Subhadeep Banerjee a common answer is 'it depends' on algorithms and software.

Nevertheless, we just published a paper where various computational parameters and their influence on thermodynamic properties are tested - https://www.mdpi.com/2073-4352/15/3/274

Here are basic conclusions:

1. Size of the unit cell (supercell) is crucial to obtain correct second order derivatives of the energy, please use at least 1-1.5 nm.

2. Test functionals and dispersion correction schemes (in case of organic crystals) to verify correct simulation of your system

3. Based on the functional chosen and software available one can use finite difference (FD) or DFPT approach. There is no much difference for VASP as only Gamma point calculations (at the center of the Brillouin zone) are supported for DFPT approach.

For other software in principle one can use primitive cell and generic q-points (also known as q-vectors), which can fasten the whole procedure in case of DFPT algorithm.

Since you are going to use VASP, supercell is needed and there is no difference if FD or DFPT scheme is used.

Hope it helps,

Best wishes

Dear, subhadeep Banerjee, If you are using phonopy or Phono3py( Phonon from finite displacement method (FDM)) , then you have to create super cell. As it is based on the above mentioned method it is necessary to create supercell. More the dimension of the supercell more the accurate but expensive your calculations will be. For example, if the primitive cell of a Si (a ~6 ang) contains 2 atoms, then 1*1*1 supercell may have 8 atoms, 2*2*2 may have 32 atoms and so on.. But in some complex structures it 1*1*1 supercell also sufficient to obtain good results with low computational cost.

Thank you

Thank you very much Denis and Niloy for the fruitful discussion.