I'm currently trying to use Gaussian to obtain ECD spectra of a fairly large molecule in a mono- and dimeric state (roughly 150/300 atoms). I'm using TD-DFT with a DZV Dunning cc basis set, the CAM-B3LYP hybrid XC functional to better account for long-range interactions and the PCM to simulate the aqueous environment. With this setup, the calculations (20 representative structures of each state) converge quite nicely, but fail to give satisfactory results when comparing with experimental measurements.
I thought this problem might be caused by the lack of diffuse basis functions, which are crucial to describe non-covalent interactions in large systems, but with the augmented Dunning set, the calculations fail miserably in terms of convergence (which AFAIK is a common issue with diffuse sets). I've already tried several "robust" approaches to the SCF procedure, with keywords Big, YQC and CDIIS, but none of them helped -- even for the smaller systems with no PCM and very few excitations. I know this is a lot to ask for anyway, but maybe there is a way to improve convergence in such difficult cases? I'd be very grateful for any piece of advice.