HF calculations, "LAPACK: Routine ZPOTRF failed" for ALGO=All, ALGO=Damped stagnant
Posted: Mon Aug 23, 2010 3:54 pm
Dear admin,
I was running many HF calculations successfully, but recently I run into the "LAPACK: Routine ZPOTRF failed!" message and program stops, after 5 or 9 CGA iterations. The set up that leads to this: Zn vacancy in ZnO, 192 atoms cell, relaxed, wavefunction from GGA at G-point only, then following tags in INCAR for HF run,
ISTART = 1
LHFCALC = .TRUE. ; HFSCREEN = 0.2 ! using HSE functional
AEXX = 0.375
NBANDS = 896
PREC = Normal
ENCUT = 300.0
ISPIN = 2
ALGO = All ; TIME = 0.4
LREAL = Auto
I tried all the remedies suggested at this forum: switched to Damped algorithm that removes the error message, but this algorithm is stagnant and almost never converges no matter what TIME is (tried 0.4, 0.3, 0.1, 0.05, 0.01); tried using lapack supplied with vasp using ALGO=All, still get the error message; tried changing TIME with ALGO=All (0.4, 0.2, 0.1, 0.05) still error message; tried changing charge mixing tags, Kereker, simple straight mixing, Chebyshev, all failed; tried adding more bands (up tp 1024) and increasing energy cutoff, most of the time it does not work either.
Now I am out of things to try. Any help with how to converge HF calculations is greatly appreciated.
I was running many HF calculations successfully, but recently I run into the "LAPACK: Routine ZPOTRF failed!" message and program stops, after 5 or 9 CGA iterations. The set up that leads to this: Zn vacancy in ZnO, 192 atoms cell, relaxed, wavefunction from GGA at G-point only, then following tags in INCAR for HF run,
ISTART = 1
LHFCALC = .TRUE. ; HFSCREEN = 0.2 ! using HSE functional
AEXX = 0.375
NBANDS = 896
PREC = Normal
ENCUT = 300.0
ISPIN = 2
ALGO = All ; TIME = 0.4
LREAL = Auto
I tried all the remedies suggested at this forum: switched to Damped algorithm that removes the error message, but this algorithm is stagnant and almost never converges no matter what TIME is (tried 0.4, 0.3, 0.1, 0.05, 0.01); tried using lapack supplied with vasp using ALGO=All, still get the error message; tried changing TIME with ALGO=All (0.4, 0.2, 0.1, 0.05) still error message; tried changing charge mixing tags, Kereker, simple straight mixing, Chebyshev, all failed; tried adding more bands (up tp 1024) and increasing energy cutoff, most of the time it does not work either.
Now I am out of things to try. Any help with how to converge HF calculations is greatly appreciated.