On singularity confinement for the pentagram map (Q372835)

The pentagram map \(T\) takes an \(n\)-sided polygon to another \(n\)-sided polygon. It was introduced by \textit{R. Schwartz} [Exp. Math. 1, No. 1, 71--81 (1992; Zbl 0765.52004)] and can be defined on the space of twisted polygons. A twisted polygon is a sequence \(A=(A_i)_{i\in { Z}}\) of points in the projective plane that is periodic, period \(n\), modulo a projective transformation \(\phi\), i.e., \(A_{i+n}=\phi(A_i)\) for all \(i\in { Z}\). Further it is required that no 4 consecutive points are collinear. Given \(A\) the pentagram map produces a twisted polygon \(T(A)=B\) defined by \(B_i=\overleftrightarrow{ A_{i-1}A_{i+1}} \cap \overleftrightarrow{ A_{i}A_{i+2}}\). Making use of the cross product, a projective invariant, Schwarz introduced generic coordinates \(x_1,\dots,x_{2n}\), \( x_i=x_{i+2n}\) for \(A\) such that the coordinates of \(T(A)\) are simple rational functions of the coordinates of \(A\). Using these coordinates it can be shown that the pentagram map is a discrete integrable system. In this paper the author studies the singular points of the pentagram map. Algebraically these are polygons \(A\) such that at least one of the denominators vanishes in the coordinates for \(T(A)\). This occurs if \(x_{2i}(A)x_{2i+1}(A)=1\) for some \(i\in Z\). Typically, such an \(A\) is a singular point of \(T^k\) for all \(k\) less than some \(m\), but not for \(T^m\). This behavior is called \textit{singularity confinement}, a feature common to many discrete integrable systems. The author analyzes singularity confinement and gives evidence that the behavior of the singularity depends on the set \(S\) of singularities \(i\) for which \(x_{2i}(A)x_{2i+1}(A)=1\). If \(n\) is odd, he has rather complete results, including an upper bound for \(m\) in terms of \(S\); for even \(n\) only conjectures. A main geometrical result is a straightedge construction of \(T^m(A)\) which is proven to be correct for certain singularity types \(S\). The author discusses at great length what needs to be done to extend the validity of the algorithm. The analysis is highly geometrical and combinatorial, including results about alternating sign matrices.