I am trying to run an NBO6 calculation on a molecular dimer of about 80 atoms.
The test in the manual for NBO runs fine, as does an H2 case.
For this however, despite setting up as closely as possible to the cases that work, nothing happens past
Can you please attach the $molecule also? The load balance issue is likely a function of NBO6 running as a serial program regardless of how many threads are used by Q-Chem.
$molecule
0 1
C -4.55889600 1.76177400 0.32729100
C -4.21009000 0.41020200 0.59884400
C -3.60795300 2.79999200 0.62472800
C -5.79793500 2.09036700 -0.23806300
C -2.95958100 0.05271700 1.12020300
C -5.16793600 -0.62255700 0.32403800
C -2.80783100 3.67967900 0.89377000
C -6.77564700 1.06744300 -0.49181000
C -6.15078500 3.43038700 -0.57086300
H -2.34505200 0.73989700 1.34694000
C -2.59064600 -1.25986500 1.31439600
C -6.45640600 -0.27045800 -0.18432600
C -4.78693700 -1.95120700 0.55872500
Si -1.62714400 4.99618200 1.41673600
C -8.03639100 1.43979200 -1.02567100
C -7.35698000 3.73747300 -1.09607400
H -5.52208000 4.12498800 -0.41990700
C -1.29693900 -1.62554600 1.78198400
C -3.53165600 -2.30299900 1.00642900
C -7.43464800 -1.30823800 -0.36970200
H -5.42280800 -2.63910100 0.40404000
H -2.35247800 6.23516600 1.79114500
H -0.67159500 5.24474700 0.32767500
H -0.95432900 4.45467000 2.62544300
H -8.69483000 0.77111000 -1.17439400
C -8.32038000 2.72257500 -1.32744700
H -7.56110100 4.63952900 -1.31297200
H -0.67500700 -0.94615100 2.01679300
C -0.94074700 -2.92461100 1.89546800
C -3.12018200 -3.64362000 1.15284200
C -8.25267800 -2.19434900 -0.47927200
H -9.16749000 2.94748300 -1.69450200
H -0.06746200 -3.14783300 2.19662500
C -1.85845900 -3.95886600 1.57004400
H -3.73394000 -4.34290000 0.95782400
Si -9.47285800 -3.57473100 -0.51673600
H -1.59476400 -4.86875000 1.64178300
H -10.80004100 -2.98935900 -0.83027900
H -9.47099700 -4.20443100 0.82726000
H -9.04937700 -4.51634100 -1.58343700
H -0.49424000 -0.20562800 -0.80835500
C 0.40597900 0.08856800 -0.73440900
C 0.69067100 1.47704700 -0.68178400
C 1.40558700 -0.81848100 -0.67869100
C 1.96541000 1.90186100 -0.57304600
H -0.01807400 2.10852500 -0.72448500
H 1.19388500 -1.74261200 -0.71982300
C 2.76964100 -0.42349400 -0.55931900
C 3.04563700 0.98567500 -0.49037900
H 2.14357600 2.83489800 -0.55080900
C 3.81693000 -1.35336000 -0.52253600
C 4.38290500 1.41178600 -0.36159800
C 3.53319200 -2.75882600 -0.64425600
C 5.16489600 -0.93573800 -0.34722100
C 5.44978400 0.46743400 -0.24964700
C 4.69681800 2.81523200 -0.35118300
C 3.29674300 -3.94877400 -0.76484600
C 6.22372100 -1.84937600 -0.26038400
C 6.77993600 0.85422000 -0.03367800
C 4.99143700 3.98944300 -0.37946000
Si 2.98141000 -5.74159400 -1.05990800
H 6.04057500 -2.77400100 -0.37351000
C 7.51992300 -1.45333200 -0.01603900
C 7.80810700 -0.05086500 0.11999700
H 6.98154900 1.78114100 0.00663800
Si 5.52629400 5.74704600 -0.52248600
H 1.79328900 -5.84842900 -1.94283000
H 2.78760200 -6.49110900 0.20601300
H 4.18770400 -6.23436300 -1.77394000
C 8.58444200 -2.38568500 0.13744500
C 9.12791800 0.33680800 0.42954200
H 4.32931400 6.55104900 -0.87371800
H 6.54813100 5.81327500 -1.59687700
H 6.06593200 6.15856800 0.79793400
H 8.41402400 -3.31360600 0.02154900
C 9.83299400 -1.96991600 0.44659300
C 10.11801100 -0.58740800 0.60538800
H 9.33329900 1.26087100 0.51700900
H 10.52706500 -2.60941200 0.55822200
H 10.99604300 -0.30596300 0.83433900
$end
Two things:
(1) I am able to reproduce this crash with the following (much shorter and simpler) sample job. Seems like a bug, I will look into it.
(2) For your job, the SCF has trouble converging unless you set THRESH=12 to deal with the linear dependencies, in which case DIIS (default SCF algorithm) converges in 9 cycles. I would not use SGM-LS unless you are trying to locate a non-Aufbau determinant.
Sure, for non-Aufbau thatâs a good algorithm (although I might plug my own studentâs STEP algorithm also), but in any case I would set THRESH=12 because otherwise there are linear dependency issues. A warning to that effect gets printed in the output file and in general for medium to large molecules you should take that warning seriously.
As for the specific NBO issue, the problem on my end is that NBO6 has to be downloaded and installed (whereas NBO5 ships with Q-Chem). It appears that the Weinhold group stopped supporting NBO6 some time in 2020, so I cannot download it to try, and I also think that we (at Q-Chem) need to update the manual to reflect this fact. I will post a bug ticket about that.
As for your calculation, can you get what you want from NBO5? To run NBO5 you just need NBO=TRUE so delete the RUN_NBO6 keyword from your $rem.
Thanks again. I think what I want to do at the moment is fully implemented in NBO5, assuming no bug fixes etc. in NBO6.
NBO changed their interface to ESS codes in NBO6, so I assume that NBO7+ would work the same unless one of the newer NBO methods needed some additional data structure that wasnât enabled in the NBO6 interface.
Anyway, looks like NBO5 ran OK to completion. Regardless, I hope the NBO6/7 interoperability with Q-Chem can get fixed, there are a number of intriguing developments in the NBO code. Thanks for your help!
My NBO contact suggests that NBO7 is quite a bit different, may be why the Weinhold group is no longer willing to provide backwards support. I put in an inquiry to see if we can get this done.
Glad to hear youâre able to use the NBO5 interface! If you want to use NBO6 in the future (or if anyone else is looking in this thread after encountering a similar problem), this hanging behavior is a recently-discovered issue with running the NBO6 interface in parallel in Q-Chem 5.4.1, and is related to a problem with Intel 17 compilers. I have submitted a ticket about it in our system, but for now, the best work-around is to manually set the environment variable âKMP_INIT_AT_FORK=FALSEâ.
I have not tried but Q-Chem staff tell me that the NBO6 interface can also run NBO7, need to set both NBO = TRUE and RUN_NBO7 = TRUE in the $rem section. Advice via Eric Glendening is: âask the user to set the NBOEXE environment variable to nbo7.i4.exe rather than to nbo6.i4.exeâ. Note that both NBO6 and NBO7 need to be obtained separately, they do not come with Q-Chem.