<br>
<style type="text/css">p { margin-bottom: 0.21cm; }</style>
<p style="margin-bottom: 0cm;">Hi,</p><p style="margin-bottom: 0cm;">I'm also working on vibrational spectra
and I have found similar problem. <br></p><p style="margin-bottom: 0cm;">I'm intersted in trying your NLCC pseudos. Can I ask you to send me the CP2K format so I can do some test?</p><p style="margin-bottom: 0cm;">regards</p><p style="margin-bottom: 0cm;">Daria<br></p><p style="margin-bottom: 0cm;"> <br></p>
<br><br>Il giorno lunedì 13 maggio 2013 11:21:48 UTC+2, Florian Schiffmann ha scritto:<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">Hi,<br><br>as always, the answer to your question and providing help is much easier (and likely to happen) if input files are provided. <br>Cp2k is a bit special for vibrational spectra. The imaginary frequencies are not necessarily connected to a badly converged geometry. Using GTH pseudopotentials the density at the core decays to 0. This is a bit problematic in the xc potentials as it leads to very sharp features. These features can cause trouble in the FFT's (see Quickstep paper, smoothing). <br>Cutting a long story short, if you use the GTH pseudos (and an insufficient CUTOFF) as they are this problem leads to a slight translational invariance of the system as the result of the FFT depends on the relative position of the atom with respect to the grid. This problem is most obvious for 1st row transition metals and the elements B to Na. Also it depends on the functional as the pseudos have different parameters (BLYP pseudos are worst from my experience). I attach a graph with some PBE test on this behaviour and cutoffs needed to achieve a good translational invariance (<1E-5 Ha). Keep in mind using different functionals the necessary cutoffs can be different.<br>At the moment there are two ways out either you have to use a hugh cutoff (600+ depending on your elements) or using NLCC pseudos ( <a href="http://arxiv.org/abs/1212.6011" target="_blank">http://arxiv.org/abs/1212.6011</a> ). The NLCC's are not yet in svn as they are not 100% tested with every feature (energies and derivs are fine) and they only exist for PBE and a limited number of elements (B - Cl). If you are intersted in trying them I can send you the CP2K format. I will add them to svn soon after testing is complete.<br><br>regards<br>Flo<br><br>Re<br></blockquote>