Selective Nitrate Recognition by a Halogen‐Bonding Four‐Station [3]Rotaxane Molecular Shuttle

Abstract The synthesis of the first halogen bonding [3]rotaxane host system containing a bis‐iodo triazolium‐bis‐naphthalene diimide four station axle component is reported. Proton NMR anion binding titration experiments revealed the halogen bonding rotaxane is selective for nitrate over the more basic acetate, hydrogen carbonate and dihydrogen phosphate oxoanions and chloride, and exhibits enhanced recognition of anions relative to a hydrogen bonding analogue. This elaborate interlocked anion receptor functions via a novel dynamic pincer mechanism where upon nitrate anion binding, both macrocycles shuttle from the naphthalene diimide stations at the periphery of the axle to the central halogen bonding iodo‐triazolium station anion recognition sites to form a unique 1:1 stoichiometric nitrate anion–rotaxane sandwich complex. Molecular dynamics simulations carried out on the nitrate and chloride halogen bonding [3]rotaxane complexes corroborate the 1H NMR anion binding results.

(100 mg, 0.0969 mmol) was dissolved in dry degassed THF (2 ml), NaI (46.1 mg, 0.307 mmol) was added followed by Cu(ClO4).6H2O, which was left stirring for 5 minutes. TBTA (4 mg, 0.0077 mmol) and DBU (11.7 mg, 0.0768 mmol) were added followed by 9 (7.07 mg, 0.038 mmol). The mixture was left to stir at room temperature for 2 days, avoiding exposure to light. The mixture was diluted with CH2Cl2 then washed with a 0.02 M EDTA/0.1 M NH4OH solution, the organic layer was dried over MgSO4, and the solvent was removed in vacuo. The resultant solid was purified by silica gel column chromatography (99: Synthesis of 5·(Cl)2 2·(Cl)2 (21.3 mg, 0.00821 mmol) and 3 (16 mg, 0.0246 mmol) were dissolved in dry CH2Cl2 (5 ml) and were left to stir for 30 minutes. Grubbs' second generation catalyst (1.6 mg, 10 wt. %) was added and the solution stirred at room temperature, monitoring by TLC and ESI mass spectrometry throughout. After two days a further addition of Grubbs' second generation catalyst (1.6 mg, 10 wt. %) was made and after three days the solvent was removed in vacuo. The crude was purified by iterative preparative silica TLC (EtOAC:MeOH 95:5) then (CH2Cl2:MeOH 95:5). ( Figure S8. Truncated 1 H NMR spectra of a) axle 1·(PF6)2 and b) [3]rotaxane 4·(PF6)2. In the absence of coordinating anions the signals for He,e' are perturbed upfield in the [3]rotaxane relative to the axle component revealing the macrocycle components interact significantly with the NDI stations (CDCl3, 298 K, 500 MHz).

Starting structure of XB [3]rotaxane
The structure of the macrocyclic components of 5 was taken from a previous work, [6] while the starting structure for the axle of 5 was generated via atomic manipulation of the crystal structure deposited with the Cambridge Crystallographic Data Centre (CCDC), [7] under RefCode ZEZDOZ. [8] The XB [3]rotaxane was obtained by the assembly of one axle and two macrocycle molecules in an interlocked fashion with the bis-iodo triazolium binding units and the macrocyclic clefts adopting an almost orthogonal relative disposition. This structure was further used to generate the 5·(Cl)2 and 11·NO3complexes for Molecular Mechanics (MM) and Molecular Dynamics (MD) study.

Quantum calculations
All quantum calculations were carried out with the Gaussian09 software, [9] and included the derivatisation of restrained electrostatic potential (RESP) charges for the [3]rotaxane's axle component, as well as the preliminary parameterisation of XB interactions on the chloride and nitrate complexes' models as detailed below.

Classical force field calculations
All MM and MD simulations were carried out with Amber14. [10] The rotaxane components (axle and macrocycle) were described with parameters taken from the Generalized Amber Force Field (GAFF) [11,12] and RESP charges. [13] The nitrate anion was described with bond terms force field parameters derived at the MP2/6-31+G* level, listed in Table S1 along with its RESP charges (vide infra). The van der Waals parameters were directly taken from GAFF. The chloride anions were described with a -1 discrete charge and van der Waals parameters developed for the TIP3P water model. [14] The all-atoms models of the methanol and chloroform solvent molecules were described with force field parameters and charges taken from refs. 19 and 20, respectively. The force field parameters and charges of the PF6counter-ion were taken from ref. [17] . The post-processing of trajectory files to obtain the structural data was performed with cpptraj. [18]

Parameterisation of XB interactions
The force field parameterisation of XB interactions was preceded by DFT optimisations of model complexes composed by a model axle incorporating the bis-iodo-triazolium motif (8Ph) and chloride or nitrate anions. In 8Ph, the bulky stoppers, the naphthalene diimide stations and the biphenyl spacer moieties of the axle were replaced by phenyl rings. These preliminary quantum calculations were carried out using the B3LYP functional, with hydrogen, carbon, nitrogen, oxygen and chlorine atoms treated with the 6-311++G** basis set, while the iodine atoms were described with the aug-cc-pVDZ-PP basis set, [19,20] obtained from the EMSL website. [21,22] The optimised structures of 8Ph·Cl2 and 8Ph·NO3are shown in Figure S12 and allowed us to ascertain, in gas-phase, the I···Cldistances of 2.638 Å and I···O distances of 2.592 Å, which guided the further parameterisation of the halogen bonding interactions. Likewise in our previous classical molecular dynamics investigations on anion recognition, [23][24][25] the halogen bonding interactions were represented by the addition of a massless extra-point (EP) of charge to GAFF, following the methodology established by Ibrahim. [26] Several I-EP distances were systematically tested in gas phase via MM energy minimisations of the complexes 8Ph·Cl2 and 8Ph·NO3 -, using RESP charges estimated for each I-EP distance. In a previously HF/6-31G* optimised structure of 8Ph, an extra point was positioned in front of each C-I bond at the desired I-EP distances and then the corresponding atomic charges were obtained via two stages RESP fitting (vide infra). The ranges of I-EP distances tested for the 8Ph·Cl2 and 8Ph·NO3model complexes are gathered in Tables S2 and S3, respectively, together with MM I···Cland I···O distances and the iodine and EP charges. Tables S2 and S3 show that several I-EP distances can be applied in the MM geometry optimisations of 8Ph·Cl2 or 8Ph·NO3 -. However, the subsequent use of the longer I-EP distances (2.31 to 2.33 Å for 8Ph·Cl2 and 2.05 to 2.12 Å for 8Ph·NO3 -) MD simulations of these complexes, carried out in gas phase at 300K for 5 ns, were shown to be instable, leading definitively to the exclusion of these distances. Therefore, these MD simulations were extended to shorter I-EP distances, and the 2.30 and 2.04 Å. These I-EP distances were further validated in gas phase MD simulations of pseudo [3]rotaxanes 5Ph·Cl2 and 5Ph·NO3 -, which were obtained by the assembly of two macrocycles with the 8Ph·Cl2 and 8Ph·NO3complexes, respectively. For 5Ph·Cl2 I···Claverage distances of 3.391 ± 0.099 and 3.393 ± 0.098 Å were calculated, while for 5Ph·NO3 -, I···O average distances of 2.930 ± 0.104 and 2.919 ± 0.099 Å were assessed. Therefore, the 2.30 and 2.04 Å I-EP distances were appointed for the subsequent calculation of the final RESP charges of the capped bis-iodo triazolium anion recognition fragment.

Calculation of RESP charges on the XB [3]rotaxane
The RESP charges for the axle component were derived using a fragment-based approach due to its large size (318 atoms). Thus, two capped entities were built as sketched in Figure S13. The 8stopper fragment comprises the bulky stopper tris(p-tertbutylphenyl)(phenyl)methane, a naphthalene diimide station and a biphenyl spacer moiety capped with a methyl group. The 8Ph fragment incorporates the axle bis-iodo triazolium anion recognition site with the bulky stoppers, the naphthalene diimide stations and the biphenyl spacer moieties capped by phenyl substituents. These hypothetical molecules were optimised using the HF method coupled with 6-31G* basis set for all atoms, apart of the iodine atoms, which were treated with the aug-cc-pVDZ-PP basis set (vide supra). The RESP atomic charges were obtained in a two RESP charge fitting stages from the electrostatic potential estimated at the same level of theory using the Gaussian IOp: 6/33=2, 6/41=4, 6/42=6. Moreover, for 8Ph the following constraints were imposed: a) each methylene group (linking the central bisiodo triazolium anion recognition site and aromatic capping groupsin red in Figure  S13), must have the same net charge as the methyl substituent (in blue in Figure S13) removed from 8stopper; b) the overall charge of the phenyl capping group and the methylene linker must amount to 0, while the net charge of the central bis-iodo triazolium anion recognition site (including the two extra-points) was set to 2.
After the charge calculation for 8stopper and 8Ph, their capping groups were removed, and the axle with RESP charges was generated attaching two capped 8stopper units to the methylene bridges of capped 8Ph central fragment. Figure S26. Capped entities used to derive the 8 axle RESP charges. The capping groups of 8Ph are identified in red and the capping group of 8stopper is identified in blue.
The RESP atomic charges for the macrocycle were the same as in our previous work. [6] The atomic charges for the nitrate anion given in Table S1 were also RESP charges determined from optimised at the HF/6-31+G* level using the number of layers and the number of points per layer quoted above.

General MD simulation methods
The starting geometries of 5·(Cl)2 or 5·NO3 -(vide supra) were minimised in gas phase by MM until the convergence criterion of 0.0001 kcal mol -1 was achieved. The MM optimised structures were then solvated in cubic boxes with 3387 chloroform molecules and 6690 methanol molecules in agreement with a 1:1 v/v solvent mixture used in 1 H NMR experimental binding studies. In addition, a PF6counter-ion was added to solvated 5·NO3to neutralise the system net charge. Each solvated system was equilibrated under periodic boundary conditions using the following multistage protocol. The system was relaxed by MM minimisation of solvent molecules and by keeping the solutes fixed with a positional restraint of 500 kcal mol -1 Å -2 . The restraint was then removed and the entire system was allowed to relax. Both minimisation stages comprised an initial set of 10000 steepest descent algorithm steps, followed by 10000 steps of conjugated gradient algorithm. The equilibration stage proceeded with heating up the system to 300 K for 100 ps using a NVT ensemble and a weak positional restraint (10 kcal mol -1 Å -2 ) on the solutes. Afterwards, each system's density was allowed to equilibrate in a NPT ensemble at 1 atm for 1 ns, at the same temperature, followed by a NPT data collection run for 100 ns. The collection run's trajectory frames were saved every 1 ps. Three independent replicates were performed for each system. The CUDA version of the PMEMD executable was used for the simulation of all solvated systems. [27][28][29] The bond lengths involving all bonds to hydrogen atoms were constrained with the SHAKE algorithm allowing the usage of 2 fs time step. [30] The Particle Mesh Ewald (PME) method was used to treat the longrange electrostatic interactions. [31] The non-bonded van der Waals interactions were truncated with a 10 Å cut-off.

Supplementary MD Movies Captions
Movie S1. Movie of the second MD run of 5·(Cl)2 (between the 1 st and the 15 th ns), showing the halfcircumrotation conversion process from co-conformation A, through B, into co-conformation C.
Movie S2. Movie of the first MD run of 5·NO3 -(between the 1 st and the 100 th ns), showing the stability of the anion association during the conversion between co-conformations A and B.