Reproducing InSe monolayer band gap with DMol³ – getting 1.847 eV vs 1.968 eV (Ni–InSe gas sensor paper)

Hi everyone,

I’m trying to reproduce the pristine InSe monolayer calculation from a recent Ni–InSe gas sensor paper that uses DMol³ in Materials Studio. In that work they report an indirect band gap of 1.968 eV for the pristine InSe monolayer.

Using Materials Studio (DMol³) I followed their computational details as closely as I can, but I consistently obtain a band gap of about 1.847 eV for the same system. I’d be very grateful if someone could help me understand whether this ~0.12 eV difference is expected, or if I’m missing some important setting.

System and structure

  • Pristine InSe monolayer
  • 4×4×1 supercell (32 In + 32 Se)
  • In-plane lattice: a = b = 4.07 Å (fixed)
  • Vacuum along z: ≈ 20 Å
  • Monolayer centered in the cell
  • Charge = 0, spin-polarized (spin unrestricted)

I first perform a full geometry optimization of atomic positions (cell fixed), then run a band-structure calculation on the relaxed structure.

DMol³ settings

Setup tab (Geometry Optimization):

  • Task: Geometry Optimization
  • Functional: GGA–PBE
  • Quality: Customized
  • Use Grimme (DFT-D dispersion) checked
  • Spin unrestricted checked, Use formal spin as initial checked
  • Metal unchecked
  • Use symmetry checked
  • Charge: 0

Electronic / k-points / cutoff:

  • Integration accuracy: Fine
  • SCF tolerance: Fine (SCF convergence set to 1×10⁻⁶ Ha)
  • Smearing: 0.005 Ha (Gaussian/Fermi)
  • k-point set: Customized 7×7×1 Monkhorst–Pack
  • Core treatment: DFT Semi-core Pseudopots (DSSP)
  • Basis set: DNP
  • Orbital cutoff radius: 5.0 Å (global)
  • Screening model: None

DFT-D (Grimme) options:

  • DFT-D: Grimme
  • Default C₆ and R₀ for In and Se (as in the Grimme parameters)
  • s₆ = 0.75, d = 20.0

I use the same electronic settings for the band-structure task.
For the band structure, I follow the standard hexagonal path Γ–M–K–Γ with special points (in fractional reciprocal coordinates):

  • Γ (0, 0, 0)
  • M (0, 0.5, 0)
  • K (−0.3333, 0.667, 0)
  • Γ (0, 0, 0)

Result

  • From the band-structure plot, I get an indirect band gap ≈ 1.847 eV.
  • The shape of the bands (VBM/CBM positions) looks qualitatively similar to the figure in the paper, just slightly compressed in energy.

Questions for the community

  1. Is a difference of ~0.12 eV (1.847 vs 1.968 eV) reasonable for DMol³ given different versions / default settings, or does it suggest I’m still missing an important parameter?
  2. Are there hidden defaults (integration grid, smearing scheme, SCF options, occupation method, etc.) that can noticeably affect the gap for this 2D system?
  3. Has anyone here successfully reproduced a band gap close to 1.968 eV for pristine InSe with DMol³? If so, could you share your key settings (especially anything beyond what I’ve listed)?
  4. Any advice on whether I should:
    • relax the lattice as well (instead of fixing a = b = 4.07 Å), or
    • increase k-point density (e.g. 9×9×1) or cutoff beyond 5.0 Å
      to get closer to the reported value?

Any insight from people who regularly use DMol³ for 2D semiconductors would be really helpful. Thanks in advance!