Thank you very much Andrei. Actually the 4th frequency is as close to zero as the 2nd and 3rd, and it is positive (as it should be)
eigenvalue # 1 omega= -40.738952961112830 eigenvalue # 2 omega= -2.05736482791876937E-002 eigenvalue # 3 omega= -1.32180970904245670E-002 eigenvalue # 4 omega= 1.02952404945284657E-002 eigenvalue # 5 omega= 13.407523547356076 Maybe the problem is not the meshcutoff but the fact that, in the relaxation, I keep fixed the 1st layer positions, so there was big forces on some atoms. Now I am trying to recalculate the phonons after relaxing all the atoms. By the way, is there a standard threshold value to judge the goodness of atomic forces? Best regards, Carlo 2012/4/7 <[email protected]> > > Andrei, > > > > thank you very much for your answer. I had already found in the ML > > database > > the nice phonon reference you indicated to me. > > By negative frequencies, I mean: > > > > eigenvalue # 1 omega= -40.738952961112830 > > eigenvalue # 2 omega= -2.05736482791876937E-002 > > eigenvalue # 3 omega= -1.32180970904245670E-002 > > Dear Carlo, > Your frequencies 2 and 3 are zero, for all practical use. > No special importance therefore that they are negative. > If you have yet the third frequency (positive, in your case?) > which is as close to zero as those two (your Nr4?) > then likely you don't have a problem with MeshCutoff. > However, your frequency Nr1 is an indication > that your relaxation is not good enough. > The negative frequency (if a stable one and not due to numerical noise) > means: as you displace the atoms in some combination, the energy > goes down. It shouldn't be like this at the equilibrium. > Have a look into p.18 of > http://www.home.uni-osnabrueck.de/apostnik/Lectures/APostnikov-Phonons.pdf > > > My system is a molecular junction. Just to explain you, at the beginning > I > > calculated the electrodes bulk lattice parameter, and then I built the > > junction geometry. If I then relax by setting MD.variableCell=T the cell > > dimensions increase by 1.5%... Do you think that I should necessarily use > > a variable cell in the relaxation? (I am suspicious about it) > > Not necessarily at all; you can calculate phonons > in a constrained system, under pressure etc. However, > atom positions have to be optimized subject to this constraint > (to be sure that trial atom displacements are around the equilibrium). > > > Finally, could you please give me some reference about the eggbox effect > > test? > > It has been discussed many many times in the main list and in tutorials. > E.g., you may have a look in > http://www.home.uni-osnabrueck.de/apostnik/Lectures/SIESTA-tuto.pdf > (pp.10-12) > > Good luck > > Andrei Postnikov > > > > > Thank you very much! > > > > Carlo > > > > > > 2012/4/7 <[email protected]> > > > >> Dear Carlo, > >> as you don't specify how negative your negative frequencies are, > >> it is not so easy to judge... Your sum of forces seems a bit too high... > >> Even if not dramatically... And some individual forces as well - > >> e.g., Fx on atom 2. So generally I'd recommend even higher MeshCutoff > >> and further relaxation. A priori 350 Ry is not safely high, > >> sometimes one needs to go to much higher. > >> > >> Before doing the phonons, I'd recommend the usual "eggbox test" > >> to be sure that fluctuations of summary force as function of > >> uniform displacement of all atoms remain within |0.1|. > >> > >> Concerning specifically the phonons, you may wish to check > >> > http://www.home.uni-osnabrueck.de/apostnik/Lectures/APostnikov-Phonons.pdf > >> - especially p.18 of it. > >> > >> Best regards > >> > >> Andrei Postnikov > >> > >> > >> > Dear Siesta users, > >> > > >> > I am trying to calculate phonon frequencies at Gamma for a molecular > >> > junction. > >> > I relaxed the system by keeping fixed the Lattice Vectors (not a > >> variable > >> > cell). When I compute the frequencies (with MeshCutoff = 200 Ry), I > >> obtain > >> > that the first three frequencies are negative, which indicates that > >> the > >> > results are not correct. > >> > Then, if I increase the MeshCutoff to 350 Ry (which seems a high value > >> to > >> > me) I obtain again negative frequencies. > >> > Could you suggest me how to fix the problem? > >> > The following are the forces of the system...do you think that they > >> are > >> > too > >> > high? Which is a good criteria to say that the forces are low enough? > >> > > >> > Thank you very much, > >> > > >> > Carlo > >> > > >> > siesta: 1 0.106233 -0.096180 0.047641 > >> > siesta: 2 -0.655838 0.034936 -0.004086 > >> > siesta: 3 0.364736 -0.100368 0.166114 > >> > siesta: 4 0.130690 0.010500 -0.101584 > >> > siesta: 5 -0.122410 0.420815 -0.012044 > >> > siesta: 6 0.120625 -0.031292 0.152270 > >> > siesta: 7 0.003319 0.004051 -0.012274 > >> > siesta: 8 0.004951 -0.015719 0.001451 > >> > siesta: 9 -0.009523 0.004610 0.013162 > >> > siesta: 10 0.019411 -0.008606 -0.009840 > >> > siesta: 11 0.002485 -0.010131 0.005577 > >> > siesta: 12 0.001815 -0.012882 -0.010601 > >> > siesta: 13 0.002293 -0.027828 0.004317 > >> > siesta: 14 0.003541 -0.004099 -0.003595 > >> > siesta: 15 -0.004545 -0.005597 0.018501 > >> > siesta: 16 -0.002375 0.005693 -0.016887 > >> > siesta: 17 0.007410 -0.017539 -0.001118 > >> > siesta: 18 0.015827 -0.012960 -0.005705 > >> > siesta: 19 -0.010629 -0.007460 -0.001509 > >> > siesta: 20 0.008748 0.000683 -0.004698 > >> > siesta: 21 -0.004943 0.029703 -0.008064 > >> > siesta: 22 -0.003202 0.004500 0.030359 > >> > siesta: 23 -0.009982 -0.003884 -0.044988 > >> > siesta: 24 0.012110 -0.008463 -0.054600 > >> > siesta: 25 0.014891 0.000114 0.030375 > >> > siesta: 26 0.011714 0.007854 -0.002981 > >> > siesta: 27 -0.006905 -0.024651 0.005841 > >> > siesta: 28 0.002399 -0.019318 0.013282 > >> > siesta: 29 -0.019852 0.006114 -0.030653 > >> > siesta: 30 -0.023687 0.004007 -0.028044 > >> > siesta: 31 0.023195 -0.022556 -0.014991 > >> > siesta: 32 0.024333 -0.033430 -0.028258 > >> > siesta: 33 -0.014937 0.006378 0.009843 > >> > siesta: 34 0.004912 -0.019004 -0.008074 > >> > siesta: 35 0.006259 -0.012323 0.009612 > >> > siesta: 36 -0.010453 0.017222 0.013853 > >> > siesta: 37 0.001281 0.001284 -0.002971 > >> > siesta: 38 0.009564 0.017869 -0.013170 > >> > siesta: 39 0.004626 -0.016715 -0.008634 > >> > siesta: 40 0.008447 -0.028871 -0.007756 > >> > siesta: 41 -0.020042 -0.027721 -0.021694 > >> > siesta: 42 0.020027 -0.019833 -0.095505 > >> > siesta: ---------------------------------------- > >> > siesta: Tot 0.016519 -0.011098 -0.032128 > >> > > >> > > >> > -- > >> > > >> > Carlo Motta > >> > * Ph.D. Student in Materials Science* > >> > University Office Phone : +39 02 6448-5183 > >> > Department of Materials Science > >> > University of Milano-Bicocca > >> > Via R. Cozzi 53, 20125 Milano, Italy > >> > > >> > >> > > > > -- Carlo Motta * Ph.D. Student in Materials Science* Email: [email protected] University Office Phone : +39 02 6448-5183 Department of Materials Science University of Milano-Bicocca Via R. Cozzi 53, 20125 Milano, Italy
