Controlling the structure and photophysics of fluorophore dimers using multiple cucurbit[8]uril clampings†

A modular strategy has been employed to develop a new class of fluorescent molecules, which generates discrete, dimeric stacked fluorophores upon complexation with multiple cucurbit[8]uril macrocycles. The multiple constraints result in a “static” complex (remaining as a single entity for more than 30 ms) and facilitate fluorophore coupling in the ground state, showing a significant bathochromic shift in absorption and emission. This modular design is surprisingly applicable and flexible and has been validated through an investigation of nine different fluorophore cores ranging in size, shape, and geometric variation of their clamping modules. All fluorescent dimers evaluated can be photo-excited to atypical excimer-like states with elongated excited lifetimes (up to 37 ns) and substantially high quantum yields (up to 1). This strategy offers a straightforward preparation of discrete fluorophore dimers, providing promising model systems with explicitly stable dimeric structures and tunable photophysical features, which can be utilized to study various intermolecular processes.


Introduction
Coupling two uorophores within a sufficiently short distance for an extended period of time is crucial for both theoretical and experimental investigation of intermolecular processes such as charge transfer, 1 excimer formation, 2,3 long-or short-range exciton coupling, 4,5 and singlet ssion. [6][7][8] Stacking together precisely two uorophores in an aqueous solution, however, remains a substantial challenge as most aromatic hydrocarbons show a tendency to aggregate unpredictably (forming clusters of arbitrary numbers of molecules). [9][10][11] To prevent uorophores from aggregation in aqueous solution, a supramolecular approach has been established to "mechanically" separate uorescent molecules through encapsulation by macrocycles. [12][13][14][15][16] A popular class of macrocyclic hosts utilized for this purpose is cucurbit[n]uril (CB[n], n ¼ 5-8, 10), which contains a cavity that enables the inclusion of various guest molecules and exhibits particularly high affinity towards positivelycharged species. 17,18 As an example, CB [7] is a promising host for the complexation of various uorescent dyes, 19 resulting in signicant changes in photophysical properties such as antiphotobleaching 20 and emission enhancement. [21][22][23] This is attributed to the hydrophobic environment provided by the CB cavity as well as mechanical protection by the macrocycle against aggregation and quenching. 12 Dimeric uorophore stacking, however, is unlikely to be realized by CB [7]-mediated complexation as its relatively small cavity only allows the complexation with one single guest molecule or, more strictly speaking, one binding moiety on a guest molecule. On the other hand, CB [8], a larger cucurbituril homologue, is capable of simultaneously encapsulating two guest moieties yielding either a heteroternary 24 or homoternary complex. 25 Although CB [8]-mediated ternary complexation may achieve stacking of two uorophores, several limitations exist as the uorophores are required to have the right shape, size and charge distribution to undergo complexation with CB [8]. 26 Moreover, they must align along the principal symmetry axis of the CB cavity limiting the way in which they stack. 27 In case of a stepwise complexation of two guests with CB [8], formation of a dynamic ternary complex is evident by the signicant signal broadening in NMR spectra. 24,25 This dynamic complex results in a short-lived coupling between the two stacked uorophores that is insufficient to allow for the investigation of specic intermolecular processes.
Recently, we have found that the dynamic exchange kinetics between the guests and CB [8] hosts are dramatically reduced through the formation of 2 : 2 quaternary complexes, 28 in which two elongated guests such as diarylviologen derivatives are "clamped" in place by two CB [8] hosts into a multicomponent complex. The simultaneous formation of two ternary motifs within a discrete complex decreases the likelihood of dissociation compared to a typical ternary complex. 28 The formation of 2 : 2 complexes opposed to elongated supramolecular polymers requires a small change in the conformational entropy during complexation, i.e. a molecule with signicant rigidity. 28,29 For instance, various rigid molecular moieties such as benzidine, 30 benzothiazole, 31,32 arylpyridinium, 28,33 arylterpyridyl, 34 bipyridinium, 35 and benzimidazole 29 have been employed to produce CB [8]-mediated 2 : 2 complexes. Herein, we present a general and modular strategy towards the dimerization of arbitrary functional components (uorophores in this work) by connecting them to multiple rigid modules that can be "clamped" together by CB [8] complexation. We use arylpyridinium moieties as the rigid "clamping" module ( Fig. 1a) and exploit the modular strategy for designing uorescent complexes in water comprised of two uorophores that are stacked in a specic conguration with a constraint applied by CB [8] macrocycles at multiple points.
As illustrated in Fig. 1a, water-soluble uorescent molecules are designed to incorporate a uorescent core between two positively charged clamping modules, which in this work are arylpyridinium motifs originating from previously studied diarylviologen derivatives. 28 When one equivalent (equiv.) of CB [8] is added to the system with one equiv. of guest molecule, two clamping modules are expected to bring together two guest molecules yielding a 2 : 2 quaternary complex. The uorophore cores from each guest molecule are brought to close proximity to each other as a consequence of the assembly, resulting in preorganized dimeric uorophore stacking. The preorganized dimer complex is stabilized by multiple CB [8] clamps, which ensures interaction between uorophores for a sufficiently long period of time, endowing the complex with emergent photophysical properties. As the uorophore modules are not encapsulated by CB [8] (Fig. 1c), a variety of uorophores, including those with sizes substantially larger than the CB [8] cavity, can be employed as functional cores in this modular strategy (Scheme S1 †). Moreover, the photophysical properties of the resultant complexes can be readily customized through altering uorophores as well as the clamping modules. As exemplied in Fig. 1b, dimeric stacking still occurs even when the two clamping modules are non-parallel to each other (separated by an angle < 180 ). The exibility offered by this modular design provides a molecular toolbox and platform in which a wide range of uorophores can be readily studied in their discrete monomeric or dimeric states facilitating future investigations of quantum optical phenomena.

Results and discussion
Fluorescent molecules are designed by bridging two arylpyridinium motifs with a central uorophore core. Nine phenyl, naphthyl, or anthracenyl homologues are investigated as the uorophore cores in this study ( Fig. 1 and Scheme S1 †). The general synthesis (Scheme S2 †) of the molecules starts with Suzuki-Miyaura cross-coupling 36 of two pyridin-4-yl groups onto the uorophore core, followed by transformation of the pyridin-4-yl groups into arylpyridinium salts through a Zincke reaction. 37-39 A complete study was carried out on Ant910Me, which contains a 9,10-anthracenyl ("Ant910") as the central core and p- Modular strategy for designing fluorescent molecules (a) by plugging in a fluorophore module between two positively charged clamping modules, which herein are arylpyridinium moieties originated from diarylviologen derivatives. Following this strategy, (b) 2,7naphthalene with non-parallel clamping modules, as an example, or (c) other fluorophores with various arrangement of clamping modules are expected to form a preorganized dimer constrained by CB [8]-mediated multiple clamping or a monomeric state of fluorophores protected by CB [7] from aggregation. The Cl À counterions are omitted for clarity. tolyl pyridiniums ("Me") as clamping modules (Fig. 1a), which is presented here as a typical case prior to a further general discussion.
Guest molecule (G) Ant910Me is found to form 1 : 2 complexes with CB [7], denoted G 1 -CB [7]  CB proton signals ranging from 4 to 6 ppm split into two sets of equivalent doublets (Fig. 2b). The signal splitting suggests that the rate of CB [7] ipping around the tolyl moieties falls in the slow exchange limit with respect to the NMR time scale (500 MHz, 298 K). This slow ipping rate enables the direct observation of the two CB portals existing in distinctly different chemical environments. 22,28,39,41 Signal splitting is observed throughout the titration of the guest into a CB [7] solution, leading to quantitative splitting at a ratio of 1 : 2, which conrms the stoichiometric formula of this CB [7] complex as G 1 -CB [7] 2 . Thus, each Ant910Me molecule is readily isolated in a monomeric state in aqueous solution when complexed by two CB [7] macrocycles.
2.2 G 2 -CB[8] 2 : preorganized p-stacked dimers Upon titration of Ant910Me into a solution of CB [8] (Fig. 2c), splitting of the CB [8] protons are observed as well as the upeld shi of H d,e,f . Both observations suggest that the CB [8] molecules remain at the tolyl moieties, with a slow ipping rate and asymmetric portal environment. Careful analysis of the signal splitting and proton integration conrms a binding stoichiometry of "1 : 1", therefore, this CB[8]-mediated complex contains an equal number of hosts and guests. As elaborated in a previous work, 28 this complex cannot be a 1 : 1 binary complex as CB [8]-mediated binary complexes exhibit much faster dynamics. An elongated polymeric G n -CB [8] n complex (n ¼ 1,2,3.), fabricated from the sequential stacking of tolyl groups, 23 is also not possible as the head-to-tail alignment of two tolyl groups would result in a symmetric portal environment contrary to the observed splitting. Therefore, the most probable binding mode is a 2 : 2 complex (G 2 -CB [8] 2 ), as illustrated in Fig. 2c, where two uorescent molecules are constrained to overlap with each other. In this binding mode, the tolyl groups are head-to-head, thus resulting in an asymmetric portal environment for each CB [8]. The observed slow ipping rate of CB [8] and the signal splitting is explained by the tightly lled CB [8] cavities as well as the electrostatic interactions between multiple positive charges on one side of the CB portals.
We have learned from previous works 28,30,39,41,42 that the diffusion coefficient (D) of a CB[n]-mediated complex is primarily determined by the number of CB macrocycles existing in the complex. Therefore, the formation of G 2 -CB[8] 2 is further conrmed through a semi-quantitative analysis of D via DOSY experiments. As shown in Fig. 3a and Table S1, † D values of unbound guests (G) in aqueous solution range from 3.49 to 4.38, showing a standard deviation (SD) of 0.3. A much narrower distribution is observed for CB [8]-mediated complexes, ranging from 1.95 to 2.07 with a SD of 0.04 (Fig. 3a). These D values are much smaller than that of free CB [8] (D ¼ 3.11) and typical  binary complexes such as dzpy 1 -CB[8] 1 (D ¼ 3.04). 41 However, the D values measured here are almost the same as for the 2 : 2 complexes produced by diarylviologen derivatives 28,41 such as (VNMe 2 ) 2 -CB[8] 2 (D ¼ 2.01). Therefore, the DOSY data fully supports the formation of a G 2 -CB[8] 2 complex involving the stacking of two uorescent molecules held together by two CB [8] hosts.
The relative orientation of the two uorophores with respect to each other is probed by proton and NOESY NMR in this dimeric system. As the proton spectrum recorded for G 2 -CB[8] 2 (G ¼ Ant910Me) exhibits more complicated signal splittings than that of G 1 -CB[7] 2 , COSY NMR (Fig. S2 †) is used to identify each proton. Both the pyridinium and anthracenyl protons in this 2 : 2 complex split into two sets of equivalent peaks corresponding to H b,b 0 ,c,c 0 and H g,g 0 ,h,h 0 in Fig. 2c. The observation of two sets of signals suggests (i) a slow dynamic process and (ii) a certain asymmetry existing for the most probable conguration of the G 2 -CB[8] 2 complex, which is consistent with a cofacial stacking and partial overlap of the two aromatic uorophores as illustrated in Fig. 2c. Partial overlap of the two uorophores with a slippage along their extended axis will result in one set of equivalent protons lying on top of or below an aromatic ring of the other molecule, while the other set of equivalent protons does not. The rst set of equivalent protons are expected to display signals in a higher-eld region on account of shielding by aromatic ring currents, compared to the latter set of equivalent protons, which is consistent with the observation of signicantly lowered chemical shis for H g 0 ,h 0 compared to H g,h . Similarly, the difference observed between H c,b and H c 0 ,b 0 is interpreted as two sets of protons that reside in different shielding and deshielding environments arising from the CB [8] portal.
The partial overlap of two aromatic uorophores is also supported by the cross-correlation signals observed in NOESY NMR, which reveals the relative position of protons located in space. Proton H b 0 ,c 0 ( Fig. S3 †), for instance, exhibits an intense cross-correlation with all anthracenyl signals (H g,g 0 ,h,h 0 ), whereas H b,c can only "feel" protons that are closer to the pyridinium protons, i.e. H g,g 0 . This observation is consistent with the partial-overlap and stacking of the uorophores where H b 0 ,c 0 rather than H b,c are closer to H h,h 0 in space ( Fig. 2c and S3 †).

Photophysics of the dimeric and monomeric Ant910Me
The well-resolved NMR spectrum of G 2 -CB[8] 2 suggests that these complexes exist as discrete preorganized uorescent dimers in aqueous solution without forming any larger aggregates. This is because the two CB [8] macrocycles mechanically block the interaction between multiple dimers. It also leads to a substantial change in the photophysical properties of Ant910Me upon complexation with CB [8] to G 2 -CB[8] 2 . As shown in Fig. 4a-c and Table 1, the anthracenyl moiety in G 2 -CB[8] 2 exhibits a bathochromic shi of its absorption maximum (l abs ¼ 469 nm) by over 50 nm compared to monomeric Ant910Me (l abs ¼ 409 nm) in G 1 -CB [7] 2 and unbound Ant910Me in pristine solution (l abs ¼ 419 nm) (all compared at a concentration of 15 mM). The emission maximum of G 2 -CB[8] 2 (l em ¼ 578 nm) is also red-shied relative to that of G 1 -CB[7] 2 (l em ¼ 537 nm), although Ant910Me in pristine solution exhibits the most bathochromic shi in emission (l em ¼ 595 nm).
Aer photoexcitation, uorescence decay as well as the corresponding lifetime are recorded from time-correlated single photon counting (TCSPC) experiments, Fig. 4d-f. Excited anthracenyl dimers in the G 2 -CB[8] 2 complexes display an excimer-like state that exhibits a lifetime (s s ) of 12.6 ns, which is much longer than s s of 8.6 ns for its monomeric counterpart in G 1 -CB[7] 2 . A biexponential decay is observed for Ant910Me in pristine solution measured at the same concentration (15 mM) as that in G 1 -CB[7] 2 and G 2 -CB[8] 2 , showing 96% of the intensity is due to a short-lived component of 1.66 ns and 4% arising from a long-lived component of 9.50 ns.
Fluorescence quantum yields (f F ) for each species are measured by an absolute method using an integrating sphere. Aqueous solution of each species with a uniform guest concentration of 15 mM is tested at 298 K. The intensity is not normalized but scaled up by the same factor except the emission of G which is enlarged by an additional 4 times for a clear vision. Quantified data can be found in Table 1. Photographs of each species with a guest concentration of 20 mM before (g) and after (h) photoexcitation at 365 nm.
The emission of Ant910Me is signicantly quenched in a pristine solution with a f F of 0.04, whereas its uorescence intensity is dramatically enhanced upon complexation with either CB [7] or CB [8], showing a f F of 0.85 for G 1 -CB [7] 2 and a f F of 0.81 for Table 1).
The negligible quantum yield, bimodal decay, and red-shied emission in a pristine solution of Ant910Me all suggest a certain extent of aggregation in aqueous solution, whose photophysical properties are highly concentration-dependent. On the other hand, complexation with CB [7] or CB [8] ensures a dispersion of discrete uorophores in solution in either monomeric or dimeric fashion, respectively. It is known that the polarity of CB cavities is lower than that of water, which will also affect photophysical properties of dye molecules. 19,43 Therefore, in the following discussion, a comparison is made between CB[7]-and CB [8]-mediated complexes on account of their similar cavity polarities. A comparison between G 1 -CB[7] 2 and G 2 -CB[8] 2 of Ant910Me shows that the stacking of anthracenyl moieties as a dimer, relative to monomer, exhibits (1) a signicant bathochromic shi in absorption and emission, (2) an elongated excited-state lifetime, and (3) comparably high uorescence efficiency.

Applicability and exibility of the modular strategy
Although individual cases have demonstrated photophysical changes upon complexation with CB [8], 31,35,39,[44][45][46][47] the beauty and power of this work stems from the simple modular design. As illustrated in Fig. 1a, any selected uorophore can be readily inserted between clamping modules, resulting in its monomeric or dimeric species through complexation with CB [7] or CB [8], respectively. Following this design strategy, a further eight uorescent molecules were successfully synthesized, with similar topology to Ant910Me but with systematic variation in their structures. For example, the uorophore cores are augmented between phenyl, naphthyl, and anthracenyl. Alternatively, the alignment between the two clamping modules is altered. While several derivatives exhibit both clamping modules in-line with one another (Ph14Me, Np14Me, Ant910Me, and Ant14Me) others have clamping modules that are not in-line but remain parallel to each other such as Np15Me and Ant15Me, or are no longer aligned in a parallel manner but with an angle < 180 (Ph13Me and Np27Me). Finally, one can readily add additional clamping motifs around the uorophore core moiety, as demonstrated in the triply clamped systems Ph135Me and Ant14Me.
Results from 1 H NMR and DOSY, as shown in Fig. 3a and in the ESI (Table S1 and Fig. S1-S17 †), demonstrates that these uorescent molecules all perform in a manner similar to Ant910Me. Despite their structural variation, they all generate a monomeric uorophore in the presence of CB [7] and dimeric stacking of uorophores with CB [8]. As CB [7] and CB [8] only bind the clamping modules (i.e. tolyl pyridinium moieties), choice of the uorophores is no longer limited by the size and shape of the macrocycle cavities. Large uorophores such as anthracenyl derivatives, which to date have only been shown to complex CB [7] or CB [8] along their principal symmetry axis, are easily incorporated using this strategy regardless of their substitution pattern. Moreover, small aromatic rings like phenyl moieties, whose binding is extremely dynamic inside a single CB [8] cavity, are now readily immobilized and constrained within a 2 : 2 complex.

Photophysical properties
In terms of photophysical properties, most uorescent molecules also behave similarly to Ant910Me, with the exception of a few outliers that are discussed later in detail.
Preorganized ground-state dimers are readily produced by the formation of G 2 -CB[8] 2 in aqueous solution, corresponding to a considerable bathochromic shi in the absorption band ( Fig. 5 and Table 1). An excimer-like emission with a broadened and structureless prole is observed for all uorophores in their G 2 -CB[8] 2 systems, exhibiting a red-shi in their emission maximum relative to their monomeric form in G 1 -CB[7] 2 systems. Solutions of G 2 -CB[8] 2 compared to their G 1 -CB[7] 2 counterparts exhibit a smaller rate constant for non-radiative deactivation (k nr ), which corresponds to their observed elongated excited-state lifetime as well as comparably high quantum yields. Molar absorption coefficients for all uorophores are slightly increased upon complexation with either CB [7] or CB [8] ( Table 1) along with their high quantum yield, leading to reasonably high brightness (3 Â f F ) in aqueous solution. 48 Considering their long uorescence lifetimes, G 2 -CB[8] 2 complexes in general should be promising candidates for timegated imaging for biological systems.  Table 1.
This journal is © The Royal Society of Chemistry 2020 Fluorescent molecules that contain anthracenyl cores all exhibit a low quantum yield in solution with a lifetime much shorter than that in their CB [7]-or CB [8]-mediated complexes. The discrete monomeric species of Ant910Me, Ant15Me, and Ant14Me in their corresponding G 1 -CB [7] 2 complexes all display a lifetime around 8 ns (Table 1), similar to the lifetime (s s ) of 9,10-diphenylanthracene (DPA), a typical anthracenyl standard. 49 This recovery of lifetime to a value similar to DPA implies that the uorescence of these CB [7] complexes is mainly contributed by their anthracenyl cores. In a solution of only free molecules (without the presence of CB), the excited anthracenyl cores are deactivated through certain pathways as evidence from the observed quenching of uorescence. In particular, a typical deactivation pathway would be the photoinduced electron transfer (PET) from the anthracenyl core to p-decient pyridinium moieties. 50,51 However, the signicant recovery of emission aer complexation suggests that these deactivation pathways are forbidden or are at least largely restricted in both the G 1 -CB[7] 2 and G 2 -CB[8] 2 complexes. Quantum yields of the naphthyl and phenyl species are generally large (0.9-1.0 for Np, 0.6-1.0 for Ph) contrary to anthracenyl analogues, regardless of complexation, implying that PET from these two uorophore cores to pyridinium moieties is not efficient.
Systematic variation in the alignment between the clamping modules also affects their photophysical properties. Np15Me and Ant15Me, with two parallel clamping modules that are not aligned, exhibit a red-shi in emission, which is not as large as for other species (Fig. 5) upon forming G 2 -CB[8] 2 complexes. This non-aligned connectivity may force the two uorophores to stack in a less J aggregate-like fashion. 52 When the clamping modules are non-parallel, the G 2 -CB[8] 2 of Np27Me displays a quantum yield of 0.55, which is almost half the value of the other naphthyl homologues (Table 1). However, this species exhibits a distinctively long-lived excited state with a s s up to 37 ns. Similar results are observed in CB [8]-mediated complexes of Ph13Me, which also possesses non-parallel clamping modules. The reduced uorescence efficiency along with the elongated lifetime suggests that dimeric stacking in species with nonparallel clamping units may signicantly suppress radiative pathways (i.e. see reduced k r values in Table 1).

Triple clamping
Ph135Me is a more complex version of non-parallel clamping, which forms dimeric stacks through triple clamping, denoted G 2 -CB[8] 3 (Fig. S16 †). Triple non-parallel clamping leads to a further decrease in k r (  (Fig. S17 †). Instead, Ph135Me molecules exist either as G 2 -CB[8] 3 complexes or as a free guest in aqueous solution.
In addition to Ph135Me, which has three uniform clamping modules, Ant14Me with a protruding uorophore core is also able to form a G 2 -CB[8] 3 complex. Isothermal titration calorimetry and UV-Vis titration both conrm a binding stoichiometry of 2 : 3 (Fig. S12 †). Its diffusion coefficient from DOSY NMR gives a D value similar to that of Ph135Me 2 -CB[8] 3 (Table  S1, Fig. S11 and S16 †). Considering its T-shape topology, a third CB [8] in the Ant14Me G 2 -CB[8] 3 complex binds with the two protruding, stacked anthracenyl cores. However, in contrast to Ph135Me, CB [8] complexation of Ant14Me (in excess) does not exhibit a self-sorting behavior. Addition of extra Ant14Me guest molecules gradually transforms the solution of G 2 -CB[8] 3 into 2 : 2 complexes, in which two CB [8] macrocycles are bound with the two tolyl pyridinium moieties rather than the protruding anthracenyl cores (Fig. S10-S12 †). This suggests that the affinity of CB [8] around the protruded binding site is substantially weaker than its binding with the clamping modules, which is conrmed by the ITC result in Fig. S12. †

Restricted intracomplex motion
Dimeric uorophore stacking in G 2 -CB[8] 2 complexes generally exhibit an enhanced uorescence efficiency, particularly in the case of employing anthracenyl motifs as cores. This observation implies that motion within the complex (intracomplex motion) of G 2 -CB[8] 2 is extremely retarded and restricted, thus effectively suppressing deactivation pathways.
2.7.1 Interconversion dynamics quantied by VT-NMR. The NOESY spectrum of G 2 -CB[8] 2 (Fig. S3 †) shows that crosscorrelations between H g and H g 0 as well as those between H h and H h 0 are much more intense than correlations caused by 3 J H-H coupling for H g -H h and H g 0 -H h 0 . As chemical exchange also contributes to NOESY signals, this observation implies the presence of a dynamic interconversion between two discrete states within the CB [8] complex. This is also the reason why anthracenyl and pyridinium proton signals in G 2 -CB[8] 2 split into two sets of equally intense peaks (Fig. 2c) that are not observed in G 1 -CB[7] 2 (Fig. 2b).
Interconversion between the two states is further conrmed and quantied by variable-temperature nuclear magnetic resonance spectroscopy (VT-NMR). As shown in Fig. 6a, four signals of H g,g 0 ,h,h 0 that correspond to two stacked anthracenyl cores broaden equally until coalescence is observed as the temperature rises from 278.6 K to 307.5 K on a high-eld NMR spectrometer (500 MHz). A subsequent increase of temperature from 306.2 K to 362.6 K on a low-eld NMR spectrometer (200 MHz) (Fig. 6b) leads to a gradual merging of the four signals into two broad peaks, which later become sharper as the temperature increases. The transition of the H g,g 0 ,h,h 0 signals from the slow exchange limit to the fast exchange limit conrms the existence of a dynamic interconversion between two discrete states for the anthracenyl pair. By analysing the temperature-dependent linebroadening in the slow exchange limit [54][55][56] (Fig. S19 †), an activation energy of 43 kJ mol À1 is obtained for this interconversion.
As lowering the magnetic eld is equivalent to severely heating the sample, the switch of VT-NMR from high-eld to low-eld enables us to witness and quantify a very slow exchange process such as that for the pyridinium pair in this study. In the spectra recorded by the high-eld spectrometer (Fig. 6a), no signicant line broadening is observed for pyridinium signals (H b,b 0 ,c,c 0 ), whereas in the low-eld VT-NMR, the signal broadening corresponding to an exchange in the slow limit is readily observed upon increase in temperature (Fig. 6b). The temperature-dependent signal broadening suggests an activation energy as large as 83 kJ mol À1 (Fig. S19 †), implying a relatively slow interconversion of the pyridinium pair within the complex. The interconversion of protons in the stacked tolyl pair is already displayed in the fast exchange limit as demonstrated by the line sharpening of the H d signal as the temperature is increased in the high-eld spectrometer, exhibiting a relatively small exchange barrier of 22 kJ mol À1 (Fig. 6a and S19 †).
2.7.2 Intracomplex motion restricted in constrained dimers. Despite the covalent bonds between the anthracenyl, pyridinium, and tolyl moieties, different interconversion barriers are observed from VT-NMR indicating three separate dynamic processes. Therefore, these three distinct processes cannot be attributed to either the back and forth shuffling of the two uorophores along the long axis of the complex or to the complexation/decomplexation process with CB [8] because these two processes require a simultaneous movement of all components at the same rate yielding uniform activation energies. Moreover, complexation/decomplexation must be slower than all three dynamic processes observed. This indicates that the G 2 -CB[8] 2 complex is fairly "static" in aqueous solution and must remain complexed longer than the dynamics for pyridinium interconversion, which is around 30 ms at room temperature.
Molecular dynamic (MD) simulations of the Ant910Me 2 -CB[8] 2 complex in a cubic water box with 4000 H 2 O molecules was carried out in order to evaluate the stability of this complex under ambient conditions (298 K, 1 atm). The simulations indicate that the Ant910Me 2 -CB[8] 2 complex (ESI † video media) remains as a single entity during the whole simulation period (>200 ns) without decomplexation or signicantly altering its structure. This result is consistent with the analysis by NMR, which shows that the dimeric stack of uorophores is constrained and stabilized by the CB[8]-mediated dual clamping. Moreover, the two stacked aromatic moieties, such as anthracenyl units, partially overlap one another and simultaneously rotate in a slow but coherent fashion during the MD simulation. For example, the two anthracenyl units (yellow) of the Ant910Me 2 -CB[8] 2 complex in Fig. 6 must rotate or swing around the central axis of the complex in a coupled manner, which we refer to here as intracomplex motion. Thus, the activation energy obtained represents the energy barrier for each intracomplex rotation.
The height of the energy barrier reects the steric hindrance present around the "rotor". As exemplied by Ant910Me in Fig. 7, the rotation of the anthracenyl group is hindered by the presence of several pairs of adjacent protons between the anthracenyl core and pyridinium units. The rotation of the  This journal is © The Royal Society of Chemistry 2020 pyridinium moieties, in addition to steric hindrance from the anthracenyl core, are also impeded by the CB [8] portals, thus showing the highest rotational barrier. On the other hand, rotation of the tolyl groups is not signicantly inuenced by CB portals as they mainly reside within the CB [8] cavities. Their motion is also not signicantly retarded by neighboring pyridinium protons, which present less steric clash than those between the anthracenyl and pyridinium units. Therefore, the tolyl groups exhibit the lowest rotational barrier and their proton signals always fall within the fast exchange limit (Fig. 6). It is worth noting that "rotation" does not necessarily refer to a full rotation. In the case of Ant910Me, it is more likely that the two anthracenyl groups swing coherently within a limited angle on account of van der Waals repulsion (Fig. 6c), where the activation energy represents the steric hindrance for swinging between two degenerate states.

Steric hindrance amplied in G 2 -CB[8] 2 complexes.
It is worth highlighting that steric hindrance between adjacent aromatic moieties is signicantly amplied in the stacked dimers (G 2 -CB [8] 2 ), which exhibit substantially slower dynamics than their monomeric counterparts (G 1 -CB [7] 2 ). As shown in Fig. 7, the uorophore and pyridinium moieties within the three molecules Ant910Me, Np14Me, and Ph13Me should experience a different degree of steric hindrance consistent with the number of clashing, neighboring protons. However, this difference is not observed in their monomeric forms (G 1 -CB [7] 2 ) where proton signals attributed to pyridinium (Fig. 7b) and uorophore units (Fig. 7d) are all sharp and are not split indicative of dynamics within the fast exchange limit.
In contrast, the dimeric complexes of these molecules (G 2 -CB [8] 2 ) display signicant differences in both their uorophore and pyridinium components in their 1 H NMR spectra. As shown in Fig. 7a and c, proton signals of Ph13Me exhibit a narrow linewidth in the fast exchange limit for both the pyridinium and 1,3-phenyl groups, which is consistent with the fact that no severe steric clash exists in this molecule. However, signal broadening occurs with an increase of steric repulsion in the dimeric complex of Np14Me. Furthermore, proton signals from the anthracenyl and pyridinium groups in the dimeric complex of Ant910Me both fall into a slow exchange limit and split into two sets of peaks, which corresponds to much slower intracomplex motions. This observation stems from a further increase in steric hindrance and is amplied for the G 2 -CB[8] 2 complexes as rotation of one moiety is not only retarded by covalently linked "neighbours" but also hindered by adjacent groups on the other stacked molecule. Careful comparison between the monomeric and dimeric systems veries that formation of a constrained system largely restricts and slows down intracomplex motions in these dimers.

2.8
Ground and excited states of p-stacked dimers 2.8.1 Preorganized p-stacked ground-state dimer. The characteristic red-shi in emission and elongated excited-state lifetime ( Fig. 4c and Table 1) suggest the formation of an excimer-like state for G 2 -CB[8] 2 upon photoexcitation. However, the formation of the excimer-like state in G 2 -CB[8] 2 complexes is quite different from those formed by pyrene derivatives or covalently linked pseudo-dimers. 2,48,57 In such cases, the generation and decay of an excimer or excimer-like state involves the excitation of one single uorophore followed by a diffusion-controlled interaction with a second ground-state uorophore and ends up with relaxation towards the ground state. 58 Therefore, the absorption band is oen similar to that of a monomeric uorophore as the excitation is rstly applied to a single molecule. 2 In the case of G 2 -CB[8] 2 (Fig. 5), however, a considerable bathochromic shi is generally observed in its steady-state absorption spectrum. Particularly, the vibronic progression is absent in the absorption of Ant910Me 2 -CB[8] 2 (Fig. 4c) indicating a strong coupling and effective delocalization of p-electrons between the dimeric anthracenyl moieties at their ground states. This preorganized p-stacked ground-state dimer is excited as a precoupled entity to an excimer-like state, which is different from an excited monomer and, more importantly, does not require an additional diffusion-controlled process aer photoexcitation. On the other hand, the excited G 2 -CB[8] 2 complex will not exhibit an energy dissipation as signicantly as during the formation of conventional excimers. This explains why Ant910Me in the G 2 -CB[8] 2 complex exhibits a Stokes shi (wavenumber difference between l abs and l em ) of 4012 cm À1 (109 nm) smaller than the value of 5828 cm À1 (128 nm) in its G 1 -CB[7] 2 complex (Table 1). Due to the absence of diffusioncontrolled steps in their excited state, one expects a monoexponential uorescence decay at pico-and nano-second timescale for G 2 -CB[8] 2 complexes aer photoexcitation, contrary to the bimodal decay of conventional excimers. 58 2.8.2 Mono-exponential decay of excited G 2 -CB[8] 2 . A mono-exponential uorescence decay is indeed observed for the Ant910Me 2 -CB[8] 2 complex in TCSPC measurements ( Fig. 4 and Table 1) and is further validated by time-resolved spectroscopies.
Femtosecond (fsTA) and nanosecond (nsTA) transient absorption were employed to monitor the dynamic relaxation of both G 2 -CB[8] 2 and G 1 -CB[7] 2 complexes of Ant910Me aer photoexcitation ( Fig. S20-S25, Table S2 †). As shown in Fig. 8a and b, two species are clearly detected in the excited state from fsTA for Ant910Me 2 -CB[8] 2 . Both exhibit a spectral feature of ground state bleaching (GSB) from 431 nm to 498 nm overlapping with an excited state absorption (ESA) from 431 nm to 800 nm and a stimulated emission (SE) from 573 nm to 654 nm. Upon photoexcitation, the rst species A relaxes to species B with a fairly short lifetime of 3.6 AE 0.3 ps (Fig. 8c, d and S21 †), and then back to its ground state with a lifetime of 12.9 AE 0.4 ns (Fig. 8e and f), consistent with the value of 12.6 ns measured from TCSPC (Table 1). Species B exhibits a similar ESA prole as species A except a slight red-shi in its absorption maximum (Fig. 8b), which suggests that the evolution from A to B with a picosecond time constant probably corresponds to excited state solvation. 59 In addition to solvation, the excited complex relaxes back to its ground state in a mono-exponential manner without observing other competitive pathways. It is worth mentioning that the excited state absorption spectra of Ant910Me 2 -CB[8] 2 (Fig. 8a) are quite broad suggesting a strong coupling also existing in the excited states.
Ant910Me 1 -CB[7] 2 aer photoexcitation (Fig. S23 †) exhibits an absorption maximum at around 475 nm in its ESA prole, which is much smaller than that of Ant910Me 2 -CB[8] 2 at around 530 nm (Fig. 8b). This observation conrms that G 2 -CB[8] 2 complexes are directly pumped up to the excited state of precoupled dimers rather than the excited state of monomers. It is worth noticing that several species are detected (Fig. S23 and S24 †) for Ant910Me 1 -CB[7] 2 upon photoexcitation, which also includes a small amount of long-lived species with a time constant of 6 AE 1 ms. This long-lived species should come from a triplet state, as its lifetime increases signicantly (>340 ms, Fig. S25 †) aer removal of oxygen from the solvent. The observed rich dynamic processes implies that aer photoexcitation the uorescent molecule within the G 1 -CB[7] 2 complex has sufficient structural freedom to relax to various low-energy excited states. On the other hand, the singular excited state dynamic observed for G 2 -CB[8] 2 complex suggests a restricted or retarded structural change even in its excited state, which may affect both the radiative and non-radiative pathways.
2.8.3 Enhanced uorescence efficiencies from constrained and discrete excited dimers. The rate constants corresponding to non-radiative (k nr ) and radiative (k r ) pathways are readily calculated from excited-state lifetime (s s ) and quantum yield (f F ) values. As shown in Table 1, the formation of preorganized G 2 -CB[8] 2 dimers always results in a reduced radiative rate constant k r corresponding to a long-lived excimer-like excited state, which is smaller than the k r values for the corresponding G 1 -CB[7] 2 complexes. Contrary to typical excimers that lead to quenched emission, 58 G 2 -CB[8] 2 complexes maintain high uorescence efficiencies on account of their substantial reduction in non-radiative rate constants k nr . This unique feature is attributed to a signicant suppression of non-radiative deactivation through the formation of a preorganized dimer in G 2 -CB[8] 2 , which strongly restricts intracomplex motions as demonstrated above.
In addition to constrained complexation, the discrete nature of uorophore dimers is also crucial to ensure high-efficiency uorescence. 60,61 The two CB [8] macrocycles that hold the uorophore dimer together will mechanically block interactions from other dimers in aqueous solution, which effectively avoids the generation of dark excited states caused by arbitrary aggregation.
2.8.4 Comparison with other dimeric systems. An advantage of forming preorganized p-stacked ground-state dimers is that the excitation wavelength for the system is shied towards the visible region (e.g. 469 nm for the G 2 -CB[8] 2 complex of Ant910Me), which is crucial for non-destructive imaging of biological systems. Importantly, the formation of a preorganized p-stacked ground-state dimer is not necessarily the same as bringing together two uorophores into spatial proximity. For instance, a red-shi in the absorption band was not observed in previous reports where two uorophores have been covalently linked together in close proximity. 2,48,57 The preorganization of p-stacked dimers of anthracene and its derivatives through non-covalent methods have been previously realized in rigid media containing small discrete cavities, such as crystalline lattices 60-63 and supramolecular capsules, 64 suggesting that the formation of a preformed p-stacked dimer requires strict spatial connement in order to: (1) isolate each dimer as a discrete entity, (2) maintain a specic p-stacked conguration, and (3) restrict interplanar spacing between the two uorophores.
The spontaneously assembled G 2 -CB[8] 2 complex satises all three requirements and facilitates the formation of preorganized p-stacked dimers. The two uorophores inside a G 2 -CB[8] 2 complex form a discrete dimeric stack with a signicant overlap of p electrons and a restricted interplanar spacing dened by the CB [8] cavities. Steric hindrance from both CB [8] macrocycles facilitates "mechanical" separation between all dimers in aqueous solution ensuring pairwise uorophores perform as a discrete entity. More importantly, the dimers are stabilized by CB [8] clamping and remain as such for a sufficiently long period of time. Finally, discrete preorganized dimers can be readily obtained through our strategy in aqueous solution at ambient temperatures, and therefore do not require formation of a specic crystal 60,61 or crystalline solvent at extremely low temperature. 62,63 Moreover, owing our modular design, a variety of uorophores are incorporated to give the corresponding p- stacked dimers without any limitation on uorophore size in direct contrast to other methods. 64 2.9 Controlling photophysics by clamping modules 2.9.1 Suppression of radiative deactivation through nonparallel clamping. Complexation enhanced uorescence is trivial for naphthyl-based guest molecules on account of their intrinsically high uorescence efficiencies, whose quantum yield is almost unity even without complexation. An exception is Np27Me whose G 2 -CB[8] 2 complex exhibits a quantum yield of 0.55, which is about 40% less than its G 1 -CB [7] 2 complex or in a non-complexed solution, and much smaller than the G 2 -CB[8] 2 complexes of other naphthyl homologues (Table 1). A reduction in quantum yield is accompanied by a dramatic decrease in the radiative rate constant, which, in turn, results in the longest uorescence lifetime observed for any species in this study of up to 36.8 ns. A similar suppression in the radiative pathway is also observed for the G 2 -CB[8] 2 complex of Ph13Me and the G 2 -CB[8] 3 complex of Ph135Me, both of which exhibit a decreased quantum yield and an elongated lifetime. All three of these uorescent molecules employ a non-parallel arrangement between their clamping modules, which suggests that non-parallel arrangements suppress the decay through radiative pathways. Clamping the dimer together in a non-parallel manner prevents any slippage of the two uorophores along their extended axis and strongly restricts any intracomplex motions. As a consequence, the structural relaxation of the complex aer photo-excitation towards a low-energy excited state is further retarded due to conformational rigidity amplied by non-parallel clamping. This restriction of motion is even more signicant in triple-clamping cases, such as Ph135Me whose k r and f F values are reduced compared to Ph13Me.
Radiative decay is practically prohibited in the case of Ant14Me when it is complexed with CB [8], either by dual clamping or by triple clamping, exhibiting negligible quantum yield in either case (Table 1). Fluorescence quenching in the G 2 -CB[8] 3 complex of Ant14Me may be readily explained by triple clamping, however, it does not explain why complete quenching is also observed for its G 2 -CB[8] 2 counterpart. One hypothesis is that the protruding anthracenyl moieties in one dimer may be long enough to interact with other protruding anthracenyl pairs located in another dimer, leading to some radiationless decay pathways that quickly deactivate the excited state. Interactions between protruding anthracenyl moieties is also supported by the diminished quantum yield observed for its G 1 -CB [7] 2 complex compared to that of other anthracenyl homologues. Another possibility leading to radiationless decay may be a transition from a singlet to a triplet state through intersystem crossing, however, this requires substantial further study of the dynamics of Ant14Me complexes in their excited states.
2.9.2 H-H and H-T stacking of the uorophore dimer. When the G 2 -CB[8] 2 complex contains non-parallel clamping modules (e.g. Np27Me, Ph13Me, and Ph135Me), the way in which the two uorophores are stacked with respect to one another is xed. However, co-facial stacking of the two uorophores may adopt either a head-to-head (H-H) or head-to-tail (H-T) conguration when the clamping modules are parallel. This is not an issue for symmetric uorophores such as Ant910Me and Ph14Me, as the H-H and H-T orientations are indistinguishable.
Interestingly, the G 2 -CB[8] 2 complex of Ant15Me also adopts a single H-H stacking conguration, as the spacing between its two off-line clamping modules is too large to allow for a feasible H-T conguration (Scheme S3 †). This specic stacking conguration is also revealed in the 1 H NMR spectrum of its G 2 -CB[8] 2 complex, in which the protons of the 1,5-anthracenyl moieties exhibit sharp and well-resolved signals (Fig. S9 †). In contrast, Np15Me with a smaller gap between its off-line clamping modules may allow for both H-H and H-T stacking congurations of the two uorophores, which leads to a signicant broadening of proton signals in the NMR of its G 2 -CB[8] 2 complex. Moreover, all the proton signals are equally broadened suggesting a dynamic process that involves the entire complex, which very likely correlates to an interconversion between H-H and H-T stacking congurations with an exchanging rate on the intermediate NMR timescale (Fig. S7 †). As a result, the aromatic uorophores in the G 2 -CB[8] 2 complexes Ant15Me and Np15Me both exhibit a substantial overlap of p-electrons in a less J aggregate-like fashion, leading to smaller bathochromic shis in their emission maxima (Fig. 5) compared to other uorescent molecules. 65 Considering their red-shi in absorption, the smaller bathochromic shis in emission maxima may also correlate to an anti-Kasha behavior as mentioned above, which requires further investigation.
Although both H-H and H-T stacking should be feasible by Np14Me and Ant14Me, the NMR spectra of their G 2 -CB[8] 2 complexes suggest a preference towards head-to-head stacking. The protons residing on the protruding ring exhibit an upeld shi due to shielding of the aromatic ring current, which is best explained by a head-to-head overlapping of the uorophores. This further suggests that the p-p interactions play a role in determining energy-favorable stacking congurations.
2.9.3 Substituents on the clamping modules. In addition to methyl (Me) groups in the para-position of the aryl clamping modules, other substituents including amino-(NH 2 ), methoxy-(OMe), dimethylamino-(NMe 2 ), isopropyl-(CMe 2 ), and methylthio-(SMe) readily form monomeric and dimeric complexes with CB [7] and CB [8], respectively, in the same manner as the parent methyl compounds. As the aryl clamping modules are both bound inside the CB cavity for the G 1 -CB[7] 2 and G 2 -CB[8] 2 complexes, they exhibit similar diffusion coefficients regardless of the variation in para-substituents (Fig. 3b).
On the other hand, the photophysical properties of the dimeric stacked uorophores are indeed affected by the size of the aryl substituents. As shown in Fig. 9, Np14H, a naphthyl uorescent molecule without any substituent on its clamping module displays the same absorption and emission spectra as those of Np14Me. However, a signicant difference of the emission maximum is observed for the G 2 -CB[8] 2 complex of Np14CMe 2 . As both the absorption and emission spectra of G and G 1 -CB[7] 2 of Np14CMe 2 are similar to those of Np14H and Np14Me, this difference observed for the G 2 -CB[8] 2 complex must stem from a certain variation in the stacking of the naphthyl pair, which is very likely caused by a signicant volume exclusion between neighboring isopropyl substituents. The resultant stacking conguration in G 2 -CB[8] 2 of Np14CMe 2 still leads to a red-shied absorption band corresponding to pelectron delocalization in the preorganized dimer. It seems that the preorganized dimer (in this case) does not result in an effective formation of an excimer-like state, as the emission maximum is very similar to that in pristine solution without an obvious bathochromic shi. This observation thus offers an additional opportunity to tune the photophysical properties of stacked uorophores by choosing appropriate substituents.

Conclusions
In summary, we have demonstrated a modular strategy to design a new class of uorescent molecules that (i) generate discrete, dimeric stacked uorophores in aqueous solution and (ii) are constrained by CB [8]-mediated multiple clamping. This modular design is surprisingly applicable and exible and has been validated by testing nine different uorophore cores ranging in size, shape, and geometric variation of their clamping modules. When complexed with CB [7], all uorescent molecules are dispersed in aqueous solution as discrete monomers, exhibiting an impressively high uorescence efficiency. On the other hand, complexation with CB[8] as 2 : 2 or 2 : 3 complexes leads to the immediate formation of discrete dimeric stacked uorophores. Multiple CB [8] clamping results in stable, preorganized ground-state dimers, which can be readily photoexcited to excimer-like states, displaying signicant bathochromic shis in absorption and emission with elongated uorescence lifetimes. Bathochromic shis in the emission spectra can be readily tuned by controlling the stacking of uorophores through specic variations in the clamping modules (through off-line alignment or altering substituents).
We demonstrate that intracomplex motion in the preorganized dimers is signicantly restricted, which suppresses both radiative and non-radiative deactivation, resulting in a substantially high quantum yield (up to 1) despite formation of excimer-like states. Some complexes are further restricted through non-parallel or triple clamping, which slows down radiative relaxation to an even greater extent, leading to elongated excited-state lifetimes up to 37 ns in aqueous solution. Moreover, complexes stabilized by multiple non-parallel clamping exhibit self-sorting in the presence of excess CB [8], which facilitates the design and fabrication of hierarchical functional structures.
While only arylpyridinium moieties have been employed as the clamping module in this study, current investigations suggest other chemical motifs with rigid structure exhibit the same clamping feature. The high rigidity ensures intrinsically low conformational entropy change during complexation, thus facilitating the formation of a long-lived, multicomponent complex in aqueous solution.
From a fundamental point of view, this study offers a model system with explicitly stable dimeric structures and tuneable features that can be utilized as a platform to study various intermolecular processes including excimer formation, charge transfer, exciton coupling, and singlet ssion. Moreover, such a modular molecular design towards quadrupolar uorescent molecules may provide a feasible toolbox in pursuit of distinct features such as large two-photon cross-sections 66,67 and non-Kasha behavior. 68 On the practical side, CB [7]-and CB [8]mediated uorescent complexes developed here are promising candidates for various (biological) imaging applications on account of their emergent photophysical properties such as long lifetimes, high emission brightness, and red-shied excitation bands.

Conflicts of interest
There are no conicts to declare.