Article Preview
TopIntroduction
Information Sciences are increasingly infiltrating the applied fields of various natural sciences. Currently such new scientific directions are disciplines such as bioinformatics, chemoinformatics, helth informatics and material informatics. In fact, all of these research areas are based on Data Mining approaches. For addressing chemical problems, machine learning methods are widely used that are implemented in a variety of quantitative structure-activity/property relationship (QSAR / QSPR) models (Roy et al., 2015). In particular, QSPR methods are being actively implemented in the materials informatics to intensify and optimize the processes of creating new materials with a set of useful properties (Senderowitz and Tropsha, 2018; Butler et al., 2018; Lu et al., 2017; Yosipof et al., 2016).
Modern methods of chemoinformatics are widely used in the study of organic compounds including potential drugs, herbicides, pesticides, etc. (Cherkasov et al., 2014; Wang et al., 2017). For inorganic compounds, one can refer a few examples of solving QSPR tasks (Seko, et al., 2014; Mu, et al., 2006; Mu, et al., 2007; Goudarzi, et al., 2013). This is due to objective reasons, ordinary descriptor schemes for describing molecules are of little use for inorganic substances. The conceptual differences in modeling the structure of organic and inorganic compounds are the reasons for this. The number of atom types is significantly less for organic compounds than for inorganic compounds. However, it should be noted that the structural diversity (molecular graphs) is much greater for organic compounds than for classical inorganic compounds. The term of “molecule” is rather conditional for many crystalline inorganic compounds; isomerism phenomenon is not as widespread for them (except for coordination compounds).
The development and implementation of new QSPR approaches for the study of inorganic compounds is quite an actual problem. The relevance and prospects of such studies are due to the wide possibilities for creating new inorganic materials.
The goal of this study was to develop QSPR models for prediction melting points (MP) and refractive indices (RI) of various inorganic compounds. Information about these properties is necessary for the creation of new promising optical materials.
Obviously, the refractive index is one of the most important parameters for optical materials. This parameter is associated with other optical characteristics of the substance, such as the region of transparency, reflection and absorption coefficients, width of the forbidden zone, etc. The refractive index has a physical meaning only for substances with ionic or covalent structure, but not for metals, and its value depends on the nature of the material, wavelength, temperature, aggregation state, pressure, etc. Fluorides of alkali and alkaline earth metals (lithium, sodium, magnesium, calcium, strontium) have one of the lowest RI, while metal chalcogenides (especially, tellurides of lead, germanium and tin) have the highest values of RI (Zinchenko, 2006). In interference optics the RI has particular importance when using optical materials for applying thin-film coatings on optical elements (lenses, prisms, etc.).