1

Consider the field $L = \mathbb{Q}_3(\alpha)$ where the minimal polynomial of $\alpha$ over $\mathbb{Q}_3$ is $f = x^4 - 3x^2 + 18$. Over $L$, the polynomial $f$ splits into $$ f = (x-\alpha)(x+\alpha)(x-\alpha')(x+\alpha') $$ where $\alpha' = \frac{(2\alpha^2-3)u}{\alpha}$ and $u = \sqrt{\frac{2}{-7}}$ (cf. this post, credits go to user KCd).

Question What is $\alpha/\alpha'$ modulo $\pi_L$ (a uniformizer in $L$)?

I know that result is only unique up to sign but I hope that does not cause any further difficulties.

I computed via Newton polygons that the valuations of $\alpha$ and $\alpha'$ are both $1/2$ (here, I assume $v(3) = 1$), so the reduction of $\alpha/\alpha'$ is non-zero.

I tried to apply a technique similar to an answer from this post:

I considered the polynomial $$ \frac{f(\alpha x)}{\alpha^4} = x^4 - \frac{3}{\alpha^2} x^2 + \frac{18}{\alpha^4} $$ whose roots are $\pm 1$ and $\pm \frac{\alpha'}{\alpha}$.

This means that the polynomial $$ \frac{x^4 - \frac{3}{\alpha^2} x^2 + \frac{18}{\alpha^4}}{x^2-1} = x^2 + \left(1-\frac{3}{\alpha^2}\right) $$ has the roots $\pm \frac{\alpha'}{\alpha}$.

Now I would like to compute the reduction of this polynomial. However, I am not sure that the reduction of $3/\alpha^2$ must be in this case. This seems to be a similar problem but it seems to be as easy an in the last mentioned post since we only have the second power of $\alpha$ (instead of the fourth power of $\alpha$) here.

Rotdat
  • 433

1 Answers1

2

Let $\beta=\frac{\alpha^2}{3}$. It follows that $\beta^2-\beta+2=0$, hence $\beta=\frac{1 \pm \sqrt{-7}}{2}$.

It follows that $(\frac{\alpha}{\alpha’})^2=\frac{-3+\sqrt{-7}}{4}$, if we choose $\beta=1+\sqrt{-7}$.

Now, note that in $\mathbb{F}_9$, $-7$ is actually a fourth power, as $(-7)^{8/4}=49=1$.

Thus the reduction of $\frac{\alpha}{\alpha’}$ in the residue field is $\frac{(-7)^{1/4}}{2}=-(-7)^{1/4}$.

If we want a more implicit description, we can check that by changing signs and considering the inverse ratio, we find that $\alpha/\alpha’$ can be any generator of $\mathbb{F}_9^{\times}$.

Aphelli
  • 34,439
  • In "if we choose" you meant $\beta=(1+\sqrt{-7})/2$ gives $(\frac{\alpha}{\alpha’})^2=\frac{-3+\sqrt{-7}}{4}$ ? Also $L= \Bbb{Q}3(\zeta_8,\zeta{16}3^{1/2})$ with the generator of the Galois group extending to $(\zeta_{16},3^{1/2})\to (\zeta_{16}^3,3^{1/2})$ might help to make those things explicit – reuns Dec 04 '20 at 19:36
  • @reuns: Well, $\alpha$ and $\alpha'$ can be switched around, so this gives two possibilities for $\beta$. So if we (if required) exchange $\alpha$ and $\alpha'$ so that $2\beta= (1+\sqrt{-7})/2$ then $\alpha^2/\alpha'^2$ has the given value. When I meant "implicit", I meant that, given the symmetries of the problem, $\alpha/\alpha'$ had four possible different values. And I showed they reduced to the four generators of $\mathbb{F}_9^{\times}$. But of course, if we define explicitly $\alpha$ and $\alpha'$ there is one well-defined result. – Aphelli Dec 05 '20 at 17:56