The production of synthesis gas (syngas) from methane has mainly been achieved through the catalytic steam reforming of methane (SRM) at 800 o C or through noncatalytic homogeneous partial oxidation of methane (POM) at >1127 o C. [1][2][3][4] Steam reforming of methane at >800 o C has been the predominant catalytic process in the industrial production of syngas since the 1930s. [2][3][4][5] Supported Ni catalysts have been the main industrial catalysts for the steam reforming of methane which suffers from a serious issue of carbon deposition despite the fact that a significant number of catalysts of SRM were reported in literature. [2][3][4] Although the addition of more H2O to the reactor is used to avoid coke formation in methane reforming, it unfortunately creates new problems in terms of increasing the H2/CO ratio through the watergas shift reaction. A high H2/CO ratio consequently limits the direct use of the CO and H2 mixture in downstream processes, mainly Fischer-Tropsch synthesis. [2][3][4] Meanwhile, the noncatalytic POM is a homogeneous reaction processed at an extremely high temperature (>1127 o C), which consumes a significant amount of energy to maintain the reaction temperatures and thus results in a large carbon footprint and a need for the costly maintenance of the reactors.
The limitations of catalytic SRM and noncatalytic POM drove explorations into catalytic POM as early as the 1920s. [2][3][4] Compared to noncatalytic POM performed at >1127 o C, catalytic POM can be performed at reaction temperatures lower than 1127 o C, significantly reducing energy costs. Due to the mildly exothermic nature of POM at room temperature, performing POM at low temperatures is thermodynamically feasible as long as its kinetics can be promoted by a catalyst. Significant efforts in this area were made in the last few decades as reviewed in literature. 2,3,[6][7][8] Ni supported on alumina were the main catalysts for POM studied in the 1920s-1940s. 9,10 In the 1980s-1990s, Ni catalysts modified with different supports including Y2O3, Co3O4, ZrO2, UO2, ThO2, and rare earth oxides were then reported. [2][3][4] The main problem of these Ni-based POM catalysts is the formation of whisker carbon which clogs reactors substantially. 11- 17 In addition, by using CaAl2O4, AlPO4-5, alkaline metal oxide-modified alumina, or Ca0.8Sr0.2TiO3 as a support of Ni, the thermal stability of Ni-based catalysts was improved to some extent although the selectivity for producing CO and H2 was sacrificed. [18][19][20][21][22] Unfortunately, the issues of deactivation of nickel catalysts by carbon deposition during POM remain unresolved so far. 2 In the 1980s, Green et al. reported Ln2M2O7 (Ln=La, Pr, Eu, La, Dy, Yb, Ba, Bi, and M=Ru, Ir, Pt, Rh) catalysts exhibiting high activity and selectivity which led to a renaissance in catalytic POM studies. 4,10,23,24 Their early reports on Ln2M2O7 inspired a large number of studies on precious metals and related catalysts since the 1980s; 2 Pt, Pd and Rh with a loading of 0.5wt%-5.0wt% on different types of supports including alumina foam, mixed oxides were investigated for POM. Through these studies, an ordering of activity impacted by a support was suggested: Y2O3 > La2O3 > MgO > Al2O3 > SiO2 and Y2O3 > ZrO2 > TiO2. 25 These supported noble metals are highly active for POM although their thermal stabilities were not reported in detail. 2 Different from these Ni-based and noble metal-based catalysts, molybdenum carbide was reported by York et al; [26][27][28] this type of catalysts has an activity similar to noble metals; however, it readily deactivates through a two-step process under reaction conditions, where molybdenum carbide is first oxidized to molybdenum oxide and then vaporization occurs at high temperatures under 1 atm pressure. 2,3,6,7,[26][27][28] Despite these significant efforts in the development of different types of POM catalysts, these reported catalysts have still suffered from either carbon deposition and/or a loss or a harmful restructuring of the active catalyst components during POM.
T h e literature has demonstrated that single-atom catalysts of many transition metals including Pt, Pd and Rh exhibit distinctly different catalytic performances for a number of reactions at relatively low temperature. [29][30][31][32][33][34][35][36][37][38][39] Compared to the continuously packed metal atoms on t he surface of a metal particle, a singly dispersed metal atom, or an isolated metal atom, anchored on surface of a nonmetallic support, typically an oxide, is in a distinctly different chemical and coordination environment and thus exhibits a very different electronic state. For instance, Rh atoms were anchored to the internal surface of a micropore of the aluminosilicate ZSM-5 through proton exchange, making single-atom catalysts, Rh1@ZSM-5; these reported single-atom sites immobilized in the 0.56 nm wide micropores of ZSM-5 are active for synthesis of acetic acid and methanol through coupling of CH4, CO and O2 in the microporous reactor where reactant molecules are delivered through the diffusion of dissolved gas molecules in aqueous solution under a high pressure gas phase at low temperatures of 150 o C. 40,41 These singly dispersed metal atoms, or single atoms, on a nonmetallic surface are typically under-coordinated, either electronically deficient or rich, which offer them a high tendency to bind to a reactant molecule or intermediate. However, one major challenge of single-atom catalysts is their short durability at high catalysis temperature and low reusability as their singly dispersed single-atom sites readily sinter at elevated temperatures 42,3,43,44 . Thus, most single-atom catalysts function well at room temperature or a relatively low temperatures (≤300 o C) but not at high temperatures (≥450 o C), although a significant portion of chemical productions of chemical and petrochemical industries are performed at high temperatures driven by thermodynamics or/and kinetics. Notably, the high temperature defined here is different from that of high temperature processes in industries which run at 800 o C or higher.
Herein we report a high-temperature single-atom catalyst, Rh1/TiO2 which is highly active for POM at 650 o C. During the high temperature catalysis at 650 o C, its Rh1 atom coordinates with five O atoms, a nearly saturated coordination shell. This single-atom catalyst exhibits long durability and extraordinary catalytic performance at 650 o C for at least 240 hrs and high reusability from 20 repeat runs in the temperature range of 450-650 o C.
The single-atom catalyst, Rh1/TiO2 was prepared through a deposition precipitate method modified for the preparation of Rh1/TiO2. Rh(III) cations were introduced to surface of TiO2 nanoparticles followed by drying at 60 o C in air overnight and then annealing at 650 o C in air for 4 hrs. Based on ICP-AES measurements, the actual loading of Rh on TiO2 is 0.037wt%. Details of the preparation can be found in the section of Experimental Methods and Computational Approaches. For convenience, 0.037wt%Rh/TiO2, Rh1/TiO2 and Rh1O5/TiO2 are interchangeably used to refer the same catalyst in this article, depending on which piece of specific information (concentration, single dispersion, or coordination) of this catalyst is emphasized.
Compared to the inactive TiO2 nanoparticles for POM in the temperature range of 300-700 o C (Figure S1a), the Rh1/TiO2 catalyst is highly active for POM in the range of 450-650 o C under a flow of a mixture of 25 ml/min 10% CH4 and 25 ml/min 5% O2 (Figure S2). At 650 o C, 50 mg of 0.037wt% Rh/TiO2 has a conversion of 80% for CH4 (Figure 1a). The selectivities for producing H2 and CO a t 6 5 0 o C are 96% and 98%, respectively (Figures 1b and1c). This catalytic activity and selectivity on Rh1/TiO2 remain after over 240 hrs of time on stream (Figure 1). There was no interphase diffusion limit since the conversion of CH4 remained nearly the same while the flow rate of the mixture of CH4 and O2 varied in the range of 30-100 ml/min (Figure S3). The observed by-products are mainly CO2 and H2O which are formed from a competing reaction channel, the complete oxidation of methane; here the selectivities for forming by-products CO2 and H2O are 4% and 2%, respectively. In the kineticscontrolled regime, the Rh1/TiO2 catalyst exhibits an extraordinarily high turnover rate of 2428 H2 molecules produced from a Rh1-based site per minute at 650 o C. Details of calculation of in catalytic activity between pure TiO2 (Figure S1) and Rh1/TiO2 (Figure 1) clearly shows the crucial role of the anchored Rh1 atom although the loading is as low as only 0.037wt%. To test the reusability of the catalyst, catalysis was performed from 300 o C to 650 o C on Rh1/TiO2, followed by cooling to 300 o C before the next cycle. for a total of 20 cycles on Rh1/TiO2. Compared to the catalytic performance of the 1 st cycle (Figure S2a), the decay in catalytic activity and selectivities at the 20 th cycle (Figure S2b) is less than 4% over a period of about 360 hrs. This suggests quite a high reusability for Rh1/TiO2. The single-atom catalyst Rh1/TiO2 also exhibited quite a long durability for POM at 650 o C for at least 240 hrs. As shown in Figure 1a, the conversion of CH4 (80%) remained without any decay at the 240 th hr compared to the 0 th hr. In addition, the selectivity for production of H2 and CO remained at 96% and 98%, respectively, under 650 o C for the whole duration with negligible changes (Figures 1b and1c). The molar ratio of produced H2/CO remained at 2.0 during the catalysis at 650 o C for 240 hrs (Figure 1d). The preservation of catalytic activity, selectivities and H2/CO molar ratio in Figure 1 confirms the high durability of the Rh1/TiO2 catalyst at the high temperature of 650 o C, suggesting that the chemical and coordination environments of Rh atoms of this catalyst remained unchanged (Figure 1), which is further supported by in-situ and operando characterizations of 0.037%Rh/TiO2 under the catalytic condition described below.
The chemical and coordination environments of Rh atoms of 0.037wt%Rh/TiO2 at the atomic scale were investigated through in-situ XANES and EXAFS studies. 20 mg of 0.037wt%Rh/TiO2 (Catalyst #1) was loaded into the XAS reactor (XANES and EXAFS); a mixture of 10 ml/min 10% CH4 and 10 ml/min 5% O2 was introduced; thus, the gas hourly space velocity (GHSV) in the XAS reactor of in-situ XAS studies at the beamline end-station was the same as that of catalytic measurements performed at our catalysis lab. The catalyst was heated to 650 o C in the flow of the aforementioned reactant mixture. High activities and selectivity of the catalyst during the in-situ XAS studies at 650 o C were confirmed through analysis with an online mass spectrometer analyzing the composition of gas received through a glass capillary installed on the outlet of the XAS reactor. During the in-situ studies, the generated fluorescence due to absorption of Rh K-edge was collected when the catalyst was in the mixture of 10% CH4 and 5% O2 upon catalysis at 650 o C for 4 hrs. The Rh K-edge of Rh1/TiO2 (Catalyst #1) at 23231.0 eV (red line in Figure 2a) is close to that of Rh2O3 nanoparticles supported on Al2O3 (blue line in Figure 2a), 48 suggesting that Rh atoms of 0.037%Rh/TiO2 are in an oxidized state. In addition, the observed energy spectrum of Rh K-edge of Rh1/TiO2 (red line in Figure 2a) is quite consistent with the energy spectrum of Rh K-edge (red line in Figure 3) theoretically simulated by using the structural model of single atom catalyst (Figure 2e) optimized in computational studies.
The proposed structural model of 0.037%Rh/TiO2 (Figure 2e) will be described and discussed in detail later in this article.
In r-space spectra of Rh K-edge of 0.037%Rh/TiO2 (Figure 2b4), a main peak at 1.50 Å was observed and readily assigned to Rh-O bonds, consistent with Rh-O bonds of the reference sample, Rh2O3 nanoparticles supported on Al2O3 (Figure 2b2) and those reported in references. 48- 50 Here 1.50 Å is the Rh-O distance in r-space spectrum before a phase correction. We note all the distances of EXAFS studies described in the following paragraphs are distances before phase correction except for a specific note. Compared to the observed Rh-O-Rh peak at 2.63Å from the reference sample (Rh2O3/Al2O3 in Figure 2b2), no such a Rh-O-Rh peak was observed in the spectra of 0.037wt%Rh/TiO2 (Figure 2b4). In addition, the spectrum of Rh1/TiO2 (Figure 2b4) does not have the peak of Rh-Rh bonds of Rh foil observed at 2.35 Å (Figure 2b1). Thus, neither peak of Rh-O-Rh bond of Rh2O3 nanoparticles (2.63 Å) nor peak of Rh-Rh bond of metal Rh nanoparticles (2.30 Å) was observed in the r-space spectrum of 0.037 wt% Rh/TiO2 (Figure 2b4). The difference in coordination environment among Rh1/TiO2, Rh2O3/Al2O3 and Rh foil was supported by wavelet transform (WT) of EXAFS data of references samples (Rh foil and Rh2O3/Al2O3) and the catalyst 0.037wt%Rh/TiO2 (Figure 4). In Figure 4a, there is one major contour maximum observed in the WT plot of Rh foil at (11.6 Å -1 , 2.30 Å), which is attributed to the first shell Rh-Rh coordination pathway in foil sample. In the WT plot of Rh2O3 reference sample (Figure 4b), there is one major contour maximum at 2e and2f); black curve is the simulated spectrum of Rh nanoparticles consisting of 79 atoms (Rh79) optimized with DFT (Figure S28b). The simulated spectra of Rh1/TiO2 and Rh nanoparticle are shown to be consistent with the observed spectra of Rh1/TiO2 (red curve in Figure 2a) and Rh foil (black curve in Figure 2a). 650 o C (Figure 4c), only one maximum at (7.6 Å -1 , 1.5 Å) is found for Rh-O coordination pathway, suggesting the lack of the second shell Rh-O-Rh on Rh1/TiO2. In addition, the contour maxima of Rh-Rh coordination which had been found in Rh foil and Rh2O3/Al2O3 are not observed in the WT plot of Rh1/TiO2, suggesting the lack of Rh-Rh coordination in the catalyst during POM. Figure 4 suggests that Rh atoms are singly dispersed in 0.037 wt% Rh/TiO2 instead of forming Rh metal NPs or Rh2O3 NPs, further evidenced in the DRIFTS, CO chemisorption measurements, and AP-XPS studies to be discussed in the following paragraphs.
It could be argued that the peak of Rh-O-Ti of Rh1/TiO2 at 2.50 Å in Figure 2b4 could not be fully distinguished from the peak of Rh-Rh of Rh foil at 2.30 Å in Figure 2b1 due to the limited resolution of these r-space spectra, the following characterizations allow us to deduce that no Rh-Rh bonds contributed to this peak of Rh-O-Ti of Rh1/TiO2 at 2.50 Å in Figure 2b4. First, compared to the vibrational feature of CO adsorbed on Rh nanoparticles at 1872 cm -1 and 2060 cm -1 in the spectra of Diffuse Reflectance Infrared Fourier Transform Spectroscopy ( DRIFTS) (Figure S10), the lack of the two peaks in the DRIFTS spectrum of the used Rh1/TiO2 (Figure 5) suggests that no Rh nanoparticles formed on surface of 0.037 wt% Rh/TiO2 during catalysis at 650 o C since 1872 cm -1 and 2060 cm -1 are attributed to stretching of CO molecules adsorbed on surface of Rh metal nanoparticles (Figure S10) or the Rh single crystal. 51 The detailed pretreatment and collection of data of DRIFT spectra were described in Section 5 in SI. Peaks at 2090 cm -1 and 2022 cm -1 in the DRIFTS spectrum of Rh1/TiO2 in Figure 5 are assigned to asymmetric and symmetric stretching of CO of two CO molecules bound to a single Rh atom; this binding configuration is called geminal-dicarbonyl binding configuration in which two CO molecules are simultaneously bound to a metal atom. The vibrational signature in Figure 5 is very similar to that of single-atom catalysts reported in literature. 8.9 For Rh nanoparticles, such a geminal-dicarbonyl binding configuration should be a quite minor or even not present since Rh atoms on the surface of a Rh nanoparticle are continuously packed. Thus, DRIFT studies suggests that no Rh NPs were formed on TiO2 of 0.037wt%Rh/TiO2 during catalysis at 650 o C. Secondly, the measurement of the dispersion of Rh atoms of 0.037%Rh/TiO2 by CO chemisorption showed that >97.5% of Rh atoms of the used 0.037%Rh/TiO2 are singly dispersed on TiO2; this further suggests that the fraction of Rh atoms of 0.037%Rh/TiO2 catalyst that could potentially form Rh nanoparticles is lower than 2.5%. The single dispersion of Rh atoms on surface of 0.037%Rh/TiO2 catalyst is suggested from CO chemisorption measurement, the vibrational signature of CO chemisorbed on 0.037 wt% Rh/TiO2 and in-situ EXAFS studies. It is further supported with the lack of Rh nanoparticles on the used catalyst (Figures 2g and2h) and the observation of singly dispersed Rh atoms (Figure 2i). The peak at 2.50 Å in r-space of Rh K-edge of 0.037wt% Rh/TiO2 (Figure 2b4 or Figure 2c) was fitted to Rh-O-Ti. Figure 2c presents the fitted r-space spectrum (in red) which matches the As most users are usually only granted beam time of 1-3 days, continuously tracking catalyst structure with EXAFS in a period of 10 days was not realistic. The following in-situ XAS studies including XANES and EXAFS of Catalyst #2 was performed. To be specific, Catalyst #2 is 0.037wt%Rh/TiO2 which has already been used for POM at 650 o C at catalysis lab before XAS studies. I n -s i t u X A N E S a n d E X A F S s t u d i e s o f Catalyst #2 were performed under the same in-situ condition as Catalyst #1. Fittings the r-space spectrum marked with red dashed lines in Figure 2b4. These periodic peaks are in good agreement with the periodic peaks in r-space spectrum of Ti K-edge of Rh1/TiO2 (Figure 2b3) and those periodic peaks of bare TiO2 (Figure S14a). Based on literature, 52,53 the appearance of the two peaks of Ti Kedge at 2.54 Å and 3.30 Å in Figure S14a was contributed from the periodic arrangement of Ti atoms (Ti-O-Ti α and Ti-O-Ti β ) in the crystal lattice of pure TiO2 (Figure 6). Compared to rspace spectrum of Ti K-edge of Ti atoms of TiO2 (Figure 2b3), a very similar periodic pattern in r-space spectrum of Rh K-edge of Rh atoms of Rh1/TiO2 (Figure 2b4) was observed. This similarity suggests that these Rh1 atoms substituted for Ti atoms in the surface lattice of TiO2
(Figure S13f).
The chemical state of Rh atoms of Rh1/TiO2 during POM at 650 o C was tracked with Ambient Pressure X-ray Photoelectron Spectroscopy (AP-XPS). 54 The formation of CO and H2 on 6 was used as the surface structure in the simulation of the catalytic mechanism of POM on the single-atom catalyst, Rh1/TiO2.
To shed light on the POM mechanism, the adsorption and transformation of CH4 on Rh1/TiO2 transfer to CH4 (Figure 7c). This specific difference can be found in the calculated Bader charges (Figure S16). The substantial charge transfer from Rh1 of Rh1/TiO2 to C atom of CH4 is supported by the elongation of C-H from 1.096 Å of free CH4 molecule to 1.115 Å of the chemisorbed CH4 molecules on Rh1 atom; however, the C-H bond distance of CH4 adsorbed on bare TiO2 (Figure S17) is 1.099 Å which is nearly the same as the free molecule.
To further understand the significance of Rh1 in orbital coupling between CH4 and Rh1, partial density of states (PDOS) of CH4 and Rh atom of Rh1/TiO2 with chemisorbed CH4 were calculated. As shown in Figure 7e, molecular orbitals of CH4 (black lines) mix most strongly with the 𝑑 z 2 orbital of the substituting Rh1 atom (red lines) since electron density of 𝑑 z 2 orbital of Rh1 can back donate to unoccupied orbital of the adsorbed CH4 molecule. This back donation is consistent with the electron-density-difference plot in Figure 7b that shows a significant charge depletion around the Rh1 atom and the corresponding charge enhancement around CH4. The electron-density difference can be evaluated through the color code of Figure 7d. Moreover, the natural bond orbital (NBO) analysis in Figure 7f shows a lower C-H σ-σ* occupancy on Rh1/TiO2 than that on bare TiO2, suggesting that Rh1 of Rh1/TiO2 is quite favorable for activating the C-H bond.
Thus, both electron density calculation and simulation of orbital coupling of CH4 and Rh1 show that Rh1 is favorable for binding CH4 and activating the C-H bond.
Extensive computational studies were performed for optimizing intermediates and searching for transition states of POM performed on the optimized structure of Rh1O5/TiO2 shown in Figure 6. These theoretical simulations proposed a complete catalytic cycle consisting of 26 elementary steps starting from chemisorption of reactants to form two CO and four H2 molecules on Rh1/TiO2 (2 CH4+O2→2CO+4H2). A complete energy profile of intermediates and transition states of the entire pathway were plotted in panel 1 of Figure 8. All intermediates are listed in Figure S18 and Figure S19, respectively. Figures S20 andS21 present the structures of transition states of this complete reaction pathway of POM on Rh1/TiO2. Notably, the activation of molecular O2 on Rh1/TiO2 is distinctly different from that on an oxygen vacancy of a transition metal oxide. The activation was not performed at the beginning of the cycle as the surface at the beginning does not present oxygen vacancies and the adsorption of molecule O2 on surface without oxygen vacancies is weak. Even if molecular O2 is dissociated, the formed O atoms can thermodynamically favorably return to gas phase. Step t to u in Figure S19 is the activation of reactant O2 through insertion of a molecular O2 between Rh1 atom and its nearest Ti atom in the intermediate t. Notably, this insertion step is kinetically barrierless and thermodynamically exothermic (panel 1 in Figure 6), suggesting a different path for activating molecular O2 on Rh1/TiO2 in contrast to dissociation of O=O on oxygen vacancy on transition metal oxides. Different from the barrierless kinetics of molecular O2 on this single atom catalyst Rh1/TiO2, kinetical barrier for activation of O2 on Co3O4, 59 TiO2, 60 CeO2, 61 is in the range of about 0.5-2.0 eV, which varies at different faces and vacancies. Compared to the activation of O2 on an oxygen vacancy of transition metal oxides, this distinctly different barrierless activation mechanism of O2 on the site of Rh1/TiO2 demonstrated the significance of Rh1 in the high activity of Rh1/TiO2 for POM. The presence of both undercoordinated Rh1 atom and its nearly oxygen vacancies make dissociative adsorption of molecular O2 significantly downhill with little or no barriers, thus stabilizing the dissociated O atoms even at high temperatures.
As can be seen from panel 2 of Figure 8
In order to understand how the singly dispersed Rh1 site could be different from its counterpart at high temperature, a surface consisting of closely packed Rh atoms was prepared for performing comparable studies on durability of catalyst under the same condition as Rh1/TiO2. Such a surface can be found on Rh nanoparticles supported on TiO2, termed Rh NP/TiO2. The precursor of Rh NP/TiO2 was prepared through impregnation which is a method different from the deposition-precipitation method used in preparation of Rh1/TiO2. In the preparation of Rh NP/TiO2, its precursor was reduced at 500 o C for 2 hrs and 650 o C for 2 hrs in 5% H2 instead of calcination in air at 650 o C used in the preparation of Rh1/TiO2. Details of preparations of Rh NP/TiO2 can be found in the section of Experimental Methods and Computational Approaches. This Rh NP/TiO2 catalyst is active for POM in the temperature range of 300-650 o C (Figure S23).
Statistical counting of over 500 Rh NPs of the used catalyst Rh NP/TiO2 shows that the average size of the Rh NPs supported on TiO2 after catalysis at 650 o C for 240 hr is approximately 3.0 nm (Figure S24). In the following sections, we use the labels Rh NP/TiO2 and 6.3wt%Rh/TiO2 interchangeably, referring to the same catalyst. Compared to 50 mg of 0.037wt%Rh/TiO2 (Rh1/TiO2) at 450 o C, 500 o C, 550 o C or 650 o C (Figure S2), 50 mg of 6.3wt%Rh/TiO2 exhibits similar conversion of CH4 at these corresponding temperatures, respectively. However, they exhibit quite different catalytic selectivities for producing H2 or CO (Figure S2 versus Figure S23). Notably, the similar conversion of CH4 exhibited by the two catalysts (50 mg Rh NP/TiO2 versus 50 mg Rh1/TiO2) at a temperature such as 650 o C does not suggests any similarity in activity of their catalytic site. As TiO2 is completely inert for oxidation of CH4 in 300-700 o C (Figure S1), the conversions of CH4 of 50 mg of 6.3wt%Rh/TiO2 and 50 mg of 0.037wt%Rh/TiO2 depend on the numbers of Rh atoms exposed on catalyst surface. For 50 mg of 0.037wt%Rh/TiO2, the amount of the exposed Rh atoms on the surface to catalyze the conversion of CH4 is 1.7510 - 7 mol; however, for 50 mg of 6.3wt%Rh/TiO2, the amount of the exposed Rh atoms on the surface to catalyze conversion of CH4 is 1.0710 -5 mol. Calculations of the amounts of exposed Rh atoms on the catalyst surfaces were described in Section 6 of SI. The ratio of Rh atoms precipitated into the POM on 50 mg of 6.3wt%Rh/TiO2 to that on 50 mg of 0.037wt%Rh/TiO2 is about 61:1. Thus, although the conversions of CH4 on the two catalysts are seemingly similar, the activities of each exposed Rh atoms of the two catalysts are in fact distinctly different. The difference in catalytic activity between the two catalysts was further confirmed by measuring the turn-over rates under the kinetics-controlled regime. As described in Section 4 of SI, the reaction rate of each catalyst was measured when the conversion of CH4 is less than 10%. With the measured reaction rate at 650 o C under kinetics-controlled regime, the turn-over rates (TOR) of the two catalysts were calculated in Section 7 of SI. The TOR of H2 on a Rh atom of single-atom catalyst 0.037wt%Rh/TiO2 is 2428 H2 molecules per minute but it is only 35.8 H2 molecules per minute on an exposed Rh atom of Rh nanoparticle catalyst, 6.3wt%Rh/TiO2. Details of the calculations of TORs using data of kinetics studies are described in Sections 7.2 and 7.3 in the SI.
Demonstrably, the activity in terms of TOR of the single-atom catalyst Rh1/TiO2 at 650 o C is much higher than the Rh nanoparticle catalyst, Rh NP/TiO2, at the same temperature.
Other than the much higher activity of Rh1/TiO2 catalyst at 650 o C than Rh NP catalyst at this temperature, the high-temperature single-atom catalyst, Rh1/TiO2 exhibits a much longer durability in catalytic performance than the Rh NP/TiO2 (Figure 1 versus Figure S27). It is necessary to highlight that the comparison of the difference in durability of catalytic performance is for obtaining a fundamental understanding of the reason of the observed high-temperature durability of single-atom catalysts; and not for developing a commercial POM catalyst for production of syngas in industries.
Compared to the nanoparticle catalyst (Rh NP/TiO2), the single-atom catalyst (Rh1/TiO2) exhibits superior resistance to carbon formation. The continuously packed Rh atoms on surface of Rh nanoparticles of Rh NP/TiO2 (Figure S28b) make CHx (x=0-2) species adsorbed on two adjacent Rh atoms couple readily toward chain propagation leading to the formation of a coke layer. As seen from XPS studies (Figures 10e, 10f and 10g), thick carbon layers were formed on the used Rh NP/TiO2 after POM at 650 o C for 240 hrs while negligible amounts of carbon were formed on Rh1/TiO2 after 240 hrs of POM at 650 o C (Figure 10a, 10b and10c). The atomic ratio of carbon atoms of surface carbon layers to all titanium atoms in surface region of used Rh NP/TiO2 is about 2.5 (Figure 10d), suggesting the thickness of carbon is at least several atomic layers. It is much larger than that of Rh1/TiO2 used at 650 o C for 240 hrs. The experimentally observed facile formation of carbon layers on Rh NPs is supported by the quite high adsorption energy of atomic carbon on Rh atoms of Rh nanoparticle surface uncovered in our computational studies (Figure S29b). Compared to the Rh atom on a Rh nanoparticle of 6.3wt%Rh/TiO2, adsorption energy of atomic carbon on the singly dispersed Rh1 atoms of 0.037wt%Rh/TiO2 is lower by 2.8 eV (Figure S29); this significant difference in adsorption energy of carbon atom to Rh atoms of the two catalysts (Rh1/TiO2 versus Rh NP/TiO2) results from the distinctly different coordination environment between Rh1 of Rh1/TiO2 and Rh of Rh NP/TiO2 (Figure S29). In terms of Rh1/TiO2, the singly dispersed Rh1 atoms on TiO2 does not provide opportunity for two CHx (x=0-2) species adsorbed on spatially isolated Rh1 atoms to couple since CHx species on Rh1/TiO2 are spatially separated. Thus, Rh1/TiO2 exhibits high resistance to formation of coke at 650 o C for a period of at least 240 hrs.
The high adsorption energy of atomic carbon on the surface of the metal Rh nanoparticle can rationalize the progressive decay of conversion of CH4 to 60% at 105 th hr and to nearly only 20% at 240 th hr (Figure S27a). Along with the decay of catalytic activity, selectivities largely decreased to 86% for H2 and 84% for CO at 105 th hr and to 42% for H2 and 29% for CO at 240 th hr, respectively (Figures S27b andS27c). This significant decay of catalytic activity of 6.3 wt% Rh/TiO2 is consistent with the observation of large amount of carbon formed on 6.3 wt% Rh/TiO2 after catalysis for 240 hrs at 650 o C (Figure 10h).
The C/Ti atomic ratio of the 6.3 wt% Rh/TiO2 catalyst used at 650 o C for 240 hrs is about 45 times of that of the single atom catalyst Rh1/TiO2 catalyst used at the same temperature for the same amount of time (Figure 10d), rationalizing the significant decay of catalytic performance of 6.3wt%Rh/TiO2 results from the formation of carbon layers on Rh nanoparticles which blocked the active sites. As reported in literature, 2,4,[11][12][13][14][15][16][17] there are two types of carbons formed on POM catalysts including encapsulated carbon which blocks catalytic sites and thus obviously degrades catalytic performance, and whisker carbon which forms on a catalyst nanoparticle but sinks to one end of the catalyst nanoparticle and thus does not directly poison the catalyst surface but results in serious reactor clogging. As the reactor of 6.3wt%Rh/TiO2 catalyst was not clogged after catalysis at 650 o C for 240 hr, likely the encapsulated carbon instead of whisker carbon was formed on Rh NPs of 6.3wt%Rh/TiO2. As these encapsulating carbon layers on the catalyst surface block the active sites, the catalytic activity clearly becomes degraded. The mechanism of deactivation of 6.3wt%Rh/TiO2 is interesting and worth further studies. As the theme of this article is the molecular-level understanding of the high-temperature single-atoms catalytic chemistry of Rh1/TiO2 for POM, we leave the further studies of the deactivation of 6.3wt%Rh/TiO2 in POM as a future task.
The high-temperature single-atom catalyst, Rh1O5/TiO2 was prepared by substituting for Ti atoms of surface lattice with Rh atoms. The singly dispersed Rh1O5 clusters exhibit extraordinary catalytic performance of POM at 650 o C including high activity, 98% selectivity for producing CO and 99% for H2, long durability of 240 hrs, and high reusability over at least 20 cycles of catalysis. The single-atom site of this catalyst exhibits a turnover frequency of 2428 H2 molecules per Rh1 atom per minute at 650 o C under kinetics-controlled regime. This extraordinary catalytic performance remains with no decay at 650 o C for at least 240 hrs.
Computational studies suggest that (1) activation barrier of the rate-determining step, dehydrogenation of CH3 to CH2 is greatly decreased while CH3 and H bind to Rh1 and O of Rh1O5 cluster of this catalyst, respectively, and (2) the long catalytic durability of Rh1/TiO2 at 650 o C results from the unique coordination environment that Rh1 atom on TiO2 is in a nearly full coordination shell of oxygen atoms since Rh1 coordinates with 5 oxygen atoms. Also benefited from the nearly saturated coordination of Rh1 on TiO2 surface, the adsorption energy of atomic carbon on Rh1 is much lower than that on Rh atom with an obviously unsaturated shell on the surface of a Rh NP. The single dispersion of Rh1 atoms on Rh1/TiO2 and the low binding energy of atomic carbon on Rh1 of the Rh1/TiO2 effectively prevents the Rh1/TiO2 from forming carbon layers at high temperature, where the formation of carbon layers in POM on currently reported catalysts remains a common phenomenon. This study suggests an avenue for single-atom catalysis for significant chemical transformations at high temperatures.
Preparation of 0.037wt% Rh/TiO 2 (Rh 1 /TiO 2 ). Single atom catalyst, termed 0.037wt% Rh/TiO2 or Rh1/TiO2 was prepared with a modified deposition-precipitation method. Rhodium (III) nitrate hydrate, Rh(NO3)3•2H2O (99.9%, Sigma-Aldrich) was used as the source of Rh cations for the preparation of the catalysts. A solution of Rh 3+ with concentration of 2.84×10 -3 mol/L was made for the following use. Crystallized TiO2 nanoparticle with a size of 20-50 nm (99.9%, Sigma-Aldrich) was used as a TiO2 support. The catalyst precursor of Rh1/TiO2 was prepared through a deposition precipitation method modified for the preparation of singleatom catalyst, Rh1/TiO2. In a typical experiment, 500 mg TiO2 was mixed with 50 mL deionized water through a vigorous stirring to form a white suspension. 3.0 mL Rh(NO3)3•2H2O aqueous solution (2.84×10 -3 mol/L) was introduced to the TiO2 suspension through a precise injection with a syringe pump while the solution was vigorously stirred. The nominal loading is 0.20wt%. The new suspension was continuously stirred for 2 hours more to allow sufficient natural adsorption of metal ions. The pH value of the mixture was carefully adjusted to 9.5 by gradually introducing ammonium hydroxide solution, followed by vigorous stirring for 6 hours more in order to reach a complete equilibrium in solution. Then, the prepared solution was centrifuged and then dried at a 60 o C in overnight with a following calcination in Muffle furnace at 650 o C in air for 4 hours. The measured loading of Rh with ICP-AES is only 0.0370.005wt%. Thus, either 0.037wt% Rh/TiO2 or Rh1/TiO2 was used in the text for referring to this catalyst.
As described above, 0.037wt% Rh/TiO2 was prepared through deposition-precipitation modified for this work. Based on the mechanism of the deposition-precipitation method, only a portion of cations were precipitated by OH -. The portion of Rh 3+ cations can be precipitated on TiO2 depends on the pH of the solution. Higher pH allows a larger portion of Rh 3+ cations to be precipitated. To make sure formation of singly dispersed Rh1 on TiO2 upon annealing at 650 o C in air instead of formation of Rh2O3 nanoparticles, the pH of the aqueous solution containing TiO2 was controlled at 9.5. Based on ICP-AES measurements, the actual loading of Rh is 0.037wt%Rh. Thus, only about 18.5% of the Rh 3+ cations dissolved in water were precipitated. The rest of the Rh 3+ cations remained in solution and then was washed out upon centrifugation to separate the TiO2 with precipitated Rh 3+ in the form of Rh(OH)3 from aqueous solution which contained the rest of Rh 3+ . Certainly, the Rh 3+ in aqueous solution can be readily reused by precipitation at high pH to form Rh(OH)3; a following calcination in air makes Rh2O3. A reduction of Rh2O3 in H2 forms Rh metal nanoparticles. Thus, the preparation of Rh1/TiO2 does not waste the precursor of Rh 3+ . Preparation of 6.3wt%Rh/TiO 2 (RhNP/TiO 2 ). The 6.3wt%Rh/TiO2 catalyst was prepared via impregnation method which is distinctly different from the modified depositionprecipitation method. 500 mg of TiO2 powder was introduced to 0.50 ml aqueous solution of Rh(NO3)3•2H2O which contains 33.36 mg Rh obtained from 85.18 mg of Rh(NO3)3•2H2O.
The concentration of Rh would be 6.25wt% if all Rh atoms could be transferred to surface of TiO2 based on the calculation, 𝟑𝟑.𝟑𝟔 𝒎𝒈 𝑹𝒉 𝟑𝟑.𝟑𝟔 𝒎𝒈 𝑹𝒉+𝟓𝟎𝟎 𝒎𝒈 𝑻𝒊𝑶 𝟐 = 𝟔. 𝟐𝟓𝒘𝒕%. The impregnated sample was dried in a 60 o C oven overnight and treated in flowing 5% H2 in a tube furnace at 500 o C for 2 hrs and then at 650 o C for 4hrs. Based on ICP measurements, the concentration of Rh loaded on TiO2 is 6.30.3wt% which suggests most of the Rh atoms were transferred to TiO2. Upon the above procedure, Rh nanoparticles were formed on TiO2. Thus, either 6.3wt%Rh/TiO2 or Rh NP/TiO2 was used in the text to refer to this catalyst which is TiO2 with loaded Rh NPs.
Compared to the preparation of single-atom catalysts described in the section of Experimental methods and computational approaches, the nanoparticle catalyst (6.3wt%Rh/TiO2 or termed Rh NP/TiO2) was prepared with the impregnation method instead of deposition precipitation used for Rh1/TiO2. The impregnated sample was dried in a 60 o C oven in air overnight.
Then, the dry catalyst precursor was treated in 5% H2 at 500 o C and 650 o C instead of calcination in air at 650 o C. In the preparation of Rh1/TiO 2 , pH of aqueous solution containing TiO2 and Rh 3+ cations was carefully controlled at pH=9.5. However, in the preparation of Rh NP/TiO2, the solution containing Rh(NO3)3 solution and TiO2 is at about 7.0. In the precipitation for preparation of Rh1/TiO2, only about 20% Rh 3+ cations were precipitated; however, in the preparation of Rh NP/TiO2, all Rh 3+ cations in solution were transferred to surface of TiO2. Thus, the preparations of single-atom catalyst (Rh1/TiO2) and nanoparticle catalyst (Rh NP/TiO2) were distinctly different. NP-0.037wt% Rh/TiO2 was prepared with the same preparation steps as 6.3wt% Rh NP/TiO2: first impregnation of Rh 3+ to TiO2 and then drying at 120 o C oven in air overnight and reduction at 650 o C in 5% H2 for 4 hrs in a tube furnace (the same protocol used to prepare 6.3wt%
Rh NP/TiO2). This preparation made Rh exist in the format of Rh nanoparticles supported on TiO2.
Thus, Rh of NP-0.037wt% Rh/TiO2 exist in the format of Rh NPs, distinctly different from Rh1 atoms on 0.037wt% Rh/TiO2. Unfortunately, the conversion of CH4 in POM on 50 mg of NP-0.037wt% Rh/TiO2 is actually much lower than detection limit of GC (<0.5% of CH4).
The density functional theory (DFT) calculations were performed with the Vienna ab initio Simulation Package (VASP). 2,3 The on-site Coulomb interaction was included with the DFT+U method by Dudarev et al. 4 in VASP using a Hubbard parameter U = 3 eV for the Ti atom, from the previous literature for TiO2. 5,6 The Perdew-Burke-Ernzerhof (PBE) 5 functional form of generalized-gradient approximation (GGA) was used to describe electron exchange and correlation. In terms of unit cell, we used a 13 supercell of the stable (101) surface facet of bulk anatase TiO2. Regarding to the distance between Rh atoms, the Rh atoms are separated by ~10 Å along each axis of the cell. About the Hubbard U correction, The Hubbard correction was applied for Ti d-orbitals of the model. As there have yet to be any studies investigating the optimal U value for Rh in TiO2, we opted to not include a correction for Rh and we do not anticipate this will change the conclusions for the DFT calculations.
All calculations were performed with spin polarization. The projector-augmented wave method was used to describe the electron-core interaction 7,8 with a kinetic energy cutoff of 450 eV. The Brillouin zone was sampled with the Monkhorst-Pack scheme of a 3×2×1 k-point mesh. 8 Transition states (TS) were found with the nudged elastic band (NEB) 9 method using a force convergence criterion of 0.05 eV/Å. To calculate the occupancies of the molecular bonds of methane, the periodic natural bond orbital (NBO) analysis implemented by Schmidt et al. was used. 10
Size and morphology of 6.3wt%Rh/TiO2; Chemisorption of CO on 0.037wt%Rh/TiO2 and 6.3wt%Rh/TiO2; Evaluation of catalytic performances of bare TiO2, 0.037wt%Rh/TiO2 and 6.3wt%Rh/TiO2; Measurement of reaction rate under kinetics-controlled regime; Ex-Situ, In-Situ and Operando Characterizations; Calculations of the amount of Rh (in mol) exposed to surface of 50 mg of 0.037wt%Rh/TiO2 and 50 mg of 6.3wt%Rh/TiO2; Calculations of turn-over rates (TOR) of 0.037wt%Rh/TiO2 (Rh1/TiO2) and 6.3wt%Rh/TiO2 (Rh NP/TiO2); Note on the difference in coordination environment of Rh atoms in 0.037wt%Rh/TiO2 (Rh1/TiO2) and 6.3wt%Rh/TiO2 (Rh NP/TiO2); Note on AP-XPS studies.