Reliable semiconductor bandgap prediction is important when assessing electronic materials before synthesis or further development. Yet standard density functional theory (DFT) approaches based on local density or generalized gradient approximations often underestimate bandgaps. A first-principles method combining a compact, full-potential basis with a correction to the exchange potential offers one route to better estimates without relying on computationally more demanding approaches.
The approach pairs full-potential Nth-order muffin-tin orbital calculations (FP-NMTO) with a van Leeuwen–Baerends correction to the local density approximation (LDA+vLB). Results for group IV and III–V semiconductors, as well as two-dimensional hexagonal boron nitride and silicon carbide, illustrate how the basis set and exchange-correlation potential both affect calculated electronic and structural properties.
Why conventional DFT can miss semiconductor bandgaps
A semiconductor’s fundamental bandgap is the energy difference between the energy required to remove an electron and the energy gained when an electron is added. In practical calculations, researchers often examine the gap between the highest occupied and lowest unoccupied Kohn–Sham states. These quantities are related, but they are not generally identical: the fundamental gap also includes a contribution associated with a discontinuous change in the exchange-correlation potential as electron number changes.
Common LDA and generalized gradient approximation (GGA) functionals do not fully capture this contribution. They consequently tend to underestimate bandgaps, sometimes substantially. This limitation matters when calculations are used to compare candidate materials, interpret electronic structure, or estimate properties ahead of experimental work.
The vLB correction modifies the LDA exchange potential using information about the electron-density gradient. Its intended benefit is improved potential behavior in low-density regions, including a more appropriate long-range decay. The correction is applied to the potential; it is not simply a uniform adjustment to the calculated conduction bands. As a result, the effects can vary across states and materials rather than acting like a constant shift.
What FP-NMTO contributes to bandgap calculations
The second part of the approach is the FP-NMTO basis set. NMTO uses energy-dependent partial-wave solutions to construct an energy-independent basis through interpolation across selected energies. In the reported calculations, a third-order correction is used. The resulting basis is compact while describing the system across both atomic regions and the interstitial space between them.
That full-potential treatment is significant. Simplified atomic-sphere methods approximate the potential using overlapping spheres, which can constrain how the interstitial region is represented. FP-NMTO is designed to handle the full potential in both regions. The study’s results underline that the choice of basis is not a minor implementation detail: a more complete basis can improve calculated electronic and structural properties, including when the exchange-correlation treatment is held in view.
Calculations were performed self-consistently, with Brillouin-zone sampling and structural parameters obtained by fitting calculated data to an equation of state. The authors compared LDA+vLB results with LDA, PBE, experimental measurements, and results from other theoretical methods.
Results across group IV and III–V semiconductors
Germanium provides a clear example of the difference made by the combined approach. FP-NMTO calculations using LDA and PBE predict gaps of 0.19 eV and 0.33 eV, respectively, and incorrectly identify the gap as direct. With LDA+vLB, the calculated gap is 0.86 eV and is indirect, consistent with the experimentally observed indirect character and closer to the measured value of 0.74 eV. The calculated lattice constant is 5.57 Å, compared with 5.66 Å experimentally, while the calculated bulk modulus is 75 GPa versus a measured 77.2 GPa.
For gallium arsenide, the LDA, PBE, and LDA+vLB calculations give bandgaps of 0.33 eV, 0.54 eV, and 1.43 eV, respectively. The LDA+vLB result is close to the reported experimental value of 1.52 eV and reproduces the material’s direct-gap character. These comparisons show why both gap size and gap location in the Brillouin zone matter when evaluating whether a calculation reflects the material’s electronic structure.
Across the wider set of group IV and III–V compounds, LDA+vLB with FP-NMTO improves many predicted gaps relative to the conventional functionals tested. It does not produce uniformly exact agreement: for example, results for indium compounds vary in their closeness to experiment. The findings therefore support the method as a useful computational approach, not as a guarantee that every material will be predicted to the same accuracy.
Extending the method to two-dimensional materials
The calculations also consider hexagonal boron nitride (h-BN) and hexagonal silicon carbide (h-SiC). For h-BN, the calculated lattice constant is 2.508 Å, close to the reported measured value of 2.504 Å. Its calculated bandgap is approximately 4.5 eV, within the reported experimental range of 3.6–6 eV. For h-SiC, the calculated lattice constant is 2.62 Å, compared with a reported measurement of 2.6 ± 0.2 Å. The calculated gap is 3.90 eV, above the reported experimental value of 2.96 eV, although the study notes that this overestimation is also seen in other theoretical results.
These examples demonstrate the scope of the approach while also showing why validation against measurements remains important. Agreement in lattice constants does not automatically establish equal accuracy in electronic properties, and results should be assessed material by material.
Practical implications for computational materials workflows
For engineers and technical teams evaluating semiconductor materials, the main lesson is methodological: bandgap accuracy depends on more than selecting an exchange-correlation correction. The representation of the wave functions and the treatment of the interstitial region also influence the result. Comparing methods on a shared material set can help identify whether an apparent improvement comes from the potential, the basis, or their combination.
- Check gap character as well as magnitude. A calculation can miss whether the fundamental gap is direct or indirect, as the germanium comparison illustrates.
- Assess structural and electronic predictions together. Lattice constants and bulk moduli provide useful checks alongside bandgaps.
- Retain material-specific validation. The reported results vary across compounds, and selected cases remain less closely aligned with experiment.
- Match comparisons carefully. Different computational methods and implementations can use different potential references and approximations, complicating direct comparison.
FP-NMTO combined with LDA+vLB offers a compact first-principles route to improved semiconductor bandgap estimates in the systems examined. Its practical value lies in combining a more complete basis with a corrected exchange potential, while keeping the calculation self-consistent. For materials screening and electronic-structure analysis, the results make a case for evaluating both ingredients—and for treating calculated gaps as evidence to be checked against structural data and experiment, rather than as standalone answers.