[QE-users] convergence problem

[alberto santonocito]

Hello qe users,
I am performing the calculation of the relaxed structure of a supercell of
Fe(CO)5 (Z=4, C2/c).
After the first scf cycle (CG style diagonalization) the new energy of the
system is calculated.
In the next cycle the scf does not converge after 1000 steps.
I believe this comes from incorrect input.
Were could the error in the input be?
The coordinates of the supercell I have extracted from the supplementary
material of the article https://doi.org/10.1038/srep15139.
Here are some parts of the output:
Self-consistent Calculation

     iteration #  1     ecut=    80.00 Ry     beta= 0.70
     CG style diagonalization
     ethr =  1.00E-02,  avg # of iterations =  3.0

     total cpu time spent up to now is     1803.9 secs

    ......

   iteration # 32     ecut=    80.00 Ry     beta= 0.70
     CG style diagonalization
     ethr =  2.04E-09,  avg # of iterations =  3.1

     total cpu time spent up to now is    43842.1 secs

     total energy              =   -1850.88412269 Ry
     Harris-Foulkes estimate   =   -1850.88412271 Ry
     estimated scf accuracy    <       0.00000137 Ry

     iteration # 33     ecut=    80.00 Ry     beta= 0.70
     CG style diagonalization
     ethr =  5.19E-10,  avg # of iterations =  3.1

     total cpu time spent up to now is    45086.7 secs

     End of self-consistent calculation

      ...

Total force =    27.930798     Total SCF correction =     0.002498


     Computing stress (Cartesian axis) and pressure

     Message from routine stres:
     noncollinear stress + GGA not implemented

     BFGS Geometry Optimization

     number of scf cycles    =   1
     number of bfgs steps    =   0

     energy   new            =   -1850.8841224519 Ry

     new trust radius        =       0.5000000000 bohr
     new conv_thr            =       0.0000010000 Ry

    ...
    Writing output data file pwscf.save/
     Message from routine punch:
     wavefunctions written to file
     NEW-OLD atomic charge density approx. for the potential

     total cpu time spent up to now is    45155.5 secs

     Self-consistent Calculation

     iteration #  1     ecut=    80.00 Ry     beta= 0.70
     CG style diagonalization
     c_bands:  4 eigenvalues not converged
     ethr =  1.00E-06,  avg # of iterations = 27.7
     ...
     iteration #999     ecut=    80.00 Ry     beta= 0.70
     CG style diagonalization
     c_bands:  4 eigenvalues not converged
     ethr =  1.04E-08,  avg # of iterations =  8.9

     total cpu time spent up to now is  2033318.7 secs

     total energy              =   -1859.84580191 Ry
     Harris-Foulkes estimate   =   -1859.84718501 Ry
     estimated scf accuracy    <       0.25486994 Ry

     iteration #***     ecut=    80.00 Ry     beta= 0.70
     CG style diagonalization
     c_bands:  2 eigenvalues not converged
     ethr =  1.04E-08,  avg # of iterations =  9.4

     total cpu time spent up to now is  2036333.6 secs

     total energy              =   -1859.85310299 Ry
     Harris-Foulkes estimate   =   -1859.84865768 Ry
     estimated scf accuracy    <       0.25711820 Ry

     End of self-consistent calculation

     convergence NOT achieved after *** iterations: stopping

     I also attach the input.
    Thanks
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20210530/eecf6860/attachment.html>