[Pw_forum] Problem in nscf calculation

李云海 liyunhai1016 at hotmail.com
Tue Nov 11 07:54:50 CET 2014


Dear Pang Rui,

Unconverged eigenvalues may indicate that the Hamiltonian is ill conditioned, i.e. it is no longer a Hermitian matrix and is not physical. This may be due to the incorrect description of the crystal, or improper magnetization settings.

First check your input using XCrysDen or other programms compatible with pwscf. Use fractional coordinates instead of cartesian coordinates, since cif files use fractional coordinates and any coordinate conversion is potential of risk. Also, add celldm(1) = 1.889726125 to the namelist 'SYSTEM' and specify 'CELL_PARAMETERS {alat}', since XCrysDen cannot recognize 'CELL_PARAMETERS {angstrom}'.

In the nscf calculation the starting magnetizaiton is read from previous scf run. Since you mentione that the scf calculation ended successfully while the nscf calculation crashed, there might be inconsistency betwen the scf and nscf calculation. Add the 'prefix' flag to the namelist 'CONTROL' in both scf and nscf input and assign the same value.


If it will not help, check carefully if you have successfully determined the groud state of a magnetic system. It requires a lot of tests to check whether the ground state is non-magnetic, ferro-magnetic or anti-ferro-magnetic by varying starting_magnetizaiton(i).

Finally, it seems that you specified a lot of empty bands in nscf calculation. The 'cg' style diagonalization is more stable in this case. And if you are generating wavefunctions for GW/BSE calculations, be aware that the GW/BSE programm may require norm-conserving pseudo-potentials.

Yunhai Li
Doctoral Student
Department of Physics, Southeast University, P.R.C.

> Date: Mon, 10 Nov 2014 10:10:36 +0800
> From: pang.r at sustc.edu.cn
> To: pw_forum at pwscf.org
> Subject: [Pw_forum] Problem in nscf calculation
> 
> Dear all
> I am trying a nscf calculation on a slab model. The scf converged well.
> But when I switched to nscf calculation, c_bands convergence warning
> appeared, and the calculation ended with an error like this
>      CG style diagonalization
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: 97 eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: 98 eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
>      c_bands: ** eigenvalues not converged
> 
> 
> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
>      Error in routine davcio (10):
>      error while reading from file
> 
> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> 
>      stopping ...
> 
> Neither switching diagonalization methods nor increasing nbnd can solve
> this problem. 
> 
> Could someone help me? The following is the input file
> scf part:
>  &control
>     pseudo_dir = "~/QE5/pseudo"
>     outdir="./",
>     calculation="scf"
>  /
>  &system
>     ibrav=  0, nat=  133, ntyp= 3,
>     ecutwfc = 40.0
>     occupations='smearing', 
>     degauss=0.01,
>     smearing='gaussian'
>     starting_magnetization(1)= 0.0
>     starting_magnetization(2)=0.1
>     starting_magnetization(3)=0.1
>     nspin=2
>     nbnd=800
>     nosym=.true.
> /
>  &electrons
>     mixing_beta = 0.05
>     electron_maxstep=300
>     startingpot='file'
>     startingwfc='file'
>  /
> ATOMIC_SPECIES
>  C 12 C.pbe-n-kjpaw_psl.0.1.UPF
>  Ni 58 Ni.pbe-n-kjpaw_psl.0.1.UPF
>  Ni2 58 Ni.pbe-n-kjpaw_psl.0.1.UPF
> ATOMIC_POSITIONS angstrom
> ......
> K_POINTS {automatic}
> 4 4 1 0 0 0
> CELL_PARAMETERS angstrom
>     9.9673770000000008    0.0000000000000000    0.0000000000000000
>     -4.9836885000000004    8.6320016909999993    0.0000000000000000
>      0.0000000000000000    0.0000000000000000   27.0000000000000000
> nscf part:
>  &control
>     pseudo_dir = "~/QE5/pseudo"
>     outdir="./",
>     calculation="nscf"
>  /
>  &system
>     ibrav=  0, nat=  133, ntyp= 3,
>     ecutwfc = 40.0
>     occupations='smearing', 
>     degauss=0.01,
>     smearing='gaussian'
>     starting_magnetization(1)= 0.0
>     starting_magnetization(2)=0.1
>     starting_magnetization(3)=0.1
>     nspin=2
>     nbnd=800
>     nosym=.true.
> /
>  &electrons
>     mixing_beta = 0.05
>     electron_maxstep=300
>     startingpot='file'
>     startingwfc='file'
>     diagonalization='cg'
>  /
> ATOMIC_SPECIES
>  C 12 C.pbe-n-kjpaw_psl.0.1.UPF
>  Ni 58 Ni.pbe-n-kjpaw_psl.0.1.UPF
>  Ni2 58 Ni.pbe-n-kjpaw_psl.0.1.UPF
> ATOMIC_POSITIONS angstrom
> ........
> K_POINTS {crystal}
> 16
> 0	0	0	0.0625
> 0	0.25	0	0.0625
> 0	0.5	0	0.0625
> 0	0.75	0	0.0625
> 0.25	0	0	0.0625
> 0.25	0.25	0	0.0625
> 0.25	0.5	0	0.0625
> 0.25	0.75	0	0.0625
> 0.5	0	0	0.0625
> 0.5	0.25	0	0.0625
> 0.5	0.5	0	0.0625
> 0.5	0.75	0	0.0625
> 0.75	0	0	0.0625
> 0.75	0.25	0	0.0625
> 0.75	0.5	0	0.0625
> 0.75	0.75	0	0.0625
> CELL_PARAMETERS angstrom
>     9.9673770000000008    0.0000000000000000    0.0000000000000000
>     -4.9836885000000004    8.6320016909999993    0.0000000000000000
>      0.0000000000000000    0.0000000000000000   27.0000000000000000
> -- 
> PostDoc
> Department of Physics, South University of Science and Technology of China
> Shenzhen, Guangdong, PRC, 518500
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20141111/bce1afd4/attachment.html>


More information about the users mailing list