Curves on surfaces, charts, and words (Q2490388)

The problem to describe in a combinatorial way the smooth immersions of a curve in the plane has a long history and goes back~-- at least~-- to Gauss. There is now a wealth of literature on this topic. Two related papers worth to be mentioned (but not done so by the author here) are [\textit{P. Rosenstiehl}, Adv. Appl. Math. 23, No. 1, 3--13 (1999; Zbl 0935.05036) and \textit{H. de Fraysseix} and \textit{P. Ossona de Mendez}, Discrete Comput. Geom. 22, No. 2, 287--295 (1999; Zbl 0932.05024)]. There are obvious variations and generalizations of the Gauss word problem. In this paper, the author studies curves immersed on orientable compact surfaces of arbitrary genus (not only zero) and allows curves to have self-tangencies (as opposed to transversal intersections) and crossings of arbitrary multiplicity (not only two). He also equips curves with a finite number of distinguished points (outside crossings) he calls corners. (They can often be formally treated as crossings of multiplicity 1.) Depending, for example, on the local behaviour near crossings (as transversality), he distinguishes between various types of curves. A curve is called filling if its complementary regions on the surface are all disks. Then he goes on to define combinatorial objects that characterize these curves. The first such object is called a chart. A chart is a permutation of the set \(\{-n,1-n,\dots,-1,1,\dots,n\}\), where \(n\) is the number of crossings (counted with their multiplicity) and corners. A detailed study of charts of the different types of curves is carried out. For self-transversal curves (all crossings are transversal) the information in a chart can be further compressed to a semichart. A semichart consists of a permutation of \(\{1,\dots, n\}\), and a subset \(S\) thereof, whose intersections with each orbit of the permutation have an odd number of elements. It is shown that these objects describe a filling curve uniquely and completely (lemmas 3.4.1 and 3.5.1 for a chart and theorem 5.2.2 for a semichart), from which an enumeration of curves depending on the number of corners or crossings of given multiplicity is possible (theorems 6.2.1 and 6.2.3). The second type of combinatorial objects describing curves are (naturally) words. Again several results (section 8) illustrate the correspondence between properties of curves (several new ones are added previously in section 7) and their words. Then follow two formulas for calculating the genus of a curve (i.e. the least genus of a surface, on which the curve is immersable, or where it becomes a filling curve). The first formula, theorem 9.1.1, using the chart, is rather easy. The second one, the author calls ``homological'', requires more work (this is the longest proof in the paper). It expresses the genus as the rank of a certain matrix. The entries can be understood as made up of linking numbers (modulo 2) with a set of generators of the homology of some surface on which the curve is immersed. It is stated in slightly different terms already in Section 8 as Theorem 8.5.4. The paper is clearly written, in the sense that the notions introduced are well explained and the arguments used are (where the reviewer checked) easy to follow. So far this makes it an interesting article, easy to read. The reviewer found only 4 typos. On p. 216 line -8, \(\sigma(A)\) should be \(\sigma(S)\), and on p. 218 line -1, the word `connection' is written in French. Corollary 8.5.5 should end with a period, and the box at the end of proof of lemma 6.1.1 should be one line below, at the display. (This lemma, by the way, is well-known to combinatorialists as Burnside's lemma, and should have been better mentioned so.) More serious drawbacks concern the level and structure of presentation. The kind of tasks posed and answers given do not contain any original or fundamental new ideas. The proofs consist mostly of straightforward calculations or observations, and do not reveal any serious degree of difficulty, neither topological nor combinatorial. It remains unclear what substantial problems this work aims to help solving. One presumably hard(er) problem, for example, is to enumerate filling (say, self-transversal and generic) curves by taking the genus into account. In genus 0 this relates to the problem to enumerate alternating \textit{knots} (as opposed to links of an arbitrary number of components) by crossing number, see [\textit{P. Zinn-Justin} and \textit{J.-B. Zuber}, J. Knot Theory Ramifications 9, No. 8, 1127--1141 (2000; Zbl 0984.57001)]. This question, however, remains untouched in the present paper. Besides, there are longer passages of commentary prosa on the relations of curves/words to various other topics like Riemannian surfaces, Coxeter groups, knots (now obtained via A'Campo's constructions), algebraic curves etc. (see page 205 of the introduction, Sections 3.7, 5.4 and 8.7 (1)). While important, and useful to a limited extent to motivate some of the particular notions studied, these relations are not substantially (in terms of concrete statements) pursued on their own, or applied (as arguments in proofs) to the main context. This, together with the lack of a central result to point the reader to, makes the presentation lose a bit of focus, and the style into some mixture between a research article and a survey article. If a partly expository purpose was in fact intended, this should have been indicated more clearly to the reader in advance.