Large-scale system analysis#
Calculation Details#
- AOF-DFT and KS-DFT calculations are performed on an fcc Pt supercell, and the computational times are compared.
Target Structure#
 fcc Pt (a=3.9Å) and its supercell |
Calculation Conditions#
The same values were used for all parameters common to AOF-DFT and KS-DFT.
| Parameter | Value |
|---|---|
| Pseudo-atomic orbital | Pt6.125_as.pao, s1p1d1 |
| Pseudopotential | Pt_PBE26_as.vps |
| Functional model (AOF-DFT only) | TF1/5vW, LKT, AS26(Crystal26-0) |
| Energy cutoff | 150 Hartree |
| k-point density (KS-DFT only) | 3.5 point/Å⁻¹ |
| Smearing width | 0.02 Hartree |
| Convergence threshold | 1.0×10⁻⁶ Hartree |
| Number of parallels* | 56 |
- Machine Specs
- CPU: Intel Xeon Gold 6330 (2.00GHz), 2 × 28 cores
- Memory: 8GB DDR4-3200 RDIMM×16 chips (=128 GB)
* AOF-DFT stands for OpenMP parallel, KS-DFT is MPI parallel.
Calculation Results#
- In contrast to KS-DFT, which exhibits 𝑂(𝑁3) scaling, AOF-DFT achieves 𝑂(𝑁) scaling for all functionals.
|