Efficient algorithms to generate candidate crystal structures with good stability properties can play a key role in data-driven materials discovery. Here, we show that a crystal diffusion variational autoencoder (CDVAE) is capable of generating two-dimensional (2D) materials of high chemical and structural diversity and formation energies mirroring the training structures. Specifically, we train the CDVAE on 2615 2D materials with energy above the convex hull ΔHhull < 0.3 eV/atom, and generate 5003 materials that we relax using density functional theory (DFT). We also generate 14192 new crystals by systematic element substitution of the training structures. We find that the generative model and lattice decoration approach are complementary and yield materials with similar stability properties but very different crystal structures and chemical compositions. In total we find 11630 predicted new 2D materials, where 8599 of these have ΔHhull < 0.3 eV/atom as the seed structures, while 2004 are within 50 meV of the convex hull and could potentially be synthesised. The relaxed atomic structures of all the materials are available in the open Computational 2D Materials Database (C2DB). Our work establishes the CDVAE as an efficient and reliable crystal generation machine, and significantly expands the space of 2D materials.

We use a generative neural network model to create thousands of new one-dimensional (1D) materials. The model is trained using 508 stable one-dimensional materials from the Computational 1D Materials Database (C1DB) database. More than 500 of the new materials are shown with density-functional theory calculations to be dynamically stable and with heats of formation within 0.2 eV of the convex hull of known materials. Some of the new materials could also have been obtained by chemical element substitution in the training materials, but completely new classes of materials are also produced. The band structures, electronic densities of states, work functions, effective masses, and phonon spectra of the new materials are calculated, and the data are added to the C1DB.

We use a generative neural network model to create thousands of new, one-dimensional materials. The model is trained using 508 stable one-dimensional materials from the Computational 1D Materials Database (C1DB) database. More than 500 of the new materials are shown with density functional theory calculations to be dynamically stable and with heats of formation within 0.2 eV of the convex hull of known materials. Some of the new materials could also have been obtained by chemical element substitution in the training materials, but completely new classes of materials are also produced. The band structures, electronic densities of states, work functions, effective masses, and phonon spectra of the new materials are calculated, and the data are added to C1DB.