This paper investigates the finite generation of symbolic Rees rings of space monomial curves with negative curves, providing criteria, examples, and computational methods to determine when these rings are Noetherian.
Contribution
It offers new conditions and criteria for the finite generation of symbolic Rees rings of space monomial curves with negative curves, including examples of infinite generation.
Findings
01
Determined minimal degree elements satisfying Huneke's criterion for Noetherian rings.
02
Provided necessary and sufficient conditions for finite generation of symbolic Rees rings.
03
Constructed an example of an infinitely generated symbolic Rees ring with a negative curve.
Abstract
In this paper, we shall study finite generation of symbolic Rees rings of the defining ideal p of the space monomial curve (ta,tb,tc) for pairwise coprime integers a, b, c. Suppose that the base field is of characteristic 0 and the above ideal p is minimally generated by three polynomials. Under the assumption that the homogeneous element ξ of the minimal degree in p is the negative curve, we determine the minimal degree of an element η such that the pair {ξ,η} satisfies Huneke's criterion in the case where the symbolic Rees ring is Noetherian. By this result, we can decide whether the symbolic Rees ring Rs(p) is Notherian using computers. We give a necessary and sufficient conditions for finite generation of the symbolic Rees ring of p under some assumptions. We give an example of an…
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsCommutative Algebra and Its Applications · Algebraic Geometry and Number Theory · Polynomial and algebraic computation
Full text
Infinitely generated symbolic Rees rings of
space monomial curves having negative curves
Kazuhiko Kurano and Koji Nishida
Abstract.
In this paper, we shall study finite generation
of symbolic Rees rings of the defining ideal p of
the space monomial curve (ta,tb,tc)
for pairwise coprime integers a, b, c.
Suppose that the base field is of characteristic [math] and
the above ideal p is minimally generated by
three polynomials.
In Theorem 1.1, under the assumption that the homogeneous element ξ of the minimal degree
in p is the negative curve,
we determine the minimal degree of an element η such that
the pair {ξ,η} satisfies Huneke’s criterion
in the case where the symbolic Rees ring is Noetherian.
By this result, we can decide whether the symbolic Rees ring
Rs(p) is Notherian using computers.
We give a necessary and sufficient conditions for finite generation of
the symbolic Rees ring of p in Proposition 4.10
under some assumptions.
We give an example of an infinitely generated symbolic Rees ring of p
in which the homogeneous element of the minimal degree
in p(2) is the negative curve in Example 5.6.
We give a simple proof to (generalized) Huneke’s criterion.
1. Introduction
Let pK(a,b,c) be the defining ideal of the space monomial
curve (ta,tb,tc) in K3, where K is a field.
The ideal pK(a,b,c) is generated by at most three binomials in K[x,y,z]
(Herzog [11]).
The symbolic Rees rings of space monomial primes are deeply studied by many authors.
Huneke [12] and Cutkosky [2] developed criterions
for finite generation of such rings.
In 1994,
Goto-Nishida-Watanabe [7] first found examples of infinitely generated
symbolic Rees rings of space monomial primes.
Recently, using toric geometry,
Gonzáles-Karu [5] found some sufficient conditions for
the symbolic Rees rings of space monomial primes to be infinitely generated.
Cutkosky [2] found the geometric meaning of the symbolic Rees rings of space monomial primes.
Let PK(a,b,c) be the weighted projective surface with degree a, b, c.
Let XK(a,b,c) be the blow-up at a point in the open orbit of the toric variety PK(a,b,c).
Then, the Cox ring of XK(a,b,c) is isomorphic to the extended symbolic Rees ring
of the space monomial prime pK(a,b,c).
Therefore, the symbolic Rees ring of the space monomial prime pK(a,b,c)
is finitely generated if and only if the Cox ring of XK(a,b,c) is finitely generated,
that is, XK(a,b,c) is a Mori dream space.
A curve C on XK(a,b,c) is called the negative curve if C2<0 and
C is different from the exceptional curve E.
Here suppose abc∈Q.
Cutkosky [2] proved that the symbolic Rees ring of the space monomial prime pK(a,b,c) is finitely generated if and only if
the following two conditions are satisfied:
(1)
There exists a negative curve C.
(2)
There exists a curve D on XK(a,b,c) such that C∩D=∅.
All known examples ([8], [5]) of the infinitely generated symbolic Rees rings of pK(a,b,c) satisfy
the following conditions:
(I)
there exists a negative curve C such that C.E=1.
(II)
the characteristic of K is [math]
In this paper, we give an example of an infinitely generated symbolic Rees ring such that there exists the negative curve C
with C.E=2.
Furthermore, in the case where both (I) and (II) as above are satisfied,
we determine the minimal value of the degree of the curve D
which satisfies the condition (2) as above.
The existence of negative curves is a very difficult problem,
that is deeply related to Nagata conjecture (Proposition 5.2 in Cutkosky-Kurano [3]).
In the rest of this section,
we state the results of this paper precisely.
Let a, b, c be pairwise coprime integers.
We regard the polynomial ring S=K[x,y,z] as a Z-graded ring
by deg(x)=a, deg(y)=b and deg(z)=c.
Let pK(a,b,c) be the kernel of the K-algebra
homomorphism
[TABLE]
given by ϕK(x)=ta, ϕK(y)=tb, ϕK(z)=tc.
If no confusion is possible, we simply denote pK(a,b,c) by p.
By a result of Herzog [11], we know that pK(a,b,c)
is generated by at most three binomials.
We define s, t, u to be
[TABLE]
where N (resp. N0) denotes the set of positive integers
(resp. non-negative integers).
Let t1, u1, s2, u2, s3, t3 be non-negative integers such that
sa=t1b+u1c, tb=s2a+u2c, uc=s3a+t3b.
Then, pK(a,b,c) is minimally generated by three elements
if and only if s, t, u≥2.
When this is the case, pK(a,b,c) is minimally generated by three elements
[TABLE]
and t1, u1, s2, u2, s3, t3 must be positive integers
satisfying s=s2+s3, t=t1+t3, u=u1+u2.
For a prime ideal P of S,
we define the symbolic Rees ring of P to be
[TABLE]
where P(n)=PnSP∩S is the nth symbolic power of P
and T is an indeterminate.
Here, Rs(P) is a Noetherian ring if and only if
Rs(P) is finitely generated over S as a ring.
In Section 2,
we give a simple proof to Huneke’s criterion [12].
We slightly generalize Huneke’s criterion here.
Furthermore, we develop the method of mod p reduction
introduced in Goto-Nishida-Watanabe [7].
In Section 3, we give a proof to the following theorem:
Theorem 1.1**.**
Let a, b, c be pairwise coprime positive integers.
Assume the following three conditions:
(i)
K* is a field of characteristic [math],*
(ii)
pK(a,b,c)* is minimally generated by
the three elements as in (2),*
(iii)
uc<abc.
Then, Rs(pK(a,b,c)) is a Noetherian ring
if and only if there exists η in [p(u)]ab
such that zu−xs3yt3 and η satisfy
Huneke’s condition [12] (see Theorem 2.5), that is,
[TABLE]
holds.
The condition (iii) as above implies that zu−xs3yu3 is the negative curve,
that is, there exists the negative curve C such that C.E=1.
Theorem 1.1 says that there exists a curve D such that D∩C=∅ and
D∼abA−uE if and only if Rs(pQ(a,b,c)) is a Noetherian ring,
where A is an Weil divisor on X satisfying OX(A)=π∗OP(1).
We emphasis that it is possible to verify weather there exists
η in [p(u)]ab satisfying (3) as above
using computers.
We shall prove this theorem using
the mod p reduction as in Goto-Nishida-Watanabe [8],
Fujita’s vanishing theorem (Theorem 1.4.35 in [13]) and Cutkosky’s method [2] in characteristic p>0.
The most important point is that the negative curve is isomorphic to
PK1 in this case.
In Section 4, we introduce the condition EU.
In Ebina [4] and Uchisawa [15],
the condition EU was defined and they proved that the condition EU is a sufficient condition
for finite generation under the assumptions (i), (ii), (iii) in Theorem 1.1.
For the convenience of the reader, we shall give a proof of it in this paper.
Furthermore, in the case where u≤6,
we show that the condition EU is a necessary and sufficient condition for the finite generation of the symbolic Rees ring of p in Propositin 4.10.
In Section 5 we give an example of infinitely generated symbolic Rees ring of p
where the homogeneous element of the minimal degree
in p(2) is the negative curve in Example 5.6.
We emphasis that
one of the minimal generators of p is the negative curve in
all known examples of infinitely generated Rs(pK(a,b,c)),
except for this example.
2. Huneke’s condition and mod p reduction
Let S=K[x,y,z],
where K is a field and x,y,z are indeterminates.
We regard S as a Z-graded ring putting suitable weights on x,y and z.
We set m=(x,y,z)S and R=Sm.
Let I be a homogeneous proper ideal of S
satisfying the following conditions;
•
(x)+I is m-primary,
•
AssSS/I=AsshSS/I:={p∈AssSS/I∣dimS/p=dimS/I}, and
•
Ip is generated by 2 elements for any p∈AsshSS/I.
Then, S/I is a Z-graded Cohen-Macaulay ring of
dimS/I=1.
If we replace x in the first assumption stated above with y or z,
it can play the same role as x in the arguments of this section.
So, homogeneous prime ideals of height 2 are typical examples of I.
For any n∈Z, we set
[TABLE]
where Ipn denotes the ideal (In)p=(Ip)n of Sp.
Then, we have AssSS/I(n)=AsshSS/I if n>0,
and the equality
I(n)=In:Sxi holds for i≫0,
which means that I(n) is a homogeneous ideal of S
and (I(n))x=Ixn, where Ixn denotes the ideal
(In)x=(Ix)n of Sx.
Moreover, we set
[TABLE]
where T is an indeterminate
and I(n)=S for n≤0.
Let us call Rs(I) the symbolic Rees ring of I.
We set a=Im=IR.
It is easy to see that R/a is a Cohen-Macaulay local ring
of dimR/a=1 and
AssRR/a=AsshRR/a={pR∣p∈AsshSS/I}.
Moreover, for p∈AsshSS/I, we have apR=Ip,
which becomes a parameter ideal of RpR=Sp.
For any n∈Z, we set
[TABLE]
Then, we have a(n)=I(n)R and
AssRR/a(n)=AsshRR/a if n>0.
As a(n)=an:Rxi holds for i≫0,
we have (a(n))x=axn.
The R-algebras Rs(a) and Gs(a) are derived from
Rs(I) and Gs(I) respectively applying R⊗S∗ .
If Rs(a) is finitely generated,
then there exists 0<m∈Z such that
a(mn)=(a(m))n for any n∈Z.
This equality implies I(mn)=(I(m))n since
I(mn)⊇(I(m))n and (I(m))n is a homogeneous ideal.
Thus we see that
Rs(I) is finitely generated if so is Rs(a).
The converse of this assertion holds obviously.
For a proper ideal J of S such that S/J is Artinian,
we have ℓS(S/J)≥ℓR(R/JR),
and the equality holds if and only if J is m-primary,
which holds if J is homogeneous.
The purpose of this section is to review the condition on I
for its symbolic Rees ring to be finitely generated,
which was originally given by Huneke [12]
in the case where I is a prime ideal of a 3-dimensional regular local ring.
Furthermore, using mod p reduction technique for
prime numbers p≫0,
we give a condition on I for
Rs(I) to be infinitely generated,
which is a modification to the method introduced in [8].
Let us begin with the following
Proposition 2.1**.**
Let 0<k,ℓ∈Z, ξ∈I(k) and η∈I(ℓ).
Then we have
[TABLE]
where the equality holds if and only if
a⊆(ξ,η)R and
[TABLE]
for all p∈AsshSS/I.
In order to prove Proposition 2.1, let us recall the following fact.
Lemma 2.2**.**
Let A be a 2-dimensional Cohen-Macaulay local ring and
Q a parameter ideal of A.
Let 0<k,ℓ∈Z,
ξ∈Qk and η∈Qℓ.
We assume that ξ,η is an sop for A.
Then, we have
[TABLE]
and the equality holds if and only if one of the following conditions,
which are equivalent to each other,
is satisfied;
(1)
QT⊆(ξTk,ηTℓ)R(Q),
where R(Q)=∑n≥0QnTn⊂A[T],
(2)
ξTk,ηTℓ* is an sop for G(Q)=R(Q)/QR(Q),*
(3)
Qk+ℓ−1=ξQℓ−1+ηQk−1,
(4)
Qn∩(ξ,η)A=ξQn−k+ηQn−ℓ*
for any n∈Z.*
Proof.
We set J=(ξℓ,ηk)A.
Then we have
[TABLE]
where eJ(A) denotes the multiplicity of A with respect to J.
Because J⊆Qkℓ, it follows that
[TABLE]
Hence we get the required inequality.
Moreover, we see that the equality holds if and only if
J is a reduction of Qkℓ,
which is a condition equivalent to (1).
The equivalence of the conditions (1) and (2) is obvious.
Let us notice that G(Q) is isomorphic to a polynomial ring
with 2 variables over A/Q,
so its homogeneous sop is always a regular sequence,
which implies the equivalence of the conditions (2) and (4).
Moreover, if the condition (2) is satisfied,
it follows that G(Q)/(ξTk,ηTℓ)G(Q) is an
Artinian Z-graded ring whose a-invariant is k+ℓ−2 (cf. [10]),
so the equality of the condition (3) holds.
Finally, if the condition (3) is satisfied, we have
[TABLE]
and hence the condition (1) is satisfied.
∎
Proof of Proposition 2.1.
We may assume that (x,ξ,η)R is mR-primary.
Then, as R/(ξ,η)R is a Cohen-Macaulay R-module,
for which x is an sop, we have
[TABLE]
Here we notice that
pR∈AsshRR/(ξ,η)R
for any p∈AsshSS/I.
Hence, using additive formula of multiplicity and Lemma 2.2, we get
[TABLE]
Thus we get the required inequality.
Moreover, we see that the equality holds if and only if
AsshRR/(ξ,η)R=AsshRR/a and
ℓSp(Sp/(ξ,η)Sp)=kℓ⋅ℓSp(Sp/Ip)
for all
p∈AsshSS/I.
Since a⊆(ξ,η)R holds if and only if
AsshRR/(ξ,η)R=AsshRR/a,
the proof is complete.
∎
Definition 2.3**.**
Let 0<k,ℓ∈Z,ξ∈I(k) and η∈I(ℓ).
We say that ξ and η
satisfy Huneke’s condition on I (with respect to x) if
[TABLE]
When this is the case,
for any 0<i,j∈Z,
ξi∈I(ki) and ηj∈I(ℓj) also satisfy
Huneke’s condition on I.
Even if there exist elements satisfying Huneke’s condition,
those elements may not be homogeneous.
Although the existance of homogeneous elements
satisfying Huneke’s condition is not clear,
but it can be verified elementary in special cases.
For example, the following remark implies that
if ξ and η satisfy Huneke’s condition and
ξ≡yi mod xS for some 0<i∈Z,
then we can choose homogeneous parts of ξ and η
so that they also satisfy Huneke’s condition.
Lemma 2.4**.**
Suppose ξ∈S and ξ≡yi mod xS, where 0<i∈Z.
Let ξ′ be the homogeneous part of ξ containing yi as a term.
Then, the following assertions hold.
(1)
(x,ξ)S=(x,ξ′)S=(x,yi)S.
(2)
For any η∈S, we can choose its homogeneous part η′ so that
[TABLE]
Proof.
The assertion (1) holds obviously.
Let us verify the assertion (2).
We may assume η∈(x,y)S.
Then, as x,y,η is an R-regular sequence, we have
[TABLE]
We write
[TABLE]
where 0<j∈Z and αj,αj+1,…
are elements of K with αj=0.
Since αj+αj+1z+⋯ is a unit of R, we have
[TABLE]
Thus we get
[TABLE]
Let η′ be the homogeneous part of
η containing αjzj as a term.
Then, as η′≡αjzj mod (x,y)S, it follows that
[TABLE]
Thus we get the required equality.
∎
Theorem 2.5**.**
The symbolic Rees algebra Rs(I) is finitely generated over R
if and only if
there exist elements in I(k) and I(ℓ)
satisfying Huneke’s condition on I for some 0<k,ℓ∈Z.
Proof.
First, let us assume that
Rs(I) is finitely generated.
Then, there exists a positive integer m such that
I(mn)=(I(m))n for any n∈Z.
We set b=a(m).
Then, for any 0<n∈Z,
we have a(mn)=bn,
which means 0ptRR/bn=1.
Hence, by Burch’s theorem (cf. [1]), we see
[TABLE]
where λ(b) denotes the Krull dimension of R/m⊗G(b),
which is called the analytic spread of b.
Thus we get λ(b)=2.
Hence, we can choose 0<i,j∈Z, ξ∈I(mi) and
η∈I(mj) such that ξTi,ηTj is an sop for R/m⊗G(b).
(Here, we notice that we don’t have to assume that K is infinite
since we don’t require i=j=1.)
Let us take r≫0.
Then, we have br=ξbr−i+ηbr−j,
which means amr⊆(ξ,η)R,
and so a⊆(ξ,η)R.
Moreover, if p∈AsshSS/I and mr<n, we have
[TABLE]
which means that ξTmi,ηTmj is an sop for G(Ip).
Therefore, by Proposition 2.1 and Lemma 2.2, it follows that
ξ and η satisfy Huneke’s condition on I.
Next, we assume that
there exist 0<k,ℓ∈Z, ξ∈I(k) and η∈I(ℓ)
such that ξ and η
satisfy Huneke’s condition on I.
We set m=kℓ, b=a(m) and
c=(ξℓ,ηk)R⊆b.
Let us look at the exact sequence
[TABLE]
of R-modules, where r is any non-negative integer.
Since ξℓ,ηk is an R-regular sequence,
cr/cr+1 is R/c-free,
so cr/bcr≅R/b⊗Rcr/cr+1
is R/b-free, which means
since ξℓ and ηk also satisfy Huneke’s condition on I.
Thus we see
[TABLE]
Now we take any P∈AsshRR/a,
and write P=pR,
where p∈AsshSS/I.
Then, by Proposition 2.1 and Lemma 2.2 we have
Ip2m−1=ξℓIpm−1+ηkIpm−1,
which means bP2=(bc)P, and so bPr+1=(bcr)P.
Hence, we get
[TABLE]
and hence a(mr+m)=bcr=br+1.
Thus we see that the m-th veronese subring of
Rs(a) is generated in degree one.
Therefore Rs(a) is
Noetherian by [6, Lemma (2.4)].
Then Rs(I) itself must be Noetherian.
∎
Lemma 2.6**.**
Let 0<k,ℓ∈Z, ξ∈I(k) and η∈I(ℓ).
Suppose that ξ and η satisfy Huneke’s condition on I.
Then the following assertions hold.
(1)
Rs(a)+=(ξTk,ηTℓ)Rs(a),
and hence
Gs(a)+⊆(ξTk,ηTℓ)Gs(a).
(2)
a(k+ℓ−1)⊆(ξ,η)R.
(3)
axn∩(ξ,η)Rx=ξaxn−k+ηaxn−ℓ*
for any n∈Z.*
(4)
a(n)∩(ξ,η)R=ξa(n−k)+ηa(n−ℓ)* if n≤k+ℓ.*
(5)
If k=1 or 2, then we have
[TABLE]
for any n∈Z, which means that
ξTk,ηTℓ is a regular sequence on Gs(a),
and hence gradeGs(a)+=2.
Proof.
(1) We set m=kℓ,b=a(m) and c=(ξℓ,ηk)R.
Then, as is stated in the proof of Theorem 2.5,
we have a(mr+m)=br+1=bcr
for any 0≤r∈Z.
Let us take any 0<n∈Z and ρ∈a(n).
Then we have ρ2m∈a(m(2n−1)+m)=b2n=bc2n−1⊆cb2n−1=ξℓb2n−1+ηkb2n−1⊆ξa(2mn−k)+ηa(2mn−ℓ), so
[TABLE]
Hence we get the assertion (1)
(2)
Let us take any P∈AsshRR/a and write P=pR,
where p∈AsshSS/I.
Then, as RP=Sp and aP=Ip,
by Proposition 2.1 and Lemma 2.2, we have
aPk+ℓ−1=ξaPℓ−1+ηaPk−1⊆(ξ,η)RP.
Therefore we get
[TABLE]
(3)
Since ξ∈Ixk and η∈Ixℓ,
the inclusion axn⊇ξaxn−k+ηaxn−ℓ holds obviously.
So, it is enough to show
[TABLE]
for any P∈SpecR satisfying
ξan−k+ηan−ℓ⊆P and x∈P.
Such P must contains a since a⊆(ξ,η)R,
so there exists p∈AsshSS/I such that P=pR.
Then, by Proposition 2.1 and Lemma 2.2, we get the required equality
as RP=Sp and aP=Ip.
(4)
Let n≤k+ℓ and φ∈a(n)∩(ξ,η)R.
We write φ=ξu+ηv, where u,v∈R.
Since φ∈axn∩(ξ,η)Rx=ξaxn−k+ηaxn−ℓ by (3),
there exist α∈axn−k and β∈axn−ℓ
such that φ=ξα+ηβ.
Here, we take i≫0 so that xiα∈an−k and xiβ∈an−ℓ.
Then we have
xi(ξu+ηv)=xiφ=xi(ξα+ηβ), so
ξ(xiu−xiα)=η(xiβ−xiv).
Since ξ,η is an R-regular sequence,
it follows that
xiu−xiα∈ηR⊆a(ℓ)⊆a(n−k) and
xiβ−xiv∈ξR⊆a(k)⊆a(n−ℓ).
Hence xiu∈a(n−k) and xiv∈a(n−ℓ),
which means u∈a(n−k) and v∈a(n−ℓ).
Thus we get φ∈ξa(n−k)+ηa(n−ℓ).
(5)
Let k=1 or 2.
By (2) and (4), it is enough to show
[TABLE]
assuming n>k+ℓ.
We take positive integers m and r such that
n−ℓ=km−r and 0≤r<k.
Then, m≥2 and r is [math] or 1.
Since ξm∈I(km) and η∈I(ℓ)
also satisfy Huneke’s condition on I and
km+ℓ−1≤km+ℓ−r=n≤km+ℓ,
we have a(n)⊆(ξm,η)R by (2) and
a(n)∩(ξm,η)R=ξma(ℓ−r)+ηa(n−ℓ) by (4).
Let us notice that ξm−1a(ℓ−r)⊆a(n−k) as
k(m−1)+(ℓ−r)=n−k.
Thus we see a(n)⊆ξa(n−k)+ηa(n−ℓ).
Since the converse inclusion is obvious, we get the required equality.
∎
Definition 2.7**.**
Let 0<k∈Z and ξ∈I(k).
We denote by HC(I;k,ξ) the set of positive integers ℓ
for which there exists η∈I(ℓ) such that
ξ and η satisfy Huneke’s condition on I.
Remark 2.8**.**
Let k and ξ be as in Definition 2.7.
If ξ≡yi mod xS, where 0<i∈Z,
and ξ′ is the homogeneous part of ξ containing yi as a term,
we have HC(I;k,ξ)=HC(I;k,ξ′) by Lemma 2.4 (1).
Proposition 2.9**.**
Let k=1 or 2, and let ξ∈I(k).
Suppose that ξ≡yimodxS
for some 0<i∈Z and HC(I;k,ξ)=ϕ.
We set m=minHC(I;k,ξ).
Then the following assertions hold.
(1)
HC(I;k,ξ)={m,2m,3m,⋯}.
(2)
S[{I(n)Tn∣1≤n≤m−1}]⊊Rs(I).
(3)
If there exist elements in I(k′) and I(ℓ′)
satisfying Huneke’s condition on I for 0<k′,ℓ′∈Z,
we have
[TABLE]
In particular,
[TABLE]
Proof.
By Remark 2.8,
we may assume that ξ is homogeneous.
Then, by Lemma 2.4 (2), we can choose a homogeneous element η∈I(m)
such that ξ and η satisfy Huneke’s condition on I.
(1)
We obviously have HC(I;k,ξ)⊇{m,2m,3m,⋯}.
In order to show the converse inclusion,
We suppose that there exists ℓ∈HC(I;k,ξ) which is not a multiple of m.
Let us choose such ℓ as small as possible.
Then, there exists a homogeneous element ρ∈I(ℓ)
such that ξ and ρ satisfy Huneke’s condition on I.
Since m<ℓ, by Lemma 2.6 (2) and (5), we have
a(ℓ)=ξa(ℓ−k)+ηa(ℓ−m), which implies
[TABLE]
as ξ and η are homogeneous.
Hence, there exists a homogeneous element ρ′∈I(ℓ−m) such that
[TABLE]
Then ρ∈(ξ,ρ′)S, and hence we get
[TABLE]
as a⊆(ξ,ρ)R by Proposition 2.1.
Now we take any p∈AsshSS/I and n≫0.
Then, by Proposition 2.1 and Lemma 2.2, we have
[TABLE]
so we get
[TABLE]
Therefore, ξ and ρ′ satisfy Huneke’s condition on I,
so ℓ−m∈HC(I;k,ξ), which contradicts to the minimality
of ℓ as ℓ−m is not a multiple of m.
Consequently, we see that any ℓ∈HC(I;k,ξ) is a multiple of m.
(2)
The assertion holds obviously if m=1,
so let us consider the case where m≥2.
Suppose
[TABLE]
Then we have
[TABLE]
We set S=S/(x,y)≅K[z].
Since any homogeneous ideal of S is a power of zS,
[TABLE]
holds for some β=1,2,…,m−1.
Moreover, we can choose homogeneous elements
ρ∈I(β) and ρ′∈I(m−β)
such that η and ρρ′ have the same class in S,
which is equivalent to
Since (x,y,η), (x,y,ρ) and (x,y,ρ′) are all
homogeneous m-primary ideals, we have
[TABLE]
Consequently, it follows that
[TABLE]
Hence we get β,m−β∈HC(I;k,ξ),
which contradicts to the minimality of m.
Thus we see
[TABLE]
(3)
Let 0<k′,ℓ′∈Z, ξ′∈I(k′) and η′∈I(ℓ′).
Suppose that ξ′ and η′ satisfy Huneke’s condition on I.
Then, by Lemma 2.6 (1), we have
[TABLE]
On the other hand, from the existance of ξ and η, we see
gradeGs(a)+=2 by Lemma 2.6 (5).
Hence, it follows that ξ′Tk′,η′Tℓ′
is a regular sequence on Gs(a).
If k′+ℓ′−1≤n, we have a(n)⊆(ξ′,η′)R
by Lemma 2.6 (2), so
[TABLE]
Thus we see
[TABLE]
which means that the first assertion of (3) holds.
We get the last assertion taking k and m as k′ and ℓ′, respectively.
∎
In the rest of this section,
let SZ=Z[x,y,z].
Moreover, for a field K,
we denote K[x,y,z] by SK instead of S
in order to emphasize that the coefficient field is K.
Putting suitable weights on x,y and z,
we regard SZ and SK as Z-graded rings.
We set mZ=(x,y,z)SZ, mK=(x,y,z)SK and RK=(SK)mK.
When we denote an ideal of SZ by JZ,
the ideal JZSK is denoted by JK.
Similarly, when we denote an element of SZ by ξZ,
its image in SK is denoted by ξK.
For a prime number p, we set Fp=Z/pZ.
Of course, SQ=(Z∖{0})−1SZ and
SFp=SZ/pSZ.
Lemma 2.10**.**
Let JZ be an ideal of SZ.
Then, we have
[TABLE]
for any prime number p≫0.
If JZ is homogeneous, we may replace
RQ,(JQ)mQ,RFp and (JFp)mFp with
SQ,JQ,SFp and JFp, respectively.
Proof.
First, let us consider the case where
RQ/(JQ)mQ is Artinian.
We prove the required equality
by induction on ℓRQ(RQ/(JQ)mQ).
If ℓRQ(RQ/(JQ)mQ)=0, then
JQ contains an element which does not belong to mQ,
so there exists ξZ∈JZ∖mZ.
Let us take a prime number p≫0 so that
the constant term of ξZ, which is non-zero,
is not a multiple of p.
Then, ξFp∈JFp∖mFp.
Hence (JFp)mFp=RFp, so
ℓRFp(RFp/(JFp)mFp)=0.
Now we suppose ℓRQ(RQ/(JQ)mQ)>0.
Then, as mZ∈AssSZSZ/JZ,
there exists ηZ∈SZ such that JZ:ηZ=mZ.
We set LZ=JZ+(ηZ).
Let us notice LQ/JQ≅SQ/mQ.
Hence we have ℓRQ((LQ/JQ)mQ)=1, so
[TABLE]
Here, we take a prime number p≫0.
Then the hypothesis of induction implies
[TABLE]
Moreover, by taking larger p if necessary,
we may assume that p is regular on SZ/LZ.
If ηFp∈JFp,
we have ηZ∈JZ+pSZ,
so there exists ρZ∈SZ such that
ηZ≡p⋅ρZ mod JZ.
Since p is regular on SZ/LZ, we have ρZ∈LZ,
so there exists σZ∈SZ such that
ρZ≡ηZσZ mod JZ.
Then, we have ηZ≡p⋅ηZσZ mod JZ,
and hence 1−p⋅σZ∈JZ:ηZ=mZ,
which is impossible.
Thus we see ηFp∈JFp.
Hence we have mFp=JFp:ηFp since
ηFp⋅mFp⊆JFp holds obviously.
Then, we get LFp/JFp≅RFp/mFp,
so ℓRFp(LFp/JFp)=1.
Consequently,
[TABLE]
Therefore, the required equality follows.
Next, we assume dimRQ/(JQ)mQ>0,
and aim to prove dimRFp/(JFp)mFp>0 for p≫0.
In this case, there exists PZ∈SpecSZ
such that JZ⊆PZ⊊mZ.
Let us take any τZ∈mZ∖PZ
and choose a prime number p≫0 so that
p is regular on SZ/(τZ)+PZ.
Then, as p,τZ is a regular sequence on (SZ/PZ)(pSZ+mZ),
it follows that τZ is regular on
(SZ/pSZ+PZ)(pSZ+mZ)≅RFp/(PFp)mFp.
Hence, we have dimRFp/(PFp)mFp>0,
and so dimRFp/(JFp)mFp>0 as JFp⊆PFp.
∎
In the rest of this section, let IZ be a homogeneous ideal of SZ contained in mZ.
We assume that the following conditions are satisfied for any field K;
•
(x)+IK is mK-primary,
•
AssSKSK/IK=AsshSKSK/IK, and
•
(IK)p is generated by 2 elements for any p∈AsshSKSK/IK.
Furthermore, for any n∈Z, we set
(I(n))Z=⋃i>0((IZ)n:SZxi),
which is a homogeneous ideal of SZ.
Let us denote (I(n))ZSK by (I(n))K for any field K.
Lemma 2.11**.**
The following assertions hold for any n∈Z.
(1)
(IQ)(n)=(I(n))Q.
(2)
(IFp)(n)=(I(n))Fp* for any prime number p≫0.*
Proof.
First, let us notice that, for any field K,
we have (IK)(n)=⋃i>0((IK)n:SKxi),
and hence (IK)(n)⊇(I(n))K holds.
The converse inclusion holds obviously if K=Q.
So, we have to prove (IFp)(n)⊆(I(n))Fp for p≫0.
Let us take a prime number p≫0 so that
p is regular on SZ/(x)+(I(n))Z.
Moreover, we take any ξZ∈SZ satisfying
ξFp∈(IFp)(n).
Then, there exists 0<i∈Z such that
xiξFp∈(IFp)n,
which means xiξZ∈pSZ+(IZ)n⊆pSZ+(I(n))Z.
Hence, there exists ηZ∈SZ such that
xiξZ≡pηZ mod (I(n))Z.
Since x,p is a regular sequence on SZ/(I(n))Z,
so is xi,p.
Hence ηZ∈(xi)+(I(n))Z, so there exists
ρZ∈SZ such that
ηZ≡xiρZ mod (I(n))Z.
Then we have xiξZ≡pxiρZ mod (I(n))Z,
which means ξZ−pρZ∈(I(n))Z.
Thus we get ξFp∈(I(n))Fp.
∎
Proposition 2.12**.**
Let 0<k,ℓ∈Z,ξZ∈(I(k))Z and
ηZ∈(I(ℓ))Z.
Suppose that ξQ∈(IQ)(k) and ηQ∈(IQ)(ℓ)
satisfy Huneke’s condition on IQ.
Then, for any prime number p≫0,
ξFp∈(IFp)(k), ηFp∈(IFp)(ℓ),
and these elements satisfy Huneke’s condition on IFp.
Proof.
Let p≫0.
Then, by Lemma 2.11, we have
ξFp∈(IFp)(k) and ηFp∈(IFp)(ℓ).
Moreover, by Lemma 2.10, we have
[TABLE]
Thus we get the required assertion.
∎
Theorem 2.13**.**
Let k=1 or 2.
Let ξZ∈(I(k))Z and
ξZ≡yimodxSZ
for some 0<i∈Z.
Suppose that there exists a positive integer r
such that, for any prime number p≫0, we have
rpep∈HC(IFp;k,ξFp)
for some 0<ep∈Z.
Then the following conditions are equivalent:
(1)
Rs(IQ)* is finitely generated.*
(2)
HC(IQ;k,ξQ)=∅.
(3)
r∈HC(IQ;k,ξQ).
(4)
r∈HC(IFp;k,ξFp)* for any prime number p≫0.*
Proof.
Let ξZ′ be the homogeneous part of ξZ
containing yi as a term.
Then, as ξZ′∈(I(k))Z,
we have ξQ′∈(IQ)(k) and ξFp′∈(IFp)(k) for p≫0
by Lemma 2.11.
Moreover, by Remark 2.8, we have
HC(IQ;k,ξQ)=HC(IQ;k,ξQ′) and
HC(IFp;k,ξFp)=HC(IFp;k,ξFp′) for p≫0.
Hence, by replacing ξZ with ξZ′,
we may assume that ξZ is homogeneous from the beginning.
It is easy to see (3)⇒(2)⇒(1).
Now, we start to prove (1)⇒(4).
By Theorem 2.5 and Lemma 2.11 (1),
there exist 0<k′,ℓ′∈Z,ζZ∈(I(k′))Z
and ρZ∈(I(ℓ′))Z such that
ζQ∈(IQ)(k′) and ρQ∈(IQ)(ℓ′)
satisfy Huneke’s condition on IQ.
Here, we take a prime number p≫0 such that
ζFp∈(IFp)(k′), ρFp∈(IFp)(ℓ′),
and these elements satisfy Huneke’s condition on IFp,
which is possible by Proposition 2.12.
By taking larger p if necessary, we may assume p>max{k′,ℓ′,k′+ℓ′−2}
and our assumption on HC(IFp;k,ξFp) is satisfied.
Then, as HC(IFp;k,ξFp)=ϕ, we have
[TABLE]
by Proposition 2.9 (3).
We set m=minHC(IFp;k,ξFp) and take 0<ep∈Z
such that rpep∈HC(IFp;k,ξFp).
Then, by Proposition 2.9 (1), there exists m′∈Z
such that rpep=mm′.
Since Proposition 2.9 (2) implies m<p,
m is not a multiple of p,
so m′ is a multiple of pep.
Hence r is a multiple of m,
which means r∈HC(IFp;k,ξFp).
Next, we shall prove (4)⇒(3).
Let us take a prime number p≫0 such that
r∈HC(IFp;k,ξFp),
ℓSQ(SQ/(x)+IQ)=ℓSFp(SFp/(x)+IFp)
and (IFp)(r)=(I(r))Fp,
which is possible by Lemma 2.10 and Lemma 2.11.
Then, by Lemma 2.4 (2) and Lemma 2.11,
there exists a homogeneous element ηZ∈(I(r))Z
such that
ξFp∈(IFp)(k) and ηFp∈(IFp)(r)
satisfy Huneke’s condition on IFp.
We write
[TABLE]
where j is a positive integer and α is an integer
which is not a multiple of p.
Let K=Q or Fp.
Then, as the image of
α in K is not vanished,
we have (x,y,ηK)SK=(x,y,zj)SK.
Hence we get
Let K be a field and a, b, c be pairwise coprime positive integers.
We regard the polynomial ring S=K[x,y,z] as a Z-graded ring
by deg(x)=a, deg(y)=b and deg(z)=c.
We denote by PK(a,b,c) the weighted projective space ProjS.
Let
[TABLE]
be the blow-up at the point corresponding to pK(a,b,c).
We remark that PK(a,b,c) is non-singular at this point
(e.g., Lemma 9 in [2]).
If no confusion is possible, we denote
pK(a,b,c) (resp. XK(a,b,c), PK(a,b,c))
simply by p (resp. X, P).
Let E be the exceptional divisor of π.
Let A be a Weil divisor on X which satisfies
OX(A)=π∗OP(1).
Since a, b, c are pairwise coprime,
we have OX(nA)=π∗OP(n)
for any n∈Z (e.g., [14]).
Then,
[TABLE]
with the intersection pairing
[TABLE]
Definition 3.1**.**
A curve C on XK(a,b,c) is called a negative curve on XK(a,b,c)
if C2<0 and C=E.
An irreducible homogeneous polynomial ξ in [pK(a,b,c)(r)]d is called
a negative curve in pK(a,b,c)(r) if d/r<abc.
If a negative curve C on XK(a,b,c) exists, then it is unique.
If a negative curve ξ in [pK(a,b,c)(r)]d exists,
then r and d are uniquely determined, and ξ
is also unique up to multiplication
by an element in K×.
The negative curve C on XK(a,b,c) is the proper transform
of V+(ξ).
Lemma 3.2**.**
Let K be a field and a, b, c be pairwise coprime positive integers.
We assume that pK(a,b,c) is minimally generated by
the three elements in (2).
Then the curve V+(zu−xs3yt3) in PK(a,b,c)
is isomorphic to PK1.
The proper transform C (in X) of this curve is also isomorphic to PK1.
Proof.
First of all, we remark that zu−xs3yt3 is an irreducible polynomial
by definition of u (see (1)).
We put v=xs2zu2/yt and w=xs3yt3/zu.
Since pK(a,b,c) is generated by the three elements as in (2),
we have
[TABLE]
Then, we have
[TABLE]
Taking the degree [math] component of
[TABLE]
we obtain
[TABLE]
Let ϕ:K[v±1,w±1]→K[v±1] be the map
given by ϕ(w)=1.
The kernel of the map ϕ is (w−1)K[v±1,w±1].
By (4), we have ϕ(S[y−1]0)=K[v].
Hence, we have
[TABLE]
In the same way, we know that
[TABLE]
is also isomorphic to a polynomial ring over K with one variable.
Hence, the curve V+(zu−xs3yt3) in PK(a,b,c)
is isomorphic to PK1.
Since the map C→V+(zu−xs3yt3) is a finite birational map,
C is also isomorphic to PK1.
∎
Lemma 3.3**.**
Let K be a field of prime characteristic p.
Let a, b, c be pairwise coprime positive integers.
We assume the conditions (ii) and (iii) in Theorem 1.1.
Then, there exist integers q1 and q2 such that
[TABLE]
for i>0, n≥q1u and m≥(ab/u)n+q2.
Proof.
First of all, remember that OX(mA−nE) is invertible if abc divides m (e.g., Lemma 1.3 in [14]).
By the condition (iii), zu−xs3yt3 is the negative curve in p.
Letting C be the proper transform of V+(zu−xs3yt3),
C is the negative curve that is linearly equivalent to cuA−E.
Therefore, mA−nE is a nef Cartier divisor
if m≥(ab/u)n≥0 and abc∣m.
Let m and n be integers such that m≥(ab/u)n≥0.
Then, there exists a nef Cartier divisor m1A−n1E such that
0≤m−m1<abc and 0≤n−n1<cu.
We use Fujita’s vanishing theorem (Theorem 1.4.35 in [13])
for finitely many coherent shaves
[TABLE]
Then, there exists an ample Cartier divisor q2A+q1(abA−uE) such that
[TABLE]
for i>0, 0≤m2<abc, 0≤n2<cu and any nef Cartier divisor m1A−n1E.
Then, q1 and q2 satisfy the requirement in Lemma 3.3.
∎
Lemma 3.4**.**
Let K be a field of prime characteristic p.
Let a, b, c be pairwise coprime positive integers.
We assume the conditions (ii) and (iii) in Theorem 1.1.
Then, there exist e>0 and η∈[p(peu)]peab such that
zu−xs3yt3 and η satisfy Huneke’s condition on p, that is,
[TABLE]
holds.
(The above integer e depends on a, b, c and p.)
Proof.
Let C be the proper transform of V+(zu−xs3yt3).
Then, C is the negative curve on X that is linearly equivalent to ucA−E.
Consider the reflexive sheaf OP(ab).
Since Sab contains both xb and ya,
OP(ab) is invertible away from the point V+(x,y).
Therefore OX(abA) is invertible away from the point
π−1(V+(x,y)).
Since C does not contain the point π−1(V+(x,y)),
OX(abA−uE)⊗OnC is
an invertible sheaf on nC for any n>0.
We choose integers q1 and q2 that satisfy Lemma 3.3.
By the condition (iii) in Theorem 1.1,
we obtain
[TABLE]
Let q be an integer which satisfies
[TABLE]
Consider the invertible sheaf OX(abA−uE)⊗OC.
Since (abcA−cuE).C=0, the degree of OX(abcA−cuE)⊗OC is [math].
Since C is isomorphic to PK1,
OX(abcA−cuE)⊗OC≃OC.
Therefore,
[TABLE]
since Pic(PK1)≃Z.
We have the exact sequences
[TABLE]
for ℓ=1,2,…,qu−1.
They induce the exact sequences 111Suppose that I is an ideal of a ring A with
I2=(0). Then, consider the map I→A× defined by a↦1+a.
It induces the exact sequence 0→I→A×→(A/I)×→1.
[TABLE]
and
[TABLE]
for ℓ=1,2,…,qu−1.
Therefore we know that the order of an element in the kernel of the map
Pic((ℓ+1)C)→Pic(ℓC) is 1 or p.
Hence, the order of OX(abA−uE)⊗OquC
(in Pic(quC)) is a power of p by (6).
Thus, there exists e>0 such that
Thus, there exists an effective Weil divisor D such that
D∼pe(abA−uE) and the support of D does not intersect with C.
Let η be the equation of π(D).
The degree of η is peab.
Since D∩C=∅, V+(zu−xs3yt3)∩V+(η)⊂V+(p) as a set.
Therefore, p is the only one minimal prime ideal of
(zu−xs3yt3,η).
Hence x, zu−xs3yt3, η form a regular sequence of S.
We obtain
Then, by Theorem 2.13, we know that
Rs(pQ(a,b,c)) is Noetherian if and only if
[TABLE]
Thus,
We have completed the proof of Theorem 1.1.
∎
Remark 3.5**.**
Let a, b, c be pairwise coprime positive integers.
Let ξ∈[pK(a,b,c)(k)]d be a negative curve with d/k<abc.
Then, Rs(pK(a,b,c)) is Noetherian if and only if
HC(pK(a,b,c);k,ξ)=∅.
Let ξ be a homogeneous element in SZ.
Then, ξQ∈[pQ(a,b,c)(k)]d is a negative curve with d/k<abc if and only if, for p≫0,
ξFp∈[pFp(a,b,c)(k)]d is a negative curve with d/k<abc.
Let ξ be a homogeneous element in SZ.
Assume that ξQ∈[pQ(a,b,c)(k)]d is a negative curve with d/k<abc, and
Rs(pQ(a,b,c)) is Noetherian.
Then, HC(pQ(a,b,c);k,ξQ)=HC(pFp(a,b,c);k,ξFp)
for p≫0.
Let ξ∈[pK(a,b,c)(k)]d be a negative curve with d/k<abc.
Assume that k=1 or 2.
Suppose that
[TABLE]
for some i.
Furthermore, assume that Rs(pK(a,b,c)) is Noetherian.
Then, there exists a positive integer m such that
In this section, we introduce a sufficient condition (which is called as “the condition EU” below)
for finite generation of Rs(p)
under the assumption in Theorem 1.1.
The condition EU was defined in Ebina [4] and Uchisawa [15].
We shall prove that, if u≤6, the condition EU is a necessary and sufficient condition
for finite generation of Rs(p) in Proposition 4.10.
Let us remember the method introduced in Gonzáles-Karu [5].
Let a, b, c be pairwise coprime positive integers and K be a field.
Let S=K[x,y,z] be a Z-graded ring with deg(x)=a, deg(y)=b and deg(z)=c.
Suppose that the prime ideal pK(a,b,c) is minimally generated by the three elements in (2).
We put
[TABLE]
Since pK(a,b,c) is generated by the three elements in (2),
we have
[TABLE]
Therefore, for each non-negative integer e, we have
[TABLE]
Let Δu be the domain (with boundary) surrounded by the following three lines
[TABLE]
Let (0,0), (u,u2), (δ1,δ2) be the vertices of Δu.
Here, δ1 and δ2 may not be integers.
[TABLE]
For a non-negative integer e, we put
[TABLE]
Then, it is easy to see that the euality (8) induces
[TABLE]
Since
[TABLE]
we have
[TABLE]
and
[TABLE]
for any n>0.
Therefore,
[TABLE]
Remark 4.1**.**
Let K be a field of characteristic [math].
Let φ(v,w) be an element in K[v±1,w±1].
Then, φ(v,w)∈(v−1,w−1)nK[v±1,w±1] if and only if
[TABLE]
for k+ℓ<n.
**
Remark 4.2**.**
Let a, b, c be pairwise coprime positive integers.
Assume the conditions (ii) and (iii) in Theorem 1.1.
Then, by Huneke’s condition on p, Rs(p) is Noetherian if and only if
[p(eu)]eab contains an element whose coefficient of yea is not [math]
for a sufficiently divisible e.
Assume the conditions (i), (ii) and (iii) in Theorem 1.1.
Then, by Theorem 1.1, Rs(p) is Noetherian
if and only if [p(u)]ab contains an element whose coefficient of ya is not [math].
By (9), it is equivalent to that
[TABLE]
contains an element whose constant term is not [math].
It is not so difficult to check whether it is satisfied or not using computers.
**
Now, we introduce the condition EU which is defined by Ebina [4] and Uchisawa [15].
Let a, b, c be pairwise coprime positive integers.
Suppose that the prime ideal p is minimally generated by the three elements in (2).
For i=1,2,…,u, we put
[TABLE]
We sort the sequence ℓ1, ℓ2, …, ℓu into assending order
[TABLE]
We say that the condition EU is satisfied for (a,b,c) if
[TABLE]
for i=1,2,…,u.
**
Example 4.4**.**
(I)
Assume (a,b,c)=(8,19,9).
Then,
[TABLE]
and the conditions (ii) and (iii) in Theorem 1.1 is satisfied.
[TABLE]
Then, u=3 and
[TABLE]
Therefore
[TABLE]
The condition EU is satisfied in this case.
(II)
Assume (a,b,c)=(25,29,72).
Then,
[TABLE]
and the conditions (ii) and (iii) in Theorem 1.1 is satisfied.
[TABLE]
Then, u=3 and
[TABLE]
Therefore
[TABLE]
The condition EU is not satisfied in this case.
(III)
Assume (a,b,c)=(17,503,169).
Then,
[TABLE]
and the conditions (ii) and (iii) in Theorem 1.1 is satisfied.
[TABLE]
Then, u=7 and
[TABLE]
Therefore
[TABLE]
The condition EU is not satisfied in this case.
In order to show that the condition EU is a sufficient condition for finite generation of Rs(p)
under some assumptions,
we need the following lemma ([4], [15]).
For the convenience of the reader, we give a proof of it here.
Let K be a field of characteristic [math] and v, w be variables.
Let u be a positive integer and α1, α2, …, αu be mutually distinct integers.
For i=1,2,…,u, consider the integers βi1, βi2, …, βii satisfying
[TABLE]
Put
[TABLE]
Then, we have
[TABLE]
Proof.
We shall prove it by induction on u.
If u=1, then #T=1.
It is easily verified in this case.
Assume u≥2.
Take
[TABLE]
Considering v−αuφ(v,w),
we may assume αu=0.
Then, ∂v∂φ satisfies all the assumptions with u−1.
Here, recall ∂v∂φ∈(v−1,w−1)u−1K[v±1,w±1] by Remark 4.1.
By induction, we obtain ∂v∂φ=0.
Therefore, we may suppose
[TABLE]
where C1,…,Cu∈K.
Since φ(v,w)∈(v−1,w−1)uK[v±1,w±1], we have
[TABLE]
for k=0,1,…,u−1.
Then, we have
[TABLE]
for k=0,1,…,u−1.
It is easy to see C1=C2=⋯=Cu=0.
∎
By this lemma, we can prove that the condition EU is a sufficient condition
for finite generation of Rs(p) under some assumptions.
By this equality, we know that [p(u)]ab is defined
by 2u(u+1) linear equations in
ya(⊕(α,β)∈Δu∩Z2Kvαwβ).
We put T′=T∪{(0,0)}.
Recall that T′⊂Δu∩Z2 and #T′=2u(u+1)+1.
Then, [p(u)]ab contains a non-zero element in the form
[TABLE]
where C(α,β)∈K.
If C(0,0)=0, we have
[TABLE]
by Lemma 4.5.
It is a contradiction.
Therefore, C(0,0)=0.
Then,
[TABLE]
holds.
Hence, Rs(p) is Noetherian by Huneke’s condition.
∎
The aim in the rest of this section is to prove the converse of Proposition 4.6 in the case u≤6.
Definition 4.7**.**
Let a, b, c be pairwise coprime positive integers.
Assume the condition (ii) in Theorem 1.1.
We define
[TABLE]
where [,] is the closed interval.
We say that the condition GK is satisfied if one of the following two conditions
is satisfied:
(I)
#{(n−1)[(u2/u),(t/t3)]∩Z}=n* and
(u2/u)n∈Z,*
(II)
#{(m−1)[−(s2/s3),(u2/u)]∩Z}=m* and
(u1/u)m∈Z.*
We remark that the above condition (I) is satisfied for a, b, c
if and only if the above condition (II) is satisfied for b, a, c.
Let a, b, c be pairwise coprime positive integers.
Assume the conditions (i), (ii), (iii) in Theorem 1.1.
If the condition GK is satisfied, then Rs(pK(a,b,c)) is not Noetherian
by Theorem 1.2 in Gonzáles-Karu [5].
Proposition 4.8**.**
Let a, b, c be pairwise coprime positive integers.
Assume the conditions (i), (ii), (iii) in Theorem 1.1.
Then, the condition GK is satisfied if and only if
one of the following five conditions is satisfied:
(GK1)
n=1,
(GK2)
m=1,
(GK3)
n=m=2<u,
(GK4)
3≤n<u, m=2 and #{(n−1)[(u2/u),(t/t3)]∩Z}=n,
(GK5)
n=2, 3≤m<u and #{(m−1)[−(s2/s3),(u2/u)]∩Z}=m.
Proof.
Let (δ1,δ2) be one of the vertices of Δu as in the beginning of this section.
First, we remark that, if 0≤i<i+1≤δ1, then ℓi+1≥ℓi+(n−1).
In the same way, if δ1≤i<i+1≤u, then ℓi≥ℓi+1+(m−1).
Thus, it is easy to see the following:
[TABLE]
[TABLE]
[TABLE]
Next, recall s=s2+s3, t=t1+t3 and u=u1+u2 by the condition (ii).
Then, we have
[TABLE]
Since a and b are coprime, u1, u2 and u are pairwise coprime.
Therefore, (u2/u)n∈Z if and only if n/u∈Z, and
(u3/u)m∈Z if and only if m/u∈Z.
It is easy to see that, if the condition (GKi) is satisfied for some i, then the condition GK is satisfied.
Conversely, assume that the condition GK is satisfied.
Then Rs(pK(a,b,c)) is not Noetherian
by Theorem 1.2 in Gonzáles-Karu [5].
By Theorem 4.6 and (10),
either n<3 or m<3 is satisfied.
If n=1 (resp. m=1), then (GK1) (resp. (GK2)) holds.
Suppose n=2 and m≥2.
Since the condition EU is not satisfied,
we have m<u by (11).
If (I) of the condition GK is satisfied, then n=m=2<u, and therefore
(GK3) is satisfied.
If (II) of the condition GK is satisfied, then (GK3) or (GK5) is satisfied.
Suppose n≥3 and m=2.
We know n<u by (12).
Then (II) is not satisfied.
If (I) is satisfied, then (GK4) is satisfied.
∎
Lemma 4.9**.**
Let a, b, c be pairwise coprime positive integers.
Assume the conditions (i), (ii), (iii) in Theorem 1.1.
Assume n=2 and 3≤m<u.
If either u1=1 or u2=1 is satisfied,
then either the condition GK or EU is satisfied.
2)
If n=2 and u>m≥(u+1)/2, then either the condition GK or EU is satisfied.
Proof.
First of all, remark that the condition EU is satisfied for a, b, c
if and only if so for b, a, c.
Furthermore, the condition GK is satisfied for a, b, c
if and only if so for b, a, c.
First, we shall prove 1).
Assume n=2, 3≤m<u and u2=1.
If (GK5) is not satisfied, then
[TABLE]
and
[TABLE]
since u2=1.
Thus, the condition EU is satisfied.
Next, assume n=2, 3≤m<u and u1=1.
Considering b, a, c, we may assume 3≤n<u, m=2 and u2=1.
Then, we have
[TABLE]
and
[TABLE]
In this case, the condition EU is always satisfied.
Here, we start to prove 2).
Assume that (GK5) is not satisfied.
Then, we have
[TABLE]
and
[TABLE]
Thus, the condition EU is satisfied.
∎
Proposition 4.10**.**
Let a, b, c be pairwise coprime positive integers.
Assume the conditions (i), (ii), (iii) in Theorem 1.1.
If u≤6, then the condition EU is a necessary and sufficient condition
for finite generation of Rs(pK(a,b,c)).
If u≤6, then the condition GK is a necessary and sufficient condition
for infinite generation of Rs(pK(a,b,c)).
Proof.
We shall prove that either the condition GK or EU is satisfied if u≤6.
If n=2 and m≥u, then then EU is satisfied by (11).
If n≥u and m=2, then then EU is satisfied by (12).
Now assume that n=2 and 3≤m<u.
If u>m≥(u+1)/2, then either the condition GK or EU is satisfied
by Lemma 4.9 2).
Assume 3≤m<(u+1)/2.
If u≤5, then such m does not exist.
Suppose u=6 and m=3.
Since u, u1, u2 are pairwise coprime, either u1 or u2 is 1.
Then, by Lemma 4.9 1), either the condition GK or EU is satisfied.
Assume 3≤n<u and m=2.
If u≤6, we can prove that either the condition GK or EU is satisfied
in the same way as above.
∎
Assume (a,b,c)=(8,19,9).
In this case, u=3, and the conditions (i), (ii) and (iii) in Theorem 1.1
are satisfied.
Since the conditin EU is satisfied, Rs(pK(a,b,c)) is Noetherian.
Assume (a,b,c)=(25,29,72).
In this case, u=3, and the conditions (i), (ii) and (iii) in Theorem 1.1
are satisfied.
Since the conditin EU is not satisfied, Rs(pK(a,b,c)) is not Noetherian.
Infinite generation of this ring was proved by Goto-Nishida-Watanabe [8].
Assume (a,b,c)=(17,503,169).
In this case, u=7, and the conditions (i), (ii) and (iii) in Theorem 1.1
are satisfied.
In this case, neither GK nor EU is satisfied.
Applying Theorem 1.1, we know that Rs(pK(a,b,c)) is not Noetherian by calculation using computers (see Remark 4.2).
**
Let a, b, c be pairwise coprime positive integers.
Assume the conditions (i), (ii), (iii) in Theorem 1.1.
We do not know any example of finitely generated Rs(pK(a,b,c))
such that the condition EU is not satisfied.
5. An example having negative curve in the second symbolic power
Let S=K[x,y,z], where K is a field and x,y,z are indeterminates.
We set m=(x,y,z)S and R=Sm.
In this section, we first take positive integers
s2,s3,t1,t3,u1,u2 arbitrarily, and set
[TABLE]
where s=s2+s3,t=t1+t3,u=u1+u2.
Moreover, we set
[TABLE]
Let us regard S as a Z-graded ring by setting
[TABLE]
Then, we can check directly that f,g,h are all homogeneous.
We set I=(f,g,h)S and a=IR.
Lemma 5.1**.**
We have the following relations;
(1)
yt3f+zu1g+xs2h=0,
(2)
zu2f+xs3g+yt3h=0.
Proof.
Since f,g,h are the maximal minors of the matrix
[TABLE]
we get the relations stated above.
∎
Lemma 5.2**.**
The following assertions hold.
(1)
(x)+I* is m-primary.*
(2)
AssSS/I=AsshSS/I.
(3)
Ip* is generated by 2 elements for any p∈AsshSS/I.*
(4)
ℓS(S/(x)+I(n))=(n(n+1)/2)⋅a*
for any 0<n∈Z.*
(5)
We have I⊆pK(a,b,c),
and the equality holds if GCD(a,b,c)=1.
Proof.
(1)
This holds as (x)+I contains x,yt and zu.
(2)
We get this assertion by Hilbert-Burch’s theorem.
(3)
Let us take any p∈AssSS/I.
Then, as x∈p, we have h∈(f,g)Sp by Lemma 5.1 (1),
so Ip=(f,g)Sp.
(4)
Since (x)+I=(x,yt,yt1zu1,zu),
we have exR(R/I)=ℓS(S/(x)+I))=a.
Let us take any 0<n∈Z.
Then,
[TABLE]
For any P∈AsshRR/a, G(aP) is isomorphic to a polynomial ring
with 2 variables over RP/aP, so
[TABLE]
Thus we get
[TABLE]
(5)
We set p=pK(a,b,c).
Since f,g,h are all homogeneous, we have I⊆p.
Hence, we have
[TABLE]
Now, we assume GCD(a,b,c)=1.
Then, as is well known, we have exR(R/pR)=a, so we see
[TABLE]
which means (x)+I=(x)+p.
Then, we have
[TABLE]
from which the equality p=I follows.
∎
Lemma 5.3**.**
Suppose s2>s3, t1=t3=1 and u1<u2.
Then, the following assertions hold.
(1)
There exists ξ∈I(2) such that
(i)
xs3ξ=zu2−u1f2−gh,
(ii)
zu1ξ=xs2−s3h2−fg, and
(iii)
ξ≡y3* mod (x).*
(2)
(x)+I(2)=(x,y3,y2z2u1,yzu+u1,z2u).
Proof.
(1)
¿From the relations (1) and (2) of Lemma 5.1, we get
[TABLE]
respectively.
Hence, we have
[TABLE]
so we get
[TABLE]
Since xs3,zu1 is a regular sequence on S,
there exists ξ∈S such that
[TABLE]
The first equality implies xs3ξ∈I2, so ξ∈I(2).
The second equality implies zu1ξ≡−fg mod (x),
so zu1ξ≡yzu1⋅y2 mod (x)
as f≡−yzu1 mod (x) and g≡y2 mod (x).
Hence, we get ξ≡y3 mod (x) since zu1 is regular on S/(x).
(2)
Since (x)+I=(x,y2,yzu1,zu) and u1<u2,
we have
[TABLE]
We set J=(ξ)+I2⊆I(2).
Then, as
[TABLE]
we have
[TABLE]
by Lemma 5.2 (4).
Hence we get the required assertion.
∎
Lemma 5.4**.**
Suppose s2>2s3, t1=t3=1 and u1<u2<2u1.
Then, the following assertions hold.
Proof.
(1)
¿From the relations (i) and (ii) of Lemma 5.3, we get
[TABLE]
respectively.
Hence, we have
[TABLE]
so we get
[TABLE]
Since xs3,zu2−u1 is a regular sequence on S,
there exists ζ∈S such that
[TABLE]
The first equality implies xs3ζ∈II(2)⊆I(3),
so ζ∈I(3) as xs3 is regular on S/I(3).
The second equality implies zu2−u1ζ≡fξ mod (x),
so zu2−u1ζ≡−yzu1⋅y3 mod (x) as
f≡−yzu1 mod (x) and ξ≡y3 mod (x).
Hence, we get ζ≡−y4z2u1−u2 mod (x)
since zu2−u1 is regular on S/(x).
as u1<u2<2u1.
We set J=(ζ)+II(2)⊆I(3).
Then, as
[TABLE]
we have
[TABLE]
by Lemma 5.2 (4).
Hence we get the required assertion.
(3)
It is enough to show
I(2)I(3)⊊I(5).
(We have (I(2))2+II(3)=I(4),
which can be verified in the same way.)
By Lemma 5.3 (2) and Lemma 5.4 (2), we have
in SFp.
We take 0≪ep∈Z and put q=pep.
Then, we have
[TABLE]
Here, we write q=3k+ℓ, where ℓ=0,1,2.
Then,
[TABLE]
The relation (i) of Lemma 5.4 means f3∈(xs3,z2u1−u2)IFp(3).
Hence, by (2), we have
[TABLE]
Thus we see
[TABLE]
so there exist σFp,τFp∈IFp(3q) such that
[TABLE]
Then, we have
[TABLE]
Since xq(s2−2s3),zq(u2−u1) is a regular sequence on SFp,
there exists ηFp∈SFp such that
[TABLE]
The first equality implies xq(s2−2s3)ηFp∈IFp(3q),
so we have ηFp∈IFp(3q) as xq(s2−s3) is regular on S/IFp(3q).
The second equality implies zq(u2−u1)ηFp≡(−1)q+1h3q mod xSFp,
so zq(u2−u1)ηFp≡(−1)q+1z3qu mod xSFp as
h≡zu mod xSFp.
Hence, we get ηFp≡(−1)z2q(u+u1) mod xSFp since
zq(u2−u1) is regular on SFp/xSFp.
Then, we have
[TABLE]
and hence 3q∈HC(IFp;2,ξFp).
(4)
If 3∈HC(IK;2,ξK), by Proposition 2.9 (3), we have
(5)
Let us notice that ξQ∈Z[x,y,z].
Then, setting k=2 and r=3 in Theorem 2.13,
we see that Rs(IQ) is not finitely generated.
∎
Example 5.6**.**
Let α=6/5 and β=49/24,
which satisfy the assumptions on α and β of Theorem 5.5.
We set
[TABLE]
where m,n are coprime positive integers such that m is odd
and n is not a multiple of 97.
Then, we have
[TABLE]
Since 683 and 97 are prime numbers, we get
GCD(a,b,c)=1.
Hence, by Lemma 5.2 (5) and Theorem 5.5 (5),
we see that Rs(pQ(a,b,c)) is infinitely generated.
If m=n=1,
then a,b,c are pairwise coprime,
and one can check directly that ξK is the negative curve
for any field K.
Bibliography15
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[1] L. Burch , Codimension and analytic spread , Proc. Camb. Phil. Soc. 72 (1972), 369–373.
2[2] S. D. Cutkosky , Symbolic algebras of monomial primes , J. reine angew. Math. 416 (1991), 71–89.
3[3] S D. Cutkosky and K. Kurano , Asymptotic regularity of powers of ideals of points in a weighted projective plane , Kyoto J. Math. 51 (2011), 25–45.
4[4] T. Ebina , Master theses, Meiji University 2017 (Japanese).
5[5] J. S. Gonzáles and K. Karu , Some non-finitely generated Cox rings , Compos. Math. 152 (2016), 984–996.
6[6] S. Goto, M. Herrmann, K. Nishida and O. Villamayor , On the structure of Noetherian symbolic Rees algebras , Manuscripta Math. 67 (1990), 197–225.
7[7] S. Goto, K. Nishida and Y. Shimoda , The Gorensteinness of symbolic Rees algebras for space curves , J. Math. Soc. Japan 43 (1991), 465–481.
8[8] S. Goto, K, Nishida and K.-i. Watanabe , Non-Cohen-Macaulay symbolic blow-ups for space monomial curves and counterexamples to Cowsik’s question , Proc. Amer. Math. Soc. 120 (1994), 383–392.