Modulating the electron-transfer properties of a mixed-valence system through host–guest chemistry

Herein, we report that the interplay between the binding properties and redox activity of a self-assembled trinuclear RuII macrocycle leads to an hitherto unreported phenomenon, in which access to specific MV states is gated by host–guest chemistry.


Introduction
Thanks to their electrochemical and photophysical properties, oligonuclear Ru II -complexes have a wide variety of possible applications from abiotic light-harvesting to luminescent DNA binding substrates. [1][2][3][4][5][6][7][8] The electron transfer properties of such complexes have been extensively studied; particularly Ru III/II mixed valence, MV, systems. The prototype MV complex is the Creutz-Taube ion, which was rst reported over forty years ago. [9][10][11][12] The original reason for interest in this complex and its numerous analogues was that they provided testable experimental and theoretical models for many biological electron transfer processes. MV systems are still much studied, not least because they oen function as key components in a broad range of single-molecule devices. [13][14][15] In a handful of reports the effect of supramolecular interactions on the electronic interactions within MV states has been explored. Two studies have shown that supramolecular interactions between individual redox-active units and specic crown ethers can modulate electronic delocalisation between centres, 16,17 whilst the Das group has shown that encapsulation of a ligand bridge within a cyclodextrin can enhance electron transfer rates. 18 In separate research, metal-ion directed self-assembly has emerged as a versatile route to supramolecular architectures. [19][20][21][22][23] Much of this work has been aimed at new hosts for ionic and molecular guests. Although the metal ion is oen just a structural motif in the nal assembly, its inclusion can enhance the physical properties and functionality of the host, yielding assemblies that function as sensors for specic molecular guests. 24,25 Strikingly, whilst a considerable number of studies have investigated electro-active self-assembled macrocycles, virtually all this work has involved the redox properties of the organic components of such systems; 26 there are very few reports focusing on the electrochemistry of metal ion components. 27 Furthermore, despite the huge activity in this area, reports on metallomacrocycles containing ruthenium moieties are relatively rare, whilst only a handful of MV systems have been reported -reecting the kinetically inert nature of such centres. [28][29][30][31][32] To combine the electron transfer properties of oligonuclear ruthenium-based MV systems with the synthetic versatility of self-assembly, we have been investigating metallomacrocycles containing embed ruthenium units. Our approach has been to either exploit the "complex-as-ligand" concept [33][34][35] or to "labilize" inert Ru II centers. Using this latter method, we have used the Ru II ( [9]aneS 3 ) centre asfor a combination of steric and electronic reasonsthis moiety is labile at high temperatures, but is kinetically inert at room temperature. 36 We have demonstrated that this building block reacts with 9-methyladenine, 9MA, and other suitably hindered adenine derivatives to form metallomacrocycles such as 1 3+ , Fig. 1, which can be reversibly oxidized into three other oxidation states, two of which are mixed valence. 37,38 Most interestingly, due to the unusual connectivity of this macrocycle, it displays unique electronic properties: optical studies revealed that whilst the Ru II 2 Ru III MV state is an electron-hopping, valence-localized, Robin and Day 39 Class II system, the Ru II Ru III 3 state is valence-delocalized, Class III. Studies by the Severin group on kinetically labile [Ru(arene)]based neutral macrocycles have resulted in analogues of 12crown-3, that bind alkali metal ions in non-aqueous solvents with micromolar affinities. 40,41 Moreover, Bedford and Tucker 42 have shown that when the [9]aneS 3 ligand is coordinated to a cationic metal center it can recognize anions through C-H/X hydrogen bonding interactions. Given that 1 3+ is cationic and possesses an array of thiacrown-based hydrogen-bonding donor groups, we reasoned that it would be a receptor for anionic guests. Herein, we describe how recognition processes involving this metallomacrocycle modulate its electronic properties in a unique manner. In particular, we report on the rst MV system to display electron transfer properties that are modulated by host-guest chemistry.

Structural studies
We attempted to crystalize the macrocycle with a variety of anions in a number of different solvent systems, nally obtaining X-ray quality crystals of [1](Br) 3 - Fig. 2. The structure's asymmetric unit is composed of two 1 3+ cations (A and B) and six bromide counter-ions (for the ORTEP diagram, see the ESI †). In cation A two [9]aneS 3 ligands are disordered over alternative positions, while B only has one disordered thiacrown. Aside from these structural features, A and B are equivalent as illustrated by the bond lengths and angles ( Table  S2 in the ESI †). The metallomacrocycle has two possible binding pockets. An a pocket dened by the thiacrown ligands and the N-H binding sites from 9MA units, and a b pocket, dened by 9MA bridging ligands projecting out to give a bowl shape aromatic surface - Fig. 2A.
Pairs of bowls, related by a crystallographic inversion centre, create capsule-like structures in which two facing b pockets dene the capsule cavity. This structure is formed by an array of weak C-H/N hydrogen bondsee Fig. 2B. In the B-based capsules, there are three independent H/N interactions. While for the A-based capsules, two H/N interactions are observed, the third H/N distance of 2.82Å being longer than the sum of atomic van der Waals radii. A space-lling representation of the capsule, Fig. 2C, shows that its equator is dened by N-methyl groups. Notably, the two a pockets of individual bowls are occupied by bromide counter-ions. These two anions are held at a distance of 5.20, 5.43 and 5.62Å respectively from the three N-H binding sites of A-based capsule and 5.24, 5.32 and 5.64Å from the analogous residues of the B-based capsule. At higher levels, the capsules form alternating linear strands (see Fig. S2 in the ESI †). Given evidence of host-guest interactions in the solid state, the interaction of 1 3+ with anionic guests in MeCN solution was investigated.

NMR spectroscopic studies
Titrations reveal that addition of specic anions produce changes in the 9MA-based protons signals of [1](PF 6 ) 3 . Whilst no shis are seen for 9-methyl hydrogens, amino group protons (NH6) display downeld shiing as depicted in Fig. S3 † for the titration with TBACl salt (see ESI †); furthermore these shis are dependent on the guest. Whilst ClO 4 À induces a maximum downeld shi of 0.115 ppm, equivalent concentrations of halide ions produce much greater effects - Fig. 3.
The largest shiof 3.651 ppmis observed with uoride ion; these shis reect the high polarizing effect of uoride anion, indeed at higher mixing ratios stillat which  precipitation begins to occur and the NH6 proton signal considerably broadensit is clear that the host is at least partly deprotonated. Job plots using the NMR changes revealed that receptor 1 3+ binds to all the anionic guests investigated in a 1 : 1 ratio - Fig. S4 in the ESI. † Given this stoichiometry, and the pattern of shiing for the 9MA-based protons, we conclude that in solution anion guests are bound within the b pocket of the host. Support for this conclusion is also provided by large shis in thiacrown-based protons.
The changes for [9]aneS 3 protons are more complex with up and downeld shis being observed. The crystal structure data shows close contacts between some ethylene protons of the thiacrown ligands and anionic guests; the strength of this interaction is reected in the downeld shiing of these protons. However, a second effect also affects the thiacrownbased signals. Due to sterics, binding to this site will reduce the conformational exibility of the coordinated thiacrowns and concomitantly increase the rigidity of the receptor; this effect will be greatest for protons on the interior of the binding pocket. A close inspection conrms this hypothesis revealing thatdue to decreased uxionalityseveral multiplets split into simpler signals as shown in Fig. S5 † for chloride. Again, these effects are dependent on the nature of the guest: for perchlorate the biggest shi in thiacrown signals is around 0.15 ppm, while changes of almost 0.5 ppm are observed in titrations with iodide.
Using NH resonances shis to t to a standard 1 : 1 binding model, association constants were calculated - Table 1. The data reveal that 1 3+ binds to Cl À and Br À up to almost three orders of magnitude more strongly than other ions and that Cl À is bound with the highest affinity (K a > 10 5 M À1 ). The values for Cl À and Br À are likely to be lower limits asdue to the concentration regime employed -NMR titrations only provide accurate estimates of K a for weak or intermediate interactions.
Absorption spectroscopy-based titrations were not possible as no guest-induced little change in macrocycle spectrum was observed.

Theoretical studies
The computed electrostatic potential of the anion-free host mapped onto the molecular electron density surface, Fig. 4, reveals the a site clearly has the most positive electrostatic potential further indicating the a pocket is the preferred anion binding site. Therefore all the experimental and theoretical data indicate that anion binding at the b site can be discounted.
Calculated structures for individual halide binding in the a binding pocket are depicted in Fig. 5 and summarized in Table  2. The distances from the centre of mass, dened by the three nitrogen atoms of the N-H binding groups (COM N ) to uoride is quite large (5.949Å) indicating absence of any N-H/F À interactions. Only intramolecular N-H/S hydrogen bonds are present in the optimised structure. Likewise for I À , no intermolecular N-H/I À hydrogen bonds are observed; since the H/I À average distances (4.80Å) and corresponding N-H/I À angles (average ¼ 141 ) are too small compared to typical H/I À values. Indeed, the coordination geometry observed in the crystal structure of [1](Br) 3 indicates that the N-H moieties of all three 9MA bridging ligands cannot point to the anion  simultaneously; hence hydrogen bond angles cannot optimise, leading to weaker receptor-anion hydrogen bonding interactions. Over these effects lead to the iodine being displaced from the binding pocket by a COM N / distance of 5.318Å. In agreement with the experimental data showing a slightly higher binding affinity for iodide compared with uoride, this distance is shorter than the equivalent COM N /F À distance. Besides the fact that iodide is intrinsically a weaker hydrogen bond acceptor, it seems the size of the anion is too large for the metallomacrocyclic cavity, preventing stronger N-H/I À hydrogen bonds. In contrast to the iodide, the uoride anion is too small to complement the binding pocket of the macrocycle and thus a lower binding affinity for this anion is also observed.
The computational studies for chloride binding to the a pocket of the host indicate that N-H/Cl À hydrogen bonds are formed with an H/Cl À average distance and N-H/Cl À angle of 2.71Å and 147 , respectively. This corresponds to a COM N /Cl À value of 3.108Å and results in a concomitant weakening of the intramolecular N-H/S bonds. The calculated H/Cl À distances are typical of N-H/Cl À hydrogen bonds, although the N-H/Cl À angles are substantially lower than the ideal (180 ). Once again, this is due to the structure of the host preventing the three bridging ligand N-H groups from simultaneously pointing at the anion. However, clearly the size of chloride anion complements the host cavity size, leading to shorter distances and stronger host-guest interactions than those reported for other halide anions.
As mentioned above, the optimized geometry of 1 3+ with Br À is in agreement with the X-ray crystal structure, although the calculated COM N /Br À distance, 5.195Å, is shorter than the experimental value (5.911Å). However, in the crystal packing the bromide anion is "shared" by two 1 3+ adjacent capsule units (see ESI †), which is not the case in the calculated solution structure. Furthermore, in accordance with recognition based on the host-guest size, tting for the H/Br À distances are intermediate between the H/Cl À and H/I À distances.

Electrochemical studies
As outlined above, previous studies have revealed that electronic interaction in the two MV states of the macrocycle is not the same: the [Ru II 2 Ru III ] valence state (1 4+ ) is electron hopping, whilst the [Ru II Ru III 2 ] state (1 5+ ) electronically delocalized. 37 This is due to the distinctive molecular architecture of the macrocycle: as metal centres are connected through peripherally arranged bridging ligands, changes in the bonds and angles at one metal centre are mechanically coupled to the other two. Since binding to anionic guests oen leads to the anodic shiing in the oxidation of electroactive hosts and structural changes within the such receptors, 43-45 the effect of anion binding on the electrochemical properties of 1 3+ was then investigated.
The general changes induced by anion addition on the electrochemistry of 1 3+ are most clearly observed using square wave voltammetry and uoride as a non-redox-active guest - Fig 6A. As expected, on addition of uoride, all three Ru III/II -based oxidation potentials are shied anodically.  5 Optimized structures of 1 3+ with F À (cyan), Cl À (green), Br À (red), and I À (purple) for binding in the a pocket (left) or the b pocket (right). Binding scenario b is disfavoured relatively to a by 3.8 (F À ), 1.9 (Cl À ), 2.5 (Br À ) and 2.2 (I À ) kcal mol À1 respectively. The N-H/S and N-H/Cl À hydrogen bonds are drawn as red and orange dashed lines, respectively. Table 2 Relevant intermolecular distances (Å) and N-H/X angles ( , italics) between 1 3+ and the anions X ¼ F À , Cl À , Br À , and I À obtained from the DFT calculations Interestingly, the response of individual couples is not identical. Up to one equivalent of uoride causes the rst two oxidations to shi into each other, however they separate on further additions, resulting in a maximum DE p of 100-120 mV, when three equivalents of anion are added, Fig 6A, Table 3. Further addition of anion produced no additional shis in oxidation potential until precipitation of the host occurs. Despite analyses for other halide being complicated by the guests' intrinsic redox activity, very different effects were still delineated.
With chloride anions, shis in host oxidation potentials are virtually over at a 1 : 1 host : guest binding ratio - Fig 6B, additions of up to a further three equivalents of chloride ion only induce very small additional shis. This is consistent with the high chloride binding affinity of the macrocycle. A "shoulder" between the second and third oxidation of the host also grows in as chloride is added, this is assigned to the oxidation of chloride, which occurs at 1.08 V in these conditions. 46 It appears that this couple is broader and slightly anodically shied compared its free valuethis perturbation is likely due to the oxidation of chloride bound to the anionic host as this would be expected to shi in this way. More notably, although the rst two Ru II oxidations of the host are anodically shied by similar amountsaround 120 mVthe shi for the third oxidation is less half this magnitude.
Any analysis of host-based potential shis induced by bromide is greatly complicated by the fact that this anion is oxidized in two one-electron steps at 0.765 V and 1.065 V respectively. 47 So whilst the rst host oxidation is clearly dened being anodically shied by 65 mV compared to the free hostthe second oxidation is difficult to deconvolute from a bromidebased couple Fig. 6c. However, most strikingly, the third oxidation is clearly cathodically shied, suggesting a complex, guest-induced, redistribution of the host's electronic structure. [48][49][50] Iodide is also oxidized in two discrete one-electron processes. 47 Although the second of these process has almost exactly the same potential as the rst oxidation of free 1 3+ it is clear that, even at >5 guest equivalents, no host-based oxidation shis are observedsee Fig. S6 in the ESI. † Using this electrochemical data, the effects of halide guest binding on the comproportionation constants, K c , for 1 4+ and 1 5+ were estimated - Table 4 as K c values are a direct measure of the thermodynamic stability of individual MV states. 51 To aid comparison, K c values for the hexauorophosphate salt of the macrocycle in the same conditions are also included. Although the redox activity of bromide makes it difficult to make conclusion on this guest, chloride and uoride clearly destabilize the 1 4+ state. Contrastingly, whilst uoride and iodide have much less effect on 1 5+ , chloride induces a large stability increases. Furthermore, since potential shis for the third oxidation of the macrocycle are markers for the stability of 1 6+ , it is clear that this redox state is most stabilized by uoride guests, whilst the cathodic shi in the third couple induced by bromide suggests that 1 6+ is destabilized by any interaction with this anion.
The host's electrochemical response to anion binding is a product of its unique combination of properties. Since they are embedded into the macrocycle, oxidation of individual ruthenium(II) units modulate the host's entire structure. The properties of this redox chain can be likened to those of a dynamic combinatorial library, DCL, of host architectures. In a DCL, differential host-guest interactions thermodynamically select the "best" host for a specic guest within a chemically equilibrating mixture; [52][53][54] in the case of 1 3+ to 1 6+ , the interactions select for, and stabilize, the best host redox state. Given that uoride is a small ion with a high charge density it is not surprising that this guest stabilizes the 1 6+ oxidation state more   than the other halide ions. However, analysis of the data also clearly indicates that chloride and bromide ions select for 1 5+ presumably as this is the best match of size and charge density between the host and guest. Finally the electrochemical interaction with iodide is weak as the size and charge density of this guest is mismatched to the host's well-dened binding site. This distinctive combination of multiple oxidation states and host-guest chemistry means that 1 3+ functions as a novel ion-triggered device. Through spectroelectrochemistry using a OTTLE cell, 1 3+ was rst oxidized into its Ru II 2 Ru III MV state (1 4+ ) by holding it at a potential just under that required for oxidation into the Ru II Ru III 2 MV state (0.980 V), generating the previously reported 34,35 characteristic absorption spectrum with structured intervalance charge transfer (IVCT) bands in the NIR - Fig. 7.
Aer the addition of one equivalent of F À , chosen as it is not itself redox active, the solution was le for 30 minutes at the same potential. This allowed time for the anion guest to diffuse from the top of the cell to the electrode and for a new equilibrium between the electrode and oxidized product to be established. Strikingly, aer this period, the IVCT bands displayed bathochromic shiing and increases in intensity; with the thiacrown(S) / Ru III ligand-to-metal charge-transfer centered at $800 nm also growing in intensity. A comparison with previously reported data, generated in the absence of a guest but at a more positive potential, conrms that this nally generated spectrum is that of 1 5+ . These observations conrm an anion-triggered change of MV state without any change in potential. Although this effect means that the host could function as a optical sensor for anions, its response can also be viewed as the operation of a Boolean logic AND gate 55,56 where the two inputs are a potential difference and uoride anion, while the output is the large NIR optical change induced by increased electronic delocalization.

Conclusions
1 3+ displays selective binding to specic halide anions, which induce characteristic shis in the host's Ru II -based oxidation potentials. This facilitates a new phenomenon: ion-triggered change in redox states. This combination of self-assembly, host-guest chemistry, and redox activity provides the potential for the creation of a range of new molecular-scale devices. Since this host is kinetically robust, guest binding in a variety of solvents can be envisaged; its properties in water will be of particular interest; the magnitude of host-guest interactions are usually highly solvent sensitive. Therefore, contrary to conventional Class III systems, it may be possible to tune electronic delocalisation through solvent mixing. The host-guest chemistry of related structures are also currently underway and these studies will form the basis of future reports; in particular the possibility of reversible switching through decomplexation is being investigated.