Molecular recognition plays a vital role in supramolecular chemistry1,2, in which specific affinity among molecules allows for the construction of high-order assemblies and stimuli response3. Usually, recognition of a sole guest with the largest affinity from the multiple-guest mixture can be achieved4, whereas specific recognition of different guests under varied environmental conditions remains challenging. A basic scientific issue is a limitation from thermodynamics, where the order of host-guest affinities hardly changes with environmental conditions in a given host-multiple guest system2. On the other hand, specific recognition of different guests is highly desired, and such “smart” host materials can be widely applied in various fields such as molecular machines5, sensors6, gas separation7, and drug delivery8. To achieve specific recognition switchable to different guests, chemists attempted to change the host-guest affinity using stimuli-responsive guests9,10, whose chemical structures or molecular conformations change with external stimuli. However, such a strategy is only limited to cyclodextrin-azobenzene11, cyclodextrin-benzimidazole12, and cyclodextrin-ferrocene systems13, whose host-guest affinities can be switched by light, pH, and redox, respectively. A simple and effective strategy for switchable molecular recognition that breaks through the limitation of thermodynamics has not been proposed so far.

Porous coordination polymers (PCPs)14,15,16,17 or metal-organic frameworks are highly designable material platforms whose structural softness and pore environment can be tailored for molecular recognition. Despite the recent progress, most of the molecular recognition in PCPs is based on thermodynamic adsorption, which is inaccessible for switchable recognition. On the other hand, the control over guest-transport kinetics allows precise discrimination of similar guests18,19, yet the strategy for switchable molecular recognition is still not proposed. Herein, we present switchable molecular recognition of two similar gaseous guests, CO2 and C2H2, only by temperature, without changing their host-guest affinity. This is achieved by regulating the gas diffusion and amplifying their rate differences using a locally dynamic PCP, in which flip-flop molecular motions of the ligand provide gate functionality. CO2 exhibits faster diffusion than C2H2, rendering the preferential adsorption of CO2 and CO2/C2H2 selectivity higher than 1 at the thermodynamic non-equilibrium state at low temperatures. By contrast, C2H2 possesses higher adsorption affinity than CO2, resulting in selective adsorption of C2H2 and C2H2/CO2 selectivity higher than 1 at the thermodynamic equilibrium state at high temperatures. Therefore, significant temperature-dependent adsorption behaviors are observed for CO2 and C2H2, with both striking CO2/C2H2 and C2H2/CO2 selectivities at low and high temperatures, respectively (Fig. 1a).

Fig. 1: The diffusion-regulatory mechanism for the temperature-switched recognition of CO2 and C2H2.
figure 1

a Schematic representation of the mechanism. Left: temperature-dependence of diffusion rates of CO2 and C2H2. Right: temperature-dependence of host-guest affinities of CO2 and C2H2. Middle: dynamics manipulation in diffusion-regulatory PCPs (this work); this mechanism involves a pore system featuring diffusion-regulatory functionality that can regulate the diffusion of gases and amplify their rate differences, thereby resulting in a temperature-switched recognition in which the gas with a high diffusion rate but a low affinity is preferentially adsorbed at low temperatures and the gas with a low diffusion rate but a high affinity is selectively adsorbed at higher temperatures. b The two-fold interpenetrated, 3, 6-connected rutile (rtl) topology of FDC–3a. The Zn2+ dual-tetrahedron cluster possesses 6 coordination sites, simplified as a 6-connected node and represented with green balls. The ligand is linked with 3 Zn2+ clusters, which is simplified as a 3-connected node and is represented with purple balls. c The void in FDC–3a visualized by a small probe radius of 0.6 Å. The void volume is 692 Å3 and corresponds to 13.3% of the unit-cell volume. The inner and outer surfaces of the pore are drawn in red and gray, respectively. The pink and blue arrows show the diffusion windows and pathways, respectively.


PCP synthesis and structural analyses

We designed a bee-type ligand comprising [1,1’:3’,1”-terphenyl]−3,3”-dicarboxylic acid and phenothiazine-5,5-dioxide (OPTz) moieties (OPTz-t3da), with the later moiety exhibiting effective waggling motion; the OPTZ moiety can waggle around its equilibrium position by ~20° with small energy increases by <25 kJ mol–1 (Supplementary Fig. 1). Such a waggling motion of ligand leads to the dynamical opening and blocking of channels in PCP crystal, which is thus termed flip-flop dynamic crystal (FDC). The as-synthesized PCP, termed FDC–3 (Supplementary Figs. 24, Supplementary Table 1), was subjected to solvent exchange and thermal activation to afford its activated phase (FDC–3a, Supplementary Figs. 510, Supplementary Table 2). The crystal structure of FDC–3a was determined by the continuous rotation electron diffraction (cRED) technique (Supplementary Fig. 5). Activation caused a structural transformation of FDC–3 into a two-fold interpenetrated, 3, 6-connected rutile (rtl) topological framework with small yet compact pores (Fig. 1, b, c, Supplementary Fig. 6). The pore aperture was surrounded by one OPTz moiety and one O atom on carboxylic acid to form an ultrasmall gate of 2.9 Å in size, which was expected to be gradually enlarged by the thermal flipping of OPTz moiety, allowing the diffusion of gases at high temperature and blocking them at low temperatures.

Gas sorption

FDC–3a adsorbed CO2 and C2H2 and showed negligible adsorption for other gases, including N2, CO, O2, Ar, C2H4, and C2H6, in a wide temperature range (Fig. 2a, Supplementary Fig. 11). The adsorption amounts for both CO2 and C2H2 substantially increased as increasing the temperature, as shown in their adsorption isotherm curves (Supplementary Fig. 12). Taking CO2 as an example (Supplementary Figs. 1214), the adsorption amount increased from 25 to 41 mL g–1 as the temperature was increased from 200 to 240 K and then decreased to 5 mL g–1 as the temperature was further increased to 370 K. Therefore, the temperature of maximum adsorption amounts (Tmax) of CO2 was 240 K, and similarly, the Tmax of C2H2 appeared at 320 K; the Tmax values for both CO2 and C2H2 were substantially higher than their boiling-point temperatures (Tbp). This is a distinctive adsorption feature in which the initial adsorption was promoted by temperature, making a sharp contrast to common gas adsorption under thermodynamic equilibrium, in which the adsorption amount monotonously decreases as increasing the temperature. Additionally, obvious desorption hysteresis was observed for CO2 (200–300 K) and C2H2 (200 to 360 K) in their sorption isotherms, which was characteristic of the diffusion-regulatory pore systems in PCPs18,19. These results further indicated that the diffusions of CO2 and C2H2 were regulated in the temperature ranges of 200 to 300 K and 200 to 360 K, respectively, showing that the adsorption of CO2 was controlled by kinetics and thermodynamics at low (200 to 300 K) and high (320 to 360 K) temperatures, respectively, whereas the adsorption of C2H2 was constantly controlled by kinetics. The temperature-assisted adsorption behavior was controlled by kinetics, in which the diffusion of gases was impeded by low temperature, whereas the diffusion was gradually promoted by raising the temperature. Remarkably, the Tmax values of CO2 and C2H2 were largely different by 80 K, although they had exactly the same kinetic diameters and very similar molecular sizes and polarizabilities (Supplementary Table 3). Therefore, the selectivity can be switched by temperature; FDC–3a preferably adsorbed CO2 in the 200 to 280 K range, whereas it reversely selected C2H2 in the 290 to 370 K range. The maximum adsorption ratios for CO2/C2H2 and C2H2/CO2 were 2.9 (at 220 K) and 3.6 (at 350 K), respectively (Fig. 2a).

Fig. 2: Gas adsorption behavior, IAST selectivities, and diffusion rates between 200 and 370 K.
figure 2

a CO2 and C2H2 adsorption isobars at 1 bar, and the CO2/C2H2 and C2H2/CO2 uptake ratios. b IAST selectivities of FDC–3a for CO2/C2H2 with different feed-gas components at various temperatures. c Global temperature–diffusion-rate–adsorption amount (TDs/R2V) landscape for CO2 adsorption, where R denotes the radius of an FDC–3a particle. d Global temperature–diffusion-rate–adsorption amount (TDs/R2V) landscape for C2H2 adsorption, where R denotes the radius of an FDC–3a particle.

Although the above-mentioned sorption curves already revealed an apparent difference in the adsorption amounts of CO2 and C2H2, they were not able to reflect the differences in the adsorption kinetics. Therefore, we performed kinetic adsorption of CO2 and C2H2 at different temperatures by FDC–3a (Supplementary Fig. 15). The adsorption amounts for both CO2 and C2H2 were lower than the amounts in their corresponding isobar curves, whereas the Tmax for both CO2 and C2H2 slightly shifted to higher temperatures. These results indicated that the kinetic factors were key to affecting the adsorption behaviors of CO2 and C2H2. On the other hand, the switching of selectivity was also observed in the kinetic adsorption, which further proved that the cooperativity of diffusion regulation and host-guest interaction caused the temperature-switchable selectivity even in the kinetic conditions.

IAST selectivities and diffusion rates

We employed the ideal adsorbed solution theory (IAST) to predict the selectivity of a CO2/C2H2 mixture for practical separation (Fig. 2b, Supplementary Fig. 16, Supplementary Tables 4, 5). The IAST selectivity in FDC–3a was temperature-manipulated: in the temperature range of 200 to 280 K, FDC–3a was CO2-selective with the CO2/C2H2 selectivity higher than 1; in the temperature range of 300 to 360 K, FDC–3a turned to C2H2-selective with the CO2/C2H2 selectivity lower than 1 (i.e., C2H2/CO2 > 1). Such a temperature-switched selectivity was not observed in other porous materials. Taking an equimolar mixture of CO2/C2H2 at 1 bar and various temperatures as an example, the maximum CO2/C2H2 selectivity was up to 18.6 at 220 K and the minimal CO2/C2H2 selectivity was as low as 0.11 (corresponding to a C2H2/CO2 selectivity of 9.5) at 360 K. Even though at a condition of very low feed-gas components (5% CO2 at 200 K and 5% C2H2 at 360 K), the CO2/C2H2 selectivity values remained to be 15.6 (200 K) and 0.2 (360 K), respectively, indicative of the capability of enriching CO2 at low temperatures and C2H2 at high temperatures. FDC–3a exhibited high IAST selectivity for CO2 and C2H2 at low and high temperatures, respectively, revealing the potential for the selective adsorption of CO2 and C2H2 in a temperature-controlled manner.

To uncover the essence of the temperature-switched adsorption, we employed the Crank theory to quantify the diffusion rate for every CO2 and C2H2 adsorption plot from the corresponding isotherms in the 200 to 360 K range, which allowed the production of global T–Ds/R2–V and P–Ds/R2–V landscapes, where T (K), P (kPa), Ds/R2 (s–1), and V (mL g–1) denote temperature, pressure, diffusion rate, and uptake volume, respectively, where R refers to the radius of an FDC–3a particle (Fig. 2, c, d, Supplementary Fig. 17). The landscapes revealed that the diffusion rates for both CO2 and C2H2 were substantially low at low temperatures, whereas they steadily increased with increasing temperature and pressure, accompanied by the enhanced uptake amounts up to the Tmax of CO2 and C2H2. The diffusion rates for CO2 and C2H2 at 240 K (Tmax of CO2) and 1 bar were 3.22 × 10–2 and 8.29 × 10–3 R2 s–1, respectively, whereas the diffusion rates for CO2 and C2H2 at 320 K (Tmax of C2H2) and 1 bar were 7.24 × 10–2 and 2.18 × 10–2 R2 s–1, respectively. The diffusion rates of CO2 were substantially higher than C2H2 at all temperatures, despite the same kinetic diameters and very similar molecular sizes. The kinetic diameter of CO2 and C2H2 (3.3 Å) was one of the smallest values in common gases, which caused a great obstacle in controlling the diffusion using porous materials for kinetic discrimination20. Nevertheless, FDC–3a could control the diffusion process of both CO2 and C2H2 and amplify their slight rate difference, thus causing substantial differences in Tmax and achieving temperature-switched selective adsorption of CO2 and C2H2.

PXRD and solid-state NMR analyses

To further understand the mechanism from a structural aspect, in-situ PXRD patterns were collected during the adsorption process. Patterns obtained during the adsorptions of CO2 (200 K) and C2H2 (300 K) did not reveal any structural changes (Supplementary Figs. 18, 19). We further performed synchrotron variable-temperature powder X-ray diffraction (VT-PXRD) measurements for FDC–3a from 100 to 380 K, which revealed that several peaks slightly shifted to lower angles upon increasing temperature (Supplementary Fig. 20). Taking the peak corresponding to the (111) plane as an example, it shifted from 2θ = 5.079° to 5.015° as the temperature increased from 100 to 380 K, which revealed the expansion of the [111] axis. Because one OPTz moiety was oriented parallel to the (111) plane (Supplementary Fig. 21), this slight expansion could be related to the extent of thermal flipping, which enlarged the gates to allow gas diffusion and controlled the diffusion rate.

Although the PXRD analysis could characterize the lattice change at different temperatures in FDC–3a, it could not show the local motions at the molecular level. To precisely reveal the flip-flop motion in FDC–3a, we performed O2-enhanced high-resolution solid-state 13C cross-polarization magic angle spinning (CPMAS) nuclear magnetic resonance (NMR) study21, which found around 25 well-distinguished sharp resonances including 11 quaternary and 15 tertiary aromatic carbon signals (Supplementary Figs. 22, 23), corresponding to the 32 total carbons (11 quaternary and 21 tertiary) of the ligand (Fig. 3a, Supplementary Table S6), in comparison with the solution 13C NMR attribution of the OPTz-t3da ligand based on calculated chemical shifts (Supplementary Fig. 24). This demonstrated the chemical purity and crystallinity of FDC–3a without observable defects. By contrast, the solution 13C NMR spectrum of OPTz-t3da ligand showed 17 distinct signals that are attributed to 17 chemically inequivalent carbons of the symmetric ligand (Supplementary Fig. 25). This contrast unambiguously revealed that the coordinated ligand in the framework is asymmetric, especially the two benzoate moieties (two distinct sets of 13C resonances for C1-C7), while the two rings of OPTz are nearly symmetric (single or two close resonances for each of C12-C17). Such asymmetry suggests rigid benzoate coordination with two distinctive environments and flexible OPTz with two similar rings, which was fully consistent with the asymmetric unit of the crystallographic structure containing monodentate and bidentate carboxylates. Moreover, the variable-temperature (VT) 13C CPMAS NMR study from 213 to 352 K (Fig. 3a, Supplementary Fig. 25) found that the resonances of C12, C13, C16, and C17 on the OPTz ring shifted by 0.3–0.8 ppm while the resonances of C14 and C15 remained almost unchanged, indicative of the partial configurational change and reoriented OPTz rings in the ligand, which was in good correlation with the finding of slight lattice expansion by VT-PXRD as well as with the thermal flipping of OPTz moiety.

Fig. 3: Variable-temperature solid-state NMR and theoretical study.
figure 3

a VT 13C CPMAS NMR spectra of FDC–3a. b Simulated structure of the CO2-adsorbed phase and schematic diagram of diffusion pathways of FDC–3a, where CO2 molecules outside the representative path are omitted for clarity. The CO2 molecules at the initial position are marked with red color, and the CO2 molecules at the other positions of the diffusion pathway are marked with orange, yellow, green, and cyan colors, respectively. Numbers 1 to 5 and the arrows represent the diffusion pathways. c Simulated structure of the C2H2-adsorbed phase and schematic diagram of diffusion pathways of FDC–3a, where C2H2 molecules outside the representative path are omitted for clarity. The C2H2 molecules at the initial position are marked with red color, and the C2H2 molecules at the other positions of the diffusion pathway are marked with orange, yellow, green, and cyan colors, respectively. Numbers 1 to 5 and the arrows represent the diffusion pathways.

Theoretical calculations

To get insight into the diffusion and adsorption process, Monte Carlo simulations and density functional theory (DFT) calculations were initially employed to optimize the adsorption positions of CO2 and C2H2 molecules in FDC–3a. The optimized cell parameters for the activated phase are barely different from those of experimental values of FDC–3a (Supplementary Fig. 26, Supplementary Table 7), suggesting the reliability of the optimized crystal structure of FDC–3a. In the optimization of CO2- and C2H2-adsorbed structures, cell parameters were kept the same as those of empty FDC–3a, because the adsorption of CO2 and C2H2 induced little structural transformation (Supplementary Fig. 10). Based on the above theoretical model, we investigated the CO2 and C2H2 adsorption and diffusion (Fig. 3, b, c). The CO2 and C2H2 diffusion barriers in FDC–3a were 35.8 and 55.9 kJ mol–1, respectively (Supplementary Fig. 27a). The large diffusion barrier differences indicated that the diffusion of CO2 was more kinetically favorable than the diffusion of C2H2. Indeed, although a self-accelerated adsorption process was demonstrated as a result of the temperature-promoted diffusion coefficients, the relative diffusion coefficient of CO2 was perpetually higher than C2H2 by 7 × 105- and 400-folds at low and high temperatures, respectively (Supplementary Fig. 27b). Such a large difference in diffusion rate at low temperatures suggests that the adsorption of C2H2 is much farther from equilibrium than that of CO2, resulting in its smaller adsorption amount than that of CO2 at low temperatures despite that the adsorption energy (‒40.4 kJ mol‒1) of C2H2 is stronger than that (‒26.5 kJ mol‒1) of CO2 (Supplementary Table 7). On the other hand, both C2H2 and CO2 adsorptions can reach adsorption equilibrium at high temperatures despite their different diffusion rates, leading to the selective adsorption of C2H2 over CO2 at high temperatures.

Mixed gas separation

The temperature-switched adsorption behavior and its diffusion-regulatory mechanism in FDC–3a inspired us to perform dynamic mixed gas separation experiments; these were carried out with temperature-programmed desorption (TPD) protocol (Supplementary Figs. 28, 29)18,19. Considering the adsorption amounts and the selectivity, we conducted the separation experiments at 240 and 320 K, respectively. At the low temperature of 240 K, FDC–3a selectively adsorbed CO2 from a nearly equimolar CO2/C2H2 mixture (CO2:C2H2 = 54.2:45.8) within a short exposure time of 1 h, leading to a remarkable CO2 enrichment with a composition up to 97.7% in the adsorbed phase (Fig. 4a, Supplementary Fig. 34) and a CO2/C2H2 separation factor of 36 (Fig. 4b). The separation factor was comparable with the IAST selectivity, though the values of the former were larger than the latter. The high CO2/C2H2 separation factor indicated that CO2 diffused much faster than C2H2, thus occupying most of the available space and excluding the C2H2 by a molecular-sieving mechanism. FDC–3a exhibited outstanding CO2 enrichment over a wide range of feed-gas components (Supplementary Figs. 3040); even though the mixture was in a composition of CO2:C2H2 = 4.0:96.0, FDC–3a enriched CO2 resulting in a CO2 concentration of 95.4% in the adsorbed phase (Fig. 4a) and a CO2/C2H2 separation factor of 498 (Fig. 4b). On the other hand, At the high temperature of 320 K, FDC–3a selectively adsorbed C2H2 from a nearly equimolar CO2/C2H2 mixture (CO2:C2H2 = 54.2:45.8) within a short exposure time of 1 h, leading to a remarkable C2H2 enrichment with a composition up to 94.1% in the adsorbed phase (Fig. 4a, Supplementary Fig. 45) and a CO2/C2H2 separation factor of 5.4 × 10–2 (i.e., C2H2/CO2 separation factor of 18, Fig. 4b). The high C2H2/CO2 separation factor suggested that FDC–3a was thermodynamically favorable to C2H2 when the diffusion-regulatory mechanism did not work at high temperatures. FDC–3a exhibited marked C2H2 enrichment over a wide range of feed-gas components (Supplementary Figs. 4151); even though the mixture was in a composition of CO2:C2H2 = 93.8:6.2, the C2H2 concentration was 92.6% in the adsorbed phase (Fig. 4a), which corresponds to a CO2/C2H2 separation factor of 5.5 × 10–3 (i.e., C2H2/CO2 separation factor of 181, Fig. 4b). Notably, the selectivities of the mixed-gas separation were higher than the ones predicted from the single-gas adsorption. At low temperatures, the diffusion of both CO2 and C2H2 was regulated, and the gas separation was governed by the diffusion-rate difference. CO2 showed a faster diffusion rate than C2H2, resulting in high CO2/C2H2 selectivities under non-equilibrium states. On the other hand, at high temperatures, the cooperativity of CO2-C2H2 interaction and gas-framework interaction amplified the selective adsorption of C2H2, rendering a higher C2H2/CO2 selectivity than the expected one from the single-gas adsorption.

Fig. 4: Mixed gas separation.
figure 4

a McCabe-Thiele diagram for CO2/C2H2 separation by FDC–3a at 240 and 320 K, with the dashed line representing the theoretical behavior of a material showing no selectivity. b The correlation between CO2 concentration in the feed gas and the CO2/C2H2 separation factor.


Our findings provide temperature-switched recognition of the CO2 and C2H2 by controlling their diffusion and amplifying the rate differences. The TPD results for kinetic gas separation of CO2/C2H2 binary mixtures demonstrate temperature-dependent high selectivities with a CO2/C2H2 separation factor of 498 at 240 K and a C2H2/CO2 separation factor of 181 at 320 K. These striking separation features should give the credit to the underlying mechanism, which is implemented by the cooperation of ultrasmall pore apertures and local dynamics of gate constituents. This design principle can be extensively adaptable with various host-guest systems for manipulatable selectivity trends by external stimuli for recognizing similar guests.


Synthesis of FDC–3

Firstly, 50 mg (0.09 mmol) OPTz-t3da was dissolved in 2 mL DMA at room temperature. A methanol solution (8 mL) of Zn(NO3)2·6H2O (54 mg, 0.18 mmol) was added to the above solution. Then the mixture was heated at 80 °C for 72 h. FDC–3 was obtained as colorless block crystals with sizes up to several hundreds of micrometers (37 mg, yield = 43%). The crystals were filtered, washed with DMA (10 mL, 3 times) and methanol (10 mL, 3 times), and dried in air. The as-synthesized FDC–3 was characterized by infrared spectra (Supplementary Fig. 3). The adsorption peak of the stretching vibration of the C = O double bond shifted to a low wavenumber, indicative of the coordination bond formation in FDC–3.

Solvent exchange and activation of FDC–3

To measure the adsorption property of FDC–3, we exchanged the guest and coordination solvents (DMA) with methanol by soaking FDC–3 in methanol at 60 °C for 7 days. Every 24 h the methanol was replaced by a new one. After the solvent exchange, the exchanged FDC–3 was dried under vacuum at 60 °C for 3 h. 1H NMR confirmed that the DMA in the exchanged FDC–3 was exchanged by methanol (Supplementary Fig. 8).

TG curve showed that the framework of the exchanged FDC–3 was thermally stable until 391 °C, whereas below 60 °C the exchanged FDC–3 lost the methanol molecules (Supplementary Fig. 9). Thus, we activated the exchanged FDC–3 at 120 °C for 11 h to afford FDC–3a; this temperature ensured the complete removal of the solvents meanwhile excluding the possibility of framework decomposition.