Aiming to greener strategies minimizing the solvents used in the synthesis of the M12L8 interlocked cages and to increase our knowledge in the solid-state self-assembling processes and structural properties of amorphous poly-[n]-catenanes, the mechanically interlocked M12L8 cages were synthesized by neat grinding TPB and ZnI2 for 15 min at room temperature.19,22 The product was amorphous (hereafter referred as a1) with a diffractogram characteristic of a short-range order solid (Figure S2). Thus, the structure of a1 is unknown but it is known that in a1, the MOCs are interlocked because the amorphous cages can uptake aromatic guests following a gas-solid reaction (i.e., indirect methods see Figure S2) and also the interlocking has been corroborated by solid-state 13C NMR.19 However, direct structural elucidation of the crystalline sample obtained from a1 after guest inclusion has never been done.
Despite not having continuous channels, single crystals of interlocked M12L8 cages can exchange aromatic molecules via single-crystal-to-single-crystal (SCSC) process in the solution-state 16 and solid-state from the gas phase.18 It has been rationalized that due to the rotor-like behaviour of the pyridine rings, and the opened nature of the M12L8 1D chains formed of opened icosahedrons, allows several aromatic guests to diffuse in the crystal whilst maintaining the crystallinity for several consecutive SCSC processes.18 However our attempts to obtain a SCSC transformation using paracetamol as incoming guest have not been successful so far. Attempts to grow single crystals of the M12L8 poly-[n]-catenane including paracetamol were carried out. Although single crystals were formed, we were unable to determine the guest content. Therefore, not being able to use single crystals containing paracetamol, guest inclusion was carried out stirring a suspension of a1 for 12h at room temperature in a concentrated methanol solution of paracetamol (ca. 1 g/15 ml). Then, the suspension was filtered and analysed by powder XRD. The diffractogram turned crystalline with Bragg reflections characteristic of the poly-[n]-catenanes of interlocked M12L8 cages but with a significant background and severe peak overlap (Figure 2a).
Microcrystalline 1 was subjected to TG analysis to quantify the paracetamol encapsulated in the M12L8 cages. The weight loss corresponds to 12.3 % as shown in Figure 2b that are ca. 6 paracetamol molecules. The mass loss fits with the paracetamol calculated from NMR (vide ante). Thus, the TG data confirms also that paracetamol has been included in the amorphous phase of the M12L8 cages following a transformation from the amorphous to the crystalline state. The temperature at which the paracetamol is released is much higher than those observed in aromatic solvents according to the different boiling points but also suggests that the paracetamol is kept in the solid structure in an efficient way. For instance, in a microcrystalline sample of TPB-ZnI2 poly-[n]-catenane, the included guests, chlorobenzene and toluene, are completely released at 190 °C and 240 °C respectively,22 while in 1 complete guest release occurs at ⁓ 325 °C as shown in Figure 2b.
Crucially, a1 can be synthesized also using ZnX2 salts (where X = Br and Cl) giving amorphous phases a2 and a3 respectively, which can be used to uptake various aromatic and non-aromatic guest molecules. Thus, the amorphous-to-crystalline transformation upon paracetamol inclusion was also monitored for the chloride and bromide counterparts of the amorphous catenanes in a liquid-solid phase. In all the experiments the short-range to long-range order transformation is observed (see ESI).
Microcrystalline 1 did not turn amorphous after being in contact with the atmosphere for over a month. This indicated that the guest is not volatile, thus keeping the poly-[n]-catenane crystalline structure after filtration. The paracetamol inclusion experiments were repeated four times, and in all the cases an amorphous-to-crystalline transformation was observed, with samples showing different degrees of crystallinity, probably indicating a different guest loading (see ESI). The powder XRD pattern corresponding to that of the M12L8 poly-[n]-catenane with the reflections with Miller indices (2-10) and (101) are clearly observed, fitting the simulated powder XRD patterns from single crystals of M12L8 poly-[n]-catenanes containing aromatic guests. The (2-10) is quite intense as it cuts the M12L8 cages including the paracetamol guests.
Importantly, poly-[n]-catenane 1 is unstable in aqueous media as it decomplexes to yield TPB and zinc halides,16 liberating the included molecules in the M12L8 nanocages that can be quantified by NMR (see ESI). From the solution 1H NMR data, the amount of paracetamol included in a1-a3 range from 2 to 6 molecules per M12L8 cage (Table S1). In all the experiments, values higher to those which the M12L8 cages can theoretically include (i.e., according to the cage volume and the guest volume) have not been seen (vide ante). This excludes that the released paracetamol measured by NMR corresponds to guest attached to the surface of a1-a3. Regarding the observed variability of included paracetamol in all the experiments, our opinion is that it might be linked to the structure of the amorphous poly-[n]-catenanes formed in the solid state, as in each reaction the product structure of the amorphous phase might be different. For instance, it is not known at the moment if the 1D chains of the interlocked M12L8 cages are interwoven structures which might cause problems to fill and pack the 1D chains after the amorphous-to-crystalline transformation and hence to fill the M12L8 empty cages (i.e., reconstruction process), giving this statistical variation. Also, if the M12L8 cages are squeezed in such a way that the opened windows are completely closed, the guest inclusion might yield different values in the various cases.
Synergistic approach for solving the crystal structure of a challenging M12L8 poly-[n]-catenane sample using powder XRD data and solid-state DFT.
There is no doubt that the most powerful technique for the 3D structure determination of a crystalline solid is SC-XRD. However, in this case, because single crystals did not include paracetamol, full structure determination of 1 was attempted from powder XRD data. It is important to notice that structure solution of 1 from powder XRD data is challenging due to the following aspects: i) Despite the inclusion of paracetamol, the sample still shows large background which often is a problem in the indexing and profile fitting process; ii) since the sample was prepared by solid-state grinding, the particle sizes are intrinsically small and Bragg reflections are overlapping considerably in some regions of the diffractogram (Figure 1a); iii) the powder XRD data has been recorded up to 30 degrees in 2θ to create a low resolution dataset (2.99 Å), as it is known that many materials obtained after solid state reactivity show poor crystallinity, so this case is also challenging from a resolution point of view. Therefore, the indexing step and space group determination in the structure solution process could not be carried out using powder XRD data.27,39 Despite of the low-quality of 1, from a crystallinity and resolution’s point of view, structure solution can still be carried out if appropriate experimental and theoretical DFT techniques are combined.
Knowing from SC-XRD data that in all the SCSC guest exchange reactions carried out in crystals of M12L8 poly-[n]-catenanes were affected by small lattice variations and the trigonal space group (R-3) was maintained, indexing and space group determination of 1 was done using the symmetry of an already known crystal structure of M12L8 poly-[n]-catenane containing nitrobenzene which was determined at room temperature.22 Thus, the unit cell introduced as a starting cell was: a = b = 38.6805 Å; c = 16.0202 Å; α = β = 90 ° and γ = 120 °; V = 20,758 Å3. The space group was (R-3). With those lattice parameters and space group, a Pawley41 refinement was carried out using DASH42 program. The refined cell had a smaller volume: a = b = 38.4795 Å; c = 15.8455 Å; α = β = 90 ° and γ = 120 °; V = 20,319 Å3 with and acceptable good fit with χ2 = 4.77. This unit cell was used to carry the structure solution from powder XRD using laboratory data.
Structure solution of 1 in the direct-space was performed using the simulated annealing (SA) method implemented in DASH42 program. For, the structure solution, a first structural fragment (SF1), comprising two ZnI2, 1.3 TPB molecules, and a second structural fragment (SF2) consisting of one paracetamol molecule43 were used as a starting structural model (Figure 3a). One paracetamol molecule was used because of the experimental data from 1H NMR and TG data, which indicated that 6 guest molecules are included per M12L8 cage, (i.e., R-3 has 6 symmetry operators). The structural model was not treated as a rigid body and only the torsion angles (τ) of the host were allowed to rotate during the structure solution process. Seven torsion angles (τ1, τ2,...,τ7) in the structural model were defined considering SF1 and SF2. In total 13 structural variables (Γ) including the three positional variables {x, y, z}, the three orientational variables {θ, Φ, Ψ} and the seven torsion angles variables {τ1-τ7} define the structural model.
In the structure solution from powder XRD, it is important to note that the process of finding the global minima (i.e., the structure with the lowest R-factor) is like exploring a N-dimensional hypersurface where N is the number of structural variables (Γ). These surfaces exhibit several minima as a function of Γ that can be divided into two groups: i) the global minimum corresponding to the fully determined crystal structure and ii) the local minima that corresponds to wrong crystal structures. In a structure refinement, it is assumed (generally) that an initial model has been obtained from a similar structure in a database (SF2), or from a Patterson or Direct methods (SF1) structure solution. Moreover, it must be noted that for a refinement to be considered successful, the initial model has to be sufficiently close to the true crystal structure for an acceptable Rietveld refinement.44,45
From all the DASH-solved structures, the one with lower χ2 was selected for the complete structural analysis (Figure 3b). The crystal structure solution showed the incomplete formation of the M12L8 nanocage with a distorted six-member ring (i.e., the six-member ring could be recognized) generated after the symmetry operators. The other structural aspect that does not yield a chemically reasonable structure involves the symmetry-generated pyridine (py) rings which are too far to be considered a reasonable py-Zn coordination bond (ca. 2.697 Å) as shown in Figure 3b. This could be due to the low-resolution powder XRD data used for the structure solution. Moreover, the paracetamol guests are too close to each other leading to a molecular arrangement without chemical sense (Figure S27b). For the rest of plausible structures solved by DASH, were very similar with different benzene geometries and pyridine-N···Zn distances and orientations.
The fact that the structure was not completely solved using powder XRD can be because the structure obtained got trapped into a local minima (hereafter this structure is referred as 1local) in the Γ-dimensional hypersurface, and by using the global optimization algorithms (simulated annealeing SA) in DASH, 1local cannot escape from the local minima in order to find the structure in the global minimum (1global). Thus, we decided to use DFT calculations to see if the structure could be further improved (by using the total energy of the model system as an additional parameter to be minimized) and converge into a final structure with reasonable geometries (i.e., the global minimum of the Γ- hypersurface) that could be, if necessary, used in a new step of Rietveld refinement.
The best structure obtained from direct-space, 1local, was used in DFT optimization procedure (DFT details, i.e. the functional and the basis set are reported in the next section). The calculations were performed in the solid state, under periodic conditions and with symmetry constrains. It is important to underline that QM optimization methods are capable, with a reasonable starting guess geometry, to reach the closest minima of the potential electrostatic surface (PES) while, in general, maintaining the overall symmetry of the investigated structure. Thus, even if many independent degrees of freedom were involved, the calculations converged with reasonable computational costs. The outcomes yield good M12L8 cage structures and guest geometries, in which, as expected, the distorted rings in the initial non optimized structure (1local) became flat benzene rings generated by the symmetry operators. Likewise, all the py-Zn coordination bonds were coherent with the previously observed SC-XRD structures with distances ranging from 2.09 Å to 2.12 Å (Figure 4). In the DFT calculations the unit cell used for the structure solution was maintained. The fixing of the unit cell is essential in the DFT geometry optimization step to match the experimental powder XRD data. The DFT optimized structure shows good paracetamol geometry and the guest-guest contacts are reasonable (Figure 4b). In fact, we can observe the good alignment of the maps of electrostatic potentials (MEPs) of paracetamol molecules (vide ante). The DFT optimized structure is hereafter named 1global.
The unit cell and space group used for the structure determination were subjected to a further LeBail (structureless) refinement using GSAS (General Structure Analysis System) software.46 The GSAS LeBail refinement led to a good profile fitting with Rwp = 2.98 %, indicating a correct indexing and space group determination (Figure 3a). Using the unit cell from the LeBail, 1global was introduced in GSAS and started the Rietveld refinement.44,45 The structural model was “fixed” using the bond distances and angles obtained in the DFT optimized structure (i.e., 1global). Standard restraints applied to bond lengths and bond angles, planar restraints were applied to the benzene and phenyl rings. A common isotropic displacement parameters was fixed for all non-hydrogen atoms, and the isotropic displacement parameters of H atoms was set at 1.2 times this value. The final Rietveld refinement was Rwp = 4.67 % for a structure with formula [(TPB)8(ZnI2)12]·6(paracetamol) (Figure S29).
1global clearly shows the poly-[n]-catenane formed of interlocked M12L8 nanocages via the π-π interactions (3.347 Å) responsible of the mechanical bond formation and the generation of the 1D chains along the c-axis. 1global contains six paracetamol molecules per M12L8 nanocage, corroborating the 1H NMR and TG analysis. Paracetamol guests are oriented differently if compared to the other M12L8 poly-[n]-catenane structures including different aromatic guests (nitrobenzene p-xylene, o-dichlorobenzene, etc,…) described so far, in which the guest were situated just below the aromatic TPB rings and stabilized through aromatic-aromatic interactions ranging from 44 to 25 kcal/mol determined by DFT.18
The crystal structure of 1gobal: A TPB-ZnI2 M12L8 poly-[n]-catenane including paracetamol.
In 1global, paracetamol is in a position like the room temperature SC-XRD structure containing two p-xy guests in the asymmetric unit. The paracetamol guest is in the type B position following the structural description in the reported structure.18 Because paracetamol is longer (8 Å) than p-xy (5.9 Å), in our opinion six paracetamol molecules cannot fit in type A position via efficient aromatic-aromatic interactions, and hence paracetamol goes to the central part of the cage (type B) optimizing the electrostatic interactions in a different way (i.e., not via host-guest π-π interactions). In fact, the included paracetamol in one M12L8 cage interacts with the two interlocked M12L8 cages that form the 1D chains of cages expanding along the c-axis (Figure 4b and 4c). The first short contact interaction (2.534 Å) occurs among the –OH group of the paracetamol that is oriented towards the N atom of the pyridine of TPB from cage A (Figure 4c). The oxygen atom in the alcohol group interacts also with the aromatic C–H group of the pyridine of TPB (2.367 Å). While the methyl group of paracetamol interacts, via C-H···π interactions,20,47 with the benzene ring of TPB (2.383 Å) from cage C (Figure 4c). Due to the contemporary presence of strong and weak van der Waals intermolecular interactions, the simulation method is of key importance, in this case, Grimme corrections are mandatory (see ESI).
Importantly, in a given M12L8 cage (cage B), the paracetamol guests mainly interact with the two TPB ligands that belong to the interlocking cages A and C as shown in Figure 4c. Within a given cage, paracetamol also interacts with the neighbouring paracetamol molecules in such a way that a well-defined cluster of guest molecules is clearly seen in the central part of the M12L8 nanocage as shown in Figure 4b and 4c. Thus, in this case also guest-guest interactions are important for the stabilization of the cages.
Packing efficiency of paracetamol in M12L8 interlocked nanocages.
The 3D elucidation of 1global is very useful to determine the number of guests that can be included in the M12L8 cages of the TPB-ZnI2 poly-[n]-catenanes. Using a probe of radius of 1.2 Å, 1global shows that the void left per unit cell is 13.4 % (2722 Å3), considering that in one cell there are 3 cages, in one single cage there are still 907 Å3 that can be filled by more paracetamol guests (Figure S34). From 1global, by manually removing the guest content and calculating the void space left, using the volume of the empty M12L8 cage 2269 Å3 (i.e., 6807/3), it can be calculated that the volume occupied by a single paracetamol molecule inside a M12L8 cage is 2269 Å3 – 907 Å3 = 1316 Å3 that divided by 6 paracetamol per cage gives 227 Å3 per paracetamol molecule. Therefore, another ⁓4 paracetamol guests (907 Å3/227 Å3) can be included in one M12L8. From these calculations we expect that a fully loaded M12L8 nanocage will contain about 10 paracetamol molecules per M12L8 cage. This is realistic because it has been observed that 12 p-xylene guests can fit in one M12L8 cage.18 It must be said that p-xylene is smaller and symmetric while paracetamol is slightly larger and asymmetric, and this has to be considered in the construction of the unit cell by the asymmetric unit by the symmetry operators.
Solid-State Density Functional Theory (DFT) calculations.
Understanding the strength of the interactions among host and guests in host guest systems involving MOCs, is fundamental to establish if there is molecular selectivity for a certain type of molecule.47 To get further insights concerning the host-guest affinity, molecular modelling analysis has been carried out. DFT calculations in the solid-phase at the PBE/DNP level (i.e., PBE functional and double zeta basis set including Grimme corrections) have been performed to determine the host-guest energy interactions considering paracetamol interacting with the TPB moiety at the internal walls of the M12L8 nanocages.48,49,50,51,52 The geometry of the host/guest complexes has been fully optimized. This paradigm showed very good results in several recent studies related to molecular modelling simulations of similar or different solid phase crystalline systems.53,54,55,56
First of all, we estimated the total stabilization energy per guest molecule by calculating the host-guest mean interaction energy (Ehost-guest): the energy difference between the overall crystalline energy (per unit cell immersed into the crystalline structure, i.e., periodical conditions) and the sum of the energy of the empty crystalline structure (per unit cell, immersed into the empty crystalline structure, i.e., periodical conditions) and the energy of the isolated paracetamol molecules (18 per unit cell). The outcome value was quite high, 46 kcal/mol per paracetamol molecule, demonstrating the importance of the guest molecules inside the M12L8 nanocages in the overall stability of the crystalline architecture. Obviously, we have also to consider the favorable guest-guest pair interaction contribution of about 2 kcal/mol, intrinsically included in the host-guest total stabilization energy. Thus the net stabilization is about 44 kcal/mol per paracetamol molecule (Figure 5) which is comparable to the value observed for the TPB-ZnI2 poly-[n]-catenane including nitrobenzene.22 Similar host-guest energy interactions results have been also obtained performing the DFT calculations in the gas phase model (see ESI). These Ehost-guest values underline the importance of the guest’s inclusion and maybe rationalises one of the main driving force in the guest absorption by the amorphous phase and in the amorphous-to-crystalline transformation processes. Interestingly, the guest-guest orientation within the cages maximizes the interactions among guests, which is done by orienting them in an antiparallel manner within the M12L8 cages along the c-axis. This is also corroborated by the MEPs analysis, shown in Figures 5a-c, where positive and negative regions of the paracetamol guests interact with the corresponding counterpart of the TPB moiety to optimize attractive forces.
Drug delivery in aqueous environment.
The solid-state reaction yielding a1 forms only the amorphous catenane made of mechanically interlocked M12L8 nanocages as no other crystalline structures are formed. This behavior is reminiscent to that observed in the solution state when the self-assembling method is the instant synthesis57 in absence of aromatic solvents which also yields an amorphous solid made of interlocked M12L8 cages. Thus, the solid-state synthesis by neat grinding selectively yields the poly-[n]-catenanes made of M12L8 nanocages. As it is a fast crystallization process, it is considered a kinetic product.15,22 We believe that the synthetic path in the solid-state by neat grinding (Figure S1) is an entropically favoured self-assembling process due to the severe disorder in a1. However, when the paracetamol guest enters in the cages the process is enthalpically favoured. Since kinetic products are metastable structures, it is possible to exploit their structural adaptability (dynamic behaviour), for instance for guest inclusion, as demonstrated by the paracetamol uptake reported here. Importantly, the M12L8 poly-[n]-catenane is stable and can be reused several times whilst maintaining the structural integrity. This is in part due to the mechanical bond that brings not only dynamic behavior but also structural stability to the M12L8 icosahedral cages which strongly interact one with the other by strong van der Waals forces (i.e., ⁓145 kcal/mol)16,18. Although toxicology tests of TPB have not been carried out yet, and since the metal used is Zn(II), which is considered non-toxic (at least within certain limits) in similar structures (MOFs),58,59 we propose here a1 as a potential drug delivery agent for the following reasons: i) clearly it can trap a drug by adapting its host structure, ii) a1 can host paracetamol for several days in the solid-state which means that it can be manipulated easily and, iii) it can release it in an aqueous media as depicted in Figure 6. The disruption of the Zn-py coordination bonds, brings about the breaking of the mechanical bond and hence the liberation of the paracetamol.
Kinetic trapping of paracetamol within the M12L8 nanocages via instant synthesis.
From the drug loading point of view, (i.e., the inclusion of paracetmaol in the M12L8 nanocages), using the amorphous phases (a1-a3), the maximum achieved loading has been 6 molecules per cage, which is the 60 % according to the theoretical limit (⁓ 10 guests). To improve the guest loading, we explored a different paracetamol inclusion approach, always under kinetic control, by using the instant synthesis method which has been shown to be very successful to trap aromatic solvent used to dissolve the TPB ligand.15 Our idea is to exploit the immediate formation of the poly-[n]-catenanes using the instant complexation of TPB ligands by Zn(II) in the presence of a concentrated solution of paracetamol. A solution of the tridentate ligand was prepared by dissolving TPB (31 mg) in 4.5 ml of chloroform and 3 ml of methanol. To this TPB solution, 186 mg of paracetamol were added. Importantly, paracetamol is very soluble in the TPB solution as it dissolved immediately without the need of stirring forming a transparent and homogenoeus solution (Figure S35). Then a methanolic solution of ZnI2 (52 mg) was immediately added to the stirring TPB and paracetamol solution (Figure S35). A white precipitate formed instantaneously (hereafter named 1IS, where IS stands for instant synthesis) and after filtration (using a low vaccum water pump), the sample was analyzed by powder XRD. The diffractogram showed to be a crystalline material but with a significant background (Figure S36) which fits well the powder XRD pattern of 1. Interestingly, the same sample measured 12 hours later showed a much better crystallinity suggesting that the 1D chains of M12L8 nanocages reorganized yielding a more packed structure. The sample was analyzed by TG analysis showing that the weight loss is ca. 14 % which corresponds to ⁓7 (6.8) paracetamol guest molecules. A second sample synthesized using the same approach gave 6.4 guest molecules included in the M12L8 nanocages with a 13.3 % of weight loss (Figure S37). In both cases the quantity of paracetamol included is below the 10 molecules per cage as estimated from the 1global X-ray structure.
Thus, by instant synthesis during the fast self-assembling process, the cages are able to trapp slightly more molecules than using the amorphous phases just by using a crystallization method that takes only few seconds to be done. Clearly, compared to the guest inclusion by single crystal formation that usually needs several days, the solid-state grinding and the instant synthesis are two different approaches providing a significant advances in the synthesis of MIMs such as polycatenanes self assembled of large M12L8 cages, and in their guest-behavior.