Defects engineering simultaneously enhances activity and recyclability of MOFs in selective hydrogenation of biomass

The development of synthetic methodologies towards enhanced performance in biomass conversion is desirable due to the growing energy demand. Here we design two types of Ru impregnated MIL-100-Cr defect engineered metal-organic frameworks (Ru@DEMOFs) by incorporating defective ligands (DLs), aiming at highly efficient catalysts for biomass hydrogenation. Our results show that Ru@DEMOFs simultaneously exhibit boosted recyclability, selectivity and activity with the turnover frequency being about 10 times higher than the reported values of polymer supported Ru towards D-glucose hydrogenation. This work provides in-depth insights into (i) the evolution of various defects in the cationic framework upon DLs incorporation and Ru impregnation, (ii) the special effect of each type of defects on the electron density of Ru nanoparticles and activation of reactants, and (iii) the respective role of defects, confined Ru particles and metal single active sites in the catalytic performance of Ru@DEMOFs for D-glucose selective hydrogenation as well as their synergistic catalytic mechanism.


/ 57
All chemicals were purchased from commercial suppliers and used without further purification. All dried samples were stored under N2 in a glovebox.

Catalysis experiment
The 50 g (25 wt%, 1.543 mol/L) D-glucose aqueous solution without any catalyst, in presence of Ru NPs, MOFs and Ru impregnated MOF catalysts (1 g), respectively, was transferred into a 100 mL stainless-steel high-pressure reactor. Before starting reaction, the reactor was purged with H2 to 5.0 MPa, then degas to 1.0 MPa at room temperature, this process was repeated for three times to remove the air. The D-glucose aqueous solution, stirred with predetermined rates (600 or 800 rpm), was heated at the desired temperature (100, 120 or 140 ºC) under 5.0 MPa H2 for a predetermined reaction time (ranging from 90 to 180 minutes), and then cooled to room temperature. Tiny amounts of aliquots were taken out every half an hour during reaction via a dip-tube inserted into the solution to test the activity of these catalysts. The separated reaction solution, obtained after removing the heterogeneous catalysts by centrifugation, was analyzed by HPLC to determine the conversion of D-glucose, selectivity and yield of sorbitol. To test the reusability of these catalysts, the catalysts were separated from the reaction solution, then washed with deionized water and ethanol, respectively, and finally dried in a vacuum oven at 80 ºC. The obtained catalysts after reaction were reused directly for the next run of biomass hydrogenation of D-glucose to form sorbitol.

Characterization methods
Powder X-ray diffraction (PXRD): PXRD patterns of all samples were recorded on a Rigaku Smartlab (3 KW) equipment with a Ni filter using Cu-Kα radiation (λ = 1.542 Å). The patterns were collected in reflectance of Bragg-Brentano geometry over a range of 2θ = 5 -50º at room temperature.

UHV-FTIR spectroscopy and XPS:
The ultra-high vacuum Fourier transformed infrared spectroscopy (UHV-FTIRS) and X-ray photoelectron spectroscopy (XPS) measurements were conducted with a sophisticated UHV apparatus combing a state-of-the-art FTIR spectrometer (Bruker Vertex 80v) and a multichamber UHV system (Prevac) [2][3][4] . This dedicated apparatus allows performing both IR transmission experiments on nanostructured powders and infrared reflection-absorption spectroscopy (IRRAS) measurements on well-defined model catalysts (single crystals and supported thin films). Both the optical path inside the IR spectrometer and the space between the UHV chamber as well as the spectrometer were evacuated to exclude the ambient molecule adoption, ensuring superior sensitivity and stability. The MOF sample (approximately 200 mg) was first pressed into an inert metal mesh which was mounted on an especially designed sample holder, and then activated in the UHV chamber at 500 K to remove all contaminants. Exposure to carbon monoxide (CO) was achieved using a leak-valve-based directional doser connected to a tube of 2 mm in diameter, which is terminated 3 cm from the sample surface and 50 cm from the hot-cathode ionization gauge. The IR experiments were carried out at temperatures as low as 110 K. All UHV-FTIR spectra were collected with 1024 scans at a resolution of 4 cm -1 in transmission mode, using a spectrum of the clean sample as a background reference.
The XPS experiments were carried out using a VG Scienta R4000 electron energy analyzer. The pass energy was fixed at 200 eV for all the measurements. A flood gun was applied to compensate for the charging effects. The binding energies were calibrated to the C1s line at 284.8 eV as a reference.
The XP spectra were deconvoluted using the software Casa XPS with a Gaussian-Lorentzian mix function and Shirley background subtraction.  D0 represents pristine MIL-100-Cr, while D1a′-c and D2a′-c represent defect engineered MIL-100-Cr incorporated DL1 and DL2, respectively, with different feeding ratios (z) of defective linker (DLx, x = 1: 3,5-pyridinedicarboxylate; x = 2: m-phthalate) to total ligands (TL = DLx + parent linkers), ranging from 5% to 50%; and Ru@D2a before and after catalysis for 12 runs indicate that they have relatively large particle sizes with good crystalline. These results confirm that the cationic framework of MIL-100-Cr has good tolerance to the incorporated DLx (x = 1: 3,5-pyridinedicarboxylate, x = 2: m-phthalate). For both types of DEMOF and Ru@DEMOFs, the fine peaks gradually disappeared and merged into broad bands along with increasing the feeding ratio (z) of DLx to TL (x = 1, z ≥ 30%; x = 2, z ≥ 50%), attributed to the decrease of particle size ( Supplementary Fig. 33) 6 . After Ru impregnation, the presence of broad bands accompanying the disappearance of fine peaks in the PXRD pattern of D2b illustrates that the process of Ru impregnation results in decreases of particle sizes of D2b 7 . Noticeably, all these Ru impregnated MOFs catalysts show overall lower thermal stability than the respective corresponding MOFs supporters, illuminating that the Ru impregnation process has critical effects on the framework of MIL-100-Cr. All TGA curves of D1a′, D2a′, Ru@D1a′ and Ru@D2a′ are quite similar to that of D0, illuminating that they maintain the framework of MIL-100-Cr. As shown in Supplementary Fig. 4a-d, the evolution trends of thermal stability of DEMOFs containing type-A defects, DEMOFs containing type-B defects, Ru NPs impregnated DEMOFs containing type-A defects and Ru NPs impregnated DEMOFs containing type-B defects with feeding ratio (z) ranging from 0% to 10% are consistent with that of feeding ratio (z) ranging from 0% to 50%, respectively (Supplementary Fig. 3). As shown in Supplementary Fig. 6a Fig. 12 The ratios of the integrated peak areas of the F 1s band to that of C 1s band in the highresolution XPS spectra of Ru@D0 and Ru@D1a-c.
The bands of F 1s are centered at the binding energy of ~684.4 eV ( Supplementary Fig. 11), and the ratios of the integral area of the F 1s band to that of C 1s band for the selected catalysts (Ru@D1a-c) are all lower than that of Ru@D0 ( Supplementary Fig. 12 The bands for Cr 3+ and Cr δ+ (δ < 3) cannot be distinguished in the high-resolution XPS spectra of Cr 2p, but the presence of Cr δ+ (δ ≤ 3) nodes could be clearly confirmed by UHV-FTIR spectra.   compared to that of D0 and Ru@D0, respectively, don't result in the increase of mesopore, primarily attributed to a certain degree of blocking pores due to local disorder of these samples.

/ 57
Supplementary   The STEM images (Supplementary Fig. 25) and the statistics of particle size distribution of Ru NPs ( Supplementary Fig. 26) show that the size evolution of confined dominate Ru NPs in these DEMOFs with feeding ratio (z) ranging from 0% to 10% is consistent with that of Ru@DEMOFs with higher content of DLx ( Supplementary Fig. 23, 24).  NPs in Ru@D1a-c is lower than that in Ru@D2a-c with the same z of DLx, illuminating that the type-A defects can stabilize Ru NPs more efficiently than type-B defects against aggregation during catalytic reaction, mainly due to the stronger anchoring effect between confined Ru NPs and basic pyridyl-N atoms at type-A defects. The above results demonstrate that a rational tuning of defects can prevent the aggregation of Ru NPs embedded in Ru@DEMOFs. The average diameters of Ru NPs in both two kinds of Ru@DEMOFs, obtained from CO pulse chemisorption measurements, also increase upon increasing the feeding ratio of DLx (x = 1, 2) to TL, being consistent with that obtained from STEM measurements. The dispersions of Ru NPs in both two kinds of Ru@DEMOFs decrease along with increasing the feeding ratio of DLx (x = 1, 2) to TL, primarily attributed to the aggregation of defects with high concentrations. These results further confirm that the sizes and distributions of Ru NPs can be controllably adjusted by the type of introduced defects as well as their concentrations. In relation to that of Ru@D0, the main bands stemming from both CO-Cr 3+ (Supplementary Fig. 27a) and CO-Cr δ+ (Supplementary Fig. 27b) in D2a-c shift slightly to lower frequency with increasing z of DL2. These results demonstrate the formation of electron-enriched Cr δ+ defects via the partial reduction of pristine Cr 3+ -CUSs along with the incorporation of DL2. (d-f) in the Cr 3+ -related CO vibration region for Ru@D0 (d), Ru@D1a (e) and Ru@D2a (f). All samples were exposed to CO (0.01 mbar) at ~110 K, and then heated to the indicated temperatures. Prior to exposure, each sample was heated to 500 K to remove all adsorbed species. The binding energy of the Ru 3d is very close to that of the more intense C 1s peaks in the XPS spectra. However, the binding energy of Au 4f is non-overlapping with all elements of the framework, and thus can be used as a solid reference for a reliable analysis of the electronic structure changes. to TL, and that of these two kinds of Ru@DEMOFs with same z is comparable (Supplementary Table   6). Generally, the larger size of metal nanoparticles results in the smaller binding energy. On consideration of these two aspects, the binding energy of Au@D1c is expected to be comparable to that of Au@D2c, and both of which should be smaller than that of Au@D0. However, the XPS spectra ( Supplementary Fig. 29) show that the binding energies of the Au 4f7/2/4f5/2 doublet for both of the 39 / 57 Au@D1c (84.1/87.8 eV) and Au@D2c (84.3/88.0 eV) are higher than that of Au@D0 (83.9/87.6 eV), revealing that the embedded Au NPs in D1c and D2c DEMOFs are slightly positively charged. This finding is attributed to the electronic interaction between the embedded Au NPs and defective Cr δ+ -CUSs (δ < 3) acting as Lewis acid sites (electron acceptors) that lose one coordinating carboxyl in DEMOFs (see Fig. 2a-b). Furthermore, the binding energy of the Au 4f7/2/4f5/2 doublet in Au@D1c is lower than that of Au@D2c. indicating the additional electronic interaction between Au NPs and pyridyl-N atoms of DL1 in Au@D1c as Lewis base sites (electron donors) that are absent for the Au@D2c DEMOF. Overall, the above results confirm that the degree of charge transfer from metal NPs to the framework with type-B defects is larger than that to the framework containing type-A defects, and both of them are higher than that to the pristine framework. proposed synergistic catalytic mechanism of D-glucose selective hydrogenation to sorbitol for these two different kinds of Ru NPs impregnated DEMOFs (Fig. 1).
As shown in Supplementary Fig. 31, the SEM images of all Ru@DEMOFs demonstrate the decrease of particle size along with increasing the feeding ratios (z) of DLx to TL (x =1, 2, z =10%, 30%, 50%).
It is a main reason that the fine peaks in the PXRD of DEMOFs and Ru@DEMOFs gradually disappeared and merged into broad bands along with increasing the feeding ratio (z) of DLx to TL (x = 1, z ≥ 30%; x = 2, z ≥ 50%).

/ 57
The impregnated amount of Ru NPs plays a significant role in catalysis of D-glucose selective hydrogenation to sorbitol, consequently, the catalytic performances of Ru NPs impregnated D0, D1a and D2a with different loading amounts of Ru element ranging from 1 to 5 wt% towards the D-glucose selective hydrogenation, have been investigated with all the other reaction conditions being fixed. As shown in Supplementary Fig. 32, the yields of sorbitol reach the maximum values when D0, D1a and D2a were impregnated by RuCl3 precursor containing 2.5 wt% Ru element, named as Ru@D0, Ru@D1a and Ru@D2a, demonstrating that the optimal impregnated content of Ru NPs of these MOFs supporters is 2.5 wt%. Consequently, all the other Ru@DEMOFs catalysts were impregnated with Ru NPs by using RuCl3 precursor containing 2.5 wt% Ru element. As shown in Supplementary Fig. 33  The maximum sorbitol yields of the selected investigated catalysts Ru@D0, Ru@D1a and Ru@D2a can be raised along with the increase of applied stirring rate when all the other reaction conditions were fixed. However, considering the tolerance of these catalysts, the operation safety and economy, all catalytic reactions in this work were conducted at 800 rpm.  Fig. 36 The curves of time-dependent D-glucose conversions (a) and selectivity to sorbitol (b) for the first four cycles of the reactions catalyzed by Ru@D0, Ru@D1a-c and Ru@D2a-c, respectively. Reaction conditions: D-glucose aqueous solution (25 wt%, 1.543 mol/L, 50 g), catalysts (1 g), hydrogen pressure (5 MPa), temperature (120 ºC) and stirring rate (800 rpm). Supplementary Fig. 36 shows the curves of time-dependent conversion of D-glucose and selectivity to sorbitol for the first four cycles of the reactions catalyzed by Ru@D0, Ru@D1a-c and Ru@D2a-c under the same optimized reaction conditions. The conversion of D-glucose and sorbitol selectivity for all these catalysts, except D1b obtaining the maximum sorbitol selectivity at 120 minutes, achieve the highest values after reacting for 150 minutes.

Samples
In order to gain an in-depth understanding of the reaction mechanism based on the role of the MIL-100-Cr MOF supporters, artificially implanted defects and impregnated Ru NPs as well as their synergetic catalytic effect on D-glucose selective hydrogenation to sorbitol, the catalytic performances of Ru NPs, pristine and defect engineered MIL-100-Cr supporters have also been investigated. The detailed discussions have been given in the section "Determination of the roles of each active species and their synergistic catalytic mechanism" of the main text.