Complete intersection dimension pdf file
(with L. Avramov and V. Gasharov)
Publications Mathématiques IHES 86 (1997), 67-114.
ABSTRACT:
In the study of minimal free resolutions over
complete intersections the key idea is to relate a given infinite resolution
to a finite resolution over a regular ring. The minimal free resolution
of the residue filed over a complete intersection was
constructed by Tate.
We introduce a new homological
dimension: CI-dimension (complete
intersection dimension).
The class of modules of finite CI-dimension contains all modules of
finite projective dimension and all modules over a complete
intersection. We
obtain sharp quantitative and
structural homological data for the modules of finite
CI-dimension. Such modules provide the
first class of modules of infinite projective
dimension with
a rich structure theory of free resolutions.
For the study of these modules
we develop some new cohomological tools and
constructions over quantum symmetric
algebras (in particular, we construct
an extension of Manin's quantum Koszul complex).
We find a place in the minimal
free resolution of a module of finite CI-dimension
after which asymptotically stable behavior develops; in
particular, beyond this place the sequence of Betti numbers
is either constant or strictly increasing, and gaps between
consecutive numbers grow polynomially.
CI-dimension localizes and
stands between the projective dimension and
the Gorenstein dimension (introduced by
Auslander and Bridger). A fundamental homological result for
modules of finite projective dimension is the
Auslander-Buchsbaum Equality; we establish an analog to it
for modules of finite CI-dimension.
We also prove that a ring R is a complete intersection if and only if
its residue field has finite CI-dimension; this result is
an analog to Serre's
criterion for regularity.
Finite regularity and Koszul algebraspdf file
(with L. Avramov)
American Journal of Mathematics 123 (2001), 275-281.
ABSTRACT:
We consider positively graded commutative algebras
finitely generated over a field, and finitely generated graded modules.
The algebras over
which all modules have finite projective dimension are characterized by the
Auslander-Buchsbaum-Serre and Hilbert Syzygy Theorems, namely,
the following three conditions are
equivalent:
every module over the algebra
has finite projective dimension
the residue field
has finite projective dimension
the algebra is a
polynomial ring in finitely many indeterminates.
Castelnuovo-Mumford regularity is an important measure of the
complexity of the differential in a graded minimal free resolution.
Vanishing of the regularity
of the residue field defines the class of Koszul algebras, which
have received considerable
attention due to their extraordinary homological properties and
to their appearance in
many cases of interest in algebraic geometry,
algebraic topology, combinatorics, commutative algebra, and
representation theory.
We characterize those algebras
over which all modules have
finite Castelnuovo-Mumford regularity. We prove that the
following three conditions are equivalent:
every module over the algebra
has finite Castelnuovo-Mumford regularity
the residue field
has finite Castelnuovo-Mumford regularity
the algebra is a
polynomial ring in finitely many indeterminates over a Koszul algebra.
In particular, this shows that a
conjecture by Avramov and Eisenbud (1990) holds.
Boundedness versus periodicity over commutative local rings pdf file
(with V. Gasharov)
Transactions of the American Mathematical Society 320
(1990), 569-580.
This paper was written while V.
Gasharov and I were undergraduate students.
ABSTRACT:
The Betti numbers in an infinite
resolution grow polynomially or exponentially
in all cases when their behavior is known.
It is expected that the
boundedness of the Betti numbers is a strong condition.
A module with bounded Betti
numbers has eventually periodic minimal free resolution
if the ground ring is a group algebra, a complete intersection
ring, or a Gorenstein ring of small codepth.
Eisenbud conjectured (1980) that this property holds over any local
noetherian
ring. We provide counterexamples to Eisenbud's conjecture.
We construct modules over a Gorenstein artinian ring with arbitrary
period of the resolution, and also modules with
constant Betti numbers and no periodicity in the resolution at all.
On Hilbert Schemes:
Connectedness of Hilbert schemes pdf
file
(with M. Stillman)
Journal of Algebraic Geometry, 14 (2005), 193-211.
ABSTRACT:
Hartshorne proved that the Hilbert scheme,
that parameterizes subschemes of Pr with a fixed Hilbert
polynomial, is connected.
Recent works by D. Eisenbud, G. Fløystad, and F.-O. Schreyer
have inspired us to study Hilbert schemes over exterior algebras.
We show that the Hilbert scheme,
that parameterizes all ideals with the same Hilbert function over
an exterior algebra, is connected. We give a new proof of Hartshorne's
Theorem that the classical Hilbert scheme is connected.
Our construction is
different than Hartshorne's:
if Q is either
a polynomial ring or an exterior algebra,
we prove that every two strongly stable ideals in Q with the same Hilbert
function
are connected by a sequence of binomial gröbner deformations.
Flips and Hilbert schemes over an exterior algebra pdf
file (with M. Stillman)
Mathematische Annalen, 339 (2007), 545-557.
ABSTRACT:
We have
showed that the Hilbert
scheme, that parameterizes all ideals with a fixed Hilbert
function
over an exterior algebra, is connected.
Now, we
introduce the notion of flips and show that the basic
flips form a basis of the tangent space at a monomial point
in the Hilbert scheme.
This implies that the
the tangent space has a basis consisting of directions tangent
to deformations built using Gröbner basis. Such a nice structure of
the tangent space is surprising since it does
not hold
in the polynomial case.
Hilbert schemes and Betti numbers over a Clements-Lindström ring pdf
file
(with S.Murai)
submitted.
ABSTRACT:
The structure of the Hilbert scheme is known to be very complicated, and it seems that this
discouraged people from studying Hilbert schemes over quotient rings (as the proofs and constructions become significantly more intricate over quotient rings). The Clements-Lindström rings are a natural class of quotient rings to consider because Macaulay's Theorem holds over them. This paper is focused on such rings.
We show that the Hilbert scheme,
that parametrizes all ideals with the same Hilbert function over
a Clements-Lindtröm ring W, is connected. More precisely, we prove that every graded ideal is
connected by a sequence of deformations
to the lex-plus-powers ideal with the same Hilbert
function. Our result is an analogue of Hartshorne's theorem that
Grothendieck's Hilbert scheme is connected; however our proof is
different, since Hartshorne's deformations (distractions)
do not work over W.
We also prove a conjecture by Gasharov, Hibi, and Peeva
that the lex ideal attains maximal Betti numbers among all graded ideals in W with a fixed Hilbert function.
Hilbert schemes and maximal Betti numbers over Veronese rings pdf
file
(with V. Gasharov and S. Murai)
submitted.
ABSTRACT:
We show that Macaulay's Theorem and Gotzmann's Persistence Theorem hold over a Veronese toric ring R.
We also prove that the Hilbert scheme over R is connected; this is an analogue of Hartshorne's
theorem that the Hilbert scheme over a polynomial ring is connected.
Deformations of codimension 2 toric varieties pdf
file (with V. Gasharov)
Compositio Mathematica 123 (2000), 225-241.
ABSTRACT:
The study of the ideals with the same multigraded Hilbert
function as a given toric ideal T was initiated by
Arnold, who showed
that the structure of such ideals is encoded in continued fractions
in the case when T defines a monomial curve
in A3.
Our main result generalizes a result of
Arnold-Korkina-Post-Roelofs and solves a Conjecture of Sturmfels'. We
prove that if T has codimension two, then the toric Hilbert
scheme
has exactly one component and this component is the closure of the
orbit of the toric ideal T under the torus action.
For monomial curves in A3 this result was proved by
Arnold, Korkina, Post, and Roelofs.
Through a systematic computer search Sturmfels found an example in
codimension 3 which
shows that the result cannot be extended to higher codimensions.
On Toric Varieties:
Generic lattice ideals pdf file
(with B. Sturmfels)
Journal of the American Mathematical Society, 11 (1998), 363-373.
ABSTRACT:
Complete intersections are ideals whose generators have
sufficiently general coefficients. So
they might be regarded as generic among all ideals with fixed
small number of generators.
We introduce a different notion of
genericity: toric ideals whose generators are generic with
respect to their exponents -- not their coefficients.
For toric rings we define a notion of genericity which ensures
nicely structured
homological behavior. Our motivation comes from a strong recent result in
Integer Programming. We construct the finite minimal free resolution of a
generic toric ring.
Toric Hilbert schemes pdf file
(with M. Stillman)
Duke Mathematical Journal 111 (2002), 419-449.
ABSTRACT:
We introduce and study the toric Hilbert scheme H
which parameterizes all ideals with the same multigraded Hilbert
function as a given toric ideal T.
We define the structure of H and
prove the universality property. The latter makes it possible to obtain the
tangent space
at any point on H.
The lex ideal is a smooth point on the classical Hilbert scheme.
Usually no lex ideal exists on H. However, there is another special point:
we show that
the toric ideal T lies on exactly one component of H and this component is
reduced. In
particular, T is a smooth point on H.
Furthermore, we consider the case when T defines a
toric variety of codimension two.
We prove that in this case
H is two dimensional and smooth; it follows that H is the toric variety of
the Gröbner fan of the toric ideal T. In sharp contrast, very
little is known
about the
structure of the classical Hilbert scheme of a codimension two toric variety;
the best
result in this direction is proved by Piene and Schlessinger and says that
the classical
Hilbert scheme of the twisted cubic curve has two components of dimensions 12
and 15, each component is smooth but the scheme is not,
the two components intersect
transversally and their intersection is smooth of dimension 11.
Rationality for generic toric rings pdf file (with V. Gasharov and V. Welker)
Mathematische Zeitschrift 233 (2000), 93-102.
ABSTRACT:
We consider the
infinite minimal free resolution of the residue field over a generic
toric ring.
Problems of rationality of Poincaré and Hilbert series were
stated by several mathematicians:
by Serre and Kaplansky for local Noetherian rings,
by Kostrikin and Shafarevich
for nilpotent algebras, by Govorov for associative graded algebras
and by Serre and Moore for simply-connected complexes.
The Serre-Kaplansky problem, ``Is the Poincaré series o
f a
finitely generated commutative
local Noetherian
ring rational?'',
was one of the central questions in
Commutative Algebra for many years.
Anick constructed a ring
with transcendental
Poincaré series which provided a counterexample to the conjecture
that
the Serre-Kaplansky problem has a positive answer.
Our main result
provides a positive answer for generic toric rings. We also show that the
rate
of a generic toric ring is the maximum degree of a minimal generator
of the
corresponding toric ideal minus one; this is an
analogue to a result (for
monomial ideals) of
Eisenbud, Reeves, and Totaro.
How to shell a monoid? pdf file
(with V. Reiner and B. Sturmfels)
Mathematische Annalen 310 (1998), 379-393.
ABSTRACT:
We introduce two ideas how to
obtain the Betti numbers of the residue field over a toric ring:
We consider the toric variety as given in non-commutative
variables in order to
identify monomials with facets of simplicial complexes.
We
show that a quadratic non-commutative Gröobner basis
describes algebraically
a uniform non-pure shelling of the toric ring.
Syzygies of codimension 2 lattice ideals pdf file
(with B. Sturmfels)
Mathematische Zeitschrift 229 (1998), 163-194.
ABSTRACT:
We introduce an approach to study
the algebraic homology of toric rings via
integer-points-free bodies.
This leads to an upper bound 2codimension-1 on the
projective dimension
of an arbitrary toric ring;
the existence of a bound in terms of the codimension is surprising
and has no analogues for other classes of rings yet.
In the codimension 2
case the integer-points-free bodies
have simple structure; this makes it possible
to
describe combinatorially the minimal free resolution of the toric ring.
A challenging central conjecture in Algebraic Geometry states that
the Castelnuovo-Mumford regularity of a complex projective variety
is bounded by the degree minus the codimension plus one.
Recently
Kraft showed that the conjecture implies important bounds in Invariant Theory.
It is known to hold for curves (by a result of
Gruson-Lazarsfeld-Peskine),
Cohen-Macaulay rings, and smooth surfaces. We prove the conjecture for
codimension 2 toric rings (note that we use a weaker definition of a
toric variety,
so such a variety is not always Cohen-Macaulay).
On Monomial Resolutions:
The lcm-lattice pdf file
(with V. Gasharov and V. Welker)
Mathematical Research Letters 6 (1999), 521--532.
ABSTRACT:
For many years, the
Stanley-Reisner correspondence, introduced by Hochster
and Reisner, was
the main progress on the problem (posed by
Kaplansky in the early 1960's)
of finding a minimal free resolution of a monomial ideal.
It has a long tradition and has led to important
results.
The Stanley-Reisner theory is based on
computing the Betti numbers of a monomial ideal by
simplicial complexes.
We introduce a new approach inspired by the topological
theory of
subspace arrangements.
We introduce the lcm-lattice of a monomial ideal. We show that it plays
the same role
in describing the homology of the ideal as the role of the intersection
lattice in describing
the cohomology of the complement of a complex subspace arrangement. Namely:
the lcm-lattice determines the Betti numbers (analog of the
Goresky-MacPherson Formula
for subspace arrangements)
the lcm-lattice determines the maps in the minimal free resolution
up to relabeling
the lcm-lattice together with the additional data, which pairs of minimal
monomial generators are relatively prime, determine the algebra structure
of the Tor-algebra (analog of results by DeConcini-Procesi and Yuzvinsky for
subspace arrangements).
Monomial resolutions pdf
file (with D. Bayer and B. Sturmfels)
Mathematical Research Letters 5 (1998), 36-41.
ABSTRACT:
We introduce an idea entirely different than
the well-known Stanley-Reisner correspondence:
we introduce an approach for
resolving a monomial ideal by
encoding the whole resolution (including the differential maps)
into a single simplicial complex.
We prove that generically such resolution
is minimal and comes from the boundary of a polytope.
For a non-generic ideal we introduce the technique of
resolving by
``deforming to the generic case". This provides a
non-minimal resolution whose
length is less or equal to the number of the variables
(the resolution is usually much shorter/smaller than Taylor's resolution).
Ideals containing the squares of the variables pdf
file (with J. Mermin and M. Stillman)
Advances in Mathematics, 217 (2008), 2206-2230.
ABSTRACT:
Denote by P the ideal generated by the squares of the variables in a
polynomial ring. It is well known how the Hilbert function changes when
we add
P to a squarefree monomial ideal J; this is given by the relation
between the
f-vector and the h-vector. It has been an open question
how the Betti numbers change. We answer this question
providing a relation between the Betti numbers of J and those of J+P.
Furthermore, we describe a basis of the minimal free resolution of
J+P in the case when J is Borel.
By Kruskal-Katona's Theorem, there exists a squarefree lex ideal L
such that L+P has the same Hilbert function as J+P. The ideal
L+P is called lex-plus-squares. It was conjectured by Herzog and Hibi
that the graded Betti numbers of L+P are greater than or equal to those
of J+P. Later, Graham Evans conjectured the more general
lex-plus-powers conjecture that, among all graded ideals with a fixed
Hilbert function and containing a homogeneous regular sequence in fixed
degrees, the lex-plus-powers ideal has greatest graded Betti numbers in
characteristic 0.
This
conjecture is very difficult and wide open.
We prove it
for graded ideals containing the
squares of the variables.
On Subspace Arrangements:
Cohomology of real diagonal subspace arrangements via resolutions pdf file
(with V. Reiner and V. Welker)
Compositio Mathematica 117 (1999), 107-123.
ABSTRACT:
Consider the topology of the complement of a
subspace arrangement. Much fewer results are known for
real subspace arrangements than for complex ones. The reason for this is
that Algebraic Geometry methods can be applied in the complex case,
but not in the real case.
We introduce an approach of
expressing the cohomology of the complement
of a real diagonal arrangement
of subspaces by the
Betti numbers of a minimal free resolution.
Our approach is completely different from the well-known method
for obtaining the cohomology of such
complement using a formula of Goresky and
MacPherson.
The new approach yields short proofs of
results
on r-equals arrangements,
opens up the possibility
to compute examples fast, and leads
to results on vanishing of cohomology.
On Hilbert Functions:
Hilbert functions and lex ideals pdf file
(with J. Mermin)
Journal of Algebra 313 (2007), 642-656.
ABSTRACT:
A well-studied and important numerical invariant of a
homogeneous ideal over a graded polynomial ring S
is its Hilbert function. It gives
the sizes of the graded components of the ideal.
In many of the recent
papers and books, Hilbert functions are described using Macaulay's representation
with binomials.
Thus, the arguments consist of
very clever computations with binomials.
One of our main goals is to go back to Macaulay's original idea in 1927:
there exist highly structured monomial ideals - lex ideals - that attain
all possible Hilbert functions.
Let W be a quotient of S modulo
powers of the variables.
We present an algebraic proof of
Clements-Lindstrom's Theorem that every Hilbert function in W is attained
by a lex ideal. We prove Green's Theorem
over W. It is an open question whether Gotzmann's Theorem holds over W.