This follows from the existence of conjugate isogenies. Given $\phi:X\rightarrow Y$, there exists $\phi^{\prime}:Y\rightarrow X$ such that $\phi^{\prime}\circ\phi=[\mathrm{deg}(\phi)]$ on $X$ and $\phi\circ\phi^{\prime}=[\mathrm{deg}(\phi)]$ on $Y$.

Now $f(\phi^{\prime})\circ f(\phi)=[\mathrm{deg}(\phi)]$, so $\mathrm{ker}(f(\phi))\leq\mathrm{ker}([\mathrm{deg}(\phi)])$. Because $|\mathrm{ker}([\mathrm{deg}(\phi)])|$ is coprime to $\ell$, the kernel has the required property. Similarly $f(\phi)\circ f(\phi^{\prime})=[\mathrm{deg}(\phi)]$, so $\mathrm{im}(f(\phi))\geq\mathrm{im}([\mathrm{deg}(\phi)])$ and $\mathrm{coker}(f(\phi))$ is a quotient of $\mathrm{coker}([\mathrm{deg}(\phi)])$. The conclusion follows. ?

We first prove that $|f(A)|=|f(B)|$. Suppose $\phi:A\rightarrow B$ is an isogeny. Then

By Lemma 3, if we pick $\phi$ of degree coprime to a prime $\ell$, the right hand side has $\ell$-adic valuation 0. Doing this for a range of isogenies, we see that it equals 1, so $|f(A)|=|f(B)|$.

Now consider the functor $X\rightarrow f(X)[n]$ for some integer $n$. This meets the required conditions, so $|f(A)[n]|=|f(B)[n]|$ for all $n$. By the structure theorem for finite abelian groups, this is enough to show $f(A)\cong f(B)$. ?

The Mordell–Weil groups and $n$-Selmer groups are isomorphic by a direct application of Lemma 4 and Remark 5, and so are the Tamagawa numbers as $c_{v}(A)=|A(k_{v})/A_{0}(k_{v})|$. The Tate–Shafarevich groups are isomorphic as they are determined by the finite groups $\Sha[\ell^{n}]$ for primes $\ell$, and we can apply Lemma 4 to these. The equality of the $L$-functions and conductors follows from the existence of an isogeny $A\rightarrow B$.

For the regulators, we will prove that given an isogeny $\phi:A\rightarrow B$, we have

where $\hat{B}$ and $\hat{\phi}$ are the duals of $B$ and $\phi$. By picking isogenies, we can then show that the right hand side is coprime to any prime by Lemma 3 and hence the regulators are equal. To do this, recall that the regulator is defined as $|\mathrm{det}\langle a_{i},\hat{a}_{j}\rangle|$, where $\{a_{i}\}$ is a lattice basis for $A(k)/A(k)_{\mathrm{tors}}$ and $\{\hat{a}_{j}\}$ for $\hat{A}(k)/\hat{A}(k)_{\mathrm{tors}}$, and $\langle\_,\_\rangle$ is the height pairing. We will define $\{b_{i}\}$ and $\{\hat{b}_{j}\}$ similarly. As in ([5] proof of Theorem I.7.3), functoriality of the height pairing implies that $|\mathrm{det}\langle a_{i},\hat{\phi}(\hat{b}_{j})\rangle|=|\mathrm{det}\langle\phi(a_{i}),\hat{b}_{j}\rangle|$. By multilinearity of the determinant and height pairing, these determinants differ from the regulators of $A$ and $B$ by factors of $|\mathrm{coker}(\hat{\phi}(\hat{B}(k)/\hat{B}(k)_{\mathrm{tors}}))|$ and $|\mathrm{coker}(\phi(A(k)/A(k)_{\mathrm{tors}}))|$ respectively and the result follows.

Finally for the Tate modules $T_{\ell}(A)$ and $T_{\ell}(B)$, pick an isogeny $\phi$ of degree coprime to $\ell$. The map $[\mathrm{deg}(\phi)]$ is an isomorphism on $T_{\ell}(A)$ and $T_{\ell}(B)$, so the proof of Lemma 3 implies that $\phi$ induces an isomorphism of Tate modules as groups. Because $\phi$ commutes with the action of $G_{k}$ on points, it does on the Tate module also, so they are isomorphic as $\mathbb{Z}[G_{k}]$-modules.

?

See Curtis–Reiner ([2] §74, p. 507). ?

$M\otimes_{\mathbb{Z}}\mathbb{Z}_{\ell}$ is a $\mathbb{Z}_{\ell}[G]$-module, and in fact it is an ideal in $\mathbb{Z}[\zeta]\otimes_{\mathbb{Z}}\mathbb{Z}_{\ell}\cong\frac{\mathbb{Z}_{\ell}[X]}{(1+X+\ldots+X^{p-1})}$. This cyclotomic polynomial factorises into distinct irreducible factors $P_{1},\ldots,P_{t}$ over $\mathbb{Z}_{\ell}$. The only prime that divides the discriminant of $1+X+\ldots+X^{p-1}$ is $p$, so for $\ell\neq p$ the polynomials $P_{1},\ldots,P_{t}$ are coprime. If $\ell=p$, then $1+X+\ldots+X^{p-1}$ is irreducible so $t=1$. Therefore in either case we have

Therefore the ideal $M\otimes_{\mathbb{Z}}\mathbb{Z}_{\ell}$ is a product of ideals $M_{i}$ in $\frac{\mathbb{Z}_{\ell}[X]}{P_{i}(X)}$. Considering the $\mathbb{Z}_{\ell}$-rank of these tells us that $M_{i}$ is never the zero ideal, as

Each ring $\frac{\mathbb{Z}_{\ell}[X]}{P_{i}(X)}$ is the ring of integers of the cyclotomic extension $\frac{{\mathbb{Q}}_{\ell}[X]}{P_{i}(X)}$ ([7] Ch. IV, §4, Prop. 16 and 17), a local field, so is a principal ideal domain. Therefore $M_{i}$ is a non-zero principal ideal so is isomorphic to $\frac{\mathbb{Z}_{\ell}[X]}{P_{i}(X)}$ as $\mathbb{Z}_{\ell}[G]$-modules, so $M\otimes_{\mathbb{Z}}\mathbb{Z}_{\ell}\cong\mathbb{Z}[\zeta]\otimes_{\mathbb{Z}}\mathbb{Z}_{\ell}$. The same is true for $N\otimes_{\mathbb{Z}}\mathbb{Z}_{\ell}$ so the result follows. ?

First let us fix some notation. Let $n:=p-1$. As in ([6] §2), we have isomorphisms $\psi_{M}:\mathbb{Z}^{n}\rightarrow M$ and $\psi_{M\otimes A}:(A_{\bar{k}})^{n}\rightarrow(M\otimes A)_{\bar{k}}$, and similarly for $N$. Denote a choice of isomorphism $\mathrm{Aut}(\mathbb{Z}^{n})\rightarrow\mathrm{Aut}_{\bar{k}}(A^{n})$, as used in ([6] §2) in the construction of $M\otimes A$ and $N\otimes A$, by $\rho$. For now, let the action of $\sigma\in G_{k}$ on a module or variety $X$ be written as $\chi_{X}(\sigma)$, and then define cocycles from $G_{k}$ to $\mathrm{Aut}(\mathbb{Z}^{n})$ and $\mathrm{Aut}_{\bar{k}}(A^{n})$ by

and similarly for $N$. Note that we view $\mathbb{Z}^{n}$ as a trivial Galois module. By construction, $\rho(s_{M})=s_{M\otimes A}$ and likewise for $N$. Henceforth, we will drop the notation $\chi_{X}$.

Now suppose there is an isomorphism $\phi_{A}:M\otimes A\rightarrow N\otimes A$ defined over $k$. We will reverse Milne’s construction, and show there is a $\mathbb{Z}[G_{k}]$-module isomorphism $\phi:M\rightarrow N$. Note that as $\phi_{A}$ is an isomorphism, $\psi_{N\otimes A}^{-1}\phi_{A}\psi_{M\otimes A}$ is an automorphism of $A^{n}$, so we can define $\phi$ by $\psi_{N}^{-1}\phi\psi_{M}=\rho^{-1}(\psi_{N\otimes A}^{-1}\phi_{A}\psi_{M\otimes A})$. As $\psi_{N}^{-1}\phi\psi_{M}$ is an isomorphism, so is $\phi$.

The fact that $\phi_{A}$ is defined over $k$ is equivalent to the fact that $\sigma\phi_{A}=\phi_{A}\sigma$ for all $\sigma\in G_{k}$. We shall prove the equivalent property for $\phi$, which implies $\phi$ is an isomorphism of $\mathbb{Z}[G_{k}]$-modules.

Note that in our case $\psi_{N\otimes A}^{-1}\phi_{A}\psi_{M\otimes A}$ is an automorphism of $A^{n}$ defined over $\bar{k}$. However all of these are given by $\mathrm{GL}_{n}(\mathbb{Z})$ and defined over $k$, so this map commutes with the action of $G_{k}$. Hence for any $\sigma\in G_{k}$, we have the following equality of maps $(A_{\bar{k}})^{n}\rightarrow(A_{\bar{k}})^{n}$:

where the second equality holds because $\sigma$ commutes with $\psi_{N\otimes A}^{-1}\phi_{A}\psi_{M\otimes A}$, and the third because it commutes with $\phi_{A}$.

Now apply $\rho^{-1}$ to this to get

Because Galois acts trivially on $\mathbb{Z}^{n}$ this implies

which tells us that $\phi$ commutes with $\sigma$. Hence $\phi$ is an isomorphism of $\mathbb{Z}[G_{k}]$-modules. ?

Let $M$ and $N$ be two ideals of $\mathbb{Z}[\zeta_{23}]$ in different ideal classes (23 can be replaced by any prime such that $\mathbb{Z}[\zeta_{23}]$ has non-trivial class group). Then, by Lemmas 7 and 8, they are not isomorphic as $\mathbb{Z}[C_{23}]$-modules but $M\otimes\mathbb{Z}_{\ell}\cong N\otimes\mathbb{Z}_{\ell}$ for all primes $\ell$. This implies $M\otimes\mathbb{Z}_{(\ell)}\cong N\otimes\mathbb{Z}_{(\ell)}$ ([2] Cor. 76.9). Therefore there is an injective homomorphism $M\rightarrow N$ with finite cokernel of order coprime to $\ell$.

Now pick an elliptic curve $E$ over $\mathbb{Q}$, with no potential complex multiplication. Pick a number field $L$ with $\mathrm{Gal}(L/{\mathbb{Q}})\cong C_{23}$, and let $G_{{\mathbb{Q}}}$ act on $M$ and $N$ via the corresponding $C_{23}$ quotient. Then $A:=M\otimes E$ and $B:=N\otimes E$ are related by an isogeny of degree coprime to $\ell$ for each $\ell$ (by Lemma 9), but are not isomorphic (by Lemma 10). ?