The structure of hexabenzotriphenylene and the problem of overcrowded "D3h" polycyclic aromatic compounds

Article abstract of DOI:10.1021/ja983471i

Hexabenzotriphenylene (1, dibenzo[f,j]phenanthro[9,10-s]picene) has been prepared in 5% yield by vacuum pyrolysis of phenanthrene-9,10-dicarboxylic anhydride, and its X-ray structure has been determined. Compound 1 is a strongly twisted, D₃-symmetric molecular propeller, in contrast to other highly substituted triphenylenes (perfluoro- and perchlorotriphenylene) which adopt C₂-symmetric conformations. Computational studies of these and other overcrowded, nominally D_3h,-symmetric, polycyclic aromatic compounds are reported, and the origins of their conformational preferences and the adequacy of various computational methods for treating these compounds are discussed.

Full text of DOI:10.1021/ja983471i

J. Am. Chem. Soc. 1999, 121, 727-733
727
The Structure of Hexabenzotriphenylene and the Problem of
Overcrowded “D3h” Polycyclic Aromatic Compounds
Lisa Barnett,† Douglas M. Ho,† Kim K. Baldridge,‡ and Robert A. Pascal Jr.*,†
Contribution from the Department of Chemistry, Princeton UniVersity, Princeton, New Jersey 08544, and
San Diego Supercomputer Center, La Jolla, California 92137
ReceiVed September 29, 1998
Abstract: Hexabenzotriphenylene (1, dibenzo[f,j]phenanthro[9,10-s]picene) has been prepared in 5% yield by
vacuum pyrolysis of phenanthrene-9,10-dicarboxylic anhydride, and its X-ray structure has been determined.
Compound 1 is a strongly twisted, D3-symmetric molecular propeller, in contrast to other highly substituted
triphenylenes (perfluoro- and perchlorotriphenylene) which adopt C2-symmetric conformations. Computational
studies of these and other overcrowded, nominally D3h-symmetric, polycyclic aromatic compounds are reported,
and the origins of their conformational preferences and the adequacy of various computational methods for
treating these compounds are discussed.
Highly symmetric polycyclic aromatic hydrocarbons (PAHs)
have commanded increasing attention since the discovery of
the fullerenes. Much effort has been spent in the synthesis of
convex, bowl-shaped PAHs akin to the fullerenes,1 but other
shapes, such as saddles and twists, have been constructed.1,2
Among members of the latter class, hexabenzotriphenylene (1,
dibenzo[f,j]phenanthro[9,10-s]picene) has a rather checkered
Scheme 1
history. At least four different syntheses of 1 have been reported
(Scheme 1),3-6 but the characterization of the products has not
always been of the highest standard, and it is clear that at least
some of the reports are incorrect. There is still no X-ray structure
of this sterically very crowded hydrocarbon, but molecular
mechanics calculations indicate that 1 should be a strongly
twisted three-bladed molecular propeller. We report herein a
new, two-step synthesis of hexabenzotriphenylene from com-
mercial starting materials and its unambiguous characterization
as a highly twisted, D3-symmetric molecular propeller by X-ray
crystallography. In addition, we report computational studies
of the conformational preferences of this and other overcrowded,
nominally D3h-symmetric polycyclic aromatics, which are
observed to display an unusual structural dichotomy.
Results and Discussion
† Princeton University.
‡ San Diego Supercomputer Center.
Synthesis and Structure of Hexabenzotriphenylene. The
first report of hexabenzotriphenylene is a patent application filed
in 1958, in which Halleux claimed to have formed 1 by
cyclodehydrogenation of hexaphenylbenzene (2) in an AlCl3-
NaCl mixture at 120-130 °C (Scheme 1), but no characteriza-
tion was offered.3 Shortly thereafter, Carey and Millar reported
the synthesis of 1 in 60% yield by treatment of 9,10-
dichlorophenanthrene (3) with Mg in boiling tetrahydrofuran.4
This material was characterized by its melting point, UV
(1) (a) Siegel, J. S.; Sieders, T. J. Chem. Br. 1995, 31, 313-316. (b)
Rabideau, P. W.; Sygula, A. Acc. Chem. Res. 1996, 29, 235-242.
(2) (a) Pascal, R. A., Jr. Pure Appl. Chem. 1993, 65, 105-110. (b) Qiao,
X.; Ho, D. M.; Pascal, R. A., Jr., Angew. Chem., Int. Ed. Engl. 1997, 36,
1531-1532.
(3) Halleux, A. L., Br. Patent 852,981 1960.
(4) Carey, J. G.; Millar, I. T. J. Chem. Soc. 1959, 3144-3146.
(5) Barton, J. W.; Grinham, A. R. J. Chem. Soc., Perkin Trans. 1 1972,
634-637.
(6) Hacker, N. P.; McOmie, J. F. W.; Meunier-Piret, J.; Van Meerssche,
M. J. Chem. Soc., Perkin Trans. 1 1982, 19-23.
10.1021/ja983471i CCC: $18.00 © 1999 American Chemical Society
Published on Web 01/14/1999

728 J. Am. Chem. Soc., Vol. 121, No. 4, 1999
Barnett et al.
spectrum, and combustion analysis,4 but in 1980 Biermann and
Schmidt demonstrated that this substance was not hexabenzo-
triphenylene (without, however, establishing its true composi-
tion).7 In 1972 Barton and Grinham found that oxidation of
1-amino-1H-phenanthro[9,10-d]triazole (4), a process known to
yield 9,10-phenanthryne, in the absence of an aryne trap gave
a small amount of a high-melting substance which exhibited a
parent ion at m/z 528 in its mass spectrum.5 More conservative
than their predecessors, they concluded that this material was
“possibly hexabenzotriphenylene” formed by trimerization of
the aryne.5 Most recently, Hacker et al. reported the formation
of 1 in 13% yield by pyrolysis of cyclobuta[l]phenanthrene-
1,2-dione (5).6 This product was characterized by melting point
1
and H NMR, IR, UV, and mass spectra,6 and these data gave
every reason to believe that the elusive 1 had been prepared.
Having recently reported the preparation of perchlorotri-
phenylene by the pyrolysis of tetrachlorophthalic anhydride,8
we conjectured that similar treatment of phenanthrene-9,10-
dicarboxylic anhydride (6) would provide a simple, if perhaps
low-yielding, synthesis of 1, since 6 may be obtained by
photolysis of commercial diphenylmaleic anhydride.9 In the
event, pyrolysis of 6 under vacuum in a quartz tube at 550-
700 °C gave compound 1 in 5% yield after purification by
1
preparative TLC. The H NMR, UV, and mass spectra of this
material were essentially the same as those of Hacker et al.,6
and, if only a small amount of 1 is required, our synthesis is
quite convenient. Indeed, the limiting factor is the preparation
of the anhydride 6, which is difficult to perform on large scale
since the photolysis is conducted on a dilute suspension in water.
Unlike many PAHs, compound 1 is quite soluble in common
organic solvents, but it proved very difficult to grow single
crystals suitable for X-ray studies. Finally, an orange plate,
obtained by the slow cooling of a very concentrated solution
of 1 in nitrobenzene, gave satisfactory diffraction. Compound
1 crystallized in the common space group P21/c, and the
structure was solved and refined without difficulty. The original
determination was carried out at 298 K, but to obtain better
atomic coordinates for a detailed comparison with computa-
tionally derived geometries, the structure was redetermined at
110 K. The molecular structures obtained from both determina-
tions are illustrated in Figure 1, and a stereoview of the molecule
appears in Figure 2. The two structures are extremely similar,
but all subsequent discussions of the experimental geometry of
1 refer to data from the 110 K determination.
Compound 1 is a steeply pitched molecular propeller with
approximate D3 symmetry. The molecule may be thought of as
three biphenyls, each of them joined at positions 2 and 2′ to a
central benzene ring. The mean planes of the three peripheral
biphenyls, C(7)-C(18), C(19)-C(30), and C(31)-C(42), make
dihedral angles of 28.5, 30.0, and 29.7°, respectively, with the
mean plane of the central ring, C(1)-C(6) (see Figure 1).
Obviously the propeller distortion of 1 results from the steric
conflict between adjacent biphenyl subunits, where the C(8)-
C(41), C(17)-C(20), and C(29)-C(32) contacts average only
3.006 Å, well within the sum of the van der Waals radii of the
carbon atoms.
Figure 1. Molecular structure of hexabenzotriphenylene (1) at 298 K
(above) and 110 K (below). Thermal ellipsoids have been drawn at
the 50% probability level.
1.434 Å. However, this is much less pronounced than the bond
alternation in the three adjoining rings (B, C, and D), which
are quite twisted. There, the six radial bonds average 1.471 Å,
and the three outer bonds 1.454 Å, but the six benzo bonds
average only 1.407 Å, and the endo bonds 1.397 Å (Table 1).
Thus, in compound 1 there are six ordinary peripheral benzene
rings linked to each other and to a somewhat distorted central
benzene ring by single bonds or bonds of only slightly higher
order.
The central ring of 1 (ring A, see Table 1) adopts a shallow
chair conformation, and it exhibits significant bond alternation,
with the three endo bonds averaging 1.397 Å and the exo bonds
Computational Studies of Hexabenzotriphenylene. How
well do modern computational methods reproduce the experi-
mental geometry of 1? Although most methods might be
expected to give structures with a “reasonable” appearance, a
truly excellent geometry is much more difficult to obtain, since
1 contains (a) strong nonbonded interactions and (b) a highly
delocalized π-system, both of which are often poorly handled
by low levels of theory. A series of molecular mechanics
(7) Biermann, D.; Schmidt, W. J. Am. Chem. Soc. 1980, 102, 3173-
3181.
(8) (a) Shibata, K.; Kulkarni, A. A.; Ho, D. M.; Pascal, R. A., Jr. J. Am.
Chem. Soc. 1994, 116, 5983-5984. (b) Shibata, K.; Kulkarni, A. A.; Ho,
D. M.; Pascal, R. A., Jr. J. Org. Chem. 1995, 60, 428-434.
(9) Sargent, M. V.; Timmons, C. J. J. Chem. Soc., Suppl. 1 1964, 5544.

Structure of Hexabenzotriphenylene
J. Am. Chem. Soc., Vol. 121, No. 4, 1999 729
Figure 2. Stereoview of the structure of hexabenzotriphenylene (298 K structure). Thermal ellipsoids have been drawn at the 50% probability
level.
Table 1. Comparison of Experimental and Calculated Geometries for Hexabenzotriphenylene
HDFT
exptl
semiempirical
ab initio
3-21G
geometry type:
metrica
mechanics
MMX
B3LYP/
cc-pVDZ
B3PW91/
cc-pVDZ
298 K
110 K
MNDO
AM1
PM3
STO-3G
6-31G*
rms deviationb
maximum dev.b
distancesc
endo
exo
radial
benzo
outer
nonbonded
0.102
0.229
0.163
0.318
0.083
0.203
0.094
0.247
0.070
0.171
0.059
0.111
0.060
0.116
0.065
0.144
0.061
0.138
1.396
1.433
1.471
1.406
1.453
3.017
1.397
1.434
1.471
1.407
1.454
3.006
1.405
1.430
1.466
1.416
1.468
3.085
1.421
1.445
1.476
1.432
1.473
3.326
1.408
1.426
1.454
1.414
1.450
2.899
1.401
1.428
1.459
1.405
1.451
2.962
1.386
1.444
1.491
1.399
1.478
2.979
1.386
1.428
1.480
1.397
1.464
3.009
1.387
1.434
1.480
1.399
1.464
3.053
1.412
1.441
1.474
1.422
1.460
3.044
1.409
1.437
1.469
1.419
1.457
3.024
a All distances are given in angstroms. b Deviations are given with respect to the 110 K structure. c For the experimental structures, the average
values for the three or six equivalent distances are given.
(MMX10), semiempirical molecular orbital (MNDO,11 AM1,12
and PM313), ab initio molecular orbital (HF/STO-3G, HF/3-
21G(*), HF/6-31G*14), and hybrid density functional (HDFT)
calculations (B3LYP/cc-pVDZ and B3PW91/cc-pVDZ15) were
performed to give fully optimized geometries for 1, and as
expected, all yielded D3-symmetric propeller conformations
generally similar to the X-ray structure. A closer examination
of the results is found in Table 1, which gives the rms and
maximum deviations of the experimental atomic positions from
those of the best fit16 of each of the computed structures, as
well as selected experimental and computational C-C bond
distances.
able to match closely the experimental bond lengths, but the
AM1 calculation yielded the best overall fit among the semi-
empirical methods. However, the nonbonded C-C contact
distances were poorly handled by these (and molecular mechan-
ics) calculations, and for this reason all of the ab initio and
HDFT methods gave significantly better geometries than the
lower levels. The HF/3-21G(*) calculation most closely agreed
with the X-ray structure, with the HF/6-31G* and HDFT
geometries only very marginally worse. Interestingly, both
HDFT methods appeared to overestimate systematically the
bond distances by a small amount. For example, all of the C-C
bond distances given by the B3LYP/cc-pVDZ calculation
(including those not listed in Table 1) were greater than the
experimental values, by 0.003 to 0.015 Å. The reason for this
overestimation is the relatively small basis set employed (cc-
pVDZ), and such a systematic error has been noted previously.18
The magnitudes of the deviations of the Hartree-Fock-
calculated bond lengths from the experimental values were no
smaller than those from the HDFT geometries, but the former
were scattered on either side of the experimental values, and
the resulting cancellation of errors led to better overall geom-
etries.
All of the tested methods indicate that the B, C, and D rings
should show strong bond alternation, with lesser but significant
alternation of the central ring. Even the PM3 calculation was
(10) The MMX force field in PCMODEL Version 5.0 was employed.
(11) Dewar, M. J. S.; Thiel, W. J. Am. Chem. Soc. 1977, 99, 4899-
4907.
(12) Dewar, M. J. S.; Zoebisch, E. G.; Healy, E. F.; Stewart, J. J. P. J.
Am. Chem. Soc. 1985, 107, 3902-3909.
(13) Stewart, J. J. P. J. Comput. Chem. 1989, 10, 209-220.
(14) Hehre, W. J.; Radom, L.; Schleyer, P. v. R.; Pople, J. A. Ab Initio
Molecular Orbital Theory; John Wiley & Sons: New York, 1986; pp 63-
100.
(15) See the Experimental Section for a description of these methods.
(16) The function OFIT in Siemens SHELXTL17 was used to determine
the best fit of the experimental and calculated geometries and the deviations
of the atomic positions; all of the non-hydrogen atoms were employed for
the fitting.
(17) Sheldrick, G. M. SHELXTL Version 5. Siemens Analytical X-ray
Instruments, Madison, Wisconsin, 1996.
(18) (a) Martin, J. M. L.; El-Yazal, J.; Francois, J.-P. Mol. Phys. 1995,
86, 1437-1450. (b) Baldridge, K. K.; Siegel, J. S. Angew. Chem., Int. Ed.
Engl. 1997, 36, 745-748.

730 J. Am. Chem. Soc., Vol. 121, No. 4, 1999
Barnett et al.
Figure 3. Schematic and perspective drawings of the D3 and C2 conformations of hexabenzotriphenylene. The HF/3-21G(*) geometries have been
used.
The C2/D3 Dichotomy. The experimentally determined
conformation of hexabenzotriphenylene provides an important
check of the principles thought to govern the shapes of similar
molecules. Consider polycyclic aromatic compounds with
“ideal” D3h symmetry (that is, when drawn flat on paper). Simple
examples such as triphenylene (7) may realize this symmetry
(at least approximately),19 but with increasing substitution such
molecules must adopt conformations of lower symmetry to
relieve steric congestion. Intuitively one might expect these
compounds to distort into D3-symmetric molecular propellers,
as observed for 1, but recent X-ray structures of several similar
molecules show this to be more the exception than the rule!
Both perfluorotriphenylene (8) and perchlorotriphenylene (9)
adopt C2-symmetric conformations which are not propellers at
all.8,20 Decacyclene (10) is a D3 propeller,21 but, in contrast,
2,5,8,11,14,17-hexa(tert-butyl)-decacyclene (11, where the tert-
possibilities that exist for the nominally D3h polycyclic aromatic
compounds 7-11 as well. Note that in the D3 structure of 1,
the interleaving of the outer benzo groups yields an alternating
“up-down-up-down-up-down” pattern, but in the C2 con-
formation the pattern is “up-down-up-up-down-down”. In
the D3 geometry, the greatest distortion is in rings B, C, and D,
which are twist-boats, whereas the A ring is a much less
distorted, shallow chair. By contrast, in the C2 geometry, the A
ring is most highly distorted (twist-boat), and the B, C, and D
rings are shallow boats.
We have previously rationalized the C2 structures of the
perhalotriphenylenes by noting that most of the distortion in
the C2 conformation is forced on the central, “nonaromatic” ring
(that is, it shows a high degree of bond alternation),23 whereas
a D3 conformation would minimize distortion in the central ring
at the expense of increased distortion in the peripheral, more
highly delocalized, “aromatic” rings.8b If this interpretation is
correct, then hexabenzotriphenylene (where the bond distances
in the X-ray structure show that the central ring is more highly
delocalized than in triphenylene but, in contrast, that the B, C,
and D rings have strong bond alternation) should indeed adopt
a D3 conformation to minimize distortion of its fully aromatic
central ring, despite intramolecular steric interactions closely
akin to those in the perhalotriphenylenes. The magnitude of the
out-of-plane distortions in each ring of 1 are fully consonant
with this idea. The rms deviation of the carbon atoms of ring A
from the mean plane of the ring is only 0.058 Å, but the rms
deviations of the carbons in the highly twisted rings B, C, and
D average 0.119 Å. The six peripheral benzo groups are nearly
planar, with rms deviations averaging only 0.010 Å.
butyl groups are not in conflict) adopts an approximate C2
conformation in the crystal.22 Why should these seemingly
similar structures fall into two distinct classes?
The two conformations of hexabenzotriphenylene are depicted
in Figure 3, and they serve to illustrate the conformational
We thus have a simple rule-of-thumb for predicting the
conformations of overcrowded “D3h” polycyclic aromatics: if
the central ring is expected to be aromatic (possessing shorter,
benzene-like bonds), then a D3 conformation should be pre-
(19) (a)Ahmed, F. R.; Trotter, J. Acta Crystallogr. 1963, 16, 503-508.
(b) Filippini, G. J. Mol. Struct. 1985, 130, 117-124.
(20) Hursthouse, M. B.; Smith, V. B.; Massey, A. G. J. Fluorine Chem.
1977, 10, 145-155.
(21) Ho, D. M.; Pascal, R. A., Jr. Chem. Mater. 1993, 5, 1358-1361.
(22) Zimmermann, K.; Goddard, R.; Kruger, C.; Haenel, M. W.
Tetrahedron Lett. 1996, 37, 8371-8374.
(23) Triphenylene is known from both experimental19 and computa-
tional24 studies to exhibit a great deal of bond alternation in the central
ring, and it is best regarded as three fully delocalized benzene rings linked
by bonds of low order. In triphenylene the endo bonds average 1.411 Å,
the exo bonds average 1.470 Å,19b and similar, or even greater, bond
alternation is seen in 2 and 3.8,19

Structure of Hexabenzotriphenylene
J. Am. Chem. Soc., Vol. 121, No. 4, 1999 731
Table 2. Calculated Energies for C2 and D3 Conformations of Overcrowded “D3h” Polycyclic Aromatic Compounds
mechanics
∆Hf (kcal/mol)
semiempirical
∆Hf (kcal/mol)
ab initio
E (HF, aua)
compd
MMX
MNDO
AM1
PM3
STO-3G
3-21G(*)
C2 hexabenzotriphenylene (1)
D3 hexabenzotriphenylene
differenceb
183.70
174.66
9.0
217.70
211.50
6.2
215.47
209.97
5.5
201.06
196.31
4.8
-1584.99208
-1585.00454
7.8
-1595.52644h
-1595.53933h
8.1h
experimental geometryc ∼D3
C2 perfluorotriphenylene (8)
D3 perfluorotriphenylene
differenceb
-396.90
-400.14
3.2
-430.39
-428.70
-1.7
-409.85
-408.24
-1.6
-415.96
-414.31
-1.7
-1849.72326
-1849.71697
-3.9
-1864.56011
-1864.54950
-6.7
experimental geometryd C2
C2 perchlorotriphenylene (9)
D3 perchlorotriphenylene
differenceb
60.32
52.47
7.8
61.57
61.22
0.4
55.29
56.25
-1.0
38.88
40.72
-1.8
-6128.13901
-6128.13071
-5.2
-6166.48582
-6166.47428
-7.2
experimental geometrye ∼C2
C2 decacyclene (10)
191.35
188.89
2.5
193.41
191.93
1.5
228.83
233.61
-4.8
195.90
195.62
0.3
-1357.10042
-1357.10135
0.6
-1366.10211
-1366.10327
0.7
D3 decacyclene
differenceb
experimental geometryf ∼D3
C2 2,5,8,11,14,17-(t-Bu)6decacyclene (11)
D3 2,5,8,11,14,17-(t-Bu)6decacyclene (11)
differenceb
34.19
31.92
2.3
171.62
176.73
-5.1
123.79
123.71
0.1
52.79
53.39
-0.6
-2283.00430
-2283.00482
0.3
-2297.76787
-2297.76856
0.4
experimental geometryg ∼C2
C2 1,4,5,8,9,12-Me6triphenylene (12)
D3 1,4,5,8,9,12-Me6triphenylene (12)
differenceb
79.13
78.83
0.3
69.42
70.83
-1.4
56.40
59.43
-3.0
-911.69406
-911.68179
-7.7
-917.65401
-917.64261
-7.2
predicted geometry: C2
C2 1,6,7,12,13,18-Me6decacyclene (13)
D3 1,6,7,12,13,18-Me6decacyclene (13)
differenceb
189.05
179.38
9.7
214.42
214.41
0.0
175.78
173.72
2.1
-1588.53257
-1588.54424
7.3
-1598.96346
-1598.97534
7.5
predicted geometry: D3
C2 hexafurotriphenylene 14
D3 hexafurotriphenylene 14
differenceb
-3.31
-4.99
1.7
98.93
98.14
0.8
31.01
31.11
-0.1
-1572.23281
-1572.23163
-0.7
-1583.19941
-1583.19716
-1.4
predicted geometry: C2
C2 hexafurotriphenylene 15
D3 hexafurotriphenylene 15
differenceb
-2.14
-7.00
4.9
89.88
86.67
3.2
36.14
33.34
2.8
-1572.16569
-1572.17430
5.4
-1583.12209
-1583.13164
6.0
predicted geometry: D3
C2 hexaphenanthrotriphenylene 16i
D3 hexaphenanthrotriphenylene 16
differenceb
459.52
457.52
2.0
predicted geometry: D3
a 1 au ) 627.503 kcal/mol. b C2 - D3; all differences are given in kcal/mol; positive values favor the D3 conformation. c This work. d Reference
20. e Reference 8. f Reference 21. g Reference 22. h At higher levels the following results were obtained: HF/6-31G*: C2 - 1604.49522, D3
-
1604.50826, difference 8.2 kcal/mol; B3LYP/cc-pVDZ: C2 - 1615.07248, D3 - 1615.08045, difference 5.0 kcal/mol. i The ∆Hf of the most stable
C2 conformation is given.
ferred, but if the central ring is nonaromatic (possessing some
very long “single” bonds and great bond alternation), then a C2
conformation will be observed. How well do the available
experimental data agree with this simple concept and with the
results of computational studies?
Table 2 lists the results of molecular mechanics, semi-
empirical molecular orbital, and ab initio molecular orbital
calculations of the energies of the fully optimized C2 and D3
conformations of the five overcrowded D3h polycyclic aromatic
compounds for which there are experimental geometries, as well
as calculations for five as-yet-unknown polycyclic aromatic
compounds. In addition, Table 3 compares some of the
experimental and calculated structures16 and provides some
selected geometric data. The known compounds are discussed
first.
All of the computational methods correctly yield a strong
preference (by 5-9 kcal/mol) for the D3 geometry of hexa-
benzotriphenylene (1), with the highest level employed, B3LYP/
cc-pVDZ, favoring the D3 conformation by 5.0 kcal/mol. For
the perhalotriphenylenes 8 and 9, however, the results are
mixed: at all ab initio levels, the experimentally observed C2
geometries are strongly preferred (by 4-7 kcal/mol), the
semiempirical methods show a weaker C2 preference (1-2 kcal/
mol), but the MMX force field favors the incorrect D3
geometries. The decacyclenes 10 and 11 are a more difficult
test, since the steric conflict in these molecules is not so great.
All of the methods but AM1 predict that decacyclene should
prefer the experimentally observed D3 conformation by a small
margin (0.3-2.5 kcal/mol), but the results for hexa(tert-butyl)-
decacyclene 11 are rather scattered. However, the higher levels
of theory indicate that the C2 and D3 conformations of 11 differ
in energy by at most a few tenths of a kcal/mol; thus, crystal
packing forces may have a significant influence on the preferred
conformation in the solid state.25 In any event, such small
differences in energy lie within the errors inherent in the
computational methods (especially the semiempirical calcula-
tions) and indicate no more than that both conformations are
accessible. Of special concern is the fact that the MMX force
field invariably prefers the D3 conformations, an apparent error
shared by the SYBYL26 and MMFF27 force fields (data not
shown).
(24) (a) Glidewell, C.; Lloyd, D. Tetrahedron 1984, 40, 4455-4472.
(b) Baldridge, K. K.; Siegel, J. S. J. Am. Chem. Soc. 1992, 114, 9583-
9587.

732 J. Am. Chem. Soc., Vol. 121, No. 4, 1999
Barnett et al.
Table 3. Comparison of Experimental and Calculated Geometries for Overcrowded “D3h” Polycyclic Aromatics
metrica
1-D3
8-C2
9-C2
10-D3
11-C2
12-C2
13-D3
14-C2
15-D3
Experimental vs Calculatedb Geometries
rms deviation
maximum dev.
0.059
0.111
0.041
0.067
0.058
0.137
0.067
0.167
0.203d
0.612
Experimental Central Ring Dimensions
endoc
1.397e
1.434
8.49
1.399f
1.495
8.68
1.414g
1.478
8.68
1.444h
1.384
8.48
1.439i
1.393
8.50
exoc
circumferencej
Calculated Central Ring Dimensions
endoc
1.386
1.428
8.44
1.407
1.470
8.63
1.406
1.486
8.68
1.440
1.369
8.43
1.444
1.370
8.44
1.417
1.491
8.72
1.448
1.369
8.45
1.400
1.464
8.59
1.405
1.405
8.43
exoc
circumferencej
circumference for other conformer 8.48 (C2) 8.62 (D3) 8.58 (D3) 8.43 (C2) 8.44 (D3) 8.67 (D3) 8.46 (C2) 8.55 (D3) 8.44 (C2)
a All distances are given in angstroms. b The HF/3-21G(*) geometries have been employed for the comparisons. c As defined in Table 1. d Because
of the relatively free rotation of the t-butyl groups, their methyl carbons were not included in the fit. e This work. f Reference 20. g Reference 8.
h Reference 21. i Reference 22. j (3 × endo) + (3 × exo).
The presently available experimental structures are perhaps
not the very best examples for an examination of the C2/D3
dichotomy. For example, the halogens in 8 and 9 may have
unforseen electronic effects, and the steric conflict in the
decacyclenes 10 and 11 is very small. For this reason the same
set of semiempirical and ab initio calculations were performed
on the hexamethyltriphenylene 12 and the hexamethyldeca-
cyclene 13 (Table 2); these molecules have comparable steric
and all of the calculations indicate that the D3 conformation is
preferred by 3-6 kcal/mol. Thus, the structures of 14 and 15
are expected to be dramatically different even though the steric
conflicts are essentially identical; the C2/D3 dichotomy is a
purely electronic effect.
As seen in Table 3, compounds 1 and 8-15 exhibit quite a
large variation in the degree of bond alternation in their central
rings; indeed, the best predictor of their conformational prefer-
ence appears to be the circumference of their central rings. Those
molecules which prefer a D3 conformation have central ring
circumferences that are less than 8.5 Å, while those preferring
a C2 conformation have central ring circumferences that are
usually greater than 8.6 Å. It is important to note that both the
C2 and D3 conformations of the same molecule are calculated
to have very similar circumferences of their central ring (Table
3); thus, the central ring geometry is not a product of the
conformation, but a determining factor. The lone exception is
11, for which the circumference is e8.5 Å by both experiment
and calculation but which is observed to adopt a C2 structure.
However, ab initio calculations indicate that the gas-phase
preference is for a D3 conformation (Table 2), and it is apparent
that crystal packing forces have distorted this molecule from
an ideal geometry.25
Finally, hexabenzotriphenylene may be considered the second-
generation “dendrimer” formed by the addition of two benzo
groups to the three outer benzene rings of triphenylene, itself
formed by the addition of three benzo groups to benzene. Few,
if any, other examples of this type of structure are known, and
the next member of the triphenylene series would be the
compound 16, which would be exceptionally crowded. Due to
its great size (C90H48) the conformation of this molecule was
explored only at the PM3 level; three separate C2 conformations
and one D3 conformation were identified which differ in the
interleaving of the twelve outer benzene rings. Of these, the D3
conformation is lowest in energy (see Table 2), a “violation”
of our central ring rule-of-thumb; however, this is hardly a
simple situation, and any experimental test of the conformation
of 16 must await its (very difficult) synthesis.
conflicts between the methyl substituents and, as hydrocarbons,
they should be among the least difficult tests for the computa-
tional methods. As expected, the triphenylene 12 is predicted
to possess a C2 conformation, whereas the decacyclene 13 is
predicted to be D3.
We have argued that the choice of the C2 and D3 conformation
is essentially governed by electronic, not steric, factors. This is
best illuminated by the isomeric hexafurotriphenylenes 14 and
15. For example, is 14 better described as a structure composed
of six furans and a central benzene ring (similar to 1) or as a
triphenylene with large peripheral substituents (similar to 8 and
9)? If the former, it would be expected to adopt a D3
conformation, if the latter, a C2. Ab initio calculations favor
the latter formulationsthere is strong bond alternation of the
central ring (see Table 3)sand the C2 conformation is preferred
by a small margin, at least by the higher levels of theory (Table
2). However, reorientation of the peripheral furans to give 15
yields a structure where there are no good “triphenylene”
resonance forms. Bond alternation in the central ring is absent,
(25) The experimental structure of compound 11 is an unusually distorted
C2 conformation. The HF/3-21G(*) geometries for 1, 8, 9, and 10 are in
excellent agreement with the experimental structures as judged by the rms
deviations of their best fits (see Table 3), but the experimental structure
for 11 shows a three times greater rms deviation from its calculated
geometry. The six tert-butyl groups of 11 must be accommodated in the
crystal, and it appears that they provide long levers for packing forces to
alter the geometry of the molecule from the gas-phase “ideal”, for which
the higher levels of theory favor a D3 conformation.
(26) Clark, M.; Cramer, R. D., III.; Van Opdenbosch, N. J. Comput.
Chem. 1989, 10, 982-1012.
(27) Halgren, T. A. J. Comput. Chem. 1996, 17, 490-519.

Structure of Hexabenzotriphenylene
J. Am. Chem. Soc., Vol. 121, No. 4, 1999 733
456 extinctions were discarded to give 6016 unique reflections in the
final data set. The structure was solved by direct methods and refined
by full-matrix least-squares on F2 (SHELXTL17). All atomic coordinates
were refined; carbon atoms were refined anisotropically, and hydrogens
isotropically. The refinement converged to R(F) ) 0.0512, wR(F2) )
0.1091, and S ) 1.095 for 3652 reflections with I > 2σ(I), and R(F) )
0.1007, wR(F2) ) 0.1327, and S ) 1.013 for 6016 unique reflections,
475 parameters, and 0 restraints. A second determination was performed
at 110 K using the same crystal: a ) 19.8971 (8) Å, b ) 6.9274 (2)
Å, c ) 19.4153 (8) Å, â ) 103.846 (1)°, V ) 2598.3 (2) Å3, Dcalcd
)
1.351 g/cm3; 57 890 reflections (θmax ) 27.5°), 5952 unique reflections;
R(F) ) 0.0563, wR(F2) ) 0.1417, and S ) 1.176 for 3809 reflections
with I > 2σ(I), and R(F) ) 0.0962, wR(F2) ) 0.1621, and S ) 1.054
for 5952 unique reflections, 475 parameters, and 0 restraints. Full details
are provided in the Supporting Information.
Computational Studies. The MMX force field implemented in
PCMODEL (Version 5.0; Serena Software, Bloomington, Indiana) was
employed for molecular mechanics calculations. All semiempirical
molecular orbital calculations (MNDO, AM1, PM3) and most conven-
tional ab initio calculations at the HF/STO-3G and HF/3-21G(*) levels
were performed by using the SPARTAN program package (Version
5.0; Wavefunction, Inc., Irvine, California), and its built-in default
thresholds for wave function and gradient convergence were employed.
Frequency calculations were performed on the AM1- and PM3-
optimized equilibrium geometries to verify that these were true potential
minima. GAUSSIAN 94 (Gaussian, Inc., Pittsburgh, Pennsylvania) was
employed for several of the larger ab initio calculations, again
employing the default convergence criteria. In addition to the conven-
tional Hartree-Fock techniques, hybrid density functional calculations
(HDFT; an improvement over DFT methods obtained by inclusion of
the exact Hartree-Fock exchange based on Kohn-Sham orbitals) were
performed for comparison in this study. The HDFT methods employed
two different exchange-correlation functionals, Becke’s three-parameter
functional29 in combination (a) with nonlocal correlation provided by
the Lee-Yang-Parr expression30,31 which contains both local and
nonlocal terms, B3LYP, and (b) with the nonlocal correlation provided
by the Perdew 91 expression,32 B3PW91. Dunning’s correlation
consistent basis set, cc-pVDZ,33 was used with the HDFT methods.
This basis set is a [3s2p1d] contraction of a [9s4p1d] primitive set.
Finally, the HF/3-21G(*) calculations for compound 11 were performed
by using the parallel version of GAMESS34 and its analytically
determined gradients and search algorithms.
Conclusion
Hexabenzotriphenylene has been unambiguously character-
ized for the first time as a D3-symmetric molecular propeller,
which is unlike the C2 structures preferred by other crystallo-
graphically characterized, crowded, highly symmetric tri-
phenylenes. After evaluation of energies and geometries of this
and a variety of similar molecules by a wide range of
computational methods, we conclude that to obtain both
reasonably accurate geometries and relative conformational
energies for such compounds, one must go beyond simple
molecular mechanics and semiempirical techniques, to, at the
very least, low level ab initio calculations.
Experimental Section
Hexabenzotriphenylene (1). Phenanthrene-9,10-dicarboxylic an-
hydride9 (6, 60 mg, 0.24 mmol) was placed in the sealed end of a quartz
tube (1 cm × 60 cm), and the tube was attached to a vacuum pump
and evacuated (∼0.2 Torr). The center section of the tube (∼20 cm)
was placed in a tube furnace, and the furnace was heated to 700 °C. A
Bunsen burner was used to heat and sublime the anhydride (with some
decomposition and gas evolution) into the center section of the tube.
During the next few minutes, a yellow-brown material condensed on
the distal, unheated portion of the quartz tube. This material was
extracted with chloroform, and the extract was filtered, concentrated,
and fractionated by preparative silica gel TLC (solvent, 4:1 hexanes-
benzene). There were three major bands with Rf 0.57, 0.32, and 0.26.
The fraction at Rf 0.57 was shown by 1H NMR analysis to be
phenanthrene. The material collected at Rf 0.32 proved to be pure
compound 1 (2.2 mg, 5% yield). Slow cooling of a solution of 1 in
Acknowledgment. This work was supported by NSF grant
CHE-9707958 (to R.A.P.) and a grant from the Fulbright
Program (to K.K.B.), which are gratefully acknowledged.
Supporting Information Available: Crystal structure re-
ports for compound 1 (including full experimental details, tables
of atomic coordinates, bond distances, bond angles, thermal
parameters, and selected figures), and an X-ray crystallographic
file, in CIF format (PDF). This material is available free of
charge via the Internet at http://pubs.acs.org.
1
nitrobenzene yielded yellow crystals suitable for X-ray analysis. H
NMR (CDCl3, 300 MHz) δ 7.22 (t, 6 H, J ) 7.5 Hz), 7.55 (t, 6 H, J
) 7.5 Hz), 8.16 (d, 6 H, J ) 7.5 Hz), 8.54 (d, 6 H, J ) 7.5 Hz); MS,
m/z 528 (M+, 30), 352 (M - C14H8, 100), 176 (M - C28H16, 22); UV
(heptane) λmax 212, 236, 250, 298 (sh), 348 (sh), 362, 382 (sh) nm.
X-ray Crystallographic Analyses of Hexbenzotriphenylene (1).
Formula C42H24; monoclinic, space group P21/c, a ) 19.9721 (5) Å, b
) 7.0005 (1) Å, c ) 19.5456 (5) Å, â ) 104.013 (1)°, V ) 2651.4 (1)
Å3, Z ) 4, Dcalcd ) 1.324 g/cm3. An orange plate with dimensions
0.04 mm × 0.25 mm × 0.28 mm was used for intensity measurements
at 298 K with a Nonius KappaCCD diffractometer and Mo KR radiation
(λ ) 0.710 74 Å). A total of 45 305 reflections (θmax ) 27.4°) were
indexed, integrated, and corrected for Lorentz and polarization effects
(using the program DENZO28), the data were scaled and merged
(SCALEPACK28) to give 6512 unique reflections (Rint ) 0.066), and
JA983471I
(28) Otwinowski, Z.; Minor, W. Methods Enzymol. 1997, 276, 307-
326.
(29) Becke, A. D. J. Chem. Phys. 1993, 98, 5648-5652.
(30) Lee, C.; Yang, W.; Parr, R. G. Phys. ReV. B 1988, 37, 785-789.
(31) Miehlich, B.; Savin, A.; Stoll, H.; Preuss, H. Chem. Phys. Lett. 1989,
157, 200-206.
(32) Perdew, J. P.; Wang, Y. Phys. ReV. B 1992, 45, 13244-13249.
(33) Woon, D. E.; Dunning, T. H., Jr. J. Chem. Phys. 1993, 98, 1358-
1371.
(34) Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert, S. T.;
Gordon, M. S.; Jensen, J. H.; Koseki, S.; Matsunaga, N.; Nguyen, K. A.;
Su, S.; Windus, T. L.; Elbert, S. T. J. Comput. Chem. 1993, 14, 1347-
1363.