[CPMD-list] TDDFT diagonalization problem
Łukasz Walewski
ljw at icm.edu.pl
Wed Dec 5 09:47:53 UTC 2007
Dear Colleagues !
I want to optimize the geometry of malonaldehyde in 2nd singlet excited
state - this is more of the calibration and learning than a serious
research. 1st singlet and 1st triplet states optimized very well using
BLYP functional, TM normconserving pseudos and the cutoff of 85 Ry. In
2nd singlet excited state the problems occur at the very beginning when
it comes to diagonalizing TDDFT matrix (ELECTRONIC SPECTRA run).
Davidson method simply does not converge and the DIIS optimizer turns
on. After some steps something goes wrong (NaN). On the other hand there
are many reports of successful calculation of the reaction pathway in
the S2 state using e.g. CASSCF method. Is there something prohibitive in
TDDFT or Tamm-Dancoff approximation that I (new to TDDFT) did not fully
understand or is it just a technical issue ?
Input:
&CPMD
ELECTRONIC SPECTRA
DIAGONALIZATION LANCZOS
ISOLATED MOLECULE
CENTER MOLECULE
&END
&TDDFT
STATES SINGLET
2
TAMM-DANCOFF
DAVIDSON PARAMETER
150 1.D-7 50
DAVIDSON RDIIS
40 3 1.D-3
&END
&SYSTEM
ANGSTROM
SYMMETRY
8
CELL ABSOLUTE
10.0 10.0 7.0 0 0 0
CUTOFF
85.0
&END
&DFT
FUNCTIONAL BLYP
GC-CUTOFF
5.D-6
&END
&ATOMS
*O_MT_BLYP.psp KLEINMAN-BYLANDER
LMAX=P
2
6.191021000000 6.698682000000 7.499957000000
8.640750000000 6.638256000000 7.499966000000
*C_MT_BLYP.psp KLEINMAN-BYLANDER
LMAX=P
3
8.695525000000 7.938102000000 7.500018000000
7.465729000000 8.689156000000 7.499907000000
6.200399000000 7.999649000000 7.500009000000
*H_MT_BLYP.psp KLEINMAN-BYLANDER
LMAX=S
4
7.444409000000 6.259053000000 7.499975000000
9.664199000000 8.401942000000 7.500090000000
7.465734000000 9.762156000000 7.499935000000
5.256652000000 8.510208000000 7.500074000000
CONSTRAINTS
FIX STRUCTURE
1
DIFFER 1 6 2 .0
END CONSTRAINTS
&END
(partial) output:
================================================================
== DAVIDSON DIAGONALISATION OF TDDFT MATRIX ==
================================================================
ITER STATES SUBSPACE RESIDUAL TCPU
1 0 2 0.14709947E-02 11.84
2 0 4 0.18604014E-03 11.77
3 0 6 0.40119050E-02 11.60
4 0 8 0.33008935E-01 11.97
5 0 10 0.96734776E-02 8.88
6 0 12 0.25567779E-02 9.94
7 0 14 0.10075854E-02 9.75
8 1 16 0.45615021E-03 10.10
9 1 17 0.37400695E-03 5.19
10 1 18 0.20701144E-03 5.19
11 1 19 0.30069471E-03 5.37
12 1 20 0.53039825E-03 5.42
>>> RESIDUAL DIIS OPTIMIZATION OF ROOT 2
ITER ENERGY SUBSPACE RESIDUAL TCPU
1 5.32330 1 0.53040762E-03 4.91
2 5.32026 2 0.14856434E-03 4.84
3 5.31755 3 0.87516146E-04 4.74
4 5.31411 4 0.62037932E-04 4.30
5 5.31152 4 0.57641858E-04 5.75
6 5.31010 4 0.55942548E-04 7.48
7 5.30919 4 0.55578369E-04 7.16
8 5.31105 4 0.54668835E-04 4.52
9 5.31164 4 0.53386220E-04 4.80
10 5.31267 4 0.51513120E-04 4.57
11 5.32171 4 0.43997184E-04 4.68
12 5.34103 4 0.32928663E-04 4.66
13 5.35106 4 0.26192807E-04 4.54
14 5.35273 4 0.20616123E-04 4.76
15 5.35421 4 0.14095987E-04 4.46
16 5.35466 4 0.81855763E-05 4.59
17 5.35440 4 0.74781079E-05 4.45
18 5.35413 4 0.57619201E-05 4.56
19 5.35396 4 0.23649238E-05 4.60
20 5.35398 4 0.43807682E-05 4.42
21 5.35394 4 0.18597910E-05 6.86
22 5.35405 4 0.50842677E-05 7.01
23 5.35394 4 0.18570448E-05 14.03
24 NaN 4 NaN 7.01
25 NaN 4 NaN 7.08
26 NaN 4 NaN 7.04
27 NaN 4 NaN 7.41
28 NaN 4 NaN 7.27
29 NaN 4 NaN 7.18
30 NaN 4 NaN 7.09
31 NaN 4 NaN 6.42
32 NaN 4 NaN 6.14
33 NaN 4 NaN 6.05
34 NaN 4 NaN 6.26
35 NaN 4 NaN 6.44
36 NaN 4 NaN 6.99
37 NaN 4 NaN 7.12
38 NaN 4 NaN 7.74
39 NaN 4 NaN 7.38
40 NaN 4 NaN 7.37
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!! DAVIDSON| NOT ALL ROOTS ARE CONVERGED !!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Thanks and regards to all!
Lukasz
--
Łukasz Walewski <ljw at icm.edu.pl>
ICM UW
More information about the CPMD-list
mailing list