Beyond dynamic scaling: rare events break universality

This paper demonstrates that surface growth driven by non-monomeric deposition with a power-law size distribution ($P(s)\sim s^{-\tau}$) breaks standard scale invariance and Family–Vicsek universality for $\tau<3$ due to the emergence of a second dynamical length scale associated with rare large clusters, causing critical exponents to vary continuously with $\tau$.

The Gist

The Big Picture: Building a City vs. Piling Up Sand

Imagine you are watching a construction site. Usually, scientists think about how cities or sand dunes grow by imagining tiny workers dropping one grain of sand at a time. If you drop enough grains, the pile gets rough, but it follows a predictable, smooth pattern. In physics, this is called the KPZ universality class. It's like a "law of nature" that says, "No matter how you build it, if you use tiny bits, the roughness will look the same."

This paper asks a simple question: What happens if, instead of tiny grains, we drop giant boulders or entire houses onto the pile? And what if the size of these objects follows a weird rule where tiny rocks are common, but occasionally, a massive mountain appears?

The authors found that when you introduce these "rare, giant events," the old laws of physics break down. The system stops behaving predictably, and a new, chaotic kind of growth emerges.

---

The Experiment: The "Eden" Game

To test this, the researchers created a computer simulation (a digital sandbox).

1. The Setup: They started with a flat line (like a calm sea).
2. The Rule: They dropped "blobs" (clusters of particles) onto the line.
3. The Twist: The size of these blobs wasn't random in a normal way. They followed a Power Law.
   • Normal Law: Most things are average size.
   • Power Law: You get lots of tiny pebbles, but every now and then, you get a massive boulder.
   • The researchers changed a knob called $\tau$ (tau).
     • High $\tau$ (e.g., 3.5): The boulders aren't too big. The system behaves normally.
     • Low $\tau$ (e.g., 2.5): The boulders can get incredibly huge. This is where things get weird.

The Discovery: Two Rulers, One Mess

In normal growth, scientists use one ruler to measure how the surface gets rough. They call this the "correlation length." It's like measuring how far a ripple travels across a pond.

But in this paper, they found a second ruler.

When the blobs are huge and rare (low $\tau$), the growth is no longer driven by the average ripple. Instead, it's driven by the single biggest blob that has landed so far.

• Analogy: Imagine you are filling a bathtub with water.
  • Normal Growth: You turn on the tap. The water level rises smoothly and evenly. You can predict exactly how fast it will fill.
  • This Paper's Growth: You have a hose, but every now and then, someone dumps a fire hydrant's worth of water into the tub.
  • The water level doesn't rise smoothly. It jumps. The "roughness" of the water surface is now determined by the last giant splash, not the steady flow of the hose.

Because there are two things fighting for control (the steady flow vs. the giant splashes), the old math (called Family-Vicsek scaling) fails. The "rulers" don't match up anymore.

The Results: When Does the Law Break?

The paper draws a clear line in the sand:

1. The Safe Zone ($\tau \ge 3$):

   • The giant blobs are rare enough that they don't dominate. The "average" behavior wins.
   • Result: The system behaves exactly like the standard KPZ model. It's boring, but predictable. It's like a city growing with standard bricks.

2. The Chaos Zone ($\tau < 3$):

   • The giant blobs are so huge and frequent that they break the rules.
   • Result: The "universality" (the idea that all rough surfaces look the same) breaks down.
   • The roughness of the surface changes continuously depending on how big the blobs are.
   • The "growth exponent" (how fast it gets rough) isn't a fixed number; it keeps shifting as the simulation runs.

Why This Matters in the Real World

You might think, "Who cares about computer blobs?" But this happens in real life!

• Cities: Cities don't grow by adding one brick at a time. They grow by adding whole neighborhoods, skyscrapers, or highways. If a city expands by dropping a massive new district on the edge, the "roughness" of the city's growth isn't smooth.
• Nature: Think of landslides, avalanches, or how bacteria colonies expand. Sometimes a single massive event changes the whole shape of the frontier.
• The Lesson: If you try to predict how a system grows using standard math, and you ignore the "rare, giant events," your prediction will be wrong. The "outliers" aren't just noise; they are the main drivers of the system.

Summary in One Sentence

This paper shows that when a growing surface is built by dropping objects of wildly different sizes (especially when massive ones appear), the usual laws of physics break, and the system becomes ruled by the single biggest "accident" rather than the average behavior.

Technical Summary

1. Problem Statement

Surface growth is a fundamental topic in non-equilibrium statistical physics, traditionally described by the Kardar–Parisi–Zhang (KPZ) universality class. The standard paradigm assumes that growth occurs via the aggregation of monomers (single particles) or small, uncorrelated clusters, leading to scale-invariant dynamics governed by a single growing correlation length $\xi(t)$.

However, many physical systems (e.g., aerosol deposition, city expansion, liquid invasion in porous media) involve the deposition of extended clusters with heavy-tailed size distributions. The central question addressed by this paper is: How does the deposition of clusters with a power-law size distribution $P(s) \sim s^{-\tau}$ affect the scaling behavior and universality of surface growth? Specifically, the authors investigate whether the standard Family–Vicsek (FV) dynamic scaling ansatz holds when the cluster size distribution lacks a characteristic scale (i.e., when the variance diverges).

2. Methodology

The authors employ a computational model based on random deposition of rigid Eden clusters on a one-dimensional substrate of size $L$ with periodic boundary conditions.

• Model Dynamics:
  • Clusters are generated as compact, stochastic aggregates (Eden clusters).
  • At each time step, a cluster of size $s$ is chosen from a power-law distribution $P(s) \sim s^{-\tau}$ (where $1 \le s \le L$).
  • The cluster is deposited vertically at a random lateral position. It adheres irreversibly upon first contact with the existing surface or a neighbor (ballistic deposition with no surface relaxation).
  • Time is defined as $t = N/L$, where $N$ is the total number of deposited blobs.
• Parameters:
  • The exponent $\tau$ is varied in the range $2 \le \tau \le 3.5$.
  • System sizes $L$ range from $10^3$ to $10^5$.
  • Simulations involve thousands of realizations to ensure statistical robustness.
• Observables:
  • Roughness ($W$): Defined as the standard deviation of the height profile.
  • Exponents: The roughness exponent $\alpha$ (saturation regime), dynamic exponent $z$ (crossover time), and growth exponent $\beta = \alpha/z$.
  • Scaling Analysis: The authors test the Family–Vicsek ansatz $W(L,t) = L^\alpha f(t/L^z)$ via data collapse and analyze the effective growth exponent $\beta_e(t) = d \ln W / d \ln t$.

3. Key Contributions

The paper makes three primary theoretical and numerical contributions:

1. Identification of a New Universality Regime: It demonstrates that for $\tau < 3$, the system departs from the KPZ universality class, exhibiting continuously varying critical exponents dependent on $\tau$.
2. Breakdown of Family–Vicsek Scaling: The authors show that the standard dynamic scaling ansatz fails for heavy-tailed distributions because the system is no longer governed by a single length scale.
3. Mechanism of Rare Events: The paper establishes that the breakdown of universality is driven by extreme value statistics. The largest deposited clusters introduce a second, competing dynamical length scale $\zeta(t)$, which dominates the interface roughness when the cluster size variance diverges.

4. Results

A. Phase Transition at $\tau = 3$

The behavior of the system undergoes a sharp transition at $\tau = 3$:

• For $\tau \ge 3$: The second moment of the cluster size distribution is finite. The system recovers the standard KPZ universality class with exponents $\alpha \approx 0.5$, $z \approx 1.5$, and $\beta \approx 1/3$. The Galilean invariance relation $\alpha + z = 2$ holds.
• For $\tau < 3$: The variance of the cluster size diverges. The system enters a non-universal regime where:
  • The roughness exponent $\alpha$ decreases continuously from $\approx 0.66$ (at $\tau=2$) to $0.5$ (at $\tau=3$).
  • The dynamic exponent $z$ increases continuously from $\approx 1.1$ to $1.5$.
  • The Galilean invariance relation $\alpha + z = 2$ is violated.

B. Failure of Dynamic Scaling

For $\tau < 3$, the standard Family–Vicsek scaling collapse fails:

• When plotting $W/L^\alpha$ vs. $t/L^z$, curves for different system sizes do not collapse onto a single master curve.
• The effective growth exponent $\beta_e(t)$ is not constant. Instead, it exhibits a slow, system-size-dependent decay from an initial value down to a crossover point before saturation.
• This indicates that the roughness does not follow a simple power law $W \sim t^\beta$ during the growth phase.

C. The Two-Length Scale Mechanism

The authors propose a physical mechanism explaining the breakdown of scaling:

1. Correlation Length ($\xi$): The standard scale $\xi(t) \sim t^{1/z}$ representing the propagation of local correlations.
2. Extremal Scale ($\zeta$): A new scale introduced by the largest cluster deposited up to time $t$. Based on extreme value statistics, for $\tau < 3$, the maximum cluster size scales as $s_{max} \sim N^{1/(\tau-1)}$. Consequently, the vertical height contribution of the largest blob scales as:
   $$\zeta(t) \sim (Lt)^{1/[2(\tau-1)]}$$
3. Competition:
   • When $\xi(t) \gg \zeta(t)$, the system behaves like standard KPZ (microscopic depositions dominate).
   • When $\zeta(t) \gtrsim \xi(t)$ (which occurs for extended times in the $\tau < 3$ regime), the interface roughness is dominated by rare, massive events (the largest blobs).
   • The general scaling form becomes $W(L,t) = L^\alpha F(L/\xi, \xi/\zeta)$, involving two variables, which prevents the reduction to the single-variable FV ansatz.

5. Significance

• Theoretical Impact: This work challenges the assumption that surface growth is universally governed by a single correlation length. It proves that rare events (heavy tails in input distributions) can fundamentally alter the universality class of a system, leading to a new phenomenology where scale invariance is broken.
• Experimental Relevance: The findings provide a theoretical framework for interpreting experimental observations in systems like turbulent liquid crystals, bacterial colonies, and city growth, where anomalous roughness exponents have been observed but not fully explained by standard KPZ theory.
• Refutation of Previous Predictions: The results contradict earlier theoretical predictions (e.g., by Krug) that attempted to map power-law noise exponents directly to roughness exponents in low dimensions, showing that the discrete nature of cluster deposition leads to different scaling behaviors than continuum noise models.

In conclusion, the paper establishes that universality in surface growth is fragile and can be broken by the presence of heavy-tailed size distributions, necessitating a new theoretical framework that accounts for the competition between correlation propagation and extreme value statistics.