Calabi-Yau algebras and weighted quiver polyhedra. (Q1936633)

This article is motivated by the construction of 3-dimensional Calabi-Yau (CY) algebras from dimer models, which was in turn motivated by theoretical physics. Previous work by Broomhead, Mozgovoy and Reineke, and by Davison had shown that certain dimer models give rise to 3-dimensional CY algebras via a quiver with potential construction. These algebras, called toric orders, are meant to be thought of as ``noncommutative toric resolutions'' of certain toric varieties. The article under review generalizes the idea of a dimer model to that of a weighted quiver polyhedron, and generalizes the idea of a toric order to that of a cancellation algebra. An algebra is called a cancellation algebra if it is the category algebra of a category in which every morphism is both epic and monic; this gives rise to a cancellation law in the associated category algebra. A quiver polyhedron is a strongly connected quiver (i.e., every vertex lies on an oriented cycle) along with two disjoint sets of cycles in the quiver, \(Q_2^+\) and \(Q_2^-\). These sets of cycles are required to satisfy: (PO) every arrow of \(Q\) appears in exactly one cycle of \(Q_2^+\) and appears in exactly one cycle of \(Q_2^-\); (PM) at each fixed vertex, the incidence graph of the cycles and arrows that meet there is connected. The author shows that this is essentially the same as drawing a strongly connected quiver on a compact, orientable surface, in such a way that erasing the quiver leaves simply connected pieces bounded by cycles. A positive weighting on a quiver polyhedron is an assignment of a positive integer \(E_c\) to each cycle \(c\) in \(Q_2^+\cup Q_2^-\) such that \(E_c|c|>2\), where \(|c|\) denotes the number of arrows in \(c\). This roughly corresponds to putting some orbifold singularities in the surface described above. Now there is a natural superpotential associated to a weighted quiver polyhedron, obtained by taking the difference of the positive and negative cycles but with some additional technicalities from the weighting. A positive grading on a weighted quiver polyhedron is an assignment of a positive real number \(R_a\) to each arrow such that \(R_cE_c\) is the same for every cycle in \(Q_2^+\) and \(Q_2^-\). The existence of such a grading means that the superpotential is homogeneous; the results of the paper do not depend on the exact choice of grading. A group acting on a weighted quiver polyhedron without fixing any vertices corresponds to a cover morphism between orbifolds and Galois cover on the level of Jacobi algebras. The author proves in Theorem 5.8 and Lemma 5.9 that the property of being a cancellation algebra, and that of admitting a positive grading are compatible with the notion of Galois covers. This is important because it allows one to reduce to the case of trivial weighting, with two exceptions (Theorem 5.10). The main theorem then states that if a positively graded cancellation algebra is 3-CY, it necessarily comes from a graded weighted quiver polyhedron (Theorem 6.1). This is used to show in Section 7 that if we restrict our cancellation algebras to the case of toric orders, the underlying manifold of the associated quiver polyhedron must be a torus. In other words, a toric order which is 3-CY must come from a dimer model on a torus (Theorem 7.7).