Let $S$ be an abelian group under an operation denoted by $+$. Suppose further that $S$ is closed under a commutative, associative law of multiplication denoted by $\cdot$. Say that $\cdot$ distributes over $+$ in the usual way. Finally, for every $s\in S$, suppose there exists some element $t$, not necessarily unique, such that $s\cdot t=s$.

Essentially, $S$ is one step removed from being a ring; the only problem is that the multiplicative identity is not unique. Here is an example.

Let $S=\{\text{Continuous functions} f: \mathbb{R}\rightarrow \mathbb{R} \ \text{with compact support}\}$ with addition and multiplication defined pointwise. It is clear that this is an abelian group with the necessary law of multiplication. Now, let $f\in S$ be supported on $[a,b]$. Let $S'\subset S$ be the set of continuous functions compactly supported on intervals containing $[a,b]$ that are identically 1 on $[a,b]$. Clearly, if $g\in S'$, then $f\cdot g=f$ for all $x$. Also, there is no unique multiplicative identity in this collection since the constant function 1 is not compactly supported.

I've observed that this example is an increasing union of rings, but I don't know if this holds for every set with the property I've defined.

  • $\begingroup$ No distributive law? $\endgroup$ – Thomas Andrews Nov 2 '12 at 19:52
  • $\begingroup$ Oops, forgot about that. Thanks. I guess we can just say that $\cdot$ distributes over $+$ in the usual way? I'll edit the question. $\endgroup$ – Alexander Sibelius Nov 2 '12 at 19:55
  • $\begingroup$ Your algebra has the added feature that if $f^n=0$ then $f=0$, for example. $\endgroup$ – Thomas Andrews Nov 2 '12 at 19:59
  • 1
    $\begingroup$ Note that for any $s\in S$, $S_s=\{r\in S: rs=r\}$ is either $\{0\}$ or a ring with $s$ the identity. You can define a pre-order on $S$ by $s\leq s' \iff \forall r\in S: rs=r \implies rs'=r$ $\endgroup$ – Thomas Andrews Nov 2 '12 at 20:02

Your example is a rng with an approximate identity.


I just wanted to throw out the most obvious and broad class of examples which has gone unmentioned so far: von Neumann regular rngs.

A ring (possibly without identity) is called von Neumann regular if for every $a$ in the ring there exists $x$ in the ring such that $axa=a$. These rings have the "local identity" property you described (on both sides in fact, although without commutativity the local identity might not be the same on both sides :) ).

One thing I find particularly interesting about VNR rings is that it seems like functional analysts have particularly natural uses for VNR rings without identity...


This is a pseudo-ring, or rng, or ring-without-unit. The article linked in fact actually mentions the example of functions with compact support. The fact that you have a per-element neutral element is probably not sufficiently useful to give a special name to pseudo-rings with this additional property.

  • 4
    $\begingroup$ Except that it is a rng with the additional property of having "local" identies. There are rngs which do not have this property. $\endgroup$ – Thomas Andrews Nov 2 '12 at 20:03
  • $\begingroup$ Yes, I've come across pseudo-rings before. I was wondering whether the "local" identities, in Thomas' words, gave this thing a special title. I suppose it would deserve a special title if pseudo-rings of this sort had something especially useful about them, which is another question. $\endgroup$ – Alexander Sibelius Nov 2 '12 at 20:07
  • $\begingroup$ @ThomasAndrews: That's what I said (calling it a per-element neutral element instead). However the fact that some rngs don't have them does not imply that having them makes a great difference, algebraically speaking. I guess if it really made a big difference, the notion would have had a special name, but I'm really quite ignorant about this. $\endgroup$ – Marc van Leeuwen Nov 2 '12 at 20:09
  • $\begingroup$ Just thought I'd add that there's a nearby example of a rng without the "local" identity property. Alexander's ring $S$ sits in the larger ring of continuous functions $f : \mathbb{R} \to \mathbb{R}$ which vanish at infinity. Any nowhere-vanishing function in the latter ring has no "local" identity. $\endgroup$ – Mike F Jul 7 '13 at 4:29

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.