On indefinite integral ternary quadratic forms

This paper resolves two longstanding problems from 1990 regarding indefinite integral ternary quadratic forms, highlighted by Margulis and initiated by Serre, by developing new tools to handle high ramification in sums weighted by diophantine invariants.

The Gist

Imagine you are a master chef trying to bake the perfect loaf of bread. In the world of mathematics, this "bread" is a ternary quadratic form. Don't let the fancy name scare you; think of it as a specific recipe for a 3D shape made of numbers.

This paper, written by a team of mathematicians (Gamburd, Ghosh, Sarnak, and Whang), solves two long-standing mysteries about these number-recipes. They are like detectives solving two different cold cases: one about shapes that never touch zero (Anisotropic), and one about shapes that do touch zero (Isotropic).

Here is the story of their discovery, broken down into simple analogies.

---

Case 1: The "Never-Zero" Shapes (The Anisotropic Mystery)

The Mystery:
Imagine you have a magical machine that takes three whole numbers (like 1, 5, or -12) and plugs them into your recipe. The machine spits out a result.

• Some recipes are "Isotropic": No matter what numbers you pick, you can eventually find a combination that makes the machine output zero.
• Other recipes are "Anisotropic": You can try every combination of numbers in the universe, but the machine never outputs zero. It always spits out a non-zero number.

The mathematicians were interested in the "Anisotropic" ones. They wanted to know: How small can the smallest non-zero number be?

Think of it like a game of "Hide and Seek." The recipe is the hiding spot. The numbers you plug in are the seekers. The "Markoff Spectrum" is a list of the best hiding spots. The question is: How many of these hiding spots are there if we only look for spots where the "seeker" can't get closer than a certain distance?

The Old Guess:
For decades, people looked at the data and guessed the answer was simple: "The number of these spots grows like $X^2$." It was like guessing that if you double the size of your search area, you find four times as many hiding spots.

The New Discovery:
The authors proved the old guess was wrong. They found that the number of these spots actually grows like $X \times \log(X)$.

• The Analogy: Imagine you are counting grains of sand on a beach. The old guess said the beach gets wider and wider very fast. The new discovery says the beach gets wider, but it also gets slightly "thicker" in a way that depends on a logarithmic curve (a slow, steady climb).
• Why it matters: They had to invent a new tool called "Packets" to solve this. Imagine trying to count a massive crowd of people. Instead of counting every single person, you group them into "packets" based on their shirt color and height. The authors realized that these number-recipes naturally group themselves into packets. By studying the packets, they could count the crowd accurately, revealing that the "thickness" of the beach (the $\log X$ factor) was the missing piece of the puzzle.

---

Case 2: The "Zero-Touching" Shapes (The Isotropic Mystery)

The Mystery:
Now, let's look at the recipes that do hit zero. These are the "Isotropic" forms.
The question here is: If I pick a random recipe from a giant box, how likely is it to be one of these "zero-touchers"?

The Old Guess:
In 1990, a famous mathematician named Serre asked this question. He used a method called "Sieve Methods" (think of it like a kitchen sieve that filters out flour to keep only the big grains). He estimated the number of zero-touchers, but he couldn't get the exact formula. He knew it was roughly proportional to $X^6 / \sqrt{\log X}$, but he didn't know the exact "constant" (the secret sauce) that multiplied that number.

The New Discovery:
The authors solved this by using Homogeneous Dynamics.

• The Analogy: Imagine a spinning carousel with horses (the number recipes). Some horses are painted red (Isotropic), some blue (Anisotropic). As the carousel spins faster and faster (as the numbers get bigger), the red horses seem to spread out evenly.
• The authors used a powerful theory (Ratner's Theorem) that says: "If you spin the carousel long enough, the red horses will distribute themselves perfectly evenly according to a specific density."
• The Result: They calculated the exact "density" of these zero-touchers. They found a precise formula involving a constant ($\varpi$) that is a product of probabilities at every prime number (2, 3, 5, 7...). It's like calculating the exact odds of rolling a specific number on a die, but doing it for an infinite number of different dice simultaneously.

The Twist:
Usually, mathematicians use "Sieves" (filtering) to count things. But for this specific problem, the authors realized that using the "spinning carousel" (dynamics) was much more powerful and accurate than the sieve. They didn't need to filter; they just needed to watch the distribution.

---

The Secret Weapon: "Packets" and "Roots"

To solve both mysteries, the authors had to deal with a problem called "High Ramification."

• The Analogy: Imagine you are organizing a library. Most books are simple (like "A" to "Z"). But some books have complex, messy titles with many special characters. These are the "highly ramified" forms.
• The authors realized that instead of looking at every single book, they could look at the Root Packet. Think of a "Root Packet" as the DNA of the recipe.
  • Every recipe has a "Root" (the messy, complex part) and a "Tame" part (the simple part).
  • They realized that if you know the "Root," you can predict the behavior of the whole recipe.
  • They treated these Roots like a map. By mapping out the "Roots," they could count the recipes without getting lost in the details of every single number.

The Big Picture

This paper is a triumph of connecting different worlds:

1. Number Theory: The study of whole numbers and their secrets.
2. Geometry: The shapes these numbers make.
3. Dynamics: The study of moving systems (like the spinning carousel).

In summary:
The authors took two old, stubborn problems about number shapes.

• For the shapes that never hit zero, they showed the count grows slightly faster than expected ($X \log X$) by grouping them into "packets."
• For the shapes that do hit zero, they used the "spinning carousel" of dynamics to find the exact probability of finding one, replacing old, rough estimates with a precise, beautiful formula.

They didn't just count the grains of sand; they figured out the exact physics of how the sand settles.

Technical Summary

Here is a detailed technical summary of the paper "On Indefinite Integral Ternary Quadratic Forms" by Alexander Gamburd, Amit Ghosh, Peter Sarnak, and Junho Peter Whang.

1. Problem Statement

The paper addresses two fundamental, long-standing problems concerning indefinite integral ternary quadratic forms (forms $F(x_1, x_2, x_3)$ with integer coefficients, non-zero determinant, and signature $(2,1)$ or $(1,2)$).

1. The Markoff Spectrum Problem (Anisotropic Forms):

   • Context: Initiated by Markoff (1902) and highlighted by Margulis (1990). It concerns the distribution of values $\mu(F)$, defined as the normalized minimum of $|F(x)|$ for non-zero integer vectors $x$.
   • Question: What is the asymptotic behavior of the counting function $MAR(X)$, which counts the number of distinct values in the Markoff spectrum greater than or equal to $1/X$?
   • Previous Status: Numerical studies suggested a quadratic growth ($X^2$), but this was misleading. Theoretical bounds were loose.

2. The Density of Isotropic Forms:

   • Context: Initiated by Serre (1990). A form is isotropic if it represents zero non-trivially over the integers ($F(x)=0$ has a non-zero integer solution).
   • Question: What is the density of primitive isotropic forms within a large family of integral forms (e.g., those with coefficients in a box of size $X$)?
   • Previous Status: Sieve methods provided upper and lower bounds of order $X^6 / \sqrt{\log X}$, but the exact asymptotic constant and the precise local density structure were unknown.

2. Methodology

The authors develop a sophisticated toolkit combining analytic number theory, algebraic number theory (genus theory), and homogeneous dynamics.

• Packet Decomposition (Root Packets):

  • The authors introduce the concept of packets and root packets. They decompose the determinant $D$ of a form into a "tame" squarefree part $s$ and a "highly ramified" part $r$ (where $p^2 | r$ for $p|r$).
  • A root packet $H$ captures the local equivalence classes of forms at primes dividing $r$ (including $p=2$). This allows the authors to partition the set of all forms into manageable families based on their local behavior at highly ramified primes.
  • This decomposition is crucial for handling the "high ramification" that complicates standard genus theory.

• Weighted Class Number Asymptotics:

  • The proofs rely on establishing asymptotic formulas for sums of class numbers and genera, weighted by invariants like the minimum $\kappa(F)$ or Siegel's mass density $\rho(F)$.
  • They utilize the Landau-Selberg-Delange method (as refined by Granville and Koukoulopoulos) to handle sums over squarefree numbers and multiplicative functions arising from the packet decomposition.

• Sieve Methods (for Anisotropic Forms):

  • To control the minimum $\kappa(F)$, the authors employ a lower-bound Brun sieve. This allows them to construct many integers that are locally represented by a form but not globally, or conversely, to show that small values are represented by many classes.
  • They refine Watson's bounds on $\kappa(F)$ using Burgess' bounds for character sums, establishing that $\kappa(F)$ is generally small relative to the determinant, except for specific "large" points in the spectrum.

• Homogeneous Dynamics (for Isotropic Forms):

  • For the isotropic case, the authors use equidistribution results (Eskin-Oh, Ratner's theorems) to localize the counting problem.
  • They relate the count of integral forms to the volume of orbits in the space of symmetric matrices, utilizing Siegel's Mass Formula to convert the sum over classes into a sum over genera.
  • A key insight is that for isotropic forms, the relevant mass functions exhibit multiplicativity in the determinant, allowing for an explicit Euler product evaluation.

3. Key Contributions and Results

A. Resolution of the Markoff Spectrum Asymptotics (Theorem 1.1)

The authors prove that the number of Markoff spectrum values $\ge 1/X$ grows as:
$$MAR(X) \sim \gamma X \log X$$

• Significance: This corrects the previous heuristic guess of $X^2$. The growth is slightly super-linear ($X \log X$).
• Constant $\gamma$: It is given by an explicit convergent infinite series of positive terms. The authors show $\gamma > \frac{19}{20} \frac{\zeta(2)}{\zeta(4)}$, reflecting that the minimum $\kappa(F)$ is often larger than 1.
• Mechanism: The proof involves showing that the sum over classes is dominated by those with small $\kappa(F)$, and using the packet decomposition to show that the contribution from "large" $\kappa(F)$ is negligible.

B. Exact Density of Isotropic Forms (Theorem 1.3)

The authors establish the precise asymptotic density of primitive isotropic forms in a dilated domain $X\Omega$:
$$ISO_{prim}(X\Omega) \sim \varpi \cdot \text{Vol}(\Omega_{iso}) \frac{X^6}{\sqrt{\log X}}$$

• Constant $\varpi$: This is an explicit constant derived from an infinite product of local $p$-adic probabilities.
  $$\varpi = \frac{2}{\Gamma(1/2)} \prod_p \lambda_p^{iso} \cdot (1-p^{-1})^{-1/2}$$
  where $\lambda_p^{iso}$ is the $p$-adic probability that a random symmetric matrix is primitive and isotropic.
• Local Density: The paper provides a refined local density formula $d\lambda_{iso}$, showing that the distribution of isotropic forms is not uniform but weighted by $1/\sqrt{\log D(F)}$.
• Significance: This resolves Serre's 1990 question, moving from sieve bounds to an exact asymptotic with a computable constant.

C. Class Number and Genus Asymptotics (Results A, B, C, D)

The paper establishes several intermediate asymptotic formulas of independent interest:

• (A) & (B): The number of genera $g(d)$ and classes $h(d)$ with determinant $\le X$ both satisfy:
  $$\sum_{d \le X} g(d) \sim \sum_{d \le X} h(d) \sim \frac{19}{20} \frac{\zeta(2)}{\zeta(4)} X \log X$$
  This implies that "almost all" genera of indefinite ternary forms have class number 1 (in the sense of density).
• (C): The weighted sum involving the minimum $\kappa(C)$ matches the growth of the Markoff spectrum.
• (D): The weighted sum of Siegel's mass $\rho(C)$ over isotropic classes yields the asymptotic needed for Theorem 1.3.

4. Significance and Impact

• Resolution of Open Problems: The paper definitively solves two major problems posed by Margulis and Serre that had remained open for over 30 years.
• Methodological Innovation: The introduction of root packets and the systematic handling of high ramification (especially at $p=2$) provides a new framework for studying quadratic forms that goes beyond classical genus theory. This is applicable to other problems involving sums over classes of forms with arithmetic invariants.
• Interplay of Fields: The work successfully bridges homogeneous dynamics (for the isotropic case) and analytic number theory/sieve methods (for the anisotropic case), demonstrating how tools from one area can solve problems in another.
• Explicit Constants: Unlike many asymptotic results in number theory where the constant is unknown or non-computable, this paper provides explicit, computable infinite products and series for the leading constants.

In summary, the paper provides a deep structural understanding of indefinite ternary quadratic forms, correcting previous numerical misconceptions about the Markoff spectrum and providing the first exact asymptotic formula for the density of isotropic forms.