[Paper Review] Band structure of NiO revisited

Luiz G. Ferreira, L. K. Teles+1 more|ArXiv.org|Oct 23, 2009

Transition Metal Oxide Nanomaterials3 citations

TL;DR

This paper re-evaluates the band structure of NiO using the LDA-1/2 and GGA-1/2 methods, demonstrating that the widely cited 4.3 eV band gap attributed to Ni 3d conduction bands is physically implausible. Instead, it identifies a narrow Ni 3d band at ~1.0 eV as the true optical excitation, with the 4.3 eV feature arising from excited atomic states, not delocalized Ni 3d bands.

ABSTRACT

The band structure of a strongly correlated semiconductor as NiO has been the object of much debate [PRL 103, 036404 (2009); PRL 102, 226401 (2009)]. Most authors, using computational techniques well beyond the simple density functional theory and the approximations GGA or LDA, claim that the band gap is about 4.0 eV and that the conduction band is of Ni-3d nature. Thus they seem to forget the results of electron energy-loss spectroscopy and inelastic x-ray scattering, both able to determine electronic transitions of only about 1.0 eV to an optically forbidden Ni-3d band. Further, a simple atomic calculation of the Ni++ spin flip energy demonstrates that a Ni-3d band at 4.0 eV is impossible. To set the issue straight, we calculated NiO with the very successful technique of PRB 78, 125116 (2008). It turns out that a band at 4.0 eV is optically accessible and made of excited atomic states, not Ni-3d. Aside from that, we also found a narrow Ni-3d band at about 1.0 eV. To confirm our procedures once again, we also calculated MnO and obtained the standard results of the good calculations as those cited above, and of experiment.

Motivation & Objective

• To resolve the long-standing controversy over the nature and position of the NiO band gap, particularly the 4.3 eV feature claimed in many advanced DFT calculations.
• To challenge the widespread assumption that the 4.3 eV excitation corresponds to a Ni 3d conduction band, as suggested by some high-level DFT methods.
• To reconcile theoretical calculations with experimental data from electron energy-loss spectroscopy (EELS) and inelastic x-ray scattering (IXS), which indicate a low-energy excitation at ~1.0 eV.
• To validate the LDA-1/2 and GGA-1/2 methods as accurate, low-cost alternatives for predicting band gaps in strongly correlated oxides like NiO.
• To demonstrate that the 4.3 eV feature is not a true conduction band but an artifact of incorrect interpretation of Kohn-Sham eigenvalues and self-energy effects.

Proposed method

• Uses all-electron full-potential linearized augmented plane wave (LAPW+lo) method with WIEN2k to avoid pseudopotential limitations in transition metal oxides.
• Applies the LDA-1/2 and GGA-1/2 formalism, which subtracts a self-energy correction derived from the classical electrostatic energy of localizing a Bloch state into an atomic-like wavefunction.
• The self-energy is calculated in the atomic limit and trimmed using a cutoff function (CUT) to prevent overlap with neighboring atoms, with CUT values optimized variationally to maximize the band gap.
• The method avoids the high computational cost of GW while capturing quasi-particle effects through a physically motivated, parameter-free self-energy correction.
• Validates the method on MnO, where it reproduces the correct 4.18 eV (GGA) and 3.98 eV (LDA) band gaps, matching experiment and high-level calculations.
• Compares Kohn-Sham bands (GGA) with corrected bands (GGA-1/2) to show that the 1.0 eV acceptor band remains stable, while the 4.3 eV feature is removed.

Experimental results

Research questions

• RQ1 Why do many advanced DFT calculations incorrectly assign a 4.3 eV band gap to Ni 3d character in NiO?
• RQ2 What is the true nature of the 4.3 eV excitation feature observed in photoemission and optical measurements?
• RQ3 Can the LDA-1/2 and GGA-1/2 methods accurately predict the band gap and electronic structure of strongly correlated NiO without expensive GW calculations?
• RQ4 How does the spin-flip energy of Ni²⁺ compare to the experimentally observed excitation thresholds, and what does this imply about the band origin?
• RQ5 Why is the 1.0 eV excitation band optically active and consistent with EELS and IXS data, while the 4.3 eV feature is not?

Key findings

• The 4.3 eV band gap, often cited as a Ni 3d conduction band in NiO, is physically impossible because the Ni²⁺ spin-flip energy is only ~1.0 eV.
• The true optical excitation at ~1.0 eV corresponds to a narrow Ni 3d band of minority spin character, consistent with EELS and IXS data.
• The GGA-1/2 calculation yields a 1.00 eV band gap, matching the spin-flip energy calculated with PBE exchange-correlation (0.99 eV), confirming the physical origin of the excitation.
• The 4.3 eV feature is not a delocalized Ni 3d band but an excited atomic state, not accessible via optical transitions to the conduction band.
• The LDA-1/2 and GGA-1/2 methods successfully reproduce the correct band gap and electronic structure of MnO (4.18 eV with GGA), validating the approach.
• The method correctly identifies the 1.0 eV band as an acceptor state, where adding an electron would reduce Ni²⁺ to Ni⁺, explaining its low conductivity and optical inactivity at higher energies.