Design and synthesisCages 1 and 2 were prepared by reacting TAPM or TABPM with 2-formylpyridine and iron(II) trifluoromethanesulfonate (Fe(OTf)2) in a molar ratio of 4:12:4 in acetonitrile at 70 oC (Fig. 1). Single crystals of 1·(OTf)8 were obtained by diffusing the 1,2-dichlorobenzene into acetonitrile solution. The cages were characterized using a variety of techniques including single-crystal X-ray diffraction, nuclear magnetic resonance (NMR), and electrospray ionization mass spectrometry (ESI-MS).Fig. 1: Schematic diagram.a Self-assembly of the metal-organic cages and covalent silane modification of cages 1 and 2. b Heterogenization of Si-1 and Si-2 in microfluidic reactor.1H and 13C NMR spectra of each cage displayed a singular set of ligand resonances in solution, indicating the formation of a single species by the reaction of each subcomponent in CH3CN (Fig. 2a, c, Supplementary Figs. 1–7). In the 1H NMR spectra of 1 and 2, the exposed amino residues were identified at 4.17 and 4.35 ppm (Supplementary Figs. 1, 5), respectively. This observation underscores the feasibility of producing the desired product by manipulating the stoichiometry of the subcomponents. Additionally, the 1H diffusion-ordered NMR spectroscopy (DOSY) of 1 and 2 displayed a single set of signals (Fig. 2a, c), providing further evidence for the formation of a single species. The measured diffusion coefficients for 1 and 2 were D = 7.94 × 10−10 and 5.01 × 10−10 m2 s−1, respectively, corresponding to hydrodynamic radius of ~7.3 and 11.6 Å. The formation of the cages was also supported by high-resolution ESI-MS. Three distinct peaks at m/z = 852.6732, 1186.1150 and 1854.3355 were observed for 1, which can be assigned to [1 – 4OTf-]4+, [1 – 3OTf-]3+ and [1 – 2OTf-]2+, respectively (Fig. 2b, Supplementary Fig. 4). 2 displayed a family of prominent signals at m/z = 503.9227, 597.1899, 721.7149, 895.8496, 1156.7988 and 1592.0614, corresponding to [2 – 8OTf-]8+, [2 – 7OTf-]7+, [2 – 6OTf-]6+, [2 – 5OTf-]5+, [2 – 4OTf-]4+and [2 – 3OTf-]3+, respectively (Fig. 2d, Supplementary Fig. 8). Moreover, all these peaks agree well with the simulated and the natural isotopic abundances.Fig. 2: NMR and ESI-MS spectra of cages 1 and 2.a, c Partial 1H NMR and 2D DOSY spectra (500 MHz, CD3CN, 298 K) of 1 and 2. b, d Experimental ESI-MS and calculated isotope patterns of 1 and 2 (Observed: Obs.; Simulated: Sim.).Single-crystal X-ray diffraction revealed that 1 crystallizes in the tetragonal P422 space group. As shown in Fig. 3a, the Fe4L4 type of cage 1 exhibits a face-capped tetrahedral structure with approximate T-symmetry. Four tridentate pyridylimine ligands bridge four facially coordinated Fe∐ centers to form a cation cage with the formula of [Fe4L4]·(OTf)8 and the positive charges are balanced by eight OTf-. The Fe-N bond length is ~1.96 Å (Supplementary Fig. 9), which is in good agreement with the typical Fe-pyridylimine complexes. The adjacent Fe···Fe separations are 11.92 Å, with an inner pore diameter of 6.0 Å. The volume of the central cavity was calculated to be 73.5 Å3 (Supplementary Fig. 10a). The three benzene rings in TAPM maintain a triangle cone configuration, giving rise to a pseudo hexahedral cage structure with a quadrilateral window (7.14 × 3.07 Å2, Fig. 3b). Notably, the solid-state structure of cage 1 also reveals that one aniline is retained in each TAPM backbone, oriented outward, the distance between two amino groups in adjacent cages 1 was measured to be 5.2 Å (Supplementary Fig. 11b). This arrangement preserves four amino groups on the cage surface, offering the potential for post-modification.Fig. 3: X-ray crystallographic structure of cage.a, b The single-crystal X-ray structures of 1 from two different directions. c, d The energy-minimized conformations of 2 from two different directions. Gray: C; white: H; dodger blue: N; purple: Fe. The cavities are highlighted by yellow balls.Replacement of constitutive ligand TAPM with the larger ligand TABPM has led to the formation of cage 2. Numerous attempts have been made to grow single crystals of cage 2 appropriate for X-ray diffraction, but none have been successful. As an alternative, a molecular model of 2 was generated through DFT calculations. Given the structural similarity between TAPM and TABPM, the energy-minimized model suggests that 2 also maintains the Fe4L4 structure with T-symmetry, and four exposed amino residues pointing outward (Fig. 3c, d). Cage 2 features six rhombus-shaped opening windows, each measuring 7.32 × 15.09 Å2 available for guest transports. The volume of the cavity was calculated to be 522.1 Å3 (Supplementary Fig. 10b).Covalent post-assembly silane modifications (PASM) of cages 1 and 2Having established the general method of preparing amine-bearing cages, we focused our investigations on the post-synthetic modifications of cages 1 and 2. Silane modification is a potent approach for enhancing the affinity between inorganic and organic compounds and is commonly employed in the surface modification of microchannels and SiO260,61. Additionally, it has been utilized for the heterogenization of homogeneous catalysts, facilitating their recycling and product separation while enhancing stability62. Thus, covalent post-assembly silane modifications (PASM) of 1 and 2 were performed by treating cages with IPTS in a ratio of 1:20 in acetonitrile at 80 oC. The crude products were obtained by adding diethyl ether to the reaction mixture.A series of NMR spectra of Si-1 and Si-2 supported the successful modification of the pristine cages, with the tetrahedral core remaining intact. For instance, in the 1H NMR spectra of Si-1, the peak area at 4.17 ppm decreased compared to that of cage 1 (Fig. 4a), indicating partial conversion of the amine into urea functionality. By carefully analyzing the peak area of Si-1, we found that two of the four amine groups were grafted with silane chains (Supplementary Fig. 12). Multiplets at 3.78, 3.07, 1.50 and 0.55 ppm were observed, which can be assigned to the alkyl group of IPTS. In order to enable the full conversion of the amine groups into urea functionality, several factors including the amounts of IPTS, reaction temperature and the kinds of catalysts were considered and regulated (Supplementary Table 3). However, in most cases, the predominant product remained Si-1. We inferred that this may be due to the comparatively poor nucleophilicity of the aniline residues of cage 1. Fortunately, the 1H NMR spectra of Si-2 showed the amino group at 4.30 was replaced with new peaks at 7.64 and 5.39 ppm, demonstrating the approximate complete conversion of the amine into urea functionality (Fig. 4c, Supplementary Fig. 18). The 13C NMR spectra of Si-1 and Si-2 also revealed the single set of signals, supporting the existence of only one species in each solution (Supplementary Figs. 13, 19). Molecular modeling calculations suggested that the diameters of the two post-modified cages were approximately 4.1 and 5.1 nm, respectively (Fig. 4b, d).Fig. 4: NMR spectra and simulated structures of Si-1 and Si-2.a, c 1H NMR spectra of cages 1, 2, Si-1 and Si-2 (500 MHz, CD3CN, 298 K). b, d The energy-minimized conformations of Si-1 and Si-2. Gray: C; white: H; dodger blue: N; red: O; tea green: Si; purple: Fe. The cavities are highlighted by yellow balls.ESI-MS data provided further evidence of the tetrahedral structures of Si-1 and Si-2. As shown in Supplementary Figs. 15–17, the peaks corresponding to 1 disappeared from the mass spectrum of Si-1 and were replaced by a family of peaks at m/z = 751.1965 – 5.6n (n = 0–12) and 976.2348 – 7n (n = 0–12), corresponding to the fragments of [M – nC2H5+ + nH+- 5OTf-]5+ and [M – nC2H5+ + nH+- 4OTf-]4+, respectively. It is noteworthy that the continuous loss of fragments of C2H5+ could only originate from silane chains. Additionally, the peaks assigned to [M – nC2H5+ + nH+- 7OTf-]7+, [M – nC2H5+ – CH3+ + (n + 1)H+- 7OTf-]7+, [M – nC2H5+ + nH+- 6OTf-]6+ and [M – nC2H5+ – CH3+ + (n + 1)H+- 6OTf-]6+ (n = 0–12) of Si-2 were isotopically resolved and in good agreement with its calculated theoretical distribution (Supplementary Figs. 21–23). These results convincingly corroborate that the silane chains can be introduced into an amine-bearing scaffold during the covalent post-assembly modification of supramolecular complexes.Immobilization of Si-1 and Si-2 in the microreactorMicroreactor, characterized by its confined microchannel size, has demonstrated its potential to enhance mass and heat transfer63,64,65,66,67,68,69,70,71. As shown in Fig. 5a, we engineered and assembled an automated microplatform, named as the catalyst loading and substrate conversion (CLSC) system. This synthesis system primarily consists of two components: a PDMS microreactor (outfitted with heating plates) and an injection system (comprising multi-position selection valves, tubing, mixing column and a peristaltic pump). The development of the CLSC system is connected with the strategic design of a six-way valve, capable of regulating the different working modules sequentially. These valves play a pivotal role in automating the processes of cage loading and catalysis. More importantly, the CLSC system is a recirculation system that extends the effective reaction time, overcoming the limitations of short residence times.Fig. 5: Characterizations of the modified microreactor.a Schematic of the CLSC model display. b Changes in microchannels after loading Si-2. c SEM images and EDS mappings of Si-1 and Si-2@PDMS. d F 1 s and N 1 s XPS spectra of Si-1 and Si-2@PDMS.Before loading the cages into the microchannel, IPTS was selected as the model compound to assess the feasibility of surface modification in microreactors. Pure IPTS was introduced into the active PDMS channel and held at 80 °C for 0.5 h, followed by a quick cleaning step using ethyl acetate as a washing solvent. Scanning electron microscopy (SEM) images revealed that the PDMS surface developed numerous nanospheres post-modification with IPTS (Supplementary Fig. 28). Energy dispersive spectroscopy (EDS) elemental mapping confirmed the uniform distribution of nitrogen across the PDMS microchannel surface (Supplementary Fig. 28). X-ray photoelectron spectroscopy (XPS) was employed to analyze surface components of PDMS after modification. Compared to pristine PDMS, a new N 1 s peak emerged at 400.1 eV in IPTS@PDMS (Supplementary Fig. 29), confirming the successful immobilization of IPTS on the PDMS microchannel surface.During the cage loading process, a CH3CN solution containing catalysts is transported into the activated PDMS microreactor by opening valves A and D. The solution is then retained in the microchannel at 80 oC for 1 h, ensuring the successful cage loading into the channel. This is followed by a cleaning step, achieved by opening valve B while closing valve A, to remove free cages. This process is automated and repeated thrice to enhance the loading efficiency of cages. Subsequently, valve C was opened to drive the substrate solution into a microchannel preloaded with cages, facilitating the conversion. The final step in the operation of the CLSC system is the cleaning of the reaction system. This is achieved by opening valve B, preparing the system for the catalytic transformation of the next substrate. The advanced synthesis system also has great potential for upgrading, for example, the above operations can be accomplished by sending commands from the computer to the electronic valves. The automation of the CLSC system extends beyond these processes. The system is designed to adapt to scenarios by modifying the modules, catalysts and reactants. This adaptability is facilitated by the system’s automation, allowing for quick and efficient changes to the system’s configuration.The successful immobilization of Si-1 and Si-2 in the microchannel was confirmed by SEM and XPS analysis. As shown in Fig. 5b, the color of the microchannel has changed from colorless to purple after loading two cages. The SEM images in Supplementary Figs. 30 and 31 showed the evolution of the surface topography of the PDMS microchannel with different processes. The SEM cross-sectional images indicated cages are well integrated and densely accumulated on the microchannel surface (Fig. 5c). Moreover, the EDS elemental mapping was also used to investigate the spatial distribution of the constituent elements of the modified microchannel, which confirmed the Si-1 and Si-2 graft uniformly on the surface of microchannel (Fig. 5c). XPS was employed to analyze the surface components of the PDMS microchannel during the modification process. Compared with the pristine PDMS, the new N 1 s (at 400.1 eV) and F 1 s (at 688.2 eV) peaks were found in Si-1@PDMS and Si-2@PDMS (Fig. 5d). Additionally, the weak peaks that appeared at 721.7 and 708.9 eV can be assigned to Fe 2p1/2 and Fe 2p3/2, respectively, indicating the Fe species is in +2 oxidation states (Supplementary Fig. 32). The above results indicated that Si-1 and Si-2 were successfully immobilized on the surface of PDMS microchannel.CatalysisThe 2,3-dihydroquinazolinone (DHQZ) has a wide range of biological activities and is often used for anticancer, antibacterial, and antiviral treatments, demonstrating unique pharmacological activities72. Thus, several catalysts have been explored to synthesize DHZQ, such as chiral Brønsted acids73 (BINOL- and SPINOL-phosphoric acids), Lewis acids74 (Sc(III)-Inda-pybox) and porous materials75. More important, the metal-organic cages with confined space have been designed and employed to promote the synthesis of DHQZ76. The present tetrahedral cages featuring rich π-electron density and hydrophobic cavities may accommodate guests, which is beneficial for chemical transformation and catalysis in their cavities. Therefore, by utilizing their distinctive and flexible cavities with tunable sizes, the two cages were used for supramolecular catalysis.Before assessing the catalytic efficiency of cages in microreactor, we conducted a preliminary evaluation of their activity in batch. To investigate the catalytic activities of cages 1 and 2, benzaldehyde (3a) and anthranilamide (4a) were chosen as substrates to screen the most efficient conditions. We systematically screened various reaction parameters, including solvent, reaction temperature, and catalyst loading, with detailed results presented in Supplementary Table 4. Notably, when 0.1 mol% of cage 2 was present, the reaction between benzaldehyde and anthranilamide proceeded smoothly in CH3CN at 25 °C. And the addition of desiccants such as MgSO4 or Na2SO4 can improve the yield of the products (Supplementary Table 4). Subsequently, under the optimized conditions, a diverse range of substrates could be efficiently converted to the targeted DHQZ, with yields of 72–93% (Fig. 6a). The findings revealed that anthranilamide bearing an electron-withdrawing group (4 f) led to a decrease in the yield of DHQZ. In contrast, benzaldehydes with electron-withdrawing groups afforded higher yields of DHQZ products compared to those with electron-rich substituents. In parallel, we explored the catalytic activity of cage 1 in the same reaction. However, the yields of DHQZ were significantly reduced (13–28%). This diminished efficiency is attributed to the small size of the windows (7.14 × 3.07 Å2) and cavity (73.5 Å3) of cage 1, limiting substrate encapsulation and exchange. Additionally, Si-2 was also conducted to the catalysis, affording the product yields of 76–90% (Fig. 6a). Furthermore, we conducted a comparative analysis of our results with those reported in the literature under identical reaction conditions. Comparative experiments were performed using 0.1 mol% of cage 2, employing 3 mL of CH3CN at 40 °C for 14 h. The results showed a comparable yield of 88% when using 3a and 4a as substrates, closely aligning with the reported literature yield of 90%76.Fig. 6: Catalytic results. a Sequential condensation and cyclization of anthranilamide with aldehydes catalyzed by the cages.a Reaction conditions: 2-aminobenzamides (1.0 mmol), aldehyde (1.0 mmol), cage: 0.1 mol %, CH3CN (6.0 mL), rt, 20 h. b Isolated yield. b Effect of flow rate of reaction mixture on the catalytic performance of cage-loaded microreactor. c Kinetic results for cyclocondensation of benzaldehyde and anthranilamide both in batch and microreactor system with the cage loading of 0.5 mol%. d Recycling tests of cage 2 and Si-2@PDMS for cyclocondensation of benzaldehyde and anthranilamide. e Schematic of the amplified continuous flow experimental system.Furthermore, a series of comprehensive experiments were conducted to illustrate that effective catalysis predominantly occurs within the confines of the cage cavity. Initially, the components of the cage framework, including the ligand of TABPM (0.4 mol%), 2-formylpyridine (1.2 mol%), Bu4NOTf (0.8 mol%), and a mono-nuclear tris(pyridylimine)iron(II) complex (0.4 mol%, Supplementary Figs. 24–27), were employed as catalysts for driving the reactions of benzaldehyde and anthranilamide under identical conditions. However, in each case, either no products or only trace amounts were observed (Supplementary Table 5). Subsequently, substrate size selectivity studies were also performed, and two sterically demanding substrates of 3 g and 3 h (Supplementary Fig. 33) were subjected to the reactions. The results revealed that as the substrate size increased, the yield of product 5 gradually diminished (Fig. 6a). Especially, for the largest substrate of 3 h (9.8 × 16.0 Å2), only a 26% yield of the targeted product was observed. The substantial contrast in conversions between small and large substrates strongly suggests that the catalytic process predominantly unfolds within the cavity, and the larger substrates may encounter hindrance in entering the cavity through the cage windows (7.3 × 15.1 Å2).Subsequently, we evaluated the catalytic performances of Si-2 in the CLSC system, which holds significant potential for industrial applications. Firstly, we evaluated the effect of the flow rate of the reaction solution of 3a and 4a on the catalytic performances of the cage-loaded CLSC system. As shown in Fig. 6b, the increased solution flow rate not only accelerated the flow velocity to improve convective mass transfer but also enhanced the total loading of the substrates. These dual advantages facilitated more efficient interactions between substrates and immobilized cages, resulting in a rapid increase in product yield. However, further increasing the flow rate, the productivity of product tended to be gentle. This can be explained by the fact that mass transfer was initially the limiting component in the catalytic process before being gradually replaced by reaction kinetics. As a result, increasing substrate loading and mass transfer does not effectively increase the reaction rate. Consequently, a flow rate of 0.3 mL min⁻¹ was chosen for subsequent continuous conversion, striking a balance between enhanced mass transfer and optimal reaction kinetics.Next, we investigated the scope of the reaction to substrate in a continuous flow system and compared the results with those of a batch system. Under optimal flow conditions, a residence time of ~20 h was required to convert substrates to product 5a–f in 79–95% yields. The Si-2@PDMS produced yields comparable to cage 2 and Si-2 for most substrates (Fig. 6a). In comparison to the results reported by Kobayashi et al.77, our designed Si-2@PDMS exhibited only a slight catalytic advantage at lower catalyst loading over the batch reactor. The sterically demanding substrates of 3 g and 3 h were also conducted in CLSC, and reduced yields of about 32% were observed. This observation strongly demonstrated that the reaction proceeded in the cavity of Si-2 under continuous flow conditions. Furthermore, we assembled two syringe pumps and a panel microreactor to evaluate the activity of Si-2 under homogeneous conditions. Si-2 (0.33 mM in CH₃CN) and a mixture of 3a (330 mM in CH₃CN) and 4a (330 mM in CH₃CN) were introduced into the microreactor through the syringe pumps. Both pumps were set to a flow rate of 0.3 mL min⁻¹. The results showed that the yield of 5a was 87%, which is comparable to the yield in the Si-2@PDMS system.The performance of the cage-loaded microreactor was investigated at various lengths (40–200 cm), corresponding to different loadings of the cage (Supplementary Fig. 34a). The results indicated that the yield of 5a increases with the increase of microreactor length. More importantly, when the microreactor length increased to 200 cm (cage loading of 0.5 mol%), substrates were efficiently converted into products within 3 h, significantly faster than with a catalyst loading of 0.1 mol%. This prompted us to investigate the reaction kinetics of the cyclocondensation of benzaldehyde and anthranilamide with different cage loadings (0.1–0.5 mol%) in the CLSC system. We observed that the reaction rate gradually accelerated with increased cage loading (Supplementary Fig. 34). Additionally, we evaluated the reaction kinetics in a batch reactor with the same range of cage loadings and found that while the reaction rate was comparable to that in the microreactor at low loadings, it was higher in the microreactor at high loadings (Supplementary Fig. 34). As shown in Fig. 6c, at a cage loading of 0.5 mol%, complete substrate conversion required only 3 h in the microreactor, compared to 8 h in the batch reactor. The TOF (h⁻¹) values for the batch and CLSC systems were calculated to be 19 and 63 h⁻¹, respectively. This can be attributed to the enhanced mass transfer capabilities of the microreactor, effectively addressing the common issue of low heterogeneous catalytic efficiency.Additionally, the reaction of 3a and 4a was selected to test the stability and recyclability of Si-2@PDMS. Remarkably, the continuous flow system exhibited consistent performance over multiple cycles, demonstrating recyclability for at least twenty successive runs, each lasting 20 h (Fig. 6d), without any obvious loss in activity. However, after 7 cycles of using cage 2 in the batch reactor, its activity decreased significantly. The TON values were determined to be 4020 in the batch reactor and 16,000 in the microreactor, respectively. Additionally, inductively coupled plasma optical emission spectrometry (ICP-OES) analysis of the product solution indicated almost no loss of Fe ions from the structure after the recycle, affirming the robustness and stability of the Si-2@PDMS. Furthermore, a comparison was conducted between our microreactor and other chip-based reactors in solid-liquid catalysis78,79,80. It was observed that the catalytic system exhibited good stability, attributed to the covalently immobilized cage on the inner walls of the microchannel. Additionally, the CLSC system was designed to be more adaptable to different production tasks, a feature that enhances its versatility.For practical application in pharmaceutical synthesis, the cage-loaded microreactor needs to be scaled up, for instance, by increasing the number of the microreactor or using a larger system such as the Corning microreactor57,81,82. In this study, we adhered to industrial microreactor standards to design and assemble a larger microreactor with outer dimensions of 400 × 400 × 15 mm (Fig. 6e and Supplementary Fig. 35). The microchannel within this reactor has a width of 5 mm, a depth of 3 mm, and a total length of 17 m. When the mixture solution of 3a (0.1 M in CH3CN) and 4a (0.1 M in CH3CN) was delivered at a flow rate of 2.0 mL/min, the substrates were fully converted into products within 2 h. After 24 h, the total TON reached as high as 2551 to yield 57.2 g of the target compound, which is 7 times higher than that of the homogeneous batch system. This observation holds significant promise for the continuous production of drugs, emphasizing the potential utility of the cage-loaded microreactor in pharmaceutical synthesis at an industrial scale.
