Local cohomology of Segre product type rings (Q473008)

Let \(A\) be a ring with a valuation \(\nu\), i.e. a map \(\nu:A\rightarrow \mathbb{Q}\cup\{\infty\}\) satisfying the following rules: {\parindent=6mm \begin{itemize} \item[(1)] \(\nu(ab)=\nu(a)+\nu(b)\) for all \(a,b\in A\). \item [(2)] \(\nu(a+b)\geq \text{min}\{\nu(a),\nu(b)\}\) for all \(a,b\in A\). \item [(3)] \(\nu(a)=\infty\) if and only if \(a=0\). \end{itemize}} An \(A\)-module \(M\) is said to be an almost zero \(A\)-module if any element of \(M\) can be killed by an element \(a\) of \(A\) with arbitrary small valuation (i.e. \(\nu(a)\) is arbitrary small). Moreover, a sequence \(a_1,\ldots,a_n\) of elements of \(A\) is said to be an almost regular sequence if the \(A\)-module, \(\big((a_1,\ldots,a_{i-1}):_Aa_i\big)/(a_1,\ldots,a_{i-1})\), is almost zero for every \(1\leq i\leq n\). Now suppose that \(A\) is an \(B\)-algebra with a valuation \(\nu\), where \(B\) is local or graded of dimension \(d\). Choose a system of parameters \(x_1,\ldots,x_d\) for \(B\). We say that \(A\) is an almost Cohen-Macaulay \(B\)-algebra if \(\text{H}_{(x_{1},\ldots,x_{d})}^{i}(A)\) is almost zero for each \(i<d\) and if \(A/\big((x_1,\ldots,x_d)A\big)\) is not almost zero. Also if the non-top local cohomologies are zero, then \(A\) is said to be a big Cohen-Macaulay \(B\)-algebra. In particular if \(x_1,\ldots,x_d\) forms a (an almost) regular sequence of \(A\) then \(A\) is a big (an almost) Cohen-Macaulay \(B\)-algebra. The preceding definitions have their motivation and origin in the homological conjectures. Many homological conjectures, including the Hochster's monomial conjecture and its equivalent forms such as canonical element conjecture and improved new intersection conjecture, follow from the existence of big-Cohen-Macaulay module. In the prime characteristic \(p>0\) it is proved that the absolute integral closure \(R^+\) of a complete local domain \(R\) is a big Cohen-Macaulay \(R\)-algebra. Unfortunately, using the trace map, we can see that the analogous result in the characteristic zero for \(\text{dim}(R)\geq 3\) does not hold. Furthermore, in [\textit{R. C. Heitmann}, Ann. Math. (2) 156, No. 2, 695--712 (2002; Zbl 1076.13511)] the author showed that \(R^+\) is an almost Cohen-Macaulay algebra provided \(R\) is a complete normal domain of dimension three. It is noteworthy to point out that the monomial conjecture and its equivalent forms follow from the existence of the almost Cohen-Macaulay algebra. This paper deals with a weaker condition than the existence of almost Cohen-Macaulay algebras. Namely, in the special case of Segre products, the aim of this paper is to find an \(B\)-algebra \(C\) satisfying, \[ \text{H}^i_{(x_1,\ldots,x_d)}(B)\rightarrow \text{H}^i_{(x_1,\ldots,x_d)}(C), \] is almost zero for each \(i\lneq \text{dim}(B)\). Consider the Segre product, \(B:=R\#S\), wherein \(R\) and \(S\) are Cohen-Macaulay \(\mathbb{N}_0\)-graded rings both torsion-free over the unramified discrete valuation ring, \(R_0=S_0=V\). Set, \[ T:=\bigcup\limits_{i=1}^{\infty}B[p^{1/p^n},\alpha_2^{1/p^n},\ldots,\alpha_s^{1/p^n},\alpha_{s+1}^{1/p^n},\ldots,\alpha_z^{1/p^n}], \] wherein \(p,\alpha_2,\ldots,\alpha_s\) is a particular system of parameters for \(B\) and \(p,\alpha_{2},\ldots,\alpha_{z}\) generates \(R\#S\) over \(V\). In his previous paper, [``The root closure of a ring of mixed characteristic'', \url{arxiv:0810.0215}], the author of the present paper introduced the integral extension, \[ C=\{t\in T_p;t^{p^n}\in T \text{\;for\;some\;} n\in \mathbb{N}_0\}, \] of \(T\), to deduce the injectivity of certain natural map from a quotient of the Fontaine ring \(\text{E}(C)\) to \(C/pC\)(respectively, a quotient of the Witt vectors ring \(\text{W}\big(\text{E}(C)\big)\) to the \(p\)-adic completion \(\widehat{C}\)) (those injective ring homomorphisms are actually isomorphisms). In the present paper, in the above case of Segre prodcut type rings, for proving the above almost vanishing of maps of local cohomology modules, it is proven that any element \(b\) of \((p,\alpha_2,\ldots,\alpha_i):_B \alpha_{i+1}\) almost belongs to \((p, \alpha_2,\ldots,\alpha_i)C\) for each \(0\leq i\leq s-1\) (i.e. for every \(\epsilon >0\) there exists \(c\in C\) such that \(\nu(c)<\epsilon\) and \(cb\in (p,\alpha_2,\ldots,\alpha_i)C\)). This too technical result which sheds some lights on the existence of almost Cohen-Macaulay algebra problem seems more interesting if we bear in mind that many examples of non-Cohen-Macaulay normal \(\mathbb{N}_0-\)graded domains are Segre product of Cohen-Macaulay normal graded domains.