[CPMD-list] Restart Option for Wave Function & Geometry Optimizatio (Shashidhar)
Shashidhar G
haishashidhar at gmail.com
Wed Apr 25 12:29:41 CEST 2007
Dear Ari P Seitsonen , Axel Kohlmeyer
Thanks for your quick response and good suggestions and pointing my
mistakes.
I did these calculation for Si, Si8, Si18, Si64, Si128, Si256. Up to Si128
in single run I achieved self consistence criterion for wave function
optimization and Idid Geometry optimization and Kohn Sham energy
calculation. But when I came across Si256 I couldn't.
Before doing CPMD , I worked in PARSEC. With PARSEC I went upto Si128, but
after it become difficult with the code.
In both Case I used same *Troullier*-*Martins pseudopotential , *LMAX=P
LOC=S & same cutoff radius.
In these two codes for Si system I got approximately equal eigenvalues
values. So I thought my different options in the Input file of CPMD, what
ever I mentioned in my first mail( to cpmd mailing list) is correct.
But after pointing my mistakes, I'm going through your suggestions.
*I didn't get Why we have to choose LMAX=D (implying LOC=D) is the
standard choice*
*even Si doesn't contain d electrons. Please give me reason or any
reference. *
Here I'm giving results of Si system for both codes. Please have a look on
these.
Psedopotentila Data :
============================================================
| Pseudopotential Report |
------------------------------------------------------------
| Atomic Symbol : SI |
| Atomic Number : 14 |
| Number of core states : 3 |
| Number of valence states : 2 |
| Exchange-Correlation Functional : |
| Slater exchange : 0.6667 |
| LDA correlation : Ceperley-Alder |
| Scalar-relativistic calculation |
| Electron Configuration : N L Occupation |
| 1 S 2.0000 |
| 2 S 2.0000 |
| 2 P 6.0000 |
| 3 S 2.0000 |
| 3 P 2.0000 |
| Full Potential Total Energy = -288.820070 |
| Trouiller-Martins normconserving PP |
| n l rc energy |
| 3 S 2.5000 -0.40000 |
| 3 P 2.5000 -0.15320 |
| 3 D 2.5000 -0.15320 |
| Number of Mesh Points : 664 |
| Pseudoatom Total Energy = -3.758534 |
============================================================
*********************************************************************************************************
* ATOM MASS RAGGIO NLCC
PSEUDOPOTENTIAL *
* Si 28.0860 1.2000 NO
KLEINMAN S LOCAL *
*
P NONLOCAL *
*********************************************************************************************************
------------------------------------------------------------------------------------------------------
Using CPMD
System : Si
CUTOFF = 48.77 Ry ( This is equivalent to grid spacing(h) = 0.45 Bohr, in
PARSEC code.
[ Plane Wave Cutoff = (pi/h)^2 ]
EIGENVALUES(EV) AND OCCUPATION:
1 -10.9506408 2.00000000 2 -4.4084689
1.00000000
3 -4.4084687 1.00000000 4 -3.7450002
0.00000000
5 -0.4819290 0.00000000 6 0.6014599
0.00000000
7 0.6135355 0.00000000 8 0.6257078
0.00000000
9 0.6257140 0.00000000 10 0.6569273
0.00000000
11 0.7385251 0.00000000 12 1.1896858NC
0.00000000
13 1.6673697NC 0.00000000 14 1.6700032NC
0.00000000
CHEMICAL POTENTIAL = -4.4084688968 EV
****************************************************************
ELECTRONIC GRADIENT:
MAX. COMPONENT = 2.33203E-02 NORM = 2.22695E+00
TOTAL INTEGRATED ELECTRONIC DENSITY
IN G-SPACE = 4.000000
IN R-SPACE = 4.000000
(B+E2+X-V) TOTAL ENERGY = -3.78889677 A.U.
(B) BAND ENERGY = -1.12884827 A.U.
(E2=I-H-S+R) ELECTROSTATIC ENERGY = -2.95663553 A.U.
(S) ESELF = 5.31923041 A.U.
(R) ESR = 0.00000000 A.U.
(X) EXCHANGE-CORRELATION ENERGY = -0.97879242 A.U.
(V) EXCHANGE-CORRELATION POTEN. = -1.27537946 A.U.
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Using PARSEC
System : Si
Grid Spacing = 0.45 Bohr ( This is equivalent Plane Wave Cutoff =
48.77Ry, in CPMD code.
State Eigenvalue [Ry] Eigenvalue [eV] Occup. Repr.
1 -0.7941280134 -10.8047469243 1.0000 1
2 -0.3024004062 -4.1143994468 1.0000 6
3 -0.3024004062 -4.1143994463 0.0000 7
4 -0.3024004062 -4.1143994460 0.0000 8
5 0.1092973937 1.4870784791 0.0000 1
6 0.1932353338 2.6291213051 0.0000 1
7 0.1932353341 2.6291213081 0.0000 1
8 0.1933878377 2.6311962419 0.0000 4
9 0.1933878377 2.6311962419 0.0000 2
10 0.1933878377 2.6311962424 0.0000 3
11 0.2926295114 3.9814586067 0.0000 7
12 0.2926295117 3.9814586101 0.0000 6
Max and min values of charge density [e/Bohr^3] 0.8548E-01 0.1320E-07
Hartree potential time [sec]: 0.10
Eigenvalue Energy = -2.19305684 [Ry]
Hartree Energy = 6.04962485 [Ry]
Integral_{Vxc*rho} = -2.61678583 [Ry]
Exc = Integral{eps_xc*rho} = -2.00811350 [Ry]
Electron-Ion energy = -13.82622026 [Ry]
Ion-Ion Energy = 0.00000000 [Ry]
(E(new)-E(old))/atom = -101.52207433 [eV]
Total Energy = -7.46167622 [Ry]
Energy/atom = -101.52207433 [eV]
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Once again Thanks, I'm looking your suggestions, I will get back you.
With Regards.
SHASHIDHAR . G
Project Intern,
Centre for Development of Advanced Computing(CDAC)-Pune,
Pune University Campus,
Pune-411 007. INDIA.
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Dear Shashidhar,
There seem to be some suspicious things in your input:
- LMAX=P LOC=S does not make sense for a pseudo potential for silicon.
LMAX=D (implying LOC=D) is the standard choice
- This is a cluster of atoms, right? Are there degeneracies which might
lead to dangling bonds being half filled? This could lead to zero
electronic gap between the occupied and unoccupied states, and would
lead to very slow or even impossible convergence
- Why do you specify 'ORTHOGONALIZATION LOWDIN' and 'SPLINE POINTS'?
- I personally find 'LBFGS' to be the fastest method for relaxing atomic
coordinates
- If you program stopped due to exceeding the computing time, it probably
didn't produce a file "RESTART.1" yet, did it? If not, then you cannot
perform a restart, and for the next job I would recommend either a
smaller 'MAXSTEP' or checking out the option 'MAXCPUTIME' (please notice
that the time specified has to be slightly smaller than the real time
allowed by some machines, IBM included; I usually subtract 200-600
seconds from the maximum time)
- You can always perform 'RESTART WAVEFUNCTION COORDINATES LATEST',
unless you have moved the file "RESTART.1" into "RESTART"
- Did you try the input files which you provided? What went wrong, or did
they work?
Greetings,
apsi
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SG> Dear All
SG>
SG> I'm running CPMD on IBM, AIX, power 4 ( 54 Nos. of 4-Way SMP nodes and 1
No.
SG> of 32-Way SMP node) for Si256 atoms.
SG> Here I'm attaching my input file for Wave Function optimization. I ran
it
SG> with 16 processor (8 X 2), but I didn't achieved the self consistence
SG> criterion for wave function optimization and program stops due to wall
clock
SG> time exceeded.
SG>
SG> I would like to Restart the calculation from last run. Here I'm giving
my
SG> 1. Input File for Wave Function Optimization
SG> 2. Input File for Restart the Wave Function Optimization.
SG>
SG> Please Suggest me If I'm in wrong direction, and any thing is incorrect
in
SG> both files 1 & 2 .
SG>
SG> -----------------------------------------------------------
SG> 1. Input For Wave Function Optimization.
SG> ------------------------------------------------------------
[...]
on top of ari's (very valid) concerns, here is a typo
that will make your performance very poor:
SG>
SG> &ATOMS
SG> *SI.OUT1 KLEIMAN-BYLANDER
this should read KLEINMAN-BYLANDER.
due to the missing 'N' cpmd assumes gauss-hermite
integration which is very costly for the large number
of particles that you have.
SG> LMAX=P LOC=S
SG> 256
generally, i was wondering, whether you have tried this
on a smaller system first. since the pseudopotential looks
as if it was newly created (why?), it would be a very good
idea to first validate it with less demanding calculations.
SG>
-------------------------------------------------------------------------------------------------------------------
SG> 2. Input For Restarting the optimization of Wave Functions From Previous
Run
SG> ------------------------------
-------------------------------------------------------------------------------------
another way to have a better chance to restart
(note that with your setup, you have to rename
your RESTART.1 file to RESTART) from an interrupted
run would be to use the STORE and RESTFILES flags.
see the manual for details.
SG> &CPMD
SG> BFGS
this has no effect for a wavefunction optimization.
what however would be helpful in your case, is to
use a different algorithm to optimize the wavefunction.
a pre-conditioned conjugate gradient with linesearch
(PCG MINIMIZE) is much more robust than the default
ODIIS (which is faster for systems where it works well).
[...]
SG> Please give your suggestion if I'm going in wrong direction
hope this helps,
axel.
SG>
SG> With Regards.
SG> SHASHIDHAR . G
SG> Project Intern,
SG> Centre for Development of Advanced Computing(CDAC)-Pune,
SG> Pune-411 007. INDIA.
SG>
--
=======================================================================
Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu http://www.cmm.upenn.edu
Center for Molecular Modeling -- University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel: 1-215-898-5425
=======================================================================
--
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://cpmd.org/pipermail/cpmd-list/attachments/20070425/5d1c6b52/attachment-0001.html
More information about the CPMD-list
mailing list