The strictly positive integers can be decomposed into a product of prime powers. Likewise, the positive rationals can be decomposed into a product of (possibly negative) prime powers.

Another way to state this is that the underlying multiplicative monoids are free and admit the primes as a basis.

If negative integers or rationals are allowed, then we no longer get a free monoid. But, we can still get a suitable prime decomposition if we create an additional "basis" element, representing -1, that takes exponents in $\Bbb Z/ 2 \Bbb Z$.

I'm wondering if a similar sort of decomposition exists for any of the more interesting structures thrown around in number theory, namely

  1. The profinite integers, and/or its group of units
  2. The ring of adeles, and/or its group of units
  3. The field of p-adic numbers for some p

I'm particularly interested in the case where the "basis" is countable, like the primes themselves.

Now, since these are all uncountable, it's clear that a decomposition into a direct sum of countable groups won't work. But perhaps we can allow infinitely many non-zero exponents, or allow the exponents to take values in some interesting group. Does anything like this work?

The supernatural numbers seem related here, but I can't tell exactly what the relationship is, or if they embed into one of the groups mentioned (or vice versa).


There is a notion of unique factorization domain, which is a commutative integral domain whose nonzero elements admit unique prime factorizations up to units. (If we don't ignore units then there's no hope of getting uniqueness.)

The $p$-adic integers $\mathbb{Z}_p$ have this property mostly by virtue of having a lot of units: every $p$-adic integer is uniquely of the form $p^n$ times a unit. In fact $p$ is the only prime in the $p$-adics! This reflects the fact that $\mathbb{Z}_p$ is a local ring.

The group of units is well understood. Let me assume that $p$ is odd because something annoying happens when $p = 2$. The group of units is isomorphic to $(\mathbb{Z}/p\mathbb{Z})^{\times}$ times the subgroup of units congruent to $1 \bmod p$. This subgroup is in turn isomorphic to the additive group $\mathbb{Z}_p$, with the isomorphism given by the exponential

$$\mathbb{Z}_p \ni a \mapsto (1 + p)^a \in \mathbb{Z}_p^{\times}.$$

In particular, it is "topologically generated" by $1 + p$.

The profinite integers and the adeles don't fall into this framework because they are not integral domains. In a ring that's not an integral domain it's very unclear what one ought to mean by unique prime factorization of a zero divisor. Nevertheless, since the multiplicative monoid and group of units functors both send products to products, once you've understood what's going on for $\mathbb{Z}_p$ you more or less understand what's going on in these more complicated cases.

  • $\begingroup$ The rationals aren't a UFD, but they admit the sort of structure I'm talking about, right? Likewise with the supernatural numbers. $\endgroup$ – Mike Battaglia Sep 22 '15 at 23:52
  • $\begingroup$ Sorry, to be technically correct, fields ARE UFD's -- but everything's a unit! Still... $\endgroup$ – Mike Battaglia Sep 22 '15 at 23:54
  • $\begingroup$ I guess the thing I'm really interested in is finding a decomposition for the group of units, as a starting point. $\endgroup$ – Mike Battaglia Sep 23 '15 at 2:36
  • $\begingroup$ @Matt: I give such a decomposition above for the $p$-adic integrs, and from there you can infer what's going on for the profinite integers, the adeles, and the $p$-adic numbers. $\endgroup$ – Qiaochu Yuan Sep 23 '15 at 2:38
  • 1
    $\begingroup$ @Mike: yes, it's different. In the usual embedding a natural number $n$ gets sent to (for each prime $p$) $p^k$ times whatever $p$-adic unit $\frac{n}{p^k}$ ends up being (where $p^k$ is the greatest power of $p$ dividing $n$). $\endgroup$ – Qiaochu Yuan Sep 23 '15 at 5:38

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.