Stability properties of powers of ideals in regular local rings of small dimension
Jürgen Herzog and Amir Mafi
J. Herzog, Fachbereich Mathematik
Universität Duisburg-Essen, Campus Essen, 45117 Essen, Germany
[email protected]
A. Mafi, Department of Mathematics, University of Kurdistan, P.O. Box: 416, Sanandaj,
Iran.
[email protected]
Abstract.
Let (R,m) be a regular local ring or a polynomial ring over a field, and let I be an ideal of R which we assume to be graded if R is a
polynomial ring. Let astab(I) resp. astab(I) be the smallest integer n for which
Ass(In) resp. Ass(In) stabilize, and dstab(I) be the smallest integer n for which depth(In) stabilizes. Here In denotes the integral closure of In.
We show that astab(I)=astab(I)=dstab(I) if dimR≤2, while already in dimension 3, astab(I) and astab(I) may differ by any amount. Moreover, we show that if dimR=4, then there exist ideals I and J such that for any positive integer c one has astab(I)−dstab(I)≥c and dstab(J)−astab(J)≥c.
Key words and phrases:
Associated primes, depth stability number.
2010 Mathematics Subject Classification:
13A15, 13A30, 13C15.
Introduction
Let (R,m) be a commutative Noetherian ring and I be an ideal of R. Brodmann [3] proved that the set of associated prime ideals Ass(Ik)
stabilizes. In other words, there exists an integer k0 such that Ass(Ik)=Ass(Ik0) for all k≥k0. The smallest such integer k0 is
called the index of Ass-stability of I, and denoted by astab(I). Moreover, Ass(Ik0) is called the stable set of associated
prime ideals of I. It is denoted by Ass∞(I). For the integral closures Ik of the powers of I, McAdam and Eakin [13]
showed that Ass(Ik) stabilizes as well. We denote the index of stability for the integral closures of the powers of I by
astab(I), and denote its stable set of associated prime ideals by Ass∞(I).
Brodmann [4] also showed that depthR/Ik stabilizes. The smallest power of I for which depth stabilizes is denoted by dstab(I). This
stable depth is called the limit depth of I, and is denoted by limk→∞depthR/Ik. These indices of stability have been studied
and compared to some extend in [8] and [9]. The purpose of this work is to compare once again these stability indices. The main result is
that if (R,m) is a regular local ring with dimR≤2, then all 3 stability indices are equal, but if dimR=3, then
we still have astab(I)=dstab(I), while astab(I) and astab(I) may differ by any amount. On the other hand, if dimR≥4, we will
show by examples that in general a comparison between these stability indices is no longer possible. In other words, any inequality between these
invariants may occur.
Quite often, but not always, depth(R/Ik) is a non-increasing function on n. In the last section we prove that if (R,m) is a 3-dimensional
regular local ring and I satisfies Ik+1:I=Ik for all k, then depthR/Ik is non-increasing.
For any unexplained notion or terminology, we refer the reader to [5].
Several explicit example were performed with help of the computer algebra systems CoCoA [1] and Macaulay2 [6], as well as with the program in
[2] which allows to compute Ass∞(I) of a monomial ideal I.
1. The case dimR≤3
In this section we study the behavior of the stability indices for regular rings of dimension ≤3. In the proofs we will use the following elementary
and well-known fact: let I⊂R be an ideal of height 1 in the regular local ring R. Then there exists f∈R such that I=fJ where either
J=R or otherwise heightJ>1. Indeed, let I=(f1,…,fm). Since R is factorial, the greatest common divisor of f1,…,fm exists.
Let f=gcd(f1,…,fm), and gi=fi/f for i=1,…,m. Then I=fJ, where J=(g1,…,gm). Suppose heightJ=1, then there exist a prime
ideal P of height 1 with J⊂P. Since R is regular, P is a principal ideal, say P=(g). It follows then that g divides all gi, but
gcd(g1,…,gm)=1, a contradiction.
We first observe
Remark 1.1**.**
Let (R,m) be a regular local ring with dimR≤2 and let I be an ideal of R. Then
[TABLE]
Proof.
If dimR≤1, then either R is a field or a principal ideal domain, and the statement is trivial. Now suppose that dimR=2 and I=0. If
heightI=2, then m belongs to Ass(Ik) and Ass(Ik) for all k, and the assertion is trivial. Hence, we may assume that
height(I)=1. Then I=fJ with J=R or heightJ=2. In the first case I is a principal ideal, and the assertion is trivial. In the second case,
Ik=fkJk for all k, and Jk is m-primary. Thus there exists g∈Jk with gm∈Jk. Then gfk∈fkJk and gfkm∈fkJk. This shows that in the second case m∈Ass(Ik) for k, so that astab(I)=dstab(I)=1.
Finally observe that in the second case, Ik=fkJk for all k. This shows that m∈Ass(Ik) for all k, so
that also in this case astab(I)=1.
Theorem 1.2**.**
Let (R,m) be a regular local ring with dimR≤3 and I be an ideal of R. Then astab(I)=dstab(I).
Proof.
By Remark 1.1, we may assume that dimR=3. If heightI≥2, then Ass(Ik)⊆Min(I)∪{m} for all k. This implies at once that
astab(I)=dstab(I). Now suppose that heightI=1. If I is a principal ideal, then the assertion is again trivial. Otherwise, I=fJ with
heightJ≥2. Since Ik is isomorphic to Jk as an R-module, it follows that projdimIk=projdimJk for all k. This implies that
projdimR/Ik=projdimR/Jk for all k, and consequently depthR/Ik=depthR/Jk, by the Auslander–Buchsbaum formula. Thus,
dstab(I)=dstab(J).
We claim that astab(I)=astab(J). Since we have already seen that astab(J)=dstab(J) if heightJ≥2, the claim then implies that
astab(I)=dstab(I), as desired.
The claim follows once we have shown that Ass(Ik)=Ass(fkJk)=Min(f)∪Ass(Jk). For that we only need to prove the second equation. So let
P∈SpecR with fkJk⊂P. Then P∈Ass(fkJk) if and only if depthRP/fkJkRP=0. If J⊂P, then fkJkRP=fkRP, and
hence depthRP/fkJkRP=0 if and only if depthRP/fkRP=0, and this is the case if and only if P∈Min(f). If J⊂P, then the
RP-modules fkJkRP and JkRP are isomorphic, so that with the arguments as above depthRP/fkJkRP=depthRP/JkRP, which shows that in
this case P∈Ass(fkJk) if and only if P∈Ass(Jk). This completes the proof.
The statements shown so far and its proofs made for ideals in a regular local ring are valid as well for any graded ideal in a polynomial ring.
We now turn to some explicit examples. In [10, Proposition 1.5] Hibi et al show that for any integer t≥2 the ideal
I=(xt,xyt−2z,yt−1z)⊂K[x,y,z] satisfies dstab(I)=t. Since by Theorem 1.2, astab(I)=dstab(I), this example shows that in a
3-dimensional graded or local ring (we may pass to K[∣x,y,z∣]) both the index of depth stability as well as the index of Ass stability may be any
given number.
The following example shows that already for an ideal I in a 3-dimensional polynomial ring the invariants astab(I) and astab(I) may
differ.
Example 1.3**.**
Let R=K[x,y,z] be a polynomial ring over a field K and let I=((xy)2,(xz)2,(yz)2)⊂R. Then
astab(I)=2 and astab(I)=1.
Proof.
We first claim that In:(xy)2=In−1+z2n(x2,y2)n−2. Indeed, let J=((xz)2,(yz)2). Then In=Jn+(xy)2In−1, and hence In:(xy)2=Jn:(xy)2+In−1. Since Jn:(xy)2=z2n(x2,y2)n:(xy)2=z2n(x2,y2)n−2,
the assertion follows.
By symmetry, we also have In:(xz)2=In−1+y2n(x2,z2)n−2 and In:(yz)2=In−1+x2n(y2,z2)n−2. Thus, for all n≥1 we obtain
[TABLE]
In other words, I satisfies strong persistence in the sense of [8]. In particular, Ass(In)⊂Ass(In+1) for all n≥1. Now since
Ass(I)={(x,y),(x,z),(y,z)} and Ass(I2)={(x,y),(x,z),(y,z),(x,y,z)}, we deduce from this that astab(I)=dstab(I)=2.
With Macaulay2 it can be checked that I={(xy)2,(xz)2,(yz)2,xyz2,xy2z,x2yz} and that
Ass(I)={(x,y),(x,z),(y,z),(x,y,z)}. By [12, Corollary 11.28], one has Ass(I)⊂Ass(I2)⊂⋯⊂Ass∞(I). Since Ass(In) is a subset of the monomial prime ideals containing I, and since this set is
{(x,y),(x,z),(y,z),(x,y,z)}, we see that Ass(I)=Ass(In) for all n. Hence, astab(I)=1.
The next result says that the difference astab(I)−astab(I) may indeed be as big as we want.
Theorem 1.4**.**
Let R=k[x,y,z] be the polynomial ring over a field K, c a positive integer and I=(x2c+2,xy2cz,y2c+2z).
Then astab(I)=c+2 and astab(I)=2.
Proof.
Note that I=(x2c+2,z)∩(x,y2c+2)∩(y2c,x2c+2), from which it follows that dimR/I=depthR/I=1.
In the next step we prove that In:I=In−1 for all n. Then [8, Theorem 1.3] implies that Ass(In)⊆Ass(In+1) for all n. In
particular, if depth(R/Ik)=0 for some k, then depth(R/Ir)=0 for all r≥0. Since depthR/Ik≤1 for all k, it then follows that
depth(R/Ik)≥depth(R/Ik+1) for all k.
In order to show that In:I=In−1, observe that
[TABLE]
and
that
[TABLE]
Similarly we have
[TABLE]
Now since
[TABLE]
it follows In:I=In−1 for all n, as desired.
Next we claim that In:x2c+2=In−1 for all n≤c+1.
If n=1, there is nothing to prove. Let 1<n≤c+1.
By a calculation as before we see that
[TABLE]
We proceed by induction n to show that depthS/In=1 for n≤c+1. We observed already that depthS/I=1, Now let 1<n≤c+1. Then, since
In:x2c+2=In−1, we obtain the exact sequence
[TABLE]
Since by induction hypothesis depthR/In−1=1, it follows that
[TABLE]
Note that (In,x2c+2)=((y2cz(x,y2)n,x2c+2), which implies that depthR/(In,x2c+2)=1. Thus we have depthR/In≥1.
On the other hand, we have seen before that depthR/In≤depthR/I=1, and so depthR/In=1 for all n≤c+1.
In the next step we show that depthR/Ic+2=0, which then implies that depthR/In=0 for all n≥c+2. In particular, it will follows that
astab(I)=c+2.
In order to prove that depthR/Ic+2=0, we show that x2c+2y(c+1)(2c+2)−1zc+1∈(Ic+2:m)∖Ic+2.
Indeed, let u=x2c+2y(c+1)(2c+2)−1zc+1. Then
[TABLE]
and
[TABLE]
This shows that u∈(Ic+2:m).
Assume that x2c+2y(c+1)(2c+2)−1zc+1∈Ic+2. Then
[TABLE]
and so y(c+1)(2c+2)−1zc+1∈Ic+1. Since Ic+1=(x2c+2,y2cz(x,y2))c+1, expansion of this power implies that
y(c+1)(2c+2)−1∈∑i=0c+1(x2c+2)i(y2c(x,y2))c+1−i. It follow that y(c+1)(2c+2)−1∈(y2c(x,y2))c+1, which is a
contradiction.
Now we compute astab(I), and first prove that
[TABLE]
Let J=(I,(x3y2c−1z,x4y2c−2z,…,x2c+1yz)). For all integers i with 3≤i≤2c+1, we have
[TABLE]
Thus J⊆I. We have Ass(I/J)⊆Ass(J). The primary decomposition of J shows that Ass(J)={(x,z),(x,y)}. Let
P=(x,z). Then (I)P=(IP)=(x2c+2,z)P=(x2c+2,z)P. The last equality follows by [11, Proposition 1.3.5],
and so (I/J)P=0. Hence
P∈/Ass(I/J). Now let P=(x,y). Then
[TABLE]
The second equality follows by [11, Exercise 1.19]. Thus we have (I/J)P=0. This shows that Ass(I/J)=∅, and hence
I=J, as desired. In particular we see that
[TABLE]
Since Ass(I)⊆Ass(Ik) for all k, it follows that {(x,z),(x,y)}⊂Ass(Ik) for all k. Suppose that
(y,z)∈Ass(Ik) for some k. Then (y,z) is a minimal prime ideal of I. However, this is not the case, as can be seen from the primary
decomposition of I.
Next we show that m=(x,y,z) belongs to Ass(I2). Then it follows that
[TABLE]
thereby showing that astab(I)=2.
In order to prove that m∈Ass(I2), we first show that the ideal
[TABLE]
is contained in I2.
Since
[TABLE]
it follows that for all integers i with 4≤i≤2c+2 the element
[TABLE]
belongs to (I2)4c. Also, for all integers i with 5≤i≤2c−2, the element
[TABLE]
belongs to (I2)4c, This shows L⊆I2.
By using primary decomposition for the ideal L, we see that
[TABLE]
On the other hand, by easy calculation, one verifies that
L:(x2c+2y2c+1z)=m. Finally we show that x2c+2y2c+1z∈/I2, which then implies that m∈Ass(I2), as desired.
In order to prove this we show by induction on n that (x2c+2y2c+1z)n∈/(I2)n for all n.
Indeed,
let n=1, and assume that x2c+2y2c+1z∈I2. Then y2c+1z∈I2:x2c+2=I+(y2cz)2=I, and this is
contradiction.
Now let n>1. By induction hypothesis we may assume that (x2c+2y2c+1z)n−1∈/(I2)n−1.
If (x2c+2y2c+1z)n∈(I2)n, then x(2c+2)(n−1)(y2c+1z)n∈(I2n:x2c+2)=I2n−1+(y2cz)2n(x,y2)2n−2(c+1), and so
x(2c+2)(n−1)(y2c+1z)n∈I2n−1.
It follows that x(2c+2)(n−1)(y2c+1z)n−1∈(I2n−1:y2c+1z). Since
[TABLE]
we see that
x(2c+2)(n−1)(y2c+1z)n−1∈y(I2)n−1, a contradiction.
Thus (x2c+2y2c+1z)n∈/(I2)n for all
n, as desired.
The theorem says that for any positive integer c there exists a monomial ideal in K[x,y,z] with astab(I)−astab(I)=c. However we do not
know whether for all ideals in I⊂K[x,y,z] one has astab(I)≤astab(I).
2. The case dimR>3
The purpose of this section is to show that for a polynomial ring S in more than 3 variables, for a graded ideal I⊂S the invariants astab(I) and dstab(I) may differ
by any amount.
We begin with two examples.
Example 2.1**.**
Let R=k[x,y,z,u] be the polynomial ring over a field k and consider the ideal I=(xy,yz,zu)
of R. Then astab(I)=1 and dstab(I)=2.
Proof.
We have Ass(I)=Min(I), and since I may be viewed as the edge ideal of a bipartite graph it follows from [7, Definition 1.4.5, Corollary
10.3.17] that Ass(I)=Ass(In) for all n∈N. Therefore astab(I)=1. By [7, Corollary 10.3.18],
limk→∞depthR/Ik=1. Moreover, it can be seen that depthR/I=2 and depthR/I2=1. Since I has a linear resolution,
[7, Theorem 10.2.6] implies that for all k≥1, Ik has a linear resolution as well. Therefore, by [9, Proposition 2.2] we have
depthR/Ik+1≤depthR/Ik for all k∈N. Hence depthR/Ik=1 for all k≥2, and so dstab(I)=2.
Example 2.2**.**
Let R=K[x,y,z,u] be the polynomial ring in 4 variables over a field K, and let I=(x2z,uyz,u3). Then astab(I)=2 and dstab(I)=1.
Proof.
Set J=(uyz,u3) and so for all n∈N, it therefore follows that In:x2z=(Jn+x2zIn−1):x2z=In−1+(Jn:x2z)=In−1. Hence, for all
n∈N, Ass(In)⊆Ass(In+1). By using Macaulay2 [6] and the program [2], we see that
Ass∞(I)=Ass(I2)={(x,u),(z,u),(x,y,u),(x,z,u)}. Therefore astab(I)=2. As Ass(In)⊆Ass(In+1) for all n∈N, it follows that
m=(x,y,z,u)∈/Ass(In) and so we have depthR/In≥1. y−z∈m is a nonzerodivisor on R/In for all n∈N. Set
R=R/(y−z). Thus by [5, Lemma 4.2.16] we have (R/In)=R/In≅K[x,z,u]/(x2z,uz2,u3)n.
Since xzu3n−1∈(In):m∖In, it follows depthR/In=0 and so depthR/In=1 for all n∈N. Therefore
dstab(I)=1.
Now we come to the main result of this section
Theorem 2.3**.**
Let R=k[x,y,z,u] be the polynomial ring over a field k. Then for any non-negative integer c, there exist two ideals I and J of R such that the
following statements hold:
astab(I)−dstab(I)≥c.
dstab(J)−astab(J)≥c.
Proof.
We may assume that c is a positive integer. Let I=(xc+1zc,u2c−1yz,u2c+1) and J=(xcyc−1,yc−1xc−1z,zcuc). We claim that
astab(I)=dstab(J)=c+1 and astab(J)=dstab(I)=1.
(i) In this case, by using Example 2.2, we can assume that c≥2. For all n∈N, we have
(In:xc+1zc)=(((u2c−1yz,u2c+1)n+xc+1zcIn−1):xc+1zc)=In−1+((u2c−1yz,u2c+1)n:xc+1zc). Since
((u2c−1yz,u2c+1)n:xc+1zc)=((u2c−1yz,u2c+1)n:zc)⊆In−1, it follows (In:xc+1zc)=In−1 and so
Ass(In)⊆Ass(In+1).
By using Macaulay2 [6] and [2], we have Ass(I)={(x,u),(z,u),(y,z,u),(x,y,u)}
and Ass∞(I)={(x,u),(z,u),(y,z,u),(x,z,u),(x,y,u)}. Set p=(x,z,u). It is easily seen that Ii:p=Ii for all i≤c and
xcyc+1zcu(2c+1)c∈(Ipc+1:p)∖Ipc+1. Hence Ass(I)=Ass(I2)=...=Ass(Ic),
Ass(Ic+1)=Ass∞(I) and so astab(I)=c+1. By the same argument as used in the proof of Example 2.2, that
m=(x,y,z,u)∈/Ass(In) for all n∈N and so we have depthR/In≥1 and x−y−z∈m is a nonzerodivisor on R/In for all n∈N. Therefore
(R/In)=R/In≅K[y,z,u]/((y+z)c+1zc,u2c−1yz,u2c+1)n, where R=R/(x−y−z). Since
z2cu(2c+1)n−1∈(In):m∖In, it follows depthR/In=0 and so depthR/In=1 for all n∈N. Therefore
dstab(I)=1.
(ii) For all n∈N, we have
(Jn:zcuc)=(((xcyc−1,yc−1xc−1z)n+zcucJn−1):zcuc)=Jn−1+((xcyc−1,yc−1xc−1z)n:zcuc)=Jn−1+((xcyc−1,yc−1xc−1z)n:zc).
Since ((xcyc−1,yc−1xc−1z)n:zc)⊆Jn−1, for all n∈N we have (Jn:zcuc)=Jn−1. Therefore, for all
n∈N, Ass(Jn)⊆Ass(Jn+1). By using [6] and [2] we have
Ass∞(J)={(x,z),(x,u),(y,z),(y,u)}=Min(J) and so astab(J)=1. Since m∈/Ass(Jn) for all n∈N, we have 2=dimR/J≥depthR/Jn≥1 and x−y∈m is a nonzerodivisor on R/Jn for all n∈N. Again by the above argument,
(R/Jn)=R/Jn≅K[x,z,u]/(x2c−1,x2c−2z,zcuc)n, where R=R/(x−y). Since
Ji:m=Ji for all i≤c and x(2c−1)nzn−1uc−1∈Jn:m∖Jn for
all n≥c+1.
It therefore follows depthR/J=depthR/J2=...=depthR/Jc=2 and depthR/Jn=1 for all n≥c+1. Hence dstab(J)=c+1.
3. Non-increasing depth functions
Theorem 3.1**.**
Let (R,m) be a regular local ring with dimR=3 and I be an ideal of R. If In+1:I=In for all n∈N, then depthR/In is
non-increasing.
Proof.
Suppose height(I)≥2. Since In+1:I=In for all n∈N, it follows that depthR/In+1≤depthR/In.
Now, let height(I)=1. Then there exists an ideal J of R and an element f∈R such that I=fJ and height(J)≥2. As in the proof of
Theorem 1.2, depthR/In=depthR/Jn for all n∈N. Since In+1:I=In for all n∈N, we have Jn+1:J=Jn. Thus
depthR/Jn+1≤depthR/Jn and so depthR/In+1≤depthR/In. This completes the proof.
Corollary 3.2**.**
Let (R,m) be regular local ring with dimR=3. Then depthR/In is non-increasing.
Let R=k[x,y,z] be a polynomial ring in 3 indeterminates over a field k. If I is an edge ideal of R, then depthR/In is non-increasing.
Example 3.3**.**
Let R=k[x,y,z,u] be a polynomial ring and I=(xy2z,yz2u,zu2(x+y+z+u),xu(x+y+z+u)2,x2y(x+y+z+u))
be an ideal of R. Then depthR/I=depthR/I4=0, depthR/I2=depthR/I3=1. Thus the depth function is neither non-increasing nor
non-decreasing.**
In view of Theorem 3.1 one may ask whether in a regular local ring (of any dimension), depthR/In is a non-increasing function of n, if In+1:I=In for all n.
Acknowledgements
The second author would like to thank Universität Duisburg-Essen, especially to the Department
of Mathematics for its hospitality during the preparation of this work.