Vector partition functions and generalized Dahmen and Micchelli spaces (Q626569): Difference between revisions

This is the first in a series of papers of the authors on vector partition functions and the index theory of transversally elliptic operators. In this paper, only algebraic and combinatorial issues related to partition functions are discussed. Let \(\Gamma\) be a lattice in a vector space \(V\) and \(X\) be a list of nonzero elements of \(\Gamma\) spanning \(V\). If \(X\) generates a pointed cone, the partition function \(\mathcal{P}_X(\gamma)\) counts the number of ways in which \(\gamma\in \Gamma\) can be written as a linear combination of elements in \(X\) with nonnegative integer coefficients. A sublist \(Y\) of \(X\) is called a cocircuit if \(X\setminus Y\) does not span \(V\) and \(Y\) is minimal with this property. Given \(a\in \Gamma,\) the difference operator \(\nabla_a\) acts on functions by \(\nabla_a(f)(b):=f(b)-f(b-a).\) Given a list \(Y\), let \(\nabla_Y:=\prod_{a\in Y} \nabla_a.\) \textit{W. Dahmen} and \textit{C. A. Micchelli} [Trans. Am. Math. Soc. 308, No.~2, 509--532 (1988; Zbl 0655.10013)] introduced the space \(DM(X)\) consisting of functions \(f\) on \(\Gamma\) satisfying \(\nabla_Y(f)=0\) for every cocircuit \(Y\). Every function in \(DM(X)\) is a quasi-polynomial, i.e., a function which coincides with a polynomial on each coset of some sublattice of finite index in \(\Gamma.\) In this paper, a space \(\mathcal{F}(X)\) generating the space \(DM(X)\) is introduced. A subspace \(\underline{r}\) of \(V\) is called rational if it is the span of a sublist of \(X\). The space \(\mathcal{F}(X)\) consists of functions \(f\in \Gamma\) such that \(\nabla_{X\setminus \underline{r}}(f)\) is supported on \(\underline{r}\) for every rational \(\underline{r}\). The main result of this paper is a localization formula for a function in \(\mathcal{F}(X)\). A tope is a connected component of the complement in \(V\) of the union of all proper rational subspaces. Given a tope \(\tau,\) a function \(f\in \mathcal{F}(X)\) can be written as a sum of a quasi-polynomial \(f^\tau\in DM(X)\) and of functions \(f_{\underline{r}}\in \mathcal{F}(X)\) supported outside the Minkowski difference \(\tau-B(X)\), where \(B(X)\) is the zonotope of \(X\), i.e., the set of linear combinations of elements in \(X\) with real coefficients in the interval \([0,1].\) Since the partition function \(\mathcal{P}_X\) is contained in \(\mathcal{F}(X)\), the localization formula, together with a wall crossing formula, allows the authors to give a short proof of the result that the partition function \(\mathcal{P}_X\) is a quasi-polynomial on the Minkowski difference \(\mathfrak{c}-B(X)\), where \(\mathfrak{c}\) is a big cell, i.e., a connected component of the complement in \(V\) of the union of all cones spanned by all sublists of \(X\) which do not span \(V\). This result was proved by Dahmen and Micchelli [loc. cit.] for topes and by \textit{A. Szenes} and \textit{M. Vergne} [Adv. Appl. Math. 30, No.~1--2, 295--342 (2003; Zbl 1067.52014)] for big cells.