ebook img

Poincaré series of algebraic links and lattice homology PDF

0.2 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Poincaré series of algebraic links and lattice homology

POINCARE´ SERIES OF ALGEBRAIC LINKS AND LATTICE HOMOLOGY EUGENEGORSKYANDANDRA´SNE´METHI ABSTRACT. We construct a version of the lattice homology for plane curve singularities us- ingthenormalizationoftheircomponents. Weprovethatthe Poincare´ seriesoftheassociated 3 graded homologiescan be identified by an algebraic procedurewith the motivic Poincare´ se- 1 ries. Hence, for a plane curve singularity the following objects carry the same information: 0 the multi-variableAlexanderpolynomial, the multi-variableHilbertseries associated with the 2 normalization,themotivicPoincare´ series,andthePoincare´seriesofthenewlyintroducedlat- n ticehomology. WealsoconjecturearelationofthislatticehomologywiththeHeegaard–Floer a homologyofthecorrespondinglink. J 1 3 1. INTRODUCTION ] LetC = ∪r C beareducedplanecurvesingularityattheoriginin C2,whereC aretheir- G i=1 i i reduciblecomponents. Letγ : (C,0) → (C ,0)besomeuniformizationsofthesecomponents. A i i Wedefine r integer-valuedfunctionson theC–algebra O = O by . C2,0 h t v (f) = Ord(f (γ (t))), a i i m and aZr-indexedfiltration [ J(v) = {f ∈ O|vi(f) ≥ vi forall i}. 1 v Campillo,DelgadoandGusein-Zadeconsideredin[4]thePoincare´ seriesofthefiltration J(v) 6 defined as theintegralwithrespect totheEulercharacteristicovertheprojectivizationof O: 3 6 7 (1.0.1) P(t ,...,t ) = tv1 ·...·tvrdχ. . 1 r 1 r 1 ZPO 0 ThePoincare´ series P can berelated withseveraltopologicaland analyticalobjects. 3 For example, Campillo, Delgado and Gusein-Zade have shown in [4] that P basically is 1 : the multi-variable Alexander polynomial of the link of C (the intersection of C with a small v i three-dimensionalspherecentered attheorigin). Fortheprecise statementseeTheorem 2.1.1. X A more analytic invariant is the Hilbert series H associated with the filtration {J(v)} , it v r a hascoefficientsh(v) = dimO/J(v),seeDefinition2.2.1. ItisknownthatP andH determine each other, see[3, 11], insection2 wereprovethisfact. Finally,weconsiderthemotivicversionofthePoincare´ series as well([5, 7, 12]): v v (1.0.2) P (t ,...,t ;q) = t 1 ·...·t rdµ. g 1 r 1 r ZPO HereµdenotesthemotivicmeasureonPO[5,7]. SinceP canbederivedfromthecoefficients g ofH aswell,seesubsection2.4,P carriesthesameinformationas H andP. Nevertheless,all g these invariants capture different aspects and highlight different geometrical structures of the local planecurvesingularities. This note introduces another object, the lattice cohomology of the singularity. In [17] the lattice cohomology of a normal surface singularity was introduced via the lattice provided by itsresolutiongraph(orplumbinggraphofthelink). Theinvariantcreatedabridgebetweenthe 1 2 EUGENEGORSKYANDANDRA´SNE´METHI analyticinvariantsof thesingularitywithseveral topologicalones ofthelink,namely Seiberg- Witten and Heegaard-Floertheories. The goal of the present construction is similar; neverthe- less here we rely on a different lattice: the needed weight function is provided by the normal- ization, by the coefficients h(v). (In order to eliminate any further confusion, we will call the present invariant“latticehomologyvianormalization”). In short, the definition runs as follows. The lattice complex L− is generated over Z[U] by elementarycubes (cid:3)ofalldimensionsinRr,withverticesinthelatticeZr. Forsuchacubewe defineh((cid:3)) = maxx∈(cid:3)∩Zr h(x). Thedifferentialisdefined as ∂ ((cid:3)) = ε Uh((cid:3))−h((cid:3)i)(cid:3) , U i i i X where (cid:3) aretheorientedboundary cubesof (cid:3), andε are thecorrespondingsigns. i i The complex L− is naturally Zr-filtered: the subcomplex L−(v) is generated by the cubes contained in the positivequadrant originatingat v. Ourmain theorem describes the homology ofthesubcomplexesL−(v)and theassociatedgraded complexesgr L− forall v. v Theorem 1.0.3. (1) The homology of L−(v) is given by H∗(L−(v)) = Z[U][−2h(v)]. Hence, thePoincare´ series forthishomologyequals t−2h(v) P (t) = . L−(v) 1−t−2 (2) ThePoincare´ polynomialofthehomologyofgr L− isgiven bytheformula v P (t) = (−t)−h(v)H (−t−1), gr L− v v where H (q)is thecoefficient inthemotivicPoincare´ series: v P (t ,...,t ;q) = H (q)tv1 ···t vr. g 1 r v 1 r v X WeprovethepartsofthistheoremasTheorem4.2.2andTheorem4.3.2. Themainingredient isthestructureoftheOrlik-Solomonalgebrasassociatedwithcentralhyperplanearrangements. Corollary1.0.4. Thepolynomial(−1)h(v)H (−t) hasonlynon-negativecoefficients. v Corollary1.0.5. ThegeneratingfunctionfortheEulercharacteristicsofgr L−, v P (−1)·tv1 ···tvr, grvL− 1 r v X equalsP(t ,...,t ). Hence, itcan beexpressed bythemulti-variableAlexander polynomial. 1 r Motivatedby Theorem1.0.3and Corollary 1.0.5, weformulatethefollowing Conjecture1.0.6. ThegradedhomologyofgrL− isisomorphictotheHeegaard-Floerhomol- ogyHF− ofthelinkof C ([21,22, 23, 24], seealso[26]). We compare various structural properties of both homology theories and show that the con- jectureholds for r = 1. We also computethehomologyof grL− in otherexamplesand match themto theHeegaard-Floerhomology. POINCARE´ SERIESOFALGEBRAICLINKSANDLATTICEHOMOLOGY 3 ACKNOWLEDGEMENTS The authors are grateful to S. Gusein-Zade, J. Moyano-Fernandez, P. Ozsva´th and J. Ras- mussen for the useful discussions. We also thank the Oberwolfach Mathematical Institute, where part of this work was done, for the hospitality. The research of E. G. is partially sup- ported by the grants RFBR-10-01-678, NSh-8462.2010.1and the Simons foundation. A. N. is partiallysupportedbyOTKA Grants81203and 100796. 2. HILBERT FUNCTION 2.1. The Poincare´ seriesofmulti-index filtrations. Let usfix alocalplanecurvesingularity withr irreduciblecomponentsasintheintroduction. Set K = {1,...,r}. Lete denotethei- 0 i thcoordinatevectorin Zr. ForasubsetK ⊂ K wewritee = e ande = e = e . 0 K i∈K i K0 i Foravectorv weset v = v e . K i∈K i i P P Furthermore, let P be the series defined in (1.0.1), and ∆(t ,...,t ) be the multi-variable 1 r P Alexanderpolynomialofthelinkof C. Then thefollowingholds. Theorem 2.1.1 ([4]). If r = 1,then P(t)(1−t) = ∆(t), whileif r > 1, then (2.1.2) P(t ,...,t ) = ∆(t ,...,t ). 1 r 1 r 2.2. The Hilbertseries ofmulti-index filtrations. Weset apartial orderon Zr by u (cid:22) v ⇐⇒ u ≤ v forall i. i i We define r integer-valued functions on O = O by v (f) = Ord (f (γ (t))), and a C2,0 i 0 i Zr-indexedfiltration J(v) = {f ∈ O | v(f) (cid:23) v}. NotethattheidealsJ(v)arealsodefinedfornegativevaluesofv. Thisfiltrationisdecreasing inthesensethatifu (cid:22) v, thenJ(u) ⊃ J(v). Definition 2.2.1. Let h(v) = dimO/J(v). Wedefine theHilbert series ofa multi-indexfiltra- tionJ as theseries (2.2.2) H(t ,...,t ) = h(v)·tv1 ···tvr. 1 r 1 r v X Furthermore, we define the set S := {v ∈ Zr : there existsf ∈ O withv(f) = v} as well. It iscalled thesemigroupoff. Fromthedefinitiononegetsthefollowingelementaryfacts. Lemma 2.2.3. S = {v ∈ Zr : h(v +e ) > h(v) forevery i = 1,...,r}. Furthermore, fix ≥0 i any0 (cid:22) v ande . Thenh(v+e ) = h(v)+1ifthereisanelementu ∈ S suchthatu = v and i i i i u ≥ v for j 6= i. Otherwiseh(v +e ) = h(v).In particular,H andS determineeach other. j j i Proof. If h(v + e ) > h(v) for all i, then there exist functions f such that v (f ) = v and i i i i i v (f ) ≥ v for j 6= i. Therefore v( r λ f ) = v for generic coefficients λ . For the second j i j i=1 i i i partnotethath(v+e )−h(v) = dimJ(v)/J(v+e ). Thisquotientspaceistrivialifthereisno i i functionf suchthatv (f) = v andPv (f) ≥ v forj 6= i. Otherwiseitisone-dimensional. (cid:3) i i j j Thefollowinglemmais avariationoftheanalogousstatementfrom[3]. Lemma 2.2.4. Theseries H andP arerelatedby (2.2.5) P(t ,...,t ) = −H(t ,...,t )· (1−t−1). 1 r 1 s i i Y 4 EUGENEGORSKYANDANDRA´SNE´METHI Atthelevel ofcoefficientsthisreads asfollows. Definetheintegersπ(v) bytheequation (2.2.6) P(t ,...,t ) = tv1 ...tvr ·π(v). 1 r 1 r v X Then (2.2.7) π(v) = (−1)|K|−1h(v +e ). K KX⊂K0 We will invert the equation (2.2.7), namely, we express the integers h(v) from the Poincare´ series P . Firstnotethat C (2.2.8) h(v) = h(max(v,0)), where max(v,0) = v if v > 0 and = 0 otherwise. Hence it is enough to determine h(v) for i i i 0 (cid:22) v only. Theseries h(v)tv willbedenotedby H(t)| . 0(cid:22)v 0(cid:22)v Next, for a subset K = {i ,...,i } ⊂ K , K 6= ∅, consider a curve C = ∪ C . As 1 |K| 0 K i∈K i P above, this germ defines a |K|-index filtration of O , hence it provides the Hilbert series C2,0 H ofC invariables {t } : CK K i i∈K H (t ,...,t ) = tvi1 ...tvi|K| ·hK(v). CK i1 i|K| i1 i|K| v X By construction,for K ⊂ K onehasH (t ,...,t ) = H (t ,...,t )| ; or 0 CK i1 i|K| C 1 r ti=0i6∈K (2.2.9) ifv = 0forall i ∈/ K, then hK(v) = h(v). i Analogously,wehavethePoincare´ series of C : K PK(t ,...,t ) = P (t ,...,t ) = tvi1 ...tvi|K| ·πK(v). 1 r CK i1 i|K| i1 i|K| v X Bydefinition,forK = ∅wetakeπ∅(v) = 0.Notethatfromh(v)onecanrecoveranyPK: first computinghK by(2.2.9) andthen recovering PK from HK using(2.2.7). Althoughby[29]P determinestheembeddedtopologicaltypeofC,hencealltheseriesPK forall subsets K ⊂ K as well, theanalogueof(2.2.9)is nottrueforthepair πK(v)and π(v). 0 Indeed, the“restrictingrelation”([28])is oftype 1 (2.2.10) PK0\{1}(t ,...,t ) = P(t ,...,t )| · , 2 r 1 r t1=1 (1−t(C1,C2))···(1−t(C1,Cr)) 2 r where (C ,C ) denotestheintersectionmultiplicityattheoriginofthecomponentsC andC , i j i j i 6= j. Thisalsoshowsthatgiventheseintersectionmultiplicitiesonecan recoverbyinduction PK from P easily. The following theorem was proved in [11] as Corollary 4.3. For the completeness of the expositionwepresentaproofofit(which isslightlydifferent from[11]). Theorem 2.2.11. Thefollowingequationholds: 1 (2.2.12) H(t ,...,t )| = (−1)|K|−1 t ·PK(t ,...,t ). 1 r 0(cid:22)v r i 1 r (1−t ) i=1 i KX⊂K0 (cid:16)iY∈K (cid:17) Proof. The identity(2.2.12) forthQe generating series is equivalentto thefollowingidentityfor theircoefficients: (2.2.13) h(v) = (−1)|K|−1 πK(u). KX⊂K0 0(cid:22)u(cid:22)XvK−eK POINCARE´ SERIESOFALGEBRAICLINKSANDLATTICEHOMOLOGY 5 Wewillprovetheidentity(2.2.13)byatwo-stepinduction: thefirstinductionisbythenumber ofcomponentsr, and thesecond one(forfixed r)is overthenorm |v| = v . i Ifr = 1,then(2.2.7)impliesthatπ(v) = h(v+1)−h(v).Therefore π(u) = h(v) P0≤u≤v−1 sinceh(0) = 0. P Letusprove(2.2.13)forthecasewhenatleastoneofcoordinatesv vanish. Wecanassume i thatv = 0. By theinductionassumptionweget r h(v) = h{1,...,r−1}(v ,...,v ) = (−1)|K|−1 πK(u). 1 r−1 K⊂{X1,...r−1} 0(cid:22)u(cid:22)XvK−eK On the other hand, in (2.2.13) for all K ⊂ K with r ∈ K we get the vacuous restriction 0 0 ≤ u ≤ −1, henceweget nontrivialcontributiononlyfromterms with K ⊂ {1,...r −1}. r Suppose now that v has no vanishing coordinates and we proved (2.2.13) for v −e for all K non-emptysubsetsK ⊂ K . Wecan rewrite(2.2.7)as alinearequationon h(v): 0 π(v −e) = (−1)r−|K|−1h(v −e ). K KX⊂K0 By theinductionassumptionforK 6= ∅ wehave h(v −e ) = (−1)|M|−1 πM(u), K MX⊂K0 0(cid:22)u(cid:22)(vMX−eK∩M−eM) and weshouldestablishthesameidentityforK = ∅. Thereforewehavetoprovethat (2.2.14) π(v−e) = (−1)r−|K|+|M| πM(u). KX⊂K0MX⊂K0 0(cid:22)u(cid:22)(vMX−eK∩M−eM) Let us fix M and u (cid:22) v − e and sum the expression (−1)|K| over all sets K ⊂ K such that 0 u ≤ v −2 for i ∈ K ∩M. This sum does not vanish iff M = K and u = v −1 for all i. i i 0 i i Thisproves(2.2.14). (cid:3) Corollary 2.2.15. The restricted Hilbert function H(t)| of a multi-component curve is a 0(cid:22)v rationalfunctionwith denominator r (1−t )2. i=1 i Example2.2.16. ConsiderthesinguQlarityA . ItsPoincare´ seriesis1+t t +···+(t t )n−1, 2n−1 1 2 1 2 and thePoincare´ series ofbothitscomponentsequals 1/(1−t). TheHilbertseries is 1 t t H(t ,t )| = 1 + 2 −t t (1+...+(t t )n−1) . 1 2 0(cid:22)v 1 2 1 2 (1−t )(1−t ) 1−t 1−t 1 2 (cid:18) 1 2 (cid:19) Therefore, fornon-negativeintegers (v ,v ) onehas 1 2 max(v ,v ), if min(v ,v ) < n 1 2 1 2 h(v) = v +v −n, otherwise. ( 1 2 Figure1illustratesthisformulafortheHilbertfunctionforA singularity. Thepointscorre- 3 spondingto thesemigroup S aremarked inbold. Example2.2.17. Considerthesingularity D ,that is,equationy ·(x2 −y3) = 0. Then 5 1 1−t +t2 P(t ,t ) = 1+t t3,P (t ) = ,P (t ) = 2 2. 1 2 1 2 1 1 1−t 2 2 1−t 1 2 6 EUGENEGORSKYANDANDRA´SNE´METHI 4 4 4 5 6 3 3 3 4 5 2 2 2 3 4 1 1 2 3 4 0 1 2 3 4 FIGURE 1. ValuesoftheHilbertfunctionfor A 3 4 4 4 5 6 7 3 3 3 4 5 6 2 2 3 4 5 6 1 1 2 3 4 5 1 1 2 3 4 5 0 1 2 3 4 5 FIGURE 2. ValuesoftheHilbertfunctionfor D 5 Onecan check that h(v ,v )fornon-negativev and v is (seeFigure2 too): 1 2 1 2 v , ifv < 3,v > 0 1 2 1 v +1, ifv = 3,v > 0  1 2 1 h(v ,v ) = v −1, ifv < 2,v ≥ 2 1 2  2 1 2  v +v −3, ifv ≥ 2,v ≥ 4 1 2 1 2 0,1,1, ifv = 0 and v = 0,1,2.  1 2      2.3. Some properties of the Hilbert function. For any i ∈ K let µ and δ (respectively 0 i i µ(C)andδ(C))betheMilnornumberandthedeltainvariantofC (respectivelyofC)([1,10]). i Then, cf. [10], µ = 2δ , andµ(C)+r −1 = 2δ(C). Definel = (l ,...,l ) by i i 1 r l = µ + (C ,C ). i i j i j6=i X It isknownthattheAlexanderpolynomialissymmetricinthe followingsense ∆(t−1,...,t−1) = t1−li ·∆(t ,...,t ) for r > 1, 1 r i 1 r (cid:16)Y (cid:17) ∆(t−1) = t−µ(C)∆(t) for r = 1. Thiscan becompared withthefollowing. POINCARE´ SERIESOFALGEBRAICLINKSANDLATTICEHOMOLOGY 7 Lemma 2.3.1. ([12, 6])TheHilbertfunctionsatisfiesthefollowingsymmetryproperty: (2.3.2) h(l −v)−h(v) = δ(C)−|v|, where |v| = r v . i=1 i Remark 2.3P.3. Consider v (cid:23) l. It follows from (2.3.2) that h(v) = |v| − δ(C). This can be verified by the identity (2.2.13) as well. Indeed, for |K| = 1 we have πK(u) = 0≤ui≤vi−1 v −δ(C ),for|K| = 2wehave π{i,j}(u ,u ) = P{i,j}(1,1) = (C ,C ),whilePK(e ) = 0 i i i j i jP K for|K| > 2. Hence h(v) = (v −δ(C ))− (C ,C ) = |v|−δ(C). i i i i,j i j P Theidentityh(v) = |v|−δ(C)andLemma2.2.3givethatv ∈ S wheneverv (cid:23) l (hence, in P P fact, l is theconductorof S). Thisfact combinedagain with2.2.3gives Corollary2.3.4. Foranybasicvectore andn ≥ l onehash(v+(n+1)e )−h(v+ne ) = 1. i i i i Thenextpropertyis notusedinthepresentnote, neverthelessweadditsinceitcontainsthe keyobservationwhichwillbeusedin aforthcomingarticlewithapplicationsindeformations. Proposition 2.3.5. (a) Let f′ be a deformation of f, where f and f′ are irreducible. Then h (v) ≤ h (v)for everyv. f f′ (b) Let a (possibly reducible) curve C′ be a deformation of an irreducible curve C. Then h (v) ≥ h (|v|)forevery v. C′ C Proof. (a) Forany function g theintersection multiplicityof g with f is greaterorequal to the intersectionmultiplicityof g withf′. Therefore J (v) ⊂ J (v)and f′ f h (v) = codimJ (v) ≤ codimJ (v) = h (v). f f f′ f′ (b)ConsiderafunctiongfromJ (v). ItsordersonthecomponentsofC′aregreaterorequalto C′ thecorrespondingcomponentsofv,hencetheintersectionmultiplicityofgwithC′isgreateror equalto|v|. Hencetheintersectionmultiplicityofg withC isgreaterorequalto |v|. Therefore J (v) ⊂ J (|v|)and h (|v|) = codimJ (|v|) ≤ codimJ (v) = h (v). (cid:3) C′ C C C C′ C′ 2.4. MotivicPoincare´ series. Following[5], wedefinethemotivicPoincare´ series ofaplane curvesingularity C = ∪r C bytheformula i=1 i 1 P (t ,...,t ;q) = tv1 ···tvr (−1)|K|qh(v+eK). g 1 r 1−q 1 r vX∈Zr KX⊂K0 It follows from the results of [5] that this definition agrees with the integral (1.0.2). In [7], and independently in [12], the following properties are proved: P (t ,...,t ;q) is a rational g 1 r functionwithdenominator r (1−t q),hence i=1 i r Q P (t ,...,t ;q) := P (t ,...,t ;q)· (1−t q) g 1 r g 1 r i i=1 Y isapolynomial. Moreover, P satisfies thefunctionalequation g 1 1 P ( ,..., ;q) = q−δ(C) t−li ·P (t ,...,t ;q). g qt qt i g 1 r 1 r i Y In [12] this equation was deduced from (2.3.2). Moreover, one can check that for q = 1 one has P (t ,...,t ;q = 1) = P(t ,...,t ). An explicit algorithm of the computation of g 1 r 1 r P (t ,...,t ;q) intermsoftheembeddedresolutiontreeof C isprovidedin[7]. g 1 r 8 EUGENEGORSKYANDANDRA´SNE´METHI 2.5. Conclusion. By the above discussions, the following objects associated with a plane curvesingularitycarry thesameamountofinformation: ∆, S, H,P andP . g 3. CENTRAL HYPERPLANE ARRANGEMENTS 3.1. Matroids and rank functions. Definition 3.1.1. (A) ([27]) Let K be a finite set. A function ρ, assigning a non-negative 0 integertoany subset K ⊂ K , iscalled arankfunction,if 0 (1) 0 ≤ ρ(K) ≤ |K|(where |K|denotesthecardinalityof K). (2) IfK ⊂ K thenρ(K ) ≤ ρ(K ). 1 2 1 2 (3) Foreverypairofsubsets K and K onehasthefollowinginequality: 1 2 ρ(K ∩K )+ρ(K ∪K ) ≤ ρ(K )+ρ(K ). 1 2 1 2 1 2 (B)Amatroidisa finiteset witharank functionon it. (C)Thecharacteristicpolynomialofamatroid M = (K ,ρ) isdefined as 0 χ (t) = (−1)|K|tρ(K0)−ρ(K). M KX⊂K0 Remark 3.1.2. Some authors define the characteristic polynomial using the Mo¨bius function ofamatroid. Thisdefinitionisequivalenttothepresent one,seee.g. [27, Theorem2.4]. Leth(v)denotetheHilbertfunctionofaplanecurvesingularity. LetusfixK = {1,...,r} 0 and forevery v considerthefollowingfunctiononsubsetsof K : 0 ρ (K) := h(v +e )−h(v) = dimJ(v)/J(v +e ). v K K Proposition3.1.3. Foreveryv thefunctionρ is arankfunctiononK . v 0 Proof. Property(1)followsfromLemma2.2.3. Next,forK ⊂ K onehasρ (K )−ρ (K ) = 1 2 v 2 v 1 dimJ(v + e )/J(v + e ) ≥ 0. To prove the third property notice that J(v + e ) = K1 K2 K1∪K2 J(v +e )∩J(v +e ), and J(v +e )+J(v+e ) ⊂ J(v +e ), hence K1 K2 K1 K2 K1∩K2 ρ (K )+ρ (K )−ρ (K ∪K ) = dimJ(v)/(J(v +e )+J(v +e )) ≥ ρ (K ∩K ). v 1 v 2 v 1 2 K1 K2 v 1 2 (cid:3) We will call ρ the rank function for a the “local matroid M ”. In the space J(v) we have r v v subspaces J(v +e ) of codimension 0 or 1. If v ∈ S, then the set of functions with valuation i v can be represented as a complement of a hyperplane arrangement (cf. [12]). If v 6∈ S, then J(v) = J(v +e ) forsomei , cf. Lemma2.2.3. Moreover,onehasthefollowinglemma. i0 0 Lemma 3.1.4. Assume that J(v) = J(v + e ) for some i ∈ K . Then J(v + e ) = J(v + i0 0 0 K e +e ) foranyK ⊂ E with K 6∋ i . Hence K i0 0 χMv(t) = (−1)|K| ·th(v+eK0)−h(v+eK) = 0. KX⊂K0 Proof. Use J(v + e + e ) = J(v + e ) ∩ J(v + e ) for the first statement, and pairwise K i0 K i0 cancellationforthesecond one. (cid:3) POINCARE´ SERIESOFALGEBRAICLINKSANDLATTICEHOMOLOGY 9 3.2. Central hyperplane arrangements andOrlik-Solomonalgebras. Letusrecallsomefactsaboutcentralhyperplanearrangements. LetV beavectorspaceand let H = {H ,...,H } be acollection oflinearhyperplanes in V. Fora set K = {α ,...,α } 1 r 1 k wedefinetherank function ρ(K) = codim(H ∩...∩H ). α1 αk SimilarlytoProposition3.1.3,onecancheckthatρisarankfunctionforacertainmatroid. Let usdenoteby χ (t) itscharacteristicpolynomial. H Lemma3.2.1. TheclassintheGrothendieckringofvarietiesofthecomplementof ∪ H inV i i equals [V \∪r H ] = LdimV−ρ(K0)χ (L), i=1 i H where Ldenotestheclassof theaffineline. Proof. Followsfrom theinclusion-exclusionformula: [V \∪r H ] = (−1)|K| [∩ H ] = (−1)|K| LdimV−ρ(K). (cid:3) i=1 i α∈K α KX⊂K0 KX⊂K0 To the arrangement H one can associate the corresponding Orlik-Solomon algebra. Con- sider the anticommutative algebra E generated by the variables z ,...,z corresponding to 1 r hyperplanes. For any set K ⊂ K we consider the monomial z = z ∧ ... ∧ z , where 0 K α1 αk K = {α ,...,α }.Wecan equipE withthenatural differential ∂ sendingz to 1,namely 1 k i k ∂(z ) = (−1)i−1z . K K\{αi} i=1 X Definition 3.2.2. We call the set K dependent, if the equations of the corresponding hyper- planesare linearlydependent. Otherwise K iscalled independent. TheOrlik-SolomonidealI istheidealinE generatedbytheelements∂z foralldependent K setsK. TheOrlik-Solomonalgebrais thequotient A = E/I. Theorem 3.2.3. ([18, Theorem 5.2]) The integral cohomology ring of the complement V \ ∪r H is isomorphic to the Orlik-Solomon algebra E/I. It has no torsion, and its Poincare´ i=1 i polynomialis givenbytheformula P(H,t) = (−t)r(E) ·χ (−t−1) = (−1)|K|(−t)ρ(K). H KX⊂K0 As a corollary, we conclude that the homology of V \ ∪r H is defined by its class in the i=1 i Grothendieckring. Thesameistrueforitsprojectivization(seebelow). Wewillneed thefollowingdeformationsofthedifferential on E. Definition 3.2.4. Let usdefine thefollowingoperator: k ∂ : E[U] → E[U], ∂ (z ) = (−1)i−1Uρ(K)−ρ(K\{αi})z , U U K K\αi i=1 X where U isaformal parameterand K = {α ,...,α }. 1 k Notethatρ(K)−ρ(K\{α }) ∈ {0,1},hence∂ decomposesintoasumoftwocomponents i U (3.2.5) ∂ = ∂ +U∂ , with ∂ +∂ = ∂. U 0 1 0 1 10 EUGENEGORSKYANDANDRA´SNE´METHI Lemma 3.2.6. The operator ∂ is a differential on E[U], that is, ∂2 = 0. In particular, the U U followingidentitieshold: ∂2 = ∂2 = 0, ∂ ∂ +∂ ∂ = 0. 0 1 0 1 1 0 Proof. Straightforward. (cid:3) Let J and J⊥ denote the subspaces of E spanned by the elements z for all dependent, K respectivelyindependentsubsets K. Clearly E = J ⊕J⊥. Lemma 3.2.7. Thefollowingstatementshold: (a) ([18], Lemma 2.7) I = J +∂J. (b) ∂ J⊥ = 0, henceIm∂ = ∂ J. 0 0 0 (c) ∂ J ⊂ J, henceI = J +∂J = J +∂ J. 1 0 (d) Ker∂ = J⊥ +Im∂ . 0 0 (e) There existsubspacesA ⊂ J,B ⊂ J⊥ suchthatIm∂ = A⊕B. 0 Proof. The claims (b) and (c) are clear. Let us prove(d). The inclusionJ⊥ +Im∂ ⊂ Ker∂ 0 0 is also clear, hence we need to prove that if ∂ (φ) = 0 then there exists φ ∈ J⊥ such that 0 φ−φ ∈ Im(∂ ). 0 Let us call z essential in a monomial z ∧ z , if ρ({i} ⊔ K) = ρ(K) +e1, and redundant i i K otherewise. Letusdecomposeφ = z ∧φ +z ∧φ +φ ,wherez isessentialineverymonomial 1 1 1 2 3 1 ofz ∧φ ,redundant ineverymonomialof z ∧φ , andφ containsno z . 1 1 1 2 3 1 Then ∂ (φ) = z ∧ψ +φ +∂ (φ ) 0 1 2 0 3 forsomeψ, and neitherφ nor∂ (φ ) containz . Hence, if∂ (φ) = 0then φ = −∂ (φ ). 2 0 3 1 0 2 0 3 Since z is redundant in every monomial in z ∧ ∂ (φ ), it is redundant in every monomial 1 1 0 3 inz ∧φ . Therefore 1 3 ∂ (z ∧φ ) = φ −z ∧∂ (φ )+z ∧η, 0 1 3 3 1 0 3 1 where z is essential in every monomial of z ∧ η. Indeed, if α ∈ K, z is redundant in 1 1 j αj K ∪{1}and essentialinK, then z is essentialinK ∪{1}\{α }. 1 j Weconcludethat φ−∂ (z ∧φ ) = z ∧(φ −η) 0 1 3 1 1 andz isessentialineverymonomialintherighthandside. Now∂ (φ) = −z ∧∂ (φ −η) = 0, 1 0 1 0 1 hence ∂ (φ − η) = 0. Then we can repeat the procedure inductively replacing φ by φ − η, 0 1 1 and z by z , etc. At the end we reduce φ modulo Im(∂ ) to an element of E where all z are 1 2 0 i essential;suchan element belongsto J⊥. Next, we prove (e). Recall that Im∂ = ∂ J and K is dependent iff ρ(K) < |K|. If 0 0 the monomial z appears in ∂ (z ) then ρ(K) = ρ(K′) and |K′| = |K| − 1. Therefore K′ 0 K ∂ (z ) ∈ J⊥ ifρ(K) = |K|−1,and ∂ (z ) ∈ J otherwise. (cid:3) 0 K 0 K Wewillneed thefollowingtwo resultsintheconstructionof latticehomology. Theorem 3.2.8. The Orlik-Solomon algebra is isomorphic to the homology of the differential ∂ : 0 H (E,∂ ) = E/I ≃ H∗(V \∪r H ). ∗ 0 i=1 i Proof. By Lemma3.2.7onehastheisomorphisms Ker∂ = J⊥ +Im∂ , Im∂ = ∂ J, 0 0 0 0 hence H (E,∂ ) = (J⊥ +∂ J)/∂ J ≃ J⊥/(∂ J ∩J⊥) ≃ E/(J +∂ J). ∗ 0 0 0 0 0

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.