5
$\begingroup$

What is $\mathrm{Hom}(\mathbb{Q},\mathbb{Z}(p^{\infty}))$?

I have a reference that says the group in question is $\mathbb{Q}_p,$ the additive group of the quotient field of the $p$-adic integers. Can anyone provide a reasonable derivation of this result?

$\endgroup$
  • $\begingroup$ I'm guessing that $\mathbb Z(p^\infty)$ means $\mathbb Z[\tfrac1p]/\mathbb Z$. $\endgroup$ – André Henriques May 17 '18 at 21:18
  • $\begingroup$ @Andre Henriques - Yes, that is correct. $\endgroup$ – Chris Leary May 18 '18 at 2:10
9
$\begingroup$

Note that this equality is as topological groups, where Hom is endowed with uniform convergence on compact subsets.

Every such homomorphism is zero on $p^n$ for some $n$. Hence the given homomorphism group $G=\mathrm{Hom}(\mathbf{Q},\mathbf{Z}(p^{\infty}))$ can be written as $\bigcup G_n$, where $G_n$ is the set of homomorphisms that vanish on $p^n$. Note that $G_n$ is open.

By definition, $$G_0=\mathrm{Hom}(\mathbf{Q}/\mathbf{Z},\mathbf{Z}(p^\infty))=\mathrm{Hom}(\bigoplus_\ell\mathbf{Z}(\ell^\infty),\mathbf{Z}(p^\infty))=\mathrm{End}(\mathbf{Z}(p^\infty)),$$ because for prime $\ell\neq p$ we have $\mathrm{Hom}(\mathbf{Z}(\ell^\infty),\mathbf{Z}(p^\infty))=0$. It is classical that $\mathrm{End}(\mathbf{Z}(p^\infty))$ is isomorphic to $\mathbf{Z}_p$ as a topological ring (details upon request).

Hence each $G_n$ is isomorphic to $\mathbf{Z}_p$. Multiplication by $p$ maps $G_n$ into $G_{n-1}$. Also, mapping $f$ to $x\mapsto f(p^{-1}x)$ maps $G_{n-1}$ into $G_n$ and is an inverse to multiplication by $p$. So indeed $pG_n=G_{n-1}$, and we deduce the result.


Edit: consider $\mathbf{Q}_p$ acting on itself as its own endomophism group (as topological group). Restricting to $\mathbf{Q}$ at the origin, and modding out by $\mathbf{Z}_p$ at the target yields a topological group homomorphism $\phi:\mathbf{Q}_p\to\mathrm{Hom}(\mathbf{Q}_p/\mathbf{Z}_p)$ (where the latter Hom is in topological groups). This homomorphism $\phi$ is injective: indeed, an element $f$ in the kernel of $\phi$ would, in restriction to $\mathbf{Q}$, satisfy $f(\mathbf{Q}_p)\subset\mathbf{Z}_p$. Since $\mathbf{Q}_p$ is divisible it has no nontrivial finite quotient and hence no nonzero homomorphism into any profinite group. So $f=0$.

Keeping in mind that there is a canonical isomorphism $\mathbf{Z}(p^\infty)\to\mathbf{Q}_p/\mathbf{Z}_p$, since we previously obtained that $\mathrm{Hom}(\mathbf{Q}_p/\mathbf{Z}_p)\simeq\mathbf{Q}_p$ as topological group, we deduce that $\phi$ is a isomorphism.

I should also add that it's important to phrase the result in terms of topological groups. Indeed, as an abstract group, $\mathbf{Q}_p$ is just isomorphic to $\mathbf{Q}^{(c)}$ (and also to $\mathbf{R}$).

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ This is a nice answer. However, I don't see the connection with $\mathrm{Hom}(\mathbb{Q},\mathbb{Z}(p^{\infty}))$. Am I being dense? $\endgroup$ – Chris Leary May 18 '18 at 2:15
  • 4
    $\begingroup$ @ChrisLeary It seems to me that $G$ is the Hom group you're looking for. $\endgroup$ – Najib Idrissi May 18 '18 at 6:51
  • $\begingroup$ @YCor - The edit helps. Thanks for your assistance. $\endgroup$ – Chris Leary May 18 '18 at 15:55
7
$\begingroup$

YCor's answer is good, but here's a slightly different way of seeing it.

Given $x \in \mathbb Q$, let $[f(x)]$ be any lift of $f(x)$ from $\mathbb Z[1/p]/\mathbb Z$ to $ \mathbb Z[1/p]$. Then $p^n [ f(1/p^n)]$ is a $p$-adic Cauchy sequence because $p^n [f(1/p^n)] - p^m [ f(1/p^m)]$ is an integer multiple of $p^{\min (n,m)}$. Changing the lift also causes a difference by an integer multiple of $p^n$ and thus does not affect the limit. This defines for each $f$ an element of $\mathbb Q_p$.

To go the other way, given $\alpha \in \mathbb Q_p$, observe that multiplication by $\alpha$ defines a map $\mathbb Q$ to $\mathbb Q_p$. Compose this with the projection to $\mathbb Q_p / \mathbb Z_p \cong \mathbb Z[1/p] / \mathbb Z$ to obtain a map $\mathbb Q \to \mathbb Z[1/p]$.

It is straightforward to check that these are inverses. Starting with $\alpha$, $[f(1/p^n)]$ is an element of $\mathbb Z_p$ plus $\alpha /p^n$ so $p^n [f(1/p^n)]$ is an approximation of $\alpha$. Starting with $f$, the limit $\alpha$ differs from $p^n [f(1/p^n)]$ by an element of $p^n \mathbb Z_p$, so $f(1/p^n)$ differs from $\alpha p^{-n}$ by an element of $\mathbb Z_p$, and then multiply these by elements of $\mathbb Z$ inverting every prime other than $p$ to get all rational numbers.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ I wonder if $\mathbb R$ can be obtained in some similar way... $\endgroup$ – მამუკა ჯიბლაძე May 18 '18 at 8:22
  • 1
    $\begingroup$ @მამუკაჯიბლაძე Hom between two discrete groups will always be totally disconnected. $\endgroup$ – YCor May 18 '18 at 12:35
  • 2
    $\begingroup$ @მამუკაჯიბლაძე There is a construction as maps $\mathbb Z \to \mathbb Z$ that are homomorphisms up to bounded error, modulo bounded functions. $\endgroup$ – Will Sawin May 18 '18 at 12:59
  • $\begingroup$ @Will Sawin - I like this answer very much as well. I appreciate your time and effort $\endgroup$ – Chris Leary May 18 '18 at 15:56

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.