I am trying to reproduce the molecule’s IE/EA calculation result by using the same basis set and hybrid functional, as mentioned in this paper, on Gaussian. Unfortunately, the IE/EA calculation results are not consistent with Material Project’s results. I speculate about whether different computational methods or using different software may lead to a discrepancy. Could you provide the computational method since this paper does not provide a clear calculation method.
Anion (-1), cation (+1), and neutral structures should each be optimized separately in vacuum with the B3LYP DFT functional and the 6-31+G* basis set. You should then run vacuum frequency calculations on all three structures to ensure that you have reached a potential energy surface minima (no imaginary frequencies) and not a saddle point (one or more imaginary frequencies). Then you should run single point calculations on each optimized structure with the same basis and functional but also with an implicit solvent environment of dielectric = 78.3553 using the IEFPCM method. You can then calculate IE = E(+1) - E(0) and EA = E(0) - E(-1). Make sure that you convert your final values into the same units used by MP (eV) for direct comparison.
One source of potential discrepancy is if you’re optimizing different structures, which may find slightly different optimized geometries. I encourage you to use the structure available on the MP website to avoid this. A further source of potential discrepancy, as you speculated, is slight algorithmic differences between Gaussian and Q-Chem. More specifically, they use different optimization algorithms and I believe may also have slightly different implementations of IEFPCM. However, I would expect these differences to very rarely cause substantially different IE or EA values to be obtained. If you are confident that you have followed the procedure that I outlined in the first paragraph and are still seeing large differences, please send me your calculation output files and a full explanation of your analysis, and I will see what I can uncover.
Thank you for your reply. I have followed the procedure you outlined and have ensured that the molecular structure does not have imaginary frequencies. But, I still cannot reproduce the result on Material Project. For example, I try to test ethylene carbonate (EC) for IE calculation. The IE value I have calculated by Gaussian is about 8.417 eV (The IE value on Material Project is 7.044 eV). I do not think that 1.37 eV is a slight difference. Could you help me figure out the large discrepancy? I have put my input files and output files for my calculation in this link. In addition, in this discussion, you mention that molecules in Explore Molecules have already converted IE value to voltage vs. Li/Li+, so perhaps my calculation is consistent with your result, if I subtract 1.4 eV from our result.
I also found that there is still a problem some molecules’ IP / EA and redox potentials on Lithium are set to the same values, such as EC.
Thanks for providing this information and comparison. You could be right on with the 1.4 eV conversion, but I’ll dig a bit deeper tomorrow and get back to you.
Hi @hsiu ,
As you thought, the culprit is the subtraction of 1.4 eV. The results are consistent within 0.03 eV. Please note that for any molecules explorer entry where the IE/EA values match the redox potential values, both displayed values are for the redox potentials vs Li/Li+. My apologies that issue still needs to be addressed.
On the subject of calculating IE and EA, I want to caution against taking the values shown on the molecules explorer as absolute truth. Given the scope of the effort, and the computational resources available at the time, approximations had to be made. I’d be happy to share more over e-mail - if you’re interested, please feel free to contact me at [email protected].