You mean you want to find the optimized lattice constant?

If yes? Than do some more calculations by varying a, b and c lattice parameters and compare the energies,

After completing Relax calculations, the structure had optimized. Why do we need to vary lattice parameters moreover the force on each atom around 0.0001 eV .

If you are working with vasp, you can change parameter in the second line of POSCAR file (0.2-1%), do ~7 points and extrapolate the min.

or simply use ISIF=3 and 4.

The SCF calculation is used to compute the electronic groundstate energy for the lattice and atomic coordinates you give. In order to predict the lattice itself, you need to instruct your DFT program to perform a variable-cell geometry optimisation, which will change the lattice to reduce the stress, and move the atoms to reduce the forces, until both are small (you can specify how small you want them to be).

The code I use is CASTEP, and there you would set "task : geometryoptimisation" in the param file, and make sure "fix_all_cell : false" in the cell file. I'm not too familiar with PWSCF, but I understand recent versions (>4.2) can perform variable-cell optimisations; check the documentation at http://www.quantum-espresso.org. The only catch is that I think PWSCF performs the optimisation with a fixed plane-wave basis, so once the geometry has been optimised you may need to start a fresh calculation from that final geometry, to ensure all the relevant plane-waves have been included.

The scf calculation is done, when you have the optimised cell parameter, to compute ground state energy which then you can calculate the density of state for exemple. But to get the cell parameter you must done a geometric optimisation, with PWscf you can do it by modification of the calculation in the &control part.

for SCF calculation:

calculation='scf'

for geometrique optimisation OPT:

calculation='vc-relax' ===> that will relax the cell and the atom position.

I have done relax as well scf calculations.

Now, Can I say the parameters from theoretically?

The accuracy depends on the number of the k-point grid and the conditions of the convergence i.e. Force, energy and the wave function. All of this criteria you can controle it from the input file.

Thank you to all..