Proof.
Fix
![$ c\in H^1(K,A)$](img32.png)
. There is a finite separable extension
![$ L$](img47.png)
of
![$ K$](img13.png)
such
that
![$ \res_L(c) = 0\in H^1(L,A)$](img48.png)
. Let
![$ J=\Res_{L/K}(A_L)$](img49.png)
be the
Weil restriction of scalars from
![$ L$](img47.png)
to
![$ K$](img13.png)
of the abelian variety
![$ A_L$](img50.png)
(see [
BLR90, §7.6]).
Thus
![$ J$](img16.png)
is an abelian variety over
![$ K$](img13.png)
of dimension
![$ [L:K]\cdot \dim(A)$](img51.png)
,
and for any scheme
![$ S$](img52.png)
over
![$ K$](img13.png)
, we have a natural (functorial)
group isomorphism
![$ A_L(S_L)\cong {}J(S)$](img53.png)
.
The functorial injection
![$ A(S) \hookrightarrow A_L(S_L) \cong {}J(S)$](img54.png)
corresponds via Yoneda's Lemma to a natural
![$ K$](img13.png)
-group scheme
map
![$ \iota: A\rightarrow J$](img34.png)
, and by construction
![$ \iota$](img17.png)
is a monomorphism.
But
![$ \iota$](img17.png)
is proper and thus
is a closed immersion (see [
Gro66, §8.11.5]).
Using the Shapiro lemma one finds, after a tedious computation, that
there is a canonical isomorphism
![$ H^1(K,J)\cong H^1(L,A)$](img55.png)
which identifies
![$ \iota_*(c)$](img56.png)
with
![$ \res_L(c)=0$](img57.png)
.