On the coordinate rings of Calogero-Moser spaces and the invariant commuting variety of a pair of matrices

The first identity was already obtained in [N], but we include it here for the sake of completion. Let us prove the second one.

Consider a quadruple (X, Y, v, w) as in (?). It is evident that the identity

also holds, where A and B are the traceless versions of the matrices X and Y, respectively. By taking traces we get that

Moreover, multiplying (Proof) by A^k, we obtain that

for any integer k. In a similar fashion, we get

On one hand, taking traces on both sides yields:

On the other hand, we can operate as follows:

which implies:

Note that

Therefore, the identity follows by equalising (Proof) and (Proof).

Similarly, multiplying (Proof) by A^2 from the left side and by AB from the right side, we obtain:

Note that

so using (Proof) together with (2.1) we deduce the desired identity.

Let us consider the formal power expansion of \operatorname{Tr}((A+tB)^6). The coefficient at t^2 is

which, making use of equations (?), (2.1) and (2.1), together with Lemma 2.1, is equal to

On the other hand, applying the Cayley-Hamilton theorem (?) first, we derive

After expanding the powers, gathering the coefficients, and further applying the already known substitutions together with equation (2.1), we spot that the coefficient at t^2 can be written as

Therefore, the expressions obtained in (Proof) and (Proof) must be equal in \mathbb{C}[\mathcal{C}_4], resulting in the desired relation.

The outline of the proof is as follows: we shall strategically obtain a set of relations by using the Poisson structure and the Fourier map (?) belonging to the non-associative ideal generated by r_1. Then, we shall define an injective morphism of \mathbb{C}[a_1, a_2, a_3, a_5, a_6, a_9, a_{10}, a_{14}]-modules from the free module generated by the monomials listed in Appendix ? to the canonical module structure of \mathbb{C}[a_1,a_2]\otimes \mathbb{C}[a_3,\dots, a_{14}]/I and verify that their Hilbert series coincide. Therefore, since \mathbb{C}[a_1,a_2]\otimes \mathbb{C}[a_3,\dots, a_{14}]/I sits inside \mathbb{C}[\mathcal{C}_4] and they are both free of the same rank, they are isomorphic.

First, let us introduce the following polynomials (whose exact expressions are written in the Appendix ?):

Applying the Fourier map, we automatically get t_1, \dots, t_6, respectively. Finally, we obtain

which are invariant under the Fourier map. Note that the coefficients placed in front of the brackets are there just to ease the notation of each corresponding relation.

Let I be the polynomial ideal generated by r_1, \dots, r_5, t_1, \dots, t_4, s_1, \dots, s_3. With the help of a computer algebra system as Macaulay2 [Mac2] we find that

is the Hilbert series of \mathbb{C}[a_1, \dots, a_{14}]/I, graded by the sum of its natural bigrading.

We know that \mathbb{C}[\mathcal{C}_4] is a free \mathbb{C}[a_1, a_2, a_3, a_5, a_6, a_9, a_{10}, a_{14}]-module of rank 24, as mentioned in Remark 3.4. We denote by M the free \mathbb{C}[a_1, a_2, a_3, a_5, a_6, a_9, a_{10}, a_{14}]-module generated by the elements of Appendix ?. There is a canonical map defined by the composition

To verify that it is injective, we compute a Gröbner basis with respect to the weighted degree reverse lexicographical order of the ideal generated by I together with a_3, a_5, a_6, a_9, a_{10}, a_{14}. Then, it is a matter of checking that the basis of generators listed in Appendix ? are normal forms with respect to the Gröbner basis. Finally, an examination of the degrees shows that the Hilbert series of M is exactly the same as (Proof).

The method we use to find the relations is the following: We compute a Gröbner basis of I (from the Theorem 3.5) with respect to the weighted degree reverse lexicographical order, obtaining the 15 polynomials listed in Appendix ?. Then, for each non-homogeneous polynomial we remove the lower degree terms, obtaining a set of homogeneous polynomials, that are listed in Appendix ?.

It is a straightforward computation to check that all the polynomials obtained are relations in the coordinate ring of the invariant commuting variety. Then, to finish we replicate the final part of the proof of Theorem 3.5.

One of the standard ways to compute this invariant is to proceed by divide and conquer. Firstly, since \mathbb{C}[{\mathcal Com_4}] = \mathbb{C}[a_1, a_2] \otimes \mathbb{C}[a_3, \dots, a_{14}]/J, it is clear that

Let us denote by \widetilde{r_i}, \widetilde{s_i} and \widetilde{t_i} the corresponding identities generating J from Appendix ?. We make the following transformations in order to ease the computations:

Assisted by Gröbner bases it is immediate to check that, substituting the prime versions of the relations by the original ones, we still generate the same ideal. Solving directly the equations for the variables a_5, a_8, a_9, a_{12}, a_{13}, a_{14}, we get a complete solution where the values of a_4, a_7 and a_{11} are free, and the following polynomial

appears in the denominators of the solutions. Note that this polynomial involves only the variables a_3, a_6 and a_{10}, which correspond to the traces of the powers of the traceless form of the first matrix X in the pair. Furthermore, it is exactly

where \lambda_1,\dots, \lambda_4 are the eigenvalues of X.

We denote by V_{{w_1}} the subvariety \operatorname{Var}(J \cup \{{w_1}\}) and by U_{w_1} the open subset of \operatorname{Var}(J) where {w_1} \neq 0 (i.e., X admits a simple spectrum), then

In order to compute the class of V_{w_1}, we add {w_1} to the set of equations, and we can now solve directly the system whenever a_6 (-4a_{10} a_3 + a_3^3 + 4 a_6^2) \neq 0. Let us denote:

Note that, as in w_1, only the variables a_3, a_6 and a_{10} appear in the polynomials w_{21} and w_{22}.

Therefore, V_{w_1} can be decomposed as the union of V_{1}, V_{21}, V_{22}, V_{23}, where

Solving directly the equations, we can find the following:

• In V_1, the variables a_4, a_8 and a_{11} are free, while a_5, a_7, a_9, a_{12}, a_{13}, a_{14} are completely determined.

• In V_{2}, the variables a_4, a_5 and a_{7} are free, while the variables a_{8}, a_{9}, a_{11}, a_{12}, a_{13}, a_{14} are completely determined.

• In V_{3}, the variables a_4, a_8 and a_{11} are free, while the variables a_{5}, a_{7}, a_{11}, a_{12}, a_{13}, a_{14} are completely determined.

• In V_{4}, the variables a_4, a_5 and a_{8} are free, while the variables a_{7}, a_{9}, a_{11}, a_{12}, a_{13}, a_{14} are completely determined.

Therefore,

Factoring out in

we realize that the sum of all the classes left to find is exactly the class of the union of a decomposition of \mathbb{C}^3, obtaining

and, consequently, [{\mathcal Com_4}] = \mathbb{L}^8.