Professor Hafid Aourag
Materials Informatics and  Computational Materials Science
Department of Physics
University of Tlemcen
email : h_aourag@mail.univ-tlemcen.dz

Over the last decade a substantial international collaborative effort has occurred  in developing a new approach for the design of a wide variety of engineering materials.  There is a well-established methodology for the design of parts,   A number of molecular simulation tools such a molecular dynamics and density functional theory have become available in recent years; however, these tools are usually restricted to idealized materials not the complex formulations that are typically present in real engineering materials.  A new design framework for complex engineering materials is needed, and this is the thrust of our research efforts.  We have recently established the Materials Genome Project to expand, formalize and disseminate this new approach for materials design leading to new engineering and consumer products.

The Materials Design process can be divided into two problem - the forward problem of determining the engineering properties from the formulation and/or molecular structure of the material and the inverse problem of determining the optimal composition/structure needed to meet a set of material requirements.  Our research objective is to develop the appropriate methodology and new computer-aided tools to address both the forward and inverse problem in materials design. The solution of the inverse problem usually has the most significant technological and economic impact, since it directly results in the formulation of new materials to meet a specific application objective.  However, solution of the inverse problem will require prediction of properties for a number of candidate material formulations and, thus, will require an accurate and robust forward model.

There are traditionally two approaches for developing forward models for predicting how changes in the molecular architecture give rise to engineering properties. First principle models try to rigorously acknowledge all the underlying chemistry and physics.  This modeling approach has the advantage that predictions can be extrapolated to new materials and application situations with more confidence, because the fundamental processes that control the material behavior are explicitly incorporated in the model. The drawbacks to a first principle's approach are (i) the time to develop the model is often exceedingly slow, (ii) the material systems being modeled are often highly idealized and (iii) model predictions are typically only available at the completion of, not during, the research program.  In contrast, data driven models are relatively easy to implement so that model predictions can be rapidly developed for complex material systems.  However, data driven models (i) often require enormous amounts of data, (ii) are limited by noisy data which is often the case in material development, (iii) have limited ability to extrapolate to composition regions that are outside the data region and (iv) require data that are uniformly spaced in composition space rather than the typical situation where data is clustered around materials of current commercial interest.  Our approach is to not use just a purely theoretical approach nor a purely data driven approach.  Rather, we use first principle information to reduce the amount of data by an order of magnitude or more, but we will also allow the data to correct for deficiencies in the first principle models. Our objective is to substantially reduce the experimental load by using physically based models, while at the same time not being held hostage to the requirement that the physical model be perfect.  In this approach, more experiments will be required when developing the framework for a new class of materials; however, for additional applications using that class of materials, the first principle models will improve and the experimental load can be decreased.