🔢 mathematics

A phase transition for the hard sphere model on the hyperbolic plane

This paper proves the existence of a phase transition for the hard sphere model in the hyperbolic plane, resolving a major open problem that remains unsolved for the same model in Euclidean spaces $\mathbb{R}^2$ and $\mathbb{R}^3$ .

Original authors: Lewis Bowen, Marcus Michelen, Will Perkins

Published 2026-07-28

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Lewis Bowen, Marcus Michelen, Will Perkins

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a vast, invisible dance floor where tiny, perfectly round marbles are trying to find a place to stand. These marbles are the "hard spheres" of a famous physics puzzle: they can't overlap, they can't squeeze through each other, and they only interact by bumping into one another. Scientists have been watching this dance in our normal, flat world (like a sheet of paper or a room) for over a century. They strongly suspect that if you pack enough marbles in, the chaos suddenly snaps into a rigid, ordered crystal, like water freezing into ice. This is called a "phase transition." But here's the kicker: despite all the math and computer simulations, no one has ever been able to prove that this snap happens in our flat world. It remains one of the biggest unsolved mysteries in statistical physics.

To understand why this is so hard, think of the difference between a flat floor and a weird, saddle-shaped floor that curves away from you in every direction. This strange, curved world is called the "hyperbolic plane." In this universe, space expands so fast that a circle's circumference grows much faster than its radius. Because of this wild geometry, the rules of packing change. The question this paper tackles is: "If we move our marble dance from the flat floor to this hyperbolic saddle, does the chaos finally snap into order?" The authors aren't just guessing; they are building a mathematical proof to show that yes, in this curved world, the marbles do undergo a phase transition, finally settling into two very different ways of organizing themselves.

The Great Marble Dance on a Curved Floor

So, what exactly did Lewis Bowen, Marcus Michelen, and Will Perkins discover? They proved that on the hyperbolic plane, the hard sphere model definitely undergoes a phase transition. In plain English, this means that as you increase the "activity" (a fancy word for how much you want to pack marbles in), the system doesn't just get denser and denser in a smooth, predictable way. Instead, it hits a tipping point where it can exist in two completely different states at the same time.

Imagine you are the referee of this marble game. You have two ways to organize the marbles:

The "Random" Crowd: You start with an empty floor and let marbles fall in randomly, bouncing off each other. They settle into a messy, jumbled pile. Even if you keep adding more marbles, they can't quite find a perfect pattern; they remain in a state of "random-like" disorder.
The "Structured" Crystal: You start with a perfect, pre-arranged grid of marbles, like soldiers standing in formation. This is the most efficient way to pack them, leaving the least amount of empty space.

The paper proves that for certain sizes of marbles (specifically, when the radius is large enough), you can have a "Random" crowd and a "Structured" crystal coexisting at the same time, even though they have different densities. One is messy and less packed; the other is orderly and tightly packed. Because the system can be in either state depending on how you start it, we say a "phase transition" has occurred.

The Two Worlds of Packing

To pull this off, the authors had to show that these two states are truly different and can't be transformed into one another. They used a clever trick involving "entropy," which is basically a measure of how much "randomness" or "surprise" is in a system.

Think of the "Random" crowd as a group of people trying to find seats in a theater by walking in randomly. They might bump into each other, but eventually, they fill the room in a way that looks like a noisy, chaotic crowd. The authors showed that this kind of randomness has a specific mathematical "signature" (called non-negative annealed entropy).

Now, look at the "Structured" crystal. This is like a group of people who have memorized a perfect seating chart. They know exactly where to sit to maximize the number of people in the room. The authors proved that this perfect, lattice-like arrangement has a completely different mathematical signature (negative infinite annealed entropy).

Because these two signatures are so different, the "Random" crowd can never magically turn into the "Structured" crystal, and vice versa, just by changing the number of marbles. They are stuck in their own distinct worlds. This proves that the system has split into two separate phases.

Why the Curved Floor Matters

You might wonder, "Why does this work on a curved floor but not on a flat one?" The secret sauce is the shape of the universe they are playing in. The hyperbolic plane is "non-amenable," which is a fancy way of saying it's too big and too weird for the usual rules of flat space to apply. In our flat world, the "Random" crowd and the "Structured" crystal might eventually look the same if you zoom out far enough, or the math gets too messy to tell them apart. But on the hyperbolic plane, the geometry is so extreme that the "Structured" crystal is uniquely special. It's the only way to achieve the absolute maximum density for certain marble sizes.

The authors found that for a specific set of marble sizes (radii), there is only one perfect, crystal-like arrangement. They then showed that the "Random" crowd, no matter how hard it tries, can never reach that same level of perfection. It gets stuck at a lower density. This gap between the best possible random packing and the best possible crystal packing is the smoking gun that proves the phase transition exists.

The Takeaway

This paper doesn't just say, "Hey, it looks like a phase transition might happen." It actually proves it. The authors constructed two different mathematical models (measures) for how the marbles behave: one that acts like a random Poisson process (the messy crowd) and one that acts like a perfect lattice (the crystal). They proved that for large enough marbles on a hyperbolic plane, these two models are distinct and cannot be the same.

This is a huge deal because it solves a problem that has stumped physicists for over a century, but in a different universe. While we still don't have a proof for our flat, Euclidean world, this work shows that the phenomenon is real and mathematically sound in a curved space. It's like finding a new species of animal in a deep ocean cave that proves a theory about how evolution works, even if we haven't seen that exact animal on land yet. The authors have shown that the "snap" from chaos to order is a fundamental feature of hard spheres, provided the stage they dance on is curved enough to let the pattern emerge.

Technical Summary: A Phase Transition for the Hard Sphere Model on the Hyperbolic Plane

1. Problem Statement and Motivation

The hard sphere model is a fundamental system in statistical physics where particles are represented as equal-sized spheres with a hard-core exclusion interaction (no overlap allowed). In Euclidean spaces ( $\mathbb{R}^d$ ), it is a longstanding prediction that the system undergoes a phase transition as the activity (fugacity) $\lambda$ increases, leading to phenomena such as crystallization in $\mathbb{R}^3$ or rotational symmetry breaking in $\mathbb{R}^2$ . However, despite over a century of study, the existence of such a phase transition in $\mathbb{R}^d$ for any dimension $d$ remains a major open problem.

This paper addresses the existence of a phase transition in the hyperbolic plane ( $\mathbb{H}^2$ ). A phase transition is defined here as the non-uniqueness of Gibbs measures for a given radius $r$ and sufficiently large activity $\lambda$ . Specifically, the authors aim to prove that there exists a threshold $\lambda_u(\mathbb{H}^2, r) < \infty$ such that for $\lambda > \lambda_u$ , there are at least two distinct Gibbs measures.

2. Methodology and Framework

2.1. Definitions and Setup

The authors define the hard sphere model on a metric measure space $X$ (specifically $\mathbb{H}^2$ ) via the Dobrushin-Lanford-Ruelle (DLR) equations. A phase transition occurs if the set of Gibbs measures, $\text{Gibbs}(X; r, \lambda)$ , contains more than one element for some $\lambda$ .

To distinguish between potential Gibbs measures, the authors utilize the concept of density. For an isometry-invariant probability measure $\mu$ on the space of sphere packings, the density is the probability that a fixed origin is covered by a sphere in a random sample from $\mu$ .

2.2. Strategy: Constructing Two Distinct Measures

The core strategy involves constructing two distinct $(r, \lambda)$ -Gibbs measures for large $\lambda$ that have different densities:

A "Structured" Measure: Derived from optimal lattice packings.
A "Random-like" Measure: Derived from spatial birth-death Glauber dynamics, shown to be a weak Poisson factor.

The proof hinges on demonstrating that for certain radii, the density achievable by weak Poisson factors is strictly less than the optimal packing density.

2.3. Key Mathematical Tools

Benjamini-Schramm Convergence: Used to lift finite-volume Gibbs measures on hyperbolic surfaces (quotients of $\mathbb{H}^2$ ) to infinite-volume Gibbs measures on $\mathbb{H}^2$ .
Glauber Dynamics: A continuous-time Markov chain (spatial birth-death process) starting from the empty configuration. The authors show that subsequential limits of this process yield Gibbs measures that are weak Poisson factors.
Annealed (Sofic) Entropy: A notion of entropy for actions of non-amenable groups (specifically free groups). This is the primary tool used to distinguish the "structured" lattice measures from the "random-like" Poisson factors.
- The authors prove that periodic measures (supported on lattice packings) have negative annealed entropy (specifically $-\infty$ ).
- Conversely, weak Poisson factors (and weak Bernoulli factors) have non-negative annealed entropy.
- This entropy gap proves that the optimal lattice measure cannot be a weak Poisson factor.

3. Key Contributions and Results

3.1. Existence of Phase Transition in $\mathbb{H}^2$

Theorem 2: There exists an unbounded open set of radii $R_2 \subset (0, \infty)$ such that for every $r \in R_2$ , the hard sphere model on $\mathbb{H}^2$ undergoes a phase transition ( $\lambda_u(\mathbb{H}^2, r) < \infty$ ).

The set $R_2$ contains an interval $(\rho, \infty)$ for some finite $\rho$ .
This result relies on the existence of "tight radii" $r_n$ (for $n \ge 7$ ) where the optimal packing is a unique lattice packing.

3.2. Separation of Densities

Theorem 6: For the radii in the set $R_2$ , the optimal density achievable by weak Poisson factors, $D_{\text{Pois}}(\mathbb{H}^2, r)$ , is strictly less than the optimal packing density, $D_{\text{opt}}(\mathbb{H}^2, r)$ .

This gap is established by showing that the unique optimal measure at tight radii is periodic and thus not a weak Poisson factor (via the entropy argument).
By continuity of the optimal density function (Theorem 24), this gap extends to an open neighborhood of the tight radii.

3.3. Construction of Gibbs Measures

Near-Optimal Density: Using finite-volume approximations on hyperbolic surfaces with optimal lattice packings, the authors construct Gibbs measures with density arbitrarily close to $D_{\text{opt}}$ for large $\lambda$ (Section 4).
Low-Density (Poisson) Measure: Using Glauber dynamics, the authors construct a Gibbs measure that is a weak Poisson factor. Its density is bounded above by $D_{\text{Pois}}$ (Section 5).
Since $D_{\text{Pois}} < D_{\text{opt}}$ for the relevant radii, these two measures are distinct, proving non-uniqueness.

3.4. Completely Saturated Packings

Theorem 8: The paper proves the existence of completely saturated sphere packings in $\mathbb{H}^2$ with sub-optimal density.

A packing is completely saturated if no finite number of spheres can be removed and replaced with a strictly larger number of spheres while maintaining the packing property.
The authors show that any weak Poisson factor achieving the density $D_{\text{Pois}}$ is supported on completely saturated packings.
This contrasts with Euclidean space, where completely saturated packings are necessarily optimal.

4. Significance and Claims

The paper claims to resolve the existence of a phase transition for the hard sphere model in a specific non-Euclidean setting ( $\mathbb{H}^2$ ), a problem that remains open in Euclidean space.

Role of Non-Amenability: The authors emphasize that their proof strategy relies essentially on the non-amenability of the hyperbolic plane. In amenable spaces like $\mathbb{R}^d$ , the optimal density of Poisson factors equals the optimal packing density ( $D_{\text{Pois}} = D_{\text{opt}}$ ), preventing this specific separation argument.
Entropy as a Distinguisher: The work introduces annealed entropy as a rigorous tool to distinguish between "random-like" (Poisson/Bernoulli) and "structured" (periodic/lattice) Gibbs measures in the context of non-amenable groups.
Limitations: The authors are modest about the scope. They do not claim to prove phase transitions for all radii in $\mathbb{H}^2$ , nor do they extend the result to $\mathbb{H}^d$ for $d > 2$ or to $\mathbb{R}^d$ . They explicitly state that the extension to higher dimensions or Euclidean space is an open question, noting that the continuity of optimal densities and the existence of unique periodic optimizers are not known in those settings.

5. Conclusion

The paper establishes that the hard sphere model on the hyperbolic plane exhibits a phase transition for a specific, unbounded set of radii. This is achieved by constructing two distinct Gibbs measures: one with near-optimal density derived from lattice packings, and another with lower density derived from Glauber dynamics. The distinction is proven via a gap in achievable densities, which is itself established using the theory of annealed entropy to show that optimal lattice measures cannot be approximated by Poisson factors. This result highlights the unique role of non-amenability in statistical mechanical phase transitions.