Some notes on Ehrhart fans

Matthieu Piquerez

2025-01

Abstract
Motivated both by lattice point counting in discrete geometry and by Euler characteristics of sheaves in algebraic geometry, Melody Chan, Emily Clader, Carly Klivans, and Dustin Ross recently introduced the notion of Ehrhart fans. Nothing is published yet. Here I study this notion with my own tools. Most results are already known by the above mentioned authors.

Main properties

Definition

Let ΣΣ be a balanced unimodular fan. We use fan conventions.

Definition 1 (Ehrhart fan)

A fan ΣΣ is an Ehrhart fan if:

  1. for any nonzero face σσ, ΣσΣ^σ is Ehrhart,
  2. there is a unique function f:A1(Σ;ℤ)→ℤf: A^1(Σ; ℤ) → ℤ, called the Ehrhart polynomial of ΣΣ, such that:
    1. f(0)=1f(0) = 1
    2. f(z+xρ)=f(z)+fρ(iρ*(z+xρ))f(z + x_ρ) = f(z) + f^ρ(i^*_ρ(z + x_ρ)) for any z∈ℤΣ1z∈ℤ^{Σ_1} and any ray ρρ, where fρf^ρ is the Ehrhart polynomial of ΣρΣ^ρ.

Note that ff is indeed a polynomial by Proposition 2 (Polynomiality). We denote it by Ehr(Σ)\mathrm{Ehr}(Σ).

Pre-Ehrhart fans and properties of the Ehrhart polynomial

Definition 2 (Pre-Ehrhart fan)
A fan ΣΣ is pre-Ehrhart if it only verifies the point 1. of the above definition.
Example 1 (One dimensional Ehrhart fans)
One dimensional balanced fans ΣΣ are Ehrhart of Ehrhart polynomial Ehr(Σ)(∑ρ∈Σ1aρxρ)=1+∑ρaρ. \mathrm{Ehr}(Σ)(∑_{ρ ∈ Σ_1} a_ρ x_ρ) = 1 + ∑_ρ a_ρ.
Proposition 1
If ΣΣ is pre-Ehrhart, then there exists a unique function f:ℤΣ1→ℤf: ℤ^{Σ_1} → ℤ which verifies points 2.a. and the following variant of 2.b: ∀z,ρ,f(z+xρ)=f(z)+fρ(iρ*(Ψ(z+xρ))) ∀z,ρ, \quad f(z + x_ρ) = f(z) + f^ρ(i^*_ρ(Ψ(z + x_ρ))) where Ψ:ℤΣ1→A1(Σ)Ψ: ℤ^{Σ_1} → A^1(Σ) is the projection.

We call this unique function the Ehrhart polynomial of ΣΣ and denote it Ehr(Σ)\mathrm{Ehr}(Σ).

Proof

The uniqueness is clear. Denote by K=⟨xρ,−xρ⟩ρ∈Σ1K = ⟨x_ρ,-x_ρ⟩_{ρ ∈ Σ_1} the free monoid generated by the xρx_ρ and there opposite. Let Φ:K→ℤΣ1Φ: K → ℤ^{Σ_1} be the projection. We denote by ;; the binary operation of the group. Let f:K→ℤf: K → ℤ by :

It suffices to prove that, for any y,z∈Ky, z ∈ K, any ρ,ρ′ρ, ρ' and any ε,ε′∈{±1}ε, ε' ∈ \{±1\},

Using Ψ(Φ(y;xρ;−xρ))=Ψ(Φ(y))Ψ(Φ(y ; x_ρ ; -x_ρ)) = Ψ(Φ(y)), the first point reduces to the case z=0z=0, and then it is trivial from the definition of ff. Then we can reduce the second case to ε=ε′=+1ε = ε' = +1; indeed, with this case, we obtain the other with algebraic manipulations similar to the following one: f(−xρ;xρ′)=f(−xρ;xρ′;xρ;−xρ)=f(−xρ;xρ;xρ′;−xρ)=f(xρ′;−xρ). f(-x_ρ ; x_{ρ'}) = f(-x_ρ ; x_{ρ'} ; x_ρ ; -x_ρ) = f(-x_ρ ; x_ρ ; x_{ρ'} ; -x_ρ) = f(x_{ρ'} ; -x_ρ).

Both cases ρ=ρ′ρ=ρ' and ρ∧ρ′=𝟏ρ∧ρ'= 𝟏 are trivial. It remains the case ρ∧ρ′=τρ ∧ ρ' = τ with |τ|=2|τ| = 2. As before, we can assume that z=0z=0. Then we get f(y;xρ;xρ′)=f(y)+fρ(iρ*(Ψ(Φ(z;xρ))))+fρ′(iρ′*(Ψ(Φ(z;xρ;xρ′))))=f(y)+fρ(iρ*(Ψ(Φ(z;xρ))))+fρ′(iρ′*(Ψ(Φ(z;xρ′)))+xπρ′(ρ))=f(y)+fρ(iρ*(Ψ(Φ(z;xρ))))+fρ′(iρ′*(Ψ(Φ(z;xρ′))))+fτ(iρ′≺τ*(iρ′*(Ψ(Φ(z;xρ′)))+xπρ′(ρ)))=f(y)+fρ(iρ*(Ψ(Φ(z;xρ))))+fρ′(iρ′*(Ψ(Φ(z;xρ′))))+fτ(iτ*(Ψ(Φ(z;xρ′;xρ)))). \begin{align} f(y ; x_ρ ; x_{ρ'}) &= f(y) + f^ρ(i^*_ρ(Ψ(Φ(z ; x_ρ)))) + f^{ρ'}(i^*_{ρ'}(Ψ(Φ(z ; x_ρ ; x_{ρ'})))) \\ &= f(y) + f^ρ(i^*_ρ(Ψ(Φ(z ; x_ρ)))) + f^{ρ'}(i^*_{ρ'}(Ψ(Φ(z ; x_{ρ'}))) + x_{π^ρ'(ρ)}) \\ &= f(y) + f^ρ(i^*_ρ(Ψ(Φ(z ; x_ρ)))) + f^{ρ'}(i^*_{ρ'}(Ψ(Φ(z ; x_{ρ'})))) + f^τ(i^*_{ρ'≺τ}(i^*_{ρ'}(Ψ(Φ(z ; x_{ρ'}))) + x_{π^ρ'(ρ)})) \\ &= f(y) + f^ρ(i^*_ρ(Ψ(Φ(z ; x_ρ)))) + f^{ρ'}(i^*_{ρ'}(Ψ(Φ(z ; x_{ρ'})))) + f^τ(i^*_{τ}(Ψ(Φ(z ; x_{ρ'} ; x_ρ)))). \end{align} The last formula is symmetric in ρρ and ρ′ρ', which concludes the proof.

Any element zz of ℤΣ1ℤ^{Σ_1} can be written in a unique way under the form ∑ρaρxρ∑_ρ a_ρ x_ρ. In the following, we will say that we work by induction on 𝐳𝐳 to mean that we work by induction on ∑ρ|aρ|∑_{ρ} |a_ρ|. Moreover, we often omit to check the step case where we remove xρx_ρ from zz, since it directly follows from the case where we add it. We also omit regularly the map i*i^*.

Lemma 1 (Pre-Ehrhart to Ehrhart fans)
Let ΣΣ be a pre-Ehrhart fan, and let ℓ∈Mℓ ∈ M. Then, for any zz, Ehr(Σ)(xℓ+z)=Ehr(Σ)(xℓ)−1+Ehr(Σ)(z). \mathrm{Ehr}(Σ)(x_ℓ + z) = \mathrm{Ehr}(Σ)(x_ℓ)-1 + \mathrm{Ehr}(Σ)(z). In particular, a pre-Ehrhart fan is Ehrhart if and only if its Ehrhart polynomial values one on any xℓx_ℓ, ℓ∈Mℓ ∈ M.
Proof
Working by induction on zz, for z=0z = 0, this is trivial, and otherwise Ehr(Σ)(xℓ+z+xρ)=Ehr(Σ)(xℓ+z)+Ehr(Σρ)(xℓ+z+xρ)=Ehr(Σ)(xℓ)−1+Ehr(Σ)(z)+Ehr(Σρ)(xℓ+z+xρ)=Ehr(Σ)(xℓ)−1+Ehr(Σ)(z)+Ehr(Σρ)(z+xρ)=Ehr(Σ)(xℓ)−1+Ehr(Σ)(z+xρ). \begin{align} \mathrm{Ehr}(Σ)(x_ℓ + z + x_ρ) &= \mathrm{Ehr}(Σ)(x_ℓ + z) + \mathrm{Ehr}(Σ^ρ)(x_ℓ + z + x_ρ) \\ &= \mathrm{Ehr}(Σ)(x_ℓ)-1 + \mathrm{Ehr}(Σ)(z) + \mathrm{Ehr}(Σ^ρ)(x_ℓ + z + x_ρ) \\ &= \mathrm{Ehr}(Σ)(x_ℓ)-1 + \mathrm{Ehr}(Σ)(z) + \mathrm{Ehr}(Σ^ρ)(z + x_ρ) \\ &= \mathrm{Ehr}(Σ)(x_ℓ)-1 + \mathrm{Ehr}(Σ)(z + x_ρ). \end{align}
Proposition 2 (Polynomiality)
Assume ΣΣ is pre-Ehrhart. Then the function f:=Ehr(Σ)f := \mathrm{Ehr}(Σ) is polynomial, that is f(∑ρ∈Σ1)aρxρ∈ℚ[(aρ)ρ∈Σ1]f(∑_{ρ ∈ Σ_1}) a_ρ x_ρ ∈ ℚ[(a_ρ)_{ρ ∈ Σ_1}].
Proof
The proof goes by induction. Assume the proposition is true for any fans of dimension d−1d-1. Let ρ1,…,ρkρ_1, …, ρ_k be the rays of ΣΣ. Then, simplifying the notations with ak=aρka_k = a_{ρ_k} and xk=xρkx_k = x_{ρ_k} and omitting the i*i^*, f(∑i=1kakxk)=∑j=1a1fρ1(jx1)+∑j=1a2fρ2(a1x1+jx2)+⋯+∑j=1akfρk(a1x1+⋯+ak−1xk−1+jxk). f(∑_{i=1}^k a_k x_k) = ∑_{j=1}^{a_1}f^{ρ_1}(j x_1) \ + \ ∑_{j=1}^{a_2}f^{ρ_2}(a_1 x_1 + j x_2) \ + \ ⋯ \ + \ ∑_{j=1}^{a_k}f^{ρ_k}(a_1 x_1 + ⋯ + a_{k-1} x_{k-1} + j x_k). We claim that each term is a polynomial. Indeed, let 1≤i≤k1 ≤ i ≤ k, and let ℓℓ be a integral linear form which is 1 on eρie_{ρ_i}. Then, for any jj fρi(a1x1+⋯+ai−1xi−1+jxi)=∑l,l<i,ρl∧ρ1≠𝟏fρi((al−jℓ(eρl))xl). f^{ρ_i} (a_1 x_1 + ⋯ + a_{i-1} x_{i-1} + j x_i) = ∑_{l, l < i, ρ_l ∧ ρ_1 ≠ 𝟏} f^{ρ_i}((a_l-jℓ(e_{ρ_l}))x_l). By induction, this is a polynomial in ℤ[(al)l,j]ℤ[(a_l)_l, j]. Since sum of rr-th powers is a polynomial for any rr, ∑j=1aifρi(a1x1+⋯+ai−1xi−1+jxi)∑_{j=1}^{a_i} f^{ρ_i} (a_1 x_1 + ⋯ + a_{i-1} x_{i-1} + j x_i) is a polynomial in ℤ[(al)l,ai]ℤ[(a_l)_l, a_i]. Summing all together, we get the conclusion follows.

Stability properties

Lemma 2 (Ehrhart on −xρ-x_\rho)
Assume ΣΣ verifies point 1. Then the function f:ℤΣ1→ℤf: ℤ^{Σ_1} → ℤ introduced in Proposition 1 verifies f(−xρ)=0f(-x_ρ) = 0 for all ρρ.
Proof
Clear from the definition.
Proposition 3 (Stability by product)
Let ΣΣ and Σ′Σ' be two Ehrhart fans in NℝN_ℝ and N′ℝN'_ℝ respectively, both verifying point 1, and let ι1:Σ→Σ×Σ′ι_1: Σ → Σ×Σ' and ι2:Σ′→Σ×Σ′ι_2: Σ' → Σ×Σ' be the two inclusions. Then, for any zz, Ehr(Σ×Σ′)(z)=Ehr(Σ)(ι1*(z))⋅Ehr(Σ′)(ι2*(z)). \mathrm{Ehr}(Σ×Σ')(z) = \mathrm{Ehr}(Σ)(ι_1^*(z)) ⋅ \mathrm{Ehr}(Σ')(ι_2^*(z)). In particular, the product of two Ehrhart fans is Ehrhart.
Proof
We get the proof by induction, both on the dimension of the product, and on the sum ∑ρ|aρ|∑_ρ|a_ρ|. Clearly Ehr(Σ×Σ′)(0)=Ehr(Σ)(0)Ehr(Σ′)(0)\mathrm{Ehr}(Σ×Σ')(0) = \mathrm{Ehr}(Σ)(0)\mathrm{Ehr}(Σ')(0). Let ρρ be a ray of ΣΣ. We also denote by ρρ the corresponding ray in Σ×Σ′Σ×Σ'. We have, for any zz and omitting the i*i^*, Ehr(Σ×Σ′)(z+xρ)=Ehr(Σ×Σ′)(z)+Ehr((Σ×Σ′)ρ)(z+xρ)=Ehr(Σ)(ι1*(z))⋅Ehr(Σ′)(ι2*(z))+Ehr(Σρ×Σ′)(z+xρ)=Ehr(Σ)(ι1*(z))⋅Ehr(Σ′)(ι2*(z))+Ehr(Σρ)(ι1*(z+xρ))⋅Ehr(Σ′)(ι2*(z+xρ))=Ehr(Σ)(ι1*(z))⋅Ehr(Σ′)(ι2*(z+xρ))+Ehr(Σρ)(ι1*(z+xρ))⋅Ehr(Σ′)(ι2*(z+xρ))=(Ehr(Σ)(ι1*(z))+Ehr(Σρ)(ι1*(z+xρ)))⋅Ehr(Σ′)(ι2*(z+xρ))=Ehr(Σ)(ι1*(z+xρ))⋅Ehr(Σ′)(ι2*(z+xρ)). \begin{align*} \mathrm{Ehr}(Σ×Σ')(z + x_ρ) &= \mathrm{Ehr}(Σ×Σ')(z) + \mathrm{Ehr}((Σ×Σ')^ρ)(z + x_ρ) \\ &= \mathrm{Ehr}(Σ)(ι_1^*(z))⋅\mathrm{Ehr}(Σ')(ι_2^*(z)) + \mathrm{Ehr}(Σ^ρ×Σ')(z + x_ρ) \\ &= \mathrm{Ehr}(Σ)(ι_1^*(z))⋅\mathrm{Ehr}(Σ')(ι_2^*(z)) + \mathrm{Ehr}(Σ^ρ)(ι_1^*(z + x_ρ))⋅\mathrm{Ehr}(Σ')(ι_2^*(z + x_ρ)) \\ &= \mathrm{Ehr}(Σ)(ι_1^*(z))⋅\mathrm{Ehr}(Σ')(ι_2^*(z + x_ρ)) + \mathrm{Ehr}(Σ^ρ)(ι_1^*(z + x_ρ))⋅\mathrm{Ehr}(Σ')(ι_2^*(z + x_ρ)) \\ &= (\mathrm{Ehr}(Σ)(ι_1^*(z)) + \mathrm{Ehr}(Σ^ρ)(ι_1^*(z + x_ρ))) \ ⋅ \ \mathrm{Ehr}(Σ')(ι_2^*(z + x_ρ)) \\ &= \mathrm{Ehr}(Σ)(ι_1^*(z + x_ρ))⋅\mathrm{Ehr}(Σ')(ι_2^*(z + x_ρ)). \end{align*} The proof of the first part easily follows. For the stability of being Ehrhart, simply note that if ℓℓ is a linear form on Nℝ×N′ℝN_ℝ × N'_ℝ, then ι1*(xℓ)=xℓ|Σι_1^*(x_ℓ) = x_{ℓ|_Σ}.
Proposition 4 (Stability by blow-up)
Let ΣΣ be a fan and let Σ′Σ' be the unimodular blow-up of ΣΣ on some face σσ, and let ρ0ρ_0 be the new ray. Then ΣΣ is pre-Ehrhart if and only if Σ′Σ' is. Moreover, let ι:Σ′→Σι: Σ' → Σ be the inclusion. Then, for any zz, Ehr(Σ′)(ι*(z))=Ehr(Σ). \mathrm{Ehr}(Σ')(ι^*(z)) = \mathrm{Ehr}(Σ). In particular, Ehrhart fans are stable by unimodular blow-ups and blow-downs.

We recall that ι*(xρ)=xρ+xρ0ι^*(x_ρ) = x_ρ + x_ρ_0 if ρ∈σρ ∈ σ.

Proof
We work by induction on both the dimension of ΣΣ, and on zz. The only nontrivial case to check is the following one. Let ρρ be a ray of σσ. Then, for any zz Ehr(Σ′)(ι*(z+xρ))=Ehr(Σ′)(ι*(z)+xρ0+xρ)=Ehr(Σ′)(ι*(z))+Ehr(Σ′ρ0)(ι*(z)+xρ0)+Ehr(Σ′ρ)(ι*(z)+xρ0+xρ)=Ehr(Σ)(z)+Ehr(Σ′ρ0)(ι*(z)+xρ0)+Ehr(Σ′ρ)(ι*(z+xρ))=Ehr(Σ)(z+xρ)+Ehr(Σ′ρ0)(ι*(z)+xρ0). \begin{align*} \mathrm{Ehr}(Σ')(ι^*(z + x_ρ)) &= \mathrm{Ehr}(Σ')(ι^*(z) + x_ρ_0 + x_ρ) \\ &= \mathrm{Ehr}(Σ')(ι^*(z)) + \mathrm{Ehr}(Σ'^{ρ_0})(ι^*(z) + x_ρ_0) + \mathrm{Ehr}(Σ'^ρ)(ι^*(z) + x_ρ_0 + x_ρ) \\ &= \mathrm{Ehr}(Σ)(z) + \mathrm{Ehr}(Σ'^{ρ_0})(ι^*(z) + x_ρ_0) + \mathrm{Ehr}(Σ'^ρ)(ι^*(z + x_ρ)) \\ &= \mathrm{Ehr}(Σ)(z + x_ρ) + \mathrm{Ehr}(Σ'^{ρ_0})(ι^*(z) + x_ρ_0). \end{align*} Hence we just have to prove that the second term is zero. Note that Σ′ρ0Σ'^{ρ_0} can be canonically identified with Σℙ|σ|−1×ΣσΣℙ^{|σ|-1} × Σ^σ, where Σℙ|σ|−1Σℙ^{|σ|-1} is the projective fan of rays corresponding to the rays of σσ. Let 𝚥1:Σℙ|σ|−1→Σ′ρ0𝚥_1: Σℙ^{|σ|-1} → Σ'^{ρ_0} be the inclusion. Then, 𝚥1*(ι*(z))𝚥_1^*(ι^*(z)) is linear of Σℙ|σ|−1Σℙ^{|σ|-1}. Since σσ is unimodular, there exists a linear form ℓℓ which is 1 on both ρ0ρ_0 and ρρ, and that is 0 on other rays of σσ. Hence, using Lemma 2 (Ehrhart on −xρ-x_\rho), Ehr(Σℙ|σ|−1)(𝚥1*(ι*(z)+xρ0))=Ehr(Σℙ|σ|−1)(𝚥1*(ι*(z)+xρ0−xℓ))=Ehr(Σℙ|σ|−1)(𝚥1*(ι*(z))−xρ)=Ehr(Σℙ|σ|−1)(−xρ)=0. \begin{align*} \mathrm{Ehr}(Σℙ^{|σ|-1})(𝚥_1^*(ι^*(z) + x_ρ_0)) &= \mathrm{Ehr}(Σℙ^{|σ|-1})(𝚥_1^*(ι^*(z) + x_ρ_0 - x_ℓ)) \\ &= \mathrm{Ehr}(Σℙ^{|σ|-1})(𝚥_1^*(ι^*(z)) - x_ρ) \\ &= \mathrm{Ehr}(Σℙ^{|σ|-1})(-x_ρ) \\ &= 0. \end{align*} With Proposition 3 (Stability by product), we get that the second term is zero, which concludes the proof.
Proposition 5 (Tropical modifications and Ehrhart fans)

Let ΣΣ be an Ehrhart fan. Let ff be a integral piecewise linear function of ΣΣ whose divisor is trivial or reduced. Set xf:=∑ρf(eρ)xρx_f := ∑_ρ f(e_ρ)x_ρ. Set Σ′:=Tropf(Σ)Σ' := \mathrm{Trop}_f(Σ) and π:Σ′→Σπ: Σ' → Σ the projection. Then Σ′Σ' is Ehrhart if and only if:

Moreover, in both cases, Ehr(Σ′)(π*(z))=Ehr(Σ)(z)\mathrm{Ehr}(Σ')(π^*(z)) = \mathrm{Ehr}(Σ)(z) for any zz.

Proof

We prove it by induction on the dimension of ΣΣ. Recall the [description of the star fan of a tropical modification 🏗️]. Assume that div(f)\mathrm{div}(f) is either trivial or Ehrhart. In both case, we get that Σ′Σ' is pre-Ehrhart by induction. The last equality now follows by a direct induction on zz. In particular, for all linear forms of the form π*(xℓ)π^*(x_ℓ), ℓ∈Mℓ ∈ M, Ehr(Σ′)(xℓ)=1\mathrm{Ehr}(Σ')(x_ℓ) = 1. It remains to check that “vertical” linear form ll, which is the new coordinate, verifies Ehr(Σ′)(xl)=1\mathrm{Ehr}(Σ')(x_l) = 1. Two cases.

which concludes the proof.

Bergman fans are Ehrhart

Theorem 1 (Complete unimodular fans are Ehrhart)
Complete unimodular fans are Ehrhart.
Proof
The point is Ehrhart. By Example 1 (One dimensional Ehrhart fans), the line is Ehrhart. The results follows from Proposition 3 (Stability by product), Proposition 4 (Stability by blow-up) and the weak factorization theorem.
Theorem 6 (Bergman fans are Ehrhart)
Any unimodular generalized Bergman fan is Ehrhart.
Proof
Using Theorem 1 (Complete unimodular fans are Ehrhart), Proposition 4 (Stability by blow-up), and Proposition 5 (Tropical modifications and Ehrhart fans), we can reduce the theorem to the following lemma.
Lemma 3
Let 𝔐𝔐 be a matroid without loop which is not free. Take an element aa of 𝔐𝔐 which is not free, and another element oo. Recall that Σ𝔐Σ_𝔐 is the tropical modification of Σ𝔐−aΣ_{𝔐-a} along the map ff such that, for any flat FF of 𝔐−a𝔐-a: f(eF)={1if o∈cl𝔐(F) and a∉cl𝔐(F),−1if a∈cl𝔐(F) and o∉cl𝔐(F),0otherwise. f(e_F) = \begin{cases} 1 & \text{if $o ∈ \mathrm{cl}_𝔐(F)$ and $a ∉ \mathrm{cl}_𝔐(F)$,} \\ -1 & \text{if $a ∈ \mathrm{cl}_𝔐(F)$ and $o ∉ \mathrm{cl}_𝔐(F)$,} \\ 0 & \text{otherwise.} \end{cases} Then Ehr(Σ𝔐−a)(−xf)=0\mathrm{Ehr}(Σ_{𝔐-a})(-x_f) = 0.
Proof

We work by induction on the rank of the matroid. Put an order on proper nontrivial flats of 𝔐−a𝔐-a which is non-decreasing for the rank : F1,…,FrF_1, \dots, F_r, with F1={o}F_1 = \{o\}. Assume by induction that we have proven that for some m≥2m ≥ 2, Ehr(Σ𝔐−a)(−∑i=1m−1f(eFk)xFk)=0 \mathrm{Ehr}(Σ_{𝔐-a})(-∑_{i=1}^{m-1}f(e_F_k)x_F_k) = 0 This is clear for m=2m=2. Denote the argument by zz. Let us prove that Ehr(Σ𝔐−a)(z−f(eFm)xFm)=0\mathrm{Ehr}(Σ_{𝔐-a})(z - f(e_F_m)x_F_m)=0.

This concludes the proof.

Counter-examples

A counter-example

Take ΣΣ the pseudo fan which is the double cover of Σℙ2Σℙ^2 with six rays ρ1,ρ2,ρ0=−ρ1−ρ2ρ_1,ρ_2, ρ_0=-ρ_1-ρ_2 and their copies ρ4,ρ5,ρ3ρ_4,ρ_5,ρ_3 and six 2-cones ρi+ρi+1ρ_i + ρ_{i+1} for i∈ℤ/6ℤi ∈ ℤ/6ℤ. It is Ehrhart. Take the piecewise-linear function ff with f(ρ0)=−1,f(ρ5)=f(ρ1)=1,f(ρ2)=f(ρ3)=f(ρ4)=0. f(ρ_0) = -1, \qquad f(ρ_5) = f(ρ_1) = 1, \qquad f(ρ_2) = f(ρ_3) = f(ρ_4) = 0. Then div(f)\mathrm{div}(f) is just the tropical line and is Ehrhart, but Ehr(Σ)(−xf)=−1≠0\mathrm{Ehr}(Σ)(-x_f) = -1 ≠ 0.

Hence, Tropf(Σ)\mathrm{Trop}_f(Σ) is just pre-Ehrhart. It is not a fan but still a pseudo-fan, but one can preform some blow-ups and a new tropical modification along a tropical line to get a classical fan. This fan cannot be Ehrhart because of the last equality of Proposition 5 (Tropical modifications and Ehrhart fans). This fan is the counter-example studied by Babaee and Huh.

Simpler counter-examples

It should be easy to find simpler counter-example. We just have to find a Ehrhart fan, for instance Σℙ2Σℙ^2, and a reduced divisor whose associated function ff verifies Ehr(Σ)(−xf)≠0\mathrm{Ehr}(Σ)(-x_f) ≠ 0.

And indeed, take Σ=ΣU3,3Σ = Σ_{U_{3,3}} with the conewise-linear function ff which is 1 on each unit vector. Then its divisor is the one-skeleton of ΣΣ and Ehr(Σ)(f)=−1\mathrm{Ehr}(Σ)(f) = -1. Hence, its tropical modification is trivial.

Actually, take any integral polytope which has a nonzero number of interior points and whose facets are all dimension one for the lattice volume (a unimodular simplex has volume 1). Take its normal fan, and subdivide it to get a unimodular fan. Take the height function ff of the polytope. Take the tropical modification of your fan along ff. Then you get a balanced fan which is not Ehrhart.

Extensions

Weighted fans

For weighted fans, and in order to preserve Example 1 (One dimensional Ehrhart fans), we could adapt Definition 1 (Ehrhart fan) asking for f(0)f(0) on a point to be the weight of the point. Not sure what to put for higher dimensional fans.

Non unimodular

For rational fans that are not unimodular, the main point which does not obviously extends is Proposition 4 (Stability by blow-up). Actually, the proof should work since one can perform unimodular blow-ups until reaching a unimodular fan.