The smearing is strongly connected with the k-points mesh you use in metals, where you need a good sampling of the occupied manifold (essentially of the Fermi surface).  In principle, the smearing should be as small as possible and the k points mesh as large as possible. Then for degauss => zero and kpoints => infty, you should get the right value, for any quantity.

As this is not computationally feasible, the crucial point is to compromise, choosing the right combination of smearing and kpoints (i.e. maximizing degauss and minimizing k-points) AND at the the same time leaving quite unaltered the quantity that you want to calculate. The point is that a very large smearing, in fact, spuriously affects the quantity you want to calculate (total energy or whatever).  

If you are using a large cell with only k=GAMMA, try to use the largest smearing, which always helps convergence, monitoring how strongly this affetcs the quantity that you want to calculate, as Roberto suggests.

In my experience, smearing helps also for insulators or molecules, greatly improving the convergence. 

Best,

Andrea

Hi,

It depends strongly on what you want to calculate. If your quantity is the energy, I would say that a value of 0.022 or 0.011 are fine, so choose 0.022 since it will be faster. On the other hand, if you are interested in another quantity, like for example the band structure, you should check if something changes if you use 0.011 instead of 0.022. 

Lesson is: check the convergence of the quantity you want to calculate. There is not "convergence" a priori. 

I hope this helps. 

Cheers,

Roberto

The smearing is strongly connected with the k-points mesh you use in metals, where you need a good sampling of the occupied manifold (essentially of the Fermi surface).  In principle, the smearing should be as small as possible and the k points mesh as large as possible. Then for degauss => zero and kpoints => infty, you should get the right value, for any quantity.

As this is not computationally feasible, the crucial point is to compromise, choosing the right combination of smearing and kpoints (i.e. maximizing degauss and minimizing k-points) AND at the the same time leaving quite unaltered the quantity that you want to calculate. The point is that a very large smearing, in fact, spuriously affects the quantity you want to calculate (total energy or whatever).  

If you are using a large cell with only k=GAMMA, try to use the largest smearing, which always helps convergence, monitoring how strongly this affetcs the quantity that you want to calculate, as Roberto suggests.

In my experience, smearing helps also for insulators or molecules, greatly improving the convergence. 

Best,

Andrea

Thanks Dr. Andrea and Dr. Roberto for such detailed answer.

The optimized degauss value too can be calculate like ecutwfc and k-point. But to determine optimized degauss first known the converged ecutwfc and k-point (which is there optimized one). Know varies the degauss value. Plot the total energy against the degauss value, the minimum energy is the point(value) of degauss.

cheer,

Akeem

you have to choose the degauss value which gives you the optimal value of total energy.

Good luck

Hello, Roberto D'Agosta Sir, Can you please specify if the value of degauss is in Ry or eV ???

Standard Quantum Espresso units for degauss is Ry and you should use that.

I have perfomed  some convergence study for W (110) +H system. I am attaching the results of that for someone can make use of this methodology later. 

Hi Ajmalghan M. which degauss value did you use?

Hilal Kucuk , Surely it depend on the system.

For W+H system i used Degauss 0.05 with 40Ry for Energy cutoff.

You can understand it from my attached files above.

Ajmalghan M. Thanks a lot