The strong monodromy conjecture for monomial ideals on toric varieties
Jen-Chieh Hsiao and Ching-Jui Lai
Department of Mathematics, National Cheng Kung University, Tainan, Taiwan
Dedicated to Professor Gennady Lyubeznik on the occasion of his 60th birthday.
ABSTRACT
We compute Denef and Loeser’s motivic zeta function associated to a monomial ideal on an affine toric variety, generalizing a result of Howald, Mustat¸a, and Yuen. We also investigate the relation between the poles of the motivic zeta function and the roots of their corresponding Bernstein–Sato polynomial defined by the first author and Matusevich.
ARTICLE HISTORY Received 8 November 2017 Revised 8 February 2018 Communicated by U.Walther
KEYWORDS
Bernstein–Sato polynomials;
motivic zeta functions;
strong monodromy conjecture; toric varieties 2010 MATHEMATICS SUBJECT
CLASSIFICATION 14M2514B0514E1814F10
1. Introduction
The strong monodromy conjecture of Igusa states that the real part of a pole of the p-adic zeta function associated to a hypersurface defined by a polynomial f is a root of the Bernstein–Sato polynomial of f [12]. A fascinating consequence of this conjecture is a link between the poles of Igusa zeta functions and the monodromy eigenvalues of Milnor fibers of f. As a generalization of Igusa zeta function, Denef and Loeser introduce their motivic zeta function using motivic integra- tion [6, 7]. We will focus on the special type of motivic zeta function Z(X, I ;s) associated to an ideal sheaf I on an algebraic variety X, whose definition will be recalled in Section 2. For more information about this subject, we refer to the surveys [5, 17, 14].
On the other hand, Budur, Mustat¸a, and Saito define a generalization of Bernstein–Sato poly- nomial bXIðsÞ for an ideal I on a smooth affine variety X [2]. Their definition is adopted in [11]
to extend the notion of Bernstein–Sato polynomial bUIrðsÞ for an ideal I on an affine (normal) toric variety Ur. It is then curious for us to see whether the analogous strong Monodromy Conjecture holds in the toric setting. In [10], this conjecture under the p-adic setting is verified for the case where I is a monomial ideal on an affine spaceAn. The key ingredient in their proof is an explicit formula of the p-adic zeta function associated to the pair (An, I). It turns out that the real parts of candidate poles of the p-adic zeta function associated to (An, I) come from the rays of the normal fanDIof I. So one can test the conjecture using the combinatorial description in [3] of the roots of bAInðsÞ in terms of the Newton polyhedron of I.
CONTACTJen-Chieh Hsiao jhsiao@mail.ncku.edu.tw National Cheng Kung University, No 1 University Road East District, Tainan 701, Taiwan.
ß 2019 Taylor & Francis Group, LLC
https://doi.org/10.1080/00927872.2018.1444172
The purpose of this paper is to test the conjecture for the motivic zeta function Z(Ur, I;s) and the Bernstein–Sato polynomial bUIrðsÞ assocaited to a monomial ideal I on an affine toric variety Ur. To do this, we need an explicit formula for Z(Ur,I;s). This relies on a change of variables formula of motivic zeta functions obtained in [7] applying to a toric log resolution h : X(D)!Urof (Ur, I) and the description of the contact locus of a monomial ideal on a toric variety in [13] or Section 3. As a result, the motivic zeta function Z(Ur, I;s) has a similar description as in the case studied in [10], except that a pole of Z(Ur, I;s) may come from a ray of the fanD that is not a ray of the normal fan DI. In fact, we show in Example 6.2 and 6.3 that such a ray does possibly produce a pole of Z(Ur, I;s) that is not a root of the b-function bUIrðsÞ. However, under a mild assumption we can still prove in Theorem 6.4 that the poles obtained from the rays of the normal fanDI are roots of bUIrðsÞ. See Section 2.4 for the definition of a pole of the motivic zeta function Z(X, I ; s) as well as the analogous strong Monodromy Conjecture in the motivic setting (Conjecture 2.1).
Our explicit computation of the toric motivic zeta functions is one of the first few examples where the ambient space is singular. We mention the work of Veys on the motivic zeta function of an effective divisor E on a Q-Gorenstein variety X [15] with the condition that the singular locus of X is contained in the support of E. See also [16] for a study of relevant zeta functions on varieties with Kawamata log terminal singularities.
2. Motivic Igusa zeta functions of Denef and Loeser We state some definitions and results from [7].
2.1.
Let k be a field of characteristic 0 and let M be the Grothendieck ring of algebraic varieties over k. SetL :
¼ ½A1k 2 M and denote ^M the completion of M[L1] with respect to the filtration fFmM½L1gm2Z, where FmM½L1 is the subgroup of M[L1] generated by f[S]LijidimS mg.
For an algebraic variety X of dimension d over k, denote L(X) and Lm(X) the arc space and the m-th jet scheme of X respectively. Let pm: LðXÞ ! LmðXÞ be the canonical morphism corre- sponding to truncation of arcs. For a semi-algebraic subset A of L(X), Denef and Loeser define the motivic measure lðAÞ 2 cM of A. We will not recall the definition of a semi-algebraic of L(X) nor the construction of the motivic measure l. Instead, we only mention some properties that will be used later.
1. If a semi-algebraic set A of L(X) is contained in L(S) for some closed subvariety S of X with dim S<dim X, then l(A) ¼ 0.
2. If AP i, i2N, are semi-algebraic and mutually disjoint, then A :¼ ⋃;i2NAi is semi-algebraic and
i2NlðAiÞ converges to l(A) in cM.
3. If BLm(X) is constructible then p1m ðBÞ is semi-algebraic. Such semi-algebraic set p1m ðBÞ will be called a cylinder or a constructible set in L(X). Moreover, if X is smooth and A ¼ p1m ðBÞ is constructible, then
lðAÞ ¼½pmðAÞ
Lðmþ1Þd2 cM which does not depend on m.
2.2.
Let I be a coherent ideal sheaf on X, and consider the function ordtI:L(X)!N[f1g given by c7! mingordtgð~cÞ, where the minimum is taken over all g in the stalk Ip0ðcÞof I at p0(c) and~c 2
LðXÞðkcÞ ¼ Xðkc½½tÞ is the corresponding rational point of c2L(X) over the residue field kc of c on L(X). The motivic Igusa zeta function associated the ideal I on X is defined as
ZðX; I; sÞ ¼Ð
LðXÞLordtIs dl:¼
X
n0
lðContnðIÞÞ Lns 2 cM½½Ls (2.1)
where the fiber ContnðIÞ :¼ fc 2 LðXÞ j ordtIðcÞ ¼ ng ¼ ðordtIÞ1ðnÞ of the function ordtI is the n-th contact locus of I .
2.3.
For a log resolution h:Y!X of the pair (X,I ), let L(h):L(Y)!L(X) be the induced map on the corresponding arc spaces. There is a change of variables formula
ð
LðXÞLordtIsdl ¼ ð
LðYÞLðordtI8LðhÞsþordtJachÞdl :¼
X
i;j2N
lðContiðI OYÞ\ContjðJachÞÞLðisþjÞ: (2.2)
Here, the Jacobian ideal sheaf Jachof h is by definition the ideal sheaf ofOYsuch that the image of the morphism hXdX! XdY induced by pulling-back d-forms can be written as JachXdY. Note thatXdX is the dth wedge product of the sheaf of K€ahler differentials XX on X. For more details, except for [7], the reader is also referred to [8] in which another proof of this formula is given.
2.4.
Now, let I be an ideal on an affine variety X. Assume X is either smooth or a normal toric var- iety. Then we have the notion of b-function bXIðsÞ for the pair (X,I) [2, 11]. The strong Monodromy Conjecture in the motivic setting can be formulated as the following.
Conjecture 2.1. There exists a finite subset S ofZ>0 Z>0 such that ZðX; I; sÞ 2 cM Ls; 1
1 Lasb
ða;bÞ2S
and such that ba is a root of bXIðsÞ.
Under this conjecture, we can express the motivic zeta function Z(X,I ;s) as a rational function
NðLsÞ
DðLsÞ for NðLsÞ; DðLsÞ 2M½L^s. A rational number ba will be called a pole of the motivic zeta function Z(X,I ;s) if 1Lasbdivides D(Ls) but not divides N(Ls).
3. Arc spaces of toric varieties
We recall some results in [13] concerning the arc spaces of toric varieties. Throughout the paper, we will use freely the results in toric geometry as described in [9].
3.1.
Let M be the free abelian groupZdand N its dual HomZ(M,Z) together with the canonical pair- ing h,i:NM!Z. We denote M ZR and N ZR by MR and NR, respectively. The pair h,i natur- ally extends to a pairing NR MR! R which we also denote by h,i. For a linear subspace WNR, the induced pairing ðNR=WÞ W?! R is also denoted by h,i. Let q : NR! NR=W be the projection. We have hv,ui ¼ hq(v), ui for v2NR and u2W?.
Let X ¼ X(D) be the toric variety over a field k corresponding to a d-dimensional fan D in NR
and denote by T: ¼ Spec k[M] the dense open torus in X. For a cone s 2D, denote XðsÞ ¼ orb s the orbit closure corresponding to s. The toric variety X(s) corresponds to the the fan Ds in NR=Rs that consists of the image r in NR=Rs of the cones r 2 D, s<r. The affine toric subvari- ety Ur XðsÞ is Spec k½s?\ rÚ\ M. Denote by Nsthe image of N in NR=Rs.
3.2.
For a field extension K k, let g be the generic point of Spec K[[t]]. The arc space L(X) naturally decomposes as L(X) ¼ ts2DL(X)(s) where
LðXÞðsÞ ¼ fc 2 LðXÞ j cðgÞ 2 orb sg
whose closure in L(X) is LðXÞðsÞ ¼ LðXðsÞÞ. This decomposition can be further refined to L(T)- orbits of L(X) under the action of L(T). In fact, for each s2D, there is a one-to-one correspond- ence
fLðTÞ c j c 2 LðXÞðsÞg $ jDsj \ Ns
that sends the L(T)-orbit of c 2 L(X)(s) to a lattice point v 2 jDsj \ Ns satisfying, for u2M, ordtcðxuÞ ¼ hv; ui. Under this correspondence, we write L(T)v for the L(T)-orbit L(T)c. Note that the orbits L(T)vare cylinders in L(X). The motivic measure
lðLðTÞvÞ ¼ 0; for 0 “ s and v 2 jDsj \ Ns; (3.1) since L(T)vL(X(s)) and since X(s) is a closed subvariety of X with dim X(s)<dim X.
3.3.
For a torus invariant ideal sheaf I on X, the n-th contact locus Contn(I ) is a disjoint union of L(T)-orbits. In fact, it follows from (3.1) above and from Theorem 4.1 and Lemma 5.9 in [13]
that we have the following proposition.
Proposition 3.1 ([13]). An orbit L(T)c ¼ L(T)v, v 2 jDsj \ Ns, is contained in Contn(I ) if and only if, for any cone r 2 Ds containing v, we have LðTÞv LðUrÞ and
n ¼ min
xu2IðUrÞordtcðxuÞ ¼ min
xu2IðUrÞhv; ui:
In particular, the motivic measure lðContnðIÞÞ ¼P
vlðLðTÞvÞ where the sum runs through all v2jDj\N satisfying n ¼ minxu2IðUrÞhv; ui for all r2D containing v.
4. Bernstein–Sato polynomials on toric varieties 4.1.
Let R be the cone Rn0 in Rn and UR¼ Ank be the corresponding affine space over a field k of characteristic 0. Denote by D the n-th Weyl algebra. In [2], Budur, Mustat¸a, and Saito define a generalization of Bernstein–Sato polynomials bUJRðsÞ for an ideal J ¼ hf1; :::; fri on URas follows.
Let s1; :::; sr be indeterminates over k, consider k½x1; :::; xn
Y
ri¼1
fi1; s1; :::; sr
" #
Y
ri¼1
fisi:
This is a D[sij]-module, where sij¼ sit1i tj, and the action of the operator ti is given by tiðsjÞ ¼ sjþ dij (the Kronecker delta). The Bernstein–Sato polynomial or b-function associated to J ¼
hf1; :::; fri is defined to be the monic polynomial bUJRðsÞ of the lowest degree in s ¼Pr
i¼1si satis- fying a relation of the form
bUJRðsÞ
Y
ri¼1
fisi ¼
X
rk¼1
Pktk
Y
ri¼1
fisi; (4.1)
where P1; :::; Pr2 D½sijj i; j 2 f1; :::; rg. It is shown in [2] that bUJRðsÞ is a nonzero polynomial and is independent of the generating set of J.
4.2.
When J is a monomial ideal on an affine space UR, the roots of bUJRðsÞ can be completely described in terms of the Newton polyhedron PJof the ideal J [3].
Denote C :¼ RÚ\ Zn and CI:¼ fu 2 C j xu2 Ig. Let Q be a face of PJ. Denote by VQ the vector subspace of Rd generated by Q. Let MQ the subset of Zn such that MQ1 is the semi- subgroup ofZngenerated by fa b j a 2CJ and b 2CJ\ Qg, where 1 ¼ (1, … ,1)2Zn. Let MQ0 :
¼ MQþ b0 be the shift of MQ by any lattice point b02 CJ\ Q, which does not depend on the choice of b0.
Theorem 4.1 ([3]). The set of roots of bUJRðsÞ is the union of the sets RQ:¼ fLQðaÞ j a 2 ðMQn M0QÞ \ VQg
for faces Q of PJthat is not contained in any coordinate hyperplane, where LQ: Rn! R is a lin- ear functional satisfying LQ(a) ¼ 1 for all a2Q.
4.3.
Now, let Ur be the affine toric variety corresponding a d-dimensional cone r in Rdand let I be an ideal on Ur. Apply the same functional equation (4.1) using the differential operators on Ur, one can define the b-function bUIrðsÞ for the pair (Ur,I). In [11], the b-function bUIrðsÞ is related to the b-function bUJRðsÞ of certain ideal J on an affine space UR. We recall the statement as follows.
Let fv1; :::; vng be the set of primitive generators of the rays of r, each of which corresponds to a facet of rÚ. Consider the linear map F: Rd! Rn that sends u2Rd to ðhv1; ui; :::; hvn; uiÞ 2 Rn. The map F induces a ring homomorphism k½rÚ\ Zd ! k½RÚ\ Zn by sending f :¼P
jkjxuj to Fðf Þ:¼P
jkjyFðujÞ, where xa:¼Q
ixaii and yb:¼Q
jybjj for a 2 rÚ\ Zd and b 2RÚ\ Zn. For an ideal I ¼ hf1; :::; fri of k½rÚ\ Zd, consider the ideal J ¼ hFðf1Þ; :::; FðfrÞi in k½RÚ\ Zn.
Theorem 4.2 ([11]). bUIrðsÞ ¼ bUJRðsÞ
We mention that Theorem 4.2 is used in [11] to relate the b-function bUIrðsÞ with multiplier ideals of the pair (Ur,I).
5. The case of monomial ideals on affine spaces
Let X ¼ Us be the d-dimensional affine space over a field k of characteristic 0 where sNR is a pointed smooth cone. Let J be a nonzero proper monomial ideal on X.
5.1.
By Proposition 3.1, the motivic zeta function in (2.1) can be express as ZðUs; J; sÞ ¼
X
n0
lðContnðJÞÞ Lns¼
X
v2s\N
lðLðTÞvÞ LJðvÞs
whereJðvÞ :¼ minxu2Jhv; ui.
5.2.
Let e1; :::; ed be the primitive generators on the rays of sÚ. Then xe1; :::; xed generate the polyno- mial algebra k[sÚ\M]. For m max1idhv; eii, consider the constructible set Cm LmðXÞ con- sisting of m-jets c with ordtcðxeiÞ ¼ hv; eii. Then p1m ðCmÞ ¼ LðTÞvis a cylinder and
½pmðLðTÞvÞ ¼ ½Cm ¼ ðL1Þd Lmd Lhv;
P
ieii
2 M:
Putting es¼P
iei, the motivic measure of L(T)v can be expressed as lðLðTÞvÞ ¼ ½Cm
Lmþ1d¼ ð1 L1Þd Lhv;esi: (5.1) Therefore, the motivic zeta function
ZðUs; J; sÞ ¼
X
v2s\N
ð1 L1Þd ðL1ÞJðvÞsþhv;esi: (5.2)
5.3.
LetDJbe the normal fan of the Newton polyhedron PJof J. The fanDJconsists of cones rQ:¼ fv 2 NRjhv; ui hv; u0i for u 2 Q and u02 PJg
corresponding to faces Q of PJ. Note that if w is a vertex of PJsuch that v2s\N is contained in rw, thenJ(v) ¼ hv,wi.
The following theorem collects the main results in [10]. Although it was originally stated for the coneR ¼ Rn0, we rewrite it for the smooth cone s by using the natural bijection s ffiR.
Theorem 5.1 ([10]). The motivic zeta function Z(Us,J;s) is a rational function ofLs. Each pole of Z(Us,J;s) can be written as the form hv;eJðvÞsifor a primitive generator v of some ray of the normal fan DJ of J. Moreover, any such number hv;eJðvÞsi is a root of the Bernstein–Sato polynomial bUJsðsÞ.
In particular, Conjecture 2.1 holds for monomial ideals on affine spaces.
6. The case of monomial ideals on affine toric varieties
Let X ¼ Urbe an affine toric variety of dimension d that corresponds to some cone r in NR. Let I be a nonzero proper monomial ideal on X. Consider a toric log resolution h:Y ¼ X(D)!X of the pair (X,I). The toric map h factors through the normalized blowing-up X(DI) of X along I. By the change of variables formula (2.2), the motivic zeta function (2.1) can be written as
ZðUr; I; sÞ :¼
X
i;j2N
lðContiðI OYÞ \ ContjðJachÞÞ LðisþjÞ: (6.1)
Let v 2r \ N and let s 2D be a d-dimensional cone containing v. The following observations are helpful in computing the motivic zeta function Z(Ur,I;s).
1. For a L(T)-orbit LðTÞv ContiðI OYÞ \ ContjðJachÞ, it follows from Proposition 3.1 that i ¼ min
xu2IOYðUsÞhv; ui and j ¼ min
xu2JachðUsÞhv; ui:
2. By (5.2) the motivic measure lðLðTÞvÞ ¼ ð1L1Þd Lhv;esi; where es is the sum of primi- tive generators on the rays of sÚ.
3. Denote by Vert(PI) the set of vertices of the Newton polyhedron PIof I, then we have
IðvÞ :¼ min
u2PIhv; ui ¼ min
u2VertðPIÞhv; ui ¼ min
xu2IOYðUsÞhv;ui:
4. Define 0ðvÞ :¼ hv; esi. Then 0ðvÞ ¼ minxu2XdYðUsÞhv; ui. Here, we identify the canonical sheaf XdY of the smooth toric variety Y with the torus invariant ideal sheaf whose local sections on Uscorrespond to lattice points in the relative interior of sÚ[9, Section 4.3].
5. Since the sheafXdX of K€ahler differential d-forms is M-graded [4, Section 3], the image of the toric map hXdXðUsÞ ! XdYðUsÞ is also M-graded. This implies that the Jacobian ideal JachðUsÞ is a monomial ideal in the polynomial ring O(Us). So we can define
00ðvÞ :¼ min
xu2JachðUsÞhv; ui
6. Notice that the two valuations 0 and 00 extend to the function field of Us, so they do not depend on the choice of s. So we put
DðvÞ :¼ 0ðvÞþ00ðvÞ ¼ min
xu2XdYJachðUsÞhv; ui:
Again, we abuse the notationXdY to denote the ideal sheaf whose local sections on Uscorres- pond to lattice points in the relative interior of sÚ. Note that the function D is linear on each cone of the fanD.
With the above observations, we conclude that the motivic zeta function (6.1) of the pair (Ur, I) can be computed as
ZðUr; I; sÞ ¼
X
v2r\N
lðLðTÞvÞ L½IðvÞsþ00ðvÞ
¼
X
v2r\N
ð1L1Þd L0ðvÞ L½IðvÞsþ00ðvÞ
¼
X
v2r\N
ð1L1Þd ðL1ÞIðvÞsþDðvÞ
(6.2)
Note that the change of variables formula is valid for any log resolution, so the zeta function Z(Ur,I;s) does not depend on the fan D. Notice also that the expression (6.2) of Z(Ur,I;s) does generalize the expression in (5.2). Indeed, if r is a pointed smooth cone, then by taking h to be the identity map
DðvÞ ¼ min
xu2XdX
hv; ui ¼ min
u2intðrÚÞ\Mhv; ui ¼ hv; eri:
Since the zeta functions (6.2) and (5.2) have similar expressions, we have the following gener- alization of the first part of Theorem 5.1 whose proof goes exactly the same as in [10] using their results of zeta functions associated to cones.
Theorem 6.1. The motivic zeta function Z(Ur,I;s) is a rational function of Ls. Each pole of Z(Ur,I;s) is of the form DIðvÞðvÞ, where v is the primitive generator on a ray of a fan D so that X(D)!Ur is a toric log resolution of (Ur,I).
Proof. Rewrite the motivic zeta function (6.2) as ZðUr; I; sÞ ¼ ð1L1Þd
X
s2DI
Zs;I;DðsÞ; where
Zs;I;DðsÞ :¼
X
v2intðsÞ\Zd
ðL1ÞIðvÞsþDðvÞ:
Since I and D are linear on each s2DI, the statement of the theorem follows from [10,
Theorem 1.3]. w
Next, we investigate whether poles of the motivic zeta function Z(Ur,I;s) are roots of the Bernstein–Sato polynomial bUIrðsÞ as defined in [11].
Example 6.2. Consider the affine toric variety Ur whose cone r ¼ Coneðe2; de1 e2Þ, d2. The dual cone rÚ ¼ Coneðe1; e1þ de2Þ. Introduce variables x,y so that the coordinate ring of Ur is represented as k½rÚ\ M ¼ k½x; xy; :::; xyd: Let I be the ideal hx3y2; x4yi in k[rÚ\M].
Let h : X(D)!Ur be the minimal log resolution of (Ur,I), namely D ¼ s1[ s2[ s3 where s1¼ Coneðe1; de1 e2Þ, s2¼ Coneðe1; e1þ e2Þ, and s3¼ Coneðe1þ e2; e2Þ. We identify the coordinate ring of the local charts Usi of X(D) as
k½sÚ1\ M ¼ k½y1; xyd; k½sÚ2 \ M ¼ k½xy1; y; and k½sÚ3 \ M ¼ k½x; x1y: The Jacobian ideals of the map h on the local charts are
JachðUs1Þ ¼ hxydi; JachðUs2Þ ¼ hxyi; and JachðUs3Þ ¼ hx2i:
Put w1¼ (2,2d1) and w2¼ ð2; 1Þ ¼ w3. Then we have
DðvÞ ¼ hv; wii for i ¼ 1; 2; 3:
Put v1 ¼ (d,1), v2 ¼ (1,0), v3 ¼ (1,1), and v4 ¼ (0,1). These are primitive generators on the rays ofD. Note that v2 comes from the ray ofD that is not a ray of the normal fan DI. We have
Dðv1Þ ¼ 1 ¼ Dðv4Þ, Dðv2Þ ¼ 2 and Dðv3Þ ¼ 3. On the other hand, we have Iðv1Þ ¼ 3d 2,
Iðv2Þ ¼ 3, Iðv3Þ ¼ 5, and Iðv4Þ ¼ 1, so the candidate poles of Z(Ur,I;s) are
1 3d 2; 2
3 ; 3
5 ; and 1:
For d ¼ 2, the b-function for the pair (Ur,I) is bUIrðsÞ ¼ ðs þ 1Þ2
s þ1
2
2 s þ1
4
s þ3 4
s þ2 5
s þ3 5
s þ4 5
s þ6 5
s þ 7
10
s þ 9 10
s þ11 10
s þ13 10
:
One sees that14 ;35 ; and 1 are roots of bUIrðsÞ, but23 is not. We point out that b-function algorithms have been developed and implemented, see [1].
In fact, Theorem 6.4 shows that a candidate pole coming from a ray of the normal fan DIis always a root of bUIrðsÞ. Moreover, it follows from [10, Lemma 3.1] that23 does appear as a pole of Z(Ur,I;s) (not just a candidate). Indeed, one can verify this by explicitly computing Z(Ur,I;s).
Set T ¼ Ls and Bi:¼ ðTÞIðviÞsþDðviÞ. Then by [10, Lemma 3.1], we have
ZðUr; I; sÞ ¼
X
1i3
1 1 Bi
1
1 Biþ1
X
i¼2;3
1 1 Bi
¼ð1B1B3Þð1B
Q
4ÞþB4ð1B2Þð1B1Þ1i4ð1 BiÞ
¼ð1T9sþ4Þð1Tsþ1ÞþTsþ1ð1T3sþ2Þð1T4sþ1Þ ð1 T4sþ1Þð1 T3sþ2Þð1 T5sþ3Þð1 Tsþ1Þ :
Example 6.3. Even for principal ideals, the situation is not as expected. Consider r ¼ Coneð3e1 2e2; e2Þ and I ¼ hx3y2i a principal ideal on Ur whose coordinate ring is represented as k½x; xy; x2y3. Let h : X(D)!Ur be the minimal log resolution of (Ur,I), namely D ¼ s1[ s2[ s3
where s1¼ Coneð3e1 2e2; 2e1 e2Þ, s2¼ Coneð2e1 e2; e1Þ, and s3¼ Coneðe1; e2Þ. We identify the coordinate ring of the local charts Usi of X(D) as
k½sÚ1 \ M ¼ k½x2y3; x1y2; k½sÚ2 \ M ¼ k½xy2; y1; and k½sÚ3 \ M ¼ k½x; y:
The Jacobian ideals of the map h on the local charts are
JachðUs1Þ ¼ hx2y3i; JachðUs2Þ ¼ hx; x2y3i; and JachðUs3Þ ¼ hxi:
One verifies that the minimal log resolution h produces three candidate poles of Z(Ur,I;s),
1
5 ;12 ;23. Only23 is not a root of bUIrðsÞ ¼ ðsþ1Þ2ðs þ12Þðs þ15Þðs þ25Þðs þ35Þðs þ45Þ.
Theorem 6.4. Let h:X(D)!Ur be a toric log resolution of (Ur,I) such that I OXðDÞ Jach XXðDÞ.
Here, we identify XX(D) with the torus invariant ideal ofOX(D) whose local sections on each chart Us, s2D, correspond to the lattice points in the interior of sÚ. If v is a primitive generator on a ray ofDIsuch thatI(v)6¼0, then DIðvÞðvÞis a root of the Bernstein–Sato polynomial bUIrðsÞ.
Proof. We use freely the notation in Section 4. Let v be a primitive generator on a ray of DIsuch thatI(v) 6¼ 0. Let Q be the facet of the Newton polyhedron PIthat corresponds to v. Note that Q is not contained in any faces of r. Note also that the assumption I OXðDÞ Jach XXðDÞ implies thatIðvÞ DðvÞ: Let s be a d-dimensional cone of D that contains v and let w 2 sÚ\M be a lat- tice point that corresponds to a monomial in the ideal Jach XXðDÞðUsÞ so that D ¼ hv,wi. We haveI(v)hv,wi1.
Let ~v 2R\Qn be such that F(~v) ¼ v where F: ðRnÞ! ðRdÞ is the dual of the map F.
Then
hv; ui ¼ hFð~vÞ; ui ¼ h~v; FðuÞi; for u 2 Zd: This implies thatIðvÞ ¼ minxu2Ih~v; FðuÞi ¼ Jð~vÞ and that
DðvÞ
IðvÞ ¼ hv; wi
Jð~vÞ¼ h~v; FðwÞi
Jð~vÞ :
Set LðaÞ:¼h~v;aiJð~vÞ. Then L(a) takes the value 1 on a face ~Q containing F(Q). By Theorems 4.1 and 4.2, it suffices to show that FðwÞ 2 ðM~Qn M0~QÞ \ V~Q. Since Q is a facet of PI, it is certainly true that w 2 VQ and hence FðwÞ 2 FðVQÞ V~Q. Moreover, pick any b 2CJ\ ~Q. Since w lies in the relative interior of r, we have F(w)12intR and a: ¼ F(w)1þb2CJ. Hence FðwÞ ¼1 þ a b 2 M~Q.
To show that FðwÞ 62 M0~Q, it suffices to observe that the linear functional L takes values that are strictly greater than 1 on M0~Q, but LðFðwÞÞ ¼hv;wiIðvÞ 1. w
Acknowledgements
The first author thanks Laura Matusevich for carefully reading the first draft of this paper. He is also grateful to Uli Walther for his comment on the expression of the poles of the motivic zeta funcions.
Funding
J.C.H. is partially supported by MOST grant 105-2115-M-006-015-MY2. C. J. L. is partially supported by MOST grant 106-2115-M-006 -019.
References
[1] Berkesch, C., Leykin, A. (2010). Algorithms for Bernstein-Sato polynomials and multiplier ideals. In: ISSAC 2010—Proceedings of the 2010 International Symposium on Symbolic and Algebraic Computation, ACM, New York, pp. 99–106. MR 2920542
[2] Budur, N., Mustat¸a, M., Saito, M. (2006). Bernstein-Sato polynomials of arbitrary varieties. Compos. Math.
142(3):779–797. MR 2231202 (2007c:32036)
[3] Budur, N., Mustat¸a, M., Saito, M. (2006). Combinatorial description of the roots of the Bernstein-Sato poly- nomials for monomial ideals. Comm. Algebra 34(11):4103–4117. MR 2267574 (2007h:32041)
[4] Corti~nas, G., Haesemeyer, C., Walker, M. E., Weibel, C. (2009). The K-theory of toric varieties. Trans. Am.
Math. Soc. 361(6):3325–3341. MR 2485429
[5] Denef, J. (1991). Report on Igusa’s local zeta function. Asterisque 1990/91(201–203):359–386 (1992).
Seminaire Bourbaki, Exp. No. 741. MR 1157848 (93g:11119)
[6] Denef, J., Loeser, F. (1998). Motivic Igusa zeta functions. J. Algebraic Geom. 7(3):505–537. MR 1618144 (99j:14021)
[7] Denef, J., Loeser, F. (1999). Germs of arcs on singular algebraic varieties and motivic integration. Invent.
Math. 135(1):201–232. MR 1664700
[8] Ein, L., Mustat¸a, M. (2009). Jet schemes and singularities. Summer Institute in Algebraic Geometry at the University of Washington, Seattle from July 25 to August 12, 2005. In: Part 2, Proc. Sympos. Pure Math., Vol. 80. Providence, RI: Am. Math. Soc., pp. 505–546. MR 2483946
[9] Fulton, W. (1993). Introduction to Toric Varieties. Annals of Mathematics Studies, Vol. 131. Princeton, NJ:
Princeton University Press (The William H. Roever Lectures in Geometry). MR 1234037
[10] Howald, J., Mustat¸a, M., Yuen, C. (2007). On Igusa zeta functions of monomial ideals. Proc. Am. Math.
Soc. 135(11):3425–3433. MR 2336554
[11] Hsiao, J.-C., Matusevich, L. F. (2016). Bernstein–Sato Polynomials on Normal Toric Varieties, ArXiv e- prints.
[12] Igusa, J.-I. (2000). An Introduction to the Theory of Local Zeta Functions. AMS/IP Studies in Advanced Mathematics, Vol. 14. Providence, RI/Cambridge, MA: American Mathematical Society/International Press.
MR 1743467
[13] Ishii, S. (2004). The arc space of a toric variety. J. Algebra 278(2):666–683. MR 2071659
[14] Nicaise, J. (2010). An Introduction to p-Adic and Motivic Zeta Functions and the Monodromy Conjecture.
Algebraic and Analytic Aspects of Zeta Functions and L-Functions, MSJ Mem., Vol. 21. Tokyo: Math. Soc.
Japan, pp. 141–166. MR 2647606
[15] Veys, W. (2001). Zeta functions and “Kontsevich invariants” on singular varieties. Canad. J. Math. 53(4):
834–865. MR 1848509 (2002e:14025)
[16] Veys, W. (2003). Stringy zeta functions for Q-Gorenstein varieties. Duke Math. J. 120(3):469–514. MR 2030094 (2004m:14075)
[17] Veys, W. (2006). Arc Spaces, Motivic Integration and Stringy Invariants. Singularity Theory and Its Applications, Adv. Stud. Pure Math., Vol. 43. Tokyo: Math. Soc. Japan, pp. 529–572. MR 2325153