Research Article

Barely porous organic cages for hydrogen isotope separation

See allHide authors and affiliations

Science  01 Nov 2019:
Vol. 366, Issue 6465, pp. 613-620
DOI: 10.1126/science.aax7427

Quantum sieves for hydrogen isotopes

One method for improving the efficiency of separation of hydrogen from deuterium (D) is to exploit kinetic quantum sieving with nanoporous solids. This method requires ultrafine pore apertures (around 3 angstroms), which usually leads to low pore volumes and low D2 adsorption capacities. Liu et al. used organic synthesis to tune the pore size of the internal cavities of organic cage molecules. A hybrid cocrystal contained both a small-pore cage that imparted high selectivity and a larger-pore cage that enabled high D2 uptake.

Science, this issue p. 613


The separation of hydrogen isotopes for applications such as nuclear fusion is a major challenge. Current technologies are energy intensive and inefficient. Nanoporous materials have the potential to separate hydrogen isotopes by kinetic quantum sieving, but high separation selectivity tends to correlate with low adsorption capacity, which can prohibit process scale-up. In this study, we use organic synthesis to modify the internal cavities of cage molecules to produce hybrid materials that are excellent quantum sieves. By combining small-pore and large-pore cages together in a single solid, we produce a material with optimal separation performance that combines an excellent deuterium/hydrogen selectivity (8.0) with a high deuterium uptake (4.7 millimoles per gram).

Deuterium (D) is used as a neutron moderator, as a nonradioactive isotopic tracer, and in neutron scattering experiments. These applications need high-purity deuterium, which is expensive because of its low natural abundance (0.0156 mol %). Typically, D2 is produced by electrolysis of heavy water and extracted through the Girdler-sulfide process (1) or by cryogenic distillation at 24 K (2). Both processes are costly and energy intensive because of the multiple enrichment steps required (3).

An attractive alternative to purify D2 from H2/D2 gas mixtures is to adsorb D2 selectively on a microporous bed (4). Kinetic quantum sieving (KQS) with nanoporous solids was first proposed by Beenakker et al. (5). The KQS effect becomes pronounced when the difference in size between a hydrogen molecule and the confining space is comparable to the thermal de Broglie wavelength of molecular hydrogen, λT=h/(2πmkBT)1/2 (h, Planck’s constant; m, mass; kB, Boltzmann constant; T, temperature). Quantum sieving has been exploited for the separation of gaseous isotope mixtures such as D2/H2 (6), but it is challenging to identify suitable porous solids. This is because KQS requires ultrafine pore apertures (~3 Å) (7, 8), and this typically leads to materials with low pore volumes and, hence, low D2 adsorption capacities, making such processes difficult to scale. An analogous selectivity-capacity trade-off—or, for membranes, a selectivity-permeance trade-off—is observed for a wide range of other gas separations that do not involve KQS (912).

Various porous materials have been studied for hydrogen isotope separation, including carbons (13, 14), carbon nanotubes (15), zeolites (16, 17), metal-organic frameworks (MOFs) (6, 18, 19), covalent organic frameworks (COFs) (20), and 2D crystals (21). MOFs and COFs have received much attention because of their crystalline nature and synthetically tunable pore size and functionality (22). However, even with MOFs or COFs, it is challenging to tune pore sizes at the exceptionally fine level required for KQS. For example, in MOFs, a common strategy for tuning the pore apertures is through systematic expansion or reduction of the number of the phenylene rings in the organic linker, with an increment or a decrement of ~2.8 Å (23); that is, too coarse for fine-tuning the aperture for optimum KQS.

Porous organic cages (POCs) (24, 25) are discrete molecules that have been used previously for the separation of xylene isomers (26), noble gases (27), and chiral molecules (27). POCs might also be promising candidates for H2/D2 separation but, unlike MOFs and COFs, it is difficult to tune the pore size in POCs simply by changing the constituent linkers. Because POCs are discrete molecules, small changes in their molecular structure can have profound effects on the solid-state packing of the cages and, hence, the pore structure (24). Modifying only the interior of the cage molecule, rather than the cage structure itself, could allow us to avoid altering the crystal packing. Internal functionalization is possible. For example, Mastalerz and co-workers described the postsynthetic modification of the interior of organic cages by using a sixfold Williamson etherification (28), but this interior modification affected the cage shape, which in turn altered the crystal packing mode.

Systematically tuning the pore size of organic cages by postsynthetic modification

We synthesized a series of internally postfunctionalized porous organic cages that crystallized in an isostructural way. Specifically, we used a protection-deprotection strategy to produce cages where five out of the six internal reaction sites in the cage cavity were functionalized by means of a formaldehyde “tying” method we reported previously (29) (Fig. 1, A and B). After deprotection, the single unreacted diamine group can be reacted with a range of aldehyde or ketone precursors to achieve ultrafine control over the cage cavity size and, hence, the pore envelope (Fig. 1C). This method allowed us to tune the size and shape of the cage pore systematically at the atomic level without affecting the external cage shape and size or its crystal packing. Using this strategy, we tuned the pore size for the series of POCs in the range 1.95 to 3.5 Å, which is a useful cut-off range for the separation of gas pairs such as H2/N2, H2/CO, CH4/N2, and Xe/Kr (7).

Fig. 1 Fine-tuning pore size by molecular organic synthesis.

(A) Scheme showing the protect-functionalize-deprotect strategy for internal cage cavity modification. (B) Full synthesis route for the modified cages, corresponding to the scheme in (A). FT, AT, ET, and PT represent products in which the diamine group(s) were tied with formaldehyde, acetone, ethanal (acetaldehyde), and propionaldehyde, respectively; x and y represent different possible tying groups (x = E, F; y = A, E, P). (C) Representative single-crystal structures of the modified cages showing accessible cavities (colored), along with calculated PLEs for each system. (D) Summary of gas adsorption properties for a range of gases (histogram; left axis) and cage cavity volumes (right axis) of a single modified cage, as calculated by the VOIDOO program on the basis of their crystal structures (probe radius = 2.0 Å) (43).

Previously, we reported that an organic cage with six diamine groups, RCC3, reacts with six formaldehyde molecules to form 6FT-RCC3 through the formation of cyclic aminal rings (29). RCC3 was reacted with six equivalents of acetaldehyde to form a new cage, 6ET-RCC3 (reaction a, Fig. 1B), where bulkier ethylidene bridges replaced the six methylene bridges in 6FT-RCC3. By contrast, RCC3 reacted with just a single acetone molecule to form 1AT-RCC3 (29) (reaction b, Fig. 1B); steric hindrance prevented further reaction in the cage cavity.

On the basis of this result, we developed a protection-deprotection strategy in which we reacted the five unfunctionalized diamine groups in 1AT-RCC3 with gaseous formaldehyde in the solid state, through a single-crystal–to–single-crystal reaction, to form a new dual “tied” cage molecule, 1AT-5FT-RCC3 (reaction d, Fig. 1B and fig. S1). Solid-state synthesis was essential, in that when 1AT-RCC3 was mixed with formaldehyde in solution (CHCl3 or MeOH), the single acetone protecting group was displaced to afford 6FT-RCC3 (reaction c in Fig. 1B and fig. S2). By exploiting the different chemical labilities of imidazolidine rings in 1AT-5FT-RCC3, we could selectively hydrolyze the single propyl “tie” by stirring in CHCl3:MeOH (1:1, v/v), thus synthesizing a new deprotected cage, 5FT-RCC3 (reaction e in Fig. 1B and fig. S3). After deprotection, we reacted the now vacant diamine group in 5FT-RCC3 with acetaldehyde and propionaldehyde to synthesize two new cages, 1ET-5FT-RCC3 and 1PT-5FT-RCC3 (reaction f, Fig. 1B). All reactions proceeded with nearly 100% conversion with no additional purification steps.

The parent cage for this study, RCC3, is derived from the shape-persistent imine precursor, CC3, which crystallized from most organic solvents to form a microporous solid with a diamondoid pore network (24, 26, 27). For guest molecules to diffuse through this pore network, they must pass through the intrinsic cage cavities that act as the tetrahedral nodes (25). Hence, internal functionalization of the cage cavities provided a route to systematically fine-tune pore size without altering the underlying shape and crystallization habit of the cage molecule (30). All cages were isostructural with CC3, as determined by powder x-ray diffraction (fig. S4) and single-crystal x-ray diffraction (fig. S5 and table S1). The high crystallographic symmetry of the structures (cubic F4132) did not allow us to determine whether the different tied groups were ordered in the microporous pore structures, but all cages had the same diamondoid pore structure.

To allow for the effect of molecular flexibility on the diffusivity of small gas molecules through these structures, we used molecular dynamics (MD) simulations to calculate a time-averaged pore-limiting envelope (PLE) (27) rather than measuring a single, static pore diameter. These calculations showed that postsynthetic modification allowed the PLE to range from 1.95 Å (6ET-RCC3) to 3.50 Å (6FT-RCC3), resulting in a tunability window of 1.55 Å (Fig. 1C). For context, this tunability window equates to the van der Waals (vdW) radius of a single nitrogen atom across the whole series of isostructural cages.

We studied the gas adsorption properties of these cages for four different gases (Fig. 1D). As expected, there is a positive correlation between the cage cavity volume and the overall gas uptake, especially under conditions in which gases are close to saturation (e.g., N2 at 77 K; orange bars in Fig. 1D). For example, 5FT-RCC3, with the largest cage cavity in this series, showed the highest uptakes for all four of the gases studied. By contrast, 1AT-5FT-RCC3 showed much lower gas uptakes, and 6ET-RCC3 size-excluded krypton (Kr) and xenon (Xe) altogether.

This synthetic strategy was also used to tune gas selectivity for hard-to-separate gas pairs, such as Xe and Kr (31, 32). Xenon is heavier and more polarizable than Kr, and it tends to form stronger vdW interactions with most sorbents unless the pores are specifically attuned for Kr adsorption. Preferential Kr adsorption is very rare. For example, just one MOF, FMOFCu, adsorbs Kr selectively over Xe at temperatures below 0°C (32).

Most of these organic cages also exhibit higher uptakes for Xe than for Kr, but 1AT-5FT-RCC3 exhibited reverse uptake for these two gases. As shown in fig. S6, the estimated ideal Kr/Xe molar selectivity could be switched below 0°C. We ascribe this effect to 1AT-5FT-RCC3 having a cavity ideally suited to accommodate Kr but too small for Xe. Inserting one additional methyl group to form 6ET-RCC3 rendered the cavity too small to allow adsorption of either gas.

Hydrogen isotope separation using an ultrafine aperture cage

Efficient KQS requires porous solids in which the difference between aperture size and gas diameter is comparable to the de Broglie wavelength. The kinetic diameter of hydrogen is 2.89 Å, and previous studies suggest that the optimum aperture size to achieve KQS with rigid frameworks lies below 3.40 Å (20). It is difficult to obtain porous materials with pore apertures in this range, and so far, only one material with pore size <3.00 Å has been reported to separate D2 and H2 (6).

MD simulations suggested that 6ET-RCC3 has a PLE centered at 1.95 Å but with a relatively broad time-averaged size distribution, with a value near the minimum molecular dimension of H2 (2.2 Å; Fig. 1C) (27). Hydrogen adsorption isotherms obtained from 6ET-RCC3 exhibited hysteresis between 30 and 60 K (Fig. 2B). The H2 uptake reached a maximum at 50 K (4.8 mmol/g at 1 bar). We observed the same phenomena for D2 adsorption (figs. S7 and S8). The hysteresis was largest at 30 K and decreased with increasing temperature, suggesting greater equilibration.

Fig. 2 Kinetic quantum sieving with ultrasmall-pore organic cages.

(A) PLE for 6ET-RCC3. The inset image shows the single-crystal structure of 6ET-RCC3 with its accessible cage cavity (colored). Ang, angstroms. (B) Hydrogen adsorption (solid symbols) and desorption (open symbols) isotherms of 6ET-RCC3 at different temperatures. (C) TDS spectra of 6ET-RCC3 obtained after exposure to a 10-mbar 1:1 D2/H2 isotope mixture at different exposure temperatures for a fixed exposure time (texp) of 10 min. The desorption spectra after evacuation at exposure temperature were measured for a heating rate of 0.1 K/s. (D) D2/H2 selectivities and gas uptakes as a function of exposure time at 30 K for CC3, 6FT-RCC3, and 6ET-RCC3; see supplementary materials for additional comparison (figs. S10 and S11).

We also compared the H2 adsorption behavior for two other organic cages, CC3 (24) and 6FT-RCC3, with calculated pore envelopes of around 4.5 and 3.4 Å, respectively. Both CC3 and 6FT-RCC3 exhibited reversible type I adsorption isotherms (fig. S9), typical of nanoporous materials without kinetic diffusion barriers. The two cages showed an increased H2 uptake with decreasing temperature, with maximum uptakes at 30 K of 8.0 and 8.2 mmol/g at 1 bar. Unlike 6ET-RCC3, no hysteresis was observed for CC3 or 6FT-RCC3. This difference suggests that the methyl groups in the 6ET-RCC3 cavities were responsible for its temperature-dependent flexibility, as this is the only structural difference between 6FT-RCC3 and 6ET-RCC3.

Cryogenic thermal desorption spectroscopy (TDS) measurements verified the hydrogen isotope separation performance of 6ET-RCC3. H2 and D2 desorption rates of 6ET-RCC3 (Fig. 2C) were collected during heating under vacuum with a rate of 0.1 K/s, after prior exposure to a 1:1 H2/D2 mixture (10 mbar) at various exposure temperatures (Texp) between 30 and 77 K for 10 min of exposure time (texp). The area under the desorption peak is proportional to the amount of the desorbing gas; hence, the selectivity for D2 over H2 could be obtained from the ratio of the peak areas (6).

TDS measurements showed that the uptake in 6ET-RCC3 increased with increasing temperature until a maximum for H2 and D2 was reached at Texp of 60 K before decreasing again at 77 K. The selectivity, SD2/H2, decreased with increasing Texp and exhibited a maximum of 3.9 at 30 K, which is a fairly good separation performance compared with previous reports of KQS (6, 33). The onset temperature of gas desorption is nearly identical to the exposure temperature, because after exposure the chamber is evacuated at Texp before cooling to 20 K. No desorption peak was observed above 60 K for Texp = 30 K, implying no deep penetration into the structure at this temperature. For higher Texp, the desorption spectra were unusually shifted to higher temperatures, which we attributed to an increased penetration depth of gas molecules into the cage structure. In agreement with observations from pure gas isotherms, the properties of 6ET-RCC3 can be attributed to the temperature-dependent opening of the pore aperture. This behavior is similar to that observed in MFU-4 reported previously (6), which exhibited different effective aperture sizes at different Texp. This flexibility enhanced the access of the isotope molecules to the pores at higher exposure temperatures.

Figure 2D shows SD2/H2 and the total quantity of gas adsorbed for 6ET-RCC3, 6FT-RCC3, and CC3 as a function of texp at Texp = 30 K (for detailed TDS spectra, see table S2). 6ET-RCC3 showed an increase in uptake for longer texp, revealing that diffusion was hindered. At 30 K, SD2/H2 decreased with longer texp, from 3.9 to 3.3 between 10 and 300 min of exposure, which is a typical KQS response. The KQS effect is based on the diffusion limitation of the lighter isotope; equilibrium is approached for both isotopes for longer exposure times. The total uptake for 6ET-RCC3 at 30 K was only 0.8 mmol/g at texp of 300 min, which is related to the hysteresis observed in the isotherm experiment at 30 K. The total gas uptake obtained at 30 K for CC3 and 6FT-RCC3 was much higher (8.0 mmol/g), consistent with the adsorption results (fig. S8). For the larger-pore materials, CC3 and 6FT-RCC3, SD2/H2 was lower and remained constant at 1.2 and 2.1, respectively, for different texp, in keeping with rapid, barrier-free diffusion of H2 and D2.

Enhanced quantum sieving performance in a two-component cage cocrystal

These single-component cage crystals exhibited either good selectivity but poor gas capacity (6ET-RCC3) or good gas capacity but poor selectivity (CC3 and 6FT-RCC3). For the optimal separation of isotopes, the ideal KQS material should combine large cavities to store more D2 along with a narrow pore aperture to increase the kinetic separation. Taking advantage of the chiral recognition assembly of discrete cage molecules (34), we designed a cocrystal structure that combined two different cages: 6ET-RCC3 was chosen as the diffusion barrier, blocking H2 diffusion and providing the KQS effect, and CC3 was chosen as the partner cage to provide enough pore space for high gas adsorption (Fig. 3A).

Fig. 3 Formation of a cocrystal enhances the D2/H2 separation performance.

(A) (Left) Scheme showing the cocrystal, Cocryst1, formed by chiral recognition between two cages to integrate capacity and selectivity in a single material. (Middle) Simplified representation of the crystal structure of Cocryst1, with pore channels shown in yellow. (Right) Single-crystal structure of Cocryst1. (B) H2 and D2 adsorption (closed symbols) and desorption (open symbols) isotherms of Cocryst1 at different temperatures. (C) D2/H2 isotherm ratio as a function of pressure for different temperatures. (D) TDS spectra for Cocryst1, obtained after exposure to a 10-mbar 1:1 H2/D2 isotope mixture for different exposure temperatures (Texp) and a fixed exposure time (texp) of 30 min. (E) D2/H2 selectivity as a function of texp at 30 K (red), 40 K (blue), and 50 K (green). (F) TDS spectra for Cocryst1, obtained after exposure to a 10-mbar 1:1 H2/D2 isotope mixture for different exposure times at 30 K. (G) The corresponding amount of adsorbed H2 (black) and D2 (red) and selectivity (blue) as functions of texp at 30 K.

The structure of the cocrystal, Cocryst1 (6ET-RCC3-R/CC3-S, 1:1), is shown in Fig. 3A (at right; see fig. S12 for structural details). Because the cage cavities formed the nodes of the diamondoid pore network, four selective 6ET-RCC3 cages surrounded each CC3 storage. Gas molecules diffusing through the cocrystal were forced to traverse the small pores in 6ET-RCC3.

Cocryst1 was studied using D2 and H2 high-resolution adsorption experiments collected between 20 and 77 K (Fig. 3B; additional isotherms are in fig. S13). The amount of gas absorbed at 1 bar showed a maximum at 40 K for both gases. As for 6ET-RCC3, the hysteresis in Cocryst1 was high at lower temperatures because of the large diffusion barrier, particularly at 30 K; at higher temperatures, a temperature-dependent opening of the aperture allowed both H2 and D2 molecules to diffuse faster.

The ratio of D2/H2 calculated for each point on the isotherms as a function of the pressure is presented in Fig. 3C. Cocryst1 exhibited a maximum D2/H2 uptake ratio of 3.5 at 30 K and 25 mbar, which is consistent with the lack of equilibrium at 30 K, implying that the cocrystal possessed good separation capabilities. To our knowledge, the D2/H2 uptake ratio of Cocryst1 is one of the largest values calculated from pure-gas adsorption isotherms (6, 20, 33, 35, 36).

We investigated actual gas separation by TDS. Cocryst1 was exposed to 10 mbar of a 1:1 H2/D2 mixture for 30 min at various exposure temperatures. The resulting TDS spectra for Texp = 30, 40, and 50 K are shown in Fig. 3D; additional TDS spectra for Cocryst1 are presented in fig. S14. The uptake after exposure to the gas mixture rose with increasing exposure temperature until a maximum for the combined gas uptake was reached at 40 K. In contrast to 6ET-RCC3, Cocryst1 shows desorption up to 80 K for Texp = 30 K, indicating gas penetration of the cages at a lower exposure temperature. The uptake then decreased until no desorption peak is observed at 77 K. SD2/H2 reached a maximum of 7.7 at 30 K and dropped to lower values at 40 and 50 K (Fig. 3E). For a constant exposure temperature of 30 K, thermally activated flexibility was again observed. As exposure time increased from 10 to 600 min, SD2/H2 remained near 8.0 (7.7 to 8.2), but the D2 uptake was enhanced from 0.4 to 4.7 mmol/g (Fig. 3, F and G), indicating hindered diffusion. X-ray diffraction data indicate that Cocryst1 remained highly crystalline and did not change its structure during the TDS measurements (fig. S18).

The kinetics of pure H2 and D2 gas uptake was studied by TDS after exposure at 30 K for exposure times varying from 10 to 1200 min (fig. S14). For both gases, Cocryst1 showed a pronounced increase in uptake for longer exposure time. Whereas the D2 uptake was almost identical for exposure times between 600 and 1200 min, the H2 amount still increased at longer exposure times. Because D2 reached saturation faster, it likely had a much different diffusion rate than H2.

The H2/D2 separation properties of Cocryst1 were excellent: SD2/H2 ~ 8.0, combined with a greatly enhanced D2 uptake with respect to 6ET-RCC3 (4.7 mmol/g). The experimental SD2/H2 values for various reported porous KQS adsorbents are compared in Fig. 4 and table S3. Of these materials, only two other porous solids have combined a selectivity greater than 3 with a gas uptake larger than 1.0 mmol/g (dashed lines in Fig. 4). The cage cocrystal displays the best combination of selectivity and D2 uptake reported to date.

Fig. 4 Summary of hydrogen isotope KQS selectivities and adsorption capacities for various porous materials (directly measured by TDS).

The list includes carbon [Takeda 3A (44)], MOFs [MFU-4 (6, 22, 44), MOF-5 (44), and IFP (45)], COFs [Py@COF-1 (46)], zeolite 5A (16), and porous organic cages (CC3, 6FT-CC3, 6ET-CC3, and Cocryst1). The practical performance of a KQS adsorbent will be a combination of its selectivity and its capacity; on this basis, the cage cocrystal shows the most promising performance.

Molecular simulations of the hydrogen isotope separation in organic cages

To further probe the separation mechanism, we simulated equilibrium adsorption isotherms for both single-component adsorption (Fig. 5, D to F) and an equimolar mixture of H2 and D2 (Fig. 5, G to I) in CC3, 6ET-RCC3, and Cocryst1 by combining grand canonical Monte Carlo (GCMC) simulations with the Feynman-Hibbs effective potential (37). We also used a hybrid GCMC-MD scheme, allowing for direct sampling of host motions in adsorption simulations, which was previously shown to be important for gas adsorption in porous organic cages (27, 38). The simulated adsorption isotherms (Fig. 5, D to I) are in good agreement with experiment, corroborating both the experimental adsorption data and the TDS results. All three cage crystals were predicted to be selective toward D2 over H2, with increased selectivity as the temperature decreases. Moreover, computation suggests, in agreement with experiment, that Cocryst1 exhibits the high D2 selectivity of 6ET-RCC3 and achieves the higher gas uptake capacities associated with the CC3 partner cage.

Fig. 5 Simulations provide insights into the mechanism for hydrogen isotope separation by the porous organic cages.

(A to C) A pair of neighboring cage molecules in 6ET-RCC3 (A), Cocryst1 (B), or CC3 (C). Gray, white, and blue atoms represent carbon, hydrogen, and nitrogen, respectively; pore spaces inside the cage molecules, defined by a spherical probe with diameter 2.2 Å, are colored in red. In (A), the colored pore space runs through a 6ET-RCC3 cage window with two methyl groups (left cage) and a second 6ET-RCC3 cage window with one methyl group (right cage). In (B), the pore runs through a 6ET-RCC3 cage window with one methyl group (left cage) and a neighboring CC3 cage window (right cage). (D to F) Predicted single-component adsorption isotherms of D2 and H2 in 6ET-RCC3 (D), Cocryst1 (E), and CC3 (F). Five independent simulations were performed to predict the adsorption uptake at each pressure at the respective temperature; the average of the five predicted uptakes was used to plot the isotherm, with error bars showing the maximum and minimum of the five predictions. (G to I) Predicted D2/H2 selectivity, calculated using competitive adsorption simulations of an equimolar D2/H2 mixture for 6ET-RCC3 (G), Cocryst1 (H), and CC3 (I). Similarly, five independent simulations were performed to predict the average uptake and to determine the error bars. (J to L) PIMD simulations predict D2/H2 selectivity and gas diffusion for a single, isolated 6ET-RCC3 cage (see supplementary text section 2.13 for simulation details). (J) Transition state for the translocation of a quantum H2 molecule. A single snapshot is shown, with an overlay of all 32 replicas in the PIMD simulation, illustrating the nuclear quantum fluctuations. Additional translucent red dots depict the fluctuations of the translocating H2 molecule, taken from 100 transition-state configurations. (K) Relative population of D2 over H2 inside the cage versus in the gas phase, as a function of temperature, compared with the observed experimental D2/H2 selectivities. (L) Free-energy profiles for a single molecule of quantum H2 (black) or quantum D2 (red) diffusing through a window of an isolated 6ET-RCC3 molecule at 50 K. Definitions of errors in (K) and (L) are given in section 2.13 of the supplementary text.

Diffusion of the hydrogen isotopes in CC3, 6ET-RCC3, and Cocryst1 was investigated and interpreted by comparison of the free-energy barriers for H2 and D2 in a given crystal structure at infinite dilution (see section 2.10.3, supplementary text). In the transition-state theory approximation, molecular diffusivity can be derived from a rate constant for the molecule hopping over the free-energy barrier. Figure S31 (supplementary text) shows the free-energy profile for a diffusion pathway between the center of mass of a cage molecule and the center of mass of a neighboring cage, travelling across two cage windows. In CC3 (fig. S31C), both H2 and D2 moved easily between the two cage molecules in a nearly barrier-free manner. By contrast, the congested cage cavities and narrowed window apertures in 6ET-RCC3 resulted in sharply peaked free-energy barriers (fig. S31A), strongly decreasing the molecular diffusivity compared with CC3.

When two methyl groups were located in a single 6ET-RCC3 cage window, the pore space became disconnected and the diffusion barrier became too large to observe any crossing (off-scale in fig. S31A; cage on the left). For windows with only one methyl group (fig. S31A, cage on the right), the pore space ran continuously between the cages, and the free-energy barriers for H2 and D2 were lowered substantially. Crucially, 6ET-RCC3 cage windows containing one methyl group provided diffusion barriers that were different for D2 and H2 and acted as a kinetic sieve. This feature was also carried over into Cocryst1 (fig. S31B). In Cocryst1, the large CC3 cavities also provided good dynamical relaxation, resulting in improved D2 kinetics and higher D2 uptakes.

We further combined path integral molecular dynamics (PIMD) simulations with quantum mechanical interaction evaluations to probe the adsorption and diffusion behaviors of H2 and D2 in the 6ET-RCC3 cage molecule, explicitly accounting for the quantum mechanical nature of both the electrons and the nuclei (Fig. 5, J to L). PIMD simulations predicted the binding free energy of D2 to be ~0.50 kJ mol−1, lower than for H2 across the temperature range 30 to 77 K. This difference in binding free energy translates into a relative population of D2 over H2 inside the cage of 4.21 at 50 K (Fig. 5H), rising to 24.65 at 30 K (see section 2.13, supplementary text). There is qualitative agreement between these simulated relative populations and the measured D2/H2 selectivity in the temperature range 30 to 77 K (Fig. 5K). We also estimated the free-energy barriers for a single molecule of quantum H2 and quantum D2 diffusing out of the 6ET-RCC3 cage molecule (Fig. 5L). These simulations show that nuclear quantum effects both destabilize H2 relative to D2 inside the cage and also provide a higher free-energy barrier for diffusion of H2 through the cage windows. The different barrier heights can be understood as the different degrees of zero-point fluctuations that increase the effective size of H2 more than D2, thereby making it more difficult for H2 to pass through the window aperture.

Analysis of the various pore structures and the simulated adsorption and diffusion results (Fig. 5 and fig. S30) may explain why the selectivity for Cocryst1 was double that of 6ET-RCC3 (SD2/H2 = 8.2 versus 3.9). The small channels in 6ET-RCC3 led to single-file diffusion. Once a molecule penetrated the pore, no passing between the isotopes in the channels is possible. Hence, the desorption after exposure to a 1:1 mixture exhibited identical maximum temperatures for both isotopes (Fig. 2C)—the gas molecules left in the same order in which they entered—whereas the maximum for pure gas desorption appeared lower for D2 than for H2 (Fig. 2B and fig. S7).

By contrast, Cocryst1 consisted of a combination of large storage pores (CC3, also selective toward D2 over H2 at 30 K) and small separation pores, separated by a differential diffusional barrier (fig. S30B and Fig. 5L). Penetration through the small apertures into the adjacent larger cavity yielded an additional sieving effect, whereby D2 molecules could pass neighboring H2 molecules, unlike in the single-file pores of 6ET-RCC3. As a result, Cocryst1 had an enhanced D2/H2 selectivity, although there were half as many small windows as in 6ET-RCC3.


When researchers began to study POCs, there was a strong emphasis on producing materials with large pore volumes and high surface areas to rival solids such as MOFs (24, 3941). Our study suggests that there are benefits in designing solids that are “barely porous” (40). By itself, this approach led to materials such as 6ET-RCC3 with low gas capacities, but by pairing small- and large-pore cages in a single cocrystal, we accessed a material with optimal gas separation properties. This strategy might be extended to produce even better performance—for example, by introducing storage cages with larger capacities than that of CC3, provided that the differential diffusional barrier between cages is preserved.

Our approach allows extremely delicate tuning of pore size—the entire tunability window for this series of cages spans the diameter a single nitrogen atom—and ideally suits applications such as KQS. Although the synthetic method involves multistep organic synthesis (Fig. 1B), including protection and deprotection steps, each step proceeded in nearly 100% yield. Because there was no need to purify intermediates, there is good potential to scale up the amount of materials made for practical separations.

Computational studies (Fig. 5) helped to explain the enhanced H2/D2 separation performance of the cocrystal. Although these calculations served to rationalize the experimental observations, the good agreement between theory and experiment suggests that the a priori design of new systems might be possible, perhaps through methods such as crystal structure prediction to identify suitable hypothetical systems (34, 42).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S44

Tables S1 to S8

References (4795)

References and Notes

Acknowledgments: We thank Diamond Light Source for access to beamlines I19 (CY21726) and I11 (EE17193). We thank the Advanced Light Source, supported by the Director, Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under contract no. DE-AC02-05CH11231 and also thank S. J. Teat for his assistance during this experiment. We thank M. Wang (University of Liverpool) for computing facilities support and A. Hill (Department of Chemical Engineering, University of Bath) for discussions on the practicability of using cage materials for hydrogen isotope separation. Funding: We gratefully acknowledge the Engineering and Physical Sciences Research Council (EP/N004884/1), the European Research Council under the European Union’s Seventh Framework Programme (FP/2007-2013)/ERC through grant agreement no. 321156 (ERC-AG-PE5-ROBOT), and the Leverhulme Trust via the Leverhulme Research Centre for Functional Materials Design for funding. S.Y. and L.D. acknowledge financial support from the Chinese Young Scholar National Science Foundation Grant (21403171), the Xi’an JiaoTong-Liverpool University (XJTLU) Research Development Fund (PGRS-13-03-08), and the Key Program Special Fund in XJTLU (KSF-E-03). D.H. thanks the Oversea Study Program of Guangzhou Elite Project from Guangzhou City, China, for financial support. V.K. and M.C. acknowledge funding from the Swiss National Science Foundation (project ID 200021-159896) and the European Research Council under the European Union’s Horizon 2020 research and innovation program (grant agreement no. 677013-HBMAP). Author contributions: M.L., L.C., and A.I.C. conceived the project. M.L. and D.H. synthesized and characterized the materials. L.C. conceived the modeling strategy and supervised the computational work, with input from L.D. and M.C. L.C., S.Y., and D.L.H. performed the classical simulations. V.K., L.C., and M.C. performed the electronic-structure, PIMD, and free-energy simulations. M.A.L. supervised the structural part of this study and refined all diffraction data. M.L. and M.A.L. cultured single crystals and performed solid-state crystal synthesis. M.A.L. and S.Y.C. carried out powder x-ray diffraction analyses. R.C. collected gas sorption data. M.H. and G.S. conceived the hydrogen isotope separation measurements. L.Z. and R.B.-X. collected the H2 and D2 isotherms. L.Z. performed TDS experiments. L.Z. and M.H. interpreted the TDS data. All authors contributed to the writing of the paper. Competing interests: The authors declare no competing interests. Data and materials availability: The crystallographic data reported in this paper are listed in the supplementary materials and archived at the Cambridge Crystallographic Data Centre under reference numbers CCDC 1910113 to 1910120. All other data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials.

Stay Connected to Science

Navigate This Article