and (iii).
Therefore to prove the equality
$\mathcal{L}=\mathbb{C}^{d\times d}$ it is sufficient to show that $\mathcal{L}\cap V_n\neq\{0\}$ for each $n=1,\dots,d$,
or equivalently, that $\mathcal{L}$ is not contained in $\sum_{j\in \{1,\dots,d\}\setminus \{n\}}V_j$.
Since $V_d$ is generated by $\rho(e)^{d-1}\in \mathcal{L}$, we have $V_d\subseteq \mathcal{L}$.
Moreover, to prove
$\mathcal{L}\nsubseteq \sum_{j\in \{1,\dots,d\}\setminus \{n+1\}}V_j$ for $n\in\{0,1,\dots,d-2\}$,
it is sufficient to present an element $L_n\in \mathcal{L}$ with
$\mathrm{Tr}(\rho(e)^n L_n)\neq 0$ by Lemma~\ref{lemma:sl2decomposition of end} (i), (ii) and (iv).
We shall give below such elements $L_n\in \mathcal{L}$ for
$n=0,1,\dots,d-2$.
For $n=1,\dots,d-1$ we have
\[\rho(e)^n=\sum_{j=1}^{d-n}j\cdot (j+1)\cdots (j+n-1)E_{j,j+n}\]
\[\rho(f)^n=\sum_{j=1}^{d-n} (d-j) \cdot (d-j-1)\cdots (d-j-n+1)E_{j+n.j}\]
and $\rho(e)^0=I_d=\rho(f)^0$, where $I_d$ is the $d\times d$ identity matrix.
It follows that for $n=1,\dots,d-1$,
\[\rho(e)^n \rho(f)^n=\sum_{j=1}^{d-n}j(j+1)\cdots (j+n-1)\cdot (d-j)(d-j-1)\cdots(d-j-n+1)E_{j,j}\]
is a diagonal matrix with non-negative integer entries, and the $(1,1)$-entry is positive. The same holds for
$\rho(e)^0\rho(f)^0=I_d$.
For $n$ with $d-1-n$ even, $\rho(h)^{d-1-n}$ is the square of a diagonal
matrix with integer entries, and its $(1,1)$-entry is positive. Hence
$\mathrm{Tr}(\rho(e)^n \rho(f)^n\rho(h)^{d-1-n})\neq 0$, being a positive integer. So in this case we may take $L_n:=\rho(f)^n\rho(h)^{d-1-n}$.
For $n<d-2$ with $d-1-n$ odd, note that $\rho(e) \rho(f)-\rho(f)\rho(e)=\rho([e,f])=\rho(h)$,
and thus
\[\rho(f)^n \rho(h)^{d-2-n}=\rho(f)^n\rho(h)^{d-3-n}(\rho(e)\rho(f)-\rho(f)\rho(e))\]
also belongs to $\mathcal{L}$. Since $\rho(h)^{d-2-n}$ is a diagonal matrix with non-negative integer entries, and with a positive $(1,1)$-entry,
we may take $L_n:=\rho(f)^n \rho(h)^{d-2-n}$ in this case.
It remains to deal with the case $n=d-2$.
Then
\[\rho(e)^{d-2}\rho(f)^{d-2}=(d-1)((d-2)!)^2\cdot (E_{1,1}+E_{2,2}),\]
hence taking $L_{d-2}:=\rho(f)^{d-2} \rho(h)$ we get
\begin{align*}\mathrm{Tr}(\rho(e)^{d-2} L_{d-2})&=\mathrm{Tr}((d-1)((d-2)!)^2\cdot((d-1)E_{11}+(d-3)E_{22})
\\ &=(2d-4)(d-1)((d-2)!)^2,
\end{align*}
which is non-zero for $d\ge 3$.
This finishes the proof of the equality $\mathcal{L}=\mathbb{C}^{d\times d}$.
\end{proof}
\section{Adjoint invariants}
Denote by $\mathrm{ad}:\mathfrak{sl}_2(\mathbb{C})\to \mathfrak{gl}(\mathfrak{sl}_2(\mathbb{C}))$ the \emph{adjoint representation} of $\mathfrak{sl}_2(\mathbb{C})$ on itself, so $\mathrm{ad}(A)(B)=[A,B]$ for $A,B\in \mathfrak{sl}_2(\mathbb{C})$. Take the $n$-fold direct sum
$\mathrm{ad}^{\oplus n}:\mathfrak{sl}_2(\mathbb{C})\to \mathfrak{gl}(\mathfrak{sl}_2(\mathbb{C})^{\oplus n})$ of the adjoint representation, and write
$\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]^{\mathfrak{sl}_2(\mathbb{C})}$ for the algebra of $\mathrm{ad}^{\oplus n}$-invariant polynomial functions on $\mathfrak{sl}_2(\mathbb{C})^{\oplus n}$.
There is a right action of
$\mathrm{GL}_n(\mathbb{C})$ on $\mathfrak{sl}_2(\mathbb{C})^n$ that commutes with $\mathrm{ad}^{\oplus n}$: for $g=(g_{ij})_{i,1=1}^n$ and $(A_1,\dots,A_n)\in
\mathfrak{sl}_2(\mathbb{C})^n$ we have
\[(A_1,\dots,A_n)\cdot g:=(\sum_{i=1}^ng_{i1}A_i,\dots,\sum_{i=1}^ng_{in}A_i).\]
This induces a left $\mathrm{GL}_n(\mathbb{C})$-action on the coordinate ring $\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]$: for $g\in \mathrm{GL}_n(\mathbb{C})$,
$f\in \mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]$ and $(A_1,\dots,A_n)\in \mathfrak{sl}_2(\mathbb{C})^n$ we have
$(g\cdot f)(A_1,\dots,A_n)=f((A_1,\dots,A_n)\cdot g)$.
\begin{lemma} \label{lemma:iota}
Consider the linear map
$\iota:F_m=\mathbb{C}\langle x_1,\dots,x_m\rangle\to \mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^{m+d-1}]$
given by
\[\iota(f)(A_1,\dots,A_{m+d-1})=\mathrm{Tr}(f(\rho^{(d)}(A_1),\dots,\rho^{(d)}(A_m))\cdot \rho^{(d)}(A_{m+1}) \cdots \rho^{(d)}(A_{m+d-1}))\]
for $f\in F_m$ and
$(A_1,\dots,A_{m+d-1})\in \mathfrak{sl}_2(\mathbb{C})^{m+d-1}$. It has the following properties:
\begin{itemize}
\item[(i)] The image of $\iota$ is contained in the subalgebra
$\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^{m+d-1}]^{\mathfrak{sl}_2(\mathbb{C})}$ of $\mathfrak{sl}_2(\mathbb{C})$-invariants.
\item[(ii)] For $d\ge 3$ the kernel of $\iota$ is the ideal $I(\mathfrak{sl}_2(\mathbb{C}),\rho^{(d)})\cap F_m$.
\item[(iii)] The map $\iota$ is $\mathrm{GL}_m(\mathbb{C})$-equivariant, where we restrict the $\mathrm{GL}_{m+d-1}(\mathbb{C})$-action on $\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^{m+d-1}]$ to the subgroup
$\mathrm{GL}_m(\mathbb{C})\cong\{\left(\begin{array}{cc}g & 0 \\0 & I_{d-1}\end{array}\right)\mid g\in \mathrm{GL}_m(\mathbb{C})\}$ in
$\mathrm{GL}_{m+d-1}(\mathbb{C})$.
\end{itemize}
\end{lemma}
\begin{proof}
For notational simplicity we shall write $\rho$ instead of $\rho^{(d)}$.
(i) By linearity of $\iota$ it is sufficient to show that $\iota(x_{i_1}\cdots x_{i_k})$ is an $\mathfrak{sl}_2(\mathbb{C})$-invariant for any $i_1,\dots,i_k\in\{1,\dots,m\}$. Setting
$n=k+d-1$, $B_1=A_{i_1},\dots,B_k=A_{i_k},B_{k+1}=A_{m+1},\dots,B_n=A_{m+d-1}$ we have
\begin{equation}\label{eq:xi1...xik}
\iota(x_{i_1}\cdots x_{i_k})(A_1,\dots,A_{m+d})=\mathrm{Tr}(\rho(B_1)\cdots\rho(B_n)).
\end{equation}
For any $X\in \mathfrak{sl}_2(\mathbb{C})$ we have
\begin{align*}
0&=\mathrm{Tr}([\rho(X),\rho(B_1)\cdots\rho(B_n)])
\\ &=\mathrm{Tr}(\sum_{j=1}^n\rho(B_1)\cdots\rho(B_{j-1})[\rho(X),\rho(B_j)]\rho(B_{j+1})\cdots\rho(B_n))
\\ &=\mathrm{Tr}(\sum_{j=1}^n\rho(B_1)\cdots\rho(B_{j-1})\rho([X,B_j])\rho(B_{j+1})\cdots\rho(B_n))
\\ &=\sum_{j=1}^n\mathrm{Tr}(\rho(B_1)\cdots\rho(B_{j-1})\rho([X,B_j])\rho(B_{j+1})\cdots\rho(B_n)).
\end{align*}
The equalities \eqref{eq:xi1...xik} and
\[\sum_{j=1}^n\mathrm{Tr}(\rho(B_1)\cdots\rho(B_{j-1})\rho([X,B_j])\rho(B_{j+1})\cdots\rho(B_n))=0\]
mean that $\iota(x_{i_1}\cdots x_{i_k})$ is $\mathfrak{sl}_2(\mathbb{C})$-invariant, so (i) holds.
(ii) Suppose that $f\in \ker(\iota)$. Then $\mathrm{Tr}(f(\rho(A_1),\dots,\rho(A_m))B)=0$ for
all $B\in \mathbb{C}^{d\times d}$ by Lemma~\ref{lemma:products of length d}. By non-degeneracy of the trace we get
$f(\rho(A_1),\dots,\rho(A_m))=0$. That is, $f\in I(\mathfrak{sl}_2(\mathbb{C}),\rho)$. Thus
$\ker(\iota)\subseteq I(\mathfrak{sl}_2(\mathbb{C}),\rho)\cap F_m$. The reverse inclusion
$ I(\mathfrak{sl}_2(\mathbb{C}),\rho)\cap F_m\subseteq \ker(\iota)$ is obvious.
(iii) Take $g=(g_{ij})_{i,j=1}^m\in \mathrm{GL}_m(\mathbb{C})$.
For $f\in F_m$ and $(A_1,\dots,A_m)\in \mathfrak{sl}_2(\mathbb{C})^m$ we have
(by linearity of $\rho$)
\begin{align*}
&\iota(g\cdot f)(A_1,\dots,A_m)\\
&=\mathrm{Tr}(f(\sum_{i=1}^mg_{i1}\rho(A_i),\dots,\sum_{i=1}^mg_{im}\rho(A_i))\cdot \rho(A_{m+1}) \cdots \rho(A_{m+d}))
\\ &=\mathrm{Tr}(f( \rho(\sum_{i=1}^mg_{i1}(A_i)),\dots, \rho(\sum_{i=1}^mg_{im}(A_i)) \cdot \rho(A_{m+1})\cdots \rho(A_{m+d}))
\\ &=(g\cdot \iota(f))(A_1,\dots,A_m).
\end{align*}
This shows (iii).
\end{proof}
Restricting the action of $\mathrm{GL}_n(\mathbb{C})$ on $\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]$ to the subgroup of diagonal matrices we get an $\mathbb{N}_0^n$-grading on $\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]$, preserved by the action of $\mathfrak{sl}_2(\mathbb{C})$. Denote by $\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]_{(1^n)}$ the multihomogeneous component of multidegree $(1,\dots,1)$; this is the space of $n$-linear functions on $\mathfrak{sl}_2(\mathbb{C})$. The spaces $\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]_{(1^n)}$ and $\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]_{(1^n)}^{\mathfrak{sl}_2(\mathbb{C})}$ are $S_n$-invariant (where we restrict the $\mathrm{GL}_n(\mathbb{C})$-action to its subgroup $S_n$ of permutation matrices).
Lemma~\ref{lemma:iota} has the following immediate consequence:
\begin{corollary}\label{cor:res-iota}
For $d\ge 3$ the restriction of $\iota$ to the multilinear component $P_m$ of $\mathbb{C}\langle x_1,\dots,x_m\rangle$
factors through an $S_m$-equivariant $\mathbb{C}$-linear embedding
\[\bar\iota:P_m/(I(\mathfrak{sl}_2(\mathbb{C}))\cap P_m)\to
\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^{m+d-1}]_{(1^{m+d-1})}^{\mathfrak{sl}_2(\mathbb{C})}\]
where on the right hand side we consider the restriction of the $S_{m+d-1}$-action to its subgroup $S_m$
(the stabilizer in $S_{m+d-1}$ of the elements $m+1,m+2,\dots,m+d-1$).
\end{corollary}
For a partition $\lambda\vdash m$ denote by $r(\lambda)$ the multiplicity of $\chi^{\lambda}$ in the restriction to $S_m$ of the $S_{m+d-1}$-module $\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^{m+d-1}]_{(1^{m+d-1})}^{\mathfrak{sl}_2(\mathbb{C})}$.
Corollary~\ref{cor:res-iota} immediately implies the following:
\begin{corollary}\label{cor:m(lambda)<r(lambda)} For $d\ge 3$ and any partition $
\lambda\vdash m$ we have the inequality
\[\mathrm{mult}_{\lambda}(\mathfrak{sl}_2(\mathbb{C}),\rho^{(d)})\le r(\lambda).\]
\end{corollary}
The $S_n$-character of
$\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]_{(1^n)}^{\mathfrak{sl}_2(\mathbb{C})}$ is known:
\begin{proposition}\label{prop:Sn-character}
For a partition $\lambda\vdash n$ denote by $\nu(\lambda)$ the multiplicity of $\chi^{\lambda}$ in the $S_n$-character of
$\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]_{(1^n)}^{\mathfrak{sl}_2(\mathbb{C})}$. Then we have
\[\nu(\lambda)=\begin{cases}1\text{ for }\lambda=(\lambda_1,\lambda_2,\lambda_3) \text{ with } \lambda_1\equiv\lambda_2\equiv\lambda_3 \text{ modulo } 2
\\ 0 \text{ otherwise.} \end{cases} \]
\end{proposition}
\begin{proof}
The $\mathrm{GL}_n(\mathbb{C})$-module structure of $\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]^{\mathfrak{sl}_2(\mathbb{C})}$ is given for example in \cite[Theorem 2.2]{Procesi:2}. The isomorphism types of the irreducible $\mathrm{GL}_n(\mathbb{C})$-module direct summands of the degree $n$ homogeneous component of $\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]$ are labeled by partitions of $n$ with at most $3$ non-zero parts. The multiplicity $\mu(\lambda)$ of the irreducible $\mathrm{GL}_n(\mathbb{C})$-module $W_{\lambda}$ in the degree $n$ homogeneous component of $\mathcal{O}[\mathfrak{sl}_2(\mathbb{C})^n]^{\mathfrak{sl}_2(\mathbb{C})}$ is $1$ if $\lambda_1,\lambda_2,\lambda_3$ have the same parity and is zero otherwise. Note finally that the multilinear component of $W_{\lambda}$ is $S_n$-stable, and its $S_n$-character is $\chi^{\lambda}$
(see for example \cite[Corollary 6.3.11]{AGPR}).
\end{proof}
Following \cite[Section I.1]{Macdonald} for partitions $\lambda \vdash n$ and $\mu\vdash k$ we write
$\lambda\subset \mu$ is $\lambda_i\le\mu_i$ for all $i$. Moreover, given $\lambda\vdash m$ and $\mu\vdash m+d-1$
with $\lambda\subset \mu$, by a \emph{standard tableau of shape} $\mu/\lambda$ we mean a sequence
$\lambda^{(0)}\subset\lambda^{(1)}\subset\dots\subset\lambda^{(d-1)}$ of partitions $\lambda^{(i)}\vdash m+i$,
where $\lambda^{(0)}=\lambda$, $\lambda^{(d-1)}=\mu$.
By the well-known branching rules for the symmetric group, for $\lambda\vdash m$ the multiplicity of
$\chi^{\lambda}$ in the restriction to $S_m$ of the irreducible $S_{m+d-1}$-character $\chi^{\mu}$ equals the number of standard tableaux of shape $\mu/\lambda$ (see for example \cite[Theorem 6.4.11]{AGPR}).
Therefore Proposition~\ref{prop:Sn-character} has the following consequence.
\begin{corollary} \label{cor:Sm-character}
We have the equality
\begin{align*}
r(\lambda)=|\{ T\mid & T \text{ is a standard skew tableau of shape }\mu/\lambda,\\
&\mu\vdash m+d-1, \ \mu=(\mu_1,\mu_2,\mu_3),\ \mu_1\equiv\mu_2\equiv\mu_3\textrm{ modulo }2\}|.\end{align*}
\end{corollary}
\begin{corollary}\label{cor:3^d}
For $d\ge 3$ we have the inequality $r(\lambda)\le 3^{d-2}$.
\end{corollary}
\begin{proof}
Associate to a standard skew tableau $T=\lambda^{(0)}\subset\lambda^{(1)}\subset\dots\subset\lambda^{(d-1)}$ of shape $\mu/\lambda$, where $\mu=(\mu_1,\mu_2,\mu_3)\vdash m+d-1$ and $\mu_1\equiv\mu_2\equiv\mu_3\textrm{ modulo }2$
the function $f_T:\{1,\dots,d-1\}\to \{1,2,3\}$, which maps $j\in \{1,\dots,d-1\}$ to the unique $i\in\{1,2,3\}$ such that the $i$th component of the partition $\lambda^{(j)}$ is $1$ greater than the $i$th component of $\lambda^{(j-1)}$.
The assignment $T\mapsto f_T$ is obviously an injective map from the set of standard skew tableaux of shape $\mu/\lambda$ into the set of functions $\{1,\dots,d-1\}\to \{1,2,3\}$. We claim that at most $3^{d-2}$ functions are contained in the image of this map. Indeed, if the three parts of $\lambda^{(d-3)}$ have the same parity, then
$(f_T(d-2),f_T(d-1))\in \{(1,1),(2,2),(3,3)\}$, since the three parts of $\mu=\lambda^{(d-1)}$ must have the same parity. If the three parts of $\lambda^{(d-3)}$ do not have the same parity, say
the first two components of $\lambda^{(d-3)}$ have the same parity, and the third part has the opposite parity, then
$(f_T(d-2),f_T(d-1))\in \{(1,2),(2,1)\}$.
Hence $r(\lambda)$ is not greater than $3$-times the number of functions from a $(d-3)$-element set to a $3$-element set.
Thus $r(\lambda)\le 3^{d-2}$.
\end{proof}
\subsection{Proof of Theorem~\ref{thm:main}}
For $d\ge 3$ the statement follows from Corollary~\ref{cor:m(lambda)<r(lambda)} and Corollary~\ref{cor:3^d}.
For the cases $d\le 3$ see Remark~\ref{remark:1} (i).
\section{A lower bound}
\begin{proposition}\label{prop:lowerbound}
For $d\ge 2$ we have the equality
\[\mathrm{mult}_{(d-1,1)}(\frak{sl}_2(\mathbb{C}),\rho^{(d)})=d-1.\]
\end{proposition}
\begin{proof}
For $k=0,1,\dots,d-2$ consider the element
\[w_k:=x_1^k[x_1,x_2]x_1^{d-2-k}\in \mathbb{C}\langle x_1,x_2 \rangle=F_2.\]
These elements are $\mathrm{GL}_2(\mathbb{C})$-highest weight vectors with weight $(d-1,1)$, hence each generates an irreducible $\mathrm{GL}_2(\mathbb{C})$-submodule isomorphic to $W_{(d-1,1)}$ (see the proof of Proposition~\ref{prop:Sn-character} for the notation $W_{\lambda}$: it is the polynomial $\mathrm{GL}_2(\mathbb{C})$-module with highest weight $\lambda=(\lambda_1,\lambda_2)$). Moreover, they are linearly independent modulo the ideal
$I(\mathfrak{sl}_2(\mathbb{C}),\rho^{(d)})$: indeed, make the substitution $x_1\mapsto \rho(h)$, $x_2\mapsto \rho(e)$.
Then we get
\begin{align*}w_k(\rho(h),\rho(e))&=
(\sum_{i=1}^d(d+1-2i)E_{i,i})^k\cdot (2\sum_{i=1}^{d-1}iE_{i,i+1})\cdot (\sum_{i=1}^d(d+1-2i)E_{i,i})^{d-2-k}
\\ &=2\sum_{i=1}^{d-1}i(d+1-2i)^k(d-1-2i)^{d-2-k}E_{i,i+1}.
\end{align*}
Denote by $Z=(Z_{i,j})_{i,j=1}^{d-1}$ the $(d-1)\times (d-1)$ matrix whose $(i,k+1)$ entry is the $(i,i+1)$-entry of
$ w_k(\rho(h),\rho(e))$ (i.e. the coefficient of $E_{i.i+1}$ on the right hand side of the above equality).
If $i\neq \frac{d-1}{2}$, then
\[Z_{i,k+1}=2(d-1-2i)^{d-2} \cdot \left(\frac{d+1-2i}{d-1-2i}\right)^k.\]
Thus when $d$ is even, $Z$ is obtained from a Vandermonde matrix via multiplying each row by a non-zero integer.
Since the numbers $\frac{d+1-2i}{d-1-2i}$, $i=1,\dots,d-1$ are distinct, we conclude that $\det(Z)\neq 0$.
When $d=2f-1$ is odd, the $(f-1)$th row of $Z$ is
\[(0,\dots,0,2(f-1)2^{d-2}).\]
Expand the determinant of $Z$ along this row; the $(d-2)\times (d-2)$ minor of $Z$ obtained by removing the
$(f-1)th$ row and the last column of $Z$ is again obtained from a Vandermonde matrix by multiplying each row by a non-zero integer. So $\det(Z)$ is non-zero also when $d$ is odd.
This shows that the elements $w_k(\rho(h),\rho(e))$, $k=0,1,\dots,d-2$ are linearly independent in $\mathbb{C}^{d\times d}$. Consequently, no non-trivial linear combination of $w_0,w_1,\dots,w_{d-2}$ belongs to
$I(\mathfrak{sl}_2(\mathbb{C}),\rho^{(d)})$.
It follows that $F_2/(I(\mathfrak{sl}_2(\mathbb{C}),\rho^{(d)})\cap F_2)$ contains the
irreducible $\mathrm{GL}_2(\mathbb{C})$-module $W_{(d-1,1)}$ with multiplicity $\ge d-1$.
This multiplicity is in fact equal to $d-1$, because $d-1$ is the multiplicity of $W_{(d-1,1)}$ as a summand in
$F_2$. Recall finally that for $\lambda=(\lambda_1,\lambda_2)\vdash m$, the multiplicity of $\chi^{\lambda}$ in the cocharacter sequence coincides with the multiplicity of $W_{\lambda}$ in
$F_2/(I(\mathfrak{sl}_2(\mathbb{C}),\rho^{(d)})\cap F_2)$.
\end{proof}
%%%%%%%%%%%%%%%%%%
\begin{thebibliography}{AAAAA}
\bibitem{AGPR} E. Aljadeff, A. Giambruno, C. Procesi, A. Regev,
Rings with Polynomial Identities and Finite Dimensional Algebras, AMS Colloquium Publications, Vol 66,
Amer. Math. Soc., Providence, RI, 2020.
\bibitem{berele-regev} A. Berele, A. Regev, Applications of hook Young diagrams to P.I. algebras,
J. Algebra 82 (1983), 559-567.
\bibitem{dSM-K} D. L. da Silva Macedo, P. Koshlukov,
Codimension growth for weak polynomial identities and non-integrality of the PI exponent,
Proc. Edinb. Math. Soc. 63 (2020), 929-949.
\bibitem{Domokos} M. Domokos,
Matrix valued concomitants of $\mathfrak{sl}_2(\mathbb{C})$, preprint.
\bibitem{DD} M. Domokos, V. Drensky,
Cocharacters for the weak polynomial identities of the Lie algebra of $3\times 3$ skew-symmetric matrices,
Adv. Math. 374, Paper no: 107343 (2020).
\bibitem{Drensky:84} V. Drensky, Codimensions of T-ideals and Hilbert series of relatively free algebras,
J. Algebra 91 (1984), 1-17.
\bibitem{Drensky:2}
V. Drensky,
Free Algebras and PI-Algebras,
Springer-Verlag, Singapore, 2000.
\bibitem{drensky} V. Drensky, Weak polynomial identities and their applications,
Communications in Mathematics 29 (2021), 291-324.
\bibitem{Formanek} E. Formanek, Invariants and the ring of generic matrices, J. Algebra 89 (1984), 178-223.
\bibitem{gordienko} A. S. Gordienko, Codimensions of polynomial identities of representations of Lie algebras,
Proc. Amer. Math. Soc. 141 (2013), 3369-3382.
\bibitem{Macdonald}
I.G. Macdonald,
Symmetric Functions and Hall Polynomials,
Oxford Univ. Press (Clarendon), Oxford, 1979, Second Edition, 1995.
\bibitem{Procesi:2}
C. Procesi,
Computing with $2\times 2$ matrices,
J. Algebra 87 (1984), 342-359.
\bibitem{Razmyslov:1}
Yu.P. Razmyslov,
Finite basing of the identities of a matrix algebra of second order
over a field of characteristic zero (Russian),
Algebra i Logika 12 (1973), 83-113.
Translation: Algebra and Logic 12 (1973), 47-63.
\bibitem{Razmyslov:2}
Yu.P. Razmyslov, On a problem of Kaplansky (Russian),
Izv. Akad. Nauk SSSR, Ser. Mat. 37 (1973), 483-501.
Translation: Math. USSR, Izv. 7 (1973), 479-496.
\bibitem{Razmyslov:3}
Yu.P. Razmyslov,
Finite basis property for identities of representations
of a simple three-dimensional Lie algebra over a field of characteristic zero (Russian),
Algebra, Work Collect., dedic. O. Yu. Shmidt, Moskva 1982, 139-150.
Translation: Transl. Am. Math. Soc. Ser. 2 140 (1988), 101-109.
\bibitem{Razmyslov:4}
Yu.P. Razmyslov,
Identities of Algebras and Their Representations (Russian),
``Sovremennaya Algebra'', ``Nauka'', Moscow, 1989.
Translation: Translations of Math. Monographs 138, AMS, Providence, R.I., 1994.
\end{thebibliography}
\end{document}
\bibitem[ATZ]{Aslaksen-Tan-Zhu}
H. Aslaksen, E.-C. Tan, C.-B. Zhu,
Invariant theory of special orthogonal groups,
Pac. J. Math. {\bf 168} (1995), No. 2, 207-215.
\bibitem[B]{Berele}
A. Berele,
Homogeneous polynomial identities,
Israel J. Math. {\bf 42} (1982), 258-272.
\bibitem[CoCoA]{cocoa} CoCoATeam,
\cocoa: a system for doing
Computations in Commutative Algebra,
Available at http://cocoa.dima.unige.it
\bibitem[DoDr]{Domokos-Drensky}
M. Domokos, V. Drensky,
Gr\"obner bases for the rings of invariants of special orthogonal and $2 \times 2$ matrix invariants,
J. Algebra {\bf 243} (2001), 706-716.
\bibitem[Dr1]{Drensky:1}
V. Drensky,
Representations of the symmetric group and varieties of linear algebras (Russian),
Mat. Sb. {\bf 115} (1981), 98-115.
Translation: Math. USSR Sb. {\bf 43} (1981), 85-101.
\bibitem[Dr2]{Drensky:2}
V. Drensky,
Free Algebras and PI-Algebras,
Springer-Verlag, Singapore, 2000.
\bibitem[DrF]{Drensky-Formanek}
V. Drensky, E. Formanek,
Polynomial Identity Rings,
Advanced Courses in Mathematics, CRM Barcelona, Birkh\"auser, Basel-Boston, 2004.
\bibitem[DrK]{Drensky-Koshlukov}
V. Drensky, P.E. Koshlukov,
Weak polynomial identities for a vector space with a symmetric bilinear form,
Math. and Education in Math., Proc. of the 16-th Spring
Conf. of the Union of Bulgar. Mathematicians, Sunny Beach, April 6-10, 1987,
Publishing House of the Bulgarian Academy of Sciences, Sofia (1987), 213-219.
arXiv:1905.08351v1 [math.RA].
\bibitem[F]{Formanek}
E. Formanek,
Central polynomials for matrix rings,
J. Algebra {\bf 23} (1972), 129-132.
\bibitem[K1]{Kaplansky:1}
I. Kaplansky,
Problems in the theory of rings,
Report of a Conference on Linear Algebras, June, 1956,
in National Acad. of Sci.--National Research Council, Washington,
Publ. {\bf 502} (1957), 1-3.
\bibitem[K2]{Kaplansky:2}
I. Kaplansky,
Problems in the theory of rings revised,
Amer. Math. Monthly {\bf 77} (1970), 445-454.
\bibitem[LB]{Le Bruyn}
L. Le Bruyn,
Trace Rings of Generic 2 by 2 Matrices,
Mem. Amer. Math. Soc. {\bf 66} (1987), No. 363.
\bibitem[Mc]{Macdonald}
I.G. Macdonald,
Symmetric Functions and Hall Polynomials,
Oxford Univ. Press (Clarendon), Oxford, 1979, Second Edition, 1995.
\bibitem[P1]{Procesi:1}
C. Procesi,
The invariant theory of $n\times n$ matrices,
Adv. in Math. {\bf 198} (1976), 306-381.
\bibitem[P3]{Procesi:3}
C. Procesi, Lie Groups (An Approach through Invariants and Representations).
Springer, New York, 2007.
\bibitem[Re]{Regev}
A. Regev,
Algebras satisfying a Capelli identity,
Israel J. Math. {\bf 33} (1979), 149-154.
\bibitem[S]{sibirskii}
K.S. Sibirskii,
Unitary and orthogonal invariants of matrices (Russian),
Dokl. Akad. Nauk SSSR {\bf 172} (1967), 40-43.
\bibitem[W]{Weyl}
H. Weyl,
The Classical Groups, Their Invariants and Representations,
Princeton Univ. Press, Princeton, N.J., 1946, New Edition, 1997.