Post by Anastasiia Novikova.
In theory, many ordinary materials can have exotic topological phases. But how can we find them? In 2018 a research group from the National Laboratory for Condensed Matter Physics in Beijing scanned 39519 materials to predict which phases of the already-known compounds might exhibit topological properties. These materials were summarised into an interactive database Materiae, where you can browse compounds containing particular elements, check if they have any topological phases and visualise their band structure.
We asked Prof. Chen Fang — one of the team members who worked on Materiae along with Prof. Hongming Weng — to give us more details of the project, which has now been published in Nature.
When did the database start? What were the main challenges of this project? What goals do you have for the future?
The database has been online since 23 July 2018; it appeared simultaneously with the posting of the corresponding paper on arXiv. By now there have been over 10000 unique visitors (1=ip*day). The most difficult part is, naturally, the calculation that was done to obtain the topological properties of about 30000 materials. The theory, the underlying work was accomplished back in late 2017 (arXiv:1711.11049 and 1711.11050), but even so, it was an effort to implement the fully automated algorithm shown in the flowchart. Currently we have the band structures plotted for topological materials only, and in the future we will add the band structure plots for all materials, topological and non-topological.
Using your algorithm, you scanned 39519 materials. How much time did the whole calculation process take?
We didn’t track the CPU hours used on this, but if we count the time spent on debugging small bugs now and then, it took us about three to four months in total for the bulk results to come out.
You mention that 8056 materials from your database are actually topological. How many of these materials were experimentally studied?
All materials have been reportedly synthesized in literature, but most of them were not studied from a “topological perspective”, but were studied for superconductivity or ferroelectricity, for example. I think at most few hundreds of these materials have been studied for potential topological properties.
What is the most “underestimated” material?
One example is Tl2Nb2O7. Oxides are seldom considered as topological materials in literature, yet our database registers it as a topological semi-metal. Surprised by this result, we further looked into this material, and realized that the mixed-valence nature of Tl ion is the origin of the nontrivial topology.
Another is Ba3Cd2As4. The layered structure made us expect it to be a weak topological insulator, but our database shows it to be a new type of topological crystalline insulator (having so-called C2-anomalous surface states). Shortly after the prediction, experimental groups have started synthesizing this material.
We expect the study of certain materials, like the ones above, may be “revived” by what we show in the database.
The database contains only non-magnetic materials. Is it possible to envision a similar type of database for magnetic materials?
The entire prediction is based on first-principle calculation, but magnetism is notoriously difficult to predict/include in any first-principle calculations. Therefore, while some theoretical work on the mapping between symmetry data and topological data has been out there for a while (arXiv:1707.01903), I do not think a similar material database can be obtained in near future because of the inherent difficulty of DFT mentioned above.