Theorem
OKBELet \star be a semistar operation of finite type on the integral domain R, and I be a quasi-\star-ideal of R. If every minimal prime ideal of I is the radical of a \star-finite ideal, then I has finitely many minimal prime ideals.
Let R be a commutative integral domain and let \star be a semistar operation of finite type on R, and I be a quasi-\star-ideal of R. We show that, if every minimal prime ideal of I is the radical of a \star-finite ideal, then the set \Min(I) of minimal prime ideals over I is finite.
Let \star be a semistar operation of finite type on the integral domain R, and I be a quasi-\star-ideal of R. If every minimal prime ideal of I is the radical of a \star-finite ideal, then I has finitely many minimal prime ideals.
Suppose that \star is a semistar operation of finite type on the integral domain R, and that I is a quasi-\star-ideal. Then \sqrt I^{\star}\cap R=\sqrt I, that is \sqrt I is also a quasi-\star-ideal of R.
Since \sqrt I\subseteq\sqrt I^{\star}\cap R, it is enough to show that \sqrt I^{\star}\cap R\subseteq\sqrt I. Let x\in\sqrt I^{\star}\cap R. Then for every P\in\operatorname{Min}(I) we have xR^{\star}\subseteq\sqrt I^{\star}\subseteq P^{\star}. Since P is a quasi-\star-prime ideal of R by [EF, Lemma 2.3(d)], we obtain that x\in P. Hence x\in\sqrt I as desired.
Suppose that \star is a semistar operation of finite type on the integral domain R, and that I is a quasi-\star-ideal. Then \sqrt{I^{\star}}\cap R=\sqrt I. Indeed suppose that x\in\sqrt{I^{\star}}\cap R, then there is a positive integer n such that x^n\in I^{\star}. Since I is a quasi-\star-ideal, and x\in R we obtain that x^n\in I. Hence x\in\sqrt I. Consequently we have \sqrt{I^{\star}}\cap R\subseteq\sqrt I\subseteq\sqrt{I^{\star}\cap R}=\sqrt{I^{\star}}\cap R, which gives us the desired equality.
Let \star be a semistar operation of finite type on the integral domain R, and I be a quasi-\star-ideal of R. If every minimal prime ideal of I is the radical of a \star-finite ideal, then I has finitely many minimal prime ideals.
Note that \sqrt I is a radical quasi-\star-ideal by Lemma 2.1. Hence since we have \operatorname{Min}(I)=\operatorname{Min}(\sqrt I), it is harmless to assume that I is a radical quasi-\star-ideal.
Let S=\{P_1\cdots P_n|\text{ each }P_i\text{ is a prime ideal minimal over }I\}. If for some C=P_1\cdots P_n\in S we have C^{\star}\subseteq I^{\star}, hence C\subseteq C^{\star}\cap R\subseteq I^{\star}\cap R=I. Then any prime ideal minimal over I contains some P_i, so \{P_1,\cdots,P_n\} is the set of minimal prime ideals of I. Hence we may assume that C^{\star}\nsubseteq I^{\star} for each C\in S. Consider the set \mathcal{A} consisting of all radical quasi-\star-ideals J of R containing I such that C^{\star}\nsubseteq J^{\star} for each C\in S. Since I\in \mathcal{A} we have \mathcal{A}\neq\emptyset. The set \mathcal{A} is partially ordered under inclusion \subseteq, and we show that it is inductive under this ordering. Let \{J_{\alpha}\}_{\alpha\in\Gamma} be a chain in \mathcal{A}. Put J=\cup J_{\alpha}. So that J is a radical quasi-\star-ideal of R containing I such that J^{\star}=(\cup J_{\alpha})^{\star}=\cup J_{\alpha}^{\star}, in which the second equality holds, since \star is a semistar operation of finite type. Now assume that C^{\star}\subseteq J^{\star} for some C\in S. Suppose that C=P_1\cdots P_n, and that P_i=\sqrt{L_i} for some \star-finite ideal L_i of R, for i=1,\cdots,n. Let F_is be finitely generated ideals of R, such that L_i^{\star}=F_i^{\star}, for i=1,\cdots,n. Now consider (F_1\cdots F_n)^{\star}=(L_1\cdots L_n)^{\star}\subseteq (P_1\cdots P_n)^{\star}\subseteq\cup J^{\star}_{\alpha}, which implies that F_1\cdots F_n\subseteq\cup J_{\alpha}. Therefore there exists an index \alpha\in\Gamma such that F_1\cdots F_n\subseteq J_{\alpha} and hence (L_1\cdots L_n)^{\star}=(F_1\cdots F_n)^{\star}\subseteq J^{\star}_{\alpha}. Thus L_1\cdots L_n\subseteq J_{\alpha}, and so we have P_1\cdots P_n\subseteq\sqrt{L_1\cdots L_n}\subseteq\sqrt{J_{\alpha}}=J_{\alpha}. Consequently (P_1\cdots P_n)^{\star}\subseteq J^{\star}_{\alpha} which is impossible. Now Zorn’s Lemma gives us a maximal element Q of \mathcal{A}. One can actually assume that Q\neq R. We show that Q is a prime ideal of R. To this end let a, b be two elements of R such that ab\in Q and assume that a, b\notin Q. Since Q\varsubsetneq(Q+aR)\subseteq\sqrt{(Q+aR)^{\star}\cap R}, and \sqrt{(Q+aR)^{\star}\cap R} is a radical quasi-\star-ideal (by Lemma 2.1) containing I, there exists an element C_1\in S such that C_1^{\star}\subseteq\sqrt{(J+aR)^{\star}\cap R}^{\star}. By the same reason there exists again an element C_2\in S such that C_2^{\star}\subseteq\sqrt{(Q+bR)^{\star}\cap R}^{\star}. Therefore we have
which is a contradiction. Therefore Q is a prime ideal of R. But since I\subseteq Q, it contains a prime ideal P minimal over I by [K, Theorem 10]. Thus P\in S and P^{\star}\subseteq Q^{\star}, a contradiction.
(Gilmer and Heinzer [GH, Theorem 1.6] and Anderson [A]) Let R be an integral domain, and let I be an ideal of R. If each minimal prime of the ideal I is the radical of a finitely generated ideal, then I has only finitely many minimal primes.
Let I be a t-ideal of the integral domain R. If each minimal prime ideal of I is the radical of a t-finite ideal, then \operatorname{Min}(I) is finite.
Suppose that R has a Noetherian overring S. If I is an ideal of R such that IS\cap R=I, then \operatorname{Min}(I) is finite.
Define a semistar operation \star, by E^{\star}=ES, for each E\in\overline{\mathcal{F}}(R). Thus I=IS\cap R=I^{\star}\cap R is a quasi-\star-ideal of R. Note that \star is a semistar operation of finite type. Let P\in\operatorname{Min}(I). Since S is a Noetherian ring, we have P^{\star}=PS=(x_1,\cdots,x_n)S=(x_1,\cdots,x_n)^{\star}, for some elements x_1,\cdots,x_n of P. This means that P is a \star-finite ideal. Now the result is clear from Theorem 2.3.
Suppose that \star is a semistar operation stable and of finite type on the integral domain R, and that I is a nonzero ideal of R, such that I^{\star}\cap R\neq R. Then \star-\operatorname{Min}(I)=\operatorname{Min}(I^{\star}\cap R). In particular
One sees easily that \star-\operatorname{Min}(I)\subseteq\operatorname{Min}(I^{\star}\cap R). For the reverse inclusion, let P\in\operatorname{Min}(I^{\star}\cap R). So that I\subseteq I^{\star}\cap R\subseteq P. Choose by [K, Theorem 10] a prime ideal Q minimal over I contained in P. Note that Q is a quasi-\star-ideal, since it is contained in P ([FH, Corollary 3.9, Lemma 4.1, and Remark 4.5]). Then I^{\star}\subseteq Q^{\star}\subseteq P^{\star} and so I\subseteq I^{\star}\cap R\subseteq Q^{\star}\cap R=Q\subseteq P^{\star}\cap R=P. Thus Q=P, since P is minimal over I^{\star}\cap R.
Let \star be a semistar operation stable and of finite type on the integral domain R, and I be a nonzero ideal of R, such that I^{\star}\cap R\neq R. If every quasi-\star-prime ideal minimal over I is the radical of a \star-finite ideal, then \star-\operatorname{Min}(I) is finite.
By the above lemma we have \star-\operatorname{Min}(I)=\operatorname{Min}(I^{\star}\cap R). Thus every prime ideal minimal over I^{\star}\cap R is the radical of a \star-finite ideal. Noting that I^{\star}\cap R is a quasi-\star-ideal of R, and using Theorem 2.3, we have \star-\operatorname{Min}(I) is a finite set.
([P, Proposition 2.4(2)]) Suppose that \star is a stable semistar operation of finite type on the integral domain R, and that R is a \star-Noetherian domain. Then \star-\operatorname{Min}(I) is finite for every ideal I of D.
Let \star be a semistar operation of finite type on the integral domain R, the following then are equivalent:
(1) Each quasi-\star-prime is the radical of a \star-finite ideal.
(2) Each radical quasi-\star-ideal is the radical of a \star-finite ideal.
(3) R satisfies the a.c.c. on radical quasi-\star-ideals.
(1)\Rightarrow(2) Consider the following set.
If \mathcal{A}\neq\emptyset, let \beta=\{I_{\alpha}\} be a chain of elements of \mathcal{A}. Put I=\cup I_{\alpha}. Hence I is a radical quasi-\star-ideal of R such that I^{\star}=(\cup I_{\alpha})^{\star}=\cup I_{\alpha}^{\star}. Suppose that I=\sqrt L for some \star-finite ideal L. Let L^{\star}=F^{\star} for some F\in f(R). So that F^{\star}=L^{\star}\subseteq I^{\star}=\cup I_{\alpha}^{\star} which implies that F\subseteq\cup I_{\alpha}. Therefore there is an index \alpha such that F\subseteq I_{\alpha}. Consequently L^{\star}=F^{\star}\subseteq I_{\alpha}^{\star} and hence L\subseteq I_{\alpha}. So we obtain that I_{\alpha}=\sqrt L, which is impossible. Hence I\in\mathcal{A}. Thus by Zorn’s Lemma \mathcal{A} has a maximal element P. Let a, b be two elements of R such that ab\in P and suppose that a, b\notin P. Since P\varsubsetneq(P+aR)\subseteq\sqrt{(P+aR)^{\star}\cap R}, and \sqrt{(P+aR)^{\star}\cap R} is a radical quasi-\star-ideal (by Lemma 2.1), we have \sqrt{(P+aR)^{\star}\cap R}=\sqrt L, for some \star-finite ideal. By the same reason \sqrt{(P+bR)^{\star}\cap R}=\sqrt N, where N is a \star-finite ideal. The same proof as Theorem 2.3 shows that P=\sqrt{LN}, which is impossible, since LN is a \star-finite ideal. Hence P is a quasi-\star-prime, a contradiction. Hence \mathcal{A}=\emptyset.
(2)\Rightarrow(3) Suppose that (I_n)_{n\in \mathbb{N}} be an ascending chain of radical quasi-\star-ideals, and set I=\bigcup_{n\in \mathbb{N}} I_n. Then I is a radical quasi-\star-ideal. Hence I=\sqrt L for some \star-finite ideal L. So, there is an integer n_0 such that I_{n_0}=\sqrt L=I. Hence (I_n)_{n\in \mathbb{N}} is stationary.
(3)\Rightarrow(1) Suppose that (1) is false. Then we can construct a chain (I_n)_{n\in \mathbb{N}} of radical quasi-\star-ideals strictly ascending. Indeed, let P be a quasi-\star-prime ideal which is not the radical of a \star-finite ideal. Set I_1=\sqrt{(x)}, where 0\neq x\in P. Given I_n=\sqrt{(x_1,\cdots,x_n)^{\star}\cap R}, where x_1,\cdots,x_n\in P, then I_{n+1}=\sqrt{(x_1,\cdots,x_n,x_{n+1})^{\star}\cap R}, where x_{n+1}\in P\backslash I_n.
Suppose that \star is a semistar operation of finite type on the integral domain R. If R satisfies the a.c.c. on radical quasi-\star-ideals, then \star-\operatorname{Min}(I) is finite for every ideal I of R.
Note \star-\operatorname{Min}(I)\subseteq\operatorname{Min}(I^{\star}\cap R) and use Theorem 2.3.
(1) One can prove Theorem 2.3 for arbitrary rings with zero divisors. Let R be a commutative ring, with total quotient ring T(R). Let \mathcal{F}(R) denote the set of all R-submodules of T(R). Suppose an operation *:\mathcal{F}(R)\rightarrow\mathcal{F}(R), E\rightarrow E^*, satisfies, for all E, F\in\mathcal{F}(R), and for all x\in T(R), the following:
*_1 xE^*\subseteq (xE)^* and if x is regular, then xE^*=(xE)^*;
*_2 E\subseteq F implies that E^*\subseteq F^*;
*_3 E\subseteq E^* and E^{**}:=(E^*)^*=E^*.
Then from these axioms the following directly follow:
(i) (EF)^*=(E^*F)^*=(EF^*)^*=(E^*F^*)^*;
(ii) (E+F)^*=(E^*+F)^*=(E+F^*)^*=(E^*+F^*)^*;
(iii) (E\cap F)^*\subseteq E^*\cap F^*=(E^*\cap F^*)^*.
It is clear that any semistar operation satisfies these axioms.
It is routine to see that [HP, Lemma 3.3] the v-operation satisfies, these axioms, where E^v=(E^{-1})^{-1}, in which E^{-1}=(R:E)=\{x\in T(R)|xE\subseteq R\}, for E\in \mathcal{F}(R).
By this operation Theorem 2.3, is true for rings with zero divisors.
(2) It is interesting to note that if we take R to be the ring of all sequences from \mathbf{Z}/2\mathbf{Z} that are eventually constant, with pointwise addition and multiplication, then R is a zero-dimensional Boolean ring with minimal prime ideals P_i=\{\{a_n\}\in R|a_i=0\} and P_{\infty}=\{\{a_n\}\in R|a_n=0\text{ for large }n\} and each P_i is principal but P_{\infty} is not finitely generated. Thus while R has infinitely many minimal prime ideals, only one is not the radical of a finitely generated ideal.
Suppose that \star is a semistar operation of finite type on the integral domain R. If R is a \star-SFT-ring, then R satisfies the a.c.c. on radical quasi-\star-ideals.
Let P be a quasi-\star-prime ideal. Since P is a \star-SFT-ideal there is a finitely generated subideal J\subseteq P such that \sqrt{P^{\star}}=\sqrt{J^{\star}}. Now consider
Since J^{\star}\cap R is an \star-finite ideal, the result follows by Proposition 2.10.
Each quasi-\star-ideal of a \star-SFT-ring R, has only finitely many minimal prime ideals.
Suppose that \star is a semistar operation of finite type on the integral domain R. Then R is a \star-SFT-ring if and only if each quasi-\star-prime ideal of R is a \star-SFT-ideal.
Suppose that R is not a \star-SFT-ring. Therefore the set
is not an empty set. The set \mathcal{A} is partially ordered under inclusion, and is inductive under this ordering. By Zorn’s lemma, \mathcal{A} contains a maximal element P. Assume that a_1, a_2 be two elements of R such that a_1a_2\in P and a_1, a_2\notin P. Since P\varsubsetneq (P+a_iR)^{\star}\cap R, (P+a_iR)^{\star}\cap R is a \star-SFT-ideal of R. Consequently there exist a finitely generated ideal L_i\subseteq (P+a_iR)^{\star}\cap R, and a positive integer k_i such that c^{k_i}\in (L_i)^{\star} for each c\in (P+a_iR)^{\star}. Let L=L_1L_2 and k=k_1+k_2. Then L is a finitely generated subideal of P such that c^k=c^{k_1}c^{k_2}\in (L_1)^{\star}(L_1)^{\star}\subseteq(L_1L_2)^{\star}, for each c\in P^{\star}. Thus P is a \star-SFT-ideal, a contradiction. Therefore, P is a quasi-\star-prime ideal which is not a \star-SFT-ideal.
The text on this page is from arXiv:0809.1307v2 by Parviz Sahandi, under the licence CC0 1.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.
The record of each item is in analysis/objects/0809/0809.1307.jsonl.