Introduction

Gas separation and purification are ongoing research topics, attracting interest from both academia and industry. In the last few years, CO2 separation, as a key component of carbon capture, utilisation and storage, has become a fast-growing research area. Carbon capture, utilisation and storage may provide a temporary solution for the continued use of fossil fuels until renewable energy becomes abundant, alleviating the emission of anthropogenic greenhouse gases1,2. Two important clean-energy-related separations involving CO2 are CO2 capture from flue-gas streams of power plants, and natural-gas upgrading. The first entails separating CO2 and N2, often under the harsh conditions of flue-gas streams from fossil-fuel-fired power plants. In natural-gas upgrading, from both natural reservoirs and landfills, the separation of CO2 from CH4 improves CH4 purity and reduces pipeline corrosion by the acidic CO2 gas. Various technologies for CO2 capture/separation from its mixture with other gases have been developed, including the relatively mature amine-scrubbing procedure3, and other techniques that are under intense exploration. Among them, adsorptive and membrane-based separations are some of the most promising because of their potential for simple and low-energy-consumption processes. The performance of porous materials, the supporting medium of adsorbent and membrane, is critical for efficient separation. Many porous solids such as zeolites and activated carbons have been widely studied for clean-energy-related separations, including CO2 capture4. However, methods for precise structural and property control of these materials, particularly at the molecular level, are still lacking. Recently, remarkable progress has been made in this regard in the development of metal-organic materials (MOMs), such as metal-organic frameworks (MOFs)5,6 and metal-organic polyhedra (MOPs)7, assembled from molecular building units. These materials possess unique properties and vast application potential8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24. The hallmark of these new materials is their modular design, which is far beyond the scope of other porous materials22,25,26,27,28,29,30,31,32,33,34,35,36 and allows for facile structure and property tuning. This therefore enables precise control of functions by accurate design of structures at the molecular level.

Although many MOFs have been prepared for CO2 separation in recent years, a precise construction implemented at the true molecular level is rare. Herein, as an experimental proof-of-concept, we demonstrate how it is possible to design a cavity specifically for the capture of a single CO2 molecule. The first step is the design and synthesis of a molecular cavity, designated ‘single-molecule trap’ (SMT) for the convenience of discussion, starting from the evaluation of the size and properties of the guest CO2 molecule (Fig. 1a). As molecular building blocks, these SMTs can be linked to construct a three-dimensional (3D) MOF to ensure efficient transport of CO2 in and out of the cavities. An alternative route is the impregnation of an existing porous material such as mesoporous silica with SMTs directly (Fig 1b). In both cases, the selectivity of the SMT for CO2 should be preserved.

Figure 1: Designing porous materials with SMTs for CO2 molecule capture.
figure 1

(a) Schematic representation of the design and construction of a SMT for CO2 adsorption. Organic linker is shown in orange, and the metal node with an open adsorption site is shown in purple and yellow; green spheres represent the cavity for CO2 adsorption. (b) Two proposed strategies for building a pre-designed SMT into extended porous solids: assembly of the SMTs by ligand extension and accommodation of SMTs into other porous materials, such as mesoporous silica.

Full size image

A CO2 molecule is linear, about 2.4 Å long, positively charged at the C atom, and negatively charged at the O atoms. The normal M–O co-ordination bond length, where M represents a transition metal, is about 2.0–2.8 Å. Derived from the above distances, two positively charged ‘anchors’13 separated by about 6.4–8.0 Å (2.4+2 × M–O) distance would be ideal for holding a CO2 molecule between them through electrostatic interactions. Negatively charged groups surrounding the C atom should strengthen the CO2–trap interaction even further. With these considerations in mind, we have prepared a molecular cage using four bridging ligands and two metal clusters (with open metal sites), which are separated by about 7.4 Å (from crystal structural determination) and expected to form strong electrostatic interactions with the negative O atoms of CO2, without the formation of a chemical bond. Such precisely designed SMTs should hold one CO2 molecule per cavity, excluding other small molecules such as N2 and CH4, allowing for the highly selective adsorption of CO2 over other gases. Herein, we describe the design and synthesis of such functional porous materials for efficient CO2 molecule capture, as well as the evaluation of the resulting materials by both molecular simulations and experiments. This exemplifies a new strategy in the highly controllable design and preparation of porous materials for specific adsorption-related applications.

Results

MOP-based SMTs for CO2 selective adsorption

To construct a MOP with fixed shape and distance between metal atoms, an organic ligand with appropriate geometry and size is a prerequisite. Our recent success in the assembly of MOPs with a variety of shapes, sizes and functional groups, by tuning the bridging ligands, enables us to synthesise a single cavity specifically designed for the accommodation of one CO2 molecule7,37. In this work, two dicarboxylic acid ligands, 3,3′-(naphthalene-2,7-diyl)dibenzoic acid (H2L1) and 3,3′-(1,8-naphthyridine-2,7-diyl)dibenzoic acid (H2L2) were designed and synthesised by Pd-catalysed coupling reactions between 3-(methoxycarbo nylphenyl)boronic acid and 2,7-dibromonaphthalene or 2,7-dibromo-1,8-naphthyridine, respectively, followed by hydrolysis of the corresponding esters. With the ‘cis and parallel’ arrangement of the two carboxylic acid groups in each ligand, both have a bridging angle of 0° and an inter-carboxylic-acid distance of about 7 Å, thus fulfilling both metal–metal distance and anchor orientation requirements in the resulting MOPs. The SMTs were assembled from H2L1 and H2L2, shown in Fig. 2a, by reacting with Cu2(OAc)4·2 H2O in N,N-dimethylacetamide (DMA) to give isostructural [Cu2(L1)2(DMA)2]2 (SMT-1-DMA) and [Cu2(L2)2(DMA)2]2 (SMT-2-DMA), at room temperature as phase-pure crystals, confirmed by a good match between the experimental and simulated powder X-ray diffraction (PXRD) patterns (Supplementary Figs S1 and S2).

Figure 2: Synthesis, structures and CO2 selective adsorption of precisely designed SMTs.
figure 2

(a) Schematic representation of the construction of SMT-1–3: S is the coordinated solvent molecule, which can be removed through sample activation to yield an empty molecular cage for gas adsorption. (b) The molecular structure of SMT-1 with (top) and without (bottom) coordinated solvent molecules. The colour schemes: Cu, cyan; O, red; N, blue; C, brown; and H, light grey. The green sphere represents the free space inside the molecular cage. (c) Gas adsorption isotherms of SMT-1, showing selective adsorption of CO2 over CH4 and N2 (inset: a picture of activated SMT-1 sample).

Full size image

The structures of the two compounds were determined by single crystal X-ray diffraction. As expected, both MOP molecules have a lantern-type structural arrangement with four bridging ligands surrounding two paddlewheel dicopper motifs to form a large paddlewheel tetragonal cage (Fig. 2b and Supplementary Figs S13 and S14). Each copper atom is coordinated by four carboxylate oxygen atoms of different ligands in the equatorial positions and one DMA molecule in the axial position, which can be removed to leave an open metal site. The dimensions of the two MOPs excluding coordinated solvents in height (Cu–Cu) and diameter (C–C between two opposite ligands) are 12.6 × 19.4 and 11.9 × 19.8 Å, respectively; and the inner cavity sizes are 7.4 × 10.7 and 6.7 × 12.5 Å (atom-to-atom distances across opposite Cu and H (or N) atoms of ligands). Particularly, the distances between two opposite copper sites in each cavity of SMT-1 and -2 are 7.4 and 6.7 Å, respectively, suitable for the accommodation of one CO2 molecule through the interaction of its two O atoms to the copper sites. Similarly, the reaction between H2L2 and Mo2(O2CCF3)4 afforded compound [Mo2(L2)2(DMA)]2·2DMA (SMT-3-DMA), with a similar structure to that of SMT-2-DMA (see Supplementary Fig. S15). The size of the molecule and dimensions of the cavity are 11.5 × 20.3 and 7.3 × 12.9 Å, respectively.

The methanol-exchanged SMT-1-DMA was further activated under a vacuum at 80 °C to give desolvated material, SMT-1, which, as a representative of these SMTs, was tested for gas adsorption. As shown in Fig. 2c, SMT-1 shows a significant adsorption for CO2 at both low and enhanced temperatures, but almost excludes N2 and CH4. At 196 K and 1 atm, the uptake of CO2 is 52 cm3 g−1 (2.32 mmol g−1), corresponding to about 4 CO2 molecules adsorbed per SMT-1 molecule, implying that besides the cavities of SMT molecules, the space between them are also accessible for CO2 adsorption. The evaluated Langmuir surface area is 153.7 m2 g−1 and the pore volume is 0.097 cm3 g−1, based on the CO2 adsorption data at 196 K. The adsorption capacity of CO2 at 273 and 296 K are 19 (0.85) and 14 (0.63) cm3 g−1 (mmol g−1), corresponding to about 1.5 and 1 CO2 molecules per SMT-1, respectively. The CO2 adsorption enthalpy at zero coverage is ca. −35 kJ mol−1 (see Supplementary Fig. S17), higher than that in most MOFs, particularly those constructed from pure carboxylate ligands and similar paddlewheel dicopper clusters5,38. However, at 196 K and higher temperatures, the adsorption of N2 and CH4 falls below the detection limit of our instrument. The lack of accurate data sets for N2 and CH4 adsorption limits the quantitative estimation of adsorption selectivities of these gases in this porous molecular material. It should also be pointed out that the desolvated SMT-1 sample is still crystalline, as confirmed by PXRD (Supplementary Fig. S1), but the diffraction peaks deviate from those of fresh SMT-1-DMA due to the axial ligand removal and the resulting overall parking structure distortion. This observation differs from those of several previously reported MOP materials, which usually became amorphous after the solvent molecules were removed, but are still porous at the molecular level7,37. In addition, SMT-2 also adsorbs CO2 up to 25 cm3 g−1 (1.11 mmol g−1) at 273 K, but the uptakes of N2 and CH4 are well below the detection limit (see Supplementary Fig. S18). It is clear that these precisely designed SMTs are efficient for the selective CO2 adsorption.

To evaluate whether the CO2 molecule prefers staying in the cavity of these SMTs, the CO2-binding energy of the SMTs with a variety of proposed metal sites (Zn, Cu, Rh and Mo) was calculated using density functional theory (DFT). The simulation indicates that it is indeed energetically favourable for a CO2 molecule to stay in the cavity of these SMTs (see Supplementary Table S9). Moreover, SMTs with different metal ions show distinct affinities towards CO2. Among them, the Zn-based SMT shows the strongest interaction with a trapped CO2 molecule, and the Mo-based one exhibits the weakest CO2 affinity. These different CO2 affinities arise from the ligand field preference of the metal ions. For example, the Mo24+ cluster contains a quadruple bond and usually interacts weakly with axial ligands/guests37. The Zn-containing paddlewheel unit, on the other hand, contains no metal–metal bond. After removal of the coordinated solvent molecules, each Zn2+ ion in the simulated compound prefers penta-coordination to square-pyramidal ligand environment through accepting an additional coordinating guest (such as CO2), or to transfer to a tetrahedral environment. Because of the instability of square planar Zn2+, it is difficult to synthesise robust Zn-based SMTs, or other MOFs built from Zn-based paddlewheel units.

The overall valid porosity of a MOP in the bulk sample is usually discounted because of the partial conglomeration of individual cages. When incorporated into a rigid porous framework, much more internal surface of MOP cages should become accessible.

Building the pre-designed SMTs into extended materials

Success in the construction of SMTs and their efficient capture of CO2 molecules motivated us to build these molecular cages into extended framework materials to fully expose their internal surface and efficiently enhance the mass transfer for CO2 adsorption and separation. The two above-mentioned strategies shown in Fig. 1 were tried initially, but only one, introducing the SMTs as building units into a MOF, has thus far been achieved. Incorporating these SMTs into other porous materials was hampered by the poor solubility of them in all tested solvents. To avoid precipitation of the SMT intermediate, we designed a ligand with two types of carboxylate groups, one to support the SMT and the other for the formation of an extended framework. Thus, an extended framework with SMTs can be assembled in a one-pot reaction. In this context, the dicarboxylic acid ligand H2L1 was functionalised by adding two additional carboxylic acid groups to generate 5,5′-(naphthalene-2,7-diyl)diisophthalic acid (H4L3, Fig 3a), which was synthesised by a procedure similar to that of H2L1, with diethyl 5-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan- 2-yl)isophthalate instead of 3-(methoxycarbonylphenyl)boronic acid. H4L3 reacted with Cu(NO3)2·2.5 H2O under solvothermal conditions (85 °C for 2 days) affording [Cu(L3)1/2(S)]n (PCN-88-S) (porous coordination network (PCN), S represents non-assignable solvent molecules), which has a 3D framework structure containing the pre-designed SMT-1 units. The phase purity of the product was confirmed by PXRD (Supplementary Fig. S4) and the structure was analysed by single crystal X-ray diffraction (see Supplementary Information).

Figure 3: Construction, structure and selective gas adsorption performance of a MOF with built-in pre-designed SMTs.
figure 3

(a) Schematic representation of the construction of PCN-88 and its 3D framework structure. Colour scheme: Cu, cyan; O, red; C, brown; and H, light grey. The green sphere represents the free space inside the built-in SMTs. (b) A SMT unit in PCN-88 with its linkages towards extending to the 3D framework. (c) Two conformers of ligand L3 observed in PCN-88. (d) Two complementary polyhedra (elongated convex/concave square dipyramid) in PCN-88 structure. Each polyhedron is formed by connecting dicopper nodes as vertices. (e) CO2, CH4 and N2 adsorption isotherms of PCN-88 at two temperatures. (f) Pressure-dependent IAST adsorption selectivities of CO2 over N2 and CH4 at 273 K (red) and 296 K (blue) in PCN-88 calculated based on single-component adsorption data.

Full size image

As shown in Fig. 3c, in PCN-88, there exist two conformers of L3 ligands, one (α-) with a dihedral angle of approximately 90° between the central and peripheral rings, whereas the other (β-) has a dihedral angle of approximately 0°, similar to a previously reported series39. In the MOF, four α-L3 ligands bridge two dicopper paddlewheel motifs to form a lantern-type unit similar to SMT-1, with eight additional carboxylate groups pointing to eight vertices of a quadrangular prism containing a SMT unit, in which the distance between two opposite copper sites in the cavity is about 7.4 Å (Fig 3b). Through these carboxylate groups, each quadrangular prism is linked to eight others by sharing eight dicopper motifs to form a 3D network (Fig 3a). In each end of a quadrangular prism, one of the four ‘blades’ of each of the four dicopper paddlewheel motifs are further capped by carboxylate groups of one β-L3 ligand. On the other hand, only two types of supramolecular building units exist if the centres of all dicopper paddlewheel motifs are connected. One is shown in Fig. 3d, top, which is an elongated square dipyramid; the other is a concave elongated square dipyramid with the two pyramids inverted (Fig. 3d, bottom). The two units are complementary to each other in space filling. Six units of one type surround one unit of the other octahedrally forming a 3D network. After removal of coordinated solvent molecules, the solvent-accessible volume of the crystal (13667.3 Å3 out of the 18544.0 Å3 per unit cell) is 73.7% (ref. 40).

After activation of the as-synthesised material by successive methanol/dichloromethane solvent exchanges and heating at 100 °C under reduced pressure, PCN-88 retained its crystallinity as confirmed by PXRD (Supplementary Fig. S4). N2 and Ar adsorption isotherms at 77 and 87 K of the activated sample (see Supplementary Figs S19 and S21, respectively) further revealed permanent porosity. A two-step adsorption was observed in N2 and Ar isotherms, with Brunauer-Emmett-Teller (BET) and Langmuir surface areas of 3308 and 3845, m2 g−1, respectively, based on the N2 adsorption data (calculated from the first step of the adsorption isotherm). The stepwise adsorption isotherm can be explained by the existence of both meso- and micro-porous cages in the structure, as observed previously in similar systems15. As shown in Fig. 3e, PCN-88 has a significant CO2 (160 (7.14) and 94 (4.20) cm3 g−1 (mmol g−1)) uptake, but little CH4 (33 and 18 cm3 g−1) and N2 uptake (2.6 and 2.1 cm3 g−1) at 273 and 296 K and 1 atm, respectively. It should be pointed out that the CO2 uptake capacity of PCN-88 at 296 K and 1 atm ranks among the highest for MOFs constructed from pure carboxylate ligands (see Supplementary Table S10) excluding the MOF-74 series, which are constructed from ligands with –OH moieties5,38. These observed selective adsorptions of CO2 over CH4 and N2 can be attributed to the much stronger interactions of the CO2 molecule with the adsorbent surface compared with those of the other two gases. The calculated Henry’s law selectivities of CO2/CH4 and CO2/N2 are 7.3 and 67.9 at 273 K, and 5.3 and 18.1 at 296 K, respectively. Furthermore, ideal adsorbed solution theory (IAST (ref. 41) selectivities are also calculated by using the adsorption data of single-component gases (Fig. 3f). The values of IAST selectivities at 273 K and 1 bar for a 15/85 and a 50/50 CO2/N2 mixture are 23.1 and 27.4, respectively; and those at 296 K and 1 bar are 15.2 and 17.6, respectively. For a 50/50 CO2/CH4 mixture, corresponding values are 7.1 at 273 K and 7.0 at 296 K and 1 atm. The highly selective adsorption of CO2 over N2 and CH4 suggests possible application of PCN-88 in capturing CO2 from flue gas or upgrading natural gas. The zero-coverage heat of adsorption of CO2 in PCN-88 is ca. −27 kJ mol−1, which is higher than that in most MOFs constructed from pure carboxylate ligands and paddlewheel dicopper clusters5,38. This intermediate adsorption enthalpy implies a low-energy consumption for adsorbent regeneration, a desired property for separation applications. In fact, CO2 capture in industry is highly process dependent, in which adsorbents are always optimised for a specific system. These equilibrium adsorption data only provide an initial step for improving the efficiency of a process. Additional experiments are still required for validating the practical value of these new adsorbents.

The CO2, CH4 and N2 adsorption in PCN-88 have also been evaluated by grand canonical Monte Carlo (GCMC) simulations. The simulated adsorption isotherms of single-component gases closely match the experimental ones as shown in Supplementary Fig. S36. To understand the adsorption behaviours of these gases in PCN-88 at the molecular level, snapshots of the structures with adsorbed gas molecules at 296 K and 1 atm were analysed (see Supplementary Fig. S37). It was found that the most preferred sites for CO2 adsorption in PCN-88 are the corner regions close to metal sites, probably because of the strong electrostatic interaction between them. On the contrary, CH4 and N2 molecules prefer to be adsorbed around the organic units because of van der Waals interactions. Furthermore, the adsorption behaviours and selectivities of gas mixture containing 15% CO2 and 85% N2 (mimicking real flue gas), as well as 50% CO2 and 50% CH4 in PCN-88, at different temperatures were also evaluated with GCMC simulations; the results are shown in Supplementary Figs S38–40. The evaluated adsorption selectivity of CO2 over N2 is about 15.3 and 10.2, and for CO2 over CH4 is 5.2 and 3.8 at 273 and 296 K and 1 atm, respectively. It is much more interesting that the snapshot at 296 K for the CO2/N2 mixture adsorption in PCN-88 shows that there are quite a few CO2 molecules but no N2 molecules absorbed in the trap cavities. In some cases, the CO2 molecule actually ‘stands up’ in the trap cavity, as shown in Fig. 4, demonstrating the validity of this designed SMT for CO2 selective capture.

Figure 4: Simulated locations of CO2 and N2 molecules in PCN-88 for CO2/N2 (15:85) mixture.
figure 4

Simulation performed with a CO2/N2 (15:85) mixture at 296 K and 3 Mpa. Viewed from different directions: (a) along the crystallographic z direction and (b) along the x direction. Colour scheme: Cu, dark orange; O, red; N, blue; C, dark grey; and H, light grey. It is clear that one CO2 molecule (shown in space-filling model) can be trapped in the centre of a SMT cavity.

Full size image

In addition, the diffusion of CO2 and its mixture with N2 in PCN-88 at 296 K has also been studied by molecular dynamic simulations (see Supplementary Figs S43 and S44). It was found that single CO2 molecules can diffuse into trap cavities but cannot be trapped at low loading; however, with increased loading, CO2 is trapped in the cavities. For CO2/N2 mixtures, it was confirmed that only CO2 molecules can be trapped in the cavities, which is consistent with the GCMC simulations discussed above. Again, this result shows that the pre-designed SMTs in PCN-88 work well in selectively capturing CO2 molecules.

SMT’s impact on CO2 adsorption in PCN-88

To further evaluate the effect of the SMTs on CO2 adsorption in PCN-88, a control experiment was designed and performed. As shown in Fig. 5a, blocking the cavities of SMTs in PCN-88 results in MOF [Cu(L3)1/2(pyrazine)1/12(S)]n (PCN-88′ -S), in which the trap cavities are occupied by pyrazine molecules by coordinating to two copper sites. PCN-88′ -S was synthesised by a one-pot reaction of H4L3, pyrazine and Cu(NO3)2·2.5 H2O under conditions similar to that for PCN-88- S. Its structure is almost identical with that of PCN-88- S except for the blocked SMTs as mentioned above. The structure contains a pore-accessible volume of 72.6% (13530.0 out of the 18646.0 Å3 per unit cell volume) after removal of the coordinated solvent molecules40. The permanent porosity of desolvated material, PCN-88′, was revealed by the N2 and Ar adsorption (see Supplementary Figs S20 and S22) at 77 and 87 K, respectively. The evaluated BET and Langmuir surface areas are 3042 and 3799, m2 g−1, respectively, based on N2 adsorption data. Similar to that of PCN-88, two-step adsorption isotherms for N2 and Ar, and different adsorption quantities of CO2 (119 (5.31) and 70 (3.13) cm3 g−1 (mmol g−1)), CH4 (27 and 15 cm3 g−1) and N2 (2.4 and 1.8 cm3 g−1) at 1 atm and 273 and 296 K, respectively, were observed (see Supplementary Fig. S31). Clearly, compared with PCN-88, PCN-88′ yields a lower adsorption capacity towards these gases at 273 and 296 K due to the blocked SMT cavities. Interestingly at 195 K, both MOFs show almost the same CO2 uptake capacities of about 850 cm3 g−1 at 1 atm; however, PCN-88 has a significantly enhanced CO2 uptake compared with PCN-88′ at pressures below 0.60 atm (450 mmHg), particularly at about 0.20 atm (150 mmHg) (Fig. 5c). This much higher adsorption at low pressure (lower than 150 mmHg) may be attributed to the contribution from SMTs in PCN-88, which is, as anticipated, very important for CO2 separation from other gases, such as N2 in flue gases, where the partial pressure of CO2 is about 0.15 atm. For PCN-88′, the calculated Henry’s law selectivities for CO2 over CH4 and N2 are 6.1 and 49.4 at 273 K, and 5.0 and 12.6 at 296 K, respectively. Calculated IAST selectivities are 16.5 and 18.5 at 273 K and 1 bar, and 12.3 and 14.6 at 296 K, for 15/85 and 50/50 CO2/N2 mixtures, respectively. The related values for a 50/50 CO2/CH4 mixture are 6.6 and 5.8 at 273 and 296 K and 1 bar, respectively. All these evaluated selectivities in PCN-88′ are lower than those in PCN-88 (see Supplementary Table S6), indicating that the SMTs in the former have an effect on the selective adsorption of CO2 over CH4 and N2. The zero-coverage heat of adsorption of CO2 in PCN-88′ is –25 kJ mol−1, again lower than that in PCN-88. In addition, it was also noted that PCN-88 has a higher H2 uptake than that of PCN-88′ at 77 and 87 K, respectively, but both show similar adsorption enthalpies (see Supplementary Figs S24–26). In conclusion, this control experiment shows that the pre-designed SMTs in PCN-88 have a significant influence on the adsorption of CO2 in the low-pressure region.

Figure 5: Control experiment to identify the effect of SMTs in the PCN-88 on CO2 adsorption.
figure 5

(a) Schematic representation of the synthesis of PCN-88′. (b) Schematic representation of introducing a pyrazine molecule into the cavity of PCN-88 to block the SMT. Colour scheme: Cu, cyan; O, red; and C, brown. The green sphere represents the free space inside the SMT. (c) CO2 adsorption isotherms of PCN-88 and -88′ at 196 K, showing significantly different uptakes at low pressure due to the contribution of SMTs in the former.

Full size image

Discussion

All results discussed above evidently demonstrate the feasibility of precisely designing a cavity in porous MOMs for specific adsorption-based applications. Clearly, designing a MOF or other composite with a higher density of SMTs will substantially improve the overall adsorption performance of a particular guest molecule, such as CO2 in this SMT-based system presented here. For many years, there has been a push towards porous materials with high surface areas and large pore sizes, particularly for MOFs. But for CO2 selective adsorption or capture (also for H2 storage), these features have not been effective. Recently, a strategy based on pore space partition in a MOF through cage-within-cage, or core-shell, type architecture has been proposed by Bu and colleagues42, as a strategy to optimise the pore space for better CO2 capture. But the cage-within-cage configuration does not address the individual fit between the partitioned pore space and the size of the gas molecule. Here we represent the first time (to our knowledge) to design a cage-type assembly to specifically address the trapping of a particular molecule based on its chemical (polarity and so on) and physical (size and shape) properties. The advantage of this method is that it not only enhances the host–guest interactions because of an optimum fit, but it also simultaneously endows the material with superior selectivity over other guest molecules because the cavity is custom-designed for a specific molecule.

In summary, a new concept, the ‘SMT’ has been utilised in designing porous materials at the molecular level for specific adsorption-based applications. As an experimental proof-of-concept, such a SMT was first designed and synthesised for CO2 molecule capture by the precise selection of organic ligand and metal node to form a MOP. Experimental and computational simulation results demonstrate the efficiency of this MOP-based SMT in the preferential adsorption of CO2 over N2 and CH4, which are important for CO2 capture from flue gas and CH4 upgrading. Building this SMT into a MOF with an extended framework structure was achieved, again, through ligand design. This resulting MOF preserves the SMT properties in capturing CO2 molecules, therefore showing good performance in the selective adsorption of CO2. In principle, by judiciously designing a SMT and thereby its extension into porous frameworks, this method can be applied in the synthesis of other new MOMs and composite materials for other specific applications, such as H2 storage and even CO2 activation by trapping one CO2 in a confined cavity with strongly binding active sites. These materials may be used in applications such as small molecule capture, activation, conversion and storage, wherever binding single guest molecule is a prerequisite.

Methods

Materials and syntheses

All general reagents and solvents (analytical grade (AR) grade) were commercially available and used as received. The protonated ligands 3,3′-(naphthalene-2,7-diyl)dibenzoic acid (H2L1), 3,3′-(1,8-naphthyridine-2,7-diyl)dibenzoic acid (H2L2), 5,5′-(naphthalene-2,7-diyl)diisophthalic acid (H4L3) and 5,5′-(1,8-naphthyridine-2,7-diyl)diisophthalic acid (H4L4, not used in the synthesis of MOFs because of its high cost) were newly designed and synthesised by using palladium-catalysed coupling reactions, followed by hydrolysis. The precursors in these reactions were purchased or synthesised according to the literature methods. SMT-1-DMA and SMT-2-DMA were synthesised by the reaction of Cu2(OAc)4·2H2O with corresponding carboxylic acid ligands in DMA at room temperature. Single crystals of SMT-1-DMA were obtained by layering a solution of H2L1 in DMA upon a DMA solution of Cu2(OAc)4·2H2O and staying for several days. For SMT-2-DMA, two solutions of H2L2 and Cu2(OAc)4·2H2O in DMA were mixed and left undisturbed for several days to obtain crystals. Single crystals of SMT-3-DMA were synthesised by heating a DMA solution of H2L2 and Mo2(O2CCF3)4 with 48% aqueous solution of HBF4 at 85 °C for 12 h. These SMT-DMA crystals slowly lose transparency as solvent molecules evaporate into the air over a longer period of time, and are insoluble in DMA, N,N-dimethylformamide, dimethyl sulfoxide, methanol and acetone. PCN-88- S and -88′- S were synthesised by similar solvothermal procedures: a DMA/water solution containing Cu(NO3)2·2.5H2O, H4L3; a few drops of HBF4 solution (and pyrazine for PCN-88′- S) was heated at 85 °C for 2 days to afford single crystals. Phase purity of all products was evaluated by PXRD; general characterisation including infrared and thermogravimetric analysis were carried out. Detailed descriptions of syntheses and characterisation are given in the Supplementary Methods.

X-ray single-crystal diffraction

Data were collected on a Bruker AXS APEX-II CCD (charge-coupled device) diffractometer with a fine-focus sealed-tube X-ray source (Mo-Kα) and the structures were solved by direct methods. Details are provided in the Supplementary Information. In all the structures except SMT-3-DMA, free solvent molecules are disordered, and attempts to locate and refine all the solvent molecule peaks were unsuccessful. Diffused electron densities resulting from these solvent molecules (except the O atoms of coordinated solvents) were removed from the data set using the SQUEEZE routine of PLATON and refined further using the data generated40. The coordinated solvent molecules in PCN-88- S and -88′- S are also disordered, only the O atoms were left and refined and other atoms were omitted. Contents of the removed solvent region are not represented in the unit cell contents in the crystal data of related compounds. Attempts to determine the final formula of these compounds from the SQUEEZE results, combined with elemental analysis and thermogravimetric analysis data, were unsuccessful because of the volatility of crystallisation solvents during measurements.

Gas adsorption

Gas adsorption measurements were performed using an ASAP 2020 volumetric adsorption analyser. A high-purity grade (99.999%) of gases was used throughout the adsorption experiments. Before adsorption, the sample was activated by solvent exchange (in several cycles using fresh methanol) to remove the coordinated DMA molecules, followed by degassing at elevated temperature and vacuum. Details are given in the Supplementary Information. To get more precise adsorption data, particularly for N2 and CH4 adsorption at higher temperatures, such as 296 K, the free space of the sample holder was minimised as much as possible. In evaluating the adsorption selectivities of these material towards different gases including CO2, N2 and CH4, Henry’s law selectivity and IAST selectivity are calculated based on the adsorption data of each single gas.

Computational simulations

DFT calculations were performed for three purposes: to optimise the geometries of CO2 interaction with SMTs; to evaluate the binding energies between CO2 and SMTs; and to calculate the partial charges of PCN-88 framework atoms that were used in GCMC simulations. All DFT calculations were carried out using Gaussian 0943. GCMC simulations were employed to calculate the uptake of single gases and gas mixtures in PCN-88, which was treated as a rigid framework, using the MUSIC program44. The potential parameters for the framework atoms were taken from DREIDING force field. Molecular dynamic simulations were performed using the DL-POLY-2.20 with the Berendsen thermostat in the NVT ensemble. Detailed descriptions and additional discussions are provided in the Supplementary Methods.

Additional information

Accession codes: The X-ray crystallographic coordinates for structures reported in this Article have been deposited at the Cambridge Crystallographic Data Centre (CCDC), under deposition number CCDC 867849 to 867853. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif.

How to cite this article: Li J-R. et al. Porous materials with pre-designed single-molecule traps for CO2 selective adsorption. Nat. Commun. 4:1538 doi: 10.1038/ncomms2552 (2013).