Dear CRYSTAL developers,
I would appreciate your help on the numerical DFT integration in a periodic calculation on an Einsteinium compound.
I am using CRYSTAL23 public 1.0.1 (October 2022), PcrystalOMP, with unrestricted PBEsol and no SOC. The primitive cell contains 2 Es and 6 Cl atoms, with 180 explicit electrons.
The basis sets are Es_ERD_SMALL_CORE_PP_4F_IN_VALENCE_NO_G_2017 and Cl_pob_TZVP_rev2, which were downloaded from the CRYSTAL website. Es uses the scalar ECP60MWB, replacing 60 core electrons per atom. Gaussian exponents, contractions and ECP terms are unchanged; CHE occupations were set for an Es3+/Cl− initial guess.
The analytical/Mulliken electron count is 180, but the DFT parameters report:
ATOM ELECTRONS NET CHARGE R(ANGSTROM)
1 299 QC 36.0000 0.0000 0.00000529
3 17 CL 18.0000 -1.0000 1.81000000
Earlier RADIUS attempts with selectors 299 and 99 did not change the printed Es radius.
I have also performed three controlled GUESSP restarts using the same checkpoint and MAXCYCLE = 1. The following values refer to the initial density, before its first SCF update:
he following values refer to the initial density, before its first SCF update:
Configuration
Earlier XLGRID reference
Numerically integrated electrons
230.5302349807
Configuration
T1: XLGRID, RADIUS 39 1.30 and Cl control 17 1.50
Numerically integrated electrons
225.2419084303
Configuration
T2: dense grid, 4 MPI × 5 OMP
Numerically integrated electrons
183.1163944507
Configuration
T3: identical dense input, 1 MPI × 1 OMP
Numerically integrated electrons
183.1163944507
In T1, the Cl radius correctly changes to 1.50 Å, while the Es radius remains unchanged. Selector 39 was only a diagnostic hypothesis.
T2/T3 use 150 radial points, an unpruned angular grid with Lebedev level 18 (1454 points), TOLLDENS 10 and TOLLGRID 18. Their initial analytical count and spin remain 180 and 8, respectively.
The cycle 0 energy changes from −4481.917602872 Eh with the reference grid to −4251.426787439 Eh with the dense grid. T2 and T3 agree in both numerical electron count and cycle 0 energy.
An ATOSCF warning occurred during the earlier Es atomic guess calculation, but is absent from these GUESSP diagnostics. The SCF cycle-limit termination is intentional.
What I would like to understand:
Is the 299 QC entry and its very small DFT radius expected for Es with this official basis/ECP? Such a small radius (well below the proton radius) looks suspicious
What is the supported way to control the integration radius or grid for this center?
Is this behavior associated with a known limitation of this build, and what numerical settings or correction would you recommend?
Exact inputs and complete outputs are attached. T2 and T3 share the same dense-grid input; the corresponding restart checkpoint is available if needed for reproduction.
Thanks in advance for your time
-MS-
EsCl3_reference_XLGRID.out EsCl3_T3_dense_1MPI_1OMP.out EsCl3_T2_dense_4MPI_5OMP.out EsCl3_T1_radius39.out EsCl3_T1_radius39.d12 EsCl3_dense_grid.d12