On Lower Bounds for sums of Fourier Coefficients of Twist-Inequivalent Newforms

This paper establishes lower bounds for the sums of Fourier coefficients of two twist-inequivalent, non-CM normalized newforms, proving that their largest prime factor grows significantly for almost all primes and that small sums on a positive-density set of primes imply the forms are twist-equivalent by a quadratic character.

The Gist

Imagine you are a detective trying to solve a mystery involving two very special, invisible musicians. Let's call them F and G.

These musicians don't play standard notes; they play "Fourier coefficients." Think of these coefficients as a secret code or a unique fingerprint that changes every time they play a new prime number (2, 3, 5, 7, 11, etc.).

The Setup: Two Different Musicians

In the world of math, these musicians are called Newforms. They are "normalized" (they follow strict rules) and "non-CM" (they don't have a special, repetitive symmetry that makes them boring).

The big question the authors ask is: What happens when F and G play together?
Specifically, if you take the note F plays for a prime number $p$ and add it to the note G plays for the same prime, what does the result look like?

The authors are looking for a specific property: How "complex" is this sum?
In math, complexity is often measured by the largest prime factor of a number.

• If the sum is 12, the largest prime factor is 3 (since $12 = 2 \times 2 \times 3$).
• If the sum is 17, the largest prime factor is 17.
• If the sum is a huge prime number, the largest prime factor is that huge number itself.

The authors want to prove that when F and G are different enough (mathematically "twist-inequivalent," meaning they aren't just playing the same song with a slight variation), their combined notes will almost always be wildly complex. They won't be made up of tiny, simple building blocks; they will contain massive, giant prime numbers.

The Main Discovery: The "Giant Prime" Guarantee

The paper proves a fascinating rule:

If you pick almost any prime number $p$ (like 1,000,003 or 999,999,937), and you add F's note to G's note, the resulting number will almost certainly have a giant prime factor.

How big?
The paper shows that this giant prime factor is at least as big as a specific formula involving logarithms: roughly $(\log p)^{1/14}$.

• Analogy: Imagine $p$ is a mountain. The authors prove that the "sum" of the two musicians' notes is a boulder sitting on that mountain, and that boulder is made of a rock so huge that it dwarfs the mountain's base. It's not a pebble; it's a massive boulder.

They also prove that this isn't just true for prime numbers. If you look at all integers (not just primes), this "giant prime factor" phenomenon happens for 99.9...% of all numbers (a set with "natural density 1").

The "Twist" (The Twist-Inequivalence Rule)

Why does this happen? It depends on the relationship between F and G.

• Scenario A: F and G are totally different (Twist-Inequivalent).
  • Result: Their sum is chaotic and huge. The "Giant Prime" rule holds.
• Scenario B: F and G are secretly related (Twist-Equivalent).
  • Result: If they are related, their notes might cancel each other out or stay small.

The authors turn this around to create a test. They say: "If you see a situation where the sum of their notes is small for a lot of prime numbers, then you know for a fact that F and G are actually the same musician in disguise (related by a quadratic character)."

• Analogy: It's like listening to two singers. If they almost always sing notes that cancel each other out into silence, you know they are singing the same song in perfect harmony. If they are singing different songs, the noise they make together is loud and messy.

The "Super Detective" Mode (Assuming GRH)

The paper also looks at what happens if we assume a famous, unproven math hypothesis called the Generalized Riemann Hypothesis (GRH).

• Without GRH: We can prove the giant prime factor exists, but it's a bit modest in size.
• With GRH: The authors can prove the sum grows exponentially.
  • Analogy: Without the hypothesis, we know the boulder is heavy. With the hypothesis, we know the boulder is actually a planet. The numbers get astronomically large, much faster than we expected.

The "Sieve" (How they found the answer)

To prove these things, the authors used a mathematical tool called Brun's Sieve.

• Analogy: Imagine you have a giant bucket of sand (all the integers). You want to find the grains of gold (numbers with giant prime factors). You pour the sand through a sieve.
  • The sieve removes the tiny grains (numbers made of small primes).
  • The authors proved that if you use the right sieve, almost all the sand that falls through is gold. The "dirt" (numbers with only small prime factors) is so rare it's practically zero.

Summary for the Everyday Reader

1. The Players: Two unique mathematical patterns (Newforms) that generate numbers based on prime numbers.
2. The Action: Adding their numbers together.
3. The Finding: If the patterns are truly different, their sum is almost never "simple." It almost always contains a massive prime number as a factor.
4. The Implication: If you ever see these sums staying small, it's a dead giveaway that the two patterns are actually related.
5. The Bonus: If a famous math guess (GRH) is true, these sums are even more explosively large than we thought.

In short, the paper proves that in the chaotic dance of numbers, two different dancers rarely end up in a simple, quiet pose. They almost always create a massive, complex explosion of prime numbers.

Technical Summary

1. Problem Statement and Context

The paper investigates the arithmetic properties of the sums of Fourier coefficients of two distinct, non-CM (Complex Multiplication) normalized newforms, $f$ and $g$, of weights $k \ge 2$. Specifically, the authors focus on the quantity $a_f(p) + a_g(p)$ for prime numbers $p$.

While the Ramanujan-Petersson conjecture (proven by Deligne) provides a sharp upper bound for individual coefficients ($|a_f(p)| \le 2p^{(k-1)/2}$), the behavior of lower bounds for these coefficients, and specifically for their sums, remains a challenging area.

• The Core Question: If $f$ and $g$ are twist-inequivalent (i.e., $f \neq \chi \otimes g$ for any Dirichlet character $\chi$), how small can the sum $a_f(p) + a_g(p)$ be?
• Specific Focus: The authors aim to establish lower bounds for the largest prime factor, denoted $P(n)$, of the integer sum $a_f(p) + a_g(p)$. They also investigate the multiplicity one theorem implications: if the sum is "small" for a positive density of primes, does it imply the forms are twist-equivalent?

2. Methodology

The authors employ a sophisticated blend of analytic number theory, algebraic geometry, and sieve methods:

• Joint Sato-Tate Theorem: They utilize the equidistribution of the normalized coefficients $(\frac{a_f(p)}{p^{(k-1)/2}}, \frac{a_g(p)}{p^{(k-1)/2}})$ in $[-2, 2] \times [-2, 2]$ with respect to the product Sato-Tate measure. This is crucial for proving that the sum $a_f(p) + a_g(p)$ is "large" for a density-one set of primes.
• Galois Representations: The proof relies heavily on the mod-$h$ Galois representations $\bar{\rho}_{f,h}$ and $\bar{\rho}_{g,h}$ associated with the newforms. By considering the product representation $\bar{\rho}_h$, they define a finite Galois extension $L_h/\mathbb{Q}$.
• Chebotarev Density Theorem: They apply effective versions of the Chebotarev Density Theorem (Lagarias-Odlyzko) to estimate the number of primes $p$ for which $a_f(p) + a_g(p) \equiv 0 \pmod h$. This involves analyzing the size of the image of the Galois group and the specific conjugacy class where the trace sum is zero.
• Brun's Sieve: To extend results from primes to integers, they use Brun's sieve to construct sets of natural numbers with density 1 that inherit the large prime factor properties from the underlying primes.
• Turán's Method (Normal Order): Under the Generalized Riemann Hypothesis (GRH), they adapt Turán's method to study the normal order of the number of distinct prime factors $\omega(a_f(p) + a_g(p))$, which allows them to derive exponential growth bounds.

3. Key Contributions and Results

A. Unconditional Lower Bounds (Theorem 1.1)

For two twist-inequivalent non-CM newforms with integer Fourier coefficients, the authors prove that for almost all primes $p$:
$$P(a_f(p) + a_g(p)) > (\log p)^{1/14} (\log \log p)^{3/7 - \epsilon}$$
This result establishes that the sum cannot be composed solely of small prime factors; it must have a "large" prime factor growing with $p$.

B. Extension to Integers (Theorem 1.2)

Using Brun's sieve, the authors extend the result to the convolution function $(a_f * a_g)(n)$. They show that the set of integers $n$ where either the sum is zero or the largest prime factor satisfies the same lower bound has natural density 1.

C. Multiplicity One Theorem Variation (Corollary 1.4)

The paper provides a converse result related to the Multiplicity One Theorem. If the set of primes where $|a_f(p) + a_g(p)|$ is "small" (bounded by the lower bound derived in Theorem 1.1) has a positive upper density, then $f$ and $g$ must be twist-equivalent by a quadratic character.

• Significance: This contrasts with Ramakrishnan's theorem (which assumes $a_f(p)^2 = a_g(p)^2$). Here, the condition is on the sum being small, yet it leads to the same structural conclusion about the forms.

D. Results under GRH (Theorem 1.5)

Assuming the Generalized Riemann Hypothesis, the bounds are significantly strengthened to exponential growth:

1. Largest Prime Factor: $\log P(a_f(p) + a_g(p)) \ge (\log p)^{1-\epsilon}$.
2. Magnitude of Sum: $\log |a_f(p) + a_g(p)| \ge (\log \log p)^{1/2+\epsilon}$.
   This implies that under GRH, the sum grows much faster than the unconditional bounds suggest.

E. Normal Order of Prime Factors (Theorem 8.1)

The authors establish that the number of distinct prime factors $\omega(a_f(p) + a_g(p))$ has a normal order of $\log \log p$ (under GRH). This is a critical step in proving the exponential growth of the largest prime factor.

4. Significance and Impact

• Advancement of Lower Bound Theory: While upper bounds for Fourier coefficients are well-understood (Deligne), lower bounds are notoriously difficult. This paper provides the first non-trivial lower bounds specifically for the sum of coefficients of two distinct forms, filling a gap in the literature.
• Refinement of Multiplicity One: The corollary offers a new perspective on the uniqueness of newforms. It demonstrates that "smallness" of the sum on a positive density set is a strong indicator of twist-equivalence, reinforcing the rigidity of the space of modular forms.
• Methodological Synthesis: The paper successfully integrates the Joint Sato-Tate distribution (probabilistic) with Chebotarev density (algebraic) and sieve methods (combinatorial). This multi-faceted approach serves as a robust template for future investigations into sums of automorphic forms.
• Conditional vs. Unconditional: The distinction between the unconditional polynomial-logarithmic bounds and the conditional exponential bounds highlights the current limitations of analytic number theory without GRH, while showing the potential depth of the problem.

In summary, the paper demonstrates that for non-twist-equivalent newforms, the sum of their Fourier coefficients is arithmetically "rich," possessing large prime factors for almost all primes, and that deviations from this behavior imply a deep algebraic relationship between the forms.