Electronic transport calculations within the non-equilibrium Green’s function framework are most efficiently performed using local-orbital implementations of density-functional theory. Although plane-wave density-functional calculations provide high accuracy, they are computationally prohibitive for large systems. In practice, the most commonly used atomic-orbital basis sets in local-orbital methods, typically constructed from Gaussian-type orbitals, are less complete than plane-wave expansions and can lead to distortions in electrode band structures, particularly in nanoscale junctions. Achieving accurate and efficient electronic transport calculations therefore requires systematic improvement of Gaussian-type orbital basis sets. In this work, we implement a multi-objective optimisation scheme that combines band-structure agreement, overlap regularisation, and total-energy improvement by modifying an existing total-energy optimiser. This formulation ensures that the optimised Gaussian-type orbital basis reproduces both the energetic and spectral characteristics of fully relativistic plane-wave reference calculations while maintaining numerical stability. Implemented using the CRYSTAL23 electronic-structure code, the workflow iteratively updates selected exponents of uncontracted primitives and converges compact, transferable basis sets with minimal manual tuning. The resulting basis sets retain the computational efficiency of local-orbital transport methods while achieving near plane-wave-level spectral agreement within the transport-relevant energy window around the Fermi level. Validation on representative systems, including metals, Bi(111) bilayers, and helical molecular junctions, demonstrates that the method reliably reproduces representative spin Hall and chirality-induced spin-selectivity transport signatures. The workflow therefore enables practical and accurate simulations of spin–orbit-coupled transport in complex nanostructures, where conventional Gaussian-type orbital approaches may be less accurate and plane-wave methods are computationally intractable. • Workflow enables accurate SOC transport in nanoscale systems. • Optimises uncontracted Gaussian primitives for band agreement. • Non-magnetic optimisation yields consistent SOC band structures. • Balances band mismatch with overlap and energy penalties. • SOC transport captures CISS trends and Bi(111) spin Hall behaviour.
Dednam et al. (Tue,) studied this question.