Dynamic molecular pockets on diffusion channel for efficient production of polymer-grade propylene

Size-based molecular sieving works well for rigid molecules with complete exclusion of larger ones, yet interaction-induced molecular sieving may offer unusual separation capability for molecules with matching physicochemical properties. Here we report a MOF material (XXU-3) featuring one-dimension channels with embedded molecular pockets opening to C 3 H 6 and C 3 H 8 at substantially different pressures. The dynamic nature of the pockets is revealed by single-crystal-to-single-crystal transformation upon exposure of the activated XXU-3 to C 3 H 6 or C 3 H 8 atmosphere. Breakthrough experiments demonstrate that XXU-3 is not only capable of separating C 3 H 6 from C 3 H 8 with record-high C 3 H 6 productivity, but also the first MOF material realizing polymer-grade C 3 H 6 production in a single adsorption-desorption cycle from an equimolar C 3 H 6 /C 3 H 8 mixture. The underlying separation mechanism, namely orthogonal-array dynamic molecular sieving, is an exemplary strategy for both large separation capacity and fast adsorption-desorption kinetics. This work presents an ideal design for next-generation sieving materials and it holds great potential for applications in adsorptive separation.

be minimized and selectivity maximized in such MOF materials, on condition that the aperture dimension is right in between the sizes of gas molecules to be separated 27,28 . For example, by employing a shorter organic linker of 4,4´-azopyridine (9.0 Å) instead of 4,4´-dipyridylacetylene (9.6 Å) in SIFSIX-2-Cu-i, an ideal C2H2 molecular sieve SIFSIX-14-Cu-i (also termed as   29 was constructed with a reduced pore size of 3.4 Å, falling right in between the kinetic diameters of C2H2 (3.3 Å) and C2H4 (4.2 Å). As a result, it displayed nearly perfect C2H2/C2H4 selectivity by only taking up C2H2. Yet the tandem array of rigid apertures along one-dimension (1D) channel is a double-edged sword, on one hand, the high selectivity could be achieved by excluding the large-sized molecules; on the other hand, the equilibrium would be difficult to reach as the matching-sized molecules have to bump through many apertures, leading to slow adsorptiondesorption kinetics and thus less energy efficiency. To alleviate the suppressed kinetics, a better design is to furnish the apertures on the channel surface instead of along the diffusion channel. Such a layout, namely orthogonal-array molecular sieving, is exemplified in the MOF material UTSA-100, which was demonstrated superior performance in removing C2H2 from a C2H2/C2H4 mixture with high efficiency 30 . Size-based molecular sieving works well for rigid molecules. Yet it can be anticipated that this approach will be less powerful for molecules with more flexibility. In such a case, interactioninduced molecular sieving (dynamic molecular sieving) would be advantageous if the molecules to be separated could induce gate opening at different pressures. In other words, the apertures preferentially interact with and take up guest molecules of particular physicochemical properties, not necessarily the small-sized ones. A good MOF example that fits right in this category is MAF-41 developed by Zhang and co-workers 31 . This flexible MOF material exhibits adsorption of styrene but no adsorption of toluene or benzene. The latter two, although smaller in size, do not have sufficient adsorption energy to open the cavity. Clearly, framework flexibility has unique advantages in adsorptive separation if properly controlled. So far the rational design and control of local flexibility in the MOF framework for dynamic molecular sieving remain difficult and challenging. Some representative works are mesh-adjustable molecular sieve by Zhou 32 , chemical-triggered pore adaptation by Rosseinsky 33 , and temperature-responsive pore aperture by Kitagawa 34 , etc. A rare example in C3H6/C3H8 separation is the fluorinated MOF NbOFFIVE-1-Ni (also termed as KAUST-7) 16 , in which the dynamic nature of the pore aperture is presumably associated with the rotation and tilting of the pyrazine rings. The internal cavities of KAUST-7 along the 1D channel are interconnected by dynamic apertures that serve as sieving sites for C3H6. Nevertheless, as a tandemarray molecular sieve, the low kinetics and associated energy penalty seem inevitable given that gas molecules have to push through many independent gates in order to get through. Thus, orthogonalarray dynamic molecular sieving (Scheme 1) would be an ideal design to achieve not only large selectivity but also fast kinetics. Herein we report such a MOF (XXU-3, XXU = XX University) featuring dynamic molecular pockets opening to 1D diffusion channel. The pockets function as interaction-induced sieving sites for desired molecules to dock during the adsorption stage, while the 1D channels facilitate fast adsorption-desorption diffusion. Single-and mixed-component gas adsorption data suggest the pockets open to C3H6 and C3H8 at substantially different partial pressures. Breakthrough experiments reveal that XXU-3 is capable of efficiently separating C3H6 from C3H8 from an equimolar C3H6/C3H8 mixture with record-high C3H6 productivity, and it is the first MOF material realizing polymer-grade C3H6 production in a single adsorption-desorption cycle, which was previously achieved by Y-abtc, but from a C3H6/C3H8 mixture of 5/95 17 .
The phase purity of bulk XXU-3 was analyzed with powder X-ray diffraction (PXRD), and its diffraction pattern is in good agreement with the simulated one from single-crystal data (Supplementary Fig. 5). XXU-3 is thermally stable up to ∼ 400 ºC, as evidenced by thermogravimetric analysis (TGA) and variable-temperature PXRD analysis (Supplementary Fig.  6). To explore the permanent porosity, the as-synthesized XXU-3 was solvent-exchanged with methanol, followed by activation at 150 ºC under a high vacuum for 24 h to afford the desolvated XXU-3 (XXU-3a). PXRD was first employed to attest to the retention of the framework of XXU-3a (Supplementary Fig. 7). Then Ar adsorption/desorption isotherms were measured at 87 K (Fig.  2a), showing a type-I sorption behavior with saturated adsorption of 198 cm 3 (STP) g -1 at 1 bar. Based on its Ar adsorption isotherm, XXU-3a was calculated to have a Brunauer-Emmett-Teller (BET) surface area of 588 m 2 g -1 , an extremely narrow pore size distribution centered at 5.6 Å, and a total pore volume of 0.27 cm 3 g -1 . All these data are consistent with the theoretical values calculated from the crystal structure, further suggesting the retention of the framework after activation.
XXU-3a retains high-quality crystallinity and the SCXRD analysis reveals a 1D channel of ca. 4.5 Å × 5.3 Å dimension along the crystallographic a-axis (Supplementary Fig. 4) with a crosssectional area of ca. 23.85 Å 2 , which is substantially larger than the minimum cross-sections of C3H6 (19.34 Å 2 ) and C3H8 (18.17 Å 2 ) (Supplementary Fig. 8). In addition, molecular-sized pockets are lined up on both sides of the 1D channel and open to the channel through a "gourd-shaped" aperture of ca. 3.7 Å in size along the a-axis (Supplementary Fig. 9). From a kinetics point of view, the "gourd-shaped" aperture seems impossible for C3H6 (kinetic diameter ca. 4.0 Å) and C3H8 (kinetic diameter ca. 4.3 Å) 35 to go through. Nevertheless, the aromatic rings on MPTBDC linker were observed flexibility in tilting and rotation through in-situ variable-temperature SCXRD studies on XXU-3a (Supplementary Fig. 10), indicating the potential of opening up the "gourd-shaped" aperture to facilitate the admission of right-sized guest molecules such as C3H6 and C3H8 into the pockets.

Gas adsorption and kinetic studies
Pure-component C3H6 and C3H8 adsorption/desorption isotherms were collected for XXU-3a at different temperatures. As shown in Figs. 2b and c, both C3H6 and C3H8 display stepwise adsorptions. As the temperature increases, the sharp adsorption steps gradually fade away, and the initiation of those steps shifts toward higher pressures. Such a stepwise adsorption isotherm may be attributed to the guest-induced gate-opening behavior, as frequently reported in flexible MOFs [36][37][38][39] . Interestingly, the gate-opening pressure for C3H8 is approximately 5 times higher than that for C3H6 at 303 K, suggesting a high selectivity for C3H6 over C3H8 in the pocket. While in the low-pressure region before gate opening (e.g., 0.1 bar), a small amount of C3H8 (0.24 mmol g -1 ) can be adsorbed, and the value is 0.46 mmol g -1 for C3H6 at 303 K, suggesting a moderate selectivity for C3H6 over C3H8 in the channel (Supplementary Figs. 11). The C3H6 uptake in XXU-3a can reach 58.6 cm 3 g -1 (2.6 mmol g -1 ) at 1 bar and 303 K, higher than those benchmark C3H6/C3H8 molecular sieves, such as zeolite 4A (1.5 ~ 2 mmol g -1 ) 40,41 , KAUST-7 16 (1.4 mmol g -1 ), Co-gallate (1.8 mmol g -1 ) 42 and Yabtc 17 (2.0 mmol g -1 ) at 1 bar and 298 K. Remarkably, based on the adsorbed C3H6 and the calculated pore volume, the density of C3H6 inside XXU-3a is estimated to be 404 g L -1 at 303 K and 1 bar, which is more than 230 times higher than the density of gaseous C3H6 (1.707 g L -1 ) under similar conditions, and close to the density of liquid C3H6 (608 g L -1 ) at 225 K and 1 bar, demonstrating a highly efficient packing of C3H6 molecules inside XXU-3a. Further, the adsorption/desorption isotherms of both C3H6 and C3H8 are reversible and exhibit no obvious hysteresis (Fig. 2d); the continuous adsorption/desorption measurement on XXU-3a at 303 K shows no loss in C3H6 adsorption capacity over 10 cycles (Supplementary Figs. 12 and 13), suggesting no elevated temperature required for regeneration.
To quantify the binding affinity toward C3H6 and C3H8, coverage-dependent adsorption enthalpy (Qst) was calculated (Supplementary Figs. 14 and 15) from their respective adsorption isotherms at 273, 283, and 298 K by the virial fitting method. The adsorption enthalpy of XXU-3a for C3H6 and C3H8 were calculated to be 38.9 kJ/mol and 34.6 kJ/mol, respectively, at zero coverage. However, the data seem inconsistent at higher pressures (Supplementary Fig. 16). We suspect the virial model may not accurately reflect the adsorption behavior for MOF materials with stepwise adsorption isotherms. Thus, we performed differential scanning calorimetry of C3H6 and C3H8 adsorption and desorption on XXU-3a at 303 K and 1 bar to obtain the experimental adsorption/desorption enthalpy. As shown in Figs. 2e-f and Supplementary Fig. 18, the adsorption enthalpy of XXU-3a for C3H8 and C3H6 were found to be 16.1 kJ mol -1 and 29.3 kJ mol -1 , respectively. The subsequent desorption enthalpy upon helium flow was found to be 11.3 kJ mol -1 and 22.6 kJ mol -1 , respectively. Note that the adsorption enthalpy value for C3H6 is markedly lower than many of the leading C3H6-selective sorbents (Supplementary Fig. 17 and Table 5), such as Fe2(m-dobdc) (65 kJ mol -1 ), KAUST-7 (57.4 kJ mol -1 ), Co-gallate (41.0 kJ mol -1 ), and Y-abtc (50 kJ mol -1 ). Such a low adsorption enthalpy of XXU-3a for C3H6 is consistent with the adsorption/desorption cycling experiment, further corroborating the potential of easy regeneration for XXU-3a in C3H6/C3H6 separation. It is worth mentioning that elevated temperatures could cause C3H6 oligomerization/polymerization in a confined space 43 .
Arguably, C3H6-induced framework transformation in flexible MOFs could enable subsequent simultaneous uptake of C3H8 molecules, compromising the C3H6/C3H8 separation efficiency. To investigate the potential co-adsorption of C3H8 after C3H6-induced gate-opening in XXU-3a, a mixed-component (C3H6/C3H8, 50/50, mol/mol) adsorption experiment was performed under similar conditions to those used in single-component adsorption experiments. The mixed-gas adsorption isotherm (partial pressure of C3H6 up to 0.5 bar) almost overlaid with the pure C3H6 adsorption isotherm (Fig. 2d), suggesting preferential adsorption of C3H6 over C3H8 for an equimolar mixture. Differential scanning calorimetry of 50/50 mixed-component (C3H6/He, C3H8/He, and C3H6/C3H8) was performed at 303 K (Figs. 2e-f and Supplementary Fig. 19). The small heat flux peaks in the case of 50/50 C3H8/He indicate a rather weak interaction of XXU-3a with C3H8 at 0.5 bar, corroborating the preferential adsorption of C3H6 over C3H8 for an equimolar mixture on XXU-3a. The ideal adsorbed solution theory (IAST) 44 was applied to calculate the adsorption selectivity for an equimolar mixture of C3H6/C3H8, and an exceptionally high C3H6 over C3H8 selectivity (513) Table 5). Compared to MOFs with global flexibility, MOFs with local flexibility appear to be more promising in minimizing co-adsorption and realizing high selectivity due to the independence of gate opening. Nevertheless, it should be pointed out that IAST selectivity is calculated based on fitting single-component adsorption isotherm to a selected model, however, models for stepwise adsorption in flexible MOFs are still lacking; therefore, the resulting IAST selectivity may not accurately reflect its separation potential. The overall kinetic behavior for a particular guest is governed not only by its binding affinity (adsorption enthalpy) but also its diffusion rate within the host 45 . From a structure point of view, molecular sieving with orthogonally arrayed apertures, as in XXU-3a, is intrinsically a better design to realize fast kinetics, given each gas molecule only has to push through one aperture in an adsorption/desorption cycle. To verify its kinetic advantage, time-dependent adsorption/desorption isotherms of C3H6 on XXU-3a were recorded at 303 K, along with two benchmark molecular sieving materials with tandemly arrayed apertures, KAUST-7 and Y-abtc (Supplementary Fig. 26). Remarkable differences in their diffusion rates are observed, and C3H6 diffuses much faster in XXU-3a than the other two materials both in adsorption and desorption stages. To further quantify the difference, the diffusion rate constant (D/r 2 ) was calculated to be 2.94E-003 for XXU-3a, which is more than one order of magnitude higher than KAUST-7 (8.66E-004,) and Y-abtc (7.64E-005) (Supplementary Fig. 23).

Dynamic sieving mechanism and breakthrough experiments
To understand the gate-opening behavior of XXU-3a, single-crystal-to-single-crystal transformation was studied. First, an XXU-3a crystal of good quality was picked and its single crystal data was collected at 150 K. Then, the same crystal was exposed to the C3H6 atmosphere at room temperature for 3 h and its single crystal data was collected again at 150 K. From the analysis of the crystal structures of XXU-3a and C3H6@XXU-3a, the pockets are confirmed to be the preferential binding sites for C3H6 molecules (Figs. 3a and b), and four C3H6 molecules are found per unit cell, which is consistent with the measured C3H6 adsorption capacity at 1 bar and 298 K. Compared with XXU-3a, the unit cell of C3H6@XXU-3a exhibits some deformation. The organic linker in C3H6@XXU-3a adopts a slightly different conformation through aromatic rings tilting and rotation in response to the adsorbed C3H6 molecules in the pockets (Supplementary Figs. 27 and  28). As a result, the unit cell volume is expanded by 6 % to 1963.72 Å 3 . Specifically, multiple interactions are found between the C3H6 molecule and surrounding organic linkers on the pocket, including two weak interactions with two carboxylate oxygen atoms (2.45 Å and 3.38 Å), one weak interaction with a triazole nitrogen atom (2.91 Å), and one CH···π interaction with pyridine ring (2.87 Å) (Fig. 3c). For comparison, we managed to obtain the crystal structure of C3H8@XXU-3a at 100 K and found the pockets are also the preferential binding sites for C3H8 molecules (Figs. 3d  and 3e). Not surprisingly, the C3H8 molecules are interacting with surrounding organic linkers on the pocket, including two weak interactions with two carboxylate oxygen atoms (2.92 Å and 3.16 Å), one weak interaction with a triazole nitrogen atom (2.68 Å), and one CH···π interaction with the pyridine ring (2.88 Å) (Fig. 3f). Connolly surface comparison of the "gourd-shaped" apertures in XXU-3a (3.7 Å), C3H8@XXU-3a (4.4 Å), and C3H6@XXU-3a (4.4 Å) indicate the opening up of the aperture along with the adsorption of C3H6 or C3H8 (Fig. 3g).
Many efforts were devoted to gas-loaded crystal structures, yet only at low temperatures did we successfully collect C3H6@XXU-3a and C3H8@XXU-3a crystal data. To further quantify the binding affinity of C3H6 and C3H8 in XXU-3a at room temperature, in the work reported here we performed density functional theory (DFT) calculations by employing Perdew-Burke-Ernzerhof (PBE) functional 46 . In the optimized host−guest structures at room temperature, the binding configurations for both C3H6 and C3H8 (Supplementary Fig. 29) are in good agreement with C3H6@XXU-3a and C3H8@XXU-3a structures determined by SCXRD at low temperatures. The static binding energy ΔE (ΔE = EMOF+gas -EMOF -Egas) at room temperature was thus calculated to be −19.96 and −16.52 kJ mol -1 for C3H6 and C3H8, respectively. The appreciable difference in binding energy (3.44 kJ mol -1 ) could be attributed to the stronger C−H···O and C-H···N interactions with C3H6 than C3H8 (Supplementary Fig. 29). Subsequently, C3H6 and C3H8 kinetic adsorption studies were carried out on XXU-3a and found that it is much faster for C3H6 to reach adsorption saturation plateau than C3H8 (Supplementary Fig. 24). Furthermore, the diffusion rate constant was calculated to be 2.94E-003, which is more than one order of magnitude higher than that for C3H8 (4.53E-004) (Supplementary Fig. 24), Thus, we conclude that the adsorption of C3H6 is both favored by thermodynamics and kinetics, leading to high C3H6/C3H8 selectivity. To further look into plausible adsorption mechanism for XXU-3a, specifically whether all the pockets open in a cooperative manner or each pocket has to further rearrange in order to accommodate C3H6, we performed DFT simulation on XXU-3a by loading C3H6 one by one. As illustrated in Fig. 4a, the first C3H6 was put in the bottom left pocket of XXU-3a and the structure (XXU-3a@1C3H6) was optimized by DFT calculations with all-atom positions and unit cell parameters allowed to vary. Starting from the optimized XXU-3a@1C3H6, the second C3H6 was put in the top left pocket and the structure (XXU-3a@2C3H6) was optimized by DFT calculations under two different conditions: (i) full relax the geometry and cell parameters (all-atom positions and unit cell parameters allowed to vary); and (ii) fix the geometry and cell parameters (all-atom positions and unit cell parameters not allowed to vary). Upon the completion of optimization, the rotation of the pyridine ring was observed in (i) with a dihedral angle of the pyridine plane and triazole plane rotating from 34 o to 28 o (Fig. 4b). The energy difference (∆E = EXXU-3a@2C3H6 -EXXU-3a@1C3H6 -EC3H6) was calculated to be 55.1 kJ mol -1 and 140.1 kJ mol -1 for scenarios (i) and (ii), respectively. The large energy difference suggests further local rearrangement is energetically favored for the pocket to accommodate the second C3H6. Thus, a cooperative framework rearrangement may occur to some extent (unit cell volume expanded by 6 %), but itself alone would be energetically disfavored. The gate-opening adsorption behavior is likely due to a combination of global flexibility (cooperative framework rearrangement) and local flexibility (e.g., pyridine ring rotation); the latter seems to be the major contributor in the case of XXU-3a. The industrial production of C3H6 by steam cracking usually contains 40 ~ 50 % of C3H8 47 . Although more than a few MOF materials have been demonstrated their outstanding performance in C3H6 purification 16,17 , hardly any exhibits separation capability of producing polymer-grade C3H6 from an equimolar C3H6/C3H8 mixture in a single adsorption/desorption cycle. To evaluate the separation capacity of XXU-3a for C3H6/C3H8 mixtures, dynamic column breakthrough experiments were conducted, in which an equimolar C3H6/C3H8 mixture flowed over a packed column of XXU-3a sample (1.4 g) at a rate of 1.0 mL min -1 and 303 K (Fig. 5a and Supplementary  Fig. 30). Breakthrough curves reveal that XXU-3a is capable of completely separating an equimolar C3H6/C3H6 mixture into individual components. As shown in Fig. 5a, C3H8 was eluted from the column first, and then the outlet C3H8 gas quickly reached high-purity with no detectable C3H6 (below the detection limit of the instrument, 0.01 %) (Supplementary Fig. 31); whereas C3H6 was preferentially adsorbed by XXU-3a and breaking through the column after a substantial time lapse.
The excellent breakthrough performance of XXU-3a encouraged us to conduct desorption experiments for C3H6 purity and productivity. For comparison, breakthrough experiments of an equimolar C3H6/C3H8 mixture were also carried out over a packed column of KAUST-7 or Y-abtc under the same condition as for XXU-3a (Figs. 5a-c). Upon reaching breakthrough equilibrium, the captured gas in the column was purged by helium gas sweeping (10 mL min -1 ) at 303 K, and the outlet gas was monitored by gas chromatography. The C3H6 purity and productivity were estimated from their desorption curves (Figs. 5a-c and Supplementary Fig. 32) by using a calibration curve of C3H6 flowrate vs. its peak area on gas chromatogram (Supplementary Table 3 and Figs. 33-39). It turns out that 34.2 L of C3H6 with over 99.5 % purity could be obtained from an equimolar C3H6/C3H8 mixture for 1 kg of XXU-3a in a single adsorption-desorption cycle, the value was significantly higher than those for KAUST-7 (16.3 L with 90.0 % purity) and Y-abtc (1.3 L with 90.0 % purity), the two benchmark MOF materials with the best performance reported so far for C3H6/C3H8 separation. For 100 g of C3H6/C3H8 (50/50) mixed gas input, XXU-3a, KAUST-7, and Y-abtc could produce 25.6 g (99.5 %), 22.8 g (90 %), and 0.8 g (90 %) of C3H6, respectively. These results demonstrate that XXU-3a is capable of efficiently producing C3H6 with both record-high purity and productivity in a single adsorption-desorption cycle from an equimolar C3H6/C3H8 mixture.
The initial slope of the desorption curve observed on XXU-3a is much lower, and it takes almost 40 min to start producing high-purity propylene, indicating co-adsorption of C3H6/C3H8 on XXU-3a in the breakthrough experiment, which is consistent with the adsorption capacity estimated from the breakthrough curves (Supplementary Figs. 40 and 41). By integrating the corresponding areas on the breakthrough curves, the dynamic adsorption capacity of C3H6 and C3H8 upon breakthrough equilibrium were estimated to be 54.9 mL g -1 and 18.7 mL g -1 , respectively, (Supplementary Figs. 40-41 and Table 4). The co-adsorption of C3H6/C3H8 could be attributed to the less selective adsorption sites such as channels and the interstices between particles. These were probably purged first during the desorption process, corresponding to the 40-min elution of a C3H6/C3H8 mixture on the desorption curve, and then, C3H6 in the pockets were subsequently purged, translating into the high-purity C3H6. To further highlight the practical industrial applications of XXU-3a as a solid adsorbent, we performed dynamic column breakthrough curves of an equimolar C3H6/C3H8 mixture at higher flow rates (up to 8.0 mL min -1 ) on XXU-3a, KAUST-7, and Y-abtc. As shown in Supplementary Figs. 42-54, the breakthrough times (from t0 to t1) were substantially reduced at higher flow rates, yet the desorption curves demonstrate that XXU-3a retains its excellent separation capability at higher flow rates. Specifically, the C3H6 productivity and purity were estimated to be 34.2 L kg -1 (99.5 %), 53.5 L kg -1 (99.5 %), and 48.2 L kg -1 (99.5 %) at flow rates of 1.0 mL min -1 , 6.0 mL min -1 , and 8.0 mL min -1 , respectively (Fig. 6). Whereas the separation performance of KAUST-7 and Y-abtc did not significantly improve upon the increasing flow rates, their C3H6 productivity and purity were estimated to be 21.0 L kg -1 (90 %) and 3.76 L kg -1 (90 %), respectively, at 8.0 mL min -1 flow rate (Fig. 6). For practical industrial applications, the adsorbent should possess good recyclability. We carried out multiple dynamic breakthrough measurements for an equimolar C3H6/C3H8 mixture on the same XXU-3a sample. In between measurements, the sample was regenerated under vacuum at room temperature. The recycling experiments reveal almost the same retention time for both C3H6 and C3H8 over eight continuous adsorption-desorption cycles (Fig. 7b), indicating no loss in C3H6 adsorption capacity and C3H6/C3H8 separation capacity. Further, continuous breakthrough experiments for pure C3H6 were carried out with helium sweeping as a regeneration method. As shown in Fig. 7a and Supplementary Fig. 55, it was demonstrated that XXU-3a could be completely regenerated under helium sweeping, and no obvious loss of adsorption capacity after 50 cycles of C3H6 adsorption-desorption. Additionally, dynamic column breakthrough experiments of an equimolar C3H6/C3H8 mixture with a flow rate of 6.0 mL min -1 were carried out on XXU-3a under either dry (0% RH) or humid conditions (50% RH) (Fig. 7c). The C3H6 purity and productivity were estimated from their desorption curves (Supplementary Fig. 57). Under 50 % RH condition, 44.9 L of C3H6 with over 99.5 % purity can be obtained from an equimolar C3H6/C3H8 mixture for 1 kg of XXU-3a, which is lower than the productivity under 0% RH condition (53.3 L with over 99.5 % purity). The slight decrease in productivity is reasonable since the partial pressure of C3H6 in the feed gas would be lowered upon the inclusion of water vapor, not to mention water vapor competing for adsorption sites. All the experiments were run three times and the averages were reported with small error bars (Fig. 7d).

Outlook
The development of porous materials for efficient separation of C3H6/C3H8 mixtures is of great importance, particularly the ones that can achieve polymer-grade C3H6 production in a single adsorption-desorption operation. This work illustrated a MOF material (XXU-3) featuring 1D channel with embedded molecular pockets opening to C3H6 and C3H8 at substantially different pressures, which was translated into efficient production of high-purity C3H6 (≥ 99.95 %) from an equimolar C3H6/C3H8 mixture in a single adsorption-desorption. Furthermore, XXU-3 has demonstrated much higher productivity and at least one order of magnitude faster diffusion kinetics compared to the two best performing materials under similar conditions. The excellent separation performance of XXU-3a was attributed to the underlying separation mechanism, namely orthogonal-array dynamic molecular sieving. The structural design depicted here represents nextgeneration molecular sieving with the potential of achieving not only large separation capacity but also fast adsorption-desorption kinetics, both of which will have significant impacts on energy conversation in adsorptive separation. Besides, the dynamic nature may offer unusual separation capability which would otherwise difficult to realize. We anticipate the orthogonal-array dynamic molecular sieving strategy be extended to a broad scope of applications for efficient separation.

Materials
All reagents and solvents were obtained commercially and used as received without further purification.

Gas adsorption measurement
At least 100 mg of sample was used for each measurement. XXU-3 was activated at 150 °C under a dynamic vacuum (below 5 μmHg) for 24 h. Single-component and mixed-component gas adsorption/desorption isotherms were measured on an ASAP 2020 PLUS Analyzer (Micromeritics). The time-dependent kinetic adsorption isotherms were collected on Micromeritics ASAP 2020 Plus in Rate of Adsorption (ROA) mode (place roa.exe in the directory where the 2020 Plus.exe is located).