A supramolecular approach to controlling the behavior of excited states

Introduction

Excited states are known to be sensitive to their environment [1, 2], a fact that is most often used in the design of emissive sensors. However, the chemical reactivity of excited states can also be controlled through interactions that reach out beyond the photoactive site. These include local solvation effects as well as molecular properties resulting from supramolecular interactions. For those applications relying on charge transport, the intermolecular orbital overlap is a critical component in determining the charge-carrier mobility of a material which depends on the molecular components and their mutual orientation over long distances [3]. Therefore, controlling the morphology of the active layer down to the molecular level is crucial for bridging the bottom-up molecular and top-down device fabrication scales.

The self-assembly of small molecules into larger, well-defined architectures can be achieved through the use of one or more designed supramolecular interactions. For example, metal ion coordination, π-stacking and hydrophobic interactions, electrostatic forces, and hydrogen-bonding (H-B) interactions have been used with varying degrees of success [2]. Hydrogen-bonding interactions are particularly interesting as they do not generally lead to binding sites that absorb in the visible region of the spectrum, making them well adapted for the fabrication of light-triggered devices and for spectroscopic investigations. Recently, halogen bonding has also been demonstrated to be an effective means for directing the self-assembly of a photoactive supramolecular cyclophane in which the intermolecular photodimerization of two stilbenes occurs quantitatively in the solid [4]. In this case, competing photoinduced dehalogenation reactions do not present a problem, perhaps due to the solid matrix that prevents diffusion of biradicals that may be formed transiently. In an another example, a contrasting lower efficiency for the enantioselective photodimerization of an anthracene carboxylate driven by multiple point hydrogen bonding was reported in solution (10% yield with 55%ee at –50°C) [5]. This may highlight the greater difficulty in using “weak” interactions in solution, where the cooperativity afforded by the crystal phase is lacking.

Numerous systems containing a photosensitive moiety have been designed to show a measurable response to light irradiation [6]. This can include morphological changes [7, 8], which then affect the size, shape, and physical properties of aggregates, but also the release [9, 10] and capture [11, 12] of substrates. Herein, we will discuss the results we have obtained with respect to the application of non-covalent interactions to the catalysis of excited-state reactions such as [2+2] and [4+4] cycloadditions and Au(III) reduction to Au(1) and Au nanoparticles.

Supramolecular catalysis of [2+2] photocycloaddition

Our early work on photoinduced cyclodimerization of styrene, cinnamate, and stilbene cycloadditions [12, 13] followed reports by Scheffer and co-workers [14] that molecular design could be used to direct the relative spatial arrangement of photoactive moieties in the solid. At the same time, Lewis et al. had demonstrated that weak steric repulsions were sufficient to guide intramolecular photoinduced hydroamination of alkenes [15]. Together, this suggested that even relatively weak non-covalent bonds might be sufficient to direct the outcome of a photoreaction based on the reactants being pre-arranged prior to the absorption of light. A previous example by Ziegler and Beak of intermolecular cyclodimerization of cinnamic esters accelerated by hydrogen-bonding provided further proof that the concept was reasonable [16].

Of the numerous systems investigated over the years, one of the most noteworthy results involved the use of multiple supramolecular interactions [11]. In compound 1 (Fig. 1), the exocyclic C=C bond of the cinnamate ester can undergo both intermolecular [2+2] photodimerization and E, Z isomerization upon excitation. The presence of 0.5 eq of 5,5′-dihexylbarbituric acid (2) increases the photodimerization quantum yield 2-fold with respect to that of 1 alone. In contrast, the presence of a cation such as Ba²⁺ known for its propensity to bind to podand structures increases the quantum yield for photodimerization by a factor of 300. What is more interesting is that there is little selectivity between the heat-to-head vs. the heat-to-tail photodimer when 1 is irradiated alone or in the presence of Ba²⁺, thereby underscoring the synergy that is observed when both 2 and Ba²⁺ are combined. Not only does the overall quantum yield for photodimerization surpass 10%, but the selectivity for the head-to-tail photodimer (generally the less favored one) is increased 2.6-fold. By comparison, one may note that the quantum yield for photodimerization in neat ethyl cinnamate (i.e. without solvent) is only about 10%.

Fig. 1:

Example of a photoactive cinnamate appended with both metal ion and hydrogen-bonding recognition sites. The presence of both dihexylbarbituric acid and barium leads to a complex (PM3 optimized structure) possessing a 1000-fold enhancement of the yield of photodimers. This demonstrates the existence of synergy between orthogonal supramolecular interactions on the excited state behavior. Adapted with permission from reference [11].

Another striking example of the breadth of using supramolecular interactions to influence photodimerization reactions was provided to us by the intermolecular [2+2] photodimerization of fullerene C₆₀ [17]. The latter is only modestly soluble in non-polar solvents (e. g. o-dichlorobenzene) and possesses a short singlet lifetime (1.5 ns) that makes bimolecular reactions between ¹C₆₀* and ground state fullerene improbable. Indeed, although fullerene photodimerization is efficient in the solid, it is reputed inexistent in fluid solution unless the two fullerene cages are linked covalently to render the reaction intramolecular. Based on this, it would appear difficult to provide a supramolecular solution to enhancing fullerene photodimerization in solution without decorating the fullerene core with a molecular recognition unit. The melamine – barbituric acid system proved to be yet again a valid approach as the repeat distance in the hydrogen-bonded ribbon is commensurate with the diameter of the fullerene cage. Therefore, it is possible to conceive linear architectures in with the fullerenes are aligned with the 6,6 C=C in close proximity. In this case, however, synthetic accessibility suggested that the barbituric acid moiety could be directly located onto C₆₀ via a modified Bingel reaction. In this case, the supramolecular catalyst is embodied by the melamine moiety appended with solubilizing octyl side-chains (3). Its presence improves the solubility of the fullerene-barbituric acid derivative 4 and induces its photodimerization in fluid solution. In the absence of 3, no photodimers are detected under the same experimental conditions or even upon prolonged irradiation, in accordance with previous reports.

A possible extension of the supramolecular catalysis of C₆₀ cycloaddition that we were only partially successful in exploiting is the synthesis of ordered, linear fulleropolymers. If one considers the product of the supramolecularly-induced photodimerization of two molecules of 4, it is evident that the two fullerenes are necessarily ordered with the barbituric acid moieties positioned along the same direction. Indeed, previous work on the photodimerization of stilbenes had revealed that the photodimers are actually very good receptors for the supramolecular catalyst used to accelerate their formation. Because 4 is symmetric, the photoproduct can bind a molecule of 3 on each side, much like 4. Therefore, one would expect that (4)₂ would also participate in hydrogen bonding to 3 and undergo photodimerization along the 6,6 C=C site antipodal to the location of the first cycloaddition. This would lead to the growth of a linear fullerene polymer in which the fullerenes are held together by cyclobutane rings and the barbiturate moieties are aligned along one side. There are, however, numerous pitfalls with this approach. First, the polymer grows step-wise similarly to a polycondensation process. Such polymerization processes are concentration dependent and do not lead to high molecular weight materials unless the reaction progresses to near completion. Then, it is know that the fullerene cyclodimers are photosensitive and undergo retro [2+2] under the same condition as which the photodimers are formed as there is little difference in the absorption spectrum of C₆₀ and its cyclodimer. With this knowledge in hand, it would be un-realistic to expect isolation of fulleropolymers from the irradiation of 3 and 4 in sufficient quantities for structural characterization.

In spite of the difficulties outlined above, we were nonetheless successful in detecting, by MALDI-TOF mass spectrometry, the presence of a covalent adduct formed by the cyclodimerization of 3, 4, and 5 fullerene derivatives (Fig. 2). The trimer, tetramer, and pentamer of 4 accompany the presence of the dimer and their relative distribution is typical of a step-growth at its early stages. Our reasoning would lead to suggest that the fullerenes are ordered as shown in Fig. 2, but there is as yet no experimental proof of this. Indirect support for this claim stems from the solid-state structure of a related fullerene-barbituric acid adduct in which the fullerene moieties are maintained in a close-packed linear arrangement thanks to the presence of complementary hydrogen-bonding interactions.

Fig. 2:

MALDI-Mass spectrum of the result of the irradiation of a fullerene barbituric acid (4) in o-dichlorobenzene in the presence of a complementary melamine derivative (3). Signals corresponding to the formation of the dimer, trimer, and tetramer are clearly visible.

Supramolecular catalysis of [4+4] photocycloaddition

All cycloaddition reactions are sensitive to the relative spatial arrangement of the reactants. In the case of [4+4] photocycloadditions, numerous examples of supramolecular control of intramolecular anthracene photodimerization were reported by Bouas-Laurent, Desvergne, and colleagues [18, 19]. The time seemed ripe to put this well-known and highly reversible reaction to use in controlling an external event such as molecular recognition. The initial results were very positive and demonstrated that appending anthracenes on a barbiturate receptor allowed efficient and reversible modulation of the binding constant over 3 orders of magnitude. This allowed us to conceive a system in which the photoactive receptors are bound to a substrate in order to fashion a read – write – erase molecular printboard.

The concept illustrated in Fig. 3 is based on a Hamilton-like receptor for barbiturates that is appended with anthracenyl chromophores on the extremities of the two arms to control the binding constant through the [4+4] dimerization reaction [9, 10]. On the opposite side, a tether terminated with an anchoring unit is used to bind the photoactive receptor to a substrate in a geometry that exposes the binding site. While relatively straightforward, the concept rests on several key assumptions, including the propensity for the anthracenes to undergo intramolecular photodimerization regardless of the geometrical constraints imparted by the surface binding. Along these lines, work by Fox [20] showed at least partial reversible photodimerization of surface-bound anthracenes. A critical point is the detection of the photoproducts as well as the characterization of the binding affinity in the absence of commonly used techniques such as NMR.

Fig. 3:

Molecular read – write – erase printboard based on supramolecular interactions governed by the photoinduced gating of a Hamilton receptor for barbituric acid. Adapted with permission from reference [10].

The binding of the photoactive receptor onto flat Au substrates was followed by ellipsometry and contact angle measurements, both of which are in agreement with the initial formation of a compact mixed monolayer composed of decanethiol and 1-dodecanethiol azide. The latter was used in a CuAAC reaction to couple onto the photoactive receptor appended with a terminal alkyne functionality. This two-step approach presents several advantages as it allows the formation of a compact monolayer that is not disrupted by the presence of the bulky receptor while avoiding surface aggregation of the receptor. Also, it makes for facile detection of the surface modification steps using both contact angle measurements (azides are polar whereas the receptor is hydrophobic) as well as using IR spectroscopy. The latter is best implemented in the form of polarization-modulated IR reflection absorption spectroscopy, which provides better signal-to-noise than transmission or reflection IR. The results presented in Fig. 4 nicely summarize the capabilities of PM-IRRAS applied to photochemically-induced surface transformations. Although only minor differences in the region 1360–1280 cm⁻¹ are evidenced between the open and closed receptor, this is sufficient to evaluate the progression of the photodimerization reaction of the anthracene moieties to be ca. 70% following irradiation at 365 nm. This result is in agreement with fluorescence studies, which indicate strong anthracene fluorescence from the receptor in solution, whereas the latter immobilized on Au substrates emits with a pronounced bathochromic shift attributed to excimer emission. Upon photodimerization, the emission from the substrate is extinguished.

Fig. 4:

Isotropic extinction coefficients k(ῡ) of the open (black) and closed 5 (red) receptors determined from ATR IR measurements (see SI for experimental details). The bands assigned to the open and closed (photodimer) receptor are labeled o and c, respectively. Right: Portion of the PM-IRRAS spectra of SAM containing the receptor before (red) and after (black) irradiation. Adapted with Permission from reference [21].

One of the principal difficulties in assessing molecular recognition between a surface-bound host and a guest in solution resides in the compartmentalization of the reaction medium. Thus, the use of association constants as understood from solution studies is not straightforward. To circumvent this problem, we performed force spectroscopy measurements using surfaces modified with receptors (open or closed) and AFM cantilevers functionalized with thiol-tethered barbituric acid (Fig. 5) [22, 23]. Multiple force spectroscopy experiments were performed on monolayers appended with open or closed receptors and using modified or unmodified tips. The force histograms obtained are characteristic of the interactions between the tip and the substrates. When a tip modified with barbituric acid is brought into contact with a substrate functionalized with an open receptor, about half of retraction curves exhibit single binding events yielding a force histogram with a dissociation force of 172 pN (Fig. 5). One may notice an additional binding event at 350 pN which corresponds to twice the force of the single event and is therefore assigned to the simultaneous breaking of two barbituric acid-receptor interactions. In contrast, no dissociation force is observed with an unmodified tip, whereas a very low probability of binding (13%) is found if the surface is grafted with the closed receptor. To study the modulation in binding upon irradiation of the surface-bound receptors, we probed the interactions on surfaces modified with the open receptor before and after photoirradiation at 350–390 nm. The dilution of the receptor on the surface was increased in these experiments to limit the probability of multiple binding events. After 15 min of irradiation, the probability of the most frequent single event force decreased from 7.8 to 6.5% and decreased to 4.5% after an additional 45 min of irradiation. This confirms the photoinduced switching of a hydrogen-binding receptor immobilized on a substrate.

Fig. 5:

Probability of single event forces plotted as a function of the interaction force measured by AFM using barbiturate-modified AFM tip and gold surface functionalized with the receptor in the open (A) or closed (B) configuration. Solid lines are the Gauss fits. The total amount of the analyzed events together with the percentage of single and multiple events are presented in each case. Adapted with permission from reference [21].

Supramolecular catalysis of an inorganic material

The interest in supramolecular applications for organic catalysis does not imply that there is little scope for supramolecular photocatalysis outside organic synthesis. Indeed, there are numerous examples of metal oxide frameworks obtained through the use of organic templates, and the possibility of using photoinduced electron transfer (PET) to prepare low-valent metal species is well-documented. Therefore, it is relatively straightforward to combine a metal ion binding event with a PET or other photosensitized reduction mechanism to form either catalytically active low-valence metal centers or an inorganic material.

To illustrate this, we conceived a diphenylanthracene chromophore appended with two podands tethered to thioether binding sites [24]. Compound 5 (Fig. 6) is highly fluorescent and is an efficient ligand for binding Au(III) ions. In fact, we observe that 5 is capable of extracting 2 eq of Au(III) from an aqueous solution of AuCl₃ into a non-polar dichloromethane or toluene phase. This is quite remarkable and allows the selective irradiation of the complex without background irradiation of unbound Au(III) salt. The intense fluorescence of 5 is quenched upon binding Au(III) with a rate of 3.7×10¹⁰ s⁻¹ as determined using femtosecond transient absorption spectroscopy.

Fig. 6:

Example of a diphenylanthracene chromophore used as a photocatalyst for the reduction of Au(III) to Au(I) upon irradiation at 385 nm.

Irradiation of degassed toluene solutions of 5·2AuCl₃ leads to a progressive discoloration of the yellow color associated with Au(III) and a return of the intense fluorescence of 5. Mass spectrometry, NMR, and catalytic activity are all in agreement with the formation of an Au(I) complex that is bound to 5. It is presumably formed following a photoreduction of the complexed Au(III) ions that is sensitized by the diphenylanthracene core. Because no evidence for the formation of 5·+ could be seen in the transient absorption experiments, we conclude that the initial photosensitization step occurs via a mechanism involving Förster energy transfer to populate an excited Au(III) species. The latter is highly oxidizing and reacts with the solvent (toluene) to form a reduced Au ion and a toluene radical cation. The photoreduction cycle is shown in Fig. 7, and may be repeated until a colorless Au(I) solution is formed,. Au(I) can no longer act as an energy acceptor from diphenylanthracene, and is therefore an end product of the transformation. Gas chromatography – MS analysis of the photolysed solutions revealed toluene decomposition products (benzyl chloride, 1,2-diphenylethane) in agreement with this mechanism [24].

Fig. 7:

Proposed mechanistic pathway involving intramolecular energy transfer from the excited 5 to bound Au(III) and subsequent oxidation of the solvent.

In the absence of water, the photogenerated Au(I) complex is stable and does not evolve further. However, addition of water following irradiation shifts the reaction towards the formation of Au nanocrystals, as commonly observed for Au(I) salts. Interestingly, the entire process can be combined by irradiating a biphasic dichloromethane/water mixture in which 3 is dissolved in the organic phase and AuCl₃ in the aqueous phase. Stirring results in the dissolution of AuCl₃ in the organic phase which, upon irradiation, generates Au nanocrystals. These are intrinsically hydrophilic and migrate to the aqueous phase, thereby releasing 5. It is possible to conceive a recirculating reactor that allows the selective irradiation of the organic phase without exposing the aqueous phase to the light. By doing so, we succeeded in determining the maximum turnover of 5 for the generation of Au nanocrystals to be 174 [24].

The Au nanocrystals produced by this supramolecular route are essentially “naked”, i.e. without specific stabilizing ligands as the photocatalyst remains confined to the organic phase. Analysis by XPS reveals that no sulfur atoms are present on the surface of the crystals, which are therefore somewhat polydisperse in size (50–60 nm, Fig. 8). This renders the material interesting in view of possible applications in medicine [25]. plasmonics, and catalysis [26, 27]. In fact, we observe that the nanocrystals are not stabilized by molecules of 5, as these remain in the organic phase.

Fig. 8:

Top: photograph and schematic diagram of a recirculating photoreactor for the production of Au nanocrystals using 5 and an aqueous solution of AuCl₃. Bottom: Transmission electron microscopy images of the as-prepared nanocrystals obtained. Adapted with permission from reference [24].

Conclusion

Our work, along with those of many other research groups, concords to show the potential of using supramolecular interactions to control excited-state reactivity. In particular, we have shown that even relatively weak interactions such as hydrogen-bonding are capable of influencing the outcome of a photochemical reaction such as [2+2] or [4+4] cycloaddition. The use of photogated receptors for designing photoactive surfaces whose surface-binding ability can be reversibly tailored is a ongoing project that has obliged us to use new approaches for studying and quantifying photochemical reactivity on surfaces. We have also found that it is possible to conceive organic photocatalysts for the preparation of inorganic materials. The example shown herein possesses a catalytic turnover that is comparatively large with respect to those reported in the literature for organic systems and may open up new possibilities for the photochemical generation of low-valence metal catalysts.