The Euler characteristic of a category as the sum of a divergent series (Q2468651)

The Euler characteristic of a finite cell complex is most often described as the alternating sum of the number of cells of each dimension. There seems little hope of extending this formula to complexes containing infinitely many cells of same dimension. However, there are interesting complexes in which there are only finitely many cells of each dimension, but infinitely many in total. The classifying space of a finite group provides an example. Writing \(C_n\) for the number of \(n\)-dimensional cells, the authors would like to find a sensible way of evaluating the divergent series \(\sum_{n\in\mathbb{N}}(-1)^n c_n\), which could then be interpreted as the Euler characteristic of the complex. To see how this might work, the authors consider a finite group \(G\). Its classifying space \(BG\) is the geometric realization of a simplicial set in which an \(n\)-simplex is an \(n\)-tuple of elements of \(G\). The nondegenerate \(n\)-simplices are \(n\)-tuples of nonidentity elements, so, writing \(o(G)\) for the order of \(G\), there are \((o(G)-1)^n\) of them. A simplicial set may be regarded as a kind of complex in which the cells are the nondegenerate simplices, so the task is to evaluate \(\sum_{n\in \mathbb{N}}(-1)^n(o(G)-1)^n\). In the Eulerian spirit of formal calculation, the authors apply the formula for the sum of a geometric series, which gives the answer \(1/o(G)\). And indeed, it has been established that this is the `right' value for the Euler characteristic of \(G\) (or \(BG\)) from several points of view, for example that of \textit{C. T. C. Wall} [Proc. Camb. Philos. Soc. 57, 182--184 (1961; Zbl 0100.25701)] or of \textit{J. Baez} and \textit{J. Dolan} [``From finite sets to Feynman diagrams'', in: Mathematics Unlimited -- 2001 and beyond, Berlin: Springer. 29--50 (2001; Zbl 1004.18001)]. Here is a first step towards making this rigorous. Take a cell complex \(X\) (in some nonspecific sense). Suppose that \(X\) has only a finite number \(C_n\) of cells of each dimension \(n\), and the authors write \(f_X(t)=\sum_{n\in \mathbb{N}}c_n t^n\) for the resulting formal power series. It may be that \(f_X\) converges in some neighborhood of \(0\). If so, it may also be that \(f_X\) can be analytically continued to \(-1\), and it may even be that all such analytic continuations take the same value at \(-1\). We could then define the Euler characteristic of \(X\) to be that value. Of course, if \(X\) has only finitely many cells then the situations is very simple: \(F_X\) is a polynomial, there is a unique analytic continuation to \(-1\), and its value there (namely, \(f_X(-1)\)) is the Euler characteristic of \(X\) in the usual sense. The purpose of this paper is to use this approach to derive a notion of the Euler characteristic of a finite category. For a ring \(K\) and a natural number \(m\), write \(\text{Mat}_m(K)\) for the ring of \(m\times m\) matrices over \(K\). Given a finite category \({\mathcal A}\) with objects \(a_1,\dots,a_m\), write \(Z_{\mathcal A}\in \text{Mat}_m(\mathbb{Q})\) for the matrix whose \((i,j)\)-entry is the number of arrows from \(a_i\) to \(a_j\). A weighting on \({\mathcal A}\) is an \(m\)-tuple \(w^\bullet=(w^1,\dots,w^m)\in \mathbb{Q}^m\) such that \(Z_{\mathcal A} {w^\bullet}^t =(1,\dots,1)^t\). A coweighting on \({\mathcal A}\) is an \(m\)-tuple \(w_\bullet=(w_1,\dots,w_m)\in \mathbb{Q}^m\) such that \(w_\bullet Z_{\mathcal A}=(1,\dots,1)\). If \(w^\bullet\) is a weighting and \(w_\bullet\) a coweighting on \({\mathcal A}\) then \(\sum_i w^i=\sum_i w_i\). A finite category \({\mathcal A}\) has Euler caracteristic if it admits both a weighting and a coweighting, and in that case its Euler characteristic \(\chi({\mathcal A})\) is \(\sum_i w^i=\sum_i w_i \in \mathbb{Q}\), for any weighting \(w^\bullet\) and coweighting \(w_\bullet\). The Euler characteristic of a category is independent of the choice of ordering of the objects. It is also independent of the composition and identities; that is, it depends only on the underlying directed graph. An important special case is when \(Z_{\mathcal A}\) is invertible. Then \({\mathcal A}\) is said to have a Möbius inversion, if there are a unique weighting and a unique coweighting, and \(\chi({\mathcal A})\) is the sum of the entries of \(Z_{\mathcal A}^{-1}\). The Euler characteristic of categories enjoys many good properties. It is invariant under equivalence and behaves predictably with respect to products, fibrations, etc. It is also compatible with Euler charcateristic of other types of structure, including topological spaces, graphs, posets, groups, manifolds, and orbifolds. Given a simplicial set \(X\) with only finitely many simplices of each dimension, let \(C_n\) be the number of nondegenerate \(n\)-simplices and \(f_X(t)=\sum_{n\in \mathbb{N}} c_n t^n \in \mathbb{Q}[[t]]\). Then for a finite category \({\mathcal A}\) is defined \(f_{\mathcal A}:=f_{N{\mathcal A}}\) where \(N{\mathcal A}\) is the nerve of \({\mathcal A}\). In the paper is proved that for any finite category \({\mathcal A}\), the formal power series \(f_{\mathcal A}\) is rational (over \(\mathbb{Q}\)). Then the author says that a finite category \({\mathcal A}\) has series Euler characteristic if \(f_{\mathcal A}(-1)\in\mathbb Q\). In that case, its series Euler characteristic is \(\chi_\Sigma ({\mathcal A})=f_{\mathcal A}(-1)\). Then the authors discuss some positive and negative results: Let \({\mathcal A}\) be a finite category equivalent to some category with Möbius inversion. Then \({\mathcal A}\) has both Euler characteristic and series Euler charcteristic, and \(\chi({\mathcal A})=\chi_\Sigma({\mathcal A})\). -- There exists a \(4\)-object category \({\mathcal A}\) which not has Euler characteristic but has series Euler characteristic. -- A category may have Euler characteristic but no series Euler characteristic. -- A finite category may have neither Euler characteristic nor series Euler characteristic. -- A series Euler charcteristic is not invariant under equivalence.