Single crystals of 1 were obtained by a solvothermal reaction of H4BINDI with Sr(CH3COO)2 in a mixed DMF-H2O solvent. Single-crystal X-ray diffraction (SCXRD) revealed that 1 crystallizes in the triclinic space group P-1, containing one Sr2+ ion, half of BINDI4− ligand, and one coordinated H2O/DMF (each with a half-occupancy) in an asymmetric unit (Figure S1). Sr2+ ion is eight-coordinated with seven oxygen atoms from six BINDI4- ligands and one H2O or DMF molecule, respectively. Each BINDI4− ligand links ten Sr2+ ions. Sr2+ ions are connected by carboxylate groups of BINDI4- into infinite 1D chains. Each 1D chain is connected by the isophthalate fragments to form 2D layers (Figures S2 and S3), which are further to a 3D framework by BINDI4- ligands (Fig. 1a). By ignoring the solvent molecules, there are 0D pores (void = 15.5%).
Interestingly, the color of 1 changes from brown to pink (denoted as 1a) after immersing in water for 30 minutes. Powder X-ray diffraction (PXRD) pattern of 1a exhibit significant change compared with 1, indicating a phase transformation (Figure S4). SCXRD revealed that the coordination mode of BINDI4- retains, but the coordinated DMF molecules have been completely replaced by H2O molecules, corresponding to the formula of [Sr2(BINDI)(H2O)2]‧G (1a), which was confirmed to be a single-crystal to single-crystal transformation by several different batches of single-crystal diffraction measurements. It is worth noting that the channels have been changed from 0D to 1D (void = 23.0% and aperture size = 5.1 × 9.1 Å2) along the a-axis (Figs. 1a and 1c). The significant change in the channel is attributed to the substitution of the coordinated DMF by smaller H2O ligands, which reduces the steric hindrance for the adjustment of the large organic ligands, and hence increases the dihedral angle (φ = 85.8°) between the phenyl and naphthalenediimide (NDI) groups in BINDI4- ligand in 1a vs that (φ = 73.1°) of 1 (Fig. 1b).
The purity of 1a was confirmed by PXRD pattern, which is consistent with the simulated one (Figure S4). Thermogravimetric analysis of 1a showed that the guests can be completely removed below 140 oC, obtaining the free guest phase ([Sr2(BINDI)(H2O)2], 1b) (Figure S5). N2 sorption isotherms were measured for 1b at 77 K, which showed a type I curve (Figure S6). The saturated N2 uptake is 4.6 mmol/g, corresponding to the pore volume of 0.16 cm3/g, which is very close to the value calculated from the crystal structure (4.7 mmol/g). The calculated Brunauer-Emmett-Teller (BET) and Langmuir surface areas are 100.6 and 103.4 m2/g, respectively. In addition, 1a retains its framework after being exposed to air for at least one month or immersed in various organic solutions or boiling water for at least one week, as well as in aqueous solutions with a wide pH range from 3 to 10 for 72 h, suggesting the high stability of 1a (Figures S7 and S8).
It has been reported that, when the electron-rich aromatic hydrocarbons interact with the electron-deficient NDI cores, there will be an excitation emission with charge-transfer (CT) characteristics.31–32 In addition, the ordered arrangement of the NDI cores benefits the CT with the electron-rich aromatic hydrocarbons. Both the guest-free phases of 1 and 1b were exposed to various aromatic hydrocarbon vapors to explore the possibilities as fluorescent sensors, respectively (Figure S9). Interestingly, there is no fluorescence enhancement for 1 under UV light, due to the small and discrete 0D pores, while the fluorescence is significantly brightened for 1b⊃VOCs (Figure S10). The fluorescence emission spectra illustrate the intensity variation and λem values (Fig. 2a). The λem values of the 1b⊃VOCs exhibit different red shifts dependent on the structures of benzene derivatives, with increasing to the electron-donating capability upon excitation at 310 nm, corresponding to 496 nm for 1b⊃benzene, 543 nm for 1b⊃toluene, 578 nm for p-xylene, 579 nm for m-xylene, 582 nm for o-xylene. These results suggest that the CT between the NDI cores and aromatic hydrocarbons dominates the fluorescence enhancement.
Very interestingly, 1b exhibits ∼116 fold fluorescence enhancement before and after exposure in saturated benzene vapor, which is the record value compared to all the MOFs fluorescent sensors reported to be responsive to organic solvents/vapors to date (Fig. 2b). Meanwhile, 1b achieves ∼73, ∼50, ∼37 and ∼23 fold fluorescence enhancement in saturated toluene, o-xylene, m-xylene and p-xylene vapors, respectively, which are also much higher than all the known MOFs. In addition, 1b exposing to other common saturated organic solvent vapors (including cyclohexane, hexane, acetone, dichloromethane and ethanol) exhibits negligible fluorescence enhancement, indicating that it possesses excellent selectivity (Figs. 2b and S11). Notably, all 1b⊃aromatic hydrocarbons can be fully regenerated by heating at 140 oC for 12 h. The fluorescence enhancement of 1b⊃aromatic hydrocarbon exhibits negligible diminishing after multiple cycle measurements, suggesting that 1b is reproducible as a sensing material (Fig. 2c).
The fluorescence response time for sensors is another key parameter for evaluating detection performance. To investigate the sensitivity of 1b for benzene, an in-situ solid-state luminescence sensing setup was performed according to the literature (Fig. 2d).30 After the addition of ∼100 µL benzene solution into a quartz cuvette, the luminescence intensity of 1b increased rapidly and reached an equilibrium after 14 s (Fig. 2e and Supplementary Movie), indicating that 1b can be used as an excellent sensor for the rapid detection of trace benzene vapor. Meanwhile, the corresponding equilibrium times of 1b for toluene, o-xylene, m-xylene and p-xylene were 40 s, 45 s, 215 s and 295 s, respectively (Figure S12). These results indicate that the equilibrium time probably relates to the concentration of vapor and the van der Waals volumes of the analytes. In fact, the absorption of 1b reaches equilibrium more quickly when exposed to the relatively high concentrations of benzene and toluene vapors of the lower boiling points (more volatile at room temperature). In contrast, when exposed in lower concentrations of vapors from o-xylene, m-xylene and p-xylene with higher boiling points, coupled with the larger van der Waals volumes, the absorption of 1b exhibits longer equilibration times (Table S1).
To explore the ultra-fast fluorescence "turn-on" behavior of 1b on benzene vapor, the adsorption curves of benzene were measured for 1 and 1b at 298 K (Figs. 3a). Interestingly, benzene molecules can diffuse into the pores of 1b, but not into 1 due to the discrete 0D pores. Specifically, there is a significant uptake from low pressure (P/P0 < 0.01) and the uptake approaches an adsorption saturation at P/P0 = 0.1, corresponding to the maximum uptake of 40.2 cm-3 g-1. This result confirms that fluorescent "turn-on" behavior of 1b is probably induced by the interactions between guest benzene molecules within the host framework, rather than the particle surface interactions. Furthermore, adsorption kinetic experiments of benzene vapor at different pressures verified that the fluorescence enhancement behavior of 1b can be attributed to the encapsulation of aromatic hydrocarbon molecules in the framework (Fig. 3b). The benzene vapor reached equilibrium in less than 1 minute at P/P0 = 0.95, and the adsorption capacity (41.9 cm-3 g-1) is consistent with the saturation uptake (40.2 cm-3 g-1) from the adsorption isotherm. It is worth noting that the required time for fluorescence enhancement of 1b to the maximum value at saturated benzene vapor coincides with that from the adsorption kinetic. The adsorption equilibrium time increases as the benzene vapor pressure decreases from P/P0 = 0.95 to P/P0 = 0.03, reaching an equilibrium within 5 minutes at P/P0 = 0.03. These results indicate strong interactions between benzene and the host framework of 1b, resulting in an extremely fast adsorption rate for benzene vapor. In addition, the PXRD pattern shows a significant shift in 1b⊃benzene compared to 1b, implying the single-crystal to single-crystal structural transformations in the process of benzene adsorption (Figure S16).
Single-crystal structure of 1b⊃benzene clearly shows the position of benzene in the pore (Fig. 3d). Compared to 1b, the space group of 1b⊃benzene is preserved, while their unit-cell parameters are similar, though the α angle being changed from 79.502(2)° to 82.704(6)°. Therefore, the fluorescent "turn-on" behavior of 1b⊃benzene can be explained by the donor-acceptor electron-transfer mechanism. The extended network of 1b⊃benzene can be considered as a giant "molecule" due to its highly localized electronic structure, allowing conduction band (CB) and valence-band (VB) to be described in the same way as molecular orbitals.33–34 Time-dependent Density Functional Theory (TDDFT) calculations were performed for 1b⊃benzene, which indicated that the lowest unoccupied molecular orbital (LUMO) of the electron-rich benzene is at a high-lying π* antibonding state with a higher energy than CB of 1b, resulting in the transfer of excited electrons from its LUMO to the CB and triggering the fluorescence "turn-on" of 1b⊃benzene (Fig. 3c). More importantly, the single-crystal structure of 1b⊃benzene clearly reveals that there are multiple interactions between benzene and 1b. Such supramolecular interactions are the main origin of the fluorescence enhancement. Each cavity can only accommodate one benzene molecule, being sandwiched by a pair of neighboring NDI macrocycles (Fig. 3d). Each benzene molecule simultaneously interacts in an offset face-to-face fashion with two adjacent NDI macrocycles in a special "bilateral π-π stacking" interactions, which is stronger than the interpenetrated frameworks (such as [Zn2(BDC)2(DPNDI)]) having "unilateral π-π stacking" interaction between the guest molecule and host framework. The distances from the benzene center to the NDI planes are 3.2 Å and 3.5 Å, respectively, and both the dihedral angles are 8.4° (Figs. 3d and S17). In addition, there exist C-H‧‧‧π interactions between the benzene molecule and NDI macrocycle with the distances at 3.0 and 3.2 Å, respectively (Fig. 3d). Considering that other aromatic hydrocarbons possess physicochemical properties similar to benzene as electron-rich molecules, hence the mechanism of fluorescence enhancement is similar. Single-crystal structures of 1b⊃VOCs verify the presence of the similar mechanism, indicating that toluene, o-xylene, m-xylene and p-xylene are able to diffuse into the pores to form the "bilateral π-π stacking" interactions with the framework (Figures S17-S19), and TDDFT calculations show that the LUMO of other aromatic hydrocarbons all have higher energies compared to CB of 1b (Fig. 3c). To better illustrate the interaction between the host and VOCs, the binding energies of the VOCs to the host framework were estimated by DFT calculations on the basis of the single-crystal structures of 1b⊃VOCs. Notably, the binding energies of benzene, toluene, o-xylene, m-xylene and p-xylene reach 161 KJ mol-1, 186 KJ mol-1, 211 KJ mol-1, 216 KJ mol-1 and 209 KJ mol-1, respectively, indicating the strong host-guest interaction. Being consistent with the strong interactions revealed by the single-crystal structure of 1b⊃benzene, the strong host-guest binding energy facilitate electrons transfer from benzene to the host framework by virtue of the formation of an ordered donor-acceptor-donor-acceptor arrangement, resulting in an extremely significant fluorescence enhancement.
Since indoor benzene vapor concentration is usually very low, existing methods (gas chromatography) for detecting trace benzene vapor are time-consuming, expensive and involving complex instruments. The highly efficient fluorescence enhancement of 1b⊃benzene makes it promising for rapid and sensitive detection of trace benzene vapor. Fluorescence spectra of 1b were measured after exposing to benzene vapor at various concentrations in a closed container for 150 s. The result shows that the fluorescence intensities of 1b exhibit a good linear relationship with the concentrations of benzene vapor (Fig. 2f). Hence, the fluorescence enhancement coefficient (Kec) can be calculated by using the fluorescence enhancement equation I/I0 = Kec[M] + 1, where I0 is the initial fluorescent intensity of 1b, I is the fluorescent intensity of the 1b exposed to benzene vapor, [M] is the concentration of benzene vapor. In addition, the limit of detection concentration (LOD) for benzene vapor can be calculated as 77 ppb based on the equation of LOD = 3σ/Kec, where σ is the standard deviation of the blank measurements (Table S2). This LOD represents the lowest record compared with all reported MOFs for detection of benzene vapor.
Although a few MOFs can be used for benzene vapor detection by fluorescence response, no report for the detection of trace benzene vapor was documented by far. Therefore, the reported MOFs of [Zn2(BDC)2(DPNDI)], [Cd3(COOA)(BIPY)2] and [Zn2(OBA)2(BPY)] (H2BDC = 1,4-benzenedicarboxylate, H2DPNDI = N,N′-di(4-pyridyl)-1,4,5,8-naphthalenediimide, H6COOA = hexa(4-(carboxy-phenyl)oxamethyl)-3-oxapentane acid, BIPY = 2,2′-bipyridine, H2OBA = 4,4′-oxybis(benzoic acid), BPY = 4,4′-bipyridine), which have been demonstrated the fluorescence enhancing behaviors to benzene,30, 34–35 were selected to measure the detection performances for comparison with 1b. The results show that, the performances of fluorescence enhancements are poor after exposing to saturated benzene vapor for 24 h, being only 3 fold for [Zn2(BDC)2(DPNDI)], 4 fold for [Cd3(COOA)(BIPY)2] and 2 fold for [Zn2(OBA)2(BPY)] (Fig. 4a), which are far below the 116 fold of 1b. In other words, 1b possesses much significant fluorescence enhancement. Furthermore, the known MOFs showed negligible fluorescence change after exposure to trace benzene vapor (40 ppm), and only 1.4 fold for [Zn2(OBA)2(BPY)], 0.8 fold for [Zn2(BDC)2(DPNDI)] and 2.2 fold for [Cd3(COOA)(BIPY)2] (Fig. 4a), which are inconspicuous to be visualized by the naked eyes. Obviously, 1b is much superior in the detection of trace benzene vapor. In addition, it is worthy of note that [Zn2(BDC)2(DPNDI)] possesses the same NDI cores as 1b, but its detection performance is significantly lower, owing to the weaker "unilateral π-π stacking" interactions between its framework and guest benzene molecules than the stronger "bilateral π-π stacking" interactions in 1b.
Also notably, 1 can be easily synthesized in gram scale by heating reflux in a flask (Fig. 4b, c), which makes it possible for practical application. Considering the feasibility of practical applications as well as portability, its fluorescence rapid test papers were prepared. A suspension was obtained by immersing the ground fine powder of 1b in ethanol for 30 minutes, then dropped on the filter paper and dried at 60 oC for 6 h. The PXRD pattern indicates that the framework of 1b retains after the ultrasonic treatment with ethanol solution (Figure S7). SEM confirmed that the sample has been loaded on the paper (Figure S23). After the prepared test paper is exposed to a trace (3 ppm) of benzene vapor for 150 s, the significant luminescence response under UV (365 nm) light can be visualized by the naked eyes, suggesting that 1b possesses a wide range of practical applications in detection of trace benzene vapor (Fig. 4d). To our best knowledge, this is the first MOF-loaded test strips for detection of trace benzene vapor, which presents promising practical application for simplicity, ease and speed of detection.
