Vol 446 | 22 March 2007 | doi:10.

1038/nature05681

LETTERS
Biasing reaction pathways with mechanical force
Charles R. Hickenboth1, Jeffrey S. Moore1,2,3, Scott R. White3,4, Nancy R. Sottos2,3, Jerome Baudry1 & Scott R. Wilson1

During the course of chemical reactions, reactant molecules need isomers, respectively3. Assuming these calculations accurately describe
to surmount an energy barrier to allow their transformation into the geometric and electronic changes that accompany the action of
products. The energy needed for this process is usually provided by extensional deformation, a formal disrotatory ring opening is pre-
heat, light, pressure or electrical potential, which act either by dicted for cis benzocyclobutene (BCB). The presence of reaction
changing the distribution of the reactants on their ground-state products that result from the conrotatory ring opening of the trans
potential energy surface or by moving them onto an excited-state isomer and from disrotatory ring opening of the cis isomer would
potential energy surface and thereby facilitate movement over strongly indicate that a mechanochemical process is in operation.
the energy barrier. A fundamentally different way of initiating To test these predictions experimentally, we incorporated bi-
or accelerating a reaction is the use of force to deform reacting functional BCB units 1 and 2 into well-defined, link-functionalized
molecules along a specific direction of the reaction coordinate. polymers7 (LFPs) (Fig. 2). Using this preparation, only a single
Mechanical force has indeed been shown to activate covalent mechanophore—the putative mechanochemically active unit—is
bonds in polymers, but the usual result is chain scission1. Here incorporated into the centre of each polymer chain. The incorpora-
we show that mechanically sensitive chemical groups make it tion of the mechanophores into polymers is necessary to study their
possible to harness the mechanical forces generated when exposing mechanochemical behaviour because small molecules are largely
polymer solutions to ultrasound2, and that this allows us to accel- inert to mechanical forces8. To carry out the synthesis, the carboxylic
erate rearrangement reactions and bias reaction pathways to yield acid functions of 1 or 2 were coupled to a-methoxy-v-amino poly
products not obtainable from purely thermal or light-induced (ethylene glycol) (mPEG)9. This synthesis yielded a series of polymers
reactions. We find that when placed within long polymer strands, 3 and 4, which contained the trans-BCB linker and the cis-BCB linker
the trans and cis isomers of a 1,2-disubstituted benzocyclobutene respectively, in weight-averaged molecular masses of Mw 5 4, 10, 20,
undergo an ultrasound-induced electrocyclic ring opening in a 40 and 60 kDa.
formally conrotatory and formally disrotatory process, respec- Having only a single BCB unit per 40 kDa polymer chain requires
tively, that yield identical products. This contrasts with reaction sensitive analytical techniques to observe reactivity at this specific
initiation by light or heat alone3, in which case the isomers follow site. N-(1-pyrene)-maleimide 5 (ref. 10) was used as a dienophile
mutually exclusive pathways to different products. Mechanical trap; the pyrene label allows the reaction to be monitored by gel
forces associated with ultrasound can thus clearly alter the shape permeation chromatography (GPC) using an ultraviolet (UV)
of potential energy surfaces4 so that otherwise forbidden or slow detector. The strength of the UV signal is normalized against that
processes proceed under mild conditions, with the directionally obtained from a refractive index (RI) detector. Each polymer mole-
specific nature of mechanical forces providing a reaction control cule contains only a single BCB unit, so cycloaddition of dienophile 5
that is fundamentally different from that achieved by adjusting and the ortho-quinodimethide (oQDM) diene generated by ring
chemical or physical parameters. Because rearrangement in our opening will result in one pyrene chromophore per polymer chain
system occurs before chain scission, the effect we describe might (Fig. 2).
allow the development of materials that are activated by mech- It has previously been demonstrated that the application of ultra-
anical stress fields. sound to dilute solutions of LFPs can be used to fragment a polymer
We used the COGEF (COnstrained Geometries simulate External chain at a single specific site9. The action of ultrasound on polymer
Force) method5 computationally to determine the effects of molecular molecules is generally thought to be mechanical, owing to frictional
deformation on trans and cis 1,2-dimethoxybenzocyclobutenes. A forces generated by the relative movement of solvent and macro-
profile of energy change relative to the ground state versus molecular molecules resulting from collapse of cavitation bubbles2. A dilute
elongation was produced for each isomer by systematically increasing solution of the 40 kDa LFP 3 or 4 was sonicated in acetonitrile in
the distance d between the methyl carbons while allowing all other the presence of a 500-fold excess of 5. After sonication, each polymer
atoms to fully relax (Fig. 1a). The energy and geometry at each defor- was analysed by GPC, monitoring the UV absorbance at 345 nm, a
mation state was calculated at the B3LYP/6-31G* level of density wavelength at which the starting polymer has no absorbance but
functional theory. The calculations predict that the two isomers are the pyrene label strongly absorbs. It was found that the UV signal
distorted in a different rotational sense. In the trans isomer, the simu- significantly increased relative to the RI signal in both polymers 3
lated extensional deformation induces a conrotatory torquing motion and 4 (Fig. 3a and b). This result suggests that the pyrene-labelled
of the substituents, consistent with the atomic motions experienced dienophile was incorporated into both LFPs, a particularly interest-
for the thermally allowed pathway for ring opening3 (Fig. 1b). In ing result in the case of 4 given the calculated predictions. Sig-
contrast, strain induces a disrotatory torquing motion of the sub- nificantly, these results were obtained for reactions conducted at
stituents in the cis isomer6, consistent with the Woodward–Hoffman approximately 6–9 uC, a temperature range for which the rate of
allowed photochemical pathway for ring opening. These simulations thermal ring opening is negligible.
predict that the mechanically activated processes are proceeding along Several control experiments were carried out. There was no increase
thermally allowed and disallowed pathways for the trans and cis in UV absorbance when LFP 3 was manipulated exactly as described
1
School of Chemical Sciences, 2Department of Materials Science and Engineering, 3The Beckman Institute, 4Department of Aerospace Engineering, University of Illinois at Urbana-
Champaign, Urbana, Illinois 61801, USA.
423
©2007 Nature Publishing Group
LETTERS NATURE | Vol 446 | 22 March 2007

above, with the ultrasound omitted (Supplementary Information). In particular, longer chains experience larger solvodynamic shear
This control experiment indicates that ultrasound is necessary for forces and accelerated reaction rates. Assuming that force reduces
the observed response. A PEG homopolymer of Mw 43 kDa containing the kinetic barrier to ring opening, larger forces will correlate to faster
no BCB linker was subjected to the same conditions as above. Al- rates. A reaction promoted purely by thermal energy is not expected to
though the UV absorbance increased slightly, it was by far less than have such a Mw dependence. The LFPs 3 and 4 with Mw 5 4, 10, 20 and
for 3 or 4 (Fig. 3c). These observations indicate that the increased UV 60 kDa were sonicated in the presence of 5. The UV/RI signal ratios at
absorbance was due to a mechanically activated ring opening reaction, the elution time corresponding to the Mw of the LFP were monitored
followed by cycloaddition of the oQDM intermediate with 5. by GPC. It was found that for both LFPs 3 and 4, the UV/RI signal ratio
Previous studies on ultrasound-induced polymer degradation have increased smoothly with increasing Mw above a threshold value of
demonstrated a chain-length dependence on the rate of reaction2,8,9. about 20 kDa (Fig. 3d). The increasing response with increasing Mw
indicates that the reaction in both polymers is activated by a mech-
anical process. Interestingly, for a given Mw, cis LFP 4 was found to
a H3C H3C react to a greater extent than trans LFP 3 for all the Mw values studied
O O above 20 kDa.
We also performed control experiments with PEG homopolymers
d d of Mw 5 5, 10, 20, 26 and 60 kDa. In this case, the UV/RI ratio of the
unbroken chains11 is independent of Mw (Fig. 3d). Furthermore, a
O O
1:1 mol ratio mixture of 4 kDa LFP 4 and 43 kDa PEG was sonicated.
H3C H3C No significant increase in the UV signal was observed (Supplemen-
trans benzocyclobutene cis benzocyclobutene tary Information). These results suggest that the linker must be
embedded within a polymer and the Mw must be above the critical
b value to observe reactivity: mixing alone is not enough.
Light
To determine the specific chemical structures of the pyrene-
modified LFPs, and to investigate the stereochemical consequences
dis E, E-isomer dis of mechanical activation, we resynthesized pyrene-labelled maleimide
5, enriched with 13C isotope in the carbonyl positions. Polymers 3 and
trans-isomer 4 (40 kDa) were subjected to ultrasound in the presence of 513C. After
sonication, each polymer was analysed by 13C nuclear magnetic res-
onance (NMR). It was found that both polymers 3 and 4 yielded a
cis-isomer E, Z-isomer single resonance at d 174.2 p.p.m. (Fig. 4a). That the polymers pro-
vided the same product was confirmed by examination of mixtures.
Heat When 1:1 mixtures of sonicated LFPs 3 and 4 were measured in either
CDCl3 or CD3OD, only a single 13C carbonyl peak was observed.
Comparison of the 13C NMR spectra obtained after sonication of
E, E-isomer LFPs 3 and 4 to model compound 6 indicated that the observed
con con NMR signal is due to the Diels–Alder adduct, which results from endo
cis-isomer addition of the maleimide to the E,E-oQDM intermediate (with
entgegen double bond geometry) (Fig. 4b). Sonication of 43 kDa
PEG under the same conditions yielded no detectable resonances in
E, Z-isomer trans-isomer
the carbonyl region.
The action of ultrasound on the trans BCB in polymer 3 induces a
formal conrotatory ring opening to the oQDM intermediate, while
Stress
the cis BCB in polymer 4 reacts via a formal disrotatory ring opening.
dis con That different modes of reactivity are observed for the two different
isomers confirms that the activation is neither thermal nor photo-
chemical but rather behaves according to COGEF simulations of
mechanical activation.
E, E-isomer
cis-isomer trans-isomer
The consequence of ultrasound-induced reactivity of LFPs 3 and 4
was compared to the products obtained under purely thermal con-
Figure 1 | Electrocyclic ring opening of benzocyclobutenes. The ditions. The thermal reaction products were generated by reacting
intermediate resulting from the electrocyclic ring opening of trap 513C with polymer 3 in toluene at 105 uC and with polymer 4 in
benzocyclobutenes depends on the method of activation and on the xylenes at 140 uC (ref. 12). 13C analysis indicated that trans LFP 3
geometry of the molecule. a, To simulate mechanical tensile force, the
distance d between the terminal methyl carbons of trans and cis 1,2-
reacted smoothly with maleimide 513C to give the same major prod-
dimethoxybenzocyclobutene was systematically increased. At each step the uct observed with polymer 3 and 4 under ultrasound conditions
geometry was minimized and the energy was calculated at the density (Fig. 4a). The thermal reaction of polymer 4 did not yield any recog-
functional theory B3LYP/6-31G* level. b, The expected pathways of nizable products. This result is consistent with previous reports12 and
photochemically activated and thermally activated electrocyclic ring opening our own experience on small-molecule model compounds that cis-
of BCB, and the computationally predicted results of mechanically activated 1,2-diacetoxybenzocyclobutenes thermally degrade to a complex
ring opening. Upon activation by light, the Woodward–Hoffman rules mixture.
predict that both cis and trans BCB undergo disrotatory (‘dis’) ring opening. To determine whether mechanical activation of the BCB linker
The cis isomer yields the E,E-oQDM intermediate, and the trans isomer yields may occur without polymer chain scission, polymer 3 was sonicated
the E,Z-oQDM intermediate (with zusammen double bond geometry). Upon
activation by heat, both isomers undergo conrotatory (‘con’) ring opening.
in the presence of maleimide trap 513C, yielding a bimodal GPC trace
The cis isomer yields the E,Z-oQDM intermediate, and the trans isomer yields with peak molecular weights of 40 and 20 kDa, presumably corres-
the E,E-oQDM intermediate. Activation by mechanical forces is predicted to ponding to unbroken and broken chains. The 40 kDa polymer, sepa-
induce a formal disrotatory ring opening in the cis isomer and a formal rated from the 20 kDa byproduct by preparatory GPC, showed the
conrotatory ring opening in the trans isomer, such that both isomers are characteristic UV increase of pyrene incorporation by cycloaddition.
predicted to yield the same E,E-oQDM intermediate. Furthermore, 13C analysis revealed that the Diels–Alder adduct was
424
©2007 Nature Publishing Group
NATURE | Vol 446 | 22 March 2007 LETTERS

Heat or
ultrasound

1 3
Mw = 4, 10, 20, 40, 60 kDa

Ultrasound

2 4
Mw = 4, 10, 20, 40, 60 kDa

Figure 2 | Preparation and reaction of mechanosensitive polymers. Shown maleimide trap 5. For 513C, the positions of the 13C labels are marked with a
are the preparation of link-functionalized polymers 3 and 4 from star.
bifunctional BCB units 1 and 2, and the reaction of these polymers with

present in the uncleaved polymer strands (Supplementary Infor-

mation). This observation demonstrates that the LFPs can be mech-
Before sonication After sonication
anically activated without fragmenting the backbone.
a The findings can be summarized by considering the reaction
coordinate diagrams shown in Fig. 4c and d. Mechanical forces per-
form work8 on the trans BCB unit by deforming bond lengths and
angles, which reduces the barrier of a formal conrotatory process and
Detector response (arbitrary units)

accelerates the reaction (Fig. 4c). For the cis isomer, mechanical force
does work on the BCB unit to reduce the barrier for a formal dis-
b rotatory process, even though it is known to be the higher energy ring
opening pathway3,6 (Fig. 4d). This explanation is consistent with the
computational studies, which suggested that elongational deforma-
tion would induce a disrotatory motion for the cis isomer rather than
the thermally allowed conrotatory ring opening (Supplementary In-
formation). The disrotatory process is favoured because the mole-
c
RI signal cular deformation induced by force is nearly congruent with the
UV signal deformation along the disrotatory reaction pathway.
In conclusion, we have shown that ultrasound can be applied to
polymer solutions to accelerate and alter the course of chemical
reactions. The effects of ultrasound are consistent with the intuitive
8 12 16 20 24 8 12 16 20 24
effects of a mechanical force, which acts to bias the reaction pathway
Retention time (min) Retention time (min)
towards products that best relieve the applied force by directly alter-
ing the molecular potential energy surface. Although there are prac-
d 3.0 tical limitations to the use of mechanochemistry for controlling the
outcome of chemical reactions, we expect that the procedures
2.5
UV/RI ratio

2.0 4
Figure 3 | Reaction of the mechanosensitive polymers. a–c, The top panels
1.5 3 show GPC traces of the polymers before (left) and after (right) sonication.
1.0 a, The 40 kDa trans polymer 3. b, The 40 kDa cis polymer 4. c, The 43 kDa
PEG PEG polymer. d, A plot of the ratio of the UV absorption to refractive index
0.5
signal intensity versus polymer Mw for 3, 4 and PEG. The increased UV
0 absorption of LFPs 3 and 4 as a function of Mw is evidence that the
0 20 40 60 80 ultrasound-induced reaction proceeds via mechanical activation. Error bars
Mw (kDa)
indicate the standard deviation observed for 2–4 independent trials.
425
©2007 Nature Publishing Group
LETTERS NATURE | Vol 446 | 22 March 2007

Stress
Stress-activated state
a c
Conrotatory
path

6 Disrotatory
path

Work by stress
Energy
4 thermal

3 thermal
E,Z-QDM

E,E-QDM
4 sonicated

3 sonicated
trans-BCB

185 180 175 170 165 Reaction coordinate

Stress-activated state
b d

Stress
CH3

Work by stress
O O O
Energy Conrotatory
path
N

E,E-QDM E,Z-QDM
O O O Disrotatory
path

CH3
6
cis-BCB

Reaction coordinate

Figure 4 | Comparison of reaction products and pathways for trans and cis red) is assigned to the ester carbonyl in 6 at natural abundance. The
benzocyclobutenes induced by ultrasound and by heating. a, A portion of resonance at d 168.0 p.p.m. (coloured green) is due to N-pyrene-2,3-
the 13C NMR spectrum comparing the products formed from polymers 3 and naphthimide, which occurs after thermal elimination of acid groups from
4 in the presence of 513C by ultrasound and by thermal activation. the Diels–Alder adducts. c, In trans-BCB, mechanical force accelerates a
Compound 6, whose structure is confirmed by X-ray crystallography formal conrotatory ring opening to the E,E-oQDM intermediate. d, In
(b), exhibits a resonance at d 174.2 p.p.m. (coloured blue) assigned to the cis-BCB, mechanical force promotes a formal disrotatory ring opening to
Diels–Alder adduct resulting from endo addition of the maleimide the E,E-oQDM intermediate even though the thermally allowed process is
dienophile to the E,E-oQDM intermediate (the carbon corresponding to the the conrotatory pathway. Additional transition states and intermediates
resonance at d 174.2 is marked with a star). This resonance is not observed in may exist and are not shown16.
the thermal reaction of polymer 4. The resonance at d 169.7 p.p.m. (coloured

reported here will be generally useful for discovering new mechano- plumbed through a Waters 2487 UV detector, and a Waters 410 RI detector. The
phores, the chemical reactivities of which are activated by external UV detector sensitivity was set to 4.00 aufs (absorbance units full scale), and the
force. absorption for all experiments was monitored at 345 nm. The RI sensitivity was
set to 8, and the temperature at the detector was kept constant with a thermostat
set between 31.9–32.5 uC. After the profile was recorded, the area under the UV
METHODS and RI curves for unbroken chains was measured using triSEC software (version
Sonication of polymers. To a solution of polymers 3 or 4 (0.75 mg ml21, 19 mM 3.0, revision B.00.05, Viscotek). The baseline and integration limits were set to
for 40 kDa polymer) in acetonitrile we added maleimide trap 5 (2.75 mg ml21). include the Mw range corresponding to the unbroken chains using the RI trace.
Approximately 12 ml of this solution was introduced into a Suslick cell13. A The UV area was then measured without changing the position of the integration
thermocouple and an argon line were threaded through the septa and placed in limits. The baseline in the UV trace was redrawn to include only the area within
contact with the solution, ensuring that they did not touch the probe. The third the integration limits.
13
side arm of the cell contained a septum. The cell was then wrapped in aluminium C NMR experiments. For 13C labelling experiments, the polymer of interest
foil to exclude light. Argon was bubbled through the solution for 30 min before was sonicated as above with 13C-labelled maleimide, 513C. The polymer from
each experiment as well as during the experiment. The entire system was placed in two such runs was isolated by preparatory GPC and combined in an NMR tube
an ice bath to maintain a temperature of 6–9 uC throughout sonication. The with the appropriate solvent. The spectra were recorded on an instrument oper-
sample was then sonicated (pulsed on 0.5 s, off 1.0 s, at a frequency of 20 kHz ating at a 13C frequency of 125 MHz, with about 10–15 mg of polymer dissolved
and at an intensity of 8.73 W cm22) for a total time of 45 min (15 min sonication in 0.5 ml of solvent and 15,000–17,000 acquisitions.
time). After sonication, the solvent was removed and the polymer sample was
isolated by preparatory GPC. The concentrations used in all experiments were Received 29 November 2006; accepted 15 February 2007.
estimated to be below the critical overlap concentration for PEG14,15. 1. Beyer, M. K. & Clausen-Schaumann, H. Mechanochemistry: The mechanical
Determination of UV/RI ratios. After isolation of the polymer, the molecular activation of covalent bonds. Chem. Rev. 105, 2921–2948 (2005).
weight profile was recorded on a GPC equipped with three Waters Ultrastyragel 2. Basedow, A. M. & Ebert, K. H. Ultrasonic degradation of polymers in solution. Adv.
preparatory GPC columns (104 Å, 103 Å and 500 Å) eluted with tetrahydrofuran, Polym. Sci. 22, 83–148 (1977).
426
©2007 Nature Publishing Group
NATURE | Vol 446 | 22 March 2007 LETTERS

3. Woodward, R. B. & Hoffmann, R. The conservation of orbital symmetry. Angew. 15. Rodd, L. E., Scott, T. P., Boger, D. V., Cooper-White, J. J. & McKinley, G. H. The
Chem. Int. Edn Engl. 8, 781–853 (1969). inertio-elastic planar entry flow of low-viscosity elastic fluids in micro-fabricated
4. Evans, E. Probing the relation between force-lifetime and chemistry in single geometries. J. Non-Newtonian Fluid Mech. 129, 1–22 (2005).
molecular bonds. Annu. Rev. Biophys. Biomol. Struct. 30, 105–128 (2001). 16. Kersey, F. R., Yount, W. C. & Craig, S. L. Single-molecule force
5. Beyer, M. K. The mechanical strength of a covalent bond calculated by density spectroscopy of bimolecular reactions: system homology in the mechanical
functional theory. J. Chem. Phys. 112, 7307–7312 (2000). activation of ligand substitution reactions. J. Am. Chem. Soc. 128, 3886–3887
6. Roth, W. R., Rekowski, V., Börner, S. & Quast, M. Die disrotative cyclobuten- (2006).
ringöffnung. Liebigs Ann. Chem. 409–430 (1996).
7. Boffa, L. S. & Novak, B. M. ‘‘Link-functionalized’’ polymers: an unusual Supplementary Information is linked to the online version of the paper at
macromolecular architecture through bifunctional initiation. Macromolecules 30, www.nature.com/nature.
3494–3506 (1997).
Acknowledgements We thank T. Martı́nez, K. Suslick and P. Beak for discussions.
8. Nguyen, T. Q. & Kausch, H.-H. Mechanochemical degradation in transient
This work was supported by grants from the Air Force Office of Scientific Research
elongational flow. Adv. Polym. Sci. 100, 73–182 (1992).
and The Petroleum Research Fund.
9. Berkowski, K. L., Potisek, S. L., Hickenboth, C. R. & Moore, J. S. Ultrasound-induced
site-specific cleavage of azo-functionalized poly(ethylene glycol). Author Contributions J.S.M. conceived the BCB experiment. C.R.H. performed all
Macromolecules 38, 8975–8978 (2005). experiments and initial computational analyses. J.B. performed the final
10. Reddy, P. Y., Kondo, S., Fujita, S. & Toru, T. Efficient synthesis of fluorophore- computational analysis. S.R. Wilson performed the crystallography. N.R.S., S.R.
linked maleimide derivatives. Synthesis 999–1002 (1998). White and J.S.M. directed the research and all authors wrote the manuscript.
11. Price, G. J. The use of ultrasound for the controlled degradation of polymer
solutions. Adv. Sonochem. 1, 231–287 (1990). Author Information The atomic coordinates for cis-1,2-bis
12. Arnold, B. J., Sammes, P. G. & Wallace, T. W. Photochemical reactions. Part IV. [(3-carboxypropanoyl)oxy]-1,2-dihydrobenzocyclobutene 2 (accession number
Thermal generation of photoenols and their derivatives from disubstituted 1,2- 628670), N-(1-pyrene)-2,3-naphthimide (accession number 628377), and
dihydrobenzocyclobutenes. J. Chem. Soc. Perkin Trans. I 415 (1974). 4,9-diacetoxy-2-(1-pyrenyl)-(3aR,4c,9c,9ac)-3a,4,9,9a-tetrahydro-benzo
13. Suslick, K. S., Goodale, J. W., Schubert, P. F. & Wang, H. H. Sonochemistry and [f]isoindole-1,3-dione 6 (accession number 628671) have been deposited in the
sonocatalysis of metal carbonyls. J. Am. Chem. Soc. 105, 5781–5785 (1983). Cambridge Crystal Structure Database. Reprints and permissions information is
14. Graessley, W. W. Polymer chain dimensions and the dependence of viscoelastic available at www.nature.com/reprints. The authors declare no competing financial
properties on concentration, molecular weight and solvent power. Polymer 21, interests. Correspondence and requests for materials should be addressed to
258–262 (1980). J.S.M. (jsmoore@uiuc.edu).

427
©2007 Nature Publishing Group