Feeding a Molecular Squid: A Pliable Nanocarbon Receptor for Electron-Poor Aromatics.

A hybrid nanocarbon receptor consisting of a calix[4]arene and a bent oligophenylene loop ("molecular squid"), was obtained in an efficient, scalable synthesis. The system contains an electron-rich cavity with an adaptable shape, which can serve as a host for electron deficient guests, such as diquat, 10-methylacridinium, and anthraquinone. The new receptor forms inclusion complexes in the solid state and in solution, showing a dependence of the observed binding strength on the shape of the guest species and its charge. The interaction with the methylacridinium cation in solution was interpreted in terms of a 2:1 binding model, with K11 = 5.92(7) × 103 M-1. The solid receptor is porous to gases and vapors, yielding an uptake of ca. 4 mmol/g for methanol at 293 K. In solution, the receptor shows cyan fluorescence (λmaxem = 485 nm, ΦF = 33%), which is partly quenched upon binding of guests. Methylacridinium and anthraquinone adducts show red-shifted emission in the solid state, attributable to the charge-transfer character of these inclusion complexes.

Introduction

Curved aromatic molecules have found diverse uses in supramolecular and nanomaterials chemistry.[1−3] In particular, carbon-rich cavities of such systems have been used to develop cylindrical,[3−11] concave,[12,13] and macrocyclic hosts[14,15] for spherical guest molecules and ions, self-assembling surfaces,[16,12,17] and porous organic materials.[18,19] In these systems, the receptor function can be precisely controlled by the type and extent of curvature and by adjusting the cavity dimensions. The curvature facilitates formation of interlocked structures, i.e., rotaxanes,[20,21] catenanes,[22−25] and molecular knots.[24] While the synthesis of curved aromatics is often challenging,[26] they provide structural rigidity, variable curvature types,[27−29] topologically nontrivial π conjugation,[23,30−32,19] chirality,[33] and unusual chromophore properties.[34−36] These features can be leveraged to enhance supramolecular interactions and to produce usable physical output upon self-assembly.[37]

Cycloparaphenylenes (CPPs) have played a major role in these advances since the development of efficient synthetic methods based on masked phenylene equivalents[38,39] and metallacycle eliminations.[40] In particular, new supramolecular functions have been produced by hybridization of oligophenylene nanohoops with other building blocks such as porphyrins,[10] perylenediimides,[41,42] electron-rich arene substructures,[43−45] perfluorinated rings,[46,47] and N-donor heterocycles.[21,23] Here we report on a calixarene–CPP hybrid (1, Chart ), in which the calixarene and oligophenylene units are directly linked via CC bonds. This squid-shaped molecule has a flexible cavity and can bind neutral and cationic guests both in solution and in the solid state.

Chart 1

Design of the Molecular Squida

π-Conjugation in 1 and its parent motifs is omitted for clarity.

Results and Discussion

Synthesis

Compound 1 was prepared from the diagonally functionalized dibromocalix[4]arene 2a, which can be obtained stereoselectively as a cone-like structure (Scheme ).[48]2a was borylated and coupled with Jasti’s masked phenylene building block 3,[38] and the resulting dibromo intermediate 4 was cyclized using Yamamoto coupling, to furnish the basket-like precursor 5. The molecular structure of 5, revealed by an X-ray crystallographic analysis, is characterized by slight bending of the lateral biphenyl sections of the loop, indicative of a small degree of internal strain. The interplanar angles between the diagonal pairs of benzene rings in the calixarene section of 5 are respectively θ1 = 33.8° and θ2 = 76.9° (Scheme ). These are angles different in the parent calix[4]arene[4]Arene Ligands and Their Chelated Complexes of Titanium, Molybdenum, and Palladium. Inorg. Chem.. 2002 ">49] (2 with X = H, θ1 = −24.2° and θ2 = 68.6°), indicating that the observed conformation of 5 is a compromise between the steric requirements of the constituent subunits. Reductive aromatization of the two masked p-phenylene units in 5 was performed using a tin(II) reagent, as reported by Yamago et al.[50] Under these conditions, 5 cleanly produced the target 1, which was isolated in an 86% yield as a yellow solid. Using the above approach we were able to prepare up to 180 mg of 1 in a single batch. The product was unambiguously identified using NMR spectroscopy and mass spectrometry (Figures S48, S49, and S54; Scheme S4 of the Supporting Information, SI), and was further characterized crystallographically in the solid state (see below).

Scheme 1

Synthesis of 1

Reagents and conditions: (a) Pd(dppf)Cl2 (0.05 equiv), [B(pin)]2 (2.4 equiv), CH3COOK (2.4 equiv), dioxane, 110 °C, 12 h; (b) Pd(OAc)2 (0.12 equiv), dppf (0.135 equiv), Ag2O (4.5 equiv), K2CO3 (2 equiv), toluene, water, 80 °C, 24 h; (c) Ni(cod)2 (2.5 equiv), 2,2′-bipyridyl (2.5 equiv), THF, DMF, 80 °C, 16 h; (d) H2SnCl4 (8 equiv), THF, rt, overnight.

Molecular Structure

1 is a flexible molecule, balancing the conformational preferences of the calixarene part with the distortion of the oligophenylene loop. An automated conformational search[51−53] performed for the simplified structure 1′ (R = ethyl), followed by a full DFT reoptimization of the resulting ensemble, revealed a structural bistability of the oligophenylene loop, which adopted either an elongated (flattened) or circular (rounded) shape (Figure ). The change of the loop shape is made possible by the flexibility of the calixarene unit, which can switch between two nonequivalent flattened cone conformations. The calculations predict the flattened geometry (1′-A) to be preferred in the gas phase, but rounded conformers are nevertheless thermally accessible with the lowest-energy structure (1′-B) with a ΔGrel298 of only 0.6 kcal/mol. Structures similar to 1′-B are characterized by a more uniform curvature of the oligophenyl substructure with POAV1 angles[54] in the range of 4.3° to 7.0°. The broader distribution of POAV1 angles found in the 1′-A conformation is similar to those found in the [16]CPP lemniscate (CPPL) and related systems.[30,19] The internal strain enthalpy of 1′-A was estimated as 43.9 kcal/mol in a homodesmotic calculation (Scheme S3). This value is less than half the enthalpy reported for CPPL (102.7 kcal/mol),[30] suggesting that the octiphenyl substructure of 1′ is somewhat less strained than each of the two lobes of CPPL.

Figure 1

Top: Lowest-energy elongated (A) and rounded (B) conformations of 1′ (R = Et) found in a gas-phase DFT calculation. The initial ensemble of 113 conformers was generated using CREST[51] with an energy cutoff of 6 kcal/mol, and reoptimized at the B3LYP-GD3BJ/6-31G(d,p) level of theory. Bottom: dependence of loop width w as a function of Gibbs free energy. θ1 and θ2 angles are defined in Scheme . POAV angles (blue, degrees) are given for quaternary phenylene carbons. θ1 and θ2 angles are defined in Scheme .

The pliable internal cavity of our molecular squid is of interest as a potential binding site for guest molecules and ions. An initial indication of the receptor capabilities of 1 was observed in its two crystalline solvates, 1·3.2CH2Cl2 and 1·3C6H6 (Figure A,B). The former of these two structures contains a benzene molecule bound in the calixarene end of the cavity. The remaining solvent molecules are located outside the loop, while the loop itself is penetrated by butyl chains of a neighboring molecule. Although not isomorphous, the dichloromethane solvate shows similar features, with an aggregate of two solvent molecules residing in the calixarene cavity, and extraneous alkyl substituents inside the oligophenylene unit. The solvation pattern observed in these two crystals resembles the reported solvates of nanotube end-caps.[55,56] In each solvate, 1 adopts a flattened conformation (w = 9.0 to 9.2 Å), similar to the 1′-A structure predicted in the gas phase. This particular conformer contains a larger free volume inside the calixarene corner of the loop, offering more space for inclusion of solvent molecules.

Figure 2

Inclusion complexes of 1 with neutral and cationic guests, observed in the solid state. One of two symmetry-independent complexes is shown for 1·3C6H6 and 1⊃AQ. Hydrogen atoms (on 1), solvent molecules (outside cavities), counteranions (for cationic guests), and minor disordered positions are omitted for clarity.

Host–Guest Chemistry in the Solid State

An initial computational search showed that electron-deficient polycyclic aromatics containing three or four fused rings may be suitable as guests for the cavity of 1. In particular, the interior of the molecular squid was expected to share some of the binding characteristics of the parent CPP and calixarene motifs, displaying an affinity for electron-deficient and positively charged π-conjugated guests. The four guests used for further study (Chart ), namely anthraquinone (AQ),[57−59] 10-methylacridinium (MA+),[60−65] diquat (DQ2+),[66−90] and its phenanthroline-derived benzologue PQ2+,[89−92] were selected on the basis of their established utility in supramolecular chemistry. As we found, crystals of an inclusion complex could be successfully grown from a dichloromethane solution of 1 and 4 equiv of anthraquinone (AQ) by slow diffusion of methanol vapors. X-ray crystallographic analysis (Figure E) revealed the formation of a 1:1 adduct, 1⊃AQ, in which the receptor cavity is filled completely with the anthraquinone molecule. As a consequence of guest inclusion, the octiphenyl loop of 1 becomes somewhat flatter than observed in the solvates (w = 8.0 to 8.2 Å), presumably to better accommodate the length of the AQ guest.

Analogous attempts to obtain solid-state adducts by cocrystallization of 1 with organic salts were unsuccessful. In an alternative approach, crystals of 1·3C6H6 were soaked[93] in an acetone–methanol solution of 6,7-dihydrodipyrido[1,2-a:2′,1′-c]pyrazine-5,8-diium hexafluorophosphate (diquat, [DQ2+][PF6–]2). The dark brown crystals obtained using this method were found to contain the desired complex, [1⊃DQ2+][PF6–]2 (Figure C). The extreme flattening of the oligophenylene loop observed in the [1⊃DQ2+] adduct (w = 7.6 Å) is likely caused by a combination of steric, electrostatic, and crystal packing contributions. An analogous crystal-to-crystal transformation could be effected when 1 was similarly treated with 5,6-dihydropyrazino[1,2,3,4-lmn][1,10]phenanthroline-4,7-diium hexafluorophosphate ([PQ2+][PF6–]2). Interestingly, even though the PQ2+ cation is flatter than DQ2+, the loop width w in the [1⊃PQ]2+ adduct (8.0 to 8.2 Å) is somewhat larger than in [1⊃PQ]2+.

The crystals formed by solvates and adducts of 1 are not isomorphous, but they nevertheless reveal striking analogies of their packing patterns (Figure ). Structures of the benzene and dichloromethane solvates consist of herringbone layers characterized by partial penetration of butyl chains into neighboring oligophenylene loops. Packing of these layers is affected by the bulk of calixarene moieties and has no direct relationship with the herringbone patterns observed in unmodified cycloparaphenylenes.[94] In each solvate, the herringbone direction is antiparallel in consecutive layers. The inclusion of molecules and ions in the adducts of 1 leads to significant expansion of the crystal lattices. Importantly, however, the antiparallel arrangement of layers is preserved in all cases. Individual molecules are collinearly aligned within each layer and the butyl chains no longer penetrate the cavities, which are now filled with the guest species (DQ2+, PQ2+ and AQ). In the salt adducts, the PF6 anions are sandwiched in between the layers and retain close contacts with the edges of the organic cations. The structural analogies between the solid-state structures of solvates and those of the adducts indicate that the incorporation of DQ2+ and PQ2+ salts in the lattice is indeed feasible via a direct crystal-to-crystal transformation, as it can occur without major reorientation of the molecules.

Figure 3

Packing diagrams of inclusion complexes of 1. Molecules of 1 in adjacent layers are colored in red and blue. Hydrogen atoms and butyl substituents (on 1), solvent molecules (outside cavities), and minor disordered positions are omitted for clarity.

Guest Binding in Solution

When solutions of 1 in acetone-d6 were titrated with hexafluorophosphate salts of DQ2+, PQ2+, and MA+, significant changes of chemical shifts were induced in the 1H NMR spectra, consistent with the formation of host–guest complexes in fast exchange with the free host (Figures , S5, S9, S13, and S17). These changes were most pronounced in the aromatic region of the spectrum, but systematic downfield relocations were also observed for all aliphatic signals of 1. The broadening of guest signals, observed in all three titrations, suggested that the chemical shifts of the bound and free guest differ considerably. This assumption was verified for a sample of 1 containing 1.5 equiv of DQ2+, for which the slow-exchange limit was observed at 174 K in acetone-d6 (Figure S19). Under these conditions, no free 1 was present in solution, whereas the signals of the bound DQ2+ could be readily identified on the basis of the exchange correlations with the free DQ2+ observed in a ROESY spectrum (Figure S22). The shifts of the bound DQ2+ were consistently upfield relative to the free DQ2+, reflecting the shielding induced by the aromatic surface of the oligophenylene loop. Interestingly, the spectral pattern of the bound DQ2+ is completely desymmetrized, with four signals corresponding to the CH2CH2 unit. This low spectral symmetry indicated that not only the “somersault” rotations of DQ2+ inside the host cavity, but also the pseudoinversion of the twisted biaryl backbone were slow on the NMR time scale at 174 K.

Figure 4

Formation of inclusion complexes of 1 in solution observed using 1H NMR spectroscopy (600 MHz, 300 K, acetone-d6 or CD2Cl2). For complete titrations, see Figures S5, S9, S13, and S17. Signals of guests are indicated with red bullets.

While the crystal structures and low temperature NMR experiments provided unambiguous evidence for the formation of binary complexes with cationic guests, binding isotherms obtained from the 1H NMR titrations produced small but systematic discrepancies when fitted using the simple 1:1 binding model. The fit could be considerably improved by assuming initial formation of a relatively unstable ternary complex [12⊃X] (where X is the cationic guest),[81] which would be converted into the [1⊃X] at higher guest concentrations (Tables and Table S1). Data obtained using such a two-step binding model showed that the formation of the ultimate 1:1 species is most efficient for MA+ (K11 = 5.92(7) × 103 M–1), and becomes weaker for PQ2+ and DQ2+ (K11 = 1.43(1)·103 and 6.03(2) × 102 M–1, respectively). In all cases, the K21 binding constant is lower by 1 order of magnitude than the respective K11. The strong binding of MA+ is likely supported by a favorable combination of the cationic charge with the good geometric match of the guest with the cavity of 1 (Figure A). The initial formation of the ternary complex [12⊃MA+], inferred from the binding isotherm, was probed computationally using a CREST conformational search. Interestingly, the resulting ensemble revealed preferential binding of the cation in a single receptor cavity (rather than across two cavities). Furthermore, in the lowest-energy conformers, the other molecule of 1 was associated with the inclusion complex in an edge-to-edge fashion (cf. Figure B). While encapsulation of hexafluorophosphate in the other receptor cavity[95] could in principle occur to produce the hypothetical species [1⊃MA+][1⊃PF6–], such a binding event was ruled out on the basis of a 19F NMR titration (1 + [MA+][PF6–], acetone-d6), which showed a negligible effect of 1 on the 19F chemical shift of the PF6– anion.

Table 1

Association Constants for Host–Guest Complexes of 1a

guest	model	K11 [M–1]	K21 [M–1]
DQ2+b	2:1	6.03(2) × 102	3.36(3) × 101
PQ2+b	2:1	1.43(1) × 103	1.78(4) × 102
MA+b	2:1	5.92(7) × 103	4.3(1) × 102
AQc	1:1	1.968(2) × 101

Based on 1H NMR titration data (300 K).

In acetone-d6.

In CD2Cl2.

Figure 5

(A) DFT-optimized lowest-energy conformer of [1′⊃MA+] and (PCM(acetone)/BGD3BJ/B3LYP/6-31G(d,p), initial conformer ensemble obtained using CREST). (B) Lowest-energy conformer of [1′2⊃MA+] found in a CREST metadynamics search.

A similar though weaker binding interaction was observed between 1 and anthraquinone (AQ) in CD2Cl2. In this case, the 1H NMR titration was fully consistent with the HG (1:1) model (K ≈ 20 M–1). The formation of the 1:1 adduct was proven using low-temperature 1H NMR spectroscopy (600 MHz, 160–270 K, CDCl2F, 4:1 molar ratio of AQ to 1). The use of the latter solvent[96] instead of CD2Cl2 was necessary for direct observation of the host–guest complex in the limit of slow exchange (Figures S20 and S21). Under these conditions, no free 1 was present, whereas the AQ molecule bound in the [1⊃AQ] complex showed four proton resonances, consistent with an effectively C2-symmetrical environment of the cavity. Additionally, the EXSY pattern observed between the resonances of free and bound AQ showed that chemical exchange was significant even at 170 K (Figure S23). However, no EXSY peaks were observed among the four resonances of the bound AQ, indicating that the guest is effectively locked inside the cavity of 1, and is not capable of “somersault” rotations at the time scale of the ROESY experiment.

Vapor and Gas Sorption

Gas adsorption analyses performed for a crystalline sample of 1 showed variable porosity toward a range of different adsorbates (Figure ). While the N2 adsorption capacity was very low, significant uptake of CO2 was observed at 195 K, reaching a maximum of 1.60 mmol/g. This value corresponds to a molar ratio of CO2 to 1 of ca. 1.8. The BET area, calculated on the basis of the adsorption branch of the CO2 isotherm, is 63.7 m2/g (Figure S25, Table S2), lower than reported for the larger [12]CPP nanohoop.[18] The isosteric heat of CO2 adsorption (Qst) was determined for 1 from isotherms measured in the temperature range of 273–293 K, using the single-site Langmuir–Freundlich model and the Clausius–Clapeyron eq (Figures S26 and S27, Table S3). The calculated Qst values reach 53.1 kJ/mol at zero coverage and then decrease to ca. 30 kJ/mol at higher CO2 uptake. The initial Qst is higher than previously reported for CO2-selective pillar[5]arene-based sorbents[5]Arene-Based Supramolecular Organic Frameworks for Highly Selective CO2-Capture at Ambient Conditions. Adv. Mater.. 2014 ">97] (up to 44 kJ/mol), implying an energetically favorable interaction between 1 and the initially adsorbed CO2. The binding enthalpy ΔH298 calculated for the inclusion complex [1⊃CO2] in the gas phase is −48.9 kJ/mol, indicating that the high initial heat of adsorption may indeed correspond to a well-defined supramolecular interaction between CO2 and 1.

Figure 6

Experimental adsorption and desorption isotherms (solid and empty circles, respectively) of N2 (77 K), CO2 (195 K), cyclohexane (293 K), MeOH (293 K), and H2O (293 K) measured on crystalline sample of 1.

At 293 K, vapor adsorption of H2O, cyclohexane, and methanol, yielded maximum uptake values of 1.14, 1.64, and 4.12 mmol/g, respectively, corresponding to approximately 1.3, 1.8, and 4.5 adsorbate molecules per one molecule of 1. The significant adsorption hysteresis observed for cyclohexane is indicative of its stronger retention in the pores of 1. On the basis of the MeOH isotherm, a pore volume of 0.167 cm3/g was estimated for 1. In comparison, when solvent molecules are removed from the crystal structure model of 1·3C6H6, the resulting virtual pores correspond to a helium volume[98,99] of 0.322 cm3/g. The comparatively lower pore volumes attainable via adsorption may indicate that either (a) only part of the virtual porosity of the crystals is available for uptake or (b) a structural reorganization of the material accompanies the sorption process.

Optical Properties

The electronic spectrum of 1 in dichloromethane (Figure ) features two absorption bands with λmaxabs = 327 and 377 nm, respectively, the latter being responsible for the yellow color of the compound. 1 displays a cyan emission with a maximum at 485 nm and a quantum yield of 33% (in dichloromethane, τF = 1.82 ns). Similar absorption and emission spectra were observed for amorphous thin films of 1 obtained by drop casting of dichloromethane solutions. Partial quenching of fluorescence was observed during titrations of 1 with molecular and ionic guests, suggesting that charge transfer (CT) may occur between the electron-rich host cavity and the electron-deficient guest molecule. The absorption spectrum of [1⊃AQ], measured for a thin film, showed a red shift of the lower energy band (λmaxabs = 396 nm vs 380 nm for free 1), and a weak tailing band above 500 nm, not observed in the free 1, which was tentatively ascribed to a CT transition. Remarkably, the film showed weak yellow-gold fluorescence (λmaxem ≈ 580 nm), red-shifted relative to the solid-state emission of the free host (λmaxem ≈ 500 nm). Similar features were observed in a thin film of [1⊃MA+][PF6–], in which an even larger red shift was recorded for the low-energy absorption band (λmaxabs = 411 nm). Again, a weak absorption tail was observed, which was complemented by an even more red-shifted emission band (λmaxem ≈ 700 nm), corresponding to the red-orange fluorescence of the film. Good quality films could not be obtained by drop-casting for complexes with DQ2+ and PQ2+; however, when 1 was dissolved in acetone containing a large excess of the corresponding guest, a weak tailing band could be identified in the 450 to 700 nm range, possibly corresponding to CT transitions of the host–guest adducts. For these solutions, there were however no visual indications of any red-shifted fluorescence.

Figure 7

Absorption and emission spectra (solid and dashed lines, respectively) of (A) 1 in dichloromethane solution (red) and as a thin film (black); (B) 1 (black), [1⊃AQ] (blue), and [1⊃MA+][PF6–] (orange) in thin films; (C) 1 in acetone solutions containing (a) no additive (black trace), (b) 178 equiv of [DQ2+][PF6–]2 (red trace), and (c) 11 equiv of [PQ2+][PF6–]2 (green trace). The latter two spectra were recorded relative to an acetone solution containing the same amount of the corresponding pure guest.

The involvement of charge transfer in the optical spectra of [1⊃DQ2+] and [1⊃MA+] was probed using time-dependent (TD) DFT. The initial geometries were again derived from a CREST metadynamics search and were reoptimized using the PCM(acetone)/CAM-B3LYP-GD3BJ/6-31G(d,p) level of theory, which was also used for the TD calculation. The Coulomb-attenuating method[100] (CAM) was chosen specifically to minimize the self-interaction error, which is known to produce spurious results for CT systems.[101] The HOMO and LUMO of 1 are mostly localized on the oligophenylene loop, with vanishing amplitudes on the calixarene subunit (Figure ). The adducts of 1 with AQ, MA+, DQ2+, and PQ2+ retain the HOMO localization of the free host, whereas the LUMO level is always localized on the electron-deficient guest (Figure ). In [1⊃DQ2+], the 10 highest occupied Kohn–Sham (KS) MOs are nearly pure orbitals of the host 1. The three lowest unoccupied MOs (LUMO through L+2) are derived from DQ2+, whereas the L+3 level corresponds to the original LUMO of the host. The calculated absorption profile obtained for the complex is very similar to the experimental one, except for the blue shift of ca. 0.6 eV, characteristic of the CAM method (Figure S34, Table S6). The calculation predicts 18 weak transitions (f < 0.02) below 3.75 eV, which may explain the emergence of the tailing band above 450 nm in the experimental spectrum. These transitions consist predominantly of excitations from host occupied levels to guest virtual levels, confirming the charge-transfer character of this band. A more intense transition at 3.76 eV (f = 0.29), which corresponds to the experimental maximum at ca. 480 nm, is dominated by the HOMO to L+3 excitation, which is accompanied by smaller CT contributions. This transition is red-shifted relative to its counterpart in the calculated spectrum of free 1 (3.88 eV), as indeed observed in the experiment. For [1⊃MA+], the majority of frontier KS orbitals were also found to be either pure host levels (HOMO through H–6, and L+1) or pure guest levels (LUMO and L+2). The TD calculation again predicted a range of weak CT transitions (below 3.70 eV), and an intense transition of the host at 3.71 eV (f = 0.38), which is again red-shifted in comparison with the guest-free 1 (Figure S33, Table S5).

Figure 8

Frontier Kohn–Sham molecular orbitals for 1 and its complexes, [1⊃DQ2+] and [1⊃MA+] (PCM(acetone)/CAM-B3LYP-GD3BJ/6-31G(d,p)).

Conclusions

The design of the molecular squid described in this work capitalizes on structural and electronic characteristics of calixarenes, linear oligophenyls, and cycloparaphenylenes, to yield an electron-rich aromatic system that is simultaneously strained and flexible. The conformational bistability of 1, predicted in the gas phase, leads to two types of energetically accessible geometries of the octiphenyl substructure. The possibility of switching between two curvature distributions is of general interest as a means of controlling supramolecular and optical properties of such “spring-loaded” molecular hybrids. The molecular squid shows promise both as a versatile supramolecular receptor, capable of providing an optical response upon binding of electron-deficient guests, and as a structurally nontrivial molecular porous material. By refining the present structural design, we are now trying to develop receptors in which the conformation and electronic structure of the curved π system of the host will be even more strongly affected by guest binding, to produce a functionally useful output.