This thread is locked.Only browsing is available.
Top Page > Browsing
nan MulP in Cluster solver
Date: 2008/05/15 23:18
Name: Jian Zhou

Dear all,

I find some problems when calculating the large size of gold cluster (for example, 30 gold atoms) using the Cluster solver, the program have problems at the first SCF
(scf.EigenvalueSolver = Cluster):

******************* MD= 1 SCF= 1 *******************
<Cluster> Eigenvalue problem...
Atom 1 MulP nan nan sum nan diff nan
Atom 2 MulP nan nan sum nan diff nan
Atom 3 MulP nan nan sum nan diff nan
Atom 4 MulP nan nan sum nan diff nan
Atom 5 MulP nan nan sum nan diff nan
Atom 6 MulP nan nan sum nan diff nan
......


But it is very strange that the program does not stop after that. It
finishes normally after 2 SCFs (Please see the log file in attachment):
[zj@material aaa]$ grep dUele log
<DFT> Uele = 0.000000000000 dUele = 1.000000000000
<DFT> Uele = 0.000000000000 dUele = 0.000000000000

At the end of the screen output, it shows:
The calculation was normally finished. (proc= 1)
The calculation was normally finished. (proc= 3)
.....

I found that If I changed the scf.EigenvalueSolver to band and also set
the 2x1x1 kpoints grid and using a proper unit cell, the programs looks
OK (at least with a proper MulP). Therefore I think it maybe has some problems with cluster solver.

In fact, I found this problem for other big systems (maybe 100-200
atoms, but without gold atoms). For the same structure, the
cluster solver is wrong but the band solver is OK.

The openMX seems have a restrict that if I use the 1x1x1 kpoint grid and
use the band solver, the program will set the solver to Cluster
automatically. Otherwise, I can use the band solver with Gamma point
instead of the cluster solver. Of course I can use the band solver by
using more than 1 kpoint, but I think it is not good and not necessary
for large systems.

Have you ever got such a problem?

Thank you in advance.

Best wishes,

Jian Zhou


The following is my atom coordinates and main settings:

Atoms.Number 30
Atoms.SpeciesAndCoordinates.Unit Ang # Ang|AU
<Atoms.SpeciesAndCoordinates # Unit=Ang.
1 Au -1.851468 -3.089829 0.955733 5.500 5.500
2 Au -2.204542 -3.484781 5.956457 5.500 5.500
3 Au -1.842557 -0.116475 2.059177 5.500 5.500
4 Au -3.714504 -0.926255 6.978288 5.500 5.500
5 Au 0.966342 -3.555008 1.457011 5.500 5.500
6 Au 3.647600 0.813239 1.794046 5.500 5.500
7 Au -2.952540 -2.234019 2.996035 5.500 5.500
8 Au -0.596209 -3.818818 3.227101 5.500 5.500
9 Au -3.721432 -0.959726 4.602283 5.500 5.500
10 Au 3.005771 -2.097924 4.941865 5.500 5.500
11 Au -3.057709 1.653883 3.985865 5.500 5.500
12 Au 3.575300 -1.335641 1.859505 5.500 5.500
13 Au -0.273627 1.189263 3.648246 5.500 5.500
14 Au 1.897724 -3.005701 3.222614 5.500 5.500
15 Au 1.767337 -3.033185 7.118495 5.500 5.500
16 Au 1.700976 -0.132073 7.985693 5.500 5.500
17 Au 0.466800 -3.940946 5.529562 5.500 5.500
18 Au -3.009616 -2.328498 8.469585 5.500 5.500
19 Au 1.856970 0.668556 5.586480 5.500 5.500
20 Au -0.834999 -1.538543 4.588304 5.500 5.500
21 Au 1.606005 -3.303303 9.243170 5.500 5.500
22 Au -2.191187 3.397278 5.579442 5.500 5.500
23 Au -3.728960 -1.429496 0.933060 5.500 5.500
24 Au 3.995744 -0.113659 3.675384 5.500 5.500
25 Au 0.942697 1.711222 9.595113 5.500 5.500
26 Au 3.735753 -0.884494 6.411868 5.500 5.500
27 Au 1.511507 -0.368247 3.408620 5.500 5.500
28 Au 0.790647 -1.380096 1.263565 5.500 5.500
29 Au 2.966570 2.259385 9.333115 5.500 5.500
30 Au 3.505382 1.628477 7.357861 5.500 5.500
Atoms.SpeciesAndCoordinates>
scf.XcType GGA-PBE # LDA|LSDA-CA|LSDA-PW|GGA-PBE
scf.SpinPolarization on # On|Off|NC
scf.partialCoreCorrection on # On|Off
scf.ElectronicTemperature 250.0 # default=300 (K)
scf.energycutoff 150.0 # default=150 (Ry)
#scf.Ngrid 100 100 150
scf.maxIter 200 # default=40
scf.EigenvalueSolver Cluster # DC|GDC|Cluster|Band
scf.Kgrid 1 1 1 # means 4x4x4
scf.Mixing.Type Rmm-Diisk # Simple|Rmm-Diis|Gr-Pulay|Kerker|Rmm-Diisk
scf.Init.Mixing.Weight 0.010 # default=0.30
scf.Min.Mixing.Weight 0.001 # default=0.001
scf.Max.Mixing.Weight 0.100 # default=0.40
scf.Kerker.factor 2.0 # default=1.0
scf.Mixing.EveryPulay 1 # default=5
scf.Mixing.History 30 # default=5
scf.Mixing.StartPulay 15 # default=6
scf.criterion 1.0e-7 # default=1.0e-6 (Hartree)
scf.lapack.dste dstevx # dstegr|dstedc|dstevx, default=dstevx











メンテ
Page: [1]

Re: nan MulP in Cluster solver ( No.1 )
Date: 2009/01/06 16:50
Name: Loc Duong Dinh  <mambom1902@yahoo.com>

I have the same problem when using Empty atoms.
メンテ
Re: nan MulP in Cluster solver ( No.2 )
Date: 2009/01/06 18:14
Name: Loc Duong Dinh  <mambom1902@yahoo.com>

Num. of grids overlapping with atom 1 = 55707
Num. of grids overlapping with atom 2 = 55707
Num. of grids overlapping with atom 3 = 55707

This is the output file when I try to test NaCl_FC example with openmx3.4. The problem is that the value is 'nan' and I can using Xcrysden to view cube file.





Num. of grids overlapping with atom 4 = 55707
Num. of grids overlapping with atom 5 = 131995
Num. of grids overlapping with atom 6 = 131995
Num. of grids overlapping with atom 7 = 131995
Num. of grids overlapping with atom 8 = 131995

*******************************************************
SCF calculation at MD = 1
*******************************************************

<MD= 1> Calculation of the overlap matrix
<MD= 1> Calculation of the nonlocal matrix
<MD= 1> Calculation of the VNA projector matrix

******************* MD= 1 SCF= 1 *******************
<Band> Eigenvalue problem...
KGrids1: -0.40000 -0.20000 0.00000 0.20000 0.40000
KGrids2: -0.40000 -0.20000 -0.00000 0.20000 0.40000
KGrids3: -0.40000 -0.20000 0.00000 0.20000 0.40000
<Band_DFT> Eigen, time=1.604980
<Band_DFT> DM, time=1.816489
Atom 1 MulP 0.0654 0.0654 sum 0.1309
Atom 2 MulP 0.1885 0.1885 sum 0.3770
Atom 3 MulP 0.1891 0.1891 sum 0.3783
Atom 4 MulP 0.1903 0.1903 sum 0.3807
Atom 5 MulP 0.1026 0.1026 sum 0.2051
Atom 6 MulP 0.0766 0.0766 sum 0.1532
Atom 7 MulP 0.0924 0.0924 sum 0.1849
Atom 8 MulP 0.0950 0.0950 sum 0.1899
<DFT> Total Spin Moment (muB) = 0.000000000000
<DFT> Mixing_weight= 0.010000000000
<DFT> Uele = 37.595314043205 dUele = 62.404685956795
<DFT> NormRD = 1.000000000000 Criterion = 0.000000001000

******************* MD= 1 SCF= 2 *******************
<Poisson> Poisson's equation using FFT...
<Set_Hamiltonian> Hamiltonian matrix for VNA+dVH+Vxc...
<Band> Eigenvalue problem...
KGrids1: -0.40000 -0.20000 0.00000 0.20000 0.40000
KGrids2: -0.40000 -0.20000 -0.00000 0.20000 0.40000
KGrids3: -0.40000 -0.20000 0.00000 0.20000 0.40000
<Band_DFT> Eigen, time=1.515009
<Band_DFT> DM, time=1.802092
Atom 1 MulP 0.0654 0.0654 sum 0.1309
Atom 2 MulP 0.1885 0.1885 sum 0.3770
Atom 3 MulP 0.1891 0.1891 sum 0.3783
Atom 4 MulP 0.1903 0.1903 sum 0.3807
Atom 5 MulP 0.1026 0.1026 sum 0.2051
Atom 6 MulP 0.0766 0.0766 sum 0.1532
Atom 7 MulP 0.0924 0.0924 sum 0.1849
Atom 8 MulP 0.0950 0.0950 sum 0.1899
<DFT> Total Spin Moment (muB) = 0.000000000000
<DFT> Mixing_weight= 0.010000000000
<DFT> Uele = 37.595314043205 dUele = 0.000000000000
<DFT> NormRD = nan Criterion = 0.000000001000

Band_DFT_MO start
kpoint=1 /1
<MD= 1> Force calculation
Force calculation #1
Force calculation #2
Force calculation #3
Force calculation #4
Force calculation #5
<MD= 1> Total Energy
Force calculation #6
Force calculation #7

*******************************************************
Dipole moment (Debye)
*******************************************************

Absolute D nan

Dx Dy Dz
Total nan nan nan
Core 378.58875986 378.58875986 378.58875986
Electron nan nan nan
Back ground 6.76051357 6.76051357 6.76051357

*******************************************************
Total Energy (Hartree) at MD = 1
*******************************************************

Uele = 37.595314043205

Ukin = 2.097095000820
UH0 = nan
UH1 = nan
Una = nan
Unl = -1.083641384921
Uxc0 = 0.000000000000
Uxc1 = 0.000000000000
Ucore = 1179.895817641656
Uhub = 0.000000000000
Ucs = 0.000000000000
Uzs = 0.000000000000
Uzo = 0.000000000000
Uef = 0.000000000000
Utot = nan

Note:

Utot = Ukin+UH0+UH1+Una+Unl+Uxc0+Uxc1+Ucore+Uhub+Ucs+Uzs+Uzo+Uef

Uele: band energy
Ukin: kinetic energy
UH0: electric part of screened Coulomb energy
UH1: difference electron-electron Coulomb energy
Una: neutral atom potential energy
Unl: non-local potential energy
Uxc0: exchange-correlation energy for alpha spin
Uxc1: exchange-correlation energy for beta spin
Ucore: core-core Coulomb energy
Uhub: LDA+U energy
Ucs: constraint energy for spin orientation
Uzs: Zeeman term for spin magnetic moment
Uzo: Zeeman term for orbital magnetic moment
Uef: electric energy by electric field

(see also PRB 72, 045121(2005) for the energy contributions)


*******************************************************
Computational times (s) at MD = 1
*******************************************************

DFT in total = 276.64199

Set_OLP_Kin = 26.08867
Set_Nonlocal = 68.68535
Set_Hamiltonian = 18.73011
Poisson = 0.01128
diagonalization = 6.76643
Mixing_DM = 0.10978
Force = 127.66870
Total_Energy = 13.11962
Set_Aden_Grid = 0.46097
Set_Orbitals_Grid = 1.32194
Set_Density_Grid = 13.34082

*******************************************************
MD or geometry opt. at MD = 1
*******************************************************


<Voronoi_Charge> calculate Voronoi charges

outputting data on grids to files...


***********************************************************
***********************************************************
Computational Time (second)
***********************************************************
***********************************************************

Min_ID Min_Time Max_ID Max_Time
Total Computational Time = 0 1451.796 0 1451.796
readfile = 0 68.350 0 68.350
truncation = 0 48.132 0 48.132
MD_pac = 0 0.013 0 0.013
DFT = 0 276.642 0 276.642

*** In DFT ***

Set_OLP_Kin = 0 26.089 0 26.089
Set_Nonlocal = 0 68.685 0 68.685
Set_Hamiltonian = 0 18.730 0 18.730
Poisson = 0 0.011 0 0.011
Diagonalization = 0 6.766 0 6.766
Mixing_DM = 0 0.110 0 0.110
Force = 0 127.669 0 127.669
Total_Energy = 0 13.120 0 13.120
Set_Aden_Grid = 0 0.461 0 0.461
Set_Orbitals_Grid = 0 1.322 0 1.322
Set_Density_Grid = 0 13.341 0 13.341
Others = 0 0.338 0 0.338

The calculation was normally finished. (proc= 0)


メンテ
Re: nan MulP in Cluster solver ( No.3 )
Date: 2009/02/08 18:24
Name: H Jeong

[1] Main article's problem

Please set "Number of scf.Mixing.History " smaller than "Mixing.StartPulay"

In input-file, 30 > 15

I expect "This comes from the exception event in RMM-DIIS routine".

[2] Problem in the " Empty Atom Scheme" remedied by the issue of patches of OpenMX 3.4
メンテ
Re: nan MulP in Cluster solver ( No.4 )
Date: 2009/02/08 19:04
Name: H Jeong

Some mistake.

The First problem starts from 1 SCF step. (before mixing process)
then other reasons in there (ex. related to initial setup routine of H matrix.
maybe minor bug)

Suggest to use the newest version of OpenMX.
メンテ
Re: nan MulP in Cluster solver ( No.5 )
Date: 2009/03/03 16:37
Name: Jian Zhou

OK, In fact, it is the compiler (ifort 9.1) proplems. A new version of ifort (11.0) is OK.
メンテ

Page: [1]