Research Projects

Thermoelectric materials can directly convert heat into electricity and vice-versa, due to which they provide unique opportunities to meet our energy requirements and reduce our dependency on fossil fuels. However, the energy conversion efficiency of the currently known thermoelectric materials is poor which limits their practical applications. The vast material space offers both challenges as well as opportunities to discover more efficient thermoelectric materials, necessitating the development of an efficient and accurate computational framework which can sample the candidate materials that can be tested later using experiments for their thermoelectric efficiency, thereby motivating this work. The objective of this research work is to develop a large-scale parallel computational framework based on first-principles electronic structure theory which can efficiently leverage modern hybrid high performance computational resources and can accurately predict the thermal and electrical conductivity of the material systems. We will develop a large-scale parallel density functional perturbation theory formulation and implementation compatible with the framework of the large-scale parallel real-space electronic structure code SPARC, to calculate the second and third order interatomic force constants. We will also extend the formulation and implementation to hybrid functionals and low dimensional nanostructures with unconventional symmetries. We will also distribute the computational load among CPU cores and GPUs to enhance the performance of the computations, especially those involving matrix-matrix multiplications. We will also develop a preconditioned iterative solver for Peierls Boltzmann transport equation which will be used to obtain non-equilibrium electron and phonon distributions, providing an accurate and efficient computation of electrical and thermal conductivities as well as thermoelectric efficiency of materials. We will also account for electron-phonon coupling through the coupling matrix which will be calculated utilizing the perturbed potential from DFPT implementation in SPARC code. Finally, we will also extend the SOAP descriptor with polynomial kernel based on-the-fly machine-learned force field implementation in SPARC to calculate the second and third order interatomic force constants at near quantum mechanical accuracy but with many orders of magnitude speedup. Overall, this research work will open avenues to accurately and efficiently study the thermoelectrical performance of material systems of varying dimensionalities and symmetries, significantly speeding up the discovery of highly efficient thermoelectric materials with technological and environmental implications.