A hybrid optimization algorithm to explore atomic configurations of TiO₂ nanoparticles

Journal Article · Tue Oct 17 00:00:00 EDT 2017 · Computational Materials Science

DOI:https://doi.org/10.1016/j.commatsci.2017.08.046· OSTI ID:1429207

Inclan, Eric J. ^[1]; Geohegan, David B.  ^[2]; Yoon, Mina  ^[3]

---

1. Georgia Inst. of Technology, Atlanta, GA (United States)
2. Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States). Center for Nanophase Materials Science (CNMS)
3. Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States). Center for Nanophase Materials Science (CNMS); Univ. of Tennessee, Knoxville, TN (United States). Dept. of Physics and Astronomy

Here in this paper we present a hybrid algorithm comprised of differential evolution, coupled with the Broyden–Fletcher–Goldfarb–Shanno quasi-Newton optimization algorithm, for the purpose of identifying a broad range of (meta)stable Ti_nO₂n nanoparticles, as an example system, described by Buckingham interatomic potential. The potential and its gradient are modified to be piece-wise continuous to enable use of these continuous-domain, unconstrained algorithms, thereby improving compatibility. To measure computational effectiveness a regression on known structures is used. This approach defines effectiveness as the ability of an algorithm to produce a set of structures whose energy distribution follows the regression as the number of Ti_nO_2n increases such that the shape of the distribution is consistent with the algorithm’s stated goals. Our calculation demonstrates that the hybrid algorithm finds global minimum configurations more effectively than the differential evolution algorithms, widely employed in the field of materials science. Specifically, the hybrid algorithm is shown to reproduce the global minimum energy structures reported in the literature up to n = 5, and retains good agreement with the regression up to n = 25. For 25 < n < 100, where literature structures are unavailable, the hybrid effectively obtains structures that are in lower energies per TiO₂ unit as the system size increases.

• Research Organization: Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)
• Sponsoring Organization: USDOE Office of Science (SC), Basic Energy Sciences (BES); National Research Foundation of Korea (NRF)
• Grant/Contract Number: AC05-00OR22725; AC02-05CH11231
• OSTI ID: 1429207
• Alternate ID(s): OSTI ID: 1496343
• Journal Information: Computational Materials Science, Vol. 141; ISSN 0927-0256
• Publisher: Elsevier
• Country of Publication: United States
• Language: English

Related Subjects

36 MATERIALS SCIENCE
Global structure search algorithm
Differential evolution
Hybrid algorithm
Atomistic simulations
Titanium dioxide (TiO2)