Hi Pouya, There is a change in the trend at about 3 ps - I'd check if there is some difference in conformation/configuration of your system at this step of trajectory. It seems to me that your system overcame some barrier and moved towards another minimum. In this case the energies should stabilize again and you should be able to observe that if you continue your run. This is the whole fun of MD trajectories analysis :)
Cheers, Bartek On Mon, 2010-11-08 at 17:37 +0100, Pouya Partovi wrote: > Dear Siesta users/experts, > > Attached is the total energy versus time (fs) after I set the > MD.TimeStepLength to 0.5 fs instead of 1.0 fs. I guess the graph looks ok in > terms of its trend etc., since it is tending to negative values (which means > lower kinetic energies for the ions, I guess). Is that true? Should I wait > longer or it has already reached the equilibrium? > > But, one more question: the siesta energy decomposition in the .out file is: > > siesta: Program's energy decomposition (eV): > siesta: Eions = 17846.210777 > siesta: Ena = 3465.061678 > siesta: Ekin = 8326.649023 > siesta: Enl = -1639.500029 > siesta: DEna = -225.505406 > siesta: DUscf = 33.391030 > siesta: DUext = 0.000000 > siesta: Exc = -3507.299949 > siesta: eta*DQ = 0.000000 > siesta: Emadel = 0.000000 > siesta: Ekinion = 5.540327 > siesta: Eharris = -11387.874138 > siesta: Etot = -11387.874102 > siesta: FreeEng = -11387.874102 > > however, the total energy for the last step in the .MDE file is -11392.56494 > eV. Shouldn't they be equal? There is a 5.3 eV difference. > > Thanks, > Pouya > > > > ----- Original Message ----- > From: "Pouya Partovi" <[email protected]> > To: [email protected] > Sent: Tuesday, November 2, 2010 12:22:24 AM > Subject: Re: [SIESTA-L] A (sort of general) MD question > > Thanks Marcos. I will do so. Right now, I'm waiting for the results for 0.5 > fs. > > ----- Original Message ----- > From: "Marcos Veríssimo Alves" <[email protected]> > To: [email protected] > Sent: Monday, November 1, 2010 9:37:30 PM > Subject: Re: [SIESTA-L] A (sort of general) MD question > > Pouya, > > More generally, you should go decreasing the value of the integration > timestep until you find the largest value that gives you a constant total > energy. If 0.5 still gives you a rise in the total energy after a while, > decrease it further, and so on, until you get a constant total energy. > > Marcos > > > On Mon, Nov 1, 2010 at 7:04 PM, Pouya Partovi < [email protected] > > wrote: > > > Thanks guys. Set it to 0.5 fs. Will post the results. > > Bests, > Pouya > > > > ----- Original Message ----- > From: "Bartek Szyja" < [email protected] > > To: [email protected] > > > > Sent: Monday, November 1, 2010 9:08:05 PM > Subject: Re: [SIESTA-L] A (sort of general) MD question > > On Mon, 2010-11-01 at 18:35 +0100, Pouya Partovi wrote: > > MD.LengthTimeStep is set to 1 fs. > > Try smaller value then. 0.5 fs should be better (depending on your > system). And - of course - it will take twice as many steps to complete > the run... > > Good luck, > Bartek > > > > > ----- Original Message ----- > > From: "Bartek Szyja" < [email protected] > > > To: [email protected] > > Sent: Monday, November 1, 2010 7:19:19 PM > > Subject: Re: [SIESTA-L] A (sort of general) MD question > > > > On Mon, 2010-11-01 at 15:26 +0100, Pouya Partovi wrote: > > > Dear SIESTA users, > > > > > > I have a problem with a MD simulation I'm performing with SIESTA at 305 > > > K. I have attached the T, KS_Energy, Tot_Energy and P versus steps > > > plots to this message. Although T, KS_Energy and P reach the > > > equilibrium quite soon, the total energy has a trend even on step > > > 15,000. > > > > What timestep do you set in your calculations? 15 thousands of steps, > > but how many picoseconds that is? > > > > Cheers, > > Bartek > > > > >
