Let $p_n$ be the $n$th prime. Define $\psi: \Bbb{Z}[X] \to \prod\limits_{n \geq 1} \Bbb{Z}/(p_n): f \mapsto \prod\limits_{n\geq 1} \overline{f(n)}$ is this a ring monomorphism or does there exist a non-zero polynomial $f$ such that $p_n | f(n), n \geq 1$ ?

Let $\psi$ be the hom. If $\ker \psi \neq 0$, then there exists nonzero $g \in \ker \psi$. But $g \in \ker \phi_k$ as well where $\phi_k : \Bbb{Z}[X] \to \Bbb{Z}/(2) \times \Bbb{Z}/(3) \times \cdots \times \Bbb{Z}/(p_k)$, for all $k \geq 1$. Clearly $\ker \phi_{k+1} \subset \ker\phi_k$ so that $\ker \phi_1 \supset \ker \phi_2 \supset \dots$ is a descending chain of ideals.

$\ker \phi_1 = \{ f \in \Bbb{Z}[X] : 2 | f(1) \} = \{ f : \sum_{i=0}^{\deg f} c_i = 0 \pmod 2, \ c_i = $ coefficients of $f\}$.

  • 1
    $\begingroup$ @JackD'Aurizio : how do I get $(1,0,0,...)$ as an integer multiple of $(1,1,...)$ in $\prod_n \Bbb Z/p_n \Bbb Z$? This product is uncountable. $\endgroup$ – Watson Feb 4 '18 at 17:36
  • 1
    $\begingroup$ @JackD'Aurizio But the set of ring homomorphisms from $\mathbb{Z}[X]$ to any commutative ring is isomorphic to that ring. So there are plenty of choice for where to map it. $\endgroup$ – Tobias Kildetoft Feb 4 '18 at 17:52
  • 1
    $\begingroup$ @JackD'Aurizio You can send $X$ to $(1,2,3,\ldots)$, i.e., to $(\overline X)_n$. $\endgroup$ – Jose Brox Feb 4 '18 at 18:07
  • 1
    $\begingroup$ Uuuuu wow, 1 vote to close and 3 favorites. This is getting interesting! :D $\endgroup$ – Shine On You Crazy Diamond Feb 4 '18 at 18:12
  • $\begingroup$ I realized I was on the wrong side of it. For any prime $p$, let $g_p$ a generator of $\mathbb{Z}/(p\mathbb{Z})^*$. We may simple send $X$ into $(g_2,g_3,g_5,\ldots)$ and $1$ into $(1,1,1,\ldots)$. $\endgroup$ – Jack D'Aurizio Feb 4 '18 at 18:16

Your question can be reworded as: if I know how some polynomial in $\mathbb{Z}[x]$ acts on each $\mathbb{Z}/(p\mathbb{Z})$, do I know the coefficients of such polynomial? The answer is yes. Assuming two different polynomials $f(x),g(x)$ act in the exactly same way over each $\mathbb{Z}/(p\mathbb{Z})$, then their difference $h(x)$ is a non-zero polynomial which is $\equiv 0\pmod{p}$ for any prime $p$ and any $x\in\mathbb{Z}$. The Chinese remainder theorem then implies $h(x)=0$ for any $x\in\mathbb{Z}$, hence $h(x)$ has an infinite number of roots and $h(x)\equiv 0$.

In particular $\mathbb{Z}[X]$ embeds in $\mathbb{Z}_2\times\mathbb{Z}_3\times\ldots$ via $1\mapsto(1,1,1,\ldots)$ and $X\mapsto(1,2,3,\ldots)$.

  • $\begingroup$ Does the chinese remainder theorem apply even for an infinite cartesian product? $\endgroup$ – Shine On You Crazy Diamond Feb 4 '18 at 23:28
  • $\begingroup$ @EnjoysMath You could probably restrict to a number of ideals big enough, since the polynomial has a finite degree, but what I do not see is that $h(n)$ is 0 modulo $p$ for every $p$ and every $n\in\mathbb{Z}$. What I think we have is only that $h(n)$ is 0 modulo $p_n$ for every $n$ (so we have only restated the problem for the kernel of a linear map, from $f=g$ to $f=0$). $\endgroup$ – Jose Brox Feb 5 '18 at 10:15
  • $\begingroup$ @EnjoysMath Btw, what can be said for an infinite number of ideals $I_i$ is that if they are pairwise coprime then there is a monomorphism from $R/\bigcap I_i$ to $R/\prod I_i$ via $f(x+\bigcap I_i):=(x+I_i)_i$, but it does not need to be surjective anymore $\endgroup$ – Jose Brox Feb 5 '18 at 10:19

A theorem of Jordan of 1872 (see On a theorem of Jordan, by Serre) states that if $f\in\mathbb{Z}[X]$ is irreducible over $\mathbb{Q}$ and of degree 2 or more, then the set of primes $p$ such that $f$ has no zeros in $\mathbb{Z}_p$ has positive density (with $\pi(X)$ as denominator), so in particular this set is infinite.

  • $\begingroup$ Do you mean $\Bbb{Z}[X]$? $\endgroup$ – Shine On You Crazy Diamond Feb 4 '18 at 18:20
  • $\begingroup$ @EnjoysMath Yeah, of course :D $\endgroup$ – Jose Brox Feb 4 '18 at 18:21
  • $\begingroup$ I'm not quite seeing if you've said that it doesn't embed that way, or it does... please explaine more. Thanks $\endgroup$ – Shine On You Crazy Diamond Feb 4 '18 at 18:22
  • $\begingroup$ Sorry, thought that it would arise from this result in several manners, but none of them actually seem to work in a obvious way. I'm still thinking about it... $\endgroup$ – Jose Brox Feb 4 '18 at 19:29

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.