Putting pressure on aromaticity along with in situ experimental electron density of a molecular crystal

When pressure is applied, the molecules inside a crystal undergo significant changes of their stereoelectronic properties. The most interesting are those enhancing the reactivity of systems that would be otherwise rather inert at ambient conditions. Before a reaction can occur, however, a molecule must be activated, which means destabilized. In aromatic compounds, molecular stability originates from the resonance between two electronic configurations. Here we show how the resonance energy can be decreased in molecular crystals on application of pressure. The focus is on syn-1,6:8,13-Biscarbonyl[14]annulene, an aromatic compound at ambient conditions that gradually localizes one of the resonant configurations on compression. This phenomenon is evident from the molecular geometries measured at several pressures and from the experimentally determined electron density distribution at 7.7 GPa; the observations presented in this work are validated by periodic DFT calculations.


Supplementary Tables
Supplementary Table 1 The refinement was carried out in two steps: 1) high order spherical atom refinement of the position and thermal parameters for C and O atoms (163 parameters); 2) refinement of the multipole coefficients using all reflections (302 parameters). A scale factor was refined in both cases.
Supplementary  (3) 2.525(4) 1 From the multipolar refinement on high resolution data. These data are not directly comparable with those of the other refinements (all based on lower resolution data only), because the thermal motion is here better de-convoluted, giving in general slightly longer C-C distances for all bonds. For sake of homogeneity these data are not included in the plot 7. Anyway the asymmetrical distribution of bonds follows the general trend observed at all pressures.  Table 6 Results of the X-ray constrained wave function calculations.

Low Temperature data collections
Single crystal X-ray diffraction data were collected at several temperatures, up to a resolution of 0.67Å, using an Agilent SuperNova diffractometer, equipped with Mo Kα microsource X-ray tube, with Al filter and mirror optics. An Oxford Cryosystem Cryostream 700 was used for cooling the crystals. The most relevant crystallographic parameters of these data collections are summarized in Supplementary Table 2.
Remark: X-ray diffraction at variable temperature did not reveal any anomalous thermal motion, which indicates that no disorder between two localized configurations occurs. The arguments proposed by Bürgi 51 to dismiss the potential disorder of two configurations in solid state form of benzene, would not be necessary here. In fact, a hypothetical disorder between two localized configurations of BCA would imply quite large distances between disordered atomic positions, very likely resolvable by a simple refinement against the X-ray data or otherwise very much visible in anomalously large atomic displacement parameters, if only an "average" configuration is refined. In addition, the calculated structure in the crystal (which is inherently ordered) is in almost perfect agreement with the experimental model.

Multipolar refinement of the data at 7.7 GPa
The multipole refinement was performed using the XD2006 program package 52 which adopts the Hansen-Coppens formalism 53 to model the electron density using atom-centred multipoles: The core and spherical valence density are computed from Slater-type orbitals using the zero order regular approximation level of theory. 52 Single-zeta orbitals with energy-optimized Slater exponents are used for the radial part of the deformation terms. 54 and are parameters that enable contraction or expansion of the density shells. Several refinement strategies were tested using the above described model parameters. Because of the lower data completeness, it was found more appropriate refining the atomic positions and ADPs of C and O atoms with high angle data (sinθ/λ > 0.7 Å -1 ; 163 parameters), then calculating the ADP's for Hydrogen atoms and then keeping the position and displacement parameters fixed for the refinement of all the multipole coefficients (302 variables, reflection/parameters = 11.8). H atoms correlated by pseudo-symmetry (see Figure 2 of the article) were constrained to have the same set of multipole coefficients. and could be independently refined for C and O atoms, using the same for dipoles, quadrupoles and octupoles. For H atoms, and were fixed at the standard value of 1.2. This refinement produced correlation coefficients below 0.7 for all pairs of variables but for and P v of O(1) and O(2) (0.82 in both cases).
Noteworthy, the preliminary high order refinement gave a quite satisfactory Hirshfeld rigid bond test, 56 with only four bonds exceeding the limit of 1.0 10 -3 Å 2 amplitude difference with a maximum of 1.4 10 -3 Å 2 for the two C=O bonds.
Because the atomic position and thermal parameters were fixed during the following refinement steps, the Hirshfeld test remain identical. On the contrary, if ADPs and atomic coordinates are simultaneously refined with multipole parameters (464 variables overall, reflection/parameters = 7.7), the Hirshfeld rigid bond test is much less satisfactory (6 bonds exceed the limit with differences up to 3.0 10 -3 Å 2 ), caused by a large correlation between u ij and parameters, without much improvement of the agreement indices. For these reasons, the model constructed with high order refinement of positions and thermal parameters and subsequent refinement of multipoles was judged to be the more adequate. Noteworthy, this procedure is always recommended to better de-convolute electron density from thermal motion (see also Dos Santos et al. 57 ).

Wave function refinement
X-ray constrained wave function calculations were carried out using the program TONTO, 58 following the procedure originally introduced by Jayatilaka and Grimwood. 59 The input geometry and thermal parameters were those obtained from the multipolar refinement. The molecular orbital calculations were carried out at Hartree Fock level, using a constraint to the experiment up to λ max = 1, after verifying that no significant improvement in χ 2 could be obtained. The wave function file obtained with this calculation was then used to compute the electron density features reported in the article, using the program AimAll. 60