Components of response properties (redundant and missing ones)
Dear developers,
I have noticed that Dalton seems to calculate multiple equivalent components, thus wasting computation time. Specifically, I am talking about the *CUBIC .IDRI option. For it, only 20 components are computed out of 81, and many are equivalent by frequency symmetry (e.g., xxzz = xzxz = zxxz = zxzx = zzxx = xzzx, and all of these are calculated explicitly!). This can be seen in results as well: their values are identical.
On the other hand, many components are not calculated at all. I have checked in Gaussian 16, and it seems that for the static case (same symmetry w.r.t. frequencies) some of these components are only 1 order of magnitude smaller than the dominant ones, and thus can have some influence in certain context.
This does not impact the average value, but, still, the calculation could be sped up considerably. I suspect the same can be true for .EFISHG and other predefined cases…
Maybe You could check this and if confirmed, perhaps it could be fixed by the next release?
Please find the illustrative calculations in both codes attached.