Columnar order in random packings of $2\times2$ squares on the square lattice

This paper proves that random packings of $2\times2$ squares on the square lattice exhibit columnar order at high densities, characterized by four extremal Gibbs measures with broken rotational symmetry and distinct correlation decay rates, and establishes that all periodic Gibbs measures are mixtures of these four states.

The Gist

Imagine you have a giant, infinite floor made of square tiles. You also have a massive supply of heavy, 2x2 square furniture pieces (let's call them "couches"). You want to arrange these couches on the floor so that they don't overlap.

Now, imagine you are playing a game where you are very eager to fill the floor. The more couches you manage to fit in, the happier you are. In physics terms, this "eagerness" is called fugacity (or chemical potential). When this eagerness is low, the couches are scattered randomly, like people at a crowded party who haven't found their seats yet.

But what happens when you are desperate to fill the floor? What happens when the "fugacity" is huge?

This paper by Daniel Hadas and Ron Peled answers that question. They prove that when you try to pack these 2x2 couches as tightly as possible on a grid, the floor doesn't just get messy; it suddenly organizes itself into a very specific, rigid pattern. They call this Columnar Order.

Here is the breakdown of their discovery using simple analogies:

1. The "Sliding" Problem

Usually, when you pack things tightly, there's only one or two perfect ways to do it. Think of a jigsaw puzzle: there's usually only one spot for a specific piece.

However, with these 2x2 couches, there is a weird quirk. Imagine you have a perfect wall of couches. You could slide an entire vertical column of couches down by one square, and they would still fit perfectly without overlapping anything. You could do this for every column independently.

This creates a "Sliding Phenomenon." It's like having a bookshelf where you can slide every single shelf up or down by an inch without the books hitting each other. Because there are infinite ways to slide these columns, mathematicians used to think the system would remain chaotic and disordered even at high density. They thought, "With so many options, the system can't decide on a pattern."

2. The Surprise: The System Does Decide

The authors prove that nature hates this indecision. Even though there are infinite ways to slide the columns, when the pressure to pack is high enough, the system spontaneously picks one specific style of organization.

They found that the floor settles into one of four distinct "phases" (states of being):

• Vertical Columns: The couches align in vertical columns.
• Horizontal Rows: The couches align in horizontal rows.
• Two Variations of Each: Within the vertical columns, they can be shifted slightly to the left or right (like shifting a zipper).

It's as if you have a crowd of people who can stand anywhere, but suddenly, they all decide to stand in perfect vertical lines, shoulder-to-shoulder, and they all agree to face the same way. This is called Symmetry Breaking. The floor was originally perfectly round (symmetric), but the final pattern is a grid (less symmetric).

3. The "Sticks" Analogy

How did they prove this? They invented a way to look at the gaps between the couches.

Imagine the spaces between the couches are like sticks.

• If the couches are organized in vertical columns, you get long vertical sticks (gaps) running up and down.
• If they are organized in horizontal rows, you get long horizontal sticks.

The authors showed that in a high-density state, you will see a forest of long vertical sticks, or a forest of long horizontal sticks. You will almost never see a mix of both in the same large area. If you try to force a vertical stick to meet a horizontal stick, they create a "traffic jam" (a defect) that is energetically too expensive to maintain.

4. The "Chessboard" Trick

To prove this mathematically, they used a tool called the Chessboard Estimate.
Imagine the floor is a giant chessboard. The authors realized that if you reflect the pattern of couches across the lines of the board, the "energy" of the system behaves in a predictable way.

They extended this trick from small, finite boards to the infinite floor. This allowed them to calculate the probability of "defects" (places where the pattern breaks). They found that defects are so rare that the system is forced to stay in one of the four ordered patterns.

5. The "One-Dimensional" Secret

Here is the most beautiful part of their discovery.
When the system organizes into vertical columns, it acts almost like a collection of independent one-dimensional lines.

• Think of each vertical column as a separate line of people waiting in a queue.
• The people in Column A don't really care what the people in Column B are doing, as long as they don't bump into each other.

The system effectively breaks down into many independent 1D problems. This makes the math much easier and explains why the order is so stable.

Summary: What does this mean?

• The Result: At high density, 2x2 squares on a grid must organize into columns or rows. They cannot stay random.
• The Implication: This refutes an old idea that "sliding" prevents order. It shows that even with infinite sliding options, entropy (disorder) can actually drive order.
• Real World Connection: This is similar to how liquid crystals work. In a liquid crystal, molecules might be disordered like a gas, but under pressure, they align into columns or rows, creating the screens on your phone. This paper proves mathematically that this "columnar" phase is a real, stable state for these specific shapes.

In a nutshell: The authors proved that when you squeeze these 2x2 squares tight enough, they stop sliding around chaotically and snap into a rigid, organized grid of columns or rows, breaking the symmetry of the floor in the process.

Technical Summary

1. Problem Statement and Context

The paper investigates the 2×2 hard-square model on the square lattice $\mathbb{Z}^2$. In this model, tiles are $2 \times 2$ squares centered at lattice vertices. A configuration is valid if the interiors of the tiles are disjoint. The system is governed by a fugacity parameter $\lambda > 0$, where the probability of a configuration is proportional to $\lambda^N$ ($N$ being the number of tiles).

The Core Challenge:

• Low Fugacity: It is well-known that for small $\lambda$, the system is disordered with a unique Gibbs measure.
• High Fugacity: The behavior at high $\lambda$ is the focus. Unlike the nearest-neighbor hard-core model (which has a finite number of fully-packed ground states), the 2×2 hard-square model exhibits a "sliding phenomenon." There exists a continuum ($2^{\aleph_0}$) of fully-packed configurations obtained by sliding columns or rows of tiles.
• The Conjecture: Due to this sliding instability, it was previously conjectured (e.g., by Mazel–Stuhl–Suhov) that the system might have a unique Gibbs measure at high fugacity, as the sliding might prevent the system from "freezing" into a specific ordered phase.
• The Goal: The authors aim to rigorously prove whether the system undergoes a phase transition to an ordered state (specifically columnar order) at high fugacity, thereby refuting the uniqueness conjecture.

2. Methodology

The authors employ a sophisticated combination of statistical mechanics techniques, adapting and extending them to handle the sliding phenomenon.

A. Reflection Positivity and Chessboard Estimates

• Standard Approach: The paper utilizes reflection positivity, a property of the 2×2 hard-square model, to derive chessboard estimates.
• Novel Extension: A key methodological contribution is the extension of the chessboard estimate from finite-volume torus measures to all infinite-volume periodic Gibbs measures. This allows the authors to apply Peierls-type arguments directly in the infinite volume, bypassing the need to construct specific finite-volume approximations for every step.

B. The "Stick" Decomposition

To handle the continuum of ground states, the authors introduce a geometric classification of configurations:

• Sticks: They define "stick edges" as boundaries between tiles of different parities. A stick is a maximal path of these edges.
• Orientation: Sticks are either vertical or horizontal. Crucially, vertical and horizontal sticks cannot intersect.
• Columnar vs. Row Order: In a columnar phase, the system is dominated by long vertical sticks (separating columns of tiles). In a row phase, it is dominated by horizontal sticks.
• Properly Divided Rectangles: They define a rectangle as "properly divided" if a stick of a specific orientation crosses both the rectangle and a slightly smaller concentric rectangle.

C. Peierls-Type Argument with Mesoscopic Scales

• The authors define a mesoscopic length scale $b(\lambda) \sim \lambda^{1/4}$ to $\lambda^{1/2}$.
• They prove that in any periodic Gibbs measure, mesoscopic rectangles are almost surely divided by sticks of a specific orientation.
• By showing that "interfaces" (regions where sticks are missing or mixed) are exponentially rare, they use a Peierls argument to demonstrate that the system must choose between a vertical-ordered phase or a horizontal-ordered phase, but cannot remain disordered.

D. Disagreement Percolation

To characterize the resulting measures and prove extremality, the authors use disagreement percolation (van den Berg–Steif method).

• They consider two independent samples from a Gibbs measure.
• They prove that the set of vertices where the configurations disagree does not percolate (i.e., no infinite disagreement paths exist).
• This implies that the measure is extremal (pure) and allows for the characterization of correlation decay.

3. Key Contributions and Results

A. Existence of Multiple Gibbs Measures (Theorem 1.1)

The paper proves that for sufficiently large $\lambda$, the model admits multiple extremal Gibbs measures. Specifically, there are at least four distinct periodic measures:

1. $\mu_{(ver, 0)}$: Vertical columnar order with even horizontal parity.
2. $\mu_{(ver, 1)}$: Vertical columnar order with odd horizontal parity.
3. $\mu_{(hor, 0)}$: Horizontal row order with even vertical parity.
4. $\mu_{(hor, 1)}$: Horizontal row order with odd vertical parity.

Key Properties of these Measures:

• Symmetry Breaking: The measures break the lattice's $90^\circ$ rotational symmetry (choosing either vertical or horizontal) and translational symmetry along one axis (periodicity of $2\mathbb{Z}$ in one direction), while preserving translational symmetry in the orthogonal direction.
• Structure: The measures are not simple perturbations of a single fully-packed configuration. Instead, they are perturbations of a "union of one-dimensional systems." In the vertical phase, columns are independent 1D hard-core systems, separated by rare vacancies.

B. Characterization of Periodic Measures (Theorem 1.2)

The authors prove that every periodic Gibbs measure is a convex combination of these four extremal measures. This establishes that the four phases exhaust all possible ways the system can order at high fugacity.

C. Anisotropic Correlation Decay

The paper quantifies the decay of correlations, revealing strong anisotropy:

• Broken Symmetry Direction (e.g., x-axis for vertical order): Correlations decay exponentially fast with a correlation length of order $O(1)$.
• Preserved Symmetry Direction (e.g., y-axis for vertical order): Correlations decay exponentially but much slower, with a correlation length of order $O(\sqrt{\lambda})$.
• This reflects the physical intuition that columns are independent (fast decay across columns) but highly correlated along the column (slow decay along the column).

D. Technical Innovation: Infinite-Volume Chessboard Estimate

The paper provides a rigorous proof that the chessboard estimate holds for infinite-volume periodic Gibbs measures. This is a significant technical advancement, as standard chessboard estimates are typically formulated for finite tori. This extension is crucial for handling the sliding phenomenon where ground states are not isolated.

4. Significance

1. Resolution of a Long-Standing Conjecture: The work definitively refutes the conjecture that sliding phenomena lead to a unique Gibbs measure at high fugacity. It proves that even with a continuum of ground states, the system undergoes a phase transition to an ordered state.
2. Rigorous Proof of Columnar Order: It provides the first mathematically rigorous proof of columnar order in a hard-core lattice model. While columnar order is a standard concept in liquid crystal physics (and predicted for this model), rigorous proofs were previously lacking for models with sliding.
3. Methodological Breakthrough: The extension of the chessboard estimate to infinite-volume periodic measures opens the door for analyzing other complex lattice models with sliding instabilities or continuous ground state manifolds.
4. Insight into Liquid Crystals: The results provide a lattice-based mathematical foundation for understanding how orientational symmetry breaking (nematic/columnar phases) can emerge from purely repulsive (hard-core) interactions, bridging the gap between statistical mechanics and liquid crystal theory.

5. Conclusion

Hadas and Peled demonstrate that the 2×2 hard-square model exhibits a rich phase diagram at high fugacity. Despite the "sliding phenomenon" which suggests a lack of rigidity, entropic effects drive the system into one of four distinct columnar or row-ordered phases. The proof relies on a novel synthesis of reflection positivity, geometric stick decomposition, and disagreement percolation, establishing a new standard for analyzing hard-core models with complex ground state structures.