Algebraic Approximation of Cohen-Macaulay Algebras

Throughout what follows $\mathbb{K}$ will denote an arbitrary field, unless specified otherwise. Also for a fixed integer $n$, $x$ will denote the $n$-tuple of variables $(x_{1},\dots,x_{n})$. The ring of formal power series in variables $x$ with coefficients in $\mathbb{K}$ will be denoted by $\mathbb{K}[[x]]$. A formal power series $F\in\mathbb{K}[[x]]$ is called algebraic if it satisfies a non-trivial polynomial relation. The set of all such power series forms a ring that is called the ring of algebraic power series and is denoted by $\mathbb{K}\langle x\rangle$. In the case when $\mathbb{K}$ is a complete real valued field, the notation $\mathbb{K}\{x\}$ is used for the ring of convergent power series in variables $x$. Further, throughout this paper, for a $\mathbb{K}[[x]]$-module $M$, the notation $\dim M$ will be used for Krull dimension of $M$.

Let $M\subseteq\mathbb{K}[[x]]^{p}$ be a module generated by $F_{1},\dots,F_{s}\in\mathbb{K}[[x]]^{p}$. A module $M_{\mu}$ is called an algebraic approximation of $M$ of order $\mu$ if $M_{\mu}$ is generated by power series vectors whose entries are algebraic power series, and that agree with the generators $F_{1},\dots,F_{s}$ up to order $\mu$. The main result proved in this paper, Theorem 4.1, establishes the existence of algebraic approximations of arbitrarily high order to modules $M\subseteq\mathbb{K}[[x]]^{p}$, that share specified algebraic properties with $M$. These algebraic properties are Hironaka’s Diagram of Initial exponents (see Section 2.2 for a precise definition), and the minimal Betti numbers of $M$. (Recall that the minimal Betti numbers are the ranks of the free modules appearing in a minimal free resolution of $M$.) Aside from standard facts in commutative algebra, the two main tools used to prove this theorem are Artin’s Approximation Theorem [2, Theorem 1.10], and the theory of standard bases of modules $M\subseteq\mathbb{K}[[x]]$, which is analogous to the theory of Gröbner bases of modules over polynomial rings and was developed by T. Becker in [3].

The motivation for the choice of the particular algebraic properties above is the generalization to arbitrary fields of the result [1, Theorem 8.1], proved by J. Adamus and the author, on the existence of arbitrarily high order algebraic approximations of Cohen-Macaulay local analytic algebras over $\mathbb{R}$ or $\mathbb{C}$ that preserve the Hilbert-Samuel function. The diagram of initial exponents of an ideal completely determines its Hilbert-Samuel function (see Theorem 2.1). Recall that for an ideal $I\subseteq\mathbb{K}[[x]]$ the Hilbert-Samuel function of it is defined as

where $\mathfrak{m}$ is the maximal ideal of $\mathbb{K}[[x]]$. Also, recall that for $\mathbb{K}=\mathbb{R}$ or $\mathbb{C}$ a local analytic algebra is a quotient of the form $\mathbb{K}\{x\}/I$ where $I\subseteq\mathbb{K}\{x\}$ is an ideal. The property of Cohen-Macaulayness is related to the minimal Betti numbers of $\mathbb{K}\{x\}/I$; specifically, it can be determined by the number of non-zero Betti numbers (see Theorem 3.2). The result proved in this paper, that achieves the generalization of [1, Theorem 8.1], is Theorem 6.1, and follows directly from Theorem 4.1. In fact, Theorem 6.1 achieves more than just the generalization of [1, Theorem 8.1] to arbitrarily fields, as the minimal Betti numbers are a finer grained property than Cohen-Macaulayness. As a consequence of this, an approximation result analogous to [1, Theorem 8.1] for Gorenstein local algebras is obtained as a corollary to Theorem 6.1. This is Corollary 6.2. It should be noted here that generalization of the finite determinacy result proved by J. Adamus and the author in [1] for algebras $\mathbb{K}\{x\}/I$ that are complete intersections, [1, Theorem 7.3], to arbitrary fields, already exists and was proved by Srinivas and Trivedi in [11] using techniques different from those used in [1]. In fact, an even stronger result exists in the special case of isolated complete intersection singularities, due to Greuel and Pham [9, Theorem 1.5].

A generalization of the algebraic counterpart of the result on flat maps from [10, Theorem 1.2] is also obtained as a by-product of the proofs of Theorem 4.1 and Theorem 6.1 (Theorem 6.4). Briefly, [10, Theorem 1.2] states that given a flat analytic map from a real or complex analytic germ whose local ring is Cohen-Macaulay into a germ of euclidean space, one can find arbitrarily high order algebraic approximations to the domain that are Cohen-Macaulay, and to the map that are flat, which preserve the Hilbert-Samuel function of the special fibre. In this paper a homomorphism of rings $\phi:A\rightarrow B$ will be called flat if it makes $B$ into a flat $A$-module, and in the case when $A$ and $B$ are local rings with maximal ideals $\mathfrak{m}_{A}$ and $\mathfrak{m}_{B}$ respectively, the ring $B/\phi(\mathfrak{m}_{A})B$ will be called the special fibre of $\phi$. With these definitions Theorem 6.4 can be stated as follows: Let $y$ denote the $m$-tuple of variables $(y_{1},\dots,y_{m})$. If one has a flat homomorphism from $\mathbb{K}[[y]]$ to $\mathbb{K}[[x]]/I$, and if $\mathbb{K}[[x]]/I$ is Cohen-Macaulay, then there exist arbitrarily high order algebraic approximations to the homomorphism and $\mathbb{K}[[x]]/I$ such that the Hilbert-Samuel function and minimal Betti numbers of the special fibre of the approximants is the same as those of the special fibre of the original homomorphism and such that the approximating homomorphisms are flat. The precise notion of algebraic approximation in the context of homomorphisms of rings is defined in a manner analogous to that for modules (see the statement of Theorem 6.4). The result [10, Theorem 1.2] is a direct consequence of Theorem 6.4, and remarkably, its proof via Theorem 4.1 and Theorem 6.1 has fewer dependencies and is conceptually simpler than that of [10, Theorem 1.2].

The structure of this paper is as follows: Section 2 presents relevant definitions and theorems used in the proofs of the main theorems of the paper. Section 4 presents the proof of Theorem 4.1. Section 5 provides justification for the validity of the results of this paper for rings of convergent power series in the case when the field $\mathbb{K}$ is a complete real valued field. Section 6 presents the applications of Theorem 4.1, specifically the proof of Theorem 6.1 in Subsection 6.1, and the proof of Theorem 6.4 in Subsection 6.2.

Throughout this section let $M\subseteq\mathbb{K}[[x]]^{p}$ be a finitely generated module. For some integer $c$, and integers $n_{0},\dots,n_{c}$ an exact sequence of the form

is called a finite free resolution of $M$. By [7, Corollary 19.6], all finitely generated $\mathbb{K}[[x]]$-modules have a finite free resolution. The number $c$ in the above is called the length of the free resolution $\mathcal{F}_{M}$. If the homomorphisms $\phi_{i}$ for $1\leq i\leq c$ in a finite free resolution $\mathcal{F}_{M}$ have the property that $\mathrm{im}(\phi_{i})\subseteq\mathfrak{m}\mathbb{K}[[x]]^{n_{i-1}}$, then $\mathcal{F}_{M}$ is called a minimal free resolution of $M$. By [7, Lemma 19.4] the minimality of $\mathcal{F}_{M}$ is equivalent to the condition that for each $1\leq i\leq c$, a basis for $\mathbb{K}[[x]]^{n_{i-1}}$ maps onto a minimal set of generators (i.e., a set of generators with minimal cardinality) of $\mathrm{coker}(\phi_{i})$. Further, by [7, Theorem 20.2], all minimal free resolutions of $M$ have the same length, which is called the projective dimension of $M$, and denoted by $\mathrm{pd}_{\mathbb{K}[[x]]}(M)$. The numbers $n_{0},n_{1},\dots,n_{c}$ in the minimal free resolution $\mathcal{F}_{M}$ of a finitely generated module $M\subseteq\mathbb{K}[[x]]^{p}$ are called the minimal Betti numbers of $M$ and are denoted by $\beta^{M}_{0},\dots,\beta^{M}_{c}$.

The following characterization of Cohen-Macaulay rings of the form $\mathbb{K}[[x]]/I$ follows directly from [7, Corollary 19.15] and the fact that $\mathbb{K}[[x]]$ is a regular local ring.

If $\mathbb{K}[[x]]/I$ is Cohen-Macaulay, and has a minimal free resolution, such as $\mathcal{F}_{\mathbb{K}[[x]]/I}$ above, then the last minimal Betti number $\beta^{\mathbb{K}[[x]]/I}_{c+1}=n_{c}$ of $\mathbb{K}[[x]]/I$ is called the Cohen-Macaulay type of $\mathbb{K}[[x]]/I$.

Given $F_{1},\dots,F_{s}\in\mathbb{K}[[x]]^{p}$, the module generated by all $H=(H_{1},\dots,H_{s})^{T}\in\mathbb{K}[[x]]^{s}$ such that $\sum_{i=1}^{s}H_{i}F_{i}=0$ is called the module of syzygies on $F_{1},\dots,F_{s}$ and is denoted by $Syz(F_{1},\dots,F_{s})$. The specification of a minimal free resolution such as (3.1) for a module $M\subseteq\mathbb{K}[[x]]^{p}$ is equivalent to the specification of certain syzygy modules as follows: If the homomorphism $\phi_{0}$ is given by,

then $S_{1,0},\dots,S_{n_{0},0}\in\mathbb{K}[[x]]^{p}$ are a minimal basis of generators of the module $M$. For $1\leq k\leq c$, if the homomorphism $\phi_{k}$ is given by the matrix,

then $S_{1,k},\dots,S_{n_{k},k}\in\mathbb{K}[[x]]^{n_{k-1}}$ are a minimal basis of generators of the module $Syz(S_{1,k-1},\dots,S_{n_{k-1},k-1})$. If $M=(F_{1},\dots,F_{s})$ and $F_{1},\dots,F_{s}$ form a standard basis, the following theorem gives us a basis of generators for $Syz(F_{1},\dots,F_{s})$.