How to restart a TD ADC(2) calculation

Questions
#1

Hello, I want to compute excitation energies of an organic compound (52 atoms). Unfortunately the calculation stopped due to the wall time limit. Thus I was wondering if it is possible to restart it. Or do you have any tips to decrease the time necessary for the calculation to be complete ? ( Changing the basis set will be the last resort option…)
I am using QChem 6.0.

I displayed below the input and the end of the output file

Have a nice day ! :slight_smile:
Input :
$rem
ADC_DAVIDSON_MAXITER 200
ADC_DAVIDSON_CONV 4
METHOD adc(2)
BASIS cc-pVDZ
aux_basis rimp2-vdz
EE_SINGLETS 5
EE_TRIPLETS 5
STS_MOM true
STATE_ANALYSIS TRUE
ADC_PROP_ES2ES TRUE
ADC_PROP_ES TRUE
solvent_method PCM
MEM_TOTAL 500000
MEM_STATIC 2000
$end

$pcm
theory IEFPCM
ChargeSeparation Marcus
StateSpecific Perturb
$end

$solvent
Dielectric 2.38
OpticalDielectric 1.9599999999999997
$end

$molecule
0 1

End of the output :

Starting Davidson for singlet states of irrep A …

It NVec Conv Avg. Norm Max. Norm Conv. states Remark

     5    0  2.939e-01  6.856e-01  0.3544 n n n n n       Guess.
1   10    0  1.694e-01  5.813e-01  0.1379 n n n n n
2   15    0  4.170e-02  1.564e-01  0.1077 n n n n n
3   20    0  2.047e-02  5.602e-02  0.1042 n n n n n
#2

Hi, if you are still interested, let me comment on this …

It looks like this job is running extremely slowly. There are only 3 iterations done, and you’ll need at least 50 to converge anything. So restarting won’t really help.

The problem with ADC is that it becomes much slower once it needs to write data out to disk. You can check the size of pagefile.0 in the QCSCRATCH folder. If that is bigger than the available memory, then Q-Chem needs to read and write to this file repeatedly. You can also check whether your jobs runs with appropriate CPU usage (e.g. 2000% if you are using 20 cores).

If it does not then you need more memory. If you don’t have more memory, you can reduce the number of vectors stored in memory via ADC_DAVIDSON_MAXSUBSPACE (or simply reduce the number of EE_SINGLETS). Otherwise you have to do something about the basis (either reducing the AO basis or freezing more core orbitals).