Hydrogen Isotope Separation Using a Metal-Organic Cage Built from Macrocycles.

Porous materials that contain ultrafine pore apertures can separate hydrogen isotopes via kinetic quantum sieving (KQS). However, it is challenging to design materials with suitably narrow pores for KQS that also show good adsorption capacities and operate at practical temperatures. Here, we investigate a metal-organic cage (MOC) assembled from organic macrocycles and ZnII ions that exhibits narrow windows (<3.0 Å). Two polymorphs, referred to as 2α and 2β, were observed. Both polymorphs exhibit D2 /H2 selectivity in the temperature range 30-100 K. At higher temperature (77 K), the D2 adsorption capacity of 2β increases to about 2.7 times that of 2α, along with a reasonable D2 /H2 selectivity. Gas sorption analysis and thermal desorption spectroscopy suggest a gate-opening effect of the MOCs pore aperture. This promotes KQS at temperatures above liquid nitrogen temperature, indicating that MOCs hold promise for hydrogen isotope separation in real industrial environments.

Introduction

Deuterium (D2) is a crucial fuel for future fusion power plants. D2 is also used as a neutron moderator, in neutron scattering experiments, and as a nonradiative isotope tracer. These applications require high purity D2, which is non‐trivial because of its low natural abundance of 0.0156 mol %. Typically, D2 is purified industrially using the Girdler sulfide process or by cryogenic distillation. However, both methods are inefficient and energy‐intensive.[ , ] An attractive alternative to separate D2 from its dominant isotope, hydrogen, is to adsorb D2 selectively on a microporous bed.

Kinetic quantum sieving (KQS), first proposed by Beenakker et al., describes the effect of lighter isotopes with larger de Broglie wavelengths encountering higher energy barriers as they diffuse through fine pores at cryogenic temperatures. KQS effects lead to differences between the diffusion rates of isotopes, making it a potential process for the separation of D2 from H2. KQS requires adsorbents with ultrafine pore apertures; typically <5 Å, with pore apertures of 3.4 Å reported as the optimal size in rigid frameworks under cryogenic conditions. Owing to their small pore sizes, several microporous materials have been investigated for KQS of hydrogen isotopes, including porous carbons, zeolites, metal‐organic frameworks, and covalent organic frameworks. We also recently reported a porous organic cage (POC) co‐crystal that combined cages with narrow pores and cages with good capacities to achieve optimal separation performance in KQS. However, despite some recent success, it remains challenging to precisely tune the pore size to the desired level required for KQS without compromising the adsorption capacity of the material. Also, the best selectivities tend to be achieved at very low temperatures (30–40 K), which is energetically costly.

Metal–organic cages (MOCs), also known as metal–organic polygons or polyhedrons (MOPs), are discrete molecules with intrinsic cavities, formed from metal cations and organic linkers. Like organic cages, many MOCs have good solubility in a range of solvents and thus can be processed into different forms to optimize their structures and functions. MOCs can also contain open metal sites, which can enhance their gas adsorption properties. To date, MOCs have been demonstrated to selectively adsorb CO2, O2, CO, NO, and C3H8. due to their specific pore sizes or open metal sites. With small pores that could be suitable for KQS and, potentially, open metal sites to enhance adsorption affinity, MOCs are interesting candidates for hydrogen isotope separation.[ , ]

The solid‐state porosity of MOCs is affected by guest accessibility to the intrinsic MOC cavity and extrinsic porosity in their structures. The intrinsic porosity of MOCs can be controlled by choosing appropriate organic linkers and metal centres, or by post‐synthetic modification. The extrinsic porosity in MOC solids can be controlled, to an extent, by using crystal engineering methods. An important consideration is that the organic units (or ligands) are key structural components of MOCs and this significantly affects their molecular flexibility and porosity.[ , ] Although linear or planar organic linkers are the most common building units for MOCs, macrocycles with intrinsic voids or cavities, such as porphyrins calixarene, and calixsalens have emerged recently as alternative MOCs building units. With their own intrinsic prefabricated cavities, macrocycles can enrich the functionality and structural diversity of MOCs. For example, calixarene‐based macrocycles have been coordinated to tetranuclear clusters to form permanently porous MOCs with surface areas as high as 1239 m2 g−1.

Calixsalen macrocycles have a bowl‐like shape and small intrinsic cavity (Figure 1). They were first reported in 1999 by Jablonski et al., who condensed 2‐hydroxyisophthalaldehyde derivatives with chiral (1R,2R)‐cyclohexane‐1,2‐diamine. In 2018, Barbour et al. found that calixsalens with Cl and Br substituents displayed remarkable sorption properties for ethylene and carbon dioxide, which suggested the potential of adjusting their sorption properties by introducing functional groups. As shown in Figure 1b, the enantiopure [3+3] calixsalen macrocycle, 1‐(R,R), has a rim and a tail. The rim consists of three cyclohexane rings and three hydroxy‐substituted aromatic rings, while the tail consists of three bulky tert‐butyl groups. In the structure of 1‐(R,R), the three salen units have the same orientation, and coordinating 1‐(R,R) to metal ions, including ZnII, has been used previously by Lisowski et al. to connect molecules of 1‐(R,R) and form the trinuclear Zn MOC, 2 (Figure 1b). The same team also reported that 2 has a Brunauer–Emmett–Teller surface area (SABET) of 610 m2 g−1 after being activated at ambient temperature, and this material can enantioselectively bind chiral alcohols. Most recently, 2 has been used for gas chromatographic separation, and chiral separation.

Figure 1

a) Structure of macrocycle 1‐(R,R). b) The self‐assembly of 2 from macrocycle 1‐(R,R) using ZnII ions. c) Window diameter and pore diameter distribution histograms of 2. The window diameter and pore diameter distribution histograms share the bottom X‐axis. Atom colours: Zn, orange; N, blue; O, red; C, grey.

The previously reported solvated single crystal structure of 2 shows that each MOC has a hollow cavity in a tubular shape with two narrow windows at both ends. As shown in Figure 1c, the window diameter and pore diameter distribution histograms were calculated by py‐window based on xTB MD trajectory of 2 (see Supporting Information Section 1.5.8). The pore diameter of 2 is around 3.1 Å. In addition, the broad window diameter distribution below 2 Å is due to the rotation of the three tert‐butyl groups. Furthermore, the window diameter of around 2.8 Å suggests that 2 has the optimal diameter for KQS. Here, we investigated the hydrogen isotope separation performance of MOCs for the first time. We successfully obtained the single‐crystal structures of the two activated polymorphs of 2, referred to here as 2α and 2β. We found there are slight differences in the activated crystal structures between the orientation of tert‐butyl groups and the crystal packing of the MOCs. These structural differences led to contrasting D2/H2 separation performance, which was evaluated by thermal desorption spectroscopy (TDS).

Results and Discussion

The [3+3] calixsalene macrocycle, 1‐(R,R), shown in Figure 1a, is synthesized by reacting (1R,2R)‐cyclohexane‐1,2‐diamine with 4‐tert‐butyl‐2,6‐diformaylphenol.[ , ] Reacting 1‐(R,R) with ZnII acetate in a 2 : 3 ratio in methanol affords 2, which comprises two triply deprotonated macrocycles 1‐(R,R)3− held together by three ZnII ions in a sandwich‐like conformation (Figure 1b).[ , ] Solvated single crystals of 2 with monoclinic Cc space group symmetry, referred to as MeOH@2, are obtained as large yellow block crystals from the reaction solvent,[ , ] (Figure 2a and Table S1) and a previous study reported that activated crystals of 2 were unsuitable for single‐crystal X‐ray diffraction analysis. However, in this study, we successfully obtained two solvent‐free single‐crystal structures of 2 after thermally activating high‐quality crystals of MeOH@2, grown in the reaction solvent at room temperature over 12 hours. To activate MeOH@2, we collected the solvated crystals by filtration and heated the sample under a vacuum at 80 °C. Thermogravimetric analysis (TGA) and NMR were used to confirm that the crystals were fully activated (Figure S2, S6). Removing the reaction solvent at 80 °C under vacuum induces MeOH@2 to transform into a new phase, referred to as 2α, which has a different crystal packing of MOCs and lower triclinic P1 space group symmetry (Figure 2a and see Table S1). Further heating of 2α at 180 °C under vacuum for 12 hours yielded another new polymorph, 2β (Figure 2a), again with triclinic P1 space group symmetry (see Table S1). We used differential scanning calorimetry (DSC) and in situ variable‐temperature powder X‐ray diffraction analysis (PXRD) to monitor the transformation between 2α to 2β (Figure S7, S8b). As shown in Figure S7, the DSC curve of 2α shows an endotherm between 168 °C to 183 °C, which we attribute to the transformation from 2α to 2β. By contrast, the DSC of 2β did not show any endotherms over this temperature range, suggesting 2β is the more thermostable phase. Furthermore, in situ variable‐temperature PXRD data (Figure. S8b) indicates that 2α is stable between room temperature and 125 °C, but at 160 °C begins to transform. At 180 °C, the PXRD pattern closely matched the simulated PXRD of 2β, which is consistent with the DSC curve of 2α.

Figure 2

Crystal packing images of MeOH@2 (top, methanol solvent is omitted for clarity), 2α (middle) and 2β (bottom) in a) top view and b) side view. H atoms are omitted for clarity (orange, Zn; blue, N; red, O; grey, C). The intrinsic cavity is highlighted in yellow. c) PSD (pore size distributions) histograms for MeOH@2 (black), 2α (blue,) and 2β (red).

As shown in Figure 2a and 2b, 2 in MeOH@2, 2α and 2β has a tubular‐shaped intrinsic cavity shape, highlighted using a yellow cylinder, with a wider cavity at the centre, highlighted using a yellow sphere. In MeOH@2, there are also large solvent‐filled extrinsic voids between aligned molecules of 2. After thermally removing solvent from MeOH@2 at 80 °C under vacuum to form 2α, the tubular‐shaped MOC cavities are preserved, but the MOCs pack more closely, and the extrinsic voids are smaller. For comparison, after deleting the solvent from MeOH@2, the solvent‐accessible volume calculated by Platon using a probe radii of 1.2 Å accounts for about 36.9 % of the unit cell volume. By contrast, the comparable void volume in 2α is 17.3 % of the unit cell volume. In 2β, which was obtained by heating 2α at 180 °C under vacuum, the void volume increased slightly to about 20.6 % of unit cell volume. This difference is mainly due to the crystal packing of 2 because the molecular overlay plot of 2α (blue) and 2β (red) shown in Figure S20 indicates that the molecular structures of 2 in 2α and 2β are very similar. In addition, we used Zeo++ to compare the pore structures in MeOH@2, 2α, and in more detail and calculate their pore size distributions (PSD, see Supporting Information Section 1.5.8 for full details). These calculations were performed after removing the solvent, including residual water, from the crystal structures. The largest free spheres (Df) in MeOH@2, 2α, and 2β, calculated by Zeo++ are 3.4, 2.0 and 2.6 Å, respectively, and these values represent their pore limiting diameters. The PSD plot of MeOH@2 shows three peaks at 3.4 Å, 4.3 Å and 4.6 Å (Figure 2c). In the PSD plots of 2α and 2β, there are also three peaks, but the centres vary from 3.6, 4.1, and 5.1 Å in 2α to 3.0 Å, 3.8 Å and 4.1 Å in 2β. Therefore, the pore sizes of 2α and 2β are around 2.0 to 5.1 Å and 2.6 to 4.1 Å, respectively, satisfy the requirement of porous materials for KQS applications.

To quantify how the PSD of 2α and 2β affected their porosity, we measured their N2, Ar, CO2 and H2 gas adsorption isotherms. As shown in Figure 3a, 2α only adsorbed a small amount of N2 (<0.6 mmol g−1) below P/P 0=0.015. However, after reaching a P/P 0 of 0.015, we observed a step in the adsorption isotherm, and the N2 uptake increased sharply to >5 mmol g−1. This result is consistent with the pressure‐induced gating effect previously reported for 2. By contrast, 2β has a Type‐I N2 isotherm at 77 K (Figure 3a), and in the absence of pressure‐induced gating effect observed for 2α, had the lower uptake at 1 bar (7.0 mmol g−1 for 2α vs. 4.2 mmol g−1 for 2β) and calculated SABET (394 m2 g−1 for 2α vs. 270 m2 g−1 for 2β). Hence, even though the calculated void volume in 2β was higher than 2α (17.3 % for 2α vs. 20.6 % for 2β), the pressure‐induced gating effect appears to have created additional porosity in 2α.

Figure 3

a) N2 sorption isotherms at 77 K for 2α (blue) and 2β (red). b) CO2 sorption isotherms at 273 and 298 K for 2α. c) CO2 sorption isotherms at 273 K and 298 K for 2β. Solid symbols: adsorption; hollow symbols: desorption.

To further investigate the flexibility of the pore and the void volume for the gas sorption, higher pressure CO2 and H2 isotherms were measured for 2α and 2β from 0 to 10 bar. As shown in Figure 3b and 3c, two successive plateaus in the high‐pressure CO2 isotherm of 2α and 2β at 273 K indicate a pressure‐induced gating effect, but this effect appeared more profound for 2α than 2β. At 298 K, the CO2 isotherm hysteresis was less extensive for 2α and 2β. A pronounced hysteresis loop was found in the H2 isotherms of 2α at 77 K from 0 to 10 bar. By contrast, no distinct hysteresis loop was observed in the H2 isotherms of 2β (Figure S14). In conclusion, the higher‐pressure isotherms are consistent with the N2 isotherms, which suggest that 2α is more structurally flexible than 2β.

H2 and D2 gas sorption isotherms for 2α and 2β recorded at various temperatures (30, 50, 77, and 100 K) are shown in Figure 4. 2α has low D2 and H2 gas uptakes at 30 and 50 K (≈1 mmol g−1 or lower at 1 bar), but much higher uptakes of 3.8 mmol g−1 for H2 and 4.9 mmol g−1 for D2 at 77 K and 1 bar. The increased H2 and D2 uptakes at the higher temperatures, where surface adsorption effects are also likely to be less profound, indicate that D2 and H2 can more easily penetrate the MOC cavities at 77 K. In addition, the 2α isotherms exhibit hysteresis, which is less pronounced at 100 K, implying faster equilibration at higher temperatures. The stronger hysteresis at lower temperature, which decreases at higher temperatures, denotes a temperature‐dependent diffusion limitation of gas molecules penetrating the cavities through different aperture sizes. Unlike typical isotherms that feature lower uptakes at increased temperature, 2α has lower H2 and D2 uptakes at 30 and 50 K compared to that observed at 77 K. This can be ascribed to a very narrow pore aperture at low temperature, which prohibits gas molecules from diffusing deeply into the 2α crystals. The gas uptakes, therefore, increase with higher temperature, denoting a larger pore aperture that is thermally opened due to the thermal vibration of the flexible windows. By contrast, the D2 and H2 adsorption isotherms for 2β reached a maximum uptake of 7.5 mmol g−1 for D2 and 5.9 mmol g−1 for H2 at 30 K and 1 bar (Figure 4b and d), indicating that this MOC crystal is fully accessible to these gases even at lower temperatures. There was again strong hysteresis at 30 K, but again this hysteresis was reduced with increasing temperature, indicating better equilibrium. For 2α and 2β, we observed higher D2 uptakes than H2 at all measurement temperatures, which we attribute to higher diffusion rates and increased heats of adsorption for D2. The D2 gas sorption isotherms show that 2β has a two‐step H2 adsorption isotherm at 30 K. Two‐step D2 adsorption isotherms for 2β were also observed at 30 and 50 K. The stepped isotherms can be attributed to the phase transition of the adsorbate. Similar phenomena have been observed before: for example, a step in the low‐pressure region of a MOF (MET‐2) Ar and N2 isotherms has been reported by Gándara et al., which was attributed to the phase transition of the adsorbate. This suggests that 2β can accommodate extra gas molecules after all the initially accessible adsorption sites are fully occupied. For 2β, when the pores are saturated with 5 mmol g−1 of H2 or 6 mmol g−1 of D2, further dosing of gases appears to open additional adsorption sites as the gas pressure increases. We used PXRD to confirm that the crystal structures of 2α and 2β did not change profoundly after the measurements. (Figure S10) The PXRD data show that the 2α and 2β remained in the same phase and were crystalline after H2 and D2 isotherms. Variable‐temperature PXRD also confirmed that 2α and 2β are stable between 30 to 100 K. (Figure S9) Hence, the stepped isotherms are due to local flexibility rather than more profound structural changes. Even though it has a similar structure, pore size, and accessible pore volume, similar steps were not observed in 2α. We note that the structure of a physisorbed layer is dependent not only on the adsorbate‐adsorbate interactions but also on the magnitude and disposition of the adsorbent‐adsorbate interactions. This difference can be ascribed to the distinctly different flexibility of 2α and 2β at very low temperatures (30 and 50 K), leading to a different accessible pore aperture.

Figure 4

H2 isotherms of a) 2α and b) 2β; D2 isotherms of c) 2α and d) 2β recorded at 30 K, 50 K, 77 K and 100 K. Solid symbols: adsorption; hollow symbols: desorption.

Encouraged by the apparent faster D2 diffusion kinetics, we evaluated the ability of 2α and 2β to perform D2/H2 separations using a laboratory‐designed cryogenic thermal‐desorption spectroscope (TDS). First, we used pure H2 and D2 atmospheres in the TDS experiments to determine the preferred H2 and D2 adsorption sites in 2α and 2β. In these TDS measurements, 2α and 2β samples were exposed at room temperature to 10 and 200 mbar of pure H2 and D2, then cooled to 20 K under a gas atmosphere, and finally, the sample chamber was evacuated at 20 K. The TDS spectra were recorded while heating the samples from 20 and 170 K (Figure S19). The TDS spectra of 2α (Figure S19a) collected after gas loading at 10 mbar shows little gas uptake. By contrast, the 2α TDS spectra obtained from the 200 mbar gas loading shows one major peak for both isotopes, centered at about 102 K for D2 and 106 K for H2, with shoulders appearing at desorption temperatures below 70 K. The presence of the single peak with a shoulder shows there are at least two adsorption sites for H2 and D2 in 2α with different adsorption potentials. The increase in gas uptake of both isotopes with higher pressure indicates the inner structure becomes more accessible at 200 mbar than 10 mbar. The TDS spectra for 2β after gas loading at 10 mbar of H2 and D2 (Figure S19b) show small peaks at around 25 K and more profound peaks at around 90 K. In addition, two shoulders were observed over the temperature range 40–80 K. The TDS spectra of 2β after gas loading at 200 mbar have one major peak at lower desorption temperatures; 86 K for D2 and 92 K for H2, compared to values of 102 K for D2 and 106 K for H2 observed for 2α. We attribute the first desorption peak at low temperature (approximately 25 K) to weakly adsorbed gas molecules on the outer surfaces of the crystals. For both 2α and 2β, the temperature of the maximum desorption peak is lower for D2 than H2, denoting a faster diffusion of D2, which is in good agreement with H2 and D2 isotherms.

In the TDS spectra, the area under the desorption peak is proportional to the quantity of desorbed gas, which can be quantified by calibrating the mass spectrometer using a Pd95Ce5 alloy (see Supporting Information Section 1.5.7). At 200 mbar, the pure H2 and D2 uptakes for 2α are 1.2 and 2.1 mmol g−1, respectively. By contrast, the pure H2 and D2 uptakes 200 mbar for 2β are higher and were 2.4 and 3.8 mmol g−1, respectively, in good agreement with gas sorption isotherms.

We next used the TDS measurements to verify the competitive separation performances of 2α and 2β. These measurements were performed after directly exposing 2α and 2β to a 1 : 1 H2 : D2 mixture (200 mbar) for 10 min at exposure temperatures (T exp) between 30 and 100 K. The D2/H2 selectivity is then calculated from the ratio of the peak areas. The TDS spectra of 2α are shown in Figure 5a, and Figure 5c shows the D2/H2 selectivity alongside the corresponding D2 uptake as a function of exposure temperature. The TDS spectra start from T exp and measure the remaining free gas molecules released during the evacuation processes that are carried out at the same temperature. The gas uptakes increase with increasing temperatures, exhibiting a maximum of 100 K. Meanwhile, the selectivity decreases with increasing T exp, exhibiting the highest S =9.1 at 30 K. Generally, the strongest adsorption site, which corresponds to the highest desorption temperature, is occupied first and at very low loadings. The weaker sites are then occupied at higher gas loadings and this results in additional low‐temperature desorption peaks. However, the H2 and D2 TDS spectra of 2α vary in shape and magnitude depending on T exp, which is contrary to the typical sequential filling behavior of accessible sites with different binding strengths. No desorption peak can be observed above 60 K for T exp=30 K, implying no deep penetration into the structure at this temperature. With increasing exposure temperature, gases can penetrate deeper into the crystals, and the desorption peaks in TDS spectra shift to higher temperatures. The gas molecules can finally penetrate the MOC crystals at T exp=100 K. These temperature‐dependent TDS spectra agree with the observation from pure gas isotherms, which is related to the temperature‐dependent gate‐opening behaviour. In contrast, 2β shows different desorption spectra under identical conditions, as shown in Figure 5b. For 2β, the desorption spectrum at 30 K shows no desorption of any isotopes occurring above 50 K, indicating the weak adsorption of gas molecules on top of the surface. However, the TDS spectra measured at 50 K exhibit two desorption maxima that last until 120 K, indicating the gas molecules can freely access the crystal pores at 50 K. The D2/H2 selectivity and its corresponding D2 uptake as a function of exposure temperature are shown in Figure 5d, in which the highest S of 8.3 is observed at T exp=30 K. There is no considerable difference in the selectivity of 2α and 2β for D2/H2 (Figure 5c, d) from 30 K to 100 K. However, the D2 uptake for 2α only slowly increases with increasing temperature, whereas the D2 for 2β is far higher at 1.1 mmol g−1 at 77 K. The difference can be ascribed to a higher temperature for the opening of the pore aperture of 2α, in which the exposed gas penetration is still limited at 77 K and needs a higher temperature for opening. The gas uptake in 2α thus reaches the maximum at 100 K. By contrast, the aperture of 2β fully opens at 77 K, allowing the gas molecules to be removed during the evacuation at exposure temperature prior to the TDS run. In addition, 2β has a higher D2 uptake maximum than 2α. The increasing isotope uptake with decreasing selectivity is related to the opening of the aperture and the sufficient kinetic energy of the molecule, where the accessibility of both isotopes is enhanced. Despite low selectivity, after exposure at T exp=100 K to an isotope mixture, the desorption maxima are centered at 115 K for both 2α and 2β, which is an unusual case exhibiting such high desorption temperature without the presence of strong adsorption sites. Therefore, both 2α (S =2.8) and 2β (S =2.2) have reasonable selectivity for D2 at 77 K compared with other porous materials without open metal sites. (Table S2) Generally, desorption above liquid N2 temperatures indicates the existence of strong binding sites, such as uncoordinated metals (Figure 6a). However, the lack of open metal sites in MOCs limits the interaction with hydrogen molecules. A similar phenomenon has been reported by Mondal et al. in a series of I−D channel ultra‐microporous MOFs, called Imidazolate Framework Potsdam (IFP), which possess smaller pore size than the kinetic diameter of hydrogen molecules, which gets accessible to hydrogen by a temperature‐dependent dynamic gate opening. IFP‐4 and IFP‐7 exhibit a selectivity around 2 above liquid N2 temperature, which can be ascribed to KQS occurring only at the outermost pore aperture, since after penetrating the narrow pore channels, no passing of gas molecules is possible, and a single‐file filling occurs (Figure 6b). Additionally, the gas uptake at 77 K in IFP only reaches 0.01 mmol g−1 in IFP‐4 and 0.05 mmol g−1 in IFP‐7, due to the narrow I−D channels. By contrast, the gas uptake is much larger for our MOCs (0.41 mmol g−1 in 2α and 1.10 mmol g−1 in 2β at 77 K). In contrast to the rigid 1D channels in the IFP series, the ultrafine narrow pores in MOC 2 allow single‐file filling initially, but the flexible pores are adaptable and become more accessible to the target gas with increasing temperature. When the system is cooled down to low temperatures under a gas atmosphere, the gas molecules are captured on the additional inner surface and can be released by heating up to this temperature again (Figure 6c). Therefore, by introducing local flexibility into the MOC system, the practical temperature for hydrogen isotope separation can be increased dramatically and become comparable to the temperatures reached by MOFs possessing open metal sites.

Figure 5

H2 (open) and D2 (close) thermal desorption spectra of 200 mbar 1 : 1 H2/D2 isotope mixture on a) 2α and b) 2β at different exposure temperatures, 30 K (green), 50 K (red), 77 K (blue), and 100 K (magenta). D2/H2 selectivity (blue) and the corresponding D2 uptake (red) as function of exposure temperature for c) 2α and d) 2β.

Figure 6

Three types of structures that could increase the practical temperature of hydrogen isotopes separation: a) Open metal sites. b) Single‐file filling in ultrafine 1D channels. c) This work: flexible narrow pores.

Conclusion

We have investigated two new polymorphs of a trinuclear Zn MOC for D2/H2 separation. The two polymorphs were isolated by activating crystals of MeOH solvate of 2, initially at 80 °C to afford 2α, and then at 180 °C to transform 2α into 2β. Surprisingly, 2β had a slightly larger extrinsic porosity than 2α. There are other differences between the crystal structures, and 2α has a more extensive range of pore sizes from about 2.0 Å to 5.1 Å and appeared more flexible in gas sorption measurements, which led to 2α having a higher BET surface area of 393.8 m2 g−1 compared to 2β (269.9 m2 g−1). 2β has a slightly higher unit cell void volume than 2α (20.6 % for 2β vs. 17.3 % for 2α), and 2β thus has a higher pure D2 capacity of 3.8 mmol g−1 than the 2α (2.1 mmol g−1) at the exposing pressure of 200 mbar. In addition, TDS measurements confirm that the D2 adsorption capacity with 2β (1.10 mmol g−1, S =2.2) is higher than for 2α (0.41 mmol g−1, S =2.8) at 77 K for 1 : 1 D2‐H2 mixture. Furthermore, the local flexibility of MOC crystals provides the additional accessible inner surface to increase the adsorption and separation of hydrogen isotopes inside the crystals via KQS. This leads to a desorption temperature of over 100 K even without existing strong adsorption sites. Hence, the selectivity and capacity of MOCs for hydrogen isotopes are sensitive to their pore size and flexibility. This study paves the way for hydrogen isotope separation above liquid nitrogen temperatures based on well‐defined pore structures and the flexibility of molecular materials.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Click here for additional data file.

Supporting Information

Click here for additional data file.

Supporting Information

Click here for additional data file.

Supporting Information

Click here for additional data file.