1991 Mathematics Subject Classification. 17B65,17B70.
<ph f="cmbx">The Lie algebra of cyclic coinvariants of a symplectic space</ph>

### Michael Larsen

Department of Mathematics, Northwestern University, 2033 Sheridan Road, Evanston, IL 60208 E-mail address : ekushnir@math.northwestern.edu Department of Mathematics, Indiana University, Bloomington, IN 47405 E-mail address : larsen@math.indiana.edu
• Abstract. We exhibit a natural Lie algebra structure on the graded space of cyclic coinvariants of a symplectic vector space.

1 Introduction

Let $V$  be a vector space endowed with a symplectic form. It is well known that the adjoint representation of $\text{Sp}\left(V\right)$  is equivalent to the symmetric square of $V$  , i.e., to the coinvariants of $\mathbb{Z}/2\mathbb{Z}$  acting by cyclic permutation on $V\otimes V$  . This allows us to endow $\left({V}^{\otimes 2}{\right)}_{\mathbb{Z}/2\mathbb{Z}}$  with the structure of a (symplectic) Lie algebra. In this note, we show that this is just the degree $0$  part of a larger object: the graded vector space ${}^{\infty }{\oplus }_{n=0}\left({V}^{\otimes n+2}{\right)}_{\mathbb{Z}/\left(n+2\right)\mathbb{Z}}$  has a natural structure of Lie algebra determined by the symplectic form on $V$  . In the special case that $dimV=2$  , for every associative algebra $A$  , this algebra acts infinitesimally on fibers of the commutator map $\left[,\right]:{A}^{2}\to A$  .
This paper developed from our interest in the fibers of the commutator map on the Lie algebra $\mathfrak{s}{\mathfrak{l}}_{n}$  . After we wrote it, B. Tsygan called our attention to a paper of M. Kontsevich [Kwhich arrived at similar conclusions, coming from a rather different (and more sophisticated) point of view. Kontsevich considered the tensor algebra of a symplectic space $V$  over $\mathbb{Q}$  and looked at the Lie algebra of derivations preserving its symplectic form. He asserted that this algebra is naturally isomorphic to ${}^{\infty }{\oplus }_{n=0}\left({V}^{\otimes n+2}{\right)}^{\mathbb{Z}/\left(n+2\right)\mathbb{Z}}.$  Of course in characteristic zero, there is a natural isomorphism between invariants and coinvariants for any finite group, which is not the case in characteristic $p$  . By contrast, we give a characteristic free formula for the Lie bracket which is naturally defined on the coinvariant space. In addition, we indicate a dihedral theory which may be useful in analyzing the commutator fibers for $\mathfrak{s}{\mathfrak{o}}_{n}$  and $\mathfrak{s}{\mathfrak{p}}_{2n}$  .

2 The Lie Algebra

Let $K$  be a field and $V$  a vector space over $K$  equipped with an alternating bilinear form $〈,〉$  . For any $l$  , $\mathbb{Z}/l\mathbb{Z}$  acts on ${V}^{\otimes l}$  in the obvious way and we form the graded $K$  -vector space $L\left(V\right){=}^{\infty }{\oplus }_{l=2}\left({V}^{\otimes l}{\right)}_{\mathbb{Z}/l\mathbb{Z}}.$  where $\left({V}^{\otimes l}{\right)}_{\mathbb{Z}/l\mathbb{Z}}$  has degree $l-2$  . We now define a Lie bracket on $L\left(V\right)$  . To do this, let us first introduce some additional notation. Fix a basis $\mathcal{ℬ}$  of $V$  . Given integers $i$  and $j$  such that $1\le i,j\le l$  , we set ${D}_{i,j}\left({\alpha }_{1}\otimes \cdots \otimes {\alpha }_{l}\right)=\left\{\begin{array}{cc}{\alpha }_{i+1}\otimes \cdots \otimes {\alpha }_{j-1}& \text{if}i+1\le j-1\text{,}\\ 0& \text{if}i=j-1\text{},\\ {\alpha }_{i+1}\otimes \cdots \otimes {\alpha }_{l}\otimes {\alpha }_{1}\otimes \cdots \otimes {\alpha }_{j-1}& \text{if}i\ge j\text{},\end{array}$  where ${\alpha }_{i}\in \mathcal{ℬ}$  . Now if $\alpha ={\alpha }_{1}\otimes \cdots \otimes {\alpha }_{l}$  and $\beta ={\beta }_{1}\otimes \cdots \otimes {\beta }_{m}$  are tensor products of basis elements, we define $\left[\alpha ,\beta \right]={\sum }_{i=1}^{l}{\sum }_{j=1}^{m}〈{\alpha }_{i},{\beta }_{j}〉\overline{{D}_{i,i}\left(\alpha \right)\otimes {D}_{j,j}\left(\beta \right)}.$  and extend by linearity. (Here $\overline{\gamma }$  denotes the class of $\gamma \in {V}^{\otimes n}$  in $\left({V}^{\otimes n}{\right)}_{\mathbb{Z}/n\mathbb{Z}}$  .) Note that if $\sigma$  denotes the canonical generator of $\mathbb{Z}/l\mathbb{Z}$  then ${D}_{i,i}\left({\sigma }^{r}\left(\alpha \right)\right)=\left\{\begin{array}{cc}{D}_{r+i,r+i}\left(\alpha \right)& \text{if}i\le l-r\text{}\\ {D}_{r+i-l,r+i-l}\left(\alpha \right)& \text{if}i>l-r\text{.}\end{array}$  Thus $\left[,\right]:\left({V}^{\otimes l}{\right)}_{\mathbb{Z}/l\mathbb{Z}}×\left({V}^{\otimes m}{\right)}_{\mathbb{Z}/m\mathbb{Z}}\mathbb{⟶}\left({V}^{\otimes \left(l+m-2\right)}{\right)}_{\mathbb{Z}/\left(l+m-2\right)\mathbb{Z}}$  is well-defined. Moreover, it does not depend on the choice of basis $\mathcal{ℬ}$  .
For each $x\in \mathcal{ℬ}$  , we also define ${D}_{x}:L\left(V\right)\to L\left(V\right)$  by ${D}_{x}\left(\alpha \right)={\sum }_{i=1}^{l}{\delta }_{{\alpha }_{i},x}\overline{{D}_{i,i}\left(\alpha \right)},$  where $\delta$  is the Kronecker delta and $\alpha$  is as before. We extend ${D}_{x}$  to a (well-defined) endomorphism of $L\left(V\right)$  .
Theorem 1. The above bracket makes $L\left(V\right)$  into a Lie algebra. Moreover each ${D}_{x}$  is a derivation.
• Proof. It is easily verified that the bracket is bilinear and antisymmetric. We now prove that the Jacobi identity holds. Suppose $\alpha ={\alpha }_{1}\otimes \cdots \otimes {\alpha }_{l}$  , $\beta ={\beta }_{1}\otimes \cdots \otimes {\beta }_{m}$  and $\gamma ={\gamma }_{1}\otimes \cdots \otimes {\gamma }_{n}$  ; then we have
$\left[\alpha ,\left[\beta ,\gamma \right]\right]+\left[\beta ,\left[\gamma ,\alpha \right]\right]+\left[\gamma ,\left[\alpha ,\beta \right]\right]={\sum }_{i=1}^{l}{\sum }_{j=1}^{m}{\sum }_{k=1}^{n}{\sum }_{r=1r\ne j}^{m}〈{\beta }_{j},{\gamma }_{k}〉〈{\alpha }_{i},{\beta }_{r}〉\overline{{D}_{i,i}\left(\alpha \right)\otimes {D}_{r,j}\left(\beta \right)\otimes {D}_{k,k}\left(\gamma \right)\otimes {D}_{j,r}\left(\beta \right)}+{\sum }_{i=1}^{l}{\sum }_{j=1}^{m}{\sum }_{k=1}^{n}{\sum }_{r=1r\ne k}^{n}〈{\beta }_{j},{\gamma }_{k}〉〈{\alpha }_{i},{\gamma }_{r}〉\overline{{D}_{i,i}\left(\alpha \right)\otimes {D}_{r,k}\left(\gamma \right)\otimes {D}_{j,j}\left(\beta \right)\otimes {D}_{k,r}\left(\gamma \right)}+{\sum }_{i=1}^{l}{\sum }_{j=1}^{m}{\sum }_{k=1}^{n}{\sum }_{r=1r\ne k}^{n}〈{\gamma }_{k},{\alpha }_{i}〉〈{\beta }_{j},{\gamma }_{r}〉\overline{{D}_{j,j}\left(\beta \right)\otimes {D}_{r,k}\left(\gamma \right)\otimes {D}_{i,i}\left(\alpha \right)\otimes {D}_{k,r}\left(\gamma \right)}+{\sum }_{i=1}^{l}{\sum }_{j=1}^{m}{\sum }_{k=1}^{n}{\sum }_{r=1r\ne i}^{l}〈{\gamma }_{k},{\alpha }_{i}〉〈{\beta }_{j},{\alpha }_{r}〉\overline{{D}_{j,j}\left(\beta \right)\otimes {D}_{r,i}\left(\alpha \right)\otimes {D}_{k,k}\left(\gamma \right)\otimes {D}_{i,r}\left(\alpha \right)}+{\sum }_{i=1}^{l}{\sum }_{j=1}^{m}{\sum }_{k=1}^{n}{\sum }_{r=1r\ne i}^{l}〈{\alpha }_{i},{\beta }_{j}〉〈{\gamma }_{k},{\alpha }_{r}〉\overline{{D}_{k,k}\left(\gamma \right)\otimes {D}_{r,i}\left(\alpha \right)\otimes {D}_{j,j}\left(\beta \right)\otimes {D}_{i,r}\left(\alpha \right)}+{\sum }_{i=1}^{l}{\sum }_{j=1}^{m}{\sum }_{k=1}^{n}{\sum }_{r=1r\ne j}^{m}〈{\alpha }_{i},{\beta }_{j}〉〈{\gamma }_{k},{\beta }_{r}〉\overline{{D}_{k,k}\left(\gamma \right)\otimes {D}_{r,j}\left(\beta \right)\otimes {D}_{i,i}\left(\alpha \right)\otimes {D}_{j,r}\left(\beta \right)}.$
Now note that $\overline{{D}_{i,i}\left(\alpha \right)\otimes {D}_{r,j}\left(\beta \right)\otimes {D}_{k,k}\left(\gamma \right)\otimes {D}_{j,r}\left(\beta \right)}=\overline{{D}_{k,k}\left(\gamma \right)\otimes {D}_{j,r}\left(\beta \right)\otimes {D}_{i,i}\left(\alpha \right)\otimes {D}_{r,j}\left(\beta \right)}$  in $\left({V}^{\otimes \left(l+m+n-4\right)}{\right)}_{\mathbb{Z}/\left(l+m+n-4\right)\mathbb{Z}}$  . Therefore, interchanging the sums over $j$  and $r$  in the last term, we see that it is exactly the negative of the first term (since $〈{\alpha }_{i},{\beta }_{j}〉〈{\gamma }_{k},{\beta }_{r}〉=-〈{\beta }_{r},{\gamma }_{k}〉〈{\alpha }_{i},{\beta }_{j}〉$  ). Similarly, the second term cancels the third and the fourth term cancels the fifth. The Jacobi identity now follows by linearity.
Finally, we have  $\begin{array}{ccc}{D}_{x}\left(\left[\alpha ,\beta \right]\right)& =& {\sum }_{i=1}^{l}{\sum }_{j=1}^{m}〈{\alpha }_{i},{\beta }_{j}〉{D}_{x}\left(\overline{{D}_{i,i}\left(\alpha \right)\otimes {D}_{j,j}\left(\beta \right)}\right)\end{array}$
 $\begin{array}{ccc}& =& {\sum }_{i=1}^{l}{\sum }_{j=1}^{m}〈{\alpha }_{i},{\beta }_{j}〉{\sum }_{k=1k\ne i}^{l}{\delta }_{{\alpha }_{k},x}\overline{{D}_{k,i}\left(\alpha \right)\otimes {D}_{j,j}\left(\beta \right)\otimes {D}_{i,k}\left(\alpha \right)}\end{array}$
 $\begin{array}{ccc}& +& {\sum }_{i=1}^{l}{\sum }_{j=1}^{m}〈{\alpha }_{i},{\beta }_{j}〉{\sum }_{k=1k\ne j}^{m}{\delta }_{{\beta }_{k},x}\overline{{D}_{k,j}\left(\beta \right)\otimes {D}_{i,i}\left(\alpha \right)\otimes {D}_{j,k}\left(\beta \right)}\end{array}$
 $\begin{array}{ccc}& =& \left[{D}_{x}\left(\alpha \right),\beta \right]+\left[\alpha ,{D}_{x}\left(\beta \right)\right]\end{array}$
If $\phi :V\to W$  is a linear transformation such that $〈\phi \left(\alpha \right),\phi \left(\beta \right){〉}_{W}=〈\alpha ,\beta {〉}_{V}$  for all $\alpha ,\beta \in V$  then we have an obvious Lie algebra homomorphism $L\left(\phi \right):L\left(V\right)\to L\left(W\right)$  induced by ${\alpha }_{1}\otimes \cdots \otimes {\alpha }_{l}↦\phi \left({\alpha }_{1}\right)\otimes \cdots \otimes \phi \left({\alpha }_{l}\right)$  .
The case $dimV=2$  is of particular interest. Let $\left\{x,y\right\}$  be a basis with $〈x,y〉=1$  .
We identify ${V}^{\otimes l}$  with the noncommutative homogeneous polynomials of degree $l$  in $x$  and $y$  .
Suppose that $A$  is an associative $K$  -algebra considered as a Lie algebra via the usual bracket $\left[X,Y\right]=XY-YX$  . Let ${N}_{l}$  denote the the trace with respect to the $\mathbb{Z}/l\mathbb{Z}$  -action on ${V}^{\otimes l}$  ; then the ${N}_{l}$  induce a well-defined map $N:L\left(V\right)\to {\oplus }_{l\ge 2}\left({V}^{\otimes l}{\right)}^{\mathbb{Z}/l\mathbb{Z}}$  . Let $K\left[\varepsilon \right]$  denote the ring of dual numbers. If $\alpha \in {V}^{\otimes l}$  represents a class in $L\left(V\right)$  , write $N\left(\alpha \right)=x{p}_{\alpha }-y{q}_{\alpha }$  and consider the vector field ${F}_{\alpha }\left(X,Y\right)=\left(X-\varepsilon {q}_{\alpha }\left(X,Y\right),Y-\varepsilon {p}_{\alpha }\left(X,Y\right)\right)$  on $A×A$  . Note that $N\left(\alpha \right)$  , ${p}_{\alpha }$  , ${q}_{\alpha }$  , and ${F}_{\alpha }$  depend only on $\overline{\alpha }$  .
Theorem 2. The map $\overline{\alpha }↦{F}_{\alpha }$  is a homomorphism of Lie algebras from $L\left(V\right)$  to the algebra of vector fields on $A×A$  tangent to the fibers of the commutator map.
• Proof. As $\sigma \left({N}_{l}\left(\alpha \right)\right)={N}_{l}\left(\alpha \right)$  , $\left(x{p}_{\alpha }-y{q}_{\alpha }\right)=\left({p}_{\alpha }x-{q}_{\alpha }y\right)$  , so $\left[X-\varepsilon {q}_{\alpha }\left(X,Y\right),Y-\varepsilon {p}_{\alpha }\left(X,Y\right)\right]\equiv \left[X,Y\right]\left(mod\varepsilon \right).$  Thus, ${F}_{\alpha }$  is always tangent to the fibers of the commutator map. To see that $\overline{\alpha }↦{F}_{\alpha }$  is a Lie algebra homomorphism, it suffices to check the case that $\alpha ={\alpha }_{1}\otimes \cdots \otimes {\alpha }_{l}$  and $\beta ={\beta }_{1}\otimes \cdots \otimes {\beta }_{m}$  are tensor products of basis vectors. Now, ${p}_{\alpha }={\sum }_{i=1}^{l}{\delta }_{{\alpha }_{i},x}{D}_{i,i}\left(\alpha \right),{q}_{\alpha }=-{\sum }_{i=1}^{l}{\delta }_{{\alpha }_{i},y}{D}_{i,i}\left(\alpha \right).$  (Note that $\overline{{p}_{\alpha }}={D}_{x}\left(\alpha \right)$  and $\overline{{q}_{\alpha }}={D}_{y}\left(\alpha \right)$  .) Regarding ${F}_{\alpha }$  and ${F}_{\beta }$  as sections $A×A\to A\left[\left[\varepsilon \right]\right]×A\left[\left[\varepsilon \right]\right]$  of the evaluation at zero map,
 $\begin{array}{cc}{F}_{\alpha }\left(& {F}_{\beta }\left(X,Y\right)\right)={F}_{\alpha }\left(X+\varepsilon {\sum }_{j=1}^{m}{\delta }_{{\beta }_{j},y}{D}_{j,j}\left(\beta \right)\left(X,Y\right),Y-\varepsilon {\sum }_{j=1}^{m}{\delta }_{{\beta }_{j},x}{D}_{j,j}\left(\beta \right)\left(X,Y\right)\right)=\end{array}$
 $\begin{array}{cc}\left(X& +\varepsilon {\sum }_{j=1}^{m}{\delta }_{{\beta }_{j},y}{D}_{j,j}\left(\beta \right)\left(X,Y\right)\end{array}$
 $\begin{array}{cc}+\varepsilon & {\sum }_{i=1}^{l}{\delta }_{{\alpha }_{i},y}{D}_{i,i}\left(\alpha \right)\left(X+\varepsilon {\sum }_{j=1}^{m}{\delta }_{{\beta }_{j},y}{D}_{j,j}\left(\beta \right)\left(X,Y\right),Y-\varepsilon {\sum }_{j=1}^{m}{\delta }_{{\beta }_{j},x}{D}_{j,j}\left(\beta \right)\left(X,Y\right)\right),\end{array}$
 $\begin{array}{cc}Y& -\varepsilon {\sum }_{j=1}^{m}{\delta }_{{\beta }_{j},x}{D}_{j,j}\left(\beta \right)\left(X,Y\right)\end{array}$
 $\begin{array}{cc}-\varepsilon & {\sum }_{i=1}^{l}{\delta }_{{\alpha }_{i},x}{D}_{i,i}\left(\alpha \right)\left(X+\varepsilon {\sum }_{j=1}^{m}{\delta }_{{\beta }_{j},y}{D}_{j,j}\left(\beta \right)\left(X,Y\right),Y-\varepsilon {\sum }_{j=1}^{m}{\delta }_{{\beta }_{j},x}{D}_{j,j}\left(\beta \right)\left(X,Y\right)\right)\right).\end{array}$
 $\begin{array}{cc}& \end{array}$
 $\begin{array}{}\end{array}$
If $\gamma ={\gamma }_{1}\otimes \cdots \otimes {\gamma }_{n}$  is a tensor monomial regarded as a noncommutative homogeneous polynomial, then
$\gamma \left(X+\varepsilon {X}_{1},Y+\varepsilon {Y}_{1}\right)\equiv \gamma \left(X,Y\right)+\varepsilon {\sum }_{k=1}^{n}{D}_{n,k}\left(\gamma \right)\left(X,Y\right)\left({\delta }_{{\gamma }_{k},x}{X}_{1}+{\delta }_{{\gamma }_{k},y}{Y}_{1}\right){D}_{k,1}\left(\gamma \right)\left(X,Y\right)\left(mod{\varepsilon }^{2}\right).$
Therefore,
 $\begin{array}{cc}{F}_{\alpha }\left({F}_{\beta }& \left(X,Y\right)\right)-{F}_{\beta }\left({F}_{\alpha }\left(X,Y\right)\right)\end{array}$
 $\begin{array}{cc}\equiv {\varepsilon }^{2}\left(& {\sum }_{i=1}^{l}{\sum }_{j=1}^{m}{\sum }_{k=1k\ne i}^{l}{\delta }_{{\alpha }_{i},y}\left({\delta }_{{\alpha }_{k},x}{\delta }_{{\beta }_{j},y}-{\delta }_{{\alpha }_{k},y}{\delta }_{{\beta }_{j},x}\right)\left({D}_{i,k}\left(\alpha \right){D}_{j,j}\left(\beta \right){D}_{k,i}\left(\alpha \right)\right)\left(X,Y\right),\end{array}$
 $\begin{array}{cc}& {\sum }_{i=1}^{l}{\sum }_{j=1}^{m}{\sum }_{k=1k\ne i}^{l}{\delta }_{{\alpha }_{i},x}\left(-{\delta }_{{\alpha }_{k},x}{\delta }_{{\beta }_{j},y}+{\delta }_{{\alpha }_{k},y}{\delta }_{{\beta }_{j},x}\right)\left({D}_{i,k}\left(\alpha \right){D}_{j,j}\left(\beta \right){D}_{k,i}\left(\alpha \right)\right)\left(X,Y\right)\right)\end{array}$
 $\begin{array}{cc}-{\varepsilon }^{2}\left(& {\sum }_{j=1}^{m}{\sum }_{i=1}^{l}{\sum }_{k=1k\ne j}^{m}{\delta }_{{\beta }_{j},y}\left({\delta }_{{\beta }_{k},x}{\delta }_{{\alpha }_{i},y}-{\delta }_{{\beta }_{k},y}{\delta }_{{\alpha }_{i},x}\right)\left({D}_{j,k}\left(\beta \right){D}_{i,i}\left(\alpha \right){D}_{k,j}\left(\beta \right)\right)\left(X,Y\right),\end{array}$
 $\begin{array}{cc}& {\sum }_{j=1}^{m}{\sum }_{i=1}^{l}{\sum }_{k=1k\ne j}^{m}{\delta }_{{\beta }_{j},x}\left(-{\delta }_{{\beta }_{k},x}{\delta }_{{\alpha }_{i},y}+{\delta }_{{\beta }_{k},y}{\delta }_{{\alpha }_{i},x}\right)\left({D}_{j,k}\left(\beta \right){D}_{i,i}\left(\alpha \right){D}_{k,j}\left(\beta \right)\right)\left(X,Y\right)\right)\end{array}$
 $\begin{array}{}\end{array}$
 $\begin{array}{cc}\equiv {\varepsilon }^{2}\left(& {\sum }_{i=1}^{l}{\sum }_{j=1}^{m}{\sum }_{k=1k\ne i}^{l}{\delta }_{{\alpha }_{i},y}〈{\alpha }_{k},{\beta }_{j}〉\left({D}_{i,k}\left(\alpha \right){D}_{j,j}\left(\beta \right){D}_{k,i}\left(\alpha \right)\right)\left(X,Y\right)\end{array}$
 $\begin{array}{cc}-& {\sum }_{j=1}^{m}{\sum }_{i=1}^{l}{\sum }_{k=1k\ne j}^{m}{\delta }_{{\beta }_{j},y}〈{\beta }_{k},{\alpha }_{i}〉\left({D}_{j,k}\left(\beta \right){D}_{i,i}\left(\alpha \right){D}_{k,j}\left(\beta \right)\right)\left(X,Y\right),\end{array}$
 $\begin{array}{cc}& {\sum }_{i=1}^{l}{\sum }_{j=1}^{m}{\sum }_{k=1k\ne i}^{l}{\delta }_{{\alpha }_{i},x}〈{\beta }_{j},{\alpha }_{k}〉\left({D}_{i,k}\left(\alpha \right){D}_{j,j}\left(\beta \right){D}_{k,i}\left(\alpha \right)\right)\left(X,Y\right)\end{array}$
 $\begin{array}{cc}-& {\sum }_{j=1}^{m}{\sum }_{i=1}^{l}{\sum }_{k=1k\ne j}^{m}{\delta }_{{\beta }_{j},x}〈{\alpha }_{i},{\beta }_{k}〉\left({D}_{j,k}\left(\beta \right){D}_{i,i}\left(\alpha \right){D}_{k,j}\left(\beta \right)\right)\left(X,Y\right)\right)\end{array}$
 $\begin{array}{cc}\equiv {\varepsilon }^{2}& \left(-{q}_{{}_{\left[\overline{\alpha },\overline{\beta }\right]}}\left(X,Y\right),-{p}_{{}_{\left[\overline{\alpha },\overline{\beta }\right]}}\left(X,Y\right)\right)\left(mod{\varepsilon }^{3}\right).\end{array}$
 $\begin{array}{cc}& \end{array}$
 $\begin{array}{}\end{array}$
Each ${V}^{\otimes l}$  has a natural involution $\iota$  induced by $\iota \left({\alpha }_{1}\otimes \cdots \otimes {\alpha }_{l}\right)=\left(-1{\right)}^{l}{\alpha }_{l}\otimes \cdots \otimes {\alpha }_{1}.$  Extending $\iota$  to all of $L\left(V\right)$  , we get a Lie algebra involution; to see this, observe that
 $\begin{array}{cc}\iota \left(\overline{{D}_{i,i}\left(\alpha \right)\otimes {D}_{j,j}\left(\beta \right)}\right)& =\left(-1{\right)}^{l+m}\overline{{D}_{m-j+1,m-j+1}\left(\iota \left(\beta \right)\right)\otimes {D}_{l-i+1,l-i+1}\left(\iota \left(\alpha \right)\right)}\end{array}$
 $\begin{array}{cc}& =\left(-1{\right)}^{l+m}\overline{{D}_{l-i+1,l-i+1}\left(\iota \left(\alpha \right)\right)\otimes {D}_{m-j+1,m-j+1}\left(\iota \left(\beta \right)\right)}\end{array}$
 $\begin{array}{cc}& \end{array}$
 $\begin{array}{}\end{array}$
and since $\left(\iota \left(\alpha \right){\right)}_{l-i+1}={\alpha }_{i}$  , $\left(\iota \left(\beta \right){\right)}_{m-j+1}={\beta }_{j}$  and $\left(-1{\right)}^{l+m-2}=\left(-1{\right)}^{l+m}$  , it follows that $\left[\iota \left(\alpha \right),\iota \left(\beta \right)\right]=\iota \left(\left[\alpha ,\beta \right]\right)$  . We will also need to consider the unsigned involution $I$  induced by $I\left({\alpha }_{1}\otimes \cdots \otimes {\alpha }_{l}\right)={\alpha }_{l}\otimes \cdots \otimes {\alpha }_{1}$  .
Denote by ${P}_{+}\left(V\right)$  and ${P}_{-}\left(V\right)$  , respectively, the $+1$  and $-1$  eigenspaces for $\iota$  .
Then of course ${P}_{+}\left(V\right)$  is a Lie subalgebra of $L\left(V\right)$  and $\left[{P}_{-}\left(V\right),{P}_{-}\left(V\right)\right]\subset {P}_{+}\left(V\right)$  .
Similarly, we let ${P}_{l,+}$  and ${P}_{l,-}$  denote the $+1$  and $-1$  eigenspaces for $\iota$  in ${V}^{\otimes l}$  . To simplify the notation, we will assume that $V$  is finite-dimensional, so that $\mathcal{ℬ}=\left\{{x}_{1},...,{x}_{r}\right\}$  .
Lemma. Suppose $\alpha \in \left({V}^{\otimes l}{\right)}^{\mathbb{Z}/l\mathbb{Z}}$  and write $\alpha =\sum {x}_{i}\otimes {p}_{i}$  , where the ${p}_{i}\in {V}^{\otimes l-1}$  . Then $\alpha \in {P}_{l,+}$  (resp., ${P}_{l,-}$  ) if and only if ${p}_{1},...,{p}_{r}\in {P}_{l-1,-}$  (resp., ${P}_{l-1,+}\right)$  .
• Proof. This is clear, as $\iota \left(\alpha \right)=-\sum \iota \left({p}_{i}\right)\otimes {x}_{i}$  and $\sigma \left(\alpha \right)=\sum {p}_{i}\otimes {x}_{i}$  .
As an application of these definitions, we turn to the case of a Lie algebra $\mathfrak{g}$  defined over $K$  and identify ${V}^{\otimes l}$  with noncommutative polynomials of degree $l$  in ${x}_{1},...,{x}_{r}$  . Our first observation is that if $\mathfrak{g}$  is either $\mathfrak{s}{\mathfrak{o}}_{n}$  or $\mathfrak{s}{\mathfrak{p}}_{n}$  and $p=p\left({x}_{1},...,{x}_{r}\right)$  is in ${\oplus }_{l}{P}_{l,-}$  then for all ${X}_{1},...,{X}_{r}\in \mathfrak{g}$  , $p\left({X}_{1},...,{X}_{r}\right)\in \mathfrak{g}$  . Indeed write $p=\sum {p}_{i}$  , where each ${p}_{i}$  is a monomial. Then if $\mathfrak{g}=\mathfrak{s}{\mathfrak{o}}_{n}$  ,
 $\begin{array}{cc}p\left({X}_{1},...,{X}_{r}{\right)}^{t}& =\sum {p}_{i}\left({X}_{1},...,{X}_{r}{\right)}^{t}=\sum I\left({p}_{i}\right)\left({X}_{1}^{t},...,{X}_{r}^{t}\right)\end{array}$
 $\begin{array}{cc}& =\sum \left(-1{\right)}^{deg{p}_{i}}I\left({p}_{i}\right)\left({X}_{1},...,{X}_{r}\right)=\iota \left(p\right)\left({X}_{1},...,{X}_{r}\right)\end{array}$
 $\begin{array}{cc}& =-p\left({X}_{1},...,{X}_{r}\right).\end{array}$
 $\begin{array}{cc}& \end{array}$
 $\begin{array}{}\end{array}$
If $\mathfrak{g}=\mathfrak{s}{\mathfrak{p}}_{n}$  let $J$  denote the matrix of the nondegenerate alternating form which defines $\mathfrak{g}$  . Then
 $\begin{array}{cc}Jp\left({X}_{1},...,{X}_{r}\right)& =\sum J{p}_{i}\left({X}_{1},...,{X}_{r}\right)=\sum \left(-1{\right)}^{deg{p}_{i}}{p}_{i}\left({X}_{1}^{t},...,{X}_{r}^{t}\right)J\end{array}$
 $\begin{array}{cc}& =\left(\iota \left(p\right)\left({X}_{1},...,{X}_{r}\right){\right)}^{t}J=-p\left({X}_{1},...,{X}_{r}{\right)}^{t}J.\end{array}$
 $\begin{array}{cc}& \end{array}$
 $\begin{array}{}\end{array}$
With $dimV=2$  and $\mathfrak{g}$  of the above type (in particular, $\mathfrak{g}\subset \mathfrak{g}{\mathfrak{l}}_{n}$  ) we consider the above map $\overline{\alpha }↦{F}_{\alpha }$  from $L\left(V\right)$  to the algebra of vector fields on $\mathfrak{g}{\mathfrak{l}}_{n}×\mathfrak{g}{\mathfrak{l}}_{n}$  .
Proposition. The image of ${P}_{+}\left(V\right)$  under this map consists of vector fields on $\mathfrak{g}×\mathfrak{g}$  tangent to the fibers of the commutator map $\mathfrak{g}×\mathfrak{g}\to \mathfrak{g}$  .
• Proof. If $\alpha \in {V}^{\otimes l}$  then $\iota \left(\sigma \left(\alpha \right)\right)={\sigma }^{-1}\left(\iota \left(\alpha \right)\right)$  , which implies that $N\left({P}_{+}\left(V\right)\right)\subset {\oplus }_{l}{P}_{l,+}$  . Moreover if $\alpha \in {P}_{+}\left(V\right)$  and $N\left(\alpha \right)=x{p}_{\alpha }-y{q}_{\alpha }$  then ${p}_{\alpha },{q}_{\alpha }\in {\oplus }_{l}{P}_{l,-}$  by the Lemma. Now we are done, as ${p}_{\alpha }\left(X,Y\right),{q}_{\alpha }\left(X,Y\right)\in \mathfrak{g}$  whenever $X,Y\in \mathfrak{g}$  by the preceding discussion.
References

1. M. Kontsevich, Formal (non)-commutative symplectic geometry, The Gelfand Mathematical Seminars, 1990–1992 (L. Corwin, I. Gelfand, J. Lepowsky, eds.), Birkhauser, 1993, pp. 173–187.

Department of Mathematics, Northwestern University, 2033 Sheridan Road, Evanston, IL 60208 E-mail address : ekushnir@math.northwestern.edu Department of Mathematics, Indiana University, Bloomington, IN 47405 E-mail address : larsen@math.indiana.edu