2403 14355
2403 14355
2403 14355
1. Introduction
Let V be an affine algebraic variety over an algebraically closed field K and let
p ∈ V be a singular point. Then we call the pair (V, p) a singularity. We denote
by A the local ring OV,p of the variety V at p, which is called the local ring of the
singularity (V, p). The multiplicity mult(V, p) of the singularity (V, p) is defined as
the (Hilbert-Samuel) multiplicity of the local ring A (see [3, §14]). The multiplicity is
a fundamental invariant playing a crucial role in singularity theory. Let m denote the
maximal ideal of A, and let d = dim(V, p) := dim A. If {f1 , . . . , fd } ⊂ A generates
an m-primary ideal, then it defines a local finite morphism (V, p) → (AdK , o) (cf. [3,
Theorem 14.5]). From a geometric point of view, mult(V, p) can be interpreted as
the minimum of the degree of such morphisms (cf. [2, Exercises 6.3.6], [3, Theorems
14.13, 14.14] and [8, Ch. VIII, §10, Corollary 2]).
In this paper, we investigate the multiplicity of singularities on branched cyclic
coverings of a given singularity. Let g be a non-zero regular function on V such that
g(p) = 0, and let y be the coordinate function of the affine line A1K . For n ∈ Z>0 ,
the natural projection φn : Vn := {y n = g} ⊂ V × A1K → V defines a cyclic covering
of degree n branched along the hypersurface {g = 0} ⊂ V . Clearly, the fiber φ−1 n (p)
consists of the point pn := p × {0} ∈ Vn . From the geometric perspective, we have
the inequality mult(Vn , pn ) ≤ n · mult(V, p). We will show the following.
Theorem 1.1. There exists a positive integer N such that the function mult(Vn , pn )
of n is constant for n > N.
2020 Mathematics Subject Classification. Primary 14B05; Secondary 13P10, 14J17, 32S05.
Key words and phrases. singularities, multiplicity, tangent cone, standard bases, cyclic
coverings.
The second-named author was partially supported by JSPS Grant-in-Aid for Scientific Research
(C) Grant Number 21K03215.
1
2 TOYA KUMAGAI AND TOMOHIRO OKUMA
This result is motivated by Tomaru’s result [7] which proves the theorem above for
normal complex analytic singularities of dimension two. Tomaru used the fact that
if M represents the maximal ideal cycle on a resolution with nice property, then the
multiplicity coincides with −M 2 . To prove the theorem, he constructed a branched
cyclic covering ψn : Ṽn → Ṽ of resolution spaces of Vn and V , and computed the
maximal ideal cycle on Ṽn from data of Ṽ via ψn . While this method is effective for
analyzing the maximal ideal cycles, it becomes somewhat intricate when we focus
just on the function mult(Vn , pn ). In this paper, we adopt the standard bases of
ideals of the localization of the polynomial rings, and prove a stronger result as
follows (see Theorem 3.1 and Remark 3.4).
of ideals in the local ring R. As an example, we introduce the negative degree reverse
lexicographical ordering >ds , which is defined as follows:
def
xα >ds xβ ⇐⇒ deg xα < deg xβ , or
α β
deg x = deg x and there exists i such that
αn = βn , . . . , αi+1 = βi+1 , αi < βi .
In the following, we fix a local degree ordering > on Mon(x).
Let R∗ denote the set of units of R. Let f ∈ R \ {0}. Then there is u ∈ R∗ ∩ K[x]
with u(0) = 1 such that uf ∈ K[x]. Let us write it as
X X n
α
uf = aα x = ai xα(i) , ai 6= 0 for 1 ≤ i ≤ n, xα(1) > xα(2) > . . .
α i=1
Then we define:
• LM(f ) := xα(1) , the leading monomial of f ,
• LT(f ) := a1 xα(1) , the leading term of f ,
• LC(f ) := a1 , the leading coefficient of f ,
• ord(f ) := minX { deg xα | aα 6= 0} = deg xα(1) , the order of f ,
• In(f ) := aα xα , the initial form of f . Let In(0) = 0.
deg xα =ord(f )
Obviously, for f ∈ K[x] \ {0}, LM(f ) appears in In(f ).
Definition 2.1. Let f, g ∈ R \ {0}, and write LM(f ) = xα and LM(g) = xβ . Let
γ = lcm(α, β) := (max(α1 , β1 ), . . . , max(αr , βr )).
Then the s-polynomial of f and g is defined to be
LC(f ) γ−β
Spoly(f, g) = xγ−α f − x g.
LC(g)
Clearly, Spoly(f, g) ∈ K[x] if f, g ∈ K[x].
Definition 2.2. Let I ⊂ R be an ideal and G ⊂ R a subset.
(1) The ideal
L(G) := hLM(g) | g ∈ G \ {0}i ⊂ K[x]
is called the leading ideal of G.
(2) If G is a finite subset of I and L(I) = L(G), then G is called a standard basis
of I.
(3) The initial ideal of I is defined to be
In(I) = hIn(f ) | f ∈ I \ {0}i ⊂ K[x].
Proposition 2.3 (cf. [2, 5.5.11–5.5.12]). Let I ⊂ hxi ⊂ K[x] be an ideal, and let
A = R/IR and m ⊂ A the maximal ideal of A.
(1) If {f1 , . . . , fs } is a standard basis
L of I, then In(I) = hIn(f1 ), . . . , In(fs )i.
(2) The tangent cone Grm (A) = n≥0 mn /mn+1 is isomorphic to K[x]/ In(I) as
graded K-algebra.
In particular, the multiplicity of the local ring A is determined by the initial forms
of a standard basis of I.
4 TOYA KUMAGAI AND TOMOHIRO OKUMA
We provide a very brief overview of the criterion for the standard bases and an
algorithms to compute them with respect to a local ordering. These are essentially
Buchberger’s algorithms but have been generalized by Mora ([4]).
Definition 2.4 (cf. [2, 1.6.4–1.6.5]). Let G denote the set of all finite subsets of R.
A map
NF : R × G → R, (f, G) 7→ NF(f | G),
is called a weak normal form on R if the following are satisfied:
(1) For all G ∈ G, NF(0 | G) = 0.
(2) For all f ∈ R and G ∈ G, NF(f | G) 6= 0 implies LM(NF(f | G)) 6∈ L(G).
(3) For all f ∈ R and G = {g1 , . . . , gs } ∈ G, there exists a unit u ∈ R∗ such that
uf − NF(f | G) (or, by abuse of notation, uf ) has a standard representation
with respect to NF(− | G), namely,
Xs
uf − NF(f | G) = ai gi (ai ∈ R)
i=1
Ps
with the property that LM( i=1 ai gi ) ≥ LM(ai gi ) for all i such that ai gi 6= 0.
A weak normal form NF is called polynomial if, whenever f ∈ K[x] and G ⊂ K[x],
in the condition (3) above, u and ai can be chosen as elements of K[x].
Definition 2.5 (cf. [4, 2.6]). Let G denote the set of all finite subsets of R. Let
h ∈ R and G ∈ G. Assume that there exist sequences {h0 = h, h1 , . . . , hm } ⊂ R and
{G0 = G, G1 , . . . , Gm } ⊂ G such that
∃gi ∈ Gi , LM(gi ) | LM(hi ), hi+1 = Spoly(hi , gi ), Gi+1 = Gi ∪ {hi },
hm = 0 or LM(g) ∤ LM(hm ) for all g ∈ Gm .
Then we call hm an s-normal form of h with respect to G. Clearly, we have {hi } ⊂
K[x] if h ∈ K[x] and G ⊂ K[x]. We write as
S-NF(h | G) = 0
to mean that an s-normal form of h with respect to G exists and is zero.
Note that in general, a sequence {h0 = h, h1 , . . . , hm } ⊂ R with the property in
Definition 2.5 does not exist; in other words, the procedure may not terminate (cf.
[2, §1.7]).
Buchberger’s criterion for Gröbner bases is also effective for standard bases with
local ordering.
Theorem 2.6 (Buchberger’s Criterion (cf. [2, 1.7.3])). Let I ⊂ R be an ideal and
G = {g1 , . . . , gs } ⊂ I. Let NF be a weak normal form on R with respect to G. Then
the following are equivalent:
(1) G is a standard basis of I.
(2) NF(f | G) = 0 for all f ∈ I.
(3) Each f ∈ I has a standard representation with respect to NF(− | G).
(4) G generates I and NF(Spoly(gi , gj ) | G) = 0 for i, j = 1, . . . , s.
(5) G generates I and NF(Spoly(gi , gj ) | Gij ) = 0 for a suitable subset Gij ⊂ G
and i, j = 1, . . . , s.
THE MULTIPLICITY OF CYCLIC COVERINGS OF A SINGULARITY 5
The following proposition is also useful in the proof of our main theorem.
Proposition 2.7 (See [4, 2.5-2.6, 3.1]). Let I ⊂ R be an ideal and G ⊂ I a finite
subset generating I. Then G is a standard basis of I if S-NF(Spoly(g, g ′) | G) = 0
for any g, g ′ ∈ G.
There is an algorithm, proposed by Mora ([4], cf. [2, 1.7.6]), that yields a polyno-
mial weak normal form. When employing a global ordering, NF(f | G) is obtained
as the last element of a sequence h0 = f, h1 , . . . , hm such that hi+1 = Spoly(hi , h′i ),
where h′i is a suitable element of G. Mora’s normal form with respect to a local
ordering is obtained in a similar manner, but h′i is an element of G ∪ {h0 , . . . , hi−1 }.
For the convenience of the readers, we state the algorithm.
α
P
For f = α aα x ∈ K[x] \ {0}, let
ecart(f ) = max { deg xα | aα 6= 0} − min { deg xα | aα 6= 0}.
Proposition 2.8 (cf. [2, 1.7.6]). Let G denote the set of all finite subsets of K[x].
Let f ∈ K[x] \ {0} and G ∈ G. We construct sequences {h0 = f, h1 , . . . } ⊂ K[x]
and {T0 = G, T1 , . . . } ⊂ G as follows: Suppose that we obtain {h0 , . . . , hi } and
{T0 , . . . , Ti }. If hi = 0 or Ti′ := {g ∈ Ti | LM(g) | LM(hi )} = ∅, then we stop here.
In case hi 6= 0 and Ti′ 6= ∅, take g ∈ Ti′ with ecart(g) = min {ecart(g ′) | g ′ ∈ Ti′ },
let Ti+1 = Ti ∪ {hi } if ecart(g) > ecart(hi ) and Ti+1 = Ti otherwise, and let hi+1 =
Spoly(hi , g). This algorithm terminates after a finite number of steps, and the last
one in the sequence {h0 = f, h1 , . . . } is a polynomial weak normal form of f with
respect to G.
Note that Mora’s normal form is also an s-normal form. The difference between
them is whether the invariant ecart is used in the process to obtain the normal
forms.
Using Mora’s normal form, we can compute the standard basis as follows (see [2,
1.7.1, 1.7.8]).
Proposition 2.9. Let I ⊂ R be an ideal generated by a finite subset G0 ⊂ K[x]
and let NF denote Mora’s polynomial weak normal form. We construct a sequence
{G0 ( G1 ( . . . } of finite subsets of K[x] as follows: if we obtain {G0 , . . . , Gi } and
if there exists f, g ∈ Gi such that h := NF(Spoly(f, g) | Gi ) 6= 0, then define Gi+1 to
be Gi ∪ {h}. This algorithm terminates after a finite number of steps, and the last
subset in the sequence is a standard basis for I.
Theorem 3.1. Let {f1 , . . . , fm } ⊂ K[x] \ {0} be a standard basis of I and let
G0 = {f1 , . . . , fm , g}. Let {g1 , . . . , gt } be a standard basis of I + hgi ⊂ K[x] which is
obtained as the last element Gk of the sequence G0 ( · · · ( Gk ⊂ K[x], as described
in Proposition 2.9. Let N = max { ord(gi ) | i = 1, . . . , t}. Then, for n > N, the
tangent cone of the local ring Bn is isomorphic to K[x, y]/hIn(g1 ), . . . , In(gt )i.
Proof. Assume that n > N. By Proposition 2.3, it is sufficient to prove that the
ideal In has a standard basis G such that { In(f ) | f ∈ G} = {In(g1 ), . . . , In(gt )}.
Note that all polynomials in Gk and those appearing in the algorithm for obtain-
ing Gk from G0 are represented as linear combinations of polynomials in G0 with
coefficients
Pm in K[x]. For each f of ¯those polynomials, we fix such a representation
f = i=1 ai fi + am+1 g, and define f ∈ K[x, y] as the polynomial obtained from f by
replacing g with g − y n in the representation; namely, f¯ = m n
P
i=1 i fi + am+1 (g − y ).
a
Let us express these procedures more precisely. For 0 ≤ i < k, we may assume
(i) (i)
that h(i) ∈ Gi+1 \ Gi is a unique element and h(i) = NF(Spoly(p1 , p2 ) | Gi ) with
(i) (i) (i) (i) (i) (i) (i)
p1 , p2 ∈ Gi . Let {h0 = Spoly(p1 , p2 ), h1 , . . . , hℓi = h(i) } be a sequence ob-
tained by the algorithm in Proposition 2.8. Let
k
! k−1
!
[ [ (i) (i) (i)
F= Gi ∪ {h0 , h1 , . . . , hℓi }
i=0 i=0
Moreover assume that Φ(fi ) = si for i = 1, . . . , m, and Φ(g) = sm+1 . For each
f ∈ F , we define a polynomial f by
f = Φ(f )(f1 , . . . , fm , g − y n ) ∈ K[x, y].
Let Gi = f¯ f ∈ Gi ⊂ K[x, y]. Then In = hG0 i.
We consider the following condition for 0 ≤ j ≤ k:
m
X
h̄t = ai fi + am+1 (g − y n ) = ht − am+1 y n .
i=1
Remark 3.2. In the proof above, the assumption that I is a prime ideal is only
applied to show that am+1 g ∈ I implies am+1 ∈ I. Therefore, this assumption
can be replaced with a weaker condition. In fact, it suffices to assume that g is
K[x]/I-regular, that is, if a ∈ K[x] and ag ∈ I, then a ∈ I.
Example 3.3. We show some simple examples with r = 2. We use the variables
(x, y, z) instead of (x1 , x2 , y) and the monomial ordering >ds .
(1) Let I = hxyi ⊂ K[x, y] and g = xα , where α is a positive integer. Then
g is not K[x, y]/I-regular. Assume that n > α ≥ 2. Then {xy, xα − z n , yz n } is
a standard basis of In = hxy, xα − z n i and In(In ) = hxy, xα , yz n i. From an exact
sequence
mult(Bn ) 6 9 12 12 12
Remark 3.4. In the situation of Theorem 3.1, we have
K[x, y]/hIn(g1 ), . . . , In(gt )i ∼
= (K[x]/hIn(g1 ), . . . , In(gt )i)[y].
Hence the affine tangent cone of Bn with n > N is isomorphic to the product of the
affine tangent cone of R/(I + hgi)R and the affine line A1K .
Remark 3.5. If A is normal and the image of g in A is reduced, then Bn are nor-
mal for every n ∈ Z>0 ([6]). Let us consider the case that Spec Bn are normal
two-dimensional singularities. Then, as n approaches infinity, the geometric genus
pg (Spec Bn ) tends towards infinity (cf. [1, §2]). Therefore, the tangent cone of a
normal two-dimensional singularity cannot provide upper bounds of the geometric
genus, even though any singularity is a small deformation of its tangent cone (cf.
[5, §5]).
References
[1] Ashikaga, T., Surface singularities on cyclic coverings and an inequality for the signature, J.
Math. Soc. Japan 51 (1999), no. 2, 485–521.
[2] Greuel, G.-M. and Pfister, G., A Singular Introduction to commutative algebra, 2nd extended
ed., Springer, 2007. With contributions by O. Bachmann, C. Lossen and H. Schönemann.
[3] Matsumura, H., Commutative ring theory, second ed., Cambridge Studies in Advanced Math-
ematics, vol. 8, Cambridge University Press, Cambridge, 1989. Translated from the Japanese
by M. Reid.
[4] Mora, F., An algorithm to compute the equations of tangent cones, Computer algebra, EURO-
CAM ’82, Conf. Marseille/France 1982, Lect. Notes Comput. Sci. 144, 158-165 (1982), 1982.
[5] Tomari, M. and Watanabe, K., Filtered rings, filtered blowing-ups and normal two-dimensional
singularities with “star-shaped” resolution, Publ. Res. Inst. Math. Sci. 25 (1989), no. 5, 681–
740.
[6] Tomari, M. and Watanabe, K., Normal Zr -graded rings and normal cyclic covers, Manuscripta
Math. 76 (1992), no. 3-4, 325–340.
[7] Tomaru, T., Maximal ideal cycles and multiplicities for cyclic coverings over normal surface sin-
gularities, Singularities – Kagoshima 2017. Proceedings of the 5th Franco-Japanese-Vietnamese
symposium on singularities, Kagoshima, Japan, October 27 – November 3, 2017, Hackensack,
NJ: World Scientific, 2020, pp. 201–214.
[8] Zariski, O. and Samuel, P., Commutative Algebra. Vol. II. Reprint of the 1958-1960 Van Nos-
trand edition, Grad. Texts Math., vol. 29, Springer, 1976.