Gautam Jha

• Jha, Gautam (2026). Convolutional Kernel-Embedded E(3)-Equivariant Networks KEN: Overcoming Architectural Rigidity in Machine Learning Interatomic Potentials. Journal of Chemical Theory and Computation. ISSN 1549-9618. Fulltekst i vitenarkiv Vis sammendrag

  Machine learning interatomic potentials (MLIPs) have emerged as scalable alternatives to first-principles methods such as density functional theory (DFT). Among them, E(3)-equivariant graph neural networks (GNNs) like PACE and MACE are highly accurate but structurally rigid, limiting the incorporation of rich, multi-channel electronic descriptors. We introduce Convolutional Kernel-Embedded E(3)-Equivariant Networks, KEN, a hybrid architecture that combines the local environmental extraction of 2D convolutional kernels with the many-body expressivity of an E(3)-equivariant GNN backbone. This design preserves symmetry while enabling flexible, multi-modal input features without increasing training data requirements. Trained and tested on heavy element 26 (HE26) dataset, KEN was compared with MACE-osaka26 model. Further validation on lattice parameter prediction through Birch–Murnaghan equation of state, magnetic moment prediction and comparison with CHGNet framework, and microcanonical molecular dynamics confirms robust energy conservation and accurate reproduction of the forces, and thus, demonstrating a viable pathway to overcoming architectural rigidity in state-of-the-art MLIPs.

Se alle arbeider i NVA

• Jha, Gautam (2026). DFTB Parameterization Framework for Actinide Molten Salts. Fulltekst i vitenarkiv Vis sammendrag

  Actinide fuel salts are a crucial part of the molten salt reactors. However, the chemistry of these salts under high temperature environments are under explored as complex electronic structure of actinides make it quite difficult to study the evolution of their molten structure for long timescale at Density Functional theory level. Semi-empirical methods such as Density Functional Tight Binding (DFTB) exist which can probe the underlying chemistry with long timescale dynamics but it lacks the required parameters. We have developed a particle swarm optimization based parametrization framework and calculated the DFTB parameter set for actinide and halide interactions which were verified with geometrical optimizations and thermodynamic reaction enthalpies spanning multiple oxidation states of actinides, and radial distribution function verifications from long timescale dynamics simulations of the fuel salts.

• Jha, Gautam (2026). Importance of Kinetic Energy Tensor in Chemistry. Fulltekst i vitenarkiv Vis sammendrag

  Understanding the bonding in chemical systems is one of the prime focus of quantum chemistry and most of the current analyzes are developed on the premise of density functional theory (DFT). DFT relies heavily on the exchange correlation (XC) functionals where the exchange energy is calculated from the Taylor expansion of a spherically averaged exchange hole which eliminates information about the anisotropy of the related properties of the exchange hole. The Taylor expansion expresses the exchange hole curvature in terms of kinetic energy density. Several formulations exist which connect the bonding scenarios to kinetic energy density, such as electron localization functions (ELF). Here, we have investigated bonding in chemical systems by taking the tensor forms of the kinetic energy density rather than the scaler form calculated by spherical averaging. This approach provides us with additional insights such as anisotropy of the bonding modes and opens the path for inclusion of exchange hole anisotropy in kinetic energy based Meta-GGA DFT functionals.

Se alle arbeider i NVA