A versatile route to fabricate single atom catalysts with high chemoselectivity and regioselectivity in hydrogenation

Preparation of single atom catalysts (SACs) is of broad interest to materials scientists and chemists but remains a formidable challenge. Herein, we develop an efficient approach to synthesize SACs via a precursor-dilution strategy, in which metalloporphyrin (MTPP) with target metals are co-polymerized with diluents (tetraphenylporphyrin, TPP), followed by pyrolysis to N-doped porous carbon supported SACs (M1/N-C). Twenty-four different SACs, including noble metals and non-noble metals, are successfully prepared. In addition, the synthesis of a series of catalysts with different surface atom densities, bi-metallic sites, and metal aggregation states are achieved. This approach shows remarkable adjustability and generality, providing sufficient freedom to design catalysts at atomic-scale and explore the unique catalytic properties of SACs. As an example, we show that the prepared Pt1/N-C exhibits superior chemoselectivity and regioselectivity in hydrogenation. It only converts terminal alkynes to alkenes while keeping other reducible functional groups such as alkenyl, nitro group, and even internal alkyne intact.

S ingle atom catalysts (SACs), with maximum atomutilization and unique electronic and geometric properties 1 , are becoming a thriving research field because of their enhanced catalytic performance in a wide scope of industrially important reactions, e.g., selective hydrogenation of nitroarenes, alkenes and carbonyl compounds [2][3][4] , catalytic transformation of methane 5,6 , aqueous reforming of methanol 7 , hydroformylation of olefins 8 , olefin metathesis 9 , and oxygen reduction 10,11 . Various approaches have been utilized to prepare SACs, including the methods of impregnation/ion-exchange/coprecipitation 6,12,13 , defect engineering 14 , iced-photochemistry 15 , atomic layer deposition 16,17 , galvanic replacement 18 , hightemperature migration 19 , and high-temperature pyrolysis 20,21 . However, developing general protocols that can be used to easily synthesize of a wide variety of SACs is still highly desirable. For example, by Jung et al., theoretical calculations were conducted to predict universal principles for the electro-catalytic performance of SACs bearing various metal sites 22 . But the difficulty arises on verifying such predictions in experiments, as there are no general routes to prepare SACs with different center metals but similar supports and coordination environment. In addition, as predicted by Beller et al., the preparation of bi-/multi-metallic SACs is regarded as a next breakthrough because of their significant importance in the domino and tandem reactions 23 , but there are few reports for their synthesis, mainly due to the huge obstacle to keep various metallic elements with obviously different physical/ chemical properties coexisting in atomically dispersed states. Furthermore, comparative studies on the catalysis of different aggregation states, e.g., single atoms (SAs), nanoclusters (NCs), and nanoparticles (NPs), like the work by Zhang et al. on the Ru catalysts for CO 2 methanation 24 , received extensive attention. But, most of these studies rely on tuning the aggregation states by changing the metal loadings 25,26 , which did not conform to the single-factor-variable research method. Thus, a facile approach to regulate the aggregation states of metal species other than altering metal loading is desired.
Inspired by our previous work on the porous porphyrin polymers 27 and the work of Jiang et al. on SACs derived from metal-organic frameworks 21 , we report here a precursor-dilution strategy to prepare N-doped porous carbon supported SACs. In brief, tetraphenylporphyrin (TPP) with chelated metal cations, acting as the metal precursor, is co-polymerized with excess amount of free TPP as the diluent. By the dilution, the mean distance between metal atoms dispersed on the as-prepared polymer matrix becomes sufficiently large, preventing their aggregation during the subsequent high-temperature pyrolysis. Thus, SACs are obtained. Specially, the high chelating ability of TPP to various metal cations [28][29][30] (Supplementary Fig. 1) empowers this method to be applicable for fabricating a wide variety of SACs. Using this strategy, we have successfully conducted the synthesis of 24 types of SACs (i.e., M 1 /N-C, M = Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Ga, Zr, Mo, Ru, Rh, Pd, Ag, Cd, In, Sn, Er, W, Ir, Pt, Au, and Bi), including noble metals and non-noble metals. Furthermore, by varying the preparation conditions, e.g., precursor concentrations, metal precursors, and pyrolysis temperatures, we can obtain various materials: SACs with different surface atom densities (0.002-0.034 Pt·nm -2 ), bi-metallic SACs (Pt 1 -Sn 1 /N-C), and Pt catalysts with different aggregation states (Pt SAs, Pt NCs, and Pt NPs), respectively.

Results
Synthesis of Pt SACs with the precursor-dilution strategy. In this work, we use Pt 1 /N-C as an example to show the precursordilution strategy for fabricating SACs (Fig. 1a). First, a mixture of tetraphenylporphyrin platinum (PtTPP) and free TPP (PtTPP: TPP = 1:40, mol:mol) was dissolved in dichloromethane, and then co-polymerized by the addition of anhydrous AlCl 3 (Friedel-Crafts alkylation reactions) 31 . The as-obtained polymers were treated at 600°C under flowing N 2 gas, and nitrogen-dopedcarbon supported Pt SACs (i.e., Pt 1 /N-C) were obtained.
The transmission electron microscopy (TEM) image (Fig. 1b) and high-angle annular dark-field scanning transmission electron microscopy (STEM) image (Fig. 1c) revealed that there were no observable Pt NPs in the prepared SACs. The image taken by aberration corrected high-angle annular dark-field scanning transmission electron microscopy (AC HAADF-STEM) showed that individual Pt atoms highlighted by yellow circles in Fig. 1d were clearly visible (no bright dots can be observed in the underlying support of nitrogen-doped carbon without metal loading (i.e., N-C), Supplementary Fig. 2), resulting from the large difference in Z contrasts of the image for Pt and N/C. Thus, this image proved the presence of atomically dispersed Pt species. The X-ray diffraction (XRD) pattern of Pt 1 /N-C exhibited no peaks at 39.8°, 46.2°, and 68.5° (Fig. 1e, PDF#04-0802). This pattern resembled that for N-C and indicated the highly dispersed state of Pt species. The aggregation state of Pt species was also probed by extended X-ray absorption fine structure spectrometry (EXAFS, Fig. 1f). There were two notable peaks at 1.7 and 2.5 Å, similar to those in the spectrum of PtTPP, which can be ascribed to the Pt-N and Pt-N-C contributions 32,33 , respectively. It should be noted that the peak at 2.5 Å cannot be ascribed to the Pt-Pt bond (2.7 Å for Pt foil), which was further confirmed by the EXAFS fitting results of Pt 1 /N-C (Supplementary Fig. 3, Supplementary Table 1). These fitting results were in good agreement with the original curves, and the coordination number of the Pt with surrounding N atoms was 3.4, indicating that the Pt atoms were connected with three or four N atoms 34,35 . These results again corroborated the dominant presence of atomically dispersed Pt species evidenced by AC HAADF-STEM. As shown from the X-ray absorption near edge structure (XANES) spectra ( Supplementary Fig. 4), the energy absorption threshold of Pt 1 /N-C located between Pt foil and PtO 2 , implying the presence of positively charged Pt δ+ stabilized by adjacent N atoms in Pt 1 /N-C. The oxidation state of Pt species was characterized by X-ray photoelectron spectroscopy (XPS, see Fig. 1g). The Pt 4f peaks located at 72.4 and 75.7 eV can be tentatively ascribed to Pt 2+ with the presence of Pt-N bonds 36 . The inductively coupled plasma optical emission spectrometry (ICP-OES) analysis revealed that the actual Pt loading was 0.43 wt% (Supplementary Table 2 Table 2). High BET area (595 m 2 g -1 ) was found for Pt 1 /N-C (Supplementary Table 2), and it was reported that high-surface-area structures could facilitate the atomic dispersion of metal species 37 .
All the characterization results above lead to a conclusion that atomically dispersed Pt species were successfully synthesized on the support of N-doped carbon, by the precursor-dilution strategy. On the contrary, when PtTPP was used to make polymers without the diluent of free TPP, e.g., under the conditions of Pt-NPs/N-C(1:0), 100% PtTPP-based polymers and 3.31 wt% Pt loading (Supplementary Table 3), Pt NPs with 3.9 nm in diameter were formed using the same synthetic scheme ( Supplementary Fig. 5). Thus, the diluent is indispensable for successful fabrication of SACs.
The versatility of the precursor-dilution strategy. The precursor-dilution strategy is of significant flexibility and generality for SAC fabrication, as demonstrated below. All of the synthesized catalysts were characterized fully by TEM, STEM, XRD, ICP-OES, EA, and BET (see Supplementary Figs. 6-28, 30-33 and Supplementary Tables 4-30). Among them, TEM/STEM images and XRD patterns were used to preliminarily identify the aggregation states of metal species on the supports. ICP-OES was used to reveal the content of metal species, and the EA and BET measurements were used to probe catalysts' texture.
First, we could extend the precursor-dilution strategy to fabricate a variety of SACs using MTPP with different metals (M = Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Ga, Zr, Mo, Ru, Rh, Pd, Ag, Cd, In, Sn, Er, W, Ir, Au, and Bi) as the precursors and free TPP as the diluent. For most of the metals, the ratio of MTPP:TPP of 1:40 was used during the catalyst synthesis. But, there were exceptions. Some MTPPs (e.g., MnTPP and FeTPP) were found to easily leach in the polymerization process (under 80°C and in AlCl 3 ). Thus, the molar ratios of MTPP:TPP were increased to obtain SACs with meaningful metal loadings (>0.05 wt%). The samples with high content of Rh or Au tended to form NPs, so the molar ratios of RhTPP:TPP and AuTPP:TPP were decreased to 1:80 and 1:160, respectively, in order to obtain atomically dispersed metal species. Details of all the synthesis were provided in Supplementary Methods. AC HAADF-STEM images (Fig. 2) showed that all of the 24 SACs featured with atomically dispersed species on the supports, which were further confirmed by the corresponding EXAFS results with the absence of metal-metal bond (Supplementary Fig. 29). Among them, SACs of Cd, Bi, and Er have never been reported before, which may underpin the exploration of intriguing applications 38 . The EA and BET results revealed some similarity of material texture among all of the catalysts, i.e., with >420 m 2 g −1 BET areas and~5.0 wt% nitrogen content, due to the utilization of the similar preparation protocols.
Third, fabricating bi-metallic SACs (e.g., Pt 1 -Sn 1 /N-C) was also achieved with the same synthesis procedure of Pt 1 /N-C and the precursor molar ratio of PtTPP:SnTPP:TPP (1:1:40). The Ndoped porous carbon-based materials with 0.48 wt% Pt loading and 0.35 wt% Sn loading were obtained (Supplementary Table 30). This ratio of Pt loading and Sn loading (1.4:1) was in good agreement with the nominal ratio (1.6:1), based on the molar ratio of PtTPP:SnTPP (1:1) and atomic weight ratio of Pt:Sn (195.1:118.7). The AC HAADF-STEM image for Pt 1 -Sn 1 /N-C revealed the metal species atomically dispersed on the porous carbon supports (Fig. 4a). Corresponding element mapping analysis of Pt 1 -Sn 1 /N-C revealed that both Pt and Sn species were homogeneously distributed (Fig. 4b). The results of EXAFS (no Pt-Pt bond and Sn-Sn bond, Fig. 4c, d) were also indicative of the dominant presence of isolated Pt and Sn atoms deposited on the carbon matrix. These mutually authenticated results provided compelling evidence for the preparation of Pt 1 -Sn 1 / N-C.
Forth, we found that the pyrolysis temperature during the materials fabrication could influence the aggregation states of dispersed metal atoms. When the samples with the same molar ratio of PtTPP:TPP (1:40) were treated in different pyrolysis temperatures (i.e., 600, 700, and 800°C), the Pt contents and BET surface areas of them were close (~0.5 wt% and~600 m 2 g -1 , respectively, see Supplementary Tables 2 and 31-32), while the aggregation states of Pt species were changed from SAs (Pt 1 /N-C) to NCs (Pt-NCs/N-C, 1.1 nm) and NPs (Pt-NPs/N-C, 6.9 nm) (Fig. 1d, Fig. 5  The chemo-/regio-selectivity of Pt SACs in hydrogenation. After illustrating the facile synthetic routes of SACs with great versatility, we show here the unique catalytic properties of SACs (Pt 1 /N-C, with 0.43 wt% Pt loading) compared with NPs (Pt-NPs/N-C, with Pt-NPs of 6.9 nm in diameter and 0.52 wt% Pt loading) in hydrogenation reactions, which were previously illustrated as a promising solution in practical applications of SACs 1 . To our great delight, Pt 1 /N-C showed excellent chemoselectivity in the hydrogenation of 1-nitro-4-ethynylbenzene (with -C≡CH and -NO 2 ) and 1-ethynyl-4-vinylbenzene (with -C≡CH and -C=CH 2 ), as it only transformed alkyne groups to alkenyl groups and kept -NO 2 and -C=CH 2 intact (99% selectivity to 1nitro-4-vinylbenzene and 99% selectivity to 1,4-divinylbenzene at 20% conversion level, and 98% selectivity to 1-nitro-4vinylbenzene and 97% selectivity to 1,4-divinylbenzene at 100% conversion level, respectively, Fig. 6a, b and Supplementary Fig. 36a, b). In contrast, similar catalysis on Pt-NPs/N-C induced the formation of multiple products, resulting from the co-hydrogenation of -C≡CH and -NO 2 , and -C≡CH and -C=C, respectively. The Pt 1 /N-C catalyst permits the distinction between -C≡CH and -NO 2 /-C=C in hydrogenation mainly because of the good match between the relatively low catalytic activity of Pt SACs and high reactivity of terminal alkynes 41 .
It is generally accepted that there are two activation pathways for semi-hydrogenation of alkyne: (i) the terminal of the -C≡CH group interacts with metal surfaces leading to deprotonation, and then the -C≡C group becomes activated; (ii) the entire -C≡C group interacts with the metal surfaces and becomes activated 42 . In our system, because of possible steric hindrance effect (1.2 Å for C≡C bond length vs. 0.8 Å for Pt 2+ radius 43 ) of Pt 1 /N-C, the first pathway is more probable. Apparently, due to its absence of terminal hydrogen, internal alkyne cannot be activated and then hydrogenated on Pt 1 /N-C. On the contrary, Pt-NPs/N-C with much larger diameters than that of Pt 1 /N-C are able to interact with substrates with less steric hindrance effect 44 and then catalyze the hydrogenation of both terminal and internal alkynes. To verify our assumption, Pt 1 /N-C, Pt-NCs/N-C (1.1 nm), and Pt-NPs/N-C (6.9 nm) were employed under the same reaction conditions (Supplementary Table 33). As expected, the catalytic activities (i.e., turnover frequency, TOF, based on the metal dispersion 13 ) for the hydrogenation of internal alkynes on Pt-NCs/N-C fell between those on Pt 1 /N-C and Pt-NPs/N-C: 1-phenyl-1-propyne (0, 132, and 2946 h −1 ), 1-phenyl-1-pentyne (0, 93, and 2556 h −1 ), and 5-decyne (0, 2860, and 13300 h −1 ) on SACs, NCs, and NPs, respectively. The observation that the activities for the hydrogenation of internal alkynes increased with the increasing size of Pt species coincided quite well with our speculation that the unique group discrimination of terminal alkynes from internal ones on Pt 1 /N-C can be attributed to the geometric effect (see Supplementary Fig. 37). In addition, the stability of the Pt 1 /N-C catalysts in the hydrogenation of the four substrates, i.e., 1-nitro-4-ethynylbenzene, 1-ethynyl-4-vinylbenzene, 1-ethynyl-4-(phenylethynyl)benzene, and 1-(dec-1-yn-1-yl)-3-ethynylbenzene, respectively, was evaluated. As shown in Supplementary Fig. 38, recycling Pt 1 /N-C catalysts for five runs exhibited no essential decrease in catalytic activities and selectivities (~98%). Furthermore, no Pt nanoparticles or nanoclusters were found in TEM and STEM images ( Supplementary Fig. 39), and the corresponding AC HAADF-STEM images revealed that the Pt species maintained the atomically dispersed states after five catalytic runs. These results above suggested that the Pt 1 /N-C catalysts exhibited excellent recyclability under the aforementioned reaction conditions.

Discussion
In summary, a precursor-dilution strategy was developed to synthesize a series of SACs on N-doped porous carbon supports. This strategy is facile and versatile, and thus meets the requirements of the in-depth research nowadays. The Pt 1 /N-C SACs prepared with this strategy showed extremely high chemo-and regioselectivity towards terminal alkynes in hydrogenation. These findings are of significant importance in broadening the application of SACs, with the implication that SACs are able to achieve superior selectivity comparable to homogeneous catalysts and enzyme catalysts, for the catalysis of complex molecules.  Catalytic performance test. Catalytic hydrogenation reactions of various substrates, including 1-phenyl-1-propyne, 1-phenyl-1-pentyne, 5-decyne, 1-nitro-4ethynylbenzene, 1-ethynyl-4-vinylbenzene, 1-ethynyl-4-(phenylethynyl)benzene, and 1-(dec-1-yn-1-yl)-3-ethynylbenzene, were carried out in 10 mL stainless autoclave. Take hydrogenation of 1-nitro-4-ethynylbenzene on Pt 1 /N-C as example. The typical reaction conditions were 0.5 mmol 1-nitro-4-ethynylbenzene, 20 mg Pt 1 /N-C catalyst (0.43 wt%, Pt:1-nitro-4-ethynylbenzene = 1:1200, mol: mol), 2.0 mL methanol as solvent, 1.0 MPa H 2 and 50°C. After cooled and filtered, the reaction products were analyzed by GC (Shimadzu 2010 GC Plus) and GC-MS (Shimadzu GCMS-QP2010 Ultra). Selectivities were reported on a carbon basis, and TOF as molar substrate conversion rates per mole of surface Pt atoms per hour (h −1 ). For Pt 1 /N-C, the Pt dispersion was estimated to be 100%; for Pt-NCs and Pt-NPs, the dispersion (D) was estimated by the metal particle size (d) according to D = 1/d pt , respectively. Detailed reaction conditions were given in the footnotes of