Skip to main content
3 of 6
added 461 characters in body
Pace Nielsen
  • 18.7k
  • 4
  • 75
  • 137

As commented below, I had made a false claim. So here is a modification. Let $f(x)=4x^2+9x+4$. The roots are $(1/8)(-9\pm \sqrt{17})$. Note that $\mathbb{Q}(\sqrt{17})$ has class number 1. If we factor $2=p p'$ then the roots looks like $p^2/p'^2, p'^2/p^2$, up to units. In particular, the extensions of the maps $\varphi_{\alpha}$ treat $x$ as an element of valuation $2$. So, any element in the kernel of the extended map (much less a purported generator) must have constant term which has valuation $2$. Thus the constant term is divisible by $4$.

This means that the constant term of $\prod_{\nu}f_{\nu}(x)$ is divisible by $16$, and thus cannot (even up to units) equal $f(x)$.


For a monic example, let $f(x)=x^{2}-5x+2$. The roots are $(1/2)(5\pm \sqrt{17})$ (which are the two prime factors of $2$ in the ring of integers of $\mathbb{Q}(\sqrt{17})$). The extensions of the maps $\varphi_{\alpha}$ treat $x$ as an element of valuation $1$, so anything in the kernel must have constant term divisible by $2$. The product of two such polynomials has constant term divisible by $4$, hence cannot equal $f(x)$ even up to units.

Pace Nielsen
  • 18.7k
  • 4
  • 75
  • 137