Machine-learning-driven Expansion of 1D van der Waals Materials Space

Machine-learning techniques were utilized to discover 1D vdW compositions that have not yet been synthesized. This model identified MoI3, a material which was experimentally confirmed to exist with wire-like subcomponents and exotic magnetic properties.

L. Bartels and A. Balandin (U. CA-Riverside) E. Reed and F. H. da Jornada (Stanford)

The material space under consideration. We start with materials in the Materials Project database with corresponding ICSD numbers, where the dimensionality of the materials is known. We seek for additional low-dimensional materials in the larger space of randomly generated compositions, targeting those that are likely conductive or magnetic.

One-dimensional (1D) van der Waals (vdW) materials display electronic and magnetic transport properties that make them uniquely suited as interconnect materials and for low-dimensional optoelectronic applications. However, there are only around 700 1D vdWstructures in general materials databases, making database curation approaches ineffective for 1D discovery. Here, machine-learning techniques were utilized to discover 1D vdWcompositions that have not yet been synthesized.

These techniques go beyond discovery efforts involving elemental substitutions and instead start with a composition space of 4741 binary and 392,342 ternary formulas. Up to 3000 binary and 10,000 ternary compounds were predicted and further classified by expected magnetic and electronic properties. This model identified MoI3, a material which was experimentally confirmed to exist with wire-like subcomponents and exotic magnetic properties. More broadly, several chalcogen-, halogen-, and pnictogen-containing compounds are expected to be synthesizable using chemical vapor deposition and chemical vapor transport.

Designing Materials to Revolutionize and Engineer our Future (DMREF)