Prediction of novel high-pressure H2O -NaCl and carbon oxide compounds with a symmetry-driven structure search algorithm

Abstract

Crystal structure prediction with theoretical methods is particularly challenging when unit cells with many atoms need to be considered. Here we employ a symmetry-driven structure search (SYDSS) method and combine it with density functional theory (DFT) to predict novel crystal structures at high pressure. We sample randomly from all 1,506 Wyckoff positions of the 230 space groups to generate a set of initial structures. During the subsequent structural relaxation with DFT, existing symmetries are preserved, but the symmetries and the space group may change as atoms move to more symmetric positions. By construction, our algorithm generates symmetric structures with high probability without excluding any configurations. This improves the search efficiency, especially for large cells with 20 atoms or more. We apply our SYDSS algorithm to identify stoichiometric (H2O)n-(NaCl)m and CnOm compounds at high pressure. We predict a novel H2O-NaCl structure with Pnma symmetry to form at 3.4 Mbar, which is within the range of diamond anvil experiments. In addition, we predict a novel C2O structure at 19.8 Mbar and C4O structure at 44.0 Mbar with Pbca and C2/m symmetry respectively.PACS numbers:

Prediction of Chlorine and Fluorine Crystal Structures at High Pressure Using Symmetry Driven Structure Search with Geometric Constraints

“…However, these probabilistic methods do not consider major aspects of the existing data for the materials that they are being employed to study, such as the structural symmetry of certain compounds. [16].…”

“…Symmetries are a key component of studying large clusters [17][18][19] and crystal structures [5,11,12,20]. The addition of symmetry constraints has been shown to improve the algorithmic efficiency for systems with many atoms [16]. In 2011, Pickard and Needs [12] presented an algorithm in which N op symmetry operations are chosen randomly, an atom is placed, and its images are generated according to symmetry.…”

“…This led to the discovery of the ground state for Mg 24 Al 16 Si 24 O 96 , a 160 atom unit cell structure which had not been found previously without symmetry constraints. In 2017 and 2018, similar approaches which rely upon the sampling of space groups and Wyckoff positions were independently developed [16,21]. The XtalOpt algorithm first selects a set of space groups, then creates a list of all possible combinations of Wyckoff positions consistent with the given structure composition for each space group [21].…”

“…In the case of TiO 2 with classical potentials and 10 chosen space groups, this work demonstrated that the symmetry constraints increased the probability of generating both low energy structures and high energy structures, with an increased average energy overall. Domingos et al's symmetry driven structure search (SYDSS) algorithm [16] samples from all 1,506 Wyckoff positions associated with the 230 space groups without generating a list at all. Since there is no need for size restriction, there is no change in the probability for how often certain combinations of Wyckoff positions are chosen.…”

“…However, the correct geometries are either sampled by chance or emerge during the structural relaxation. The novel H 2 O-NaCl structure [16] is an example of the latter case. The structure has Pnma symmetry and the typical geometry of a H 2 O molecule is well preserved.…”