• Multiwfn forum

    Multiwfn official website: http://www.shanxitv.org/multiwfn. Multiwfn forum in Chinese: http://bbs.keinsci.com/wfn

    You are not logged in.

    #1 2024-03-05 12:03:42

    valeriy.shendrikov
    Member
    Registered: 2024-03-05
    Posts: 2

    Segmentation Fault upon LIBRETA ESP Calculation

    Dear Tian Lu,

    I have run into a problem with ESP calculation by LIBRETA on my system of interest (PO4 3-) in some bases. For example, in ma-def2-QZVP it works particularly well, but switching to aug-cc-pV5Z results in crash (SEGFAULT) upon LIBRETA initialization. At the same time, old slow algorithm works well. I use Linux Multiwfn 3.8 dev version from March 1 2024, numthreads=8, OMP_STACKSIZE=400MB (I tried various), RAM=32GB. Could you please advise how to solve this problem?

    Best regards,
    Valeriy

    Last edited by valeriy.shendrikov (2024-03-05 12:11:12)

    Offline

    #2 2024-03-05 17:14:13

    sobereva
    Tian Lu (Multiwfn developer)
    From: Beijing
    Registered: 2017-09-11
    Posts: 2,013
    Website

    Re: Segmentation Fault upon LIBRETA ESP Calculation

    Dear Valeriy,

    I don't find this problem on my machine

    My Gaussian input file for generating .fchk file:

    #p hf/aug-cc-pv5z
    
    Title Card Required
    
    -3 1
     P                  0.00000000    0.00000000    0.00000000
     O                 -0.00000000    0.00000000    1.71000005
     O                 -0.00000000   -1.61220351   -0.57000002
     O                  1.39620919    0.80610175   -0.57000002
     O                 -1.39620919    0.80610175   -0.57000002

    Calculation of RESP charge is very normal using latest version of Multiwfn 3.8(dev) on RockyLinux 9: (OMP_STACKSIZE=5000M, with 96 cores)

     Atomic radii used:
     Element:O      vdW radius (Angstrom): 1.400
     Element:P      vdW radius (Angstrom): 1.800
     Number of MK fitting points used:      6013
     Calculating ESP at fitting points, please wait...
     Note: Albeit current file type is fch/fchk/chk and "cubegenpath" parameter in settings.ini has been defined, the cubegen cannot be found, therefore electrostatic potential will still be calculated using internal code of Multiwfn
    
     Initializing LIBRETA library (fast version) for ESP evaluation ...
     LIBRETA library has been successfully initialized!
    
     NOTE: The ESP evaluation code based on LIBRETA library is being used. Please in your work cite both Multiwfn original paper (J. Comput. Chem., 33, 580-592 (2012)) and the paper describing the efficient ESP evaluation algorithm adopted by Multiwfn: Phys. Chem. Chem. Phys., 23, 20323 (2021)
    
     Progress: [##################################################]   100.0 %     \
     Calculation of ESP took up wall clock time         1 s
    
     No charge constraint is imposed in this stage
     No atom equivalence constraint is imposed in this fitting stage
    
     **** Stage 1: RESP fitting under weak hyperbolic penalty
     Convergence criterion:  0.0000010000
     Hyperbolic restraint strength (a): 0.000500    Tightness (b): 0.100000
     Iter:   1   Maximum charge variation:    2.2499461951
     Iter:   2   Maximum charge variation:    0.0623520725
     Iter:   3   Maximum charge variation:    0.0000730711
     Iter:   4   Maximum charge variation:    0.0000000837
     Successfully converged!
    
     **** Stage 2: RESP fitting under strong hyperbolic penalty
     Stage 2 of standard RESP fitting is skipped since no atom needs to be fitted
    
       Center       Charge
         1(P )   2.3123714224
         2(O )  -1.3278780518
         3(O )  -1.3283076594
         4(O )  -1.3278780518
         5(O )  -1.3283076594
     Sum of charges:  -3.0000000000
     RMSE:    0.005017   RRMSE:    0.010958

    Also I can normally perform the calculation using Windows version of Multiwfn 3.8(dev)

    Best,

    Tian

    Offline

    #3 2024-03-12 12:52:30

    valeriy.shendrikov
    Member
    Registered: 2024-03-05
    Posts: 2

    Re: Segmentation Fault upon LIBRETA ESP Calculation

    Dear Tian,
    Thank you very much for your reply. Sorry, but I forgot to clarify an important detail, the calculation was carried out with the Orca 5.0.3 program. I suspect that the problem may be in the gbw-files it produces, considering that the problem is not observed with Gaussian.

    Offline

    Board footer

    Powered by FluxBB

    久久精品国产99久久香蕉