Why Local Independence Fails for Elliptic Curves Ordered by Conductor: Cubic Fields, Small Values and Central Zeros

For elliptic curves ordered by naive height, Frobenius traces at distinct primes are asymptotically independent. We show numerically that this fails for isogeny classes ordered by conductor (permutation p-value at most 6 · 10⁻⁴, also within ranks 0, 1 and 2), and ask why. First, the parities of a_p and a_q are negatively correlated. Without rational 2-torsion the parity detects inertness of p in the cubic 2-division field, and we prove, as a quantitative consequence of the counting theorem of Bhargava, Taniguchi and Thorne, that inertness at distinct primes is negatively correlated among cubic fields, with leading term κ± Δ_p Δ_q |d|^(-1/6) (the Cubic-Field Inertness Correlation Theorem). A census of 68,454 fields agrees with this term beyond the range where its dominance is proved. Removing the complex fields with a root of the reduced cubic form at its four smallest points brings the correlation within errors of zero (3 · 10⁴ ≤ |d_K| < 3 · 10⁵, p, q ∈ {7, 11, 13}), a comparison of selections, not a calibrated test. The transfer to elliptic curves is conditional on an open hypothesis on their 2-division fields, consistent with the data in four windows below 5 · 10⁴; an excess near 5 · 10⁵, nominally significant only in a post hoc family of eight tests and not over all 36 shells, needs independent confirmation. Second, the values are negatively correlated in positive rank. In rank 0 the logarithm of the first non-central zero accounts linearly for 80% of the variance of the logarithm of the partial Euler product at the centre, but correlates at 0.90 with log L(E, 1), so no causal role is identified. We certify the central order of vanishing for 5,400 curves and the first zero for 1,080; other zeros are floating-point. An excised orthogonal ensemble agrees with the data only qualitatively and fails at N ≈ 10⁷. Results are labelled as proved, conditional, numerical or open.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-16
DOI
https://doi.org/10.5281/zenodo.24285041
Primary Topic
Analytic Number Theory Research
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Why Local Independence Fails for Elliptic Curves Ordered by Conductor: Cubic Fields, Small Values and Central Zeros

Igor Merlini
Zenodo (CERN European Organization for Nuclear Research)
Analytic Number Theory Research
preprint

Why Local Independence Fails for Elliptic Curves Ordered by Conductor: Cubic Fields, Small Values and Central Zeros

Igor Merlini
preprint en

Abstract

For elliptic curves ordered by naive height, Frobenius traces at distinct primes are asymptotically independent. We show numerically that this fails for isogeny classes ordered by conductor (permutation p-value at most 6 · 10⁻⁴, also within ranks 0, 1 and 2), and ask why. First, the parities of a_p and a_q are negatively correlated. Without rational 2-torsion the parity detects inertness of p in the cubic 2-division field, and we prove, as a quantitative consequence of the counting theorem of Bhargava, Taniguchi and Thorne, that inertness at distinct primes is negatively correlated among cubic fields, with leading term κ± Δ_p Δ_q |d|^(-1/6) (the Cubic-Field Inertness Correlation Theorem). A census of 68,454 fields agrees with this term beyond the range where its dominance is proved. Removing the complex fields with a root of the reduced cubic form at its four smallest points brings the correlation within errors of zero (3 · 10⁴ ≤ |d_K| < 3 · 10⁵, p, q ∈ {7, 11, 13}), a comparison of selections, not a calibrated test. The transfer to elliptic curves is conditional on an open hypothesis on their 2-division fields, consistent with the data in four windows below 5 · 10⁴; an excess near 5 · 10⁵, nominally significant only in a post hoc family of eight tests and not over all 36 shells, needs independent confirmation. Second, the values are negatively correlated in positive rank. In rank 0 the logarithm of the first non-central zero accounts linearly for 80% of the variance of the logarithm of the partial Euler product at the centre, but correlates at 0.90 with log L(E, 1), so no causal role is identified. We certify the central order of vanishing for 5,400 curves and the first zero for 1,080; other zeros are floating-point. An excised orthogonal ensemble agrees with the data only qualitatively and fails at N ≈ 10⁷. Results are labelled as proved, conditional, numerical or open.

Zenodo (CERN European Organization for Nuclear Research)
Analytic Number Theory Research
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon