Quantum Espresso is your friend and has a calculation mode called vc-relax that perform the optimisation of the unit cell parameters using, if I am not mistaken the Parrinello Raman scheme.

This in the process relaxes also the atomic positions to reduce residual force between the atoms.

Optimising simultaneously the 3 constants would be admittedly impossible without having any idea where to go. Changing them one at a time is similarly difficult as you need to check that the system stays in a local minimum when you change the other parameters.

Clearly this does not offer any potential to study elastic properties, but that was not your question.

I hope this helps,

Roberto

Dear Mohammad Ishfaq ,

From a mathematical point of view, you are searching the local minimum of a 3-variable function. Clearly, you can do that by fixing 2 variables and changing the 3rd to look for a minimum in this third variable. Then you lock another pair and minimise again. When you have finished, you need to start again with minimising the function with respect to the first variable, as you might have moved away from its local minimum.

Unless you are somewhat closer to a true local minimum of your function, this scheme might never converge. So to answer your question, yes you could in principle do the way you suggest but it might be not as efficient as other methods.

To have a feeling of the complexity of the problem, I might suggest reading for example a book like “Numerical Recipes” or on numerical methods as it might provide a glimpse of the different methods and their associated difficulties.

For the elastic coefficients, you do fix 2 variables and change the third (in fact it is a bit more complex than this) but in that case you want 1) to move slightly away from the minimum, 2) you know already you are starting from a local minimum with respect to all the lattice parameters.

Finally, you can again do that with QE — there is a flag, cell_dofree that allows you to specify how you want the unit cell to be optimised. There are different options.

I hope this helps,

Roberto

hello sir.Thanks for ur precious answers.Can u please tell me can i go for the vc relax at the very first step.Then i begin the ecut and k point convergence.

Thanks.OR follow the sequence ecut>>k points>>vc relax.

Moreover,i am using the pseudopotential generated by the QWE input generator.Are those reliable,if not then where can i judge that i am not using the accurate pseudopotentials.Thanks.

In the first lines of your input file for Quantum Espresso, you should see a line

calculation_mode

you have to set it to

calculation_mode=‘vc-relax’

Then, after the electron card, you should add

&cell

/

&ions

/

That means you have leaving QE devise the way for performing the relaxation.

For the pseudopotentials, those generated by Quantum Espresso are quite reliable, but you can also try those from Pseudodojo.

Regards,

Roberto

Hi Sir

how are you and your family

Sir i have an issue to ask

i have compound and i know that it has A type ferromagnetic nature

how can i set the scf for it right from the start

e.g., i have four Mn atoms which are magnetic and render magneisation to it

regards