Confirmation of Matheron's conjecture on the covariogram of a planar convex body (Q1041255): Difference between revisions

Let \(K\) be a compact set in the Euclidean space \({\mathbb R}^d\), \(d\geq 2\) and \(K+x\) be the translation of \(K\) to the vector \(x\), \(x\in {\mathbb R}^d\). The covariogram \(g_K\) of \(K\) is the function on \({\mathbb R}^d\) defined by \(g_K(x)= V_d(K\cap (K+x))\), where \(V_d\) stands for the \(d\)-dimensional Lebesgue measure. This function, which was introduced by \textit{G. Matheron} in his book [1] [Random Sets and Integral Geometry. Wiley Series in Probability and Mathematical Statistics. New York etc.: John Wiley\&Sons (1975; Zbl 0321.60009)] on random sets, is also called set covariance. The covariogram \(g_K\) coincides with the autocorrelation of the characteristic function \({\mathbf 1}_K\) of \(K\), i.e., \(g_K={\mathbf 1}_K*{\mathbf 1}_{-K}\). The covariogram \(g_K\) is clearly unchanged with respect to translations and reflections in a point of \(K\). A convex body in \({\mathbb R}^d\) is a convex compact set with nonempty interior. In 1986, \textit{G. Matheron} [2] [``Le covariogramme géométrique des compacts convexes des \({\mathbb R}^2\)'', Technical report N-2/86/G, Centre de Géostatistique, Ecole Nationale Supérieure des Mines de Paris (1986)] asked the following question and conjectured a positive answer for the case \(d=2\). Covariogram Problem. Does the covariogram determine a convex body in \({\mathbb R}^d\), among all convex bodies, up to translations and reflections? The authors solve Matheron's conjecture for \(d=2\) completely: Theorem 1. Every planar convex body is determined within all planar convex bodies by its covariogram, up to translations and reflections. The covariogram problem is equivalent to any of the following three problems: \textbf{I)} Determine a convex body \(K\) by the knowledge, for each unit vector \(u\) in \({\mathbb R}^d\), of the distribution of the lengths of the chords of \(K\) parallel to \(u\); \textbf{II)} Determine a convex body \(K\) by the distribution of \(X-Y\), where \(X\) and \(Y\) are independent random variables uniformly distributed over \(K\). \textbf{III)} Determine the characteristic function \({\mathbf I}_K\) of a convex body \(K\) from the modulus of its Fourier transform \(\widehat{{\mathbf I}_K}\). In each of the three problems the determination is unique up to translations and reflections of the body. In Problem \textbf{I)} a random chord parallel to \(u\) is obtained by taking the intersection of \(K\) with a random invariant line \(L_u\) parallel to \(u\), conditioned on \(K\cap L_u\neq \emptyset\). Matheron in [1] explained the relation between \textbf{I)} and the covariogram of a set. The first partial solution of Matheron's conjecture was given by \textit{W. Nagel} [3] [J. Appl. Probab. 30, No.3, 730--736 (1993; Zbl 0781.60018)], who confirmed it for all convex polygons. In [4] [J. Differential Geom. 60, 177--198 (2002; Zbl 1047.52002)], \textit{G. Bianchi, F. Segala} and \textit{A. Volčič} gave a positive answer to the covariogram problem for all planar convex bodies whose boundary has strictly positive continuous curvature. In [5] [J. Lond. Math. Soc., II. Ser. 71, 203--220 (2005; Zbl 1067.52001)], \textit{G. Bianchi} proved a common generalization of results of [4] and [3]. In [6] [Adv. Appl. Probab. 39, 613--629 (2007; Zbl 1127.52004)], the authors of the present paper studied how much of the covariogram data is needed for the uniqueness of the determination, and also extended the class of bodies for which the conjecture was confirmed. In view of results from [5], to prove the main result (Theorem 1) it suffices to derive the following statement: Proposition 1. Let \(K\) and \(L\) be planar strictly convex and \(C^1\) regular bodies with equal covariograms. Then \(L\) possesses a non-degenerate boundary arc whose translation or reflection lies in the boundary of \(K\). Theorem 1 follows directly from Proposition 1 and the following two statements. Theorem 2 (see [5]). Let \(K\) and \(L\) be planar convex bodies with equal covariograms. Assume that one of them is not strictly convex or not \({\mathbb C}^1\) regular. Then \(K\) and \(L\) are translations or reflections of each other. Proposition 2 (see [5]). Let \(K\) and \(L\) be planar convex bodies with equal covariograms and a common non-degenerate boundary arc. Then \(K\) and \(L\) coincide up to translations and reflections. The covariogram problem in the general setting has a negative answer, as G. Bianchi (see [5]) proved by finding counterexamples in \({\mathbb R}^d\) for every \(d\geq 4\). In [7] [\textit{G. Bianchi}, Adv. Math., 220, 1771--1808 (2009; Zbl 1159.60012)], it is proved that a convex three-dimensional polytope is determined by its covariogram. The authors of the present paper mention three open problems: 1. Which four-dimensional convex polytopes are determined by their covariogram? 2. All known examples of convex bodies that are not determined by their covariogram are Cartesian products. Do there exist other examples? 3. Is the answer to the covariogram problem positive for all three-dimensional convex bodies whose boundary has continuous and strictly positive principal curvatures?