The in-situ anchoring method for atomically dispersed Pt1/HAP includes only one step. In brief (Figure 1a), Pt and P precursors were premixed and precipitated by NH3·H2O at 0oC, and then crystallized at a certain temperature together with Ca precursor. The as-formed sample was calcined in air and reduced (in 10 vol% hydrogen flow) to obtain the Pt1/HAP
catalyst. X-ray diffraction (XRD) patterns show that standard HAP material [Ca10(PO4)6(OH)2] is successfully synthesized38, and no characteristic peak of Pt is observed, suggesting a highly dispersed Pt species (Figure S1). Energy dispersive X-ray spectroscopy (EDX) mapping shows that O, P, Ca and Pt elements are uniformly distributed in the HAP support (Figure 1b). Furthermore, aberration-corrected scanning transmission electron microscopy with high-angle annular dark-field (HAADF-STEM) in Figure 1c reveals that although few tiny Pt clusters are generated on rod-shaped HAP support, atomically dispersed Pt is absolutely dominant in Pt1/HAP. Control experiments confirm that the atomically dispersed Pt is formed on Pt1/HAP with the metal loading of 0.14 wt% (Table S1), while Pt nanoparticles are formed with higher Pt loading (Figure S2). Meanwhile, N2 physisorption results show that both HAP and Pt1/HAP possess typical isotherms of type-IV with the D-shaped tapered hole and similar Brunauer-Emmett-Teller surface area (approximately 50 m2/g) and pore volume (approximately 0.4 cm3/g), indicating that the incorporation of Pt does not affect the physical structure of HAP support (Figure S3 and Table S1).39 On these foundations, it can be preliminarily concluded that atomically dispersed Pt atoms are formed in Pt1/HAP via the in-situ anchoring method.
Coordination environment of Pt species on HAP was studied by extended X-ray absorption fine structure (EXAFS). From the k3-weighted Fourier transform (FT) spectra (Figure 2a), only one main peak located at approximately 1.8 Å is observed, corresponding to the Pt-O shell.40 Compared with Pt foil and PtO2, no Pt-Pt coordination peak (located at approximately 2.7 Å) is detected.40 The EXAFS fitting results including the first shell coordination parameters, structure and bond length were summarized in Table S2. Pt1/HAP catalyst displays a Pt-O coordination number of 5 in the first shell of the Pt atom, which was different from the results on standard Pt foil (12) and PtO2 (6) samples.41 Density functional theory (DFT) calculations were employed to better understand the coordination state of isolated Pt atoms. Periodical HAP (001) was selected as the support in the calculation model due to its easily exposed crystal surface (Figure S4-S5 and Figure 2b). The optimized structure (Figure S4) shows that an isolated Pt atom replaces Ca2+ and is surrounded by five O atoms in three PO43− ligand with the formation of five Pt-O-P linkages. Moreover, Pt-O bond lengths range from 1.96 to 2.09 Å, which is in good agreement with the EXAFS fitting results. In addition, the Pt and surrounding PO43− ligands display a quasi-square-planar
geometry, and the binding energy of Pt atom on the HAP (001) is as high as -3.51 eV.42 These indicate that PO43− ligand in the HAP support plays a vital role in the stabilization of isolated Pt atom. Figure 2c shows that the intensity maxima in a k space of 6-10 Å−1 and 4-8 Å−1 are assigned to Pt–Pt (in Pt foil) and Pt–O contributions (in PtO2), respectively. Pt1/HAP exhibits only one maximum intensity lobe near 6 Å−1 (Pt–O). The WT representations obtained for Pt1/HAP catalyst show a main lobe at low R values in the k space of 2-6 Å−1. However, the R and k values of Pt1/HAP catalyst are slightly different from that of PtO2. The evident elongation of the first lobe toward small R and k-values should be related to O atoms coordinated with isolated Pt atom in HAP structure. The suber lob in the large R values (2-3 Å−1) in the k space of 2-4 Å−1 can be assigned to the lighter element scatting such as O in PtO2. The suber lob is found in a similar k space for Pt1/HAP, but the lob shows much weaker and lower R value compared to PtO2, due to the different O environment in PO43− ligand compared to the one of PtO2. Consequently, EXAFS and HAADF-STEM strongly confirm that atomically dispersed Pt1/HAP catalyst is successfully prepared by in-situ anchoring method.
In the synthesis process of Pt1/HAP, low-temperature in-situ doping method was employed to avoid the introduction of other environmentally unfriendly or expensive solvents. We envisaged that HAP support could stabilize isolated Pt atoms for several reasons: 1) HAP is constructed by the phosphate tetrahedral group (PO43−), Ca2+ (surrounded by PO43−) and OH- (neutralizing positive charge). Part of the internal cation (mainly the Ca2+) can be replaced by partial positively charged Pt to keep the balance of charge and space structure, thus the stable crystal structure [Ca10−xPtx(PO4)6(OH)2] remains unchanged. Therefore, Pt precursor can be uniformly doped into the crystal of HAP during crystallization43,44. Meanwhile, due to the smaller radial size of Pt (94 Å) [compared with Ca2+ (100 Å)], the overall XRD peak position of Pt1/HAP shifts to a lower angle, resulting in the smaller lattice constant (Figure S1). 2) Fourier Transform infrared spectroscopy (FT-IR) spectra strongly confirm the typical and stable PO43− structure in the HAP support and Pt1/HAP (Figure S6). The existence of PO43− ligand with stable tetrahedral characteristics could provide electrostatic stabilization for single-atom species, leading to the formation of five Pt-O-P linkages.39 This PO43− coordination effect of HAP with the formation of Pt-O-P linkages creates an ideal prerequisite for the formation of atomically dispersed Pt. Under the influence of these two factors, the isolated Pt atoms can exist stably in Pt1/HAP.
X-ray absorption near edge structure (XANES) was further carried out to determine the electronic interactions between Pt atom and HAP (Figure 2d). The Pt white-line intensity for the Pt1/HAP catalyst near 11568 eV45 (1.59 eV) is lower than that of PtO2 (2.23 eV), but higher than that of Pt foil (1.24 eV). This indicates that isolated Pt atom is partial positively charged. Mulliken charge obtained by DFT calculation (Figure 2b) also confirms the electron transfer from Pt to PO43− ligands, resulting in a positive charge of 0.41 |e| for Pt atom and the negative charge of ~-1.0 |e| for PO43− ligands.40 Therefore, XANES and DFT calculation reveal that the stable PO43− ligand processes strong electronic-withdrawing function in Pt1/HAP. This partial positively charge state of Pt atoms caused by covalently bonded with PO43− may exhibit distinctive catalytic performance due to the formation of electronic coupled Pt-(O-P) linkages.
Subsequently, the structure of atomically dispersed Pt1/HAP catalyst on the reduction properties and chemical adsorption was analyzed. H2-temperature-programmed reduction (TPR) profile of unreduced Pt1/HAP only exhibits an extremely weak reduction peak for metallic Pt at approximately 200oC (Figure 2e-1 and Figure S7). Table S3 shows that the H2 consumption of Pt in Pt1/HAP is only 0.7 µmol/g, and the ratio of H2 to Pt (< 300oC) is only 0.05, which is much lower than that of Pt/AC (2.56). This indicates that the isolated Pt atoms have entered into the framework of HAP, and Pt-O-P bond is strong in the HAP structure.46 Meanwhile, Pt1/HAP and HAP exhibit almost the same NH3 temperature-programmed desorption (NH3-TPD) profiles (Figure <link rid="fig2">2</link>e-2 and Figure S8), suggesting that the introduction of atomically dispersed Pt atoms have no significant effect on the acidity of HAP. Nevertheless, Figure 2e-3 shows that the desorption peak of CO2-TPD of Pt1/HAP catalyst at ~220oC (medium basic sites) is lower than that of HAP, while the desorption peak at ~150oC (weak basic sites) increases. The chemisorption of CO2 is realized through the interaction with Lewis base of O sites.47,48 It is highly possible that the replacement of Ca2+ could significantly lower the basicity of O near Pt atoms, leading to the increase of weak basic sites on Pt1/HAP (Figure S8 and Table S4). On these foundations, it can be concluded that partial positively charged Pt atoms are coordinated and stabilized by PO43− ligand of HAP, resulting in the significant change of reduction property and basic site of Pt1/HAP.
To further understand the interaction between Pt atom and the structure of HAP, in-situ diffuse reflectance infrared fourier transform spectroscopy (DRIFTS) and CO-TPD were performed. The DRIFTS spectra of Pt/AC (Figure 2f-1) reveal two sets of CO absorption bands centered at 2100-2200 cm−1 and 2060-2100 cm−1, which could be assigned to the CO gas and CO molecules linearly adsorbed in an a-top geometry on Pt nanoparticles (NPs) respectively. In contrast, Figure <link rid="fig2">2</link>f-2 shows that Pt1/HAP displays weak CO adsorption peaks. Meanwhile, CO-TPD in Figure 2g clearly shows that Pt/AC has a stronger CO desorption peak located at ~110oC, while Pt1/HAP has almost the same CO adsorption quantity as the HAP support (Table S5). Commonly, the coordination environment near positively charged Pt could result in the suppression of adsorption of CO molecule along with the electron transfer between metal and support.49–51 XANES and DFT calculation in this work also demonstrated the electron transfer from Pt1 atom to PO43− ligand in Pt1/HAP. Consequently, DRIFTS and CO-TPD suggest the strong PO43−-coordination effect in Pt1/HAP.
Selective oxidation of primary hydroxyl group in polyols, which is schematically illustrated in Figure 3a, was conducted. Initially, oxidation of glycerol was performed under mild conditions, given that good catalytic performance was previously obtained on Pt based catalysts.52 Negligible products are obtained using the HAP support, while both Pt/AC and Pt1/HAP are active for glycerol oxidation. The final conversion and selectivity are presented in Figure 3b. The Pt mass based initial reaction rate of Pt1/HAP (1.8 mol/gPt/h) is almost four times higher than that of Pt/AC (0.4 mol/gPt/h). However, the turnover frequency (TOF) values on Pt/AC and Pt1/HAP are almost identical (Table S6), indicating that the site activity for glycerol oxidation is similar for the two catalysts, although the active sites are very different. The high initial reaction rate of Pt1/HAP based on Pt mass is mostly attributed to the high accessible Pt atom. Furthermore, the glyceric acid selectivity of Pt1/HAP (86.7%) is nearly twice as high as that of Pt/AC (47.8%), and a record high yield of glyceric acid up to 80.2% is obtained on Pt1/HAP. Considering the difference of by-product selectivity between Pt/AC and Pt1/HAP (Table S6), the remarkable glyceric acid yield is probably because that the isolated Pt atom active sites could selectively oxidize glycerol to carboxylic acids by
suppressing the C-C bond cleavage to form C1 and C2 acids. We further extended the application of Pt1/HAP to the oxidation of ethylene glycol, 1,2-propylene glycol and 1,2-butanediol (Figure 3c-e and Table S7-S9). Obviously, the selectivity of the corresponding hydroxy acid products is also significantly improved on Pt1/HAP compared to Pt nanoparticles on Pt/AC for all the reactions tested here. The results illustrate the atomically dispersed catalyst Pt1/HAP is active and selective for low carbon polyol oxidation to the corresponding acids in general. To the best of our knowledge, this is the first successful example of atomically dispersed catalysts with excellent catalytic performance in polyol oxidation system.
Moreover, the Pt1/HAP catalyst has substantially better stability in glycerol oxidation compared to Pt/AC. As shown in Figure 3f, the yield of glyceric acid over Pt/AC rapidly decreases with the increase of recycle times. In sharp contrast, the yield of glyceric acid over Pt1/HAP is almost constant (~ 80%) in the first ten recycles, and then it decreases slowly. It is worth mentioning that after 20 recycles (320 h), the yield of glyceric acid is still higher than 65%, which is even higher yields achieved than some fresh catalysts reported in literature (Table S10). Moreover, the deactivation factor of φ on Pt1/HAP (0.81) is almost twice that of Pt/AC (0.44) after 20 cycles. This indicates that Pt1/HAP effect could be used as a robust and efficient catalyst for the enhanced selective oxidation of low carbon polyols. The deactivation mechanism of catalysts was studied by characterization of used catalysts. The Pt particle size of Pt/AC increased significantly from 3.9 nm to 10.0 nm after 20 cycles, while only very small nanoparticles formed on Pt1/HAP (Figure S9). Inductively coupled plasma optical emission spectrometry (ICP-OES) revealed that the Pt metal in Pt/AC is leached seriously during the recycle test (the metal loading decreased from 0.84 wt% to 0.43 wt%), while the Pt metal loading remains almost unchanged (~ 0.14 wt%) before and after recycle test. The dissolution and re-desorption of Pt from the nanoparticles during the leaching process could be the main reason for the particle growth because the reaction temperature is too low for sintering.53 FT-IR spectra show that used Pt1/HAP (after 20 recycles) still exhibits almost the same transmission characteristic peaks belonging to PO43− as fresh Pt1/HAP. It suggests that the strong PO43−-coordination effect stabilizes atomically dispersed Pt atoms, thus avoiding the Pt metal leaching. Therefore, Pt1/HAP successfully realized noble metal free leaching under long-term test conditions in the polyol aqueous-phase oxidation system.
To further rationalize the excellent catalytic performance of the Pt1/HAP catalyst for the selective oxidation of low carbon polyols, DFT calculation, kinetic isotope effect (KIE) measurements, in-situ FT-IR and reaction kinetics were performed (Table S11, Table S12, Figure 4-5 and Figure S10-S16). Potential active sites for exploring the reaction mechanism were considered. Firstly, the active site formed by atomically dispersed Pt and O (belonging to PO43− ligand) is defined as Pt1-OPO43− switch (Figure 4a). The concept of switch in Pt1-OPO43− originates from the Pt-(O-P) linkage cyclic catalysis that partial bonding states between Pt1 and five O atoms could continuously break and regenerate during the oxidation reaction due to the stable tetrahedral characteristics of PO43− ligand (Figure S10a). On this basis, surface
O* from O2 direct dissociation and OH* from water assisted O2 activation (all derived from Pt1-OPO43− switch) could generate Pt1-O* and Pt1-OH* active sites. Subsequently, four types of reaction pathways, defined based on the active sites involved such as Pt1-O* assisted pathway I, Pt1-OH* assisted pathway II, Pt1-OPO43− switch assisted pathway III and coupled Pt1-OH*/Pt1-OPO43− switch assisted pathways IV were explored to reveal the formation of glyceric acid. As shown in Figure 4(b-1), three successive steps of glycerol activation namely dehydrogenation of O-H and C-H bonds as well as OH oxidation are involved. Obviously, pathway IV promoted by Pt1-OH*/Pt1-OPO43− switch is much more energetically favorable than that of pathway I, II or III. In this optimal pathway IV, the H2O assisted O2 activation exhibits lower transition state free energy than the direct O2 dissociative adsorption, inducing the formation of Pt1-OH* instead of Pt1-O* participating in the oxidation reaction. Subsequently, Pt1-OPO43− switch and Pt1-OH* are responsible for the activation of C-H and O-H bonds respectively, promoting the oxidation of glycerol to glyceric acid.
Notably, C-H bond activation in oxygen-containing intermediate (especially RCH2O* intermediate) with the highest free energy is the rate determining step (RDS) in each oxidation reaction pathways. Among the active sites tested, the Pt1-OPO43− switch significantly reduces the transition state free energy of C-H bond activation in RDS. To further verify the reliability of the RDS obtained by DFT calculation, KIE experiment was performed to study the C-H bond cleavage rate by investigating the initial reaction rates in the solution of H2O/D2O and CH2OHCHOHCH2OH (glycerol) /CD2OHCDOHCD2OH (deuterated glycerol). Figure 4c shows that a small KIE value of 1.03 is obtained by adjusting H2O/D2O, suggesting that the isotopically substituted atom broken in the reactant is the non-kinetic relevant step.54 In sharp contrast, a high KIE value (KH/KD = 2.70) for an exchange of glycerol to deuterated glycerol is obtained. It strongly proves that the activation of C-H bond of glycerol is the RDS, which fully supports the results from DFT calculations. Further analysis on reaction configurations and Mulliken charge in RDS (Figure S10) shows that the Pt-(O-P) linkage in Pt1-OPO43− switch plays a vital role in the activation of C-H bond. When the oxygenated substrate (O-R) is adsorbed on Pt1-OPO43− switch, previous bottom Pt-(O-P) linkage breaks with the formation of new top Pt-(O-R) linkage (switching on). Transition state (TS2) in RDS clearly shows that atomically dispersed Pt atom preferentially bonds to O in the primary OH group of the oxygenated intermediate, while the OPO43− in the Pt1-OPO43− switch tends to remove the H of C-H bond. Meanwhile, there is an obvious electron transfer between oxygenated intermediate and Pt1-OPO43− switch. C-H bond is acidic (positively charged) and OPO43− is basic (negatively charged). The basic OPO43− attacks the acidic C-H band leading to an attraction of H. Essentially, the strong interaction between atomically dispersed Pt and PO43− ligand triggers partial negatively charged O, thus significantly reconstructing the electronic local environment of Pt1-OPO43− switch. Due to this synergistic effect, Pt1-OPO43− switch is endowed with the dehydrogenation activity for the C-H bond activation (similar to oxidative dehydrogenation). With the desorption of product, top Pt-(O-R) linkage is broken and bottom Pt-(O-P) linkage is re-bonded (switching off), thus realizing the Pt-(O-P) linkage cyclic catalysis of Pt1-OPO43− switch (Figure S10a).
In addition, compared to the glycerol oxidation to dihydroxyacetone, the glycerol oxidation to glyceric acid is energetically preferred over Pt1/HAP (001) because of its lower transition state free energy (Figure 4(b-2)). Glycerol tends to oxidize primary hydroxyl rather than secondary hydroxyl to produce glyceric acid over Pt1/HAP. Furthermore, free energy diagram for the oxidation of glyceric acid shows that the existence of Pt1-OPO43− switch greatly inhibits the occurrence of deep oxidation reaction, especially the C-C bond cleavage (Figure S12). Compared to the Pt (111) model, the atomically dispersed Pt with partial positively charge could supply much fewer electrons for the formation of Pt-C bond during the cleavage of C-C bond due to the lack of Pt-Pt active sites (Figure S11), hence weakening the back-donating interaction between Pt and glyceric acid. In other words, the presence of strong PO43−-coordination effect in Pt1/HAP induces the formation of deficient electron state of Pt1. This Pt1 repulses the adsorbed C-C bond from Pt1-OPO43− switch, suppressing the C-C bond cleavage and leading to the significant improvement of hydroxyl acid product selectivity compared with Pt/AC. This is in good agreement with the experimental observation (Table S6-10).
In-situ FT-IR of glycerol oxidation over Pt1/HAP further confirmed the above conclusions of DFT calculation. Figure 5a shows that the peak of the intermolecular hydrogen bonds of glycerol in the 1660-1640 cm−1 gradually disappears with the increase of temperature, and the peak of C=O bond is dominant at high temperature (≥80oC), proving that glycerol converts into carboxylic acid products.55 Meanwhile, the νCO region in Figure 5b shows that the characteristic peaks of α interaction in the 1125-1075 cm−1 (belonging to alkoxy bond from primary hydroxyl and active sites) and γ interaction in the 1075-1000 cm−1 (belonging to primary hydroxyl) shift to the high wavelength with the increase of temperature, indicating the activation of primary hydroxyl group of glycerol.55 In contrast, the position of the characteristic peak of β interaction in the 1200-1125 cm−1 (belonging to the activation of secondary hydroxyl) remains unchanged. This further confirms that glycerol is mainly oxidized to glyceric acid via the activation of primary hydroxyl rather than secondary hydroxyl on Pt1/HAP (Figure 5c), which is consistent with the results of DFT calculation. In addition, the in-situ FT-IR results of glycerol conversion on Pt/AC are similar to those on Pt1/HAP (Figure S13), indicating that Pt-Pt and Pt1-OPO43− switch active sites have the same glycerol acid generation pathway.
Kinetic study of glycerol selective oxidation was further carried out on Pt/AC and Pt1/HAP at different temperatures. The reaction rates of Reaction 1-3 were estimated based on the formation rate of each compound (supplementary information 2.11). The Arrhenius plots for the Reaction 1-3 were presented in Figure S15. The activation energies and pre-exponential factors estimated based on the reaction rate as well as the TOF for Pt/AC and Pt1/HAP were summarized in Table S12. The apparent entropy changes (ΔS) were estimated from the pre-exponential factors based on the transition state theory (Table S12). The apparent entropy change reflects the freedom loss of the transition states of the rate determining step of each reaction compared to the reactants caused by the adsorption on the surface.56 Larger ΔS (absolute value) represents the stronger the bound state of adsorbed intermediate over the active center.57 Table S12 shows that Pt/AC and Pt1/HAP exhibit similar ΔS (~ -148 kJ/mol) for glycerol to glyceric acid (Reaction 1), suggesting that the adsorption strength of glycerol on Pt1-OPO43− switch active site is close to that of Pt-Pt active site. It resulted in a similar activation energy. In addition, the estimated apparent activation energy (Ea) for the oxidation of glycerol to glyceric acid over Pt1/HAP (Ea1 = 34.5 kJ/mol) and Pt/AC (Ea1 = 34.9 kJ/mol) is well consistent with that estimated apparent activation energy based on the Gibbes free energy profiles using the DFT calculation results (Figure S16). In a sharp contrast, the apparent ΔS of glycerol to glycolic acid and formic acid involving a C-C bond cleavage (Reaction 2) is significantly lower on Pt1/HAP (-35.8 kJ/mol) than that on Pt/AC (-122.6 kJ/mol), while the Ea is just the opposite. It means that the atomically dispersed Pt on HAP significantly destabilizes the transition state from glyceric acid toward C-C cleavage reaction, resulted in a significantly high activation energy. Combining the lower desorption energy of glyceric acid and higher activation energy of its conversion, a higher selectivity of glyceric acid is obtained. Consequently, reaction kinetics and DFT calculation fully indicate that the Pt1-OPO43− switch active site in Pt1/HAP not only displays similar catalytic activity to Pt-Pt active site in Pt/AC (for C-H activation), but also shows excellent selectivity for hydroxyl acid products via inhibiting the cleavage of C-C bond. This approach introduced in this contribution could effectively bridge the relationship between the microstructure of Pt1/HAP and the catalytic performance of low-carbon polyol oxidation to hydroxyl acid.
In summary, a novel in-situ synthesis method is established to stabilize isolated Pt atom on the HAP support. Significantly, this Pt1/HAP catalyst is efficient for the selective oxidation of low carbon polyols with excellent hydroxyl acids selectivity and catalyst stability. Multi-characterizations together with detailed kinetic study and DFT calculation demonstrated that partial negatively charged PO43− stably coordinates the atomically dispersed Pt atom, leading to an excellent resistance against leaching. Furthermore, the strong PO43−-coordination effect of Pt1/HAP constructs the switchable Pt-(O-P) linkage, i.e., Pt1-OPO43− active site. Due to this electronic coupled Pt1-OPO43− switch, the activation of C-H bond of the key oxygen-containing intermediate is promoted and the C-C bond cleavage of hydroxy acid product is well suppressed in the oxidation process. This work opens a new paradigm of rational design of atomically dispersed leaching free catalysts in production of fine chemicals, pharmaceutical intermediates, as well as in catalytic biomass conversion processes.