No label defined (Q4530755)

The notion of the global dimension of an abelian category \(\mathcal A\) generalizes naturally the global dimension of the category of modules over a ring \(R\) which is in turn equal to the global dimension of the ring \(R\). Formally the global dimension \(\text{gl.dim}_{\mathcal P} \mathcal A\) of an abelian category \(\mathcal A\) with respect to a proper class of short exact sequences \({\mathcal P}\) [see chapter IX, section 4 in the book of \textit{S. MacLane}, ``Homology'', Springer, Berlin (1963; Zbl 0133.26502)] may be defined as the maximum of such \(n\) that \(\text{Ext}_{\mathcal P}^n\) is not a zero functor. When \({\mathcal P}={\mathcal E}\) is the class of all short exact sequences this is simply the global dimension of \(\mathcal A\). Now let \(\mathcal C\) be a partially ordered set and let \({\mathcal A}^{\mathcal C}\) be the category of functors from \(\mathcal C\) into \(\mathcal A\). The author is interested in the study of relations between the global dimension of \(\mathcal A\) and that of the category \({\mathcal A}^{\mathcal C}\) of functors from \(\mathcal C\) into \(\mathcal A\) (the category of diagrams). As a result, he characterizes finite partially ordered sets such that \[ \text{gl.dim}_{\mathcal C\mathcal P} {\mathcal A}^\mathcal C= \text{gl.dim}_{\mathcal P} \mathcal A, \] where \({\mathcal C \mathcal P}\) is the proper class of short exact sequences of functors generated by the class \({ \mathcal P}\); he also characterizes the sets such that \[ \text{gl.dim}_{\mathcal C\mathcal P} {\mathcal A}^\mathcal E = 1+ \text{gl.dim}_{\mathcal P} \mathcal A, \] proves that \[ \text{gl.dim}_{\mathcal C\mathcal P} {\mathcal A}^\mathcal C \leq \dim \mathcal C + \text{gl.dim}_{\mathcal P} \mathcal A, \] and studies the cases in which \[ \text{gl.dim}_{\mathcal C\mathcal P} {\mathcal A}^\mathcal C \leq 2+ \text{gl.dim}_{\mathcal P} \mathcal A \] and \[ \text{gl.dim}_{\mathcal C\mathcal P} {\mathcal A}^\mathcal C \leq 3+ \text{gl.dim}_{\mathcal P} \mathcal A. \] Finally, he characterizes the situation in which \[ \text{gl.dim}_{\mathcal C\mathcal P} {\mathcal A}^\mathcal E < \dim {\mathcal C} + \text{gl.dim}_{\mathcal P} \mathcal A. \] The methods of the proofs are based on the spectral sequence with the term \(E^{p,q}_2\) isomorphic to \[ \widetilde{ H}^{p-2}(]a,b[, \text{Ext}_{\mathcal P}^q(A,B)) \] and converging to \(\text{Ext}_{\mathcal C \mathcal P}^q(A[a],B[b]))\), where \(]a,b[\) is the interval \(\{x\in\mathcal C : a<x<b\}\) and the functor \(A[a]\) from \(\mathcal C\) into \(\mathcal A\) is defined by the following rule: \(A[a](x)= a\) for \(x=a\) and zero otherwise. This spectral sequence is a generalization of the spectral sequence constructed by \textit{M. Dzhibladze} and \textit{T. Pirashvili} [J. Algebra 137, No. 2, 253-296 (1991; Zbl 0724.18005)] for functors with values in the category of \(R\)-modules. In the article under review the author refers to the theorem of existence of this spectral sequence as the main theorem.