[CPMD-list] fail to read RESTART?
Nisanth Nair
nisanth.nair at theochem.ruhr-uni-bochum.de
Wed Dec 20 11:16:10 CET 2006
Dear Lan-Feng,
In the RESTART file, restart informations are written as sections. For
example, coordinates are in a section, velocities are in another
section etc.. Some of these sections are always written in the RESTART file,
while the rest depends on the various options in the input. You can have a
look at the rv30.F wv30.F setirec.F files for more details.
RESTART-sections are read sequentially, depending on the options you specified
for RESTART. But unfortunately, before reading any information, there is no
check implemented on the type of sections or section number; rather the check
is on the number of records. Thus there is NO GUARANTEE that the restart
file is read correctly and/or CPMD initialized every thing correctly if you
have WRONG restart options in the input file.
In your case the following might have happened:
RESTART VELOCITIES read the velocities 'correctly',
which were just zero. Since it correspond to zero temperatue, CPMD will
initialize the nuclear velocities for the temperature you specified.
For the wavefunction, it will _always_ initialize with zero kinetic energy,
unless there is non-zero velocities present in the RESTART file.
NOSEP option passed the number_of_record_check and read something else from
the restart, which might the reason for the NaN you saw.
To be on the safe side, look always for any warnings after reading the RESTART
files.
When you restart the wavefunction from a RESTART file having a converged
wavefunction, you don't need to use QUENCH BO ELECTRONS; Velocities of the
plane-wave coefficients are set to zero by default.
Hope it helps,
Nisanth
On Tuesday 19 December 2006 23:49, Lan-Feng Yuan wrote:
> Dear Nisanth,
>
> Thank you very much, your suggestion is correct! What seems strange is
> that with VELOCITES and NOSEP in RESTART keyword the program runs just fine
> in many other jobs, so this is an unpredictable problem ... Without reading
> VELOCITIES, the system starts with the target temperature, which is a bad
> thing as I understand. Therefore, maybe we should use QUENCH IONS,
> ELECTRONS to make the initial velocities zero. Thanks again!
>
> Lan-Feng
>
> In your mail:
> >From: Nisanth Nair <nisanth.nair at theochem.ruhr-uni-bochum.de>
> >Reply-To:
> >To: Lan-Feng Yuan <yuanlf at ustc.edu.cn>
> >Subject: Re: [CPMD-list] fail to read RESTART?
> >Date:Tue, 19 Dec 2006 23:05:32 +0100 (CET)
> >
> >Hi Lan,
> >
> > You are trying to read some information which is not in the
> > restart file. Don't restart the velocities and Nose information from a
> > restart file after the wavefunction or geometry optimization!
> >
> > With best regards,
> > Nisanth
> >
> > On Wed, 20 Dec 2006, Lan-Feng Yuan wrote:
> > > Dear CPMD people,
> > >
> > > Sometimes I meet strange cases where a RESTART file cannot be read
> > > into a
>
> MD
>
> > > run. For example, for the molecule H2, a RESTART file is generated by:
> > >
> > > &CPMD
> > > CENTER MOLECULE
> > > ISOLATED MOLECULE
> > > OPTIMIZE WAVEFUNCTION
> > > CONVERGENCE ORBITALS
> > > 1.D-8
> > > &END
> > >
> > > &DFT
> > > FUNCTIONAL PBE
> > > &END
> > >
> > > &SYSTEM
> > > ANGSTROM
> > > SYMMETRY
> > > 1
> > > CELL
> > > 5.000 1. 1. 0.0 0.0 0.0
> > > CUTOFF
> > > 80.00
> > > &END
> > >
> > > &ATOMS
> > > ISOTOPE
> > > 1.007825
> > > *H_MT_PBE.psp KLEINMAN-BYLANDER
> > > LMAX=S
> > > 2
> > > 2.290887532260 2.270887532267 2.280887532298
> > > 2.719112467740 2.739112467733 2.711312467702
> > > &END
> > >
> > > Then I try to read it to start a MD simulation:
> > >
> > > &CPMD
> > > CENTER MOLECULE
> > > MOLECULAR DYNAMICS CP
> > > CONVERGENCE ORBITALS
> > > 1.D-7
> > > EMASS
> > > 100
> > > TIMESTEP
> > > 1.5
> > > NOSE IONS MASSIVE
> > > 300.0 5000.0
> > > NOSE PARAMETER
> > > 20 0 0 6. 625 4
> > > MAXSTEP
> > > 20000
> > > TRAJECTORY XYZ
> > > RESTART COORDINATES CELL WAVEFUNCTIONS VELOCITIES LATEST NOSEP
> > > STORE
> > > 500
> > > &END
> > >
> > > &DFT
> > > NEWCODE
> > > FUNCTIONAL PBE
> > > &END
> > >
> > > &SYSTEM
> > > ANGSTROM
> > > SYMMETRY
> > > 1
> > > CELL
> > > 5.000 1. 1. 0.0 0.0 0.0
> > > CUTOFF
> > > 80.00
> > > &END
> > >
> > > But the program outputs a lot of "NaN":
> > >
> > > CPU TIME FOR INITIALIZATION: 0.44 SECONDS
> > >
> > > ATOM COORDINATES GRADIENTS (-FORCES)
> > > 1 H 4.3292 4.2914 4.3103 NaN NaN NaN
> > > 2 H 5.1384 5.1762 5.1236 NaN NaN NaN
> > >
> > > TOTAL INTEGRATED ELECTRONIC DENSITY
> > > IN G-SPACE = Infinity
> > > IN R-SPACE = NaN
> > >
> > > (K+E1+L+N+X) TOTAL ENERGY = NaN A.U.
> > > (K) KINETIC ENERGY = NaN A.U.
> > > (E1=A-S+R) ELECTROSTATIC ENERGY = NaN A.U.
> > > (S) ESELF = 0.66490380 A.U.
> > > (R) ESR = 0.15685511 A.U.
> > > (L) LOCAL PSEUDOPOTENTIAL ENERGY = NaN A.U.
> > > (N) N-L PSEUDOPOTENTIAL ENERGY = 0.00000000 A.U.
> > > (X) EXCHANGE-CORRELATION ENERGY = 0.00000000 A.U.
> > > GRADIENT CORRECTION ENERGY = 0.00000000 A.U.
> > >
> > > NFI EKINC TEMPP EKS ECLASSIC EHAM
>
> DIS
>
> > > TCPU
> > > 1 NaN NaN NaN NaN NaN
>
> 0.204E-11
>
> > > 0.47
> > > 2 NaN NaN NaN NaN NaN
> > > NaN
> > >
> > > 0.40
> > > 3 NaN NaN NaN NaN NaN
> > > NaN
> > >
> > > 0.43
> > > 4 NaN NaN NaN NaN NaN
> > > NaN
> > >
> > > 0.41
> > > 5 NaN NaN NaN NaN NaN
> > > NaN
> > >
> > > 0.40
> > > 6 NaN NaN NaN NaN NaN
> > > NaN
> > >
> > > 0.41
> > >
> > > In some other cases of the same problem, I can make the RESTART read
>
> correctly
>
> > > by adjusting some parameters (like Nose frequency or number of CPUs;
> > > funny,
>
> isn't
>
> > > it?). However, for this H2 thing, I get no luck. This same trouble
> > > occurs in
>
> both
>
> > > 3.10 and 3.11.
> > >
> > > Did anyone see this problem before? Your help will be greatly
> > > appreciated.
>
> Thank
>
> > > you very much!
> > >
> > > Lan-Feng
> > >
> > >
> > >
> > >
> > > ------------------------------
> > > -------------------------------------------------
> > > Dr. Lan-Feng Yuan,
> > > Associate Professor,
> > > Hefei National Laboratory for Physical Sciences at the Microscale,
> > > University of Science and Technology of China,
> > > Hefei, Anhui 230026, People's Republic of China
> > > Email: yuanlf at ustc.edu.cn
> > > Telephone: 86-551-3600994
> > > -------------------------------------------------
> > > _______________________________________________
> > > CPMD-list mailing list
> > > CPMD-list at cpmd.org
> > > http://cpmd.org/mailman/listinfo/cpmd-list
> >
> > =========================================================================
> >== Dr. Nisanth N. Nair
> > Lehrstuhl fuer Theoretische Chemie Tel: ++49 (0)234 32 22121
> > Ruhr-Universitaet Bochum Fax: ++49 (0)234 32 14045
> > Universitaetstrasse 150
> > D-44801 Bochum, Germany
> > Email: nisanth.nair at theochem.ruhr-uni-bochum.de
> > Web : http://www.theochem.ruhr-uni-bochum.de
> > =========================================================================
> >==
--
===========================================================================
Dr. Nisanth N. Nair
Lehrstuhl fuer Theoretische Chemie Tel: ++49 (0)234 32 22121
Ruhr-Universitaet Bochum Fax: ++49 (0)234 32 14045
Universitaetstrasse 150
D-44801 Bochum, Germany
Email: nisanth.nair at theochem.ruhr-uni-bochum.de
Web : http://www.theochem.ruhr-uni-bochum.de
===========================================================================
More information about the CPMD-list
mailing list