BSD Invariants and Murmurations of Elliptic Curves

This paper demonstrates that while BSD invariants themselves do not exhibit murmuration oscillations, the analytic order of the Tate-Shafarevich group uniquely modulates Frobenius trace murmurations and low-lying L-function zero distributions through a pure mean shift concentrated at small primes, revealing information not captured by other standard invariants.

The Gist

Imagine you are listening to a massive choir of singers. Each singer represents an elliptic curve, a complex mathematical object that pops up in everything from cryptography to the study of prime numbers.

For a long time, mathematicians noticed something strange about this choir. When they grouped the singers by the "size" of their notes (a property called the conductor) and listened to their average pitch at different moments, the choir didn't just hum a steady tone. Instead, they swayed back and forth in a rhythmic, wave-like pattern. They called this phenomenon "Murmuration," named after the mesmerizing, synchronized flocking patterns of starlings in the sky.

This paper, by Dane Wachs, asks a very specific question: Does the "soul" of the singer affect how they murmur?

In the world of elliptic curves, the "soul" is described by a famous recipe called the Birch and Swinnerton-Dyer (BSD) formula. This formula links the local behavior of the curve (how it acts at specific prime numbers) to five global "invariants" (big-picture numbers):

1. Real Period: How much "space" the curve takes up.
2. Tamagawa Product: A measure of how the curve behaves at its "bad" spots.
3. Torsion Order: How many points loop back on themselves.
4. Regulator: A measure of the curve's complexity.
5. The Tate–Shafarevich Group ($\text{X}$): A mysterious number that measures how much the curve fails a specific "local-to-global" rule.

Here is what the paper discovered, translated into everyday language:

1. The Invariants Don't Sing the Song Themselves

First, the author checked if these five "soul" numbers (the invariants) themselves had a murmuring rhythm.

• The Analogy: Imagine checking if the singers' heights or ages sway back and forth in a wave as you look at larger groups of singers.
• The Result: No. The invariants just drift smoothly up or down as the curves get bigger. They don't have the wave-like "murmur." The rhythm belongs strictly to the local notes (the Frobenius traces), not the global summary numbers.

2. But the "Soul" Changes the Shape of the Song

This is the big discovery. Even though the invariants don't murmur themselves, they act like conductors that change how the choir murmurs.

• The Analogy: Imagine two groups of singers. One group has "soul number A" (e.g., a small Tamagawa product) and the other has "soul number B" (a large one). Even though they are singing the same song, Group A sways gently, while Group B sways wildly.
• The Result: The author found that if you sort the curves by these invariants, their murmuration patterns look significantly different.
  • Curves with a large Tamagawa product (lots of "bad spots") murmur differently than those with none.
  • Curves with a large Tate–Shafarevich group ($\text{X}$) have a completely different wave shape than those with a small one.

3. The Mystery of the "X" Group

The most exciting finding concerns the Tate–Shafarevich group ($\text{X}$). This is a notoriously difficult number to calculate and understand.

• The Control Test: The author was careful to make sure this wasn't a trick. They held everything else constant (the L-value, the period, the size of the curve) and only changed the size of $\text{X}$.
• The Result: The difference in the murmuration pattern persisted. This means the size of $\text{X}$ contains secret information about how the curve behaves at prime numbers that no other standard number can tell you. It's like finding out that a singer's hidden emotional state changes the pitch of their voice, even if you've already measured their lung capacity and age.

4. Why Does This Happen? (The "Ghost" in the Machine)

The paper digs deep to explain why the size of $\text{X}$ changes the song.

• The Mechanism: The author looked at the zeros of the curve's L-function. Think of these zeros as the "resonant frequencies" of the curve, like the specific notes a guitar string naturally wants to vibrate at.
• The Discovery: Curves with a large $\text{X}$ have their first "resonant note" shifted slightly higher than curves with a small $\text{X}$.
• The Connection: There is a mathematical rule (the Explicit Formula) that says these resonant notes dictate how the curve behaves at prime numbers.
  • The Analogy: If you shift the resonant frequency of a guitar string, the way the string vibrates when you pluck it changes. Similarly, shifting the "zero" changes the murmuration wave.
  • The paper shows that the shift in the first zero perfectly explains the "crossover" pattern seen in the data (where the wave goes up for small primes and down for large primes).

Summary: The Big Picture

This paper connects two worlds that were previously thought to be separate:

1. Local Data: How the curve behaves at individual prime numbers (the "notes").
2. Global Data: The deep, abstract structure of the curve (the "soul" or $\text{X}$).

The Takeaway: The "soul" of the curve (specifically the size of the Tate–Shafarevich group) leaves a fingerprint on the local notes. It's as if the global history of a number system whispers instructions to the local primes, changing the rhythm of the murmuration.

This is a massive step forward because it proves that the mysterious Tate–Shafarevich group isn't just an abstract number; it actively shapes the distribution of prime numbers in a way we can now see and measure. It's like discovering that the shape of a mountain (global) changes the way the wind whistles through the trees at the bottom (local), and we finally have the microphone to hear it.

Technical Summary

Here is a detailed technical summary of the paper "BSD Invariants and Murmurations of Elliptic Curves" by Dane Wachs.

1. Problem Statement and Motivation

The paper investigates the interaction between the Birch and Swinnerton-Dyer (BSD) conjecture and the recently discovered "murmuration" phenomenon in elliptic curves over $\mathbb{Q}$.

• Murmurations: Discovered by He et al. (2024), this phenomenon describes striking oscillatory patterns in the average Frobenius traces ($a_p$) of elliptic curves when plotted against primes $p$ within sliding conductor windows. The oscillation shape depends on the analytic rank (e.g., rank 0 and rank 1 curves oscillate in anti-phase).
• The Gap: While murmurations are driven by local arithmetic data ($a_p$), the BSD formula links these local traces to global invariants (Tate–Shafarevich group order $|\text{X}|$, real period $\Omega_E$, regulator $R_E$, Tamagawa product, torsion).
• Core Questions:
  1. Do the global BSD invariants themselves exhibit murmuration-type oscillations?
  2. Conversely, do specific values of BSD invariants (e.g., $|\text{X}|$, Tamagawa product) modulate the shape of the Frobenius trace murmurations?
  3. If so, what is the mechanism (e.g., via L-function zeros)?

2. Methodology and Dataset

• Dataset: The study utilizes a massive dataset of 3,064,705 elliptic curves from the Cremona database with conductors $N \le 499,998$.
• Data Processing:
  • BSD invariants were extracted for all curves.
  • Frobenius traces $a_p$ were computed for the first 500 primes ($p \le 3571$) for 657,396 curves with conductor $< 100,000$.
  • Isogeny classes were handled by using one representative per class to avoid bias from shared L-functions.
• Statistical Methods:
  • Sliding Window Analysis: Averaging invariants over windows of width $W=5000$ to detect oscillations.
  • Stratification: Partitioning curves by rank and specific BSD invariants (e.g., $|\text{X}| \ge 4$ vs. $|\text{X}|=1$) to compare murmuration profiles.
  • Null Models: Permutation tests (10,000 shuffles) to determine statistical significance ($p < 0.001$) against random groupings.
  • Confounder Control: Rigorous matching of curves by conductor, number of prime factors $\omega(N)$, L-value $L(E,1)$, and real period $\Omega_E$ to isolate the effect of specific invariants.
  • Spectral Analysis: Welch method for power spectra and moment analysis (mean, variance, skewness, kurtosis).
  • L-function Zeros: Numerical computation of the first five nontrivial zeros for 2,000 curves to test mediation via the explicit formula.

3. Key Contributions and Results

A. BSD Invariants Do Not Murmur (Null Result)

• Finding: The global BSD invariants themselves (real period, Tamagawa product, $|\text{X}|$, regulator, torsion) do not exhibit oscillatory murmuration behavior.
• Evidence: Sliding-window averages show monotone trends with respect to the conductor. Detrended residuals for rank 0 and rank 1 curves are positively correlated (in-phase), unlike the anti-phase correlation seen in Frobenius traces.
• Implication: The oscillatory behavior of local data ($a_p$) cancels out when integrated into global quantities.

B. BSD Invariants Modulate Murmuration Shape (Positive Result)

• Finding: While invariants don't murmur, they modulate the shape of the Frobenius trace murmurations.
• Evidence: Within a fixed rank (specifically rank 0), curves stratified by Tamagawa product, analytic $|\text{X}|$, or real period display significantly different murmuration profiles ($p < 0.001$).
  • Tamagawa Effect: Curves with a Tamagawa product $\ge 5$ have higher amplitude murmurations than those with product $= 1$.
  • X Effect: Curves with $|\text{X}| \ge 4$ exhibit a qualitatively different shape compared to $|\text{X}| = 1$, characterized by a "crossover" pattern (higher amplitude at small primes, lower at large primes).
• Scale Invariance: These effects persist across different conductor ranges ($5,000$to$100,000$), decaying slowly as$N^{-1/4}$, confirming they are genuine arithmetic signals.

C. The X Modulation is Genuine and Independent

• Finding: The modulation by $|\text{X}|$ is not an artifact of correlations with other invariants (like $L(E,1)$ or $\Omega_E$).
• Evidence: The effect survives a "triple control" test where curves are matched simultaneously for:
  1. $L(E,1)$ value (narrow band).
  2. Real period $\Omega_E$ (above/below median).
  3. Conductor range.
• Conclusion: The order of the Tate–Shafarevich group encodes information about the distribution of Frobenius traces at good primes that is not captured by any other standard invariant.

D. Mechanism: Pure Mean Shift and Zero Distribution

• Nature of the Effect: Diagnostic analysis reveals the $|\text{X}|$ effect is a pure mean shift in the $a_p$ distribution. Variance, skewness, and kurtosis are identical between groups; only the mean changes. The shift changes sign around $p \approx 200$.
• Mediation by L-function Zeros:
  • Curves with $|\text{X}| \ge 4$ have a systematically higher first nontrivial zero ($\gamma_1$) and lower subsequent zeros ($\gamma_2, \gamma_3$) compared to $|\text{X}|=1$ curves, even at fixed $L(E,1)$.
  • Joint Test: Hotelling's $T^2$ test on the first five zeros yields $p = 5.4 \times 10^{-9}$.
  • Explicit Formula Connection: The displacement of $\gamma_1$ explains the crossover pattern via the explicit formula. The term $\cos(\gamma_1 \log p)$ oscillates faster for larger $\gamma_1$, creating the observed sign change in the mean shift.
  • Random Matrix Theory: Both groups share the same symmetry type ($SO(\text{even})$), indicating the effect is a sub-leading arithmetic correction within a fixed universality class, not a change in symmetry type.

4. Significance and Implications

1. First Empirical Link: This is the first empirical evidence that the order of the Tate–Shafarevich group ($|\text{X}|$) influences the distribution of local Frobenius traces at good primes in a way independent of other BSD invariants.
2. Global-Local Bridge: It establishes a concrete mechanism connecting the global arithmetic of the Tate–Shafarevich group (failure of the local-global principle) to local data (Frobenius traces) via the analytic structure of the L-function (zero distribution).
3. Refinement of Murmuration Theory: The results suggest that murmuration phenomena are sensitive to the fine structure of the L-function's zero distribution, not just the analytic rank.
4. Validation of BSD: The study validates the BSD formula at a granular level, showing that constraints on the product of invariants (via $L(E,1)$) force specific adjustments in the distribution of local traces.

5. Future Directions

The paper outlines several avenues for further research:

• Computing the full Sawin–Sutherland murmuration density formula for Tamagawa strata.
• Applying the Ratios Conjecture to predict the $|\text{X}|$ modulation.
• Computing more L-function zeros (50–100) to see if they fully account for the amplitude of the murmuration difference.
• Investigating the $p$-adic structure of $|\text{X}|$ (e.g., 2-primary vs. 3-primary parts) and its distinct effects on murmurations.
• Extending the analysis to rank 1 curves and larger conductor ranges.