Identifying structural changes with unsupervised machine learning methods

Unsupervised machine learning methods are used to identify structural changes using the melting point transition in classical molecular dynamics simulations as an example application of the approach. Dimensionality reduction and clustering methods are applied to instantaneous radial distributions of atomic configurations from classical molecular dynamics simulations of metallic systems over a large temperature range. Principal component analysis is used to dramatically reduce the dimensionality of the feature space across the samples using an orthogonal linear transformation that preserves the statistical variance of the data under the condition that the new feature space is linearly independent. From there, k-means clustering is used to partition the samples into solid and liquid phases through a criterion motivated by the geometry of the reduced feature space of the samples, allowing for an estimation of the melting point transition. This pattern criterion is conceptually similar to how humans interpret the data but with far greater throughput, as the shapes of the radial distributions are different for each phase and easily distinguishable by humans. The transition temperature estimates derived from this machine learning approach produce comparable results to other methods on similarly small system sizes. These results show that machine learning approaches can be applied to structural changes in physical systems.

[1]  Jean-Pierre Hansen,et al.  Phase Transitions of the Lennard-Jones System , 1969 .

[2]  M. J. Gillan,et al.  Iron under Earth’s core conditions: Liquid-state thermodynamics and high-pressure melting curve from ab initio calculations , 2001, cond-mat/0107307.

[3]  Roger G. Melko,et al.  Machine learning phases of matter , 2016, Nature Physics.

[4]  M. Gillan,et al.  First-order phase transitions by first-principles free-energy calculations: the melting of Al , 1998 .

[5]  C. Varma,et al.  Helicity order: hidden order parameter in URu2Si2. , 2005, Physical review letters.

[6]  Gaël Varoquaux,et al.  Scikit-learn: Machine Learning in Python , 2011, J. Mach. Learn. Res..

[7]  H. Diep,et al.  Melting of rare-gas crystals: Monte Carlo simulation versus experiments. , 2012, The Journal of chemical physics.

[8]  Wenjian Hu,et al.  Discovering phases, phase transitions, and crossovers through unsupervised machine learning: A critical examination. , 2017, Physical review. E.

[9]  S. Luo,et al.  Melting of Cu under hydrostatic and shock wave loading to high pressures , 2008 .

[10]  Steve Plimpton,et al.  Fast parallel algorithms for short-range molecular dynamics , 1993 .

[11]  James E. Gubernatis,et al.  Structure classification and melting temperature prediction in octet AB solids via machine learning , 2015 .

[12]  L. Onsager Crystal statistics. I. A two-dimensional model with an order-disorder transition , 1944 .

[13]  Mark Jarrell,et al.  Bayesian inference and the analytic continuation of imaginary-time quantum Monte Carlo data , 1996 .

[14]  Li Huang,et al.  Accelerated Monte Carlo simulations with restricted Boltzmann machines , 2016, 1610.02746.

[15]  Börje Johansson,et al.  Melting and critical superheating , 2006 .

[16]  Karl Pearson F.R.S. LIII. On lines and planes of closest fit to systems of points in space , 1901 .

[17]  Shuaichuang Wang,et al.  Modified Z method to calculate melting curve by molecular dynamics. , 2013, The Journal of chemical physics.

[18]  M. Baskes,et al.  Phase-Field Crystal Model for Fe Connected to MEAM Molecular Dynamics Simulations , 2014 .

[19]  Wang,et al.  Melting line of aluminum from simulations of coexisting phases. , 1994, Physical review. B, Condensed matter.

[20]  A. Rosengren,et al.  High-pressure melting curve of platinum from ab initio Z method , 2012 .

[21]  Sebastian Johann Wetzel,et al.  Unsupervised learning of phase transitions: from principal component analysis to variational autoencoders , 2017, Physical review. E.

[22]  M. Baskes,et al.  Semiempirical atomic potentials for the fcc metals Cu, Ag, Au, Ni, Pd, Pt, Al, and Pb based on first and second nearest-neighbor modified embedded atom method , 2003 .

[23]  H. Kim,et al.  Atomistic modeling of the Ti–Al binary system , 2016 .

[24]  Davenport,et al.  Free-energy calculations and the melting point of Al. , 1992, Physical review. B, Condensed matter.

[25]  John D. Hunter,et al.  Matplotlib: A 2D Graphics Environment , 2007, Computing in Science & Engineering.

[26]  Ranga Komanduri,et al.  Monte Carlo simulations of void-nucleated melting of silicon via modification in the Tersoff potential parameters , 2005 .

[27]  S. P. Lloyd,et al.  Least squares quantization in PCM , 1982, IEEE Trans. Inf. Theory.

[28]  A. van de Walle,et al.  Solid-liquid coexistence in small systems: A statistical method to calculate melting temperatures. , 2013, The Journal of chemical physics.

[29]  Lei Wang,et al.  Discovering phase transitions with unsupervised learning , 2016, 1606.00318.

[30]  A. Strachan,et al.  Nonequilibrium melting and crystallization of a model Lennard-Jones system. , 2004, The Journal of chemical physics.

[31]  R. Hennig,et al.  Classical potential describes martensitic phase transformations between the α, β, and ω titanium phases , 2007, 0706.1764.

[32]  Jarrell,et al.  Quantum Monte Carlo simulations and maximum entropy: Dynamics from imaginary-time data. , 1991, Physical review. B, Condensed matter.

[33]  D. Alfé,et al.  The kinetics of homogeneous melting beyond the limit of superheating. , 2011, The Journal of chemical physics.

[34]  D. Alfé Temperature of the inner-core boundary of the Earth: Melting of iron at high pressure from first-principles coexistence simulations , 2009, 0902.1831.

[35]  S. Davis,et al.  Z method calculations to determine the melting curve of silica at high pressures , 2016 .