70
Delaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work with Mathieu Dutour Sikiric and Frank Vallentin ) Berlin, October 2013 ERC Workshop Delaunay Geometry Polytopes, Triangulations and Spheres

Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Delaunay Tessellations of Point Lattices Theory, Algorithms and Applications

Achill Schürmann(University of Rostock)

( based on work with Mathieu Dutour Sikiric and Frank Vallentin )

Berlin, October 2013

ERC Workshop Delaunay Geometry

Polytopes, Triangulations and Spheres

Page 2: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Делоне

Boris N. Delone1890-1980

( Delone )French: Delaunay

Page 3: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Delone tessellations

Page 4: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Delaunay subdivisions of lattices

L = hexagonal lattice 2-periodic (m=2)

Page 5: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Delone star and DV-cell

( ) =�

∈ R : � � ≤ � − � ∈�

Up to translation, there is only on DV-cell in a lattice:

Page 6: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Delone star and DV-cell

( ) =�

∈ R : � � ≤ � − � ∈�

Up to translation, there is only on DV-cell in a lattice:

all Delone polyhedra incident to a given vertex

It is dual to a Delone star

Page 7: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

How do I compute

the DV-cell(or Delone star)?

Page 8: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computing DV-cells(first approach)

THM (Voronoi, 1908):

∈ ( )± �= +

Page 9: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computing DV-cells(first approach)

THM (Voronoi, 1908):

⇒ ( ) ( − )

∈ ( )± �= +

Page 10: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computing DV-cells(first approach)

THM (Voronoi, 1908):

⇒ ( ) ( − )

0

∈ ( )± �= +

Page 11: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computing DV-cells(first approach)

THM (Voronoi, 1908):

⇒ ( ) ( − )

0

PLAN

• compute facets

• obtain vertices

∈ ( )± �= +

Page 12: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computing DV-cells

STEP 1: Compute an initial vertex (an initial Delone Polyhedron)

(practical approach)

Page 13: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computing DV-cells

STEP 1: Compute an initial vertex (an initial Delone Polyhedron)

(practical approach)

Page 14: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computing DV-cells

STEP 1: Compute an initial vertex (an initial Delone Polyhedron)

(practical approach)

Page 15: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computing DV-cells (continued...)

STEP x: Compute neighboring vertices (neighboring Delone Polyhedra)

• Compute facets of Delone polyhedron • For each facet find neighboring Delone polyhedron• Test if they are equivalent to one of the known ones

Page 16: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computing DV-cells (continued...)

STEP x: Compute neighboring vertices (neighboring Delone Polyhedra)

• Compute facets of Delone polyhedron • For each facet find neighboring Delone polyhedron• Test if they are equivalent to one of the known ones

Page 17: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computing DV-cells (continued...)

STEP x: Compute neighboring vertices (neighboring Delone Polyhedra)

• Compute facets of Delone polyhedron • For each facet find neighboring Delone polyhedron• Test if they are equivalent to one of the known ones

( up to translation, central inversion, ...)

Page 18: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computing DV-cells (continued...)

STEP x: Compute neighboring vertices (neighboring Delone Polyhedra)

• Compute facets of Delone polyhedron • For each facet find neighboring Delone polyhedron• Test if they are equivalent to one of the known ones

( up to translation, central inversion, ...)

Well suited for exploiting symmetry!

Page 19: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computational Results

obtained by Mathieu using polyhedral

Mathieu

Page 20: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Computational Results

obtained by Mathieu using polyhedral

Mathieu

IN: Complexity and algorithms for computing Voronoi cells of lattices, Math. Comp. (2009)

• computation of vertices for many different DV-cells of lattices (in particular for Coxeter-, Laminated and Cut-Lattices) • verified that Leech Lattice cell has 307 vertex orbits (Conway, Borcherds, et. al.)

Page 21: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Application I: Covering Constants

Page 22: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Application I: Covering Constants

µ( ) = sup∈ ( )

( )

Page 23: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Application II: Quantizer Constants

Page 24: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

What happensif we vary the lattice?

Page 25: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Two views

Instead of varying the lattice we can equivalently vary the norm

Z � � =√

=

= Z � � =√

Page 26: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Quadratic Forms

Page 27: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Arithmetical Equivalence

Page 28: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Dictionary

Page 29: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Dictionary

Page 30: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Delone tessellations revisited

Page 31: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Delone tessellations revisited

Page 32: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Secondary cones

D R Z

∆(D) =�

∈ S> : ( ) = D�

Page 33: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Secondary cones

D R Z

∆(D) =�

∈ S> : ( ) = D�

Page 34: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Baranovskii Cones

Z

∆( ) =�

∈ S> : ∈ ( )�

DEF:

Page 35: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Baranovskii Cones

Z

∆( ) =�

∈ S> : ∈ ( )�

DEF:

THM:∆( )

Page 36: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Baranovskii Cones

Z

∆( ) =�

∈ S> : ∈ ( )�

DEF:

THM:∆( )

D ∆(D) =�

∈D∆(P)

Note:

Page 37: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Baranovskii Cones

Z

∆( ) =�

∈ S> : ∈ ( )�

DEF:

THM:∆( )

D ∆(D) =�

∈D∆(P)

Note:

⇒ ∆(D)

Page 38: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Application:Finding best lattice coverings

Page 39: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Application:Finding best lattice coverings

�∈ S> : ( ( )) ≤

THM (Barnes, Dickson; 1968): Among PQFs in the closure of a secondary cone

Θ

Page 40: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Application:Finding best lattice coverings

�∈ S> : ( ( )) ≤

THM (Barnes, Dickson; 1968): Among PQFs in the closure of a secondary cone

Θ

Note:

Page 41: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Voronoi’s second reduction

Page 42: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Voronoi’s second reduction

THM (Voronoi, 1908):

• there exist finitely many inequivalent secondary cones

• inclusion of faces corresponds to coarsening of subdivisions

• closures of secondary cones tesselate S>

Page 43: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Voronoi’s second reduction

THM (Voronoi, 1908):

• there exist finitely many inequivalent secondary cones

• inclusion of faces corresponds to coarsening of subdivisions

• closures of secondary cones tesselate S>

=> top-dimensional cones come from triangulations

Page 44: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Voronoi’s second reduction

THM (Voronoi, 1908):

• there exist finitely many inequivalent secondary cones

• inclusion of faces corresponds to coarsening of subdivisions

• closures of secondary cones tesselate S>

=> top-dimensional cones come from triangulations

Page 45: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Already known...

Page 46: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Already known...

IDEA: In higher dimensions, determine the best lattice coverings with a given group of symmetries!? (obtaining all Delone subdivisons with a given symmetry)

Page 47: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

G-Theory?

Page 48: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

G-Theory?

Page 49: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

G-Theory?

IDEA: Intersect secondary cones with a linear subspace T

Page 50: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

G-Theory?

IDEA: Intersect secondary cones with a linear subspace T

DEF: ∩∆(D)

Page 51: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

T-secondary cones• T-secondary cones tesselate S> ∩

Page 52: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

T-secondary cones• T-secondary cones tesselate S> ∩

:⇔ ∃ ∈ (Z) = � ⊆DEF:

Page 53: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

T-secondary cones• T-secondary cones tesselate S> ∩

:⇔ ∃ ∈ (Z) = � ⊆DEF:

THM: ⊂ (Z)

Page 54: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

T-secondary cones• T-secondary cones tesselate S> ∩

:⇔ ∃ ∈ (Z) = � ⊆DEF:

• Delone subdivision of a neighboring T-secondary cone can be obtained by a T-flip in repartitioning polytopes

THM: ⊂ (Z)

Page 55: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

T-secondary cones• T-secondary cones tesselate S> ∩

:⇔ ∃ ∈ (Z) = � ⊆DEF:

• Delone subdivision of a neighboring T-secondary cone can be obtained by a T-flip in repartitioning polytopes

THM: ⊂ (Z)

Page 56: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Application to Lattice Coverings

Page 57: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

What about the nice lattices?

Page 58: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

What about the nice lattices?

Prominent Example:

(E )

Θ

Page 59: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

What about the nice lattices?

Question: Are there local maxima??

Prominent Example:

(E )

Θ

Page 60: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

What about the nice lattices?

Question: Are there local maxima??

YES! E Θ

Prominent Example:

(E )

Θ

Page 61: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

What about the nice lattices?

Question: Are there local maxima??

YES! E Θ= , . . . ,

Prominent Example:

(E )

Θ

Page 62: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

What about the nice lattices?

Question: Are there local maxima??

THM: A necessary condition for a local maximum is thatevery Delone polytope attaining the covering radius

is an extreme Delone polytopedim∆( ) =

YES! E Θ= , . . . ,

Prominent Example:

(E )

Θ

Page 63: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

i-eutaxy and i-perfectness

DEF: Q is i-perfect if

DEF: Q is i-eutactic if

Page 64: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

i-eutaxy and i-perfectness

DEF: Q is i-perfect if

DEF: Q is i-eutactic if

THM: Θ⇔

Page 65: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Bahavior of nice lattices

lattice covering densityZ global minimumA2 global minimumD4 almost local maximumE6 local maximumE7 local maximumE8 almost local maximumK12 almost local maximumBW16 local maximumΛ24 local minimum

Page 66: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Application: Minkowski Conjecture

Conjecture:

( ) = | · · · | ⊂ R det =

sup∈R

inf∈

( − ) ≤ −

= ( , . . . , )Z ( ) =

Page 67: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Covering Conjecture

Page 68: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

Covering Conjecture

Local covering maxima among well rounded lattices are attained by T-extreme Delone Polyhedra and there are only

finitely many of them in every dimension.(with T = space of well rounded lattices)

THM:

Page 69: Delaunay Tessellations of Point Lattices fileDelaunay Tessellations of Point Lattices Theory, Algorithms and Applications Achill Schürmann (University of Rostock) ( based on work

References

• Complexity and algorithms for computing Voronoi cells of lattices, Math. Comp. (2009)

• computation of covering radius and Delone subdivisions for many lattices• verified that Leech lattice cell has 307 vertex orbits (Conway, Borcherds, et. al.)

• A generalization of Voronoi’s reduction theory and its application, Duke Math. J. (2008)

• Inhomogeneous extreme forms, Annales de l'institut Fourier (2012)

• generalized Voronoi’s reduction for L-type domains to a G- and T-invariant setting• obtained new best known covering lattices and classified totally real thin number fields

• characterization of locally extreme forms for the sphere covering problem