Open kirk0830 opened 5 days ago
I confirm this with tests by testing with both MGGA_X_SCAN+MGGA_C_SCAN
and MGGA_X_SCAN+MGGA_C_SCAN_RVV10
, test results are below
------------------------------------------------------------
+ ABACUSTEST
+ version: v0.4.17
+ GITHUB: https://github.com/pxlxingliang/abacus-test/tree/develop
+ BohriumApp: https://bohrium.dp.tech/apps/abacustest
------------------------------------------------------------
Read the parameters from /lustre/home/2201110432/example/abacus/param.json
Handle no-rvv10
Handle rvv10
Write the results to metrics.json
version natom kpt \
no-rvv10 v3.7.5(732c2b877 (Mon Sep 9 14:41:02 2024 +0800)) 120 [3, 1, 2]
rvv10 v3.7.5(732c2b877 (Mon Sep 9 14:41:02 2024 +0800)) 120 [3, 1, 2]
ibzk element_list nelec \
no-rvv10 4 [H, H, C, C, C, C, C, C, C, C, C, C, C, C, C, ... 1436.0
rvv10 4 [H, H, C, C, C, C, C, C, C, C, C, C, C, C, C, ... 1436.0
normal_end converge scf_steps relax_converge relax_steps \
no-rvv10 True True 112 None 1
rvv10 True True 113 None 1
largest_gradient total_mag absolute_mag energy \
no-rvv10 None None 216.08 -263726.421731
rvv10 None None 216.08 -263726.421731
force \
no-rvv10 [0.031132097, 0.1075834643, 0.0308904571, -0.0...
rvv10 [0.0308552057, 0.1048753775, 0.0307628442, -0....
stress band_gap \
no-rvv10 [-48.0553958072, -3.1049321799, -2.9041373665,... 0
rvv10 [-48.0655416562, -3.0936104961, -2.9052967238,... 0
efermi ncore total_time scf_time force_time stress_time \
no-rvv10 5.702932 16 6357.64 6233.13 None 111.642774
rvv10 5.702932 16 6398.13 6272.98 None 112.167654
INPUT/ks_solver INPUT/mixing_beta INPUT/scf_thr \
no-rvv10 genelpa 0.4 1.000000e-07
rvv10 genelpa 0.4 1.000000e-07
and the energy table in running_scf.log
from SCAN+rvv10 do not have any vdw correction information
It seems that rVV10 did not affect DFT calculation
Attachments:
For users who work in chemistry, they often care more about dispersion, which will play a crucial role when evaluating weak interaction between molecules. However, for functionals like BEEF-vdw, SCAN_rVV10, etc, involving dispersion correction term distinct with Grimme-D2/D3 series, need additional implementation of nonlocal dispersion correction support. Presently the use like
will not trigger any warnings from ABACUS, but there will be dispersion correction term in total energy absent, thus user cannot get correct results!