ScalingStacks

4. The essential skeleton of a Calabi-Yau variety [04VW]

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. The essential skeleton of a Calabi-Yau variety

4.1. The skeleton is a pseudo-manifold

(4.1.1) The case where XX is a family of Calabi-Yau varieties over CC is of particular interest; the connections with homological mirror symmetry were the main motivation for Kontsevich and Soibelman to define the skeleton in [KS06]. If ω\omega is a volume form on XKX_{K} (i.e., a nowhere vanishing differential form of maximal degree) then Sk⁡(XK)=Sk⁡(XK,ω)\mathrm{Sk}(X_{K})=\mathrm{Sk}(X_{K},\omega) by [MN13, 4.6.4]. We will now prove that the underlying topological space of the essential skeleton Sk⁡(XK)\mathrm{Sk}(X_{K}) is a pseudo-manifold with boundary; this result is implicitly contained in [KK10, Ko11].

(4.1.2) A topological space TT is called an nn-dimensional pseudo-manifold with boundary if it admits a triangulation 𝒯\mathscr{T} satisfying the following conditions:

  1. (1)

    (dimensional homogeneity) T=|𝒯|T=|\mathscr{T}| is the union of all nn-simplices.

  2. (2)

    (non-branching) Every (n−1)(n-1)-simplex is a face of precisely one or two nn-simplices.

  3. (3)

    (strong connectedness) For every pair of nn-simplices σ\sigma and σ′\sigma^{\prime} in 𝒯\mathscr{T}, there is a sequence of nn-simplices

    σ=σ0,σ1,…,σℓ=σ′\sigma=\sigma_{0},\sigma_{1},\ldots,\sigma_{\ell}=\sigma^{\prime}

    such that the intersection σi∩σi+1\sigma_{i}\cap\sigma_{i+1} is an (n−1)(n-1)-simplex for all ii.

We say that TT is a closed pseudo-manifold if we can replace condition (2) by the property that every (n−1)(n-1)-simplex is a face of precisely two nn-simplices. A typical example of a 2-dimensional closed pseudo-manifold which is not a manifold is the pinched torus.

(4.1.3) For the reader’s convenience, we include some basic facts about adjunction for d​l​tdlt-pairs. We refer to Chapter 4 of [Ko13] for more background. Let (Y,Δ)(Y,\Delta) be a d​l​tdlt-pair over kk, and let DD be a log canonical center of (Y,Δ)(Y,\Delta). Then DD is normal, by [Ko13, 4.16]. There is a well defined ℚ\mathbb{Q}-divisor ΔD\Delta_{D} on DD, called the different of Δ\Delta on DD [Ko13, 4.18], which is induced by the Poincaré map and satisfies the equation

(KY+Δ)|D=KD+ΔD.(K_{Y}+\Delta)|_{D}=K_{D}+\Delta_{D}.

In the sequel, whenever we write such an equation it will be understood that ΔD\Delta_{D} is the different of Δ\Delta on DD. The pair (D,ΔD)(D,\Delta_{D}) is again a d​l​tdlt-pair, by [Ko13, 4.19]. Write ⌊Δ⌋=∑i∈IDi\lfloor\Delta\rfloor=\sum_{i\in I}D_{i}. If JJ is a subset of II and DD is a component of ⋂j∈JΔj\bigcap_{j\in J}{\Delta}_{j}, it is not hard to see that for every non-empty subset J′J^{\prime} of I∖JI\setminus J, every irreducible component of the intersection

D∩⋂j∈J′ΔjD\cap\bigcap_{j\in J^{\prime}}\Delta_{j}

is a log canonical center of (D,ΔD)(D,\Delta_{D}) (see [Ko13, 4.19]). Conversely, by repeatedly using inversion of adjunction [Ko13, 4.9], one sees that any log canonical center of (D,ΔD)(D,\Delta_{D}) is a log canonical center of (Y,Δ)(Y,\Delta), and thus an irreducible component of an intersection D∩⋂j∈J′ΔjD\cap\bigcap_{j\in J^{\prime}}\Delta_{j} for some non-empty subset J′J^{\prime} of I∖JI\setminus J.

Theorem 4.1.4.

Assume that KXK_{X} is ℚ\mathbb{Q}-linearly equivalent to 00 over CC. Then the underlying topological space of Sk⁡(XK)\mathrm{Sk}(X_{K}) is a a pseudo-manifold with boundary.

Proof.

As we mentioned above, this result is essentially contained in [KK10, Ko11]. Using the terminology there, properties (1)-(3) of a pseudo-manifold all follow from the fact that two minimal log canonical centers of a log crepant structure are ℙ1\mathbb{P}^{1}-linked in the sense of Definition 9 in [Ko11]. We will now explain this in more detail. We denote by nn the relative dimension of XX over CC.

By Theorem 2.2.6(1), there exists a a good minimal d​l​tdlt-model 𝒳\mathscr{X} of XX over 𝒞\mathscr{C}. By Theorem 3.3.4, we have Sk⁡(XK)=Sk⁡(𝒳)\mathrm{Sk}(X_{K})=\mathrm{Sk}(\mathscr{X}). As a triangulation on Sk⁡(𝒳)\mathrm{Sk}(\mathscr{X}), we take the first barycentric subdivision of the simplicial structure on Sk⁡(𝒳)\mathrm{Sk}(\mathscr{X}). This barycentric subdivision is necessary to guarantee that the intersection of two faces is a codimension one face of both, rather than a union of faces (think of a type I2I_{2} degeneration of elliptic curves, whose skeleton consists of two vertices joined by two edges).

We choose an integer m>0m>0 such that m​KX∼0mK_{X}\sim 0. Since the divisor m​K𝒳+m​(𝒳s)redmK_{\mathscr{X}}+m(\mathscr{X}_{s})_{\mathrm{red}} is semi-ample over 𝒞\mathscr{C} and trivial over CC, we see that m​K𝒳+m​(𝒳s)redmK_{\mathscr{X}}+m(\mathscr{X}_{s})_{\mathrm{red}} must be a multiple of 𝒳s\mathscr{X}_{s} and thus trivial over 𝒞\mathscr{C}. Thus we can apply Theorem 10 in [Ko11] to the d​l​tdlt-pair (𝒳,(𝒳s)red)(\mathscr{X},(\mathscr{X}_{s})_{\mathrm{red}}) over 𝒞\mathscr{C}. It states that every two minimal log canonical centers DD and D∗D^{*} of (𝒳,(𝒳s)red)(\mathscr{X},(\mathscr{X}_{s})_{\mathrm{red}}) are ℙ1\mathbb{P}^{1}-linked. This means, in particular, that they have the same dimension, say n−dn-d, and that there exist a sequence of (n−d+1)(n-d+1)-dimensional log canonical centers E1,E2,…,EℓE_{1},E_{2},\ldots,E_{\ell} and a sequence of (n−d)(n-d)-dimensional log canonical centers D=D0,D1,…,Dℓ=D∗D=D_{0},D_{1},\ldots,D_{\ell}=D^{*} such that Di−1,Di⊂EiD_{i-1},D_{i}\subset E_{i} for 1≤i≤ℓ1\leq i\leq\ell. In this way, we obtain properties (1) and (3) of a pseudo-manifold with boundary.

If we have two minimal log canonical centers D1,D2D_{1},\,D_{2} of (𝒳,(𝒳s)red)(\mathscr{X},(\mathscr{X}_{s})_{\mathrm{red}}), contained in an (n−d+1)(n-d+1)-dimensional log canonical center EE, and if we write

(K𝒳+(𝒳s)red)|E=KE+D1+D2+Δ(K_{\mathscr{X}}+(\mathscr{X}_{s})_{\mathrm{red}})|_{E}=K_{E}+D_{1}+D_{2}+\Delta

for some Δ≥0\Delta\geq 0, then (E,D1+D2+Δ)(E,D_{1}+D_{2}+\Delta) is again a d​l​tdlt-pair [Ko13, 4.19]. Moreover, D1D_{1} cannot intersect D2D_{2} or ⌊Δ⌋\lfloor\Delta\rfloor because the intersection would be a union of log canonical centers of (𝒳,(𝒳s)red)(\mathscr{X},(\mathscr{X}_{s})_{\mathrm{red}}), which contradicts the minimality of D1D_{1}. Thus we are in the situation of the second part of the proof of Theorem 10 in [Ko11]. That proof shows that D1D_{1} and D2D_{2} are the only log canonical centers of (E,D1+D2+Δ)(E,D_{1}+D_{2}+\Delta). Property (2) follows. ∎

(4.1.5) We can say more in the case where Sk⁡(𝒳)\mathrm{Sk}(\mathscr{X}) has maximal dimension, that is, dimension equal to n=dim(XK)n=\dim(X_{K}). First, we need a lemma.

Lemma 4.1.6.

Let (Z,Δ=∑i=1jΔi)(Z,\Delta=\sum^{j}_{i=1}\Delta_{i}) be a reduced d​l​tdlt-pair over kk such that KZ+ΔK_{Z}+\Delta is Cartier. Let DD be a log canonical center and let Δ1,…,Δℓ\Delta_{1},\ldots,\Delta_{\ell} be the irreducible components of Δ\Delta that contain DD. Let UU be the maximal open subset of ZZ where ZZ is smooth and Δ\Delta is a divisor with strict normal crossings. If we write (KZ+Δ)|D=KD+ΔD(K_{Z}+\Delta)|_{D}=K_{D}+\Delta_{D}, then ΔD\Delta_{D} is equal to the closure of the restriction of ∑i=ℓ+1jΔi|U\sum^{j}_{i=\ell+1}\Delta_{i}|_{U} to U∩DU\cap D.

Proof.

We first notice that in the above statement, UU can be replaced by any smaller open set that meets all the log canonical centers: the closure of the restriction of ∑i=ℓ+1jΔi|U\sum^{j}_{i=\ell+1}\Delta_{i}|_{U} to U∩DU\cap D will yield the same divisor on DD.

Then by induction, we only need to treat the case where DD is a component of Δ\Delta, say Δ1\Delta_{1}. If we take a log resolution f:Y→(Z,Δ)f:Y\to(Z,\Delta) and let D′=Δ1′D^{\prime}=\Delta^{\prime}_{1} be the birational transform of DD, then ΔD\Delta_{D} can be computed as follows: if we write f∗​(KZ+Δ)|D′=KD′+ΔD′f^{*}(K_{Z}+\Delta)|_{D^{\prime}}=K_{D^{\prime}}+\Delta_{D^{\prime}} then ΔD=(f|D′)∗​(ΔD′)\Delta_{D}=(f|_{D^{\prime}})_{*}(\Delta_{D^{\prime}}). In particular, as KZ+ΔK_{Z}+\Delta is Cartier, we know that ΔD\Delta_{D} is a integral divisor. Since it is effective, and all the components of ΔD\Delta_{D} are log canonical centers of (X,Δ)(X,\Delta), we see that ΔD\Delta_{D} must be equal to the closure of the restriction of ∑i=2jΔi|U\sum^{j}_{i=2}\Delta_{i}|_{U}. ∎

Theorem 4.1.7.

Assume that kk is algebraically closed, KXK_{X} is trivial over CC and XX has an s​n​csnc-model 𝒴\mathscr{Y} with reduced special fiber 𝒴s\mathscr{Y}_{s}. Assume moreover that Sk⁡(XK)\mathrm{Sk}(X_{K}) is of dimension n=dim(XK)n=\dim(X_{K}). Then Sk⁡(XK)\mathrm{Sk}(X_{K}) is an nn-dimensional closed pseudo-manifold.

Proof.

By running MMP for 𝒴\mathscr{Y} over 𝒞\mathscr{C}, we know that XX has a good minimal dlt model 𝒳\mathscr{X} with reduced special fiber (see [Fu11] or [HX13]). Then one sees as in the proof of Theorem 4.1.4 that K𝒳K_{\mathscr{X}} is trivial over 𝒞\mathscr{C}. Our assumption on the dimension of Sk⁡(XK)\mathrm{Sk}(X_{K}) implies that the minimal log canonical centers of (𝒳,𝒳s)(\mathscr{X},\mathscr{X}_{s}) are points. Let DD be a one-dimensional log canonical center, and let DiD_{i} (1≤i≤ℓ1\leq i\leq\ell) be the 0-dimensional log canonical centers contained in DD. From Lemma 4.1.6, we know that

(K𝒳+𝒳s)|D=KD+∑i=1ℓDi∼0.(K_{\mathscr{X}}+\mathscr{X}_{s})|_{D}=K_{D}+\sum^{\ell}_{i=1}D_{i}\sim 0.

Thus DD is a rational curve and ℓ=2\ell=2, which means that Sk⁡(XK)\mathrm{Sk}(X_{K}) is closed. ∎

(4.1.8) In Theorem 4.1.7, the condition that Sk⁡(X)\mathrm{Sk}(X) has maximal dimension can not be omitted; for instance, there are examples of semi-stable degenerations of K3-surfaces with trivial relative canonical sheaf where the special fiber is a chain of surfaces, so that the skeleton is homeomorphic to a closed interval. We will now give an interpretation of this condition in terms of the monodromy around s∈𝒞s\in\mathscr{C}.

Lemma 4.1.9.

Let YY be a connected smooth and proper KK-variety and let ω\omega be a non-zero mm-pluricanonical form on YY, for some m>0m>0. Let K′K^{\prime} be a finite extension of KK, set Y′=Y×KK′Y^{\prime}=Y\times_{K}K^{\prime} and denote by ω′\omega^{\prime} the pullback of ω\omega to Y′Y^{\prime}. Then the skeleton Sk⁡(Y,ω)\mathrm{Sk}(Y,\omega) is the image of Sk⁡(Y′,ω′)\mathrm{Sk}(Y^{\prime},\omega^{\prime}) under the projection morphism π:(Y′)an→Yan\pi:(Y^{\prime})^{\mathrm{an}}\to Y^{\mathrm{an}}.

Proof.

We may assume that K′K^{\prime} is Galois over KK. Let dd be the ramification index of K′K^{\prime} over KK. We will prove that

wtω′​(y)=d⋅wtω​(π⁡(y))−d+1\mathrm{wt}_{\omega^{\prime}}(y)=d\cdot\mathrm{wt}_{\omega}(\pi(y))-d+1

for every divisorial point yy on (Y′)an(Y^{\prime})^{\mathrm{an}} (see [MN13, 2.4.10] for the notion of divisorial point). This immediately implies the statement in the lemma, since Sk⁡(X,ω)\mathrm{Sk}(X,\omega) is the closure of the set of divisorial points where the weight function reaches its minimal value [MN13, 4.5.1].

We denote by R′R^{\prime} the integral closure of RR in K′K^{\prime}. Let 𝒴′\mathscr{Y}^{\prime} be a regular separated R′R^{\prime}-scheme of finite type with irreducible special fiber 𝒴k′\mathscr{Y}^{\prime}_{k}, endowed with an isomorphism of K′K^{\prime}-schemes 𝒴K′′→Y′\mathscr{Y}^{\prime}_{K^{\prime}}\to Y^{\prime}. Let yy be the unique point in red𝒴′−1​(ξ)\mathrm{red}_{\mathscr{Y}^{\prime}}^{-1}(\xi), where ξ\xi denotes the generic point of 𝒴k′\mathscr{Y}^{\prime}_{k}. Removing a closed subset of 𝒴k′\mathscr{Y}^{\prime}_{k} if necessary, we can find a regular separated RR-scheme of finite type 𝒴\mathscr{Y} and an isomorphism 𝒴K→Y\mathscr{Y}_{K}\to Y such that 𝒴′\mathscr{Y}^{\prime} is an open subscheme of the normalization of 𝒴×RR′\mathscr{Y}\times_{R}R^{\prime}. Then red𝒴​(π​(y))\mathrm{red}_{\mathscr{Y}}(\pi(y)) is a generic point of 𝒴k\mathscr{Y}_{k}.

If we use the notations from (3.2) and denote by (S′)+(S^{\prime})^{+} the log scheme associated to R′∖{0}→R′R^{\prime}\setminus\{0\}\to R^{\prime}, then the (S′)+(S^{\prime})^{+}-log scheme (𝒴′)+(\mathscr{Y}^{\prime})^{+} is isomorphic to an open log subscheme of the f​sfs base change of 𝒴+\mathscr{Y}^{+} from S+S^{+} to (S′)+(S^{\prime})^{+}. Since log differentials are compatible with f​sfs base change, we can deduce from the description of the weight function in (3.2) that

wtω′​(y)=d⋅wtω​(π⁡(y))−d+1\mathrm{wt}_{\omega^{\prime}}(y)=d\cdot\mathrm{wt}_{\omega}(\pi(y))-d+1

(the scaling factor dd is caused by the renormalization of the discrete valuation on K′K^{\prime}). ∎

Theorem 4.1.10.

Assume that k=ℂk=\mathbb{C} and denote by nn the relative dimension of XX over CC. Suppose that XX is projective over CC and that KXK_{X} is trivial over CC. Let FF be a general fiber of the morphism X→CX\to C. Then Sk⁡(XK)\mathrm{Sk}(X_{K}) has dimension nn if and only if the monodromy transformation around s∈𝒞s\in\mathscr{C} on Hn​(F​(ℂ),ℚ)H^{n}(F(\mathbb{C}),\mathbb{Q}) has a Jordan block of size n+1n+1. If this holds, and hi,0​(F)=0h^{i,0}(F)=0 for 0<i<n0<i<n, then Sk⁡(XK)\mathrm{Sk}(X_{K}) is a ℚ\mathbb{Q}-homology sphere.

Proof.

By Lemma 4.1.9 and the Semi-Stable Reduction Theorem we can assume that XX has a projective s​n​csnc-model 𝒴\mathscr{Y} over 𝒞\mathscr{C} such that 𝒴s\mathscr{Y}_{s} is reduced. For every integer i≥0i\geq 0, we denote by

𝐇i=ℍi​(𝒴s,R​ψ𝒴​(ℤ))≅Hi​(F⁡(ℂ),ℤ)\mathbf{H}^{i}=\mathbb{H}^{i}(\mathscr{Y}_{s},R\psi_{\mathscr{Y}}(\mathbb{Z}))\cong H^{i}(F(\mathbb{C}),\mathbb{Z})

the degree ii nearby cohomology of 𝒴\mathscr{Y} at ss; here R​ψ𝒴​(ℤ)R\psi_{\mathscr{Y}}(\mathbb{Z}) denotes the complex of nearby cycles with ℤ\mathbb{Z}-coefficients associated to 𝒴\mathscr{Y}. By [St76], the spaces 𝐇i\mathbf{H}^{i} carry a canonical mixed Hodge structure, whose weight filtration coincides with the monodromy filtration. In particular, there exists a Jordan block of monodromy of size n+1n+1 on 𝐇ℚn\mathbf{H}^{n}_{\mathbb{Q}} if and only if W0​𝐇ℚn≠0W_{0}\mathbf{H}^{n}_{\mathbb{Q}}\neq 0.

By [Be09, 5.1] and its proof, the ℚ\mathbb{Q}-vector space W0​𝐇ℚiW_{0}\mathbf{H}^{i}_{\mathbb{Q}} is canonically isomorphic to the degree ii singular cohomology of XKanX_{K}^{\mathrm{an}}, for every i≥0i\geq 0. Since XKanX_{K}^{\mathrm{an}} is homotopy equivalent to Sk⁡(XK)\mathrm{Sk}(X_{K}) by Corollary 3.3.6, we see that W0​𝐇ℚnW_{0}\mathbf{H}^{n}_{\mathbb{Q}} can only be different from zero if the dimension of Sk⁡(XK)\mathrm{Sk}(X_{K}) is equal to nn. We will now prove the converse implication. Suppose that Sk⁡(XK)\mathrm{Sk}(X_{K}) has dimension nn and let ω\omega be a relative volume form on XX over CC such that ω\omega extends to a global section of ω𝒴/𝒞​(log⁡𝒴s)\omega_{\mathscr{Y}/\mathscr{C}}(\log\mathscr{Y}_{s}) that generates ω𝒴/𝒞​(log⁡𝒴s)\omega_{\mathscr{Y}/\mathscr{C}}(\log\mathscr{Y}_{s}) at at least one generic point of 𝒴s\mathscr{Y}_{s} (modulo shrinking 𝒞\mathscr{C}, such ω\omega always exists). Then it follows from [MN13, 4.5.5] that Sk⁡(XK)\mathrm{Sk}(X_{K}) is the simplicial subspace of Sk⁡(𝒴)\mathrm{Sk}(\mathscr{Y}) spanned by the vertices corresponding to the irreducible components EE of 𝒴s\mathscr{Y}_{s} such that ω\omega generates ω𝒴/𝒞​(log⁡𝒴s)\omega_{\mathscr{Y}/\mathscr{C}}(\log\mathscr{Y}_{s}) at the generic point of EE. Since Sk⁡(XK)\mathrm{Sk}(X_{K}) has dimension nn, we can find such components E1,…,EnE_{1},\ldots,E_{n} that intersect in a point. Denote by DD the union of nn-fold intersection points of components of 𝒴s\mathscr{Y}_{s}. Then by reduction modulo tt, ω\omega induces an element of

H0​(𝒴s,ω𝒴/𝒞​(log⁡𝒴s)⊗𝒪𝒴s)H^{0}(\mathscr{Y}_{s},\omega_{\mathscr{Y}/\mathscr{C}}(\log\mathscr{Y}_{s})\otimes\mathcal{O}_{\mathscr{Y}_{s}})

whose image under the Poincaré residue map

ℛ:H0​(𝒴s,ω𝒴/𝒞​(log⁡𝒴s)⊗𝒪𝒴s)→H0​(𝒴s,Gr−nW​(ω𝒴/𝒞​(log⁡𝒴s)⊗𝒪𝒴s))≅H0​(D,𝒪D)\mathcal{R}:H^{0}(\mathscr{Y}_{s},\omega_{\mathscr{Y}/\mathscr{C}}(\log\mathscr{Y}_{s})\otimes\mathcal{O}_{\mathscr{Y}_{s}})\to H^{0}(\mathscr{Y}_{s},\mathrm{Gr}_{-n}^{W}(\omega_{\mathscr{Y}/\mathscr{C}}(\log\mathscr{Y}_{s})\otimes\mathcal{O}_{\mathscr{Y}_{s}}))\cong H^{0}(D,\mathcal{O}_{D})

is different from zero. However, by the degeneration of the Hodge and weight spectral sequences, the image of ℛ\mathcal{R} injects into W0​𝐇ℂnW_{0}\mathbf{H}^{n}_{\mathbb{C}}. Thus W0​𝐇ℚnW_{0}\mathbf{H}^{n}_{\mathbb{Q}} is non-trivial.

Finally, assume that Sk⁡(XK)\mathrm{Sk}(X_{K}) has dimension nn and that hi,0​(Xgen)=0h^{i,0}(X_{\mathrm{gen}})=0 for 0<i<n0<i<n. Then

GrF0​𝐇ℂi≅Hi​(𝒴s,𝒪𝒴s)=0\mathrm{Gr}_{F}^{0}\mathbf{H}^{i}_{\mathbb{C}}\cong H^{i}(\mathscr{Y}_{s},\mathcal{O}_{\mathscr{Y}_{s}})=0

for 0<i<n0<i<n and

GrF0​𝐇ℂi≅Hi​(𝒴s,𝒪𝒴s)≅ℂ\mathrm{Gr}_{F}^{0}\mathbf{H}^{i}_{\mathbb{C}}\cong H^{i}(\mathscr{Y}_{s},\mathcal{O}_{\mathscr{Y}_{s}})\cong\mathbb{C}

for i=0,ni=0,n by the degeneration of the Hodge spectral sequence for the limit mixed Hodge structure. Thus W0​𝐇ℚi=0W_{0}\mathbf{H}^{i}_{\mathbb{Q}}=0 for 0<i<n0<i<n, W0​𝐇ℚ0≅ℚW_{0}\mathbf{H}^{0}_{\mathbb{Q}}\cong\mathbb{Q} and W0​𝐇ℚnW_{0}\mathbf{H}^{n}_{\mathbb{Q}} has dimension at most one; it must have dimension one since we have already proven that it is non-zero. It follows that Sk⁡(XK)\mathrm{Sk}(X_{K}) is a ℚ\mathbb{Q}-homology sphere. ∎

4.2. Removing the algebraicity condition

(4.2.1) In this section, we will extend Theorems 3.2.8, 4.1.4, 4.1.7 and 4.1.10 to the case where XX is a Calabi-Yau variety over K=k⁡((t))K=k(\negthinspace(t)\negthinspace) instead of over the curve CC. The crucial point is that the skeleton of a Calabi-Yau variety can be computed from the logarithmic structure on the special fiber of any s​n​csnc-model.

Proposition 4.2.2.

Let 𝒴\mathscr{Y} be a connected regular flat proper RR-scheme such that 𝒴k\mathscr{Y}_{k} is a strict normal crossings divisor. Then for every connected flat proper RR-scheme 𝒵\mathscr{Z} and every isomorphism of R/(t2)R/(t^{2})-schemes

f:𝒴×RR/(t2)→𝒵×RR/(t2),f:\mathscr{Y}\times_{R}R/(t^{2})\to\mathscr{Z}\times_{R}R/(t^{2}),

the following properties hold.

  1. (1)

    The scheme 𝒵\mathscr{Z} is regular and 𝒵k\mathscr{Z}_{k} is a divisor with strict normal crossings.

  2. (2)

    Denote by S+S^{+} the log scheme associated to R∖{0}→RR\setminus\{0\}\to R and by 𝒴+\mathscr{Y}^{+} and 𝒵+\mathscr{Z}^{+} the schemes 𝒴\mathscr{Y} and 𝒵\mathscr{Z} endowed with the divisorial log structures associated to their special fibers. For every integer d>0d>0 we denote by sd+s^{+}_{d} the standard log point (Spec​k,k∗⊕ℕ)(\mathrm{Spec}\,k,k^{*}\oplus\mathbb{N}) viewed as a log scheme over S+S^{+} via the morphism of charts ℕ→ℕ:n↦d​n\mathbb{N}\to\mathbb{N}:n\mapsto dn. If we denote by ee the least common multiple of the multiplicities of the components of 𝒴k\mathscr{Y}_{k}, then there exists an isomorphism of log schemes

    g:𝒴+×S+se+→𝒵+×S+se+g:\mathscr{Y}^{+}\times_{S^{+}}s^{+}_{e}\to\mathscr{Z}^{+}\times_{S^{+}}s^{+}_{e}

    over se+s^{+}_{e}, such that gg is compatible with the reduction of ff modulo tt (meaning that the obvious square in the category of kk-schemes commutes).

Proof.

It is easy to see that (1) holds, since we can detect regularity by looking at the dimensions of the Zariski tangent spaces at the points of

𝒴×RR/(t2)≅𝒵×RR/(t2).\mathscr{Y}\times_{R}R/(t^{2})\cong\mathscr{Z}\times_{R}R/(t^{2}).

Moreover, the special fibers of 𝒴\mathscr{Y} and 𝒵\mathscr{Z} are isomorphic so that 𝒵k\mathscr{Z}_{k} is a divisor with strict normal crossings. Point (2) is more subtle and follows from [Ki03, 2.6(2)]. ∎

Proposition 4.2.3.

Let 𝒴\mathscr{Y} be a connected regular flat proper RR-scheme such that 𝒴K\mathscr{Y}_{K} has trivial canonical sheaf and 𝒴k\mathscr{Y}_{k} is a strict normal crossings divisor. Then the skeleta Sk⁡(𝒴)\mathrm{Sk}(\mathscr{Y}) and Sk⁡(𝒴K)\mathrm{Sk}(\mathscr{Y}_{K}) only depend on 𝒴×RR/(t2),\mathscr{Y}\times_{R}R/(t^{2}), in the following sense. Assume that 𝒵\mathscr{Z} is a regular flat proper RR-scheme such that 𝒵K\mathscr{Z}_{K} has trivial canonical sheaf and there exists an isomorphism of R/(t2)R/(t^{2})-schemes

f:𝒴×RR/(t2)→𝒵×RR/(t2).f:\mathscr{Y}\times_{R}R/(t^{2})\to\mathscr{Z}\times_{R}R/(t^{2}).

Then there exists an isomorphism of simplicial spaces Sk⁡(𝒴)→Sk⁡(𝒵)\mathrm{Sk}(\mathscr{Y})\to\mathrm{Sk}(\mathscr{Z}) that maps Sk⁡(𝒴K)\mathrm{Sk}(\mathscr{Y}_{K}) onto Sk⁡(𝒵K)\mathrm{Sk}(\mathscr{Z}_{K}).

Proof.

Reducing ff modulo tt, we obtain an isomorphism of kk-schemes 𝒴k→𝒵k\mathscr{Y}_{k}\to\mathscr{Z}_{k} and, by taking the dual intersection complexes, an isomorphism of simplicial spaces with piecewise ℤ\mathbb{Z}-affine structure Sk⁡(𝒴)→Sk⁡(𝒵)\mathrm{Sk}(\mathscr{Y})\to\mathrm{Sk}(\mathscr{Z}). We will prove that this isomorphism maps Sk⁡(𝒴K)\mathrm{Sk}(\mathscr{Y}_{K}) onto Sk⁡(𝒵K)\mathrm{Sk}(\mathscr{Z}_{K}).

We use the notations from Proposition 4.2.2(2) and we set s+=s1+s^{+}=s_{1}^{+}. We denote by 𝒴k+\mathscr{Y}^{+}_{k} the log scheme 𝒴+×S+s+\mathscr{Y}^{+}\times_{S^{+}}s^{+} obtained by restricting the log structure on 𝒴+\mathscr{Y}^{+} to the special fiber 𝒴k\mathscr{Y}_{k} of 𝒴\mathscr{Y}. It follows from [IKN05, 7.1] that

Ω:=H0​(𝒴,ω𝒴+/S+)\Omega:=H^{0}(\mathscr{Y},\omega_{\mathscr{Y}^{+}/S^{+}})

is a free RR-module of rank one and that the reduction map

Ω⊗Rk→Ωk:=H0​(𝒴k,ω𝒴k+/s+)\Omega\otimes_{R}k\to\Omega_{k}:=H^{0}(\mathscr{Y}_{k},\omega_{\mathscr{Y}^{+}_{k}/s^{+}})

is an isomorphism. Let ω\omega be a generator of the RR-module Ω\Omega and denote by ωk\omega_{k} its image in Ωk\Omega_{k}. By (3.2), the generic point ξ\xi of an irreducible component EE of 𝒴k\mathscr{Y}_{k} is ω\omega-essential in the sense of [MN13, 4.5.4] if and only if ωk\omega_{k} generates ω𝒴k+/s+\omega_{\mathscr{Y}^{+}_{k}/s^{+}} at the point ξ\xi. Moreover, the skeleton Sk⁡(𝒴K)=Sk⁡(𝒴K,ω)\mathrm{Sk}(\mathscr{Y}_{K})=\mathrm{Sk}(\mathscr{Y}_{K},\omega) is the simplicial subspace of Sk⁡(𝒴)\mathrm{Sk}(\mathscr{Y}) spanned by the vertices corresponding to such points ξ\xi [MN13, 4.5.5]. However, for every integer d>0d>0, the stalk of ω𝒴k+/s+\omega_{\mathscr{Y}^{+}_{k}/s^{+}} at ξ\xi is generated by global sections if and only if ω𝒴+×S+sd+/sd+\omega_{\mathscr{Y}^{+}\times_{S^{+}}s_{d}^{+}/s_{d}^{+}} is generated by global sections at any point lying above ξ\xi, by the base change property in [IKN05, 7.1]. The analogous statements hold for 𝒵\mathscr{Z}. Thus it follows from Proposition 4.2.2(2) that the isomorphism Sk⁡(𝒴)→Sk⁡(𝒵)\mathrm{Sk}(\mathscr{Y})\to\mathrm{Sk}(\mathscr{Z}) maps Sk⁡(𝒴K)\mathrm{Sk}(\mathscr{Y}_{K}) onto Sk⁡(𝒵K)\mathrm{Sk}(\mathscr{Z}_{K}). ∎

Theorem 4.2.4.

Let XX be a geometrically connected, smooth and proper KK-variety with trivial canonical sheaf. Then the following properties hold.

  1. (1)

    The essential skeleton Sk⁡(X)\mathrm{Sk}(X) is a strong deformation retract of XanX^{\mathrm{an}}.

  2. (2)

    If 𝒳\mathscr{X} is a proper s​n​csnc-model of XX over RR, then Sk⁡(X)\mathrm{Sk}(X) is contained in Sk⁡(𝒳)\mathrm{Sk}(\mathscr{X}) and can be obtained from Sk⁡(𝒳)\mathrm{Sk}(\mathscr{X}) (as a topological subspace of Sk⁡(𝒳)\mathrm{Sk}(\mathscr{X}) with piecewise affine structure) by a finite number of elementary collapses.

  3. (3)

    The essential skeleton Sk⁡(X)\mathrm{Sk}(X) is a pseudo-manifold with boundary. If kk is algebraically closed and Sk⁡(X)\mathrm{Sk}(X) has dimension dim⁡(X)\mathrm{dim}(X), then it is a closed pseudo-manifold.

  4. (4)

    Assume that kk is algebraically closed and XX is projective. Let σ\sigma be a topological generator of the absolute Galois group G⁡(Ka/K)G(K^{a}/K) and let ℓ\ell be a prime. Then Sk⁡(XK)\mathrm{Sk}(X_{K}) has dimension n=dim(X)n=\dim(X) if and only if the action of σ\sigma on

    He´​tn​(X×KKa,ℚℓ)H^{n}_{\mathrm{\acute{e}t}}(X\times_{K}K^{a},\mathbb{Q}_{\ell})

    has a Jordan block of size n+1n+1. If this holds, and hi,0​(X)=0h^{i,0}(X)=0 for 0<i<n0<i<n, then Sk⁡(XK)\mathrm{Sk}(X_{K}) is a ℚ\mathbb{Q}-homology sphere.

Proof.

Let 𝒳\mathscr{X} be a proper s​n​csnc-model of XX over RR. By a standard argument based on spreading out and Greenberg Approximation (as explained in [MN13, 5.1.2], for instance) we can find a connected smooth kk-curve 𝒞\mathscr{C}, a kk-rational point ss on 𝒞\mathscr{C}, a uniformizer tt in 𝒪𝒞,s\mathcal{O}_{\mathscr{C},s} and a smooth and proper 𝒞\mathscr{C}-scheme 𝒳′\mathscr{X}^{\prime} with geometrically connected fibers such that there exists an isomorphism

𝒳×RR/(t2)→𝒳′×𝒞Spec​𝒪𝒞,s/(t2)\mathscr{X}\times_{R}R/(t^{2})\to\mathscr{X}^{\prime}\times_{\mathscr{C}}\mathrm{Spec}\,\mathcal{O}_{\mathscr{C},s}/(t^{2})

over R/(t2)≅O𝒞,s/(t2)R/(t^{2})\cong{O}_{\mathscr{C},s}/(t^{2}). Inspecting the proof of [MN13, 5.1.2], we see that we can also assume that the relative canonical sheaf of 𝒳′\mathscr{X}^{\prime} is trivial over C=𝒞∖{s}C=\mathscr{C}\setminus\{s\} (if the generic fiber of a smooth and proper family over an integral scheme has trivial canonical sheaf, then this holds for all fibers over some dense open subscheme of the base).

By (3.1) we know that Sk⁡(𝒳)\mathrm{Sk}(\mathscr{X}) is a strong deformation retract of XKanX_{K}^{\mathrm{an}}. Thus by Proposition 4.2.3, it suffices to prove assertions (1)–(3) for 𝒳′×𝒞Spec​K\mathscr{X}^{\prime}\times_{\mathscr{C}}\mathrm{Spec}\,K instead of XX. In this case, they follow from Corollary 3.3.6 and Theorems 3.2.8, 3.3.4, 4.1.4 and 4.1.7.

It remains to prove (4). Invoking the Lefschetz Principle, we may assume that k=ℂk=\mathbb{C}. Taking for 𝒳\mathscr{X} a projective s​n​csnc-model over RR, we can arrange that 𝒳′\mathscr{X}^{\prime} is projective over 𝒞\mathscr{C}. By Proposition 4.2.2 and the theory of logarithmic nearby cycles [Na98, 3.3] the action of σ\sigma on

He´​tn​(X×KKa,ℚℓ)H^{n}_{\mathrm{\acute{e}t}}(X\times_{K}K^{a},\mathbb{Q}_{\ell})

has a Jordan block of size n+1n+1 if and only if the corresponding statement holds for 𝒳K′\mathscr{X}^{\prime}_{K}. By Deligne’s comparison theorem for étale and complex analytic nearby cycles in [SGA7b, Exp.XIV], it is also equivalent to the property that the monodromy action on the degree nn singular cohomology of a general fiber of 𝒳′\mathscr{X}^{\prime} has a Jordan block of size n+1n+1. If hi,0​(X)=0h^{i,0}(X)=0 for 0<i<n0<i<n, then we can assume that this also holds for a general fiber of 𝒳′\mathscr{X}^{\prime}, by the proof of [MN13, 5.1.2] (if this property is satisfied by the generic fiber of a smooth and proper family over an integral scheme, then it holds for all fibers over a dense open subscheme of the base, by semi-continuity). Thus the assertion (4) follows from Theorem 4.1.10. ∎

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