ebook img

Regularity and b-functions for D-modules PDF

0.18 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 Regularity and b-functions for D-modules

REGULARITY AND B-FUNCTIONS FOR D-MODULES YVES LAURENT 6 1 0 2 Abstract n AholonomicD-moduleonacomplexanalyticmanifoldadmitsalwaysab-functionalong u any submanifold. If the module is regular, it admits also a regular b-function, that is a J 8 b-function with a condition on the order of the lower terms of the equation. There is a weaker condition of regularity: regularity along a submanifold. We prove that a module ] which is regular along a submanifold admits a regular b-function along this submanifold. P A . h t a m [ 2 v 2 3 9 6 0 . 1 0 5 1 : v i X r a 2000 Mathematics Subject Classification. 35A27. Key words and phrases. D-module, regular, b-function. 2 YVESLAURENT Introduction Let f be a holomorphic function on an open set X of Cn. A b-function for f is a polynomial b ∈ C[s] such that there exists a differential operator P on U with parameter s satisfying an equation: (1) b(s)f(x)s = P(s,x,D )f(x)s+1 x Thegenerator oftheidealoftheb-functionsassociated tof isusuallycalled theBernstein- Sato polynomial of f (see[1] for details). This definition has been extended by Kashiwara [3] to holonomic D-modules. Let X be a complex manifold and Y a smooth hypersurface. Let D be the sheaf of differential X operators on X. Let M be a holonomic D -module and u a section of M. A b-function X for u along Y is an equation (2) b(tD )u= tP(t,x,tD ,D )u t t x satisfied by u. Here (t,x) are local coordinates of X such that t is an equation for Y. A similar definition exits for submanifolds of X of any codimension. The b-function is called regular if P may be chosen so that its order as a differential operator is not greater than the degree of b. It has been proved by Kashiwara [3] that a holonomic D-module admits a b-function along any smooth hypersurface and by Kashiwara-Kawa¨ı [5] that a regular holonomic D-module admits a regular b-function along any smooth hypersurface. Thereis another notion of regularity, theregularity of aD-modulealong asubmanifold. The definition will be given in definitions 1.3 and 3.1. In the case of a hypersurface, this is equivalent to the fact that solutions in the formal completion transversally to the hypersurface are convergent. It has been proved that a regular holonomic module is regular along any submanifold in [6]. It is also a direct consequence of the definition that if a D-module admits a regular b-function along a submanifold, it is regular along it. We will prove here that a module which is regular along a submanifold Y admits a regular b-function along Y. The problem is better understood after microlocalization, that is for modules over the sheaf E of microdifferential operators. Regularity and b-function may be defined for a X coherent E -module M along a conic lagrangian submanifold Λ of the cotangent bundle X [8]. All definitions are invariant under quantized canonical transformation. Concerning D-modules, this will prove that the result is true not only for hypersurfaces but may be extended to any smooth subvariety of X. AregularholonomicE -moduleisregularalonganysmoothconiclagrangian subvariety X of T∗X. Conversely, in the definition of Kashiwara-Kawa¨ı [6], a holonomic E -module is X regular if it is regular along the smooth part of its characteristic variety. In the first section of this paper, we briefly recall the definitions of b-functions and state the result for D -modules. In the second section, we study the microlocal case. We X define the filtrations and bifiltrations on holonomic E -modules In the third section, we X study the equivalent definitions of regularity along a lagrangian manifold and we prove the existence of a regular b-function as a product of θd by the b-function b(θ). 1. The case of differential equations In this first section, we briefly recall the definitions of b-functions and regularity in the framework of D -modules. X REGULARITY AND B-FUNCTIONS FOR D-MODULES 3 1.1. Filtrations and b-functions. Let X be a complex analytic manifold, O the sheaf X of holomorphic functions on X and D the sheaf of differential operators on X with X coefficients in O . Let T∗X be the cotangent bundle to X with canonical projection X π : T∗X → X. Let Y be a submanifold of X. The sheaf D is provided with two filtrations. The first one is the filtration (D ) X X,m m≥0 bytheusualorder,thatisthedegreeinthederivations. Here, wewilldenotethisfiltration by FD , i.e. F D = D . The corresponding graded ring grFD is identified to X m X X,m X π∗O[T∗X] the sheaf of holomorphic functions on T∗X with polynomial coefficients in the fibers of π. The second one is the V-filtration which has has been defined by Kashiwara in [4] as: (1.1) V D = {P ∈ D | | ∀ℓ ∈Z,PIℓ ⊂ Iℓ−k} k X X Y Y Y where I is the ideal of definition of Y and Iℓ = O if ℓ ≤ 0. Y Y X Let τ : T X → Y bethe normal bundleto Y in X and O the sheaf of holomorphic Y [TYX] functions on T X which are polynomial in the fibers of τ. Let O [k] be the subsheaf Y [TYX] of O of homogeneous functions of degree k in the fibers of τ. There are canonical iso- [TYX] morphisms between Ik/Ik−1 and τ O [k], between Ik/Ik−1 and τ O . Hence Y Y ∗ [TYX] L Y Y ∗ [TYX] the graded ring grVD associated to the V-filtration on D acts naturally on O . An X X [TYX] easy calculation [14] shows that as a subring of End(τ O ) it is identified to τ D ∗ [TYX] ∗ [TYX] the sheaf of differential operators on T X with coefficients in O . Y [TYX] The Euler vector field θ of T X is the vector field which acts on O [k] by multipli- Y [TYX] cation by k. Let ϑ be any differential operator in V D whose image in grVD is θ. Let 0 X 0 X M be a coherent D -module and u a section of M. X Definition 1.1. A polynomial b is a b-function for u along Y if there exists a differential operator Q in V D such that (b(ϑ)+Q)u = 0. −1 X Definition 1.2. A polynomial b of degree m is a regular b-function for u along Y if there exists a differential operator Q in V D ∩F D such that (b(ϑ)+Q)u = 0. −1 X m X The set of b-functions is an ideal of C[T], if it is not zero we call a generator of this ideal ”the” b-function of u along Y. Remark that the set of regular b-function for u is not always an ideal of C[T]. 1.2. Regularity. A holonomic module is regular its formal solutions converge at each point. More precisely, if x is a point of X, O the ring of germs of O at x and m X,x X the maximal ideal of O , let us denote by O the formal completion of O for the X,x X,x X,x m-topology. The holonomic D -module M isbregular if and only if: X j j ∀j ≥ 0, ∀x ∈ X, Ext (M,O )= Ext (M,O ) DX bX,x DX X,x This is the direct generalization of the definition in dimension 1 (Ramis [12]). An algebraic and microlocal definition is given by Kashiwara-Kawa¨ı in [6] and they prove that, for a D -module, it is equivalent to this one. X A weaker notion is the regularity along a submanifold. This has been studied in several papers [7],[8],[11]. The definition will be given in section 3 using microcharacteristic va- rieties. Here we give a more elementary definition for D -modules which works only for X hypersurfaces. 4 YVESLAURENT Let Y be a smooth hypersurface of X, ϑ is the vector field of section 1.1 and M be a holonomic D -module M defined in a neighborhood of Y. X Definition 1.3. Theholonomic D -moduleMis regularalong Y ifany section uof Mis X annihilatedbyadifferentialoperatoroftheformϑN+P+QwhereP isinF D ∩V D N−1 X 0 X and Q is in F D ∩V D N X −1 X Let us denote by O the formal completion of O along Y, that is X|Y X b O = limO /Ik X|Y ←− X Y b k We proved in [9] and [11] that M is regular along Y if and only if j j (1.2) ∀j ≥ 0, Ext (M,O ) = Ext (M,O ) DX bX|Y DX X We can still state definition 1.3 and equation (1.2) when Y is not a hypersurface but they are not equivalent. Correct statements will be given in section 2 in the microlocal framework. It has been proved by Kashiwara in [3], that if M is a holonomic D -module, there X exists a b-function for any section u of M along any submanifold Y of X. Kashiwara and Kawa¨ı proved in [5] that if M is a regular holonomic D -module, there exists a regular X b-function for any section u of M along any submanifold Y of X. We will prove in section 3.2 that if M is a holonomic D -module which is regular along X a submanifold Y then it admits a regular b-function along Y. 1.3. Explicit formulas in local coordinates. Consider local coordinates (x ,...,x ,t) 1 p ofX suchthatY isthehypersurfaceY = {(x,t) ∈ X | t = 0}. ThenT X hascoordinates Y (x ,...,x ,t˜). 1 p The Euler vector field θ of T X is θ = t˜D and we can choose ϑ = tD . Then a Y t˜ t b-function is an equation b(tD )+tQ(x,t,D ,tD ) t x t and it is a regular b-function if the order of Q is less or equal to the degree of b. The module M is regular along Y if any section of M is annihilated by an equation (tD )N +P(x,D ,tD )+tQ(x,t,D ,tD ) t x t x t with P of order less or equal to N −1 and Q of order less or equal to N. 2. Microdifferential equations In this section, we review basic definitions of filtration, V-filtration and bifiltrations on E -modules. Details may be founded in [1] or [14], and [8]. X 2.1. V-filtration on microdifferential operators. We denote by E the sheaf of mi- X crodifferential operators of [13]. It is filtered by the order, we will denote that filtration by E = F E and call it the usual filtration. X S k X Let O(T∗X) be the sheaf of holomorphic functions on T∗X which are finite sums of homogeneousfunctionsinthefibersofπ :T∗X → X. ThegradedringgrFE isisomorphic X to π∗O(T∗X) [14]. In [8], we extended the definitions of V-filtrations and b-functions to microdifferential equations and lagrangian subvarieties of the cotangent bundle. REGULARITY AND B-FUNCTIONS FOR D-MODULES 5 Let Λ be a lagrangian conic submanifold of the cotangent bundle T∗X and M be a Λ holonomic E -module supported by Λ. A section u of MΛ is non degenerate if the ideal X of OT∗X generated by the principalsymbols of the microdifferential operators annihilating u is the ideal of definition of Λ. The module MΛ is a simple holonomic E -module if it is X generated by a non degenerate section u . Such a module always exists locally [13]. Λ Let M = E u . The V-filtration on E along Λ is defined by: Λ,k X,k Λ X (2.1) V E = {P ∈E | | ∀ℓ ∈Z,PM ⊂ M } k X X Λ Λ,ℓ Λ,ℓ+k This filtration is independent of the choices of M and u , so it is globally defined [8, Λ Λ Prop. 2.1.1.]. Let O [k] be the sheaf of holomorphic functions on Λ homogeneous of degree k in the Λ fibersofΛ→ X andO = O [k]. ThereisanisomorphismbetweenM /M (Λ) Lk∈Z Λ Λ,k Λ,k−1 and O . By this isomorphism the graded ringgrVE acts on O and may beidentified Λ,k X (Λ) to the sheaf D of differential operators on Λ with coefficients in O . (Λ) (Λ) All these definitions are invariant under quantized canonical transformations [8]. The restriction to the zero section T∗X of T∗X of the sheaf E is the sheaf D of X X X differential operators. If Λ is the conormal bundle T∗X to a submanifold Y of X then Y the V-filtration induced on D by the V-filtration of E is the same than the V-filtration X X defined in section 1.1. The associated graded ring is the restriction to Y of D which is (Λ) the sheaf D of differential operators with coefficients polynomial in the fibersof Λ → X. [Λ] The correspondence between the isomorphism grVD ≃ D of section 1.1 and the X [TYX] isomorphism grVDX ≃ D[T∗X] is given by the partial Fourier transform associated to the Y duality between the normal bundle T X and the conormal bundle T∗X. Y Y Remark 2.1. Consider local coordinates (x ,...,x ,t ,...,t ) of X and Y = {(x,t) ∈X | 1 p 1 d t = 0}. Let (x,t˜) be the corresponding coordinates of T X and (x,τ) the corresponding Y coordinates of T∗X. Let θ = d t˜D be the Euler vector field of the fiber bundleT X Y Pi=1 i t˜i Y and θ be the Euler vector field of Λ= T∗X. Λ Y The Fourier transform is given by t˜ 7→ −D and D 7→ τ . So θ is mapped to −d−θ i τi t˜i i Λ where d is the codimension of Y. 2.2. Filtrations on E-modules. Let M be a (left) coherent E -module. A F-filtration X (resp. a V-filtration) of M is a filtration compatible with the F-filtration (resp. V- filtration) of E . X N A good filtration of M is a filtration which is locally of the form M = E u k Pi=1 X,k−ki i for (u ,...,u ) local sections of M and (k ,...,k ) integers. In the same way, a good 1 N 1 N N V-filtration of M is locally of the form V M= V E u . k Pi=1 k−ki X i If M is provided with a good F-filtration, the associated graded module grFM is a coherent module over grFEX = π∗O(T∗X). Then OT∗X ⊗π−1grFEX π−1grFM is a coherent OT∗X-module which defines a positive analytic cycle on T∗X independent of the good filtration. This cycle is the characteristic cycle of M and denoted by Ch(M). Its support is thecharacteristic variety ofMdenotedCh(M), itis equaltothesufpportofthemodule M itself. The characteristic variety of a coherent E -module is a homogeneous involutive subva- X riety of T∗X. When its dimension is minimal, that is when it is a lagrangian, the module is holonomic. 6 YVESLAURENT If M is provided with a good V-filtration, the graded module grVM is a coherent module over grVE = D hence grVM is a coherent D -module. The characteristic X (Λ) (Λ) cycle of this module is a positive analytic cycle on T∗Λ and its support is involutive. According to [8] and [11], this characteristic cycle is called the microcharacteristic cycle of M of type ∞ and is denoted by ChΛ(∞)(M). The corresponding microcharacteristic f variety is ChΛ(∞)(M). We have the following fundamental result: Theorem 2.2 (Theorem 4.1.1 of [8]). The dimension of the characteristic variety of grVM is less or equal to the dimension of the characteristic variety of M. In particular, grVM is holonomic if M is holonomic. As grVE is identified to the subsheaf of D of differential operators P satisfying k X (Λ) [θ,P] = kP, the Euler vector field θ acts on grVM, so we may define a morphism Θ by Θ = θ−k on grVM. As [θ,P]= +kP on grVE , Θ commutes with the action of grVE k k X X that is defines a section of End (grVM). grVEX Definition 2.3. The set of polynomials b satisfying b(Θ)grV(M) = 0 is an ideal of C[T]. When this ideal is not zero, a generator is called the b-function of M for the V-filtration along Λ. The b-function depends on the good V-filtration, its roots are shifted by integers by change of V-filtration. Theb-functionisinvariantunderquantizedcanonicaltransform. LetY beasubmanifold of X of codimension d. If we want to identify the b-function of definition 2.3 with that of definition 1.1, we have to replace ϑ in definition 1.1 by −d−ϑ because of remark 2.1. Theorem 2.4. If M is a holonomic E -module, there exists a b-function for M along X any lagrangian submanifold Λ of T∗X. This theorem has been proved first by Kashiwara for D-modules in [3]. It has been proved for E -module in [8] as a corollary of theorem 2.2. In fact, as grVM is holonomic, X End (grVM) is finite dimensional and the b-function is the minimal polynomial of Θ grVEX on grVM. 2.3. Bifiltration. From the two filtrations on E , we get a bifiltration: X (2.2) ∀(k,j) ∈ Z×Z W E = F E ∩V E k,j X j X k X To this bifiltration is associated the bigraded ring: (2.3) grWEX = MgrkW,jEX with grkW,jEX = Wk,jEX/(Wk−1,jEX +Wk,j−1EX) (k,j) This bigraded ring is equal to the graded ring of grVE = D (this latter with the X (Λ) standard filtration). So it is isomorphic to πΛ∗O[T∗Λ], the sheaf of holomorphic functions on T∗Λ which are polynomial in the fibers of π : T∗Λ → Λ and sum of homogeneous Λ functions for the second action of C∗ on T∗Λ. ”Second action” means action induced on T∗Λ by the action of C∗ on Λ. Let M be a (left) coherent E -module. A good bifiltration of M is a bifiltration com- X patible with the bifiltration of E which is locally generated by local sections (u ,...,u ) X 1 N REGULARITY AND B-FUNCTIONS FOR D-MODULES 7 ofMThismeansthatthereareintegers (µ ,...,µ )and(ν ,...,ν )suchthatW M = 1 N 1 N k,j N W E u . Pν=1 k−kν,j−jν X ν The bigraded module associated to the bifiltration is defined by (2.4) grWM = MgrkW,jM with grkW,jM= Wk,jM/(Wk−1,jM+Wk,j−1M) (k,j) Proposition 2.5. (1) The bigraded module associated to a good bifiltration of a coherent E -module is a X coherent grWEX = πΛ∗O[T∗Λ] module. (2) The bifiltration induced on a coherent submodule by a good bifiltration is still a good bifiltration. This result may be found in [2] or [8], let us briefly recall its proof. We denote by E the sheaf grVE provided with the filtration induced by that of E V 0 X X (cf [6]). Lemma 2.6. Outside of the zero section of Λ, there isan equivalence between the following data: • a good bifiltration of M • a coherent sub-E -module of M which generates M and is provided with a good V filtration. Proof. Outsideofitszerosection, wemaytransformΛintotheconormaltoahypersurface by a canonical transformation, that is in local coordinates: Λ= {(x,t,ξ,τ) ∈ T∗X | t = 0,ξ = 0,τ 6= 0} Then the microdifferential operator D is invertible. It is an element of W E . t 1,1 X ConsideragoodbifiltrationWMofMwhichislocallygeneratedbysectionsu ,...,u 1 N ofM. MultiplyingthembysuitablepowersofD wemay assumethat(u ,...,u )belong t 1 N to N = [ W0,jM j∈Z Then N is the sub-E -module of M generated by (u ,...,u ) and, by [6, Prop. 1.1.10.], V 1 N N is an E -coherent submodule of M. The sections (u ,...,u ) define a good filtration V 1 N of N and M= E ⊗N. X Conversely, let N be coherent sub-E -module of M which generates M and is provided V with a good filtration. The good filtration is given locally by sections (u ,...,u ) which 1 N define a good bifiltration of M. (cid:3) Proof of proposition 2.5. From [7, §2.6.] (see also [2]), we know that he filtration of E is V ”a good noetherian filtration” that is a zariskian filtration in the notation of [14]. Then the results of [14] show that good E -filtrations induce good E -filtrations on submodules V V and that the graded ring associated to a good E -filtration is a coherent grE -module. V V The result may be then deduce from the previous lemma out of the zero section of Λ. We get the result on the zero section of Λ by adding a dummy variable as usual (cf [6, §A1]). (cid:3) 8 YVESLAURENT Let OT∗Λ be the sheaf of holomorphic functions on T∗Λ. As grWM is a πΛ∗O[T∗Λ] coherent module, OT∗Λ ⊗O[T∗Λ] πΛ−1grM is a coherent OT∗Λ-module. Its support is in- dependent of the good bifiltration. It is called the microcharacteristic variety of type (∞,1) and is denoted by ChΛ(∞,1)(M). This microcharacteristic variety is an involutive bihomogeneous subvariety of T∗Λ [7],[8]. Let us remark that the same microcharacteristic variety has been defined by Teresa Monteiro Fernandes in [2], where it is denoted by C (M), and in [7] as the support of the Λ tensor product of the module by a sheaf of 2-microdifferential operators. 2.4. Local coordinates. Assume that the conic lagrangian manifold Λ is the conormal to a submanifold Y of X. (By a canonical transformation we may always transform Λ into a conormal). Let (x ,...,x ,t ,...t ) be local coordinates of X such that Y = {(x,t) ∈ X | t = 0}. 1 p 1 d Then Λ = {(x,t,ξ,τ) ∈ T∗X | t = 0,ξ = 0} and T∗Λ has coordinates (x,τ,x∗,τ∗). N The Euler vector field of Λ is θ = τ D and its principal symbol is the function Pi=1 i τi ϕ = N τ τ∗ =< τ,τ∗ >. The hypersurface S of T∗Λ defined by the equation ϕ is a Pi=1 i i Λ canonical hypersurface which will be considered in next section. For the W bifiltration, the operator x , i = 1,...,p, is of bi-order (0,0), t , j = 1,...,d i j is of bi-order (−1,0), D is of bi-order (0,1), D is of bi-order (1,1). xi tj The operator ϑ may be chosen here as t D , it is of bi-order (0,1). P j tj 3. Regularity along a lagrangian conic submanifold 3.1. Equivalentdefinitionsofregularity. TheEulervectorfieldθ ofΛisadifferential Λ operator on Λ, its characteristic variety is a canonical subvariety of T∗Λ which will be denoted by S . As in section 2.1, M is a simple holonomic E -module supported by Λ. Λ Λ X Definition 3.1. Let Λ be a lagrangian conic submanifold of T∗X and M be a holonomic E -module. The module M is regular along Λ if and only if it satisfies the following X equivalent properties: i) The microcharacteristic variety ChΛ(∞,1)(M) is contained in SΛ. ii) The microcharacteristic variety ChΛ(∞,1)(M) is lagrangian. iii) The microcharacteristic variety ChΛ(∞,1)(M) is equal to ChΛ(∞)(M) iv) ∀j ≥ 0, Extj (M,M )= Extj (M,E∞ ⊗ M ) EX Λ EX X EX Λ We recall that E∞ is the sheaf of microdifferential operators of infinite order. X If Λ is the conormal bundle to a submanifold Y of X, we may take M = C the Λ Y|X sheaf of holomorphic microfunctions of [13]. Then (iv) is reformulated as: (iv)’ ∀j ≥ 0, Extj (M,C ) = Extj (M,C∞ ) EX Y|X EX Y|X Let us now briefly remain how the equivalence between the items of definition 3.1 is proved. By theorem 2.2, if M is holonomic, grVM is holonomic hence its characteristic variety ChΛ(∞)(M)is lagrangian. Soif ChΛ(∞,1)(M)is equal to ChΛ(∞)(M) itis lagrangian that is (iii)⇒(ii). The microcharacteristic variety ChΛ(∞,1)(M) is defined by the coherent grBM-module πΛ∗O[T∗Λ] hence is bihomogeneous, that is homogeneous in the fibers of T∗Λ and homo- geneous for the action of C∗ induced by the action of C∗ on Λ. By section 4.3. of [8], REGULARITY AND B-FUNCTIONS FOR D-MODULES 9 any lagrangian bihomogeneous submanifold of T∗Λ is contained in S . In fact, if Σ is Λ lagrangian the canonical 2-form Ω of T∗Λ vanishes on Σ and if it is bihomogeneous the vectorfieldsv andv associatedtothetwoactionsofC∗ aretangenttoΣ. HenceΩ(v ,v ) 1 2 1 2 vanishes on Σ and an easy calculation of [8] shows that Ω(v ,v ) is an equation of S . So 1 2 Λ (ii)⇒ (i). Finally, the main parts of the result are (i) ⇒(iv) which is given by corollary 4.4.2. of [10] and (iv) ⇒(iii) given by theorem 2.4.2 of [11]. 3.2. Regular b-function. Let Λ be a conic lagrangian submanifold of the cotangent bundle T∗X. As before, θ is the Euler vector field of Λ and ϑ is a microdifferential Λ Λ operator in V E whose image in grVE is θ . 0 X 0 X Λ Definition 3.2. Let M be a coherent E -module with a good bifiltration WM relative X to Λ. A regular b-function for WM is a polynomial b such that: ∀(k,j) ∈ Z×Z, b(ϑ −k)W M⊂ W M Λ k,j k−1,j+n where n is the degree of b. if u is a section of M, a regular b-function for u is a regular b-function for the canonical bifiltration of the module N = E u that is W N = (Wk,jE )u. X k,j X Remark that if WM is a good bifiltration, we get a good V-filtration by setting: VkM = [ Wk,jM j∈Z So, consideringnonregular b-function,wemay defineab-functionforW tobeab-function for the corresponding V-filtration. Proposition 3.3. Let M be a holonomic E -module which a good bifiltration admitting X a regular b-function, then M is regular along Λ. Indeed, the existence of a regular b-function implies immediately the point (i) of defini- tion 3.1. By the following theorem and its corollary, the converse is true which gives another condition equivalent to the regularity along Λ. Theorem 3.4. Let M be holonomic E -module regular along a the conic lagrangian man- X ifold Λ. Any section u of M admits locally a regular b-function. If b(θ) is the b-function of u, there exits an integer d such that θdb(θ) is a regular b-function for u. Proof. Let θ be the Euler vector field of Λ and ϑ an operator of E whose image in grVE X X isθ. Bydefinition, theprincipalsymbolϕofθ isanequation ofthecanonicalhypersurface S of T∗Λ. Λ If M is regular along Λ, the submodule N = E u is regular along Λ. We consider on X N the canonical bifiltration, that is W N = (W E )u. By definition 3.1 of regularity, kj kj X ChΛ(∞,1)(N) is contained in SΛ hence grWM is annihilated by a power of ϕ. As the operator ϑ is of bi-order (0,1), there is an integer q such that ϑqW N ⊂ W N +W N k,j k−1,j+q k,j+q−1 Hence there is an operator P ∈ W E and an operator Q ∈ W E such that −1,q X 0,q−1 X ϑqu= Pu+Qu 10 YVESLAURENT We have (ϑq −P)2u = Q(ϑq − P)u+ [ϑq − P,Q]u = Q2u+[ϑq − P,Q]u. The com- mutator [ϑq − P,Q] belongs to W E as well as Q2 so (ϑq − P)2u = Q u with 0,2(q−1) X 2 Q ∈ W E . By induction we find that (ϑq −P)ru= Q u with Q ∈ W E . 2 0,2(q−1) X r r 0,r(q−1) X Let b be the b-function of u. If p is the degree of b, there is some integer r such that b(ϑ)W N ⊂ W N. 0,j −1,j+p+r We get b(ϑ)(ϑq −A)ru⊂ b(ϑ)W N ⊂ W N 0,rq−r −1,p+rq The operator (ϑq − A)r is equal to ϑrq − A where A belongs to W E . hence r r −1,rq X b(ϑ)ϑrqu ⊂ W N and as p+rq is the degree of b(ϑ)ϑrq, this is a regular b-function −1,p+rq for u. (cid:3) Remark 3.5. It is clear in the proof that θdb(θ) may be replaced by (θ−α)db(θ) for any complex number α. Corollary 3.6. Let M be holonomic E -module regular along the conic lagrangian man- X ifold Λ. Any good bifiltration WM relative to Λ admits locally a regular b-function. If b(θ) is the b-function associated to this bifiltration, there exits an integer d such that θdb(θ) is a regular b-function. Proof. Let (u ,...,u ) be local generators of the bifiltration. By definition, there are 1 N integers (λ ,...,λ ) and (ν ,...,ν ) such that 1 N 1 N N WkjM= XWk−λi,j−νiEXui i=1 If b(θ) is the b-function of the bifiltration, b(θ−λ ) is a b-function for the section u . By i i theorem 3.4 and remark 3.5, there exists an integer di such that (θ −λi)dib(θ −λi) is a regular b-function for d . Let d be the maximum of d for i= 1,...,N. i i As remarked in section 2.2, [θ,P] = +kP on grVE hence if P is an operator of k X W E we have ϑP = P(ϑ+k−λ )+R with R in W E . k−λi,j−νi X i k−λi−1,j−νi X Ifuisasection of W Mwehaveu= P u withP inW E and(θ−k)db(θ− kj P i i i k−λi,j−νi X k) is a regular b-function for u. (cid:3) 3.3. ApplicationtoD -modules. PreviousresultshavebeenestablishedforE -modules X X on the whole of T∗X including the zero section that is also for D -modules. X If x is a point of the zero section X of T∗X, a conic lagrangian submanifold of T∗X defined in a neighborhood of x is the conormal T∗X to a submanifold Y of X. Let Y B∞ = H1(O ) be the cohomology of O and B = H1 (O ) the corresponding Y|X Y X X Y|X [Y] X algebraic cohomology. If Y has codimension 1 we have: C∞ /C = π−1(B∞ /B ) with π :T∗X → Y Y|X Y|X Y|X Y|X Y So if M is a holonomic D -module condition (iv) of definition 3.1 is equivalent to X (iv)” ∀j ≥ 0, Extj (M,B )= Extj (M,B∞ ) DX Y|X DX Y|X and by [11] this is equivalent to formula 1.2. If Y has codimension greater than 1, the situation is different. By a canonical trans- formation, the situation is microlocally equivalent to the situation of codimension 1. But in the definition 3.1 we have to consider points of T˙∗X = T∗X −Y the complementary Y Y

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.