<div dir="ltr">Dear all, I was trying to perform Vibrational Analysis for my system containing 590 atoms. &VIBRATIONAL_ANALYSIS DX [bohr] 0.04 NPROC_REP 288 FULLY_PERIODIC .TRUE. INTENSITIES .TRUE. &PRINT &PROGRAM_RUN_INFO &END &MOLDEN_VIB FILENAME MOLVIB &END &END PRINT &END VIBRATIONAL_ANALYSIS For this calculation I fixed 324 atoms and let the other 266 to be optimized. &CONSTRAINT &FIXED_ATOMS LIST 1..324 &END FIXED_ATOMS &END CONSTRAINT However instead of 266*6 SCF cycles that I expected cp2k did more than 2500: > grep "SCF WAVEFUNCTION OPTIMIZATION" *-r-*.out | wc -l 2520 At some point the execution reached the WALLTIME 86000 and stopped. What surprised me again that cp2k was still able to compute the normal modes. So my first question would be: can I trust the results of these calculations, taking into account that the last SCF cycles did not converge due to the time limit? In order to narrow down the problem I updated cp2k to the latest trunk version(svn:17932). Then I took a water molecule and slightly simplified the VIBRATIONAL_ANALYSIS section. Then I did the same calculations using 1 and 2 replicas. It turned out that in the first case cp2k did 19 scf cycles, while in the second case 24. So my second question: what cp2k is doing? is it a bug? 1 replica: VIB_no_intens_1replica> grep "SCF WAVEFUNCTION OPTIMIZATION" *-r-*.out | wc -l 19 VIB_no_intens_1replica> grep "VIB| REPLI" out | wc -l 18 2 replicas: ....VIB_no_intens_2replicas> grep "VIB| REPLI" out | wc -l 18 ....VIB_no_intens_2replicas> grep "SCF WAVEFUNCTION OPTIMIZATION" *-r-*.out | wc -l 24 I attached the compressed running directories for the both water cases. I would appreciate any help. Best regards, Sasha </div>