Materials Informatics

Designing novel materials with the desired physical and chemical properties is recognized as an outstanding problem in materials research. This problem is extremely challenging because materials’ properties depend on a large number of variables such as constitutive elements, crystal forms, and geometrical or electronic characteristics. Traditionally, computational design of materials has been accomplished by means of quantum mechanical and simulations approaches. However, rapid growth of material research has led to the accumulation of a huge body of empirical data on crystal structures and properties of materials stored in specialized databases. These developments create new opportunities to explore the materials knowledge base using statistical and data mining approaches and enable further development and enhancement of the nascent field of materials informatics.


We develop and deploy, ultimately in the form of user-friendly software, novel materials informatics approaches and computational tools to empower material scientists with efficient technologies for analyzing, modeling, and imputing properties of materials. These tools will enable:

  • mining of specialized databases to identify materials with specific constitutional, geometric, electronic, and other measured or computed characteristics;
  • visualizing very large sets of diverse materials to detect clusters of materials with both similar characteristics and properties and determine patterns of materials characteristics directly responsible for their properties; and
  • establishing reliable Materials Quantitative Structure-Property Relationship (MQSPR) models for predicting materials’ properties from their computed characteristics.
Data Mining and Structure-Activity Relationships (SAR)
Empty section. Edit page to add content here.
GPU Computing
Empty section. Edit page to add content here.
Chemical Accuracy & First-Principles Simulations

During 2010-2012 focus of my work brings together a hierarchy of state-of-the-art computational methods to address both applications and requisite theoretical methodology development in relation to the properties of nanoparticles and bio-nano interfaces, thus facilitating the pipeline from fundamental understanding to perspective practical solutions. In particular, we have carried out theoretical simulations of tensile tests on Si<001> nanowires adapting constrained Car–Parrinello molecular dynamics.

I have a long-standing research interest in trying to push ab initio methods to their limit of accuracy. This involves such issues as: basis set convergence and extrapolation to the complete basis set limit, accounting for higher-order correlation effects, and the rotational-vibrational anharmonicity. For small and medium polyatomic molecules we are presently able, to predict binding free energies and enthalpies to an accuracy of 1 kcal/mol or better, very accurate geometries and vibrational modes.

Our research efforts have manifested into two “black box” composite com-putational thermochemistry protocols. First provides the procedure for calculations of interaction Gibbs free energies for intermolecular complexes of biological interests. The second one offers very robust vey to predict redox potential for organic aromatic compounds. Aside from applications of extremely demanding correlated methods composite protocols usually include less CPU-intensive methods (like DFT) that provide reasonable trade-off between desired accuracy and time to solution.