|
|
|
## Tutorial example on crystal-structure prediction I: The case of octet-binary zincblend-vs.-rocksalt semiconductors
|
|
|
|
|
|
|
|
# Introduction
|
|
|
|
|
|
In this tutorial we present a tool for predicting the crystal structure of octet binary compounds, by using a set of descriptive parameters (a descriptor) based on free-atom data of the atomic species constituting the binary material.
|
|
In this tutorial we present a tool for predicting the crystal structure of octet binary compounds, by using a set of descriptive parameters (a descriptor) based on free-atom data of the atomic species constituting the binary material.
|
|
|
|
|
|
In this example, we address only Rocksalt (RS) and Zincblende (ZB) crystal structures, that are the most common for the material class of octet binaries.
|
|
In this example, we address only Rocksalt (RS) and Zincblende (ZB) crystal structures, that are the most common for the material class of octet binaries.
|
|
Specifically, the tool predicts the difference in total energy between RS and ZB equilibrated structures (i.e., each structure is relaxed to its local minimum).
|
|
Specifically, the tool predicts the difference in total energy between RS and ZB equilibrated structures (i.e., each structure is relaxed to its local minimum).
|
|
|
|
|
|
The prediction of RS vs ZB structure from a simple descriptor has a notable history in materials science [1-7], were descriptors were designed by chemically/physically-inspired intuition. The tool presented here allows for the machine-learning-aided automatic discovery of a descriptor and a model for the prediction of the difference in energy between RS and ZB for 82 octet binary materials.
|
|
The prediction of RS vs ZB structure from a simple descriptor has a notable history in materials science [1-7], were descriptors were designed by chemically/physically-inspired intuition. The tool presented here allows for the machine-learning-aided automatic discovery of a descriptor and a model for the prediction of the difference in energy between RS and ZB for 82 octet binary materials.
|
|
The tool is based on Compressed-sensing based method (LASSO performed on a tailor-made feature space, followed by L0-regularized minimization, click here for more info on the LASSO+L0 method),
|
|
|
|
as introduced in:
|
|
|
|
"Big Data of Materials Science: Critical Role of the Descriptor". L. M. Ghiringhelli, J. Vybiral, S. V. Levchenko, C. Draxl, and M. Scheffler Phys. Rev. Lett. 114, 105503 (2015)
|
|
|
|
|
|
|
|
By running the tutorial with the default setting, the results of the PRL 2015 can be recovered. In particular, by clicking on “View interactive 2D plot”, an interactive structure-map (a chart where different structures are located in different regions of a low-dimensional representation, here two dimensional) will be opened in a new tab, similar to the following (a revised version of Fig. 2 in PRL 2015):
|
|
The tool is based on Compressed-sensing based method (LASSO performed on a tailor-made feature space, followed by L0-regularized minimization, click here for more info on the LASSO+L0 method), as introduced in:
|
|
|
|
|
|
|
|
"Big Data of Materials Science: Critical Role of the Descriptor". L. M. Ghiringhelli, J. Vybiral, S. V. Levchenko, C. Draxl, and M. Scheffler Phys. Rev. Lett. 114, 105503 (2015) [(Click here for the free access pdf)](http://journals.aps.org/prl/abstract/10.1103/PhysRevLett.114.105503)
|
|
|
|
|
|
|
|
By running the tutorial with the default setting, the results of the [PRL 2015](http://journals.aps.org/prl/abstract/10.1103/PhysRevLett.114.105503) can be recovered. In particular, by clicking on “View interactive 2D plot”, an interactive structure-map (a chart where different structures are located in different regions of a low-dimensional representation, here two dimensional) will be opened in a new tab, similar to the following (a revised version of Fig. 2 in [PRL 2015](http://journals.aps.org/prl/abstract/10.1103/PhysRevLett.114.105503)):
|
|
|
|
|
|
![2016-08-01_ZB_RS3-1](/uploads/44186947a58c4babc13cc9716bd0968d/2016-08-01_ZB_RS3-1.png)
|
|
![2016-08-01_ZB_RS3-4](/uploads/692807af92fb7ebc3fd139e4f1a08bca/2016-08-01_ZB_RS3-4.png)
|
|
|
|
|
|
In this map the octet binaries are located via the descriptor found by our LASSO+L0 approach. The descriptor is based purely on free-atom data, namely radii of the s and p valence orbitals (rs and rp) of the atomic species and their Ionizaiton Potential and Electron Affinity (IP and EA).
|
|
In this map the octet binaries are located via the descriptor found by our LASSO+L0 approach. The descriptor is based purely on free-atom data, namely radii of the s and p valence orbitals (rs and rp) of the atomic species and their Ionizaiton Potential and Electron Affinity (IP and EA).
|
|
Mateirals in the red (bue) region crystallize preferably in the zincblende (rocksalt) structure. The distance to the green line is proportional to the difference in energy between the two structures.
|
|
Mateirals in the red (bue) region crystallize preferably in the zincblende (rocksalt) structure. The distance to the green line is proportional to the difference in energy between the two structures.
|
... | | ... | |