JOURNAL OF ALGEBRA 143, 156172 (1991)

On the Cohen-Macaulayness of
the Fiber Cone of an Ideal
KISHOR SHAH

Deparlment of Mathemarics and Computer Science, Utiiversity qf California,

Riverside, California 92521

Communicated by Melvin Hochster

Received February 27, 1989

Let R always be a commutative, associative, Noetherian local ring con-

taining an identity and maximal ideal M. Let Z be an ideal in R. Define
F(Z) = @ Sm=O (I”/Z”M) with I0 = R to be the fiber cone of Z or simply the
fiber cone. Define f(s) = length of Z”/Z”M to be the .liber function of ideal
Z or simply the fiber function.
F(Z) is evidently a basic geometric object but it has scarcely been studied
algebraically for its own sake. f(s) on the other hand is nothing but the
Hilbert function of F(Z). Its study would serve to funnel insight into the
classical Hilbert polynomial as well as into the Hilbert coefficients. In this
paper we propose to begin developing the theory of F(Z) andf(s) for their
own sake by answering to some extent the following questions:
(1) When is F(Z) Cohen-Macaulay?
(2) How does Cohen-Macaulayness affect F(Z) and G(Z) =
@ Fzo (II/Z”+ ‘) with I0 = R?
(3) How does Cohen-Macaulayness of F(Z) affect f(s)?
(4) How does normal flatness along Z relate with Cohen-Macaulay-
nessof F(Z)?
Consider (1). If Z is generated by a regular sequencethen F(Z) is a poly-
nomial ring over the field R/M. If Z is integral over a regular sequence x
then Z” = I”- Ix for some s and F(Z) is a finitely generated module over
F(x) in particular. Thus F(Z) is Cohen-Macaulay iff Z’(Z) is free over F(x)
by Hironaka’s freeness lemma [13, 25.161. We show that F(Z) is indeed
Cohen-Macaulay if Z* = I& (Theorem 1). We also study the case I3 = Z*&
(Theorem 2).
Consider (2). *This is a natural but difficult question to answer. We
obtain some insight into the Cohen-Macaulay phenomenon of F(Z) and
G(Z), however: If Z is M-primary then, using a new formula for multiplicity
156
0021-8693/91 53.00
Copyright 0 1991 by Academx Press, Inc.
All rights of reproduction in any form reserved
 COHEN-MACAULAYNESSOFFIBER CONEOFIDEALS 157

of G(Z) (Theorem 3) we deduce the Cohen-Macaulayness of G(Z) by

showing that it depends on whether @,“=, (xl”-’ @R/Z) canonically
embeds into G(Z) or not (Theorem 4). The Cohen-Macaulayness of F(Z)
for any Z depends but only partially on whether @,“=O (xZs-’ @R/M)
canonically embeds into F(Z) or not (Theorem 5). Thus we infer that
@ FzO (xl”+ ’ 0 R/Z) to quite an extent controls the Cohen-Macaulayness
of both G(Z) and F(Z) (Corollary 5(b)).
Consider (3). If F(Z) is CohenMacaulay then we calculate the fiber-
functionf(s); (Theorem 6). This contrasts well with the difficulty of under-
standing the classical Hilbert polynomial and the Hilbert coefficients. We
record f(s) in the basic case of Z2= Ix (Theorem 7).
Consider (4). The notion of normal flatness was introduced by Hironaka
[3]. If R is regular then R is normally flat along Z iff Z is a complete inter-
section.,If R is CohenMacaulay and Z is M-primary then we relate norm.al
flatness of Z with the multiplicities of F(Z), G(Z) in the case Z2=I&
(Theorem 8). We conclude with a startling corollary (8(a)) that if the mul-
tiplicity of Z is prime then normal flatness along Z is equivalent to Z being
a complete intersection or Z= M.
We deduce many classical and well-known results from our theorems. In
particular, Sally’s famous results on the Cohen-Macaulayness of the tangent
cone, [ 16, Theorem 2; 17, Theorem 2.11, are deduced several different
times. Abhyankar’s classical inequality [ 1, Theorem 1] is also deduced. We
also raise many questions and conjectures concerning the fiber cone F(Z).
Our general references for the paper are [4, 11, 14, 151.

SOME PROPERTIES OF THE FIBER CONE F(Z)

We shall list some relevant properties of F(Z) which serve also to reveal
its nature. The statements are either referenced or left as easy exercises.
Now, let x be any minimal reduction of ideal I. By ~(1) we mean the mini-
mum number of generators for Z. We say that Z is equimultiple if dim F(Z)
equals the height of Z. We shall always assume throughout that R/M is
infinite.
(1) If x form a regular sequencethen F(s) is a polynomial ring over
R/M in &x) variables.
(2) Z is generated by x iff F(Z) is a polynomial ring over R/M in p(Z)
variables.
(3) F(Z) is Cohen-Macaulay need not imply R is Cohen-Macaulay.
(4) F(Z) is integral over F(X).
(5) Suppose Z is integral over J with F(J) a domain. Then F(J) is a
subring of F(Z).
158 KISHOR SHAH

Proo$ Let t be transcendental over R. Set A = R[Zt] and B= R[Jt].

Then F(Z) = A/MA. Now B c A implies that B/Bn MA c A/MA. Since
F(J) is a domain we obtain that B/Bn MA = B/MB = F(J) This means
that Z”M n J” = J”M, V’s3 1. Indeed we shall be using this fact often when
J equals a minimum reduction of I.
(6) F(Z) is CohenMacaulay iff F(Z) is free over F(x) (apply [13,
25.161).
(7) Assume that R is an integrally closed domain and that all powers
of Z are integrally closed. Then F(Z), also integrally closed implies that the
ideal (x, Z”M) is integrally closed for all x E I”. (In particular, we may apply
this to the case when R is regular and Z= M to deduce a new class of
integrally closed ideals.)
(8) Assume R is Cohen-Macaulay containing a system of parameters
Xl, x2, .... Xd. Assume Z is generated by monomials m, , m,, .... m, in x, ,
x2 > .“, xd. Assume R contains a field. Set E to be the n x d index matrix
obtained naturally by writing the exponents of every monomial mi as rows.
Then
(a) dim F(Z) d rank E
(b) dim F(Z) = rank E if all the monomials are of equal degree. (see
[ 19, Chap. 33).
(9) Assume R to be any quasi-unmixed ring containing a field
instead of Cohen-Macaulay, in (8) above. Then (a) of (8) continues to
hold [20, Theorem 11. (Note: Properties (7) (8) and (9) are not used in
the paper.)

PROOFS OF THEOREMS 1 AND 2

Our main purpose in this section is to prove Theorems 1 and 2. We also

sketch the proofs of some of their various corollaries. Some questions and
conjectures are also raised.

THEOREM 1. Let R be a local ring with maximal ideal M. SupposeZ is

an ideal which is integral over a regular sequencex such that Z2= Ix. Then
F(Z) is Cohen-Macaulay.

Proof: Let xi, x2, .... xd minimally generate J. Extend xi, x2, .... xd to
a minimal generating set xi, x2, .... xd, y,, .... y, of Z. F(Z) is generated as
a module over F(x) by T, jl, .... p,, where 7 is 1 modulo M, and each j?,
is yi modulo ZM. Here 1 is the identity of R. We may assume that Z does
not equal (x). We shall show that 1, pi, .... J,, are free over F(x). We need
 COHEN-MACAULAYNESS OF FIBER CONE OF IDEALS 159

only show the following: If YE x’+ ‘/sr+ ‘44 and if Si E x’/x’M for i = 1, .... n
and if

i=l

in F(Z) then7=g1= . . . =g,=n in F(x).

Suppose that y.i+ C:=, gi . jj = 0 in F(Z). Also, suppose that t > 0. Lift
7 to fE$+l, 1 to 1 E R, and gi to gi E x’ for i = 1, .... IZ. Write f =
CctCCf= I uicrxi)xct9 where x, is a general degree t monomial in xi, x2, .... xd
and<;.,; R, Vi VCC Write gi = C, h, x,, where x, is a general degree t
monomial in x,, x2, .... xd and b, E R, Vi Vcc Thus (1) reduces to

c ( j, ais,) x, + ,gl (1 b.,x,) Y, E Z’+lM. (2)

a a
Since I’+ ‘M = (Zx’)M = (ZM)$ we obtain

(3)

where c, E ZM and x, is a general degree t monomial in xi, x2, .... xd.

Simplification yields

x[ i aixXi+ i bi,yi-c,]x,=O. 64)

a i= I i= 1

Since x1, x2, .... xd form a regular sequence in R, we obtain [ 15, p. 273,
Ex. 91 that Vcc

i$, ajaxi + i bix Yi -C,EX. (5)

i= 1

Writing this inclusion explicitly, we obtain

ig, ‘izxi + f birx yi - c, = i dixi, (6)

i= 1 i=l

where di E R. Simplification yields

2 [aim-di]xi+ f: b,yi=c,. (7)

i= 1 i= 1

If we consider this equation modulo ZM we deduce that 6, EM, ViVcl

because xi, x2, .... xd, y, , .... y, minimally generate I. Thus gi E $M for
i = 1, .... n. Hence gi = 0 for i = 1, .... n. Hence 7. T + C;=, gi ji = 0 reduces

481;‘143.‘1-11
160 KISHORSHAH

to ?.I= 0 in F(Z). But since F(x) is a subring of F(Z) we obtain that T= 0

in F(x).
Hence we have shown that

jY+ i gj;=a implies f=gl= . . =&=a

i=l

under the assumption that t > 0. The case t = 0 is too easy. Thus F(Z) is free
over F(x).
Now, let Jf be the unique graded irrelevant maximal ideal of F(Z). Of
course, JV” n F(x) is the unique graded irrelevant maximal ideal of F(x).
Also, nilradical of (J+” n F(x)) F(Z) is JV. Hence if W= F(x) - JV n F(x)
then E;(Z), = F(Z),n-. Therefore F(Z) is free over F(x) implies that F(Z),- is
free over F(x). +/n nrj. Our situation is this: F(x),,+-,~(~) is a regular local
ring which is a subring of F(Z),.-. Moreover F(Z), +Ais -a finitely generated
free module over F(x),,. n F(Xj. By Hironaka’s freeness lemma [ 13, 15.161
we deduce that F(Z),,,- is a-Cohen-Macaulay local ring. Finally by [ 12,
p. 125 J, we deduce that F(Z) itself is Cohen-Macaulay. This completes the
proof of Theorem 1.

COROLLARY l(a). Let S be a Cohen-Macaulay local ring and Z be an

equimultiple ideal satisfying Z2 = Ix for someminimal reduction x of I. Then
F(Z) is Cohen-Macaulay.
Proof x must be a regular sequence for Z is equimultiple and S is
Cohen-Macaulay. This completes the proof.

COROLLARY l(b). Fix the set-up of Theorem 1. Assumefurther that S is

a local overring such that S is integral and flat over R. Then the fiber cone
F(ZS) of ideal IS in S is Cohen-Macaulay.
Proof: The hypothesis implies that ht. (xR) = ht. (xS). Since S is flat
over R we obtain IS is integral over a regular sequence xS with (IS)* =
(ZS)(xS). It follows that F(ZS) is Cohen-Macaulay. This completes the
proof.

COROLLARY l(c). Fix the set-up of Theorem 1. Assumefurther that R is

Cohen-Macaulay, and R is normally Cohen-Macaulay along Z, and R/Z is
regular. Then G(Z) is Cohen-Macaulay. (Note: R is normally Cohen-
Macaulay along Z if depth Z,/Z”+ I = dim R/Z for all n 2 0.)
Proof: This follows immediately from [2, 4.51.
COROLLARY l(d) (Sally [ 16, Th. 21). Let S be a CohenMacaulay ring
with maximal ideal M. Assume M* = MT for some minimal reduction x of
M. Then F(M) is Cohen-Macaulay.
 COHEN-MACAULAYNESSOFFIBER CONEOFIDEALS 161

COROLLARY l(e) (Huneke-Sally [S, Th. 3.43). Let S he a Cohen-

Macaulay ring with maximal ideal M. Let I he a M-primary ideal such that
I2 = Ix for some minimal reduction x of I. Then F(I) is Cohen-Macaulay.

Conjecture 1 (f) (Based on Theorem 1). Let R be a local ring with

maximal ideal M. Let I be an equimultiple ideal. Let I be integral over ht .I
elements x such that Z2 = Ix. Suppose grade of I is t for some t > 1. Then
depth of F(I) is at least t.

THEOREM 2. Let R be a local ring with maximal ideal M. SupposeI is

an ideal which is integral over a regular sequence x such that I3 =12x.
Suppose that I2 n x = Ix and I’M = IxM. Then F(I) is Cohen-Macaulay.

Proof We shall freely use notation as in the proof of Theorem 1. Let

XI, x2, .... xd minimally generate x. Extend xi, x2, .... xd to a minimal
generating set xi , x2, .... xd, y,, .... y, of I. If I2 =I$ then Theorem 1
finishes our result. Assume therefore that I2 #Ix. Let e,, e2, .... e, be
elements in I2 whose images in Z2/(xZ+ Z2M) form a basis. We shall show
that 7, j,, .... j,, c?,, .... g,,, are free over F(x). We need only show the
following: If TE $+2/x’+2M and if ii E$+‘/~‘+‘M for i= 1, .... n and if
h”, E x’/x’M for i = 1, .... m and if

jT+ i gi.ji+ f Ki.zi=O (1)

i=1 i=I
in F(Z) thenT=g,= . . . =g,,=z,= . . . =h”,,,=o in F(x).
Suppose that (1) holds. Also suppose that t > 0. Lift f to f E $+ 2. Lift Sj
to gE$+‘. Lift zi to hi E x’ for all appropriate i. Let x, denote a general
degree t monomial in x,, x2, .... xd. Write f = C, a,~,, where a, E x2. Write
gi = Cm biczxz, where b,, ES. Write hi = C, ciuxz, where cia E R. Thus (1)
reduces to

Za.X.+~,(~h.,x,)Yi+~~(~c,x.)ei~I’+’M. (7-I
il
Since I* +’ M= (Z’$)M= (Z2M)x’ we obtain

Eazx=+~(~l biaYi)xx+f(;l cj~e~).r,=T;4xzt (3)

a
where d, E Z2M. Simplification of (3) yields

1 [a,+ i b,y,+ f c.,e,-d,]x,=O. (4)

a i= 1 i= 1
162 KISHOR SHAH

Since xi, x2, .... xd form a regular sequencein R, (4) implies that Va
n m
a,+ c b,y;+ c c,e,-d,E&nZ2=Z~. (5)
i= I i= 1
Since a, E x2 and b, E3, the inclusion (5) implies

jg, ci,ei EIx + Z2M. (6)

Since the images of e, in 1*/(1x + Z’ti) form a basis, (6) implies that
cia EM, Vi Var. So hi EZ’M. This shows that Ej = 0, Vi. Let xB denote a
general degree t + 1 monomial in x,, x2, .... xd. Write f= & raxg, where
rp E x. Write g, = & tipxB, where tig E R. Thus (1) reduces to

Craxa+ i 1 tiDxg y,6Z’+*M. (7)

P i=l ( /3 >
Since Z’+*M= (Z*M)$= (Z&M)?‘= (ZM)$+’ we obtain

1 [rF+ i ti/3Yi-Ufl]*fl=03 (8)

B i=l
where up E ZM. Since x1, x2, .... xd form a regular sequencein R, (8) implies
that VP

r,+ C ti@yiYUfiE&. (9)

i=l

Since rg Ex the inclusion (9) implies that

n
c tipy, - UgEX. (10)
i= 1

If we consider this equation modulo ZM it follows that tjfi E M, ViVg.

Hence ii = 0, Vi. Hence (1) reduces to 7. ? = 0. It follows that T= 0.
Thus we have shown that if t > 0 then all the coefftcients in (1) are zero.
The case t=O is too easy. Thus F(Z) is free over F(x). It follows that F(Z)
is Cohen-Macaulay. This completes the proof of Theorem 2.

COROLLARY 2(a) (Sally [ 17, Th. 2.1I). Let S be a Cohen-Macaulay

local ring with maximal ideal M. Assume M2 = Mx or M3 = M2x for some
minimal reduction x of M. Then F(M) is Cohen-Macaulay.
Proof: If M2 = Mx then the corollary follows from Theorem 1. Assume
therefore that M2 # Mx. From [ 14, pp. 1471 it follows that x n M2 = xM.
 COHEN-MACAULAYNESSOFFIBERCONEOFIDEALS 163

Since S is Cohen-Macaulay, x forms a regular sequence. It follows that

F(M) is Cohen-Macaulay. This completes the proof.
Remark 2(b). We note that the condition 1’ n x = Ix always holds in a
quasi-unmixed local ring if I is integrally closed by the results of Huneke
[6] and Itoh [7, S].
Remark 2(c). The technique used in the proof suggests a conjecture
that if F(Z) is CohenMacaulay then F(Z’) is Cohen-Macaulay Vt 3 1.

QUESTION 2(d) (CohenMacaulayness of the Bifiber Cone). Let R be a

local ring with maximal ideal M. Let Z and J be any two ideals in R. Define
the biliber cone I;(Z, J) of Z and J as follows:

F(Z, J) = @ Z’J”/Z’J”M, Vr, s > 0.

When is F(Z, J) Cohen-Macaulay?

PROOFS OF THEOREMS 3, 4, AND 5

Our main purpose in this section is to prove Theorems 3, 4, and 5. We

also sketch proofs of some of their various corollaries. First, a series of
notation statements:
e(Z) is the HilberttSamuel multiplicity of an M-primary ideal I.
y(Z) is the leading coefficient of the polynomialf(s) for large s.
f(Z) = (d- l)! [f(Z)] where d= dim F(Z). We call f(Z) as the fiber
multiplicity of I.
Z(N) is always the length of R-module N.
h(s) = l(F/Is+l).
p(J) is the minimal number of generators of any ideal J in R.
All graded rings are non-negatively graded.
a(Z) is the analytic spread of ideal I.

THEOREM 3. Let R be a local ring with maximal ideal M. Let Z be a

M-primary ideal. Suppose x is some minimal reduction of I. Then the
following assertionshold:
(a) e(Z) Q l(R/Z) + C,“= 1 [Z(Z”/Z”+‘) - /(xl”-‘/xl”)]. Furthermore,
equality holds iff R is Cohen-Macaulay.
(b) e(Z) 3 l(R/Z) + p(Z) - dim R provided R is Cohen-Macaulay.
(cl 402fW.
164 KISHORSHAH

ProoJ: Consider (a). We filter R in two different ways as follows:

R~~~~I~~12~~13~~~ zxl”-‘=,xI”... (1)
R=,ITI~~I~~I~... zI”=,I”+‘.... (2)

Since I” + I = xl”’ for some m, we obtain l(R/I” + ‘) = l(R/xI”) for all n 3 m.

Simplifying these lengths with respect to (1) and (2) we obtain

l(R/x)+ i @I”-‘/~I”)=I(R/I)+ i l(I”/I”+‘). (3)

J= I s=l ’

Since e(I) = e(x) d l(R/g) we obtain from (3) that

e(I) < l(R/I) + i [l(I”/I”+ ‘) - l(xI”- ‘/?I”)]. (4)

.s=I

Since (4) holds for all n > m and since e(I) = e(x) = l(R/x) iff R is Cohen
Macaulay, our assertion (a) follows.
Consider (b). Since R 2 I? IM and R 2 x 2 TM we obtain that

l(R/x) + 1(&M) 3 l(R/I) + l(I/IM). (5)

But l(R/x) =e(I) as R is Cohen-Macaulay and l(I/IM)=p(I) and

1(&/&M) = dim R, we deduce our assertion (b).
Consider (c). Since F(I) is a quotient of G(I) we deduce that f(s) Q h(s)
for all s. Comparing the leading coefficients yields our assertion (c).

COROLLARY 3(a) (Abhyankar [ 11). e(R) 2 1 + p(M) - dim R, prouided

R is Cohen-Macaulay.
Proof: Set I=M in (b) of the theorem. Since e(M) =e(R), the proof is
completed.

COROLLARY 3(b). f(I) 3 1 + flu(I) -a(I), provided F(I) is Cohen-

Macaulay.
Proof Apply Corollary 3(a) to the ring F(I)-,, where JV is the irrele-
vant maximal ideal of F(I) noting that dim F(I) is nothing but the analytic
spread a(I) of I and the embedding dimension of F(I),y is nothing but p(I).
This completes the proof.

COROLLARY 3(c). e(R) = I+ C,t’Y=1 [p(MS) - p(M”- ‘&)I, provided R is

Cohen-Macaulay.
ProoJ: Set I= M in (a) of the theorem to complete the proof.
 COHEN--MACAULAYNESSOFFIBER CONEOFIDEALS 165

QUESTION 3(d). Consider 3(c) above. When are each of the summation
terms non-negative? (It can be conjectured that if F(M) is Cohen-
Macaulay then every summation term is non-negative.)

Note 4. We give some notation concerning our next Theorems 4 and 5.

Consider @ ,“=0 (xl”- ’ @ R/Z). Its degree 0 component, by definition,
is R/I. Its degree 1 component, by definition, is x 0 (R/Z). Consider
@,==, (xl”- ’ @R/M). Its degree 0 component, by definition, is RIM. We
shall always take Ia = R, by definition.

THEOREM 4. Let R be a Cohen-Macaulay local ring with maximal ideal

M. Let Z be a M-primary ideal which is integral over a regular sequence3.
Then the following statements are equivalent:
(a) G(Z) is Cohen-Macaulay.
(b) @,“=, (XI”-‘@ R/Z) canonically embedsin G(Z).
(c) ~z”-‘nz”+‘=&z”,Vs3 1.

Proof We use multiplicity of Z to deduce a proof of our result. Now,

G(Z) is Cohen-Macaulay iff e(Z) = l[G(Z)/xG(Z)]. Hence from Theorem 3
we obtain that G(Z) is Cohen-Macaulay iff

l(R/Z)+ f [l(F/Z’+‘)-l(xZ” -‘/xl’)]

I =I

=l(R/Z)+ f [l(Z”/Z”+‘)-l(xP-‘+I”+‘/Z”+‘)].
s=l
Simplifying this equation we obtain that G(Z) is Cohen-Macaulay iff

f. l(xZ’-‘/xl”)= f l(xZ”-‘/xl”-InI”+‘).
.s= 1 s=l

Since 1(&Z”- ‘/xZS) > 1(&Z”- ‘/xl”- ’ n I”+‘) for all s > 1 we deduce that R
is Cohen-Macaulay iff xZ” ~ ’ n I” + ’ = xl” for all sB 1. This proves (a) is
equivalent to (c). On the other hand (b) is easily seen to be equivalent to
(c). This completes the proof of the theorem.

COROLLARY 4(a) (Sally [16, Th. 21). Assume further that I= M and
M2 = Mx. Then G(M) is Cohen-Macaulay.

COROLLARY 4(b) (Sally [ 17, Th. 2.11). Assumefurther that Z = M and

M3=M2~. Then G(M) is Cohen-Macaulay.
166 KISHOR SHAH

COROLLARY 4(c) (Valabrega-Valla [21, 3.11). Assume further that

I’ = Ix. Then G(Z) is Cohen-Macaulay.
Proof of Corollary 4(c). The hypothesis I2 = Ix implies that (c) of the
theorem holds. This completes the proof.

COROLLARY 4(d). Assume further that Z3 =I’&. Then G(Z) is Cohen-

Macaulay iff Z2 n x = Ix.
Proof The hypothesis Z3 = Z2x implies that the equation in (c) of the
theorem holds Vs 2 2. This completes the proof.

COROLLARY 4(e). Assume further that J is an ideal such that xc Zc J

and J2 = Jx. Then G(Z) is Cohen-Macaulay iff Z2 = Ix.
Proof: Assume that G(Z) is Cohen-Macaulay. Then x n Z2 = Ix holds in
particular by (c) of the theorem. But x 1 J2 2 I2 follows as J2 = Jx. Thus
Z2 = Ix. The converse is nothing but Corollary 4(c). This completes the
proof.

COROLLARY 4(f) (More Insight into the Classical Lipman-Teissier

Result). Let R be a 2-dimensional regular local ring with maximal ideal M.
Let Z be a M-primary ideal. Let x be any minimal reduction of I. Then
I2 = Ix iff G(Z) is CohenMacaulay.
Proof Let J be the integral closure of I. It is a well-known Lipman-
Teissier result that J2 = J&. Corollary 4(e) completes the proof.
Remark 4(g). Let R be analytically unramilied. Then the method in the
proof of Theorem
-- 4 also helps decide the Cohen-Macaulayness of the ring
G(Z) = @,“=, zs/zs+‘, where J denotes integral closure of the ideal J.
Remark 4(h). Theorem 4 may perhaps be considered a folklore result
although the proof outlined is new and applicable to rings such as
mentioned in 4(g) above.

THEOREM 5. Let R be a local ring with maximal ideal M. Let Z be an

ideal with a minimal reduction x. Then any two of the following statements
imply the third:
(a) F(Z) is CohenMacaulay.
(b) @,“=, (xZ’+ I @R/M) canonically embeds in F(Z).
Cc) fU)= 1+c,“=1 cA~“)-P(x~s-l)l.
Furthermore, (b) is equivalent to the statement: xl”- ’ n Z”M = xl”- ‘M,
vsa 1.
 COHEN-MACAULAYNESSOFFIBER CONEOFIDEALS 167

Proof Assume (c). Now, F(Z) is Cohen-Macaulay iff f(Z) =

KW/xW)l ifl

1+ 5 [~(Is)-Z@-ix)] = 1+ f [~(Z~)-~(Z~~l~+Z~M/ZSM)].
s=l S=l

Since p(ZsP ‘x) 3 Z(P+ ‘x/Z’-‘x n Z”M) for all .sa 1 we deduce that R is
Cohen-Macaulay iff &ZsP1 n Z”M=xZ”-‘A4 for all s > 1. Hence (a) and
(b) are equivalent. Assume (a). Then

f(Z) = l[F(Z)/xF(Z)] = 1 + 2 Z[p(Z”) - Z(xZ”-’ + Z”M/Z”M)].

*=l

But the latter expression is simply 1 + C,“= i Z[p(Z”) - E(xZ”- ‘/xZYP1 n Z”M)].
Hence we obtain that f(Z) = 1 + I,“= i Z[p(Zs) - p(xZ’+ ‘)I iff xZ’+ ‘Mn
Z”M= xl”- ‘M for all s > 1. It follows that (b) and (c) are equivalent. This
completes the proof of Theorem 5.

COROLLARY 5(a) (Sally [ 16, Th. 23). Assume further that R is Cohen-
Macaulay, and Z= M, and M2 = Mx. Then F(M) is Cohen-Macaulay.

COROLLARY 5(b). Let R be a Cohen-Macaulay local ring with maximal

ideal M. Let Z be a M-primary ideal which is integral over a regular x.
Assume that the canonical mapfrom @ SmzO (xl”- ‘/XI’) to G(Z) isfaithfully
flat. Then the following statements hold:
(a) G(Z) is Cohen-Mucauluy
(b) F(Z) is Cohen-Mucauluy ijjf

f(Z)= 1+ f [~(z3)-/L(&z~-1)].
r=l

Proof. Consider (a). Now @,“=, (xPP ‘/xl”) canonically embeds in

G(Z) because of faithful flatness. Hence (a) follows from Theorem 4.
Consider (b). Set A to be the ideal generated by M/Z in G(Z). Set ~9’” to
be the ideal generated by M/Z in @,“=O (xZsP ‘/XI’). Clearly JV extends to
J? in G(Z). On the other hand faithful flatness implies that

1
= G
S=O
(xZ”-‘nMl”)/~Z”. (1)
It follows that xl”- ‘M= xl”+ ’ n MI” for all s >, 1. From Theorem 5 we
deduce our assertion (b).
168 KISHORSHAH

PROOFS OF THEOREMS 6 AND 7

Our main purpose in this section is to prove Theorems 6 and 7. We also

state the various corollaries leaving the easy proofs for the perceptive
reader. Set D(n) = C:=, f(s). We refer to o(n) as the fiber polynomial of I.

LEMMA 6. Let R be local ring with maximal ideal M. Let I be an ideal

with a minimal reduction x. Set V(x) = rank of F(Z) over F(x). Assume F(Z)
to be Cohen-Macaulay. Then V(x) is independent of x. Moreover
V(x) = f(O
Prooj Fix any minimal reduction 5 of I. Let ~2’ be the irrelevant maxi-
mal ideal of F(x). Let .N be the irrelevant maximal ideal of F(Z). Then
V(x) also equals rank of F(Z),,. over F(x)&. However, the latter rank is
nothing but the multiplicity of J# F(Z),,*-. Since x forms a reduction of Z we
obtain that &F(Z)., is a reduction of JV”. Hence V(x) is but the multi-
plicity of F(Z). This shows that V(x) is independent of x. Furthermore
V(x) =f(Z). This completes the proof of the lemma.

THEOREM 6. Let R be a local ring with maximal ideal M. Let Z be an

ideal with a minimal reduction x. Set p(x) = d. Assume F(Z) to be Cohen-
Macaulay. Then

f(n)= f [~(ZS)-Z(~ZS~l/~ZS-‘,Z”M)] n+fIS-l), Vn>O.

s= 0 (
ProoJ: For each s >O, pick a basis from the vectorspace
(Z”/Z”M)/((xZ”-’ + I”M)/Z”M). All these basis elements together with
1 E R/M collectively form a generating set for F(Z) considered as a F(g)-
module. The number of these basis elements equal l(F(Z)/&F(Z)). The latter
equals f(Z) as F(Z) is Cohen-Macaulay. We deduce, from the lemma that
this generating set is as a matter of fact a free generating set. Let b,Y,,bs2,....
b, be the basis corresponding to each s. Decompose F(Z) as follows:

WJ=,+o (+, b,,F(x)). (1)

In (1) above,f,=~(Z”)-I(xl”-‘/LrZ”~‘nZ”M) for all s> 1 andf,= 1. The

dimension of the sth graded piece of F(x) is the number of degree s
monomials in d variables. Equating the vectorspace dimension of the nth
graded piece of both sidesin (1) we obtain

f(n)= f [~(I”)-I(xZ”~‘/xZ”~‘nZ”M)] n+~~~-l), Vn>O. (2)

S=O (

This completes the proof of our theorem.

 COHEN-MACAULAYNESSOFFIBER CONEOFIDEALS 169

COROLLARY 6(a). Assume further that I= M. Then the Hilbert function

of R equals

n+d-s-l
H(n)= c [/@I”)+#+‘)] VnBO.
A=0
d-l >’
COROLLARY 6(b). Assume further that I= M. Then

e(R)= f [p(M”)-p(xM”-‘)I.
s= 0

Remark 6(c). The proof of Theorem 6 suggests that the classical

Hilbert polynomial of an M-primary ideal I may best be analysed if G(I)
is viewed as a module over some suitable graded ring 2. It certainly works
out for I= M as Corollary 6(a) shows.

QUESTION 6(d). Consider the biliber cone F(I, .I). Set f(r, s) =
p(I’J”/I’J”M). It can be shown that f(r, s) is actually a polynomial in r, s
for all large r, s. Assume now that F(Z, J) is Cohen-Macaulay. Describe the
bitiber function f(r, s). Perhaps a multidimensional analogue of Noether
Normalisation would help analyse questions concerning the bifiber.
Conjecture 6(e). The converse of Theorem 6 holds.

THEOREM 7. Let R be a local ring with maximal ideal M. Let I be an

ideal which is integral over a regular sequence x such that I2 = Ix. Set
u(x) = d. Then

Proof If s b 3 then xl”- ’ = I”. Hence

f,=CL(IS)--(XI1~‘/X~~‘nI”M)
= p(I’) - I( P/I’ n I”M) = 0, t/s 3 3.

Thus (2) in the proof of Theorem 6 reduces to the assertion in our theorem.
This completes the proof of our theorem.

COROLLARY 7(a). The fiber polynomial of I equals

COROLLARY 7(b). f(I)= 1 +p(I)- d

 170 KISHOR SHAH

COROLLARY 7(c) (Sally [ 18, Th. 11). Assume further that I= M. Then
the Hilbert function h(n) of the ring R is

PROOF OF THEOREM 8

Our purpose in this section is to prove Theorem 8. We refer to [2, 31 for

the notion of normal flatness. We conclude this paper with a simple but a
startling Corollary 8(a).

LEMMA 8. Let R be a local ring with maximal ideal M. Let I be a

M-primary ideal. Then R is normally flat along I iff H(n) = l(R/I) . D(n),
VnbO.

Proof Suppose that R is normally flat along I. Then IS/IS+’ is a free

R/I-module for all s. Fix some s. So there is an isomorphism from a direct
sum of t copies of R/I-modules onto IS/I”+ ‘, where t depends on s. Tensor
the resulting short exact sequence with R/M. We thus obtain an
isomorphism from a direct sum of t copies of R/M-modules onto I”/I”M.
Consequently, the R/M-dimension of I”/I”M equals t. But l(I”/I”+‘) =
l(R/I) . t. Summing up the terms in this equation from s = 0 to s = n we
obtain a part of the lemma.
Suppose that H(n) = l(R/I) . D(n), V/n 3 0. Consider any fixed s. There is
a natural R/I-module map from a direct sum of u(I) copies of R/I onto
IS/P+ l. Our supposition implies that this map must be an isomorphism. So
R is normally flat along I. This also finishes the proof of our lemma.

THEOREM 8. Let R be a Cohen-Macaulay local ring with maximal ideal

M. Let I be an M-primary ideal which is integral over a regular sequence
x such that I2 = Ix. Then R is normally flat along I iff e(I) =
l(R/I) . (1 + u(I) - d).
Proof: According to the lemma, R is normally flat along I iff H(n) =
l(R/I) . D(n), Vn 2 0. But a well-known result of Kubota [ 10, p. 4401 and
Corollary 7(a) give concrete expressions for H(n) and D(n). They are

II(n)=e(I)(n+~-l)+l(R/I)(n+dd~l), Vn>O

D(n)=(l+p(l)-d)(n+~-l)+l.(n~d~l), Vn>O. (2)

 COHEN-MACAULAYNESSOFFIBER CONEOF IDEALS 171

It follow that R is normally flat along I iff e(Z) = l(R/Z) . (1 + ~(1) - d). This
completes the proof of our theorem.

COROLLARY 8(a). Assumefurther that e(Z) is prime. Then R is normally

flat along I $I is a complete intersection or I= AL
Proof: If R is normally flat along I then by the theorem l(R/Z) = 1 or
1 + ~(1) - d= 1 as e(Z) is prime. Hence I is a complete intersection or
I= A4. The converse is well known.

ACKNOWLEDGMENTS

It is a pleasure to thank Bill Heinzer, Craig Huneke, Jack Ratliff, and Dave Rush for their
help at various times.

REFERENCES

1. S. S. ABHYANKAR, Local rings of high embedding dimension, Amer. J. Math. 89 (1967)

1073-1077.
2. U. GROTHE, M. HERMANN, AND U. ORBANZ, Graded Cohen-Macaulay rings associated to
equimultiple ideals, Mafh. Z. 186 (1986), 531-556.
3. H. HIRONAKA, Resolution of singularities of an algebraic variety over a field of charac-
teristic 0, Ann. of Math. 79 (1964) 1099326.
4. M. HOCHSTER, Some applications of the Frobenius in characteristic 0, Bull. Amer. Mufh.
Sot. 84 (1978) 886-912.
5. C. HUNEKE AND J. SALLY, Birational extensions in dimension two and integrally closed
ideals, J. Algebra 115 (1988), 481-500.
6. C. HUNEKE, Hilbert functions and symbolic powers, Michigan Math. .I. 34 (1987)
293-318.
7. S. ITOH, Integral closures of ideals of the principal class, Hiroshima Mafh. J. 17 (1987),
373-375.
8. S. ITOH, Integral closures of ideals generated by regular sequences, J. Algebra 117 (1988),
39&401.
9. J. LIPMAN AND B. TEISSIER, Pseudo-rational local rings and a theorem of Briancon-Skoda
about integral closures of ideals, Michigan Math. J. 28 (1981), 97-116.
10. K. KUBOTA, On the Hilbert-Samuel function, Tokyo J. Math. 8, No. 2 (1985), 439448.
11. H. MATSUMURA, “Commutative Algebra,” Benjamin, New York, 1970.
12. MATIJEVIC AND P. ROBERTS, A conjecture of Nagata on graded Cohen-Macaulay rings,
J. Math. Kyofo Univ. 14, No. 1 (1974), 1255128.
13. M. NAGATA, “Local Rings,” Interscience, New York, 1962.
14. D. G. NORTHCOTT AND D. REES, Reductions of ideals in local rings, Proc. Cumb. Phil. Sot.
50 (1954), 145-158.
15. D. G. NORTHCOTT, “Lessons on Rings, Modules and Multiplicities,” Cambridge Univ.
Press, Cambridge, 1968.
16. J. SALLY, On the associated graded ring of a local Cohen-Macaulay ring, J. Mafh. Kyofo
Univ. 17, No. 1 (1977), 19-21.
17. J. SALLY, Tangent cones at Gorenstein singularities, Composifio Mufh. 40 (1980), 167-175.
172 KISHOR SHAH

18. J. SALLY, Cohen-Macaulay local rings of maximal embedding dimension, J. Algebra 56

(1979), 168-183.
19. K. SHAH, “Coefftcient Ideals of the Hilbert Polynomial and Integral Closures of
Parameter Ideals,” Ph.D. thesis, Purdue University, 1988.
20. K. SHAH, On the dimension of the libcr cone of monomial ideals, preprint, 1989.
21. P. VALABREGA AND GWSEPPE VALLA, Form rings and regular sequences, Nagoya Math. J.
72 (1978), 93-101.
22. 0. ZARISKI AND P. SAMUEL, “Commutative Algebra,” Vol. II, Van Nostrand, New York,
1960.