\input amstex
\loadbold
\documentstyle{amsppt}
\magnification=\magstephalf
\voffset=-20pt
\baselineskip=14pt
\overfullrule=0pt
\font\footfont=cmr8
\font\footitfont=cmti8
\font\sc=cmcsc10 at 12pt
\font\large=cmbx12 at 14pt
\font\bb=msbm10
%nuove definizioni
\define\Spec{\hbox{Spec}\hskip 0.3em}
\define\Proj{\hbox{Proj}\hskip 0.3em}
\define\Ker{\hbox{Ker}\hskip 0.3em}
\define\hht{\hbox{ht}\hskip 0.3em}
\define\arith-deg{\hbox{arith-deg}\hskip 0.3em}
\define\br{\bigpagebreak}
\define\nt{\noindent}
\define\th{\theta}
\define\Th{\Theta}
\define\Del{\Delta}
\define\la{\lambda}
\define\La{\Lambda}
\define\al{\alpha}
\define\be{\beta}
\define\si{\sigma}
\define\Si{\Sigma}
\define\un{\underline}
%\define\me{\medspace}
\define\me{\hskip 0.3em}
\define\de{\displaystyle}
\define\lw{\longrightarrow}
\define\rw{\rightarrow}
\define\ten{\otimes}
\define\Ten{\bigotimes}
\define\ds{\oplus}
\define\DS{\bigoplus}
\define\Un{\bigcup}
\define\ssq{\subseteq}
\define\intR{\overline R}
\define \m{\frak m}
\define \n{\frak n}
\define\Q{\Bbb Q}
\define\N{\Bbb N}
\define\C{\frak C}
\define\I{\widetilde I}
\newcount\sectno \sectno=0
\newcount\thmno \thmno=0
\define\section#1{\smallskip
\global\advance\sectno by 1 \global\thmno=0
\noindent{\bf \the\sectno. #1} \smallskip}
\define\thmline#1{\vskip 2pt
\global\advance\thmno by 1
\noindent{\bf #1\ \the\sectno.\the\thmno:}\ \ %
\bgroup \advance\baselineskip by -1pt \it
\abovedisplayskip =2pt
\belowdisplayskip =2pt
\parskip=0pt
}
\define \dthmline#1{\vskip 2pt
\global\advance\thmno by 1
%\noindent{\bf \the\sectno.\the\thmno\ #1:}\ \ }
\noindent{\bf #1\ \the\sectno.\the\thmno:}\ \ }
\define\defin{\dthmline{Definition}}
\define\thm{\thmline{Theorem}}
\define\endb{\egroup \vskip 1.2ex}
\define\cor{\thmline{Corollary}}
\define\lemma{\thmline{Lemma}}
\define\rmk{\dthmline{Remark}}
\define\dis{\dthmline{Discussion}}
\define\ques{\dthmline{Question}}
\define\nota{\dthmline{Notation}}
\define\question{\thmline{Question}}
\define\ex{\dthmline{Example}}
\define\prop{\thmline{Proposition}}
\define\proof{\smallskip\noindent {\sl Proof:\ \ }}
\define\proofof#1{\smallskip\noindent {\sl Proof of #1:\ }}
\define\claim{\smallskip\noindent {\sl Claim:\ }}
\define\proofclaim{\smallskip\noindent {\sl Proof of the claim:\ }}
\define\addots{\mathinner{\mkern1mu\raise1pt\hbox{.}\mkern2mu\raise4pt\hbox{.}
\mkern2mu\raise7pt\vbox{\kern7pt\hbox{.}}\mkern1mu}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Forward referencing
% The following will put labels also in a file so I can do forward referencing.
% I need an open file for that -- see below.
\define\label#1{\unskip\immediate\write\isauxout{\noexpand\global\noexpand\edef
\noexpand#1{{\the\sectno.\the\thmno}}} {\global\edef#1{\the\sectno.\the\thmno}}
\unskip}
\define\sectlabel#1{\unskip\immediate\write\isauxout{\noexpand\global\noexpand\edef
\noexpand#1{{\the\sectno}}} {\global\edef#1{\the\sectno}} \unskip}
% Opening a file for forward referencing:
\newwrite\isauxout
\openin1\jobname.aux
\ifeof1\message{No file \jobname.aux}
\else\closein1\relax\input\jobname.aux
\fi
\immediate\openout\isauxout=\jobname.aux
\immediate\write\isauxout{\relax}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\vskip 4ex
\baselineskip = 18pt
\centerline{\bf IDEALS HAVING A ONE-DIMENSIONAL}
\vskip 3ex
\centerline{\bf FIBER CONE}
\vskip 4ex
\centerline{ Marco D'Anna, Anna Guerrieri and William Heinzer}
\vskip 3ex
\centerline{Dedicated to Jim Huckaba on the occasion of his retirement}
\vskip 1cm
\document
\baselineskip=19pt
\noindent
{\bf Abstract.}
For a regular ideal $I$ having a principal reduction
in a Noetherian local ring $(R,\m)$ we consider
properties of the powers of $I$ as reflected in
the fiber cone $F(I)$ and
the associated graded ring $G(I)$ of $I$. In particular,
we examine the postulation number of $F(I)$ and compare it
with the reduction number of $I$, and the postulation
number of $G(I)$ when the latter is meaningful. We discuss
a sufficient condition for $F(I)$ to be Cohen--Macaulay and
consider for a fixed $R$ what is possible for the reduction
number $r(I)$ of $I$ and the multiplicity of $F(I)$.
\vskip 0.4cm
\section{Introduction.}
\vskip 0.4cm
Given an ideal $I$ in a Noetherian local ring $(R,\m)$,
information on properties of $I^n$ as $n$ grows
is encoded in various graded rings related to $I$. These
graded rings are built by using and manipulating the
$I$-adic filtration, $\{I^n\}_{n\ge 0}$ on $R$. Among these
rings are:
\noindent
(i)
the {\it Rees algebra} of $I$,
$\de R[It] = \oplus_{n \ge 0}I^n$,
\noindent
(ii)
the {\it associated graded ring} of $I$,
$\de G(I) = \oplus_{n \ge 0}I^n/I^{n+1} = \oplus_{n \ge 0}G_n
\cong R[It]/IR[It]$,
and
\noindent
(iii)
the {\it fiber cone} of $I$,
$\de F(I) = \oplus_{n \ge 0}I^n/\m I^n = \oplus_{n \ge 0}F_n
\cong R[It]/\m R[It]$.
These are a few of the {\it blowup algebras} of $I$,
(see \cite{V1}),
where by this term one refers to those algebraic objects that
are related to the concept of blowing up a variety along a subvariety.
The graded rings $G(I)$ and $F(I)$ are both {\it homogeneous, }or
{\it standard, }
in the sense that they are generated by their forms of
degree one over their subring of elements of degree zero.
If $I$ is $\m$-primary, the Hilbert function giving the length,
$\la (I^n/I^{n+1})$, of $I^n/I^{n+1}$ as an $R$-module
is $H_G(n)$, where $H_G(X)$ is the Hilbert function
of the associated graded ring $G(I)$.
For an arbitrary ideal $I \subseteq \m$ of $(R,\m)$,
the fiber cone $F(I)$ has the attractive property of
being a finitely generated graded ring over the
residue field $K := R/\m$.
It is well known in this setting that
the Hilbert function $H_F(n)$
giving the dimension of $I^n / \m I^n$ as a vector space
over $K$ is defined for $n$ sufficiently large by
a polynomial $h_F(X) \in \Q[X]$, the {\it Hilbert polynomial}
of $F(I)$ \cite{Mat, Corollary, page 95}, \cite{AM, Corollary 11.2}.
A simple application of Nakayama's lemma,
\cite{Mat, Theorem~2.2}, shows that
the cardinality of a minimal set of generators of $I^n$,
$\mu(I^n)$, is equal to $\lambda(I^n/\m I^n)$, the Hilbert function
$H_F(n)$ of $F(I)$. For these reasons $G(I)$ and $F(I)$ are good
objects to analyze and compare when studying the asymptotic
properties of $I$.
If $I$ is $\m$-primary it is natural
to ask about the relationship of
the Hilbert function $H_G(X)$ and
Hilbert polynomial $h_G(X)$ of $G(I)$
with the Hilbert function $H_F(X)$ and
Hilbert polynomial $h_F(X)$ of $F(I)$. We begin
such an investigation here in the one-dimensional case.
We also consider, with no restriction
on the dimension of $R$, properties of the fiber cone $F(I)$ of
ideals $I$ that have principal reductions in $R$.
Thus, in certain aspects, this paper is a continuation
of our work in \cite{DGH}. If
$I$ is a regular ideal having a principal reduction,
we analyze in \cite{DGH}
the mutual relations
among the following, where $\N$ denotes the nonnegative integers:
\roster
\item
$r = r(I)
= \text{min}\{ n \in \N \ | \ I^{n+1}=xI^n \text{ for some x } \in I\}$,
the {\it reduction number} of $I$,
\item
$k = k(I)
= \text{min}\{ n \in \N \ | \ \I= (I^{n+1}: I^n )\}$,
the {\it Ratliff--Rush number} of $I$,
\item
$h = h(I) =
\text{min}\{ n \in \N \ | \ \widetilde {I^m}
= I^m\ \text{for all}\ m\ge n\}$,
the {\it asymptotic Ratliff--Rush number} of $I$.
\endroster
In defining $k$ and $h$, we are using the
{\it Ratliff--Rush closure} of $I$ and of its powers,
namely
$ \de\widetilde{I^k} := \cup_{n \in \N}(I^{n+k}:_R I^n) $.
This concept was introduced by L. J. Ratliff and D. E. Rush in \cite{RR}
where it was also observed that
$\widetilde {I^m} = I^m$
for all sufficiently large integers $m$ \cite{RR, Remark~(2.3)}.
This motivates our
definition of $h$. We have $h(I) = 0$ if and only if
$G(I)$ contains a regular homogeneous element of
positive degree. If $h(I) > 0$, then $h(I) \ge 2$.
It is shown in \cite{RV} in the case of the maximal ideal of a
one-dimensional local Noetherian ring and in a more general setting
in \cite{DGH, (2.1) and (2.2)} that if $I$ has a
principal reduction, then $h \le r$ and $k \le \max\{0,r-1\}$.
If $R$ is a reduced Noetherian ring with
total ring of fractions $Q(R)$ and if the integral
closure $\overline R$ of $R$ in $Q(R)$ is a finitely generated $R$-module,
it is shown in \cite{DGH, (3.2) and (3.10)} that if $I$ is contained
in the conductor of $\overline R$ into $R$, then
$k = \max\{0, r-1\}$, and, if $I \ne \widetilde{I}$,
then $h=r$.
%With the same hypotheses, one also obtains
%$\widetilde{I^{n+1}} \cap I^n = I^n \I$, for all $n \ge 0$
%see \cite{DGH, (3.6)}.
%This gives an analogue for the Ratliff--Rush filtration
%to the equalities shown by Itoh, \cite{I},
%and by Huneke, \cite{H1}, for the filtration of the integral closures
%of the powers of $I$.
%Further, it is shown in \cite{DGH, (3.9)} that, in this context,
%the reduction number of the
%filtration $\{\widetilde{I^n}\}_{n \ge 0}$ is less than or
%equal to $1$.
Suppose $I$ is an $\m$-primary ideal of a Noetherian
local ring $(R,\m)$. The {\it postulation
number} $n(I)$ of $I$ is the largest integer $n$ such that
$H_G(n) \ne h_G(n)$, where $H_G(X)$ and $h_G(X)$ are,
respectively, the Hilbert function and Hilbert polynomial
of $G(I)$. Thus $H_G(n)$ is the length of $I^n / I^{n+1}$ as an
$R$-module. T. Marley shows in \cite{Mar, Theorem 2}
that if $(R,\m)$ is a $d$-dimensional Cohen--Macaulay local ring
with infinite residue field and $I$ is an $\m$-primary ideal
such that $G(I)$ has depth at least $d-1$, then
$r(I) = n(I) + d$. In the case where $I = \m$
this had been shown by J. Sally \cite{Sa, Proposition~3}.
Note that the condition on the depth of
$G(I)$ is vacuous if $d =1$, a case also considered by
A. Ooishi in \cite{O, Proposition 4.10}. Thus if $I$ is an
$\m$-primary ideal in a one-dimensional Cohen--Macaulay local
ring $(R,\m)$ with $R/\m$ infinite, then $n(I) = r(I) - 1$.
We define the {\it fiber postulation
number} $fp(I)$ of $I$ to be the largest integer $n$ such
that $H_F(n) \ne h_F(n)$, where $H_F(X)$ and $h_F(X)$ are, respectively,
the Hilbert function and Hilbert polynomial of the fiber cone $F(I)$.
Suppose $I$ is a nonprincipal regular ideal having a principal reduction
in a Noetherian local ring $(R,\m)$. We observe in
Proposition \FPnoRedno \me that $fp(I) \le r(I) - 1$,
where $r(I)$ is the reduction number of $I$. Thus
if $I$ is a nonprincipal $\m$-primary ideal of a one-dimensional
Cohen--Macaulay local ring, then the fiber postulation number
$fp(I)$ is less than or equal to the postulation number $n(I)$ of $I$.
For each integer $n \ge 3$, we exhibit
in Example \FPnoIneq \me the
existence of a
one-dimensional Cohen--Macaulay local domain
$R_n$ and an ideal $I_n$ primary for the
maximal ideal of $R_n$ such that $fp(I_n) = 0$
while $r(I_n) = n - 1$.
This shows that the difference $r(I) - fp(I)$ can
be arbitrarily large for ideals $I$ and rings
$R$ as in Propositon \FPnoRedno. \me For the
ideals $I_n$ of Example \FPnoIneq \me the
fiber postulation number $fp(I_n)$ is
strictly smaller than the postulation number $n(I)$.
It follows that the fiber cone
$F(I_n)$ is not Cohen--Macaulay.
Indeed, as we observe in Proposition~\FPnoeq :\me
if $F(I)$ is Cohen--Macaulay, then $fp(I) = r(I) - 1$.
This and a higher dimensional analogue stating that
if $F(I)$ is Cohen--Macaulay, then
$fp(I) = r(I) - \ell$, where $\ell$ is the analytic spread of $I$,
are consequences of results of T. Cortadellas and S. Zarzuela \cite{CZ}
and C. D'Cruz, K. N. Raghavan and J. Verma \cite{DRV} on the structure of
the Hilbert function of a Cohen--Macaulay fiber cone $F(I)$.
One can use \cite{BH, Proposition 4.1.12} to obtain the
asserted relation between the
reduction number and fiber postulation number.
In Example \negative, \me we exhibit a 3-dimensional
Cohen--Macaulay local domain $(R,\m)$ of multiplicity 3
having, for each positive integer $k$, an ideal $I_k$
such that: (i) $I_k$ has a principal reduction, (ii) $I_k$ is
minimally generated by $k + 3$ elements, and (iii) the
fiber cone $F(I_k)$ has multiplicity 3 and is not Cohen--Macaulay.
We observe in Remark \tangent \me that
the ideal $I_3$ of Example \FPnoIneq \me
provides an example of a one-dimensional
Cohen--Macaulay local domain $(R,\m)$
and an $\m$-primary ideal $I$ such that
the associated graded ring
$G(I)$ is Cohen--Macaulay, while the fiber cone
$F(I)$ is not Cohen--Macaulay.
In Example \highmult \me we exhibit a 3-dimensional
Cohen--Macaulay local domain $(R,\m)$ of multiplicity 2
with the following property. For
each positive integer $k$, the ring $R$ has a
stable ideal $I_k$ having a principal reduction
such that the fiber cone $F(I_k)$
of $I_k$ is Cohen--Macaulay and has multiplicity $k + 2$.
Therefore, in contrast with the case where $\dim R = 1$, for
$R$ of higher dimension there may exist no upper
bound on the multiplicity of $F(I)$ as $I$ varies
over the ideals of $R$ that have a principal reduction.
In \S 5 we consider a question raised by S. Huckaba
\cite{Hu1,Question 2.6} as to whether the multiplicity
of a quasi-unmixed analytically unramified Noetherian
local ring containing an infinite field
of characteristic different from 2 is strictly larger than the
reduction number of each regular ideal of
analytic spread one of the ring. Using an interesting
observation shown to us by Craig Huneke
(Proposition \huneke \me) we obtain a positive answer
to Huckaba's question in several special cases.
For example, in Corollary \analytictwodim \me we confirm a
positive answer to the question for the
case of a 2-dimensional Noetherian analytically irreducible
local domain. However, the general case of Huckaba's
question remains open.
\vskip 3ex
The authors thank Craig Huneke for interesting and helpful conversations
on the topic described in \S 5.
\vskip 3ex
\section{The postulation number of a one-dimensional fiber cone.}
\vskip 3ex
Suppose $I$ is a regular ideal of a Noetherian local ring $(R,\m)$.
If there exists $x \in I$ and an integer $n \ge 0$ such
that $xI^n = I^{n+1}$, then $xR$ is said to be a
{\it principal reduction of } $I$. In terms of the
fiber cone $F(I)$ this is reflected in the fact that if
$\overline x \in I/\m I$ denotes the image of $x$, then
$F(I)$ is a finitely generated integral extension of
its polynomial subring $K[\overline x]$, where
$K := R/\m$. Indeed, if $K$ is infinite, the
converse also holds: if the
fiber cone $F(I)$ is one-dimensional, then using
Noether normalization, there exists $\overline x \in I/\m I$
such that $F(I)$ is integral over $K[\overline x]$. If
$x \in I$ is a preimage of $\overline x$, then $xR$ is a
principal reduction of $I$.
D. G. Northcott and D. Rees in their famous paper \cite{NR}
introduce reductions of ideals and associate to an
ideal $I$ a homogeneous ideal in a polynomial ring
over $K$, the {\it null form ideal of } $I$.
The null form ideal of $I$ is the kernel in a
presentation of the fiber cone $F(I)$ as a homomorphic
image of a polynomial ring over $K$.
K. Shah in \cite{S} coined the term `fiber cone' for
this ring.
If $I$ is a regular ideal having a principal
reduction, the reduction number
$r(I)$ is known to be independent of the principal
reduction $x$ \cite{Hu1, page~504}. In this
situation, $r(I)$ is also equal to the reduction
number of the algebra $F(I)$. If
$m_1, \dots,m_s$ is a minimal set of homogeneous module
generators for $F(I)$ over $(R/\m)[\overline x]$, then
$r(I) = \max\{\deg m_i \}$, see \cite{V2}.
The Hilbert function $H_F(X)$ defined so that
$H_F(n) = \la (I^n /mI^n) $
is given for all sufficiently large $n$ by a polynomial
$h_F(X)$. Since, in this case, $\de F(I) = \oplus_{n \ge 0}I^n /mI^n$ is a
one-dimensional homogeneous graded ring over a field,
$h_F(X)$ is a constant $f_0$. Here
$f_0$ is a positive integer. It is the multiplicity
of the graded ring $F(I)$ and the number of
elements in a minimal
generating set of $I^n$ for all sufficiently large $n$.
It is of interest to also consider
the first iterated Hilbert function of $F(I)$. This is
the function
$\de H^1_F(n) = \sum_{j=0}^n H_F(j) = \sum_{j=0}^n \la(I^j/ \m I^j)$.
It is a polynomial function for $n > fp(I)$.
The associated Hilbert polynomial is
$h^1_F(X) = f_0 (X+1) - f_1= f_0 X - (f_1 - f_0)$.
In analogy with the Hilbert coefficients of $G(I)$ in
the case where $I$ is $\m$-primary, it is of interest
to have sufficient conditions in order that
$H^1_F(0) - h^1_F(0) = 1 - f_0 + f_1 \ge 0$. In other words,
sufficient conditions in order that $f_1 \ge f_0 - 1$.
Notice that for $n\ge r-1$, where $r=r(I)$ one has
$$
H^1_F(n) = \sum_{j=0}^n \la(I^j/ \m I^j) =
H^1_F(r-1) + (n-(r-1)) f_0= nf_0 - ((r-1)f_0 - H^1_F(r-1)).
$$
It follows that $f_1 - f_0 = (r-1)f_0 - H^1_F(r-1)$
and $f_1 = rf_0 - H^1_F(r-1)$. We observe in
Example \negative \me below that this last
equality implies that $f_1$ can be negative
and that for a fixed ring $R$ there may be no
lower bound on the value of $f_1$ as we vary over
ideals of $R$ having a principal reduction.
In the case, however, where the fiber cone $F(I)$
is Cohen--Macaulay, as we observe in Remark \northcott, \me
it follows from classical results
that $f_1 \ge f_0 -1$.
\vskip 2ex
\rmk
\label{\northcott}
Suppose $(R,\m)$ is a Noetherian
local ring and $I$ is a regular ideal of $R$ having
a principal reduction. If the fiber cone $F(I)$ is
Cohen--Macaulay, then a result of Northcott
implies that $f_1 \ge f_0 -1$. For the one-dimensional
homogeneous graded ring $F(I)$ has the same Hilbert function
as the local ring $(S,\n)$ obtained by localizing $F(I)$ at its
homogeneous maximal ideal. If $F(I)$ is Cohen--Macaulay,
then $(S,\n)$ is a one-dimensional Cohen--Macaulay local ring.
Northcott defines normalized Hilbert coefficients $e_0$ and $e_1$
of $\n$ such that $\lambda(S/\n^{n+1}) = e_0(n+1) - e_1$ for
$n >> 0$,
and proves \cite{No, page~211} that $e_1 \ge e_0 - 1$.
Since $\lambda(S/\n^{n+1}) = H^1_F(n) = f_0(n+1) - f_1$
for $n >> 0$, we have $e_0 = f_0, e_1 = f_1$ and thus
$f_1 \ge f_0 - 1$.
\vskip 2ex
We are interested in comparing the fiber postulation
number $fp(I)$ of $I$ with other invariants associated to
$I$. We start by observing in Proposition \FPnoRedno \me an upper
bound for $fp(I)$ in terms of the reduction number of $I$.
\prop
\label{\FPnoRedno}
Suppose $(R,\m)$ is a Noetherian
local ring and $I$ is a nonprincipal regular ideal of $R$ having
a principal reduction $xR$. Then $fp(I) \le r(I) - 1$,
where $r(I) = r$ is the reduction number of $I$.
\endb
\proof
Notice that $I^{s}/ \m I^{s} = x^{s-r}I^r/\m x^{s-r}I^r
\cong I^r/\m I^r$ for all $s \ge r = r(I)$. Thus
$H_F(n) = \la(I^r/\m I^r)$ for every $n \ge r$.
It follows that $h_F(X) = f_0 = \la(I^r/\m I^r)$
and that $H_F(n) = h_F(n)$ for every $n \ge r$.
Therefore $fp(I) \le r(I) - 1$.
\qed
\vskip 3ex
With notation as in the proof of Proposition \FPnoRedno, \me observe
that $fp(I) < r(I)-1$ if and only if
$h_F(r-1) = H_F(r-1) = f_0$. Thus
$fp(I) < r(I) - 1$ if and only if
$I^{r-1}$ is minimally generated by $f_0$ elements.
We use this observation in order to construct in Example~\FPnoIneq \me
for each integer $n \ge 3$ an example of a
one-dimensional Cohen--Macaulay local domain
$R_n$ and an ideal $I_n$ primary for the
maximal ideal of $R_n$ such that $fp(I_n) = 0$
while $r(I_n) = n - 1$.
As is the case for our examples in \S 4 of
\cite{DGH}, the rings $R_n$
in Example \FPnoIneq \me are
complete one-dimensional local Cohen--Macaulay domains of the form
$R_n = K[\![t^s : s \in S_n]\!]$, i.e. formal power
series in the indeterminate $t$ with coefficients in
a field $K$ and exponents from an additive submonoid $S_n$ of
the nonnegative integers that contains all sufficiently large
integers. The formal power series ring
$K[\![t]\!]$ is a finitely generated $R_n$-module having the
same fraction field as $R_n$ and is thus the integral closure
of $R_n$. To establish the asserted properties of the
rings $R_n$ and ideals $I_n$ in Example \FPnoIneq \me we
work directly with the additive monoid $S_n$ and the
corresponding semigroup ideal $I_n$ of $S_n$.
\vskip 3ex
\ex
\label{\FPnoIneq}
Fix an integer $n \geq 3$, and consider the
numerical semigroup $S_n$ generated by the
elements $a := 2n$, $b := 4n-1$, $d := n(2n-1)$,
and $c_h := (n+h)(2n-1) + 1$, where $h = 3 \ldots n$.
Thus $S_n = $. We prove
that these $n+1$ elements are the minimal generating
set for $S_n$. In considering the general case, it
is useful to keep in mind the first few examples,
$S_3 = < 6, 11, 15, 31>$, $S_4 = <8, 15, 28, 50, 57>$
and $S_5 = <10, 19, 45, 73, 82, 91>$.
\vskip 2ex
\noindent
{\bf Claim 2.3.1: } The monoid $S_n$ is minimally generated by
$a$, $b$, $d$, $c_3$, $c_4, \dots, c_n$.
\proof
We have, modulo $2n$, that $b \equiv -1 \equiv 2n-1$,
$d \equiv -n \equiv n$ and $c_h \equiv -n-h+1 \equiv n-(h-1)$.
The smallest integer $x$ in $$ such that $x \equiv n$
is $x=nb= 4n^2-n$, which is larger than $d$; hence $d \notin $
and $a,b,d$ are minimal generators of the submonoid $$.
Now we consider $c_h$.
If $h > 3$ and if $3 \le i $.
In fact, we have $c_h-c_i =(h-i)(2n-1) \iff c_h-c_i \notin $.
But $c_h-c_i \equiv -h+i$ (mod $2n$) \ and
the smallest integer $x$ in $$ such that $x \equiv -h+i$
is $x=(h-i)b = (h-i)(4n-1)$ which is larger than $c_h-c_i=(h-i)(2n-1)$.
It follows that \ $c_h-c_i \notin $ \ for every $i =3,\dots,h-1$.
Thus, for every $h > 3$, \ $c_h \notin $
\ if and only if \ $c_h \notin $.
Therefore, in order to prove that $S_n$ is minimally generated
by $a,b,d,c_3,c_4,\dots,c_n$, it suffices to show that
$c_h \notin $, for every $h =3,\dots,n$.
Since $c_h \equiv n-(h-1)$ (mod $2n$), we compute the smallest
element in $$ that is congruent to $n - (h-1)$ modulo $2n$
and observe that it is larger than $c_h$.
We have that $a\alpha+b\beta+d\delta \equiv -\beta+n\delta$ (mod $2n$).
Hence we want to compute the minimum of the set
$H=\{b\beta+d\delta \ | \ \beta, \delta \geq 0,
\ -\beta+n\delta \equiv n-(h-1)\}$.
We claim that $\min H=(h-1)b+d$ (i.e. we have the minimum for
$\beta=(h-1)$ and $\delta = 1$).
Assume, by way of contradiction, that there exists an
integer $b\beta+d\delta \equiv n-(h-1)$
such that $b\beta+d\delta < (h-1)b+d$.
This implies $\beta < (h-1)$ or $\delta =0$.
If $\delta=0$, then $b\beta \equiv -\beta \equiv n-(h-1)$, that is
$\beta \equiv -n+(h-1) \equiv n+(h-1)$; this implies
$\beta = n+h-1 +m(2n)$, with $m \in \Bbb N$. But $nb > d$,
thus $ (n+h-1 +m(2n))b > (h-1)b+d$, a contradiction.
If $\beta < (h-1)$, then $b\beta+d\delta \equiv -\beta+n\delta
\equiv n-(h-1)$. But this implies that
$n(\delta-1) \equiv \beta-(h-1)$, which is a contradiction,
since $n(\delta-1) \equiv n$ or $n(\delta-1) \equiv 0$
and $-n <\beta-(h-1)<0$.
We conclude that $(h-1)b+d$ is the minimum element $x \in
$ such that $x \equiv n-(h-1)$. Since for every $h=3,\dots,n$,
we have
$(h-1)b+d= (h-1)(4n-1)+n(2n-1) = 2n^2+n(4(h-1)-1)-h+1
>2n^2+n(2(h-1)+1)-h+1=c_h$, it follows that
$c_h \notin $. Therefore $S_n$ is minimally
generated by $a$, $b$, $d$, $c_3, \dots,c_n$. \qed
\medskip
The {\it Frobenius number } $g(S_n)$ of $S_n$ is by definition
the largest integer not belonging to $S_n$.
We next compute $g(S_n)$ :
since $c_n$ is a generator of $S_n$, we have $c_n -a \notin S_n$.
Moreover all the elements of the set $\{c_n-a+1, c_n-a+2, \dots , c_n \}$
belong to $S_n$:
\roster
\item $c_n -a+1=(2n-1)(2n-1)+1 = c_{n-1}$;
\item for all $m =2, \dots n-3$,
$c_n-a+m= 2n(2n-1)+1-2n+m \equiv m+1 \equiv
n-(n-m)+1 \equiv c_{n-m}$ (mod $a$) \ and \
$c_n-a+m > c_{n-m}$;
\item $c_n-a+n-2= 2n(2n-1)-n-1 \equiv b+d$ (mod $a$) \ and \
$c_n-a+n-2 >b+d$;
\item $c_n-a+n-1= 2n(2n-1)-n \equiv d$ (mod $a$) \ and
\ $c_n-a+n-2 >d$;
\item for every $m =0,\dots , n-2$, $c_n-a+n+m = 2n(2n-1)+1-n+m
= (2m+1))a+(n-m-1)b$;
\item $c_n-a+2n-1= 2n(2n-1)=2d$;
\item $c_n-a+2n=c_n$.
\endroster
Therefore $S_n$ contains all integers greater than or equal to
$c_n-a+1 =c_{n-1}$ and $g(S_n) = c_n-a$.
\vskip 2ex
Consider the semigroup ideal $I_n : = $ of
$S_n$.
\vskip 2ex
\noindent
{\bf Claim 2.3.2: } For every integer $m \ge 1$,
the ideal $mI_n$ is minimally generated by $n$ elements.
\proof
Since the elements $a, b, c_3, \dots, c_n$ are part of the
minimal set of generators of $S_n$, it follows at once that
$I_n$ is minimally generated by $a$, $b$, $c_3,\dots c_n$.
We prove that $2I_n$ is minimally generated by
$2a$, $a+b$, $2b$, $a+c_4,\dots, a+c_n$.
We have the following relations:
$$
a+c_3=2b+d \tag{$R_1$}
$$
and, for every $h=4, \dots n$,
$$
a+c_h=b+c_{h-1}. \tag{$R_2$}
$$
Moreover, since $b+c_n -2a >c_n -a =g(S_n)$, for some $s \in S_n$ we have
$$
b+c_n =2a+s. \tag{$R_3$}
$$
The same is true for $c_h+c_k$, for all $h,k =3,\dots, n$,
since $c_h+c_k \geq 2c_3 >b+c_n$: for some $t_{h,k} \in S_n$,
$$
c_h+c_k = 2a+t_{h,k}. \tag{$R_4$}
$$
It follows that $2I_n =<2a, a+b, 2b, a+c_4,\dots, a+c_n>$.
We check that this is the minimal set of generators
(i.e. there are not other relations).
Clearly we have $a+b-2a=2b-(a+b) =b-a \notin S_n$.
Notice that $y(b-a) \notin S_n$ for every $y $,
since $y(b-a) $
such that $x \equiv -y$ (mod $a$).
It follows that $2b-2a \notin S_n$.
Moreover, for every $h=4,\dots, n$,
since $c_h$ is part of the minimal
set of generators of $I_n$, we have $a+c_h - 2a= c_h-a \notin S_n$,
$a+c_h-(a+b)=c_h -b \notin S_n$,
$a+c_h-2b=b+c_{h-1}-2b = c_{h-1}-b \notin S_n$ and,
for $kn$ means that
$(m+1)I_n =(n-1)I_n$ is minimally generated by
$(n-1)a$, $(n-2)a+b$, \dots $(n-1)b$).
Using relations $(R_1)$ and $(R_2)$,
notice that $a+(m-1)a+c_{m+2}= a+(m-1)b+c_3 = (m+1)b+d$.
Moreover, for every $h= m+2, \dots, n-1$, we have
$b+(m-1)a+c_h= ma+c_{h+1}$ (using $(R_2)$) and $b+(m-1)a+c_n=(m+1)a+s$
(using $(R_3)$).
Now we consider the elements $c_k+ra+sb$ and $c_k+(m-1)a+c_h$
(where $k=3,\dots n$, $r+s=m$ and $s>0$).
We easily get, by $(R_4)$, that $c_k+(m-1)a+c_h= (m+1)a+t_{k,h}$;
moreover, if $k+s \leq n$, using $(R_2)$, we get $c_k +ra+sb= c_{k+s}+ma$,
while, if $k+s > n$, using $(R_2)$ and $(R_3)$,
$c_k+ra+sb = c_n+(r+n-k)a+(s-(n-k))b=
(r+n-k+2)a+(s-(n-k)-1)b+s$.
Hence $(m+1)I_n =$.
In order to prove that $(m+1)I_n$ is minimally generated by these
elements, we show that there are no other relations.
If $r+s=u+v=m+1$ and $s>v$, the difference $ra+sb-(ua+vb)=
(s-v)(b-a) \notin S_n$, since $s-v $.
Now, since
$$
nb=na+d, \tag{$R_5$}
$$
we have, with similar arguments as above, that
$nI_n$ is minimally generated by $na$, $(n-1)a+b$, \dots, $nb$,
that is $nI_n= (n-1)I_n+a$.
Hence, for every $m \geq n$, also $mI_n$ is minimally $n$-generated;
moreover the reduction number of $I_n$ is $r(I_n)=n-1$.
This completes the proof of Claim 2.3.2 and
the presentation of Example \FPnoIneq. \qed
\vskip 2ex
In Example \negative, \me we exhibit a 3-dimensional
Cohen--Macaulay local domain $(R,\m)$ of multiplicity 3
having, for each positive integer $k$, an ideal $I_k$
such that $I_k$ has a principal reduction, $I_k$ is
minimally generated by $k + 3$ elements, and the
fiber cone $F(I_k)$ has multiplicity 3 and is not Cohen--Macaulay.
\vskip 3ex
\ex
\label{\negative}
Let $(S, \n)$ be a two-dimensional regular local ring and let
$R= S[t^3, t^4, t^5]_{(\n, t^3, t^4, t^5)}$, where $t$ is an
indeterminate over $S$.
For $k$ a fixed positive integer, let
$I_k =(t^3, t^4, \n^k t^5)R$. Since $\n^k$ is minimally generated
by $k+1$ elements as an ideal in $S$, we see that $I_k$ is
minimally generated by $k + 3$ elements.
Notice that $t^3I_k \subsetneq I_k^2 = (t^6, t^7, t^8)$ and that
$I_k^3 = t^3 I_k^2$. Thus $t^3$ is a principal reduction of $I_k$
and $I_k$ has reduction number
$r(I_k) = 2$. Moreover, $I_k^2$ is minimally generated by
$t^6, t^7, t^8$, so the multiplicity of the
fiber cone $F(I_k)$ is $f_0 = 3$ and $f_1 = 6 - (1 + k + 3)$.
Since $k$ can be arbitrarily large, there is no lower bound
on $f_1$ as we vary over ideals of $R$ having a principal reduction.
Since $t^3\n^k t^5 \subseteq \m I_k^2$, where $\m$ is the maximal
ideal of $R$, we see that the image of $t^3$ in $I_k/\m I_k = F_1$
in $F(I_k)$ is a zero-divisor. Therefore
Remark \cmcheck \me as given below implies that $F(I_k)$ is not
Cohen--Macaulay and thus has depth zero.
\vskip 3ex
\section{The Cohen--Macaulay property of the fiber cone.}
\vskip 3ex
Interesting work on the Cohen--Macaulay property of the
fiber cone $F(I)$ has been done by K. Shah in
\cite{S}, by T. Cortadellas and S. Zarzuela in \cite{CZ}
and by C. D'Cruz, K. N. Raghavan and J. K. Verma in \cite{DRV}.
In particular, we use the following:
\vskip 2ex
\rmk
\label{\cmcheck}
As Shah notes in \cite{S}, the freeness lemma of
Hironaka \cite{N, (25.16)} implies that $F(I)$ is
Cohen--Macaulay if and only if $F(I)$ is a free module
over one (or equivalently every) Noether normalization
subring. If $x$ is a principal reduction of $I$
and $\overline x$ denotes the image of $x$ in
$I/\m I = F_1$, then $K[\overline x]$ is a
Noether normalization of $F(I)$. Since $K[\overline x]$ is
a principal ideal domain, $F(I)$ is a free $K[\overline x]$-module
if and only if it is a torsionfree $K[\overline x]$-module.
Since $F(I)$ is graded and $\overline x$ is homogeneous, we
see that $F(I)$ is torsionfree as a $K[\overline x]$-module if
and only if $\overline x$ is a regular element of $F(I)$.
Therefore $F(I)$ is Cohen--Macaulay if and only if
$\overline x$ is a regular element of $F(I)$.
\vskip 2ex
It follows from Remark \cmcheck \me that for the ideal
$I_n=(t^a, t^b, t^{c_3}, \dots, t^{c_n})$
of the ring $R_n=K[\![t^s:s\in S_n ]\!]$ given in Example \FPnoIneq, \me
the fiber cone $F(I_n)$ is not Cohen-Macaulay:
the relation $(R_1)$ of Example \FPnoIneq \me
implies that $t^at^{c_3} \in I_n^2\m$. Therefore
the image in $F(I_n)$ of $t^a$
(which is a principal reduction of $I_n$) is a zerodivisor in $F(I_n)$.
The fact that $F(I_n)$ is not Cohen-Macaulay also follows from
Proposition \FPnoeq.
\vskip 2ex
Proposition \FPnoeq \me as given below
is a consequence of results of Cortadellas and Zarzuela \cite{CZ}
and D'Cruz, Raghavan and Verma \cite{DRV} on the structure of
the Hilbert function of a Cohen--Macaulay fiber cone $F(I)$.
One can use \cite{BH, Proposition 4.1.12} to see the
asserted relation between the fiber postulation number and
reduction number from their results. We give a direct elementary
proof of the result.
\vskip 2ex
\prop
\label{\FPnoeq}
Suppose $(R,\m)$ is a Noetherian
local ring and $I$ is a nonprincipal regular ideal of $R$ having
a principal reduction $xR$. If $F(I)$ is
Cohen--Macaulay, then $fp(I) = r(I) - 1$.
\endb
\proof
Let $\overline x$ denote the image of $x$ in
$I/\m I = F_1$. As noted in Remark \cmcheck, \me
$F(I)$ is Cohen--Macaulay if and only if $\overline{x}$ is
a regular element of $F(I)$. By
Proposition \FPnoRedno, \me $fp(I) \le r(I) - 1$, and
by definition of
$r(I) = r$, we have $xI^{r-1} \subsetneq I^r$.
Assume, by way of contradiction, that $fp(I) < r - 1$. This means
that $\dim F_{r-1} = f_0 = \dim F_r$. Since
$\overline x$ is a regular element of $F(I)$,
$f_0 = \dim \overline{x}F_{r-1}$. Since
$\overline{x}F_{r-1} \subseteq F_r$, this
implies $\overline{x}F_{r-1} = F_r$. However,
this implies $xI^{r-1} = I^r$,
a contradiction. \qed
\vskip 3ex
\rmk
\label{\converse}
The converse of Proposition \FPnoeq \me is not true in general.
For example, let $R = K[t^4, t^5, t^{11}]_{(t^4, t^5, t^{11})}$,
where $t$ is an indeterminate over the field $K$.
Let $I = \m$, the maximal ideal of $R$. Since the image of
$t^4 \in I/I^2$ is a zero-divisor, $F(I) = G(I)$ is not
Cohen--Macaulay. On the other hand, $fp(I) = n(I) = 2$
and $r(I) = 3$, so $fp(I) = r(I) - 1$.
Other examples illustrating the failure of the converse of
Proposition \FPnoeq \me are the ideals $I_k$ of
Example \negative. \me In these examples one
has $fp(I_k) = 1$ and $r(I_k) = 2$.
\vskip 3ex
\dis
\label{\fibercm}
Shah in \cite{S} proves several interesting
results on the Cohen--Macaulay property of $F(I)$. In \cite{S, Theorem 1}
he proves that if $I$ is an ideal of a Noetherian local ring $(R,\m)$ and
if $I$ is integral over an ideal generated by a regular sequence
$\underline x$ such that $I^2 = (\underline{x})I$, then $F(I)$ is
Cohen--Macaulay. As Shah notes, it was proved earlier by C. Huneke
and J. Sally in \cite{HS, Proposition 3.3} that if $(R,\m)$ is a
Cohen--Macaulay local ring and $I$ is an $\m$-primary ideal
such that $I^2 = (\underline{x})I$, where $\underline x$ is a
regular sequence, then $F(I)$ is Cohen--Macaulay.
It is immediate from these results that, if $I$ is a regular
ideal having a principal reduction $xR$ and if $r(I) \le 1$,
then the fiber cone $F(I)$ is Cohen--Macaulay. In \cite{S, Theorem 2},
Shah proves that if $I$ is an ideal of a Noetherian local ring $(R,\m)$ and
if $I$ is integral over an ideal generated by a regular sequence
$\underline x$ such that $I^3 = (\underline{x})I^2$,
$I^2 \cap (\underline{x}) = I(\underline{x})$, and
$I^2\m = I(\underline{x})\m$, then $F(I)$ is Cohen--Macaulay.
In \cite{CZ; Theorem 3.2}, Cortadellas and Zarzuela generalize Shah's
result and show that if $I$ is an ideal and $J$ a minimal reduction of $I$
generated by a regular sequence with $J \cap I^n = J I^{n-1}$ for all
$1 \le n \le r_J (I)$, then $F(I)$ is Cohen--Macaulay if and only if
$J \cap \m I^n = J \m I^{n-1}$ for all
$1 \le n \le r_J (I)$.
\vskip 2ex
\prop
\label{\twogen}
Suppose $I$ is a regular ideal having a principal reduction $(x)$
in a Noetherian local ring $(R,\m)$. If
$I$ is 2-generated, then $F(I)$ is Cohen--Macaulay.
\endb
\proof
Consider a presentation $\phi$ of $F(I)$ as a
graded $K$-algebra homomorphic
image of the polynomial ring $K[X,Y]$, where
$\phi(X)$ and $\phi(Y)$ are the images in $I/\m I = F_1$
of two generators of $I$. To show $F(I)$ is Cohen--Macaulay
it suffices to show that $\ker \phi$ is principal.
By a result of P. Eakin and A. Sathaye \cite{ES, page 440},
if $I^n$ is generated by less than $n+1$ elements, then
$xI^n = I^{n+1}$. Hence if $r = r(I)$ is the reduction
number of $I$, then for each positive integer $n \le r$
the ideal $I^n$ is minimally generated by $n+1$
elements. Thus the multiplicity of $F(I)$ is $r+1$ and
$\dim F_n = r+1$ for each $n \ge r$.
Therefore the minimal degree of a form
$f \in \ker \phi$ is $r+1$ and there exists a form
$f \in \ker \phi$ with $\deg f = r+1$. It
follows that $\ker \phi$ is generated by $f$.
Therefore $F(I)$ is a complete intersection and
hence Cohen--Macaulay.
\qed
\vskip 2ex
\cor
\label{\mult2}
Suppose $(R, \m)$ is a one-dimensonal Cohen--Macaulay local ring
of multiplicity 2.
For each $\m$-primary ideal $I$ of $R$,
the fiber cone $F(I)$ is Cohen--Macaulay.
\endb
\proof
Since $R$ has multiplicity 2, the integral closure of $R$
has at most 2 maximal ideals and each $\m$-primary ideal $I$
of $R$ is 2-generated. Therefore $I$
has a principal reduction and by Proposition~\twogen, \me
$F(I)$ is Cohen--Macaulay. \qed
\vskip 2ex
\ques
For which one-dimensional Cohen--Macaulay local rings
$(R, \m)$ is it true that $F(I)$ is Cohen--Macaulay
for every $\m$-primary ideal $I$? Is this true if
$R$ has multiplicity 3 ?
\vskip 2ex
\ex
\label{\all}
Let $R = K[\![t^3, t^4, t^5]\!]$. Then $K[\![t]\!]$
is the integral closure $\overline R$ of $R$
and the maximal ideal $\m = (t^3, t^4, t^5)R$ is the
conductor of $\overline R$ into $R$.
If $I$ is an $\m$ primary ideal of $R$, then either
(i) $I$ is principal, or (ii) $I$ is minimally 2-generated,
or (iii) $I$ is minimally 3-generated and $I = I\overline R$.
In this third case, $I = t^s\overline R$ for some integer
$s \ge 3$. It follows that $t^sI = I^2$ and $I$ has reduction
number one as an ideal of $R$. Therefore the fiber cone
$F(I)$ is Cohen--Macaulay for each $\m$-primary ideal $I$
of $R$.
\vskip 2ex
\rmk
For $R$ as in Example \all, \me there exist $\m$-primary ideals
$I$ of $R$ such that the associated graded ring $G(I)$ is not
Cohen--Macaulay. For example, $I = (t^3, t^4)$ is not a
Ratliff-Rush ideal since $I \subsetneq \m$ and $I^2 = \m^2$.
Therefore $I$ is an example of an ideal
for which $F(I)$ is Cohen--Macaulay and $G(I)$ is not
Cohen--Macaulay.
\vskip 2ex
\rmk
\label{\numerical}
Suppose $(R,\m)$ is a one-dimensional Cohen--Macaulay local
ring and $I$ is an $\m$-primary ideal having a principal
reduction $xR = J$ with reduction number $r$.
As a special case of \cite{DRV, Theorem 2.1}
it follows that the fiber cone $F(I)$ is Cohen--Macaulay
if and only if the Hilbert series
$P(t) = \sum_{n=0}^\infty \lambda(I^n/\m I^n)t^n$ has the
form $h(t)/(1-t)$, where
$h(t) = \sum_{i=0}^r\lambda(I^i/(JI^{i-1} + \m I^i)$. It is
not true, however, that the Cohen--Macaulay property of
the fiber cone $F(I)$ is determined by its Hilbert series
$P(t)$. For example, if $R=K[\![t^6,t^{11},t^{15}, t^{31} ]\!]$
and $I=(t^6, t^{11}, t^{31})$ are as
in Example \FPnoIneq, where $R=K[\![t^s : s \in S_3]\!]$
and $I$ corresponds to the semigroup ideal $I_3$ of $S_3$,
\me $F(I)$ is not Cohen--Macaulay
and the Hilbert series for $I$ is $P(t) = (1+2t)/(1-t)$.
This is also the Hilbert series of a Cohen--Macaulay
fiber cone; for example it is the Hilbert series of
the maximal ideal of $R = K[\![ x,y,z ]\!]$, where $y^2 = yz = z^2 = 0$.
\vskip 3ex
\section{The multiplicity of the fiber cone. }
\vskip 3ex
Suppose $I$ is a regular ideal of $R$ having a principal
reduction $xR$. In this situation, the blowing up
scheme $\Proj R[It]$ of $\Spec R$ with respect to $I$
is affine and has the form $\Spec R[I/x]$.
Therefore we refer to the ring $R^I : = R[I/x]$ as
the {\it blowing--up ring}
of $I$. It is simple to prove that
$R^I := R[I/x] = \bigcup_{n\ge 0} I^n/x^n = I^r / x^r$
where $r = r(I)$. Moreover, the blowing up ring $R^I$ of $I$
is also the blowing up ring of each power $I^n$ of $I$.
\vskip 3ex
\lemma
\label{\mult}
The multiplicity $f_0$ of $F(I)$ is the
minimal number of generators of $R^I$ as an $R$-module.
Thus $f_0 = \la(R^I/\m R^I)$.
\endb
\proof
We know that $f_0 = \la(I^r/ \m I^r)$. In other words the multiplicity of $F(I)$
is given by the cardinality of a minimal set of generators for $I^r$
as an $R$-module. Moreover, the fact that $x$ is a regular
element of $R$ implies that if $\{a_1, \ldots, a_{f_0}\}$
is a minimal set of generators for $I^r$, then
$\{a_1/x^r, \ldots, a_{f_0}/x^r\}$ is a minimal set
of generators of $I^r/x^r$ as an $R$-module. \qed
\vskip 2ex
\rmk
\label{\tangent}
Suppose $(R,\m)$ is a one-dimensional Cohen--Macaulay local
ring and $I$ is an $\m$-primary ideal having a principal
reduction $xR$. The associated graded ring $G(I)$ is
Cohen--Macaulay if and only if $I^n = \widetilde{I^n}$ for
each $n \in \N$. Moreover, $\widetilde{I^n} = I^nR^I \cap R$,
where $R^I = R[I/x]$ is the blowing up ring of $I$. We use this
to observe that $G(I)$ is Cohen--Macaulay if
$R=K[\![t^6,t^{11},t^{15},t^{31}]\!]$ and $I=(t^6,t^{11},t^{31})$
are as in Example \FPnoIneq, \me where $R = K[\![t^s : s \in S_3]\!]$
and $I$ corresponds to the semigroup ideal $I_3$ of $S_3$.
Since $r(I) = 2$, it follows from \cite{DGH, Proposition 2.2}
that the asymptotic Ratliff-Rush number $h(I) \le 2$.
Thus $\widetilde{I^n} = I^n$ for each $n \ge 2$. Hence it
suffices to show that $I = IR^I \cap R$. Observe that
$R^I = R[t^5] = K[\![t^5, t^6]\!]$, $IR^I = t^6R^I$,
and $\lambda(R/I) = 2$.
Thus it suffices to show that $t^{15} \not\in t^6K[\![t^5, t^6]\!]$,
and this is clear since $t^9 \not\in K[\![t^5, t^6]\!]$.
We remark that for this ring $R = K[\![t^6, t^{11}, t^{15}, t^{31}]\!]$,
the associated graded ring $G(\m) = F(\m)$ of the maximal ideal
$\m$ is not Cohen--Macaulay. This is readily seen from the fact
that the image of $t^6$ in $\m/\m^2$ in $G(\m)$ is a zero divisor
of $G(\m)$ since $t^{31} \not\in \m^2$ and $t^6t^{31} =
t^{15}(t^{11})^2 \in \m^3$.
\vskip 3ex
\dis
\label{\fibermult}
Suppose $(R,\m)$ is a one-dimensional Cohen--Macaulay local
ring of multiplicity $e$. It is well known that every
$\m$-primary ideal $I$ of $R$ can be generated by $e$
elements \cite{SV, Theorem 1.1}.
Therefore $F(I)$ has multiplicity at most $e$ for each
$\m$-primary ideal $I$ of $R$. Moreover,
high powers of $\m$ require $e$ generators and
$F(\m) = G(\m)$ has multiplicity $e$. Also
there exist $\m$-primary principal
ideals of $R$ and it is clear that if $I$ is a principal
$\m$-primary ideal, then $F(I)$ has multiplicity one.
Thus the integers $1$ and $e$ are the multiplicity of
$F(I)$ for $\m$-primary ideals $I$ of $R$.
It is natural to ask what integers $f$ with
$1 \le f \le e$ are realized as the
multiplicity of $F(I)$ for some
$\m$-primary ideal $I$ of $R$. We exhibit in
Example \onlytwo \me the existence, for each positive
integer $e \ge 2$, of a one-dimensional
Cohen--Macaulay local domain $(R,\m)$ having
multiplicity $e$ such that for each $\m$-primary
ideal $I$ of $R$, either $I$ is principal and $F(I)$ has
multiplicity one, or else $F(I)$ has multiplicity $e$.
On the other hand, we observe in Example \every \me
the existence of a one-dimensional Cohen--Macaulay
local domain $(R, \m)$ of multiplicity $e$
such that for every integer $f$ with $1 \le f \le e$
there exists an $\m$-primary ideal $I$ with
$F(I)$ having multiplicity $f$.
\vskip 2ex
\ex
\label{\onlytwo}
Fix a positive integer $e \ge 2$ and let $K/E$ be an algebraic
extension of fields such that $[K:E] = e$ and such that
there are no fields properly between $E$ and $K$. Let
$t$ be an indeterminate over $K$, let $S = K[t]_{(t)}$,
and $R = E + tS$. Then $S$ is the
integral closure of $R$ and $S$ is the only subring of $S$
that properly contains $R$. Hence if $I$ is a nonprincipal
$\m$-primary ideal of $R$, then $R^I = S$, so $F(I)$ has
multiplicity $e$.
\vskip 2ex
\ex
\label{\every}
Fix a positive integer $e \ge 2$ and let $K$ be a field.
Let $t$ be an indeterminate over $K$ and let
$R = K[t^e, t^{e+1}, \ldots, t^{2e-1}]_{(t^e, t^{e+1}, \ldots,t^{2e-1})}$.
Fix an integer $k$ with $1 \le k < e$. Let
$I = (t^e, t^{2e-k}, t^{2e-k+1}, \ldots, t^{2e-1})$.
Then $R^I$ as an $R$-module is minimally generated by
the $k+1$ elements $1, t^{e-k}, t^{e-k+1}, \ldots, t^{e-1}$.
Therefore by Lemma \mult \me $F(I)$ has multiplicity $f := k+1$
\vskip 2ex
In Example \highmult \me we exhibit a 3-dimensional
Cohen--Macaulay local domain $(R,\m)$ of multiplicity 2
with the following property. For
each positive integer $k$, the ring $R$ has a
stable ideal $I_k$ having a principal reduction
such that the fiber cone $F(I_k)$
of $I_k$ is Cohen--Macaulay and has multiplicity $k + 2$.
Therefore, in contrast with the case where $\dim R = 1$, for
$R$ of higher dimension there may exist no upper
bound on the multiplicity of $F(I)$ as $I$ varies
over the ideals of $R$ that have a principal reduction.
\vskip 3ex
\ex
\label{\highmult}
Let $(S, \n)$ be a two-dimensional regular local ring and let
$R= S[t^2, t^3]_{(\n, t^2, t^3)}$, where $t$ is an
indeterminate over $S$.
Fix a positive integer $k$, and let
$I=(t^2, \n^k t^3)R$. Since $\n^k$ is minimally generated
by $k+1$ elements as an ideal in $S$, we see that $I$ is
minimally generated by $k + 2$ elements. Moreover,
$I^2 = (t^4, \n^kt^5) = t^2I$. Thus $I$ has a principal
reduction and is a stable ideal, i.e., $r(I) = 1$. It
follows that $F(I)$ is Cohen--Macaulay and has multiplicity
$f_0 = k + 2$.
\vskip 2ex
\section{A bound for the reduction number.}
\vskip 2ex
Suppose $(R,\m)$ is a one-dimensional Cohen--Macaulay
local ring of multiplicity $e$. It is well known from
results of Sally-Vasconcelos \cite{SV} and
Eakin-Sathaye \cite{ES} that $e-1$ is then an upper
bound for the reduction number $r(I)$ of $\m$-primary
ideals of $R$ having a principal reduction. In a
higher dimensional situation, S. Huckaba proves in
\cite{Hu1, Theorem 2.5} that if $(R,\m)$ is a quasi-unmixed
analytically unramified local ring containing an infinite
field of characteristic $\ne 2$ and if the multiplicity of
$R$ is 2, then
each regular ideal $I$ of $R$ of analytic spread one
has reduction number $r(I) \le 1$. Now the two-dimensional
regular local domain $S$ of Example \highmult \me can
be chosen so that $R$ as in Example \highmult \me
satisfies all these properties. It then follows that
for ideals $I$ of $R$ having a principal reduction there
is no upper bound on the multiplicity of $F(I)$, but the
reduction number $r(I) \le 1$ for each $I$.
Huckaba \cite{Hu1, Question 2.6} raises the interesting
question of
whether an analogous result to \cite{Hu1, Theorem 2.5}
holds in the case where $R$ has
multiplicity $e > 2$. To illustrate a special case of
this question, suppose $(R,\m)$ is a complete
Noetherian local domain of dimension $d$ containing
an infinite coefficient field $K$. Then there
exist $x_1, \dots x_d \in \m$ that form a system
of parameters for $R$ and generate a reduction
of $\m$. Since $R$ is complete $A := K[[x_1, \dots ,x_d]]$
is a $d$-dimensional regular local subring of $R$
and $R$ is a finitely generated $A$-module. Moreover,
the multiplicity $e$ of $R$ is precisely the
degree of the fraction field extension $[Q(R):Q(A)]$
\cite{ZS, Corollary 2, page 300}. In this situation,
$\lambda(R/(x_1, \dots,x_d)R) \ge e$ is the number
of generators for $R$ as an $A$-module and $R$ is
Cohen--Macaulay if and only if $R$ is a free
$A$-module if and only if $\lambda(R/(x_1,\dots,x_d)R) = e$.
Since $R$ is complete, the integral closure $\overline R$
of $R$ is again local and is a finitely generated $R$-module
and thus also a finitely generated $A$-module.
The freeness lemma of Hironaka \cite{N, (25.16)}
implies that $\overline R$ is a free $A$-module if and only
if $\overline R$ is Cohen--Macaulay if and only if
$\lambda(\overline R/(x_1,\dots,x_d)\overline R) = e$. If
$\overline R$ has residue field $K$, then $e$ is also the
multiplicity of $\overline R$, but if the residue field
of $\overline R$ is a proper extension of the residue
field $K$ of $R$ with respect to the canonical inclusion map
of $R$ into $\overline R$, then $\overline R$ has multiplicity
less than $e$.
The following interesting observation shown to us by Craig
Huneke proves the existence of a global bound
on the reduction number $r(I)$ of ideals having a
principal reduction in certain rings $R$.
\vskip 2ex
\prop
\label{\huneke}
Suppose $I$ is a regular ideal of a ring $R$ and that
$xR$ is a principal reduction of $I$. If there
exists an $n$-generated faithful $R$-module
$M = $ such that $IM = xM$, then
$I$ has reduction number $r(I) \le n-1$.
\endb
\proof
Let $u_1, \ldots, u_n$ be (not necessarily distinct) elements
of $I$. Since $IM = xM$, we obtain $n$ equations
$u_i m_i = x\sum_{j=1}^n a_{ij} m_j$, where the $a_{ij} \in R$.
Rearranging to make the system of equations homogeneous,
we obtain a coefficient matrix $A$
that applied to the column vector $(m_1, \ldots, m_n)^T$ gives the zero vector.
By multiplying $A$ by its adjoint, one sees that $\det(A) \in R$ is
in $(0:_R M) = (0)$. Thus $\det(A) = 0$.
The explicit expression of $\det(A)$
shows that $u_1 \cdots u_n \in x I^{n-1}$. In conclusion
$I^n = xI^{n-1}$ and $r(I) \le n - 1$. \qed
\vskip 2ex
In general, if $\overline R$ is the integral closure
of an integral domain $R$, then principal ideals of $\overline R$ are
integrally closed. Thus if $I$ is an ideal of $R$
having a principal reduction $xR$, then
$x\overline R = I\overline R$. Thus
the following corollary to Proposition \huneke \me is immediate.
\vskip 2ex
\cor
\label{\globalbound}
Suppose $R$ is an integral domain and that the integral closure
$\overline R$ of $R$ in its fraction field is $n$-generated as
an $R$-module. Then each ideal $I$ of $R$ having a
principal reduction has reduction number $r(I) \le n - 1$
\endb
We record several other results that follow from
Proposition \huneke.
\vskip 2ex
\cor
\label{\multbound}
Suppose $(R,\m)$ is a Noetherian local ring and $I$ is
a regular ideal of $R$ having a principal reduction
$xR$. Let $R^I = R[I/x]$ denote the blowing up ring of $I$.
Then the reduction number $r(I) \le \lambda(R^I/\m R^I) - 1$,
so $r(I) \le f_0 - 1$, where $f_0$ is the multiplicity of
the fiber ring $F(I)$.
\endb
\proof
Since $R^I$ is a faithful $R$-module and $IR^I = xR^I$,
Proposition \huneke \me implies the first assertion.
The second assertion follows from Lemma \mult. \qed
\vskip 2ex
\cor
\label{\cohenint}
Suppose $(R,\m)$ is a complete local
domain of multiplicity $e$ containing an infinite field.
If the integral closure $\overline R$ of $R$ is Cohen--Macaulay,
then every ideal $I$ of $R$ of analytic spread one has
reduction number $r(I) \le e - 1$.
\endb
\proof
Since $R$ is complete, $R$ contains a coefficient field $K$.
If $(x_1, \ldots,x_d)$ is a minimal reduction of $\m$, then
$A := K[[x_1, \ldots, x_d]]$ is a $d$-dimensional regular
local subring of $R$, $R$ is a finitely generated $A$-module
and $[Q(R): Q(A)] = e$. Since $\overline R$ is Cohen--Macaulay,
$\overline R$ is a free $A$-module on $e$ generators.
Therefore $\overline R$ is $e$-generated as
an $R$-module. By Proposition \huneke, \me
the reduction number $r(I) \le e-1$
for each ideal $I$ of $R$ having a principal reduction,
or equivalently in our setting, each ideal of analytic
spread one. \qed
\vskip 2ex
\rmk
\label{\analyticirr}
There are other situations to which
Proposition \huneke \me applies.
Suppose $(R,\m)$ is a Noetherian local ring and let
$\widehat R$ denote the $\m$-adic completion of $R$.
If $I$ is a regular ideal of $R$ having a principal
reduction $xR$, then $x\widehat R$ is a principal
reduction of $I\widehat R$. Moreover,
$xI^n = I^{n+1}$ if and only if $xI^n\widehat R =
I^{n+1}\widehat R$. Thus $r(I) = r(I\widehat R)$.
Also $R$ and $\widehat R$ have the same multiplicity.
Thus if $(R,\m)$ is an analytically irreducible
Noetherian local domain of multiplicity $e$ and if
its completion $\widehat R$ contains an infinite
field and has a Cohen--Macaulay normalization,
then each ideal $I$ of $R$ of analytic spread one
has reduction number $r(I) \le e-1$.
\vskip 2ex
\cor
\label{\twodim}
Suppose $(R,\m)$ is a $2$-dimensional complete local
domain of multiplicity $e$ containing an infinite field.
Then every ideal $I$ of $R$ of analytic spread one has
reduction number $r(I) \le e - 1$.
\endb
\proof
Since a $2$-dimensional integrally closed Noetherian
domain is Cohen--Macaulay, this follows from
Corollary \cohenint. \me \qed
\vskip 2ex
\cor
\label{\analytictwodim}
Suppose $(R,\m)$ is a 2-dimensional Noetherian analytically
irreducible local domain containing an infinite field.
If $R$ has multiplicity $e$, then
every ideal $I$ of $R$ of analytic spread one has
reduction number $r(I) \le e - 1$.
\endb
\proof
This follows from Remark \analyticirr \me and Corollary \twodim.
\vskip 2ex
\rmk
\label{\serre}
It is known that a complete local ring that contains
a field and satisfies Serre's condition $S_n$ is
Cohen--Macaulay if it has multiplicity $\le n$ \cite{H2}.
Thus Corollary \cohenint \me yields in a special case
the result of Huckaba \cite{Hu1, Theorem 2.5} mentioned
above.
\vskip 3ex
\Refs
\widestnumber\key{1-10-A}
\ref
\key AM
\by M. F. Atiyah and I. G. Macdonald
\book Introduction to Commutative Algebra
\publ Addison-Wesley, London
\yr 1969
\endref
\ref
\key BH
\by W. Bruns and J. Herzog
\book Cohen--Macaulay rings, revised edition
\publ Cambridge University Press, Cambridge
\yr 1998
\endref
\ref
\key CZ
\by T. Cortadellas and S. Zarzuela
\paper On the depth of the fiber cone of filtrations
\jour J. Algebra
\vol 198
\pages 428--445
\yr 1997
\endref
\ref
\key DGH
\by M D'Anna, A. Guerrieri and W. Heinzer
\paper Invariants of ideals having principal reductions
\jour Comm. in Algebra (to appear)
\endref
\ref
\key DRV
\by C. D'Cruz, K. N. Raghavan and J. K. Verma
\paper Cohen--Macaulay fiber cones
\inbook Commutative algebra, algebraic geometry and computational methods
(Hanoi, 1996)
\pages 233--246
\publ Springer, Singapore
\yr 1999
\endref
\ref
\key ES
\by P. Eakin and A. Sathaye
\paper Prestable ideals
\jour J. Algebra
\vol 41
\pages 439--454
\yr 1976
\endref
\ref
\key HJLS
\by W. Heinzer, B. Johnston, D Lantz and K. Shah
\paper Coefficient ideals in and blowups of a commutative Noetherian domain
\jour J. Algebra
\vol 162
\pages 355--
\yr 1993
\endref
\ref
\key HJL
\by W. Heinzer, B. Johnston and D Lantz
\paper First coefficient domains and ideals of reduction number one
\jour Comm. Alg.
\vol 21(10)
\pages 3797--3827
\yr 1993
\endref
\ref
\key HLS
\by W. Heinzer, D Lantz and K. Shah
\paper The Ratliff--Rush ideals in a Noetherian ring
\jour Comm. Alg.
\vol 20(2)
\pages 591--622
\yr 1992
\endref
\ref
\key HZ
\by L.T. Hoa and S. Zarzuela
\paper Reduction number and a-invariant of good filtrations
\jour Comm. Alg.
\vol 22(14)
\pages 5635--5656
\yr 1994
\endref
\ref
\key HuS
\by R. H\"ubl and I. Swanson
\paper Normal cones of monomial primes
\paperinfo preprint
\yr 2000
\endref
\ref
\key Hu1
\by S. Huckaba
\paper Reduction numbers for ideals of analytic spread one
\jour J. Algebra
\vol 108
\pages 503--512
\yr 1987
\endref
\ref
\key Hu2
\by S. Huckaba
\paper Reduction numbers for ideals of higher analytic spread
\jour Proc. Camb. Phil. Soc.
\vol 102
\issue 49
\pages 49--57
\yr 1987
\endref
\ref
\key H1
\by C. Huneke
\paper Hilbert functions and symbolic powers
\jour Michigan Math. J.
\vol 34
\pages 293--318
\yr 1987
\endref
\ref
\key H2
\by C. Huneke
\paper A remark concerning multiplicities
\jour Proc. Amer. Math. Soc.
\vol 85
\pages 331--332
\yr 1982
\endref
\ref
\key HS
\by C. Huneke and J. Sally
\paper Birational extensions in dimension two and
integrally closed ideals
\jour J. Algebra
\vol 115
\pages 481--500
\yr 1988
\endref
\ref
\key I
\by S. Itoh
\paper Integral closures of ideals generated
by regular sequences
\jour J. Algebra
\vol 117
\pages 390--401
\yr 1988
\endref
\ref
\key L
\by J. Lipman
\paper Stable ideals and Arf rings
\jour Amer. J. Math.
\vol 93
\pages 649--685
\yr 1971
\endref
\ref
\key Mar
\by T. Marley
\paper The coefficients of the Hilbert polynomial and
the reduction number of an ideal
\jour J. London Math. Soc.
\vol 40
\pages 1--8
\yr 1989
\endref
\ref
\key M
\by E. Matlis
\book One-dimensional Cohen--Macaulay Rings,
Lecture Notes in Mathematics 327
\publ Springer-Verlag, New York
\yr 1973
\endref
\ref
\key Mat
\by H. Matsumura
\book Commutative ring theory
\publ Cambridge Univ. Press, Cambridge
\yr 1986
\endref
\ref
\key N
\by M. Nagata
\book Local Rings
\publ John Wiley, New York
\yr 1962
\endref
\ref
\key No
\by D. G. Northcott
\paper A note on the coefficients of the abstract Hilbert function
\jour J. London Math. Soc.
\vol 35
\yr 1960
\pages 209--214
\endref
\ref
\key NR
\by D. G. Northcott and D. Rees
\paper Reductions of ideals in local rings
\jour Proc. Camb. Phil. Soc.
\vol 50
\yr 1954
\pages 145--158
\endref
\ref
\key O
\by A. Ooishi
\paper Genera and arithmetic genera of commutative rings
\jour Hiroshima Math. J.
\vol 17
\yr 1987
\pages 47--66
\endref
\ref
\key RR
\by L. J. Ratliff, Jr. and D. E. Rush
\paper Two notes on reductions of ideals
\jour Indiana Univ. Math. J.
\vol 27
\yr 1978
\pages 929--934
\endref
\ref
\key RV
\by M.E. Rossi and G. Valla
\paper Cohen--Macaulay local rings of
embedding dimension $e+d-3$
\jour Proc. London Math. Soc.
\vol 80
\yr 2000
\pages 107--126
\endref
\ref
\key Sa
\by J. Sally
\paper Reductions, local cohomology and Hilbert functions
of local rings
\book Commutative Algebra: Durham 1981 - LMS 72
\publ Cambridge University Press
\pages 231--241
\endref
\ref
\key SV
\by J. Sally and W. Vasconcelos
\paper Stable rings
\jour J. Pure Appl. Algebra
\vol 4
\yr 1974
\pages 319--336
\endref
\ref
\key S
\by K. Shah
\paper On the Cohen--Macaulayness of the fiber cone of an ideal
\jour J. of Algebra
\vol 143
\yr 1991
\pages 156--172
\endref
\ref
\key VV
\by P. Valabrega and G. Valla
\paper Form rings and regular sequences
\jour Nagoya Math. J.
\vol 72
\yr 1978
\pages 93--101
\endref
\ref
\key V1
\by W.V. Vasconcelos
\book Arithmetic of Blowup Algebras - LMS 195
\publ Cambridge University Press
\yr 1994
\endref
\ref
\key V2
\by W. Vasconcelos
\paper The reduction number of an algebra
\jour Composito Math.
\vol 104
\yr 1996
\pages 189--197
\endref
\ref
\key W
\by Y. Wu
\paper Hilbert functions, reduction numbers and relation types
\jour Ph.D. Thesis, Purdue Univ.
\yr 1992
\endref
\ref
\key ZS
\by O. Zariski and P. Samuel
\book Commutative Algebra vol II
\publ D. Van Nostrand, Princeton
\yr 1960
\endref
\endRefs
\vskip 4ex
\centerline{\sl Universit\`a di Catania - Dipartimento di Matematica,}
\centerline{\sl Viale Andrea Doria, 6 - 95125 Catania, Italy}
\centerline{\sl E-mail: {mdanna\@dipmat.unict.it}}
\vskip 2ex
\centerline{\sl Universit\`a di L'Aquila - Dipartimento di Matematica,}
\centerline{\sl Via Vetoio, Loc. Coppito - 67100 L'Aquila, Italy}
\centerline{\sl E-mail: {guerran\@univaq.it}}
\vskip 2ex
\centerline{\sl Purdue University - Department of Mathematics,}
\centerline{\sl West Lafayette, Indiana 47907, USA}
\centerline{\sl E-mail: {heinzer\@math.purdue.edu}}
\footnote{ }{{\footitfont 1991 Mathematics Subject Classification.}
13A30, 13C05, 13E05, 13H15}
\footnote{ }{{\footitfont Key words and phrases.}
\footfont principal reduction, fiber cone, fiber postulation number,
multiplicity, Cohen--Macaulay.}
\enddocument