Resources
Web applications
Twisted Multilayer Builder
Open the Twisted Multilayer Builder →
An interactive web application for constructing and analyzing twisted multilayer (three or more layers) two-dimensional structures. Stacks can use one material throughout (alternating-twist trilayers, helical multilayers, twisted double bilayers) or mix materials, with layers drawn from tens of thousands of DFT-relaxed monolayers in public 2D materials databases or imported from structure files.
- Commensurate supercells with minimal strain for any stacking and twist pattern, with the moiré period and per-layer strain
- Interactive 3D, top (moiré pattern), and side views, with figure export (PNG, SVG)
- Symmetry analysis of every stack (point and space group, chirality, polarity) and the symmetry-allowed second-harmonic generation (χ⁽²⁾) and circular-dichroism (gyration) tensor components
- Structure export: POSCAR, CIF, XSF, extended XYZ, and metadata
Hosted on Hugging Face Spaces. If the app has been idle, the first visit can take up to a minute to start.
Where the 2D materials come from. The builder’s catalog of 29,855 monolayers is a snapshot of four public databases (monolayers with at most 60 atoms per cell that form a single slab). Every exported structure records its source, license, and citation.
| Database | Monolayers | License | Reference |
|---|---|---|---|
| JARVIS-DFT 2D (NIST) | 1,102 | CC BY 4.0 | K. Choudhary et al., npj Comput. Mater. 6, 173 (2020) |
| 2DMatPedia | 6,345 | open, citation requested | J. Zhou et al., Sci. Data 6, 86 (2019) |
| Materials Cloud MC2D (EPFL) | 2,730 | CC BY 4.0 | D. Campi et al., Materials Cloud Archive 2022.84 (2022) |
| Alexandria 2D (PBE; Ehull ≤ 0.10 eV/atom) | 19,678 | CC BY 4.0 | J. Schmidt et al., 2D Mater. 10, 035009 (2023) |
Users can also import their own structures (POSCAR, CIF, XSF, XYZ, Quantum ESPRESSO, ASE JSON), including individual structures downloaded from C2DB or the Materials Project.
Datasets
- Tensorial optical and transport properties of materials from the Wannier function method. Frequency-dependent optical and transport tensors for 7,301 materials from automated Wannierization. Data: doi:10.6084/m9.figshare.28689059 · Paper: Scientific Data 12, 1092 (2025)
- TSENN models and dataset. Frequency-dependent dielectric tensors and pretrained equivariant models. Data: doi:10.6084/m9.figshare.31180054 · Paper: Nature Communications (2026) · Code: TSENN
- Band structure dataset for Bandformer. Electronic band structures of about 27,000 materials from the Materials Project, prepared for end-to-end band-structure learning. Data: doi:10.6084/m9.figshare.30502967 · Paper: arXiv:2411.16483 · Code: Bandformer
Software
Open-source codes for our machine learning models and workflows are listed on the Code page and on GitHub.