Line Transversals to Disjoint Balls (Joint with C. Borcea, O. Cheong, X. Goaoc, A. Holmsen) Sylvain Petitjean LORIA, Vegas team O. Schwarzkopf
A long time ago… E. Helly (1923): n convex sets in R d have a point in common iff every d+1 have a point in common. Basic combinatorial result on convex sets E. Helly (1923): n convex sets in R d have a point in common iff every d+1 have a point in common. Basic combinatorial result on convex sets
Transversals Reformulate: points hitting convex sets Raises the obvious question: can one generalize to lines hitting convex sets? (line transversals) No! Bummer. Raises the obvious question: can one generalize to lines hitting convex sets? (line transversals) No! Bummer.
Milestones Danzer (1957): n disjoint unit discs in R 2 have a line transversal if and only if every 5 discs have a line transversal. shape is important, convexity not enough Danzer (1957): n disjoint unit discs in R 2 have a line transversal if and only if every 5 discs have a line transversal. shape is important, convexity not enough Hadwiger (1957): n disjoint convex sets in R 2 have a line transversal if and only if every triple has a transversal consistent with some fixed order order is important Hadwiger (1957): n disjoint convex sets in R 2 have a line transversal if and only if every triple has a transversal consistent with some fixed order order is important
In 3D: bummer again! Holmsen-Matousek (2004): No Helly-type theorem for translates of convex sets, not even with a restriction on the ordering (à la Hadwiger) geometric permutations ≠ isotopy equiv. induced by ordering ≠ equiv. induced by connected components
What about balls? Danzer’s conjecture: Helly for disjoint balls in nD typeHadwiger numberHelly number Hadwiger (1957) & Grünbaum (1960) thinly distributed in R d d 2 2d-1 Holmsen et al. (2003) & Cheong et al. (2005) disjoint unit in R 3 12 646 11 Cheong et al. (2006)pairwise-inflatable in R d 2d4d-1 Borcea et al. (2007)disjoint in R d 2d
Convexity of cone of directions Borcea, Goaoc, P. (2007): Directions of oriented lines stabbing a finite family of disjoint balls in R d in a given order form a strictly convex subset of S d-1 Instrumental in most proofs in transversal theory Previously known for thinly distributed balls (Hadwiger), pairwise-inflatable balls
3D case: 3 disjoint balls
New proof technique Write down equations conics and sextic Write down equations conics and sextic Identify the border arcs Prove Hessian does not meet them local convexity Prove Hessian does not meet them local convexity Argue that cone is contractible
Disjointness is a natural boundary
Extension to higher dimensions
Implications: disjoint balls Isotopy = geometric permutations Smorodinsky et al. (2000): n disjoint balls in R d have (n d-1 ) geometric permutations in the worst case same bound for connected components, previous was O(n 2d-4 ); also better bound in R 3 Isotopy = geometric permutations Smorodinsky et al. (2000): n disjoint balls in R d have (n d-1 ) geometric permutations in the worst case same bound for connected components, previous was O(n 2d-4 ); also better bound in R 3 Hadwiger-type theorem with constant ≤ 2d But no Helly-type! (need constant bound on geometric permutations) Hadwiger-type theorem with constant ≤ 2d But no Helly-type! (need constant bound on geometric permutations)
Conclusions and perspectives Disjoint balls are nice wrt line transversals! … but undoubtedly exceptions Disjoint balls are nice wrt line transversals! … but undoubtedly exceptions Optimality (gap between lower and upper bounds) congruent balls in 3D: Hadwiger between 5 and 6, Helly between 5 and 11 Number of geometric permutations of disjoint unit balls in R 3 : 2 or 3? Algorithmic perspectives: GLP Optimality (gap between lower and upper bounds) congruent balls in 3D: Hadwiger between 5 and 6, Helly between 5 and 11 Number of geometric permutations of disjoint unit balls in R 3 : 2 or 3? Algorithmic perspectives: GLP Thanks for your attention!