Technical Comments

Response to Comments on “Single-Crystal X-ray Structure of 1,3-Dimethylcyclobutadiene by Confinement in a Crystalline Matrix”

See allHide authors and affiliations

Science  19 Nov 2010:
Vol. 330, Issue 6007, pp. 1047
DOI: 10.1126/science.1195846


Scheschkewitz and Alabugin et al. suggest that photolysis under confinement in a crystalline matrix of 4,6-dimethyl-α-pyrone does not yield the crystal structure of 1,3-dimethylcyclobutadiene (Me2CBD) as we reported, but rather that of a 4,6-dimethyl-β-lactone intermediate. We provide arguments that the square-planar Me2CBDS/CO2 complex and the rectangular-bent Me2CBDR molecule are stabilized under confinement by the guanidinium-sulfonate-calixarene host matrix used in our study.

We recently reported the single-crystal x-ray structure of 1,3-dimethylcyclobutadiene synthesized within a guanidinium-sulfonate-calixarene (G4C) crystalline matrix (1). Scheschkewitz (2) and Alabugin et al. (3) contend that our structures instead correspond with a bicyclic β-lactone intermediate. Our view on the importance of the quality of the x-ray data on structure resolution and the view of the Technical Comment authors coincide. In their comments, Scheschkewitz and Alabugin et al. consider our work as an unconvincing attempt, mainly due to the presumed poor quality of the experimental x-ray data. They fail, however, to place the crystallographic work in the particular context of highly advanced and challenging chemistry of unstable or reactive host-guest crystalline systems. A number of very comparable studies, albeit on other systems, have preceded our work (47), and the problems encountered are virtually the same as ours: resolved and unresolved disorder leading to results far from optimal data quality. Data and model quality are not expressed in a satisfactory way by just one or two numbers or parameters. Instead, a more or less complete set of parameters needs to be given to characterize the quality of data and model. Table S1 summarizes the different quality parameters for these very comparable studies (1, 47) in the field of reactive host-guest crystalline systems. Our data are no worse than these studies, if not better, and in all cases the conclusion would be that the data warrants the structural model described in the paper. Our opinion is that all previous important contributions related to unstable molecules sequestered in porous crystals can reach their full potential.

The difficulties of experiments involving host-guest crystalline systems are related to the following factors: (i) in the case of nonreactive molecules (4, 5), the multiple orientations of the guest in a host cavity induce disorder; (ii) in the case of reactional processes, inconsistent diffusion of the reagents or unequal distribution of the irradiation energy over the whole volume of the crystal also leads to disorder (6, 7); and (iii) the preservation of the crystallinity of the host-matrix may be affected during the reactional process by a loss of water or solvent molecules. All these factors lead to what is called “poor” data quality. Data quality can indeed lead to alternative interpretations, as illustrated by the comments of Scheschkewitz (2) and Alabugin et al. (3). However, we show below that our interpretation is more likely.

Scheschkewitz and Alabugin et al. contest the atomistic interpretation we have derived from the electron density maps (1). Two general considerations are worth noting. First, an atomistic interpretation usually starts by assigning element types to the maxima observed in the phased electron density map. The electron density maps need subsequently to be interpreted in terms of a structurally logical interconnected set of bonds and angles, providing a full three-dimensional picture of unstable species as sequential snapshots of chemical reactions. It is indeed possible to get the desired structured model by using suitable restraints in the refinement, but this is not what we have done. Second, ab initio calculations and theoretical molecular models are highly valuable methods, but in order to be fairly compared with experimental data, they should be performed by considering the whole (crystalline host-matrix and guest molecule) as a unitary system, involving supplementary noncovalent interactions and supramolecular stabilization energies. This is very difficult, and we have therefore limited ourselves to the analysis of the pure experimental data.

One of the most challenging aspects of our study (1) was to fit the appropriate host matrix with the active guest molecule, which underscores the importance of subtle noncovalent interactions for this kind of study. To be more precise, we were unsuccessful with cyclodextrine host systems or nonmethylated guest molecules until we imagined the {G4C ⊂ Me21} system. The G4C host matrix offers a robust crystalline network, optimally immobilizing the 4,6-dimethyl-α-pyrone Me21 molecule during the assembly of the crystal lattice (CCDC 764865). This fit is clearly almost optimal, considering that the guest Me21 molecule is anchored via H-bonding and –CH-π interactions. Upon irradiation of G4C{Me21}, separate density maxima in the electronic density map were observed on both sides of the Me21 ring (CCDC 764867), and the quality of the x-ray diffraction data set decreased, due to the appearance of resolved and nonresolved disorder within the crystal. Further irradiation induced conversion into G4C{Me23&Me2CBDR} (CCDC 764866). We agree that the bridging bond of 1.79 Å deviates substantially from the theoretical value. The cigar-shaped thermal ellipsoids for atom C3 in the structures of Me21 and Me23 could be real or indeed an artifact, as suggested by Scheschkewitz (2). Reexamination of the Fourier maps does not show separate maxima, but this does not mean per se that there is only one equilibrium site. Similar effects have been observed in (7). We only note that, if it really concerns dynamic disorder/thermal vibrations, the real C3–C6 distance (corrected for thermal motion effects) is even longer than that based on the distance between equilibrium sites (8). Based on the evolutive coherence between the different structures, we argue that the latter situation, that is, the one presented in (7), is the most likely one. Additionally, separate density maxima were detected on both sides of Me23, corresponding to 22.7% conversion into Me2CBDR. Further irradiation led to the new geometry G4C{Me2CBDS&Me2CBDR}:Me2CBDS/CO2 complex (62.7%) and Me2CBDR (37.3%) (CCDC 764868).

When unstable intermediates or products form in highly confined conditions, a complex random distribution of electron density maps inside the cavity is observed. Our experiments (1) resulted in the formation of a limited number of peaks showing specific relative geometrical positions into the cavity aligning along the sulfonate groups, as shown in fig. S2. Regarding the question of whether Me23 transforms into the Me2CBDS/CO2 complex, we argue that the latter complex, that is, the one presented in (1), is forming in such confined conditions. First, we observe two important geometrical tendencies when Me23 transforms into Me2CBDS/CO2: (i) the C2–C3 and C6–O1 bonds expand to 1.50 Å (1.54 Å after new refinements) and 1.61 Å, respectively, and (ii) the C3C4C5C6 ring tends toward a square geometry with a mean side length of 1.46 ± 0.04 Å. Another element showing the tendency of CO2 to separate from CBD is the shape of the electronic clouds of the CO2 atoms, which, in the Me2CBDS structure, are becoming oval and slightly bigger (Fig. 1B). Meanwhile, the electronic clouds of the C3C4C5C6 ring atoms are more localized. Linked by covalent bonds, the CO2 and the C3C4C5C6 ring electronic clouds would be modified in a similar manner. Consequently, the CO2 molecule is tending to leave the system, and simultaneously the C3C4C5C6 ring adopts a square geometry, different from the calculated and experimental trapezoidal geometry of the Dewar-β-lactone corresponding ring (Fig. 1A). Second, bending of CO2 when interacting with other molecular units has previously been shown. The CO2–IV phase structure confirms that at high pressure (a scenario not very different from confined conditions), CO2 molecules are nonlinear (9). Interactions with metals (10) or metal ions (11), or ultraviolet photoreactions (1214) can also produce bent states of CO2. Third, Maier et al. (15) and Pong et al. (16) independently showed the tendency of cyclobutadiene to undergo strong association with ligands (including CO2) constrained to remain in close proximity by a solid matrix. The model proposed by Pong et al. (15), based on asymmetric vibrational bands of a strongly interacting CO2/cyclobutadiene system in which the CO2 molecule bends in either a perpendicular or parallel manner relative to the plane of the cyclobutadiene, can unambiguously correlate with our experimental structure. Finally, under photoexcitation conditions, the formation of a bent CO2 radical anion (12, 13) in interaction with the Me2CBDS radical cation via strong ionic bonds—more in line with the observed C2–C3 and C6–O1 distances—seems reasonable because the host matrix is ionic and polar. We agree with Scheschkewitz (2) and Alabugin et al. (3) that much research remains to be done to experimentally characterize the Me2CBDS/CO2 complex structure.

Fig. 1

(A) The bond lengths of Me23 (left) and Me2CBDS (right) reproduced from Alabugin et al. (3). (B) Electron density maps for the x-ray crystal structure of Me23 and Me2CBDS/CO2 complex. Separate density maxima present on both sides of Me23 and Me2CBDS correspond to conversion to Me2CBDR.

Regarding the Me2CBDR rectangular geometry, the CO2 is perpendicularly oriented with respect to the C3C4C5C6 ring of Me2CBDS, and a proposed Dewar-β-lactone enantiomer Me2CBDR (2) does not fit with the distorted structure as presented in Fig. 2. Moreover, one might argue that the CO2 molecule could not be in such vicinity with Me2CBDR and therefore is not present in the elementary cell where Me2CBDR is present, because of some fairly short distances as outlined in (2, 3). Indeed, alternative refinements show that if the occupancy of the O7–C2–O1 part is refined freely, it drops down from 1.0 to 0.83, as compared with an occupation of 0.63 for Me2CBDS. Moreover, the coordinates and thermal parameter (Ueq) values of O7 and O1 are relatively high both in the initial model (both 0.14 Å2) and the first alternative one (~0.11 Å2). However, if the occupancy of the CO2 part is constrained to be equal to the occupancy of Me2CBDS, it comes out at 0.73, with Ueq values of 0.09 Å2 for O1 and O7, nearly equal to that of C3 (0.10 Å2). Crystallographic agreement factors are very close to each other, but slightly lower for the first and second alternative models (RF = 0.0797 and 0.0799, respectively) than for the initial model (RF = 0.0809).

Fig. 2

Based on observed x-ray structures, the overall penalty for structural distortion during the enantiomerization reaction of Dewar-β-lactone as proposed by Scheschkewitz (2) is 212 kcal/mol (calculated in the gas phase).

For a more lengthy discussion about the quality of the x-ray data questioned by Alabugin et al. (3) and more comments about table S1, see the Supporting Online Material of this Response.

In conclusion, we observed two distinct geometries: the square-planar Me2CBDS/CO2 complex and the rectangular-bent Me2CBDR molecule, stabilized under confinement by the G4C host matrix. We believe that the key to making additional progress in understanding the formation of CBD structures in confined conditions is to further improve the supramolecular design of the crystalline hosts and to develop new analytical capabilities. Reducing the uncertainty in CBD formation in the solid state under confined conditions will principally require the design of new complementary host-guest systems that reduce the disorder problem that has so far been an unavoidable challenge.

Supporting Online Material

SOM Text

Figs. S1 to S3

Table S1


Stay Connected to Science

Navigate This Article