Hello, i need to generate a few parameters of excitation states in a series of molecules. I am using g09. I have a opensuse (linux) with 24 cores, Dell setup (i believe it configures a supercomputer). In order to do this, i was drawing the molecules in gaussivew and performing a pre optimisation in PM6 and using the .log file to optimise the structure in the b3lyp level of theory at a cc-pvdz basis set with scrf=(dichloromethane) and the opt freq keywords.If the results of this second optimisation showed complete convergence, Max force, RMS force,Max displacement and RMS displacement, as well as presenting no negative frequencies and gaussian told me stationary point found then i would run a job with TD b3lyp/cc-pvdz scrf=(dichloromethane). However how can i make sure the stationary point found by this b3lyp optimisation is the real global minimum and not just a local one ? Also, in order to obtain the HOMO and LUMO orbitals of the excited states in these molecules as well as force constraints, excitation energies, transitions orbitals, absorption wavelenghts and UV-VIS spectra is it absolute necessary to start at the global minimum before performing the TDDFT?
How many Number of States should i set the calculations ? i am using the default NS=3 is it enough? how do i make sure it is enough? and finally, is this basis set with doublezeta enough to accurately obtain these parameters or do i need to run a cc-pvtz?
Obs:i have attatched one of the molecules i need to obtain the required parameters.
Hi!
Most times, if your starting geometry is reasonable, you should get a good guess about the global minimum (here the chemical intuition is rather useful). Every aspect you mentioned are clues that you found a minimum, but you can't know if it's the global one for sure, because scanning the whole potential energy surface is impossible for most compounds. I'd look at the output and try to rationalize it, just to check if it looks like a stable and reasonable geometry.
About the number of states: it depends on what kind of information you need, for some applications only a small number of states are needed, but sometimes it is not that simple, if you are trying to get information about the whole absorption spectrum of some complex structures, for example, you might need even 50 states, it's quite elusive. (here I recommend some reading about the way TD-DFT is calculated, to clarify how you should approach your problems).
Most times it's useful to read some papers on benchmarks of basis sets, so you might get an overview of which ones are suitable for your application, TD-DFT is highly dependent on the level of theory, so, be careful about what you choose.
Hope it helps!
Everything Lucas said and: You should not use B3LYP (for geometries and even more importantly) for excited states. At least use CAM-B3LYP or some other kind of long-range corrected functional, because otherwise, you will get a plethora of artificial low-lying charge-transfer states instead of the "real" physical pipi* excitations. For geometries, I'd suggest to include some kind of dispersion correction (e.g. B3LYP-D3) and a triple-zeta basis to reduce BSSE.
Good luck with your calculations!
Everything Lucas Gian Fachini and Jan-Michael Mewes said is on point, but I would like to add a few additional notes as well: use proper basis sets. cc basis sets are designed for coupled cluster and extrapolation techniques, they are very large and not very efficient. Also, note that CAM-B3LYP still improves with D3 dispersion corrections, at least when it comes to non covalent interactions and geometries, quite significantly.
Thank you Lucas Gian Fachini, Jan-Michael Mewes and Rubén Laplaza for your time and valuable advices. I will run my calculations with the CAM-B3LYP level of theory and look for reasonable geometries in the output files of Gaussian.
I have only two last questions:
Which basis set is more appropriate for such calculations?
I was looking at some papers and the /6-31+G(2d,2p) seems accurate, as mentioned by Yanai et al (Chemical Physics Letters 393 (2004) 51–57). Would it be a good basis set?
Regarding the Number of States for the required properties, should I just use a higher number as 50 or a lower number can do the work? do you recommend some study explaining how to properly set a number of states for TDDFT calculations?
Once more, thank you all for the valuable help.
It seems that basis set convergence depends mostly in the nature of the transition(s) of interest. Which you should understand before you can choose the state. I mean, if you dont know what you are looking for, why bother doing TD-DFT? There might be several excitations of different kinds, and you need to understand this. TD-DFT is NOT a black box method!
You can search the literature for TD-DFT benchmark studies of different types of molecules, e.g. Article Choosing an atomic basis set for TD-DFT, SOPPA, ADC(2), CIS(...
- Badges
- Science method
- Similar topics
- Methodology
- Absorption Spectroscopy
- UV-Visible Spectroscopy