Heteroleptic metallosupramolecular aggregates/complexation for supramolecular catalysis

Introduction

Supramolecular catalysis [1-3] for most chemists is associated with a catalytically active capsule providing either activating groups or surface/volume properties for catalytic activation [4-12]. In the present short review, limited to discrete heteroleptic metallo-supramolecular ensembles [13,14], we will show that, in addition to the above-mentioned way, there are other diverse possibilities to profit from supramolecular protocols in catalysis [9].

Metal–ligand-based 2D and 3D self-assembled architectures have been extensively studied over the past decades [15-21]. While in the early years, the focus has been on the exploration of structural features, more recent advancements have led to a multitude of extremely useful functional applications (molecular recognition, ion sensing, catalysis, etc.) [9,22]. Since most of these structures are homoleptic in nature, i.e., they are constructed from a single type of ligand [4-22], the idea of breaking/reducing symmetry has ignited a vivid search for heteroleptic aggregates that are made up by different components. Their higher degree of functional diversity and structural heterogeneity should allow, as amply demonstrated by Nature, for instance, by the multicomponent ATP synthase motor [23], a more detailed and refined configuration of purposeful machinery [24].

For the preparation of heteroleptic aggregates, one must differentiate between dynamic (rapidly exchanging) and kinetically inert heteroleptic metal–ligand interactions. While the inert heteroleptic metal–ligand motifs often center about iridium, ruthenium, rhodium etc. [25], the dynamic ones are constructed using copper(I), zinc(II), cadmium(II), iron(II), palladium(II), etc. as metal ions due to their more rapid ligand exchange rates [24-26].

The strategies to prepare inert vs dynamic heteroleptic aggregates are quite different. While in the former often a step-by-step attachment of the different ligands to the metal centers under kinetic control is dominating, the formation of dynamic aggregates relies on effective self-sorting protocols under thermodynamic control [24].

In its initial definition, self-sorting describes the capability to distinguish "self" from "non-self" in a mixture of constituents [27-30], i.e., the formation of well-defined homomeric aggregates [31] instead of a random combination of constituents in the product. This definition was later extended to heteromeric complexes by Issacs’ classification [32] of two main categories: (a) social self-sorting, which involves the assembly of different species, and (b) narcissistic self-sorting, which only involves aggregation of the same component. Over the past few years, a variety of social self-sorting protocols has led to a significant number of self-sorted cages/assemblies that have demonstrated their potential to act as functional materials [30].

Various self-sorting protocols leading to quantitative formation of heteroaggregates under thermodynamic control have recently proven their capacity. A prominent procedure developed by Sauvage on the basis of topological control [33] has found ample use in the preparation of rotaxane-based machines and devices [34]. A key element is a macrocyclic phenanthroline with an endotopic binding site as it precludes homoleptic complex formation. A further principle, introduced by Lehn, uses maximum site occupancy to afford heteroleptic aggregates [35]. While this principle is limited, the charge-separation approach by Stang is of much wider use [36]. Probably, most contributions in the literature, though, are based on using steric constraints in heteroleptic aggregation, since a variety of heteroleptic aggregation protocols have been developed by Schmittel [37] (for copper(I), zinc(II), cadmium(II), mercury(II) ions) and Yoshizawa/Fujita [38] (for palladium(II) ion) that involve pyridine-derived ligands. Highly innovative are the approaches for terpyridine-based complexes by Chan using complementary ligand binding [39], sometimes combined with conformational regulation [40], and of Newkome/Li [41] applying mainly geometric complementarity [42]. Clever utilized shape complementarity [21] for building heteroleptic palladium(II) cages whereas Crowley developed a procedure to kinetically metastable cages using naked Pd²⁺ [43,44].

The following selected structures (Figure 1) shall give a flavor of recent achievements in making fascinating heteroleptic structures using dynamic binding motifs.

The availability of powerful tools for building heteroleptic aggregates has led to a multitude of fascinating structures, accompanied by bright and confident prospects for interesting future applications. Concentrating on the topic of catalysis, however, one must confess that despite ample promises, so far, the number of established cases is rather small. For this brief account, we have identified four different categories under which the current examples describing supramolecular catalysis profiting from heteroleptic binding motifs may be summarized:

Review

Catalysis using heteroleptic discrete supramolecular architectures

In its early years, supramolecular chemistry mainly focused on host–guest interactions, primarily on the electrostatic interaction of crown ethers and alkali metals [4]. While, in the beginning, crown ethers were an excellent choice for metal ion complexation, they later received ample recognition as supramolecular catalysts [49].

The majority of host capsules, however, has been constructed using aromatic walls that offer van-der-Waals and π–π stacking interactions in order to compensate for the absence of a strong electrostatic interaction. These non-covalent/ionic interactions play an important role in encapsulating aromatic organic molecules, especially in aqueous medium. Along this rationale, organic host molecules such as cyclodextrin [50], pillararenes, and cucurbiturils [51] have been developed in the last decades. Although, these exhibit excellent host–guest encapsulation properties with a variety of organic molecules, there are major drawbacks associated with them: variation of the shape and size of their cavity requires often tedious de-novo synthesis.

Eventually, these drawbacks have been tackled by synthesizing discrete supramolecular hosts based on metal–ligand coordination-driven self-assembly [15-22,52]. This approach not only solved the issue with low overall yields, but it also provided chemists with a superior control over the shape and size of the host’s cavity. Since the coordination bonds which define the host structure are labile in nature, they allow the formation of thermodynamically controlled architectures using a self-correcting mechanism [15,53]. Generally, the construction is spontaneous and highly selective with quantitative conversion.

In search of supramolecular cavity-induced catalysis, chemists have become fascinated toward the design of large and sophisticated molecular vessels. In this context, Stang [54], Nitschke [55], Fujita [56], and others [57] have reported several template-free assemblies giving access to novel structures. The primary objective of these nanovessels as supramolecular catalysts is to encapsulate organic reactant/s to lower the activation barrier, thereby mimicking the functions of enzymes (without replicating their structures). The structural dissimilarity between the reactants and the subsequent product often contributes to the successful release of the product from the reaction vessel, thus, reducing product inhibition. Hence, it can be envisioned that the introduction of functionality within the building blocks to decorate the inner cavity of the host would produce efficient types of catalyst, where the substrates get activated for a particular reaction upon interaction with the functional group inside the cavity.

Mukherjee et al. have demonstrated the construction of 3D nanocages employing imidazole-based multidentate donors [58]. The conformational asymmetry of the imidazole units opened the venue to nanocages of different shapes and sizes with ease. To take the directional self-assembly to the next level, a three-component self-assembly of the tetra- and tri-imidazole donors 6 and 7, respectively, was carried out with 5 via social self-sorting to form the Pd₇ molecular cage 8 (Figure 2) as driven by the directionality of the donor nitrogen of the building ligands. The unique three-component Pd₇ molecular boat has a proper internal nanocavity showing preferential affinity towards aromatic molecules through π–π stacking with the hydrophobic aromatic wall of the host. Finally, the boat was investigated as a catalyst for the Knoevenagel condensation reaction (Figure 2) of a series of aromatic aldehydes with 1,3-dimethylbarbituric acid and Meldrum’s acid in aqueous media.

One of the primary advantages of utilizing self-assembly to construct discrete nanocages with predetermined geometry and function is the use of a one-pot reaction employing complementary organic linkers with inorganic metal ions. Although the one-pot synthesis of homoleptic metallacages has been thoroughly investigated over the years [15-22] a more accurate understanding of the self-assembly of diverse components and the development of functionally integrated smart architectures for catalysis as “artificial enzymes” certainly demands further attention. Enticed by this idea, Mukherjee and co-worker have demonstrated the design and synthesis of urea-functionalized 2D/3D architectures and their catalytic activity. The urea moieties were incorporated within the building blocks and were meant to serve as binding sites for appropriate substrates, therefore, promoting selectivity and reactivity. For this purpose, they have chosen the ditopic bisurea “strut” 12 (Figure 3), which generated a 2D discrete molecular triangle 14 in the presence of an equimolar amount of the cis-(tmen)Pd(NO₃)₂ acceptor 5 [59].

In order to prevent the urea moieties on the triangle to get engaged in intermolecular H-bonding, which would lead to catalytic quenching, a unique design strategy was applied. Instead of using a cis-blocked palladium(II) unit for self-assembly, Pd(NO₃)₂ was employed along with the triazole-based 0° clip 13. In a one-pot reaction, 12 and 13 in DMSO were treated with Pd(NO₃)₂ in a 1:1:1 ratio, which entailed the quantitative formation of the edge-directed molecular prism 15 (Figure 3). The adequate length of the ditopic clip 13 constrained the urea triangle from intersupramolecular H-bonding. Thus, the urea moieties in this newly assembled 3D architecture were freely available for interactions with appropriate guest molecules through H-bonding. Different organic molecules, such as nitroolefins, capable of forming H-bonding with the urea moieties inside the cavity were investigated in water for their ability to encapsulate in the cavity under heterogeneous conditions. Successful binding of the guest molecule was proven by UV–vis and IR spectroscopy. The multicomponent prism (Figure 3) was finally utilized as a heterogeneous catalyst for Michael and Diels–Alder (DA) reactions in water, representing an uncommon hydrogen-bond donating heterogeneous catalyst [59]. Intrigued by the successful guest-inclusion, Michael reactions were performed with prism 15 (Figure 4). Generally, most of the organocatalysts get destroyed during the work-up procedure, so that recovery is often difficult. However, in the present case the catalyst could be easily recovered, and multiple catalytic cycles could be performed.

Supramolecular systems based on non-covalent interactions have drawn considerable attention in assembling efficient light harvesting systems (LHSs) in the last decade [60,61]. Significant attention has been centered to construct artificial LHSs via FRET (fluorescence resonance energy transfer) that include organic materials and supramolecular assemblies. Recently, FRET phenomena have been successfully demonstrated in supramolecular architectures based on metal organic frameworks and covalent organic frameworks [62-64]. However, poor solubility of such polymeric systems in common solvents restricts their use in potential applications. In this aspect, coordination-driven discrete architectures provide a promising future due to their facile one-pot synthesis and high solubility in common solvents.

Mukherjee and co-workers have developed supramolecular architectures containing tetraphenylethene (TPE) units which act as an aggregation-induced emissive (AIE) fluorophore [65]. The newly designed TPE-based tetraimidazole donor 19 has been treated with 180°/120° trans-Pt(II) acceptors which led to the coordination-driven self-assembly of 3D discrete molecular cages in aqueous medium (Figure 5). The 180° acceptor without an organic spacer, trans-[Pt(PEt₃)₂(ONO₂)₂] 20, provided the cage 23a with the molecular composition (20)₄(19)₂(NO₃)₈, whereas 21 (with a spacer unit) led to the formation of the Pt₈ cage 24a = (21)₄(19)₂(NO₃)₈. In a similar fashion the bent 120° acceptor 22 also assembled into the Pt₈ cage 25a = (22)₄(19)₂(NO₃)₈ [65]. Counter anion exchange from NO₃⁻ to PF₆⁻ made the cages soluble in acetonitrile. All three cages formed spherical supramolecular nano-aggregates in a water/acetonitrile (9:1) mixture and showed increased emission in the aggregated state [66]. Rhodamine B (26) was chosen as a FRET acceptor as there is a considerable overlap in energy of the donor emission and acceptor absorption. Then, artificial LHSs were constructed with aggregates from 24b or 25b at a donor/acceptor ratio of 5:1 (Figure 6). Finally, the light harvesting materials (24b + 26) and (25b + 26), respectively, were successfully employed as visible-light photocatalysts for a cross-coupling cyclization of N,N-dimethylaniline (27) and N-alkyl/aryl maleimides 28 (Figure 6). Notably, the systems showed much higher catalytic activity compared to similar reactions with the dye or cages alone.

Transition-metal catalysts play an important role for the development of intricate pharmaceutical drugs. Although transition-metal catalysts based on rhodium, cobalt, and palladium have been intensively studied, gold catalysis has received encouraging attention only recently [67,68]. Since selectivity of gold-catalyzed reactions is still a concern, the catalytic transformation is often controlled by introducing a ligand as a first coordination sphere of the active gold species. In a supramolecular approach, introduction of the ligated gold complex inside a hollow cage may significantly improve the reactivity and selectivity because the cage provides a second coordination sphere around the catalyst thus controlling the catalytic reaction. This idea has been impressively demonstrated by Reek [69], Ballester [70], and Raymond [71], where a single gold-based catalytic unit is encapsulated inside a molecular host to modulate the reactivity.

Recently, Reek and co-workers have constructed an M₁₂L₂₄ nanosphere by treating the bispyridyl 120° ligand 30 with a Pd(II) precursor [72]. Here, the ligand 30 is optimally functionalized with a phosphine gold(I) chloride moiety so that the metal catalyst will reside inside the sphere (Figure 7). In order to vary the local gold concentration inside the cavity, heteroleptic cages were assembled from a multicomponent one-pot reaction of Pd(II) with 30 and the analogous non-functionalized ligand 31. By controlling the ratio of 30 and 31, spheres with varying concentrations of AuCl could be constructed. Finally, the various spheres were investigated regarding the catalytic effect of the local gold concentration on the hydroalkoxylation of γ-allenol 34 (Figure 8). Since all catalytic tests were carried out at the same overall gold concentration of 5 mM, it was very interesting to observe that nanospheres with a very low local gold concentration could not catalyze the reaction at all. Actually, product formation was observed only with nanospheres having a higher 30/31 ratio (>6:18). Surprisingly, only the formation of the five-membered ring product 35 was observed, with a maximum yield of 88% at the highest 30/31 ratio of 24:0. The system was benchmarked with the building unit 30 itself and Ph₃PAuCl where they observed a negligible conversion, which points toward the importance of local catalyst concentration.

A wide range of cavity-based artificial supramolecular catalysts has been successfully developed employing dynamic metal–ligand coordination bonds. While only a small fraction of these 3D architectures was useful for chiral catalysis, an even smaller fraction was able to provide a high stereoselectivity during asymmetric catalysis [5]. For instance, an enantiopure tetrahedral Pt₁₂ cage has been previously studied for catalytic Michael addition reactions, but no enantioselectivity was detected because the chiral building blocks were located at peripheral positions thus not sufficiently breaking symmetry within the cavity [73]. Therefore, it was envisioned that the chiral moiety should be incorporated in the ligand unit in order to provide an enantiopure assembly with an asymmetric cavity. 1,1'-Binaphthol (BINOL) is one such chiral building block, which has been successfully utilized to carry out numerous asymmetric catalytic reactions [74]. Keeping this in mind, Stang and co-workers constructed heteroleptic triangles via the assembly of a BINOL-based ditopic ligand and 180° trans-Pt(II) acceptors. The 3,3'-dipyridyl-substituted chiral BINOL donor (S)-37 has a bite angle of 60° and when treated with linear 180° acceptors 38 and 39, it produced the differently sized triangles (S)-40 and (S)-41 depending on the length of the organic spacer in the acceptor unit (Figure 9) [75].

Since the interior cavity of the homochiral macrocycles was equipped with BINOL units, they were utilized as catalysts for the asymmetric conjugate addition of chalcone 42 with trans-styrylboronic acid (43, Figure 10). The catalytic reaction inside the chiral cavity of (S)-40 provided a yield up to 91% with a very high enantioselectivity (94% ee). In contrast, the larger chiral macrocycle (S)-41 afforded a slightly lower catalytic activity (87%), however, at a similar enantioselectivity (94% ee). Using similar reaction conditions, the non-assembled BINOL derivative (S)-3,3'-dibromo-[1,1'-binaphthyl]-2,2'-diol acted as a superior catalyst (99% yield) but achieved a lower enantioselectivity (84% ee). Therefore, this result indicates that the incorporation of multiple catalytic sites and an appropriate asymmetric cavity is the key for the enhancement of catalytic activity and stereoselectivity [75].

Strictly speaking, the above example does not contain fully dynamic heteroleptic metal fragments, as the aryl–Pt bond is not kinetically labile. Nevertheless, the example illustrates the opportunities in running enantioselective catalysis in mixed-ligand frameworks.

Instead of constructing supramolecular catalysts by functionalization of the linker units, a different approach can be adopted where a catalytically active molecule is encapsulated inside a confined space, as demonstrated by Reek and co-workers by using a previously reported heteroleptic bisporphyrin cage [76]. The tetragonal prismatic nanocage 47 consisted of two zinc-porphyrin units along the two tetragonal faces (Figure 11), which allowed encapsulation of the chiral phosphoramidite 48 as a precursor for the final catalyst. In the next step, a transition-metal-ion based catalyst was prepared in situ by addition of one equiv of [Rh(acac)(CO)₂], which generated the monoligated rhodium complex. Finally, catalysis was carried out at 5 bar of H₂/CO (1:1) that converted the [Rh(acac)(CO)₂] complex into the active hydride species 49 that is well known for hydroformylation reactions [77]. Incorporation of the monoligated catalyst into the confined cavity of the capsule showed very good catalytic activity towards the hydroformylation of styrene (50, Figure 11) with a high stereoselectivity (65% ee) at 32% conversion compared to the non-encapsulated catalyst, which only managed to yield 8% ee at 4% of conversion. Thus, the molecular capsule 47 can be viewed as a second coordination sphere of the catalyst, reminiscent of enzymatic active sites.

In summary, the above discrete molecular architectures containing cavity and sometimes smart functionality describe a new class of supramolecular catalysts that are effective tools to control the activity and selectivity of organic transformations. Clearly, building heteroleptic assemblies provides an unprecedented flexibility towards controlling the dimension as well as functionality. For instance, a particular catalytic reaction can be efficiently carried out by choosing a specific functionality. The functional entity responsible for catalysis can either be incorporated with the building blocks, or the catalyst itself may be encapsulated into the cavity. Unfortunately, the examples of heteroleptic cages acting as catalysts are limited so that further development is urgently needed.

Catalytic effects due to nanomechanical motion

The above examples have demonstrated the potential of heteroleptic cages to enable catalysis under various conditions. In this subchapter, the role of the cage/architecture depends on its dynamics: the catalytic activity will correlate with the rate of thermal motion. As illustrated below, there are presently two phenomena known where continuous nanomechanical motion influences the catalytic activity: a) Increasing nanomechanical speed reduces product inhibition, and b) higher nanomechanical speed enlarges catalyst liberation.

While the development of multicomponent rotors has started almost 20 years ago with seminal works by Shionoya [78-80] and Kume [81,82], the fascinating prospects of discrete nanomechanical motion was impressively demonstrated by Aida with the development of multicomponent tweezers [83,84]. Using a variety of orthogonal complexation motifs, the Schmittel group has developed over the past five years a general approach to multicomponent rotors that relies on the binding difference found in the HETPYP (HETeroleptic PYridine and Phenanthroline complexes [85,86]), HETPHEN (HETeroleptic bisPHENanthroline complexes [87]) and HETTAP (HETeroleptic Terpyridine And Phenanthroline complexes [88]) interactions (Figure 12). Due to the different amount of donor atoms about the metal ion, the binding strength will decrease in the series of HETTAP > HETPHEN > HETPYP [37]. In any dynamic system designed for exchange motion predictably the weakest interaction will be the most dynamic one. This protocol is quite generally applicable and is readily illustrated by the ensuing example. The three-component rotor 52 = [Zn(53)(54)]²⁺ was developed on the basis of two complexation events, i.e., the zinc(II) HETTAP (log β = 14, log K₁ ≥ 6) and N_py → ZnPor (log K = 4.45) binding motifs (Figure 12) [89]. Due to the design of 52, the pyridine terminal oscillates between the two degenerate zinc porphyrin (ZnPor) stations of 53 at k₂₉₈ = 24 kHz (at 298 K).

A similar exchange process was observed in the four-component rotors [90] developed along two orthogonal self-sorting motifs (HETPYP = N_py → [Cu(phenAr₂)]⁺ and N_DABCO → ZnPor interactions). Again, the synthetic approach is a straightforward multicomponent self-sorting assembly. Accordingly, the distinct zinc porphyrins 55 and 57 were positioned at a defined distance through two N_DABCO → ZnPor interactions. Clearly, without additional measures, homo- and heteromeric assemblies would form. However, due to the additional HETPYP interaction(s) in the presence of copper(I) ions, thermodynamic stabilization quantitatively drives the reaction to the hetero-assembly 59, simply by mixing the components in the correct stoichiometric ratio (Figure 13). Various nanorotor assemblies are possible by this approach.

The dynamics of the four-component rotor 59 = [Cu₂(55)(57)(60)]²⁺ cannot be resolved due to the high symmetry in both the rotator and stator [91]. However, when rotator 58 with two different terminals was used then the rotational speed could be measured using VT ¹H NMR [90]. Over the years, both the stator and the rotators were varied over a wide range, i.e., see 61, changing the geometrical or constitutional situation at the binding sites [92]. In a detailed recent study, the finding of a Hammett correlation in such nanorotors corroborated that a rate-determining dissociation at the rotator–metal binding interaction dictated the rotational speed [93].

Reducing Product Inhibition (RPI) through nanomechanical motion. Recently, the suitability of the four-component rotors to act as catalysts in various click reactions was investigated having a look at nanorotors [Cu₂(55)(60)(X)]²⁺ (with X = 62, 63 or 64), revealing an unexpected correlation between their rotational speed and catalytic activity [94] (Figure 14). Because in any moment of the rotation there should at least one copper(I) phenanthroline be freed from contact with the monodentate rotator X, one would expect that the temporarily exposed copper(I) ions are catalytically active. It is important to note that this copper ion due to steric impediments at the phenanthroline site will not engage in complexation with a second phenanthroline (see HETPYP concept [85,86]).

The concept was probed by using nanorotors [Cu₂(55)(60)(X)]²⁺ as catalyst (10 mol %) for the click reaction of 9-(azidomethyl)anthracene (65) and (prop-2-yn-1-yloxy)benzene (66) at 55 °C (4 h) [94]. Notably, the fastest nanorotor [Cu₂(55)(60)(64)]²⁺ afforded the highest yield of the click product 67 (62%) followed by nanorotors [Cu₂(55)(60)(63)]²⁺ (44%) and [Cu₂(55)(60)(62)]²⁺ (20%) (Figure 15). The analogous tendency was recognized in a second click reaction, using now reactants 68 and 69 furnishing 70. Markedly, the yield of both click reactions was linearly correlated with the exchange speed of the catalytic nanorotors (Figure 15, right). With faster rotational exchange of the nanorotor both the measured rate of catalysis at time = zero, v₀, and the catalytic yield increased in a linear fashion. Such a clear correlation asks for a convincing theory. A straightforward explanation is that higher rotational speed should lead to a reduction of product inhibition by kicking out the product bound at the catalytic copper center. As shown experimentally, only the copper(I) phenanthroline unit which is temporarily not occupied by the rotator head is catalytically active. Thus, when the rotator dissociates from the copper(I) phenanthroline and moves to the product-filled site it should liberate the product into solution. The increased liberation of product with increasing speed of the nanorotor was proven by independent experiments, and additionally it was demonstrated that a “static” reference catalyst showed only a turnover of 1 due to product inhibition.

The capacity of the four-component nanorotors to act as catalyst in click reactions was more recently utilized for setting up a multicomponent logic AND gate that required the networking of altogether twelve components (Figure 16) [95]. At the heart of the logic operation, the ensemble of the copper(I)-loaded nanoswitch [Cu(71)]⁺ and ligand 72 was actuated by two metal-ion inputs (Zn²⁺ and Hg²⁺) and generated a stoichiometric Cu⁺ output according to the AND gate logic only in truth table state (1,1). The released copper(I) ions self-assembled the four-component rotor [Cu₂(55)(60)(73)]²⁺ which enabled catalysis of a click reaction. In summary, copper(I) ions as stoichiometric output of the AND gate finally generated a catalytic output due to the assembly of a rotating four-component catalyst. Moreover, the AND gate could reversibly be reset into truth table state (0,0).

Like the nanorotors above, domino nanorotors with two exchanging rotational axes showed catalytic action that depended on the rotational speed [96]. Ligands 53 and 76 were conceived based on geometric complementarity at their coordination sites using the HETPYP interaction in both aggregates [Cu₄(76)₂]⁴⁺ and 77 = [Cu₂(53)(76)]²⁺ (Figure 17). The structure of 76 suggested that the dimeric parallelogram-type double rotor [Cu₄(76)₂]⁴⁺ with two antiparallel pyridyl head groups operating as axles would form. Again, it is important to stress that the coordinatively frustrated terminal [Cu(phenAr₂)]⁺ units will not engage in complexation with a second phenanthroline due to steric control in the HETPYP concept [85,86]. Since at a given time, only one of both HETPYP-bound pyridines in [Cu₄(76)₂]⁴⁺ can serve as axle, the rotor undergoes a domino rotation that was measured to occur at k₂₉₈ = 142 kHz. For the heteromeric rotor [Cu₂(53)(76)]²⁺, two orthogonal dynamic interactions are relevant, i.e., the weak N_py → ZnPor binding (log K = 4.3) and the stronger copper(I) HETTAP linkage (log β = 9.3). Ligand 53 with its terpyridine (tpy) and two ZnPor sites was designed in the way that the tpy should connect with 76 via a HETTAP binding motif while simultaneously allowing binding of the pyridine terminus of 76 to one of both ZnPor units of 53. Now there are two motions possible that can only occur in a domino fashion. When the strong HETTAP complexation is intact, the exchange of the pyridine head of 76 between both ZnPor sites of 53 occurs at 64 kHz at rt. When the much stronger HETTAP complex dissociates, the much weaker N_py → ZnPor interaction remains intact and serves as a rotational axle, but now the exchange is slower by several orders of magnitude (k₂₉₈ = 0.55 Hz).

The two rotors and reference complex [Cu(78)]⁺ (Figure 17) were compared for their catalytic activity in the click reaction of 79 + 80 (Table 1), after normalizing for their different content of copper [96]. Reactants 79 and 80 were chosen, because their click product 81 should be a rather good chelate ligand leading to product inhibition. The data given in Table 1 indicate that the rotor [Cu₂(53)(76)]²⁺ has basically the same catalytic activity as a non-dynamic reference catalyst (turnover close to 1), because the exchange at the copper(I) sites preventing product inhibition is extremely slow (k₂₉₈ = 0.55 Hz). The much faster rotation leading to an exchange of the N_py → ZnPor interaction is irrelevant for the catalysis. In contrast, the faster domino rotor [Cu₄(76)₂]⁴⁺ produced far superior yields (Table 1). The different yields correlate with the distinct v₀ of the catalytic reaction. In conclusion, we see a clear trend in several rotating catalysts that with higher speed product inhibition is reduced (RPI).

Increased Liberation of Catalyst (ILC) through nanomechanical motion. While the previous and related examples [97] are based on a reduction of product inhibition (RPI) with increasing speed of the rotating catalyst, there is a second general concept, i.e., ILC, for linking catalytic activity with the speed of a nanomechanical device. It is based on the increasing liberation of a bound organocatalyst with rising speed of the catalytic machinery. This concept was first realized in the slider-on-deck systems (82•X) (X = 83, 84, or 85) (Figure 18) that were simply generated by mixing the tris-ZnPor deck 82 with one of the bipeds 83–85 (1:1) [98]. The thermal sliding speed of the biped across the deck 82 depends on the thermodynamic strength of the pyridine (or pyrimidine, methylpyridine) → ZnPor interactions of the biped’s feet with the ZnPor units. Obviously, weaker binding to ZnPor should lead to faster sliding [17].

With one of the three ZnPor units being available for the attachment (immobilization) of an organocatalyst, we wondered about the catalytic activity of the dynamic three-component ensembles 89•(82•X) using N-methylpyrrolidine (89) as organocatalyst. For assessment, the conjugate addition of 86 and 87 was studied at different temperatures for 4 h (Figure 19). In comparison with the static reference system 89•90 (no yield at 50 °C), there was a clear trend for higher yields (Table 2) the faster the exchange process in the slider-on-deck is: system 89•(82•83) furnished (18 ± 2)%, 89•(82•84) afforded (32 ± 2)%, and 89•(82•85) provided (50 ± 2%) of 88 at 50 °C after 4 h. Control experiments uncovered that product formation was kinetically controlled and that the slider-on-deck systems on their own were catalytically silent. With various controls one could demonstrate that the effect on catalysis in the catalytic machinery was due to kinetic and not thermodynamic reasons (Figure 19).

While the concept of increased liberation of an organocatalyst (ILC) has been demonstrated in other dynamic nanomechanical systems as well [99], a particular highlight was recently realized with a catalytic nanorotor that was able to build a new catalytic machinery [100]. Hereunto, the two concepts RPI and ILC were combined in a synergistic manner, starting with rotor [Cu₂(83)(91)]²⁺ (k₂₉₈ = 46.0 kHz) that catalyzed the click reaction between the zinc porphyrin ligand 92 and the azide 93 furnishing triazole 94 (Figure 20). As it was possible to drive this reaction to completion and as the click product 94 proved to be a good chelate ligand for the two copper(I) phenanthroline sites of [Cu₂(83)(91)]²⁺, the formation of the dynamic four-component slider-on-deck [Cu₂(83)(91)(94)₂]²⁺ was warranted. The exchange of the biped 83 across the newly generated deck [Cu₂(83)(91)(94)₂]²⁺ occurred at a rate of k₂₉₈ = 65.0 kHz.

As known from earlier ILC work (vide supra ref. [98]), a carefully chosen organocatalyst can be immobilized at a ZnPor unit in a way that no catalysis will result (at a specific temperature and in a defined time). We thus started with the rotor [Cu₂(83)(91)]²⁺ in the presence of ligand 92 as well as organocatalyst N-methylpyrrolidine (89) and the substrates 95 and 96 for a Michael addition (Figure 21). No formation of product 97 was observed, because the complex between the catalyst 89 and the zinc porphyrin 92 is catalytically inactive.

However, upon addition of azide 93, the formation of [Cu₂(83)(91)(94)₂]²⁺ was detected. Since the organocatalyst originally firmly bound at the ZnPor of ligand 92 is now part of the dynamic slider-on-deck [Cu₂(83)(91)(94)₂]²⁺, the motion of biped 83 will dynamically release 89 into solution. Indeed, now the catalysis of addition product 97 was turned ON. While supramolecular transformations are widely recognized [101-103], the present example illustrates how a supramolecular catalyst (three-component rotor) transforms itself into a new catalytic machinery and turns on the respective catalytic process. All in all, this protocol is remotely reminiscent of gradual evolutionary processes [100].

In summary, both the RPI and ILC concepts using variable speed of nanomechanical machinery for catalytic effects are novel and innovative not only for supramolecular, but certainly also for common catalysis. In particular, product inhibition in catalysis is a frequently encountered challenge during catalyst development. Moreover, the RPI concept may point the way towards the development of endergonic catalytic transformations, because in many of those the product needs to be stabilized within the catalytic cavity. Release from the active site then requires destruction of the stabilizing interactions. For instance, Nature has chosen in the ATP-synthase to use “fueled” nanomechanical motion to release ATP from the active site [23].

Switchable catalysis due to reversible assembly/disassembly

The common modus operandi to set up switchable catalysis usually relies on systems that can be toggled between two (or more) distinct switching states within a molecule [104-107]. In contrast, supramolecular approaches allow the shuffling and reshuffling of components to switch ON/OFF catalytic processes, a topic that has not yet found adequate attention, but links supramolecular catalysis to systems chemistry [108,109].

The following information system utilizes a seven-component mixture that is reversibly reconfigured through fully reversible assembly and disassembly thereby tuning ON/OFF two diverse catalytic reactions [99]. Addition and removal of zinc(II) ions triggered altogether three diverse processes: i) mutual re-shuffling of components leading to two different nanorotors, ii) catalysis depending on decisively different exchange rates in the nanorotors, and iii) two different catalytic processes (Figure 22).

The key challenge in the toggling process was to interconvert two nanorotors by exchange of two components but with a single-input trigger from outside. Eventually, the two-component reshuffling was solved by providing zinc as trigger (component 1) and by using a ligand (component 2) from a ligand reservoir. In the initial self-sorted State-I, the rotor [Cu(53)(98)]⁺ was paired with [Cu₂(100)₂]²⁺, the latter representing a reservoir for the rotator arm 100. Addition of zinc(II) ions induced a second self-sorting that encompassed transfer of two components along the following equation: 2 × [Cu(53)(98)]⁺ + [Cu₂(100)₂]²⁺ + 2 × Zn²⁺ → 2 × [Zn(53)(100)]²⁺ + [Cu₂(98)₂]²⁺ + 2 × Cu⁺. Accordingly, the added zinc(II) ions and ligand 100 from the reservoir [Cu₂(100)₂]²⁺ enabled the formation of the zinc(II)-based rotor [Zn(53)(100)]²⁺ (Figure 22a). The liberated ligand 98 reacted with the copper(I) ions to afford [Cu₂(98)₂]²⁺ representing again a reservoir for a rotator. Based on the stoichiometry of the above equation, two equiv of copper(I) ions were liberated. Since the released free copper(I) ions generated problems with reversibility of the transformation, 1-aza-18-crown-6 (99) (2.0 equiv) had to be added as a receptor for Cu⁺.

As State-I contains the azacrown ether 99, a potential organocatalyst, and State-II harbors the copper complex [Cu(99)]⁺ a likely click catalyst, both networked states were expected to be catalytically active, possibly even in an ON/OFF manner [99]. To test for dual catalysis (Figure 22b), State-I was reacted at 50 °C with 1.0 equiv of catalyst 99 (with respect to rotor) and 10.0 equiv (with respect to rotor) of substrates 75, 86, 95, and 102 in CD₂Cl₂/CD₃CN 5:1 for 2 h. Analysis demonstrated that 30% of product 101 but no click product 103 had formed. State-II was furnished by addition of 1.0 equiv of zinc(II) ions (with respect to the rotor) and heated under identical conditions. Finally, 55% of the click product 103 was revealed, but without further conversion of product 101. A notable reproducibility of the yields was identified in two consecutive catalytic cycles. As a result, an astounding switchable catalytic system could be based on information processing (Figure 22).

The following example of switchable catalysis involves the interconversion of the closed dimeric parallelogram [Cu₂(104)₂]²⁺ and the bishomoleptic complex [FeCu₂(104)₂]⁴⁺ (Figure 23), the latter controlling a double-click catalytic access to rotaxanes 109, by addition/removal of iron(II) ions [110]. Although [FeCu₂(104)₂]⁴⁺ is an open and flexible structure, the availability of two catalytic copper(I) centers positioned at 34 Å in the transition state of the second click reaction, leads to an astounding synthetic efficiency, although there is a major distance mismatch between the copper(I) ions in [FeCu₂(104)₂]⁴⁺ and two triazole units of the rotaxane.

For instance, the system [FeCu₂(104)₂]⁴⁺ (d_Cu–Cu = 34 Å) afforded the formation of both 109a (73%) and 109b (82%) in high yield (Figure 23b) [110], although the relevant distances in the pseudo-rotaxane 110a,b (prior to the second click reaction) are 14.3 and 21.1 Å. Actually, the yield with [FeCu₂(104)₂]⁴⁺ was far better than with a dicopper reference catalyst, where the separation of the copper(I) ions was optimal (d_Cu–Cu = 14.6 Å) for the formation of 109a. The findings were explained based on a model in which the monotriazoles 110a,b were bound to both copper centers prior to the second click reaction. Aside of one copper(I)–triazole interaction, the model suggested a chelate cooperative effect resulting from an additional η²-binding of the second copper(I) at the acetylene unit at which the second click reaction would take place. In case of distance mismatch, it is the formation of this copper–alkyne η²-complex that compensates for the build-up of strain as demonstrated by DFT computations [110]. Once the second click reaction has occurred, the rotaxane is liberated under release of strain in the catalyst if there was a distance mismatch. For cases with a close match of distances, product inhibition reduced the yield. Due to the high relevance of the CuAAC approach [111] for the preparation of rotaxanes, the exploitation of cooperative and strain effects in double-click strategies is a promising strategy.

A completely different approach to switchable supramolecular catalysis made use of a supramolecular cage-to-device transformation under dissipative conditions (Figure 24) [112]. Notably, the dissipative conditions were realized by addition of a fuel acid [113,114] that surprisingly ignited a base-catalyzed Knoevenagel addition reaction (Figure 25).

As DABCO is a stronger binding ligand (log β = 7.20) [115] towards zinc porphyrin (ZnPor) than pyridine (log K = 4.45) [90] the reaction of ligand 111, deck 82, DABCO (60), and [Cu(CH₃CN)₄](PF₆) (4:2:3:2) furnished the supramolecular cage [(82)₂(60)₃] with 60 = DABCO acting as pillars whereas biped [Cu(111)₂]⁺ remained uncoordinated [112]. In an analogous self-assembly with 83 instead of [Cu(111)₂]⁺, ligand 83 was left uncoordinated. When State-I was treated with TFA, then the reshuffling of the components afforded the slider-on-deck [Cu(82)(111)₂]⁺ (or [(82)(83)]) and the monoprotonated DABCO (60-H⁺). Addition of DBU reversed the process. The bipeds in the slider-on-deck systems [Cu(82)(111)₂]⁺ (k₂₉₈ = 42.2 kHz) and [(82)(83)] (k₂₉₈ = 32.2 kHz) move across the deck 82 and prevent binding of the protonated DABCO at the ZnPor binding sites. Use of the fuel acid 112 or 113 instead of applying the TFA/DBU acid/base combination leads initially to protonation of DABCO but due to the decarboxylation of 114 the resulting strong base 115 reclaims the proton back (Figure 25, top). As a result, State-II may be afforded under dissipative conditions.

A surprising facet of this State-I/State-II interconversion (Figure 24) [112] was the finding that the protonated DABCO (60-H⁺) was a rather efficient base catalyst for a Knoevenagel addition (Figure 25, bottom). Several control experiments excluded the possibility that the reaction was triggered by acid or by alternative pathways. The remaining unprotonated nitrogen in monoprotonated DABCO (60-H⁺) hence is a sufficiently strong base for the reaction of 116 and 117. Finally, the State-I/State-II interconversion in the presence of 116 and 117 was triggered by addition of the fuel acid. The traces in Figure 26 show the amount of Knoevenagel addition product 118 within two subsequent pulses of the fuel.

In summary, the examples listed in this subchapter shed some light on the manifold opportunities for switchable ON/OFF catalysis controlled by supramolecular interconversions. Switchable catalysis decided by an ensemble of communicating molecules, which act as an information system and control the switching states, remodels somehow the situation in a living cell, where a desired event is only ignited when several signaling parameters agree. Explicitly, the subchapter demonstrates three distinct examples, in which metal–ligand coordination was shuffled by addition/removal of either metal ions or acid. To (re)shuffle metal–ligand coordination by acid/base generates a number of further options. In particular, the protocol of acid addition and removal (upon base addition) may be elegantly replaced by using a suitable fuel acid thus opening the field of supramolecular catalysis for out-of-equilibrium processing.

Toggling between intra- and intermolecular complexation in nanoswitches

Heteroleptic complexation does not only open the way to multicomponent assembly but also to switchable catalysis in coordination-based toggles. An important and widely applicable protocol is the weak-link approach (WLA) that was developed and exploited by Mirkin [116]. It is based on the association/dissociation of hemilabile ligands bound to a metal center. Through the addition of secondary ligands, the weakly coordinated donor sites are substituted which allows an opening of the switch from a rigid-closed to a flexible semi-open form [117]. A common protocol used addition/removal of strongly binding monodentate ligands (CO or Cl^‒) at rhodium centers to dissociate/reassociate the weakly coordinated donor, for instance in tweezer-type structures [118].

An early contribution by Mirkin (Figure 27) described the opening of the closed structure 119²⁺ to the open form [119(CO)₂Cl₂] that showed a two-fold activity in the catalytic opening of epoxide 121 to 122 [119].

In another example by Mirkin, a well-known aluminum(III) salen catalyst was hidden in switch 125²⁺ [118] between sterically demanding biphenyl rings preventing ε-caprolactone from accessing the catalytic site (Figure 28). As a result, the polymerization of ε-caprolactone (123) was switched OFF. Addition of chloride anions from n-Bu₄NCl led to the substitution of the tertiary amine ligand at both rhodium centers generating the semi-open form [125(Cl)₂]. Since the aluminum(III) salen center became now exposed, the catalytic polymerization of ε-caprolactone (123) was turned ON. The chloride was readily removed by adding sodium tetrakis[(3,5-trifluoromethyl)phenyl]borate (NaBArF) and as result catalysis was turned OFF again.

Along this protocol, Mirkin et al. have developed a variety of allosterically modulated catalysts that allow ON/OFF reaction control in photoredox catalysis [120], phosphate diester transesterification [121], Friedel–Crafts reaction, ring opening of epoxides, oligomerization [116], and acyl-transfer reactions [122,123].

While there are further examples by Mirkin [124-126], Schmittel [127,128], and others [129,130] that operate at the borderline of dynamic heteroleptic complexation events, the supramolecular cases selected in the following are characterized by the reshuffling of inter- and intramolecular coordination events to realize different toggling states for catalysis [131].

Using addition and removal of chloride, the Mirkin group reversibly and quantitatively toggled the platinum(II)-based switch 126²⁺ between a homo- and heteroligated form (Figure 29) [132]. In the closed platinum(II) complex 126²⁺, the urea units were available for activation of butenone (127) by hydrogen bonding. As a result, the Diels–Alder reaction of cyclopentadiene (128) and 127 was catalyzed. Upon addition of n-Bu₄NCl, the open form was afforded that aggregated to oligomers [(126•Cl)_n]ⁿ⁺ through intermolecular hydrogen bonding at the urea moieties. Now, activation of 127 stopped and catalysis was turned OFF. Catalysis was turned back ON after trapping of the chloride ions with NaBArF.

The next example bridges in some way the gap between the preceding (Figure 23) and the present chapter because it addresses again catalysis with two catalytically active copper(I) centers held at a defined distance, now though for proximity catalysis using acyl transfer. The utility of this approach was demonstrated when a duo of catalysts was used for achieving substrate selectivity [133].

In detail, nanoswitch [Cu(130)]⁺ was transformed into the slow rotor [Cu₂(130)]²⁺ (k₂₉₈ = 1.34 s⁻¹) upon the addition of a second equiv of copper(I) ions (Figure 30). When 0.5 equiv of iron(II) was added, the rotator arm got involved in iron(II) bis(terpyridine) complexation affording [Fe(Cu₂(130))₂]⁶⁺. An analogous transformation was seen for [Cu(131)]⁺ → [Cu₂(131)]²⁺ → [Fe(Cu₂(131))₂]⁶⁺.

When hydroxymethylpyridine 132 and the acetylation agent 133 (Figure 30) were reacted in the presence of the copper(I)-loaded rotor [Cu₂(130)]²⁺ no reaction was detected. In contrast, the dimeric complex [Fe(Cu₂(130))₂]⁶⁺, formed after addition of iron(II) ions, offers two cavities each with two copper(I) ions to pre-bind both 132 and 133 at an optimal reaction distance. It was no surprise, that due to the increased local concentration, the acetylation 132 + 133 → 134 did rapidly take place. Similarly, the bigger dimeric nanoswitch [Fe(Cu₂(131))₂]⁶⁺ catalyzed the acetylation of the larger substrate 136 due to size-matching [133]. In order to test the reversibility, both switches were toggled by adding and removing the iron(II) ions over 2.5 cycles demonstrating that ON/OFF catalysis was reversible and reproducible.

Interestingly, the nanoswitches [Cu₂(130)]²⁺ and [Cu₂(131)]²⁺ (1:1) could be selectively addressed in the presence of all substrates and reagents (132, 133, and 136). Upon addition of substoichiometric amounts of iron(II), predominantly nanoswitch [Cu₂(131)]²⁺ reacted to afford [Fe(Cu₂(131))₂]⁶⁺ which selectively turned on the reaction affording 137. As a result, this example highlights a sophisticated case of a multicatalyst system that can select between substrates of essentially identical reactivity but different size.

A spectacular and unique example of toggling catalysis was demonstrated in nanoswitch 138 (Figure 31) with its four distinct switching states [134]. In State-I, i.e. [Cu(138)]⁺, an intramolecular HETTAP complex between a copper(I) phenanthroline and a terpyridine site was realized. Upon addition of 0.5 equiv of iron(II) ions the HETTAP interaction opened up and the “dimeric” bishomoleptic nanoswitch [Fe(Cu(138))₂]⁴⁺ was furnished in State-II held together by a bis(tpy) iron(II) complexation. Due to the eradication of the HETTAP binding, the copper(I) ions were left coordinatively frustrated, which made them potentially available for catalysis of a click reaction. In the next switching step, the removal of copper(I) ion led to State-III, i.e. [Fe(138)₂]²⁺. The ensuing removal of iron(II) ions broke down the bis(tpy) complex [Fe(138)₂]²⁺ and afforded State-IV. The latter state was characterized by a closed structure of the nanoswitch 138 due to the intramolecular N_pym → ZnPor binding (pym = pyrimidine). The full switching cycle using addition/removal of ions was repeated without fatigue.

The above sequence of four switching states has been demonstrated to activate sequential catalysis (Figure 32), once the protocol is executed in the presence of piperidine (139) as a catalyst and the reactants 79, 80, and 95 [134]. Due to the “free” ZnPor unit in State I, catalyst 139 was strongly bound at the ZnPor unit of [Cu(138)]⁺, while the copper(I) ion was firmly encapsulated in a HETTAP binding site. Thus, State-I should be catalytically inactive. Upon addition of iron(II), the tpy unit of the HETTAP unit became involved in the bishomoleptic iron(II) terpyridine complex [Fe(Cu(138))₂]⁴⁺, while the copper(I) ions were exposed for catalysis in State-II. Indeed, in this state 50% of the click product 81 was formed. Removal of copper(I) and formation of State-III stopped the click catalysis. The ensuing removal of iron(II) ions finally afforded nanoswitch 138 (= State-IV). Due to the intramolecular linkage of the azaterpyridine to the ZnPor unit, the piperidine that was firmly bound in States I → III, was now released from the ZnPor unit. Remarkably, product 81 formed in State-II now underwent a catalyzed Michael addition to provide 140 in 28% yield. When the catalytic cycle was repeated, it fully reproduced the yields of the first cycle, thus demonstrating that a catalytic eleven-component machinery may work without destructive interference despite the large number of functional groups in the switches, reagents, and products [134].

The following example (Figure 33) does not reach the complexity of the above nanoswitch system, but it is remarkable as it involves a remote control of catalysis [135]. Here, the complex [Cu(141)]⁺ with its heteroleptic HETTAP binding site controlled the switching state of nanoswitch 142 by using fully reversible communication via ion signaling. State-I was characterized by a clean self-sorting of the copper(I) ions resulting in [Cu(141)]⁺ + switch 142. In the presence of N-methylpyrrolidine (89), a conjugate addition was observed in State-I. However, upon addition of iron(II) (0.5 equiv) ligand 141 became involved in the bishomoleptic iron(II) terpyridine complex [Fe(141)₂]²⁺ with the effect that now weakly bound copper(I) ions travelled to nanoswitch 142. In the thus formed complex [Cu(142)]⁺ the azabipyridine dissociated from the ZnPor binding site to generate a HETPHEN complexation site for copper(I). With the ZnPor unit being freed from intramolecular binding, it now served as coordination site for the N-methylpyrrolidine (89), preventing the latter from catalyzing the conjugate addition [135]. As a result, catalysis was stopped in State-II. The cycle could be repeated two more times, however, generating a lower yield (Figure 33). The reduction of the yield in cycles two and three was traced to the slow switching of State-II to State-I.

In summary, this subchapter highlighted switching systems that were built using dynamic heteroleptic complexation together with supramolecular dissociation/association events for regulating catalysis. It is noteworthy that switching in all cases went along with a major nanomechanical reorganization of one (or all) constituents.

Conclusion

Supramolecular catalysis is an emerging field that has attracted a lot of attention recently. Its comprehensive coverage encompasses several fields [3], however, still with a main focus on catalysis that is promoted by discrete cages, capsules, and in a variety of confined environments. The present short account with a focus on dynamic heteroleptic metal complexation seeks to demonstrate that (metallo)supramolecular catalysis has many more opportunities to offer. While the examples in the first subchapter illustrate the much higher diversity of heteroleptic over homoleptic cages, the following subchapters above all exemplify that supramolecular chemistry extends beyond the single "supramolecular structure", as complex it may be, thus reaching far into the field and concepts of systems chemistry [30,136] and information science [137,138]. For instance, toggling between and dynamic exchange within supramolecular structures add features to catalysis as ON/OFF or UP/DOWN regulation of catalytic activity. If these factors will be consequently developed, they should eventually open opportunities to autonomous catalytic systems.

A look at systems biology, where enzymatic catalysis is tightly regulated by various chemical messengers to enable life, can serve as a guide. In this regard, we expect that upcoming endeavors will increasingly focus on the development of artificial systems that can make autonomous decisions and that energetically operate out-of-equilibrium.