Our current work is focused on structure-based atomistic approach, where the information about the atomistic arrangement is known a priori (see data conventions).
Our data convention has support for materials data where no structural information is available, however this topic is beyond the content of the current documentation.
For examples of JSON representation of materials and structure-based descriptors see materials data section.
Features, Fingerprints, Targets¶
On many occasions terms "Features", "Fingerprints", "Targets" are used for materials informatics purposes. For example, when constructing a machine learning model a dataset containing information about multiple materials is used in order to find regular patterns and inter-dependencies. Such dataset is usually split into properties that represent the known data, or features, and the properties to be predicted, or targets. First we clarify this terminology as follows:
- Feature: any property of a material, eg. density or electronic band gap.
Fingerprint: property of a material used as an input for a (statistical modeling) Workflow, equivalent of Descriptive property by definition
By default, we only store the minimal amount of information required to identify a material enough to reveal a set of its Fingerprints. Such minimal set of properties is called Identity Fingerprints, and the rest - Derived Fingerprints.
Target: property of a material used as an input for a (statistical modeling) Workflow, equivalent of a Characteristic property by definition.
Thus a property-descriptive-characteristic triad is equivalent to feature-fingerprint-target.
Exact set of Materials properties that have to be supplied to and can be extracted as a result of a Workflow vary based on the type of Workflow and models/methods included therein.
Below we provide example properties extracted by the using Density Functional Theory (DFT) as simulation results:
|Total Energy||The ground state energy (free energy) of the system|
|Fermi Energy||The highest energy level occupied by electrons in a system|
|Fermi Surface||Surface of constant energy (Fermi) in reciprocal space|
|Atomic forces||Force exerted on each atom by its surrounding|
|Stress tensor||3x3 matrix expressing stresses in x, y and z dimensions|
|Pressure||Scalar average pressure|
|Charge density||Spatial function of charge distribution|
|Band Structure||Electronic Band Structure|
|Band Gap||Electronic Band Gap (direct / indirect)|
|Density of States||Electronic Density of States (including partial contributions)|
|Zero Point Energy||Energy of the lowest vibrational level wrt to vacuum|
|Final Structure||Visualization of the final computed crystal structure|
|Total Energy Contributions||Ewald, Exchange correlation and Hartree contributions to the total energy|
|Magnetic Moments||The magnetic moment of ferromagnetic materials when the "Magnetism" modifier is activated|
|Total Force||Sum of the atomic forces|
These are the data points that can be monitored during the course of a DFT calculation:
|Standard Output||Standard output of an execution unit in UNIX sense|
|Ionic Convergence||Convergence information on ionic moves in relaxation or molecular dynamics calculations|
|Electronic Convergence||Convergence information on self-consistent electronic calculation steps|