[Pw_forum] something wrong with vc-relax & relax on my clusters
lewvega
vegalew at hotmail.com
Tue Jun 24 11:12:12 CEST 2008
Dear Users
I'm a new user of pwscf, and I have some experience of castep, which seems to be similar with pwscf.
First I tested paralell vision of pwscf with anatase-TiO2 lattice cell optimization calculation. the input file as follows:
&CONTROL
title = 'Anatase lattice BFGS' ,
calculation = 'vc-relax' ,
restart_mode = 'from_scratch' ,
outdir = '/home/vega/tmp/' ,
pseudo_dir = '/home/vega/espresso-4.0/pseudo/' ,
prefix = 'Anatase lattice default' ,
etot_conv_thr = 0.000000735 ,
forc_conv_thr = 0.0011668141375 ,
nstep = 1000 ,
/
&SYSTEM
ibrav = 6,
celldm(1) = 7.135605333,
celldm(3) = 2.5121822033898305084745762711864,
nat = 12,
ntyp = 2,
ecutwfc = 25 ,
ecutrho = 200 ,
/
&ELECTRONS
conv_thr = 7.3D-8 ,
/
&IONS
/
&CELL
/
ATOMIC_SPECIES
Ti 47.86700 Ti.pw91-sp-van_ak.UPF
O 15.99940 O.pw91-van_ak.UPF
ATOMIC_POSITIONS angstrom
Ti 0.000000000 0.000000000 0.000000000
Ti 1.888000000 1.888000000 4.743000000
Ti 0.000000000 1.888000000 2.372000000
Ti 1.888000000 0.000000000 7.115000000
O 0.000000000 0.000000000 1.973000000
O 1.888000000 1.888000000 6.716000000
O 0.000000000 1.888000000 4.345000000
O 1.888000000 0.000000000 9.088000000
O 1.888000000 0.000000000 5.141000000
O 0.000000000 1.888000000 0.398000000
O 1.888000000 1.888000000 2.770000000
O 0.000000000 0.000000000 7.513000000
K_POINTS automatic
7 7 3 1 1 1
then I checked the initial structure in Xcrysden, and type paralell command like this: mpiexec -n 20 pw.x < inputfile > outputfile
the calculation ran immediately. After calculation finished, the last CELL_PARAMETERS section in outputfile is
CELL_PARAMETERS (alat)
1.000000000 0.000000000 0.000000000
0.000000000 1.000000000 0.000000000
0.000000000 0.000000000 2.512182203
this indicate the 'vc-relax' didn't take effect, no cell parameters was not optimized during the vc-relax strategy.
And the atom positons also changed a lot, the ATOMIC_POSITIONS sections like this:
ATOMIC_POSITIONS (angstrom)
Ti 0.000000000 0.000000000 0.000108235
Ti 1.888000000 1.888000000 4.743111514
Ti 0.000000000 1.888000000 2.371065034
Ti 1.888000000 0.000000000 7.115503582
O 0.000000000 0.000000000 1.981862393
O 1.888000000 1.888000000 6.726003352
O 0.000000000 1.888000000 4.353977562
O 1.888000000 0.000000000 9.096778202
O 1.888000000 0.000000000 5.132185229
O 0.000000000 1.888000000 0.389375062
O 1.888000000 1.888000000 2.760157286
O 0.000000000 0.000000000 7.504294082
Both x,y remain as the initial value.
All of this indicate my above calculation is a 'relax' calculation not a 'vc-relax' strategy.
Also if I set the ionic and cell optimization both using BFGS schedule as the castep did, the cell parameters changed so much that an error occured said 'non orthogonal operation'.
Does anyone encounter the same problem?
Second, I uesed the optimized anatase cell via pwscf's default vc-relax damped MD schedule to cleave the 101 surface and make the further calculation. But another serious problem happened. the input file as follows:
&CONTROL
title = 'Anatase101_2x' ,
calculation = 'relax' ,
restart_mode = 'from_scratch' ,
outdir = '/home/vega/tmp/' ,
pseudo_dir = '/home/vega/espresso-4.0/pseudo/' ,
prefix = 'Anatase101_2x' ,
etot_conv_thr = 0.000000735 ,
forc_conv_thr = 0.0011668141375 ,
nstep = 1000 ,
/
&SYSTEM
ibrav = 8,
celldm(1) = 19.293941717,
celldm(2) = 1.109509628630600744472407458828,
celldm(3) = 1.7484889932972017886534804648227,
nat = 72,
ntyp = 2,
ecutwfc = 30 ,
ecutrho = 240 ,
/
&ELECTRONS
conv_thr = 7.3D-8 ,
/
&IONS
ion_dynamics = 'bfgs' ,
/
ATOMIC_SPECIES
Ti 47.86700 Ti.pw91-sp-van_ak.UPF
O 15.99940 O.pw91-van_ak.UPF
ATOMIC_POSITIONS angstrom
Ti 2.649000000 -1.888000000 0.733000000
Ti -1.060000000 0.000000000 4.242000000
Ti 0.446000000 0.000000000 1.610000000
Ti -3.264000000 -1.888000000 5.119000000
O 0.807000000 -1.888000000 1.466000000
O -2.902000000 0.000000000 4.975000000
O -1.397000000 0.000000000 2.344000000
O -5.105000000 -1.888000000 5.852000000
O -1.421000000 -1.888000000 4.386000000
O 2.287000000 0.000000000 0.877000000
O 0.783000000 0.000000000 3.508000000
O -4.323000000 -1.888000000 3.509000000
Ti 2.649000000 5.664000000 0.733000000
Ti -1.060000000 7.552000000 4.242000000
Ti 0.446000000 7.552000000 1.610000000
Ti -3.264000000 5.664000000 5.119000000
O 0.807000000 5.664000000 1.466000000
O -2.902000000 7.552000000 4.975000000
O -1.397000000 7.552000000 2.344000000
O -5.105000000 5.664000000 5.852000000
O -1.421000000 5.664000000 4.386000000
O 2.287000000 7.552000000 0.877000000
O 0.783000000 7.552000000 3.508000000
O -4.323000000 5.664000000 3.509000000
Ti 2.649000000 1.888000000 0.733000000
Ti -1.060000000 3.776000000 4.242000000
Ti 0.446000000 3.776000000 1.610000000
Ti -3.264000000 1.888000000 5.119000000
O 0.807000000 1.888000000 1.466000000
O -2.902000000 3.776000000 4.975000000
O -1.397000000 3.776000000 2.344000000
O -5.105000000 1.888000000 5.852000000
O -1.421000000 1.888000000 4.386000000
O 2.287000000 3.776000000 0.877000000
O 0.783000000 3.776000000 3.508000000
O -4.323000000 1.888000000 3.509000000
Ti 4.045000000 -1.888000000 4.242000000
Ti 1.842000000 0.000000000 5.119000000
O 2.204000000 -1.888000000 4.975000000
O 0.000000000 0.000000000 5.852000000
O 3.684000000 0.000000000 4.386000000
Ti 4.045000000 5.664000000 4.242000000
Ti 1.842000000 7.552000000 5.119000000
O 2.204000000 5.664000000 4.975000000
O 0.000000000 7.552000000 5.852000000
O 3.684000000 7.552000000 4.386000000
Ti 4.045000000 1.888000000 4.242000000
Ti 1.842000000 3.776000000 5.119000000
O 2.204000000 1.888000000 4.975000000
O 0.000000000 3.776000000 5.852000000
O 3.684000000 3.776000000 4.386000000
Ti 7.754000000 3.776000000 0.733000000
Ti 5.549000000 1.888000000 1.611000000
O 5.911000000 3.776000000 1.467000000
O 3.708000000 1.888000000 2.344000000
O 7.392000000 1.888000000 0.877000000
O 9.596000000 3.776000000 -0.000000000
O 4.490000000 1.888000000 0.000000000
Ti 7.754000000 7.552000000 0.733000000
Ti 5.549000000 5.664000000 1.611000000
O 5.911000000 7.552000000 1.467000000
O 3.708000000 5.664000000 2.344000000
O 7.392000000 5.664000000 0.877000000
O 9.596000000 7.552000000 -0.000000000
O 4.490000000 5.664000000 0.000000000
Ti 7.754000000 0.000000000 0.733000000
Ti 5.549000000 -1.888000000 1.611000000
O 5.911000000 0.000000000 1.467000000
O 3.708000000 -1.888000000 2.344000000
O 7.392000000 -1.888000000 0.877000000
O 9.596000000 0.000000000 -0.000000000
O 4.490000000 -1.888000000 0.000000000
K_POINTS gamma
after checking the initial structure via Xcrysden, I typed 'mpiexec -n 20 pw.x < inputfile > outputfile'. But unfortunately nothing happened, even the first few lines of pwscf was not printed, and all the processes of pw.x triggered by mpich2 are sleeping. No calculation is running at all. Then I run this task in serial version of pwscf by typing 'pw.x < inputfile > outputfile', the calculation ran immediately. then I tried again and again, the paralell calculation of the input file always failed with no message out. And to make sure whether the environment is suitable for paralell calculation, I ran the first input file again in 20 paralell process. Everything seems to be ok with the paralell environment.
I had tested the same task(same number of atoms, same atom, same convergence threshold et al.) in CASTEP successfully, so the error is not because of hardware.
Is there any one can try my input file, and give me some suggestions. thank you so much for reading.
by the way, I also will tell you something about my hardware and softeware:
hardware: five computers with intel coreII Q6600 CPU (quadcore) and 40G menmory (8G memory per computer) on intel S3000AH system board. Computers linked with 1Gbit/s network by only one Swich( TP-Link )
software: Readhat linux enterprise 4 Advanced Server Update 4, mpich2, intel fortran compiler 10.1.008 (64 bit-edition), intel c++ 10.1.008 (64 bit-edition), intel MKL 10.0.011 (64 bit-edition), fftw2.1.
And I the fftw in MKL can't be recognized by 'configure' of espresso, even the lib dictionary is included in LD_LIBRARY_PATH.
thank you all again for reading and I'm looking forward to your suggestions.
Vega Lew
PH.D Candidate in Chemical Engineering
College of Chemistry and Chemical Engineering
Nanjing University of Technology, 210009, Nanjing, Jiangsu, China
_________________________________________________________________
多个邮箱同步管理，live mail客户端万人抢用中
http://get.live.cn/product/mail.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.democritos.it/pipermail/pw_forum/attachments/20080624/51ea2d1e/attachment-0001.htm
More information about the Pw_forum
mailing list