Modelling Realistic TiO2 Nanospheres: a Benchmark Study of SCC-DFTB against Hybrid DFT

TL;DR

This study compares SCC-DFTB and hybrid DFT methods for modeling TiO2 nanospheres, demonstrating that SCC-DFTB offers accurate structural and electronic insights at a significantly lower computational cost for large systems.

Contribution

It provides a systematic benchmark of SCC-DFTB against hybrid DFT for realistic TiO2 nanospheres, highlighting its efficiency and accuracy.

Findings

DFTB accurately predicts geometrical properties of TiO2 nanospheres.

Electronic properties from DFTB closely match those from hybrid DFT.

DFTB enables efficient simulation of large nanostructures.

Abstract

TiO2 nanoparticles (NPs) are nowadays considered fundamental building blocks for many technological applications. Morphology is found to play a key role with spherical NPs presenting higher binding properties and chemical activity. From the experimental point of view, the characterization of these nano-objects is extremely complex, opening a large room for computational investigations. In this work, TiO2 spherical NPs of different size (from 300 to 4000 atoms) have been studied with a two-scale computational approach. Global optimization to obtain stable and equilibrated NSs was performed with a self-consistent charge density functional tight-binding (SCC-DFTB) simulated annealing process, causing a considerable atomic rearrangement within the nanospheres. Those SCC-DFTB relaxed structures have been then optimized at DFT(B3LYP) level of theory. We present a systematic and comparative SCC-DFTB vs DFT(B3LYP) study of the structural properties, with particular emphasis on the surface-to-bulk sites ratio, coordination distribution of surface sites and surface energy. From the electronic point of view, we compare HOMO-LUMO and Kohn-Sham gaps, total and projected density of states. Overall, the comparisons between DFTB and hybrid DFT show that DFTB provides a rather accurate geometrical and electronic description of these nanospheres of realistic size (up to a diameter of 4.4 nm) at an extremely reduced computational cost. This opens for new challenges in simulations of very large systems and more extended molecular dynamics.