Journal of Algebra 252 (2002) 161–166

www.academicpress.com

An example of an infinite set of associated

primes of a local cohomology module
Mordechai Katzman
Department of Pure Mathematics, University of Sheffield, Hicks Building,
Sheffield S3 7RH, UK
Received 14 November 2001
Communicated by Craig Huneke

Keywords: Graded commutative Noetherian ring; Graded local cohomology module; Infinite set of
associated primes

0. Introduction

Let (R, m) be a local Noetherian ring, let I ⊂ R be any ideal and let M
be a finitely generated R-module. It has been long conjectured that the local
cohomology modules HIi (M) have finitely many associated primes for all i (see
[1, Conjecture 5.1] and [2]).
If R is not required to be local these sets of associated primes may be infinite,
as shown by Anurag Singh in [3], where he constructed an example of a local
cohomology module of a finitely generated module over a finitely generated
Z-algebra with infinitely many associated primes. This local cohomology module
has p-torsion for all primes p ∈ Z.
However, the question of the finiteness of the set of associated primes of
local cohomology modules defined over local rings and over k-algebras (where
k is a field) has remained open until now. In this paper I settle this question by
constructing a local cohomology module of a local finitely generated k-algebra
with an infinite set of associated primes, and I do this for any field k.

E-mail address: [email protected].

0021-8693/02/$ – see front matter  2002 Elsevier Science (USA). All rights reserved.
PII: S 0 0 2 1 - 8 6 9 3 ( 0 2 ) 0 0 0 3 2 - 7
162 M. Katzman / Journal of Algebra 252 (2002) 161–166

1. The example

Let k be any field, R0 = k[x, y, s, t] and S = R0 [u, v]. Define a grading on S

by declaring deg(x) = deg(y) = deg(s) = deg(t) = 0 and deg(u) = deg(v) = 1.
Let f = sx 2 v 2 − (t + s)xyuv + ty 2 u2 and R = S/f S. Notice that f is
homogeneous and hence R is graded. Let S+ be the ideal of S generated by u
and v and let R+ be the ideal of R generated by the images of u and v.
Consider the local cohomology module HR2+ (R): it is homogeneously isomor-
phic to HS2+ (S/f S) and we can use the exact sequence

f
HS2+ (S)(−2) −→ HS2+ (S) −→ HS2+ (S/f S) −→ 0

of graded R-modules and homogeneous homomorphisms (induced from the exact

sequence

f
0 −→ S(−2) −→ S −→ S/f S −→ 0)

to study HR2+ (R). Furthermore, we can realize HS2+ (S) as the module R0 [u− , v − ]
of inverse polynomials described in [4, 12.4.1]: this graded S-module vanishes
beyond degree −2, and, for each d
2, its (−d)th component is a free
R0 -module of rank d − 1 with base (u−α v −β )α,β>0,α+β=−d . We will study
the graded components of HS2+ (S/f S) by considering the cokernels of the
R0 -homomorphisms

f−d : R0 [u− , v − ]−d−2 −→ R0 [u− , v − ]−d (d
2)

given by multiplication by f . In order to represent these R0 -homomorphisms

between free R0 -modules by matrices, we specify an ordering for each of the
above-mentioned bases by declaring that

uα1 v β1 < uα2 v β2

(where α1 , β1 , α2 , β2 < 0 and α1 + β1 = α2 + β2 ) precisely when α1 > α2 . If we

use this ordering for both the source and target of each fd , we can see that each
fd (d
2) is given by multiplication on the left by the tridiagonal d − 1 by d + 1
matrix
 sx 2 −xy(t + s) ty 2 0 ... 0 
 0 sx 2 −xy(t + s) ty 2 0... 0 
 
Ad−1 := 

0 0 sx 2 −xy(t + s) ty 2 . . . 0 .

 ..
. 
0 ... sx 2 −xy(t + s) ty 2
 M. Katzman / Journal of Algebra 252 (2002) 161–166 163

We also define
s −(t + s) t 0 ... 0
0 s −(t + s) t 0... 0
0 −(t + s) 0
Ad−1 := 

0 s t ... 

 .. 
.
0 ... s −(t + s) t
obtained by substituting x = y = 1 in Ad−1 .
Let also τi = (−1)i (t i + st i−1 + · · · + s i−1 t + s i ).

Lemma 1.1.
(i) Let Bi be the submatrix of Ai obtained by deleting its first and last columns.
Then det Bi = τi for all i
1.
(ii) Let S be an infinite set of positive integers. Suppose that either k has
characteristic zero or that k has prime characteristic p and S contains
infinitely many integers of the form pm − 2. The (k[s, t])-irreducible factors
of {τi }i∈S form an infinite set.

Proof. We prove the first statement by induction on i. Since

det B1 = det(−t − s) = −t − s and

 
−t − s t
det B2 = det = t 2 + st + s 2 ,
s −t − s
the lemma holds for i = 1 and i = 2. Assume now that i
3. Expanding the
determinant of Bi by its first row and applying the induction hypothesis, we obtain

det Bi = (−t − s) det Bi−1 − st det Bi−2

= (−1)i−1 (−t − s) t i−1 + · · · + s i−2 t + s i−1

− (−1)i−2 st t i−2 + · · · + s i−3 t + s i−2



= (−1)i t i + · · · + s i−2 t 2 + s i−1 t + st i−1 + · · · + s i−1 t + s i


− st i−1 + · · · + s i−2 t 2 + s i−1 t

= (−1)i t i + st i−1 + · · · + s i−1 t + s i .

We now prove the second statement. Define σi = t i + t i−1 + · · · + t + 1 and

notice that it is enough to show that the set of irreducible factors of {τi }i∈S is
infinite. Let I be the set of irreducible factors of {τi }i∈S . If k has characteristic
zero consider Q[I ] ⊇ Q, the splitting field of this set of irreducible factors. If I
is finite, Q[I ] ⊇ Q is finite extension which contains all ith roots of unity for all
i ∈ S, which is impossible.
164 M. Katzman / Journal of Algebra 252 (2002) 161–166

Assume now that k has prime characteristic p. Let F be the algebraic closure
of the prime field of k. For any positive integer m,
d pm −1

t t − 1 = −1;
dt
so τpm −2 = (t p −1 − 1)/(t − 1) has pm − 2 distinct roots in F and, therefore, the
m

roots of {τs }s∈S form an infinite set. ✷

Theorem 1.2. For every d
2 the R0 -module HR2+ (R)−d has τd−1 -torsion.
Hence HR2+ (R) has infinitely many associated primes.

Proof. For the purpose of this proof we introduce a bigrading in R0 by declaring

deg(x) = (1, 0), deg(y) = (1, 1) and deg(t) = deg(s) = (0, 0).
We also introduce a bigrading on the free R0 -modules R0n by declaring
deg(x α y β s a t b ej ) = (α + β, β + j ) for all non-negative integers α, β, a, b and
all 1  j  n. Notice that R0n is a bigraded R0 -module when R0 is equipped with
the bigrading mentioned above.
Consider the R0 -module Coker Ad−1 ; the columns of Ad−1 are bihomoge-
neous of bidegrees
(2, 1), (2, 2), . . . , (2, d + 1).
We can now consider Coker Ad−1 as a k[s, t] module generated by the natural
images of x α y β ej for all non-negative integers α, β and all 1  j  d − 1. The
k[s, t]-module of relations among these generators is generated by k[x, y]-linear
combinations of the columns of Ad−1 , and since these columns are bigraded, the
k[s, t]-module of relations will be bihomogeneous and we can write

Coker Ad−1 = (Coker Ad−1 )(D,j ) .
0
D, 1
j

Consider the k[s, t]-module (Coker Ad−1 )(d,d) the bihomogeneous component of
Coker Ad−1 of bidegree (d, d). It is generated by the images of
xy d−1e1 , x 2 y d−2 e2 , . . . , x d−2 y 2 ed−2 , x d−1 yed−1
and the relations among these generators are given by k[s, t]-linear combinations
of
y d−2 c2 , xy d−3c3 , . . . , x d−3 ycd−1, x d−2 cd
where c1 , . . . , cd+1 are the columns of Ad−1 . So we have
(Coker Ad−1)(d,d) = Coker Bd−1
where Bd−1 is viewed as a k[s, t]-homomorphism k[s, t]d−1 → k[s, t]d−1 .
Using Lemma 1.1(i) we deduce that for all d
2 the direct summand
(Coker Ad−1 )(d,d) of Coker Ad−1 has τd−1 torsion, and so does Coker Ad−1 itself.
 M. Katzman / Journal of Algebra 252 (2002) 161–166 165

Lemma 1.1(ii) applied with S = N now shows that there exist infinitely
many irreducible homogeneous polynomials {pi ∈ k[s, t]: i
1} each one of
them contained in some associated prime of the R0 -module d2 Coker Ad−1 .
Clearly, if i = j then any prime ideal P ⊂ R0 which contains both pi and pj must
contain both s and t.
Since the localisation of (Coker Ad−1 )(d,d) at s does not vanish, there exist
Pi , Pj ∈ AssR0 Coker Ad−1 which do not contain s and such that pi ⊂ Pi ,
pj ⊂ Pj , and the previous paragraph shows that Pi = Pj .
The second statement now follows from the fact that HR2+ (R) is R0 -isomorphic

to d2 Coker Ad−1 . ✷

Corollary 1.3. Let T be the localisation of R at the irrelevant maximal ideal

m = s, t, x, y, u, v. Then H(u,v)T
2 (T ) has infinitely many associated primes.

Proof. Since τi ∈ m for all i
1, H(u,v)T

2 (T ) ∼
= (H(u,v)R
2 (R))m has τi -torsion for
all i
1. ✷

2. A connection with associated primes of Frobenius powers

In this section we apply a technique similar to the one used in Section 1 to give
a proof of a slightly more general statement of Theorem 12 in [5]. The new proof
is simpler, open to generalisations and it gives a connection between associated
primes of Frobenius powers of ideals and of local cohomology modules, at least
on a purely formal level.
Let k be any field, let S = k[x, y, s, t], let F = xy(x − y)(sx − ty) = sx 3 y −
(t + s)x 2 y 2 + txy 3 and let R = S/F S.

Theorem 2.1. Let S be an infinite set positive integers and suppose that either k
has characteristic zero or that k has characteristic p and that S contains infinitely
many powers of p. The set
  
R
AssR
x n , y n 
n∈S

is infinite.

Proof. We introduce a grading in S by setting deg(x) = deg(y) = 1 and deg(s) =

deg(t) = 0. Since F is homogeneous, R is also graded.
Fix some n > 0 and consider the graded R-module T = R/x n , y n . For each
d > 4 consider Td , the degree d homogeneous component of T , as a k[s, t]-
module. If d < n, Td is generated by the images of y d , xy d−1, . . . , x d−1 y, x d
and the relations among these generators are obtained from y d−4 F, xy d−5F, . . . ,
166 M. Katzman / Journal of Algebra 252 (2002) 161–166

x d−5yF, x d−4 F . Using these generators and relations, in the given order, we
write Td = Coker Md where
 
0 0 ... 0
 t 
 
 −t − s t 
 
 s −t − s 
 
 .. 
Md =   s . .

 t 
 
 −t − s t 
 −t − s 
 s 
 s 
0 0 ... 0
When d = n, Td is isomorphic to the cokernel of the submatrix of Md obtained by
deleting the first and last rows which correspond to the generators y n , x n of Tn .
When d = n + 1, Td is isomorphic to the cokernel of the submatrix of Md
obtained by deleting the first two rows and last two rows which correspond to the
generators y n+1 , xy n , x n y, x n+1 of Tn+1 , and the resulting submatrix is Bn−2
defined in Lemma 1.1; the result now follows from that lemma. ✷

This technique for finding associated primes of non-finitely generated graded

modules and of sequences of graded modules has been applied in [6,7] to yield
further new and surprising properties of top local cohomology modules.

Acknowledgments

I thank Rodney Sharp, Anurag Singh, and Gennady Lyubeznik for reading a
first draft of this paper and for their helpful suggestions.

References

[1] C. Huneke, Problems on local cohomology, in: Free Resolutions in Commutative Algebra and
Algebraic Geometry, Sundance, UT, 1990, Res. Notes Math., Vol. 2, Jones and Bartlett, Boston,
MA, 1992, pp. 93–108.
[2] G. Lyubeznik, A partial survey of local cohomology, in: G. Lyubeznik (Ed.), Local Cohomology
and Its Applications, Marcel Dekker, 2001.
[3] A.K. Singh, p-torsion elements in local cohomology modules, Math. Res. Lett. 7 (2000) 165–176.
[4] M.P. Brodmann, R.Y. Sharp, Local Cohomology: An Algebraic Introduction with Geometric
Applications, Cambridge University
 Press, 1998.
[5] M. Katzman, Finiteness of e Ass F e (M) and its connections to tight closure, Illinois J. Math. 40
(1996) 330–337.
[6] M. Brodmann, M. Katzman, R.Y. Sharp, Associated primes of graded components of local
cohomology modules, Trans. Amer. Math. Soc., to appear.
[7] M. Katzman, R.Y. Sharp, Some properties of top graded local cohomology modules, preprint.