🔬 materials science

First-principles calculations of the thermodynamic factor of multicomponent alloys

This paper introduces two efficient computational routes to calculate the thermodynamic factor for non-dilute diffusion in multicomponent alloys by accounting for vacancy-mediated free energy curvature, thereby enabling accurate and resource-efficient predictions of diffusion matrices for use in mesoscale mass transport simulations.

Original authors: Damien K. J. Lee, Shashank Saxena, Anton Van der Ven, Anirudh Raju Natarajan

Published 2026-09-28

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Damien K. J. Lee, Shashank Saxena, Anton Van der Ven, Anirudh Raju Natarajan

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

In the heart of every metal, atoms are not frozen in place but are constantly swapping positions, a slow, silent shuffle that allows materials to change shape, heal cracks, or degrade over time. This movement, known as diffusion, is the engine behind the aging of bridges, the hardening of jet engines, and the reliability of the electronics in our pockets. In pure metals, this process is straightforward, but in the complex alloys used for modern technology—mixtures containing several different types of metal atoms—the rules become murky. These mixtures, often called high-entropy alloys, are prized for their strength and resistance to heat, yet scientists struggle to predict how fast atoms will move through them. The difficulty lies in a subtle but critical detail: atoms in these solids do not move through empty space. Instead, they hop into tiny, empty spots called vacancies. While these empty spots are rare, appearing in only a few out of every million lattice sites, they are the only way the material can flow. To understand how the alloy behaves, researchers must calculate how the energy of the system changes as the composition shifts, a measurement known as the thermodynamic factor. However, because the empty spots are so scarce, trying to count them directly in a computer simulation is like trying to find a specific grain of sand on a beach by scanning the entire shore one grain at a time; the process is so slow and prone to error that it often fails to give a clear answer.

A team of researchers has now developed a new way to solve this problem, allowing them to predict how atoms move in these complex mixtures without getting lost in the noise of the rare empty spots. Working with a specific five-element alloy made of hafnium, molybdenum, niobium, titanium, and zirconium, the scientists realized they did not need to simulate the entire chaotic dance of atoms and empty spots together. Instead, they found they could separate the problem into two distinct parts. First, they calculated how the energy of the material changes when the atoms are packed tightly together with no empty spots at all. Second, they simply measured how many empty spots exist in the mixture. By combining these two pieces of information, they could reconstruct the full picture of how the material diffuses. This approach is like knowing the weight of a fully loaded truck and the exact number of missing passengers to determine the total load, rather than trying to weigh every single person and every empty seat individually.

The researchers tested two different methods to achieve this separation. The first method, which they call an approximation, uses the energy curvature of the perfect, vacancy-free alloy and multiplies it by the concentration of the empty spots. The second method is more rigorous, treating the empty spots as a single, specific entity within the mathematical framework to recover the exact behavior of the real material. When they compared these new methods against traditional computer simulations that tried to track every single atom and every empty spot, the results were striking. The traditional simulations, even when run for a very long time, struggled to converge on a stable answer because the empty spots were so rare. In contrast, the new methods produced accurate results with roughly one-tenth of the computational effort. The new approach correctly predicted the speed at which the material would diffuse, matching the performance of the much slower, fully converged simulations.

The study also revealed a deeper truth about how these materials move. The researchers found that the fastest way the material can change is governed entirely by how the empty spots themselves move, a rate that is independent of the complex interactions between the different metal atoms. The slower movements, which involve the different atoms swapping places with one another, are directly proportional to how many empty spots are available. This means that the most important speed limit for the material is set by the scarcity of the empty spots, while the specific chemistry of the alloy only fine-tunes the slower, secondary movements. By proving that the complex interactions of the metal atoms can be calculated separately from the rare empty spots, the researchers have provided a reliable shortcut. This allows scientists to take detailed atomic models of new alloys and immediately use them to simulate how the material will behave over years or decades, without needing to run simulations that would take an impractical amount of time to complete. The work confirms that for these complex metals, the key to understanding their long-term stability lies not in tracking every single defect, but in understanding the simple, underlying geometry of how atoms fill the space around them.

Technical Summary: First-principles calculations of the thermodynamic factor of multicomponent alloys

Problem Statement
Accurately predicting mass transport in multicomponent alloys, particularly Multi-Principal Element Alloys (MPEAs), requires the calculation of non-dilute diffusion coefficients ( $D$ ). These coefficients are the product of Onsager transport coefficients ( $\tilde{L}$ ) and the thermodynamic factor ( $\tilde{\Theta}$ ). While $\tilde{L}$ can be derived from kinetic Monte Carlo (kMC) simulations, computing $\tilde{\Theta}$ presents a significant bottleneck. $\tilde{\Theta}$ represents the curvature of the free energy with respect to composition. In metallic alloys, diffusion is vacancy-mediated, necessitating that $\tilde{\Theta}$ be evaluated in the dilute-vacancy limit. However, the low concentration of vacancies renders the covariance matrix of composition fluctuations nearly singular. Consequently, obtaining accurate $\tilde{\Theta}$ via standard semi-grand canonical Monte Carlo (GCMC) simulations requires prohibitively long sampling times to converge, as composition fluctuations involving vacancies occur rarely. Furthermore, standard CALPHAD assessments typically treat alloys as vacancy-free, failing to capture the specific curvature required for the dilute-vacancy limit.

Methodology
The authors develop two distinct computational routes to determine the thermodynamic factor for multicomponent alloys in the dilute-vacancy limit, avoiding the need to sample vacancy composition fluctuations directly in a fully converged manner.

Coordinate Transformation and Approximation:
The first route utilizes a linear coordinate transformation of the composition space. By transforming the elemental composition axes ( $x$ ) into a new basis ( $y$ ) where one coordinate tracks the vacancy concentration and the remaining coordinates track intermixing parallel to the vacancy-free hyperplane, the thermodynamic factor matrix ( $\tilde{\Theta}$ ) is decoupled.
- In this transformed basis, the matrix separates into a block dominated by the divergent term $1/x_{Va}$ (vacancy thermodynamics) and a block representing the curvature of the vacancy-free alloy ( $\Theta_{mix}$ ).
- The authors demonstrate that the off-diagonal coupling terms and specific vacancy-dependent derivatives do not significantly affect the eigenspectrum of the diffusion matrix in the dilute limit.
- This allows for an approximation where $\tilde{\Theta}$ is constructed using only the equilibrium vacancy concentration and the thermodynamic factor of the vacancy-free alloy (obtained from standard GCMC simulations without vacancies).
Single-Vacancy Expansion:
The second route provides a numerically exact benchmark by truncating the semi-grand canonical partition function at a single vacancy.
- Instead of sampling the full ensemble of vacancy concentrations, this method samples vacancy-free configurations and explicitly calculates the energy cost of introducing a single vacancy at every lattice site.
- The covariance matrix is reconstructed using a weighted sum of these single-vacancy configurations relative to the vacancy-free ensemble.
- This approach yields a non-singular covariance matrix that can be inverted to obtain the full thermodynamic factor of the vacancy-containing alloy without the slow convergence issues associated with direct vacancy sampling.

Key Results
The methods were validated on the Hf–Mo–Nb–Ti–Zr (Senkov) alloy system, ranging from binary to quinary compositions at 2000 K.

Convergence and Efficiency: Direct GCMC simulations of the vacancy-containing system required roughly an order of magnitude more Monte Carlo steps to converge covariance values compared to the vacancy-free system. The single-vacancy expansion achieved accuracy comparable to fully converged GCMC simulations with approximately one-tenth of the sampling effort.
Diffusion Matrix Eigenspectrum: Both routes successfully reproduced the eigenvalues and eigenvectors of the diffusion matrix ( $D$ $D$ ) obtained from fully converged simulations.
- The largest eigenvalue of $D$ corresponds to the vacancy tracer diffusion coefficient ( $D^*_{Va}$ ) and is independent of the specific approximation used, provided the vacancy concentration is known.
- The remaining eigenvalues, which govern interdiffusion, scale with the vacancy concentration.
Validation of Approximation: The approximate method (Route 1), which relies solely on the vacancy-free alloy curvature and vacancy concentration, reproduced the eigenspectrum of the diffusion matrix with high fidelity. The cosine distance between eigenvectors calculated via the approximation and the single-vacancy expansion was near zero across all tested systems.
Scaling Relation: The authors derived a scaling relation that extends these results to local vacancy chemical potentials away from equilibrium, allowing the methods to be applied in mesoscale simulations without additional atomistic calculations.

Significance and Claims
The paper claims to resolve the computational bottleneck associated with calculating thermodynamic factors for non-dilute diffusion in multicomponent alloys. By demonstrating that the complex thermodynamics of vacancy-containing alloys can be effectively separated into vacancy concentration and vacancy-free alloy curvature, the authors provide a pathway to integrate atomistic models (such as Cluster Expansions) and CALPHAD assessments directly into mesoscale mass transport simulations.

The work establishes that:

The single-vacancy expansion is a robust, efficient benchmark that recovers the full thermodynamic factor without the sampling cost of direct vacancy simulation.
A simplified approximation, using only the vacancy-free alloy's free energy curvature and the equilibrium vacancy concentration, is sufficient to accurately predict the diffusion matrix's eigenspectrum.
These methods enable the direct use of first-principles-derived thermodynamic descriptions in mesoscale simulations, facilitating the prediction of microstructural evolution and high-temperature performance in complex MPEAs.

More like this