Abstract
Macroscopic theories of nucleation such as classical nucleation theory envision that clusters of the bulk stable phase form inside the bulk metastable phase. Molecular dynamics simulations are often used to elucidate nucleation mechanisms, by capturing the microscopic configurations of all the atoms. In this paper, we introduce a thermodynamic model that links macroscopic theories and atomic-scale simulations and thus provide a simple and elegant framework for testing the limits of classical nucleation theory.
The Conceptual Gap in Nucleation Modeling
Nucleation—the process where a new phase begins to form from a metastable one—is a fundamental phenomenon in physics and materials science. For decades, researchers have relied on Classical Nucleation Theory (CNT), which treats nuclei as macroscopic droplets with sharp interfaces.
However, at the atomic scale, Molecular Dynamics (MD) simulations reveal a more complex picture. Instead of a single, well-defined "droplet," simulations show a sea of short-lived fluctuations alongside potentially one larger growing cluster.
To map these simulations back to theory, researchers typically employ heuristic rules to classify atoms as "solid" or "liquid." These rules are often arbitrary and can introduce significant bias into calculated nucleation barriers. This paper addresses this "conceptual gap" by establishing a mathematically rigorous link between fluctuating atomic configurations and macroscopic thermodynamic definitions.
Methodology: A Probabilistic Approach
The authors propose a probabilistic definition of Gibbs dividing surfaces. Rather than labeling individual atoms, they utilize extensive order parameters ($Φ$)—global quantities representing the state of the entire system, such as total magnetization or integrated structural descriptors.
Key elements of the framework include:
- Idealized Multiple-Cluster Model: In a metastable system, the total population of the new phase is modeled as the sum of many small clusters following a Poisson distribution.
- Fluctuating Reference Framework: The model accounts for the inherent fluctuations of pure bulk phases, defining the probability distribution of the order parameter for a reference system containing a cluster of size $n$.
- Self-Consistent Iteration: A central thermodynamic equation relates the total system free energy (measurable in simulation) to the free energy of a single cluster, $G(n)$. This is solved iteratively to converge on the true nucleation profile.
This "top-down" approach uses global properties to deduce the characteristics of the largest cluster, bypassing the need for arbitrary bottom-up clustering algorithms.
Validation and Results
The framework was validated using two benchmark physical models: a Lennard-Jones system (23,328 atoms) for solidification and a 2D Square-Lattice Ising Model for magnetic phase transitions.
A primary insight is the "One Big Cluster" Rule: once the total amount of the new phase exceeds a threshold ($n_{cut}$), it becomes statistically certain that the system contains exactly one large cluster plus a background of sub-critical fluctuations.
Key results include:
- Perfect Reconstruction: For the Lennard-Jones system, the reconstructed free energy profile $G(n)$ perfectly matched the total free energy observed in simulations.
- Analytical Agreement: In the 2D Ising model, the method produced a free energy curve matching exact analytical predictions without any fitting parameters.
- Robustness: The method proved effective across diverse transitions, suggesting its general utility in phase transition research.
Discussion and Future Work
By using global extensive quantities, this method removes the ambiguity of choosing cutoffs for clustering. While the current derivation assumes a regime dominated by a single large cluster—typical for standard nucleation simulations—future work may extend this to systems near the spinodal limit or to heterogeneous nucleation involving surfaces or impurities. This framework provides a robust tool for systematically testing the physical boundaries where Classical Nucleation Theory begins to fail.
Summary updated