If $R$ is a commutative ring and $I$ and $J$ are ideals in $R$ such that $I+J=R$ then $I \cap J=IJ$. This is not generally true in noncommutative rings, e.g. let $R$ be the lower triangular 2 x 2 matrices over $\mathbb{Z}$ and let $I$ be the ideal generated by $E_{1,1}$ and $J$ be the ideal generated by $E_{2,2}$.

Is this ever possible in a noncommutative ring? If so, what are rings satisfying this property called? (Where can we find them in the literature?)


This is more a remark, since I do not directly answer the question. The statement in the question is true for all ideals $I,J$ (without the condition $I+J=R$) if and only if all ideals a idempotent, i.e. $I^2=I$. Indeed, $$I \cap J = (I \cap J)^2 \subset IJ \subset I \cap J.$$ The converse is obvious by taking $I=J$.

This holds for instance if $R$ is a $C^\star$-algebra.


I don't know non-commutative ring theory that well yet, but I ended up working this problem out anyway partially for my own edification. Let us suppose that I, J are both left ideals in R. Observe,

$(I \cap J) (I + J) = (I \cap J) I + (I \cap J) J \subseteq IJ + JI$

If I + J = (1), then we get

$I \cap J \subseteq IJ + JI$

Now it can be shown that

$I J \subseteq comm(I,J) \cap J$

where comm(I,J) denotes the set of all elements in I which commute with elements in J;

$comm(I, J) = ${$ x \in I : \forall y \in J; x y \in I $}


$I \cap J \subseteq IJ + JI \subseteq comm(I,J) \cap J + comm(J,I) \cap I$

If it is true that all of I commutes with J and vice-versa, then $comm(I,J) = I$, $comm(J,I) = J$ and we get:

$I \cap J = I J$

Now let us suppose that these bounds are not tight, or in other words there is some element $x \in I \setminus comm(I,J)$. This implies that there exists some $y \in J$ such that $x y \not \in I$, and so $x J \not \subseteq I \cap J$.

Therefore if $I + J = (1)$, then $I \cap J = I J + J I$ if and only if $comm(I,J) = I$ and $comm(J,I) = J$.

(Of course there is probably an easier way to say this, so maybe it is also good to get an expert to weigh in on the issue.)

  • 1
    $\begingroup$ Since the original question was asked about (two-sided) ideals rather than one-sided ideals, you might as well restrict your argument to that case. Then your observation above greatly simplifies: $I \cap J = IJ + JI$ for ANY pair of two-sided ideals $I$ and $J$. This makes it clear that the original question is asking for situations in which $IJ + JI = IJ$ - that is, situations in which $JI \subseteq IJ$. $\endgroup$ Sep 18 '10 at 17:11
  • $\begingroup$ Hmm, you have a good point. I must admit to being somewhat unfamiliar with the conventions regarding non-commutative rings. My initial reading was that he just said `ideals' without any further qualification and so I chose to examine the most restrictive instance of this class. However, making the assumption that the ideals are two-sided does greatly simplify things. $\endgroup$
    – Mikola
    Sep 18 '10 at 22:07

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.