[Pw_forum] About 'qcutz', 'q2sigma' and 'ecfixed' in 'vc-relax'

[Madan Mithra. L. M]

Dear All,

I have some doubts regarding setting the parameter values of 'qcutz', 'q2sigma' and 'ecfixed' in 'vc-relax'. I went through the original paper of M. Bernasconi et al, J. Phys. Chem. Solids 56, 501 (1995) and arrived at these conclusions : 

To mimic a constant cut-off run at 40Ry, they used the following settings for the vc-relax :

A = 70Ry , sigma = 8Ry and E_0 = 35 Ry

ecfixed = 2*E_0 = 70 Ry (ie, 1.75 times Ecut(constant) = 40Ry)
q2sigma = 2*sigma = 16 Ry (ie, 0.4 times Ecut(constant) = 40Ry)
qcutz = A = 70Ry (ie, 1.75 times Ecut(constant) = 40Ry)
ecutwfc = 35 Ry (0.875 times Ecut(constant) = 40Ry)

Can we generalize these findings and arrive at a conclusion that for an arbitrary value of Ecut(constant), one has to set :

ecfixed (and qcutz) at 1.75*Ecut(constant)
q2sigma at 0.4*Ecut(constant)
ecutwfc at  0.875*Ecut(constant)

I would also like to know whether these is any systematic way of improving the quality of vc-relax calculations by varying the three parameters (qcutz, ecfixed and ecutwfc).

Thank you

- Madan

                         ----------------------------------------
                         Madan Mithra .L.M
                         Senior Research Fellow
                         Dept.of Physics
                         Indian Institute of Science
                         Bangalore - 560 012
                         INDIA
                         .........................................
                         e-mail : mit at physics.iisc.ernet.in
                                : snd2mra at yahoo.com
                         Phone  : 91-80-22932313    (Lab)
                         FAX    : (080) 360 2602
                         -----------------------------------------