Dear Sir
I'm beginner for Quantum Espresso?
I conducted this calculation using Quantum Espresso, but I observed that the final total energy is more than the etot_conv_thr but the total force is less than the forc_conv_thr. In addition I observed one more criterion, i.e., cell < 5.0E-01, but I know neither how it arises nor its meaning (see line 8181 of output the file).
Please share me any Idea on (1) the third criterion (2) shall all these criteria be fulfilled in order to finish the 'relax' or 'VC-relax' calculation?
I used a tetragonal supercell. The input and output files are attached bellow.
Dear Endale Abebe Gudeta,
first, about the criteria, they must all be fulfilled for the calculations to stop. If you do only ionic relaxation, only etot_conv_thr, forc_conv_thr are considered. If you also relax the cell, another criterion is considered: press_conv_thr (In the output file, it is written 'cell'). So in your case, all three criteria have to be met for the calculations to stop.
This is the case if you look at line 8181 of your output file. Then, after the cell+ion relaxation, Pwscf makes another self-consistent calculations (only electrons) with the new cell parameters. The result at the end appears to be not well converged (maximum force is high for instance). This is because you start with starting cell parameters too far from the converged solution. As a consequence, your final energy and forces are not well converged. Now you should restart your calculation (vc-relax), but with the final cell parameters in your output file (line 8267-8269). This will greatly improve your convergence
Best regards
L.P
well,you can attempt to change the electrictron minxing ridio or the converge method..
Thank You Laurent Pizzagalli
After taking the final cell parameters for 'v-c relax' I'm observing better convergence. But I have some doubts about the following points?
1. Does spin-orbit-coupling (SOC) has any effect on vc-relax or relax calculations?
2. What about the effect of k-points mesh on vc-relax or relax calculations?
3. From this calculation I'm able to find Fermi level and the energy band. Is it necessary to do other scf or non-scf calculation to find these parameters?
Sorry for asking so many questions
Best Regards
Dear Endale Abebe Gudeta ,
regarding the first question, it depends on the materials you are computing, not on the calculation method itself. I am not familiar with SnTe, so I cannot tell you if spin-orbit coupling is weak or not for this material. If it is not, it may change your final output values. I recommend to first check in the literature about this point, then you can make your own calculations to check it.
Best
For the second one,
the k-points mesh is indeed very important. The accuracy of your electronic structure calculation critically depends on it (so also your ionic and cell relaxations obviously). So you should redo your calculations with refined grids (starting with the cell parameters you already obtained with a coarse grid), to attain convergence.
Best
Finally, for the third point, no you do not need to make an extra calculation. At the end of the vc-relax, pwscf makes a final scf calculation. If you are converged at this point, and also if you output the needed information (converged wavefunctions for instance) for computing the band structure (by doing a non-scf calculation), you should not need to make the extra calculation.
One advice: train yourself with a simple system that can be computed fast, like a simple metal, and do all these steps, playing with the different parameters and procedures. It is better than trying to directly tackle more complicated systems if you are not completely familiar with the code and the method.
Best
- Badges
- Science topic
- Similar topics
- Filing