[Pw_forum] cp.x temperature decreases

Nicola Marzari marzari at MIT.EDU
Mon Jul 21 20:10:52 CEST 2008



Dear Tim,


quick comment - does the CP constant of motion stay flat (econt -
electronic energy plus ionic kinetic plus ficititious kinetic) ?

If it does, you have a properly executed CP simulation, where, depending
on the gap, you observe some amount of "equilibration" between the
real (ions) and fictitious degrees of freedom. I forgot if the
2x1 in C is slighlty metallic - of course, molten Si is definitely 
metallic, so it all seems to make sense.

If the case, have a close look at the Hutter Marx paper review we often
mention, and available on the web.

If the drift is small, you can add a thermostat on ions (and maybe also
on electrons, to avoid them warming up). If the drift is large, you need
to move to born-oppenheimer simulations, of which cp has the 
"ensemble-dft" recipe. If you do that, remember that only the current
cvs (and not the public release) has a fix for an important bug in
parallel calculations.

			nicola

Tim Teatro wrote:
> Hello All,
> 
> Thank you in advance for your time and any help you can offer.
> 
> We are running on HP/XC clusters, Opteron CPUs and pathscale
> compilers, MPI, etc. The clusters have version 3.2.3 of the
> Quantum-ESPRESSO code.
> 
> We are experiencing an issue with two MD simulations using cp.x in
> certain atomic systems. The two systems of note are a C(111) 2x1 slab
> type structure (12 layers with 96 atoms), and the other is a simple
> cubic 64 atom Si system. The problem is that over a 40 000 time step
> run, there is a very noticeable decrease in average temperature. Of
> course, the temperature oscillates over the run, but the C111 system
> appears to have decreasing temperature over time (from about 800K to
> about 600K for example). The temperature is not controlled and thermal
> motion is achieved by initial random displacement of atoms within the
> run (using the tranp switch). At the end of this e-mail, I will attach
> the set of input files used in the run. It appears that the rate of
> energy loss is related to the energy of the system. A system which at
> a higher temperature appears to lose energy more quickly. For C111,
> over 40 000 time steps, this visually resembles a linear decrease, but
> as I mention in a moment, the Si system suggests that the relationship
> is exponential.
> 
> The Si system seems to lose temperature in a way that resembles
> (visually) an exponential decrease. The temperature eventually levels
> off after falling from some initial value. This only appears to happen
> at higher temperatures (I've only seen it in one of my colleagues
> runs, and the temperature was above 1500K)
> 
> Sixty-four atom GaAs, which has the same structure as Si, shows a very
> stable temperature.
> 
> Best regards,
> 
> 
> Timothy A.V Teatro
> 
> --
> Timothy A.V. Teatro <timothy.teatro at uoit.ca>
> University of Ontario Institute of Technology
> Faculty of Science
> 2000 Simcoe Street North
> Oshawa, ON, CANADA, L1H 7K4
> 
> 
> Input files for C(111) 2x1 (96 atom, 12 layer slab)_________________
> **************************************************************
> 
> &control
> title = 'c96 GS',
> calculation='cp',
> restart_mode='from_scratch',
> prefix='c96',
> outdir = '/scratch/timtro/c96/'
> ndr = 51,
> ndw = 51,
> dt = 5.0,
> nstep = 100,
> isave=100,
> tprnfor=.TRUE.,
> forc_conv_thr=1.0d-3,
> etot_conv_thr = 1.d-6,
> ekin_conv_thr = 1.d-5,
> wf_collect=.TRUE.,
> /
> &system
> ibrav = 8,
> celldm(1)=16.510372468,
> celldm(2)=0.577350269,
> celldm(3)=2.558782523,
> nat=96,
> ntyp=1,
> xc_type = 'BLYP',
> ecutwfc = 30.0,
> ecutrho = 300.0,
> /
> &electrons
> emass=400.d0,
> emass_cutoff=2.5d0,
> electron_dynamics='cg',
> /
> &ions
> ion_positions='default',
> ion_velocities='zero',
> ion_dynamics='none'
> /
> ATOMIC_SPECIES
>  C   12.0107  C.pz-vbc.UPF
> ATOMIC_POSITIONS (bohr)
> ...
> 
> 
> 
> &control
> title = 'c96 SD',
> calculation='cp',
> restart_mode='restart',
> prefix='c96',
> outdir = '/scratch/timtro/c96/'
> ndr = 51,
> ndw = 51,
> dt = 5.0,
> nstep = 500,
> isave=100,
> tprnfor=.TRUE.,
> forc_conv_thr=1.0d-6,
> etot_conv_thr = 1.d-6,
> ekin_conv_thr = 1.d-5,
> wf_collect=.TRUE.,
> /
> &system
> ibrav = 8,
> celldm(1)=16.510372468,
> celldm(2)=0.577350269,
> celldm(3)=2.558782523,
> nat=96,
> ntyp=1,
> xc_type = 'BLYP',
> ecutwfc = 30.0,
> ecutrho = 300.0,
> /
> &electrons
> emass=400.d0,
> emass_cutoff=2.5d0,
> electron_dynamics='sd',
> electron_velocities='zero',
> /
> &ions
> ion_positions='default',
> ion_velocities='zero',
> ion_dynamics='none'
> /
> ATOMIC_SPECIES
>  C   12.0107  C.pz-vbc.UPF
> ATOMIC_POSITIONS (bohr)
> ...
> 
> 
> &control
> title = 'c96 RAND',
> calculation='cp',
> restart_mode='restart',
> prefix='c96',
> outdir = '/scratch/timtro/c96/'
> ndr = 51,
> ndw = 51,
> dt = 5.0,
> nstep = 100,
> isave=100,
> tprnfor=.TRUE.,
> forc_conv_thr=1.0d-6,
> etot_conv_thr = 1.d-6,
> ekin_conv_thr = 1.d-5,
> wf_collect=.TRUE.,
> /
> &system
> ibrav = 8,
> celldm(1)=16.510372468,
> celldm(2)=0.577350269,
> celldm(3)=2.558782523,
> nat=96,
> ntyp=1,
> xc_type = 'BLYP',
> ecutwfc = 30.0,
> ecutrho = 300.0,
> /
> &electrons
> emass=400.d0,
> emass_cutoff=2.5d0,
> electron_dynamics='sd',
> electron_velocities='zero',
> /
> &ions
> ion_positions='default',
> ion_dynamics='none',
> tranp(1)=.TRUE.,
> amprp(1)=0.42898,
> /
> ATOMIC_SPECIES
>  C   12.0107  C.pz-vbc.UPF
> ATOMIC_POSITIONS (bohr)
> ...
> 
> 
> &control
> title = 'c96 MD',
> calculation='cp',
> restart_mode='reset_counters',
> prefix='c96',
> outdir = '/scratch/timtro/c96/'
> ndr = 51,
> ndw = 51,
> dt = 5.0,
> nstep = 40000,
> isave=200,
> tprnfor=.TRUE.,
> forc_conv_thr=1.0d-6,
> etot_conv_thr = 1.d-6,
> ekin_conv_thr = 1.d-5,
> wf_collect=.TRUE.,
> /
> &system
> ibrav = 8,
> celldm(1)=16.510372468,
> celldm(2)=0.577350269,
> celldm(3)=2.558782523,
> nat=96,
> ntyp=1,
> xc_type = 'BLYP',
> ecutwfc = 30.0,
> ecutrho = 300.0,
> /
> &electrons
> emass=400.d0,
> emass_cutoff=2.5d0,
> electron_dynamics='verlet',
> electron_velocities='zero',
> /
> &ions
> ion_positions='default',
> ion_dynamics='verlet',
> ion_velocities='zero',
> /
> ATOMIC_SPECIES
>  C   12.0107  C.pz-vbc.UPF
> ATOMIC_POSITIONS (bohr)
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum


-- 
---------------------------------------------------------------------
Prof Nicola Marzari   Department of Materials Science and Engineering
13-5066   MIT   77 Massachusetts Avenue   Cambridge MA 02139-4307 USA
tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu


More information about the Pw_forum mailing list