<div dir="ltr">Hello all,<div>I am putting in practice RESPA calculations, and I need some advice with some details that I found.</div><div>My system is liquid Deep Eutectic Solvent, which I run a long MD with GGA functional, then it is converged for GGA at least. I start from there to run the MD. I ran 3 MD: BLYP, B3LYP and B3LYP(BLYP)-RESPA. All of them with 0.5 fs of time step, and a frequency of 5 for the RESPA. I attach the input file for the RESPA.</div><div><br></div><div>I run these 3 MDs to compare and check the quality of the RESPA MD in a "realistic" system, instead over a single molecule MD. First of all I try to check the potential energy, and all it is ok, RESPA obtains pretty good values. (see Potential.eps, all energy values are relative to the first snapshot energy of each dynamic).</div><div>But the conserved quantity (Kinetic+Potential) is not conserved in the case of RESPA, but in the pure B3LYP calculation there is no problem about it. (see Conserved.eps). There is a tendency to get lower energies along the MD (~5mHa for 20 fs)). I do not know if this behavior is "normal" in RESPA calculations with respect to the standard hybrid calculations, the papers do not say anything about that. Searching for the error, I plot the kinetic energy (Kinetic.eps), which shows that the RESPA calculation evolves "losing" kinetic energy with respect to the standard B3LYP MD... Finally to see the values of mismatch between both MD, I compute the delta kinetic and potential energy (Delta.eps), it is pretty clear that potential energy difference is very low (<2 mHa) and it is sometime positive and sometime negative which means there is no bias in the potential computation, but for kinetic the differences is always negative and ~4 mHa, which explains that conserved quantity is not conserved.</div><div><br></div><div>After that, I have some questions... </div><div>1- Is it normal this behavior on kinetic energy for RESPA calculations?</div><div>2- How can I improve the MD to get a conserved total energy in this case? Because for the standard B3LYP MD there is no problem with it.</div><div>3- Are there any criteria that should be tight in the calculation? I calculate potential energy for the first snapshot with EPS_SCHWARZ 1E-12 and SCREENING FALSE, and the energy decreases just 0.3 mHa, and the energy convergence is about 1E-6...</div><div>4- Could I be calm about this and proceed to a longer dynamic without worrying?</div><div><br></div><div>Usually I run calculations by trial and error to understand the problems and improve the calculations, but these calculations are no cheaper and I will need one or two weeks to search with a couple of variotions on the keywords, that is beacuse I am here.</div><div><br></div><div>Regards!</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">El mar, 15 dic 2020 a las 14:17, Lucas Lodeiro (<<a href="mailto:eluni...@gmail.com">eluni...@gmail.com</a>>) escribió:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr">Thanks professor Hutter,<div>It is so clarifier this example, with it I can to compute RESPA calculations and reproduce the figure 4 in the implementation paper. As you mention, the first comment in the example input is wrong, The order is inverted.</div><div>Just to clarify the RESPA calculations for other users:</div><div>The principal task is to set a multiple force_eval section as:</div><div><br></div><div>&MULTIPLE_FORCE_EVALS<br> FORCE_EVAL_ORDER 1 2 <br>&END<br></div><div><br></div><div>Then you have to define two FORCE_EVAL sections with all typical information for a calculation, if you use ADMM, put in both force eval sections the auxiliary basis set file, if not, there will be problems. The first FORCE_EVAL is the "outer" calculation, which is computed in the "long step time", meanwhile, the second FORCE_EVAL is the "inner" and computed in the "short step time". The hybrid ave to be the first, and the GGA the second, but if you put in reverse order, it works but compute a GGA MD, and a hybrid calculation for he short steps.</div><div>Finally, you have to put in MOTION/MD section:</div><div><br></div><div> &RESPA<br> FREQUENCY 5<br> &END RESPA<br></div><div><br></div><div>And define the time step, take care with it, because the time step defined is the "long step time", which apply for the first (hybrid) force eval. The second is computed in time steps of dt/n, where n is the frequency defined above. For example to reproduce the figure 4 in the paper, I set time step = 2.5 and frequency 5, then the hybrid is computed at each 2.5 fs and the GGA evolves with a time step of 0.5 fs.</div><div><br></div><div>Regards!</div><div>Lucas Lodeiro</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">El lun, 14 dic 2020 a las 12:47, <<a href="mailto:hut...@chem.uzh.ch" target="_blank">hut...@chem.uzh.ch</a>> escribió:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hi<br> <br> did you have a look into<br> <br> tests/QS/regtest-hfx-periodic/h2o-respa.inp<br> <br> I'm not sure if the comment in the input is correct.<br> <br> regards<br> <br> Juerg Hutter<br> --------------------------------------------------------------<br> Juerg Hutter Phone : ++41 44 635 4491<br> Institut für Chemie C FAX : ++41 44 635 6838<br> Universität Zürich E-mail: <a href="mailto:hut...@chem.uzh.ch" target="_blank">hut...@chem.uzh.ch</a><br> Winterthurerstrasse 190<br> CH-8057 Zürich, Switzerland<br> ---------------------------------------------------------------<br> <br> -----<a href="mailto:cp...@googlegroups.com" target="_blank">cp...@googlegroups.com</a> wrote: -----<br> To: <a href="mailto:cp...@googlegroups.com" target="_blank">cp...@googlegroups.com</a><br> From: "Lucas Lodeiro" <br> Sent by: <a href="mailto:cp...@googlegroups.com" target="_blank">cp...@googlegroups.com</a><br> Date: 12/13/2020 06:19AM<br> Subject: [CP2K:14381] AIMD+Hybrid-Functionals, How to set the MTS and E_box?<br> <br> Hello,<br> I am reading about how to speed up AIMD+hybrid-functional calculations. I explore different settings of the CP2K-HF section to accelerate the ERIs calculations. But reading this paper (<a href="https://aip.scitation.org/doi/10.1063/1.2931945" rel="noreferrer" target="_blank">https://aip.scitation.org/doi/10.1063/1.2931945</a>), I see one more eps keyword, the epsilon_box, and I cannot find it in the manual (the Eps_schwarz and eps_storage are pretty clear).<br> <br> Also, the MTS section (multiple time step) shows that a very hge speed up using a mix between GGA and hybrid calculations for different frames. I want to do it, but it is not clear how to set up the MTS in the AIMD.In the HF sections there is no key for it. The only key related is RESPA, but is not clear to me how to activate and use it... I cannot find any example for this type of calculation.<br> <br> I am trying PBE0 calculation, and i do not know how MTS uses the XC section, because the X-GGA portion is 0.75, and in the non-hybrid frames, does it use the same X-GGA portion or it uses the full one?<br> <br> If somebody gives me some advice about it, I will appreciate it.<br> <br> Regards<br> Lucas Lodeiro<br> <br> <br> <br> -- <br> You received this message because you are subscribed to the Google Groups "cp2k" group.<br> To unsubscribe from this group and stop receiving emails from it, send an email to <a href="mailto:cp2k%2Bun...@googlegroups.com" target="_blank">cp...@googlegroups.com</a>.<br> To view this discussion on the web visit <a href="https://groups.google.com/d/msgid/cp2k/CAOFT4PKNF01%2BExKAKSNmfKEZQ8%2BN3-RSnSymo-Ec5CFmkDOzUg%40mail.gmail.com" rel="noreferrer" target="_blank">https://groups.google.com/d/msgid/cp2k/CAOFT4PKNF01%2BExKAKSNmfKEZQ8%2BN3-RSnSymo-Ec5CFmkDOzUg%40mail.gmail.com</a>.<br> <br> <br> -- <br> You received this message because you are subscribed to the Google Groups "cp2k" group.<br> To unsubscribe from this group and stop receiving emails from it, send an email to <a href="mailto:cp2k%2Bun...@googlegroups.com" target="_blank">cp...@googlegroups.com</a>.<br> To view this discussion on the web visit <a href="https://groups.google.com/d/msgid/cp2k/OF875C48FD.CB9C8C1F-ONC125863E.00558360-C125863E.00559F76%40lotus.uzh.ch" rel="noreferrer" target="_blank">https://groups.google.com/d/msgid/cp2k/OF875C48FD.CB9C8C1F-ONC125863E.00558360-C125863E.00559F76%40lotus.uzh.ch</a>.<br> </blockquote></div> </blockquote></div>