Consider a commutative noetherian local ring $R$ of dimension $d$ and define

$$c_R\colon=\min_{(x_1,\ldots,x_d)} \{\mathrm{length}\ R/(x_1,\ldots,x_d)R\mid (x_1,\ldots,x_d)\ \mathrm{is\ a\ system\ of parameters\ of\ }R\}.$$

1) Does $c_R$ always give information about the ring $R$? For instance, clearly $c_R=1$ if and only if $R$ is regular. When is $c_R=2$? Does $R$ have to be any particular type of ring for $c_R$ to be $2$?

2) In a given local ring $R$ can one characterize those system of parameters $(x_1,\ldots,x_d)$ for which $c_R=\mathrm{length}\ R/(x_1,\ldots,x_d)R$? For instance, one could say that a necessary condition for $c_R=\mathrm{length}\ R/(x_1,\ldots,x_d)R$ is that $(x_1,\ldots,x_d)$ has to be "maximal", that is, not contained in any larger parameter ideal. Is this condition also sufficient?


Here is answer of (1), $c_R=2$. I assume that $R$ is complete (as $\mathfrak{m}$-adic topology) and the residue field $A/\mathfrak{m}$ is infinite (if necessary).

Let $v(R) = \dim_{R/\mathfrak{m}} \mathfrak{m}/\mathfrak{m}^2$ be the embedded dimension of $R$. Since $c_R = 2$ we have $R$ is not a regular local ring, so $v(R) \ge d+1$. Let $\mathfrak{q} = (x_1, ..., x_d)$ be a parameter ideal such that $\ell(R/\mathfrak{q}) = c_R=2$. We have $\ell(\mathfrak{m}/\mathfrak{q})= 1$, so $v(R) = d+1$. Furthermore $\mathfrak{m}^2 \subseteq \mathfrak{q}$ and $\mathfrak{m} = (x_1, ..., x_d, y)$ for some $y$.

First we consider the case $(R, \mathfrak{m})$ is unmixed i.e. $\dim R/\mathfrak{p} = d$ for all $\mathfrak{p} \in \mathrm{Ass}R$.

We note that the multiplicity $e(R) = e_R(\mathfrak{m}) \ge 2$, since $R$ is not regular.

Let $\mathfrak{q} = (x_1, ..., x_d)$ be a parameter ideal such that $\ell(R/\mathfrak{q}) = c_R=2$. It is well known that the multiplicity $e_R(\mathfrak{q}) \le \ell(R/\mathfrak{q})$. Thus we have two cases:

Case 1: $e_R(\mathfrak{q}) = 1$, so the multiplicity of $R$, $e(R) =1$. It is a contradiction.

Thus we have

Case 2: $e_R(\mathfrak{q}) = 2$, so $e(R) = 2$. Moreover $e_R(\mathfrak{q}) = \ell(R/\mathfrak{q})$, hence $(R, \mathfrak{m})$ is Cohen-Macaulay.

By Abhyankar's inequality we have $v(R) \le e(R) + d -1 = d+1$. Thus $R$ is a Cohen-Macaulay local ring of maximal embedded dimension. Furthermore, we have $\mathfrak{q}$ is a minimial reduction of $\mathfrak{m}$ and $\mathfrak{m}^2 = \mathfrak{q}\mathfrak{m}$ (by Sally?). The converse is true. Thus we have

Conclusion. Suppose $(R, \mathfrak{m})$ is unmixed. Then $c_R = 2$ if and only if $R$ is a Cohen-Macaulay (in fact, Gorenstein) of maximal embedded dimension with $v(R) = d+1$.

The case $R$ is not unmixed. Let $U$ be the largest ideal of $R$ of dimension least than $d$ (as $R$-module). Let $S = R/U$, we have $S$ in unmixed. It is clear that $c_S \le c_R = 2$. As above we have $S$ is Cohen-Macalay. So $x_1, ..., x_d$ is an regular sequence on $S$, we have $U \cap \mathfrak{q} = \mathfrak{q}U$. Therefore we have the following short exact sequence $$0 \to U/\mathfrak{q}U \to R/\mathfrak{q} \to S/\mathfrak{q}S \to 0.$$ Since $\ell(R/\mathfrak{q}) = 2$ we have $\ell(U/\mathfrak{q}U) = \ell(S/\mathfrak{q}S) = 1$. Therefore $R$ is a regular local ring and $U = (a)$ is a principal ideal. Moreover we have $\mathfrak{m} = (x_1, ..., x_d, a)$.

Examples. Take $S = K[[X_1, ..., X_d]]$ and $T$ is a non-zero quotient ring of $S$. Let $R = S \ltimes T$ be the idealization (or trivial extension). We can check that the maximal ideal of $R$ generated by $d+1$ elements $(X_1, 0), ..., (X_d, 0), (0, 1)$. If $T = S$ we have $R$ is unmixed. If $T \neq S$ we have $R$ is not unmixed.

Edit: I give a counter-example for (2). Let $R$ be the localization of $\mathbb{C}[X,Y]/(Y^2) = \mathbb{C}[x, y]$ at the maximal ideal $\mathfrak{m} = (x,y)$. We have $R$ is a Cohen-Macaulay local ring of dimension one, and $e(R) = 2$, $v(R) = 2$, $c_R=2$. Thus $R$ is maximal embedded dimension.

Consider the parameter ideal $\mathfrak{q} = (x^2+y)$. Since $\mathfrak{q} \nsubseteq \mathfrak{m}^2$ we have there is no parameter ideal containing $\mathfrak{q}$. Therefore $\mathfrak{q}$ is a maximal parameter ideal. One can check that $$\ell(R/\mathfrak{q}) = \ell(\mathbb{C}[X,Y]/(X^2+Y, Y^2)) = 4.$$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.