Reversible transformations between the non-porous phases of a flexible coordination network enabled by transient porosity

Flexible metal–organic materials that exhibit stimulus-responsive switching between closed (non-porous) and open (porous) structures induced by gas molecules are of potential utility in gas storage and separation. Such behaviour is currently limited to a few dozen physisorbents that typically switch through a breathing mechanism requiring structural contortions. Here we show a clathrate (non-porous) coordination network that undergoes gas-induced switching between multiple non-porous phases through transient porosity, which involves the diffusion of guests between discrete voids through intra-network distortions. This material is synthesized as a clathrate phase with solvent-filled cavities; evacuation affords a single-crystal to single-crystal transformation to a phase with smaller cavities. At 298 K, carbon dioxide, acetylene, ethylene and ethane induce reversible switching between guest-free and gas-loaded clathrate phases. For carbon dioxide and acetylene at cryogenic temperatures, phases showing progressively higher loadings were observed and characterized using in situ X-ray diffraction, and the mechanism of diffusion was computationally elucidated.


General information
Commercially available starting materials and solvents were purchased from Sigma Aldrich, Merck and Fluorochem. All reactions were monitored using aluminium backed silica gel Merck 60 F254 TLC plates and visualised using UV irradiation. Column chromatography was carried out with Merck silica gel 230-400 mesh silica gel.
Synthesis of 1,3-bib was carried out in a single step, according to a previously reported procedure with minor modifications. 1 1,3-dibromobenzene (10.0 g, 42.8 mol, 1.0 equiv), imidazole (14.5 g, 213.0 mmol, 5.0 equiv), CuI (1.63 g, 8.6 mmol, 20 mol%) and K2CO3 (29.4 g, 213.0 mmol, 5.0 equiv) were all added to anhydrous DMF (150 ml) under an inert N2 atmosphere. The resulting reaction mixture was heated to reflux under an inert atmosphere for 72 h. After cooling to room temperature, the reaction mixture was filtered. The filtered residue was washed with DCM (2 × 150 ml) and the filtrate was transferred to a large separating funnel. The organic layer was washed with water (3 × 250 ml), separated and dried over MgSO4. The organic layer was concentrated under reduced pressure and the resulting solid material was finally purified by trituration from a DCM/hexane mixture. The resulting 1,3-bib was isolated as a white solid (8.41 g, 94%).
Synthesis of dpt was carried out in two steps, according to a previously reported procedure. 2 Scheme 1: Two-step synthesis of dpt ligand.
Step 2: 2',5'-dimethyl-p-terphenyl (700 mg, 2.71 mmol, 1.0 equiv) was added to 20 ml of pyridine. 2.2 g of KMnO4 in 2.0 ml of H2O was then added and the reaction mixture was heated to reflux for 2 hours. After reaching reflux, every 30 min, an additional 1.0 g of KMnO4 in 2.0 ml was added (a total of 4 times). After 6 hours at reflux, a final 10 ml of water was added to the reaction mixture, which was allowed to reflux overnight. The MnO2 precipitate was hot filtered from the reaction mixture and washed with near boiling water (100 ml). The filtrate was acidified (pH 3-4) using conc. HCl, precipitating the dpt product as a white solid, which was collected by filtration, washed with 0.2 M HCl and finally dried in a 378 K oven overnight (732 mg, 85%).

Single-Crystal X-ray Diffraction
X-ray intensity data were recorded on a Bruker SMART APEX II 3 and a Bruker Quest APEX III equipped with a Mo or Cu sealed tube source. Both diffractometers employ an Oxford Cryosystems 700 Plus cryostat to control the temperature of the sample. Data reduction was carried out by means of standard procedures using the Bruker software package SAINT. 4 Absorption corrections and correction of other systematic errors were carried out using SADABS. 5 All structures were solved by direct methods using SHELXS-16 and refined using SHELXL-16. 6 X-Seed 7 was used as the graphical interface for the SHELX program suite. 5 Solvent-accessible voids can be visualised by calculating Connolly surfaces using MS- ROLL,8 another program incorporated into X-Seed. Hydrogen atoms were placed in calculated positions using riding models.

Activation Procedure
A suitable crystal of the as-synthesised material was selected and glued onto a glass fibre with cyanoacrylate glue. The glass fibre was then inserted into an environmental gas cell (EGC), which consists of a 0.3 mm Lindemann capillary secured to a steel nut with epoxy that is screwed into a valve body. The EGC allows for evacuation/pressurisation of the immediate crystal environment and transportation to a diffractometer. The EGC was then connected to a Pfeiffer Hi-Cube vacuum pump (pressure: ~ 3 x 10 -3 mbar) and immersed in oil, which was heated to 393 K overnight. The valve was then closed and the EGC removed from the activation apparatus.

Apohost structure determination
The evacuated crystal in a EGC was mounted onto a conventional goniometer and SCXRD data was recorded at 298 K. showing the effect of evacuation on the cavity dimensions. Hydrogen atoms have been omitted for clarity.

Gas loading
The activated crystal in the EGC was attached to a CO2 cylinder via a gas manifold (regulator). The system was pressurised to 1 bar and left to equilibrate under static pressure 8 for 3 h (multiple equilibration times were tested and this was found to be the best) after which the EGC was closed and loaded onto the diffractometer.

bar CO2 loaded structure determination
In the case of the gas loaded structure (1′CO2), the crystal was non-merohedrally twinned (split) and was therefore treated as a twinned crystal. Two crystal domains could be separated manually in the reciprocal lattice viewer and were processed as a non-merohedral twin during integration and scaling (TWINABS). Refinement was carried out using only the HKLF4 file, the HKLF5 file did not improve the refinement. An additional TWIN card was used in the refinement (BASF 0.39) to account for minor twinning still present. Thermal parameters were restrained using RIGU and SIMU cards.
The Checkcif contains the following A and B alerts:

Alert level A:
PLAT601_ALERT_2_A Unit Cell Contains Solvent Accessible VOIDS of 278 Ang**3 Response: Owing to the structural change imparted by the guest on the host, the structure undergoes a gate-opening phenomenon wherein the previously negligible cavities expand to accommodate CO2 guest molecules. Response: The crystal of 1′CO2 was non-merohedrally twinned. We believe this to be as a result of the structural transformations that occur as 1DMF converts to 1Apohost and then to 1′CO2. This coupled with a reduction in peak intensity resulted in a high R1 value. Response: As the 1apohost crystal undergoes progressive gas loading, it first converts to 1′CO2 and then to 1′′CO2. These phase transformations induce striation within the crystal resulting in reduced peak intensity, an elevated Rint and reduced bond precision. Our goal for this experiment was to attain the 1′′′CO2 phase however at 56 bar CO2 and 298 K, the crystal of 1apohost has converted to 1′′CO2. showing the effect of gas loading on the cavity dimensions. Guest molecules and hydrogen atoms have been omitted for clarity.

Thermogravimetric Analysis (TGA)
Thermogravimetric analyses (TGA) were performed under N2 using a TA Instruments Q50 system. A sample was loaded into an aluminium sample pan and heated at 283 K min -1 from room temperature to 773 K.

Differential Scanning Calorimetry (DSC)
Differential scanning calorimetry was carried out using a TA Instruments Q2000 differential scanning calorimeter. Samples were prepared by crimping the sample pan and lid (a pin hole was placed in the lid to prevent pressure build-up). A reference pan was prepared in the same manner for each analysis. Analyses were generally carried out in the temperature range 253 K -523 K and a general experimental procedure consisted of two heating/cooling cycles while the heat flow into or out of the sample, relative to the reference, was measured as a function of time and temperature in a controlled atmosphere. N2 gas, flowing at a rate of 50 ml min -1 was used to purge the furnace. The resulting thermograms were analysed

BET Surface Area
BET surface area corresponding to 516.6 m²/g was determined from the 77 K N2 sorption isotherm (Fig. S15). 11  Eq.1 The determined volumes are presented with 5% error bars (Table S3) and show reasonable agreement with the assignment of phases based on in situ PXRD, and SCXRD determined solvent-accessible volumes. This is an essential step as the sample adsorbs water from the atmosphere. A CO2 sorption experiment was carried out up to 100 kPa at 195 K. In-situ PXRD patterns were measured simultaneously at each equilibrium point of the adsorption and desorption isotherm. In addition to CO2, in situ PXRD patterns were measured at each equilibrium point of the adsorption isotherm for N2 at 77 K and C2H2 at 189 K.           The PXRD pattern of 1′′N2 is in good agreement with that of 1′′CO2. 9. Switching materials reference table   Potential for transient transport between neighbouring isolated cavities was examined along two potential pathways in the 1′CO2 structure. An initial position was obtained from canonical Monte Carlo runs of CO2 sorption. A single sorbate position settled near the posited transport window was selected from each of the trajectories and settled into the local minima via simulated annealing. A single unit cell was then taken to probe transport between diagonally adjacent voids as each 1′CO2 unit cell contains two cavities which have that relative orientation. For the stacked cavities a unit cell was replicated along the c-axis to create a two unit cell system with a single sorbate molecule (two cells were used as the stacked relative orientation is not observed within a single unit cell). These systems were then fully relaxed using CP2K. Using these relaxed initial positions, a series of sequential relaxations were performed wherein the optimized structure was modified, shifting the sorbate in the direction of the postulated transport channel and re-relaxed subject to the constraint that a single sorbate oxygen atom was constrained in one direction (the b-axis for the diagonally adjacent and the a-axis for the stacked). This forced proximity to the channel while allowing the atom to shift in the remaining two dimensions in order to find the optimal route through. No further constraints were put on the other sorbate atoms or the crystal, permitting sorbate orientation and geometry to relax along with the structure. The constrained optimized structures were then taken, the CO2 shifted further along the channel and re-relaxed. These iterations were repeated until a trajectory traversing each channel was obtained. The resultant trajectories are shown in animations 1-4 below. Having determined the pathways, the energies of each sequential configuration were then compared with the initial position energy to determine the energy barrier as reported in the "Computational insight into the mechanism of guest transport" section of the main text.
Dynamic molecular motion between cavities in the 1′CO2 structure was modelled using a single unit cell with one cavity loaded with 12 CO2 molecules and the second cavity left empty. Initial positions of sorbate occupants was taken from grand canonical Monte Carlo.
The single cavity loaded structure was subjected to DFTMD simulations allowing all atomic positions to fluctuate at 473 K and 623 K. Sorbate loading differential between cavities was found to provide sufficient motive impetus for sorbate molecules to diffuse through the barrier between diagonally adjacent cavities. Simulation across both temperatures reproduced the more energetically favorable mechanism of diffusion via rotation of the dpt linker.