Tunable molecular assembly codes direct reaction pathways

Article abstract of DOI:10.1002/anie.200802606

In the twist: The extent of twisting from planarity in a series of perylenetetracarboxylic diimides (PDIs) modulates the attractive π-π stacking force, revealing an array of inherent molecular recognition codes. Such coded self-assembly directs specific reaction pathways so that a mixture of reactive monomers with different codes results in identical products as when the reactions were carried out in separate flasks (see picture).

Full text of DOI:10.1002/anie.200802606

Angewandte
Chemie
DOI: 10.1002/anie.200802606
Molecular Codes
Tunable Molecular Assembly Codes Direct Reaction Pathways**
Andrew D. Shaller, Wei Wang, Haiyang Gan, and Alexander D. Q. Li*
The physical barrier of a single reaction flask corrals the
reaction pathway between a pair of reactants at high
concentrations (> 1 mm). In living cells, however, self-organ-
ization directs and regulates specific reactions, often at very
dilute (ca. nm) concentrations while competing with a myriad
of other possible reactants without physical barriers.[1] Nature
accomplishes this multifarious task by using dynamic molec-
ular recognition to sort reactive centers according to inherent
molecular codes, thus augmenting their effective molarity
both efficiently and specifically.[2] Weak secondary forces
available for molecular encoding include dispersion forces,
coulombic interactions, hydrogen bonding, solvophobic
forces, metal–ligand interactions, and p–p stacking interac-
tions.[3] Complex molecular codes can be created from
relatively few factors such as size, shape, and charge. The
interactive forces are often used simultaneously and synerg-
istically to balance molecular rigidity and flexibility, thereby
producing self-sorting effects that control reaction pathways.
Self-assembly by p–p stacking forces, which result from
the molecular overlap of p orbitals between planar aromatic
systems, has recently attracted much interest.[4] Perylene
diimides (PDIs) are particularly interesting because of their
efficient assembly,[5] which is often augmented by solvophobic
forces, and thus steric groups are often added in either the
imide or bay positions to inhibit their strong propensity to
self-assemble or ultimately precipitate.[6] However, function-
alization of the PDIs with flexible tetraethylene glycol (TEG)
chains balances rigidity and flexibility in a wide range of
solvents while still allowing dynamic self-assembly (DSA) to
manifest its power.[7]
Bay substitutions twist the perylene unit dihedrally out of
Scheme 1. a) Bay substitutions encode compounds 1–4, twisting them
the plane, with the dihedral angle tunable from 0 to 378.[8]
from 0 to 378. b) Heterogeneous mixtures of encoded PDIs self-sort
Herein, we examine DSA between bay-substituted mono-
into homogeneous stacks.
mers 1–4 (Scheme 1a) to broadly establish the role of the
dihedral twist angle in molecular encoding.[9] Remarkably,
monomers with different twist angles preferentially self-
assemble into segregated nanostructures even in the presence
of other twisted units, thus revealing their unique molecular
encryption (Scheme 1b).
Self-assembly in solution can be readily detected by using
complementary spectroscopic methods, especially UV/Vis
and NMR spectroscopy.[10] The extent of PDI self-assembly
can be monitored by absorbance, as DSA results in a
displaced harmonic oscillator that favors excitation to the
second vibrational level (v’ = 1) of the excited state instead of
the lowest vibrational level (v’ = 0). This results in a diagnostic
change in the Franck–Condon factors, which is manifested by
a dramatic alteration in the ratio of the absorbance vibra-
[*] A. D. Shaller, Dr. W. Wang, Dr. H. Gan, Prof. A. D. Q. Li
Department of Chemistry
Washington State University
Pullman, WA (USA)
Fax: (+1)509-335-8867
E-mail: dequan@wsu.edu
[**] We acknowledge support from the National Institute of General
Medicine Sciences (Grant GM065306), the Mass Spectrometry Core
Laboratories at WSU for discussions concerning MALDI data, and
the WSU Center for NMR spectroscopy for assistance with variable
temperature NMR studies. A.D.Q.L. is a former Beckman Young
Investigator (BYI).
tional peaks robs = A0ꢀ0/A0ꢀ1 [9–12]
.
NMR spectroscopy was used as a complementary tech-
nique, and shows the changes in the environment of the
aromatic perylene protons, which experience upfield chem-
ical shifts Ddobs upon assembly as the p-stack ring current
increases the shielding of the neighboring protons.[7,9]
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.200802606.
Angew. Chem. Int. Ed. 2008, 47, 7705 –7709
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
7705

Communications
Figure 1. Concentration-dependent a) A0ꢀ0/A0ꢀ1 ratio (robs) and b) NMR chemical shift (dobs) for compounds: 1 ( ), 2 ( ), 3 ( ), and 4 ( ) in
CHCl3/CDCl3 at RT. Values calculated by isodesmic modeling studies are represented by solid lines. Critical assembly concentrations are marked
by vertical dashed lines. Typical concentration-dependent spectra for compound 1 are shown as insets. c) Calculated DGSA values vary nearly
linearly with the dihedral twist angle q. d) Chemical shifts of the perylene protons for compounds 1, 3, and 4 in a 1:1:1 mixture (symbols)
compared to the same shifts for separate homogeneous solutions (lines) at equal concentrations. Nearly identical values between the
homogeneous and heterogeneous solutions demonstrate almost exclusively independent molecular codes. 1H NMR spectra of the same mixture
at four sample concentrations are shown as an inset.
~
&
*
^
Although all the aromatic protons undergo a change in
chemical shift upon stacking, similar non-bay-position pro-
tons were compared for consistency between different twisted
species.[9]
(KSA) for indefinite self-association is not trivial, as self-
assembled dimers, trimers, and higher stacks typically coexist
with the monomer and lead to absorbance ratios and chemical
shifts beyond a simplified dimerization model.[7] The value of
KSA is determined using the isodesmic equal K model
summarized by Martin.[13] In this model, the p-stacking
force is assumed to be the same between the ith and
(i+1)th monomer, and K2 = K3 = …K1 = KSA. The calculated
r and d values, determined from least-squares fitting to the
observed values, are plotted as solid lines in Figure 1a and b.[9]
In agreement with Würthner and co-workers,[5b] the free
energy for self-assembly (DGSA = ꢀRTlnKSA) exhibits nearly
linear dependence on the twist angle (Figure 1c). Variable
solvent composition and temperature also drive the same
stacking interactions, and Vanꢀt Hoff analysis corroborates
the concentration-driven DGSA values.[9]
The tunable twisted monomers assemble because of their
unique encoding. Even more remarkably, monomers with
different codes are able to assemble simultaneously and
independently from a mixture. To test code specificity, we
compared the concentration-dependent dobs values in a
heterogeneous equimolar 1:1:1 mixture of compounds 1, 3,
and 4 to those of the individual homogeneous solutions
(Figure 1d).[9] Remarkably, the mixture self-organizes so that
Both UV/Vis and NMR spectroscopy show that the extent
of DSA increases with concentration (Figure 1a,b); however
the monomers 1–4 with different twist angles organize at
distinct rates and the more highly twisted monomers assemble
at increasingly higher concentrations. At low concentrations,
the monomers remain mostly free and the observed physical
property Pobs (robs or dobs) shows little change over several
orders of magnitude of concentration. Once the concentra-
tion reaches the molecular-coded critical concentration
dictated by the twist angle, a significant percentage of the
monomers form stacked dimers, and Pobs, which indicates the
collective molecular behavior, begins to change noticeably.
Longer stacks form at even higher concentrations, although
the rate of change in Pobs decreases as the number of
assembled units increases. The critical concentration onset
covers a broad range (greater than two orders of magnitude),
which implies the existence of unique molecular codes that
can be exploited to control reaction pathways.
While DSA can be clearly detected by UV/Vis and NMR
spectroscopy, determinination of the equilibrium constant
7706 www.angewandte.org
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2008, 47, 7705 –7709

Angewandte
Chemie
the monomers with different twist angles selectively self-
assemble with identical species, and very little cross-assembly
apparently occurs. If the molecular codes were interchange-
able, the three independent critical self-assembly concentra-
tions would collapse into one critical concentration and the
highly twisted compound 4 would assemble at much lower
concentrations than in the homogeneous case.
We have defined a new physical quantity cCA to describe
the extent of cross-assembly. The dependence of cCA on the
PDI code-enabling twist angles is shown in Table 1, with the
Table 1: Extent of cross-assembly between two compounds A and B
Compound[a]
1
2
3
4
1
2
3
4
1.00[b]
–
–
–
0.57
1.00
–
0.18
0.36
1.00
–
0.02
0.04
0.15
1.00
–
[a] Concentrated compound A in left column, dilute compound B along
top row. [b] cCA =1 for indistinguishable codes (no selectivity) and cCA
=
0 for completely independent codes (absolute selectivity).
difference in the homogenous and heterogeneous assembly
situations particularly evident. A trace amount of compound
B was added to a concentrated homogeneous solution of
monomer A, and the chemical shifts of each monomer in the
mixture were recorded (dobs) and compared to their chemical
shifts in infinitely dilute homogeneous solutions (ddil).[9] This
provides a numerical value for the extent of cross-assembly
(cCA) between compounds with different twist, where cCA
=
(dB,dilꢀdB,obs)/(dA,dilꢀdA,obs). Thus, cCA = 1 for indistinguishable
(redundant) codes and cCA = 0 for completely independent
codes. These cCA values suggest that 1, 3, and 4 establish
mostly unique molecular codes, while 2 exhibits significant
cross-assembly with 1 and 3 (see Table 1). The cCA value is
generally independent of the concentration of monomer A.
Remarkably, cross-assembly is limited except for compounds
with similar twist angles. A cCA value of less than 0.2 results in
essentially separate codes, which corresponds to a Dq value of
approximately 158. The twist angle alone, which ranges from 0
to 378, provides at least three distinguishable codes. There-
fore, molecular shapes act as selective molecular codes that
control self-organization.
Phosphoramidite coupling chemistry was used to relate
the value of Dq to the yield of a cyclization reaction between
differently twisted PDIs (Scheme 2) Different reaction yields
were obtained because similarly coded molecules effectively
reduce the activation entropy of coupling. Reaction of the
bifunctionalized planar monomer 5 with various twisted
monomers 1–4 in separate reactions afforded bichromophoric
cyclic compounds 6–10, which were isolated in all cases in
addition to other linear products. However, the reaction yield
of the desired binary cyclic compounds decreased as the twist
angle of the starting monomers 1–4 increased. The yield of
each reaction and data supporting the existence of molecular
encoding effects during the reactions is summarized in
Scheme 2.
Scheme 2. The planar bisphosphoramidite 5 was treated separately
withvarious twisted monomers to form cyclic structures 6–10.[9] As the
value of Dq increases, reaction yields diminishbecause of deteriorat-
ing compatible assembly codes. Concomitantly, the upfield chemical
shift (Dd) induced by the p stack decreases and the absorbance
vibronic peak ratio (robs) increases providing reliable metrics of code
compatibility. Reagents: a) CH2Cl2, N-phenylimidazolium triflate;
b) 0.2m I2 (CH2Cl2/pyridine/H2O 1:3:1).
Stronger attraction between planar precursors drives the
formation of a self-assembled stack, which reduces the
activation entropy for coupling and favorably orients the
reaction centers for a cooperative cyclization reaction that
preferentially yields cyclic products over other possible linear
polymer structures. As the difference in twist angle (Dq)
increases, the intermolecular assembly codes no longer hold
and the formation of heterochromophoric cyclic compounds
becomes less probable than other random collision products.
In all cases, the distance between the two cooperative
reaction centers is the same, roughly 39 Š. The nanoenviron-
ment of the reaction centers is also the same because both are
bound to tetraethylene glycol linkers. Electronic effects such
as electron induction or hyperconjugation from the perylene
unit (through-bond effects) cannot contribute to the reactivity
difference or alter the reaction course because the reaction
center is approximately 14 Š away from the perylene unit. At
Angew. Chem. Int. Ed. 2008, 47, 7705 –7709
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 7707

Communications
such an isolated position, perylene can hardly impart any
steric repulsion (through-space effects). The experimental
results demonstrate that molecular self-assembly invokes
selective coding phenomena and the reactions follow the
molecular codes programmed by the design of the PDIs.
The restrictive cyclic architectures force the chromo-
phores into proximity irrespective of the concentration. In a
similar way as the extent of homogeneous self-assembly can
be calculated by UV/Vis and NMR spectroscopy, the extent of
folding in covalently bound heterodimers can also be derived
using these spectroscopic techniques.[9] The robs and Dd values
for compounds 6–10 are summarized in Scheme 2. Despite
the restrictive architectures, cyclic dimers with incompatibly
coded chromophores (10) are essentially unfolded, whereas
compatibly coded structures (6) are strongly folded. Both
UV/Vis and NMR spectroscopy show that molecular shapes
act as codes for self-assembly.
Of course, monomers with similar twist angles tend to self-
organize irrespective of the presence of other coded struc-
tures. The next natural step was to test whether such
molecular codes can direct chemical reactions to form specific
products. Homochromophoric exotic catenane and ring
structures were previously synthesized from compounds 1
and 4 by using disulfide linkages.[14,15] From homogenous
mixtures of reactive monomers, planar 11 formed monomer
ring 13, dimer ring 14, and concatenated rings 15, while highly
twisted compound 12 only yielded homochiral or heterochiral
dimer ring 16.
To validate the molecular code concept, we employed a
1:1 heterogeneous reaction mixture of compounds 11 and 12,
which have identical reactive functional groups and can
undergo the same disulfide bond formation (Scheme 3). If the
reaction pathway was not driven by a molecular code, the
reaction products would contain an indistinguishable statis-
tical mixture of two building blocks 11 and 12. Convincingly,
only the homoreaction products 13–16, also formed in the
separate homogeneous reactions, could be isolated; no cross-
assembled product such as heterogeneous ring compound 17
was detected, regardless of when the reaction was quenched.
The heterogeneous reaction proceeded as if the reactions had
occurred in separate flasks, thus demonstrating that the
molecular codes determined by the structures with different
twist angles effectively direct the reaction products. There-
fore, it is self-assembly and not a random collision mechanism
that dominates the reaction course.
In contrast, the heterogeneous mixture of similarly coded
planar compounds 11 and 18 resulted in a new cross-
assembled compound, the heterochromophoric cyclic dimer
22 (Scheme 3). NMR and UV/Vis spectroscopy demonstrate
that both planar compounds independently display strong
homogeneous p-stacking interactions while jointly exhibiting
strong heterogeneous p-stacking interactions.[9] Even though
the monosulfur compound 18 has a larger potential p-stacking
area and the addition of such a five-membered ring enlarges
its size and generates a slightly higher p-stacking force, the
difference is not sufficient to form a unique code, thereby
providing an example of code redundancy.
Scheme 3. Reaction between independently coded monomers 11 and
12 results in only homochromophoric products 13–16, the identical
products observed in homogeneous reactions in separate flasks. No
heterochromophoric cyclic product 17 was isolated. However, reaction
between redundantly coded planar monomers 11 and 18 resulted in
new heterochromophoric compound 22, manifesting cross-assembly
between compatible codes. Conditions: a) NaOMe, CH2Cl2, air; H+
quench.
out in separate flasks, with molecular codes effectively
directing their distinct separate reaction pathways. When
two species co-self-assemble (cCA ꢁ 1.0), experimental results
show that molecular codes become redundant and the same
or very similar reaction pathways are followed. The imprinted
twist-angle (structural architecture) determines the thermo-
dynamic parameters and directs solvophobic interactions,
thereby establishing useful molecular codes. The similarity
limits in the twist angle for PDI code compatibility, Dqtwist
>
158, effectively distinguishes unique molecular codes, which
correspond to a DGSA value of about 1 kcalmolꢀ1. Differences
in chemical shift (Dd) and absorbance ratio (robs) effectively
report the extent of p stacking, thus providing metrics for
code deciphering.
Understanding the forces that direct self-assembly repre-
sents a crucial step forward in recreating the mechanisms
nature employs for efficient and specific synthesis of macro-
molecules. Twisting the PDI cores over a range of angles
effectively tunes the p-stacking force, which imprints distinct
molecular shapes and provides at least three independent
codes capable of organizing reaction centers and directing
separate reaction outcomes. Thus, a new hypothesis emerges:
matching molecular size, shape, and charge produces syner-
gistic weak intermolecular attractive forces, which impart
molecular codes; such molecular codes effectively control
chemical reaction pathways and products.
Received: June 3, 2008
Published online: September 2, 2008
In summary, when two species barely interact (cCA = 0.02),
experimental results show their reactions proceed as if carried
Keywords: molecular recognition · perylenes · selectivity ·
self-assembly · supramolecular chemistry
.
7708 www.angewandte.org
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2008, 47, 7705 –7709

Angewandte
Chemie
Liu, M. Yuan, H. Liu, S. Wang, Chem. Eur. J. 2006, 12, 8378 –
8385.
[6] a) X. Zhang, Z. Chen, F. Wurthner, J. Am. Chem. Soc. 2007, 129,
4886 – 4887; b) H. Langhals, R. Ismael, O. Yuruk, Tetrahedron
2000, 56, 5435 – 5441; c) C. Kohl, T. Weil, J. Qu, K. Mullen,
Chem. Eur. J. 2004, 10, 5297 – 5310.
[1] a) G. M. Whitesides, J. P. Mathias, C. T. Seto, Science 1991, 254,
1312; b) X. Li, D. R. Liu, Angew. Chem. 2004, 116, 4956 – 4979;
Angew. Chem. Int. Ed. 2004, 43, 4848 – 4870; c) B. Hess, A.
Mikhailov, Science 1994, 264, 223 – 224; d) T. Misteli, J. Cell Biol.
2001, 155, 181 – 185.
[7] a) W. Wang, J. J. Han, L.-Q. Wang, L.-S. Li, W. Shaw, A. D. Q. Li,
Nano Lett. 2003, 3, 455 – 458; b) W. Wang, L.-S. Li, G. Helms, H.
Zhou, A. D. Q. Li, J. Am. Chem. Soc. 2003, 125, 1120 – 1121.
[8] a) F. Würthner, Chem. Commun. 2004, 1564 – 1579; b) P. Oss-
wald, F. Wurthner, J. Am. Chem. Soc. 2007, 129, 14319 – 14326;
c) W. Wang, A. D. Bain, L.-Q. Wang, G. J. Exarhos, A. D. Q. Li,
J. Phys. Chem. A 2008, 112, 3094 – 3103.
[2] a) S. J. Rowan, S. J. Cantrill, G. R. L. Cousins, J. K. M. Sanders,
J. F. Stoddart, Angew. Chem. 2002, 114, 938 – 993; Angew. Chem.
Int. Ed. 2002, 41, 898 – 952; b) P. Mukhopadhyay, P. Zavalij, L.
Isaacs, J. Am. Chem. Soc. 2006, 128, 14093 – 14102.
[3] a) G. M. Whitesides, M. Boncheva, Proc. Natl. Acad. Sci. USA
2002, 99, 4769 – 4774; b) F. J. M. Hoeben, P. Jonkheijm, E. W.
Meijer, A. P. H. J. Schenning, Chem. Rev. 2005, 105, 1491 – 1546.
[4] a) M. M. J. Smulders, P. H. J. Schenning, E. W. Meijer, J. Am.
Chem. Soc. 2008, 130, 606 – 611; b) E. A. Meyer, R. K. Castel-
lano, F. Diederich, Angew. Chem. 2003, 115, 1244 – 1287; Angew.
Chem. Int. Ed. 2003, 42, 1210 – 1250; c) D. H. Busch, Top. Curr.
Chem. 2005, 249, 1 – 65; d) G. J. Gabriel, S. Sorey, B. L. Iverson,
J. Am. Chem. Soc. 2005, 127, 2637 – 2640.
[5] a) W. Su, Y. Zhang, C. Zhao, X. Li, J. Jiang, ChemPhysChem
2007, 8, 1857 – 1862; b) Z. Chen, U. Baumeister, C. Tschierske, F.
Wurthner, Chem. Eur. J. 2007, 13, 450 – 465; c) M. J. Ahrens,
L. E. Sinks, B. Rybtchinski, W. Liu, B. A. Jones, J. M. Giaimo,
A. V. Gusev, A. J. Goshe, D. M. Tiede, M. R. Wasielewski, J. Am.
Chem. Soc. 2004, 126, 8284 – 8294; d) Y. Li, Y. Li, J. Li, C. Li, X.
[9] See the Supporting Information for complete experimental
details, synthesis, and characterization of compounds 2–3, 5–10,
and 18–22, and additional supporting figures and tables.
[10] A. D. Q. Li, W. Wang, L. Q. Wang, Chem. Eur. J. 2003, 9, 4594 –
4601.
[11] robs ꢁ Huang–Rhys factorꢀ1
.
[12] F. C. Spano, S. C. J. Meskers, E. Hennebicq, D. Beljonne, J. Am.
Chem. Soc. 2007, 129, 7044 – 7054.
[13] R. B. Martin, Chem. Rev. 1996, 96, 3043 – 3064.
[14] W. Wang, L. Q. Wang, J. Palmer, G. Exarhos, A. D. Q. Li, J. Am.
Chem. Soc. 2006, 128, 11150 – 11159.
[15] W. Wang, A. D. Shaller, A. D. Q. Li, J. Am. Chem. Soc. 2008, 130,
8271 – 8279.
Angew. Chem. Int. Ed. 2008, 47, 7705 –7709
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 7709