ScalingStacks

4.4. Positivity properties of 𝕋 -Cartier divisors [02QE]

Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.

Complete original source context · Original author HTML

4.4. Positivity properties of 𝕋\mathbb{T}-Cartier divisors

Let Σ\Sigma be a fan in NℝN_{\mathbb{R}} and Ψ\Psi a virtual support function on Σ\Sigma. In this section, we will assume that Σ\Sigma is complete or, equivalently, that the variety XΣX_{\Sigma} is proper.

Many geometric properties of the pair (XΣ,DΨ)(X_{\Sigma},D_{\Psi}) can be read directly from Ψ\Psi. For instance, 𝒪⁡(DΨ)\mathcal{O}(D_{\Psi}) is generated by global sections if and only if the function Ψ\Psi is concave, and the line bundle 𝒪⁡(DΨ){\mathcal{O}}(D_{\Psi}) is ample if and only if Ψ\Psi is strictly concave on Σ\Sigma. In the latter case, the fan Σ\Sigma agrees with the polyhedral complex Π⁡(Ψ)\Pi(\Psi) (Definition 3.34) and the pair (XΣ,DΨ)(X_{\Sigma},D_{\Psi}) is completely determined by Ψ\Psi. Thus, the variety XΣX_{\Sigma} is projective if and only if the fan Σ\Sigma is complete and regular (Definition 3.60).

We associate to Ψ\Psi the subset of MℝM_{\mathbb{R}}

ΔΨ={x∈Mℝ∣⟨x,u⟩≥Ψ(u) for all u∈Nℝ}.\Delta_{\Psi}=\{x\in M_{\mathbb{R}}\mid\langle x,u\rangle\geq\Psi(u)\mbox{ for all }u\in N_{\mathbb{R}}\}.

This set is either empty or a lattice polytope. When 𝒪⁡(DΨ){\mathcal{O}}(D_{\Psi}) is generated by global sections, the polytope ΔΨ\Delta_{\Psi} agrees with stab⁡(Ψ)\operatorname{stab}(\Psi), and Ψ\Psi is the support function of ΔΨ\Delta_{\Psi}.

The polytope ΔΨ\Delta_{\Psi} encodes a lot of information about the pair (XΣ,DΨ)(X_{\Sigma},D_{\Psi}). For instance, we can read from it the space of global sections of 𝒪⁡(DΨ)\mathcal{O}(D_{\Psi}). A monomial rational section χm∈𝒦XΣ\chi^{m}\in\mathcal{K}_{X_{\Sigma}}, m∈Mm\in M, is a regular global section of 𝒪⁡(DΨ)\mathcal{O}(D_{\Psi}) if and only if m∈ΔΨm\in\Delta_{\Psi}. Moreover, the set {χm}m∈M∩ΔΨ\{\chi^{m}\}_{m\in M\cap\Delta_{\Psi}} is a KK-basis of the space of global sections Γ⁡(XΣ,𝒪⁡(DΨ))\Gamma(X_{\Sigma},\mathcal{O}(D_{\Psi})). In the sequel we will see many more examples of this principle.

Proposition 4.37.

Let DΨiD_{\Psi_{i}}, i=1,…,ni=1,\dots,n, be 𝕋\mathbb{T}-Cartier divisors on XΣX_{\Sigma} generated by their global sections. Then

(4.38) (DΨ1⋅⋯⋅DΨn)=MVM⁡(ΔΨ1,…,ΔΨn).(D_{\Psi_{1}}\cdot\dots\cdot D_{\Psi_{n}})=\operatorname{MV}_{M}(\Delta_{\Psi_{1}},\dots,\Delta_{\Psi_{n}}).

where MVM\operatorname{MV}_{M} denotes the mixed volume function associated to the Haar measure volM\operatorname{vol}_{M} on MℝM_{\mathbb{R}} (Definition 3.109). In particular, for a 𝕋\mathbb{T}-Cartier divisor DΨD_{\Psi} generated by its global sections,

(4.39) degDΨ⁡(XΣ)=(DΨn)=n!​volM⁡(ΔΨ).\deg_{D_{\Psi}}(X_{\Sigma})=(D_{\Psi}^{n})=n!\operatorname{vol}_{M}(\Delta_{\Psi}).
Proof.

This follows from [Oda88, Proposition 2.10]. ∎

Remark 4.40.

The intersection multiplicity and the degree in the above Proposition only depend on the isomorphism class of the line bundles 𝒪⁡(DΨi){\mathcal{O}}(D_{\Psi_{i}}) and not on the 𝕋\mathbb{T}-Cartier divisors themselves. It is easy to check directly that the right-hand sides of (4.38) and (4.39) only depends on the isomorphism class of the line bundles. In fact, let LL be a toric line bundle generated by global sections and s1s_{1}, s2s_{2} two toric sections. For i=1,2i=1,2, set Di=div⁡(si)D_{i}=\operatorname{div}(s_{i}) and let Ψi\Psi_{i} be the corresponding support function and Δi\Delta_{i} the associated polytope. Then s2=χm​s1s_{2}=\chi^{m}s_{1} for some m∈Mm\in M. Thus Ψ2=Ψ1−m\Psi_{2}=\Psi_{1}-m and Δ2=Δ1−m\Delta_{2}=\Delta_{1}-m. Since the volume and the mixed volume are invariant under translation, we see that these formulae do not depend on the choice of sections.

Definition 4.41.

A polarized toric variety is a pair (XΣ,DΨ)(X_{\Sigma},D_{\Psi}), where XΣX_{\Sigma} is a toric variety and DΨD_{\Psi} is an ample 𝕋\mathbb{T}-Cartier divisor.

Polarized toric varieties can be classified in terms of their polytopes.

Theorem 4.42.

The correspondence (XΣ,DΨ)↦ΔΨ(X_{\Sigma},D_{\Psi})\mapsto\Delta_{\Psi} is a bijection between the set of polarized toric varieties and the set of lattice polytopes of dimension nn of MM. Two ample 𝕋\mathbb{T}-Cartier divisors DΨD_{\Psi} and DΨ′D_{\Psi^{\prime}} on a toric variety XΣX_{\Sigma} are rationally equivalent if and only if ΔΨ′\Delta_{\Psi^{\prime}} is the translated of ΔΨ\Delta_{\Psi} by an element of MM.

Proof.

If Ψ\Psi is a strictly concave function on Σ\Sigma, then ΔΨ\Delta_{\Psi} is an nn-dimensional lattice polytope. Conversely, if Δ\Delta is a lattice polytope in MℝM_{\mathbb{R}}, then ΨΔ\Psi_{\Delta}, the support function of Δ\Delta, is a strictly concave function on the complete fan ΣΔ=Π⁡(ΨΔ)\Sigma_{\Delta}=\Pi(\Psi_{\Delta}) (see examples 3.71 and 3.76). Therefore, the result follows from Theorem 4.18 and the construction of Remark 4.40. ∎

Remark 4.43.

When DΨD_{\Psi} is only generated by its global sections, the polytope ΔΨ\Delta_{\Psi} may not determine the variety XΣX_{\Sigma}, but it does determine a polarized toric variety that is the image of XΣX_{\Sigma} by a toric morphism. Write Δ=ΔΨ\Delta=\Delta_{\Psi} for short. Let M⁡(Δ)M(\Delta) be as in Notation 3.103 and choose m∈aff⁡(Δ)∩Mm\in\operatorname{aff}(\Delta)\cap M. Set N⁡(Δ)=M​(Δ)∨N(\Delta)=M(\Delta)^{\vee}. The translated polytope Δ−m\Delta-m has the same dimension as its ambient space LΔ=M​(Δ)ℝL_{\Delta}=M(\Delta)_{\mathbb{R}}. By the theorem above, it defines a complete fan ΣΔ\Sigma_{\Delta} in N​(Δ)ℝN(\Delta)_{\mathbb{R}} together with a support function ΨΔ:N⁡(Δ)→ℝ\Psi_{\Delta}\colon N(\Delta)\to\mathbb{R}. The projection N→N⁡(Δ)N\to N(\Delta) induces a toric morphism

φ:XΣ⟶XΣΔ,\varphi\colon X_{\Sigma}\longrightarrow X_{\Sigma_{\Delta}},

the divisor DΨΔD_{\Psi_{\Delta}} is ample, and DΨ=φ∗​DΨΔ+div⁡(χ−m)D_{\Psi}=\varphi^{*}D_{\Psi_{\Delta}}+\operatorname{div}(\chi^{-m}).

Example 4.44.

The projective morphisms associated to 𝕋\mathbb{T}-Cartier divisors generated by global sections can also be made explicit in terms of the lattice points of the associated polytope. Consider a complete toric variety XΣX_{\Sigma} of dimension nn equipped with a 𝕋\mathbb{T}-Cartier divisor DΨD_{\Psi} generated by global sections. Let m0,…,mr∈ΔΨ∩Mm_{0},\dots,m_{r}\in\Delta_{\Psi}\cap M be such that conv⁡(m0,…,mr)=ΔΨ\operatorname{conv}(m_{0},\dots,m_{r})=\Delta_{\Psi}. These vectors determine an H-representation Ψ=mini=0,…,r⁡mi\Psi=\min_{i=0,\dots,r}m_{i}. Let H:Nℝ→ℝrH\colon N_{\mathbb{R}}\to\mathbb{R}^{r} be the linear map defined by H⁡(u)=(mi​(u)−m0​(u))i=1,…,rH(u)=(m_{i}(u)-m_{0}(u))_{i=1,\dots,r}. By Lemma 3.79, Ψ=H∗​ΨΔr+m0\Psi=H^{\ast}\Psi_{\Delta^{r}}+m_{0}.

In ℝr\mathbb{R}^{r} we consider the fan ΣΔr\Sigma_{\Delta^{r}}, whose associated toric variety is ℙr\mathbb{P}^{r}. One easily verifies that, for each σ∈Σ\sigma\in\Sigma, there is σ′∈ΣΔr\sigma^{\prime}\in\Sigma_{\Delta^{r}} with H⁡(σ)⊂σ′H(\sigma)\subset\sigma^{\prime}. Let p=(p0:…:pr)p=(p_{0}:\dots:p_{r}) be an arbitrary rational point of the principal open subset of ℙr\mathbb{P}^{r}. The equivariant morphism φp,H:X→ℙKr\varphi_{p,H}\colon X\to\mathbb{P}^{r}_{K} can be written explicitly as (p0χm0:…:prχmr)(p_{0}\chi^{m_{0}}:\dots:p_{r}\chi^{m_{r}}). Moreover, DΨ=φp,H∗​DΨΔr+div⁡(χ−m0)D_{\Psi}=\varphi_{p,H}^{\ast}D_{\Psi_{\Delta^{r}}}+\operatorname{div}(\chi^{-m_{0}}).

The orbits of a polarized toric variety (XΣ,DΨ)(X_{\Sigma},D_{\Psi}) are in one-to-one correspondence with the faces of ΔΨ\Delta_{\Psi}.

Proposition 4.45.

Let Σ\Sigma be a complete fan in NℝN_{\mathbb{R}} and Ψ\Psi a strictly concave function on Σ\Sigma. The correspondence F↦O⁡(σF)F\mapsto O(\sigma_{F}) is a bijection between the set of faces of ΔΨ\Delta_{\Psi} and the set of the orbits under the action of 𝕋\mathbb{T} on XΣX_{\Sigma}.

Proof.

This follows from Example 3.71. ∎

Equation (4.25) gives a formula for the Weil divisor [DΨ][D_{\Psi}] in terms of the virtual support function Ψ\Psi. When the line bundle 𝒪⁡(DΨ)\mathcal{O}(D_{\Psi}) is ample, we can interpret this formula in terms of the facets of the polytope ΔΨ\Delta_{\Psi}.

Let DΨD_{\Psi} be an ample line bundle on XΣX_{\Sigma}. The polytope ΔΨ\Delta_{\Psi} has maximal dimension nn. For each facet FF of ΔΨ\Delta_{\Psi}, let vFv_{F} be as in Notation 3.103. The ray τF=ℝ≥0​vF\tau_{F}=\mathbb{R}_{\geq 0}v_{F} is a cone of Σ\Sigma.

Proposition 4.46.

With the previous hypothesis,

div(sΨ)=[DΨ]=∑F−⟨vF,F⟩V(τF),\operatorname{div}(s_{\Psi})=[D_{\Psi}]=\sum_{F}-\langle v_{F},F\rangle V(\tau_{F}),

where the sum is over the facets FF of Δ\Delta.

Proof.

Since Ψ\Psi is strictly concave on Σ\Sigma, the Legendre-Fenchel correspondence shows that the set of rays of the form τF\tau_{F} agrees with the set Σ1\Sigma^{1}. Moreover, Ψ⁡(vF)=⟨vF,F⟩\Psi(v_{F})=\langle v_{F},F\rangle, because Ψ\Psi is the support function of Δ\Delta. The proposition then follows from (4.25). ∎

For a 𝕋\mathbb{T}-Cartier divisor generated by global sections, we can interpret its intersection with the closure of an orbit, and its inverse image with respect to an equivariant morphism, in terms of direct and inverse images of concave functions.

Proposition 4.47.

Let Σ\Sigma be a complete fan in NℝN_{\mathbb{R}} and Ψ:Nℝ→ℝ\Psi\colon N_{\mathbb{R}}\to\mathbb{R} a support function on Σ\Sigma.

  1. (1)

    Let σ∈Σ\sigma\in\Sigma, FσF_{\sigma} the associated face of ΔΨ\Delta_{\Psi}, and mσ′∈Fσ∩Mm_{\sigma}^{\prime}\in F_{\sigma}\cap M. Let πσ:Nℝ→N​(σ)ℝ\pi_{\sigma}\colon N_{\mathbb{R}}\to N(\sigma)_{\mathbb{R}} be the natural projection. Then

    (4.48) (Ψ−mσ′)​(σ)=(πσ)∗​(Ψ−mσ′).(\Psi-m^{\prime}_{\sigma})(\sigma)=(\pi_{\sigma})_{\ast}(\Psi-m_{\sigma}^{\prime}).

    In particular, the restriction of DΨ−mσ′D_{\Psi-m^{\prime}_{\sigma}} to V⁡(σ)V(\sigma) is given by the concave function (πσ)∗​(Ψ−mσ′)(\pi_{\sigma})_{\ast}(\Psi-m_{\sigma}^{\prime}). Moreover, the associated polytope is

    (4.49) Δ(Ψ−mσ′)​(σ)=Fσ−mσ′⊂M​(σ)ℝ=σ⊥.\Delta_{(\Psi-m_{\sigma}^{\prime})(\sigma)}=F_{\sigma}-m_{\sigma}^{\prime}\subset M(\sigma)_{\mathbb{R}}=\sigma^{\bot}.
  2. (2)

    Let H:N′→NH\colon N^{\prime}\to N be a linear map and H∨:M→M′H^{\vee}\colon M\to M^{\prime} its dual map, where M′=(N′)∨M^{\prime}=(N^{\prime})^{\vee}. Let Σ′\Sigma^{\prime} be a fan in Nℝ′N^{\prime}_{\mathbb{R}} such that, for each σ′∈Σ′\sigma^{\prime}\in\Sigma^{\prime} there is σ∈Σ\sigma\in\Sigma with H⁡(σ′)⊂σH(\sigma^{\prime})\subset\sigma, and let p∈XΣ,0​(K)p\in X_{\Sigma,0}(K). Then

    (4.50) φp,H∗​DΨ=DH∗​Ψ,\varphi_{p,H}^{\ast}D_{\Psi}=D_{H^{\ast}\Psi},

    and the associated polytope is

    (4.51) ΔH∗​Ψ=H∨​(ΔΨ)⊂Mℝ′.\Delta_{H^{\ast}\Psi}=H^{\vee}(\Delta_{\Psi})\subset M^{\prime}_{\mathbb{R}}.
Proof.

Equation (4.48) follows from (4.30), while equation (4.50) follows from Proposition 4.35. Then (4.49) and (4.51) follow from Proposition 3.78. ∎

As a consequence of the above construction, we can compute easily the degree of any orbit.

Corollary 4.52.

Let Σ\Sigma be a complete fan in NℝN_{\mathbb{R}}, Ψ:Nℝ→ℝ\Psi\colon N_{\mathbb{R}}\to\mathbb{R} a support function on Σ\Sigma, and σ∈Σ\sigma\in\Sigma a cone of dimension n−kn-k. Then

degDΨ⁡(V⁡(σ))=k!​volM⁡(Fσ)⁡(Fσ).\deg_{D_{\Psi}}(V(\sigma))=k!\operatorname{vol}_{M(F_{\sigma})}(F_{\sigma}).
Proof.

In view of equations (4.49) and (4.39), it is enough to prove that M⁡(σ)=M⁡(Fσ)M(\sigma)=M(F_{\sigma}). But this follows from the fact that LFσ=σ⟂L_{F_{\sigma}}=\sigma^{\perp} (see Notation 3.103). ∎

Example 4.53.

Let τ∈Σn−1\tau\in\Sigma^{n-1}. The degree of the curve V⁡(τ)V(\tau) agrees with the lattice length of FτF_{\tau}.

We will also need the toric version of the Nakai-Moishezon criterion.

Theorem 4.54.

Let XΣX_{\Sigma} be a proper toric variety and DΨD_{\Psi} a 𝕋\mathbb{T}-Cartier divisor on XΣX_{\Sigma}.

  1. (1)

    The following properties are equivalent:

    1. (a)

      DΨD_{\Psi} is ample;

    2. (b)

      (DΨ⋅C)>0(D_{\Psi}\cdot C)>0 for every curve CC in XΣX_{\Sigma};

    3. (c)

      (DΨ⋅V⁡(τ))>0(D_{\Psi}\cdot V(\tau))>0 for every τ∈Σn−1\tau\in\Sigma^{n-1}.

  2. (2)

    The following properties are equivalent:

    1. (a)

      DΨD_{\Psi} is generated by its global sections;

    2. (b)

      (DΨ⋅C)≥0(D_{\Psi}\cdot C)\geq 0 for every curve CC in XΣX_{\Sigma};

    3. (c)

      (DΨ⋅V⁡(τ))≥0(D_{\Psi}\cdot V(\tau))\geq 0 for every τ∈Σn−1\tau\in\Sigma^{n-1}.

Proof.

This follows from [Oda88, Theorem 2.18] for non singular toric varieties, and from [Mav00] for the general case. ∎

Original mathematics by the credited authors. Source-backed reader collection; mathematical self-containment is not assessed.