🔬 condensed matter

Modified Family-Vicsek Scaling and Probability Distributions for Brownian Castle Interfaces

This paper numerically verifies that the Brownian Castle interface growth model belongs to a new universality class characterized by modified Family-Vicsek scaling with specific growth and roughness exponents, alongside non-Gaussian height distributions and Cauchy-Lorentz distributed height changes.

Original authors: Noah Sublett, Charles Sutton, Benjamin Long, Daniel B. Dougherty

Published 2026-07-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Noah Sublett, Charles Sutton, Benjamin Long, Daniel B. Dougherty

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 the world is full of surfaces that are never perfectly smooth. Think of a sand dune shifting in the wind, the edge of a melting ice cube, or even the jagged border of a growing bacterial colony. In physics, there is a fascinating idea called "universality." It suggests that even though these surfaces are made of different stuff and grow in different ways, they often follow the same hidden mathematical rules. Scientists call these rules "scaling laws." They act like a secret code that describes how rough a surface gets over time and how that roughness changes if you look at a bigger or smaller piece of the surface. For decades, scientists have known a few of these codes, like the famous "Kardar-Parisi-Zhang" (KPZ) class, which describes how things grow when particles just pile up and stick. But what happens if the rules change? What if, instead of just sticking, particles can sometimes vanish or jump around wildly? That is the question this paper tackles, exploring a new, weird, and wonderful way that surfaces can grow.

The researchers in this study are investigating a new model they call the "Brownian Castle." To understand it, picture a game of digital Tetris. In the standard version, called "Ballistic Deposition," blocks fall from the sky and stick to the first block they hit, creating jagged towers and overhangs. This is a well-known way surfaces grow. But the Brownian Castle changes the rules of the game. In this version, when a block falls, it doesn't just stick where it lands. Instead, it has a chance to randomly choose to grow taller, stay the same, or even shrink! It's as if the blocks are playing a game of chance where they can sometimes disappear or jump to a different height. The scientists wanted to see what kind of surface this chaotic, "infinite temperature" game would create.

Using powerful computer simulations, the team built these digital castles on strips of land ranging from 16 to 512 blocks wide. They watched how the "height" of the castle walls changed over millions of steps. What they found was that these castles do follow a pattern, but it's a modified one. They measured how fast the roughness of the castle grew and found a growth rate of 0.472 ± 0.012. They also measured how rough the castle gets when it stops changing and found a "roughness exponent" of 1.01 ± 0.018. These numbers are slightly different from the standard rules, suggesting that the Brownian Castle belongs to a brand-new "universality class"—a new category of growth that no one had fully mapped out before.

However, the story gets even more interesting when they looked at the details. In the standard models, the math usually works out perfectly if you zoom in or out. But for these castles, the math gets a bit wobbly at small sizes. The researchers discovered that the way the castle grows depends on the size of the strip in a tricky way, requiring a new "size scaling exponent" of 0.50 ± 0.05 to make the numbers line up. It's like the castle's growth speed changes depending on how big the playground is, a feature that hints at a hidden "memory" in the system, similar to other complex growth models.

The team also looked at the shape of the castle's height distribution. In many natural systems, heights follow a "bell curve" (Gaussian distribution), where most heights are average and extreme heights are rare. But the Brownian Castle is rebellious. The heights don't follow a nice bell curve; instead, they have "heavy tails," meaning extreme jumps up and down happen much more often than you'd expect. When they looked at the changes in height from one moment to the next, the pattern didn't look like a bell curve at all. It looked like a "Cauchy-Lorentz" distribution. This is a shape known for having very long, heavy tails, which fits perfectly with a system where giant, random jumps are a regular feature.

So, where do we see these Brownian Castles in the real world? The authors suggest they probably aren't found in simple thermal systems like melting ice, because those systems don't usually allow for such dramatic, random removal of material. Instead, they suspect these patterns might appear in living systems far from equilibrium, like the edges of bacterial colonies or fungal networks. In these living systems, cells are constantly being born and dying based on local conditions, creating a chaotic mix of growth and shrinkage that mimics the Brownian Castle rules. While the researchers haven't found a real-world castle yet, their simulations show that this new model creates a unique, jump-filled landscape that challenges our old ideas about how rough surfaces form. They suggest that if we look closely at the edges of living colonies, we might just find these digital castles hiding in nature.

Problem Statement
The paper addresses the characterization of the "Brownian Castle" (BC), a recently discovered interface growth model introduced by Cannizzaro and Hairer [2]. The BC model is a variation of the standard Ballistic Deposition (BD) model, designed to simulate interfaces where material removal is possible alongside deposition. While Cannizzaro and Hairer identified that BC interfaces belong to a new universality class with specific scaling exponents ( $\beta=1/2, \alpha=1$ ), this report aims to provide numerical verification of these properties, investigate finite-size scaling corrections, and analyze the statistical distributions of interface heights and height changes. The goal is to bring this new model to the attention of the statistical physics community and determine if its behavior aligns with established scaling theories or requires modifications.

Methodology
The authors performed numerical simulations of BC interfaces in one dimension (1+1 dimensions) using personal computers. The simulations utilized the following parameters and techniques:

System Sizes: Interfaces with lengths $L$ ranging from 16 to 512 sites.
Evolution: Simulations started from flat initial conditions and evolved for $10^6$ time steps.
Averaging: To improve signal-to-noise ratios, interface widths were averaged over 20 independent evolutions, a process repeated four times.
Probability Distributions:
- Heights: An $L=128$ system was evolved for $10^6$ steps across $10^6$ independent trials to generate $1.28 \times 10^8$ height data points in the saturation regime.
- Height Changes: Height differences ($dh$) were recorded at every time step for every site where a change occurred during a $10^6$ step evolution on an $L=128$ interface.
Analysis: Growth and roughness exponents were extracted via linear fits to double-logarithmic plots of interface width versus time and system size, respectively. Probability distributions were analyzed using Anderson-Darling tests and curve fitting.

Key Results

Scaling Exponents: The study confirms the existence of a new universality class for BC interfaces.
- The growth exponent was measured as $\beta = 0.472 \pm 0.012$ , consistent with the theoretical prediction of $1/2$ but slightly lower due to finite-size effects.
- The roughness exponent was measured as $\alpha = 1.01 \pm 0.018$ , consistent with the predicted value of $1$.
- The interface width exhibits modified Family-Vicsek scaling. While data collapse is observed, the standard dynamic exponent relationship $z = \alpha/\beta$ does not hold directly for the observed data collapse.
Modified Scaling and Finite-Size Effects:
- The authors observed that the standard scaling law requires modification at short times due to a size scaling exponent $\gamma$ .
- By analyzing the early-time growth regime, they extracted a size scaling exponent $\gamma = 0.50 \pm 0.05$ .
- This leads to an effective dynamic exponent $z = (\alpha + \gamma)/\beta = 3$ , which successfully describes the data collapse observed in the simulations. This suggests the BC model shares features with BD models that include memory effects, where the probability of material loss creates an effectively reduced propensity for deposition.
Probability Distributions:
- Height Distribution: The probability density of normalized interface heights, $p(\tilde{h})$ , shows significant deviations from a Gaussian distribution. An Anderson-Darling test rejects the Gaussian hypothesis at the 99% confidence level. The tails exhibit richness similar to the Tracy-Widom distribution observed in standard BD (KPZ class), though the specific functional form for BC tails could not be precisely determined.
- Height Change Distribution: The distribution of local height changes, $p(dh)$, follows a Cauchy-Lorentz form near the center. This heavy-tailed distribution is consistent with a process involving large jumps (Levy flights), reflecting the model's allowance for material removal and large height fluctuations.

Significance and Claims
The paper claims to provide the first detailed numerical verification of the Brownian Castle universality class, confirming the exponents reported by Cannizzaro and Hairer while highlighting the necessity of modified Family-Vicsek scaling to account for finite-size effects. The identification of the size scaling exponent $\gamma$ and the resulting effective dynamic exponent $z=3$ offers new insight into the model's connection to other growth models with memory.

Furthermore, the characterization of the non-Gaussian height distribution and the Cauchy-Lorentz form of height changes suggests that the BC model cannot be easily described by standard stochastic partial differential equations (SPDEs) derived via Kramers-Moyal expansions, which typically assume small, continuous jumps. The authors suggest that the BC universality class may be relevant to natural systems far from equilibrium, such as step fluctuations on solid surfaces at high temperatures (e.g., Si(100) B-type steps) or the growth of biological networks (bacterial or fungal colonies) where spatially correlated birth and death events occur. The paper concludes that Brownian Castles represent a distinct and important addition to the landscape of interface growth models, warranting further exploration of their continuum limits and potential real-world applications.

More like this