CIRCUMCENTER OF MASS Emmanuel Tsukerman University of California, Berkeley Based on joint work with Prof.
Download ReportTranscript CIRCUMCENTER OF MASS Emmanuel Tsukerman University of California, Berkeley Based on joint work with Prof.
CIRCUMCENTER OF MASS Emmanuel Tsukerman University of California, Berkeley Based on joint work with Prof. Sergei Tabachnikov Based on joint work with Prof. Sergei Tabachnikov Also A. Akopyan Based on joint work with Prof. Sergei Tabachnikov Also A. Akopyan Both papers in Discr. Comp. Geom., 2014. OUTLINE I. Motivation II. Definition III. Properties IV. Open problem I. MOTIVATION Motivation comes from integrable systems Motivation comes from integrable systems CCM is an integral of two interesting discrete conjecturally completely integrable systems 1. POLYGON RECUTTING 1. POLYGON RECUTTING 1. POLYGON RECUTTING V. Adler, Recuttings of polygons. Funct. Anal. Appl. 27 (1993), 141-143 V. Adler, Integrable deformations of a polygon. Phys. D 87 (1995), 52-57 2. DISCRETE DARBOUX (BICYCLE) TRANSFORMATION 2. DISCRETE DARBOUX (BICYCLE) TRANSFORMATION Bicycle transformation: imagine you have the rear track of a bike whose length you know 2. DISCRETE DARBOUX (BICYCLE) TRANSFORMATION Bicycle transformation: imagine you have the rear track of a bike whose length you know You can place the bike so that its back wheel matches the track and trace out the front track 2. DISCRETE DARBOUX (BICYCLE) TRANSFORMATION Bicycle transformation: imagine you have the rear track of a bike whose length you know You can place the bike so that its back wheel matches the track and trace out the front track Actually you can do this in two ways 2. DISCRETE DARBOUX (BICYCLE) TRANSFORMATION Bicycle transformation: imagine you have the rear track of a bike whose length you know You can place the bike so that its back wheel matches the track and trace out the front track Actually you can do this in two ways Related to Sherlock Holmes story 2. DISCRETE DARBOUX (BICYCLE) TRANSFORMATION 2. DISCRETE DARBOUX (BICYCLE) TRANSFORMATION V1W1V2W2 is an isosceles trapezoid 2. DISCRETE DARBOUX (BICYCLE) TRANSFORMATION V1W1V2W2 is an isosceles trapezoid 2. DISCRETE DARBOUX (BICYCLE) TRANSFORMATION V1W1V2W2 is an isosceles trapezoid U. Pinkall, B. Springborn, S. Weissmann, A new doubly discrete analogue of smoke ring flow and the real time simulation of fluid flow. J. Phys. A 40 (2007), 12563-12576 S. Tabachnikov, E. T., On the discrete bicycle transformation. Publ. Math. Uruguay, 14 (2013), 201-220. II. DEFINITION Construction: Construction: - π arbitrary point (off the sides) Construction: - π arbitrary point (off the sides) - Triangulate radially via π Construction: - π arbitrary point (off the sides) - Triangulate radially via π - πΆπ are the circumcenters with weights π΄π , the signed areas of triangles Construction: - π arbitrary point (off the sides) - Triangulate radially via π - πΆπ are the circumcenters with weights π΄π , the signed areas of triangles - πΆπΆπ is center of mass of the πΆπ Theorem: The circumcenter of mass π πͺπͺπ΄ π· = is well-defined. π=π π¨π πͺ π¨(π·) π Theorem: The circumcenter of mass π πͺπͺπ΄ π· = π=π is well-defined. π¨π πͺ π¨(π·) π Formula: If the vertices are (π₯π , π¦π ), then πβ1 1 πΆπΆπ π = ( 4π΄(π) πβ1 2 2 2 2 π¦π π₯πβ1 + π¦πβ1 β π₯π+1 β π¦π+1 , π=0 2 2 2 2 βπ₯π (π₯πβ1 + π¦πβ1 β π₯π+1 β π¦π+1 )) π=0 Theorem: The circumcenter of mass π πͺπͺπ΄ π· = π=π is well-defined. π¨π πͺ π¨(π·) π Formula: If the vertices are (π₯π , π¦π ), then πβ1 1 πΆπΆπ π = ( 4π΄(π) 2 2 2 2 π¦π π₯πβ1 + π¦πβ1 β π₯π+1 β π¦π+1 , π=0 πβ1 1 ( 6π΄(π) 2 2 2 2 βπ₯π (π₯πβ1 + π¦πβ1 β π₯π+1 β π¦π+1 )) π=0 Cf. the center of mass (of the lamina): πΆπ π = πβ1 πβ1 (π₯π +π₯π+1 )(π₯π π¦π+1 β π₯π+1 π¦π ), π=0 (π¦π + π¦π+1 )(π₯π π¦π+1 β π₯π+1 π¦π ) π=0 The πΆπΆπ can be defined also for simplicial polytopes and in spherical and hyperbolic geometries The πΆπΆπ can be defined also for simplicial polytopes and in spherical and hyperbolic geometries My current understanding is that the βproperβ setting for πΆπΆπ is geometric simplicial complexes III. PROPERTIES ARCHIMEDESβ LEMMA ARCHIMEDESβ LEMMA Classically: if an object is divided two smaller objects, then the center of mass of the compound object is the weighted sum of the centers of mass of the two smaller objects. ARCHIMEDESβ LEMMA Classically: if an object is divided two smaller objects, then the center of mass of the compound object is the weighted sum of the centers of mass of the two smaller objects. Applies also to πΆπΆπ ARCHIMEDESβ LEMMA Classically: if an object is divided two smaller objects, then the center of mass of the compound object is the weighted sum of the centers of mass of the two smaller objects. Applies also to πΆπΆπ (assuming two simplicial complexes) ARCHIMEDESβ LEMMA Classically: if an object is divided two smaller objects, then the center of mass of the compound object is the weighted sum of the centers of mass of the two smaller objects. Applies also to πΆπΆπ (assuming two simplicial complexes) valuation ARCHIMEDESβ LEMMA Classically: if an object is divided two smaller objects, then the center of mass of the compound object is the weighted sum of the centers of mass of the two smaller objects. Applies also to πΆπΆπ (assuming two simplicial complexes) valuation (isometry covariant) ARCHIMEDESβ LEMMA Classically: if an object is divided two smaller objects, then the center of mass of the compound object is the weighted sum of the centers of mass of the two smaller objects. Applies also to πΆπΆπ (assuming two simplicial complexes) valuation (isometry covariant) Consequence: πΆπΆπ can be computed via any nondegenerate triangulation ARCHIMEDESβ LEMMA Classically: if an object is divided two smaller objects, then the center of mass of the compound object is the weighted sum of the centers of mass of the two smaller objects. Applies also to πΆπΆπ (assuming two simplicial complexes) valuation (isometry covariant) Consequence: πΆπΆπ can be computed via any nondegenerate triangulation Consequence: πΆπΆπ of an inscribed polygon is its circumcenter I will now discuss subtleties involving triangulations βDANGEROUSβ TRIANGLES Flattening a triangle so its area goes to zero does not prevent the circumcenter from escaping to infinity βDANGEROUSβ TRIANGLES Flattening a triangle so its area goes to zero does not prevent the circumcenter from escaping to infinity βDANGEROUSβ TRIANGLES Flattening a triangle so its area goes to zero does not prevent the circumcenter from escaping to infinity The left is βdangerousβ, the right is βsafeβ PARADOX? Compute the CCM of a right triangle in two different ways: PARADOX? Compute the CCM of a right triangle in two different ways: counterexample? PARADOX? Compute the CCM of a right triangle in two different ways: counterexample? No, because on the left we have introduced a vertex. PARADOX? Compute the CCM of a right triangle in two different ways: counterexample? No, because on the left we have introduced a vertex. This gives us the πΆπΆπ of a degenerate quadrilateral: Another example where we must be careful Another example where we must be careful is when we try to compute the CCM of a nonsimplicial polytope Another example where we must be careful is when we try to compute the CCM of a nonsimplicial polytope Another example where we must be careful is when we try to compute the CCM of a nonsimplicial polytope Different triangulations will yield different βCCMβs Another example where we must be careful is when we try to compute the CCM of a nonsimplicial polytope Different triangulations will yield different βCCMβs Back to properties of CCM Proposition: If π is an equilateral polygon, then πͺπͺπ΄ π· = πͺπ΄ π· . EULER LINE EULER LINE It is a classical fact that the circumcenter, centroid, orthocenter, and many other notable points of a triangle are collinear EULER LINE It is a classical fact that the circumcenter, centroid, orthocenter, and many other notable points of a triangle are collinear This line is called the Euler line EULER LINE It is a classical fact that the circumcenter, centroid, orthocenter, and many other notable points of a triangle are collinear This line is called the Euler line We can now extend the notion to polygons (more generally, simplicial polytopes) by taking affine combinations of πΆπΆπ and πΆπ: πΆπ‘ = π‘πΆπ + 1 β π‘ πΆπΆπ EULER LINE It is a classical fact that the circumcenter, centroid, orthocenter, and many other notable points of a triangle are collinear This line is called the Euler line We can now extend the notion to polygons (more generally, simplicial polytopes) by taking affine combinations of πΆπΆπ and πΆπ: πΆπ‘ = π‘πΆπ + 1 β π‘ πΆπΆπ (orthocenter: π‘ = 3) EULER LINE - EXAMPLE EULER LINE - EXAMPLE For a simplex, it is the line through the centroid and circumcenter EULER LINE - EXAMPLE For a simplex, it is the line through the centroid and circumcenter This line passes through the Monge point EULER LINE - EXAMPLE For a simplex, it is the line through the centroid and circumcenter This line passes through the Monge point (the six planes through the midpoints of the edges of a tetrahedron and perpendicular to the opposite edges concur in a point known as the Monge point β there is also a higher dimensional analog) EULER LINE - EXAMPLE For a simplex, it is the line through the centroid and circumcenter This line passes through the Monge point (the six planes through the midpoints of the edges of a tetrahedron and perpendicular to the opposite edges concur in a point known as the Monge point β there is also a higher dimensional analog) The Monge point is πΆπ+1 and can now be defined for simplicial polytopes πβ1 EULER LINE - PROPERTIES EULER LINE - PROPERTIES Theorem: Suppose that a center: - depends analytically on the polygon - commutes with dilations - satisfies Archimedesβ lemma Then it is an affine combination of πΆπ and πΆπΆπ EULER LINE - PROPERTIES Theorem: Suppose that a center: - depends analytically on the polygon - commutes with dilations - satisfies Archimedesβ lemma Then it is an affine combination of πΆπ and πΆπΆπ Proposition: assume that the π-gon is almost equilateral: the sides satisfy π1 = π2 = β― = ππβ1 . Then the Euler line is orthogonal to side ππ . Proof: reflect in side ππ to obtain an equilateral polygon and use the fact that πΆπΆπ = πΆπ for equilateral polygons EULER LINE β FURTHER PROPERTIES If π has a line of reflection symmetry πΏ, then πΈ is either πΏ or a point on πΏ EULER LINE β FURTHER PROPERTIES If π has a line of reflection symmetry πΏ, then πΈ is either πΏ or a point on πΏ If π has a center of rotational symmetry πΆ and none of the (extended) sides passes through πΆ, then πΈ = πΆ CONTINUOUS LIMIT Let πΎ be a curve enclosing a star-shaped domain. CONTINUOUS LIMIT Let πΎ be a curve enclosing a star-shaped domain. Define πΆπΆπ πΎ = πΆ π‘ ππ΄ ππ΄ CONTINUOUS LIMIT Let πΎ be a curve enclosing a star-shaped domain. Define πΆπΆπ πΎ = Then πΆπΆπ πΎ = πΆπ πΎ . πΆ π‘ ππ΄ ππ΄ CONTINUOUS LIMIT Let πΎ be a curve enclosing a star-shaped domain. Define πΆπΆπ πΎ = Then πΆπΆπ πΎ = πΆπ πΎ . In general, should be an intrinsic moment. πΆ π‘ ππ΄ ππ΄ RELATION TO OTHER VALUATIONS RELATION TO OTHER VALUATIONS βπ -valued continuous (wrt Hausdorff metric) isometry-covariant valuations on convex compact sets are classified as linear combinations of intrinsic moments RELATION TO OTHER VALUATIONS βπ -valued continuous (wrt Hausdorff metric) isometry-covariant valuations on convex compact sets are classified as linear combinations of intrinsic moments Namely, given convex πΎ, let πΎπ be its π-neighborhood. RELATION TO OTHER VALUATIONS βπ -valued continuous (wrt Hausdorff metric) isometry-covariant valuations on convex compact sets are classified as linear combinations of intrinsic moments Namely, given convex πΎ, let πΎπ be its π-neighborhood. The moment vector π₯ ππ₯ πΎπ Is a polynomial in π and its coefficients span the space of continuous isometry-covariant valuations. RELATION TO OTHER VALUATIONS βπ -valued continuous (wrt Hausdorff metric) isometry-covariant valuations on convex compact sets are classified as linear combinations of intrinsic moments Namely, given convex πΎ, let πΎπ be its π-neighborhood. The moment vector π₯ ππ₯ πΎπ Is a polynomial in π and its coefficients span the space of continuous isometry-covariant valuations. The map π: π βΌ πππ π πΆπΆπ π is not an element of this space RELATION TO OTHER VALUATIONS βπ -valued continuous (wrt Hausdorff metric) isometry-covariant valuations on convex compact sets are classified as linear combinations of intrinsic moments Namely, given convex πΎ, let πΎπ be its π-neighborhood. The moment vector π₯ ππ₯ πΎπ Is a polynomial in π and its coefficients span the space of continuous isometry-covariant valuations. The map π: π βΌ πππ π πΆπΆπ π is not an element of this space Proof: (in dim 2, general argument similar) RELATION TO OTHER VALUATIONS βπ -valued continuous (wrt Hausdorff metric) isometry-covariant valuations on convex compact sets are classified as linear combinations of intrinsic moments Namely, given convex πΎ, let πΎπ be its π-neighborhood. The moment vector π₯ ππ₯ πΎπ Is a polynomial in π and its coefficients span the space of continuous isometry-covariant valuations. The map π: π βΌ πππ π πΆπΆπ π is not an element of this space Proof: (in dim 2, general argument similar) Consider a βdangerousβ isosceles triangle whose base is of length 2, aligned with the x-axis, and which is symmetric about the y-axis. RELATION TO OTHER VALUATIONS βπ -valued continuous (wrt Hausdorff metric) isometry-covariant valuations on convex compact sets are classified as linear combinations of intrinsic moments Namely, given convex πΎ, let πΎπ be its π-neighborhood. The moment vector π₯ ππ₯ πΎπ Is a polynomial in π and its coefficients span the space of continuous isometry-covariant valuations. The map π: π βΌ πππ π πΆπΆπ π is not an element of this space Proof: (in dim 2, general argument similar) Consider a βdangerousβ isosceles triangle whose base is of length 2, aligned with the x-axis, and which is symmetric about the y-axis. Flattening gives a 0-moment vector. RELATION TO OTHER VALUATIONS βπ -valued continuous (wrt Hausdorff metric) isometry-covariant valuations on convex compact sets are classified as linear combinations of intrinsic moments Namely, given convex πΎ, let πΎπ be its π-neighborhood. The moment vector π₯ ππ₯ πΎπ Is a polynomial in π and its coefficients span the space of continuous isometry-covariant valuations. The map π: π βΌ πππ π πΆπΆπ π is not an element of this space Proof: (in dim 2, general argument similar) Consider a βdangerousβ isosceles triangle whose base is of length 2, aligned with the x-axis, and which is symmetric about the y-axis. Flattening gives a 0-moment vector. However, a calculation shows that π=(0,-1/2). RELATION TO OTHER VALUATIONS βπ -valued continuous (wrt Hausdorff metric) isometry-covariant valuations on convex compact sets are classified as linear combinations of intrinsic moments Namely, given convex πΎ, let πΎπ be its π-neighborhood. The moment vector π₯ ππ₯ πΎπ Is a polynomial in π and its coefficients span the space of continuous isometry-covariant valuations. The map π: π βΌ πππ π πΆπΆπ π is not an element of this space Proof: (in dim 2, general argument similar) Consider a βdangerousβ isosceles triangle whose base is of length 2, aligned with the x-axis, and which is symmetric about the y-axis. Flattening gives a 0-moment vector. However, a calculation shows that π=(0,-1/2). We can also see that it has a contribution from the βparadoxβ HISTORICAL NOTE Once we had understood CCM well-enough, we started asking other researchers if they had seen this object (CCM) before HISTORICAL NOTE Once we had understood CCM well-enough, we started asking other researchers if they had seen this object (CCM) before β’It seems that Giusto Bellavitis was the first who noted the existence of the circumcenter of mass of a planar polygon in 1834 HISTORICAL NOTE Once we had understood CCM well-enough, we started asking other researchers if they had seen this object (CCM) before β’It seems that Giusto Bellavitis was the first who noted the existence of the circumcenter of mass of a planar polygon in 1834 β’In 1993, it was independently noticed by Adler for the case of triangulation of planar polygon by diagonals HISTORICAL NOTE Once we had understood CCM well-enough, we started asking other researchers if they had seen this object (CCM) before β’It seems that Giusto Bellavitis was the first who noted the existence of the circumcenter of mass of a planar polygon in 1834 β’In 1993, it was independently noticed by Adler for the case of triangulation of planar polygon by diagonals β’in the private correspondence of G.C. Shephard and B. Grünbaum. They also noted that the circumcenter could be replaced by any point on the Euler line, that is, by a fixed affine combination of the centroid and the Circumcenter. HISTORICAL NOTE Once we had understood CCM well-enough, we started asking other researchers if they had seen this object (CCM) before β’It seems that Giusto Bellavitis was the first who noted the existence of the circumcenter of mass of a planar polygon in 1834 β’In 1993, it was independently noticed by Adler for the case of triangulation of planar polygon by diagonals β’in the private correspondence of G.C. Shephard and B. Grünbaum. They also noted that the circumcenter could be replaced by any point on the Euler line, that is, by a fixed affine combination of the centroid and the Circumcenter. β’Myakishev proved the existence of Euler line for a quadrilateral. QUESTION Assign to every nondegenerate simplex β a center πΆ(β) so that QUESTION Assign to every nondegenerate simplex β a center πΆ(β) so that - The map ββΌ πΆ β commutes with similarities QUESTION Assign to every nondegenerate simplex β a center πΆ(β) so that - The map ββΌ πΆ β commutes with similarities - The map ββΌ πΆ(β) is invariant under permutations of the vertices of β QUESTION Assign to every nondegenerate simplex β a center πΆ(β) so that - The map ββΌ πΆ β commutes with similarities - The map ββΌ πΆ(β) is invariant under permutations of the vertices of β - The map ββΌ πππ(β)πΆ(β) is a polynomial in the coordinates of the vertices of the simplex β QUESTION Assign to every nondegenerate simplex β a center πΆ(β) so that - The map ββΌ πΆ β commutes with similarities - The map ββΌ πΆ(β) is invariant under permutations of the vertices of β - The map ββΌ πππ(β)πΆ(β) is a polynomial in the coordinates of the vertices of the simplex β Theorem: πΆ(β) is an affine combination of πΆπ and πΆπΆπ QUESTION Assign to every nondegenerate simplex β a center πΆ(β) so that - The map ββΌ πΆ β commutes with similarities - The map ββΌ πΆ(β) is invariant under permutations of the vertices of β - The map ββΌ πππ(β)πΆ(β) is a polynomial in the coordinates of the vertices of the simplex β Theorem: πΆ(β) is an affine combination of πΆπ and πΆπΆπ Question: what are the right axioms for πΆπΆπ? Summary Summary β’ CCM is a cousin of the CM Summary β’ β’ CCM is a cousin of the CM Arises in discrete integrable systems Summary β’ β’ β’ CCM is a cousin of the CM Arises in discrete integrable systems Satisfies Archimedesβ Lemma Summary β’ β’ β’ β’ CCM is a cousin of the CM Arises in discrete integrable systems Satisfies Archimedesβ Lemma Allows to define an Euler line Summary β’ β’ β’ β’ β’ CCM is a cousin of the CM Arises in discrete integrable systems Satisfies Archimedesβ Lemma Allows to define an Euler line Have several results showing Euler line centers are the only centers satisfying Archimedesβ Lemma Summary β’ β’ β’ β’ β’ β’ CCM is a cousin of the CM Arises in discrete integrable systems Satisfies Archimedesβ Lemma Allows to define an Euler line Have several results showing Euler line centers are the only centers satisfying Archimedesβ Lemma What are the right axioms? Thank you Questions?