Customization of functional MOFs by a modular design strategy for target applications

Customization of FM-MOFs by single substitution

A schematic illustration of the construction of FM-MOFs is shown in Scheme 1. Theoretically, in the assembly process, various desired functional modules could be imparted into the host framework to customize FM-MOFs according to arbitrary target applications. Thus, it is imperative to ensure that the various functional molecules can be modularized as much as possible. Considering that labile formate groups on MOF-808 are easy to be substituted, a carboxyl group can serve as a connector to impart functional molecules into MOF-808. To specify this concept, several applications including Hg²⁺ capture, Ag⁺ recovery, proton conductivity, detection of nitro explosives, and the control of surface wettability were taken as examples. We subsequently used thioglycollic acid (M1, M refers to module), propiolic acid (M2), oxalic acid (M3), pyrenecarboxylic acid (M4), and perfluorooctanoic acid (M5) to assemble series of FM-MOFs based on MOF-808 according to respective target applications [Figure 1 and Supplementary Figure 1]. Supplementary Figure 2 shows the PXRD patterns of MOF-808 and its functional derivatives. It is obvious that the related peaks of these assembled FM-MOFs are in good agreement with that of pristine MOF-808, confirming that the crystal structure of these materials remains intact after functional modules insertion. N₂ adsorption-desorption isotherms of FM-MOFs measured at 77 K display a typical I isotherm, suggesting the reservation of microporous features after introducing functional units into the framework, despite a decrease in their BET surface areas [Supplementary Figure 3]. The scanning electron microscopy (SEM) images demonstrate that our FM-MOFs possess octahedral morphology, with no apparent morphology change observed compared with MOF-808 [Supplementary Figure 4]. Individually, given that the thiolate ligand possesses a strong coordination ability to Hg²⁺, thioglycollic acid (M1) was used as a functional modular to construct FM-MOF-1 for Hg²⁺ capture. The successful insertion of M1 into MOF-808 was also proved by Fourier transform infrared (FT-IR) spectroscopy, ¹H nuclear magnetic resonance (NMR) spectroscopy, and elemental analysis. The FT-IR spectra of FM-MOF-1 reveal that there is a new peak at 2564 cm⁻¹ [Supplementary Figure 5], which can be attributed to the S-H stretching of free -SH groups. Supplementary Figure 6 shows the ¹H NMR spectra of FM-MOF-1 and MOF-808 after dissolving the samples in KOH/D₂O solution. Obviously, the peak intensity at δ = 8.3 ppm corresponding to the hydrogen of formate of FM-MOF-1 significantly decreased and a new peak at δ = 2.8 ppm for the hydrogen signal of -CH₂- in M1 appeared, indicating that formate ligands on Zr₆ clusters were substituted by M1 molecules. All these observations demonstrate that M1 was successfully inserted into the framework of MOF-808. To evaluate the effectiveness of FM-MOF-1 for Hg²⁺ capture, adsorption kinetics and isotherms were measured. As shown in Figure 1B and Supplementary Figure 7, an extremely quick process was observed for FM-MOF-1. Over 99% of Hg²⁺ was removed within 5 min. Supplementary Figure 8 shows the adsorption isotherms of FM-MOF-1 towards Hg²⁺. By fitting with Langmuir model, the saturated Hg²⁺ removal capacity of FM-MOF-1 was estimated to be 758 mg g⁻¹, suggesting the great potential of FM-MOF-1 for Hg²⁺ capture. Similarly, considering the strong affinity of the terminal alkynyl group toward Ag⁺, propiolic acid (M2) was selected as another functional module to assemble FM-MOF-2 for silver recovery. FT-IR and ¹H NMR spectra were used to demonstrate the successful insertion of M2 into the framework of MOF-808 [Supplementary Figures 9 and 10]. Supplementary Figure 11 shows the adsorption kinetics of FM-MOF-2 towards Ag⁺; the removal efficiency increased sharply during the initial stage and then approached equilibrium after 2 h of contact. The adsorption isotherm revealed that FM-MOF-2 exhibits high silver uptake [Figure 1C]. By fitting with Langmuir model, the saturated Ag⁺ uptake capacity was calculated to be 806 mg g⁻¹, surpassing most reported porous materials [Supplementary Table 1]^[25-34]. This is also the first study of introducing alkynyl groups into the MOF framework for Ag⁺ recovery. Besides, to construct the proton-conducting MOF, we used oxalic acid as the functional module to assemble FM-MOF-3. The detailed characterizations are provided in the Supplementary Materials [Supplementary Figures 12-14]. Proton conductivities were then checked on a pressed plate at varying relative humidity (RH) from 33% to 100% at 298 K for FM-MOF-3. As shown in Supplementary Figure 15, all of the Nyquist plots display circular arcs at high frequencies. The proton conductivity of FM-MOF-3 increased with the elevated moisture [Figure 1D], suggesting water molecules play a vital role in the proton transportation of this MOF. Meanwhile, FM-MOF-3 exhibited a striking proton conductivity of 1.2 10^-2 S cm^-1 at 298 K and 100% RH, whereas the original MOF-808 displayed an inferior proton conductivity of 2.2 10^-6 S cm^-1 under the same condition [Supplementary Figure 16]. It should be noted that only a few MOFs featuring high proton conductivity over 10^-2 S cm^-1 have been reported^[35-37]. Moreover, to achieve the goals of detecting nitro explosives and controlling surface wettability, conjugated pyrenecarboxylic acid (M4) and perfluorooctanoic acid (M5) were utilized as modules to construct FM-MOF-4 and FM-MOF-5, respectively. The related characterizations can be found in the Supplementary Materials [Supplementary Figures 17-20]. To investigate the explosive sensing ability of FM-MOF-4, fluorescence quenching titrations were conducted by the increasing amount of nitro aromatics. As shown in Figure 1E, fast and significant fluorescence quenching was observed with an increase in the amount of TNP (2,4,6-trinitrophenol). The fluorescence quenching could be discerned at a low concentration (0.5 µM; Supplementary Figure 21) and reached nearly 83% when the concentration of TNP increases to 1 mM. By contrast, other nitro compounds have little effect on the fluorescence intensity [Supplementary Figure 22]. These results demonstrate that FM-MOF-4 could selectively and sensitively detect TNP over other nitro explosives. On the other hand, the surface wettability of FM-MOF-5 and MOF-808 was also examined by contact angle measurements. As shown in Figure 1F, the contact angle of water on MOF-808 was estimated to be 37°, whereas FM-MOF-5 gave the water contact angle at as high as 142°. The surface character of MOF-808 transformed from hydrophilic to hydrophobic after M5 insertion into the framework, confirming the coating of perfluoroalkyl groups on the surface of material can dramatically enhance its hydrophobicity. This hydrophobic behavior of FM-MOF-5 was also illustrated by water vapor adsorption experiment: the water uptakes of MOF-808 reduced significantly after the incorporation of M5 [Supplementary Figure 23].

Scheme 1. Illustration of customization of FM-MOFs toward different applications.

Figure 1. Customization of FM-MOFs through single substitution. (A) Schematic illustration of constructing FM-MOFs by functional modules with carboxyl connector. (B) Hg²⁺ adsorption kinetics of FM-MOF-1 at the initial concentration of 10 ppm. (C) Ag⁺ adsorption isotherm for FM-MOF-2. The inset shows the linear regression by fitting the experimental data with the Langmuir model. (D) Humidity dependence of the proton conductivities in FM-MOF-3 at room temperature. (E) The effect on the emission spectra of FM-MOF-4 dispersed in DMF upon incremental addition of a TNP solution (1 mM). The inset shows the original fluorescence and the decreased fluorescence upon the addition of TNP solution (1 mM). (F) Contact angle images of FM-MOF-5 and MOF-808.

Customization of FM-MOFs through tandem postsynthetic modification

The above proof-of-concept experiments revealed that functional molecules with carboxyl groups can be used as modules to insert into MOF-808 directly according to numerous target applications. However, to expand the scope of our strategy, some useful functional molecules without carboxyl groups should also be modularized. To solve this problem, we installed a pre-designed convertor into MOF-808 before functional modules insertion through a tandem postsynthetic modification [Figure 2]. Such convertors should have two terminals that can combine modules and the host framework, respectively. Here, the aforementioned Hg²⁺ capture and CO₂ conversion were taken as target applications to fully illustrate this. Firstly, we tried to use thioglycollic acid (M1) to construct FM-MOF-1 for mercury adsorption. Considering a functional unit with more chelating sites is needed to achieve higher Hg²⁺ uptake capacity, 1,2-ethanedithiol (M6) containing higher sulfur content yet without carboxyl group compared with M1 was adopted. To impart M6 on MOF-808, we chose propenoic acid (M7) as a convertor and inserted it into the host framework firstly to construct FM-MOF-7, where the accessible vinyl groups could allow for further chemical modifications. As shown in Supplementary Figures 24-28, FM-MOF-7 was characterized by PXRD, N₂ adsorption, SEM, FT-IR, and ¹H NMR. FT-IR spectra reveal the appearance of a new characteristic peak at 1639 cm^-1 compared with MOF-808, which can be attributed to the C=C stretching band of vinyl groups [Supplementary Figure 27]. ¹H NMR spectra demonstrate that the peak at 8.3 ppm assigned to the formate groups disappeared and several strong peaks around 5.0-7.0 ppm corresponding to the vinyl groups emerged, indicating the successful insertion of M7 into MOF-808 [Supplementary Figure 28]. FM-MOF-7 was then treated with 1,2-ethanedithiol (M6) to obtain FM-MOF-6 through the thiol-ene “click” reaction between thiol compounds and the vinyl groups in FM-MOF-7. The resulting material was also carefully characterized by PXRD, N₂ adsorption, FT-IR, SEM, and NMR [Supplementary Figures 29-33]. Compared with FM-MOF-7, FT-IR spectra of FM-MOF-6 show a new peak at 2555 cm^-1 despite the absence of the band at 1639 cm^-1, indicating the elimination of vinyl groups and the appearance of free-standing thiol group after modification [Supplementary Figure 32]. The thio-ene transformation was also confirmed by the disappearance of the peaks around 5.0-7.0 ppm corresponding to the hydrogen signals of vinyl groups in the ¹H NMR spectrum of FM-MOF-6 and the concomitant emergence of new peaks around 2.0 and 3.0 ppm attributed to the different hydrogen of -CH₂- in M6 [Supplementary Figure 33]. To assess the overall capacity of FM-MOF-6 towards Hg²⁺, the adsorption isotherm was then measured. As shown in Figure 2C, FM-MOF-6 displayed extremely high Hg²⁺ uptake. By fitting with Langmuir model, the saturated adsorption capacity of FM-MOF-6 for Hg²⁺ was determined to be 1077 mg g^-1, which is about 1.4 times higher than that of FM-MOF-1, surpassing various MOF adsorbents Supplementary Table 2^[38-48]. In addition, we also investigated the adsorption kinetics of FM-MOF-6 toward Hg²⁺. As shown in Supplementary Figure 34, a quick purification process was observed. The removal efficiency displayed a steep profile at the initial stage and then tended to equilibrium with about 99.84% of Hg²⁺ ions being removed. The excellent performance of FM-MOF-6 in mercury capture can be traced to the strong binding interactions between the Hg and sulfur species in FM-MOF-6, as evidenced by the enlarged S 2p binding energy of XPS analysis and the absence of S-H stretching mode in FT-IR spectra after Hg²⁺ adsorption [Supplementary Figures 35-37]. These results demonstrate the superiority of the utilization of FM-MOF-6 as a promising candidate for mercury capture from aqueous solutions.

Figure 2. Customization of FM-MOFs through tandem postsynthetic modification. (A and B) Scheme illustration of FM-MOFs and their pore structures. (C) Hg²⁺ adsorption isotherm of FM-MOF-6. The inset shows the linear regression by fitting the experimental data with the Langmuir model. (D) Yields of various cyclic carbonates prepared from the cycloaddition of CO₂ with related epoxides catalyzed by FM-MOF-8 (red) and FM-MOF-9 (blue).

On the other hand, the cycloaddition of CO₂ with epoxides into cyclic carbonates is an attractive strategy for addressing anthropogenic CO₂ emission issues, and imidazole-based functional molecules are regarded as efficient catalysts for this purpose. Considering the issues of homogeneous catalysts in product purification and catalyst recycling, it is advisable to graft imidazole-based functional molecules onto host materials for use as heterogeneous catalysts. Here, we chose 1-allylimidazole (M8) as a functional module to construct FM-MOF-8. Because of the absence of carboxylic connector in M8, a suitable convertor was needed. Therefore, bromoacetic acid (M9), which contains both a carboxylic connector and an accessible bromomethyl group for further chemical modification, was selected to assemble FM-MOF-9 first. The detailed characterizations are provided in Supplementary Figures 38-42. FM-MOF-9 was then treated with 1-allylimidazole (M8) through a tandem postsynthetic modification to obtain FM-MOF-8, as characterized by PXRD, N₂ adsorption, SEM, FT-IR, and NMR [Supplementary Figures 43-47]. Interestingly, the two incorporated modules of M8 and M9 in FM-MOF-8 combined together to form an imidazolium-based ionic liquid (IL) unit. Because of the proper distribution of formate groups in MOF-808, theoretically, such IL unit in FM-MOF-8 would also disperse uniformly within the pore, thus avoiding agglomeration and providing more accessible catalytic sites compared to the conventional dipping method. To evaluate the effectiveness of FM-MOF-9 as a catalyst, experiments were carried out. As shown in Figure 2D, the reaction yields catalyzed by FM-MOF-8 from related epoxides were determined to be 97.1% for epichlorohydrin, 95.2% for epibromohydrin, 86.2% for 1,2-epoxybutane, and 84.1% for allyl glycidyl ether, indicating that FM-MOF-8 exhibited a highly efficient catalytic performance for a variety of epoxides. It should be noted that the good performance of FM-MOF-8 was achieved without using any co-catalysts, whereas most reported MOF catalysts employ other substances, such as TBAB, to improve activity, which is considered uneconomical and complicated^[49]. In contrast, FM-MOF-9 showed an extremely low activity, with a yield of < 10% under similar conditions, manifesting that the grafted M8 played a key role during the catalytic process. The above two proof-of-concept experiments fully demonstrated the reliability of this tandem postsynthetic modification for inserting functional modules without a carboxyl connector into MOF-808, which greatly expands the scope of FM-MOFs for functional targeting customization.

Switching modules on FM-MOFs

Another important aspect to consider is whether the functional modules on FM-MOFs could be freely switched in a simple manner, which allows for changing user-defined units in accordance with different needs, thereby affording matrix recycling and cost saving. The mutual transformation among FM-MOF-2, FM-MOF-7, and FM-MOF-9 was taken as a representative sample to confirm this [Figure 3]. These FM-MOFs were added into the solution containing the corresponding functional module, and the detailed experiments can be found in the Supplementary Materials. This conversion process was repeated several times, and the as-transformed FM-MOFs were determined by ¹H NMR analysis. As shown in Supplementary Figures 48-53, for all three FM-MOFs, the signals of the original functional module disappeared, whereas the peaks attributed to the new unit emerged, demonstrating full switching of functional modules on MOF-808. Besides, PXRD measurements indicate no apparent loss of crystallinity during the conversion process [Supplementary Figures 54-59]. To further illustrate this, we also symmetrically investigated the conversion process from FM-MOF-9 to FM-MOF-7 with the increase of time. As shown in Figure 3C and D, the conservation rate reached 65% in the initial 1 h and steadily increased over time. After 24 h of reaction, the ¹H NMR spectrum reveals that δ at 3.8 ppm corresponding to the hydrogen signals of -CH₂- in M9 significantly decreased, whereas several new peaks around 5-7 ppm attributed to the vinyl group in M7 enhanced dramatically, achieving a high conservation of 95%. The peak of M9 further reduced when we repeated this conservation process several times, demonstrating essentially full conversion from FM-MOF-9 to FM-MOF-7.

Figure 3. Full conversion process among FM-MOFs: (A and B) scheme illustration of mutual transformation among FM-MOFs and the associated pore structures of FM-MOF-2, FM-MOF-7, and FM-MOF-9; (C) the conversion process characterized by ¹H NMR spectra; and (D) the conversion rate from FM-MOF-9 to FM-MOF-7 with the increase of time.

Interestingly, we found that two functional modules could coexist in one host matrix during the switching process, which would form multivariate MOFs (MTV-MOFs) via partial conversion [Figure 4]. For full illustration, Hg²⁺ detection was selected as target application. We combined a mercury capture module and fluorophore moiety into MOF-808 through partial conservation and supposed that the resulting MTV-MOF would show improved performance in mercury detection. As such, the aforementioned FM-MOF-1 was immersed in DMF solution containing M4 for a period of time to obtain FM-MOF-1-4 (see Supplementary Figures 60-62). Compared to FM-MOF-1, the ¹H NMR spectra of this MTV-MOF reveal that δ at 2.8 ppm corresponding to M1 decreased and new peaks assigned to the hydrogen signals of pyrene in M4 emerged [Figure 4C]. The relevant signals of M1 in FM-MOF-1 and FM-MOF-1-4 were also integrated against that of the BTC ligand, resulting in peak ratios of 6:7 and 6:2, respectively, demonstrating that 71% of M1 was substituted by M4 during module switching process [Supplementary Figures 63 and 64]. FM-MOF-1-4 was then used to sense a trace quantity of Hg²⁺ in aqueous solutions. As shown in Figure 4D, obvious fluorescence quenching was observed as the concentration of Hg²⁺ increased from 10 to 100 µM, specially, the quenching effect of Hg²⁺ was quantified by the Stern-Volmer equation (I₀/I = 1 + K_sv[M]), where I₀ and I are emission intensity of MOF material before and after reaction with Hg²⁺, K_sv is the Stern-Volmer quenching effect constant of Hg²⁺, and [M] is the concentration of Hg²⁺. As shown in Figure 4E, FM-MOF-1-4 showed a distinguishable K_sv value for Hg²⁺ at 6260 M⁻¹, which is 2.6 times higher than that of FM-MOF-4 at 2410 M^-1. It is worth noting that FM-MOF-1 exhibited very weak fluorescence, and no visible quenching effect was observed towards Hg²⁺. With M4 anchored in the pores, the thiol group within MOF and the fluorophore part of pyrene can cooperatively interact to significantly improve sensing performance. Specifically, the thiol groups of M1 in MOF captured traces of Hg²⁺, and then the fluorophore pyrene of M4, which is in close proximity, responded to mercury ions. This process allows two functional modules to work in a cooperative manner. The above representative sample, therefore, demonstrates the feasibility of constructing MTV-MOF via partial conversion, which may further enhance the performance of MOFs. In principle, two arbitrary functional modules can exist in one host MOFs, thus constructing numerous of MTV-MOFs for target applications.

Figure 4. Partial conversion process among FM-MOFs: (A) scheme illustration of constructing MTV-MOF through a partial conversion process; (B) constructing FM-MOF-1-4 from FM-MOF-1; (C) ¹H NMR spectra of (A) FM-MOF-1-4, (B) FM-MOF-1, and (C) FM-MOF-4 in D₂O/KOH solution; (D) effect on the emission spectra of FM-MOF-1-4 with different concentration of Hg²⁺ in aqueous solution; and (E) Stern-Volmer plots of Hg²⁺ in aqueous solution for FM-MOF-1-4 (red) and FM-MOF-4 (blue).