Hybrid HSE06 Functional Overview

• Hybrid HSE06 is a screened exchange–correlation functional that replaces part of PBE exchange with exact Hartree–Fock exchange to balance accuracy and efficiency.
• It confines the computationally expensive Hartree–Fock exchange to short interelectronic distances, reducing self-interaction errors and convergence issues compared to global hybrids.
• Parameter tuning, such as adjusting the Fock exchange fraction, enables closer experimental agreement in band gaps and optical properties for materials like anatase TiO2.

The hybrid HSE06 functional is a screened exchange–correlation (XC) functional in the Heyd–Scuseria–Ernzerhof (HSE) family, designed for Kohn–Sham density functional theory (DFT). It replaces a fixed fraction of short-range semilocal exchange (Perdew–Burke–Ernzerhof, PBE) with screened Hartree–Fock (exact) exchange, while retaining PBE at long range and for correlation. The range separation reduces the computational cost and convergence issues of global hybrids, offering improved band gaps and electronic structure properties in extended systems, particularly solids and low-dimensional materials. Modifications to the standard exact-exchange fraction have also been explored to further optimize agreement with experimental properties in specific materials.

1. Mathematical Formulation of the HSE06 Functional

The total exchange–correlation energy in HSE06 is given by

$$E_{\rm xc}^{\rm HSE06}(a,\mu) = a\,E_x^{\rm HF,SR}(\mu) + (1-a)\,E_x^{\rm PBE,SR}(\mu) + E_x^{\rm PBE,LR}(\mu) + E_c^{\rm PBE}$$

where:

• $E_x^{\rm HF,SR}(\mu)$: Short-range Fock (Hartree–Fock) exchange, screened via the error-function kernel $\textrm{erfc}(\mu r)/r$.
• $E_x^{\rm PBE,SR}(\mu)$ and $E_x^{\rm PBE,LR}(\mu)$: PBE exchange, partitioned into short- and long-range components using the same screening parameter $\mu$.
• $E_c^{\rm PBE}$: Full PBE correlation energy (unmodified).
• $a$: Fraction of exact Fock exchange entering the short-range part (standard HSE06: $a=0.25$).
• $\mu$: Range-separation parameter that controls the decay of the short-range kernel (standard HSE06: $\mu=0.20\, \text{\AA}^{-1}$, corresponding to $\omega=0.11\, \text{bohr}^{-1}$).

The underlying kernel splits the Coulomb operator: $$\frac{1}{r} = \frac{\textrm{erfc}(\mu r)}{r} + \frac{\textrm{erf}(\mu r)}{r}$$ This structure confines the computationally demanding Hartree–Fock exchange to small interelectronic distances while recovering the PBE form at long range.

2. Parameter Choices and the Standard vs. Modified HSE06

The canonical HSE06 parameters are $a=0.25$ and $\mu=0.20\,\text{\AA}^{-1}$, optimizing performance for a wide range of semiconductors and insulators. However, recent research has demonstrated that adjusting $a$ can yield better quantitative agreement for specific material classes. For anatase TiO$_2$, reducing $a$ from 25% to 20% (denoted HSE06(20)) brings the DFT band gap (3.34 eV) closer to experiment and ensures that subsequent $G_0W_0$ and Bethe–Salpeter Equation (BSE) corrections yield optical excitation energies, oscillator strengths, and exciton binding energies in near-experimental agreement. The table below summarizes the systematic impact of $a$ on electronic and optical gaps for anatase TiO$_2$ (B et al., 2021):

$a$ (%)	$E_{\rm HSE06}^{I}$ (eV)	$E_{\rm QP}^{I}$ (eV)	$\Delta_{\rm QP}$ (eV)	$E_{\rm QP}^{D}$ (eV)	$E_{\rm opt}^{D}$ (eV)	$E_B$ (meV)
20	3.34	4.10	0.76	4.14	3.911	229±10
25	3.68	4.17	0.49	4.21	3.949	261±10
30	4.01	4.25	0.24	4.29	3.992	298±10

Lowering $a$ leads to a smaller DFT gap but larger $GW$ corrections, and ultimately to exciton binding energies and optical gaps that more closely match experimental benchmarks.

3. Physical Principles and Rationale for Range Separation

The rationale for range separation in HSE06 is to combine the strengths of Hartree–Fock exchange for short-range electronic correlations, especially essential in localized and correlated electrons, with the computational efficiency and screening properties of GGA functionals at longer range. The screening parameter $\mu$ effectively controls the spatial extent over which exact exchange is included. The PBE0 functional, which lacks range separation, mixes Fock exchange globally and converges more slowly in $k$-space and is more sensitive to periodic image interactions in extended systems (Schlipf et al., 2011).

The choice of $a$ can be empirically tied to the inverse dielectric constant, $\epsilon_\infty^{-1}$, for semiconductors and insulators, but in practice is often optimized by direct fitting to reference properties, notably the experimental band gap (Moussa et al., 2012, B et al., 2021). For transition metal oxides, such calibration is essential to prevent systematic overestimation or underestimation of band gaps and derived quantities.

4. Performance Across Materials Classes

The HSE06 functional, both in its standard and material-specific forms, has been extensively benchmarked:

• Semiconductors and Insulators: HSE06 systematically corrects the severe band-gap underestimation of standard GGA functionals (e.g., PBE) for prototypical semiconductors (Si, GaAs) and insulators (MgO), with typical mean absolute errors (MAEs) of 0.32–0.5 eV, versus 1–2 eV for GGA (Schlipf et al., 2011, Moussa et al., 2012, Karlický et al., 2012).
• Transition Metal Oxides: For anatase TiO$_2$, the modified HSE06(20) yields DFT and quasiparticle gaps aligning with experimental optical band edges and exciton properties (direct BSE gap and binding energy) to within 0.1–0.2 eV (B et al., 2021).
• Defects and Strongly Correlated States: In transition metal defects in Si or SiC, HSE06 reduces but does not eliminate self-interaction errors for localized $d$-states (as quantified by deviations from generalized Koopmans' theorem); explicit corrections may be required for quantitative accuracy in charge-transition levels and defect states (Ivády et al., 2013).
• Rare-Earth Compounds: In monopnictides (e.g., LaAs, GdSb), HSE06 yields carrier densities, lattice constants, and band topologies closely matching experiment, outperforming GGA which overestimates band overlap and $n$ by factors of 2–4 (Khalid et al., 2019).

5. Computational Implementation and Efficiency

HSE06's screening mechanism enables efficient implementations in both plane-wave and localized-orbital codes for periodic solids:

• In FLAPW codes, the screened non-local exchange is evaluated using auxiliary mixed-product bases and fast real/reciprocal space algorithms, allowing favorable scaling ($\sim N^2$–$N^3$) for 100-atom supercells. The screened LR term converges rapidly in $k$-space due to the analytic decay of the error function kernel (Schlipf et al., 2011).
• In plane-wave codes with pseudopotentials (e.g., VASP + PAW), typical settings are a plane-wave cutoff of 400–800 eV, $k$-meshes converged for the chosen supercell, and explicit treatment of SOC and $4f$ states for heavy or correlated elements (Khalid et al., 2019, Kaewmaraya et al., 2012).
• With localized orbital basis sets (GAUSSIAN, CRYSTAL), HSE06 achieves rapid gap convergence and robust band structure predictions for 2D derivatives, with the computational cost only $\sim$5× that of GGA and lower parallel overhead than global hybrids (Karlický et al., 2012).

6. Guidelines for Parameter Selection and Material-Specific Optimization

Parameter selection in HSE06 is context-dependent:

• For band-gap–driven applications, set $a$ such that the DFT fundamental gap is within $\sim 0.1$ eV of the low-temperature experimental value (B et al., 2021).
• For TM oxides: $a \approx 0.20$ is optimal for anatase TiO$_2$. Keeping $\mu=0.20\,\text{\AA}^{-1}$ maintains the correct balance between short-range accuracy and computational efficiency.
• For rare-earth and correlated semimetals, include 4$f$ electrons in the valence and tune $a$ and $\mu$ to ensure correct Fermi surface topology and carrier concentration (Khalid et al., 2019).
• For defects, monitor deviations from the generalized Koopmans' condition as a diagnostic for residual self-interaction and add an occupation-dependent $V_w$ term to enforce it, if necessary (Ivády et al., 2013).

The following table summarizes recommended $a$ and $\mu$ values for representative systems:

System	$a$ (Fock exchange)	$\mu$ (Å⁻¹)	Rationale
Standard HSE06	0.25	0.20	General semiconductors
Anatase TiO$_2$ (HSE06(20))	0.20	0.20	TM oxides, gap calibration
III–V semiconductors	0.25–0.30	0.20	Band-gap tuning
Rare-earth pnictides	0.25	0.20	4$f$ valence, SOC

7. Applications and Limitations

The hybrid HSE06 functional is widely used for:

• Accurate prediction of band structures, band edges, and band offsets in heterostructures and alloys (Wadehra et al., 2010).
• Modeling optical gaps and excitonic properties via $GW$ and BSE, starting from a reliable single-particle Hamiltonian (B et al., 2021).
• Addressing self-interaction errors and gap underestimation for systems with localized states (e.g., $d$, $f$, or defect-derived states).
• Structural and optical property determination in complex chalcogenides and mixed-anion compounds (Kaewmaraya et al., 2012).
• Electronic structure calculations in 2D materials, where conventional GGA and meta-GGA fail to capture the correct band-gap trends (Karlický et al., 2012).

However, HSE06 is not free of limitations. For highly correlated or polaronic systems, or where strong inhomogeneity in electronic screening is present, the uniform parametrization of the erfc kernel may not suffice; further corrections or fully nonlocal functionals may be required to achieve quantitative accuracy (Ivády et al., 2013, Moussa et al., 2012). The computational overhead, while reduced compared to global hybrids, remains significant relative to GGA, especially for large supercells or superlattice calculations.

• (B et al., 2021) Quasiparticle electronic structure and optical response ($G_0W_0$+BSE) of anatase TiO$_2$ starting from modified HSE06 functionals
• (Schlipf et al., 2011) The HSE hybrid functional within the FLAPW method and its application to GdN
• (Kaewmaraya et al., 2012) Hybrid density functional study of electronic and optical properties of phase change memory material: $\mathrm{Ge_{2}Sb_{2}Te_{5}}$
• (Karlický et al., 2012) Band gaps and structural properties of graphene halides and their derivates: A hybrid functional study with localized orbital basis sets
• (Ivády et al., 2013) The role of screening in the density functional applied on transition metal defects in semiconductors
• (Moussa et al., 2012) Analysis of the Heyd-Scuseria-Ernzerhof density functional parameter space
• (Khalid et al., 2019) Hybrid functional calculations of electronic structure and carrier densities in rare-earth monopnictides
• (Wadehra et al., 2010) Band offsets of semiconductor heterostructures: a hybrid density functional study