Modularity for every elliptic curve over an imaginary quadratic field

← All digests

Family 030 · Number theory
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-10

1. The question

An elliptic curve is given by a cubic equation such as $y^2=x^3+Ax+B$, together with a point at infinity and a nonsingularity condition. Reducing its equation modulo different primes produces a sequence of point counts. Modularity predicts that these arithmetic data come from an automorphic form, an analytic object with strong symmetries.

Over the rational numbers this connection is the modularity theorem associated with Wiles’s work and its successors. The analogous question over an imaginary quadratic field is more difficult. Such a field has the form $K=\mathbb Q(\sqrt{-d})$, with $d>0$ squarefree, so its arithmetic has a complex place rather than a real one.

The manuscript claims modularity for every elliptic curve over every imaginary quadratic field. It also claims agreement of the full local parameters at every place, including primes of bad reduction. This is stronger than matching point counts at almost all good primes, and stronger than obtaining modularity only after extending the ground field.

2. Background and essential definitions

Point counts and local factors

A finite place $v$ of $K$ corresponds to a prime ideal of its ring of integers. Its residue field $k_v$ has finite size $q_v$. At a place of good reduction, the reduced elliptic curve $E_v$ is still nonsingular. Define

$$a_v=q_v+1-\#E_v(k_v).$$

The local factor of the curve’s $L$-function is

$$L_v(E/K,s)=\frac{1}{1-a_vq_v^{-s}+q_v^{1-2s}}.$$

The global $L$-function multiplies local factors over finite places, with appropriately modified factors at bad reduction. It packages the arithmetic of the curve into one analytic expression.

An automorphic representation is a systematic way to package a space of automorphic forms and its local symmetries. In this setting the group is $\operatorname{GL}_2$ over the adeles $\mathbb A_K$, which combine all completions of $K$. For understanding modularity, the essential point is that the representation supplies local polynomials and local parameters to compare with those of $E$. Over imaginary quadratic fields, the relevant forms are often called Bianchi forms.

Galois representations connect the two sides

The absolute Galois group $G_K$ consists of automorphisms of an algebraic closure fixing $K$. It acts on torsion points of $E$. Keeping all powers of a prime $\ell$ produces a two-dimensional $\ell$-adic representation; the manuscript uses its cohomological counterpart $H^1_{\mathrm{\acute et}}(E_{\overline K},\mathbb Q_\ell)$.

At good places, Frobenius is a distinguished Galois action reflecting the finite-field power map. Its polynomial encodes $a_v$. At bad places, the local parameter also records inertia, which describes ramification, and a monodromy operator measuring a further part of the degeneration. These extra data are why matching full local parameters says more than matching good-prime counts.

Reducing a stable lattice in an $\ell$-adic representation modulo $\ell$ gives a residual representation. A modularity-lifting theorem says, under specified local and image hypotheses, that a characteristic-zero representation is automorphic if its residual representation already comes from an automorphic object. A compatible system is a family of representations for different coefficient primes with common Frobenius polynomials away from a finite set.

An elliptic curve has complex multiplication, or CM, if its endomorphism ring over an algebraic closure is larger than $\mathbb Z$. Such curves have extra symmetries and a separate description by characters. The non-CM case is the main geometric construction in this manuscript.

3. Development history and previous partial results

Bianchi’s nineteenth-century work introduced the arithmetic groups associated with imaginary quadratic fields and their actions on hyperbolic three-space. Swan and Harder developed their group structure and cohomology, while Grunewald–Helling–Mennicke, Cremona and Cremona–Whitley established extensive computational links between these forms and elliptic curves.

The general expectation belongs to the Langlands program, which connects Galois representations with automorphic representations. Over $\mathbb Q$, Wiles and Taylor–Wiles proved modularity for semistable elliptic curves. Breuil, Conrad, Diamond and Taylor then completed the theorem for all elliptic curves over $\mathbb Q$.

Transferring the method to imaginary quadratic fields involves new difficulties. Relevant cohomology occurs in more than one degree, and constructing the associated Galois representations requires different tools. Work of Harris–Soudry–Taylor, Taylor and Berger–Harcos supplied important earlier cases. Harris–Lan–Taylor–Thorne established general constructions over CM fields, and Scholze incorporated torsion cohomology. Calegari–Geraghty’s patching method addressed the multiple-degree setting.

Potential automorphy theorems of Allen and collaborators, and work of Boxer–Calegari–Gee–Pilloni, imply modularity after suitable extensions in broad situations. Potential modularity permits changing the ground field; descending to the original field is an additional issue.

Allen–Khare–Thorne proved small-characteristic residual results and modularity for positive proportions of curves. Caraiani–Newton obtained lifting and local-global compatibility theorems that cover small coefficient primes and ramified CM fields. Their results include modularity for 100% of Weierstrass equations in the stated height ordering over suitable Galois CM fields, and every elliptic curve over an imaginary quadratic field $K$ for which $X_0(15)(K)$ has rank zero.

Here $X_0(15)$ is the modular curve classifying elliptic curves with a cyclic subgroup of order fifteen, and rank zero means its rational-point group has no free part. A density-one theorem still allows exceptional curves, and the rank condition still excludes some fields. The manuscript claims to remove these remaining restrictions while using Caraiani–Newton’s lifting theorem as an essential input.

4. What the new manuscripts claim

For every imaginary quadratic $K$ and every elliptic curve $E/K$, M1 claims an isobaric automorphic representation $\pi_E$ of $\operatorname{GL}_2(\mathbb A_K)$ with trivial central character and weight zero, realizing the curve’s cohomology for every coefficient prime $\ell$.

“Isobaric” allows both a cuspidal representation and, where appropriate, an automorphic sum of characters. A cuspidal representation is the discrete, genuinely two-dimensional kind relevant to non-CM curves. The theorem claims cuspidality when $E$ is non-CM. In the CM case, it uses automorphic induction from a Hecke character, or an isobaric sum when the CM field is already $K$.

At every finite place away from the chosen coefficient prime, the Frobenius-semisimplified Weil–Deligne parameter matches the automorphic local parameter with the manuscript’s Tate normalization. Frobenius semisimplification is the standard convention for this comparison; it retains the inertia and monodromy data being asserted. Varying the coefficient prime covers all finite places.

The paper uses unitary normalization on the automorphic side. Accordingly, its identity of $L$-functions is

$$L(E/K,s)=L(\pi_E,s-\tfrac12).$$

The shift is a normalization, not a discrepancy in arithmetic data. At the complex place, the unitary parameter is

$$z\longmapsto \begin{pmatrix} z/|z|&0\\ 0&\overline z/|z| \end{pmatrix},$$

corresponding to the two Hodge types $(1,0)$ and $(0,1)$ of an elliptic curve. The theorem thus specifies both the finite and archimedean behavior.

5. Why the advance matters

Importance: 5/5, provisionally. The claim completes a major case of the modularity program over a natural class of number fields. It replaces results with field restrictions, residual-image restrictions or possible exceptional curves by a statement for every elliptic curve in the category.

Full local compatibility matters for applications. Bad-prime factors and ramification are part of the arithmetic $L$-function, so identifying only almost all good-prime polynomials leaves important local information unstated. The manuscript explicitly addresses that information.

The proposed construction also offers a geometric way to move between coefficient primes. Rather than finding a single easy residual representation for the original curve, it builds a chain of related systems inside one family of Jacobians. The rating reflects this combination of a central theorem and a potentially reusable method, assuming the stated claims.

6. The main idea

Build one family containing several useful systems

Fix a non-CM curve $E/K$. Choose a sufficiently large prime $p$ for which the action on $E[p]$ has large image, and set $n=3p$. Over $F=K(\mu_n)$, where $\mu_n$ denotes the $n$th roots of unity, construct cyclic covers of quadratic twists $B_t$ of $E$.

A quadratic twist becomes isomorphic to $E$ after a quadratic extension. Each cover has genus $n$, and its Jacobian is an $n$-dimensional abelian variety encoding the curve’s first cohomology. The cyclic rotations split that cohomology into rank-two character pieces $V_\theta$.

Choose characters $\eta$ and $\psi$ of orders three and $p$, respectively. Their product has order $3p$. The manuscript constructs the chain

$$V_\eta\equiv V_{\eta\psi}\pmod p, \qquad V_{\eta\psi}\equiv V_\psi\pmod3, \qquad V_\psi\equiv V_1\pmod p.$$

These congruences identify semisimplified residual representations. Character values of order $p$ become trivial after reduction modulo $p$, and the analogous statement holds for order three. The chain’s first piece comes from an auxiliary elliptic curve $A/F$; the last is the cohomology of $B_t$.

Make each lifting step applicable

Congruences alone do not transfer automorphy. The residual images and the local behavior at coefficient primes must satisfy the lifting theorem.

Two degenerations of the covering family supply complementary controls. In one, the Jacobian approaches a product $E^n$. This gives an integral description of the cohomology and permits simultaneous control of reduction behavior. In the other, vanishing cycles supply opposite unipotent monodromy operators, helping ensure the required irreducibility and residual genericity.

Local constancy of sufficiently deep finite torsion, together with semistability and the finite possibilities for Frobenius polynomials, preserves the needed local types near the first degeneration. A specialization argument based on Hilbert irreducibility chooses one parameter satisfying the global image requirements and these local conditions at once. It treats the two conjugate families together to control all places above the relevant split rational primes.

Lift three times, then descend

The prime five is used to establish initial modularity of the auxiliary curve $A$. Automorphy then passes through the displayed chain using coefficient primes $p$, three and $p$, in that order. Compatibility across coefficient primes supplies the automorphic information needed for each successive step.

This first yields modularity of $B_t$ over $F$. Since $F/K$ is abelian, the manuscript descends through a tower of cyclic prime-degree extensions. A finite-order character adjustment ensures the descended Galois representation is the prescribed one, rather than merely a twist with the same restriction upstairs. Undoing the original quadratic twist then gives modularity of $E/K$.

Finally, local-global compatibility and the geometry of the reduction types identify the full finite parameters. The CM cases are handled separately through the classical theory of complex multiplication and automorphic induction.

7. Scope, limitations, and open questions

The theorem concerns imaginary quadratic ground fields. It does not assert modularity for every elliptic curve over every number field, or a general automorphy theorem for all higher-dimensional abelian varieties.

Its use of an extension during the proof should not obscure the final conclusion: the claimed automorphic object is over the original field $K$. Conversely, the existence theorem does not provide an efficient algorithm for computing every associated Bianchi form or its coefficients.

Modularity also does not by itself settle the Birch–Swinnerton-Dyer formula for these curves. Analytic continuation and automorphic structure are different from an exact formula relating ranks, leading values and arithmetic invariants.

The non-CM and CM alternatives must remain distinct, as must the unitary and cohomological normalizations. The digest reports the manuscript’s statement with these conventions rather than treating every curve as giving the same kind of cuspidal object.

8. Sources and further reading

Caraiani and Newton’s On the modularity of elliptic curves over imaginary quadratic fields is the key primary predecessor and a central input to M1. The original rational-field history appears in Wiles, Modular elliptic curves and Fermat’s last theorem, Taylor–Wiles, Ring-theoretic properties of certain Hecke algebras, and Breuil–Conrad–Diamond–Taylor, On the modularity of elliptic curves over Q: wild 3-adic exercises, all cited in M1.

For the distinction between potential and actual automorphy, M1 also cites Allen and collaborators, Potential automorphy over CM fields. Its own Sections 3–9 explain the covering family, specialization, three lifts, descent and final local comparisons.

Source manuscripts

• M1. OpenAI, Modularity of elliptic curves over imaginary quadratic fields (October 4, 2026).

Sources consulted on 10 October 2026 at repository revision fd4aeeb. Download the manuscript citations.