[gpaw-users] Changes in the new PAW setups bundle of 26/10 ("gpaw-setups-0.9.9672")
Marcin Dulak
Marcin.Dulak at fysik.dtu.dk
Tue Nov 27 14:32:29 CET 2012
Hi,
On 11/27/12 13:40, Niels Theis Bendtsen wrote:
> So I did my own little test with the new setups on rutile RuO2. I was unable to converge using RM-DIIS (not surprised anymore) and switching to cg works. I got RM-DIIS to converge using basis = dzp
>
> My calculation is just a single calculitation so no structure optimization.
>
> Using the old setups I get a total energy of -39.895 eV with 34 bands and 40 valence electrons. It took only 6 m 42s and 44 iterations
>
> Using new setups:
>
> cg and default parameters:
>
> -42.983 eV, 42 bands, 56 valence electrons and 9 min 14 s and 24 iterations.
>
> RM-DIIS and basis = 'dzp' try 1:
>
> -42.983 eV, 102 bands, 56 valence electrons and 11 min 52 s and 28 iterations.
>
> RM-DIIS and basis = 'dzp' try 2:
> -42.983 eV, 78 bands, 56 valence electrons and 14 min 18 s and 46 iterations.
>
> The total energies does deviate on the fourth decimal. In this small set cg is actually faster than RM-DIIS??? I saw that Ask did something with basis set that I cannot replicate here. Can it have something to do with the structure optimisation?
this is great news that you are able to converge the system with rmm.
If you compare the same number of iterations of cg and rmm for the same,
large enough system, you will see that rmm is faster.
Measuring time on a Linux system has 5-10% precision, so in order to
make any conclusions
you need a large system and run it several times.
Marcin
>
> Script is provided below together with the output file.
>
> #!/usr/bin/env python
> #PBS -N RuO2cg
> #PBS -m ae
> #PBS -l nodes=1:ppn=8:xeon
> #PBS -l walltime=20:00:00
>
> from ase.parallel import paropen
> from gpaw import GPAW, Mixer
> #from ase.data import reference_states, atomic_numbers
> from ase.utils.eos import EquationOfState
> from ase.io import read, write
> from ase import units
> from ase.lattice.spacegroup import crystal
> import numpy as np
> from scipy.optimize import fmin
> from ase.optimize import BFGS
> import sys
> import os
> from gpaw import setup_paths
>
> setup_paths.insert(0, '/home/camp/ntben/localnewsetup/gpaw-setups-0.9.9672')
>
> element = 'RuO2dzp'
>
> a = 4.49
> c = 3.10
>
> atoms =crystal(['Ru', 'O'], basis=[(0, 0, 0),(0.3, 0.3, 0.0)],
> spacegroup=136, cellpar=[a, a, c, 90, 90, 90])
>
> uc = atoms.get_cell()
> print "Suggested k points:", np.round(30 / uc.diagonal())
> kpts = [7, 7, 10]
> print "Used k points:", kpts
>
>
> calc = GPAW(h=0.15,
> kpts=kpts,
> basis='dzp',
> xc='RPBE',
> txt='test' + element+'.txt')
>
> atoms.set_calculator(calc)
> atoms.get_potential_energy()
>
> Output:
>
>
> ___ ___ ___ _ _ _
> | | |_ | | | |
> | | | | | . | | | |
> |__ | _|___|_____| 0.9.0.8965
> |___|_|
>
> User: ntben at a032.dcsc.fysik.dtu.dk
> Date: Tue Nov 27 11:12:54 2012
> Arch: x86_64
> Pid: 26527
> Dir: /opt/campos-gpaw/0.9.0.8965/1.el5.fys.open64.4.2.3.python2.4.without_ase.openmpi.1.3.3.acml.4.4.0.acml.4.4.0/lib64/python2.4/site-packages/gpaw
> ase: /opt/campos-ase3/3.6.0.2515/1.el5.fys.python2.4/lib64/python2.4/site-packages/ase (version 3.6.0.2515)
> numpy: /opt/numpy/1.3.0/1.el5.fys.gfortran.4.1.2.python2.4.acml.4.0.1.acml.4.0.1/lib64/python2.4/site-packages/numpy (version 1.3.0)
> units: Angstrom and eV
> cores: 8
>
> Memory estimate
> ---------------
> Process memory now: 27.33 MiB
> Calculator 333.36 MiB
> Density 29.54 MiB
> Arrays 3.11 MiB
> Localized functions 25.71 MiB
> Mixer 0.72 MiB
> Hamiltonian 6.05 MiB
> Arrays 2.03 MiB
> XC 0.00 MiB
> Poisson 2.32 MiB
> vbar 1.70 MiB
> Wavefunctions 297.77 MiB
> Arrays psit_nG 244.04 MiB
> Eigensolver 25.60 MiB
> Projectors 3.57 MiB
> Overlap op 24.56 MiB
>
> Positions:
> 0 Ru 0.0000 0.0000 0.0000
> 1 Ru 2.2450 2.2450 1.5500
> 2 O 1.3470 1.3470 0.0000
> 3 O 3.1430 3.1430 0.0000
> 4 O 0.8980 3.5920 1.5500
> 5 O 3.5920 0.8980 1.5500
>
>
>
>
>
> O
> Ru
> O
> O
> Ru
>
>
>
>
>
> Unit Cell:
> Periodic X Y Z Points Spacing
> --------------------------------------------------------------------
> 1. axis: yes 4.490000 0.000000 0.000000 28 0.1604
> 2. axis: yes 0.000000 4.490000 0.000000 28 0.1604
> 3. axis: yes 0.000000 0.000000 3.100000 20 0.1550
>
> Ru-setup:
> name : Ruthenium
> id : 76477906e28f51dc1b29884d53556fe9
> Z : 44
> valence: 16
> core : 28
> charge : 0.0
> file : /home/camp/ntben/localnewsetup/gpaw-setups-0.9.9672/Ru.RPBE.gz
> cutoffs: 1.23(comp), 2.16(filt), 1.30(core), lmax=2
> valence states:
> energy radius
> 4s(2) -76.255 1.281
> 5s(1) -4.126 1.281
> 4p(6) -46.371 1.286
> 5p(0) -0.913 1.286
> 4d(7) -5.146 1.254
> *d 22.066 1.254
>
> LCAO basis set for Ru:
> Name: dzp
> Basis set was loaded from file /home/camp/ntben/localnewsetup/gpaw-setups-0.9.9672/Ru.dzp.basis.gz
> Number of radial functions: 9
> Number of spherical harmonics: 25
> l=0, rc=3.2969 Bohr: 4s-sz confined orbital
> l=0, rc=9.6875 Bohr: 5s-sz confined orbital
> l=1, rc=3.7656 Bohr: 4p-sz confined orbital
> l=2, rc=6.3281 Bohr: 4d-sz confined orbital
> l=0, rc=2.0781 Bohr: 4s-dz split-valence wave
> l=0, rc=5.8281 Bohr: 5s-dz split-valence wave
> l=1, rc=2.4219 Bohr: 4p-dz split-valence wave
> l=2, rc=3.7500 Bohr: 4d-dz split-valence wave
> l=2, rc=3.7656 Bohr: d-type Gaussian polarization
>
> O-setup:
> name : Oxygen
> id : 08071ca1eed670e7821b24b7eb4d558c
> Z : 8
> valence: 6
> core : 2
> charge : 0.0
> file : /home/camp/ntben/localnewsetup/gpaw-setups-0.9.9672/O.RPBE.gz
> cutoffs: 0.67(comp), 1.17(filt), 0.83(core), lmax=2
> valence states:
> energy radius
> 2s(2) -24.041 0.688
> 2p(4) -8.984 0.598
> *s 3.170 0.688
> *p 18.228 0.598
> *d 0.000 0.619
>
> LCAO basis set for O:
> Name: dzp
> Basis set was loaded from file /home/camp/ntben/localnewsetup/gpaw-setups-0.9.9672/O.dzp.basis.gz
> Number of radial functions: 5
> Number of spherical harmonics: 13
> l=0, rc=4.3438 Bohr: 2s-sz confined orbital
> l=1, rc=5.3906 Bohr: 2p-sz confined orbital
> l=0, rc=2.2969 Bohr: 2s-dz split-valence wave
> l=1, rc=2.8906 Bohr: 2p-dz split-valence wave
> l=2, rc=5.3906 Bohr: d-type Gaussian polarization
>
> Using the RPBE Exchange-Correlation Functional.
> Spin-Paired Calculation
> Total Charge: 0.000000
> Fermi Temperature: 0.100000
> Wave functions: Uniform real-space grid
> Kinetic energy operator: 6*3+1=19 point O(h^6) finite-difference Laplacian
> Eigensolver: rmm-diis
> XC and Coulomb potentials evaluated on a 56*56*40 grid
> Interpolation: tri-quintic (5. degree polynomial)
> Poisson solver: Jacobi solver with 4 multi-grid levels
> Stencil: 6*3+1=19 point O(h^6) finite-difference Laplacian
> Reference Energy: -254678.132380
>
> Total number of cores used: 8
> Parallelization over k-points: 8
> MatrixOperator buffer_size: default value or
> see value of nblock in input file
> Diagonalizer layout: Serial LAPACK
> Orthonormalizer layout: Serial LAPACK
>
> Symmetries present: 8
> 490 k-points (7 x 7 x 10 Monkhorst-Pack grid)
> 80 k-points in the Irreducible Part of the Brillouin Zone
> Linear Mixing Parameter: 0.1
> Pulay Mixing with 3 Old Densities
> Damping of Long Wave Oscillations: 50
>
> Convergence Criteria:
> Total Energy Change: 0.0005 eV / electron
> Integral of Absolute Density Change: 0.0001 electrons
> Integral of Absolute Eigenstate Change: 4e-08 eV^2
> Number of Atoms: 6
> Number of Atomic Orbitals: 102
> Number of Bands in Calculation: 102
> Bands to Converge: Occupied States Only
> Number of Valence Electrons: 56
> log10-error: Total Iterations:
> Time WFS Density Energy Fermi Poisson
> iter: 1 11:13:33 +3.6 -45.656478 4 15
> iter: 2 11:13:57 -0.2 -48.491408 4
> iter: 3 11:14:21 -1.0 -48.580999 2
> iter: 4 11:14:46 -1.1 -1.2 -44.853351 9 7
> iter: 5 11:15:11 -1.0 -1.3 -42.964989 9 8
> iter: 6 11:15:36 -1.3 -1.6 -43.635732 5 7
> iter: 7 11:16:01 -1.5 -1.6 -43.108654 7 8
> iter: 8 11:16:26 -2.2 -2.0 -43.030008 3 7
> iter: 9 11:16:51 -2.7 -2.1 -42.981454 3 6
> iter: 10 11:17:16 -3.3 -2.4 -42.980521 3 5
> iter: 11 11:17:41 -3.3 -2.4 -42.981560 3 6
> iter: 12 11:18:06 -3.5 -2.6 -42.984099 3 6
> iter: 13 11:18:31 -3.8 -2.7 -42.983144 3 5
> iter: 14 11:18:56 -4.1 -3.0 -42.983235 2 5
> iter: 15 11:19:20 -4.8 -3.1 -42.983202 2 4
> iter: 16 11:19:45 -4.0 -3.2 -42.982700 3 5
> iter: 17 11:20:10 -4.8 -3.7 -42.982800 2 3
> iter: 18 11:20:35 -5.6 -3.7 -42.982379 2 3
> iter: 19 11:21:00 -6.0 -3.8 -42.982337 2 3
> iter: 20 11:21:25 -5.7 -3.8 -42.982226 2 3
> iter: 21 11:21:50 -6.5 -4.0 -42.982674 2 2
> iter: 22 11:22:14 -6.5 -4.0 -42.982677 2 3
> iter: 23 11:22:39 -6.0 -4.1 -42.982764 2 3
> iter: 24 11:23:04 -6.8 -4.2 -42.982445 2 2
> iter: 25 11:23:29 -7.3 -4.3 -42.982215 2 2
> iter: 26 11:23:54 -6.8 -4.4 -42.982439 2 3
> iter: 27 11:24:19 -7.0 -4.6 -42.982253 2 2
> iter: 28 11:24:43 -7.8 -4.7 -42.982604 1 1
> ------------------------------------
> Converged After 28 Iterations.
>
> Energy Contributions Relative to Reference Atoms:(reference = -254678.132380)
> -------------------------
> Kinetic: +41.038599
> Potential: -50.993219
> External: +0.000000
> XC: -33.726018
> Entropy (-ST): -0.121683
> Local: +0.758877
> -------------------------
> Free Energy: -43.043445
> Zero Kelvin: -42.982604
>
> Fermi Level: 8.06960
>
>
> Total Charge: -0.000000 electrons
> Dipole Moment: [-0.95402944 -0.95402944 -2.94283037]
> Memory usage: 400.61 MB
>
> ============================================================
> Timing: incl. excl.
> ============================================================
> Initialization: 8.716 0.942 0.1% |
> Hamiltonian: 0.852 0.011 0.0% |
> Atomic: 0.338 0.001 0.0% |
> XC Correction: 0.337 0.337 0.0% |
> Communicate energies: 0.000 0.000 0.0% |
> Hartree integrate/restrict: 0.002 0.002 0.0% |
> Initialize Hamiltonian: 0.002 0.002 0.0% |
> Poisson: 0.394 0.394 0.1% |
> XC 3D grid: 0.104 0.104 0.0% |
> vbar: 0.001 0.001 0.0% |
> LCAO initialization: 6.922 0.795 0.1% |
> LCAO eigensolver: 1.035 0.001 0.0% |
> Atomic Hamiltonian: 0.012 0.012 0.0% |
> Calculate projections: 0.010 0.010 0.0% |
> Distribute overlap matrix: 0.001 0.001 0.0% |
> Orbital Layouts: 0.161 0.161 0.0% |
> Potential matrix: 0.656 0.656 0.1% |
> Sum over cells: 0.193 0.193 0.0% |
> LCAO to grid: 2.848 2.848 0.4% |
> Set positions (LCAO WFS): 2.244 0.002 0.0% |
> Basic WFS set positions: 0.001 0.001 0.0% |
> Basis functions set positions: 0.000 0.000 0.0% |
> TCI: Calculate S, T, P: 2.241 2.241 0.3% |
> SCF-cycle: 700.769 0.216 0.0% |
> Density: 4.911 0.000 0.0% |
> Atomic density matrices: 0.153 0.153 0.0% |
> Mix: 0.464 0.464 0.1% |
> Multipole moments: 0.002 0.002 0.0% |
> Pseudo density: 4.291 4.075 0.6% |
> Symmetrize density: 0.216 0.216 0.0% |
> Hamiltonian: 14.386 0.283 0.0% |
> Atomic: 8.506 0.017 0.0% |
> XC Correction: 8.489 8.489 1.2% |
> Communicate energies: 0.000 0.000 0.0% |
> Hartree integrate/restrict: 0.071 0.071 0.0% |
> Poisson: 2.945 2.945 0.4% |
> XC 3D grid: 2.559 2.559 0.4% |
> vbar: 0.021 0.021 0.0% |
> Orthonormalize: 169.089 0.017 0.0% |
> Band Layouts: 0.358 0.003 0.0% |
> Inverse Cholesky: 0.355 0.355 0.1% |
> calc_s_matrix: 54.200 54.200 7.6% |--|
> projections: 76.031 76.031 10.7% |---|
> rotate_psi: 38.483 38.483 5.4% |-|
> RMM-DIIS: 368.106 171.084 24.1% |---------|
> Apply hamiltonian: 27.076 27.076 3.8% |-|
> precondition: 95.610 95.610 13.5% |----|
> projections: 74.335 74.335 10.5% |---|
> Subspace diag: 144.062 0.013 0.0% |
> Band Layouts: 3.706 0.007 0.0% |
> Diagonalize: 3.690 3.690 0.5% |
> Distribute results: 0.009 0.009 0.0% |
> calc_h_matrix: 65.289 38.585 5.4% |-|
> Apply hamiltonian: 26.704 26.704 3.8% |-|
> rotate_psi: 75.054 75.054 10.6% |---|
> Other: 0.019 0.019 0.0% |
> ============================================================
> Total: 709.504 100.0%
> ============================================================
> date: Tue Nov 27 11:24:43 2012
>
> Niels Theis Bendtsen
> Ph.d.-studerende
> FYS-CAMD
> DTU Fysik
> Bygning 307
> 2800 Kgs. Lyngby
> Direkte telefon 45253177
> ntben at fysik.dtu.dk
> www.fysik.dtu.dk
>
>
>
> _______________________________________________
> gpaw-users mailing list
> gpaw-users at listserv.fysik.dtu.dk
> https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users
>
>
--
***********************************
Marcin Dulak
Technical University of Denmark
Department of Physics
Building 307, Room 229
DK-2800 Kongens Lyngby
Denmark
Tel.: (+45) 4525 3157
Fax.: (+45) 4593 2399
email: Marcin.Dulak at fysik.dtu.dk
***********************************
More information about the gpaw-users
mailing list