Can a point on a curve always be mapped to another point with trivial residue map?

57 Views Asked by At

The following construction is taken from the book Central simple algebras and Galois cohomology by Gille and Szamuely in Chapter 6.4:

Let $C$ be a smooth projective curve over a perfect field $k$ with function field $K$. For an algebraic closure $\bar{k}$ of $k$, let $\bar{C} := C \times_k \bar{k}$ and let $G := \mathrm{Gal}(\bar{k}/k)$. The function field of $\bar{C}$ is by definition the composite $K\bar{k}$.

Let $C_0$ denote the closed points of $C$. For $P \in C_0$, $G$ permutes the closed points $Q$ in $\bar{C}_0$ lying over $P$, so we get a decomposition $$\mathrm{Div}(\bar{C}) = \bigoplus _{P \in C_0} \left(\bigoplus_{Q \mapsto P}\mathbb{Z} \right).$$ Therefore for each $i \geq 0$, the divisor map $(K\bar{k})^\times \rightarrow \mathrm{Div}(\bar{C})$ induces the map $$H^i(G,(K\bar{k})^\times) \rightarrow \bigoplus _{P \in C_0} H^i \left(G,\bigoplus_{Q \mapsto P}\mathbb{Z} \right) \cong \bigoplus _{P \in C_0}H^i(G_P,\mathbb{Z}),$$ where $G_P \subset G$ denotes the stabilizer of a fixed $Q$ in $\bar{C}$ and the last map follows from Shapiro's lemma. For $i \geq 2$, it is well-known that $H^i(G_P,\mathbb{Z}) \cong H^{i-1}(G_P,\mathbb{Q}/\mathbb{Z})$. For our case, we fix $i=2$ and a $P \in C_0$ to obtain the residue map $$r_P: H^2(G, (K\bar{k})^\times) \rightarrow H^1(G_P,\mathbb{Q}/\mathbb{Z}) = \mathrm{Hom}(G_P,\mathbb{Q}/\mathbb{Z}).$$ By our construction, we know that such maps are trivial for all but finitely many $P$.

Question. If $r_P$ is non-trivial, can we find an automorphism $\tau: C \rightarrow C$ and a $P' \in C_0$ such that $\tau(P) = P'$ and $r_{P'}$ is trivial?

Note that this is automatically true when $C$ is an elliptic curve. Also, for a curve of genus $\geq 2$, it is known that it has finitely many automorphisms, so if the answer to the question is affirmative, it would not be straightforward (to me).