How to choose the IterationControl criteria??

[hhspace]

Hello,
         I am confused about the iterationControl criteria and want to know their differences.
         Especially I want to know which one to choose in order to execute an efficient and reliable self-consistent calculation. 

[Anders Blom]

The IterationControlParameters actually do not influence the self-consistent cycle in any way. Perhaps a more proper name for it should have been "convergence criterion parameters", because that's what it's all about: determining the criterion for when the scf cycle should be considered to have converged. The default TotalEnergy criterion is, in basically all cases, most suitable, while you can choose a higher threshold than default for testing purposes (or lower, to make sure you have good quality results). Finally it's sometimes convenient to increase the maximum number of steps if you expect that you need more, to avoid having to restart after 100 steps.

[hhspace]

I'm afraid that I can't understand your explanation very well. So let us take an example.

The default criterion is TotalEnergy (Case 1.) and its corresponding tolerance is 1.0e-5. If we choose TotalEnergyAndDensityMatrix (Case 2.) or even Strict (Case 3.) as our criterion and the tolerance is still 1.0e-5. Of course, we test them in the same system. Could you tell me which one will give a more accurate result, Case 1., Case 2. or Case 3? My intuition tells me Case 3. does it and am I right?

In the mannual, the statement appears for the IterationControl criteria that "When performing calculations using a different criterion than the default one, an accuracy which is equivalent to the one provided by the default choice can be often obtained by increasing the tolerance of one (or even more, in some cases) order of magnitude."
Does that mean the accuracy provided by the tolerance 1.0e-7 or even lower for Case 1 could be equivalent to that provided by the tolerance 1.0e-5 for Case 2. or Case 3.?  

[Nordland]

The default criterion is TotalEnergy (Case 1.) and its corresponding tolerance is 1.0e-5. If we choose TotalEnergyAndDensityMatrix (Case 2.) or even Strict (Case 3.) as our criterion and the tolerance is still 1.0e-5. Of course, we test them in the same system. Could you tell me which one will give a more accurate result, Case 1., Case 2. or Case 3? My intuition tells me Case 3. does it and am I right?

Yes, the Case 3 will give a more accurate result.

Does that mean the accuracy provided by the tolerance 1.0e-7 or even lower for Case 1 could be equivalent to that provided by the tolerance 1.0e-5 for Case 2. or Case 3.?  

Yes ! ( for Case 3 that is )