I am stuck on the proof for this. This is what I manage to conclude so far: we have two trivial ideals, (0) and (1). Since it is not a PID, the ring R cannot be a field, so it must have some element that isn't a unit. Let it be x, so (x) is another ideal. Then, as it is principal, we must have another ideal, say I, which is not principal. (x) is not equal to I, so we have to have some element y that: either belongs to I but not to (x) (if I is not a subset of (x)), or y is in (x). Whatever the case may be, (y) is another ideal. That gives 5 ideals, but I am struggling to find a 6th one. My ideas were (x, y) or (x+y) but I cannot conclude that those ones are different from all the rest.
-
" so (x) is another ideal. then, as it is principal, we must have another ideal," What does that mean? What is principal? – John Douma Oct 12 '23 at 11:37
-
3Welcome to MSE. A question should be written in such a way that it can be understood even by someone who did not read its title. – José Carlos Santos Oct 12 '23 at 11:39
-
I'd go at it like this: since it's not a PID, it must have a nonprincipal ideal, say, $(a,b)$. Then see whether you can show that the ideals $(a,b)$, $(a)$, $(b)$, $(a+b)$, $(0)$, and $(1)$ must all be different. – Gerry Myerson Oct 12 '23 at 12:36
-
Please use MathJax. Here is a tutorial. For exactly six ideals see this post. – Dietrich Burde Oct 12 '23 at 12:39
-
@GerryMyerson That doesn't quite work, as "nonprincipal" does not mean "2-generated". There are domains that are not PIDs but where every finitely generated ideal is principal. – Arturo Magidin Oct 12 '23 at 18:57
-
@Arturo, you are, of course, quite correct, and I see you have posted a workaround in your answer. – Gerry Myerson Oct 12 '23 at 21:39
1 Answers
Let $R$ be a commutative ring with unity that is not a PID. Let $I$ be a nonprincipal ideal.
If $I$ is not finitely generated, let $X$ be an infinite generating set; we may assume no element of $X$ is zero. Let $x_1\in X$. Then $(x_1)\neq (0)$, $(x_1)\neq R$, and $(x_1)\neq I$. If for every $x\in X$ we have $(x_1,x)=(x_1)$, then $X\subseteq (x_1)$, so $I=(x_1)$, which is not possible. Thus, there exists $x_2\in X$ such that $(x_1,x_2)\neq (x_1)$. But also $(x_1,x_2)\neq I$. Continuing this way we obtain an infinite chain of ideals contained in $I$, so you get at least six different ideals.
So we may assume that all ideals of $R$ are finitely generated. Note that this means that there must be a $2$-generated ideal that is not principal, for if every $2$-generated ideal is principal, then every finitely generated ideal is principal (do induction). Thus, there is an ideal $I$, $I\neq R$, $I\neq (0)$, and elements $a,b\in R$, $a\neq b$, such that $I=(a,b)$, $I\neq (a)$, $I\neq (b)$. In particular we have the ideals $(0)$, $(a)$, $(b)$, $(a,b)$, and $R$, pairwise distinct. We just need one more ideal.
Consider $(a+b)$. Note that because $(a)\neq(b)$, we cannot have $a+b=0$, so $(a+b)\neq 0$. Because $a+b\in I$, we also cannot have $(a+b)=R$. If $a\in (a+b)$, then $a,b\in (a+b)$, so $I=(a,b)\subseteq (a+b)\subseteq I$, and we would have $I$ principal, so $a\notin (a+b)$. Therefore, $(a)\neq(a+b)$. Symmetrically, $(b)\neq (a+b)$. So now we have the ideals $(0)$, $(a)$, $(b)$, $(a+b)$, $(a,b)=I$, and $R$, pairwise distinct. That makes six.
In fact, six is the best you can do, as there are non-PIDs with exactly six ideals.
Note that there are rings in which every finitely generated ideal is principal (they are called Bézout domains, and a standard example is the ring of all algebraic integers), so if you are looking for a nonprincipal ideal you cannot just assume that there is one which is finitely generated. But if you know there are finitely generated non-principal ideals, then there must be a 2-generated non-principal ideal.

- 398,050
-
I guess that the comment by user rschwieb, at the question to which you link, is faulty for the reasons you have given here. – Gerry Myerson Oct 12 '23 at 21:44
-
@GerryMyerson it is correct in that you cannot assume the existence of a 2-generated non-principal ideal from the existence of a non-principal ideal. It is faulty in that there maybe no "minimal" generating set if there is no finite generating set (consider the ideal generated by all real $n$th roots of $2$ in the ring of algebraic integers). – Arturo Magidin Oct 12 '23 at 21:50
-
@GerryMyerson Of course, he may have been working in the setting of looking for a ring with exactly six ideals, which would immediately disqualify any ring whichnis not noetherian (by ACC). I would say it's correct to point out that one must argue that a non-principal 2-generated ideal exists, rather than just posit it does. – Arturo Magidin Oct 12 '23 at 22:01