About this Abstract |
| Meeting |
MS&T26: Materials Science & Technology
|
| Symposium
|
Progress in High Entropy Materials: Integrating Experiments, Computation, and Machine Learning
|
| Presentation Title |
Entropy, Zentropy and ZENN |
| Author(s) |
Zi-Kui Liu |
| On-Site Speaker (Planned) |
Zi-Kui Liu |
| Abstract Scope |
Entropy governs the evolution of systems through interactions with their surroundings. Classical statistical mechanics defines entropy based on all configurations embraced by the system, yielding the Gibbs entropy. Zentropy theory expands this perspective by defining the total system entropy as the sum of the Gibbs entropy arising from the probability distribution over configurations and the intrinsic entropy associated with each configuration(1). When all configurations can be characterized through density functional theory, or other accurate computational methods, zentropy provides a predictive framework for emergent behavior without empirical models and fitting parameters. For complex systems where such simulations are intractable, a zentropy-enhanced neural network (ZENN) AI framework has been developed to learn the key configurations along with their internal energies and intrinsic entropies(2) based on the recursive feature of entropy(3).
1 J. Phase Equilibria Diffus. 43, 598 (2022)
2 Proc. Natl. Acad. Sci. 123, e2511227122 (2026)
3 http://arxiv.org/abs/2511.04950 (2025) |
