CO oxidation over a ligand coordinated single site Rh catalyst: identification of the active complex†

Fereshteh Rezvani‡ ^a, Xuemei Zhou‡^ab, Debora Motta Meira^c, George E. Sterbinsky^d and Steven L. Tait*^a
aDepartment of Chemistry, Indiana University, 800 E. Kirkwood Ave., Bloomington, Indiana 47405, USA. E-mail: tait@indiana.edu
^bSchool of Chemical Engineering, Sichuan University, No. 24 South Section 1, Yihuan Road, Chengdu, 610065, P.R. China
^cCanadian Light Source, Saskatoon, Saskatchewan, Canada
^dAdvanced Photon Source, Argonne National Laboratory, 9700 South Cass Avenue, Lemont, Illinois 60439, USA

First published on 31st July 2024

---

---

1. Introduction

The oxidation of carbon monoxide (CO) plays a crucial role in eliminating CO from gas streams, for instance, in automotive emission control and the purification of hydrogen fuel cell feeds. This reaction, one of the earliest¹ and most extensively studied heterogeneous reactions, can be catalyzed by noble metal catalysts in the form of supported metal particles, metal single atom catalysts (SACs), or metal single crystals in ultra high vacuum (UHV) conditions. While noble metal catalysts investigated over the years are dominated by Au, Pt, and Pd,^2–8 less attention has been directed to Rh. Supported Rh catalysts have shown to be promising for use in the CO oxidation reaction once the cost of Rh can be significantly decreased.^9,10

The study of SACs for the CO oxidation reaction dates back to 2011, when one of the first studies demonstrated that a single atom Pt catalyst supported on FeO_x could catalyze CO oxidation at a relatively low temperature.⁷ In their study, the single Pt atoms exhibited efficient atom utilization and strong interactions with the support. It has been reported that the support impacts the activity of metal centers through mechanisms such as strong metal–support interaction (SMSI) between metals and TiO₂ (ref. 11 and 12) or by varying the defect density in TiO₂ supports for Pt single atoms in hydrosilylation reactions.¹³ Jiang et al.¹⁴ showed that the presence of TiO_x species significantly improves the stability of the Pt–TiO_x–Al₂O₃ catalyst. This improvement is attributed to negative charge transfer from TiO_x to Pt, leading to less propylene adsorption on Pt sites and thus avoiding further dehydrogenation and hydrogenolysis in their study of propane dehydrogenation.¹⁴ In the CO oxidation reaction, the activity of Pt₇ clusters is higher on a slightly reduced TiO₂ (110) surface than on a highly reduced surface.⁵ The Ti³⁺ species at high concentration (highly reduced surface) will diffuse from the interstitial positions in the bulk to the surface and deplete adsorbed oxygen.⁵

Li's group reports that the reaction rate of CO oxidation for single Au atoms on defective TiO₂ is higher than that of Au on pristine TiO₂ at the same reaction temperature.¹⁵ For Au on a defective TiO₂ surface, competitive adsorption is easier, and the energy barrier for CO oxidation is lower compared to Au on a pristine TiO₂ surface. However, this promotion is not sustained in oxygen-rich (near stoichiometric oxygen) conditions because the oxygen vacancies heal in seconds. The Ti–Au–Ti bond structure on the defective surface could alleviate the competition in the adsorption of CO and O₂ on a single atomic Au site.¹⁵ The Mars–van Krevelen (MvK) mechanism is related to the oxygen vacancy (O_v) formation energy of different oxide supports,¹⁶ but the reaction in this work does not follow the MvK mechanism because the energy barrier is higher for CO molecules reacting with surface lattice oxygen compared to CO molecules reacting with adsorbed molecular oxygen.¹⁵

Over the years, investigations on Rh-based catalysts for CO oxidation have not significantly developed, and even less has been done on the corresponding reaction mechanism. Only a few studies have used in situ spectroscopy to engineer Rh single atoms on TiO₂.¹⁷ A MvK mechanism is proposed in the case of Rh single atoms supported on a reducible support, such as heteropolyacid¹⁶ or ceria.¹⁰ In general, several mechanisms have been proposed for the CO oxidation reaction, including Langmuir–Hinshelwood (LH), Eley–Rideal (ER), and MvK. In the LH mechanism, CO and O₂ molecules are first co-adsorbed and then react to form an OOCO intermediate, which is the rate-determining step for the oxidation progress. In the ER mechanism, the activated O₂ molecules, as O atoms, directly react with free CO molecule from the gas phase, where the activation of O₂ is the rate-determining step. In the MvK mechanism, the CO reacts with lattice oxygen from the support to form CO₂, and subsequently, molecular O₂ replenishes the oxygen to the support.^18,19

In this work, we adopted the metal–ligand approach that has been reported previously by our group to synthesize single site Rh on TiO₂ (Fig. 1).²⁰ The oxygen defects were introduced on the oxide support before the Rh–ligand loading. These defects play a role in the assembly of Rh with the ligand during deposition. The experiment will be performed under oxygen-rich reaction conditions, similar to those in current internal combustion engines (high air/fuel ratio) and catalytic converters. Hence, it is essential to investigate the reaction under oxidative conditions. In our previous work, we found that ligands coordinated to Pt kept Pt resistant to the reductive conditions.²¹ However, the stability of metal–ligand catalysts under oxidative reaction atmospheres is not clear. In addition, the reaction serves as an ideal model reaction to investigate the reaction mechanism and identify active sites since it involves molecular rearrangements with CO- and oxygen-intermediates. Here, the Rh–ligand coordination of the catalyst will be resolved, and the active sites for the reaction will be identified using in situ spectroscopic and microscopic tools. The study of active sites under in situ conditions remains one of the essential challenges in heterogeneous catalysis.

2. Experimental section

2.1 Materials and chemical reagents

Titanium(IV) oxide (anatase, nanopowder, <25 nm, 99.7% trace metals basis), Rh(III) chloride hydrate (Rh 38–40%), 1-butanol (99.9%), 1,10-phenanthroline-5,6-dione (PDO, 97%), 3,6-di(2-pyridyl)-1,2,4,5-tetrazine (DPTZ, 96%), and 2,2′-bipyridine-5,5′-dicarboxylic acid (BPDCA, 97%) were purchased from Sigma-Aldrich. Dimethyl sulfoxide (DMSO, 99.90%) was purchased from Fisher Chemical. All chemicals were used without further purification. Compressed oxygen (UN 1072), ultra-high pure argon, carbon monoxide, and carbon dioxide were purchased from Airgas.

2.2 The preparation of Rh–ligand catalysts on titania by a modified wet-impregnation method

Titania (TiO₂) was used as the catalyst support, either in its pristine form (as received) or after an annealing treatment at a specific temperature to generate surface defects, particularly oxygen vacancies (O_v). This treatment, extensively explained in our previous work,¹³ involved the characterization of Ti³⁺/O_v states in the annealed TiO₂ using continuous wave (CW) electron paramagnetic resonance (EPR) spectroscopy (Fig. S1†), which is a widely employed technique for examining paramagnetic centers. For the annealed TiO₂ supports, 0.4 g TiO₂ powder was loaded into a ceramic crucible and placed in the center of a tube furnace. The tube furnace was first purged with pure lab nitrogen gas (99.998%, <3 ppm H₂O, <5 ppm O₂) for 30 min. Then, a flow of forming gas (5 vol% hydrogen, balanced in nitrogen gas, Airgas, 99.99%) at 2 L h⁻¹ was started. Under this gas flow, the temperature was increased from room temperature to 500 °C at a ramp rate of 10 °C min⁻¹. After holding at this temperature for 1 h, the furnace was cooled down to room temperature in the forming gas flow.

The synthesis of titania supported Rh–ligand single atom catalysts is based on a modified wet impregnation method previously reported by our group.²⁰ To synthesize Rh–PDO on TiO₂, 0.3 g TiO₂ powder (pristine or annealed) was dispersed in 15 mL deionized (DI) water, sonicated for 5 minutes, and then stirred for 30 minutes. 0.0198 g PDO powder was dissolved in 15 mL DI water to form a yellow transparent solution and added to the dispersed TiO₂ under stirring. RhCl₃ aqueous solution was then added dropwise to the PDO/TiO₂/DI water mixture under stirring, aiming for a nominal Rh loading of 1.0 wt% and a molar ratio of Rh to PDO of 1:6. A Rh:PDO molar ratio of 1:9 was also synthesized but found to have poor metal loading (Table S1†) and catalytic performance (Fig. S2†). The entire solution was kept at room temperature and stirred overnight. The synthesis of Rh–DPTZ and Rh–BPDCA on TiO₂ followed a similar process using 1-butanol as a solvent for DPTZ and RhCl₃ precursor and DMSO for BPDCA. The molar ratios of Rh:DPTZ and Rh:BPDCA were ideally calculated to be 1:3 with a nominal Rh loading of 1.0 wt%. After overnight stirring, the suspension was dried in an air stream, and the resulting precipitate was collected, washed with DI water, and finally dried in air. Rh/TiO₂ (without ligand) was prepared by adding RhCl₃ precursor into a TiO₂ suspension with 1 mL of 0.1 M NaOH, aiming for a nominal Rh loading of 2.0 wt%. After preparation, the solution was dried at 100 °C in air and washed with DI water. The Rh NPs were obtained by reducing the Rh/TiO₂ sample in Ar/H₂ (50% H₂) at 400 °C for 3 h before the CO oxidation reaction. Throughout this paper, we will use “TiO₂” to refer to the as-received TiO₂ powder and “def-TiO₂” to refer to TiO₂ powder annealed under forming gas at 500 °C. The chemical structures of the PDO, DPTZ, and BPDCA ligands are shown in Fig. 2a–c, respectively.

2.3 Characterizations of the as-prepared catalysts

X-ray Photoelectron Spectroscopy (XPS) measurements were conducted using a PHI Versaprobe II XP spectrometer, with a monochromated Al X-ray source, located at the Indiana University (IU) Nanoscale Characterization Facility (NCF). Rh 3d, Ti 2p, N 1s, C 1s, O 1s, and Cl 2p regions were collected. The binding energy of each spectrum was calibrated to the C 1s peak (284.8 eV). The intensity of each spectrum was normalized to that of the Ti 2p spectrum. XPS peak fitting was performed using CasaXPS software.

Scanning Transmission Electron Microscopy (STEM) imaging was carried out using a JEOL JEM NEOARM and JEOL JEM 3200FS microscopes, at the Indiana University Electron Microscopy Center. Energy dispersive X-ray spectroscopy (EDX) was used to analyze the quantity of each element, and the data were collected using the accessory of the same microscope.

Inductively Coupled Plasma Mass Spectrometry (ICP-MS) measurements were conducted using an Agilent 7700 quadrupole ICP-MS instrument, at the IU Department of Earth & Atmospheric Sciences. All solid catalysts were treated in aqua regia solution at 60 °C to dissolve all Rh, and the resulting solutions were analyzed after centrifugation and dilution. The calibration curve for Rh is performed using Rh standard for ICP (Sigma-Aldrich, in HCl) diluted with trace-free HCl for different concentrations.

X-ray Absorption Spectroscopy (XAS) was measured at the Rh K-edge, from 23025 to 24460 eV. The spectra consist of two main regions: the X-ray absorption near edge structure (XANES), near the absorption edge, and the extended X-ray absorption fine structure (EXAFS) region. The XAS measurements were performed at the 20-BM beamline at the Advanced Photon Source (APS) of Argonne National Laboratory using a double crystal Si (111) monochromator. The energy scale of the monochromator was calibrated using a Rh metal foil at the Rh K-edge. For all experiments, both fluorescence and transmission data were collected, and fluorescence data were used for analysis due to a better signal-to-noise ratio. Solid sample powder was pressed into a 7 mm diameter pellet and fixed to a mobile sample holder. Rh foil and Rh₂O₃ were used as standard samples for the XAS measurement (samples provided by the APS).

Analysis of the EXAFS data was carried out using the software ATHENA (part of the IFEFFIT package), including data alignment, edge calibration, deglitching, data normalization, and background subtraction. Fitting of EXAFS spectra was performed using ARTEMIS.²² The EXAFS spectra were Fourier transformed after application of a Hanning window with a minimum k of 3.0 Å⁻¹, a maximum k between 13.4 and 15.0 Å⁻¹, and with a transition region of dk = 1.0 Å⁻¹. The Fourier transform range for each specific sample was selected based on the signal to noise ratio at high k and is given in Table S2.† Fitting of the Fourier transform was carried out over an R range from 1 to 3 Å. The amplitude reduction factor, S₀², and the energy shift between the measured and the theoretical spectra (ΔE₀) were abstracted from the fitting of standards (Table S2†).

CO adsorption was performed using Diffuse Reflectance Infrared Fourier Transform Spectroscopy (DRIFTS) at 30 °C under a 1 bar atmosphere. The DRIFTS cell was first purged with an Ar stream, followed by exposure to a CO/Ar mixture (CO 10%) for 10 min. After that, physically adsorbed CO was removed by flowing Ar (40 sccm) for 10 min. The background spectrum was collected after the Ar purge and before the addition of CO/Ar gas. In situ DRIFTS experiments were performed under Ar/O₂ (30 sccm/10 sccm), or Ar/CO (36 sccm/4 sccm) conditions, reflecting the feed gas conditions for CO oxidation in the fixed-bed reactor, to reveal the intermediates under reaction conditions. Each of the mass flow controllers was operated at standard temperature (25 °C), regardless of the reactor temperature, thus, flow rates are reported in units of standard cubic centimeters per minute (sccm). The background spectrum for the in situ DRIFTS was collected after the adsorption of CO at 30 °C. For the characterization of the surface functional groups, IR spectroscopy was performed with a diffuse reflectance IR environmental chamber (PIKE Technologies, 162-4160, HTV) at room temperature under flowing air. The Background spectrum was collected with pristine TiO₂. Each IR spectrum was an average of 500 scans and was converted into Kubelka–Munk units.

2.4 The catalytic testing of supported Rh–ligand catalysts

The CO oxidation reaction in the fixed-bed reactor was performed using a gas phase mixture of carbon monoxide and oxygen in a 1:25 ratio, balanced in Ar, with a flow ratio of 1% CO: 25% O₂: 74% Ar. The Rh–ligand catalyst loading was 100 mg in a quartz tube. The two ends of the catalyst bed were fixed with glass wool. The total height of the catalyst bed was approximately 1 cm, and the inner diameter of the quartz tube was 4 mm. The quartz tube containing the catalyst was covered with a thermal jacket. A K-type thermocouple was placed from the bottom of the tube, contacting the catalyst bed, and the gas mixture flowed from the top of the tube. Reactants and products were detected using a mass spectrometer. Fragments at m/z 28 (CO), 32 (O₂), and 44 (CO₂), were analyzed while the background was subtracted by flowing Ar/O₂ at a 75:25 ratio using an empty tube. The temperature of the catalyst bed was controlled by a temperature program, increasing from 30 °C to 200 °C in 10 °C intervals with a ramping rate of 1 °C min⁻¹. At each temperature, the reaction was held for 20 minutes. The CO conversion (X_CO) was calculated by comparing the CO concentration in the post-reactor gas flow during the reaction to the concentration when the reaction was not running, according to the following equation:

3. Results and discussion

3.1 Identification of the as-synthesized Rh–ligand catalysts

The synthesis of each Rh–ligand catalyst follows a modified wet impregnation method that has been reported previously²⁰ and is described in detail in the Experimental section.

The Rh 3d spectra (Fig. 3a) show a doublet for all the synthesized samples. The binding energy (BE) position of Rh suggests that the Rh in Rh–PDO coordination is charged, possibly at a 2+ chemical state.²³ For Rh–PDO on pristine TiO₂, the spectra are located at 308.2 eV and 313.0 eV for Rh 3d_5/2 and Rh3d_3/2, respectively. On the defective TiO₂ surface, the BE positions of Rh in Rh–PDO assembly are similar to the former, as well as the full width at half maximum (FWHM) values, indicating similar Rh species in the two samples (FWHM = 2.5 eV).

However, for reduced Rh NPs on TiO₂, an additional doublet peak at a lower BE position is observed (307.2 eV for Rh 3d_5/2 and 311.9 eV for Rh3d_3/2), which is ascribed to metallic Rh.²⁴ It is noted that under our reductive treatment conditions, the chemical state of Rh is not entirely rhodium zero; i.e., partial Rh remains as Rh³⁺ from the precursor. The binding energy of such unreduced Rh is higher than that of Rh in the Rh–PDO ligand structure, indicating an electron transfer from the PDO ligand to Rh during coordination. The electron transfer between the PDO ligand and Rh can be observed from the N 1s spectra (Fig. 3b), where the peak position of N 1s for PDO deposited on TiO₂ (uncoordinated) is lower than that of PDO coordinated with Rh. Nevertheless, an insignificant amount of PDO adsorbs on pristine TiO₂ in the absence of Rh precursor. Cl 2p XP spectra are located at anion Cl⁻ positions (Fig. 3c). The Cl 2p intensity with PDO presence is noticeably lower, possibly due to PDO replacement of Cl in the Rh coordination sphere. The above results show that the support surface has no significant impact on the oxidation state of Rh 3d, N 1s, and Cl 2p.

The elemental ratio of each element, based on the XP spectrum calculation, is presented in Table 1. The Rh/Ti ratio on the defective TiO₂ support is higher than that on a pristine surface, likely attributed to the higher concentration of surface defects offering more anchor sites for Rh. The Rh loading on the defective surface is twice the amount of that on the pristine surface, as measured by ICP-MS (Table 1). Moreover, Rh NPs/TiO₂ exhibit a higher Rh loading compared to the Rh–PDO/def-TiO₂ catalyst, as shown in Table 1 and Fig. 4c and f. The PDO/Rh ratio is 2.31 on the pristine surface and 1.14 on the defective surface, indicating that on the defective surface, Rh is less coordinated with the PDO ligand. Further coordination information has been revealed using EXAFS spectroscopy (Table S2†). The Cl/Rh ratios show a higher content of Cl on the Rh nanoparticles compared to that of Rh–PDO catalysts.

---

Infrared spectroscopy (Fig. S3†) confirms the successful deposition of the ligand (PDO) on Rh–PDO/def-TiO₂, as all PDO-derived features are observed. These features include ν(CO) between 1710 cm⁻¹ and 1720 cm⁻¹, aromatic overtones between 1800 cm⁻¹ and 2000 cm⁻¹, ν(aromatic CC) between 1500 cm⁻¹ and 1600 cm⁻¹, and O–H/C–H bending (1455 cm⁻¹ and 1000–1150 cm⁻¹ for in-plane modes, 800–1000 cm⁻¹ for out-of-plan modes).

3.2 The morphologies of as-prepared Rh–PDO catalysts and the dispersion of Rh nanoparticles

STEM images show the presence of Rh nanoparticles on the pristine titania support (Fig. 4a–c). These particles are measured to have an average diameter of 1.4 (± 0.2) nm. In the presence of the PDO ligand, the structure of Rh is significantly different that without the ligand, i.e., hardly any evidence for Rh nanoparticle formation is observed in STEM imaging on either defective TiO₂ (Fig. 4d–f) or pristine TiO₂ (Fig. S4†).

In the characterization of metal single-atom catalysts, carbon monoxide is commonly employed as a gas probe molecule to assess the dispersion and oxidation state of the metal. In this study, we employ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) to characterize the distribution of the Rh–ligand catalyst through CO adsorption.

In the reduced Rh NPs sample, two CO adsorption bands are observed: one located at a higher wavenumber of ∼2084 cm⁻¹ and the other at a lower wavenumber of ∼2012 cm⁻¹ (Fig. 5, red curve). These bands are ascribed to the gem-dicarbonyl species of Rh⁰(CO)₂, corresponding to the symmetric and anti-symmetric CO-stretching modes, respectively.²⁵ The intensity ratio of the doublet is related to the angle between carbonyl groups.¹⁷ However, there is a lack of a bridged CO band at 1860 cm⁻¹, as well as the linearly adsorbed CO on Rh_NPs⁰ (2060–2070 cm⁻¹), indicating the absence of large/bulk metallic Rh species for Rh NPs.²⁶ This observation aligns with the Rh 3d XP spectrum (Fig. 3a), where a mixture of Rh⁰ and Rh³⁺ is observed. It is possible that the Rh NPs exhibit a core–shell structure, with a core of Rh oxide on the oxide support and metallic Rh on the surface.

For Rh–PDO on the pristine surface, the doublet is located at relatively higher wavenumbers, approximately 2093 cm⁻¹ and ∼2025 cm⁻¹, compared to that on Rh NPs. This difference can be ascribed to the weaker back donation from the filled d orbital of Rh²⁺ to the anti-bonding orbital of CO compared to the d orbital of Rh⁰. In the literature, the doublet of CO frequencies in Rh (CO)₂³⁺ configuration is located at 2110 and 2039 cm⁻¹,^17,27 indicating that Rh in the Rh–PDO structure is at 2+ chemical state. In addition, the CO adsorption doublets of Rh–PDO on defective titania are located at 2090 cm⁻¹ and 2021 cm⁻¹, which are slightly lower than those on pristine titania.

More importantly, for both Rh–PDO samples, an additional peak at 2110 cm⁻¹ is attributed to linearly adsorbed CO on the cationic Rh site.²⁸ The shoulder above the highest peak, located at 2124 cm⁻¹, is assigned as a third CO insertion into the Rh²⁺(CO)₂.²⁶ Overall, the CO adsorption peak is more intensive on the defective surface than on the pristine surface (black curve, compared to the blue curve), indicating a higher density of CO adsorption sites on the Rh–PDO/def-TiO₂.

The Rh K-edge XANES reveals that the Rh in the Rh–PDO structure is oxidized on both the defective and pristine titania support (Fig. 6a), based on comparison to metallic Rh foil.

The reduced Rh nanoparticles on pristine TiO₂ are not entirely in the metallic state but remain partially oxidized. We note that XPS measurements also show that Rh is not exclusively in a metallic state; some Rh 3d signals correspond to an oxidized state (Fig. 3a). A potential reason for this incomplete reduction could be the sample treatment after Rh deposition. The sample underwent an annealing step in Ar/H₂ at 400 °C for 3 h, which may induce a strong metal support interaction (SMSI).²⁹ This interaction may result in the embedding of Rh particles in the oxide support and simultaneously creating vacancies on the support surface.²⁹ This process can act as a barrier, preventing some Rh particles from being fully reduced to the zero-oxidation state, thereby leaving a portion in an oxidized state.

The EXAFS spectrum for Rh NPs shows the presence of both Rh–O and Rh–Rh paths (Fig. 6b), resulting in an N (Rh–Rh) of 1.7 and N (Rh–N/O) of 5.2 (Table S2†). The relatively low N (Rh–Rh) value, compared to metallic Rh on oxide supports in the literature,^30–32 may be ascribed to the incomplete reduction of Rh NPs or the very small particle size (Fig. 4a) observed in our samples, which may allow for significant Rh–O interaction. This value is also lower than that reported for Rh NPs on TiO₂ (0.5 wt%),³⁰ likely due to the high dispersion of Rh NPs and ex situ treatment conditions. More importantly, for the Rh–ligand catalysts, no Rh–Rh path is observed. The N (Rh–Cl) is 1.4 for Rh–PDO on defective TiO₂ and 1.0 for Rh–PDO/TiO₂. The N (Rh–N/O) is 4.8 for Rh–PDO on defective TiO₂ and 5.1 for Rh–PDO/TiO₂. Compared to the single site Rh on metal oxide in the literature,^10,17,33 the loading of single site Rh in our work is relatively high, indicating that the presence of the PDO ligand allows the high dispersion of Rh.

3.3 The catalytic performance of Rh–ligand catalysts

We further explore how the coordination of Rh affects the reaction pathway of CO oxidation by investigating the conversion of CO at different reaction temperatures. CO oxidation was tested in a fixed-bed flow reactor with a reaction gas composition of 1 vol% CO and 25 vol% O₂, balanced in Ar. We selected a high O₂/CO ratio in our experiments to achieve the so-called “hyperactive” states, that require an O₂/CO ratio > 10 for Rh metal, as reported in the literature.³⁴ Under stoichiometric O₂/CO conditions, the Rh surface will be essentially saturated with chemisorbed CO, resulting in far less activity for CO oxidation.³⁴

The profiles of CO conversion as a function of reaction temperature upon heating (Fig. 7) for different Rh–ligand catalysts and a Rh NPs catalyst show distinct reaction paths. The CO conversion over Rh–PDO/def-TiO₂ exhibits an onset of CO oxidation reactivity at 70 °C and reaches 90% conversion by 190 °C. The T₅₀ (50% conversion of CO) for this catalyst is 115 °C, which is significantly lower than that for Rh–PDO on pristine TiO₂ surface (183 °C), and for reduced Rh nanoparticles on TiO₂ (198 °C). The T₅₀ for our best catalyst is comparable to the Rh₁/CeO₂ and lower than the T₅₀ (150 °C) for the commercial three-way-catalyst.¹⁰ Except for the phenanthroline dione ligand, we employed two other ligands: bipyridyl with dicarboxylic acid groups and tetrazine structure, to coordinate with Rh. These two catalysts (Rh–BPDCA and Rh–DPTZ) exhibited reluctance to convert CO in the examined temperature range, likely due to their very low Rh loading (Fig. 7a). Characterizations of the Rh–BPDCA and Rh–DPTZ catalysts are shown in the ESI† (Fig. S6 and S7, and Table S1).

At 140 °C, the performance of the catalyst remains stable for a certain period of reaction time (Fig. 7b). The CO conversion rate at the Rh site is calculated for each catalyst, and the results show that at 140 °C, the CO conversion rates are 2.15 mol g_Rh⁻¹ h⁻¹ for Rh–PDO/def-TiO₂, 1.23 mol g_Rh⁻¹ h⁻¹ for Rh–PDO/TiO₂, 0.15 mol g_Rh⁻¹ h⁻¹ for Rh NPs/TiO₂, 0.31 mol g_Rh⁻¹ h⁻¹ for Rh–DPTZ/def-TiO₂, and 0.15 mol g_Rh⁻¹ h⁻¹ for Rh–BPDCA/def-TiO₂. The atom utilization efficiency for Rh coordinated with the ligand is an order of magnitude higher than that in nanoparticles.

In addition, our results show no oxygen poisoning observed with Rh–ligand catalysts, even under oxygen-rich conditions (Fig. 7b). Therefore, from a fundamental perspective, it is important to study in detail the nature of the active surfaces for Pt-group metals in oxygen-rich conditions using in situ characterization techniques. Sautet et al.³³ demonstrated that single site Rh₁ can adapt to the environment using in situ STEM imaging combined with DRIFTS.

3.4 The evolution of CO at Rh sites under O₂-rich conditions: impact of support surface

To identify adsorbed CO species on both the Rh site and TiO₂ support during CO oxidation, in situ DRIFTS analyses were performed using Rh–ligand and Rh NPs/TiO₂ catalysts (Fig. 8).^33,35 In each experiment, the temperature was increased in 10% O₂ after CO adsorption at 30 °C.

For Rh–PDO on defective TiO₂ (Fig. 8a), as the temperature rises, the intensity of gem-dicarbonyl CO in Rh²⁺(CO)₂ (2090 cm⁻¹, 2021 cm⁻¹) significantly decreases when the temperature reaches 90 °C, completely vanishing at 120 °C. A gradual increase in the region around 1660 cm⁻¹ is ascribed to the desorption of the bending vibration of coordinated H₂O as well as Ti–OH, which is also observed for the pure defective titania (Fig. 8d). Linearly adsorbed CO persists up to 120 °C, and a shoulder at a higher wavenumber (∼2136 cm⁻¹) appears, attributed to Rh²⁺–CO–O–O (a bridged O₂ on Rh²⁺).³⁶ For Rh–PDO on a pristine TiO₂ surface (Fig. 8b), the evolution of CO molecules takes place differently. There is no clear decrease in the gem-dicarbonyl CO in Rh²⁺(CO)₂ (∼2093 cm⁻¹ and ∼2025 cm⁻¹) until 120 °C, which may explain the relatively high onset temperature in Fig. 7, as the strong adsorption of reactant molecules may hinder the subsequent heterogeneous reaction.

More predominantly, for Rh–PDO and Rh NPs on pristine surface, we observe an increase in the peak at ∼1323 cm⁻¹, starting from 30 °C, which further increases significantly at 90 °C. This feature at FTIR has been ascribed to carbonate-like species on TiO₂,^37,38 originating from the product CO₂ further reacting with surface oxygen and forming carbonate species. It could be attributed to CO₃ in a bidentate binding position on the Ti site (v_SS = 1330 cm⁻¹ for asymmetric vCOO).³⁷ These species are not observed for Rh–PDO on defective TiO₂, which may result from the rapid desorption of the generated CO₂ from the surface. We attribute the difference in CO₂ desorption to the specific configuration of the Rh–ligand on the defective surface, which blocks the formation of carbonates.

For Rh NPs on TiO₂ (Fig. 8c), a clear decrease in the gem-dicarbonyl CO in Rh⁰(CO)₂ (2084 cm⁻¹, 2012 cm⁻¹) indicates that the CO molecule was consumed/desorbed at 60 °C. However, it is likely that the CO molecules are converted to carbonate-like species on TiO₂, not to gas phase CO₂, as the peak intensity for carbonate-like species increases. Note that this catalyst was reduced in Ar/H₂ (50% H₂) at 400 °C for 3 h; considering the spillover effect of hydrogen onto the surface,³⁹ a strong metal support interaction may be introduced. Despite this, for this catalyst, below 120 °C, there is no plausible CO₂ generation. Above 90 °C, the CO adsorption changes from gem-dicarbonyl CO to a single CO adsorption band at 2100 cm⁻¹, which corresponds to linearly adsorbed CO on single cationic Rh. It is thus deduced that the supported Rh NPs catalyst experiences a different mechanism for CO oxidation reaction, which may need to overcome a higher energy barrier compared to Rh–PDO on defective TiO₂.

3.5 Thermodynamic analysis of the reaction pathway based on DRIFTS

While the literature indicates that, with reducible support, single site metal catalysts often undergo a Mars–van Krevelen mechanism for CO oxidation, exceptions exist.^10,40 It has been reported that, in the case of a single Rh atom on a ceria surface, the CO reacts with CeO₂ lattice oxygen via the Mars–van Krevelen mechanism, while Rh nanoparticles on CeO₂ follow a Langmuir–Hinshelwood mechanism.¹⁰ Nevertheless, recent studies suggest that the Mars–van Krevelen CO oxidation mechanism is barely possible on single site Au/TiO₂ due to a higher energy barrier for CO reacting with lattice oxygen (2.73 eV of O_2c and 2.79 eV of O_3c). This is in contrast to the free energy of CO reacting with molecular oxygen species in air (0.6 eV of O_2c and 0.72 eV of O_3c).¹⁵ The co-adsorption of CO and O₂ shows greater favorability on single site Au on a defective TiO₂ (001) compared to a defective TiO₂ surface (without Au).¹⁵ Furthermore, under oxygen-rich conditions, the probability of participation of support oxygen is low.⁴¹

We conducted DRIFTS measurements for Rh NPs/TiO₂ in a pure Ar flow at elevated temperatures after CO adsorption at 30 °C (Fig. 9a). The results reveal no significant changes in the position of the CO adsorption bands, and any changes in peak shape can be due to variations in the background peak of titania. Importantly, there is no obvious formation of carbonate-like species on the TiO₂ surface when there is no oxygen gas flow. Therefore, we can exclude the possibility of a Mars–van Krevelen mechanism (CO reaction with support oxygen) for Rh NPs/TiO₂. We were not able to conduct the same experiment for Rh–ligand catalysts, as these tend to aggregate at high temperatures (200 °C) in an Ar environment without oxygen.

However, to further elucidate the mechanism of carbonate species formation, we carried out DRIFTS experiments using Rh NPs/TiO₂, Rh–PDO/TiO₂, and Rh–PDO/def-TiO₂ without pre-adsorbed CO molecules (Fig. 9b–d). In each case, the background DRIFT spectra obtained at 30 °C was subtracted from the data. In Fig. 9b, without CO pre-adsorption, no carbonate-like species formation is observed at high temperatures, which further suggests that carbonates are generated during the CO oxidation reaction. Fig. 9c shows a weak intensity of the carbonate peak (1315 cm⁻¹) and the CO adsorption peak (2124 cm⁻¹) for Rh–PDO/TiO₂ (Fig. 9c), possibly indicating the oxidation of surface carbon species. For Rh–PDO/def-TiO₂ (Fig. 9d), only a weak CO adsorption peak is observed. However, according to the IR spectra of the fresh catalyst (Fig. S3†), it is evident that the carbonate does not originate from the catalyst ligand. Overall, the results in Fig. 9 supplement the findings in Fig. 8 indicating that carbonate formation occurs during CO oxidation for the less active two catalysts (Rh NPs/TiO₂ and Rh–PDO/TiO₂). However, for Rh–PDO/def-TiO₂, CO is converted to gaseous CO₂ at a relatively low temperature through a Rh²⁺(CO)₂ intermediate.

3.6 Kinetic analysis of the reaction mechanism on ligand-coordinated single-site Rh

To further reveal the role of oxygen during the CO oxidation reaction, we performed kinetic measurements to determine the oxygen reaction order at different CO partial pressures (Fig. 10 and Table 2) at 140 °C. The kinetic study revealed reaction orders between 0–0.5 for oxygen in Rh NPs/TiO₂, 0.4–1.0 for Rh–PDO/def-TiO₂, and above 1 for Rh–PDO/TiO₂. A first-order dependence on O₂ at a fixed CO partial pressure indicates that oxygen activation is involved in the rate-determining step, while a zero-order dependence suggests facile O₂ activation. The reaction order of oxygen that we reported for the single site Rh–ligand catalyst is different from that of single site metals for CO oxidation,¹⁰ possibly due to the oxygen-rich conditions applied in our experiments.

---

In our experimental setup, we observed that the oxygen dependence (around 1.2) of Rh–PDO/TiO₂ remains unaffected by variations in CO partial pressure. This suggests the exclusion of the competitive Langmuir–Hinshelwood mechanism. Instead, the reaction on Rh–PDO/TiO₂ may follow the Eley–Rideal mechanism. Despite the excess oxygen in this experiment, oxygen molecules participate in the reaction as intermediate states, either as single oxygen atoms, as two individual oxygen atoms, or bridged oxygen species.³⁶ However, for Rh NPs/TiO₂, the low oxygen reaction order indicates facile oxygen dissociation on Rh NPs. The oxygen dependence decreases (from 0.53 to 0.26) as the CO partial pressure drops (Fig. 10a and b), suggesting a competing Langmuir–Hinshelwood (L–H) mechanism.

In the case of Rh–PDO on defective TiO₂, a similar CO dependence and oxygen reaction order are observed, with the oxygen reaction order decreasing from 0.80 to 0.43 as the CO pressure decreases. Although the oxygen reaction order for single site Rh on defective support is 0.4–0.8, indicating a less facile oxygen activation process compared to Rh NPs, the kinetic analysis for this catalyst at 140 °C, which shows a CO conversion of about 70% (Fig. 7), falls beyond the rational region. However, DRIFTS results demonstrate that for Rh–PDO/def-TiO₂, the conversion of CO with oxygen generates gaseous CO₂ rather than carbonate intermediates, which significantly facilitates the reaction thermodynamically.

Therefore, kinetically, for the single-site Rh coordinated with PDO on the pristine TiO₂ support, the CO reaction may occur via an Eley–Rideal (E–R) mechanism, where CO coordinates with Rh, and the dissociated oxygen reacts with the coordinated CO molecule. In previous studies, it has been reported that supported Rh single atom catalysts follow the Mars–van Krevelen (MvK) mechanism for CO oxidation.^10,17,42 For instance, in a study conducted by Han, et al.,¹⁰ the MvK mechanism was proposed based on CO-DRIFTS experiments and reaction rates as a function of CO and O₂ concentrations. We did not find other prior examples of supported Rh single atoms following the E–R mechanism; however, it may be useful to consider supported Ir single atoms, which are reported to follow the E–R mechanism. In a study by Lu, et al.,⁴⁰ it was found that the coordination ability of Ir single atoms to multiple ligands allows the CO molecule to coordinate at the Ir site, forming an Ir(CO) interfacial site. This facilitates O₂ activation between Ir and aluminum from the support. They observed that the reaction barrier is reduced between gas-phase CO and adsorbed O in Ir(CO)(O) via an Eley–Rideal mechanism.⁴⁰

For Rh–PDO/def-TiO₂ and Rh NPs/TiO₂, we propose a Langmuir–Hinshelwood (L–H) mechanism. In the former, the reaction occurs without carbonate formation, possibly due to the rapid desorption of the generated CO₂ from the defective TiO₂ surface.

3.7 Characterization of ligand-coordinated single site Rh catalysts after reaction

The post-reaction catalysts were analyzed using XPS (Fig. 11 and S8†), EXAFS (Fig. S9†), and DRIFTS (Fig. S3†). For Rh–PDO on defective TiO₂, the Rh 3d spectrum showed virtually no change in binding energy after the CO oxidation reaction (Fig. 11b). The Rh 3d spectra (Fig. 11a) for Rh NPs show that the Rh species are oxidized after the reaction. XPS and DRIFTS measurements confirm the presence of the ligand before and after the reaction (Fig. S3 and S8†). In EXAFS measurements, the N (Rh–O/N) and N (Rh–Cl) for both Rh–PDO catalysts do not change significantly after the reaction (Table S2†). The Rh–Rh shell contribution decreases in the EXAFS spectrum for Rh NPs after the CO oxidation reaction. After fitting the spectra, we found that N (Rh–Rh) drops from 1.7 ± 0.9 to 1.2 ± 0.7, while N (Rh–O) increases from 5.2 ± 0.4 to 5.5 ± 0.3.

4. Conclusion

Rh–ligand coordinated single atom catalysts have been successfully synthesized, demonstrating enhanced catalytic activity for the CO oxidation reaction at relatively low temperatures. The use of a defective TiO₂ surface as the support affects the configuration of Rh with the ligand, resulting in a different reaction pathway for CO oxidation compared to Rh–ligand on a pristine TiO₂ surface. Additionally, the ligand coordinated single site Rh catalyst exhibits high stability in an oxygen-rich environment, preventing the formation of carbonate-like species during the reaction. This behavior makes these catalysts promising for applications such as CO removal in controlling automobile emissions and in hydrogen fuel cell technologies.

Data availability

The authors declare that the data supporting the findings of this study are available within the paper and its ESI† files. Should any raw data files be needed in another format they are available from the corresponding author upon reasonable request.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the U. S. Department of Energy, Office of Basic Energy Sciences, Chemical Sciences program, DE-SC0021390. XPS measurements were carried out at the Indiana University (IU) Nanoscale Characterization Facility, with assistance from Dr. Yaroslav Lasovyj. Scanning transmission electron microscopy (STEM) and EDS measurements were performed at the IU Electron Microscopy Center, operated by Dr. Xun Zhan and Dr. Jun Chen. XAS measurements were performed at beamline 20-BM at the Advanced Photon Source, an Office of Science User Facility operated for the U.S. Department of Energy (DOE) Office of Science by Argonne National Laboratory, under Contract No. DE-AC02-06CH11357. ICP-MS measurements were performed in the IU Department of Earth and Atmospheric Sciences by Dr. Shelby Rader.

References

1. M. Haruta, S. Tsubota, T. Kobayashi, H. Kageyama, M. J. Genet and B. Delmon, J. Catal., 1993, 144, 175–192 CrossRef CAS .
2. T. Kropp, Z. Lu, Z. Li, Y.-H. C. Chin and M. Mavrikakis, ACS Catal., 2019, 9, 1595–1604 CrossRef CAS .
3. H. Xu, C.-Q. Xu, D. Cheng and J. Li, Catal. Sci. Technol., 2017, 7, 5860–5871 RSC .
4. X. Liu, M. Xu, L. Wan, H. Zhu, K. Yao, R. Linguerri, G. Chambaud, Y. Han and C. Meng, ACS Catal., 2020, 10, 3084–3093 CrossRef CAS .
5. S. Bonanni, K. Ait-Mansour, W. Harbich and H. Brune, J. Am. Chem. Soc., 2012, 134, 3445–3450 CrossRef CAS PubMed .
6. A. D. Allian, K. Takanabe, K. L. Fujdala, X. Hao, T. J. Truex, J. Cai, C. Buda, M. Neurock and E. Iglesia, J. Am. Chem. Soc., 2011, 133, 4498–4517 CrossRef CAS PubMed .
7. B. Qiao, A. Wang, X. Yang, L. F. Allard, Z. Jiang, Y. Cui, J. Liu, J. Li and T. Zhang, Nat. Chem., 2011, 3, 634–641 CrossRef CAS PubMed .
8. K. Ding, A. Gulec, A. M. Johnson, N. M. Schweitzer, G. D. Stucky, L. D. Marks and P. C. Stair, Science, 2015, 350, 189–192 CrossRef CAS PubMed .
9. H. Guan, J. Lin, B. Qiao, X. Yang, L. Li, S. Miao, J. Liu, A. Wang, X. Wang and T. Zhang, Angew. Chem., Int. Ed., 2016, 55, 2820–2824 CrossRef CAS PubMed .
10. B. Han, T. Li, J. Zhang, C. Zeng, H. Matsumoto, Y. Su, B. Qiao and T. Zhang, Chem. Commun., 2020, 56, 4870–4873 RSC .
11. S. J. Tauster, S. C. Fung, R. T. K. Baker and J. A. Horsley, Science, 1981, 211, 1121–1125 CrossRef CAS PubMed .
12. S. J. Tauster, S. C. Fung and R. L. Garten, J. Am. Chem. Soc., 1978, 100, 170–175 CrossRef CAS .
13. X. Zhou, L. Chen, G. E. Sterbinsky, D. Mukherjee, R. R. Unocic and S. L. Tait, Catal. Sci. Technol., 2020, 10, 3353–3365 RSC .
14. F. Jiang, L. Zeng, S. Li, G. Liu, S. Wang and J. Gong, ACS Catal., 2015, 5, 438–447 CrossRef CAS .
15. J. Wan, W. Chen, C. Jia, L. Zheng, J. Dong, X. Zheng, Y. Wang, W. Yan, C. Chen, Q. Peng, D. Wang and Y. Li, Adv. Mater., 2018, 30, 1705369 CrossRef PubMed .
16. N. J. O'Connor, A. S. M. Jonayat, M. J. Janik and T. P. Senftle, Nat. Catal., 2018, 1, 531–539 CrossRef .
17. M. J. Hülsey, B. Zhang, Z. Ma, H. Asakura, D. A. Do, W. Chen, T. Tanaka, P. Zhang, Z. Wu and N. Yan, Nat. Commun., 2019, 10, 1330 CrossRef PubMed .
18. R. M. Al Soubaihi, K. M. Saoud and J. Dutta, Catalysts, 2018, 8, 660 CrossRef .
19. S. Dey, G. C. Dhal, D. Mohan and R. Prasad, Adv. Compos. Hybrid Mater., 2019, 2, 626–656 CrossRef CAS .
20. L. Chen, G. E. Sterbinsky and S. L. Tait, J. Catal., 2018, 365, 303–312 CrossRef CAS .
21. L. Chen, I. S. Ali, G. E. Sterbinsky, X. Zhou, E. Wasim and S. L. Tait, Catal. Sci. Technol., 2021, 11, 2081–2093 RSC .
22. B. Ravel and M. Newville, J. Synchrotron Radiat., 2005, 12, 537–541 CrossRef CAS PubMed .
23. L. S. Kibis, A. I. Stadnichenko, S. V. Koscheev, V. I. Zaikovskii and A. I. Boronin, J. Phys. Chem. C, 2016, 120, 19142–19150 CrossRef CAS .
24. L. Óvári and J. Kiss, Appl. Surf. Sci., 2006, 252, 8624–8629 CrossRef .
25. J. Evans, B. Hayden, F. Mosselmans and A. Murray, J. Am. Chem. Soc., 1992, 114, 6912–6913 CrossRef CAS .
26. E. Ivanova, M. Mihaylov, F. Thibault-Starzyk, M. Daturi and K. Hadjiivanov, J. Catal., 2005, 236, 168–171 CrossRef CAS .
27. R. Lang, T. Li, D. Matsumura, S. Miao, Y. Ren, Y.-T. Cui, Y. Tan, B. Qiao, L. Li, A. Wang, X. Wang and T. Zhang, Angew. Chem., Int. Ed., 2016, 55, 16054–16058 CrossRef CAS PubMed .
28. J. C. Matsubu, V. N. Yang and P. Christopher, J. Am. Chem. Soc., 2015, 137, 3076–3084 CrossRef CAS PubMed .
29. S. Tauster, S. Fung and R. L. Garten, J. Am. Chem. Soc., 1978, 100, 170–175 CrossRef CAS .
30. S. Yokota, K. Okumura and M. Niwa, Catal. Lett., 2002, 84, 131–134 CrossRef CAS .
31. M. A. Newton, S. G. Fiddy, G. Guilera, B. Jyoti and J. Evans, Chem. Commun., 2005, 118–120 RSC .
32. J.-D. Grunwaldt, L. Basini and B. S. Clausen, J. Catal., 2001, 200, 321–329 CrossRef CAS .
33. Y. Tang, C. Asokan, M. Xu, G. W. Graham, X. Pan, P. Christopher, J. Li and P. Sautet, Nat. Commun., 2019, 10, 4488 CrossRef PubMed .
34. M. S. Chen, Y. Cai, Z. Yan, K. K. Gath, S. Axnanda and D. W. Goodman, Surf. Sci., 2007, 601, 5326–5331 CrossRef CAS .
35. Y. Zhou, D. E. Doronkin, M. Chen, S. Wei and J.-D. Grunwaldt, ACS Catal., 2016, 6, 7799–7809 CrossRef CAS .
36. E. A. Wovchko and J. T. Yates, J. Am. Chem. Soc., 1998, 120, 10523–10527 CrossRef CAS .
37. L. F. Liao, C. F. Lien, D. L. Shieh, M. T. Chen and J. L. Lin, J. Phys. Chem. B, 2002, 106, 11240–11245 CrossRef CAS .
38. L. Mino, G. Spoto and A. M. Ferrari, J. Phys. Chem. C, 2014, 118, 25016–25026 CrossRef CAS .
39. Y. Zhu, D. Liu and M. Meng, Chem. Commun., 2014, 50, 6049–6051 RSC .
40. Y. Lu, J. Wang, L. Yu, L. Kovarik, X. Zhang, A. S. Hoffman, A. Gallo, S. R. Bare, D. Sokaras, T. Kroll, V. Dagle, H. Xin and A. M. Karim, Nat. Catal., 2019, 2, 149–156 CrossRef CAS .
41. H. Wang, J.-X. Liu, L. F. Allard, S. Lee, J. Liu, H. Li, J. Wang, J. Wang, S. H. Oh, W. Li, M. Flytzani-Stephanopoulos, M. Shen, B. R. Goldsmith and M. Yang, Nat. Commun., 2019, 10, 3808 CrossRef PubMed .
42. J. Xu, Y. Wang, K. Wang, M. Zhao, R. Zhang, W. Cui, L. Liu, M. S. Bootharaju, J. H. Kim and T. Hyeon, Angew. Chem., Int. Ed., 2023, 62, e202302877 CrossRef CAS PubMed .

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4cy00507d

‡ These authors contributed equally to this publication.

This journal is © The Royal Society of Chemistry 2024