Ehrhart Positivity Federico Castillo University of California, - - PowerPoint PPT Presentation

▶
ehrhart positivity
SMART_READER_LITE
LIVE PREVIEW

Ehrhart Positivity Federico Castillo University of California, - - PowerPoint PPT Presentation

Feb 15, 2024 114 likes •1.13k views

Ehrhart Positivity Federico Castillo University of California, Davis Joint work with Fu Liu December 15, 2016 Federico Castillo UC Davis Ehrhart Positivity Lattice points of a polytope A (convex) polytope is a bounded solution set of a

slide-1
SLIDE 1

Ehrhart Positivity

Federico Castillo

University of California, Davis Joint work with Fu Liu

December 15, 2016

Federico Castillo UC Davis Ehrhart Positivity

slide-2
SLIDE 2

Lattice points of a polytope

A (convex) polytope is a bounded solution set of a finite system of linear inequalities, or is the convex hull of a finite set of points.

Federico Castillo UC Davis Ehrhart Positivity

slide-3
SLIDE 3

Lattice points of a polytope

A (convex) polytope is a bounded solution set of a finite system of linear inequalities, or is the convex hull of a finite set of points. An integral polytope is a polytope whose vertices are all lattice points. i.e., points with integer coordinates.

Federico Castillo UC Davis Ehrhart Positivity

slide-4
SLIDE 4

Lattice points of a polytope

A (convex) polytope is a bounded solution set of a finite system of linear inequalities, or is the convex hull of a finite set of points. An integral polytope is a polytope whose vertices are all lattice points. i.e., points with integer coordinates. Definition For any polytope P ⊂ Rd and positive integer m ∈ N, the mth dilation

  • f P is mP = {mx : x ∈ P}. We define

i(P, m) = |mP ∩ Zd| to be the number of lattice points in the mP.

Federico Castillo UC Davis Ehrhart Positivity

slide-5
SLIDE 5

Example

P 3P

Federico Castillo UC Davis Ehrhart Positivity

slide-6
SLIDE 6

Example

P 3P

In this example we can see that i(P, m) = (m + 1)2

Federico Castillo UC Davis Ehrhart Positivity

slide-7
SLIDE 7

Theorem of Ehrhart (on integral polytopes)

Figure: Eugene Ehrhart.

Federico Castillo UC Davis Ehrhart Positivity

slide-8
SLIDE 8

Theorem of Ehrhart (on integral polytopes)

Figure: Eugene Ehrhart.

Theorem[Ehrhart] Let P be a d-dimensional integral polytope. Then i(P, m) is a polynomial in m of degree d.

Federico Castillo UC Davis Ehrhart Positivity

slide-9
SLIDE 9

Theorem of Ehrhart (on integral polytopes)

Figure: Eugene Ehrhart.

Theorem[Ehrhart] Let P be a d-dimensional integral polytope. Then i(P, m) is a polynomial in m of degree d.

Federico Castillo UC Davis Ehrhart Positivity

slide-10
SLIDE 10

The h∗ or δ vector.

Therefore, we call i(P, m) the Ehrhart polynomial of P.

Federico Castillo UC Davis Ehrhart Positivity

slide-11
SLIDE 11

The h∗ or δ vector.

Therefore, we call i(P, m) the Ehrhart polynomial of P.We study its coefficients.

Federico Castillo UC Davis Ehrhart Positivity

slide-12
SLIDE 12

The h∗ or δ vector.

Therefore, we call i(P, m) the Ehrhart polynomial of P.We study its

  • coefficients. ... however, there is another popular point of view.

Federico Castillo UC Davis Ehrhart Positivity

slide-13
SLIDE 13

The h∗ or δ vector.

Therefore, we call i(P, m) the Ehrhart polynomial of P.We study its

  • coefficients. ... however, there is another popular point of view.

The fact that i(P, m) is a polynomial with integer values at integer points suggests other forms of expanding it. An alternative basis We can write: i(P, m) = h∗

0(P)

m + d d

  • + h∗

1(P)

m + d − 1 d

  • + · · · + h∗

d(P)

m d

  • .

Federico Castillo UC Davis Ehrhart Positivity

slide-14
SLIDE 14

More on the the h∗ or δ vector.

The vector (h∗

0, h∗ 1, · · · , h∗ d) has many good properties.

Theorem(Stanley) For any lattice polytope P, h∗

i (P) is nonnegative integer.

Federico Castillo UC Davis Ehrhart Positivity

slide-15
SLIDE 15

More on the the h∗ or δ vector.

The vector (h∗

0, h∗ 1, · · · , h∗ d) has many good properties.

Theorem(Stanley) For any lattice polytope P, h∗

i (P) is nonnegative integer.

Additionally it has an algebraic meaning.

Federico Castillo UC Davis Ehrhart Positivity

slide-16
SLIDE 16

Back to coefficients of Ehrhart polynomials

Federico Castillo UC Davis Ehrhart Positivity

slide-17
SLIDE 17

Back to coefficients of Ehrhart polynomials

What is known?

Federico Castillo UC Davis Ehrhart Positivity

slide-18
SLIDE 18

Back to coefficients of Ehrhart polynomials

What is known?

1 The leading coefficient of i(P, m) is the volume vol(P) of P.

Federico Castillo UC Davis Ehrhart Positivity

slide-19
SLIDE 19

Back to coefficients of Ehrhart polynomials

What is known?

1 The leading coefficient of i(P, m) is the volume vol(P) of P. 2 The second coefficient equals 1/2 of the sum of the normalized

volumes of each facet.

Federico Castillo UC Davis Ehrhart Positivity

slide-20
SLIDE 20

Back to coefficients of Ehrhart polynomials

What is known?

1 The leading coefficient of i(P, m) is the volume vol(P) of P. 2 The second coefficient equals 1/2 of the sum of the normalized

volumes of each facet.

3 The constant term of i(P, m) is always 1.

Federico Castillo UC Davis Ehrhart Positivity

slide-21
SLIDE 21

Back to coefficients of Ehrhart polynomials

What is known?

1 The leading coefficient of i(P, m) is the volume vol(P) of P. 2 The second coefficient equals 1/2 of the sum of the normalized

volumes of each facet.

3 The constant term of i(P, m) is always 1.

No simple forms known for other coefficients for general polytopes.

Federico Castillo UC Davis Ehrhart Positivity

slide-22
SLIDE 22

Back to coefficients of Ehrhart polynomials

What is known?

1 The leading coefficient of i(P, m) is the volume vol(P) of P. 2 The second coefficient equals 1/2 of the sum of the normalized

volumes of each facet.

3 The constant term of i(P, m) is always 1.

No simple forms known for other coefficients for general polytopes. Warning It is NOT even true that all the coefficients are positive. For example, for the polytope P with vertices (0, 0, 0), (1, 0, 0), (0, 1, 0) and (1, 1, 13), its Ehrhart polynomial is i(P, n) = 13 6 n3 + n2−1 6n + 1.

Federico Castillo UC Davis Ehrhart Positivity

slide-23
SLIDE 23

General philosophy.

They are related to volumes.

Federico Castillo UC Davis Ehrhart Positivity

slide-24
SLIDE 24

Ehrhart Positivity

Main Definition. We say an integral polytope is Ehrhart positive (or just positive for this talk) if it has positive coefficients in its Ehrhart polynomial.

Federico Castillo UC Davis Ehrhart Positivity

slide-25
SLIDE 25

Ehrhart Positivity

Main Definition. We say an integral polytope is Ehrhart positive (or just positive for this talk) if it has positive coefficients in its Ehrhart polynomial. In the literature, different techniques have been used to proved positivity.

Federico Castillo UC Davis Ehrhart Positivity

slide-26
SLIDE 26

Example I

Polytope: Standard simplex.

Federico Castillo UC Davis Ehrhart Positivity

slide-27
SLIDE 27

Example I

Polytope: Standard simplex. Reason: Explicit verification.

Federico Castillo UC Davis Ehrhart Positivity

slide-28
SLIDE 28

Standard simplex.

In the case of ∆d = {x ∈ Rd+1 : x1 + x2 + · · · + xd+1 = 1, xi ≥ 0},

Federico Castillo UC Davis Ehrhart Positivity

slide-29
SLIDE 29

Standard simplex.

In the case of ∆d = {x ∈ Rd+1 : x1 + x2 + · · · + xd+1 = 1, xi ≥ 0}, It can be computed that its Ehrhart polynomial is m + d d

  • .

(Notice how simple this h∗ vector is).

Federico Castillo UC Davis Ehrhart Positivity

slide-30
SLIDE 30

Standard simplex.

In the case of ∆d = {x ∈ Rd+1 : x1 + x2 + · · · + xd+1 = 1, xi ≥ 0}, It can be computed that its Ehrhart polynomial is m + d d

  • .

(Notice how simple this h∗ vector is). More explicitly we have m + d d

  • = (m + d)(m + d − 1) · · · (m + 1)

d!

Federico Castillo UC Davis Ehrhart Positivity

slide-31
SLIDE 31

Standard simplex.

In the case of ∆d = {x ∈ Rd+1 : x1 + x2 + · · · + xd+1 = 1, xi ≥ 0}, It can be computed that its Ehrhart polynomial is m + d d

  • .

(Notice how simple this h∗ vector is). More explicitly we have m + d d

  • = (m + d)(m + d − 1) · · · (m + 1)

d! which expands positively in powers of m.

Federico Castillo UC Davis Ehrhart Positivity

slide-32
SLIDE 32

Hypersimplices.

In the case of ∆d+1,k = conv{x ∈ {0, 1}d+1 : x1 + x2 + · · · + xd+1 = k},

Federico Castillo UC Davis Ehrhart Positivity

slide-33
SLIDE 33

Hypersimplices.

In the case of ∆d+1,k = conv{x ∈ {0, 1}d+1 : x1 + x2 + · · · + xd+1 = k}, it can be computed that its Ehrhart polynomial is

Federico Castillo UC Davis Ehrhart Positivity

slide-34
SLIDE 34

Hypersimplices.

In the case of ∆d+1,k = conv{x ∈ {0, 1}d+1 : x1 + x2 + · · · + xd+1 = k}, it can be computed that its Ehrhart polynomial is

d+1

  • i=0

d + 1 i d + 1 + mk − (m + 1)i − 1 d

  • (−1)i

Federico Castillo UC Davis Ehrhart Positivity

slide-35
SLIDE 35

Hypersimplices.

In the case of ∆d+1,k = conv{x ∈ {0, 1}d+1 : x1 + x2 + · · · + xd+1 = k}, it can be computed that its Ehrhart polynomial is

d+1

  • i=0

d + 1 i d + 1 + mk − (m + 1)i − 1 d

  • (−1)i

Not clear if the coefficients are positive.

Federico Castillo UC Davis Ehrhart Positivity

slide-36
SLIDE 36

Example II

Polytope: Crosspolytope

Federico Castillo UC Davis Ehrhart Positivity

slide-37
SLIDE 37

Example II

Polytope: Crosspolytope Reason: Roots have negative real part.

Federico Castillo UC Davis Ehrhart Positivity

slide-38
SLIDE 38

Crosspolytope.

In the case of the crosspolytope: ♦d = conv{±ei : 1 ≤ i ≤ d}, It can be computed that its Ehrhart polynomial is

d

  • k=0

2k d k m k

  • ,

which is not clear if it expands positively in powers of m.

Federico Castillo UC Davis Ehrhart Positivity

slide-39
SLIDE 39

Crosspolytope.

However

Federico Castillo UC Davis Ehrhart Positivity

slide-40
SLIDE 40

Crosspolytope.

However, according to EC1, Exercise 4.61(b), every zero of the Ehrhart polynomial has real part −1/2. Thus it is a product of factors (n + 1/2)

  • r

(n + 1/2 + ia)(n + 1/2 − ia) = n2 + n + 1/4 + a2, where a is real, so positivity follows.

Federico Castillo UC Davis Ehrhart Positivity

slide-41
SLIDE 41

Crosspolytope.

However, according to EC1, Exercise 4.61(b), every zero of the Ehrhart polynomial has real part −1/2. Thus it is a product of factors (n + 1/2)

  • r

(n + 1/2 + ia)(n + 1/2 − ia) = n2 + n + 1/4 + a2, where a is real, so positivity follows. What are the roots about? This opens more questions.

Federico Castillo UC Davis Ehrhart Positivity

slide-42
SLIDE 42

Birkhoff Poytope

The following is the graph (Beck-DeLoera-Pfeifle-Stanley) of zeros for the Birkhoff polytope of 8 × 8 doubly stochastic matrices.

Federico Castillo UC Davis Ehrhart Positivity

slide-43
SLIDE 43

Example III

Polytope: Zonotopes.

Federico Castillo UC Davis Ehrhart Positivity

slide-44
SLIDE 44

Example III

Polytope: Zonotopes. Reason: Formula for them.

Federico Castillo UC Davis Ehrhart Positivity

slide-45
SLIDE 45

Example III

Polytope: Zonotopes. Reason: Formula for them.

One of the few examples in which the formula is explicit on the coefficients.

Federico Castillo UC Davis Ehrhart Positivity

slide-46
SLIDE 46

Zonotopes.

Definition The Minkowski sum of vectors Z(v1, · · · , vk) = v1 + v2 + · · · + vk. The Ehrhart polynomial i(Z(v1, · · · , vk), m) = admd + ad−1md−1 + · · · a0m0, has a coefficient by coefficient interpretation.

Federico Castillo UC Davis Ehrhart Positivity

slide-47
SLIDE 47

Zonotopes.

Theorem(Stanley) In the above expression, ai is equal to (absolute value of) the greatest common divisor (g.c.d.) of all i × i minors of the matrix M =   · · · v1 v2 · · · vk · · ·  

Federico Castillo UC Davis Ehrhart Positivity

slide-48
SLIDE 48

Zonotopes.

This includes the unit cube [0, 1]d which has Ehrhart polynomial i(d, m) = (m + 1)d.

Federico Castillo UC Davis Ehrhart Positivity

slide-49
SLIDE 49

Zonotopes.

This includes the unit cube [0, 1]d which has Ehrhart polynomial i(d, m) = (m + 1)d. And also the regular permutohedron Πn =

  • 1≤i<j≤n+1

[ei, ej], = conv{

  • σ(1), σ(2), · · · , σ(n + 1)
  • ∈ Rn+1 : σ ∈ Sn+1}.

Federico Castillo UC Davis Ehrhart Positivity

slide-50
SLIDE 50

Permutohedron.

Figure: A permutohedron in dimension 3.

The Ehrhart polynomial is 1 + 6m + 15m2 + 16m3.

Federico Castillo UC Davis Ehrhart Positivity

slide-51
SLIDE 51

Example IV

Polytope: Cyclic polytopes.

Federico Castillo UC Davis Ehrhart Positivity

slide-52
SLIDE 52

Example IV

Polytope: Cyclic polytopes. Reason: Higher integrality conditions.

Federico Castillo UC Davis Ehrhart Positivity

slide-53
SLIDE 53

Cyclic polytopes.

Consider the moment map m : R → Rd that sends x → (x, x2, · · · , xd). The convex hull of any(!) n points on that curve is what is called a cyclic polytope C(n, d).

Federico Castillo UC Davis Ehrhart Positivity

slide-54
SLIDE 54

Cyclic polytopes.

Consider the moment map m : R → Rd that sends x → (x, x2, · · · , xd). The convex hull of any(!) n points on that curve is what is called a cyclic polytope C(n, d). Ehrhart Polynomial. Fu Liu proved that under certain integrality conditions, the coefficient

  • f tk in the Ehrhart polynomal of P is given by the volume of the

projection that forgets the last k coordinates.

Federico Castillo UC Davis Ehrhart Positivity

slide-55
SLIDE 55

Not a combinatorial property

Theorem (Liu) For any polytope P there is a polytope P′ with the same face lattice and Ehrhart positivity.

Federico Castillo UC Davis Ehrhart Positivity

slide-56
SLIDE 56

Plus many unknowns.

Other polytopes have been observed to be positive. CRY (Chan-Robbins-Yuen). Tesler matrices (Mezaros-Morales-Rhoades). Birkhoff polytopes (Beck-DeLoera-Pfeifle-Stanley). Matroid polytopes (De Loera - Haws- Koeppe).

Federico Castillo UC Davis Ehrhart Positivity

slide-57
SLIDE 57

Plus many unknowns.

Other polytopes have been observed to be positive. CRY (Chan-Robbins-Yuen). Tesler matrices (Mezaros-Morales-Rhoades). Birkhoff polytopes (Beck-DeLoera-Pfeifle-Stanley). Matroid polytopes (De Loera - Haws- Koeppe). Also: Littlewood Richardson Ronald King conjecture that the stretch littlewood richardson coefficients ctν

tλ,tµ are polynomials in N[t]. This polynomials are known

to be Ehrhart polynomials.

Federico Castillo UC Davis Ehrhart Positivity

slide-58
SLIDE 58

General approach?

Federico Castillo UC Davis Ehrhart Positivity

slide-59
SLIDE 59

General approach?

“Though it be madness, yet there’s method in’t...” Hamlet, Act II.

Federico Castillo UC Davis Ehrhart Positivity

slide-60
SLIDE 60

Method in the madness.

Coming from the theory of toric varieties, we have Definition A McMullen formula is a function α such that |P ∩ Zd| =

  • F⊆P

α(F, P)nvol(F). where the sum is over all faces and α depends locally on F and P. More precisely, it is defined on the normal cone of F in P.

Federico Castillo UC Davis Ehrhart Positivity

slide-61
SLIDE 61

Method in the madness.

Coming from the theory of toric varieties, we have Definition A McMullen formula is a function α such that |P ∩ Zd| =

  • F⊆P

α(F, P)nvol(F). where the sum is over all faces and α depends locally on F and P. More precisely, it is defined on the normal cone of F in P. McMullen proved the existence of such α in a nonconstructive and nonunique way.

Federico Castillo UC Davis Ehrhart Positivity

slide-62
SLIDE 62

Constructions

There are at least three different constructions

1 Pommersheim-Thomas. Need to choose a flag of subspaces. 2 Berline-Vergne. No choices, invariant under On(Z). This is what

we use.

3 Schurmann-Ring. Need to choose a fundamental cell.

Federico Castillo UC Davis Ehrhart Positivity

slide-63
SLIDE 63

Example

Berline-Vergne Pommersheim-Thomas Schurmann-Ring

5 12 9 20 1 2 5 12 1 4 1 4 3 10 1 6 1 4

McMullen Formula: |P ∩ Z| = (Area of P) + 1

2(Perimeter of P) + 1.

The way one gets the +1 is different.

Federico Castillo UC Davis Ehrhart Positivity

slide-64
SLIDE 64

Refinement of positivity.

This gives expressions for the coefficients. |nP ∩ Zd| =

  • F⊂nP

α(F, nP)nvol(F) =

  • F⊂P

α(F, P)nvol(F)ndim(F)

Federico Castillo UC Davis Ehrhart Positivity

slide-65
SLIDE 65

Refinement of positivity.

This gives expressions for the coefficients. |nP ∩ Zd| =

  • F⊂nP

α(F, nP)nvol(F) =

  • F⊂P

α(F, P)nvol(F)ndim(F) We see that Coefficient The coefficient of nk is

  • F:dim(F)=k

α(P, F)vol(F).

Federico Castillo UC Davis Ehrhart Positivity

slide-66
SLIDE 66

Refinement of positivity.

This gives expressions for the coefficients. |nP ∩ Zd| =

  • F⊂nP

α(F, nP)nvol(F) =

  • F⊂P

α(F, P)nvol(F)ndim(F) We see that Coefficient The coefficient of nk is

  • F:dim(F)=k

α(P, F)vol(F). As long as all α are positive, then the coefficients will be positive.

Federico Castillo UC Davis Ehrhart Positivity

slide-67
SLIDE 67

Main properties.

The important facts about the Berline-Vergne construction are It exists. Symmetric under rearranging coordinates. It is a valuation. We exploit these.

Federico Castillo UC Davis Ehrhart Positivity

slide-68
SLIDE 68

A refined conjecture.

We pose the following. Conjecture. The regular permutohedron is (Berline-Vergne) α positive.

Federico Castillo UC Davis Ehrhart Positivity

slide-69
SLIDE 69

A refined conjecture.

We pose the following. Conjecture. The regular permutohedron is (Berline-Vergne) α positive. We care because this imply Ehrhart positivity for a family of polytopes.

Federico Castillo UC Davis Ehrhart Positivity

slide-70
SLIDE 70

A refined conjecture.

We pose the following. Conjecture. The regular permutohedron is (Berline-Vergne) α positive. We care because this imply Ehrhart positivity for a family of polytopes. Proposition. The above conjecture implies that Generalized Permutohedra are positive.

Federico Castillo UC Davis Ehrhart Positivity

slide-71
SLIDE 71

A refined conjecture.

We pose the following. Conjecture. The regular permutohedron is (Berline-Vergne) α positive. We care because this imply Ehrhart positivity for a family of polytopes. Proposition. The above conjecture implies that Generalized Permutohedra are positive. This would expand on previous results from Postnikov, and a conjecture of De Loera-Haws-Koeppe stating that matroid polytopes are positive.

Federico Castillo UC Davis Ehrhart Positivity

slide-72
SLIDE 72

Partial results.

We’ve checked the conjecture in the cases:

1 The linear term (corresponding to edges) in dimensions up to

100.

2 The third and fourth coefficients. 3 Up to dimension 6.

Federico Castillo UC Davis Ehrhart Positivity

slide-73
SLIDE 73

Regular permutohedra revisited.

Figure: A permutohedron in dimension 3.

Federico Castillo UC Davis Ehrhart Positivity

slide-74
SLIDE 74

Regular permutohedra revisited.

Figure: A permutohedron in dimension 3.

For example, α(v, Π3) =

1 24 for any vertex. Since they are all

symmetric and they add up to 1.

Federico Castillo UC Davis Ehrhart Positivity

slide-75
SLIDE 75

A deformation.

Figure: Truncated octahedron

Federico Castillo UC Davis Ehrhart Positivity

slide-76
SLIDE 76

A deformation.

Figure: Truncated octahedron

Federico Castillo UC Davis Ehrhart Positivity

slide-77
SLIDE 77

Computing with the properties.

Note that we have just two types of edges (with normalized volume 1). From the permutohedron we get 24α1 + 12α2 = 6. Now looking at the octhaedron, the alpha values are the same, since the normal cones didn’t change. In this case we get 12α2 = 7/3

Federico Castillo UC Davis Ehrhart Positivity

slide-78
SLIDE 78

Computing with the properties.

Note that we have just two types of edges (with normalized volume 1). From the permutohedron we get 24α1 + 12α2 = 6. Now looking at the octhaedron, the alpha values are the same, since the normal cones didn’t change. In this case we get 12α2 = 7/3 This information is enough to conclude that α1 = 11/72 α2 = 14/72

Federico Castillo UC Davis Ehrhart Positivity

slide-79
SLIDE 79

Computing with the properties.

Note that we have just two types of edges (with normalized volume 1). From the permutohedron we get 24α1 + 12α2 = 6. Now looking at the octhaedron, the alpha values are the same, since the normal cones didn’t change. In this case we get 12α2 = 7/3 This information is enough to conclude that α1 = 11/72 α2 = 14/72 Remark. We did not use the explicit construction at all, just existence and

  • properties. This line of thought is the one we generalize.

Federico Castillo UC Davis Ehrhart Positivity

slide-80
SLIDE 80

Main result.

We have a combinatorial formula for the α values of faces of regular

  • permutohedra. This formula involves mixed Ehrhart coefficients of
  • hypersimplices. The takeaway from this is

Uniqueness theorem. Any McMullen formula that is symmetric under the coordinates is uniquely determined on the faces of permutohedra.

Federico Castillo UC Davis Ehrhart Positivity

slide-81
SLIDE 81

Main result.

We have a combinatorial formula for the α values of faces of regular

  • permutohedra. This formula involves mixed Ehrhart coefficients of
  • hypersimplices. The takeaway from this is

Uniqueness theorem. Any McMullen formula that is symmetric under the coordinates is uniquely determined on the faces of permutohedra. Which leads to the question. Question. Is Berline and Vergne the only construction that satisfies additivity and symmetry?

Federico Castillo UC Davis Ehrhart Positivity

slide-82
SLIDE 82

Warning

We want to remark that it is not true that zonotopes are BV α positive, even though they are Ehrhart positive.

Federico Castillo UC Davis Ehrhart Positivity

slide-83
SLIDE 83

A bit about the formula

Let P1, · · · , Pm be a list of polytopes of dimension n, then Mixed Valuations The expression Lat(w1P1 + · · · + wmPm) is a polynomial on the wi

  • variables. The coefficients are called mixed Ehrhart coefficients.

Federico Castillo UC Davis Ehrhart Positivity

slide-84
SLIDE 84

A bit about the formula

Let P1, · · · , Pm be a list of polytopes of dimension n, then Mixed Valuations The expression Lat(w1P1 + · · · + wmPm) is a polynomial on the wi

  • variables. The coefficients are called mixed Ehrhart coefficients.

On the top degree we have the mixed volumes.

Federico Castillo UC Davis Ehrhart Positivity

slide-85
SLIDE 85

A bit about the formula

Let P1, · · · , Pm be a list of polytopes of dimension n, then Mixed Valuations The expression Lat(w1P1 + · · · + wmPm) is a polynomial on the wi

  • variables. The coefficients are called mixed Ehrhart coefficients.

On the top degree we have the mixed volumes. Volumes are always positive and mixed volumes are too, although this is not clear from the above definition.

Federico Castillo UC Davis Ehrhart Positivity

slide-86
SLIDE 86

Permutohedra

We define a permutohedron for any vector x) = (x1, · · · , xn+1) ∈ Rn+1. Let’s assume x1 ≤ · · · ≤ xn+1.

Federico Castillo UC Davis Ehrhart Positivity

slide-87
SLIDE 87

Permutohedra

We define a permutohedron for any vector x) = (x1, · · · , xn+1) ∈ Rn+1. Let’s assume x1 ≤ · · · ≤ xn+1. Perm(x) := conv{

  • xσ(1), xσ(2), · · · , xσ(n+1)
  • ∈ Rn+1 : σ ∈ Sn+1}.

Federico Castillo UC Davis Ehrhart Positivity

slide-88
SLIDE 88

Permutohedra

We define a permutohedron for any vector x) = (x1, · · · , xn+1) ∈ Rn+1. Let’s assume x1 ≤ · · · ≤ xn+1. Perm(x) := conv{

  • xσ(1), xσ(2), · · · , xσ(n+1)
  • ∈ Rn+1 : σ ∈ Sn+1}.

If we define wi := xi+1 − xi, for i = 1, · · · , n, then Perm(x) = w1∆1,n+1 + w2∆2,n+1 + · · · + wn∆n,n+1.

Federico Castillo UC Davis Ehrhart Positivity

slide-89
SLIDE 89

Permutohedra

We define a permutohedron for any vector x) = (x1, · · · , xn+1) ∈ Rn+1. Let’s assume x1 ≤ · · · ≤ xn+1. Perm(x) := conv{

  • xσ(1), xσ(2), · · · , xσ(n+1)
  • ∈ Rn+1 : σ ∈ Sn+1}.

If we define wi := xi+1 − xi, for i = 1, · · · , n, then Perm(x) = w1∆1,n+1 + w2∆2,n+1 + · · · + wn∆n,n+1. So the number of integer points depends polynomially on the parameters wi.

Federico Castillo UC Davis Ehrhart Positivity

slide-90
SLIDE 90

Permutohedra

We define a permutohedron for any vector x) = (x1, · · · , xn+1) ∈ Rn+1. Let’s assume x1 ≤ · · · ≤ xn+1. Perm(x) := conv{

  • xσ(1), xσ(2), · · · , xσ(n+1)
  • ∈ Rn+1 : σ ∈ Sn+1}.

If we define wi := xi+1 − xi, for i = 1, · · · , n, then Perm(x) = w1∆1,n+1 + w2∆2,n+1 + · · · + wn∆n,n+1. So the number of integer points depends polynomially on the parameters wi.These parameters are the lenghts of the edges in Perm(x).

Federico Castillo UC Davis Ehrhart Positivity

slide-91
SLIDE 91

Permutohedra

We define a permutohedron for any vector x) = (x1, · · · , xn+1) ∈ Rn+1. Let’s assume x1 ≤ · · · ≤ xn+1. Perm(x) := conv{

  • xσ(1), xσ(2), · · · , xσ(n+1)
  • ∈ Rn+1 : σ ∈ Sn+1}.

If we define wi := xi+1 − xi, for i = 1, · · · , n, then Perm(x) = w1∆1,n+1 + w2∆2,n+1 + · · · + wn∆n,n+1. So the number of integer points depends polynomially on the parameters wi.These parameters are the lenghts of the edges in Perm(x). For instance, the coefficient of w1w2 is, by definition, 2!MLat2(∆1,n+1, ∆2,n+1)

Federico Castillo UC Davis Ehrhart Positivity

slide-92
SLIDE 92

Formula

Roughly What we have looks like α(F, P) = A × B. Where A is some combinatorial expression, evidently positive. And B is one (depending of F) mixed Ehrhart coefficient of hypersimplices.

Federico Castillo UC Davis Ehrhart Positivity

slide-93
SLIDE 93

Formula

Roughly What we have looks like α(F, P) = A × B. Where A is some combinatorial expression, evidently positive. And B is one (depending of F) mixed Ehrhart coefficient of hypersimplices. In particular, our conjecture is equivalent to the positivity of such coefficients.

Federico Castillo UC Davis Ehrhart Positivity

slide-94
SLIDE 94

Formula

Roughly What we have looks like α(F, P) = A × B. Where A is some combinatorial expression, evidently positive. And B is one (depending of F) mixed Ehrhart coefficient of hypersimplices. In particular, our conjecture is equivalent to the positivity of such

  • coefficients. It is not even clear if hypersimplices themselves (without

any mixing) are Ehrhart positive.

Federico Castillo UC Davis Ehrhart Positivity

slide-95
SLIDE 95

Example

An instance of the formula looks like: A facet in Π3 Formula would say it is equal to 2 · 2 24 2!MLat2(∆1,4, ∆3,4). where M stands for mixed and Lat2 is the quadratic coefficient of Ehrhart polynomial.

Federico Castillo UC Davis Ehrhart Positivity

slide-96
SLIDE 96

Example

An instance of the formula looks like: A facet in Π3 Formula would say it is equal to 2 · 2 24 2!MLat2(∆1,4, ∆3,4). where M stands for mixed and Lat2 is the quadratic coefficient of Ehrhart polynomial. Remark: The value at facets is always 1

2.

Federico Castillo UC Davis Ehrhart Positivity

slide-97
SLIDE 97

Example

An instance of the formula looks like: A facet in Π3 Formula would say it is equal to 2 · 2 24 2!MLat2(∆1,4, ∆3,4). where M stands for mixed and Lat2 is the quadratic coefficient of Ehrhart polynomial. Remark: The value at facets is always 1

2.

This mixed valuations can be evaluated in the usual alternating form. We can check if the above expression is right. Let’s do it!

Federico Castillo UC Davis Ehrhart Positivity

slide-98
SLIDE 98

Example

i (∆14 + ∆34, t) = 10 3 t3 + 5t2 + 11 3 t + 1, i (∆14, t) = 1 6t3 + t2 + 11 6 t + 1, i (∆34, t) = 1 6t3 + t2 + 11 6 t + 1. Therefore, 2!MLat2(∆1,4, ∆3,4) =5 − 1 − 1 = 3 So we get 2 · 2 24 2!MLat2(∆1,4, ∆3,4) = 4 24 · 3 = 1 2

Federico Castillo UC Davis Ehrhart Positivity

slide-99
SLIDE 99

Further direction.

Some observations lead to the very natural question: Sum of positives. If P and Q are positive, is it true that P + Q is positive?

Federico Castillo UC Davis Ehrhart Positivity

slide-100
SLIDE 100

Thank you!

Gracias! Danke!

Federico Castillo UC Davis Ehrhart Positivity