Using Molecular Modeling To Understand Some of the More Subtle

Mar 10, 2011 - Craig D. Montgomery. Journal of Chemical Education 2015 92 (10), ... Mike T. Springer. Journal of Chemical Education 2014 91 (8), 1162-...
2 downloads 0 Views 1MB Size
ARTICLE pubs.acs.org/jchemeduc

Using Molecular Modeling To Understand Some of the More Subtle Aspects of Aromaticity and Antiaromaticity Vernon G. S. Box* Department of Chemistry, City College of The City University of New York, New York, New York 10031, United States ABSTRACT: π-Εlectron delocalization exerts one of the most significant structure or energy influences in organic chemistry. Apart from determining the shapes of alkenes and alkynes, the planarity of aromatic molecules is a hallmark of π-electron delocalization. Huckel’s rules for aromaticity are easily applied in the teaching of undergraduates, but occasionally, some interesting discussions arise when questions are asked about some potentially aromatic nonbenzenoid molecules and antiaromaticity. Molecular modeling, using the X-ray crystallographic coordinate data for a representative sample of these interesting molecules, will lead to a greater understanding of just how these π-systems achieve stability. KEYWORDS: Upper-Division Undergraduate, Curriculum, Organic Chemistry, Computer-Based Learning, Molecular Modeling, X-ray Crystallography

T

he normal undergraduate organic chemistry curriculum strongly emphasizes aromaticity, antiaromaticity, and Huckel’s rules. Indeed, aromaticity and antiaromaticity are some of the most powerful determinants of molecular structure and stability in all of chemistry. The normal curriculum does not, however, provide many examples of the structural consequences of antiaromaticity in π-systems, nor does it spend time on nonbenzenoid molecules that could be, but are not obviously so, aromatic. It would be nice to be able to show advanced organic chemistry undergraduates some actual examples of these interesting πsystems, and how the stereoelectronic puzzles are resolved.

atom, except the hydrogen atom. All of the attempts to use the QM methods to arrive at solutions for wave equations of more complex molecular systems must be overlaid by numerous assumptions and approximations. Thus, none of these QM calculations can be guaranteed to produce a perfect molecular simulation. Each of the many flavors of QM calculations that have evolved (because the ab initio methods were pioneered) seeks to address some deficiency of the ab initio method, especially intramolecular electronic interactions. None of the QM methods can handle intermolecular electronic interactions. However, to some extent, all of these methods are useful in one scenario or the other. Similarly, the MM methods are all heavily parametrized to enable them to generate acceptable molecular simulations. Some of these MM methods are more advanced in that they use force fields that intimately incorporate all of the possible bond dipolar interactions, as well as lone pair, and bond pair interactions, into the fabric of their calculations. These high-level MM force fields obviously generate more realistic simulations than other MM force fields that are not designed to perform those calculations. One of the most important differences between the MM and the QM methods is that the MM methods that handle delocalization do so for explicitly defined π-systems, while the QM methods, because of the assumption that an electron’s density extends to infinity, perceive global delocalization as the norm in almost all molecules. The consequences of these two radically different philosophies can be dramatic. Thus, there are going to be controversies about which molecular simulations are better, and which can be relied on, and these controversies will dilute the real value of any discussion on the structural consequences of aromaticity, or antiaromaticity, in a suitable π-system. The way to avoid these issues (which are often more political than scientific) is to simply not use any theoretical methods to generate the structure of the molecular systems to be examined, but instead to find the actual experimentally generated X-ray

’ STRUCTURE SIMULATION BY MOLECULAR MODELING Molecular modeling is seen by many as an exercise in which a realistic, and plausible, molecular model is generated in silico. A relatively crude molecular model is generated in memory, usually by using a molecular mechanics (MM) program, without regard for the consequences of stereoelectronic effects on the structure. Then some theoretical method, usually either quantum mechanics (QM), or molecular mechanics (MM), is used to adjust the geometry of this crude structure to arrive at an “energy minimized” structure, in which all of the various structural and stereoelectronic forces are made minimal. However, it is also well-known that there is still substantial controversy about the relative merits of the various theoretical methods used in the structure energy minimization process that leads to a stable conformer, or structure, of a given organic molecule. Although all of the better known, QM and MM, theoretical methods eventually produce acceptable molecular simulations, there is often conflict as to which simulations are better. In reality, most QM studies begin with MM-generated structure models. Further, most QM studies on simple molecules eventually leave the initially MM-generated structures essentially unchanged, indicating that the MM-generated structures were very plausible and realistic. It is impossible for any QM calculation to arrive at an exact solution to the Schrodinger wave equation of any molecule, or Copyright r 2011 American Chemical Society and Division of Chemical Education, Inc.

Published: March 10, 2011 898

dx.doi.org/10.1021/ed100408s | J. Chem. Educ. 2011, 88, 898–906

Journal of Chemical Education

ARTICLE

coordinate data for the molecule, or close analogues, and then to examine that experimental data carefully. Hopefully, all theoreticians will defer to the experimental reality of a molecule’s geometry, even if their favored methods of calculations suggest other scenarios. The conclusions that are arrived at by studying the X-ray crystallographic coordinate data of molecules ought to be the best reflections of reality and should override any purely theoretical molecular simulation study. This work will not depend on, or use, any theoretical calculations that involve structure energy minimizations. The analysis of the X-ray coordinate data will be based solely on the interatomic distances, bond lengths, that are implicit in the data. The data will be presented as a graphical display of the molecules along with the measured distances, and any dihedral angles, that are interesting.

’ BOND TYPES Because we are about to engage in a discussion on bonds commonly found in organic molecules, and especially about bonds found in the solid phase of organic molecules, we will need to specify some criteria about bond types that have been cherished during several hundred years of experimental organic chemistry. In a previously reported study done on the CN bond, we convincingly demonstrated that the CN bond length was tightly linked to its bond order.1 A series of molecules were selected that had, in each molecule, both single CN bonds and delocalized CN π-bonds. This was done so that the experimental accuracy for measuring any bond length in any molecule, in this study, would be the same for all of the bonds in that molecule. This study clearly demonstrated that there were no overlaps between the lengths of the CN single bonds and the CN π-bonds and that simply determining the length of a CN bond almost always points to the order of that bond. Indeed, similar studies on other bond lengths, in a wide variety of molecules, have suggested that we ought to consider bond types, and their associated lengths, as quantized. The range of bond lengths associated with any bond type would then be analogous to the range of energy values associated with any simple electronic energy level transition, as seen for example in ultraviolet spectroscopy.2 The stability of a molecule in solution is a critically important criterion in this discussion because the crystals used in X-ray crystallography are normally made by the deposition of material from solution onto a solid phase. The physical and chemical properties of a molecule in solution provide the baseline for the assessment of its bonds. Although we might occasionally encounter unusual molecular clusters or associations in the solid phase, which might influence the conformations of the molecules, these associations usually do not persist in solution and so do not represent the true bonding features of the molecule. Molecules that are stable in solution have traditionally, and normally, been regarded as having double bonds and single bonds based on the following criteria: • Double bonds do not undergo free rotation in solution, as can be verified by solution NMR and other experimental evidence. • Single bonds can freely rotate and do so even in the solid phase, so leading to the possibility of conformational changes. • Double bonds are highly reactive entities with respect to addition reactions. • Single bonds participate in substitution reactions.

There is no experimentally verified instance of a CdC double bond, in any organic molecule that is stable in solution, that is longer than 143 pm. Similarly, there is no experimentally verified instance of a single bond, in any organic molecule that is stable in solution, that is shorter than 143 pm. However, a survey of the X-ray crystallographic coordinate data will reveal many instances in which bonds of different types come quite close to that 143 pm length, especially in delocalized π-systems.2 The mean length of a simple, isolated carboncarbon double bond in an alkene is about 134 pm. Delocalized bonds in truly aromatic entities, such as benzene, are about 139 pm long. The mean length of a simple, isolated carboncarbon single bond in an alkane is about 154 pm. Carboncarbon single bonds, that exhibit free rotation, show significant shortening due to electronegative substituents while retaining the ability to rotate freely.3 For example, the CC bond length in 1,2-difluoroethane is about 150.3 pm, and the bond rotates freely. As more fluorine atoms are added to this molecule, the CC bond get shorter, whereas free rotation is preserved, indicating that no π-bonding is involved in the bond shortening. Carboncarbon single bonds that join two sp2 carbons, in nonconjugated π-systems, are normally about 146 pm long, as will be seen with cyclooctatetraene below. After reviewing the bond lengths of literally thousands of benzenoid compounds, it became obvious that the length of the bonds in the benzene ring is almost unaffected by the electronic effects of the substituents. Both electron-donating and electronwithdrawing substituents left the length of the benzene ring bonds almost unchanged at about 139 pm.2 The bonds in congested benzenoid ring, and especially those in the polycyclic aromatics, are, however, susceptible to steric effects, which cause predictable bond lengthening. Examinations of the bond lengths in polycyclic aromatic hydrocarbons, even the simple naphthalenes, demonstrate this fact. Whereas steric effects can cause the lengthening of bonds in aromatic entities, under no circumstances do these bonds lengthen to beyond 143 pm.2

’ REQUIRED FEATURES FOR THE MOLECULAR MODELER X-ray crystallographers usually produce acceptable data from their studies. However, occasionally we encounter X-ray crystallographic coordinate data that have errors, leading to molecular structures that are flawed (unrealistic) and which, if not identified, could lead us in the wrong direction. To a large extent these errors escape the crystallographers’ attentions because of the software that they use. Most molecular modeling software requires the user to generate an atom connectivity list that specifies how the atoms in the molecule are connected, by what types of bonds, and their coordinate data. In most cases, given that the crystallographer has some idea of the origin, and maybe the structure, of the molecule being examine, that atom connectivity list is based on what the molecular structure “should be”, rather than on what the experimental coordinate data actually indicates. If the author of the atom connectivity list then unwittingly specifies an erroneous bond type, in defiance of the actual experimental bond length, the molecular modeling program will show the molecule with the type of bond mandated by the atom connectivity list, even though this is incorrect. Ideally, the entity that generates the connectivity list should do so after carefully examining the bond 899

dx.doi.org/10.1021/ed100408s |J. Chem. Educ. 2011, 88, 898–906

Journal of Chemical Education length data that is implicit in the X-ray coordinate data. This would require many hours of time if done by a human. If the molecular modeling program can generate the atom connectivity list guided the actual bond lengths implicit in the experimentally determined X-ray coordinate data and by sophisticated molecular structural analytical algorithms, it should correctly identify all bond types in the molecule. This would provide an independent evaluation of the molecular structure, which could be compared to what the scientist “believed” the structure should be. This would also avoid other accidentally introduced connectivity list errors, such as omitted hydrogen bonds and so forth. One example of such a program is STR3DI32.4 Programs that depend solely on user-generated connectivity lists will always be useful, but should be used carefully, especially if you always believe what you see. Indeed, the point of this manuscript is to encourage you to unleash your skepticism and to examine all the information presented below, using whatever molecular modeling program you currently employ, as careful scientists should. Most molecular modelers that do not show bond order data will show almost flawless structures for the examples below, but if you do measure the actual bond lengths in the molecules, using any molecular modeler, startling errors show themselves. The graphics presented below were generated by an examination of the relevant structures using STR3DI32. It is important to be aware that STR3DI32 shows delocalized π-systems as being continuously bonded, each atom to its neighbors involved in the delocalization, as is considered correct by resonance and QM orbital theory. The display of benzene rings illustrates this. The Cambridge Structural Database (CSD) reference codes will be used for several of the molecules presented below. This will enable readers that have subscriptions to the Cambridge Crystallographic Data Center (CCDC) to get the coordinate data and to examine them. The CIF files for several of these molecules can also be requested from the CCDC, at http://www. ccdc.cam.ac.uk, or from the relevant journals’ Web sites.

ARTICLE

Figure 1. Triostin A: carbons are blue, oxygens are red, nitrogens are purple, and sulfurs are mustard.

Figure 2. FAKFIG: carbons are blue, oxygens are red, nitrogens are purple, and hydrogens are brown.

’ EXAMPLES OF CRYSTAL STRUCTURE DETERMINATIONS WITH ERRONEOUS COORDINATE DATA Triostin A

Triostin A (Figure 1) is a DNA-binding antitumor antibiotic.5 The X-ray structural determination was marred by several unusual bond lengths that, when closely examined, were obviously errors due to flawed atomic coordinate data. Although several errors exist in this structure determination, only a few will be highlighted. The N-methylvaline unit, at the bottom right “looks” like an odd allene in STR3DI32, because the two methyl groups of the isopropyl moiety have bond lengths of 139 and 131 pm, respectively, even though the central (R-) carbon is obviously pyramidal. Notice that a bond is shown as being “missing” in the alanine unit at the top because that bond length would have had to be 166.4 pm, which is much longer than is found in any alanine derivative.2 The aromatic ring to the right has a series of bonds that are longer than the maximum 143 pm found in benzenoids, and, in fact, one of the ring bonds is 149.2 pm long, almost as long as the CC single bond in 1,2difluoroethane.

’ THE MOLECULES AND METHODS Although there are literally hundreds of molecules whose structures have been determined by X-ray crystallography and whose data have atomic coordinate data errors, the discussion obviously must be limited to a few examples. The presence of erroneous X-ray crystallographic atomic coordinate data in seven interesting molecules will be presented. The discussion of the significance of bond lengths in the structures of aromatic, and especially in presumed antiaromatic, molecules is not normally found in most undergraduate textbooks, except to mention the length of the benzene bond. Although there are hundreds of these molecules, the discussion of these systems will be limited to 10 interesting, but representative, nonbenzenoid molecules that show aromaticity or should show antiaromaticity. Although too few examples could initiate a discussion as to the generality of the data to be presented, too many examples could also be criticized. It is important to remember that no structural simulations are presented. All of the data was obtained from X-ray crystallographic studies and so no molecular modeling program, or method, was used to generate any of the data below. No calculations, other than the measurement of the lengths of bonds (r2 = x2 þ y2 þ z2) from the X-ray crystallographic coordinate data, were used.

FAKFIG

FAKFIG (Figure 2) is a 1-azidogibberellic acid derivative,6 and the crystal unit cell has one molecule of diethyl ether in it, to which the crystallographer has obviously added the required 10 hydrogens. However, STR3DI32 shows that the CC bonds in the ether molecule are too short, in fact, as short as expected of CdC double bonds and on measurement proved to have lengths of 139.7 and 143.0 pm, respectively. So whereas the ether’s structure is obvious from the number of hydrogens attached to the carbons, showing that the crystallographer knows that these carbons are sp3 hybridized, the CC bond lengths are similar to those in benzene, instead of being about 152 pm. These errors are 900

dx.doi.org/10.1021/ed100408s |J. Chem. Educ. 2011, 88, 898–906

Journal of Chemical Education

ARTICLE

Figure 5. GIKJEP: carbons are blue, oxygens are red, nitrogen is purple, and hydrogens are brown.

Figure 3. BRMNAC: carbons are blue, oxygens are red, and bromine is reddish-brown.

Figure 6. OXNAPO: carbons are blue and oxygens are red.

Figure 4. FUVGEI: carbons are blue and hydrogens are brown.

length of 166.7 pm, and in the other, the CC bond is too short with a length of 138.7 pm (similar to benzene). Notice too the distorted bond lengths in the naphthalene nucleus, where three bonds indicated are easily long enough to be CC single bonds, and one, in particular, is longer than a ring bond in cyclohexane.

due to flawed atomic coordinate data. A normal OH bond should be about 96 pm long (similar to the sum of the covalent radii of oxygen and hydrogen), but the OH bond indicated, among others, is short at 87.2 pm. Unreasonably short OH, NH, and often CH bonds are commonly found features in crystallography. Notice how STR3DI32 shows the correct bond orders in the azide group and the delocalization in the carboxylic acid group. Notice also the π-bond localization in the ester group.

OXNAPO

The crystal structure determination of OXNAPO (Figure 6), racemic-trans-1a,7b-dihydro-oxireno(a)naphthalene-3-spiro-20 oxiran-2(3H)-one,10 shows obvious problems that are due to flawed atomic coordinate data. The diepoxide has bond lengths as shown, above. Notice the long benzenoid ring bond of 152.2 pm. One epoxide unit has a missing bond, because STR3DI32 flags it as being too long for an epoxide CO bond, and at 154.6 pm it is longer than even a CC bond in cyclohexane. The bond angle of 59.9 degrees confirms that we are looking at an epoxide. The flagged benzylic CC bond is erroneously short at 139.3 pm, and is short enough to be a benzene ring bond.

BRMNAC

BRMNAC (Figure 3) is (þ)-trans-9S-bromo-8S-(menthoxy acetoxy)-8,9,10,11-tetrahydrobenz(a)anthracene.7 The transdiaxially substituted cyclohexane ring shows a bond length error due to flawed atomic coordinate data leading to the flagged CC bond having a length of 141.7 pm (similar to that of a naphthalene ring bond). Notice also that three of the “phenanthrenoid” ring bonds are too long, and two are long enough to be CC single bonds, at 148 pm each.

KOVFEG

FUVGEI

The crystal structure determination of KOVFEG (Figure 7), dl-3-chloro-6-isopropyl-5-oxo-5,6,7,8-tetrahydrophenanthrene, gave the data shown in Figure 7.11 The crystallographer’s assignments of hydrogens and the bond angles in the cyclohexyl ring clearly attest to the notion that the C2C3 bond ought to be a single bond. However, the measured length of that single bond was 138.6 pm, as short as a benzene ring bond. This error is due to flawed atomic coordinate data.

FUVGEI (Figure 4) is 4-methylphenanthrene.8 There are three CdC bond length errors in this structure, due to flawed atomic coordinate data, one of which is notably large. The CdC ring bond that measures 153.2 pm is especially notable. This bond is as long as a CC bond in cyclohexane when its length ought to be about 136 pm. GIKJEP

The crystal structure of GIKJEP (Figure 5), a highly substituted 8-acetyl-3-acetylamino-1,7-di-isopropoxy-4,5-dimethoxynapthalene,9 shows numerous errors that are due to flawed atomic coordinate data. Among the most obvious are those with its two isopropyl ether groups. In one case, the CC bond that is missing would have had a

’ SUMMARY OF ERRONEOUS COORDINATE DATA Notice that in all of these cases described, STR3DI32 provides visual clues that immediately attract the user’s attention to the atomic coordinate data errors that enable the generation of 901

dx.doi.org/10.1021/ed100408s |J. Chem. Educ. 2011, 88, 898–906

Journal of Chemical Education

ARTICLE

Scheme 1. The Anthrone and 9-Hydroxyanthracene Equilibrium

Figure 7. KOVFEG: carbons are blue, oxygen is red, chlorine is green, and hydrogens are brown.

Scheme 2. Two Examples of the Pyridone Equilibrium

theses odd structures. Molecular modeling programs that do not show bond order data or do not automatically perform a bond length analysis on the entire molecule will not show these problems. These atomic coordinate data errors can result in severe problems because they generate an incorrect appreciation and understanding of “actual” molecular structure. Indeed, in a molecule like KOVFEG, if there were no hydrogen present in the structure, an unsuspecting user of that data would believe that the molecule was an unsaturated ketone. These examples provided convincing evidence that it is critically important that the detailed bond length analysis of all of the bonds in a molecular model generated from X-ray derived coordinate data be performed to spot these rare, but interesting, problems. These kinds of errors, that are due to flawed atomic coordinate data, already have and will continue to lead unsuspecting chemists to dramatic misinterpretations of reality. If bond length data will be used to arrive at definitive conclusions concerning stereoelectronic effects, then the X-ray crystallographic coordinate data must be screened for coordinate data errors. Once the data is tainted by several coordinate data errors, one cannot guarantee that any bond length derived from the data will be definitive and meaningful. The business of manually hunting through X-ray crystallographic coordinate data for these issues can be time-consuming if you do not know where to focus your attention, especially for the structural determinations of large molecules. This will not be a real problem in this work because your attention will be directed to the “interesting” bonds. Once you have obtained the relevant atomic coordinate data, any molecular modeling program that you decide to use can then be employed to measure the bond lengths to which attention will be drawn. If you are not using STR3DI32, it will be instructive to visually compare the molecular models generated by your molecular modeler with the graphics in this manuscript.

Scheme 3. Conformations of Cyclooctatetraene

equilibrium between anthrone and 9-hydroxyanthracene, the molecule assumes the structure of anthrone as the major component. Obviously, the stability of the carbonyl-form is greater than that of the phenol-form, notwithstanding its implied aromaticity. Instead of regarding this situation as anomalous, maybe we need to present Huckel’s rules with more cautious reservations because there are also many other molecules that opt for nonaromatic forms over aromatic forms, such as the pyridones13 shown in Scheme 2. Huckel’s rules always need to be used carefully and maybe should be phrased; IF a molecule is cyclic, planar, has only π-atoms, and has 4n þ 2 π-electrons, then it MIGHT be aromatic. The actual experimental evidence of the molecule’s chemistry and stability would then be the deciding factors for the existence of aromaticity. In the same way, Huckel’s rules suggest that IF a molecule is, cyclic, planar, has only π-atoms, and has 4n π-electrons, then it MIGHT be antiaromatic. The option here is that the molecule could also assume a structure that removed, or prohibited, the π-electron delocalization that would otherwise have generated the antiaromaticity, and so become nonaromatic, and much more stable.

’ HUCKEL’S RULES Aromaticity has been simplistically presented as an implacable force that drives some molecules into greater stability once the molecule possesses the attributes embraced by Huckel’s rules. Indeed, many of us believe that if a molecule has an option of being aromatic or nonaromatic, then we always must assume that it will always opt for the aromatic form. However, this is not always true.

Cyclooctatetraene

The classical example of the molecule selecting a nonaromatic structural option, over the less stable antiaromatic one, is cyclooctatetraene, shown in Scheme 3. The molecule puckers, instead of remaining planar, and so becomes nonaromatic and more stable. Interestingly, there is considerable evidence that the molecule can pass through the planar conformation in its flipping from one tub conformer to the other. Cyclooctatetraene also provides us with two other important yardsticks. The bond lengths of the single bonds, which obviously have no π-character because if its molecular geometry (nonplanar, dihedral angle of about 71° between any two adjacent double

Anthrone and the Pyridones

The simple case of the anthrone and 9-hydroxyanthracene equilibrium12 shown in Scheme 1 serves as a good example of the aromaticity misconception. 9-Hydroxyanthracene does not exist in the solid phase, nor as the major component of this equilibrium in solution in a nonpolar medium. Indeed, in almost any 902

dx.doi.org/10.1021/ed100408s |J. Chem. Educ. 2011, 88, 898–906

Journal of Chemical Education

ARTICLE

Scheme 4. Structure of Biphenylene

Figure 9. 2,5-Di(trimethylsilyl)-,3,4-diphenyl-cyclopentadienone: carbons are blue, oxygen is red, silicons are peach, and hydrogens are brown.

The Cyclopentadienones

Cyclopentadienone could be antiaromatic, by Huckel’s rules, if the carbonyl group polarized to generate the cyclopentadienyl cation. The X-ray crystal structure (Figure 9) of 2,5-di(trimethylsilyl)-,3,4-diphenylcyclopentadienone18 shows that the C1C2, C1C5, and the C3C4 bonds in the cyclopentadienone ring have elongated to normal CC single bond lengths, much longer than the single bond length found in cyclooctatetraene, so avoiding the instability of antiaromaticity. Notice that the lengths of the ring CdC bonds of the cyclopentadienone ring are almost as short as those of isolated, nonconjugated alkenes, as opposed to the 139 pm length expected for a conjugated, delocalized π-system. These lengths are indications of π-electron localization, as opposed to π-electron delocalization, and we will see more examples below. Notice further that the phenyl substituents have large dihedral angles (about 56°) with the cyclopentadienone ring and so do not affect the π-system in that ring. Thus, for carbocycles, the obvious molecular strategies for dealing with antiaromaticity are bond elongation and puckering because both lead to the disruption of π-electron delocalization. An important feature of these molecules, that will be readily apparent during molecular modeling, will be bond lengths that indicate π-electron localization. Heterocycles also have the option of rehybridization at the heteroatom. Indeed, for most cyclic conjugated π-molecules, the options are being either aromatic or nonaromatic, and we should teach Huckel’s rules in that context.

Figure 8. BIPHEC: carbons are blue.

bonds π-system), are 146 pm. The lengths of the CdC double bonds, which are conjugated but are not delocalized, show that these bonds have the same lengths as localized, isolated CdC double bonds. We shall see that π-bond localization, as reflected by the length of the bond, is a powerful tool to help us to assess the presence, or absence, of aromaticity or antiaromaticity. It is especially important that we can comfortably acknowledge that CC single bonds as short as 146 pm need not be that length because of any π-character in that bond. It could simply be the result of the smaller covalent radii of the atoms participating in the single bond. The Cyclobutadienes

Cyclooctatetraene has a ring that is large enough to pucker and become nonplanar, so what do the cyclobutadienes do, because they cannot pucker or become nonplanar? If a cyclobutadiene molecules was truly delocalized, one could assume that the molecule would be square, and the bond lengths would all be about 139 pm, similar to those of benzene. However, theoretical calculations suggest that cyclobutadiene avoids the possibility of unstable antiaromatic π-systems by stretching two opposing ring bonds to about 156 pm, longer than unstrained CC single bonds and much longer than the single bond length found in cyclooctatetraene.14 This geometry would result in the uncoupling any π-electron overlap and delocalization that could lead to instability (antiaromaticity), and the molecule would assume the rectangular geometry shown. The energy of this highly strained system (bond, torsional, and angle strain) is obviously thought to be lower than that of the antiaromatic alternative structure. The enormous strain factors in this molecule lead to its high reactivity, especially in any reaction that results in the conversion of its ring carbons to sp3 hybridization (addition reactions). This structural feature is also shown by biphenylene15 (Scheme 4), benzo[b]biphenylene,16 and the helicene, BIPHEC17 (Figure 8). The consistency with which the dibenzocyclobutadiene bonds in these example tend to 151 pm also suggests that the bond lengths in rectangular, unsubstituted, cyclobutadiene might also be 151 and 134 pm, respectively, instead of 156 and 134 pm.

’ SOME INTRIGUING POTENTIALLY-AROMATIC Π-SYSTEMS The Pyridones

Much attention has been drawn to the remarkable chemistry of 2-pyridone, which is nonaromatic in its monomeric form but aromatic when it in dimeric, as shown in Scheme 5. The fact that this strong intermolecular hydrogen bonding pattern facilitates the delocalization in the 2-pyridone ring was seen as an instance in which aromaticity could be switched on or off simply by modifying the environment of a molecule.13 This kind of mutually synergistic paired hydrogen bonding, known as CrickWatson hydrogen bonding, is of the greatest importance in considerations of the stabilities of DNA double helices.19 903

dx.doi.org/10.1021/ed100408s |J. Chem. Educ. 2011, 88, 898–906

Journal of Chemical Education

ARTICLE

Scheme 5. Monomeric and Dimeric forms of 2-Pyridone

Figure 11. Bond length alternation in tropone.

Scheme 7. Bond Lengths in Tropolone Hydrochloride

Scheme 6. Structure of 2,3-Diphenylcyclopropenone

lowered the energy required for the polarization of the carbonyl group, greatly facilitated the transition of the cyclopropenone into its aromatic state. Another example of a Lewis acid promoted aromatization of a diphenylcyclopropenone, by the polarization of the carbonyl group, is encountered in the X-ray coordinate data of bromo(diphenylcyclopropenone)triphenyllead,21 shown in Figure 10. The cyclopropenone ring shows full π-electron delocalization and that the carbonyl group is significantly elongated. The dihedral angles between the phenyl groups and the cyclopropenone ring are about 9.3°, but there is no π-electron delocalization between these rings; a nice example of a trigonal bipyramidal complex. The indisputable conclusions from these experimentally determined crystal structures have shed much more light on the chemistry of this interesting molecule. Thus, the cyclopropenone ring of diphenylcyclopropenone is not normally aromatic, and this is another example of a molecule whose aromaticity can be enhanced by coordination to a Lewis acid.

Figure 10. Bromo(diphenylcyclopropenone)triphenyllead: carbons are blue, oxygen is red, lead is gray, bromine is reddish-brown, and hydrogens are brown.

The Cyclopropenones

In an interesting study, Tsukada et al.20 determined the crystal structure of anhydrous diphenylcyclopropenone and compared this data with the structural features of the hydrated diphenylcyclopropenone (Scheme 6). They discovered that as the number of water molecules coordinated (hydrogen bonded) to the carbonyl oxygen went from zero (anhydrous) to two (the dehydrate), the bond length features of the diphenylcyclopropenone progressively became “more aromatic” in character. In the dihydrate, the C1C2 bond and the C2C3 bond were about 139.2 and 136.8 pm, respectively, similar to the lengths of benzene’s ring bonds, whereas in the anhydrous molecule, the lengths were 141.7 and 134.9 pm, respectively. The pronounced π-electron localization of the C2C3 double bond in the anhydrous form is obvious. The carbonyl bond length went from 122.5 pm in the anhydrous form to 124.4 pm in the dihydrate. Diphenylcyclopropenone should be an example of a molecule that obeys Huckel’s rules, IF the carbonyl group had spontaneously polarized. However, Tsukada’s study showed that the energy required to spontaneously polarize the carbonyl group was not offset by the stability gained after the induction of aromaticity. The polarization of the carbonyl group was therefore too small to enable full aromaticity to be generated in this system. The experimental reality, described above, shows that the coordination of a Lewis acid to the carbonyl group, which

The Tropones

Another intriguing molecule, in this context, is cycloheptatrienone (tropone). Similar to cyclopropenone, tropone (Figure 11) could become aromatic simply by polarizing its carbonyl (CdO) group. The tropylium ion is thought to be delocalized and aromatic, notwithstanding its average CC bond length of 135 pm, which might be due to its electron deficiency.22 If the carbonyl group of tropone was polarized, then tropone should also obey the Huckel’s rules and be aromatic. However, tropone shows pronounced bond length alternation,23 and significant π-electron localization, thereby suggesting that it has limited stabilization from resonance. Notice the C1C2 single bond that is longer than found in cyclooctatetraene (146 pm). On the other hand, tropolone hydrochloride (protonated tropolone, Scheme 7) has all of the classical features of an aromatic molecule, with bond lengths as shown.24 These bond lengths are very similar to those found in benzene. Here again, a Lewis acid protonation has dramatically enhanced the observed aromaticity that was only, at best, weakly aromatic. The fact that tropolone is more acidic that phenol is also very interesting and significant. If we examine the resonance forms of tropolone in the context of its acidity (Scheme 8), the proton movement in the equilibrium process could be either an intermolecular or an intramolecular event. However, neither the 904

dx.doi.org/10.1021/ed100408s |J. Chem. Educ. 2011, 88, 898–906

Journal of Chemical Education

ARTICLE

Scheme 8. Ionization of Tropolone

Scheme 9. Cyclobutenediones and Transformations into 1,2-Bisketene

Figure 12. 5-Acetoxytropolone: carbons are blue, oxygen is red, and hydrogens are brown.

Figure 14. BELLUZ: carbons are blue, oxygen is red, and hydrogens are brown.

The Cyclobutendiones

A much less contentious, but still intriguing, molecule is cyclobutenedione, shown in Scheme 9. Cyclobutenediones are well-known and are extensively used in organic syntheses.27 The ease of their transformations into 1,2-bisketenes is well-known. This high reactivity of the monocyclic system suggests a lack of aromatic, resonance stability. If the unlikely event of the spontaneous polarization of both carbonyl groups were to occur, then cyclobutenedione would be a candidate for aromaticity, according to Huckel’s rules. There are many obvious reasons that this will not occur, especially the problem of having two positive charges in the same small ring, and on adjacent carbons, but what evidence do we have for either position? The X-ray crystal structure of BELLUZ28 (Figure 14) helps us to examine several aspects of the bond lengths in cyclobutenediones, and their involvement in π-electron delocalization. The bond between the two carbonyl groups is obviously single and elongated, most probably in reaction to the repulsion between the two carbonyl dipoles. The cyclobutenedione CdC bond is short, reminiscent of an isolated alkene, even though it is clearly involved in a delocalized π-system with the phenanthrene entity. The 134.8 pm length probably reflects more the presence of the two electron withdrawing groups attached to the CdC, rather than any π -electron localization. So, we should conclude that cyclobutenediones are not aromatic, nor could they become aromatic by simple Lewis acid complexation.

Figure 13. 2-(4-Fluoroanilino)tropone: carbons are blue, oxygen is red, fluorine is light green, nitrogens are purple, and hydrogens are brown.

resonance forms depicted for the neural molecule nor for the anion need invoke any increase in the bond order of the C1C2 bond. Why is the tropolone anion quite stable? The X-ray structure of 5-acetoxytropolone25 (Figure 12), in which there is an intramolecular hydrogen bond, shows a C1C2 bond length that is longer than expected in an aromatic molecule. We saw that in tropolone hydrochloride the length is 138 pm. Notice that the way STR3DI32 shows the π-electron delocalization here reflects faithfully the situation shown in the resonance diagram above. As in tropolone, the single intramolecular hydrogen bond is not sufficient to polarize the carbonyl group enough to convert the molecule into its totally delocalized, aromatic, form. Another example of this kind of intramolecular hydrogenbonded tropolonoid, and the partial delocalization puzzle, is shown in the X-ray structure of 2-(4-fluoroanilino)tropone26 (Figure 13). Here again the C1C2 bond length is longer than expected in an aromatic molecule and is certainly longer than the CC single bond found in cyclooctatetraene (146 pm). These structures reinforce the conclusion that the protonation of tropolone is indeed responsible for its aromatization. The question of the greater acidity of tropolone over phenol might well have nothing to do with aromaticity in the tropolone anion. We must therefore conclude that tropolone is truly a molecule that hovers on the edge of being aromatic, but needs some additional Lewis acid complexation to its carbonyl group to induce the required degree of polarization of its carbonyl group that would lead to its conversion into its aromatic form.

’ CONCLUSION Molecular modeling can help us to understand the π-electron delocalization features of molecules that have simple, or controversial, π-electron systems. The geometrical features of these molecules indicate whether they are aromatic or nonaromatic. The degree of confidence offered by these reviews and analyses of experimentally generated X-ray diffraction coordinate data far 905

dx.doi.org/10.1021/ed100408s |J. Chem. Educ. 2011, 88, 898–906

Journal of Chemical Education

ARTICLE

surpasses that from any potentially controversial theoretical molecular simulation exercise. Although these exercises can be performed using any available molecular modeling program, using a molecular modeling program that automates the generation of the atom connectivity list, and the classification of the bond types, from X-ray diffraction generated coordinate data, greatly facilitates these exercises. The ability of the molecular modeler to display the atomic connectivity and bond order data is very valuable. Some other interesting structures whose X-ray crystallographic coordinate data have errors and whose data would form the bases of interesting classroom discussions are suggested29 using their Cambridge Crystallographic Data Center codes.

(23) Ogasawara, M.; Iijima, T.; Kimura, M. Bull. Chem. Soc. Jpn. 1972, 45, 3277–3282. (24) Sasada, Y.; Shimanouchi, H. Acta Crystallogr. 1973, B29, 81–90. Sasada, Y.; Nitta, I. Bull. Chem. Soc. Jpn. 1957, 30, 62–68. (25) Kubo, K.; Yamamoto, E.; Mori, A. Acta Crystallogr. 2006, 62, 2988–2990. (26) Steyl, G. Acta Crystallogr. 2007, 63, 4353–4353. (27) Tidwell, T. T.; Allen, A. D.; Colomvakos, J. D.; Egle, I.; Ma, J.; Marra, R. M.; McAllister, M. A. Can. J. Chem. 1996, 74, 457–464. (28) Hacker, N. P.; McOmie, J. F. W.; Meunier-Piret, J.; Van Meerssche, M. J. Chem. Soc., Perkin Trans. 1 1982, 19–23. (29) SUMZUV, DUVGOQ, CAHKUR, FIXWUE, FIXXAL, GAHGEB, JURKEM, ZEHDEV, JOHGOC, GEBLII, PBXPCH, FIXVUD, ABIMUT, AFADUH, BARVOF, BPENCE10, CIZPUX, DUMXIS, GERPUO, GIVCET

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ REFERENCES (1) Box, V. G. S.; Yu, H. W. J. Chem. Educ. 1997, 74, 1293–1296. (2) Box, V. G. S. J. Mol. Model. 1997, 3, 124–141. (3) Fodi, B.; McKean, D. C.; Palmer, M. H. J. Mol. Struct.: THEOCHEM 2000, 500, 195–223. (4) The 200 atom version of STR3DI32 is free, is suitable for most undergraduate labs and general usage, and can be downloaded from http://www.Exorga.com (accessed Mar 2011). (5) Shelldrick, G. M.; Guy, J. J.; Kennard, O.; Rivera, V.; Waring, M. J. J. Chem. Soc., Perkin Trans. 2 1984, 1601–1605. (6) Kutschabsky, L.; Voigt, B.; Adam, G. Cryst. Res. Technol. 1985, 20, 1157–1162. (7) Boyd, D. R.; Dawson, K. A.; Gadaginamath, G. S.; Hamiliton, J. G.; Malone, J. F.; Sharma, N. D. J. Chem. Soc., Perkin Trans. 1 1981, 94–97. (8) Imashiro, F.; Saika, A.; Taira, Z. J. Org. Chem. 1987, 52, 5727–5729. (9) De Koning, C. B.; Giles, R. G. F.; Knight, L. S.; Niven, M. L.; Yorke, S. C. J. Chem. Soc., Perkin Trans. 1 1988, 2477–2483. (10) Gatehouse, B. M.; Lloyd, D. J. J. Chem. Soc., Perkin Trans. 2 1972, 932–934. (11) Venugopalan, P.; Weiss, R. G.; Venkatesan, K. J. Org. Chem. 1992, 57, 276–279. (12) Almdal, K.; Eggert, H.; Hammerich, O. Acta Chem. Scand. 1986, B40, 230232; also http://actachemscand.dk/pdf/acta_vol_40b_ p02300232.pdf (accessed Mar 2011). (13) Box, V. G. S. Heterocycles 1992, 34, 1631–1659. (14) Kollmar, H.; Staemmier, V. J. Am. Chem. Soc. 1977, 99, 3583–3587. Kollmar, H.; Staemmier, V J. Am. Chem. Soc. 1978, 100, 4304–4305. Balci, M.; McKee, M. L.; Schleyer, P. J. Phys. Chem. 2000, 104, 1246–1255. (15) Fawcett, J. K.; Trotter, J. Acta Crystallogr. 1966, 20, 87–93. (16) Ferrara, J. D.; Solooki, D.; Tessier-Youngs, C.; Youngs, W. J. Acta Crystallogr. 1989, C45, 57–60. (17) Van Meerssche, M.; Germain, G.; Declerq, J. P.; SoubrierPayen, B.; Figeys, H. P.; Vanommeslaeghe, P. Acta Crystallogr. 1981, B37, 1218–1224. (18) Saito, M.; Ito, T. Acta Crystallogr. 2008, E64, 21212121; http://journals.iucr.org/e/issues/2008/11/00/issconts.html (accessed Mar 2011). (19) Box, V. G. S.; Jean-Mary, F. J. Mol. Model. 2001, 7, 334–342. (20) Tsukada, H.; Shimanouchi, H.; Sasada, Y. Chem. Lett. 1974, 639–642. (21) Ng, S. W. Acta Crystallogr. 1995, C51, 2042–2044. (22) Lamsa, M.; Suorsa, T.; Pursiainen, J.; Huuskonen, J.; Rissanen, K. Chem. Commun. 1996, 1443–1444. 906

dx.doi.org/10.1021/ed100408s |J. Chem. Educ. 2011, 88, 898–906