Construction of Rh–POPs catalystsThe Rh–POPs samples were synthesized using the impregnation method by introducing rhodium precursor on the POPs–PPh3 support which referred to our previous work20,24. Multiple characterizations including ICP, XRD, TG, N2 sorption, and SEM images indicate the Rh–POPs catalysts processing high specific surface area, hierarchical porosity, and relative good thermostability (Supplementary Figs. 1–5 and Supplementary Tables 1 and 2). High-angle annular dark field-scanning transmission electron microscopy (HAADF-STEM) and the corresponding STEM-energy dispersive spectroscopy (STEM-EDS) elemental mapping images show that isolated individual Rh atoms are uniformly dispersed within the POPs framework without any nanoparticles or clusters of Rh species with the Rh loading from 0.25 to 5 wt% (Fig. 2a–c and Supplementary Figs. 6 and 7).Fig. 2: Structure characterization and catalytic performance of Rh–POPs catalysts.HAADF-STEM images of a 0.25%Rh–POPs, b 2%Rh–POPs, c 5%Rh–POPs; d Rh 3d XPS of Rh–POPs catalysts; e Normalized Rh K-edge XANEs spectra, f FT k3-weighted EXAFS spectra in R-space, and g Wavelet Transforms EXAFS (WT-EXAFS) spectra of Rh–POPs catalysts with different Rh loadings; Reaction pathway (h) and catalytic performance (i, j) of propylene hydroformylation over Rh–POPs catalysts, Conv. represented the conversion of propylene, Sel. represented the selectivity of product aldehydes, including linear aldehydes (blue color) and branched aldehydes (yellow color), l/b represented the ratio of linear and branched aldehydes. TOF was calculated with the lower propylene conversion (<10%) conditions, detailed reaction conditions are shown in Supplementary Table 4.X-ray photoelectron spectroscopy (XPS) was utilized to investigate the electronic states and coordination environment of the unique Rh single sites of Rh–POPs (Fig. 2d and Supplementary Fig. 8). The spectrum of POPs–PPh3 support shows two peaks at 132.0 and 130.2 eV, corresponding to P 2p3/2 of the oxidized and uncoordinated phosphine species, respectively (Supplementary Fig. 8)35. With the introduction of Rh species, a new signal appears at 131.5 eV, which can represent the P 2p3/2 of the phosphine ligands that coordinated to Rh atoms. The positive shift from 130.2 to 131.5 eV of P 2p3/2 can be ascribed to the lone pair electrons filling the empty orbital of the Rh atom to form the Rh–P coordination bonds, decreasing the electron density of uncoordinated phosphine. The Rh 3d spectrum of every Rh–POPs sample represents two peaks, which are attributed to the 3d5/2 and 3d3/2 of Rh+, suggesting the Rh species existed in the form of +1 oxidation state for all the catalysts (Fig. 2d)24. As the Rh contents increased from 0.25 wt% to 5 wt%, the binding energy (B.E.) of Rh 3d5/2 shifts to the higher energy (from 307.7 to 308.7 eV), indicating that the density of the Rh electronic states decreases along with the increase of Rh loading.The Rh K-edge X-ray absorption near edge structure (XANES), extended X-ray absorption fine structure (EXAFS) and Wavelet Transforms EXAFS (WT-EXAFS) spectroscopy were employed to further determine the local coordination and electronic structures of the Rh–POPs samples (Fig. 2e–g). The Rh K-edge XANES spectra indicate that all POPs support Rh species process a higher oxidation state than the Rh foil and lower than that of the Rh2O3, which are very close to the Rh(acac)(CO)2. According to the rising edge overlapping that of Rh(acac)(CO)2, Rh species in all three samples are mainly maintained +1 oxidation state, which is in good consistent with XPS characterization36. In the EXAFS spectra, the major scattering peak at around 1.5–1.6 Å is ascribed to the first coordination shell of Rh–C/O and the shoulder peak appeared at about 1.9 Å is ascribed to the first coordination shell of Rh–P19. The Rh–Rh shell at 2.4 Å cannot be found in the EXAFS spectra and WT-EXAFS spectra of all the Rh–POPs samples, in good consistent with the HAADF-STEM results, indicating that the phosphine ligand of POPs supports could provide a unique coordination environment to immobilize the Rh species as single sites even the Rh loading up to 5 wt%.The fitted EXAFS results of Rh–POPs in the first shell are shown in Supplementary Fig. 9 and Supplementary Table 3. The 0.25%Rh–POPs shows a well-defined structure with a 3.0 coordination number (CN) of Rh–P (2.29 Å) and 2.0 CN of Rh–C/O (2.03 Å), suggesting the P atoms of the POPs framework completely replaced the CO group in the precursor Rh(acac)(CO)2 to form the active center at low Rh content. However, the coordination states of the 2% and 5%Rh–POPs samples are significantly different from the 0.25%Rh–POPs sample with a decreased CN (2.0) of Rh–P and increased CN (3.0) of Rh–C/O. This indicates that a number of the Rh–CO in the Rh–POPs with higher Rh content are retained during the impregnation process. The reduced Rh–P coordination number and the emerging Rh–C coordination number induce the d electrons of the Rh active center to fill into the 2π anti-bonding orbital of CO, resulting in a decrease in the electron cloud density.The catalytic performance of propylene hydroformylation over Rh–POPs catalysts was tested in a fixed-bed reactor with the 0.5 MPa reactant gas at 363 K. The propylene conversion, butyraldehyde selectivity, the ratio of linear and branched product (l/b), and TOF value were evaluated (Fig. 2h–j and Supplementary Table 4). All three samples represent perfect butyraldehyde selectivity (>99%), and the propylene conversion is increased from 7.86 to 75.8% with the Rh content from 0.25 to 5 wt%. However, it is notable that the l/b ratio decreases continuously from 5.78 to 1.6. To exclude the effects of secondary reaction and diffusion, we tested these three samples at lower conversion (<10%, Fig. 2h). All the samples possess a similar TOF value of ~1150 h−1 but a decreased l/b ratio with the increased Rh contents. The noticeable distinction in regioselectivity is most likely owing to the peculiar microenvironment of the single Rh active site.Precise structures of the active centersSolid-state nuclear magnetic resonance (ssNMR) has the unique advantage of characterizing the microstructure of the active center. The framework of POPs is mainly composed of triphenylphosphine polymer, and 1H-13C Cross-Polarization (CP) MAS NMR could provide insight into the microenvironment of the Rh–POPs before and after syngas treatment (Fig. 3a). The main peaks appear in the range from 126 to 146 ppm, which is attributed to the aromatic carbons of PPh3-framework. The signals appeared at 40 and 45 ppm are ascribed to the polymerized vinyl groups, and the peak at 112 ppm is assigned to unpolymerized residual vinyl functional groups23. Before the syngas treatment, with the increase of Rh loading, three signals are gradually highlighted at 31, 100, and 187 ppm, which are attributed to the signal of acetylacetone on the Rh (I) precursor37. After activation under syngas atmosphere, two new signals can be observed at 201 and 199 ppm, which are the characteristic signals of the linear and branched aldehyde groups, respectively. The intensity of the residual unpolymerized vinyl groups corresponding to 113 ppm decreases to a certain extent, demonstrating the hydroformylation of the unpolymerized vinyl functional group on the POPs support. An extraordinary signal appeared at around -10 ppm in the 1H MAS NMR of 5%Rh–POPs is unquestionably ascribed to the proton bonded to the Rh active center, the so-called Rh–H species (Supplementary Fig. 10)38,39. Combined with the decrease of the 13C NMR signal at 31, 100, and 187 ppm, it can be concluded that the acetylacetone are dissociated accompany by the formation of Rh–H bond during the activation process. Simultaneously emerged signals at 15, 28, 45, and 52 ppm of the 13C MAS NMR confirmed the formation of aldehyde chains after hydroformylation of vinyl functional groups, indicating high hydroformylation activity of the single Rh active site (Supplementary Fig. 11). The signals of tetrahydrofuran appeared at 26 and 67 ppm indicate that some solvent remains in the catalysts. The detailed structures were recognized by complementary 2D heteronuclear correlation spectroscopy, such as 13C{1H} and 1H{13C} HETCOR MAS NMR spectra (Supplementary Figs. 12 and 13), and an overview of 13C species is displayed in the Supplementary Fig. 14.Fig. 3: The structural identification of Rh–POPs catalysts before and after activation.a 1H-13C CP MAS NMR spectra of Rh–POPs catalysts before and after activation by syngas. b 13C{1H} 2D R-SLF NMR spectra of 5%Rh–POPs-active sample. c 31P MAS NMR spectra and the fitness of Rh–POPs catalysts before and after activation by syngas. d 1H{31P} HETCOR MAS NMR spectrum of 5%Rh–POPs-active catalyst, the corresponding 1H MAS NMR and 1H-31P CP MAS NMR are displayed on the top and left of the 2D spectrum. In situ time-resolved FT-IR spectroscopy study of e 0.25%Rh–POPs and f 5%Rh–POPs in syngas feeding at 363 K and purging with N2 subsequently.13C{1H} 2D separated local field (R-SLF) NMR experiments were performed to disclose the microenvironment of a series of Rh–POPs-active samples (Fig. 3b and Supplementary Fig. 15). The residual 13C–1H dipolar coupling can be used as an index to evaluate the motility of molecules or intramolecular segments of the POPs framework. The 13C–1H dipole coupling of CH and CH2 groups in rigid molecules is ~22 kHz. If the molecular or carbon chain segment is relatively mobile, the molecular motion will average the residual dipole coupling, resulting in the final measured 13C–1H dipole coupling less than 22 kHz40. Two peaks appear around 130 ppm and 40 ppm in the F2 dimension of the spectra, which can be attributed to the CH species in the aromatic hydrocarbon and polyvinyl segment of the POPs framework, respectively. The residual 13C–1H dipole coupling of all the samples maintains at 22 kHz with the increase of Rh doping from 0.25 to 5 wt%, indicating the robust structural stability under syngas activation.Resolving the local structure of the active site by detecting the signal of P is undoubtedly very straightforward because Rh and P are directly coordinated. Herein, 31P MAS NMR and 1H–31P CP MAS NMR spectra are used to monitor the precise structure and evolution of the active sites before and after activation of the catalysts (Fig. 3c and Supplementary Fig. 16). There are two obvious peaks at −6 and 25 ppm of all the samples, representing uncoordinated P atoms and slightly oxidized P = O species of the POPs framework21. In the 0.25%Rh–POPs sample, a shoulder signal around 30 ppm appears at the lower field of the P = O peak, which is attributed to P atoms with multiple coordination bonds connected with the Rh atom41,42. As the content of Rh increases to 2 wt% and 5 wt%, a new signal appears at 47 ppm and the peak intensity increases with the Rh loading, which is attributed to Rh(acac)(CO)(PPh3-frame)43. All samples were inevitably mildly oxidized during syngas activation and subsequent NMR testing, resulting in a slight increase in the signal at 26 ppm. In the 0.25%Rh–POPs, the signal at 30 ppm corresponding to the Rh–P multi-coordination bonds keeps unchanged after activation, indicating superior structure stability. It is worth noting that in the remaining two samples, the spectra change dramatically before and after the syngas activation, with the signal at 47 ppm decreased accompanied by an increased peak at 33 ppm, indicating that the state of P that coordinated with Rh is changed from monophosphate to polyphosphate during this process. 2D 1H{31P} heteronuclear correlation spectroscopy was adopted to further assist the attribution of the 31P NMR signal as shown in Fig. 3d. The arresting correlated signal at (−10, 33) of 5%Rh–POPs sample proves the spatial proximity of P species at 33 ppm and the proton of the Rh–H bond. The standard sample HRh(CO)(PPh3)3 was used as a reference and the analogous correlated signal proved the correctness of attribution that the 33 ppm is the P species directly bonded to the Rh atom (Supplementary Figs. 17 and 18).For deeper insight into the coordination state of the single Rh active site and how the active center dynamically changes during activation through syngas, in situ time-resolved FT-IR spectroscopy was conducted (Fig. 3e, f and Supplementary Fig. 19). Two distinct absorption bands at 2173 and 2117 cm−1, gradually increase with the injection of syngas and decrease until vanish with the purge of nitrogen, which is attributed to the CO gas. In the spectrum of 0.25%Rh–POPs (Fig. 3e), the peak at 2069 cm−1 is ascribed to the ν(Rh–CO) of the HRh(CO)(PPh3-frame)3 species44. The bands at 2000 and 1950 cm−1 are belonged to the ν(Rh–CO) stretching vibration of HRh(CO)2(PPh3-frame)2, and the remaining critical signal at 2038 cm−1 is ascribed to the stretching vibration of ν(Rh–H) species24,29,45. In the 5%Rh–POPs sample (Fig. 3f), a peak attributed to the Rh–H species analogously appeared at 2040 cm−1 and this attribution can be proved by the H-D exchange experiments (Supplementary Fig. 20). The ν(Rh–CO) stretching vibrations of HRh(CO)2(PPh3-frame)2 are appear at 2008 and 1948 cm−1, representing a slight shift in wavenumber compared with the sample of 0.25%Rh–POPs (2000 and 1950 cm−1). This result indicates that the electron state of Rh is different in these two samples, which can affect the vibration wavenumber of Rh–CO species. The active center of the 2%Rh–POPs sample (Supplementary Fig. 19) shows a similar spectrum to that of the 5%Rh–POPs sample. Hence, the 0.25%Rh–POPs and 5%Rh–POPs were selected as representative samples for further comparison. It is worth noting that two penta-coordinate HRh(CO)(PPh3-frame)3 and HRh(CO)2(PPh3-frame)2 could be simultaneously observed in the IR spectrum, but in the real reaction process, a Rh–P or Rh–CO bond will be dissociated to form the tetradentate HRh(CO)(PPh3-frame)2 species for receiving the olefin coordination during the hydroformylation reaction20,39,46. For better understanding, HRh(CO)(PPh3-frame)2 is used to represent the real active center, in which the bite angle specifically refers to the P–Rh–P angle of this active center.In the 5%Rh–POPs (Fig. 3f), the Rh–H species appeared at 2040 cm−1 and the intensity is significantly stronger than that of the 0.25%Rh–POPs sample (Supplementary Fig. 21), which is in consistent with the 1H MAS NMR (Supplementary Fig. 10), while the signal at −10 ppm cannot be observed in 0.25%Rh–POPs sample mainly because the content of Rh–H species in this sample is too low to detect by NMR. The signal at 2070 cm−1 gradually increase with the introduction of syngas but gradually disappear with the purge of N2, indicating the HRh(CO)(PPh3-frame)3 species cannot be stabilized in this sample. Meanwhile, after the purging of N2, the signal intensity at 2070 cm−1 of the 5%Rh–POPs is much lower than that of the 0.25%Rh–POPs. This illustrates that the bite angle of P–Rh–P of 5%Rh–POPs is larger than that of 0.25%Rh–POPs, and the large steric hindrance makes the third PPh3-frame difficult to coordinate with Rh.The P–Rh–P bite angle of 0.25%Rh–POPs has been well demonstrated with the region between 90° and 120° that was referred to the homogeneous active center HRh(CO)2(PPh3)2 with ee (120°) and ea (90°) isomer in hydroformylation20,23,24. Therefore, we hypothesize that the bite angle could surpass the range of 90–120°, particularly with an increase in Rh content. To determine the possible range of P–Rh–P bite angles, DFT calculations were employed for detailed analysis (Fig. 4a, b, Supplementary Fig. 22, and Supplementary Tables 5 and 6). The phosphine ligands in Rh–POPs materials are immobilized on the POPs framework, which are very different in homogeneous situations where the phosphine ligands exhibit high flexibility in solvents. Although the POPs framework possesses a certain level of flexibility, it restricts the range of movement for coordinated P ligands compared to that observed in a homogeneous phase. To streamline the computational model and make it closer to the real heterogeneous conditions, triphenylphosphine is selected to replace the PPh3-frame, and the three farthest protons of the three benzene rings opposite to the P atom are immobilized to restrict the mobility of coordinated P atoms. The geometric center of the plane created by three protons serves as a representation of the coordination site for the P ligand, while the spatial distance between the two geometric centers is designated as the ligand coordination distance (Fig. 4a). Hence the different bite angles and P–P distance are determined by adjusting the ligand coordination distance.Fig. 4: Evolutions of the active centers in the Rh–POPs catalysts.a Schematic diagram described the gravity distance of the single Rh active center. b Optimized structures of the Rh active center by regulating gravity distance and related DFT analysis for the relative energy and P–Rh–P bite angle, i–viii represent the structures with Rh–H and Rh–CO coordinated on the same side, corresponding i’–viii’ represented the opposite side. c Dynamic evolving trajectories of Rh active center before and after activation by syngas. d Schematic illustration of the probable Rh positions under the certain circumstance of lower and higher Rh content, for ease of viewing, Rh–H and Rh–CO are not drawn in the figure.In the case of only two PPh3 coordinated with Rh atom, a series of possible stabilized Rh–POPs structures were optimized by regulating the ligand coordination distance from 9.9 Å to 12.3 Å (Supplementary Fig. 22). The range of P–Rh–P angles can be obtained from 103.2° to 175°, correspondingly the P–P distance from 3.41 to 5.01 Å. It is interesting to find that the angles show a completely different tendency in different regions as the ligand coordination distance increased from 9.9 Å to 12.3 Å. In the range from 9.9 Å to 10.8 Å, the angles increase linearly from 103.2° to 124.7°. However, the ligand coordination distance raised from 10.8 Å to 11.1 Å, the angle changes with a huge jump from 124.7° to 168.7°. As the ligand coordination distance continues to increase from 11.1 Å to 12.3 Å, the angles increase linearly again from 168.7° to 175.0°. The notable difference observed in these two regions (below 124.7° and above 168.7°) likely stems from the variation in hybridization between rhodium d-orbital and phosphorus p-orbital, a phenomenon influenced by the compound’s geometry.Due to the HRh(CO)(PPh3-frame)2 being the actual active center, Rh–H and Rh–CO were also taken into account as shown in Fig. 4b. The Rh–H and Rh–CO could be coordinated on the same side or opposite side, which also affects the P–Rh–P bite angle. In the range of ligand coordination distance from 9.9 Å to 10.4 Å, the Rh–H and Rh–CO are more inclined to coordinate on the same side, with the P–Rh–P bite angles between 95.4° and 104.1°. As the ligand coordination distance increases from 10.8 Å to 12.1 Å, it can be found that the active center structures formed by the opposite side coordination of Rh–H and Rh–CO have the lower energy, and the bite angel is in a certain range between 158.3° and 168.1°. Based on the above theoretical calculation, we can infer that two distinct active centers could exist with a discrete P–Rh–P angle range. Therefore, compared with the 0.25%Rh–POPs sample with the bite angel between 90° and 120°, we speculate that the enlarged bite angle may exist in the range from 158° to 168° with the increase of Rh content to 5 wt%.The possible step-by-step evolution of the Rh–POPs active centers based on experiments and calculations are depicted in Fig. 4c. In the process of synthesis of Rh–POPs catalyst by impregnation method of Rh(acac)(CO)2 precursor on the POPs framework, Rh(acac)(PPh3-frame)2 (so-called Rh–POPs) is easily formed with the lower Rh content (0.25 wt%), then convert to HRh(CO)(PPh3-frame)3 and HRh(CO)2(PPh3-frame)2 species (so-called Rh–POPs-active) during the process of syngas activation in which these two species could be transformed each other under the CO atmosphere. Subsequently, a Rh–P or Rh–CO bond dissociated accompanied by the formation of HRh(CO)(PPh3-frame)2 species with the P–Rh–P bite angle between 90° and 120° for further hydroformylation reaction. When the Rh loading increased to 5 wt%, Rh(acac)(CO)(PPh3-frame) species were formed. Then the acetylacetone is removed under the treatment of syngas, along with the formation of the Rh–H bond. Besides, Rh forms a new coordination bond with the neighboring PPh3-frame to generate HRh(CO)2(PPh3-frame)2 species. Finally, a CO is dissociated to form the truly active center HRh(CO)(PPh3-frame)2 with the P–Rh–P bite angle between 158° and 168°.We have hypothesized the possible reasons for the distinct active centers caused by different Rh loading as shown in Fig. 4d. In the case of lower Rh content, Rh (I) precursor is more inclined to locate in the rich P area during the impregnation process and the P atom on the POPs framework completely replaces all the CO of the precursor Rh(acac)(CO)2 to form Rh(acac)(PPh3-frame)2. After activation by syngas, Rh–H bond formation is accompanied by the removal of acetylacetone. At the same time, one of the surrounding superfluous P atoms will coordinate with a single Rh active center to form the HRh(CO)(PPh3-frame)3 species. With the increase of Rh loading, the Rh (I) precursor has to settle in the region of lower P concentration. Due to the steric hindrance of acetylacetone, the P atom only replaces one of the CO groups to form Rh(acac)(CO)(PPh3-frame) species. With the activation of syngas, Rh will coordinate with the surrounding P atom after acetylacetone desorption. The lower p concentration and the large bite angle of P–Rh–P in the range from 158° to 168° make it difficult for a third P to coordinate with the Rh active center.Dynamic evolution of hemilabile coordinationThe above characterizations revealed the microstructure of the truly active center HRh(CO)(PPh3-frame)2 with distinct P–Rh–P bite angles under different rhodium content. To understand how single Rh active centers regulate the distribution of product aldehydes, a variety of characterizations such as in situ time-resolved XAFS and FT-IR, combined with quasi-in situ NMR have been used to explore the dynamic changes of active centers and coordinated species (Fig. 5a–e). In situ XAFS were used to identify the change of Rh valence state and electron cloud density of the 5%Rh–POPs during the reaction process as shown in Fig. 5a. It can be seen from the time-resolved spectra with the injection of reaction gas, a drop in white line intensity at 23,247 eV indicates that the coordination state of Rh changes dynamically during the reaction process13. The adsorption edge position at around 23,220 eV slightly shifts to the lower energy, indicating a gradual decrease in the electron cloud density of the Rh active center and this maybe ascribe to the CO replaces the coordinated PPh3-frame in the reaction process47. However, the slight change in the electron cloud density is insufficient to affect the valence state of Rh, suggesting that the valence state of Rh is almost unchanged during the whole reaction process.Fig. 5: Mapping the dynamic evolution of the active center and hemilabile coordination.a In situ time-resolved Rh K-edge XAFS spectra of 5%Rh–POPs sample treated by mixture reactant (C3H6/CO/H2 = 1:1:1) for 20 min at 363 K. b In situ time-resolved FT-IR spectroscopy study of 5%Rh–POPs-active in mixture reactant (C3H6/CO/H2 = 1:1:1) feeding for 30 min at 363 K. c 31P MAS NMR spectra and the fitness of sample SI, SII, and SIII (sample SI: 5%Rh–POPs-active; sample SII: sample SI treated with C3H6 at 363 K; sample SIII: sample SII treated with syngas at 363 K). d Comparison of 1H{31P} HETCOR MAS NMR spectrum of SII and SIII, the corresponding 1H MAS NMR and 1H-31P CP MAS NMR are displayed on the top and left of the 2D spectrum (black line: sample SII, red line: sample SIII). e 1H-13C CP MAS NMR spectra of sample SI, SII, and SIII. f Schematic diagram described the dynamic evolution of the active center and hemilabile coordination under reaction conditions.In situ time-resolved FT-IR spectroscopy was used to intensively explore the reaction process of 5%Rh–POPs-active sample with the reactants (C3H6/CO/H2 = 1:1:1) and then purging with N2 (Fig. 5b, Supplementary Figs. 23 and 24, and Supplementary Table 7). The peaks at 1664 cm−1 and 1642 cm−1 are attributed to the ν(C = C) of C3H6. The peaks at 991 cm−1 and 912 cm−1 are ascribed to the ν(C–H) out-of-plane bending vibration on unsaturated carbon atoms of propylene48,49. The antisymmetric bending vibrations of ν(−CH2−) and ν(−CH3) are presented at 1442 cm−1 and 1473 cm−1, respectively. The multi-peaks between 2880 cm−1 and 3104 cm−1 are ascribed to the symmetrical and antisymmetric stretching ν(C–H) vibration of −CH2− and −CH3 groups. The characteristic adsorption peak at 1728 cm−1 is attributed to the stretching vibration of ν(C = O) in the product aldehyde, and the corresponding ν(C–H) bending and stretching vibration of aldehyde group appeared at 2714 cm−1 and 2811 cm−1, respectively50. These three characteristic peaks gradually increase with the introduction of reactant, accompanied by the ν(Rh–H) and ν(Rh–CO) species, indicating the occurrences of hydroformylation at the active center.In addition to the above in situ characterization techniques, quasi-in situ ssNMR method was used to monitor the dynamic changes of the coordinated ligands and reaction active intermediates. The 5%Rh–POPs-active sample (labeled as “SI”) was stepwise treated with C3H6 (labeled as “SII”) and syngas (labeled as “SIII”) at reaction conditions and then quenched by liquid nitrogen for subsequent 31P and 13C NMR experiments (Fig. 5c–e). The truly active center of the 5%Rh–POPs-active sample has been verified with the specific structure of HRh(CO)(PPh3-frame)2, and the bite angle of P–Rh–P is between 158° and 168°. With the introduction of propylene, it is interesting to find that the signal of Rh connected with gemini PPh3-frame at 33 ppm significantly decreases, corresponding to the prominent increase of characteristic signal of Rh coordinated with mono PPh3-frame at 47 ppm. This suggests the coordination ability of propylene to Rh is stronger than that of PPh3-frame, which results in one Rh–P coordination bond dissociation in the coordination process. As to sample SII, it can be seen the signal at 47 ppm reduces accompanied by the signal at 33 ppm increasing with the introduction of syngas. The compared 1H{31P} HETCOR spectra of sample SII and SIII were also adopted to further explain this attribution as shown in Fig. 5d. The cross peak of sample SII at (6.8, 47) disappears after the syngas introduction, along with the cross peak emerged at (−10, 33), indicating the syngas processing hydroformylation reaction with the coordinated propylene. With the Rh–H species formation and the product aldehyde desorption, the coordinate state of Rh will change from a single PPh3-frame to a gemini PPh3-frame. 1H-13C CP and 13C MAS NMR spectra were adopted to further explore the intermediates in the reaction process (Fig. 5e and Supplementary Fig. 25). As compared with the initial 5%Rh–POPs-active sample, the signals at 18, 115, and 137 ppm generated with the introduction of propylene are attributed to the −CH3, −CH = , and =CH2 of propylene, respectively. Two notable signals at 184 and 187 ppm are ascribed to the C = O species of the Rh-acyl group, indicating that propylene coordinated on the active center and subsequently interacted with the Rh–H bond to form Rh-alkyl, then CO inserted into the Rh-alkyl group to form Rh-acyl species. Based on this state (sample SII) with the continuous injecting of syngas, the signals at 115, 184, and 187 ppm disappeared, indicating that the hydroformylation reaction continued and the coordinated propylene was completely consumed. After the hydroformylation reaction, some new signals appear mainly in the range of 13 to 45 ppm (13, 16, 40, and 45 ppm), which are attributed to the product of butyraldehyde. These signals are sharp in the 13C CP and MAS NMR spectra, indicating the residual butyraldehyde has strong mobility and can be easily desorbed from the Rh–POPs framework.Based on the above characterizations and analysis, the dynamic dissociation and re-coordination between the metal center and coordinated ligands can be demonstrated during the adsorption, reaction, and desorption of guest molecules in the reaction process, which can be the so-called hemilability. Hemilability is an important concept in homogeneous catalysis, that is, the activation of reactants and the formation of products can occur simultaneously through the reversible opening and closing of the coordination state between metal and ligand (Supplementary Fig. 26). However, the dynamic change of this coordination state and its effect on the reaction is rarely discussed in heterogeneous catalysis. Meanwhile, the complexity of the active centers in heterogeneous catalysis makes it more difficult to give explicit experimental evidence for hemilabile coordination. Herein, the Rh–POPs catalyst system contains a well-defined structure with flexible framework and dynamic evolution active sites, which can be used as an ideal model to understand the hemilabile coordination property. Combined with the deep understanding of the microenvironment of the truly active center and the capture of the active intermediate under in situ conditions, we use schematic diagram to describe the dynamic evolution between the active center and hemilabile coordination in the process of guest molecules adsorption, activation, formation of intermediates and finally desorption (Fig. 5f).In the 0.25%Rh–POPs sample with lower Rh content, the bite angle of metal center and the coordinated ligands is between 90° and 120°, which makes the ligands connected to the metal center more stable, causing reactant molecule coordination and activation without breaking the metal–ligand coordination bond, as proved by in situ time-resolved FT-IR spectroscopy and 31P MAS NMR (Supplementary Figs. 27 and 28). However, in the 5%Rh–POPs sample with higher Rh content, the bite angle is between 158° and 168°, and thereby makes the ligand hemilabile. With the coordination and activation of guest molecules, the coordination bond between a certain metal and the ligand will be broken into an open state, and then the hemilabile ligand will re-coordinate with the metal center and become to a closed state, accompanied by the generation and desorption of products.Mechanism of the propylene hydroformylation over Rh–POPsThrough the comprehensive understanding of hemilability in heterogeneous catalysis and the precise analysis of the microenvironment of the active center, the reaction mechanism of propylene hydroformylation is proposed (Fig. 6a). DFT calculation was also used to confirm the feasibility of the reaction path (Fig. 6b–d). In the process of propylene hydroformylation catalyzed by Rh–POPs material with a lower Rh content such as 0.25 wt%, one PPh3-frame or CO of the HRh(CO)(PPh3-frame)3 and HRh(CO)2(PPh3-frame)2 (I and Irev in Fig. 6a) will be dissociated, resulting in a real active center of HRh(CO)(PPh3-frame)2 with the P–Rh–P bite angle between 90° and 120° (II in Fig. 6a). In the higher Rh content such as 2 wt% and 5 wt%, one CO of the HRh(CO)2(PPh3-frame)2 (I’ in Fig. 6a) will be dissociated to result in HRh(CO)(PPh3-frame)2 with an enlarged P–Rh–P bite angle from 158° to 168° (II’ in Fig. 6a). Two typical models with the ligand coordination distance of 10.1 Å and 12.1 Å were chosen to represent the truly active center II and II’ of which the P–Rh–P bite angle at 100° and 168°, respectively.Fig. 6: Mechanism investigation of the propylene hydroformylation over Rh–POPs catalysts.a Proposed reaction cycle of the propylene hydroformylation over Rh active centers with lower and higher Rh content. b Optimized structures of the active center II (top) and II’ (bottom) with the coordination of propylene. c 2D color distribution of the localized-orbital locator (LOL) projection along the P–Rh–P plane of the II (top) and II’ (bottom). d Relative energy between linear and branched products of the Rh active center with lower and higher Rh contents following the reaction route of (a).Prior to the coordinating with propylene, the Rh centers in the intermediate structures maintain a tetrahedral coordination configuration (Supplementary Fig. 29). When the hydroformylation reaction starts, propylene is activated by filling π electrons into the Rh empty d-orbital, and the corresponding optimized structures are shown in Fig. 6b. As proposed in the mechanism outlined in Fig. 6a, the propylene will be coordinated at the active center II to form a penta-coordinated intermediate (III in Fig. 6a) and the Rh still coordinated with two P atoms throughout the process, in which one Rh–P bond is stretched from 2.35 to 2.41 Å, while another Rh–P bond is remaining with the bond length of 2.41 Å. However, in the process of coordination and activation of propylene on the active center II’, the hemilabile coordination phosphine ligand will be dissociated from the single Rh active center to form a tetra-coordinate intermediate (III’ in Fig. 6a), resulting in the coordination state of Rh from bisphosphate ligand to monophosphate ligand with the open state. The DFT calculation shows that the distance of one Rh–P bond is extended from 2.44 to 3.55 Å, which is much longer than that of one Rh–P bond, indicating a Rh–P bond fracturing during the coordination of propylene, meanwhile the distance of residual Rh–P bond is shortened from 2.44 to 2.34 Å.In order to gain a deeper understanding of the differential molecular coordination following propylene coordination in the two scenarios. A bond characteristic relying on the kinetic-energy density, known as the localized-orbital locator (LOL), was employed for characterizing the chemical bond nature of the structures (Fig. 6c). One can clearly see that the localization of orbitals between the Rh and P bonds is more pronounced near P in the structure II’; hence, the bonding between Rh and P in structure II’ is weaker as compared to that in structure II. In addition, the average Mayer bond order of the Rh–P in II’ was 1.14 which is smaller than that of II with the value of 1.32, indicating that the degree of the electron cloud between Rh and P atom of II’ is less overlapping. This is also the reason why, after propylene coordinated in structure II’, one of Rh–P bond breaks, leading to a stable tetrahedral coordination structure formation.Subsequently, the different configurations of Rh-alkyl are formed when the coordinated propylene inserted into the Rh–H bond, which determines whether the final aldehydes are linear or branched. It is more inclined to form the linear Rh–C3H7 active intermediate (V in Fig. 6a) due to the crowded steric hindrance provided by the gemini PPh3–ligands with lower Rh content. However, the branched Rh–C3H7 (V’ in Fig. 6a) is easier to form due to the large reaction space provided by the monophosphate-coordinated ligand with higher Rh content. The energy differences between linear and branched intermediates were also calculated by the DFT calculation (Fig. 6d). It can be seen from the relative energy diagram that the energy of the products formed by linear aldehydes is lower, so the l/b ratio of the products will be larger than 1, which is also in consistent with the results of the hydroformylation of propylene (Fig. 2i). At lower Rh content, the energy difference between the linear and branched product is 0.72 eV, which is much higher than that of higher Rh content sample (0.23 eV). This indicates that the possibility of the branched product will increase accompanied by the Rh loading, which is also in consistent with the results that the l/b ratio of the 0.25%Rh–POPs and 5%Rh–POPs are 5.78 and 1.6, respectively.Following the coordination of CO, it is inserted into the Rh–C3H7 group to form (C3H7CO)Rh(CO)(PPh3-frame)2 active species (VI and VI’ in Fig. 6a) through carbonylation. Finally, linear and branched butyraldehyde is eliminated through a hydrogenation reaction, and the catalyst returns to the initial HRh(CO)(PPh3-frame)2 state (II and II’ in Fig. 6a) to complete the catalytic cycle. In the whole catalytic cycle, Rh keeps coordinated with the gemini PPh3-frame at lower Rh content, which makes the higher l/b ratio of the aldehydes. In the higher Rh content, the dissociation and re-coordination between the hemilabile coordination PPh3-frame ligands and the single Rh active center are the keys to the formation of branched aldehydes, which is also the reason for the decline of the l/b ratio.
