epw.x stops when DEALLOCATE rwork

[hellolori]

Dear all,

I use epw.x (v5.2) with QEv6.5 to calculate electron-phonon coupling of a 2D material including spin-orbit coupling. Recently, I run into a problem with strange notification (see below). I look into the code and find that that epw.x stops at the 168th line of the "wan2bloch.f90", which reads "DEALLOCATE(rwork, STAT = ierr)". But I cannot solve the problem. Could anyone help me?

The last several lines of the standard output:

Code: Select all

    1 *** Error in `epw.x': free(): invalid next size (normal): 0x000000000f43a280 ***
    2 *** Error in `epw.x': free(): invalid next size (normal): 0x00000000056a9e20 ***
    3 *** Error in `epw.x': double free or corruption (!prev): 0x00000000062aae20 ***
    4 *** Error in `epw.x': free(): invalid next size (normal): 0x0000000006122c90 ***
    ...

The last several lines of the outfile:

Code: Select all

594      ===================================================================
595      irreducible q point #    7
596      ===================================================================
597
598      Symmetries of small group of q:  6
599
600      Number of q in the star =    2
601      List of q in the star:
602           1   0.333333333   0.577350269   0.000000000
603           2  -0.333333333  -0.577350269   0.000000000
604      Dyn mat calculated from ifcs
605
606         q(   35 ) = (   0.3333333   0.5773503   0.0000000 )
607         q(   36 ) = (  -0.3333333  -0.5773503   0.0000000 )
608
609      Band disentanglement is used: nbndsub =   40
610      Use zone-centred Wigner-Seitz cells
611      Number of WS vectors for electrons      157
612      Number of WS vectors for phonons       43
613      Number of WS vectors for electron-phonon       43
614      Maximum number of cores for efficient parallelization      903
615      Results may improve by using use_ws == .TRUE.
616
617      Velocity matrix elements calculated
618
619
620      Writing Hamiltonian, Dynamical matrix and EP vertex in Wann rep to file
621
622      ===================================================================
623      Memory usage:  VmHWM =      2328Mb
624                    VmPeak =      6426Mb
625      ===================================================================
626
627      Using uniform q-mesh:  120 120   1
628      Size of q point mesh for interpolation:      14400
629      Using uniform k-mesh:  120 120   1
630      Size of k point mesh for interpolation:      28800
631      Max number of k points per pool:              226
           [the end] 

I find that the epw.x stops at the 168th line of the file "wan2bloch.f90", by printing "hello".

Code: Select all

 
 163     ! Diagonalization routine
 164     cz(:, :) = chf(:, :)
 165     CALL ZHEEVD('V', 'L', nbnd, cz, nbnd, w, cwork, 2 * nbnd + nbnd**2, &
 166             rwork, 1 + 5 * nbnd + 2 * (nbnd**2), iwork, 3 + 5 * nbnd, info)
 167     !
 168    DEALLOCATE(rwork, STAT = ierr)
 169     IF (ierr /= 0) CALL errore('hamwan2bloch', 'Error deallocating rwork', 1)
 170     DEALLOCATE(iwork, STAT = ierr)
 171     IF (ierr /= 0) CALL errore('hamwan2bloch', 'Error deallocating iwork', 1)
 172     DEALLOCATE(cwork, STAT = ierr)
 173     IF (ierr /= 0) CALL errore('hamwan2bloch', 'Error deallocating cwork', 1)
 174     !

[hlee]

Dear hellolori:

I think that I already fixed this bug in the development version of EPW:
Please check this page:
https://gitlab.com/QEF/q-e/-/merge_requests/824

If you still have a problem after applying this fix, please get back to us with all inputs including pseudopotentials.

Sincerely,

H. Lee

[hellolori]

Dear Lee,

The method you provided seems valid, and the calculation can proceed using the recompiled program. Thanks.

Best regards!