1 Introduction
1.1 Context and motivation
Hilbert’s tenth problem inquires whether there exists a universal algorithm to determine the solvability of any given Diophantine equation over the integers
$\mathbb {Z}$
. This problem was resolved negatively by Matijasevich [Reference MatijasevichMat70], who demonstrated that no such algorithm exists. Extending this inquiry to number fields, we can ask whether a procedure can be devised to decide the solvability of Diophantine equations over the ring of integers
$\mathcal {O}_K$
of a number field K.
It is established that if the ring of rational integers
$\mathbb {Z}$
can be described as a Diophantine set within
$\mathcal {O}_K$
, then the analogue of Hilbert’s tenth problem for
$\mathcal {O}_K$
also has a negative solution. The Denef–Lipshitz conjecture posits that
$\mathbb {Z}$
is indeed a Diophantine subset of
$\mathcal {O}_K$
for any number field K. This conjecture has been validated in several specific instances:
-
• when K is totally real [Reference DenefDen80];
-
• when K is a quadratic extension of a totally real number field [Reference Denef and LipshitzDL78];
-
• when K has exactly one complex place [Reference PheidasPhe88, Reference ShlapentokhShl89];
-
• when $K/\mathbb {Q}$
is abelian [Reference Shapiro and ShlapentokhSS89].
It is natural to seek to prove the conjecture for families of number fields for which the above results do not apply. Mazur et al. [Reference Mazur, Rubin and LarsenMRL18] introduced the notion of Diophantine stability for a variety defined over a number field. By studying this notion for abelian varieties, they were able to prove that Hilbert’s tenth problem has a negative answer for certain large families of number fields. Garcia-Fritz and Pasten [Reference Garcia-Fritz and PastenGFP20] confirmed the Denef–Lipshitz conjecture for fields of the form
$\mathbb {Q}(p^{1/3}, \sqrt {-q})$
, with p and q being primes from specific sets characterized by positive Tchebotarev density. Recently, Kundu et al. [Reference Kundu, Lei and SprungKLS24] extended these results, proving the conjecture for a broader family of number fields
$\mathbb {Q}(p^{1/3}, \sqrt {Dq})$
, where D ranges over a finite set of integers, and p and q are primes within explicit sets with positive Tchebotarev density.
The number fields examined in these studies are all of degree less than 6. This raises the question of whether the conjecture can be proven for families of non-abelian number fields with unbounded degrees. Partial progress has been made by the second-named author [Reference RayRay23]. Consider a number field K and its cyclotomic
$\mathbb {Z}_p$
-extension
$K_{\mathrm {cyc}}$
, with
$K_n$
as the intermediate extension of degree
$p^n$
, where p is an odd prime. If an elliptic curve
$E/K$
satisfies certain local conditions at p, then
$\mathbb {Z}$
is a Diophantine subset of
$\mathcal {O}_{K_n}$
for all n. Starting from an imaginary quadratic field K and examining the anticyclotomic
$\mathbb {Z}_p$
-extension, similar results were obtained by the second-named author and Weston for the anticyclotomic
$\mathbb {Z}_p$
-extension of K [Reference Ray and WestonRW24]. This used the theory of congruences between Galois representations, and the techniques of Greenberg and Vatsal [Reference Greenberg and VatsalGV00], who study the effect of such congruences in Iwasawa theory. This leads to explicit examples; for instance, it is shown that in all finite layers of the anticyclotomic
$\mathbb {Z}_3$
-extension of
$\mathbb {Q}(\!\sqrt {-5})$
, Hilbert’s tenth problem has a negative answer.
1.2 Main results
The goal of the present paper is to start with an elliptic curve over an imaginary quadratic field K. Since K is abelian, Hilbert’s tenth problem has a negative solution in all number fields contained in the cyclotomic
$\mathbb {Z}_p$
-extension of K. Note that to obtain interesting new results, we wish to consider a family of
$\mathbb {Z}_p$
-extensions of K, other than the cyclotomic one. Before presenting our main results, we introduce some additional notation. Let
$ E_{/\mathbb {Q}} $
be an elliptic curve and
$ K_\infty $
the
$\mathbb {Z}_p^2$
-extension of
$ K $
. Define
$ \Gamma _\infty := \mathrm {Gal}(K_\infty /K) $
and
$ \Gamma _{\mathrm {cyc}} := \mathrm {Gal}(K_{\mathrm {cyc}}/K) $
. Let
$ \sigma , \tau \in \Gamma _\infty $
be topological generators with
Set
$ X+1 = \sigma $
and
$ Y+1 = \tau $
, allowing us to identify
$ \Lambda _\infty $
with the power series ring
$ \mathbb {Z}_p[\![ X,Y]\!] $
. Consider
$ \mathbb {P}^1(\mathbb {Z}_p) $
, the space of lines in
$ \mathbb {Z}_p^2 $
. Each line has a representative
$ (a,b) \in \mathbb {Z}_p^2 $
with
$ p \nmid a $
or
$ p \nmid b $
, uniquely determined up to a unit in
$ \mathbb {Z}_p $
. For
${ (a,b) \in \mathbb {P}^1(\mathbb {Z}_p) }$
, there is a unique
$\mathbb {Z}_p$
-extension
$ K_{a,b}/K $
such that
For
$n\in \mathbb {Z}_{\geq 0}$
, let
$K_{a,b}^{(n)}/K$
be the unique extension contained in
$K_{a,b}$
such that
${[K_{a,b}^{(n)}:K]=p^n}$
. Denote by
$E^{(K)}$
the twist of E over K. The p-primary Selmer group of E over
$\mathbb {Q}$
is denoted
$\mathrm {Sel}_{p^\infty }(E/\mathbb {Q})$
. Given a prime
$\ell $
, set
$c_\ell (E/\mathbb {Q})$
to be the Tamagawa number of
$E_{/\mathbb {Q}_\ell }$
. Assuming that E has good reduction at p, take
$\widetilde {E}(\mathbb {F}_p)$
to be the group of
$\mathbb {F}_p$
-points of the reduction of E at p.
Theorem A (Theorems 4.13 and 4.14).
Let
$E_{/\mathbb {Q}}$
be an elliptic curve, p an odd prime and K be an imaginary quadratic field. Assume that the following conditions hold:
-
(1) $\mathrm {corank}_{\mathbb {Z}_p} \mathrm {Sel}_{p^\infty }(E/\mathbb {Q})=1$
and
$\mathrm {corank}_{\mathbb {Z}_p} \mathrm {Sel}_{p^\infty }(E^{(K)}/\mathbb {Q})=0$
; -
(2) E and $E^{(K)}$
have good ordinary reduction at p; -
(3) $\widetilde {E}(\mathbb {F}_p)[p]=0$
and
$\widetilde {E}^{(K)}(\mathbb {F}_p)[p]=0$
; -
(4) $p\nmid \prod _{\ell \neq p} c_\ell (E/\mathbb {Q})\times \prod _{\ell \neq p} c_\ell (E^{(K)}/\mathbb {Q})$
; -
(5) the normalized p-adic regulator of E over $\mathbb {Q}$
is a p-adic unit; -
(6) $\mathrm{III}(E/\mathbb {Q})[p^\infty ]=0$
and
$\mathrm{III}(E^{(K)}/\mathbb {Q})[p^\infty ]=0$
.
There is a line
$(a_0, b_0)\in \mathbb {P}^1(\mathbb {F}_p)$
such that for all
$(a,b)\not \equiv (a_0, b_0)\ (\text {mod}\ {p})$
and for all
$n\in \mathbb {Z}_{\geq 0}$
, Hilbert’s tenth problem has a negative answer for
$K_{a,b}^{(n)}$
. Moreover, one can take
$(a_0, b_0)\neq (1, 0)$
.
In many of the cases we consider, the rank of the elliptic curve in question is unbounded in the anticyclotomic
$\mathbb {Z}_p$
-extension, due to the presence of Heegner points. Thus, we are not in general able to obtain a result that applies to all
$\mathbb {Z}_p$
-extensions of K. For
$p=3$
, we show unconditionally that there is a large family of imaginary quadratic fields for which an elliptic curve
$E_{/\mathbb {Q}}$
satisfies the above conditions; the corresponding assertions for
$p=11,13,31,37$
are conditional as specified in Theorem B. It is convenient to introduce a shorthand for the assertions made in the above-mentioned results.
Definition 1.1. We say that the pair
$(K, p)$
satisfies Criterion
$(\textbf {H10-gen})$
if there exists a line
$(a_0,b_0)\in \mathbb {P}^1(\mathbb {F}_p)$
such that
$K_{a,b}^{(n)}/\mathbb {Q}$
is integrally Diophantine for all
$(a,b)$
that are not congruent to
$(a_0,b_0)$
modulo p. Moreover,
$(a_0, b_0)$
can be taken to not be equal to
$(1, 0)$
.
The criterion stands for ‘Hilbert’s tenth problem in generic
$\mathbb {Z}_p$
-extensions’. Endow
$\mathbb {P}^1(\mathbb {Z}_p)$
with the natural measure induced from p-adic valuation on
$\mathbb {Z}_p$
. The condition implies that Hilbert’s tenth problem is negative for all finite layers in
$K_{(a,b)}$
for a density of at least
$(1-(p+1)^{-1})$
of all
$(a,b)\in \mathbb {P}^1(\mathbb {Z}_p)$
. In this section, we fix an odd prime number p and write
$K=\mathbb {Q}(\!\sqrt {d})$
, where d is a squarefree negative integer. Set
${\mathrm {ht}(K):=|d|}$
and consider the following natural question.
Question 1.2. Fix an odd prime p. When ordered by their height, is there a positive density of imaginary quadratic fields K for which the Criterion ( H10-gen ) is satisfied?
If the answer to the above question is yes, it shows that there are indeed large families of imaginary quadratic fields for which our results hold, thus giving rise to many new families of number fields for which Hilbert’s tenth problem has a negative answer.
Theorem B (Theorems 5.5 and 5.14).
-
(1) For $p=3$
, there is unconditionally a positive density of imaginary quadratic fields for which Criterion (
H10-gen
) holds. -
(2) Let $p\in \{11,13,31,37\}$
. Suppose that, for at least one of the triples
$(E,K_0,p)$
listed in Example 5.4, the twists in the associated positive-density rank-zero family for which $$ \begin{align*} \mathrm{III}(E^{(d)}/\mathbb{Q})[p^\infty]=0 \end{align*} $$have positive relative lower density in that family. Then, there is a positive density of imaginary quadratic fields for which Criterion ( H10-gen ) holds.
1.3 Organization and methodology
Including the introduction, the article consists of five sections. Section 2 is dedicated to establishing preliminary notions and setting up relevant notation. In this section, we discuss the algebraic structure of Selmer groups over infinite towers of number fields. Moreover, we discuss basic notions related to Hilbert’s tenth problem and the notion of an integrally Diophantine extension. A criterion of Shlapentokh (Theorem 2.6) relates the stability of ranks of elliptic curves to Hilbert’s tenth problem. Given an extension of number fields
$L/K$
, if there is an elliptic curve
$E_{/K}$
such that
$\mathrm {rank}\ E(L)=\mathrm {rank}\ E(K)>0$
, then,
$L/K$
is an integrally Diophantine extension. This is the main tool that we use in establishing our results in subsequent sections. Iwasawa theory is used to give explicit conditions for the rank of an elliptic curve to remain constant in a
$\mathbb {Z}_p$
-extension. In Section 3, we discuss the role of the Euler characteristic in Iwasawa theory and its relevance to the study of ranks of elliptic curves. We discuss an explicit formula due to Perrin-Riou and Schneider, which is used to calculate Iwasawa invariants of Selmer groups. In Section 4, we prove the main results of this article, namely Theorems 4.12, 4.13 and 4.14. Given an imaginary quadratic field K and an odd prime p, we give explicit conditions for there to exist an elliptic curve
$E_{/\mathbb {Q}}$
with positive rank, such that the rank remains constant in many
$\mathbb {Z}_p$
-extensions
$K_{a,b}/K$
. This construction relies on understanding the structure of the Selmer group over the
$\mathbb {Z}_p^2$
-extension of K and relating it to that over
$K_{a,b}$
. Explicit examples are provided at the end of this section; see Examples 4.15 and 4.16. In Section 5, we study the conditions of Theorems 4.12, 4.13 and 4.14 in families of imaginary quadratic fields. The strategy used here is motivated (in part) by the method of Gajek-Leonard et al. [Reference Gajek-Leonard, Hatley, Kundu and LeiGLHKL25], who prove Mazur’s growth-number conjecture for the
$\mathbb {Z}_p^2$
-extension of an imaginary quadratic field, provided a number of additional conditions are satisfied. Results of Bhargava et al. [Reference Bhargava, Klagsbrun, Lemke Oliver and ShnidmanBKLOS19] give the unconditional
$p=3$
assertion in Theorem B. The results obtained from Kriz–Li [Reference Kriz and LiKL19] for
$p=11,13,31,37$
are conditional on a positive-relative-density vanishing statement for the relevant p-primary Tate–Shafarevich groups.
1.4 Outlook
Our results illustrate that Iwasawa theory can be employed as a novel tool to establish new cases of the Denef–Lipshitz conjecture. When combined with arithmetic statistics, these techniques give unconditional positive-density families for
$p=3$
and conditional positive-density families in the Kriz–Li setting described above. The Iwasawa theory of elliptic curves was introduced by Mazur [Reference MazurMaz72], whose original motivation was to study the growth of ranks of elliptic curves in
$\mathbb {Z}_p$
-towers. We consider a broader application of Iwasawa theory, which is in keeping with its traditional scope. One may seek to explore more generalized contexts, specifically those involving p-adic Lie extensions of dimension
$d> 1$
. Harris [Reference HarrisHar00] provided general upper bounds for the asymptotic growth of the rank of the Mordell–Weil group of an abelian variety in a tower of number fields contained within a p-adic Lie extension. These upper bounds are crucial as they give a framework for understanding how the rank behaves as we move up through the layers of the extension. Despite this progress, the precise order of growth in Mordell–Weil ranks remains an open question and is not fully understood, as highlighted in [Reference Hung and LimHL20, Question following Proposition 4.7].
2 Preliminaries
In this section, we examine foundational concepts. In Section 2.1, we introduce Selmer groups associated with elliptic curves and their connections to the Mordell–Weil group and the Tate-Shafarevich group. These Selmer groups are analysed over different
$\mathbb {Z}_p$
-extensions of an imaginary quadratic field K. In Section 2.2, we address Hilbert’s tenth problem and the relative notion of integrally Diophantine extensions of number fields.
2.1 Selmer groups and their Iwasawa theory
Throughout, d is a squarefree negative integer,
$K := \mathbb {Q}(\!\sqrt {d})$
and
$ p $
is an odd prime number. Set
$ K_\infty /K$
to denote the
$\mathbb {Z}_p^2$
-extension of
$ K $
. The cyclotomic
$\mathbb {Z}_p$
-extension
$ K_{\mathrm {cyc}}/K $
is the unique
$\mathbb {Z}_p$
-extension of K contained within
$ K(\mu _{p^\infty }) $
. However, the anticyclotomic
$\mathbb {Z}_p$
-extension
$ K_{\mathrm {ac}}/K $
is a Galois extension over
$\mathbb {Q}$
, where the induced action of
$\mathrm {Gal}(K/\mathbb {Q})$
on
$\mathrm {Gal}(K_{\mathrm {ac}}/K)$
is nontrivial. We note that
$ K_{\mathrm {cyc}} $
and
$ K_{\mathrm {ac}} $
are both subfields of
$ K_\infty $
and that
$ K_{\infty }=K_{\mathrm {cyc}} \cdot K_{\mathrm {ac}} $
. Set
$\Gamma _\infty := \mathrm {Gal}(K_\infty /K)$
,
$\Gamma _{\mathrm {cyc}} := \mathrm {Gal}(K_{\mathrm {cyc}}/K)$
and
$\Gamma _{\mathrm {ac}} := \mathrm {Gal}(K_{\mathrm {ac}}/K)$
.
Given a pro-p group
$\mathcal {G}$
, the Iwasawa algebra is defined as the inverse limit
where U runs over all finite index normal subgroups of
$\mathcal {G}$
. The Iwasawa algebras associated with
$\Gamma _\infty $
,
$\Gamma _{\mathrm {cyc}}$
and
$\Gamma _{\mathrm {ac}}$
are denoted by
$\Lambda _\infty $
,
$\Lambda _{\mathrm {cyc}}$
and
$\Lambda _{\mathrm {ac}}$
, respectively. Note that
$\Gamma _{\mathrm {cyc}}$
and
$\Gamma _{\mathrm {ac}}$
are both quotients of
$\Gamma _\infty $
, and that the natural quotient maps induce surjections at the level of Iwasawa algebras:
The Iwasawa theory of elliptic curves concerns itself with the structure of Selmer groups, viewed over infinite Galois extensions. These Selmer groups naturally give rise to modules over Iwasawa algebras and it is the structure theory of such modules that gives rise to Iwasawa-theoretic arithmetic invariants. The elliptic curves
$E_{/\mathbb {Q}}$
on which we focus in this article are defined over
$\mathbb {Q}$
and have good ordinary reduction at p.
Let
$E_{/\mathbb {Q}} $
be an elliptic curve and denote by
$ E^{(K)} $
the quadratic twist of
$ E $
by
$ K $
. Assume that both E and
$E^{(K)}$
have good ordinary reduction at p. Let
$ N_E $
denote the conductor of
$ E $
and let
$\Sigma $
be a finite set of rational primes containing the primes that divide
$ N_E p $
. Denote by
$\mathbb {Q}_\Sigma $
the maximal algebraic extension of
$\mathbb {Q}$
in which the finite primes outside
$\Sigma $
are unramified. We introduce the p-primary Selmer group of E over an algebraic extension of
$\mathbb {Q}$
. For any number field
$ F $
, let
$\Sigma (F)$
be the set of primes
$ v $
of
$ F $
that lie above
$\Sigma $
. Assume that F is contained in
$\mathbb {Q}_\Sigma $
, that is,
$\Sigma $
contains the primes that ramify in F. For
$\ell \in \Sigma $
, set
Let
$ \mathcal {F}/F$
be an infinite pro-p extension of F and denote
where
$ F'/F$
ranges over number field extensions of F contained in
$\mathcal {F}$
. The
$ p $
-primary Selmer group of
$ E $
over
$ \mathcal {F} $
is then defined as follows:
We take note of a useful statement about Selmer groups and change of base.
Proposition 2.1. Suppose
$ E/\mathbb {Q} $
is an elliptic curve and let
$ E^{(K)} $
be the twist of
$ E $
corresponding to
$ K $
. Then,
Proof. From Shapiro’s lemma, we deduce that
and
These isomorphisms are compatible with the restriction maps and we thus find that
Let
$ \sigma , \tau \in \Gamma _\infty $
be topological generators satisfying
Denote
$ X+1 $
as
$ \sigma $
and
$ Y+1 $
as
$ \tau $
. Then, we can identify
$ \Lambda _\infty $
with the ring of power series
$\mathbb {Z}_p[\![ X,Y]\!] $
. Let
$\mathbb {P}^1(\mathbb {Z}_p)$
be the space of lines in
$\mathbb {Z}_p^2$
. Any line has a representative
$ (a,b) \in \mathbb {Z}_p^2 $
such that
$ p \nmid a $
or
$ p \nmid b $
, which is uniquely determined up to multiplication by a unit in
$\mathbb {Z}_p$
. Given
$ (a,b) \in \mathbb {P}^1(\mathbb {Z}_p) $
, there is a unique
$\mathbb {Z}_p$
-extension
$ K_{a,b}/K $
such that
For
$ (a,b) \in \mathbb {P}^1(\mathbb {Z}_p) $
, set
$\Gamma _{a,b} = \mathrm {Gal}(K_{a,b}/K)$
,
$ H_{a,b} = \overline {\langle \sigma ^a \tau ^b \rangle } $
, and
$\Lambda _{a,b} = \mathbb {Z}_p[\![ \Gamma _{a,b} ]\!] $
. Define
$\pi _{a,b} : \Lambda _\infty \rightarrow \Lambda _{a,b}$
to be the map induced by the natural projection
$ \Gamma _\infty \rightarrow \Gamma _{a,b} $
. Set
$\omega _{a,b}:= (1+X)^a(1+Y)^b - 1 $
. For
$(a,b) = (1:0)$
, use
$\Gamma _{\mathrm {cyc}}$
,
$ H_{\mathrm {cyc}}$
and
$\pi _{\mathrm {cyc}}$
instead of
$\Gamma _{(1:0)}$
,
$ H_{(1:0)}$
and
$\pi _{(1:0)}$
, respectively. Choose a topological generator
$\gamma =\gamma _{a,b}$
of
$\Gamma _{a,b}$
and set
$T=T_{a,b}:=\gamma -1$
. Then, we identify
$\Lambda _{a,b}$
with the formal power series ring
$\mathbb {Z}_p[\![ T]\!] $
. In particular, identify
$\Lambda _{\mathrm {{cyc}}}$
(respectively
$\Lambda _{\mathrm {{ac}}}$
) with
$\mathbb {Z}_p[\![ T]\!] $
.
The Pontryagin dual of a module M over
$\mathbb {Z}_p[\![ T]\!] $
is defined by
We say that M is cofinitely generated (respectively cotorsion) over
$\mathbb {Z}_p[\![ T]\!] $
if
$M^\vee $
is finitely generated (respectively torsion). A monic polynomial
$ f(T) \in \mathbb {Z}_p[\![ T ]\!] $
is called a distinguished polynomial if all its coefficients, except for the leading one, are divisible by
$ p $
. Let
$ M $
and
$ M' $
be cofinitely generated and cotorsion
$\mathbb {Z}_p[\![ T]\!] $
-modules. Then,
$ M $
and
$ M' $
are pseudo-isomorphic if there is a
$\mathbb {Z}_p[\![ T ]\!] $
-module map
$\phi : M \rightarrow M'$
whose kernel and cokernel are finite.
According to the structure theory of
$\mathbb {Z}_p[\![ T]\!] $
-modules, any cofinitely generated and cotorsion
$\mathbb {Z}_p[\![ T]\!] $
-module
$ M $
is pseudo-isomorphic to a module
$ M' $
whose Pontryagin dual is a direct sum of cyclic torsion
$\mathbb {Z}_p[\![ T]\!] $
-modules:
Here,
$ s $
and
$ t $
are natural numbers (possibly
$ 0 $
),
$ n_i \in \mathbb {Z}_{\geq 1} $
, and
$ f_j(T) $
are distinguished polynomials. The Iwasawa
$\mu $
- and
$\lambda $
-invariants are then defined as follows:
If
$ s = 0 $
(respectively
$ t = 0 $
), the sum
$\sum _{i=1}^s n_i$
(respectively
$\sum _{j=1}^t \deg f_j$
) is taken to be
$ 0 $
.
Lemma 2.2. Let M be a cofinitely generated and cotorsion
$\mathbb {Z}_p[\![ T]\!] $
-module. Then, the following conditions are equivalent.
-
(1) The module M is a cotorsion $\mathbb {Z}_p[\![ T]\!] $
-module and
$\mu _p(M)=0$
. -
(2) The module M is cofinitely generated as a $\mathbb {Z}_p$
-module.
Moreover, if the above (equivalent) conditions are satisfied, then,
where A is a finite
$\mathbb {Z}_p$
-module.
Proof. The result is an easy consequence of the structure isomorphism (2.1), so the details are omitted.
It then follows from results of Kato [Reference KatoKat04] and Rubin [Reference RubinRub81] that
$\mathrm {Sel}_{p^\infty }(E/K_{\mathrm {cyc}})$
is cofinitely generated and cotorsion over
$\Lambda _{\mathrm {cyc}}$
.
2.2 Hilbert’s tenth problem for number rings
Let
$ A $
be a commutative ring with an identity element and consider
$ A^n $
, which is the free
$ A $
-module of rank
$ n $
. This module consists of tuples
$ a = (a_1, \ldots , a_n) $
, where each
$ a_i $
is an element of
$ A $
. Suppose
$ m $
and
$ n $
are positive integers. For
$ a = (a_1, \ldots , a_n) \in A^n $
and
$ b = (b_1, \ldots , b_m) \in A^m $
, we denote by
$ (a,b) \in A^{n+m} $
the concatenated tuple
$ (a_1, \ldots , a_n, b_1, \ldots , b_m) $
. Given a finite collection of polynomials
$ F_1, \ldots , F_k $
, we define the set
Definition 2.3. A subset
$ S $
of
$ A^n $
is called a Diophantine subset of
$ A^n $
if there exists an integer
$ m \geq 1 $
and a collection of polynomials
such that
$ S $
consists of all
$ a \in A^n $
for which the set
$ \mathcal {F}(a; F_1, \ldots , F_k) $
is nonempty.
Definition 2.4. An extension of number fields
$ L/K $
is said to be integrally Diophantine if the ring of integers
$ \mathcal {O}_K $
is a Diophantine subset of the ring of integers
$ \mathcal {O}_L $
.
Let
$ L' $
,
$ L $
and
$ K $
be number fields such that
We assume that both
$ L'/L $
and
$ L/K $
are integrally Diophantine extensions. It is a basic fact in the theory of Diophantine sets and fields that, under these conditions, the composite extension
$ L'/K $
is also integrally Diophantine. This assertion follows from [Reference ShlapentokhShl07, Theorem 2.1.15]. In fact, [Reference ShlapentokhShl07] provides a comprehensive treatment of Hilbert’s tenth problem for rings of integers in number fields.
In the context of integrally Diophantine extensions, we recall a conjecture proposed by Denef and Lipshitz [Reference Denef and LipshitzDL78].
Conjecture 2.5 (Denef–Lipshitz).
For any number field
$ L $
, the extension
$ L/\mathbb {Q} $
is integrally Diophantine.
This conjecture posits that for any number field
$ L $
, the ring of integers
$\mathcal {O}_{\mathbb {Q}} = \mathbb {Z}$
is a Diophantine subset of the ring of integers
$\mathcal {O}_L$
. The veracity of this conjecture would imply a significant expansion of the negative resolution of Hilbert’s tenth problem.
The renowned result by Matijasevich, building on work by Davis, Putnam and Robinson, demonstrated that Hilbert’s tenth problem, which asks whether there is an algorithm to determine the solvability of arbitrary Diophantine equations over
$\mathbb {Z}$
, has a negative answer. Specifically, there is no such algorithm. If Conjecture 2.5 holds for a number field
$ L $
, it would imply that Hilbert’s tenth problem also has a negative answer for the ring of integers
$\mathcal {O}_L$
. In other words, there would be no algorithm to determine the solvability of Diophantine equations over
$\mathcal {O}_L$
. In addition to the cases discussed in Section 1, there have been significant advancements in establishing new criteria for the validity of Conjecture 2.5. Notable contributions have been made by Poonen [Reference PoonenPoo02], Cornelissen et al. [Reference Cornelissen, Pheidas and ZahidiCPZ05] and Shlapentokh [Reference ShlapentokhShl08]. These researchers have introduced criteria linked to the stability of the Mordell–Weil rank of elliptic curves over number fields.
One particularly influential result is Shlapentokh’s criterion, which relates Diophantine stability for elliptic curves with positive rank to Hilbert’s tenth problem for rings of integers in number fields.
Theorem 2.6 (Shlapentokh).
Let
$ L/K $
be an extension of number fields and suppose there exists an elliptic curve
$ E $
defined over
$ K $
such that
$\mathrm {rank}\ E(L) = \mathrm {rank}\ E(K)> 0$
. Then,
$ L/K $
is integrally Diophantine. Consequently, if Hilbert’s tenth problem has a negative answer for
$\mathcal {O}_K$
, then Hilbert’s tenth problem also has a negative answer for
$\mathcal {O}_L$
.
This theorem demonstrates that the existence of an elliptic curve
$ E $
over
$ K $
with equal positive rank over both
$ K $
and
$ L $
ensures that the ring of integers
$\mathcal {O}_K$
is a Diophantine subset of
$\mathcal {O}_L$
. This result provides a powerful tool for transferring the negative solution of Hilbert’s tenth problem from
$\mathcal {O}_K$
to
$\mathcal {O}_L$
, thereby expanding the range of number fields for which the problem is known to have no algorithmic solution.
3 Control theory for Selmer groups and the Euler characteristic formula
In this section, we briefly review Mazur’s control theorem as well as the Euler characteristic formula for the Selmer group, due to Perrin-Riou and Schneider. Throughout, we let p be an odd prime number, E be an elliptic curve over a number field F and let
$F_{\mathrm {cyc}}$
be the cyclotomic
$\mathbb {Z}_p$
-extension of F. Fix an algebraic closure
$\bar {F}$
of F. Set
$F_{\mathrm {cyc}}^{(n)}/F$
to be the subextension for which
$[F_{\mathrm {cyc}}^{(n)}:F]=p^n$
. Let
$\Gamma _{F_{\mathrm {cyc}}}:=\mathrm {Gal}(F_{\mathrm {cyc}}/F)$
and
$\Gamma _{F_{\mathrm {cyc}}, n}:=\mathrm {Gal}(F_{\mathrm {cyc}}/F_n)=\Gamma _{F_{\mathrm {cyc}}}^{p^n}$
. Let S be a set of primes of F containing the primes
$v|p$
and the primes at which E has bad reduction. Denote by
$F_S$
the maximal extension of F that is contained in
$\bar {F}$
, in which the primes
$v\notin S$
are unramified.
3.1 Mazur’s control theorem
There is a natural map
which fits into the following commutative diagram.
![Commutative diagram. Top row: 0 → Sel p^infinity (E/F cyc^n) → H^1 (F S/F cyc^n, E[p^infinity]) → direct sum v in S J v (E/F cyc^n). Bottom row: 0 → Sel p^infinity (E/F cyc)^Gamma F cyc,n → H^1 (F S/F cyc, E[p^infinity])^Gamma F cyc,n → (direct sum).](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20261111231602584-0724:S1446788726101748:S1446788726101748_eqnu22.png?pub-status=live)
Here, the maps
$\beta _n$
and
$\gamma _n$
are induced by the restriction maps in Galois cohomology.
Theorem 3.1 (Mazur).
The above-mentioned maps
$\alpha _n$
have finite kernels and cokernels whose orders are uniformly bounded, independent of n.
We now present a criterion for
$\mathrm {Sel}_{p^\infty }(E/F_{\mathrm {cyc}})$
to be a cofinitely generated and cotorsion
$\mathbb {Z}_p[\![ T]\!] $
-module.
Theorem 3.2. If
$\mathrm {Sel}_{p^\infty }(E/F)$
is finite, then
$\mathrm {Sel}_{p^\infty }(E/F_{\mathrm {cyc}})$
is cofinitely generated and cotorsion.
Proof. For the proof, see [Reference GreenbergGre99, Theorem 1.4].
Theorem 3.3 (Kato–Rohrlich).
Assume E is defined over
$\mathbb {Q}$
and F is an abelian extension of
$\mathbb {Q}$
. Then,
$\mathrm {Sel}_{p^\infty }(E/F_{\mathrm {cyc}})$
is cofinitely generated and cotorsion.
3.2 Generalities on the Euler characteristic
In this section, let
$\Gamma \simeq \mathbb {Z}_p$
and
${\Lambda :=\mathbb {Z}_p[\![ \Gamma ]\!] }$
. Let
$ M $
be a
$\Lambda $
-module that is cofinitely generated and cotorsion. We focus on two derived modules: the module of invariants
$ H^0(\Gamma , M) = M^\Gamma $
and the module of co-invariants
$ H^1(\Gamma , M) = M_\Gamma = M / TM $
. There is a natural map between these modules:
sending an element
$ x \in M^\Gamma $
to
$ x \ \pmod{TM} $
in
$ M_\Gamma $
. Since
$ M $
is cofinitely generated over
$\Lambda $
, both
$ M^\Gamma $
and
$ M_\Gamma $
are cofinitely generated over
$\mathbb {Z}_p$
. Additionally, since
$\Gamma \cong \mathbb {Z}_p$
has cohomological dimension 1,
$ H^i(\Gamma , \cdot ) = 0 $
for
$ i \geq 2 $
. For such a module
$ M $
, the coranks of
$ M^\Gamma $
and
$ M_\Gamma $
as
$\mathbb {Z}_p$
-modules are equal:
see [Reference Kundu and RayKR21, Lemma 2.1]. A key consequence of this lemma is that
$ M^\Gamma $
is finite if and only if
$ M_\Gamma $
is finite.
Definition 3.4. Let
$ M $
be a
$\Lambda $
-module that is cofinitely generated and cotorsion. The Euler characteristic of
$ M $
is well defined if
$ M^\Gamma $
(and equivalently
$ M_\Gamma $
) is finite. In this case, the Euler characteristic of
$ M $
is defined as
There is indeed a generalization of the above notion when
$M^\Gamma $
is infinite, which we treat in the next subsection. We define the characteristic series of
$ M $
as follows.
Definition 3.5. Consider the numbers
$n_i$
and polynomials
$f_j$
in (2.1). Then, the characteristic series
$ f_M(T) $
is defined by
and can be expressed as
Proposition 3.6. For a
$\Lambda $
-module
$ M $
that is cofinitely generated and cotorsion, the following conditions are equivalent:
-
(1) the Euler characteristic $\chi (\Gamma , M)$
is well defined; -
(2) the constant term $ a_0 $
of the characteristic series
$ f_M(T) $
is nonzero.
Furthermore, if these conditions are satisfied, then
$\chi (\Gamma , M)$
is an integer and it holds that
where
$ a \sim b $
means
$ a = ub $
for some unit
$ u \in \mathbb {Z}_p^\times $
.
Proof. For a proof of the result, see [Reference Ray and ShingavekarRS24, Proposition 4.4].
Lemma 3.7. For a cofinitely generated and cotorsion
$\Lambda $
-module
$ M $
with a well-defined Euler characteristic, the following conditions are equivalent:
-
(1) $\mu _p(M) = 0$
and
$\lambda _p(M) = 0$
; -
(2) $ a_0 $
is not divisible by
$ p $
; -
(3) $\chi (\Gamma , M) = 1$
.
Proof. This result is [Reference Ray and ShingavekarRS24, Lemma 4.5].
3.3 The truncated Euler characteristic
When the cohomology groups
$ H^i(\Gamma , M) $
are not finite, a generalized version called the truncated Euler characteristic, denoted
$ \chi _t(\Gamma , M) $
, is used. Given that
$ H^1(\Gamma , M) $
is isomorphic to the group of coinvariants
$ H_0(\Gamma , M) = M_{\Gamma } $
, there exists a natural map
which sends
$ x \in M^{\Gamma } $
to its residue class in
$ M_{\Gamma } $
. The truncated Euler characteristic
$ \chi _t(\Gamma , M) $
is defined when both the kernel and the cokernel of
$ \Phi _{M} $
are finite. It is given by the formula
It can be shown that if the classical Euler characteristic
$ \chi (\Gamma , M) $
is defined, then
$ \chi _t(\Gamma , M) $
is also defined, and in fact,
Let
$ r_{M} $
denote the order of vanishing of
$ f_M(T) $
at
$ T = 0 $
. For
$ a, b \in \mathbb {Q}_p $
, recall that
$ a \sim b $
if there exists a unit
$ u \in \mathbb {Z}_p^{\times } $
such that
$ a = bu $
.
Lemma 3.8. Let
$ M $
be a cofinitely generated cotorsion
$ \mathbb {Z}_p[\![ T]\!] $
-module. Assume that the kernel and cokernel of
$ \Phi _{M} $
are finite. Then:
-
(1) $ r_{M} = \mathrm {corank}_{\mathbb {Z}_p}(M^{\Gamma }) = \mathrm {corank}_{\mathbb {Z}_p}(M_{\Gamma }) $
; -
(2) $ a_{r_{M}} \neq 0 $
; -
(3) $ a_{r_{M}} \sim \chi _t(\Gamma , M) $
.
Proof. For the proof, see [Reference ZerbesZer09, Lemma 2.11].
In particular, the classical Euler characteristic
$ \chi (\Gamma , M) $
is defined if and only if
${ r_{M} = 0 }$
. Next, we give a criterion for the truncated Euler characteristic to be defined.
Lemma 3.9. Let M be a cofinitely generated and cotorsion
$\mathbb {Z}_p[\![ T]\!] $
-module. Let
${f_1(T), \ldots , f_t(T)}$
be distinguished polynomials in (2.1). Assume that
$T^2\nmid f_j$
for all
${j=1, \ldots , t}$
. Then, the truncated Euler characteristic
$\chi _t(\Gamma , M)$
is well defined. In particular,
$\chi _t(\Gamma , M)$
is defined when
$r_M\leq 1$
.
Proof. The assertion follows from the proof of [Reference ZerbesZer09, Lemma 2.11].
Lemma 3.10. Let M be a cofinitely generated and cotorsion
$\mathbb {Z}_p[\![ T]\!] $
-module for which
$\mathrm {corank}_{\mathbb {Z}_p}(M^\Gamma )= 1$
. Then, one has that
$r_M=1$
. Moreover, the truncated Euler characteristic is defined and
3.4 A formula for the truncated Euler characteristic
Let E be an elliptic curve over a number field F. Assume that E has good ordinary reduction at the primes of F that lie above p. We set
$\Gamma :=\mathrm {Gal}(F_{\mathrm {cyc}}/F)$
. Assume that the Selmer group
$\mathrm {Sel}_{p^\infty }(E/F_{\mathrm {cyc}})$
is cofinitely generated and cotorsion as a
$\Lambda _{\mathrm {cyc}}$
-module. Note that this is indeed true when
$F=\mathbb {Q}$
, or if F is an abelian extension of
$\mathbb {Q}$
and E is defined over
$\mathbb {Q}$
. Assume that the
$\mathbb {Z}_p$
-corank of
$\mathrm {Sel}_{p^\infty }(E/F)$
is
$1$
. Consider the natural map
sending an element
$x\in \mathrm {Sel}_{p^\infty } (E/F_{\mathrm {cyc}})^\Gamma $
to its residue class
$\Phi (x)=\bar {x}\in \mathrm {Sel}_{p^\infty } (E/F_{\mathrm {cyc}})_{\Gamma }$
. It follows from Mazur’s control theorem that
$\mathrm {corank}_{\mathbb {Z}_p}\mathrm {Sel}_{p^\infty }(E/F_{\mathrm {cyc}})^{\Gamma _{\mathrm {cyc}}}=1$
. The truncated Euler characteristic
$\chi _t(\Gamma , E[p^\infty ])$
is defined as the following quotient:
Note that by Lemma 3.10,
$\mathrm {ker}\Phi $
and
$\mathrm {cok}\Phi $
are finite, and thus, the truncated Euler characteristic is well defined.
Let
$f(T)$
represent the characteristic element of
$\mathrm {Sel}_{p^\infty }(E/F_{\mathrm {cyc}})$
. Let
$r\in \mathbb {Z}_{\geq 0}$
be the order of vanishing of
$f(T)$
at
$T=0$
and express
$f(T)$
as
$T^r \cdot g(T)$
. According to Lemma 3.10, r is equal to 1. Note that
$g(0) \neq 0$
. The p-adic regulator of E over F, denoted as
$\mathrm {Reg}_p(E/F)$
, is defined as the determinant of the p-adic height pairing (see [Reference SchneiderSch82]). It is conjectured that the p-adic regulator is nonzero (see [Reference SchneiderSch85]). If
$r_E:=\mathrm {rank}\ E(F)$
, let
$\mathcal {R}_p(E/F):=p^{-r_E}\mathrm {Reg}_p(E/F)$
denote the normalized regulator. Thus,
$\mathcal {R}_p(E/F)=\mathrm {Reg}_p(E/F)/p$
in the rank-one situation considered above, while in rank zero, we use the usual empty-determinant convention
$\mathrm {Reg}_p(E/F)=\mathcal {R}_p(E/F)=1$
. Given a prime v of K, denote by
$c_v(E/F)$
the Tamagawa number of E over
$F_v$
. Set
$\kappa _v$
to denote the residue field at v. Suppose E has good reduction at v, set
$\widetilde {E}(\kappa _v)$
to be the group of
$\kappa _v$
-rational points of the reduction of a regular model of E at the prime v.
The following result gives the formula for the truncated Euler characteristic of the p-primary Selmer group when it is defined. In the CM case, this was proven by B. Perrin-Riou (see [Reference Perrin-RiouPR83]) and in the general case by P. Schneider (see [Reference SchneiderSch85]).
Theorem 3.11 (Perrin-Riou and Schneider).
Let
$E/F$
be an elliptic curve over a number field F and let p be an odd prime. Assume that the following conditions are satisfied:
-
(1) $\mathrm {Sel}_{p^\infty }(E/F_{\mathrm {cyc}})$
is cofinitely generated and cotorsion over
$\Lambda _{\mathrm {cyc}}$
, and
$\mathrm {rank}\ E(F)\in \{0,1\}$
; -
(2) $E(F)[p]=0$
; -
(3) the p-primary part of the Tate–Shafarevich group $\mathrm{III}(E/F)[p^\infty ]$
is finite; -
(4) the p-adic height pairing on $E(F)$
is nondegenerate.
Then,
Here,
$\sim $
denotes equality up to multiplication by an element of
$\mathbb {Z}_p^\times $
. When
$\mathrm {rank}\ E(F)=0$
, the truncated Euler characteristic in this formula is the classical Euler characteristic. When
$\mathrm {rank}\ E(F)=1$
, it is the rank-one truncated Euler characteristic defined above.
The following result gives a relationship between the Iwasawa invariants and the (truncated) Euler characteristic.
Corollary 3.12. Let
$E/F$
be an elliptic curve satisfying the conditions of the above theorem. Assume in addition that
$\mathrm {corank}_{\mathbb {Z}_p}\mathrm {Sel}_{p^\infty }(E/F)=1$
. Then, with respect to the notation above, the following conditions are equivalent.
-
(1) The Iwasawa invariants of $\mathrm {Sel}_{p^\infty }(E/F_{\mathrm {cyc}})$
are given by $$ \begin{align*}\mu(\mathrm{Sel}_{p^\infty}(E/F_{\mathrm{cyc}}))=0\quad\text{and}\quad\lambda(\mathrm{Sel}_{p^\infty}(E/F_{\mathrm{cyc}}))=1.\end{align*} $$
-
(2) We have $\chi _t(\Gamma , E[p^\infty ])=1$
. -
(3) The following are all satisfied:
-
• $p\nmid \mathcal {R}_p(E/F)$
; -
• $\#\, \mathrm{III}(E/F)[p^\infty ]=1$
; -
• $p\nmid c_v(E/F)$
for all primes
$v\nmid p$
of F; and -
• $\widetilde {E}(\kappa _v)[p^\infty ]=0$
for all primes
$v|p$
of F.
-
Proof. Recall that
$f(T)=T g(T)$
. Thus,
$\mu =0$
and
$\lambda =1$
if and only if
$g(T)$
is a unit in
$\mathbb {Z}_p[\![ T]\!] $
. However,
$\chi _t(\Gamma , E[p^\infty ])\sim g(0)$
(see Lemma 3.8(3)). Thus,
$\mu =0$
and
$\lambda =1$
if and only if
$\chi _t(\Gamma , E[p^\infty ])=1$
. This proves that Conditions (1) and (2) are equivalent. The equivalence of Conditions (2) and (3) follows from Theorem 3.11.
Corollary 3.13 (Rank-zero criterion).
Let
$E/F$
satisfy the hypotheses of Theorem 3.11 and assume that
$\mathrm {corank}_{\mathbb {Z}_p}\mathrm {Sel}_{p^\infty }(E/F)=0$
. Then, the following conditions are equivalent:
-
(1) $\mu(\mathrm{Sel}_{p^\infty}(E/F_{\mathrm{cyc}}))=0 \quad\text{and}\quad \lambda(\mathrm{Sel}_{p^\infty}(E/F_{\mathrm{cyc}}))=0$
; -
(2) $\chi(\Gamma,\mathrm{Sel}_{p^\infty}(E/F_{\mathrm{cyc}}))=1$
; -
(3) $\mathrm{III}(E/F)[p^\infty ]=0$
,
$p\nmid c_v(E/F)$
for every
$v\nmid p$
, and
$\widetilde E(\kappa _v)[p^\infty ]=0$
for every
$v\mid p$
.
Proof. Since
$\mathrm{III}(E/F)[p^\infty ]$
is finite, the Selmer exact sequence shows that
$E(F)$
has rank zero. Mazur’s control theorem therefore makes the
$\Gamma $
-invariants of
$\mathrm {Sel}_{p^\infty }(E/F_{\mathrm {cyc}})$
finite, so its classical Euler characteristic is defined. The equivalence of Conditions (1) and (2) is Lemma 3.7. Theorem 3.11, together with
$\mathcal {R}_p(E/F)=1$
in rank zero, gives the equivalence of Conditions (2) and (3).
4 Hilbert’s tenth problem in towers of number fields
In this section, we apply ideas from Iwasawa theory to prove Hilbert’s tenth problem in the finite layers of various
$\mathbb {Z}_p$
-extensions of interest. Our results require that there exist elliptic curves for which certain additional conditions are satisfied. We present examples to illustrate our results. The main results of this section are Theorems 4.12, 4.13 and 4.14. In the next section, we verify these conditions unconditionally for a positive-density family when
$p=3$
. For the families obtained from Kriz–Li, the required vanishing of the p-primary Tate–Shafarevich group of the rank-zero twists is retained as an explicit density hypothesis.
As in the previous section, fix an elliptic curve
$E_{/\mathbb {Q}}$
and an odd prime number p. We let
$K=K_d=\mathbb {Q}(\!\sqrt {d})$
be an imaginary quadratic field. Here, d is a negative squarefree integer. Denote the twist of E over K by
$E^{(K)}$
or, equivalently, by
$E^{(d)}$
. Recall that Proposition 2.1 asserts that a p-primary Selmer group over
$K_{\mathrm {cyc}}$
decomposes as
Assumption 4.1. We assume throughout that:
-
(1) E and $E^{(K)}$
have good ordinary reduction at p; -
(2) $\mathrm {corank}_{\mathbb {Z}_p} \mathrm {Sel}_{p^\infty }(E/\mathbb {Q})=0$
; -
(3) $E(K)[p]=0$
; -
(4) $\mathrm{III}(E/K)[p^\infty ]$
is finite.
It follows from the structure theory of
$\Lambda _{\mathrm {cyc}}$
-modules that
$\mu _p(E/K_{\mathrm {cyc}})=0$
and
$\lambda _p(E/K_{\mathrm {cyc}})=1$
if and only if as a
$\mathbb {Z}_p$
-module,
for some finite abelian group A. Let
$(a,b)\in \mathbb {P}^1(\mathbb {Z}_p)$
and
$K_{a,b}$
be the associated
$\mathbb {Z}_p$
-extension of K. Let
$K_{a,b}^{(n)}$
be the nth layer of
$K_{a,b}/K$
, that is,
$K_{a,b}^{(n)}$
is the subextension for which
$[K_{a,b}^{(n)}:K]=p^n$
. Identify the Iwasawa algebra
$\Lambda _{a,b}$
with a formal power series ring
$\mathbb {Z}_p[\![ T]\!] $
, with respect to some choice of topological generator of
$\Gamma _{a,b}$
. Denote by
$f_{a,b}(T)$
the characteristic element associated with
$\mathrm {Sel}_{p^\infty }(E/K_{a,b})$
, with the understanding that
$f_{a,b}(T)$
is
$0$
if
$\mathrm {Sel}_{p^\infty }(E/K_{a,b})$
is not cotorsion over
$\Lambda _{a,b}$
. Set
$\mu _p(E/K_{a,b})$
(respectively
$\lambda _p(E/K_{a,b})$
), the
$\mu $
-invariant (respectively
$\lambda $
-invariant) associated to
$\mathrm {Sel}_{p^\infty }(E/K_{a,b})$
.
Lemma 4.2. Consider a finite-dimensional
$\mathbb {F}_p$
-vector space V and the finite p-group G acting on V. If
$V^G=0$
, then it follows that
$V=0$
.
Proof. The result follows from Nakayama’s lemma, but we give a more elementary proof. Consider the relation
$\#\, V = \sum _i \#\, \mathcal {O}_i$
, where
$\mathcal {O}_i$
runs through the G-orbits of V. Since it is assumed that
$V^G = 0$
, it follows that
$\{0\}$
is a G-orbit consisting of just one element, and all other G-orbits have sizes that are powers of p. This implies in particular that
$\#\, V\equiv 1\ \pmod{\ p}$
. Therefore,
$\#\,V=1$
, that is,
$V=0$
.
Lemma 4.3. Assume that
$E(K)[p^\infty ]=0$
. Then, it follows that
$E(K_{a,b})[p^\infty ]=0$
.
Proof. It suffices to show that for all n,
$E(K_{a,b}^{(n)})[p^\infty ]=0$
. Let
$\Gamma _n:=\mathrm {Gal}(K_{a,b}^{(n)}/K)$
. We set
$V:=E(K_{a,b}^{(n)})[p]$
and note that
Lemma 4.2 then implies that
$V=0$
. Since
$E(K_{a,b}^{(n)})[p]=0$
, it is clear that
${E(K_{a,b}^{(n)})[p^\infty ]=0}$
. This concludes the proof.
Proposition 4.4. Let
$K_{a, b}$
be a
$\mathbb {Z}_p$
-extension of K and assume that:
-
(1) $\mathrm {rank}\ E(K)=1$
; -
(2) $\mathrm {Sel}_{p^\infty }(E/K_{a,b})$
is cotorsion as a
$\Lambda _{a,b}$
-module; -
(3) $\mu _p(E/K_{a,b})=0$
and
$\lambda _p(E/K_{a,b})=1$
.
For all
$n\in \mathbb {Z}_{\geq 0}$
, the extension
$K_{a,b}^{(n)}/\mathbb {Q}$
is integrally Diophantine. Thus, Conjecture 2.5 is satisfied for
$K_{a,b}^{(n)}$
. In particular, Hilbert’s tenth problem has a negative answer for
$K_{a,b}^{(n)}$
.
Proof. It follows from Lemma 2.2 that
where A is a finite p-group. It is easy to see (from the inflation-restriction sequence) that the kernel of the natural map
injects into
$H^1(K_{a,b}/K_{a,b}^{(n)},E(K_{a,b})[p^\infty ])$
. It follows from Lemma 4.3 that
${E(K_{a,b})[p^\infty ]{\kern-1.5pt}={\kern-1.5pt}0}$
. Thus,
$\alpha _n$
is injective. Therefore, it follows that
In particular, this implies that
However, since
$\mathrm {rank}\ E(K)=1$
, it follows that for all n,
Thus, by Theorem 2.6, we have that
$K_{a,b}^{(n)}/K$
is integrally Diophantine. Since
$K/\mathbb {Q}$
is abelian, it is an integrally Diophantine extension. Thus, it follows that
$K_{a, b}^{(n)}/\mathbb {Q}$
is integrally Diophantine; thus, verifying Conjecture 2.5. In particular, for all n, Hilbert’s tenth problem has a negative answer for
$K_{a,b}^{(n)}$
.
We give conditions for the assumptions of Proposition 4.4 to hold. Let
${M {\kern-1pt}:={\kern-1pt} \mathrm {Sel}_{p^\infty }(E/K_\infty )^\vee }$
, viewed as a module over
$\Lambda _\infty {\kern-1pt}={\kern-1pt} \mathbb {Z}_p[\![ X, Y]\!] $
. Following [Reference Gajek-Leonard, Hatley, Kundu and LeiGLHKL25], we introduce a condition on the algebraic structure of M.
Definition 4.5. The module M is said to satisfy Criterion (S-C) if it is a direct sum of cyclic torsion
$\Lambda _\infty $
-modules.
The module M over
$\mathbb {Z}_p[\![ X, Y]\!] $
is said to be pseudonull if for all prime ideals
$\mathfrak {p}$
of
$\mathbb {Z}_p[\![ X, Y]\!] $
of height
$\leq 1$
, one has that
$M_{\mathfrak {p}}=0$
. The maximal pseudonull submodule of M is denoted by
$M_o$
.
Proposition 4.6. With respect to the notation above, we have that
$M_o=0$
.
Proof. The result follows from [Reference Kundu, Lei and RayKLR22, Proposition 6.1].
Lemma 4.7. Suppose that M satisfies Criterion (
S-C
). Then, there exist principal ideals
$I_1,\ldots , I_m$
of
$\Lambda _\infty $
such that
$M\simeq \bigoplus _{i=1}^m\Lambda _\infty /I_i$
.
Proof. The result follows directly from Proposition 4.6 and [Reference Gajek-Leonard, Hatley, Kundu and LeiGLHKL25, Theorem 2.4].
Definition 4.8. Suppose that M satisfies Criterion ( S-C ). The characteristic ideal of M is defined to be
where
$I_i$
are the principal ideals prescribed in Lemma 4.7.
We recall an explicit condition for Criterion ( S-C ) to be satisfied.
Proposition 4.9. With respect to the notation above, suppose that
${\lambda (\mathrm {Sel}_{p^\infty }(E/K_{\mathrm {cyc}})^\vee )\le 1}$
and
$\mu (\mathrm {Sel}_{p^\infty }(E/K_{\mathrm {cyc}})^\vee )=0$
, then
$\mathrm {Sel}_{p^\infty }(E/K_\infty )^\vee $
satisfies Criterion (
S-C
).
Proof. The result is [Reference Gajek-Leonard, Hatley, Kundu and LeiGLHKL25, Proposition 5.1].
Definition 4.10. Criterion (N-a) is satisfied if for all primes
$v|p$
of K, we have that
$p\nmid \#\,\widetilde {E}(\kappa _v)$
.
Proposition 4.11. Suppose that E satisfies Criterion ( N-a ). Then, the natural restriction map induces an isomorphism
Proof. The result is [Reference Gajek-Leonard, Hatley, Kundu and LeiGLHKL25, Proposition 3.2].
Theorem 4.12. Let
$(a,b)$
and
$(1, 0)$
be p-congruent, that is,
$a\equiv 1\ (\text {mod}\ {p})$
and
${b\equiv 0\ (\text {mod}\ {p})}$
. Suppose that the following conditions hold:
-
(1) Criterion ( N-a ) is satisfied;
-
(2) $\mathrm {Sel}_{p^\infty }(E/K_{\mathrm {cyc}})$
is a cofinitely generated and cotorsion
$\Lambda _{\mathrm {cyc}}$
-module; -
(3) $\mu _p(E/K_{\mathrm {cyc}})=0$
and
$\lambda _p(E/K_{\mathrm {cyc}})=1$
.
Then, the following assertions hold:
-
(1) $\mathrm {Sel}_{p^\infty }(E/K_{a,b})$
is a cofinitely generated and cotorsion
$\Lambda _{a,b}$
-module; -
(2) $\mu _p(E/K_{a,b})=0$
and
$\lambda _p(E/K_{a,b})=1$
; -
(3) for all $n\in \mathbb {Z}_{\geq 0}$
, Conjecture 2.5 is satisfied for
$K_{a,b}^{(n)}$
; in particular, Hilbert’s tenth problem has a negative answer for
$K_{a,b}^{(n)}$
.
Proof. Note that it follows from Proposition 4.9 that Criterion (
S-C
) is satisfied. Lemma 4.7 then implies that
$\mathrm {Sel}_{p^\infty }(E/K_\infty )^\vee $
decomposes as a direct sum of cyclic modules as follows:
for
$g_1, \ldots , g_m \in \Lambda _\infty $
. We note that
$\Lambda _\infty =\mathbb {Z}_p[\![ X, Y]\!] $
and
$\Lambda _{a,b}=\mathbb {Z}_p[\![ X, Y]\!] /(\omega _{a,b})$
, where
$\omega _{a,b}=(1+X)^a(1+Y)^b-1$
. By Proposition 4.11,
We identify
$\Lambda _{\mathrm {cyc}}$
with
$\mathbb {Z}_p[\![ X, Y]\!] /(X)=\mathbb {Z}_p[\![ Y]\!] $
. Observe that
and that there is a uniquely determined power series
$g(Y)\in \mathbb {Z}_p[\![ Y]\!] $
such that
The function
is constructed by inductively solving for
$c_i$
, noting that
$c_1=-b/a$
. There is a natural identification
where the isomorphism takes
$X\mapsto g(Y)$
and
$Y\mapsto Y$
. Let
$F(X,Y)=\prod _i g_i$
be a generator of the characteristic ideal of
$\mathrm {Sel}_{p^\infty }(E/K_\infty )$
. Let
$h(Y)$
and
$h_{a, b}(Y)$
be generators of the characteristic ideals of
$\mathrm {Sel}_{p^\infty }(E/K_{\mathrm {cyc}})$
and
$\mathrm {Sel}_{p^\infty }(E/K_{a,b})$
, respectively. Note that
$\pi _{1,0}(X)=0$
and
$\pi _{1,0}(Y)=Y$
. Thus, we find that
Write
$h(Y)=a_0+a_1 Y+a_2Y^2+\cdots $
and
$h_{a,b}(Y)=b_0+b_1 Y+\cdots $
. By construction,
$g(Y)=-a^{-1}b Y+c_2 Y^2+c_3Y^3+\cdots $
and note that
$p|a^{-1}b$
; we find that
$a_0=b_0$
and
$a_1\equiv b_1\ (\text {mod}\ {p})$
. Since
$\mu _p(E/K_{\mathrm {cyc}})=0$
and
$\lambda _p(E/K_{\mathrm {cyc}})=1$
, we have
$p\mid a_0$
and
$p\nmid a_1$
. Since
$a_0=b_0$
and
$a_1\equiv b_1\ (\text {mod}\ p)$
, it follows that
$p\mid b_0$
and
$p\nmid b_1$
. In particular,
$h_{a,b}(T)\neq 0$
, and hence,
$\mathrm {Sel}_{p^\infty }(E/K_{a,b})$
is cotorsion. Moreover, the conditions
$p\mid b_0$
and
$p\nmid b_1$
show that
$\mu _p(E/K_{a,b})=0$
and
$\lambda _p(E/K_{a,b})=1$
. This proves part (2). Part (3) then follows from Proposition 4.4.
Theorem 4.13. Assume that we have the following conditions:
-
(1) Criterion ( N-a ) is satisfied;
-
(2) $\mathrm {Sel}_{p^\infty }(E/K_{\mathrm {cyc}})$
is a cofinitely generated and cotorsion
$\Lambda _{\mathrm {cyc}}$
-module; -
(3) $\mu _p(E/K_{\mathrm {cyc}})=0$
and
$\lambda _p(E/K_{\mathrm {cyc}})=1$
.
Then, there exists exactly one line
$(\bar {a}_0: \bar {b}_0)\in \mathbb {P}^1(\mathbb {F}_p)$
such that for all
$(a,b)\in \mathbb {P}^1(\mathbb {Z}_p)$
with
$(\bar {a}, \bar {b})\neq (\bar {a}_0: \bar {b}_0)$
:
-
(1) $\mathrm {Sel}_{p^\infty }(E/K_{a,b})$
is a cofinitely generated and cotorsion
$\Lambda _{a,b}$
-module; -
(2) $\mu _p(E/K_{a,b})=0$
and
$\lambda _p(E/K_{a,b})=1$
; -
(3) for all $n\in \mathbb {Z}_{\geq 0}$
, Conjecture 2.5 is satisfied for
$K_{a,b}^{(n)}$
; in particular, Hilbert’s tenth problem has a negative answer for
$K_{a,b}^{(n)}$
.
Proof. Let
$g_i$
be as in the proof of Theorem 4.12 and write
Consider as before the power series
$\omega _{a,b}$
(but this time, we do not assume that
${a\equiv 1\ (\text {mod}\ {p})}$
,
$b\equiv 0\ (\text {mod}\ {p})$
).
We first consider the case where a is a p-adic unit. As the cyclotomic line is already covered in Theorem 4.12, we can assume that b is coprime to p as well. As in the proof of Theorem 4.12, there is a uniquely determined power series
$g(Y)\in \mathbb {Z}_p[\![ Y]\!] $
such that
$\omega _{a,b}(g(Y), Y)=0$
and we find that
$h_{a,b}(Y)=F(g(Y),Y)$
. By construction,
$g(Y)$
is divisible by Y and the linear coefficient is
$-b/a$
, which is a p-adic unit. We obtain
As
$\mu _p(E/K_{\mathrm {cyc}})=0$
and
$\lambda _p(E/K_{\mathrm {cyc}})=1$
, we know that
$a_{0,0}$
is divisible by p and that
$a_{0,1}$
is a p-adic unit. If
$a_{1,0}$
is coprime to p, there exists exactly one value for
$b\ (\text {mod}\ {p})$
such that
$(-a_{1,0}b/a+a_{0,1})$
is divisible by p, that is, such that
$\mu _p(E/K_{a,b})\neq 0$
or
${\lambda _p(E/K_{a,b})\neq 1}$
. If p divides
$a_{1,0}$
, then the coefficient
$(-a_{1,0}b/a+a_{0,1})$
is always coprime to p, and we obtain
$\lambda _p(E/K_{a,b})=1$
and
$\mu _p(E/K_{a,b})=0$
for all values of
$(a,b)$
.
It remains to consider the case
$a=0$
and
$b=1$
, that is,
$\omega _{a,b}=Y$
. In this case, we obtain the anticyclotomic
$\mathbb {Z}_p$
-extension and the characteristic power series associated to the Selmer group is
Recall that p divides
$a_{0,0}$
. We find that
$p\nmid a_{1,0}$
if and only if
$\mu _p(E/K_{a,b})=0$
and
$\lambda _p(E/K_{a,b})=1$
.
Therefore, we find that there is exactly one line
$(\bar {a}_0, \bar {b}_0)\in \mathbb {P}^1(\mathbb {F}_p)$
such that
$f_{a,b}\neq 0$
,
$\mu _p(E/K_{a,b})=0$
and
$\lambda _p(E/K_{a,b})=1$
for all
$(a,b)\not \equiv (\bar {a}_0, \bar {b}_0)\ (\text {mod}\ {p})$
. This proves Parts (1) and (2). Part (3) then follows from the proof of Proposition 4.4.
We now give conditions for the conditions of Theorem 4.12 (or equivalently, Theorem 4.13) to be satisfied. Note that if E is defined over
$\mathbb {Q}$
, it follows from the results of Kato and Rohrlich that Part (2) is satisfied. Recall that Proposition 2.1 asserts that
Note that
$\mu _p(E/K_{\mathrm {cyc}})=0$
if and only if
$\mathrm {Sel}(E/K_{\mathrm {cyc}})$
is cofinitely generated as a
$\mathbb {Z}_p$
-module. Likewise,
$\mu _p(E/\mathbb {Q}_{\mathrm {cyc}})=0$
(respectively
$\mu _p(E^{(K)}/\mathbb {Q}_{\mathrm {cyc}})=0$
) if and only if
$\mathrm {Sel}_{p^\infty }(E/\mathbb {Q}_{\mathrm {cyc}})$
(respectively
$\mathrm {Sel}_{p^\infty }(E^{(K)}/\mathbb {Q}_{\mathrm {cyc}})$
) is cofinitely generated as a
$\mathbb {Z}_p$
-module. Thus, we find that
If the above equivalent conditions are satisfied, then
In order to establish Condition (3), we require that
and that the Iwasawa invariants are prescribed as follows:
Theorem 4.14. Let
$E_{/\mathbb {Q}}$
be an elliptic curve, p an odd prime and K be an imaginary quadratic field. Assume that the following conditions hold:
-
(1) $\mathrm {corank}_{\mathbb {Z}_p} \mathrm {Sel}_{p^\infty }(E/\mathbb {Q})=1$
and
$\mathrm {corank}_{\mathbb {Z}_p} \mathrm {Sel}_{p^\infty }(E^{(K)}/\mathbb {Q})=0$
; -
(2) E and $E^{(K)}$
have good ordinary reduction at p; -
(3) $\widetilde {E}(\mathbb {F}_p)[p]=0$
and
$\widetilde {E}^{(K)}(\mathbb {F}_p)[p]=0$
; -
(4) $p\nmid \prod _{\ell \neq p} c_\ell (E/\mathbb {Q})\times \prod _{\ell \neq p} c_\ell (E^{(K)}/\mathbb {Q})$
; -
(5) the normalized p-adic regulator of E over $\mathbb {Q}$
is a p-adic unit; -
(6) $\mathrm{III}(E/\mathbb {Q})[p^\infty ]=0$
and
$\mathrm{III}(E^{(K)}/\mathbb {Q})[p^\infty ]=0$
.
Let
$(a_0,b_0)$
be a representative for the line excluded in Theorem 4.13. Then, for all
$(a,b)\not \equiv (a_0, b_0)\ (\text {mod}\ {p})$
, the following assertions hold:
-
(1) $\mathrm {Sel}_{p^\infty }(E/K_{a,b})$
is a cofinitely generated and cotorsion
$\Lambda _{a,b}$
-module; -
(2) $\mu _p(E/K_{a,b})=0$
and
$\lambda _p(E/K_{a,b})=1$
; -
(3) for all $n\in \mathbb {Z}_{\geq 0}$
, Conjecture 2.5 is satisfied for
$K_{a,b}^{(n)}$
; in particular, Hilbert’s tenth problem has a negative answer for
$K_{a,b}^{(n)}$
.
Proof. Note that the Selmer groups
$\mathrm {Sel}_{p^\infty }(E/\mathbb {Q}_{\mathrm {cyc}})$
and
$\mathrm {Sel}_{p^\infty }(E^{(K)}/\mathbb {Q}_{\mathrm {cyc}})$
are cofinitely generated and cotorsion over
$\Lambda _{\mathrm {cyc}}$
. The kernel of the reduction map
consists of the
$\mathbb {Z}_p$
-points on the formal group of E at p and is, in particular, torsion free. Condition (3) thus implies in particular that
$E(\mathbb {Q}_p)[p]=0$
and
$E^{(K)}(\mathbb {Q}_p)[p]=0$
. Thus, we find that
$E(\mathbb {Q})[p]=0$
and
$E^{(K)}(\mathbb {Q})[p]=0$
.
We apply the two Euler-characteristic criteria separately. For
$E/\mathbb {Q}$
, Condition (1) is the rank-one Selmer-corank hypothesis, while Conditions (2)–(6) give all the remaining hypotheses of Corollary 3.12. Hence,
For
$E^{(K)}/\mathbb {Q}$
, Condition (1) gives Selmer corank zero. Its p-adic height pairing is therefore vacuously nondegenerate, and Conditions (2), (3), (4) and (6) give the other hypotheses of the rank-zero criterion, Corollary 3.13. Consequently,
From the previous discussion, it follows that
The assertions thus follow from Theorem 4.13.
We conclude this section by illustrating Theorems 4.12, 4.13 and 4.14.
Example 4.15. Consider the elliptic curve
$E/\mathbb {Q}$
with LMFDB label
$58.a1$
and set
$p=17$
. The curve E has rank
$1$
and analytic rank
$1$
. Moreover, the following properties hold:
-
• the Tamagawa product of E is equal to $2$
; -
• E has good ordinary reduction at $17$
; -
• $a_{17}(E)=-4$
and, hence,
$\widetilde {E}(\mathbb {F}_{17})[17]=0$
; -
• the normalized $17$
-adic regulator of E is a
$17$
-adic unit; -
• $E(\mathbb {Q})[17]=0$
; -
• $\mu _{17}(E/\mathbb {Q}_{\mathrm {cyc}})=0$
and
$\lambda _{17}(E/\mathbb {Q}_{\mathrm {cyc}})=1$
.
Since E has analytic rank
$1$
, the Gross–Zagier–Kolyvagin theorem implies that
$\mathrm{III}(E/\mathbb {Q})$
is finite. The
$17$
-adic regulator is nonzero, so the
$17$
-adic height pairing is nondegenerate. We may therefore apply Corollary 3.12. Since the Tamagawa product is prime to
$17$
,
$\widetilde {E}(\mathbb {F}_{17})[17]=0$
, and the Iwasawa invariants are
$\mu _{17}=0$
and
$\lambda _{17}=1$
, it follows that
Now, set
$K=\mathbb {Q}(\!\sqrt {-1})$
and let
$E'=E^{(K)}=E^{(-1)}$
. Then,
$E'$
has LMFDB label
$464.f1$
. Since
$-1$
is a square in
$\mathbb {Q}_{17}$
, the curves E and
$E'$
are isomorphic over
$\mathbb {Q}_{17}$
. In particular,
$E'$
has good ordinary reduction at
$17$
and
The curve
$E'$
has rank
$0$
and analytic rank
$0$
, its Tamagawa product is equal to
$2$
, and
The analytic rank-zero condition and the Gross–Zagier–Kolyvagin theorem imply that
$\mathrm{III}(E'/\mathbb {Q})$
is finite. Since
$\mu _{17}(E'/\mathbb {Q}_{\mathrm {cyc}}) =\lambda _{17}(E'/\mathbb {Q}_{\mathrm {cyc}})=0$
, the classical Euler characteristic is a
$17$
-adic unit. The rank-zero Euler characteristic criterion, Corollary 3.13, together with the fact that the Tamagawa product is prime to
$17$
and
${\widetilde {E}'(\mathbb {F}_{17})[17]=0}$
, therefore give
Thus,
and all the hypotheses of Theorem 4.14 are satisfied. Consequently, apart from the single exceptional residue class of lines prescribed by Theorem 4.13, the finite layers of the corresponding
$\mathbb {Z}_{17}$
-extensions of
$K=\mathbb {Q}(\!\sqrt {-1})$
give negative answers to Hilbert’s tenth problem.
Example 4.16. Consider the curve with label
$61.a1$
and
$p=11$
. Then:
-
• the Tamagawa product is equal to $1$
; -
• $E/\mathbb {Q}$
is of rank
$1$
; -
• the analytic rank is $1$
; -
• the p-adic regulator is coprime to p;
-
• $a_p=-5$
is not congruent to
$1$
modulo p.
If we twist this curve by
$-3$
, we obtain the curve with label 549.c1. Setting
$E':=E^{(-3)}$
, we find that the pair
$(E',p)$
has the properties:
-
• the Tamagawa product is equal to $2$
; -
• $E'/\mathbb {Q}$
is of rank
$0$
; -
• the analytic rank is $0$
; -
• the p-adic regulator is coprime to p;
-
• $a_p=5$
is not congruent to
$1$
modulo p; -
• $\mu _{11}(E'/\mathbb {Q}_{\mathrm {cyc}})=\lambda _{11}(E'/\mathbb {Q}_{\mathrm {cyc}})=0$
.
Moreover,
Both curves have analytic rank at most one, so the Gross–Zagier–Kolyvagin theorem gives the finiteness of their Tate–Shafarevich groups. Corollary 3.12, applied to E, and Corollary 3.13, applied to
$E'$
, now give
Thus, the strengthened hypotheses of Theorem 4.14 hold. Theorems 4.12, 4.13 and 4.14 thus apply in this case for
$p=11$
and
$K:=\mathbb {Q}(\!\sqrt {-3})$
. This gives rise to many
$\mathbb {Z}_{11}$
-extensions of K for which Hilbert’s tenth problem has a negative answer.
5 Density results
In this section, we prove the unconditional
$p=3$
part of Theorem B and the conditional Kriz–Li parts for
$p=11,13,31,37$
. The idea is to fix
$(E,p)$
and vary d over negative squarefree integers. Recall that
$E^{(d)}:=E^{(K_d)}$
, where
$K_d:=\mathbb {Q}(\!\sqrt {d})$
. Thus, we assume that the following conditions are satisfied for E:
-
(1) E has good ordinary reduction at p;
-
(2) $\mathrm {corank}_{\mathbb {Z}_p}\mathrm {Sel}_{p^\infty }(E/\mathbb {Q})=1$
; -
(3) the prime p is nonanomalous for E;
-
(4) $p\nmid \prod _{\ell \neq p} c_\ell (E/\mathbb {Q})$
; -
(5) the normalized p-adic regulator of E over $\mathbb {Q}$
is a p-adic unit; -
(6) $\mathrm{III}(E/\mathbb {Q})[p^\infty ]=0$
.
These conditions imply that
$\mathrm {Sel}_{p^\infty }(E/\mathbb {Q}_{\mathrm {cyc}})$
is cotorsion over
$\Lambda _{\mathrm {cyc}}$
and, moreover, the Iwasawa invariants are given by
This follows from the formula for the truncated Euler characteristic; see Corollary 3.12. As d ranges over all negative squarefree integers,
$K_d$
ranges over all imaginary quadratic fields. We wish to show that there is a positive proportion of such values d for which
$(K_d,p)$
satisfies Criterion (
H10-gen
). This follows as a consequence of Theorem 4.14 by showing that there is a positive density set of d for which the following assertions hold for
$E^{(d)}$
:
-
(1) $E^{(d)}$
has good ordinary reduction at p; -
(2) $\mathrm {corank}_{\mathbb {Z}_p}\mathrm {Sel}_{p^\infty }(E^{(d)}/\mathbb {Q})=0$
; -
(3) $\widetilde {E}^{(d)}(\mathbb {F}_p)[p]=0$
; -
(4) $p\nmid \prod _{\ell \neq p} c_\ell (E^{(d)}/\mathbb {Q})$
; -
(5) $\mathrm{III}(E^{(d)}/\mathbb {Q})[p^\infty ]=0$
.
The last condition is not a consequence of analytic rank zero. The Gross–Zagier–Kolyvagin theorem gives only the finiteness of
$\mathrm{III}(E^{(d)}/\mathbb {Q})$
. To record precisely the additional input needed below, let
$\mathcal D$
be a positive-density family of negative squarefree twisting parameters and put
We refer to the assertion
$\eta _{E,p}(\mathcal D)>0$
as the p-primary Tate–Shafarevich density hypothesis for
$\mathcal D$
. The positive-density conclusions obtained below from Kriz and Li are conditional on this hypothesis. The argument for
$p=3$
instead produces twists with trivial
$3$
-Selmer group and, therefore, proves the required Tate–Shafarevich vanishing unconditionally.
5.1 Density results via the method of Kriz and Li
First, we apply the results of Kriz and Li [Reference Kriz and LiKL19] to obtain conditional density theorems for
$p=11,13,31,37$
. Their results supply positive-density families of analytic rank-zero twists, but they do not by themselves imply the vanishing of the p-primary Tate–Shafarevich groups.
Lemma 5.1. Let
$E/\mathbb {Q}$
be an elliptic curve and let
$p\geq 5$
be a prime such that
${E(\mathbb {Q})[2]=\{0\}}$
. Let N be the conductor of E and let
$K_0$
be an imaginary quadratic field. Assume that the following conditions hold:
-
(1) $\mathrm {corank}_{\mathbb {Z}_p}\mathrm {Sel}_{p^\infty }(E/\mathbb {Q})=1$
and
$\mathrm{III}(E/\mathbb {Q})[p^\infty ]=0$
; -
(2) E has good ordinary reduction at p;
-
(3) the prime p is nonanomalous for E;
-
(4) $p\nmid \prod _{\ell \neq p}c_\ell (E/\mathbb {Q})$
; -
(5) the normalized p-adic regulator of E over $\mathbb {Q}$
is a p-adic unit; -
(6) $2$
splits in
$K_0$
; -
(7) all primes dividing $pN$
split in
$K_0$
; -
(8) let P be the Heegner point for $E/K_0$
; let
$E^{\mathrm {ns}}(\mathbb {F}_2)$
be the set of nonsingular points of E over
$\mathbb {F}_2$
and let
$\log _{\omega _E}$
be the logarithm associated to the differential
$\omega _E$
; then $$ \begin{align*} \frac{|E^{\mathrm{ns}}(\mathbb{F}_2)|\log_{\omega_E}(P)}{2} \not\equiv 0\ (\text{mod}\ {2}). \end{align*} $$
Let S be the set of primes
$\ell $
coprime to
$2N$
such that:
-
(1) $\ell $
splits in
$K_0$
; -
(2) $\ell $
is a square modulo every prime dividing
$pN$
; -
(3) $\ell \equiv \begin {cases} 1\ (\text {mod}\ {4})&\text {if }2\nmid N,\\ 1\ (\text {mod}\ {8})&\text {if }2\mid N; \end {cases}$
-
(4) $\mathrm {Frob}_\ell \in \mathrm {Gal}(\mathbb {Q}(E[2])/\mathbb {Q})$
has order
$3$
.
Let
$E'$
be the twist of E by
$dd_{K_0}$
, where
$d>0$
is squarefree and only divisible by primes in S. If, in addition,
then
$(E,E',p)$
satisfies the hypotheses of Theorem 4.14. Consequently, the conclusions of Theorems 4.12 and 4.13 apply. In particular, Criterion (
H10-gen
) is satisfied for
Proof. We need to check the assumptions of Theorem 4.14. We first verify the condition required to apply [Reference Kriz and LiKL19, Theorem 4.3]. Let
$\ell \mid d$
. If q is an odd prime dividing N, Condition (2) in the definition of S gives
$({\ell }/{q})=1$
. Since
${\ell \equiv 1\ (\text {mod}\ {4})}$
, quadratic reciprocity gives
$({q}/{\ell })=1$
. Moreover,
$({-1}/{\ell })=1$
. If
$2\mid N$
, then by Condition (3), we have
$\ell \equiv 1\ (\text {mod}\ {8})$
and, hence,
$({2}/{\ell })=1$
. It follows that
$({-N}/{\ell })=1$
for every prime
$\ell \mid d$
. Since d is squarefree, the Chinese remainder theorem shows that
$-N$
is a square modulo d. We therefore obtain from [Reference Kriz and LiKL19, Theorem 4.3] that
$E'$
has algebraic and analytic rank zero.
The result of Gross, Zagier and Kolyvagin [Reference Gross and ZagierGZ86, Reference KolyvaginKol90] (see also [Reference BurungaleBur22, Theorem 2.4]) implies that
$\mathrm{III}(E'/\mathbb {Q})$
is finite. Consequently,
$\mathrm {corank}_{\mathbb {Z}_p}\mathrm {Sel}_{p^\infty }(E'/\mathbb {Q})=0$
. This yields the Selmer-corank assertion in Condition (1) of Theorem 4.14. The additional hypothesis on
$E'$
and the vanishing hypothesis on E in Condition (1) of the present lemma together give Condition (6) of Theorem 4.14. Since every prime dividing d is a square modulo p and p splits in
$K_0$
, the integer
$dd_{K_0}$
is a square modulo p. Hence, E and
$E'$
are isomorphic over
$\mathbb {Q}_p$
, which implies Conditions (2) and (3).
Note that all primes of bad reduction split in
$\mathbb {Q}(\!\sqrt {dd_{K_0}})$
. Thus, the Tamagawa factors at these primes remain the same. At all primes dividing
$dd_{K_0}$
, the curve E has good reduction and
$E'$
has additive reduction. Thus, the Tamagawa factors at these primes are bounded by
$4$
. As p is odd, the whole Tamagawa product is not divisible by p, which proves Condition (4). The regulator Condition (5) does not depend on
$E'$
and is therefore satisfied. Thus, all six conditions of Theorem 4.14 hold.
Lemma 5.2. Let k be the number of distinct prime factors dividing N. Let
$\mathfrak {d}(S)$
denote the Dirichlet density of S. If
we have that
If
$\mathrm {Gal}(\mathbb {Q}(E[2])/\mathbb {Q})\cong S_3$
, we obtain a density of
$\mathfrak {d}(S)=(2/3)({1}/{2^{k+1}})$
.
Proof. We begin by considering the case when
$\mathrm {Gal}(\mathbb {Q}(E[2])/\mathbb {Q})\cong \mathbb {Z}/3\mathbb {Z}$
. Let
${L=K_0(\!\sqrt {-1},\sqrt {q_1},\ldots ,\sqrt {q_k})}$
, where the
$q_i$
are the prime factors of N. As all the primes dividing N are split in
$K_0$
, we see that
$d_{K_0}$
is coprime to
$q_i$
for
$1\le i\le k$
. Thus,
$[L:\mathbb {Q}]=2^{k+1}$
if
$\sqrt {-1}\in K_0$
and
$[L:\mathbb {Q}]=2^{k+2}$
otherwise. Therefore, the set of primes that are totally split in
$L/\mathbb {Q}$
has density
${1}/{2^{k+1}}$
if
$K_0=\mathbb {Q}(\!\sqrt {-1})$
and equal to
${1}/{2^{k+2}}$
otherwise. The density of the primes having Frobenius of order
$3$
in
$\mathrm {Gal}(\mathbb {Q}(E[2])/\mathbb {Q})$
is equal to
$\tfrac 23$
. As
$L\cap \mathbb {Q}(E[2])=\mathbb {Q}$
, the Tchebotarev density of S is just the product of the two densities we computed, which gives the desired result.
It remains to consider the case that
$\mathrm {Gal}(\mathbb {Q}(E[2])/\mathbb {Q})\cong S_3$
. The extension
$\mathbb {Q}(E[2])$
contains the quadratic extension
$\mathbb {Q}(\!\sqrt {\Delta (E)})$
, where
$\Delta (E)$
is the discriminant of E. As
$\mathbb {Q}(\!\sqrt {\Delta (E)})\subset \mathbb {Q}(\!\sqrt {-1},\sqrt {q_1},\ldots ,\sqrt {q_k})$
, we see that every prime that splits in
$\mathbb {Q}(\!\sqrt {-1},\sqrt {q_1},\ldots ,\sqrt {q_k})$
also splits in
$\mathbb {Q}(\!\sqrt {\Delta })/\mathbb {Q}$
. In particular, the Frobenius in
$\mathrm {Gal}(\mathbb {Q}(E[2])/\mathbb {Q})$
is either trivial or of order
$3$
. Thus, the density of S is equal to
${2}/({3\cdot 2^{k+1}})$
.
Remark 5.3. Kriz and Li give a list of imaginary quadratic fields and elliptic curves satisfying the rank assertion in Conditions (1), (6), (7) and (8) for all primes that divide N. Then, there is a positive-density set of primes p that split in
$K_0$
and satisfy Conditions (2), (3) and (4). The regulator Condition (5) and the vanishing
$\mathrm{III}(E/\mathbb {Q})[p^\infty ]=0$
in Condition (1) must be checked separately for the fixed curve.
Example 5.4. Consider the curve E with Cremona label 37a1 and the imaginary quadratic field
$K_0=\mathbb {Q}(\!\sqrt {-7})$
. We are looking for a prime p such that
$(E,K_0,p)$
satisfies the conditions of Lemma 5.1. As
$E/\mathbb {Q}$
has trivial torsion, the condition
$E(\mathbb {Q})[2]=0$
is satisfied. Furthermore, E has rank
$1$
over
$\mathbb {Q}$
such that Condition (1) is satisfied. As
$2$
splits in
$K_0$
, Condition (6) is satisfied. The only prime dividing N is
$37$
, which splits in
$K_0$
. Kriz and Li checked that Condition (8) is satisfied (see [Reference Kriz and LiKL19, Example 6.1]). Thus, we need to find a prime p with the following properties:
-
• E has good ordinary reduction at p;
-
• p is nonanomalous for E;
-
• $p\nmid \prod _{l\neq p}c_l(E/\mathbb {Q})$
; -
• the normalized p-adic regulator of E over $\mathbb {Q}$
is a p-adic unit; -
• p splits in $K_0$
.
The smallest prime
$p\ge 5$
that splits in
$K_0$
is
$11$
. It follows from the LMFDB data that
$p=11$
also satisfies all the other conditions. In particular, the Iwasawa invariants are
Since E has analytic rank one, Gross, Zagier and Kolyvagin imply that
$\mathrm{III}(E/\mathbb {Q})$
is finite; Corollary 3.12 then gives
Thus, the triple
$(E,K_0,p)$
satisfies indeed all assumptions of Lemma 5.1. The first few primes in the set S are
$\{53,617,641,1409,1453\}$
.
Kriz and Li provide a table of pairs
$(E,K_0)$
satisfying Conditions (1), (6), (7) only for primes dividing N and Condition (8). The table below records these pairs of elliptic curves and auxiliary imaginary quadratic fields
$K_0$
. The third column lists one prime such that the triple
$(E,K_0,p)$
satisfies Conditions (1)–(8) of Lemma 5.1 (compare also with the list in [Reference Kriz and LiKL19, Example 6.1]).

For every row, the same rank-one Euler-characteristic check, using the displayed prime and the corresponding Iwasawa data, verifies the fixed-curve condition
Theorem 5.5. Fix
$p\in \{11,13,31,37\}$
. For a row
$(E,K_0,p)$
of the table in Example 5.4, let
$\mathcal D_{E,K_0,p}$
denote the positive-density family of negative fundamental discriminants supplied by the Kriz–Li construction used above. Suppose that at least one row with third entry p satisfies
Then,
$(K,p)$
satisfies Criterion (
H10-gen
) for a positive density of imaginary quadratic fields K.
Proof. Fix such a row and retain those
$d\in \mathcal D_{E,K_0,p}$
for which
By the definition of
$\eta _{E,p}$
, these parameters form a positive-density subfamily. Lemma 5.1 applies to every member of this subfamily and gives Criterion (
H10-gen
).
Our computations lead us to make the following conjecture.
Conjecture 5.6. For all primes large enough, there exists an elliptic curve
$E/\mathbb {Q}$
and an imaginary quadratic field
$K_0$
such that the triple
$(E,K_0,p)$
satisfies the conditions of Lemma 5.1.
Indeed, given an odd prime number p, one expects that each of the Conditions (1)–(5) be satisfied by a positive density of elliptic curves (ordered by their height [Reference BrumerBru92]). Given an elliptic curve E that satisfies Conditions (1)–(5), it is natural to expect that each of the Conditions (6)–(8) be satisfied by a positive density of imaginary quadratic fields
$K_0$
.
In applications, it can be cumbersome to check the Heegner point conditions in Lemma 5.1. If we work with elliptic curves that admit a
$3$
-isogeny, we obtain the following result.
Theorem 5.7. Let
$E/\mathbb {Q}$
be a semistable elliptic curve of conductor N that admits a rational
$3$
-isogeny. Assume that E has rank
$1$
and let
$p>3$
be a prime such that:
-
(1) $\mathrm {corank}_{\mathbb {Z}_p}\mathrm {Sel}_{p^\infty }(E/\mathbb {Q})=1$
and
$\mathrm{III}(E/\mathbb {Q})[p^\infty ]=0$
; -
(2) E has good ordinary reduction at p;
-
(3) p is nonanomalous for E;
-
(4) $p\nmid \prod _{\ell \neq p}c_\ell (E/\mathbb {Q})$
; -
(5) the normalized p-adic regulator of E over $\mathbb {Q}$
is a p-adic unit.
Put
and let m be the number of odd primes
$\ell \neq 3$
dividing N. For a prime
$\ell $
, put
$q_\ell =\ell $
if
$\ell $
is odd and
$q_2=4$
. Finally, set
Then, there is a set
$\mathcal D_{\mathrm {KL}}$
of negative fundamental discriminants of density at least
for which the analytic rank-zero and local conditions required in Theorem 4.14 hold. Put
Then,
$(\mathbb {Q}(\!\sqrt d),p)$
satisfies Criterion (
H10-gen
) for every
$d\in \mathcal D_{\mathrm {KL}}^{\mathrm{III}}$
. In particular, if the p-primary Tate–Shafarevich density hypothesis
$\eta _{E,p}(\mathcal D_{\mathrm {KL}})>0$
holds, then the set of negative fundamental discriminants d for which
$(\mathbb {Q}(\!\sqrt d),p)$
satisfies Criterion (
H10-gen
) has positive lower density. More precisely, its lower density is at least
$\eta _{E,p}(\mathcal D_{\mathrm {KL}})$
times the displayed lower bound.
Proof. Let
$N_1$
be the product of the primes at which E has split multiplicative reduction and let
$N_2$
be the product of the primes at which E has nonsplit multiplicative reduction. Since E is semistable,
$N=N_1N_2$
. We apply [Reference Kriz and LiKL19, Theorem 9.5] to the triple
$(N_1,N_2,1)$
.
We impose in addition the local condition
$\psi _d(p)=1$
. Since
$p\nmid 3N$
, this can be incorporated into the congruence conditions used in the proof of [Reference Kriz and LiKL19, Theorem 9.5]: one adjoins p to the modulus and restricts d modulo p to the nonzero quadratic residues. The valid-pair argument of [Reference Kriz and LiKL19, Proposition 9.3] is unchanged. The additional condition contributes the factor
$p/(2(p+1))$
to the density.
We now use [Reference Kriz and LiKL19, Proposition 9.7]. If
$3\nmid N$
, or if E has split multiplicative reduction at
$3$
, that proposition gives a proportion at least
of negative fundamental discriminants for which
$E^{(d)}$
has analytic rank zero. If E has nonsplit multiplicative reduction at
$3$
, the corresponding lower bound for negative rank-zero twists is
Together with the additional condition at p, these are precisely the density bounds stated above. Denote the resulting family by
$\mathcal D_{\mathrm {KL}}$
.
Let
$d\in \mathcal D_{\mathrm {KL}}$
and put
$K=\mathbb {Q}(\!\sqrt d)$
. Since
$E^{(d)}$
has analytic rank zero, the Gross–Zagier–Kolyvagin theorem implies that
$E^{(d)}(\mathbb {Q})$
has rank zero and
$\mathrm{III}(E^{(d)}/\mathbb {Q})$
is finite. Consequently,
Moreover,
$\psi _d(p)=1$
means that p splits in K. Hence, d is a square in
$\mathbb {Q}_p^\times $
, so E and
$E^{(d)}$
are isomorphic over
$\mathbb {Q}_p$
. Thus,
$E^{(d)}$
has good ordinary reduction at p and is nonanomalous at p.
It remains to check the Tamagawa condition. If
$\ell \mid N$
and
$\ell \neq 3$
, the local conditions in [Reference Kriz and LiKL19, Theorem 9.5] ensure that
$E^{(d)}$
has nonsplit multiplicative reduction at
$\ell $
and, hence, its Tamagawa number is at most
$2$
. At a prime dividing d but not N, the twist has additive reduction and its Tamagawa number is at most
$4$
. At
$3$
, if
$3\mid N$
, the possibilities occurring in [Reference Kriz and LiKL19, Proposition 9.7] give either the original multiplicative Tamagawa factor, a nonsplit multiplicative factor of order at most
$2$
or an additive factor of order at most
$4$
. Since
$p>3$
and the original Tamagawa product is prime to p, it follows that
Thus, the pair
$(E,E^{(d)})$
satisfies Conditions (1)–(5) of Theorem 4.14. If
$d\in \mathcal D_{\mathrm {KL}}^{\mathrm{III}}$
, the fixed-curve hypothesis in Condition (1) above and the definition of
$\mathcal D_{\mathrm {KL}}^{\mathrm{III}}$
give Condition (6) as well. Theorem 4.14 therefore proves Criterion (
H10-gen
) for every such d. The final density assertion follows immediately from the definition of
$\eta _{E,p}(\mathcal D_{\mathrm {KL}})$
.
Remark 5.8. The explicit expression in Theorem 5.7 is the density of the underlying Kriz–Li family. The lower density of the subfamily for which Criterion (
H10-gen
) is proved is obtained by multiplying it by
$\eta _{E,p}(\mathcal D_{\mathrm {KL}})$
.
Example 5.9. Consider the curve with Cremona label 209a1 and
$p=5$
. The rank-one Euler-characteristic criterion verifies
$\mathrm{III}(E/\mathbb {Q})[5^\infty ]=0$
from the corresponding Iwasawa and local data. Thus,
$(E,p)$
satisfies the conditions of the above theorem. The positive-density conclusion for this example remains conditional on
${\eta _{E,5}(\mathcal D_{\mathrm {KL}})>0}$
.
5.2 Density results for
$p=3$
In this section, we fix the prime
$p=3$
and an elliptic curve
$E_{/\mathbb {Q}}$
with conductor
$N_E$
satisfying the following conditions:
-
(1) E admits a $3$
-isogeny over
$\mathbb {Q}$
; -
(2) E has good ordinary reduction at $3$
; -
(3) $\mathrm {corank}_{\mathbb {Z}_3} \mathrm {Sel}_{3^\infty }(E/\mathbb {Q})=1$
; -
(4) $\widetilde {E}(\mathbb {F}_3)[3]=0$
; -
(5) $3\nmid \prod _{\ell \neq 3} c_\ell (E/\mathbb {Q})$
; -
(6) the normalized $3$
-adic regulator of E over
$\mathbb {Q}$
is a
$3$
-adic unit; -
(7) $\mathrm{III}(E/\mathbb {Q})[3^\infty ]=0$
.
We wish to apply the results of Bhargava et al. [Reference Bhargava, Klagsbrun, Lemke Oliver and ShnidmanBKLOS19] to prove that there is a positive density of negative squarefree numbers d for which the following conditions hold for
$E^{(d)}$
:
-
(1) $\mathrm {Sel}_{3^\infty }(E^{(d)}/\mathbb {Q})=0$
; -
(2) $E^{(d)}$
has good ordinary reduction at
$3$
; -
(3) $3$
is nonanomalous for
$E^{(d)}$
, that is,
$\widetilde {E}^{(d)}(\mathbb {F}_3)[3]=0$
; -
(4) $3\nmid \prod _{\ell \neq 3} c_\ell (E^{(d)}/\mathbb {Q})$
.
Let
$\phi : E\rightarrow \widehat {E}$
be a
$3$
-isogeny defined over
$\mathbb {Q}$
. Note that this induces a
$3$
-isogeny
$\phi _d: E^{(d)}\rightarrow \widehat {E}^{(d)}$
.
Definition 5.10. Let
$\ell $
be a prime, define the local Selmer ratio as follows:
The global Selmer ratio is defined as the product of all local Selmer ratios:
We recall some of the results of [Reference Bhargava, Klagsbrun, Lemke Oliver and ShnidmanBKLOS19]. A subset
$\Sigma \subset \mathbb {Q}^*/\mathbb {Q}^{*2}$
is said to be determined by local conditions if it can be expressed as
$\Sigma = \mathbb {Q}^*/\mathbb {Q}^{*2} \cap \prod _\ell \Sigma _\ell $
, where each
$\Sigma _\ell $
is a subset of
$\mathbb {Q}_\ell ^*/\mathbb {Q}_\ell ^{*2}$
representing a local condition. If
$\Sigma _\ell = \mathbb {Q}_\ell ^*/\mathbb {Q}_\ell ^{*2}$
for all but a finite number of places
$\ell $
, we then say that
$\Sigma $
is defined by finitely many local conditions. The height of an element
$d \in \mathbb {Q}^*/\mathbb {Q}^{*2}$
is defined as
$H(d) := |d|$
, and for
$X>0$
, set
$\Sigma (X):=\{ d\in \Sigma \mid |d| < X \}$
. Define, for each
$m \geq 0$
, the subset
where
$t(\phi _d) := \mathrm {ord}_3 \, c(\phi _d)$
, and let
$\mu (T_m(\phi ))$
denote the density of
$T_m(\phi )$
within
$\mathbb {Q}^*/\mathbb {Q}^{*2}$
. The set
$T_m(\phi )$
is defined by local conditions at the primes
$\ell $
that divide
$6N\infty $
.
Theorem 5.11. Let E be an elliptic curve over
$\mathbb {Q}$
admitting a
$3$
-isogeny
$\phi \colon E \to \widehat {E}$
. Let
$\Sigma $
be a subset of
$T_0(\phi )$
defined by finitely many local conditions. Then, the proportion of twists
$E^{(d)}$
for which
$\mathrm {Sel}_3(E^{(d)})=0$
is at least
$\tfrac 12\mu (\Sigma )$
.
Proof. The above result follows from [Reference Bhargava, Klagsbrun, Lemke Oliver and ShnidmanBKLOS19, Theorem 2.5] and its proof.
We recall results in loc. cit., which provide a description for the local Selmer ratios
$c_\ell (\phi )$
. We start with
$\ell =\infty $
.
Proposition 5.12. With respect to the notation above, we have that
Proposition 5.13. For
$\ell \notin \{3, \infty \}$
, we have that
$c_\ell (\phi _d) = c_\ell (\widehat {E}^{(d)})/c_\ell (E^{(d)})$
.
Proof. This result is [Reference Bhargava, Klagsbrun, Lemke Oliver and ShnidmanBKLOS19, Corollary 10.3].
We now come to the main application of this section.
Theorem 5.14. There is unconditionally a positive density of negative squarefree integers such that
$(\mathbb {Q}(\!\sqrt d),3)$
satisfies Criterion (
H10-gen
).
Proof. The strategy is to show that there is an explicit elliptic curve with
$3$
-isogeny
$\phi $
for which the set
$T_0(\phi )$
has positive density. The values of d range in a certain subset of
$T_0(\phi )$
subject to conditions at
$3$
and
$\infty $
.
Let
$E_{/\mathbb {Q}}$
be the elliptic curve
$y^2=x^3+216x-54$
(LMFDB label 1216.o3) and set
$p:=3$
. There is a cyclic
$3$
-isogeny
$\phi :E\rightarrow \widehat {E}$
defined over
$\mathbb {Q}$
. The Iwasawa invariants are given by
Since E has analytic rank
$1$
, the Gross–Zagier–Kolyvagin theorem implies that
$\mathrm{III}(E/\mathbb {Q})$
is finite. The normalized
$3$
-adic regulator is a unit, so the
$3$
-adic height pairing is nondegenerate; moreover, the nonanomalous condition implies
${E(\mathbb {Q})[3]=0}$
. Corollary 3.12 therefore applies and the displayed Iwasawa invariants imply that
We find that the twist
$E^{(-2)}$
(LMFDB label 304.f3) has
$\mu _3(E^{(-2)}/\mathbb {Q}_{\mathrm {cyc}})=0$
and
$\lambda _3(E^{(-2)}/\mathbb {Q}_{\mathrm {cyc}})=0$
. The Tamagawa product for all curves in the isogeny class 304.f is
$1$
. It follows from Proposition 5.13 that
$c_\ell (\phi _{-2})=1$
for
$\ell \notin \{3,\infty \}$
. We find that
$c(\phi _{-2})=c_3(\phi _{-2})c_\infty (\phi _{-2})$
. According to Proposition 5.12,
$c_\infty (\phi _{-2})\in \{1,1/3\}$
. However, it follows from [Reference Bhargava, Klagsbrun, Lemke Oliver and ShnidmanBKLOS19, Theorem 10.5] that
$c_3(\phi _{-2})\in \{1,3\}$
. Therefore,
$c(\phi _{-2})\in \{1/3,1,3\}$
. Alternatively, it follows from [Reference Bhargava, Elkies and ShnidmanBES20, Proposition 42(ii)] that
Thus, we deduce that
$c(\phi _{-2})=1$
, that is,
$-2\in T_0(\phi )$
.
We now introduce local conditions at
$3$
and
$\infty $
. Consider the subset
$T_0'(\phi )$
of
$T_0(\phi )$
consisting of those d for which
$d\equiv 1\ (\text {mod}\ {3})$
and
$d<0$
. The set
$T_0'(\phi )$
is nonempty since
$-2\in T_0'(\phi )$
. The set
$T_0'(\phi )$
is defined by finitely many congruence conditions (see [Reference Bhargava, Klagsbrun, Lemke Oliver and ShnidmanBKLOS19]) and is nonempty. This implies that
$T_0'(\phi )$
has positive density. Set
By Theorem 5.11,
$\mathcal D_3$
has positive density. For every
$d\in \mathcal D_3$
, the Kummer exact sequence
shows that
$E^{(d)}(\mathbb {Q})$
has rank zero and that
$\mathrm{III}(E^{(d)}/\mathbb {Q})[3]=0$
. Since every nonzero
$3$
-primary torsion group contains an element of order
$3$
, it follows that
$\mathrm{III}(E^{(d)}/\mathbb {Q})[3^\infty ]=0$
and, consequently,
$\mathrm {Sel}_{3^\infty }(E^{(d)}/\mathbb {Q})=0$
.
For
$d\in T_0'(\phi )$
, the prime
$3$
splits in
$\mathbb {Q}(\!\sqrt d)$
because
$d\equiv 1\ (\text {mod}\ {3})$
. Since
${\widetilde E(\mathbb {F}_3)[3]=0}$
, we have
$\widetilde E^{(d)}(\mathbb {F}_3)[3]=0$
. For every prime
$\ell \neq 3$
, the Kodaira type of
$E^{(d)}$
is
$I_0$
,
$I_0^*$
,
$I_1$
, or
$I_1^*$
(see [Reference Bhargava, Klagsbrun, Lemke Oliver and ShnidmanBKLOS19, Table 3]); in particular,
Moreover, E and
$E^{(d)}$
have the same reduction type at
$3$
, so
$E^{(d)}$
has good ordinary reduction there. Thus, all six hypotheses of Theorem 4.14 hold for every
$d\in \mathcal D_3$
, and
Theorems 4.12 and 4.13 therefore show that
$(\mathbb {Q}(\!\sqrt d),3)$
satisfies Criterion (
H10-gen
) for every
$d\in \mathcal D_3$
. This proves the assertion unconditionally.











