Eliminating Delocalization Error through Localized Orbital Scaling Correction with Orbital Relaxation from Linear Response

Despite the great success Kohn-Sham density functional theory (KS-DFT) has achieved, the delocalization error remains a major challenge for commonly used density functional approximations (DFAs), resulting in systematic errors in ionization energies, electron affinities, band structures, and charge distributions. A recently developed localized orbital scaling correction (LOSC) method, namely linear response LOSC (lrLOSC), addresses these challenges by incorporating a functional correction that includes the screening effect and orbital localization within the LOSC framework. The method has been shown to provide accurate descriptions of bulk systems and core-level binding energies in small molecular systems. In this work, we extend the applicability of lrLOSC to a broader range of molecular systems, spanning various sizes, with a focus on the corrections to valence orbital energies and total energies. To enable the calculation of large chemical systems, we developed an efficient implementation of lrLOSC with computational costs comparable to standard KS-DFT calculations. Numerical results show that, while screening provides modest improvements for small molecules, it becomes critical for achieving high accuracy in larger molecules, from linear to three-dimensional systems. With the screening effect well captured in a unified way, lrLOSC provides accurate descriptions for a wide range of chemical systems, including organic molecular systems of varying sizes and transition-metal oxide complexes, establishing it as a powerful tool for enhancing the reliability of computational simulations of chemical systems.

💡 Research Summary

This paper addresses the persistent delocalization error (DE) that plagues conventional Kohn‑Sham density‑functional approximations (DFAs). The authors extend the previously introduced localized‑orbital scaling correction (LOSC) by incorporating linear‑response theory, resulting in the linear‑response LOSC (lrLOSC). The method combines two essential ingredients: (1) orbital localization in both real‑space and energy‑space, achieved through a unitary transformation of all occupied and virtual canonical molecular orbitals into “orbitalets,” and (2) orbital relaxation (screening) captured via a curvature matrix derived from linear‑response equations.

In the localization step, a cost function Fσ that balances spatial compactness and energy spread is minimized with the weight γ set to zero, forcing the procedure to prioritize spatial localization. This yields orbitalets that are highly localized for large or stretched systems, leading to fractional occupations in the localized‑orbital (LO) occupation matrix λσ. For small molecules the orbitalets remain close to canonical orbitals, producing integer diagonal λσ elements.

The screening component is encoded in the curvature matrix κσ. Unlike the frozen‑orbital approximation used in earlier global‑scaling corrections (GSC), lrLOSC constructs κσ from the Hartree‑exchange‑correlation kernel K and the orbital energy differences, forming a large matrix M that couples occupied–virtual pairs. Direct inversion of M would scale as O(N⁶), prohibiting applications to large systems. To overcome this, the authors employ a Resolution‑of‑Identity (RI‑V) decomposition of the four‑center integrals and rewrite M as A + X Yᵀ, where A is diagonal and X, Y are low‑rank three‑center matrices built from auxiliary basis functions. Using the Sherman‑Morrison‑Woodbury formula, the inverse is obtained with O(N³) operations and modest memory requirements.

The implementation is benchmarked on four categories of systems: (i) small molecules (≤ 6 atoms), (ii) larger non‑linear organic molecules, (iii) polyacetylene chains of varying length, and (iv) transition‑metal monoxide complexes. All calculations are performed as post‑SCF corrections on top of PBE. The authors compare ionization potentials (IP) and electron affinities (EA) obtained from lrLOSC with those from GSC2 (screened correction using canonical orbitals), LOSC2 (localized orbitals without screening), and Δ‑SCF.

Results show that lrLOSC consistently yields the lowest mean absolute errors across all test sets. For small molecules, the three methods perform similarly, but as molecular size grows, lrLOSC outperforms GSC2 because the orbitalets capture the shift from delocalized canonical orbitals to localized ones. Compared with LOSC2, the inclusion of the screening term in κσ reduces errors for every system, confirming the importance of orbital relaxation. In transition‑metal oxides, where Δ‑SCF is often ill‑defined, lrLOSC still provides accurate IP and EA values and halves the error relative to the original LOSC.

Beyond accuracy, the RI‑V/SMW implementation brings the computational cost of lrLOSC close to that of standard KS‑DFT, enabling routine application to systems with hundreds of atoms. The authors therefore demonstrate that lrLOSC offers a universal, efficient, and systematically improvable correction for DE, applicable to molecules, polymers, and strongly correlated materials alike.

In summary, the paper presents a theoretically sound and computationally tractable framework that eliminates delocalization error by (i) dynamically constructing localized orbitalets, (ii) incorporating linear‑response screening into the curvature matrix, and (iii) exploiting low‑rank RI techniques for scalability. This positions lrLOSC as a powerful post‑DFT tool for reliable prediction of frontier orbital energies, total energies, and related properties in a broad spectrum of chemical and material systems.