Dear members,
Recently, I have calculated the emission energy from T1--> S0 state of molecule A. First I optimize the structure in triplet state multiplicity (0 3). Then I calculate single point calculation of ground state in T1 optimization energy multiplicity (0 1).
E (emission) = E(T1,T1)-E(S0,T1)
The output file are attached. However, the excited energy is so large... around 690 nm but the experimental result is only 457 nm. I am very confused and wonder if I did something wrong with the calculation. Please help
First do S0 optimization then use that geometry for T1 opt. Next you can calculate the vertical excitation . it corresponds to phosphorescence emission energy .
As far as I can see, your general methodology seems more or the less okay. For emission (phosphorescence) and when the lifetime of the T1-state is long enough to allow full structure relaxation (which one can often safely assume), it is in fact correct to optimize the T1-state and then calculate the vertical excitation from S0.
However, since you employed the DeltaSCF method, i.e. calculating the excitation energy based on conventional unrestricted KS-DFT (no TDDFT), your excitation is no pure triplet excitation due to spin contamination (in your case apparently only the T1-state). The spin contamination of your T1-state is, however, only very small (and the state very close to a pure triplet state), so that this is very unlikely to be the main error source. On the contrary, DeltaSCF is often the better choice (compared to TDDFT and TDA-TDDFT) when calculating T1-excitations.
A second possible pitfall might be, that you only used a quite small basis set (something like 6-31G* for the light atoms and also only a small double-zeta basis set for Pt). Calculation times seem short enough to employ at least a triple-zeta basis such as def2-TZVP. However, basis set errors should also not be in the range of 1 eV.
The largest error source is, in fact, the employed XC functional. Although simple global hybrid functionals as B3PW91 (20% exact exchange) are usually okay in many cases, there are several cases, for which they fail gloriously, e.g. Rydberg, longer-range charge-transfer (CT), core and triplet valence excitations (especially T1 due to static correlation issues). For CT and Rydberg excitations, one should usually use range-separated hybrid functionals such as wB97X-D or LC-wPBE. For Rydberg, core and triplet valence (especially T1) excitations, local hybrid functionals turned out to give satisfyingly low errors. However, local hybrids seem not to be in the official version of Gaussian, so that one would have to use another program (such as Turbomole, which also supports local hybrid TDDFT). For triplet valence excitations, also the M06-2X functional gave relatively low errors.
In your case, I would thus first check, whether the S0-T1 excitation exhibits long-range CT character, by taking a look on the difference density or even simpler just comparing the MOs which are changing upon excitation (simple visualization e.g. by molden or Molekel, etc.). In case of a CT excitation, which I honestly would not assume, you should take a range-separated hybrid such as wB97X-D or a global hybrid with a larger amount of exact-exchange like M06-2X (but only for CT with less long-range character).
If that is not the case, your error is most-likely caused by the standard static correlation problem for low-lying T1-states, e.g. well known for singlet-fission chromophores. If you need to stay with a DFT procedure, the only solution is to employ local hybrid functionals (e.g. with Turbomole) or highly parametrized global hybrids such as M06-2X (with Gaussian). However, S0-T1-excitation energies will be most likely still significantly underestimated, although errors should be below 0.5 eV. At least you should try some of these functionals.
If you can afford to spend a bit more computation time, excitation energies (and geometries) could be calculated with LR-CC2 (LR-CCSD with approximated singles), which usually give quite low errors. I'm not sure whether Gaussian is able to do these calculations (and if so, how long they will take...), but with Turbomole this kind of calculation should be still possible.
Last but not least, relativistic effects (which are not covered by the pseudopotential) such as spin-orbit coupling might become significant for Pt. But I would neglect these effects for the time being.
In summary, my suggestion for your problem is:
- stay with DeltaSCF within the DFT framework
- perhaps larger basis like def2-TZVP for higher accuracy
- check for long-range charge-transfer chracter of T1-excitation
- long-range CT: range-separated hybrids such as wB97X-D
- otherwise: M06-2X or local hybrid functionals
- alternatively: linear-response CC2
By the way, your calculated excitation energy is not too large but significantly too small (energy scale is anti-proportional to wavelength scale).
I hope this will help you solving your problem (and please forgive my suggestion of a certain commercial program ;) ).
Best regards
Article ChemInform Abstract: The Calculations of Excited-State Prope...
- Badges
- Science method
- Similar topics
- Methodology
- Absorption Spectroscopy