Finding the twin state for hydrogen bonding in malonaldehyde

I was quite excited when I saw the picture above when I recently visited Susanta Mahapatra.

One of the key predictions of the diabatic state picture of hydrogen bonding is that there should be an excited electronic state (a twin state) which is the "anti-bonding" combination of the two diabatic states associated with the ground state H-bond.
Recently, I posted about how this state is seen in quantum chemistry calculations for the Zundel cation.

The figure above is taken from
Optimal initiation of electronic excited state mediated intramolecular H-transfer in malonaldehyde by UV-laser pulses 
K. R. Nandipati, H. Singh, S. Nagaprasad Reddy, K. A. Kumar, S. Mahapatra

The figure hinted to me that for malonaldehyde the twin state is the S2 excited state, because of the valence bond pictures shown at the bottom of the figure and because the shape of the two potential energy curves is similar to that given by the diabatic state model.
Below I have plotted the curves for a donor-acceptor distance of R=2.5 Angstroms, comparable to that in malonaldehyde.

The vertical scale is such that D=120 kcal/mol, leading to an energy gap between the ground and excited state of about 4 eV, comparable to that in the top figure [which is in atomic units, where Energy = 1 Hartree = 27.2 eV].
Note that in the first figure, there is a gap on the vertical scale and the top and bottom part of the figures involve a different linear scale.
Hence, to make a more meaningful comparison I took the potential energy curves and replotted them on a linear scale using the polynomial fits given in the paper. The results are below.

There is reasonable agreement, but only at the semi-quantitative level.
[With regard to units the distances are in atomic units (Bohr radius = 0.5 Angstroms] and comparable].

The figure below shows how the calculated transition dipole moment between the ground state and the first excited state. I was surprised that it only varied by about five per cent with change in nuclear geometry.
However, Seth Olsen pointed out to me that this small variation reflects the Franck- Condon approximation [which is very important and robust in molecular spectroscopy].

I also calculated this with the diabatic state model, assuming that the dipole moment of diabatic states did not vary with geometry and the Mulliken-Hush diabaticity condition [that the dipole operator is diagonal in the diabatic state basis] held (see Nitzan for an extensive discussion).
[The units here are 1 atomic unit = 2.6 Debye].

The vertical scale here is the magnitude of the dipole moment in one of the diabatic states. This is also the approximate value of the dipole moment of the molecule at "high" temperatures above which there is no coherent tunnelling between the two isomers, i.e. the proton is localised on the left or right side of the value. The experimental value reported here is about 2.6 Debye.
This is consistent with the claim that the diabatic state model gets the essential physics of the electronic transition.

But, of course the real molecule is more complex. For example, between the S0 and S1 states there is a "dark" S1 state, characterised as a n to pi* transition where n is the lone pair orbital on the oxygen. The three states and their associated conical intersections are discussed in a nice paper by Joshua Coe and Todd Martinez.

I think a good way to more rigorously test/establish/disprove the diabatic state picture is to use one of the "unbiased" recipes to construct diabatic states, such as that discussed here, from high-level computational chemistry.