Ok many thanks for the reply, and sorry for the delay. 1- To check this problem I made a simple scf calculation for H atom in a cubic cell with 8 k points and when I run wfck2r now I get nks = 8, so I do not think it is a bug, I will try to figure out what is wrong with my other calculation but I think there is a problem with my input 2- So you are saying that the wave functions I get are the coefficients u_nk(r) in Psi_nk(r) = u_nk(r) e^ikr . These cannot be used to reconstruct the density in real space because they are on the smooth grid . This smooth grid is always given by the expression r_ijk(:) = (i-1)/nr1x * a_1(:) + (j-1)/nr2x * a_2(:) + (k-1)/nr3x * a_3(:) ? This function u_nk(r) does not seem to be normalized, what is the normalization factor ? are these the same wave functions used in the routine hpsi to compute the product with the local part of the potential ?
I am sorry for all these questions, but before using this object I want to be sure what it is exactly. So I really appreciate your effort in answering these question. Jacopo Simoni Los Alamos National Laboratory, Theoretical Division ---------- Forwarded message ---------- From: Lorenzo Paulatto <[email protected]> Date: 27 September 2017 at 01:36 Subject: Re: [Pw_forum] shape of real space wave functions in wfc2kr To: Jacopo Simoni <[email protected]> > 1- why the number of k points (nks) is 1, even if I have in the restart folder 8 k points ? the fact that in the code there is a cycle over nks suggests that I can have more k points, I do not understand. > It isn't.. can you provide a simple example where you noticed this problem? > 2- in my case dffts%nnr = dffts%nr1x * dffts%nr2x * dffts%nr3x, does this mean that the total number of points in the real grid is equal to the total number of points in the FFT grid ? > in particular I have nr1x=nr2x=nr3x=160, with nnr=4096000, while npwx=256119. The total number of points nnr is considerably higher than the number of plane waves, does this mean that the plane wave grid is supplemented with 4096000-256119=3839881 zeros, am I right ? > Yes, as long as we stick to g-space. Keep in mind that what is really stored i "u" the periodic part of the wavefunction, without the exp(ikr) envelope. Only the plane waves such that |G+k|^2 < ecutwfc (i.e. a sphere) are used for u_k, but more are needed when doing the Fourier transform (because you need a box, not a sphere), and even more when doing charge density the square of wavefunctions is always computed in real space, on a dense grid). Also, npwx is a per-processor maximum. All these values are printed at the beginning: Parallelization info -------------------- sticks: dense smooth PW G-vecs: dense smooth PW Min 530 269 81 8627 3055 539 Max 531 270 82 8628 3056 542 Sum 1061 539 163 17255 6111 1081 "dense" is the charge-density grid "smooth" is the grid of wavefunction squared (4*ecutwfc) "PW" is the grid of wavefunctions > 3- the real space wave function evc_r(:,1) that I obtain in output from the invfft routine can be reconstructed on the unit cell according to the following index notation: > index = i + (j-1) * nr1x + (k-1) * nr2x * nr1x > r_ijk(:) = (i-1)/nr1x * a_1(:) + (j-1)/nr2x * a_2(:) + (k-1)/nr3x * a_3(:) > is this true or I am missing something ? > > I really appreciate any help about these questions. Many thanks in advance. > > Jacopo Simoni > Los Alamos National Laboratory, Theoretical division. > > > _______________________________________________ > Pw_forum mailing list > [email protected] > http://pwscf.org/mailman/listinfo/pw_forum > --
_______________________________________________ Pw_forum mailing list [email protected] http://pwscf.org/mailman/listinfo/pw_forum
