This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Complexes, Residues and obstructions
for log-symplectic manifolds

Ziv Ran
UC Math Dept.
Big Springs Road Surge Facility
Riverside CA 92521 US
ziv.ran @ ucr.edu
http://math.ucr.edu/~ziv/
(Date: \DTMnow)
Abstract.

We consider compact Kählerian manifolds XX of even dimension 4 or more, endowed with a log-symplectic structure Φ\Phi, a generically nondegenerate closed 2-form with simple poles on a divisor DD with local normal crossings. A simple linear inequality involving the iterated Poincaré residues of Φ\Phi at components of the double locus of DD ensures that the pair (X,Φ)(X,\Phi) has unobstructed deformations and that DD deforms locally trivially.

Key words and phrases:
Poisson structure, log-symplectic manifold, deformation theory, log complex, mixed Hodge theory
2010 Mathematics Subject Classification:
14J40, 32G07, 32J27, 53D17

Data availaibility statement

There is no data set associated with this paper.

Introduction

A log-symplectic manifold is a pair consisting of a complex manifold XX, usually compact and Kählerian, together with a log-symplectic structure. A log-symplectic structure can be defined either as a generically nongegenerate meromorphic closed 2-form Φ\Phi with normal-crossing (anticanonical) polar divisor DD, or equivalently as a generically nondegenerate holomorphic tangential 2-vector Π\Pi such that [Π,Π]=0[\Pi,\Pi]=0 with normal-crossing degeneracy divisor DD. The two structures are related via Π=Φ1\Pi=\Phi^{-1}. See [4] or [11] or [3] or [12] for basic facts on Poisson and log-symplectic manifolds and [5] (especially the appendix), [6], [1], [8] or [10], and references therein, for deformations.

Understanding log-symplectic manifolds unavoidably involves understanding their deformations. In the very special case of symplectic manifolds, where D=0D=0, the classical theorem of Bogomolov [2] shows that the pair (X,Φ)(X,\Phi) has unobstructed deformations. In [14] we obtained a generalization of this result which holds when Φ\Phi satisfied a certain ’very general position’ condition with respect to DD (the original statement is corrected in the subsequent erratum/corrigendum). Namely, we showed in this case that (X,Φ)(X,\Phi) has ’strongly unobstructed’ deformations, in the sense that it has unobstructed deformations and DD deforms locally trivially.

Further results on unobstructed deformations (in the sense of Hitchin’s generalized geometry [7]) and Torelli theorems in the case where DD has global normal crossings were obtained by Matviichuk, Pym and Schedler [9], based on their notion of holonomicity.

Our purpose here is to prove a more precise strong unobstructedness result compared to [14], nailing down the generality required: we will show in Theorem 6 that strong unobstructedness can fail only when the log-symplectic structure Φ\Phi, more precisely its (iterated Poincaré) residues at codimension-2 strata of the polar divisor DD (which are essentially the (locally constant) coefficients of Φ\Phi with respect to a suitable basis of the log forms adapted to DD) satisfy certain special linear relations with integer coefficients. Explicitly, at a triple point of DD with branches labelled 1,2,3 and associated residues c12,c23,c31c_{12},c_{23},c_{31}, the condition is

c23+c31c12.c_{23}+c_{31}\in\mathbb{N}c_{12}.

Essentially, if this never happens over the entire triple locus then (X,Φ)(X,\Phi) has strongly unobstructed deformations.

The strategy of the proof as in [14] is to study the inclusion of complexes

(TXlogD,[.,Π])(TX,[.,Π]),(T^{\bullet}_{X}\langle-\log{D}\rangle,[\ .\ ,\Pi])\to(T^{\bullet}_{X},[\ .\ ,\Pi])\ ,

albeit from a more global viewpoint. In fact as in [14] it turns out to be more convenient to transport the situation over to the De Rham side where it becomes an inclusion

(ΩXlogD,d)(ΩXlogD+,d)(\Omega^{\bullet}_{X}\langle\log{D}\rangle,d)\to(\Omega^{\bullet}_{X}\langle\log{{}^{+}D}\rangle,d)

where the latter ’log-plus’ complex is a certain complex of meromorphic forms with poles on DD. We study a filtration, introduced in [14], interpolating between the two complexes, especially its first two graded pieces. As we show, the first piece is automatically exact, while 0-acyclicity for the second piece leads to the above cocycle condition. See §3 for details.

We begin the paper with a couple of auxiliary, independent sections. In §1 we construct a ’principal parts complex’ associated to an invertible sheaf LL on a smooth variety, extending the principal parts sheaf P(L)P(L) together with the universal derivation LP(L)L\to P(L). We show this complex is always exact. In §2 we show that, for any normal-crossing divisor DXD\subset X on any smooth variety, the log complex ΩXlogD\Omega^{\bullet}_{X}\langle\log{D}\rangle- unlike ΩX\Omega_{X}^{\bullet} itself- can be pulled back to a complex of vector bundles on the normalization of DD. These complexes play a role in our analysis of the aforementioned inclusion map.

I am grateful to Brent Pym for helpful communications, in particular for communicating Example 8.

1. Principal parts complex

In this section XX denotes an arbitrary nn-dimensional smooth complex variety and LL denotes an invertible sheaf on XX.

1.1. Principal parts

The Grothendieck principal parts sheaf P(L)P(L) (see EGA) is a rank-(n+1)(n+1) bundle on XX defined as

P(L)=p1(p2L(𝒪X×X/Δ2))P(L)=p_{1*}(p_{2}^{*}L\otimes(\mathcal{O}_{X\times X}/\mathcal{I}_{\Delta}^{2}))

where ΔX×X\Delta\subset X\times X is the diagonal and p1,p2:X×XXp_{1},p_{2}:X\times X\to X are the projections. We have a short exact sequence

0ΩX1LP(L)L00\to\Omega^{1}_{X}\otimes L\to P(L)\to L\to 0

whose corresponding extension class in Ext1(L,ΩX1L)=H1(X,ΩX1)\operatorname{Ext}^{1}(L,\Omega^{1}_{X}\otimes L)=H^{1}(X,\Omega^{1}_{X}) coincides with c1(L)c_{1}(L). The sheaf

P0(L)=P(L)L1,P_{0}(L)=P(L)\otimes L^{-1},

which likewise has extension class c1(L)c_{1}(L), is called the normalized principal parts sheaf. The map P(L)LP(L)\to L admits a splitting dL:LP(L)d_{L}:L\to P(L) that is a derivation, i.e.

dL(fu)=fdLu+dfu.d_{L}(fu)=fd_{L}u+df\otimes u.

In fact, dLd_{L} the universal derivation on LL. Moreover P(L)P(L) is generated over 𝒪X\mathcal{O}_{X} by the image of dLd_{L}. Likewise, P0(L)P_{0}(L) is generated by elements of the form dlog(u):=dLuu1\operatorname{dlog}(u):=d_{L}u\otimes u^{-1} where uu is a local generator of LL.

1.2. Complex

It is well known that P(Lm+1)P(L)Lm,m0P(L^{m+1})\simeq P(L)\otimes L^{m},m\geq 0 which in particular yields a derivation Ln+1P(L)Ln,n0L^{n+1}\to P(L)\otimes L^{n},n\geq 0. In fact, This map extends to a complex that we denote by Pn+1(L)P_{n+1}^{\bullet}(L) or just P(L)P^{\bullet}(L) and call the ( (n+1)(n+1)st) principal parts complex of LL:

(1) P(L):Ln+1P(L)Ln2P(L)Ln1n+1P(L)=ΩXnLn+1.\begin{split}P^{\bullet}(L):L^{n+1}\to P(L)L^{n}\to\wedge^{2}P(L)L^{n-1}\to...\wedge^{n+1}P(L)=\Omega^{n}_{X}\otimes L^{n+1}.\end{split}

The differential is given, in terms of local 𝒪X\mathcal{O}_{X}-generators u1,,uk,v1,,vu_{1},...,u_{k},v_{1},...,v_{\ell} of L, by

d(u1ukdL(v1)dL(v)=u1ui^ukdL(ui)dL(v1)dL(v)d(u_{1}...u_{k}d_{L}(v_{1})\wedge...d_{L}(v_{\ell})=\sum u_{1}...\hat{u_{i}}...u_{k}d_{L}(u_{i})\wedge d_{L}(v_{1})\wedge...\wedge d_{L}(v_{\ell})

and extending using additivity and the derivation property. There are also similar shorter complexes

LmP(L)Lm1mP(L).L^{m}\to P(L)L^{m-1}\to...\to\wedge^{m}P(L).

Note the exact sequences

0ΩXmLmmP(L)ΩXm1Lm0.0\to\Omega^{m}_{X}L^{m}\to\wedge^{m}P(L)\to\Omega^{m-1}_{X}L^{m}\to 0.

These sequences splits locally and also split globally whenever LL is a flat line bundle. In such cases, we get a short exact sequence

(2) 0ΩXLn+1[1]P(L)ΩXLn+10\begin{split}0\to\Omega^{\bullet}_{X}L^{n+1}[-1]\to P^{\bullet}(L)\to\Omega^{\bullet}_{X}L^{n+1}\to 0\end{split}

The principal parts complex P(L)P^{\bullet}(L) may be tensored with Ljn1L^{j-n-1}, for any j>0j>0, yielding the jj-th principal parts complex:

(3) Pj(L):LjP0(L)Lj2P0(L)Ljn+1P0(L)Lj\begin{split}P^{\bullet}_{j}(L):L^{j}\to P_{0}(L)L^{j}\to\wedge^{2}P_{0}(L)L^{j}\to...\to\wedge^{n+1}P_{0}(L)L^{j}\end{split}

The differential is defined by setting

d(dlog(u1)dlog(ui)vj)=jdlog(u1)dlog(ui)dlog(v)vjd(\operatorname{dlog}(u_{1})\wedge...\wedge\operatorname{dlog}(u_{i})v^{j})=j\operatorname{dlog}(u_{1})\wedge...\wedge\operatorname{dlog}(u_{i})\operatorname{dlog}(v)v^{j}

where u1,,ui,vu_{1},...,u_{i},v are local generators for LL, and extending by additivity and the derivation property. Thus, P(L)=Pn+1(L)P^{\bullet}(L)=P^{\bullet}_{n+1}(L).

An important property of principal parts complexes is the following:

Proposition 1.

For any local system SS, the complexes Pj(L)SP^{\bullet}_{j}(L)\otimes S are null-homotopic and exact for all j>0j>0.

Proof.

The assertion being local, we may assume LL is trivial and S=S=\mathbb{C} so the ii-th term of Pj(L)SP^{\bullet}_{j}(L)\otimes S is just ΩXi1ΩXi\Omega^{i-1}_{X}\oplus\Omega^{i}_{X} and the differential is (did0d).\left(\begin{matrix}d&\text{id}\\ 0&d\end{matrix}\right). Then a homotopy is given by (0id00).\left(\begin{matrix}0&\text{id}\\ 0&0\end{matrix}\right). Thus, Pj(L)P^{\bullet}_{j}(L) is null-homotopic, hence exact. ∎

1.3. Log version

The above constructions have an obvious extension to the log situation. Thus let DD be a divisor with normal crossings on XX. We define P(L)logDP(L)\langle\log{D}\rangle as the image of P(L)P(L) under the inclusion ΩXΩXlogD\Omega_{X}\to\Omega_{X}\langle\log{D}\rangle, and likewise for P0(L)logDP_{0}(L)\langle\log{D}\rangle. Then as above we get complexes

(4) Pj(L)logD:LjP0(L)logDLjn+1P0(L)logDLj.\begin{split}P^{\bullet}_{j}(L)\langle\log{D}\rangle:L^{j}\to P_{0}(L)\langle\log{D}\rangle L^{j}\to...\to\wedge^{n+1}P_{0}(L)\langle\log{D}\rangle L^{j}.\end{split}

1.4. Foliated version

Let FΩXlogDF\subset\Omega_{X}\langle\log{D}\rangle be an integrable subbundle of rank mm. Then FF gives rise to a foliated De Rham complex (ΩXlogD/F)\wedge^{\bullet}(\Omega_{X}\langle\log{D}\rangle/F), we well as a foliated principal parts sheaf PF1(L)logD=P1(L)logD/FLP^{1}_{F}(L)\langle\log{D}\rangle=P^{1}(L)\langle\log{D}\rangle/F\otimes L. Putting these together, we obtain the foliated principal parts complexes (where P0,F(L)logD:=P0(L)logD/FP_{0,F}(L)\langle\log{D}\rangle:=P_{0}(L)\langle\log{D}\rangle/F):

(5) Pj,F(L)logD:LjP0,F(L)logDLjnm+1P0,F(L)logD\begin{split}P^{\bullet}_{j,F}(L)\langle\log{D}\rangle:L^{j}\to P_{0,F}(L)\langle\log{D}\rangle L^{j}\to...\to\wedge^{n-m+1}P_{0,F}(L)\langle\log{D}\rangle\end{split}

Note that the proof of Proposition 1 made no use of of the acyclicity of the De Rham complex. Hence the same proof applies verbatim to yield

Proposition 2.

For any local system SS, the complexes Pj,F(L)logDSP^{\bullet}_{j,F}(L)\langle\log{D}\rangle\otimes S are null-homotopic and exact for all j>0j>0.

2. Calculus on normal crossing divisors

In this section XX denotes a smooth variety or complex manifold and DD denotes a locally normal-crossing divisor on XX. Our aim is to show that the log complex on XX, unlike its De Rham analogue, can be pulled back to the normalization of DD.

2.1. Branch normal

Let fi:XiXf_{i}:X_{i}\to X be the normalization of the ii-fold locus of DD. A point on XiX_{i} consists of a point on DD together with a choice of ii distinct local branches of DD at it. There is a canonical induced normal-crossing divisor DiD_{i} on XiX_{i}: at a point where x1xmx_{1}...x_{m} is an equation for DD and x1,,xix_{1},...,x_{i} are the chosen branches, the equation of DiD_{i} is xi+1xmx_{i+1}...x_{m}. Note the exact sequence

(6) 0TXlogDTXf1Nf10\begin{split}0\to T_{X}\langle-\log{D}\rangle\to T_{X}\to f_{1*}N_{f_{1}}\to 0\end{split}

where Nf1N_{f_{1}} is the normal bundle to f1f_{1} which fits in an exact sequence

0TX1f1TXNf10.0\to T_{X_{1}}\to f^{*}_{1}T_{X}\to N_{f_{1}}\to 0.

Locally, Nf1N_{f_{1}} coincides with x11𝒪X/𝒪Xx_{1}^{-1}\mathcal{O}_{X}/\mathcal{O}_{X} where x1x_{1} is a ’branch equation’: to be precise, if KK denotes the kernel of the natural surjection f11𝒪X𝒪X1f_{1}^{-1}\mathcal{O}_{X}\to\mathcal{O}_{X_{1}}, then J=K/K2=Kf1𝒪X𝒪X1J=K/K^{2}=K\otimes_{f^{-1}\mathcal{O}_{X}}\mathcal{O}_{X_{1}} is an invertible 𝒪X1\mathcal{O}_{X_{1}}-module locally generated by x1x_{1} and Nf1=J1N_{f_{1}}=J^{-1}. Note that

Nf1𝒪X1(D1)=f1(𝒪X(D)).N_{f_{1}}\otimes\mathcal{O}_{X_{1}}(D_{1})=f_{1}^{*}(\mathcal{O}_{X}(D)).

2.2. Pulling back log complexes

Interestingly, even though the differential on the pullback De Rham complex f11ΩXf_{1}^{-1}\Omega_{X}^{\bullet} does not extend to f1ΩX𝒪X1f^{-1}\Omega_{X}^{\bullet}\otimes\mathcal{O}_{X_{1}}, the analogous assertion for the log complex does hold: the differential on f11ΩXlogDf_{1}^{-1}\Omega_{X}^{\bullet}\langle\log{D}\rangle extends to what might be called the restricted log complex:

f1ΩXlogD=f11ΩXlogD𝒪X1.f_{1}^{*}\Omega_{X}^{\bullet}\langle\log{D}\rangle=f_{1}^{-1}\Omega_{X}^{\bullet}\langle\log{D}\rangle\otimes\mathcal{O}_{X_{1}}.

This is due to the identity (where x1x_{1} denotes a branch equation)

dx1=x1dlog(x1).dx_{1}=x_{1}\operatorname{dlog}(x_{1}).

Note that the residue map yields an exact sequence

(7) 0ΩX11logD1jf1ΩX1logDRes𝒪X10.\begin{split}0\to\Omega^{1}_{X_{1}}\langle\log{D_{1}}\rangle\stackrel{{\scriptstyle j}}{{\to}}f_{1}^{*}\Omega^{1}_{X}\langle\log{D}\rangle\stackrel{{\scriptstyle\mathrm{Res}}}{{\to}}\mathcal{O}_{X_{1}}\to 0.\end{split}

Note that the rsidue map commutes with exterior derivative. Therefore this sequence induces a short exact sequence of complexes

(8) 0ΩX1logD1f1ΩXlogDΩX1logD1[1]0.\begin{split}0\to\Omega^{\bullet}_{X_{1}}\langle\log{D_{1}}\rangle\to f_{1}^{*}\Omega^{\bullet}_{X}\langle\log{D}\rangle\to\Omega_{X_{1}}^{\bullet}{\langle\log{D_{1}}\rangle}[-1]\to 0.\end{split}

Furthermore, a twisted form of the restricted log complex, called the normal log complex, also exists:

(9) Nf1f1ΩXlogD:Nf1Nf1f1ΩX1logD\begin{split}N_{f_{1}}\otimes f_{1}^{*}\Omega^{\bullet}_{X}\langle\log{D}\rangle:N_{f_{1}}\to N_{f_{1}}\otimes f_{1}^{*}\Omega^{1}_{X}\langle\log{D}\rangle\to...\end{split}

this is thanks to the identity, where ω\omega is any log form,

d(ω/x1)=(dω)/x1dlog(x1)ω/x1.d(\omega/x_{1})=(d\omega)/x_{1}-\operatorname{dlog}(x_{1})\wedge\omega/x_{1}.

Now recall the exact sequence coming from the residue map

0ΩX1logD1f1ΩXlogD𝒪X100\to\Omega_{X_{1}}\langle\log{D_{1}}\rangle\to f_{1}^{*}\Omega_{X}\langle\log{D}\rangle\to\mathcal{O}_{X_{1}}\to 0

In fact, it is easy to check that this exact sequence has extension class c1(Nf1)c_{1}(N_{f_{1}}) hence identifies f1ΩXlogDf_{1}^{*}\Omega_{X}\langle\log{D}\rangle with P0(Nf1)P_{0}(N_{f_{1}}) so that the normal log complex (9) may be identified with the principal parts complex P(Nf1)P^{\bullet}(N_{f_{1}}):

Lemma 3.

The normal log complex Nf1f1ΩXlogDN_{f_{1}}\otimes f_{1}^{*}\Omega_{X}\langle\log{D}\rangle is isomorphic to P(Nf1)P^{\bullet}(N_{f_{1}}), hence is exact.

Similarly, a pull back log complex fkΩXlogD=fk1ΩXlogD𝒪Xkf_{k}^{*}\Omega^{\bullet}_{X}\langle\log{D}\rangle=f_{k}^{-1}\Omega_{X}^{\bullet}\langle\log{D}\rangle\otimes\mathcal{O}_{X_{k}} exists for all k1k\geq 1. A similar twisted log complex also exists the determinant of the normal bundle NfkN_{f_{k}}:

(10) detNfkfkΩXlogD:detNfkdetNfkΩX1logD\begin{split}\det N_{f_{k}}\otimes f_{k}^{*}\Omega_{X}^{\bullet}\langle\log{D}\rangle:\det N_{f_{k}}\to\det N_{f_{k}}\otimes\Omega^{1}_{X}\langle\log{D}\rangle\to...\end{split}

This comes from (where x1,,xkx_{1},...,x_{k} are the branch equations at a given point of XkX_{k}):

d(ω/x1xk)=dω/x1xkdlog(x1xk)ω/x1xk).d(\omega/x_{1}...x_{k})=d\omega/x_{1}...x_{k}-\operatorname{dlog}(x_{1}...x_{k})\omega/x_{1}...x_{k}).

2.3. Iterated residue

We have a short exact sequence of vector bundles on XkX_{k}:

(11) 0ΩXklogDkfkΩXlogDνk𝒪Xk0\begin{split}0\to\Omega_{X_{k}}\langle\log{D_{k}}\rangle\to f_{k}^{*}\Omega_{X}\langle\log{D}\rangle\to\nu_{k}\otimes\mathcal{O}_{X_{k}}\to 0\end{split}

where νk\nu_{k} is the local system of branches of DD along XkX_{k} and the right map is multiple residue. Taking exterior powers, we get various exact Eagon-Northcott complexes. In particular, we get surjections, called iterated Poincaé residue:

(12) fkΩXilogDΩXkiklogDkdet(νk),ik,\begin{split}f_{k}^{*}\Omega_{X}^{i}\langle\log{D}\rangle\to\Omega_{X_{k}}^{i-k}\langle\log{D_{k}}\rangle\otimes{\det}_{\mathbb{C}}(\nu_{k}),i\geq k,\end{split}
(13) fkΩXilogDiνk𝒪Xk,ik.\begin{split}f_{k}^{*}\Omega_{X}^{i}\langle\log{D}\rangle\to\wedge^{i}_{\mathbb{C}}\nu_{k}\otimes\mathcal{O}_{X_{k}},i\leq k.\end{split}

det(νk){\det}_{\mathbb{C}}(\nu_{k}) is a rank-1 local system on XkX_{k} which may be called the ’normal orientation sheaf’. The maps for iki\geq k together yield a surjection

(14) fkΩXlogDΩXklogDk[k]det(νk).\begin{split}f_{k}^{*}\Omega_{X}^{\bullet}\langle\log{D}\rangle\to\Omega_{X_{k}}^{\bullet}\langle\log{D_{k}}\rangle[-k]\otimes\det(\nu_{k}).\end{split}

3. Comparing log and log plus complexes

In this section XX denotes a log-symplectic smooth variety with log-symplectic form Φ\Phi and corresponding Poisson vector Π=Φ1\Pi=\Phi^{-1}, and DD denotes the degeneracy divisor of Π\Pi or polar divisor of Φ\Phi. Our aim is to prove Theorem 6 which shows that deformations of (X,Φ)(X,\Phi) coincide with locally trivial deformations of (X,Φ,D)(X,\Phi,D) and are unobstructed.

3.1. Setting up

We will use ΩX+\Omega_{X}^{+\bullet} to denote i>0ΩXi\bigoplus\limits_{i>0}\Omega^{i}_{X} and similarly for the log versions. This to match with the Lichnerowicz-Poisson complex TXT^{\bullet}_{X} and TXlogDT^{\bullet}_{X}\langle-\log{D}\rangle. Thus, interior multiplication by Φ\Phi induces and isomorphism TXlogDΩXlogDT^{\bullet}_{X}\langle-\log{D}\rangle\to\Omega^{\bullet}_{X}\langle\log{D}\rangle. Equivalently, Φ\Phi itself is a form im ΩX2logD\Omega^{2}_{X}\langle\log{D}\rangle inducing a nondegenerate pairing on TXlogDT_{X}\langle-\log{D}\rangle. In terms of local coordinates, at a point of multiplicity mm on DD, we have a basis for ΩXlogD\Omega_{X}\langle\log{D}\rangle of the form

η1=dlog(x1),,ηm=dlog(xm),ηm+1=dlog(xm+1),\eta_{1}=\operatorname{dlog}(x_{1}),...,\eta_{m}=\operatorname{dlog}(x_{m}),\eta_{m+1}=\operatorname{dlog}(x_{m+1}),...

and then

Φ=bijηiηj.\Phi=\sum b_{ij}\eta_{i}\wedge\eta_{j}.

We have an inclusion of complexes

TXlogDTXT_{X}^{\bullet}\langle-\log{D}\rangle\to T^{\bullet}_{X}

where, for XX compact Kähler, the first complex controls ’locally trivial’ deformations of (X,Π)(X,\Pi), i.e. deformations of (X,Π)(X,\Pi) inducing a locally trivial deformation of D=[Πn]D=[\Pi^{n}], and the second complex controls all deformations of (X,Π)(X,\Pi). It is known (see e.g. [14]) that locally trivial deformations of (X,Π)(X,\Pi) are always unobstructed and have an essentially Hodge-theoretic (hence topological) character, so one is interested in conditions to ensure that the above inclusion induces an isomorphism on deformation spaces; as is well known, the latter would follow if one can show that the cokernel of this inclusion has vanishing 1\mathbb{H}^{1}.

Our approach to this question starts with the above ’multiplication by Φ\Phi’ isomorphism

(TXlogD,[.,Π])(ΩX+logD,d).(T_{X}^{\bullet}\langle-\log{D}\rangle,[\ .\ ,\Pi])\to(\Omega^{+\bullet}_{X}\langle\log{D}\rangle,d).

This isomorphism extends to an isomorphism to TXT^{\bullet}_{X} with a certain subcomplex of ΩX+(D)\Omega^{+\bullet}_{X}(*D), the meromorphic forms regular off DD, that we call the log plus complex and denote by ΩX+logD+\Omega_{X}^{+\bullet}\langle\log{{}^{+}D}\rangle.

Our goal then becomes that of comparing the log and log-plus complexes. To this end we introduce a filtration on ΩX+logD+\Omega^{+\bullet}_{X}\langle\log{{}^{+}D}\rangle, essentially the filtration induced by the exact sequence

0TXlogDTXf1Nf100\to T_{X}\langle-\log{D}\rangle\to T_{X}\to f_{1*}N_{f_{1}}\to 0

and its isomorphic copy

0ΩXlogDΩXlogD+fNf100\to\Omega_{X}\langle\log{D}\rangle\to\Omega_{X}\langle\log{{}^{+}D}\rangle\to f_{*}N_{f_{1}}\to 0

where f1:X1DXf_{1}:X_{1}\to D\subset X is the normalization of DD and Nf1N_{f_{1}} is the associated normal bundle (’branch normal bundle’). We will show that the first graded piece is always an exact complex. The second graded piece is much more subtle. We will show that it is locally exact in degree 0 unless the log-symplectic form Φ\Phi, i.e. the matrix (bij)(b_{ij}) above satisfies some special relations with integer coefficients.

The computations of this section are all local in character, though the applications are global.

3.2. Residues and duality

Let fi:XiXf_{i}:X_{i}\to X be the normalization of the ii-fold locus of DD, DiD_{i} the induced normal-crossing divisor on XiX_{i}. Thus a point of XiX_{i} consists of a point pp of DD together with a choice of an unordered set SS of ii branches of DD through pp and DiD_{i} is the union of the branches of DD not in SS. We consider first the codimension-1 situation. As above, we have a residue exact sequence

(15) 0ΩX11logD1jf1ΩX1logDRes𝒪X10\begin{split}0\to\Omega^{1}_{X_{1}}\langle\log{D_{1}}\rangle\stackrel{{\scriptstyle j}}{{\to}}f_{1}^{*}\Omega^{1}_{X}\langle\log{D}\rangle\stackrel{{\scriptstyle\mathrm{Res}}}{{\to}}\mathcal{O}_{X_{1}}\to 0\end{split}

(the right-hand map given by residue is locally evaluation on x1x1x_{1}\operatorname{\partial}_{x_{1}} where x1x_{1} is a local equation for the branch of DD through the given point of X1X_{1} ). Note that if η\eta comes from a closed form on XX near DD then Res(η)\mathrm{Res}(\eta) is a constant.

Dualizing (15), we get

(16) 0𝒪X1R1ˇf1TXlogDjˇTX1logD10,\begin{split}0\to\mathcal{O}_{X_{1}}\stackrel{{\scriptstyle\check{R_{1}}}}{{\to}}f_{1}^{*}T_{X}\langle-\log{D}\rangle\stackrel{{\scriptstyle\check{j}}}{{\to}}T_{X_{1}}\langle-\log{D_{1}}\rangle\to 0,\end{split}

where the left-hand map, the ’co-residue’, is locally multiplication by x1x1x_{1}\operatorname{\partial}_{x_{1}} where x1x_{1} is a branch equation). Set

v1=x1x1.v_{1}=x_{1}\operatorname{\partial}_{x_{1}}.

Then v1v_{1} is canonical as section of f1TXlogDf_{1}^{*}T_{X}\langle-\log{D}\rangle , independent of the choice of local equation x1x_{1}. By contrast, x1\operatorname{\partial}_{x_{1}} as section of f1TXf_{1}^{*}T_{X} is canonical only up to a tangential field to X1X_{1}, and generates f1TXf_{1}^{*}T_{X} modulo TX1logDT_{X_{1}}\langle-\log{D}\rangle.

Now f1ΩX1logDf_{1}^{*}\Omega^{1}_{X}\langle\log{D}\rangle and f1TXlogDf^{*}_{1}T_{X}\langle-\log{D}\rangle admit mutually inverse isomorphisms

iX1Π:=Π,.X1=f1Π,.,iX1Φ:=Φ,.X1=f1Φ,..i_{X_{1}}\Pi:=\langle\Pi,.\rangle_{X_{1}}=f_{1}^{*}\langle\Pi,.\rangle,i_{X_{1}}\Phi:=\langle\Phi,.\rangle_{X_{1}}=f_{1}^{*}\langle\Phi,.\rangle.

The composite

jˇiX1Πj:ΩX11logD1TX1logD1\check{j}\circ i_{X_{1}}\Pi\circ j:\Omega^{1}_{X_{1}}\langle\log{D_{1}}\rangle\to T_{X_{1}}\langle-\log{D_{1}}\rangle

has a rank-1 kernel that is the kernel of the Poisson vector on X1X_{1} induced by Π\Pi, aka the conormal to the symplectic foliation on X1X_{1}. Now set

ψ1=iX1(Φ)(v1)=Φ,v1X1.\psi_{1}=i_{X_{1}}(\Phi)(v_{1})=\langle\Phi,v_{1}\rangle_{X_{1}}.

Then ψ1\psi_{1} is locally the form in ΩX1logD1\Omega_{X_{1}}\langle\log{D_{1}}\rangle denoted by x1ϕ1x_{1}\phi_{1} in [14]. Again ψ1\psi_{1} is canonically defined, independent of choices and corresponds to the first column of the B=(bij)B=(b_{ij}) matrix for a local coordinate system x1,x2,x_{1},x_{2},... compatible with the normal-crossing divisor DD. By contrast, ϕ1\phi_{1}, which depends on the choice of local equation x1x_{1}, is canonical up to a log form in ΩX1logD1\Omega_{X_{1}}\langle\log{D_{1}}\rangle and generates ΩX1logD1+\Omega_{X_{1}}\langle\log{{}^{+}D_{1}}\rangle modulo the latter.

In X1D1X_{1}\setminus D_{1}, Φ\Phi is locally of the form dlog(x1)dx2+\operatorname{dlog}(x_{1})\wedge dx_{2}+(symplectic), so there ψ1=dx2\psi_{1}=dx_{2}. Note that by skew-symmetry we have

ResiX1(Φ)R1ˇ=0.\mathrm{Res}\circ i_{X_{1}}(\Phi)\circ\check{R_{1}}=0.

Thus, locally ψ1ΩX1logD1\psi_{1}\in\Omega_{X_{1}}\langle\log{D_{1}}\rangle. In terms of the matrix BB above, ψ1=j>1b1jdlog(xj)\psi_{1}=\sum\limits_{j>1}b_{1j}\operatorname{dlog}(x_{j}). Note that ψ1\psi_{1} which corresponds to the Hamiltonian vector field v1v_{1}, is a closed form. Consequently, ψ1\psi_{1} defines a foliation on X1X_{1}. Let Q1=ψ1ΩX1Q_{1}^{\bullet}=\psi_{1}\Omega_{X_{1}}^{\bullet} be the associated foliated De Rham complex ψ1ΩX1\psi_{1}\Omega_{X_{1}}^{\bullet}:

Q10=𝒪X1ϕ1Q11=ψ1ΩX11ΩX11/𝒪X1ψ1Q1i=iQ11Q_{1}^{0}=\mathcal{O}_{X_{1}}\phi_{1}\to Q_{1}^{1}=\psi_{1}\Omega^{1}_{X_{1}}\simeq\Omega^{1}_{X_{1}}/\mathcal{O}_{X_{1}}\psi_{1}\to...\to Q_{1}^{i}=\wedge^{i}Q_{1}^{1}\to...

endowed with the foliated differential.

Note that the residue exact sequence (15) induces the Poincaré residue sequence

0ΩX1logD1f1ΩXlogDΩX1logD1[1]0.0\to\Omega^{\bullet}_{X_{1}}\langle\log{D_{1}}\rangle\to f_{1}^{*}\Omega^{\bullet}_{X}\langle\log{D}\rangle\to\Omega^{\bullet}_{X_{1}}\langle\log{D_{1}}\rangle[-1]\to 0.

Again the Poincaré residue of a closed form is closed. Now the exact sequence

0TXlogDTXf1Nf100\to T_{X}\langle-\log{D}\rangle\to T_{X}\to f_{1*}N_{f_{1}}\to 0

yields

(17) 0ΩXlogDΩXlogD+f1Nf10.\begin{split}0\to\Omega_{X}\langle\log{D}\rangle\to\Omega_{X}\langle\log{{}^{+}D}\rangle\to f_{1*}N_{f_{1}}\to 0.\end{split}

and this sequence induces the \mathcal{F}_{\bullet} filtration on the log-plus complex ΩXlogD+\Omega_{X}^{\bullet}\langle\log{{}^{+}D}\rangle.

3.3. First graded piece

Now consider first the first graded 𝒢1=(1/0)[1]\mathcal{G}^{\bullet}_{1}=(\mathcal{F}^{\bullet}_{1}/\mathcal{F}^{\bullet}_{0})[1] which is supported in codimension 1. (the shift is so that 𝒢\mathcal{G}^{\bullet} starts in degree 0). Then 𝒢1\mathcal{G}_{1}^{\bullet} is a (finite) direct image of a complex of X1X_{1} modules:

1:Nf1Nf1Q1Nf1Q12\mathcal{E}_{1}:N_{f_{1}}\to N_{f_{1}}\otimes Q_{1}\to N_{f_{1}}\otimes Q_{1}^{2}\to...

Using Lemma 3, we can easily show:

Proposition 4.

1\mathcal{E}_{1} is isomorphic to PR1(Nf1)P^{\bullet}_{R_{1}^{\prime}}(N_{f_{1}}), hence is null-homotopic and exact, hence 𝒢1\mathcal{G}^{\bullet}_{1} is exact.

3.4. Second graded piece

Next we study 𝒢2\mathcal{G}_{2}, which is supported on X2X_{2}. We consider a connected, nonempty open subset WX2W\subset X_{2}, for example an entire component, over which the ’normal orientation sheaf’ ν2:X2,1X2\nu_{2}:X_{2,1}\to X_{2}, i.e. the local 2\mathbb{Z}_{2}-system of branches of X1X_{1} along X2X_{2}, is trivial (we can take W=X2W=X_{2} if, e.g. DD has global normal crossings). Such a subset WW of X2X_{2} is said to be a normally split subset of X2X_{2} and a normal splitting of WW is an ordering of the branches is specified. Obviously X2X_{2} is covered by such subsets WW. Likewise, for a subset ZXkZ\subset X_{k}.

3.4.1. Iterated residue

Over a normally split subset WW, we have a diagram

(18) 02𝒪WRˇ2f2TXlogD|WTX2logD2|W00ΩWlogD2f2ΩXlogD|WR22𝒪W0\begin{split}\begin{matrix}0\to 2\mathcal{O}_{W}\stackrel{{\scriptstyle\check{R}_{2}}}{{\to}}&f_{2}^{*}T_{X}\langle-\log{D}\rangle|_{W}&\to T_{X_{2}}\langle-\log{D_{2}}\rangle|_{W}\to 0\\ &\downarrow&\\ 0\to\Omega_{W}\langle\log{D_{2}}\rangle\to&f_{2}^{*}\Omega_{X}\langle\log{D}\rangle|_{W}&\stackrel{{\scriptstyle R_{2}}}{{\to}}2\mathcal{O}_{W}\to 0\end{matrix}\end{split}

where Rˇ2\check{R}_{2} is the map induced by Rˇ1\check{R}_{1}. The composite map R2Rˇ2:2𝒪W2𝒪WR_{2}\check{R}_{2}:2\mathcal{O}_{W}\to 2\mathcal{O}_{W} is just the alternating form induced by Φ\Phi, and has the form cWH2c_{W}H_{2} where H2H_{2} is the hyperbolic plane (0110)\left(\begin{matrix}0&1\\ -1&0\end{matrix}\right). In terms of a local frame for ΩXlogD\Omega_{X}\langle\log{D}\rangle containing dlog(x1),dlog(x2)\operatorname{dlog}(x_{1}),\operatorname{dlog}(x_{2}), cWc_{W} is the coefficient of dlog(x1)dlog(x2)\operatorname{dlog}(x_{1})\wedge\operatorname{dlog}(x_{2}) in Φ\Phi. Note cWc_{W} must be constant because Φ\Phi is closed. In fact we have

cW=Res1Res2(Φ)c_{W}=\mathrm{Res}_{1}\mathrm{Res}_{2}(\Phi)

where Resi\mathrm{Res}_{i} denote the (Poincaré) residues along the branches of X1X_{1} over X2X_{2}. Set

ResW(Φ):=cW.\mathrm{Res}_{W}(\Phi):=c_{W}.

This is essentially what is called the biresidue by Matviichuk et al., see [9]. Thus, when cW0c_{W}\neq 0, we have a basis for the log forms

η1=dlog(x1),,ηm=dlog(xm),ηm+1=dxm+1,,η2n=dx2n\eta_{1}=\operatorname{dlog}(x_{1}),...,\eta_{m}=\operatorname{dlog}(x_{m}),\eta_{m+1}=dx_{m+1},...,\eta_{2}n=dx_{2n}

m=m= multiplicity of DD, m2m\geq 2, and then

Φ=bijηiηj\Phi=\sum b_{ij}\eta_{i}\wedge\eta_{j}

where

b12=b21=cW.b_{12}=-b_{21}=c_{W}.

If WW may be not be normally orientable (e.g. an entire component of X2X_{2}) then cWc_{W} is defined only up to sign; if cW=0c_{W}=0 we say that WW is non-residual, otherwise it is residual.

3.4.2. Non-residual case

Here we consider the case cW=0c_{W}=0.

Note that in that case we may express Φ\Phi along WW in the form

Φ=dlog(x1)γ3+dlog(x2)γ4+γ5\Phi=\operatorname{dlog}(x_{1})\gamma_{3}+\operatorname{dlog}(x_{2})\gamma_{4}+\gamma_{5}

where the gammas are closed log forms in the coordinates on WW, i.e. x3,,x2nx_{3},...,x_{2n}. Moreover, γ3γ40\gamma_{3}\wedge\gamma_{4}\neq 0 because Φn\Phi^{n} is divisible by dlog(x1)dlog(x2)\operatorname{dlog}(x_{1})\operatorname{dlog}(x_{2}). Also, unless γ3,γ4\gamma_{3},\gamma_{4} are both holomorphic (pole-free), there is another component WW^{\prime} of X2X_{2} such that cW0c_{W^{\prime}}\neq 0 (in particular, WD2W\cap D_{2}\neq\emptyset). Hence if no such WW^{\prime} exists, we may by suitably modifying coordinates, assume locally that γ3=dx3,γ4=dx4\gamma_{3}=dx_{3},\gamma_{4}=dx_{4}. A similar argument, or induction, applies to γ5\gamma_{5}. This means we are essentially in the P-normal case considered in [13]. This we conclude:

Lemma 5.

Unless Π\Pi is P-normal, there exists a nonempty residual open subset WW of X2X_{2}.

3.4.3. Residual case: identifying 𝒢2\mathcal{G}_{2}

Next we analyze a residual normally oriented open subset WX2W\subset X_{2}. As above, we get a composite map of R2:2𝒪Wf2ΩXlogD|WR^{\prime}_{2}:2\mathcal{O}_{W}\to f_{2}^{*}\Omega_{X}\langle\log{D}\rangle|_{W} , whose image we denote by M2WM_{2W}. It has a local basis (ψ11=x1ϕ1,ψ12=x2ϕ2)(\psi_{11}=x_{1}\phi_{1},\psi_{12}=x_{2}\phi_{2}) corresponding to the basis (e1,e2)(e_{1},e_{2}) of 2𝒪W2\mathcal{O}_{W}. In term of the BB-matrix, we have

ψ11=b1jηj=bj1ηj,ψ12=b2jηj=bj2ηj.\psi_{11}=\sum b_{1j}\eta_{j}=-\sum b_{j1}\eta_{j},\psi_{12}=-\sum b_{2j}\eta_{j}=\sum b_{j2}\eta_{j}.

As ψ11,ψ12\psi_{11},\psi_{12} are closed, M2M_{2} is integrable. Let Ω¯\bar{\Omega} denote the quotient f2ΩXlogD|W/M2Wf_{2}^{*}\Omega_{X}\langle\log{D}\rangle|_{W}/M_{2W}. Then we have an isomorphism

(19) Ω¯ΩWlogD2\begin{split}\bar{\Omega}\to\Omega_{W}\langle\log{D_{2}}\rangle\end{split}

given explicitly by

ω¯ωRes1(ω)ψ12/cWRes2(ω)ψ11/cW\bar{\omega}\mapsto\omega-\operatorname{Res}_{1}(\omega)\psi_{12}/c_{W}-\operatorname{Res}_{2}(\omega)\psi_{11}/c_{W}

(because Res2(ψ11)=Res1(ψ12)=cW\operatorname{Res}_{2}(\psi_{11})=\operatorname{Res}_{1}(\psi_{12})=c_{W}, residues with respect to the two branches of DD). Now set N2=detNf2N_{2}=\det N_{f_{2}}, an invertible sheaf on X2X_{2}. Then 𝒢2=(2/1)[2]\mathcal{G}^{\bullet}_{2}=(\mathcal{F}^{\bullet}_{2}/\mathcal{F}^{\bullet}_{1})[2] is the direct image of a complex on X2X_{2}:

(20) 2:N2N2Ω¯N22Ω¯\begin{split}\mathcal{E}^{\bullet}_{2}:N_{2}\to N_{2}\otimes\bar{\Omega}\to N_{2}\otimes\wedge^{2}\bar{\Omega}\to...\end{split}

where a local generator of N2N_{2} has the form 1/x1x21/x_{1}x_{2} and the differential has the form

ω¯/x1x2dω¯/x1x1±(ω¯/x1x2)dlog(x1x2).\bar{\omega}/x_{1}x_{2}\mapsto d\bar{\omega}/x_{1}x_{1}\pm(\bar{\omega}/x_{1}x_{2})\operatorname{dlog}(x_{1}x_{2}).

3.4.4. Zeroth differential

Using the identification (19), the zeroth differential has the form

(21) d~(g/x1x2)=1x1x2(dg+g(dlog(x1x2)(ψ11+ψ12)/cW)),g𝒪X2.\begin{split}\tilde{d}(g/x_{1}x_{2})=\frac{1}{x_{1}x_{2}}(dg+g(\operatorname{dlog}(x_{1}x_{2})-(\psi_{11}+\psi_{12})/c_{W})),g\in\mathcal{O}_{X_{2}}.\end{split}

The form ψ2=dlog(x1x2)+(ψ11+ψ12)/cW\psi_{2}=-\operatorname{dlog}(x_{1}x_{2})+(\psi_{11}+\psi_{12})/c_{W} has zero residues with respect to x1,x2x_{1},x_{2}, hence yields a form in ΩX2logD2\Omega_{X_{2}}\langle\log{D_{2}}\rangle. Changing the local equations x1,x2x_{1},x_{2} changes ψ\psi by adding a holomorphic (pole-free) form on X2X_{2}.

For gg nonzero (21) can be rewritten

(22) d~(g/x1x2)=gx2x2(dlog(g)ψ2)\begin{split}\tilde{d}(g/x_{1}x_{2})=\frac{g}{x_{2}x_{2}}(\operatorname{dlog}(g)-\psi_{2})\end{split}

When does this operator have a nontrivial kernel? First, if gg is constant then ψ2=0\psi_{2}=0 on WW which is im[possible if WW meets D2D_{2}. Next, locally at a point xWD2Wx\in W\setminus D_{2}\cap W, clearly g/x1x2g/x_{1}x_{2} holomorphic and nonzero in the kernel exists locally since ψ2\psi_{2} is closed and holomorphic so ψ2=dh\psi_{2}=dh for a holomorphic function hh and we can take g=ehg=e^{h}. Moreover nonzero solutions to d(g/x1x2)=0d(g/x_{1}x_{2})=0 differ by a multiplicative constant. The condition that the local solutions patch is clearly that 12πiγψ2\frac{1}{2\pi i}\int\limits_{\gamma}\psi_{2} be an integer for any loop γ\gamma in WD2WW\setminus D_{2}\cap W. Now ψ2\psi_{2} is defined only modulo a holomorphic form on X2X_{2} while H1(WD2W)H_{1}(W\setminus D_{2}\cap W) is generated modulo H1(W)H_{1}(W) by small loops normal to components of D2D_{2}, So the relevant condition is just integrality over such loops γ\gamma.

At a simple point of D2WD_{2}\cap W, the condition that gg exist locally as a holomorphic function with no pole on D2D_{2} is clearly that for γ\gamma as above, oriented positively, the integer 12πiγψ2\frac{1}{2\pi i}\int\limits_{\gamma}\psi_{2} is nonnegative, so that gg has no pole on D2D_{2}. In other words, that the sum of the first 2 columns of the BB matrix, normalized so that b12=b21=1b_{12}=-b_{21}=1, should be a nonnegative integer vector. Finally by Hartogs, if gg is holomorphic off the singular locus of D2WD_{2}\cap W, it extends holomorphically to WW.

3.4.5. Special components

Now let ZZ be a component of D2WD_{2}\cap W and assume WW and ZZ are both normally split so that the branches of DD along WW may be labelled 12 while those along ZZ may be labelled 123. Thus branches of X2X_{2} over ZZ are labelled 12, 23, 31 and the preceding discussion shows that the zeroth differential has nontrivial kernel along ZZ only if the iterated residues of Φ\Phi along these branches, denoted c21,c23,c31c_{21},c_{23},c_{31}, assuming c120c_{12}\neq 0, satisfy

(23) c23+c31=kc21,k.\begin{split}c_{23}+c_{31}=kc_{21},k\in\mathbb{N}.\end{split}

We call such a component ZZ special; then WW is said to be special if every (normally split) component of D2WD_{2}\cap W is special.

What about the normally split hypothesis? Suppose first WW is contained in a connected open set WW^{\prime} which is not normally split. Then as c12c_{12} is locally constant in WW^{\prime} it follows that c12=0c_{12}=0, i.e. WW is not residual. Now suppose ZZ is contained in ZZ^{\prime} open connected and not normally split. Then monodromy acts on the branches of X2X_{2} along ZZ^{\prime} cyclically and consequently the cijc_{ij} above are all equal. Then (23) holds automatically with k=2k=2 so ZZ is special.

3.4.6. Conclusion

What we have so far proven is the following: if WW is a normally oriented residual open subset of of X2X_{2} then the stalk of the zeroth cohomology 0(𝒢2)\mathcal{H}^{0}(\mathcal{G}_{2}^{\bullet}) vanishes somewhere on WW unless either

(i) WD2=W\cap D_{2}=\emptyset, or

(ii) WW is special.

Note that if the stalk of 0(𝒢2)\mathcal{H}^{0}(\mathcal{G}_{2}^{\bullet}) vanishes somewhere in WW, then because 𝒢20\mathcal{G}^{0}_{2} is coherent and torsion-free, it follows that H0(𝒢2)|W=0H^{0}(\mathcal{G}^{\bullet}_{2})|_{W}=0, hence a similar vanishing holds for the entire component of X2X_{2} containing WW. Now recall that, minding the index shift, if H0(𝒢2)=0H^{0}(\mathcal{G}^{\bullet}_{2})=0 then the cokernel of the inclusion ΩX+logDΩX+logD+\Omega^{+\bullet}_{X}\langle\log{D}\rangle\to\Omega^{+\bullet}_{X}\langle\log{{}^{+}D}\rangle has vanishing 1\mathbb{H}^{1} (and 0\mathbb{H}^{0}). On the other hand, it is well known (see e.g. [14]) that ΩX+logDTXlogD\Omega_{X}^{+\bullet}\langle\log{D}\rangle\simeq T_{X}^{\bullet}\langle-\log{D}\rangle controls deformations of (X,Φ)(X,\Phi) or (X,Π)(X,\Pi) where DD deforms locally trivially, and those deformations are unobstructed thanks to Hodge theory.

Summarizing this discussion, we conclude:

Theorem 6.

Let (X,Φ)(X,\Phi) be a log-symplectic manifold with polar divisor DD. With notations as above, let

ΩX+logD=i>0ΩXilogD,ΩX+logD+=i>0ΩXilogD+.\Omega^{+\bullet}_{X}\langle\log{D}\rangle=\bigoplus\limits_{i>0}\Omega^{i}_{X}\langle\log{D}\rangle,\Omega^{+\bullet}_{X}\langle\log{{}^{+}D}\rangle=\bigoplus\limits_{i>0}\Omega^{i}_{X}\langle\log{{}^{+}D}\rangle.

Then the inclusions

ΩX+logDΩX+logD+,\Omega_{X}^{+\bullet}\langle\log{D}\rangle\to\Omega_{X}^{+\bullet}\langle\log{{}^{+}D}\rangle,
TXlogDTXT_{X}^{\bullet}\langle-\log{D}\rangle\to T^{\bullet}_{X}

induce isomorphisms on 2\mathcal{H}^{2} and injections on 3\mathcal{H}^{3}, hence isomorphisms on 1\mathbb{H}^{1} and injections on 2\mathbb{H}^{2}, unless either

(i) X2X_{2} has a non-residual component; or

(ii) X2X_{2} has a special component.

As noted above, any component of X2X_{2} that is disjoint from D2D_{2}, i.e. contains no triple points of DD, is automatically non-residual.

Corollary 7.

Notations as above, if XX is compact and Kählerian and conditions (i), (ii) both fail, then the pair (X,Φ)(X,\Phi) has unobstructed deformations and the polar divisor of Φ\Phi deforms locally trivially.

In the case where DD has global normal crossings, i.e. is a union of smooth divisors, this result also follows from results in [9], which also states a partial converse: when TXlogDTXT_{X}^{\bullet}\langle-\log{D}\rangle\to T^{\bullet}_{X} is not a quasi-isomorphism, (X,Φ)(X,\Phi) has obstructed deformations and admits deformations where DD either smooths or deforms locally trivially.

Example 8.

(Due to M. Matviichuk, B. Pym, T. Schedler, see [9], communicated by  B. Pym) Consider the matrix

(24) B=(bij)=(0124103523064510)\begin{split}B=(b_{ij})=\left(\begin{matrix}0&1&2&4\\ -1&0&3&5\\ -2&-3&0&6\\ -4&-5&1&0\end{matrix}\right)\end{split}

and the corresponding log-symplectic form on 4\mathbb{C}^{4}, Φ=i<jbijdzizidzjzj\Phi=\sum\limits_{i<j}b_{ij}\frac{dz_{i}}{z_{i}}\wedge\frac{dz_{j}}{z_{j}} and corresponding Poisson structure Π=Φ1\Pi=\Phi^{-1}, both of which extend to 4\mathbb{P}^{4} with Pfaffian divisor D=(z0z1z2z3z4)D=(z_{0}z_{1}z_{2}z_{3}z_{4}), z0=z_{0}= hyperplane at infinity. Then Π\Pi admits the 1st order Poisson deformation with bivector z3z4z1z2z_{3}z_{4}\operatorname{\partial}_{z_{1}}\operatorname{\partial}_{z_{2}}, which in fact extends to a Poisson deformation of (4,Π)(\mathbb{P}^{4},\Pi) over the affine line \mathbb{C}, and the Pfaffian divisor deforms as (z3z4z0(z1z2tz3z4))(z_{3}z_{4}z_{0}(z_{1}z_{2}-tz_{3}z_{4})), hence non locally-trivially. Correspondingly, the log-plus form z3z4ϕ1ϕ2z_{3}z_{4}\phi_{1}\phi_{2} is closed ( and not exact). That d(z3z4ϕ1ϕ2)=0d(z_{3}z_{4}\phi_{1}\phi_{2})=0 corresponds to the integral column relation

k1k2+(e1+e2)(e3+e4)=0k_{1}-k_{2}+(e_{1}+e_{2})-(e_{3}+e_{4})=0

where the kik_{i} and eje_{j} are the columns of the BB matrix and the identity, respectively, showing that (z1z2z3)(z_{1}z_{2}z_{3}) and (z1z2z4)(z_{1}z_{2}z_{4}) are residual triples of type II and (12), i.e. (x1)(x2)(x_{1})\cap(x_{2}) is a special component of X2X_{2}.

Remark 9.

As we saw above, the presence of monodromy on the branches of DD is related to non-residual or special components. This suggests that log-symplectic manifolds with irreducible polar divisor may often be obstructed. However we don’t have specific examples.

References

  • [1] I. Mǎrcut and B. Osorno Torres, Deformations of log symplectic structures, J. London Math. Soc. (2014).
  • [2] F. A. Bogomolov, Hamiltonian Kählerian manifolds, Dokl. Akad. Nauk. SSSR 243 (1978), no. 5, 1101–1104.
  • [3] N. Ciccoli, From Poisson to quantum geometry, Notes taken by P. Witkowski, avaliable on http://toknotes.mimuw.edu.pl/sem4/files/Ciccoli_fpqg.pdf.
  • [4] J.-P. Dufour and N. T. Zung, Poisson structures and their normal forms, Prog. Math., vol. 242, Birkhauser, Basel- Boston- Berlin, 2005.
  • [5] V. Ginzburg and D. Kaledin, Poisson deformations of symplectic quotient singularities, Adv. Math. 186 (2004), 1–57, arxiv.org/0212279v5.
  • [6] R. Goto, Rozanski- Witten invariants of log-symplectic manifolds, Integrable systems, topology, and Physics, Tokyo 2000, Contemp. Math., vol. 309, 2002.
  • [7] N. Hitchin, Generalized Calabi-Yau manifolds, Quarterly J. Math. 54 (2003), no. 3, 281–308.
  • [8] R. Lima and J. V. Pereira, A characterization of diagonal Poisson structures, Bull. Lond. Math. Soc. 46 (2014), 1203–1217.
  • [9] M. Matviichuk, B. Pym, and T. Schedler, A local torelli for log symplectic manifolds, Arxiv.math (2020), no. 2010.08692.
  • [10] Y. Namikawa, Flops and poisson deformations of symplectic varieties, Publ. RIMS, Kyoto Univ. 44 (2008), 259–314.
  • [11] A. Polishchuk, Algebraic geometry of Poisson brackets, J. Math. Sci. 84 (1997), 1413–1444.
  • [12] B. Pym, Constructions and classifications of projective Poisson varieties, Arxiv:1701.08852v1 (2017).
  • [13] Z. Ran, Deformations of holomorphic pseudo-symplectic Poisson manifolds, Adv. Math. 304 (2017), 1156–1175, arxiv.org/1308.2442.
  • [14] by same author, A Bogomolov unobstructedness theorem for log-symplectic manifolds in general position., J. Inst. Math. Jussieu (2018), Erratum/corrigendum (2021):; arxiv.org/1705.08366.