Photoswitchable catalysts: Correlating structure and conformational dynamics with reactivity by a combined Experimental and computational approach

Article abstract of DOI:10.1021/ja807694s

Photocontrol of a piperidine's Bronsted basicity was achieved by incorporation of a bulky azobenzene group and could be translated into pronounced reactivity differences between ON- and OFF- states in general base catalysis. This enabled successful photomodulation of the catalyst's activity in the nitroaldol reaction (Henry reaction). A modular synthetic route to the photoswitchable catalysts was developed and allowed for preparation and characterization of three azobenzene-derived bases as well as one stilbene-derived base. Solid-state structures obtained by X-ray crystal structure analysis confirmed efficient blocking of the active site in the E isomer representing the OFF-states, whereas a freely accessible active site was revealed for a representative Z isomer in the crystal. To correlate structure with reactivity of the catalysts, conformational dynamics were thoroughly studied in solution by NMR spectroscopy, taking advantage of residual dipolar couplings (RDCs), in combination with comprehensive DFT computational investigations of conformations and proton affinities.

Full text of DOI:10.1021/ja807694s

Published on Web 12/05/2008
Photoswitchable Catalysts: Correlating Structure and
Conformational Dynamics with Reactivity by a Combined
Experimental and Computational Approach
Ragnar S. Stoll,† Maike V. Peters,† Andreas Kuhn,† Sven Heiles,‡
Richard Goddard,§ Michael Bu¨hl,*, Christina M. Thiele,*,‡ and Stefan Hecht*,†
Department of Chemistry, Humboldt-UniVersita¨t zu Berlin, Brook-Taylor-Strasse 2, 12489
Berlin, Germany, Max-Planck-Institut fu¨r Kohlenforschung, Kaiser-Wilhelm-Platz 1, 45470
Mu¨lheim an der Ruhr, Germany, School of Chemistry, North Haugh, UniVersity of St. Andrews,
St. Andrews, Fife KY16 9ST, U.K., and Clemens Scho¨pf Institut fu¨r Organische Chemie and
Biochemie, Technische UniVersita¨t Darmstadt, Petersenstrasse 22, 64287 Darmstadt, Germany
Received September 29, 2008; E-mail: buehl@st-andrews.ac.uk; cmt@punk.oc.chemie.tu-darmstadt.de; sh@chemie.hu-berlin.de
Abstract: Photocontrol of a piperidine’s Brønsted basicity was achieved by incorporation of a bulky
azobenzene group and could be translated into pronounced reactivity differences between ON- and OFF-
states in general base catalysis. This enabled successful photomodulation of the catalyst’s activity in the
nitroaldol reaction (Henry reaction). A modular synthetic route to the photoswitchable catalysts was
developed and allowed for preparation and characterization of three azobenzene-derived bases as well as
one stilbene-derived base. Solid-state structures obtained by X-ray crystal structure analysis confirmed
efficient blocking of the active site in the E isomer representing the OFF-states, whereas a freely accessible
active site was revealed for a representative Z isomer in the crystal. To correlate structure with reactivity
of the catalysts, conformational dynamics were thoroughly studied in solution by NMR spectroscopy, taking
advantage of residual dipolar couplings (RDCs), in combination with comprehensive DFT computational
investigations of conformations and proton affinities.
Introduction
tion with the exceptional temporal and spatial control attainable,
distinguishes it as an ideal external stimulus.
Photocontrol over reactivity is an attractive task since a
number of possible applications can be envisioned.1 Catalysis
is of particular interest due to the potential of amplifying the
light stimulus by exploiting the catalytic cycle to enhance the
efficiency of the overall chemical process. Reversible control
can only be achieved by incorporating suitable photochromic
gates, to allow driving the photoswitchable catalyst between
(at least) two states of different ground-state reactivity. It is
important to note that this approach is conceptually different to
irreversibly photoactivated, i.e. caged, catalysts as well as
photocatalysts.2 The noninvasive character of light, in combina-
Almost all examples describing the photocontrol of catalytic
processes rely on geometrical changes of the catalyst’s backbone
that alter its interaction with the substrate.3,4 Although some
photoswitchable catalyst systems have been described, no
general and broadly applicable systems have yet been reported.
We therefore decided to develop a conceptually new approach
toward photoswitchable catalysts (Figure 1a).5 Our concept is
based on the reversible steric shielding of the catalyst’s active
site by a suitable blocking group, which can be moved away
from the reactive center using a photochromic linker. A two-
(2) (a) Parmon, V. N. Catal. Today 1997, 39, 137–144. (b) Esswein, A. J.;
Nocera, D. G. Chem. ReV. 2007, 107, 4022–4047. (c) Palmisano, G.;
Augugliaro, V.; Pagliaro, M.; Palmisano, L. Chem. Commun. 2007,
3425–3437. (d) Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahne-
mannt, D. W. Chem. ReV. 1995, 95, 69–96. (e) Linsebigler, A. L.;
Lu, G.; Yates, J. T., Jr. Chem. ReV. 1995, 95, 735–758.
† Humboldt-Universita¨t zu Berlin.
‡ Technische Universita¨t Darmstadt.
§ Max-Planck-Institut fu¨r Kohlenforschung.
University of St. Andrews.
(1) (a) Molecular Switches; Feringa, B. L., Ed.; Wiley-VCH: Weinheim,
2001. (b) Balzani, V.; Venturi, M.; Credi, A. Molecular DeVices and
Machines; Wiley-VCH: Weinheim, 2008. (c) Kay, E. R.; Leigh, D. A.;
Zerbetto, F. Angew. Chem., Int. Ed. 2007, 46, 72–191. (d) Photo-
chromism: Molecules and Systems; Du¨rr, H., Bouas-Laurent, H., Eds.;
Elsevier: Amsterdam, 2003. (e) Special Issue: Photochromism:
Memories and Switches. Irie, M. Chem. ReV. 2000, 100, 1683–1890.
(f) Organic Photochromic and Thermochromic Compounds; Crano,
J. C., Guglielmetti, R. J., Eds.; Plenum Press: New York, 1999. (g)
Organic Photochromes; El’tsov, A. V., Ed.; Consultants Bureau: New
York, 1990. (h) Mayer, G.; Heckel, A. Angew. Chem., Int. Ed. 2006,
45, 4900–4921. (i) Dynamic Studies in Biology; Goeldner, M., Givens,
R., Eds.; Wiley-VCH: Weinheim, 2005. (j) Willner, I. Acc. Chem.
Res. 1997, 30, 347–356. (k) Willner, I.; Rubin, S. Angew. Chem., Int.
Ed. Engl. 1996, 35, 367–385. (l) Hecht, S. Small 2005, 1, 26–28.
(3) (a) Ueno, A.; Takahashi, K.; Osa, T. Chem. Commun. 1980, 837–
838. (b) Ueno, A.; Takahashi, K.; Osa, T. Chem. Commun. 1981, 94–
95. (c) Wu¨rthner, F.; Rebek, J., Jr. Angew. Chem., Int. Ed. Engl. 1995,
34, 446–448. (d) Wu¨rthner, F.; Rebek, J., Jr. J. Chem. Soc., Perkin
Trans. 2 1995, 172, 7–1734. (e) Sugimoto, H.; Kimura, T.; Inoue, S.
J. Am. Chem. Soc. 1999, 121, 2325–2326. (f) Cacciapaglia, R.; Di
Stefano, S.; Mandolini, L. J. Am. Chem. Soc. 2003, 125, 2224–2227.
(g) Sud, D.; Norsten, T. B.; Branda, N. R. Angew. Chem., Int. Ed.
2005, 44, 2019–2021.
(4) The only exception relates to a photoswitchable Lewis acid; however,
no catalysis was reported. Lemieux, V.; Spantulescu, M. D.; Baldridge,
K. K.; Branda, N. R. Angew. Chem., Int. Ed. 2008, 47, 5034–5037.
(5) Peters, M. V.; Stoll, R. S.; Ku¨hn, A.; Hecht, S. Angew. Chem., Int.
Ed. 2008, 47, 5968–5972.
9
10.1021/ja807694s CCC: $40.75
2009 American Chemical Society
J. AM. CHEM. SOC. 2009, 131, 357–367 357

A R T I C L E S
Stoll et al.
Figure 1. (a) Concept of reversible steric shielding of a catalyst’s active site; (b) schematic representation of a photoswitchable piperidine base.
state catalyst system is obtained that can be driven between the
ON-state, having an active site freely accessible for substrate
molecules, and the OFF-state, which is sterically blocked and
therefore inactive. To maintain the desired photochromic
reactivity, other chromophores, potentially leading to detrimental
energy transfer processes, should be excluded.6
3a,b with BOC-protected hydrazines 4a,b followed by oxidative
deprotection, while stilbene-based catalyst 2 was prepared via
Suzuki cross-coupling of spiroannulated piperidine 3a with
styryl boronic ester 5.7
Structure and Conformational Dynamics. If the concept of
reversible steric shielding is to be applicable to the photocontrol
of the activity of a catalyst, then a large structural change
between ON- and OFF-states is required. Single crystals suitable
for X-ray structural analysis of E-1, representing the OFF-states,
and crystals of Z-1b, representing the ON-state were obtained,
and their structures were determined. Figure 2 shows a
superposition of the benzofuranone moieties of E-1a-c as well
as Z-1b. Remarkable is the similarity of the structures E-1a-c,
despite completely different crystal environments (root-mean-
square deviation of the diazo-piperidyl-phthalide units: 0.1 Å).
In the case of the crystal structure of E-1b there are three
independent molecules in the asymmetric unit which differ only
in the conformation of the methyl groups (root-mean-square
deviation of non-methyl atoms: 0.2 Å). The common structural
features of all structures E-1 are (1) the placement of the
blocking group directly in front of the lone pair of the nitrogen
atom, with shortest intramolecular N-C distances 4.1-4.5 Å
in the three structures; (2) the chair conformation of the
piperidine ring, to within 0.02 Å; and (3) the perpendicular
relationship between the benzofuranone unit and the piperidine
mean ring plane (mean angle: 89°). Clearly, the blocking group
takes up a position, which efficiently shields the lone pair of
the piperidine N-atom.
As a first step toward this ambitious goal, we recently
described the first prototype of a photoswitchable base (Figure
1b),5 potentially enabling photocontrol over a large variety of
general base-catalyzed reactions. Significant changes of Brønsted
basicity upon isomerization of an azobenzene blocking group
fused to a rigidified piperidine framework were demonstrated
and exploited to control the activity as a general base catalyst.
The accessibility to the basic site was conformationally restricted
using a locked piperidine chair in combination with a spiroan-
nulated azobenzene photochrome carrying a bulky shield.
Herein, we give a detailed experimental and theoretical analysis
of the conformational behavior of various generations of our
catalyst system, enabling a comprehensive discussion of their
photochromic behavior, structural dynamics, and resulting
structure-reactivity relationships.
Results and Discussion
Synthesis. A modular synthesis of photoswitchable bases
1a-c and 2 was developed to facilitate variation of key
substituents at both the piperidine’s N-atom and the terminal
phenyl moiety of the azobenzene in order to tune the properties
of the catalyst (Scheme 1). Azobenzene-based piperidine
catalysts 1a-c were readily assembled by a two-step procedure
involving Pd-catalyzed N-arylation of spiroannulated piperidines
In the case of the Z isomer (green in Figure 2), the azobenzene
moiety has a torsion angle 7.1°, and the blocking group is moved
away from the active site of the catalyst, leaving the active site
(6) (a) Peters, M. V.; Goddard, R.; Hecht, S. J. Org. Chem. 2006, 71,
7846–7849. (b) Peters, M. V.; Stoll, R. S.; Goddard, R.; Buth, G.;
Hecht, S. J. Org. Chem. 2006, 71, 7840–7845.
(7) See Supporting Information.
9
358 J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009

Photoswitchable Catalysts
A R T I C L E S
Scheme 1. Synthesis of Azobenzene-Derived Catalysts 1a-c and Stilbene-Derived Catalyst 2
freely accessible for reaction as intended in our design. The
intramolecular N-C distance of the lone pair of the piperidine
to the nearest tert-butyl group of the phenyl moiety is 7.1 Å,
and therefore the substituents on the phenyl ring no longer
provide a barrier. Otherwise, the conformation of the piperidine-
benzofuranone moiety is essentially similar to those seen in
E-1a-c. The conformation of the piperidine in all four crystal
structures is such that the N-lone pair points away from the
lone pairs on the furanone ring O-atom, in line with the less
destabilizing 1,3-diaxial interactions of the two piperidine 2,6-
H-atoms with the 4-carbonyloxy substituent of the lactone as
compared to the alternative larger phenyl substituent of the
benzofuranone. X-ray analysis, however, only provides a static
picture of the structural situation in the solid state; thus, for a
thorough understanding of the reactivity of catalysts E-1a-c
and Z-1 it is necessary to investigate their structure and
conformational behavior in solution, hence gaining detailed
insights into the dynamics of the systems.
To achieve the desired control over the accessibility of the
basic/nucleophilic piperidine N-atom’s lone pair, it is necessary
to dictate both configuration (path a in Scheme 2) as well as
conformation (paths b-d in Scheme 2) of the interconvertible
E and Z isomers. Control of the latter one is a much more
difficult task due to the complex dynamics under the reaction
conditions. We therefore investigated 1 in solution by NMR
spectroscopic methods, in particular, variable-temperature
NMR,8 EXSY,9 and the recently introduced residual dipolar
couplings (RDCs),10 in combination with computational studies
(Table 1).
Inversion at the piperidine’s N-atom (blue path b in Scheme
2) is of particular interest, because the concomitant exchange
of equatorial alkyl group and axial lone pair in E-1 exposes the
latter to the environment, potentially bestowing residual catalytic
activity on the formal OFF-state. The computed N-inversion
barrier via 1-TSi is rather similar for all bases, amounting to
approximately 7-8 kcal/mol (Table 1). As expected, the
N-inverted isomers i-1 carrying an axial substituent R are
significantly higher in energy than the equatorial ground state.
The species with the smallest R, i-1a, is least disfavored by ∼4
kcal/mol, according to DFT,11 whereas the increased bulk of
that group in i-1b and i-1c results in further destabilization by
approximately 1-3 kcal/mol (Table 1).
(8) Kessler, H. Angew. Chem., Int. Ed. Engl. 1970, 9, 219–235.
(9) Perrin, L.; Dwyer, T. J. Chem. ReV. 1990, 90, 935–967.
(10) (a) Thiele, C. M.; Berger, S. Org. Lett. 2003, 5, 705–708. (b) Thiele,
C. M. Concepts Magn. Reson. 2007, 30A, 65–80. (c) Thiele, C. M.
Eur. J. Org. Chem. 2008, 24, 5673–5685.
Figure 2. Overlay of X-ray crystal structures of catalysts E-1a-c and Z-1b
(green): (a) side view (projection onto the plane of the benzofuranone
moiety); (b) rear view.
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 359

A R T I C L E S
Stoll et al.
Scheme 2. Schematic Representation of Selected Local Minima
and Transition States Investigated by DFT Calculations
Table 1. Potential Energies ∆E (in parentheses: Free Energies
∆G)a of Selected Isomers and Transition States of 1a-c
a
b
c
E-1
1-TSEZ
Z-1
1-TSi
i-1
1-TSboat
fax-1
0.0 (0.0)
39.8 (37.6)
15.8 (14.8)
8.5 (7.4)
3.8 (4.0)
11.6 (12.0)
7.2 (7.8)
0.0 (0.0)
39.6 (37.6)
15.4 (15.8)
7.0 (8.2)
0.0 (0.0)
40.1 (38.0)
15.3 (15.0)
5.8 (7.2)
5.5 (5.4)
n.a.
6.0 (7.1)
11.5 (12.7)
9.0 (10.9)
7.4 (8.7)
n.a.
n.a.
feq-1
7.0 (7.6)
1-TSrot
7.7 (7.0)
7.5 (6.9)
6.9 (5.6)
a In kcal/mol relative to E-1 isomer, B3LYP/6-31G* level.
orientation of the N-tert-butyl group with respect to the
piperidine ring, thus yielding information on N-inversion.
1
Usually DC-N cannot be measured easily, yet in the case of
1b, it can be determined indirectly from the 1DC-H coupling of
the rotatable tert-butyl groups (attached to the piperidine N-atom
and the steric blocker).7 The structural proposal, fitting the
experimentally derived RDCs best, can be considered as accurate
representation of the spatial structure in solution.10b,c
Compound E-1b was oriented in a lyotropic liquid crystalline
matrix of high-molecular weight PBLG13 with CDCl3 as organic
cosolvent, and variable angle sample spinning (VASS)14 was
used to reduce RDCs to a suitable range, as strong coupling
was observed between the diastereotopic protons of the piperi-
1
dine methylene groups. This enabled us to measure all DC-H
and even some long-range DH-H. The geometry-optimized DFT
structures of E-1b and i-1b were subsequently used to probe
N-inversion. In order to distinguish between the two conformers,
1
a scaling procedure had to be used such that DC-N gets
approximately the same weight as 1DC-H 15 Two approaches of
.
RDC data analysis were applied. First the respective coupling
1DC-N, describing the orientation of the tert-butyl group with
respect to the piperidine ring, was included in the fitting step
using the DFT geometries of E-1b and i-1b and all RDCs except
those of the steric blocker (Vide infra).7 Using this procedure,
a significantly better fit was obtained for E-1b (quality factor q
) 0.131) than for i-1b (q ) 2.046). This already suggests that
the N-inverted isomer i-1b is not significantly populated,
whereas the proposed structure E-1b is matched extremely well
by the experimental RDCs (Figure 3b). To independently
confirm our findings, a second kind of analysis was performed
It is not straightforward to investigate N-inversion experi-
mentally. For N-methyl-substituted piperidines, 13C chemical
shifts for the N-methyl carbon for axial and equatorial methyl
groups have been reported.12 However, the chemical shift of
the only N-methyl signal observed for E-1a, δ (13C) ) 45.6
ppm, lies exactly in between those two values and does not
change position or line shape even when lowering the temper-
ature to 178 K. This observation can have two possible
explanations: either only one conformer is mainly populated
(no change in signal position and line shape) or N-inversion is
fast on the NMR time scale even at low temperatures (chemical
shift in between those reported for axial and equatorial N-methyl
groups). As we did not obtain conclusive experimental results
for 1a, we decided to conduct a more detailed structural and
dynamical NMR investigation on the E isomer of the catalyti-
cally more active system 1b.
1
by excluding DC-N from the fitting procedure and predicting
its value from the orienting tensor of the rest of the compound16
(again without using the data of the steric blocker). The predicted
1DC-N value of -24.1 Hz for E-1b is in good agreement with
the measured coupling of -36 ( 18 Hz as compared to the
We used residual dipolar couplings (RDCs) for this purpose,
as the other conformationally relevant NMR-parameters (NOE,
J couplings) are not expected to lead to unequivocal results. If
only dipolar couplings of directly bonded nuclei (1D) are used
for the determination of the spatial structure of rigid compounds,
RDCs can be used as angular restraints.10 In this case we use
1DC-N (coupling between the quaternary carbon of the N-tert-
butyl group and the piperidine N-atom) to determine the
(13) Marx, A.; Thiele, C. M. Chem. Eur. J. 2008, http://dx.doi.org/10.1002/
chem.200801147.
(14) Thiele, C. M. Angew. Chem., Int. Ed. 2005, 44, 2787–2790.
(15) The two conformers could initially not be distinguished because of a
1
fitting problem as DC-N becomes very small (because of the small
gyromagnetic ratios) and the PALES program (Zweckstetter, M.; Bax,
A. J. Am. Chem. Soc. 2000, 122, 3791-3792) used for fitting the data
does not allow for weighing of couplings within the fitting procedure,
resulting in equally good fits for E-1b and i-1b. To obtain conclusive
results, the corresponding coupling was scaled by a factor of 1000,
while the bond length was reduced in the coordinate file to 10% of its
original value using the program MolArch+. (Immel, S. MolArch+,
Molecular architecture modelling program; Universita¨t Leipzig:
Leipzig, 2008.) As 1D is proportional to 1/r3 the physical information
(11) A similar energy difference has been obtained between equatorial and
axial N-methyl piperidine, 3.9 kcal/mol at the MP2/cc-pVTZ level: Dos
Santos, F. P.; Tormena, C. F J. Mol. Struct. (THEOCHEM) 2006,
763, 145–148.
1
of DC-N is unchanged by this procedure.
(12) Crowley, P. J.; Robinson, M. J. T.; Ward, M. G. Tetrahedron 1977,
33, 915–925.
(16) This method is described in: Thiele, C. M. J. Org. Chem. 2004, 69,
7403-7413.
9
360 J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009

Photoswitchable Catalysts
A R T I C L E S
These observations were further supported by EXSY experi-
ments, revealing no exchange between these diastereotopic
protons and thus excluding any significant population of ring-
flipped isomers feq-1a,b. Final confirmation of the absence of
chair inversion was obtained by fitting the RDC data for E-1b
to the DFT geometries of E-1b and its corresponding ring-
flipped conformer feq-1b. As expected, the computed structure
E-1b is in excellent agreement with the RDC data (q ) 0.067,
see Figure 3a, blue diamonds), whereas alternative structure feq-
1b results in a poor fit (q ) 0.823, see Figure 3a, green squares).
Finally, rotation of the aromatic blocking group C6H3R′2,
about the N-C axis (green path d in Scheme 2) represents a
fluxional process that could potentially affect the accessibility
of the piperidine lone pair in E-1. The fact that this process is
rapid at ambient temperature, at least on the NMR time scale,
is already suggested by the equivalence of the two R′ moieties
in the NMR spectra. As the signals could be isochronous19 by
coincidence, detailed analysis of our collected RDC data for
E-1b was carried out, and conformational flexibility became
apparent. Observed RDC values are usually smaller for flexible
parts of the compound (due to averaging), and RDCs do not fit
any of the structural proposals,20 as observed when fitting the
complete set of RDCs to the DFT-structure of E-1b (q ) 0.215,
1
see Figure 3a, red dots). Furthermore, the calculated DC-H’s
for the two ortho protons of the steric blocker in the case of a
Figure 3. Comparison of RDCs calculated (Dcalc) for the different possible
conformational states of E-1b with the ones observed experimentally (Dexp).
In (a) the fit for the equilibrium structure E-1b (blue diamonds, without
using the data from the steric blocker) is depicted together with the ring-
flipped feq-1b (green squares). The red dots show the fit for all data
(including the RDCs of the steric blocker) to E-1b, the black triangles to
the transition structure for the rotation of the steric blocker 1b-TSrot. In
(b) the fit for the equilibrium structure E-1b (blue diamonds, without using
the data from the steric blocker, the C-N bond length reduced to 10% of
its original length, see text) is compared to the N-inverted i-1b (red squares).
rigid structure (-13.1 and -2.6 Hz) deviate largely from the
1
measured averaged DC-H value (-6.1 Hz), confirming fast
rotation on the NMR time scale.7,21
In accordance with these observations, only modest barriers
ranging between 6-8 kcal/mol (Table 1) were computed for
this rotation involving 1-TSrot (green path d in Scheme 2).
Interestingly, the lowest barrier was computed for the largest
substituent. Indeed, visual inspection of the E-1c ground state
shows that no serious clash between the R and R′ groups is to
be expected upon rigid rotation of the blocking group. Despite
the comparably low hindrance toward blocking-group rotation
in E-1a, E-1b, and E-1c, the bulky R′ groups in the latter appear
to be more effective in shielding the lone pair than those in the
former (see discussion of reactivities below).
value of -248.8 Hz predicted for i-1b.15 Therefore, from
analysis of our RDC data we can unequivocally conclude that
no significant population of the N-inverted conformer i-1b could
be detected by NMR spectroscopy.
Inversion of the six-membered piperidine chair, i.e. ring “flip”
(red path c in Scheme 2), represents another possible mechanism
to render the axial lone pair in E-1 accessible to the environment,
thereby constituting a potential source of residual catalytic
activity of the formal OFF-state. This ring flip is expected to
proceed via one or more twist-boat intermediates to the ring-
inverted form fax-1, which via N-inversion can subsequently
rearrange to isomer feq-1 with equatorial R and an exposed
N-lone pair. After conformational analysis of the parent N-
methyl piperidine at the B3LYP/6-31G* level,7 selected station-
ary points along this path were located for 1a and 1b, focusing
on the final product and a representative transition state leading
to a twist-boat (or twist) isomer. As expected, the ring-flipped
isomers feq-1a,b are strongly disfavored, by ∼7-9 kcal/mol,
and are accessible only via significant barriers, up to ∼12-13
kcal/mol (Table 1).17 Incidentally, these numbers are close to
the corresponding barrier for cyclohexane.18 Overall, there is
no indication from the computations that ring-flipped isomers
could be populated to any significant extent. Very similar results
are to be expected for the bulky derivative 1c, which was not
further studied computationally.
Photochromism. The photochromic behavior of photoswit-
chable piperidine bases 1a-c and 2 was investigated in detail
(Table 2). Azobenzene-based catalysts 1a-c showed the
expected isomerization behavior upon irradiation of the ther-
mally stable E isomers. As evident from UV/vis spectra,
irradiation with light of 365 nm wavelength leads to a decrease
in intensity of the absorption band at about 330 nm, which is
(17) There are many more possible twist or twist-boat intermediates and
transition states, depending on which C-C or C-N bond in the ring
is rotated relative to its opposite counterpart and in what direction.
We have refrained from locating all of them, but from the present
results it is already clear that the overall barrier must be in the interval
between ∼7-13 kcal/mol, implying notable activation for this process.
(18) For example: ∆E ) 11.4 kcal/mol at B3LYP/6-31G* for the
C2-symmetric twist-chair transition state (or half-chair), cf. structure
10 in the following reference, where a value of 11.9 kcal/mol has
been obtained at MP2/6-31G* Leong, M. K.; Mastrryukov, V. S.;
Boggs, J. E. J. Phys. Chem. 1994, 98, 6961–6966.
(19) Leutha¨usser, S.; Schmidts, V.; Thiele, C. M.; Plenio, H. Chem. Eur.
J. 2008, 14, 5465–5481.
(20) Another indication is given by the fact that we did not observe
differentiation of enantiotopic faces, which is to be expected in a chiral
orienting medium if no rotation of the steric blocker (converting one
enantiotopic face into the other) is occurring.
Compounds E-1a and E-1b were investigated experimentally,
1
and inspection of their H NMR spectra shows that both pairs
(21) If the transition-state structure of this rotation, 1-TSrot, is used, the
quality of the fit is further reduced (q ) 0.426, Figure 3a, black
triangles).
of diastereotopic protons of the piperidine methylene groups
are anisochronous, indicative for locked chair conformations.
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 361

A R T I C L E S
Stoll et al.
Table 2. Photochromic Properties and Activation Parameters for Thermal Z f E isomerization of 1a-c and 2
absorption
PSS composition
thermal Z f E isomerization
d
d
λ
max(E)a [nm]
λ
max(Z)b [nm]
EfZ (Z/E)
Z fE (Z/E)
τ1/2c [h]
268
286
466
-
∆Gq
d [kcal mol-1
]
∆Hq [kcal mol-1
]
∆Sq [cal mol-1 K-1
]
298 K
1a
1b
1c
2
335
335
333
298
444
444
440
279
90:10
90:10
95:5e
93:7
15:85
15:85
10:90
78:22
26
25
25
-
23
22
23
-
-9
-11
-8
-
a π,π*-Absorption. b n,π*-Absorption. c At 20 °C (293.15 K). d Obtained from kinetic data of the thermal Z f E isomerization at various
temperatures and Eyring analysis of the data. e No E isomer was detected by UPLC analysis.7
different temperatures using UV/vis spectroscopy (Table 2).23
All activation parameters obtained are in reasonable agreement
with parameters determined for other azobenzene derivatives.24
Thermal isomerization was also investigated computationally.
For 1a, the only E/Z-transition state that could be located is
one that is associated with inversion at the diazo-N-atom bound
to the spiro-annulated core (1a-TSEZ, see black path a in
Scheme 2).25 That transition state was obtained even when the
search started from a structure that would correspond to a
rotation about the NdN axis. This finding is in line with earlier
DFT results for azobenzene, according to which inversion has
a lower barrier than rotation.26 For the other derivatives, similar
inversion transition states 1b,c-TSEZ were found. In all of these,
the 3,5-disubsituted benzene ring is oriented almost perpen-
dicular to the plane of the spiro-annulated benzene ring having
a C-NdN-C torsional angle around 100° and NdN-C-C
Figure 4. Absorbance spectra of 1a-c in acetonitrile taken at varying
torsional angles close to 0° and 180°, respectively. Activation
intervals during irradiation with light of 365 nm wavelength (1a: 2.8 ×
parameters for Z f E isomerization in the gas phase, calculated
as the free energy of 1-TSEZ relative to that of Z-1 (Table 1),
10-5 M, 12:48 min irradiation time; 1b: 3.1 × 10-5 M, 12:46 min irradiation
time; 1c: 3.0 × 10-5 M, 14:00 min irradiation time). Spectra are normalized
to the maximum n,π-absorption around 335 nm.
are 22.8, 21.8, and 23.0 kcal/mol for 1a, 1b, and 1c, respectively.
These numbers are slightly lower than the experimental
estimates for ∆Gq in Table 2 and show no particular trend.
Enthalpies of activations ∆Hq are fairly equal when comparing
theoretical27 and experimental values, indicating a comparable
enthalpic loss when going from the strained Z isomer to the
twisted transition state. Activation entropies ∆Sq for the gas-
phase reactions are small in all cases, suggesting that contribu-
tions to ∆Sq in the condensed phase are mostly associated with
solvation effects. From the negative values found experimentally
for ∆Sq, it can be deduced that the degree of order of solvated
Z-1a-c increases by going to its respective transition state.
With regard to the prolonged half-life of Z-1c, it is reasonable
to assume a larger entropic contribution resulting in a decrease
of ∆Gq (∆Gq ) ∆Hq - T∆Sq) compared to 1a,b, because the
larger, more hydrophobic blocking group is more weakly
solvated therefore requiring a smaller amount of reorganization
in the solvation sphere. The small increase in half-life observed
experimentally for 1a and 1b is most likely due to a similar
attributed to the π,π*-transition of the azobenzene chromophor,
along with an increase of the band around 450 nm, which is
attributed to the n,π*-absorption (Figure 4).1d,g Irradiation of
larger quantities of catalysts 1a-c in acetonitrile solution
allowed for isolation of samples containing mainly the Z
isomers. X-ray crystal structure analysis of Z-1b shows a
pronounced deviation from planarity of the azobenzene core,
leading to significant electronic decoupling of both phenyl rings
in the Z isomer (Figure 2) and the resulting hypsochromic shift
of the π,π*-absorption constitutes the basis of photochromism
in azobenzenes. Photostationary states obtained by irradiation
of analytical samples of 1a-c show a remarkably high content
of Z isomer, amounting to at least 90% according to UPLC
analysis. The Z isomer can be converted to the E isomer
photochemically by irradiation with visible light (λ g 400 nm).
However, the extent of switching is smaller due to the non-
negligible absorption of the E isomer above 400 nm and yields
photostationary-state mixtures containing up to 90% E isomer
in the case of 1c and around 85% for 1a and 1b.
As expected for azobenzenes, catalysts 1a-c thermally revert
from the Z to the E isomer. Thermal half-lives τ1/2 observed for
Z-1a-c were found to be extraordinarily extended as compared
to the parent azobenzene (τ1/2 ) 16 h in benzene).22 Thermal
isomerization of Z-1a and Z-1b was observed to occur with half-
lives of 268 and 286 h at 20 °C, respectively. Change of the
steric demand of the blocking group by replacing the 3,5-di-
tert-butylphenyl fragment with the 3,5-bis(2,6-dimethylphe-
nyl)phenyl fragment further prolonged half-life to 466 h at 20
°C for Z-1c. Activation parameters for the thermal isomerization
of Z-1a-c were determined by measuring the rate constants at
(23) Please note, half lives reported in Table 2 were directly determined
from the rate constants at the given temperature. Estimated errors for
activation parameters are in the range of (1 kcal mol-1 for ∆Hq and
(1 cal mol-1 K-1 for ∆Sq. Rate constants and half lives, which can
be calculated from ∆Hq and ∆Sq by the thermodynamic formulation
of the Eyring equation, are extremely sensitive to these quantities.
(24) (a) Gegiou, D.; Muszkat, K. A.; Fischer, E. J. Am. Chem. Soc. 1968,
90, 3907–3918. (b) Le Fe`vre, R. J. W.; Northcott, J. J. Chem. Soc.
1953, 867–870.
(25) That this transition state indeed connects Z-1a and E-1a was verified
by reoptimization at B3LYP/3-21G and following the intrinsic reaction
coordinate (see: (a) Gonzalez, C.; Schlegel, H. B. J. Chem. Phys. 1989,
90, 2154–2161. (b) Gonzalez, C.; Schlegel, H. B. J. Phys. Chem. 1990,
94, 55235527) at that level.
(26) Crecca, C. R.; Roitberg, A. E. J. Phys. Chem. A 2006, 110, 8188–
8203.
(27) The corresponding DFT values for ∆Hq (not included in Table 1) are
23.0, 22.9, and 23.6 kcal/mol for 1a, 1b, and 1c, respectively.
(22) Schulte-Fro¨hlinde, D. Liebigs Ann. Chem. 1958, 612, 131–138.
9
362 J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009

Photoswitchable Catalysts
A R T I C L E S
Table 3. Basicities and Rate Constants of the Henry Reaction Obtained for Catalysts 1a-c and 2.
basicities
Henry reaction
e
f
g
pKa(OFF)a
pKa(ON)b
-
16.7
16.7
-
∆pKac
-
0.8
0.7
-
PSSd(Z/E)
80:20
90:10
90:10
93:7
kOFF [10-6 s-1
]
kobs [10-6 s-1
]
kON [10-6 s-1
]
krel (kON/kOFF
4.3
)
1a
1b
1c
2
-
4.96
0.963
0.391
12.9
15.0
8.87
10.6
-
21.5
12.7
13.9
25.4
15.9
16.0
-
13.2
35.5
2.0
a pKa-value of the E isomer representing the OFF-state. b pKa-value of the photostationary state mixture representing the ON-state. c Difference of
pKa-values, i.e. pKa(PSS) - pKa(E). d Photostationary state obtained by preparative irradiations at λ ) 365 nm. e Rate constant of Henry reaction using
pure E isomer. f Rate constant of Henry reaction using the photostationary state mixture obtained by preparative irradiations at λ ) 365 nm. g Rate
constant of Henry reaction extrapolated to 100% Z isomer.
effect associated with the increase in hydrophobicity when
exchanging the methyl substituent for the tert-butyl group. The
effect is smaller, because the site of substituent exchange is
more remote from the azo fragment causing differences in
solvation having a smaller influence on the transition state of
E f Z isomerization. In fact, experimentally determined values
for ∆Sq increase by about 3 cal mol-1 K-1 going from 1a to 1c.
However, this can only be seen as a trend and experimental
uncertainties preclude any accurate quantification.
To avoid potential problems originating from the thermal
instability of the Z isomer of the azobenzene-based catalysts
1a-c, stilbene-based catalyst 2 was investigated. In contrast to
azobenzene and its derivatives, stilbenes exhibit a photochromic
Figure 5. Catalytic performance of catalyst 1c in the Henry reaction
behavior with both switching states being thermally stable
calculated from product concentrations determined from 1H NMR spec-
(photochromism of the P type).28 As expected, irradiation of 2
troscopy (red 2: E-1c, blue 2: PSS:Z-1c:E-1c ) 90:10, purple 4: PSS with
thermal correction for Z f E isomerization, green 3: extrapolation to 100%
led to conversion of the E isomer to the Z isomer accompanied
by a decrease of the absorption band around 300 nm and a small
increase of absorption around 250 nm.7 Remarkably, the
photostationary state mixture obtained by irradiation of analytical
samples in acetonitrile with light of wavelength greater than
330 nm contained almost exclusively Z isomer, i.e. 97% as
determined by HPLC. As expected, thermal reversion of Z-2 to
E-2 was not observed. Irradiation of the photostationary state
mixture with light of 250 nm wavelength induced Z f E
isomerization, yet only 22% E isomer could be recovered due
to the marked absorption of the E isomer at 250 nm. No spectral
region can be identified that would allow for satisfactory Z f
E photoisomerization, since the E isomer’s absorption is more
significant throughout the entire wavelength range as compared
to the Z isomer, precluding the use of 2 as a practical
photoswitch. Implications of the small extent of Z f E switching
on the catalytic performance of the stilbene catalyst 2 are
discussed in the next section.
Z-1c, black 9: no catalyst, orange b: E-azobenzene).
and protonated forms, necessary for a direct evaluation of the
pKa-values. However, indirect determination of these concentra-
tions was possible by using Neutral Red29 (3-amino-7-dimethy-
lamino-2-methylphenazine) as the reference base, having a pKa
of 15.6 for the protonated form in acetonitrile.7 The choice of
the reference base is crucial, since a too large difference in
basicity of catalyst and reference base precludes meaningful
measurements.30 Indeed, pKa-values of both isomers of 1b differ
significantly (Table 3). The basicity of Z-1b (ON-state) increased
by almost 1 order of magnitude amounting to pKa ) 15.9 as
compared to pKa ) 16.7 for the corresponding E-1b (OFF-state).
For E-1c and Z-1c pKa-values of 16.0 and 16.7, respectively,
were found and are essentially identical to those of the respective
1b analogues, when considering the uncertainty of these
measurements of approximately (0.1 pKa units.
General Base Catalysis. Reactivity differences associated with
both switching states of compounds 1a-c as well as 2 could
be exploited to photocontrol the conversion in a general base-
catalyzed reaction. For this purpose, the nitroaldol reaction,
commonly referred to as Henry reaction,31 of nitroethane with
4-nitrobenzaldehyde was investigated as a model reaction
(Figure 5) due to its low background rate in the absence of base
catalyst. Using an excess of the nitroethane substrate to ensure
Reactivity.
Basicity. The photochromic behavior of catalysts 1a-c is
associated with pronounced structural changes, which translate
into differences in chemical reactivity of both isomers. In context
with the desired general base catalysis, the basicity of the amines
and the extent to which it can be altered between the ON- and
OFF-states are of great interest. A change in basicity upon
isomerization could be expected due to the different stabilities
of the corresponding acid-base adducts involved. However, it
is important to note that the realization of a photoswitchable
Brønsted base constitutes a formidable task since the proton
represents the smallest possible electrophile and precludes strong
steric interactions.
To compare the basicities of different switching states of
compounds 1b,c, titration experiments with trifluoromethane-
sulfonic acid were carried out. Negligible changes in the
absorption spectra of 1b and 1c upon protonation precluded a
direct determination of the concentrations of the nonprotonated
(28) Bouas-Laurent, H.; Du¨rr, H. Pure Appl. Chem. 2001, 73, 639–665.
(29) (a) Leito, I.; Kaljurand, I.; Koppel, I. A.; Yagupolskii, L. M.; Vlasov,
V. M. J. Org. Chem. 1998, 63, 7868–7874. (b) Kaljurand, I.; Rodima,
T.; Leito, I.; Koppel, I. A.; Schwesinger, R. J. Org. Chem. 2000, 65,
6202–6208. (c) Biondic, M. C.; Erra-Balsells, R. J. Chem. Soc. Perkin
Trans. 1997, 7, 1323–1327.
(30) For optimal results, the pKa of the reference base should not differ
more than 2 pKa units from the pKa of the base to be measured.
(31) (a) Henry, L. C. R. Hebd. Seances Acad. Sci. 1895, 120, 1265–1268.
(b) For a recent review, consult: Luzzio, F. A Tetrahedron 2001, 57,
915–945.
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 363

A R T I C L E S
Stoll et al.
pseudo-first-order kinetics, the buildup of syn- and anti-aldol
which estimate a significant destabilization of i-1b over i-1a
(Table 1). This leads to a roughly 200-fold decrease in
population of the N-inverted isomer of catalyst 1b, correlating
with a significantly diminished residual activity, as compared
to catalyst 1a. In addition, the kinetic stability of i-1b is
decreased due to the lowering of the N-inversion barrier as
compared to that of i-1a (Table 1). Not unexpectedly, flipping
of the piperidine chair is even more unlikely for 1b. Investigation
of the catalytic activity of catalyst 1b in its ON- and OFF-state
under identical conditions as used for 1a revealed a significant
improvement of the ON/OFF-ratio, reaching a value of
krel ) 13.2, corresponding to a reactivity difference of both
switching states exceeding 1 order of magnitude (Table 3). The
enhanced reactivity difference mainly originates from the
pronounced decrease of the OFF-state’s activity overcompensat-
ing the slightly reduced reactivity of the ON-state.34
Assuming the reactivity of Z-1b to be the maximum intrinsic
reactivity of a N-tert-butyl-substituted piperidine, the catalyst’s
ON/OFF-ratio can only be improved by further lowering the
reactivity of the OFF-state. This could be realized by improving
the efficiency of blocking, provided by the azobenzene shield.
Inspection of models of E-1b as well as 1b-TSrot, obtained from
DFT calculation, led to the initial conclusion that a twist of the
blocking group around the C-N bond leads to a less shielded
E isomer, potentially causing some residual reactivity. To
address this problem, the R′ substituents of the blocking group
responsible for shielding were increased in size, i.e. from tert-
butyl (1b) to 2,6-dimethylphenyl (1c), taking advantage of our
modular synthetic route (Scheme 1). Indeed, no rate acceleration
was observed by using the catalyst’s OFF-state E-1c as
compared to the noncatalyzed reaction, while the reactivity of
the ON-state Z-1c remained practically unaffected. Both effects
led to an increased overall ON/OFF-ratio krel ) 35.5.
It is interesting to note, that according to DFT calculations
N-inversion of 1c is energetically more favorable than inversion
of 1b, indicated by a decrease of the relative free enthalpy of
the inverted isomers from 7.1 to 5.4 kcal/mol on going from
1b to 1c in the gas phase.35 Obviously, the reactivities of the
OFF-states cannot simply be explained by differences in
populating the N-inverted conformers i-1 and show that ad-
ditional effects have to be taken into account. In order to address
this rather complicated issue in more detail, we investigated
the proton affinities (PAs) of the bases 1a-c by computational
methods. Representative calculated protonation energies and
enthalpies are summarized in Table 4. Noting that the trends of
the protonation energies in the gas phase are comparable for
the two basis sets employed (compare ∆E/6-31G* and ∆E/cc-
pVTZ values), we will discuss primarily the PAs obtained with
the lower basis set. The latter are defined as the negative
enthalpies of protonation of the free bases, i.e. the absolute value
of the ∆H/6-31G* values of the protonated species in Table 4.
PAs of the global minimum conformation show a constant
increase with increasing steric demand of the molecule on going
from 1a to 1c. While the PA increases strongly on going from
1a to 1b, the difference between 1b and 1c is small, in line
with the similar substitution at the piperidine nitrogen atom.
Isomerization of the double bond leads to an increase of the
products in the presence of 10 mol % of catalyst in [D8]THF
1
solution was monitored by H NMR spectroscopy. Reactivity
originating from the much less basic azochromophore could be
ruled out by a control experiment using azobenzene itself as
catalyst, leading to no acceleration of product formation (Figure
5). The catalysts employed in the kinetic experiments were used
in their isolated form, i.e. the E isomer was used in its pure
form, while the Z isomer was used as the photostationary-state
mixture containing minor amounts of residual E isomer.32 For
comparison of the reactivity associated with the two switching
states of a catalyst, krel was defined as the ratio of rate constants
for use of either pure Z or pure E isomer, that is krel ) kON/kOFF
.
The rate constant for use of the pure Z isomer is not directly
accessible by kinetic measurements since the Z isomer can only
be obtained as a mixture containing residual E isomer represent-
ing the photostationary state and thermal reversion to the
thermodynamically more stable E isomer starts to contribute
after prolonged reaction times. Therefore, initial kinetic data
were corrected for reactivity attributed to residual E isomer and
for thermal reversion of Z isomer to E isomer (Figure 5, Table
3).
In all cases, use of the reactive Z isomer of catalysts 1a-c
as well as 2 led to a rate enhancement as compared to the use
of E isomer (Table 3). Catalyst 1a displayed only a moderate
difference in activity between its ON- and OFF-state, resulting
in a relatively small value of krel ) 4.3. It is evident from the
kinetic data that use of the E isomer led to some residual activity
not desirable for the OFF-state of a switchable catalyst. Most
likely, the E isomer’s reactivity originates from deprotonation
of the nitroalkane, pointing to unwanted accessibility of the basic
piperidine lone pair. Three pathways can be considered for the
deprotonation of the nitroalkane by the catalyst’s OFF-state:
(1) inversion of the piperidine N-atom placing the lone pair in
an equatorial position (i-1a-c), (2) chair flip leading to
piperidine conformations with the basic lone pair in nonshielded
equatorial or axial positions (feq-1a-c or fax-1a-c), and (3) a
rotation of the blocking group allowing the nitroalkane to be
deprotonated by the piperidine’s most stable (chair) conforma-
tion with the lone pair in axial position. On the basis of
computational and NMR studies (Vide supra) the chair flip can
be ruled out for catalysts 1a-c. However, inversion of the
nitrogen atom and/or rotation of the blocking group could in
principle be responsible for residual catalytic activity of the
OFF-state E-1a (Table 1).
In order to suppress N-inversion, the methyl group R was
replaced with a more demanding tert-butyl group, known to
act as an efficient conformational anchor for six-membered
rings.33 Indeed, N-inversion of the E isomer of catalyst 1b could
be ruled out by NMR spectroscopy as well as DFT calculations,
(32) Note that the higher concentration used in preparative irradiation
experiments increased the absorbance of the solution such that usually
even after prolonged irradiation times the photostationary-state
mixtures contained a larger amount of E isomer as compared to
irradiating analytical samples of much lower concentration. For these
reasons, the conditions necessary to accurately determine the kinetics,
i.e. the rather high catalyst concentrations employed in the NMR
experiments, precluded switching of the catalysts in situ. However,
this does only constitute a drawback in bulk or solution-phase catalysis
associated with high optical densitiessa drawback that photochemistry
suffers from in generalsyet an application utilizing mono-or multi-
layers of immobilized catalysts should be feasible.
(34) The reactivity of the Z isomer Z-1b was also lowered, most likely
because the steric demand of the tert-butyl group hinders the
nitroalkane to approach the basic catalyst site.
(33) For a discussion of the conformational behavior of piperidines, see:
Eliel, E. L.; Wilen, S. H. Stereochemistry of Organic Compounds;
Wiley: New York, 1994, pp 740 and references therein.
(35) At the B3LYP/cc-pVTZ/PCM(acetonitrile) level, the relative energies
of i-1a, i-1b, and i-1c are 3.9, 7.1, and 5.8 kcal/mol, respectively, i.e.
quite similar to the gas-phase ∆E values in Table 4.
9
364 J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009

Photoswitchable Catalysts
A R T I C L E S
Table 4. Relative Energies ∆E and Enthalpies ∆H Obtained at
Selected Levels for Key Isomers of 1a-c (with Respect to the
Global Minimum, E-1), and Energies and Enthalpies of Protonation
(Relative to the Respective Minima E-1, Z-1, i-1)a
be a stronger base than E-1a, whereas the opposite is found for
i-1b,c. Steric reasons are likely to be responsible for this result,
because the repulsion between the R and R′ groups is so strong
in i-1b,c that the amine nitrogen atom is significantly planarized,
i.e. the angle sum at N amounts to 352° in both cases. When
going from i-1a (angle sum of 339°) to i-1b,c, the s-character
of the nitrogen lone pair and hence its basicity are decreased.
In the polar continuum, this effect is overcompensated by the
better stabilization of the i-1H+ isomers with their solvent-
accessible cationic NH centers, and all i-1 forms are computed
to be stronger bases than their E-1 counterparts.37
∆E
cc-pVTZ
PCM(THF)c
∆E
cc-pVTZ
PCM(MeCN)d
∆E
6-31G*
∆H
∆E
cc-pVTZ
(-PA)b6-31G*
Z-1a
15.8
3.8
-244.1
-245.5
-245.1
15.2
3.8
-234.5
-235.6
-236.1
15.5
4.4
-242.4
-243.6
-243.8
13.7
4.0
-276.0
-281.8
-282.5
13.3
3.9
-280.9
-287.7
-288.4
i-1a
E-1aH+
Z-1aH+
i-1aH+
To summarize this part, the higher basicity of the inverted
isomer of 1a might account to some extent for the significant
reactivity of the OFF-state, since a thermal population of the
N-inverted isomer is possible. Even if the population of i-1a is
small (at 298 K, a relative free energy of 4 kcal/mol would
correspond to 0.1% in an equilibrium mixture), its contribution
to the overall catalytic turnover would be noticeable because
of the expected enhanced reactivity. Freezing the N-inversion
by incorporation of the tert-butyl group lowers the population
of the N-inverted isomers of 1b and 1c even more, thus strongly
reducing their potential contribution to the residual activity. This
interpretation is consistent with the experimental results,
provided the relative reactivities reflect the respective thermo-
dynamic driving forces (for a discussion of possible kinetic
effects see below). By comparing catalysts 1b and 1c, it is
evident that no further improvement of the activity ratio krel
seems possible for the particular system under investigation.
As mentioned, the reactivity of Z-1b and Z-1c represents the
intrinsic reactivity of a N-tert-butyl-substituted piperidine
incorporated into the catalysts framework toward nitroethane,
leaving no room for further improvement since a greater extent
of deshielding is not possible. The lower limit of reactivity is
given by E-1c displaying a reactivity not distinguishable from
the reaction’s background reactivity.
Since slow thermal reversion of the Z isomers of catalysts
1a-c leads to unwanted catalyst deactivation upon prolonged
reaction times, stilbene-based catalyst 2 was devised to suppress
thermal Z f E isomerization. Comparison of DFT-optimized
structures of azobenzene-based catalysts E-1a and E-2 reveals
interesting differences.7 On going from the azobenzene to the
stilbene, the elongation of the XdX distance and the opening
of the XdX-C angle (115° and 126° for X ) N and C,
respectively) strongly increases the separation between the amine
N-atom and the tert-butyl substituent of the blocking group,
and is expected to lead to less efficient shielding and a higher
reactivity of the E isomer. This was confirmed experimentally,
observing a 2.6 times higher reactivity for E-2 as compared to
E-1a under otherwise identical conditions. Switching stilbene
2 to its corresponding Z isomer only moderately accelerated
the rate of the reaction, i.e. krel ) 2. In addition to this low
ON/OFF-ratio the applicability of stilbene 2 as a photoswitch-
able catalyst is further narrowed by the inefficient photochemical
Z f E isomerization (Vide supra).
Z-1b
15.4
6.0
-250.5
-251.7
-250.0
15.2
5.8
-241.7
-242.3
-240.3
15.0
6.3
-248.5
-249.5
-248.4
13.6
5.9
-278.3
-281.3
-282.5
13.2
5.8
-282.8
-286.5
-288.0
i-1b
E-1bH+
Z-1bH+
i-1bH+
Z-1c
15.3
5.5
-252.0
-251.0
-250.6
15.0
5.4
-242.4
-241.5
-240.8
15.2
6.0
-249.6
-248.9
-249.1
14.0
5.9
-278.9
-281.2
-282.6
13.5
5.8
-283.3
-286.9
-287.7
i-1c
E-1cH+
Z-1cH+
i-1cH+
a In kcal/mol, employing the B3LYP functional and B3LYP/6-31G*
optimized geometries. b Negative PA for the protonated species (298.15
K, 1 atm). c Continuum model employing the parameters of THF.
d Continuum model employing the parameters of acetonitrile.
PAs of 1a and 1b by 1.1 and 0.6 kcal/mol, respectively, whereas
that of 1c decreases by 0.9 kcal/mol upon E f Z isomerization.
As the latter result cannot easily be rationalized on steric grounds
alone, a potential explanation involves ”internal solvation” of
the protonated nitrogen center by the proximal R′ groups of
the blocking moiety, which in the case of E-1cH+ is somewhat
more efficient due to a stabilizing interaction with the aromatic
ring.36 Upon solvation (modeled by a simple polarizable
continuum), protonation of the Z isomers becomes much more
favorable than that of the corresponding E isomers. Clearly,
the protonated nitrogen center bearing the charge is more
exposed to the solvent (or the continuum) in Z-1H+, increasing
their stabilization by favorable interactions with the solvent. This
computed increase in basicity of Z vs E isomers is in qualitatve
agreement with the experimental pKa measurements discussed
above, where both 1b and 1c show an increase in pKa upon
E f Z isomerization (Table 3). Quantitatively, this increase in
basicity is significantly overestimated by the DFT/PCM calcula-
tions for acetonitrile as the computed differences in protonation
energies are on the order of -4 kcal/mol (e.g., from -282.8 to
-286.5 kcal/mol for 1b in Table 4), which would translate into
changes by several pKa units. Evidently, the continuum solvation
model is too simplistic and specific interactions with the solvent
(or its dynamics) would have to be taken into account for a
better quantitative accuracy.
When assessing the PAs of the N-inverted isomers, we
assume that protonation equilibria are fast relative to N-inversion
and discuss energetic differences between i-1 and i-1H+ (i.e.,
N-protonated i-1). Although N-inversion is so rapid that
practically only equilibria involving the global minima E-1
would be observable, the question of the inherent basicity of
i-1 is of interest in the context of catalysis, as the more reactive
species (even when present in small amounts) might constitute
the more potent catalysts. In the gas phase, i-1a is indicated to
In their OFF-state, catalysts 1b and 1c display a distinct
difference in reactivity in the Henry reaction, even though they
(37) For i-1a, this stabilization is predicted to be so strong that the
protonated inverted isomer i-1aH+ is even more stable than E-1aH+.
This particular result should be taken with care in light of the
performance of the continuum model for predicting the basicities of
Z vs E isomers (Vide supra). However, the qualitative finding that the
N-inverted isomers i-1 are more basic than their more stable chair
conformers E-1 forms should be reliable.
(36) (a) Hunter, C. A.; Lawson, K. R.; Perkins, J.; Urch, C. J. J. Chem.
Soc., Perkin Trans. 2 2001, 651–669. (b) Meyer, E. A.; Castellano,
R. K.; Diederich, F. Angew. Chem., Int. Ed. 2003, 42, 1210–1250.
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 365

A R T I C L E S
Stoll et al.
have almost equal basicities in acetonitrile. This observation
suggests that under catalytic turnover, dictated by the kinetics
of the rate-limiting step, the reaction barriers are influenced by
other factors in addition to those solely grounded on thermo-
dynamic driving forces for substrate deprotonation. The rate-
determining step in the general-base catalyzed Henry reaction
is the deprotonation of the nitroalkane by the amine base. The
conformational flexibility of the systems under scrutiny as well
as the need to properly account for solvation effects during the
concomitant charge separation make a computational search for
the relevant transition states for deprotonation very involved.38
As a first step, representative adducts formed from E-1a-c and
the simplest nitroalkane, nitromethane, which are likely to be
involved in the early stages of deprotonation, were located
(Figure 6). The raw nitromethane-complexation energies, uncor-
rected for zero-point energies and basis-set superposition error,
are -4.0, -3.1, and -2.4 kcal/mol for 1a, 1b, and 1c,
respectively. As can be seen in the optimized structures in Figure
6, the blocking group has to rotate significantly about the C-N
bond in order to allow coordination of the nitromethane to the
lone pair. For a more quantitative analysis, the energies required
to distort the free bases from their conformations in the fully
optimized global minima to those in the nitromethane complexes
were evaluated. These energies should be a good indication of
the kinetic hindrance imposed by the substituents R and R′
during the deprotonation step and amount to 1.4, 2.3, and 3.6
kcal/mol for 1a, 1b, and 1c, respectively. These data are in very
good qualitative agreement with the experimentally observed
reactivity trend for the OFF-states of 1a-c. The rather low
energetic “penalty” for coordination of the nitroalkane to E-1a
is noteworthy, as it implies that a substantial part of the residual
activity of 1a in its OFF-state is in fact due to the global E
minimum itself, in addition to the part that may arise from traces
of the N-inverted isomer (Vide supra). Even though it is beyond
the theoretical methods presently available to relate these rather
modest changes in the energetics quantitatively to specific rate
constants, the overall qualitative agreement between theory and
experiment provides a solid basis for our interpretation of the
observations.
In the general context of comparing the competing reactivities
of an equilibrating mixture of conformers, we most likely deal
with a scenario in which the activation barriers for intercon-
verting the participating conformers, i.e. E-1, i-1, fax-1, and feq-
1, are comparable to the activation barriers for proton abstraction
from a nitroalkane.38 For this reason, the simplest case of the
reactivity schemes, developed by Winstein and Holness on the
basis of kinetic observations as well as Curtin and Hammett on
the basis of product distributions,39 probably does not apply
and our system would be more appropriately described by the
general treatment given by Seeman and Farone.40
Figure 6. Adducts between nitromethane and E-1a-c illustrating the degree
of twisting of the aromatic blocker group (left: front view along the N-R
axis, right: side view; B3LYP/6-31G*optimized).
synthesized and differences in basicity of the switching states
involved were successfully exploited to photocontrol the
catalysts’ activities in general base catalysis, exemplified by
photocontrolling conversion of the Henry reaction. A careful
analysis of the dynamic structural behavior in solution using
the recently introduced residual dipolar couplings enabled us
to establish the structure-reactivity relation for the catalyst
system, thereby facilitating tuning of the catalysts to enhance
ON/OFF-ratios. Current work in our laboratory is directed
toward surface tethering of the catalyst, potentially offering
unprecedented applications in surface patterning and function-
alization. Extension of our concept to intrinsically more reactive
catalysts with better ON/OFF-ratios, potentially allowing for
photocontrol of chemical selectivities and polymerization
processes, constitutes a challenging target for future research.
With such “smart” catalysts in hand, new responsive and
adaptive materials can be envisioned.
Conclusion and Outlook
In conclusion, we have developed a new conceptual design
for photoswitchable catalyst systems. Photoswitchable catalysts
on the basis of conformationally locked piperidine bases were
(38) In a recent B3LYP study of the Henry reaction catalyzed by a cinchona
alkaloid, a barrier of 10.8 kcal/mol for deprotonation of nitromethane
has been computed in a THF continuum: Hammar, P.; Marcelli, T.;
Hiemstra, H.; Himo, F. AdV. Synth. Catal. 2007, 349, 2537–2548.
(39) For comprehensive reviews on the effect of conformational changes
on reactivity in organic chemistry, see: (a) Seeman, J. I. Chem. ReV.
1983, 83, 83–134. (b) Reference 33, pp 647-655.
(40) Seeman, J. I.; Farone, W. A. J. Org. Chem. 1978, 43, 1854–1864.
9
366 J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009

Photoswitchable Catalysts
A R T I C L E S
Acknowledgment. We thank Petra Neubauer and Dr. Burkhard
Ziemer (HU Berlin) for carrying out the single crystal X-ray structure
analyses. Generous support by the German Research Foundation (DFG
via SPP 1179 and Emmy Noether Program TH1115/3-1) and the Fonds
der Chemischen Industrie is gratefully acknowledged. M.B. thanks
EaStChem for support. RSS is indebted to the Studienstiftung der
deutschen Volkes for providing a doctoral fellowship. Wacker Chemie
AG, BASF AG, Bayer Industry Services, and Sasol Germany are
thanked for generous donations of chemicals.
Supporting Information Available: Synthetic and computa-
tional details, compound characterization data, NMR experi-
ments and data analysis including measured RDCs and fitting
details as well as optimized Cartesian coordinates of all
structures and selected three-dimensional plots thereof. This
material is available free of charge via the Internet at http://
pubs.acs.org.
JA807694S
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 367