Convolution algebras for relational groupoids and reduction (Q2234342)

This paper is concerned with the convolution algebra of a so-called relational groupoid. Ultimately, the authors show that there is a reduced convolution algebra (Definition 4.21) of a relational groupoid is isomorphic to the (groupoid) convolution algebra of its reduction (Theorem 4.22). Very briefly, a Haar system on a Lie groupoid is a continuous field of measures on its source fibers that behaves well with the groupoid structure (Definition 2.4). Once a Lie groupoid admits a Haar system, there is a natural convolution product that turns its space of compactly supported functions into an associative algebra (Definition 2.5 and Proposition 2.6). In the present article, the authors go about extending these notions to the realm of relational groupoids. A relational groupoid is a generalization of a groupoid, where, roughly, the structural maps are relaxed to be relations rather than functions (Definition 3.1). These objects appeared naturally when studying the infinite-dimensional space of cotangent paths on a non-integrable Poisson manifold, and, in fact, they can always be ``reduced'', i.e., quoted out by certain redundancies to yield an actual groupoid (Theorem 3.16). In the smooth case, the necessary notions to define the relational convolution algebra are defined: A relational Haar system (Definition 4.2) and a subspace on which the convolution product is defined (Definition 4.9 and Proposition 4.10). As it turns out, the relational convolution algebra is not always associative (Example 4.17), so this question is studied identifying: a subspace on which the convolution is always associative (Proposition 4.13), and a condition on the relational Haar system that ensures the associativity (Proposition 4.15). The article is absolutely well-written. Definitions are in general well motivated, mostly accompanied with (lots of) meaningful examples, as well as with reinterpretations to help the reader move along. In addition, the introduction gives a fine overview on the topic of Groupoid Quantization (see, e.g., [\textit{E. Hawkins}, J. Symplectic Geom. 6, No. 1, 61--125 (2008; Zbl 1154.46041)]) and its relation with the larger quantization landscape. The paper closes with a series of future directions and an appendix with some results on relational groupoids. An innocuous typo in the last word in the statement of Proposition 4.10 ``\(\mathcal{A})(\mathcal{G})\)'' should read ``\(\mathcal{A}(\mathcal{G})\)''.