TDDFT cis energy converg not met

I am trying to do a TDDFT calculation on a tetramer system containing total of 160 atoms, the atoms are C, N, O, H. My aim is to get the excitation energy of 80 excited state. Here is my input part:

$rem
EXCHANGE wb97xd
BASIS def2-svp
CIS_N_ROOTS 80
SYMMETRY false
SYM_IGNORE true
CIS_SINGLETS true
CIS_TRIPLETS false
SOLVENT_METHOD pcm
CIS_AMPL_PRINT -1
gui 2
$end

$pcm
Theory CPCM
Method SWIG
Solver Inversion
Radii Bondi
$end

$solvent
Dielectric 14
OpticalDielectric 4
$end

My problem is that the cis energy converge not met, and I don’t know why. Here is the problem part:

Direct TDDFT/TDA calculation will be performed
** Exchange: 0.2220 Hartree-Fock + 1.0000 wB97X-D + LR-HF**
** Correlation: 1.0000 wB97X-D**
** Using SG-2 standard quadrature grid**
** Dispersion: Grimme D**
** Singlet excitation energies requested**
** Adding PCM contribution to the XC response**
** CIS energy converged when residual is below 10e- 6**
** ---------------------------------------------------**
** Iter Rts Conv Rts Left Ttl Dev Max Dev**
** ---------------------------------------------------**
** 1 0 80 0.037216 0.000870**
** 2 0 80 7.246293 0.457001**
** 3 0 80 89.592577 11.679501**

Can someone explain why is this happening?

Is there an error printed at the end of the output? The default maximum number of CIS iterative cycles by default is 30, your post only shows to 3, so I am curious if your job ended abruptly or if there are other errors printed at the end of the output. In addition, the job should have 30 cycles to try and converge the CIS energy and in the 3 cycles none of the roots are converged. Therefore this job should take additional iterations.

It would also be useful if you add the $molecule section so I can try running this job to better assist with this issue.

It’s also really helpful if you can construct the minimal example that fails. For example, if you turn off the PCM, does it still fail? Request few roots? Try a trivially small basis set? This helps to narrow down what is causing the problem.

Thanks for your reply. And, yes, there is an error printed at the end of the output, because the Ttl Dev continued to become bigger, and can’t convergence. I tried “MAX_CIS_CYCLES 50”, the Ttl Dev still to become bigger and the roots can’t converged (I use 16 cores to run my job). Now I am tring to use a single core to run my job (the same job), it still running, and the Ttl Dev have a trend to become smaller, it may met convergence in the end, but it still need time to get the final result.

So I am very confused about that.

Here is my $molecule part:
$molecule
0 1
O 10.74713407 13.570009 11.37115993
C 13.95845326 13.7261046 15.28628522
C 13.71270617 12.8013118 14.27568466
C 12.77082528 13.0163114 13.2998949
C 12.04270483 14.1575764 13.27161113
C 11.00588546 14.387302 12.2555714
H 14.21884136 11.9854914 14.25936711
H 12.6249547 12.3462784 12.62869948
C 14.97313318 13.5155228 16.32952088
C 15.14279276 14.5139456 17.34338494
C 15.75350585 12.3639496 16.37194652
C 16.15685426 14.2827474 18.3539855
C 16.69747607 12.1680938 17.34229711
H 15.62660796 11.6880262 15.70183894
C 16.90380678 13.0943592 18.33222876
H 17.22588193 11.3669994 17.33141874
C 17.96565138 12.841072 19.3602347
N 18.13904915 13.8851454 20.28598397
O 18.67672922 11.8721012 19.40701169
H 18.79679821 13.6716184 20.90713888
O 9.71208987 17.5798988 13.22809765
N 10.24976995 15.5668546 12.34912538
C 13.41568591 15.9364772 16.30558847
C 13.24602633 14.9380544 15.2917244
C 12.23196484 15.1692526 14.28112385
C 11.48501232 16.3576408 14.30288059
C 11.69134303 17.2839062 15.29281224
C 12.63531325 17.0880504 16.26316282
C 10.42316771 16.610928 13.27487464
H 9.59202089 15.7803816 11.72797046
H 11.16293717 18.0850006 15.30369061
H 12.76221114 17.7639738 16.9332704
C 14.43036583 15.7258954 17.34882413
C 14.67611293 16.6506882 18.35942468
C 15.61799382 16.4356886 19.33521445
H 14.16997774 17.4665086 18.37574224
C 16.34611426 15.2944236 19.36349821
H 15.7638644 17.1057216 20.00640987
C 17.38293363 15.064698 20.37953795
O 17.64168503 15.881991 21.26394941
O 15.61913407 13.570009 11.37115993
C 18.83045326 13.7261046 15.28628522
C 18.58470617 12.8013118 14.27568466
C 17.64282528 13.0163114 13.2998949
C 16.91470483 14.1575764 13.27161113
C 15.87788546 14.387302 12.2555714
H 19.09084136 11.9854914 14.25936711
H 17.4969547 12.3462784 12.62869948
C 19.84513318 13.5155228 16.32952088
C 20.01479276 14.5139456 17.34338494
C 20.62550585 12.3639496 16.37194652
C 21.02885426 14.2827474 18.3539855
C 21.56947607 12.1680938 17.34229711
H 20.49860796 11.6880262 15.70183894
C 21.77580678 13.0943592 18.33222876
H 22.09788193 11.3669994 17.33141874
C 22.83765138 12.841072 19.3602347
N 23.01104915 13.8851454 20.28598397
O 23.54872922 11.8721012 19.40701169
H 23.66879821 13.6716184 20.90713888
O 14.58408987 17.5798988 13.22809765
N 15.12176995 15.5668546 12.34912538
C 18.28768591 15.9364772 16.30558847
C 18.11802633 14.9380544 15.2917244
C 17.10396484 15.1692526 14.28112385
C 16.35701232 16.3576408 14.30288059
C 16.56334303 17.2839062 15.29281224
C 17.50731325 17.0880504 16.26316282
C 15.29516771 16.610928 13.27487464
H 14.46402089 15.7803816 11.72797046
H 16.03493717 18.0850006 15.30369061
H 17.63421114 17.7639738 16.9332704
C 19.30236583 15.7258954 17.34882413
C 19.54811293 16.6506882 18.35942468
C 20.48999382 16.4356886 19.33521445
H 19.04197774 17.4665086 18.37574224
C 21.21811426 15.2944236 19.36349821
H 20.6358644 17.1057216 20.00640987
C 22.25493363 15.064698 20.37953795
O 22.51368503 15.881991 21.26394941
O 5.87513407 13.570009 11.37115993
C 9.08645326 13.7261046 15.28628522
C 8.84070617 12.8013118 14.27568466
C 7.89882528 13.0163114 13.2998949
C 7.17070483 14.1575764 13.27161113
C 6.13388546 14.387302 12.2555714
H 9.34684136 11.9854914 14.25936711
H 7.7529547 12.3462784 12.62869948
C 10.10113318 13.5155228 16.32952088
C 10.27079276 14.5139456 17.34338494
C 10.88150585 12.3639496 16.37194652
C 11.28485426 14.2827474 18.3539855
C 11.82547607 12.1680938 17.34229711
H 10.75460796 11.6880262 15.70183894
C 12.03180678 13.0943592 18.33222876
H 12.35388193 11.3669994 17.33141874
C 13.09365138 12.841072 19.3602347
N 13.26704915 13.8851454 20.28598397
O 13.80472922 11.8721012 19.40701169
H 13.92479821 13.6716184 20.90713888
O 4.84008987 17.5798988 13.22809765
N 5.37776995 15.5668546 12.34912538
C 8.54368591 15.9364772 16.30558847
C 8.37402633 14.9380544 15.2917244
C 7.35996484 15.1692526 14.28112385
C 6.61301232 16.3576408 14.30288059
C 6.81934303 17.2839062 15.29281224
C 7.76331325 17.0880504 16.26316282
C 5.55116771 16.610928 13.27487464
H 4.72002089 15.7803816 11.72797046
H 6.29093717 18.0850006 15.30369061
H 7.89021114 17.7639738 16.9332704
C 9.55836583 15.7258954 17.34882413
C 9.80411293 16.6506882 18.35942468
C 10.74599382 16.4356886 19.33521445
H 9.29797774 17.4665086 18.37574224
C 11.47411426 15.2944236 19.36349821
H 10.8918644 17.1057216 20.00640987
C 12.51093363 15.064698 20.37953795
O 12.76968503 15.881991 21.26394941
O 1.00313407 13.570009 11.37115993
C 4.21445326 13.7261046 15.28628522
C 3.96870617 12.8013118 14.27568466
C 3.02682528 13.0163114 13.2998949
C 2.29870483 14.1575764 13.27161113
C 1.26188546 14.387302 12.2555714
H 4.47484136 11.9854914 14.25936711
H 2.8809547 12.3462784 12.62869948
C 5.22913318 13.5155228 16.32952088
C 5.39879276 14.5139456 17.34338494
C 6.00950585 12.3639496 16.37194652
C 6.41285426 14.2827474 18.3539855
C 6.95347607 12.1680938 17.34229711
H 5.88260796 11.6880262 15.70183894
C 7.15980678 13.0943592 18.33222876
H 7.48188193 11.3669994 17.33141874
C 8.22165138 12.841072 19.3602347
N 8.39504915 13.8851454 20.28598397
O 8.93272922 11.8721012 19.40701169
H 9.05279821 13.6716184 20.90713888
O -0.03191013 17.5798988 13.22809765
N 0.50576995 15.5668546 12.34912538
C 3.67168591 15.9364772 16.30558847
C 3.50202633 14.9380544 15.2917244
C 2.48796484 15.1692526 14.28112385
C 1.74101232 16.3576408 14.30288059
C 1.94734303 17.2839062 15.29281224
C 2.89131325 17.0880504 16.26316282
C 0.67916771 16.610928 13.27487464
H -0.15197911 15.7803816 11.72797046
H 1.41893717 18.0850006 15.30369061
H 3.01821114 17.7639738 16.9332704
C 4.68636583 15.7258954 17.34882413
C 4.93211293 16.6506882 18.35942468
C 5.87399382 16.4356886 19.33521445
H 4.42597774 17.4665086 18.37574224
C 6.60211426 15.2944236 19.36349821
H 6.0198644 17.1057216 20.00640987
C 7.63893363 15.064698 20.37953795
O 7.89768503 15.881991 21.26394941
$end

Thank you for your help.

Thank you for your reply and thanks for your advice, I will give it a try. I have tried to change the “EXCHANGE” to wb97xd, still the Ttl Dev continued to become bigger, and can’t convergence.I will try something else.

I am not quite sure what is happening because I’m unable to reproduce the behavior that you describe with Q-Chem 5.4. However, I do run into various memory issues on my hardware, using your input file. I was able to avoid these by selecting a smaller number of CIS roots. For PCM calculations on large molecules, you might also considering reducing the number of grid points used to discretize the solute cavity surface, keywords “HPoints” (for H atoms) and “HeavyPoints” (for other atoms) in the $pcm section

Yes, when I reduce the number of CIS ROOTS, the job will not report an error. When I work with multiple cores in parallel (CIS N ROOTS=80) and use “Heavypoints 110” and “Hpoints 50” I also run out of memory and report an error at SCF stage. But when I run my job with a single core, this problem and the aforementioned situation don’t occur. When I use Gaussian to calculate the same job, it will not report an error (28 cores).
Is there a problem with the parallelism of QChem?

I am confused about what problem you are reporting. Is it the the calculation fails with a memory problem, or that it runs okay but fails to converge? I am also slightly skeptical of the claim that you ran an 1880 basis function TDDFT calculation requesting 80 excited states on a single processor.

There are a couple of things worth noting about this calculation:
(1) This is a good candidate for SCF_GUESS=FRAGMO, since your system consists of a supramolecular complex of four 40-atom chromophores.
(2) The number of PCM discretization points (called “tesserae”, for historical reasons) is printed out early on: 8894 for the first job that you posted. With Solver = Inversion that means you need to diagonalize a 8894 x 8894 matrix in the first SCF cycle. Conjugate gradient solvers are available as well and it looks like they activate by default for a system this large, except that you turned that off.

Thank you for your advice.I am confused too. When I parallelize the job, I do have CIS convergence failure, and I get an out of memory error in ".err ". But when I do it with a single processor, it converges. So I would like to ask what causes it. And it takes days to compute on a single processor, so is there any way to parallel my work?

This is the result of my single processor running::

This is a parallel run on 1 processors
/public/software/qchem-5.0/bin/parallel.csh, , pbi4.in, 1, 0, /tmp/76122.console/qchem6226/
MPIRUN in parallel.csh is mpirun
P4_RSHCOMMAND in parallel.csh is ssh
QCOUTFILE is stdout
Q-Chem machineFile is /opt/gridview//pbs/dispatcher/aux//76122.console
Process 0 of 1 is on node7 - thread support 0
initial socket setup …start
initial socket setup …done
now start server 0 …
node7 11.11.11

Here is the CIS part:

CIS energy converged when residual is below 10e- 6

Iter Rts Conv Rts Left Ttl Dev Max Dev

1 0 85 0.003973 0.000093
2 0 85 0.002073 0.000045
3 0 85 0.002309 0.000097
4 1 84 0.001112 0.000042
5 12 73 0.000893 0.000039
6 22 63 0.000760 0.000042

The job is still running and tends to converge.

Hi Walker,
I think there are two separate issues here.
(1) First, it appears that you are using MPI-parallelized version of Q-Chem, which is sub-optimal for most types of calculation and the OpenMP-parallelized version is greatly preferred. In particular, the MPI version is a memory hog. Please contact Q-Chem support if you are unfamiliar with the distinction. This also perhaps explains why I was unable to replicate precisely the behavior that you describe, because I use the OpenMP version.
(2) That said, I still have memory issues with your calculation that are related to the fact that Q-Chem’s memory usage for TDDFT could be improved. I am successfully able to run your calculation by requesting 40 roots instead of 80. I ran it first without PCM to test whether that was the issue, and that calculation ran overnight (~7 hours on 28 cores). With PCM, the 40-root calculation runs in ~15 hours on 28 cores. Most of that added cost is probably not the PCM cost per se but the fact that the PCM code turns off symmetry and your molecule has Ci symmetry, so the integrals cost goes up.

Input file is pasted below. I haven’t tried anything in between 40 and 80 roots.

> $rem
> method wb97x-d
> BASIS def2-svp
> CIS_N_ROOTS 40
> CIS_SINGLETS true
> CIS_TRIPLETS false
> SOLVENT_METHOD pcm
> cis_dynamic_mem 1
> Mem_total 100000
> mem_static 4000
> scf_guess fragmo
> !pcm_print 3
> !scf_print_frgm true
> xc_grid 1
> sym_ignore true
> $end
> 
> $rem_frgm
> solvent_method 0
> $end
> 
> $pcm
> Theory CPCM
> Method SWIG
> Radii Bondi
> hpoints 50
> heavypoints 110
> $end
> 
> $solvent
> Dielectric 14
> OpticalDielectric 4
> $end
> 
> 
> $molecule
> 0 1
> --
> 0 1
> O 10.74713407 13.570009 11.37115993
> C 13.95845326 13.7261046 15.28628522
> C 13.71270617 12.8013118 14.27568466
> C 12.77082528 13.0163114 13.2998949
> C 12.04270483 14.1575764 13.27161113
> C 11.00588546 14.387302 12.2555714
> H 14.21884136 11.9854914 14.25936711
> H 12.6249547 12.3462784 12.62869948
> C 14.97313318 13.5155228 16.32952088
> C 15.14279276 14.5139456 17.34338494
> C 15.75350585 12.3639496 16.37194652
> C 16.15685426 14.2827474 18.3539855
> C 16.69747607 12.1680938 17.34229711
> H 15.62660796 11.6880262 15.70183894
> C 16.90380678 13.0943592 18.33222876
> H 17.22588193 11.3669994 17.33141874
> C 17.96565138 12.841072 19.3602347
> N 18.13904915 13.8851454 20.28598397
> O 18.67672922 11.8721012 19.40701169
> H 18.79679821 13.6716184 20.90713888
> O 9.71208987 17.5798988 13.22809765
> N 10.24976995 15.5668546 12.34912538
> C 13.41568591 15.9364772 16.30558847
> C 13.24602633 14.9380544 15.2917244
> C 12.23196484 15.1692526 14.28112385
> C 11.48501232 16.3576408 14.30288059
> C 11.69134303 17.2839062 15.29281224
> C 12.63531325 17.0880504 16.26316282
> C 10.42316771 16.610928 13.27487464
> H 9.59202089 15.7803816 11.72797046
> H 11.16293717 18.0850006 15.30369061
> H 12.76221114 17.7639738 16.9332704
> C 14.43036583 15.7258954 17.34882413
> C 14.67611293 16.6506882 18.35942468
> C 15.61799382 16.4356886 19.33521445
> H 14.16997774 17.4665086 18.37574224
> C 16.34611426 15.2944236 19.36349821
> H 15.7638644 17.1057216 20.00640987
> C 17.38293363 15.064698 20.37953795
> O 17.64168503 15.881991 21.26394941
> --
> 0 1
> O 15.61913407 13.570009 11.37115993
> C 18.83045326 13.7261046 15.28628522
> C 18.58470617 12.8013118 14.27568466
> C 17.64282528 13.0163114 13.2998949
> C 16.91470483 14.1575764 13.27161113
> C 15.87788546 14.387302 12.2555714
> H 19.09084136 11.9854914 14.25936711
> H 17.4969547 12.3462784 12.62869948
> C 19.84513318 13.5155228 16.32952088
> C 20.01479276 14.5139456 17.34338494
> C 20.62550585 12.3639496 16.37194652
> C 21.02885426 14.2827474 18.3539855
> C 21.56947607 12.1680938 17.34229711
> H 20.49860796 11.6880262 15.70183894
> C 21.77580678 13.0943592 18.33222876
> H 22.09788193 11.3669994 17.33141874
> C 22.83765138 12.841072 19.3602347
> N 23.01104915 13.8851454 20.28598397
> O 23.54872922 11.8721012 19.40701169
> H 23.66879821 13.6716184 20.90713888
> O 14.58408987 17.5798988 13.22809765
> N 15.12176995 15.5668546 12.34912538
> C 18.28768591 15.9364772 16.30558847
> C 18.11802633 14.9380544 15.2917244
> C 17.10396484 15.1692526 14.28112385
> C 16.35701232 16.3576408 14.30288059
> C 16.56334303 17.2839062 15.29281224
> C 17.50731325 17.0880504 16.26316282
> C 15.29516771 16.610928 13.27487464
> H 14.46402089 15.7803816 11.72797046
> H 16.03493717 18.0850006 15.30369061
> H 17.63421114 17.7639738 16.9332704
> C 19.30236583 15.7258954 17.34882413
> C 19.54811293 16.6506882 18.35942468
> C 20.48999382 16.4356886 19.33521445
> H 19.04197774 17.4665086 18.37574224
> C 21.21811426 15.2944236 19.36349821
> H 20.6358644 17.1057216 20.00640987
> C 22.25493363 15.064698 20.37953795
> O 22.51368503 15.881991 21.26394941
> --
> 0 1
> O 5.87513407 13.570009 11.37115993
> C 9.08645326 13.7261046 15.28628522
> C 8.84070617 12.8013118 14.27568466
> C 7.89882528 13.0163114 13.2998949
> C 7.17070483 14.1575764 13.27161113
> C 6.13388546 14.387302 12.2555714
> H 9.34684136 11.9854914 14.25936711
> H 7.7529547 12.3462784 12.62869948
> C 10.10113318 13.5155228 16.32952088
> C 10.27079276 14.5139456 17.34338494
> C 10.88150585 12.3639496 16.37194652
> C 11.28485426 14.2827474 18.3539855
> C 11.82547607 12.1680938 17.34229711
> H 10.75460796 11.6880262 15.70183894
> C 12.03180678 13.0943592 18.33222876
> H 12.35388193 11.3669994 17.33141874
> C 13.09365138 12.841072 19.3602347
> N 13.26704915 13.8851454 20.28598397
> O 13.80472922 11.8721012 19.40701169
> H 13.92479821 13.6716184 20.90713888
> O 4.84008987 17.5798988 13.22809765
> N 5.37776995 15.5668546 12.34912538
> C 8.54368591 15.9364772 16.30558847
> C 8.37402633 14.9380544 15.2917244
> C 7.35996484 15.1692526 14.28112385
> C 6.61301232 16.3576408 14.30288059
> C 6.81934303 17.2839062 15.29281224
> C 7.76331325 17.0880504 16.26316282
> C 5.55116771 16.610928 13.27487464
> H 4.72002089 15.7803816 11.72797046
> H 6.29093717 18.0850006 15.30369061
> H 7.89021114 17.7639738 16.9332704
> C 9.55836583 15.7258954 17.34882413
> C 9.80411293 16.6506882 18.35942468
> C 10.74599382 16.4356886 19.33521445
> H 9.29797774 17.4665086 18.37574224
> C 11.47411426 15.2944236 19.36349821
> H 10.8918644 17.1057216 20.00640987
> C 12.51093363 15.064698 20.37953795
> O 12.76968503 15.881991 21.26394941
> --
> 0 1
> O 1.00313407 13.570009 11.37115993
> C 4.21445326 13.7261046 15.28628522
> C 3.96870617 12.8013118 14.27568466
> C 3.02682528 13.0163114 13.2998949
> C 2.29870483 14.1575764 13.27161113
> C 1.26188546 14.387302 12.2555714
> H 4.47484136 11.9854914 14.25936711
> H 2.8809547 12.3462784 12.62869948
> C 5.22913318 13.5155228 16.32952088
> C 5.39879276 14.5139456 17.34338494
> C 6.00950585 12.3639496 16.37194652
> C 6.41285426 14.2827474 18.3539855
> C 6.95347607 12.1680938 17.34229711
> H 5.88260796 11.6880262 15.70183894
> C 7.15980678 13.0943592 18.33222876
> H 7.48188193 11.3669994 17.33141874
> C 8.22165138 12.841072 19.3602347
> N 8.39504915 13.8851454 20.28598397
> O 8.93272922 11.8721012 19.40701169
> H 9.05279821 13.6716184 20.90713888
> O -0.03191013 17.5798988 13.22809765
> N 0.50576995 15.5668546 12.34912538
> C 3.67168591 15.9364772 16.30558847
> C 3.50202633 14.9380544 15.2917244
> C 2.48796484 15.1692526 14.28112385
> C 1.74101232 16.3576408 14.30288059
> C 1.94734303 17.2839062 15.29281224
> C 2.89131325 17.0880504 16.26316282
> C 0.67916771 16.610928 13.27487464
> H -0.15197911 15.7803816 11.72797046
> H 1.41893717 18.0850006 15.30369061
> H 3.01821114 17.7639738 16.9332704
> C 4.68636583 15.7258954 17.34882413
> C 4.93211293 16.6506882 18.35942468
> C 5.87399382 16.4356886 19.33521445
> H 4.42597774 17.4665086 18.37574224
> C 6.60211426 15.2944236 19.36349821
> H 6.0198644 17.1057216 20.00640987
> C 7.63893363 15.064698 20.37953795
> O 7.89768503 15.881991 21.26394941
> $end

The memory requirement for TDDFT calculations is 2 x NRoots x NOcc x NVirt x NIter, where
NRoots = CIS_N_ROOTS (no. of states)
NOcc, NVirt = no. of occ and virtual orbitals
NIter = no. of CIS/TDDFT iterations

That is to say, the memory requirement gets larger as it iterates, because more subspace vectors are added. This info is contained in the manual, although I have opened a ticket and will make some code changes to try to get this information into the Q-Chem output file in a future release. (It is present in a version that uses MEM_STATIC for TDDFT calculations but the warnings about memory usage did not get updated to the current version of the TDDFT code, which uses MEM_TOTAL by default.)

Update: I have checked in changes that will appear in Q-Chem 5.4.2, which will issue a warning (with some advice about what to do) in the case that the subspace vectors may not fit in memory. This warning doesn’t necessarily mean there will be a problem, because the memory estimate is based on an assumption that the CIS/TDDFT calculation requires the max number of iterations to converge (=30, by default), whereas in my experience 10-15 iterations is more typical.

Briefly, the steps to take are (in order of priority):
(1) increase MEM_TOTAL so that the subspace vectors will fit in memory, or
(2) decrease CIS_N_ROOTS (i.e., request fewer states), as this means fewer subspace vectors, or
(3) decrease MAX_CIS_SUBSPACE, which limits the max no. of vectors and can be adjusted to fit within whatever memory is available.

The subspace collapses when MAX_CIS_SUBSPACE is reached, which can hamper convergence, so the preferred option is to increase memory. The new code provides recommendations.