Experimental and Computational Evidence for “Double Pancake Bonds”: The Role of Dispersion-Corrected DFT Methods in Strongly Dimerized 5-Aryl-1λ2,3λ2-dithia-2,4,6-triazines

Crystal structures are reported for bicyclic 3-CF3C6H4CN5S3 and monocyclic 3-CF3C6H4CN3S2, the latter of which is strongly dimerized in a cis-cofacial geometry [3-CF3C6H4CN3S2]2. The title compounds have previously been characterized in solution by NMR, displaying spectra that are consistent with the structure of [3-CF3C6H4CN3S2]2 in the crystal with anti-oriented CF3 substituents. The interannular binding was investigated using density functional theory (DFT) methods. However, the DFT-optimized geometry spreads the aryl rings too far apart (centroid–centroid distances of ≥4.353 Å versus experimental distance of 3.850 Å). Significant improvements are obtained with dispersion-corrected DFT functionals B3LYP-D3, B3LYP-D3BJ, M062X, and APFD using the 6-311+G(2d,p) basis set. However, all of these overbind the aryl rings with centroid–centroid distances of 3.612, 3.570, 3.526, and 3.511 Å, respectively. After selecting B3LYP-D3BJ/6-311+G(2d,p) as the best method, five alternative dimer geometries were tested, and all were found to be binding; however, anti cofacial-4 (matching the structure in the solid state) is the most stable. Computed energies of the remainder are as follows: +7.0 kJ mol–1 (syn-cofacial-5), +26.7 kJ mol–1 (anti-cofacial-64), +27.0 kJ mol–1 (syn-cofacial-150), +102.0 kJ mol–1 (S,S-antarafacial), and +103.7 kJ mol–1 (S,N-antarafacial), where the suffixes are torsional angles around the CN3S2 thiazyl ring centroids. The binding in the four most stable cofacial dimers may be described by “double pancake bonding”.


Figure S1
Full view of a displacement ellipsoids plot of 3 as found in the crystal lattice and showing the disorder model for the C18 CF3 group (refined to an occupancy ratio of 0.58:0.42).

Figure S2
View of the molecular structure of the dimer 3 orthogonal to the plane defined by the C12 > C17 phenyl ring (H-atoms omitted). The upper C2 > C7 ring is displaced towards optimal π-stacking with the ring centroid centred over the C13-C14 bond of the lower ring. Some inter-annular C-C distances are shown.  Fig. 3 are further connected by contacts between the apical N atoms and the ring C atoms to form walls of DTTA dimers. Cross-links between these walls have the [CN3S2] units off-register by half resulting in rather long contacts; the metrically shortest contacts are consequently between ortho H atoms of the phenyl ring and sulfur and F to S contacts.

Lattice interactions in the structure of 3
Discuss the short contacts in the crystal. Strongly-associated dimers of dimers. Extended network of contacts less than the sums of the v.d.Waals radii form into parallel 'walls' of dimers that are out of register by half the vertical distance. This long-range organization of the crystal lattice found in 3 is contrasted with that in the only other structurally characterized DTTAs, 1 and 2 (refcodes: DESSID and PAFLAJ), in Fig. S3. 1 shows a layer structure of weakly-associated dimers oriented along the cell b axis, resulting in columns of CN3S2 rings isolated by phenyl rings. The extended structure of 2 shares many similarities with that of 1 and both are significantly different from the more complex pattern of interactions found in 3.

Lattice interactions in the structure of 4
The shortest contacts in the crystal lattice are those that connect S3 on one molecule with the N2 of the next molecule which are shorter than (∑rvdW-0.30 Å). This leads to 'nesting' of one cage with the next above and below it in an infinite row or 'stack'. At (∑rvdW-0.20 Å) the nesting is augmented by an N1 to C1 contact. The stacks of nested cages are aligned with the crystallographic b axis (see Fig. S5). At (∑rvdW-0.20 Å) additional short contacts link the rows of nested cages with a second row via centrosymmetric S1•••S1" and N1•••N1" contacts, so that within the C2/c unit cells, there are four such "pairs" of cages grouped around 1 ̅ locations.

DFT Computational Results
The general approach was to do a full optimization at the DFT/6-31+G(2d,p) level with frequency checks, followed by DFT/6-311+G(2d,p). The latter repeatedly displayed imaginary frequencies that correspond to deformations towards the 'correct' geometries from the high-compliance methods. Re-optimization starting from the statically deformed geometries then led to fully converged DFT/6-311+G(2d,p) geometries without imaginary frequencies. A computed structure of 3 was first conducted with the B3LYP functional; although this optimized fully, the geometry indicated excessive repulsion between aryl substituents. Next, a series of functionals with differing approaches for inclusion of dispersion effects were tested, using primarily methods already validated for DTTA dimers in the work of Mou et al.: B3LYP-D3, B3LYP-D3BJ, M062X, O3LYP, and also the new APFD method built into GW16. 62. The most tractable method (good compromise between accuracy and efficiency) was B3LYP-D3BJ which was thenceforth used for all other calculations in conjunction with the above mentioned double and triple-ζ Pople basis sets. Cartesians coordinates of all the optimized geometries reported in this work are presented in Table S14.
Although it has a similar structure to the basal DTTA ring, the DFT calculations indicate that the FMOs of the cage compound are really quite different from those of the (monomeric or dimeric) DTTA rings. Notably, the LUMO of 4 is quite low-lying at -3.80 eV, compared to the HOMO at -7.65 eV. This LUMO is a π* fragment on the bridging "-N=S=N-" moiety. The HOMO by contrast is dominated by a p orbital framework mostly on the DTDA ring (at the two N atoms).