I'm looking for solutions $t$ of an equation of the form $$ t \sigma(t) \cdots \sigma^{n-1}(t) = v $$ in a field equipped with an automorphism $\sigma$ of order $n$. In this case, I call $t$ a "$\sigma$-semilinear $n$-th root of $v$".

I have a particular situation in mind, coming from the study of $\varphi$-modules (see the motivation provided at the end of the question). This means that we are working over a $p$-adic field, and $\sigma$ is the Frobenius automorphism.
Allow me then to make the question more specific and precise.

Let $K / \mathbb{Q}_p$ be a finite extension, with ring of integers $\mathcal{O}_K$. Choose a uniformiser $\varpi$ of $\mathcal{O}_K$.
Let $K_0$ denote the maximal unramified intermediate extension $K / K_0 / \mathbb{Q}_p$. Write $f = [K_0 : \mathbb{Q}_p]$ for the inertial degree of $K/\mathbb{Q}_p$.

We know that $K_0 / \mathbb{Q}_p$ is a cyclic extension of degree $f$. Its Galois group is generated by the Frobenius automorphism, which we'll call $\sigma \in \text{Gal}(K_0 / \mathbb{Q}_p)$.

Question: I am looking for an explicit formula for $t \in K_0 \otimes_{\mathbb{Q}_p} K$, say $$ t = \sum_i x_i \otimes y_i, $$ such that $$ \sum_i x_i y_i = \varpi$$ and $$ \sum_i \sigma^j(x_i) y_i = 1,$$ for all $j = 1, \ldots, f-1$.

Such an element is then necessarily a $\sigma$-semilinear $f$-th root of $1 \otimes \varpi$. There are others (for instance, $\sigma(t)$ is another one; or one could throw in some units), but this particular one seems to be the element that comes up (see the motivation at the end).

Is there a general formula for $t$ that works for all $K$? Note that such $t$ is in fact integral, meaning that it lives in $\mathcal{O}_{K_0} \otimes_{\mathbb{Z}_p} \mathcal{O}_K$. So there should be no need to introduce denominators.
If you know of a formula for the special case when $K$ is unramified, so that $K = K_0$ and $\varpi = p$, I'd also love to hear it.

Motivation: The motivation for the question comes from $\varphi$-modules. The $\varphi$-module attached (contravariantly) to the Lubin–Tate formal group $\text{LT}_{K,\varpi}$ is a free $K_0 \otimes_{\mathbb{Q}_p} K$-module of rank $1$ on which $\varphi^f$ acts by multiplication by $1 \otimes \varpi$. The element $t$ then provides a natural ($\sigma$-semilinear) $f$-th root of $\varpi$, i.e. $$ t \sigma(t) \cdots \sigma^{f-1}(t) = \varpi.$$ It should then be the case that the ($\sigma$-semilinear) action of $\varphi$ on the $\varphi$-module attached to $\text{LT}_{K,\varpi}$ is given by $t$. The question is thus of finding an explicit description of this $\varphi$-module.

  • 1
    $\begingroup$ What happens in the extreme cases, namely when $K$ is either totally ramified or unramified over $\Bbb Q_p$? $\endgroup$ – Lubin Apr 5 '16 at 17:41
  • $\begingroup$ @Lubin If $K/\mathbb{Q}_p$ is totally ramified, then $f = 1$ and we can take $t = \varpi$. I don't know how it works if $K$ is unramified. I am hoping there is some expression involving $(p^f-1)$-th roots of unity that could work, but I haven't found a sensible way to approach it yet. $\endgroup$ – Will Apr 5 '16 at 18:28

I have worked out an explicit formula in the special case that $K_0/\mathbb{Q}_p$ is quadratic, i.e. $f=2$.
Let $\zeta = \zeta_{p^2-1}$ be a primitive $(p^2-1)$-th root of unity in $K_0$, so that $K_0 = \mathbb{Q}_p(\zeta)$.

Define then $t \in K_0 \otimes_{\mathbb{Q}_p} K$ by: $$ t = 1 \otimes \left ( \varpi + (1-\varpi) \frac{2 \zeta^{p+1} - \zeta (\zeta+\zeta^p)}{4\zeta^{p+1}-(\zeta+\zeta^p)^2 }\right ) + \zeta \otimes \left ( (1-\varpi) \frac{ -(\zeta+\zeta^p)+2\zeta}{4\zeta^{p+1}-(\zeta+\zeta^p)^2 }\right ).$$ Note that $\zeta^{p+1}$ and $(\zeta+\zeta^p)$ are both elements of $\mathbb{Z}_p$, because they are the norm (respectively, the trace) of $\zeta$. Moreover, $(\zeta+\zeta^p)^2-4\zeta^{p+1}$ is the discriminant of the extension $\mathbb{Q}_p(\zeta)/\mathbb{Q}_p$; as this extension is unramified, it is a unit. This means that the above formula in fact defines an integral element $t \in \mathcal{O}_{K_0} \otimes_{\mathbb{Z}_p} \mathcal{O}_K$.

Using the identities $\zeta^2 = (\zeta+\zeta^p)\zeta - \zeta^{p+1}$ and $\zeta^p = (\zeta+\zeta^p)-\zeta$, it is straightforward to check that indeed $t$ satisfies the formulas of the question.


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.