Hse06 Vasp -
If your system has less than 50 atoms and you care about the band gap to 0.1 eV accuracy, pay the cost. If you're studying a metal or a giant interface, stick with PBE+U. Have you had success (or nightmares) running HSE06? Let me know in the comments below. And yes, your SCF will oscillate on the first try—check your mixing parameters.
If you have spent any time running density functional theory (DFT) calculations, you know the drill: PBE (Perdew-Burke-Ernzerhof) is fast, reliable, and often... wrong. It systematically underestimates band gaps, over-delocalizes electrons, and struggles with strongly correlated materials. hse06 vasp
In this post, I’ll break down what HSE06 is, how to set it up in VASP, and when it’s actually worth the pain. HSE06 is a screened hybrid functional. It mixes 25% exact (Hartree-Fock) exchange with 75% PBE exchange at short range, while keeping PBE correlation. If your system has less than 50 atoms
KPOINTS: Automatic generation 0 Gamma 4 4 2 0 0 0 Let me know in the comments below
ALGO = Damped # Damped algorithm (often more stable than Normal) TIME = 0.4 # Mixing time (increase from default 0.1) BMIX = 0.0001 # Small mixing parameter AMIN = 0.01 # Avoid Pulay collapse If still failing, try ALGO = All (fast but memory hungry) or IALGO = 53 (very stable but slow). | System | PBE wall time | HSE06 wall time | Memory | | :--- | :--- | :--- | :--- | | Si (8 atoms, 6x6x6 kpoints) | 2 min | 20 min | 2x | | TiO₂ (12 atoms, 4x4x4) | 5 min | 1.5 hours | 3x | | NiO (8 atoms, 8x8x8) | 3 min | Fails / 4 hours | 5x |
In older VASP versions (pre-6), you needed LHFCALC = .TRUE. and HFSCREEN = 0.2 . In VASP 6+, you can also use HSE06 as a pseudopotential flag, but the manual INCAR approach is safer. Step 2: FFT grids and precision Hybrid functionals are sensitive to the real-space grid. Use high precision:
