Dear Gabriele, Thanks very much. I understand better now the idea behind including states that are outside the energy window defined by Ef and Ef+sample_bias. I'm still confused about something, though. Let's say we're simulating a room temperature STM image of a metal or semimetal. Here's my logic-- please correct me if I'm wrong. Typical smearing widths of the order of 0.01 Ry correspond to non-negligible populations of excited states (for Fermi-Dirac smearing, 0.01 is equivalent to ~1500K). It seems to me that when you add in states to the LDOS that are, say, above the Tersoff-Hamann energy window, then you could very well be adding in charge density from states composed of higher-index periodic functions [psi=planewave * periodic function] that really shouldn't contribute much to the LDOS at 300K. The shapes of these higher-index periodic functions could distort the STM image. So, to minimize the distortion, you'd want to run the PW calculation at a rather low smearing width (~0.002 Ry), which of course would require a finer k-point mesh. Does this argument make sense?
Thanks, David At 08:31 AM 12/10/2012, you wrote: >Dear David, > > I don't think the algorithm is wrong, it is > (more or less) consistent with the way the > charge density is computed in presence of a > smearing of the electronic occupations. >The energy window for the integral of the local >density of states is the one prescribed by the >Tersoff-Hamann method, but one also needs to >consider the "tails" of the electronic levels >just above and below that window. The code does >this by including extra states outside the >window, but their charge is weighted with a >"smeared" delta function w0gauss( ) that falls off exponentially or so. >The extend range is defined to spare time by >considering only eigenvalues not too far from the window edges. > >This is not so bad, but in my opinion one should >instead use the wgauss functions (integral of >the smeared delta, or generalized step function, >if you prefer), in order to be consistent with >the charge integration in the rest of the code. Something like: >wg(ibnd,ik) = wgauss(up-et(ibnd,ik)) - wgauss(down-et(ibnd,ik)) >would do the job, consistently with the weights >wg computed in PW/src/gweights.f90, and used in >sum_band.f90 (I am correct, Paolo?). >Probably this solution would give similar results > >HTH > >GS > > >>I have a question about QE's implementation of the the Tersoff-Hamann >>formalism for simulating STM images. If I understand the stm.f90 >>code correctly, the energy sampling window does not range from Ef to >>Ef+sample_bias (which is what Tersoff-Hamann says it should >>be). Rather, the code increases the upper limit by 3*degauss >>(degauss=smearing width) and also decreases the lower limit by >>3*degauss. In the case of metals, the value of degauss is taken from >>the prior PW run. I believe the subsequent lines of code modify the >>weights of the states that are outside the Tersoff-Hamann window. >> >>So, as an example, if a metal has a bias of -0.1 eV and the smearing >>width from the prior PW run was 0.01 Ry (or 0.136 eV), then states >>from -0.5 eV to +0.4 eV (with respect to Ef) are included in >>calculating the LDOS. This strikes me as a rather broad range, even >>if temperature and energy linewidths are considered, and could alter >>the appearance of the computed images. >> >>Why do the STM energy limits take into account the smearing width >>from the PW output? And is it best to use as small a width as >>possible if you intend to run STM simulations? >> >>Thanks, >> >>David Pullman >>Department of Chemistry and Biochemistry >>San Diego State University >>San Diego, CA 92182-1030 > > >? Gabriele Sclauzero, EPFL SB ITP CSEA > PH H2 462, Station 3, CH-1015 Lausanne > > > > > > > >_______________________________________________ >Pw_forum mailing list >Pw_forum at pwscf.org >http://pwscf.org/mailman/listinfo/pw_forum -------------- next part -------------- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20121211/9a876cbe/attachment.html
