Crystallographic Confusion

Crystallographic Confusion

Author: David Bradley

A Controversial Crystal Structure

Can X-ray crystal structures obtained from supramolecular species be entirely relied upon when the species being studied is so unstable as to have eluded chemists for decades? A controversial X-ray crystal structure published in Science in July would suggest so, but British and US researchers have carried out calculations on the proposed structure that suggest otherwise.

Yves-Marie Legrand, Arie van der Lee and Mihail Barboiu of the University of Montpellier in France, published work on cyclobutadiene, an intriguingly strained and unstable molecule: the smallest of the conjugated cyclic hydrocarbons. The molecule’s strained geometry and electronic structure have meant that its unperturbed crystal structure has remained elusive.

The French team reasoned that trapping a precursor, 4,6-dimethyl-α-pyrone, in the vase-like cavity of a calixarene and then irradiating it with ultraviolet light would generate 4,6-dimethyl-β-lactone, a Dewar intermediate. This, they reasoned, would be stable enough at low temperature in the confined space of the calixarene, and could be pushed further with continued irradiation to 1,3-dimethylcyclobutadiene. The team carried out X-ray diffraction studies of both the intermediate and the final product.

Calculated Inconsistencies

However, according to calculations by three independent teams all is not as it seems with the trapped cyclobutadiene and its CO2 companion. Barboiu and colleagues reported four distinct species trapped in the calixarene cavity – the activated precursor, isomeric Dewar β-lactone, and square and rectangular isomers of 1,3-dimethylcyclobutadiene. However, David Scheschkewitz of Imperial College London, UK, has inspected the CIF (crystallographic information format) file, which represents a refinement of the raw data from the diffractometer. He found that the bond lengths were inconsistent with the published interpretation. Scheschkewitz suggests that a new refinement of the model is needed that uses both enantiomers of the Dewar β-lactone. He has requested the raw data file, the hkl, to allow a new refinement to be carried out independently.

Separate calculations carried out by Imperial College’s Henry Rzepa and Igor Alabugin of Florida State University (USA) and their co-workers essentially corroborate the suggestion of Scheschkewitz that the original interpretation is incorrect. Overall, their studies suggest that only the bicyclic β-lactone intermediate in which carbon dioxide remains covalently bound to the cyclobutadiene unit is present in the calixarene, rather than there being two distinct compounds within. Rzepa says that the original research and various chemists’ attempts to explain the findings have stirred up “quite a hornet’s nest”. He explains further: “At issue is whether the two bonds that link the cyclobutadiene and the CO2 moieties are covalent in the normal sense of the word, or whether they are in fact very much weaker, van der Waals contacts, or indeed whether they are essentially non-interacting atoms.” If the latter, then Barboiu and colleagues would be vindicated; the two molecules exist as separate species in the calixarene cavity. Further calculations by Rzepa on the spectra of the species thought to be involved, show that an enormous energy shift would be required to selectively stabilize the excited state, with the absorption moving from 230 to 320 nm. This makes the photochemical formation of cyclobutadiene within the calixarene even more implausible.

Unreasonable Claim?

Barboiu and colleagues have defended their interpretation in a published riposte in Science in mid-November, stating that the guanidinium-sulfonate-calixarene host matrix used in their original study plays an essential role in stabilizing the rectangular-bent cyclobutadiene derivative under confinement and that their interpretation is indeed correct. However, Scheschkewitz is on record as saying that, “It is fully justified to report unsatisfactory X-ray data as long as it is not over-interpreted and not the one and only basis for a totally unexpected claim.” Barboiu and colleagues concede that their X-ray data have large error bars: Scheschkewitz points out that this is precisely why claiming the presence of 1,3-dimethylcyclobutadiene as well as two distinct forms of this highly reactive compound in the calixarene is so unreasonable.

Limited Data

Rzepa is convinced that part of the problem that has given rise to this controversy, and allowed it to persist, lies in the way crystallographic data are currently handled. The system is not serving the chemical community well: erroneous interpretations are slipping through the peer-review filter and entering the research literature. “Normally, it is mandatory for authors to supply a CIF file,” he explains, “but a CIF file represents only one analysis using a given model, and does not allow alternative models to be evaluated.” If the raw hkl data were available then alternative refinement models could be tested. Unfortunately, journals, peer reviewers and crystallographic data centers very rarely request the hkl files and authors almost never submit them.

Rzepa adds that according to his calculations, if the 1,3-dimethyl cyclobutadiene and CO2 were to be formed, they would in fact very quickly recombine, too quickly for their X-ray structure to be determined. Rzepa emphasizes through his paper published in Chemical Communications, that “theory is nowadays sophisticated enough to provide reliable estimates of the lifetimes of two species co-constrained in a cavity, and that this general approach could be used for any systems inside a cavity.”

Regardless of the merits or otherwise of the Barboiu work itself, the tale reveals that more data needs to be routinely provided by authors who are proposing structural refinements of molecules. Two decades ago when the first crystal data on cyclobutadiene derivatives were deposited at the Cambridge Crystallographic Data Centre (CCDC), information storage was a problem. Today, computer servers provide scientists with essentially limitless space in which all the available data might be stored, allowing others to assess a given crystal structure interpretation in the raw.



  1. Mihail Barboiu

    Our opinion is that some experimental details can be certainly very useful to add to those deliberately chosen to create confusion on crystal structures described in our Science paper concerning the structure of cyclobutadiene and its intermediates: 1.These crystals can exist at 175K for at least several days without experiencing damage. An explanation for the “twin” superposed structures can arise from the lack of homogeneity in the irradiation process. Although the single crystal was continuously rotating during irradiation, two faces of the crystal were possibly less exposed to light, as well as the inside of the crystal, creating zones of intensely exposed species and zones of partially exposed species. If the “twining” of the structure was arising from evolution of the transient species over 15 hours, it would not be possible to record twice the same data with 72 hours interval. This experimental evidence in contradiction with theoretical calculations showed in a very clear manner that the photochemical rearrangements are occurring thoroughly very short irradiation times and no further evolution of the system can be observed in the absence of the irradiation procedure. More details on irradiation procedure will be described soon. 2.It is known that the Dewar-β-lactone intermediate gives rise to CO2 and CBD when irradiated with light of higher energy (λ<290) (J. Am. Chem. Soc. 95, 244-246 (1973). But also important, Chapman et al. showed that the same irradiation energy is equally unfavorable for the reaction completion, since the concentration of CBD rise to a maximum and then decrease as irradiation is continued. Most of the examples in literature concerning the CBD multistep chemistry showed at each photoreactional step a mixture of compounds, co-existing in the reactional mixture and depending on the used protocol. This is also the case of our experiments since we claimed to observe the β-Dewar-lactone, the Me2CBDS/CO2 &Me2CBDS+-CO2- and the Me2CBDR as superposed photoproduct structures in the G4C cage in the solid state. Under confined conditions the formation of the Me2CBDS+-CO2- O-zwitterion with a bent geometry for the CO2- moiety is more in line with the observed distances between Me2CBDS and CO2 while the square geometry in the C3-C6 of the ring is more in line with Me2CBDS/CO2 complex. These two structures Me2CBDS+-CO2- and Me2CBDS/CO2 are probably superposed in the crystal. (see our Response to Technical comments). 3. Further, close inspection of the IR spectra (not yet published), will show that the carbonyl vibration bands at 1710 and 1557 cm-1 strongly decrease in intensity at the same rate while a new CO2 asymmetric stretching vibration of the CO2 band at 2328 cm-1 increases. These IR data leave no doubt that the CO2 molecule can be considered as a discrete entity during the photochemical process (λ=320-500 nm at 175 K). Further irradiation with higher energy light, λ=190-500 nm, does not lead to an important increase of the intensity of the CO2 vibration band. 4.These systems operate collective information stored in sets of reactionally connected constituents and as such go beyond the properties of any of the singular (former) system-component. See for example J. Am. Chem. Soc. 95, 247-248 (1973) that shows that a more complex experimental mechanism has to be take into account before considering the CBD+CO2 reaction a singular decisive process in the system! Our recent experimental results will confirm that considering such complex systems as the sum of individual processes may not be warranted since we miss the complete information about each individual system components. 5.Finally, in the G4C host matrix, a distinct Me2CBDR structure can be unequivocally identified (22.7% and 37.4% conversion during the last two irradiation steps) and independently existing within G4C cage. Its structure is reminiscent with the previously reported the rectangular CBD geometry and deliberately not discussed by the authors of these comments. Certainly, the Me2CBDR ring together with the CO2 fragment cannot be considered as a “presumed Dewar-β-lactone enantiomer” because the overall penalty to be paid for structural distortion to form the observed confined isomer (with angles of 54.5° and 61.0° inside 90° in the Dewar isomer) is 212 kcal/mol. The O=C-O moiety is not free to spatially migrate on the cavity because it is asymmetrically sequestered via H-bonding within G4C host matrix. At this end, I completely agree with Prof. Rzepa project, that I consider very interesting for a peer review process. In this context, we have provided the primary data before the publication of our paper in Science and then to Drs Scheschkewitz and Alabugin. However we have not received any request to provide these data before the submission of two “independent contributions by two researchers working in the same Department”…. Our position is that direct exchanges are more useful for the progress of science, before submitting subjective considerations not useful in the real experimental context. 20 December 2010, Mihail Barboiu

  2. Henry Rzepa

    In his reply to the above article, Mihail Barboiu adds some interesting new experimental data. I address myself to the detection of a new IR band at 2328 cm-1 upon irradiation, quite reasonably attributed to discrete CO2. A calculation of discrete CO2 residing next to cyclobutadiene in the calixarene host (see: 10042/to-5119 ) at the reasonably high dispersion corrected wB97XD/6-311G(d,p) DFT level (and as included in the interactive table published at 10.1039/c0cc04023a) has an intense CO2 stretch predicted at 2440 cm-1, which is in good agreement with the quoted experimental value (see also here for a visual view of the spectrum). However, the geometry of the species associated with that vibration exhibits linear CO2, and a distance of 3.1A between the C of the CO2 and the closest C of the cyclobutadiene (a typical van der Waals contact). This is very different from the structure determined from the X-ray data as originally reported by Barboiu and colleagues. It is unlikely that DFT could be so erroneous in its structural prediction. It remains difficult to see how these two structures can be reconciled.

  3. Mihail Barboiu

    I found very interesting comments on Prof. Rzepa reply and I would like to assure him that I am in the favor of the “Open access data “ project. However I was not able to extract any information from the web-links proposed in his comment and his publication. This is because I am not familiar how to handle this kind of data (files). As a normal reader it will be useful to have more familiar information. Within this context, I would like to provide us the related calculated CBD-CO2 structure in an appropriate file format. In the point 3. of my previous comment and in our Response to TC published Science we have considered O-zwiterionic carboxylate form which might be considered in this discussion. Its calculated structure file will be also very useful for analysis.

  4. Henry Rzepa

    I am delighted that this important project by Prof Barboiu and his team is actively illustrating the importance of scientists being able to collaboratively both access, exchange and re-apply data. Prof Barboiu and I will also go “off-line” to discuss his request below, but because he does in fact raise important issues, I am also posting a more general reply here. For example, the syntactical form of data can often be a considerable barrier for its use in areas outside the original application. Thus crystallographers have necessarily quite specialised formats for their data, such as .CIF and e.g. the binary .dsc format. The molecular modelling equivalent is in this case the Gaussian logfile and checkpoint file (.log, .fchk). A common format which in principle applies equally to both areas is the .cml (chemical markup language) file, which is sometimes said to be a “self-describing” format. These presence of such files is noted in the “Interactive table” associated with the Web links referred to. Of course one needs to identify programs which can recognise these formats, and there a wide range of programs, some commercial, many open source, are available. It can however be a challenge to find a comprehensive list associating a file format with a program or processing environment that recognises it (ultimately, achieving this automatically is part of a project called the chemical semantic web, but I digress). One project with such objectives can be seen at Quixote. Another is Open Babel. There are various others. In parallel to the above, Prof Barboiu and I will go “off-line” to discuss what formats I can provide which his group can use in this context. I am quite confident such will be found.

  5. Henry Rzepa

    As one part of responding to Mihail Barboiu’s request for data, I have produced the following page on my own blog, which hopefully will help readers understand what might be done.

  6. Henry Rzepa

    The files which can be retrieved at this link contain a vibrational analysis of the O-zwiterionic carboxylate form, which Prof Barboiu referred to in an earlier post here. There is an interesting intense mode at 1751 cm-1 which is a combination of an asymmetric OCO stretch and H-N wags from an adjacent guanidinium cation. However, the free energy barrier for closure of this ionic form to the Dewar lactone is very small indeed (< 1 kcal/mol). If this calculation is accurate, then the lifetime of such a species would be expected to be very short.

  7. Mihail Barboiu

    Thank you Prof. Rzepa for this very useful information. We will attempt to compare our crystallographic and your calculated structures along the lines of what you suggested in your paper.

  8. Gary Battle

    Whilst not wishing to discuss the original work, it is perhaps appropriate to state the position of the Cambridge Crystallographic Data Centre with respect to experimental data. We have already been in discussion with leading publishers, explaining our view; that structure factors should be deposited alongside (or as part of) the CIF. This is already the case for some publishers. At CCDC we are implementing improved workflows to allow the more efficient processing of crystal structures. These changes include the ability for us to better handle experimental data. We will continue to work with publishers over the coming year to encourage the deposition of structures factors. It is our view that this will rapidly gain acceptance in the community and depositing experimental data will soon become the norm, before becoming mandatory. As with individual crystal structures, accompanying experimental data will, of course, be made freely available.


Leave a Reply

Kindly review our community guidelines before leaving a comment.

Your email address will not be published. Required fields are marked *