Dear all, I am using QE version 7.3.1 on an Ubuntu 22.04 environment. I am also using Intel oneAPI for the compiler, MPI, etc.
I am performing neb.x calculations under a homogeneous finite electric field based on the modern theory of polarization (lelfield = .true.). When calculating images one at a time using commands like “mpirun -np 16 -inp xxxxx.in > xxxxx.out”, the calculation proceeds without issues. However, when calculating images simultaneously using commands like “mpirun -np 48 -ni 3 -inp xxxxx.in > xxxxx.out”, the following problem occurs. Below is an excerpt from PW.out in image1 , ..... End of self-consistent calculation k = 0.0000 0.0000 0.0000 ( 13805 PWs) bands (ev): -46.0910 -46.0650 -45.9979 -45.9869 -45.9276 -45.9010 -45.8950 -45.8401 -22.7314 -22.6129 -22.6032 -22.5686 -22.5587 -22.5497 -22.5272 -22.5158 ..... 8.8268 8.9163 8.9827 9.3431 9.4554 9.4892 9.5100 9.5884 highest occupied level (ev): 9.5884 ! total energy = -2303.12093350 Ry estimated scf accuracy < 0.00000735 Ry The total energy is the sum of the following terms: one-electron contribution = -624.09930643 Ry hartree contribution = 489.01939044 Ry xc contribution = -463.47927866 Ry ewald contribution = -1696.38943348 Ry convergence has been achieved in 11 iterations Writing all to output data dir xxxxx/xxx_1/xxx.save/ : XML data file, charge density, pseudopotentials, collected wavefunctions negative rho (up, down): 5.389E-01 0.000E+00 OH BOY -2086101920 OH BOY -1107549545 OH BOY 189234244 OH BOY -1106211815 OH BOY -1813111808 OH BOY -1114285954 ..... Reviewing the source code, it appears there may be issues with the values in `aux_rcv_ind(ig,iproc)` or `aux_g_mpi_ind(ig,mpime+1)` within `forces_bp_efield.f90`, but I couldn't understand it further. I would appreciate your advice on this issue. Masahito YOSHINO JAPAN
_______________________________________________________________________________ The Quantum ESPRESSO Foundation stands in solidarity with all civilians worldwide who are victims of terrorism, military aggression, and indiscriminate warfare. -------------------------------------------------------------------------------- Quantum ESPRESSO is supported by MaX (www.max-centre.eu) users mailing list users@lists.quantum-espresso.org https://lists.quantum-espresso.org/mailman/listinfo/users
