Interfacial compatibility critically controls Ru/TiO₂ metal-support interaction modes in CO₂ hydrogenation

Abstract

Supports can widely affect or even dominate the catalytic activity, selectivity, and stability of metal nanoparticles through various metal-support interactions (MSIs). However, underlying principles have not been fully understood yet, because MSIs are influenced by the composition, size, and facet of both metals and supports. Using Ru/TiO₂ supported on rutile and anatase as model catalysts, we demonstrate that metal-support interfacial compatibility can critically control MSI modes and catalytic performances in CO₂ hydrogenation. Annealing Ru/rutile-TiO₂ in air can enhance CO₂ conversion to methane resulting from enhanced interfacial coupling driven by matched lattices of RuO_x with rutile-TiO₂; annealing Ru/anatase-TiO₂ in air decreases CO₂ conversion and converts the product into CO owing to strong metal-support interaction (SMSI). Although rutile and anatase share the same chemical composition, we show that interfacial compatibility can basically modify metal-support coupling strength, catalyst morphology, surface atomic configuration, MSI mode, and catalytic performances of Ru/TiO₂ in heterogeneous catalysis.

Supports can largely affect the catalytic performance of metal nanoparticles, but the underlying principles are not yet fully understood. Here the authors demonstrate that metal-support interfacial compatibility of Ru/TiO₂ can critically control the metal-support interaction modes and the catalytic performances in CO₂ hydrogenation.

Introduction

Supported metal nanoparticles (NPs) dominate practical catalysts in producing bulk and fine chemicals, reducing environmental emissions, energy conversions, etc.^1–3. Their catalytic performances (activity, selectivity, and stability) not only depend on the composition, size, shape, and ligation state of metal NPs, but are also highly affected or even dominated by supports^4–11. Metal-support interaction (MSI) has therefore become a central topic in heterogeneous catalysis^1,3,12,13. Typical MSI modes include strong metal-support interaction (SMSI), interfacial charge transfer, interfacial perimeter, spillover, etc.¹. In particular, SMSI phenomena, featuring in encapsulating metallic NPs (such as Pt, Au, Pd, Rh, Ru, and Ni) by reducible oxide supports like TiO₂, can widely modify catalytic activity and selectivity of hydrogenation reactions^9,14–20. Although various MSI forms have been reported, the underlying tuning principles still remain elusive, because all structural factors can affect MSIs, such as the composition, size, and shape of metals, the composition, phase, facet, and size of supports, as well as adsorbates and reaction atmospheres^1,12,21–25. Moreover, their simultaneous interactions extremely complicate MSI phenomena and challenge the study on the mechanisms. Exploring the principles dominating support effects and MSIs is therefore crucial for rational design, optimization, and understanding of heterogeneous catalysis.

Among all structural factors, metal-support interface should play the primary role, because all MSI modes occur based on the direct contacts of catalysts with supports^26,27. For solid–solid contacts, coupling strength is the most fundamental parameter determining the property and stability of their interfaces, which is thermodynamically described with adhesion energy (Φ_adh)^28,29. The relative value of Φ_adh to bulk cohesion energy (Φ_coh) basically determines interfacial contact angles and thermal stability of supported particles³⁰. In catalysis science, adhesion energy widely controls the morphologies and sintering rates of supported-metal NPs^31,32. Many post-treatment methods such as thermal annealing and reduction–oxidation cycles can modify interfacial adhesion and catalytic performances^{1,9,16,22,33,34}. Furthermore, at the atomic scale, interfacial coupling occurs through forming chemical bonds, thus, interfacial bonding strength basically determines metal-support adhesions³⁵. For example, Campbell et al. theoretically studied the trends in the adhesion energies of metal NPs on various oxide surfaces, and found that higher metal oxophilicity and more active surface-oxygen atoms could lead to stronger metal-support adhesions²⁹. Senftle et al. further revealed that interfacial binding strengths between single-metal atoms and oxide supports depended on the oxophilicity of supported metals and reducibility of oxide supports³⁶. Despite these understandings on the surface stability of metal NPs on oxides, the structure–function relationships on how interfacial structure features modify MSI modes and catalytic performances are still not fully revealed yet^9,33,34,37.

At the atomic scale, strong catalyst-support contacts result from interfacial bonds. The strength parameters, macroscale adhesion energy, and microscale bonding energy can be correlated following:

$${\mathrm{\Phi }}_{\mathrm{adh}}=k{E}_{\mathrm{IB}}{N}_s$$ 1

where E_IB and N_s denote the average energy and surface density of interfacial bonds, and k is a coefficient. Thus, Φ_adh can be enhanced by increasing E_IB or N_s. Both E_IB and N_s further depend on the atomic configurations of contacting surfaces, because the positions of interfacial atoms intrinsically affect the length, angle, and density of interfacial bonds. Therefore, interfacial configurations of catalysts and supports are intrinsic structural factors controlling catalyst states and MSI effects. The matching degree of interfacial configurations is also referred to as interfacial compatibility, which measures the strength of interfacial bonding and adhesion^27,30,35. Catalysts weakly wet supports at misfit interfaces with low interfacial compatibilities, which leads to ready phase separations at the interfaces and catalyst sintering. While high interfacial compatibility can increase both the strength and density of interfacial bonds^13,35. The highest interfacial compatibility exists in epitaxial interfaces between lattice-matched materials. In general, lattice misfit less than 5% can form epitaxial overlayers with zero contact angles. This principle has widely guided the growth of semiconductor and oxide heterogeneous structures with minimized interfacial defects³⁸. However, the mechanism of how metal-support contacts and their interfacial compatibilities affect MSI modes and catalytic performances has been rarely revealed yet.

Using Ru/TiO₂ as a model catalyst, here we demonstrate the critical role of interfacial compatibility in controlling MSI modes, surface atomic states, and catalytic activity and selectivity in CO₂ hydrogenation reaction. Rutile (R–TiO₂) and anatase (A–TiO₂) are used as supports to vary interfacial structures. RuO₂ shares the same lattice structure with R–TiO₂, which can lead to a high interfacial compatibility to form epitaxial overlayers³⁹. The interfacial RuO_x species can act as anchoring layers to strengthen interfacial bonding between Ru and R–TiO₂. For CO₂ hydrogenation reaction, we find that both catalytic activity and selectivity can be oppositely modified by annealing Ru/TiO₂ catalysts in air. On R–TiO₂, Ru shows enhanced activity and dominant selectivity of CH₄; on A–TiO₂, Ru shows decreased activity and dominant selectivity of CO (Fig. 1a). Such opposite catalytic performances clearly indicate that interfacial compatibility can vary MSI modes—Ru/A–TiO₂ shows normal SMSI effect, while Ru/R–TiO₂ displays strong interfacial coupling.

Fig. 1. Opposite catalytic performances of Ru/TiO₂ supported on rutile (R–TiO₂) and anatase (A–TiO₂) for CO₂ hydrogenation.

a Summary scheme of varied activity and selectivity by direct H₂ reduction and after annealing in air. Ru/TiO₂–H₂ refers to directly reduced catalysts by H₂, while Ru/TiO₂–air–H₂ refers to catalysts by annealing in air at 400 °C and further reduction by H₂. b Temperature-dependent CO₂ conversions. c Temperature-dependent CH₄ selectivity. d Temperature-dependent CO₂ conversions of 2% Ru/R–TiO₂ catalysts on other R–TiO₂ supports that were prepared from TiN and TiCl₃, and P25 TiO₂. Temperature-dependent CO₂ conversions by annealing (e) 2% Ru/R–TiO₂–H₂ and (f) 4% Ru/R–TiO₂–H₂ catalysts in Ar and air. Ru/R–TiO₂–H₂–air–H₂ means the catalyst was first reduced by H₂ at 300 °C, then annealed in air at 400 °C, and reduced with H₂ at 300 °C again.

Results

Support effects on activity and selectivity of Ru/TiO₂ for CO₂ hydrogenation

All Ru/TiO₂ catalysts were prepared following the same protocol to minimize the interference from synthetic conditions. To stabilize the surfaces and prevent further sintering during catalyst preparations and reactions, the supports were first annealed in air at 500 °C for 10 h (Supplementary Fig. 1). The annealed supports and RuCl₃ were uniformly mixed in water, rapidly frozen with liquid nitrogen, and dried in a freezing drier (Supplementary Fig. 2). The fully mixed RuCl₃–TiO₂ precursors were reduced by H₂ at 300 °C either directly (Ru/TiO₂–H₂) or after pre-annealing in air at 400 °C (Ru/TiO₂–air–H₂). CO₂ hydrogenation was conducted at normal pressure with gas hourly space velocity (GHSVs) of 12,000 mL·g⁻¹·h⁻¹, in which the reaction gas composed of 60 vol% H₂/15 vol% CO₂/25 vol% Ar. At normal pressure, the products of CO₂ hydrogenation are CH₄ and CO. Therefore, CO₂ conversion denotes catalytic activity, and the ratio of n(CH₄)/(n(CO) + n(CH₄)) denotes selectivity, while the varied activity and selectivity further reflect different support effects and MSI modes^5,12,40–43.

Annealing RuCl₃–TiO₂ precursors in air can effectively modify catalytic performances of Ru/TiO₂ catalysts (Fig. 1a, b). Figure 1b presents temperature-dependent CO₂ conversions by 2% Ru/TiO₂. Ru/R–TiO₂–H₂ and Ru/A–TiO₂–H₂ show similar CO₂ conversions between 120 °C and 320 °C, suggesting that R–TiO₂ and A–TiO₂ apply similar support effects on directly reduced Ru/TiO₂ catalysts. However, the conversions dramatically differentiate by pre-annealing RuCl₃–TiO₂ precursors in air at 400 °C. Ru/R–TiO₂–air–H₂ displays an enhanced catalytic performance, with CO₂ conversion at 300 °C increasing from 31.4% to 89.2%; Ru/A–TiO₂–air–H₂ shows a highly decreased activity, with CO₂ conversion at 300 °C reducing from 29.4% to 1.7%. At each reaction temperature, Ru/R–TiO₂–air–H₂ shows the highest CO₂ conversions among four catalysts. The results indicate that R–TiO₂ and A–TiO₂ apply opposite support effects on the activity of Ru NPs.

In addition to activity, annealing RuCl₃–TiO₂ precursors in air can also dramatically modify the catalytic selectivity of Ru on R–TiO₂ and A–TiO₂ (Fig. 1c). CH₄ dominates both the products of Ru/R–TiO₂–H₂ and Ru/A–TiO₂–H₂ between 200 °C and 320 °C (>98%). However, the product on Ru/A–TiO₂–air–H₂ converts into 100% CO between 220 °C and 320 °C, while on Ru/R–TiO₂–air–H₂ is still 100% CH₄ between 140 °C and 320 °C. The different products show another effect of supports on the catalytic performances of Ru NPs^18,41.

The enhancement effect of air-annealing on the activity of Ru/R–TiO₂ is a general trend. We verified the phenomena by varying the loading amounts of Ru, rutile supports, and post-processing procedures. The activities of 1% Ru/R–TiO₂ and 4% Ru/R–TiO₂ can also be enhanced by pre-annealing in air, showing the same trend with 2%-Ru/R–TiO₂ (Supplementary Fig. 3). Moreover, the activities of 2% Ru/R–TiO₂ supported on other R–TiO₂ materials prepared using TiCl₃ and TiN (Fig. 1d) as precursors, and P25 TiO₂, a commercial TiO₂ product composed of 4/5 anatase and 1/5 rutile, can also be effectively enhanced by pre-annealing in air. It is noted that CO₂ conversions by Ru/R–TiO₂–H₂ and Ru/R–TiO₂–air–H₂ are almost the same on these three supports, confirming the stable reproducibility of this enhancement effect.

We further annealed 2% Ru/R–TiO₂–H₂ (Figs. 1e) and 4% Ru/R–TiO₂–H₂ (Fig. 1f) catalysts at 400 °C in 25-sccm air or Ar flows, and then reduced with H₂ at 300 °C. CO₂ conversion by the annealed catalyst in Ar (2% Ru/R–TiO₂–H₂–Ar–H₂) is similar to that of 2% Ru/R–TiO₂–H₂. In contrast, CO₂ conversions by the annealed catalyst in air (2% Ru/R–TiO₂–H₂–air–H₂) notably increase between 160 °C and 320 °C. In particular, CO₂ conversion increases from 1.9% to 8.8% at 200 °C, and 31.5% to 48.7% at 300 °C. This promotion effect can also be supported by the reduced apparent activation energies (E_a, Supplementary Fig. 4). E_a of 2% Ru/R–TiO₂–H₂ is 66.5 kJ·mol⁻¹, while the values are 52.3 kJ·mol⁻¹ and 40.4 kJ·mol⁻¹ for 2% Ru/R–TiO₂–H₂–Ar–H₂ and 2% Ru/R–TiO₂–H₂–air–H₂, respectively⁵. The enhancement effect on 4% Ru/R–TiO₂–H₂ is more apparent than 2% Ru/R–TiO₂ between 140 °C and 320 °C after annealing in Ar and air. This is because Ru nanoparticles on 4% Ru/R–TiO₂–H₂ are larger, and annealing can more effectively increase their contacts with R–TiO₂ supports.

Geometric states of Ru NPs on TiO₂ supports

Given that R–TiO₂ and A–TiO₂ share the same chemical compositions, supports and catalysts were annealed and prepared following the same procedures, and reaction conditions were controlled the same, Ru–TiO₂ interfacial interactions should dominate the opposite support effects. RuO₂ shares the same lattice structure with R–TiO₂, and their lattice misfit is less than 3.0%, thus RuO₂ can form epitaxial overlayers on R–TiO₂ with zero contact angles³⁹. Figure 2a, b and Supplementary Fig. 5 present transmission electron microscopy (TEM) and scanning TEM (STEM) images of RuO₂/R–TiO₂ prepared by annealing RuCl₃–R–TiO₂ mixture at 400 °C in air. RuO₂ encapsulates R–TiO₂ nanorods as epitaxial overlayers and forms core–shell structures. Such epitaxial structures can enhance the activity and stability of RuO₂ in catalytic oxidation reactions, such as Deacon reaction⁴⁴. In contrast, RuO₂ supported on A–TiO₂ are NPs rather than epitaxial overlayers due to their different lattice types (Supplementary Fig. 6). Therefore, RuO₂ shows a much higher interfacial compatibility with R–TiO₂ than A–TiO₂—this intrinsically determines their opposite support effects on the catalytic performances of Ru NPs.

Fig. 2. Structure characterizations of Ru/TiO₂ catalysts.

a TEM and b STEM images of RuO₂ overlayers on R–TiO₂ nanorods. c Ru/R–TiO₂ reduced from RuO₂/R–TiO₂ by H₂. d STEM, e elemental mapping, and f TEM images of Ru/R–TiO₂–air–H₂ after CO₂ hydrogenation. STEM images of g Ru/R–TiO₂–H₂, h Ru/A–TiO₂–air–H₂, and i Ru/A–TiO₂–H₂ after reactions.

The different interfacial compatibilities modify the surface states of Ru NPs on TiO₂ supports first. For Ru/R–TiO₂–air–H₂, Ru can still partly show epitaxial structures after H₂ reduction (Fig. 2c). In particular, after CO₂ hydrogenation reaction, Ru presents as flat NPs on Ru/R–TiO₂–air–H₂ as shown by the images of STEM (Fig. 2d), elemental mapping (Fig. 2e), and high-resolution TEM (Fig. 2f). In contrast, for Ru/R–TiO₂–H₂, Ru/A–TiO₂–H₂, and Ru/A–TiO₂–air–H₂, Ru exists as NPs with average sizes around 2.0–3.0 nm after reactions (Fig. 2g, i, Supplementary Fig. 7). Specifically, size distributions of Ru nanoparticles on Ru–TiO₂–air–H₂ and Ru–TiO₂–H₂ are 2.4 ± 0.4 nm and 2.5 ± 0.5 nm, respectively (Supplementary Fig. 7). The high interfacial compatibility can intrinsically increase interfacial coupling strength, and modify the chemical and surface states of Ru species, MSI modes, and catalytic performances of Ru/TiO₂ catalysts.

Interfacial bonding states of Ru/TiO₂ catalysts

We use H₂ temperature-programmed reduction (H₂-TPR) to probe the effects of interfacial compatibility on Ru–TiO₂ coupling strengths (Fig. 3a). H₂-TPR is an effective method to characterize the reducibility of oxides and their interfacial interaction strengths with supports^37,45. We prepared RuO₂/TiO₂ materials by annealing RuCl₃–TiO₂ mixtures at 400 °C in air. H₂-TPR results show that RuO₂/A–TiO₂ can be reduced between 110 °C and 175 °C; the peaks at 128 °C, 150 °C, and 300 °C correspond to surface RuO₂, interfacial RuO_x, and surface A–TiO₂ species, respectively. While RuO₂/R–TiO₂ can be reduced between 100 °C and 290 °C, and shows three states at 138, 185, and 270 °C, corresponding to surface RuO₂, interfacial RuO_x species⁴⁵. The higher reduction temperature indicates that RuO_x is more stable on R–TiO₂, which further confirms the stronger interfacial coupling due to matched lattices. Furthermore, X-ray diffraction (XRD) patterns of Ru/A–TiO₂ display clear Ru signals, while no Ru peaks appear on Ru/R–TiO₂–air–H₂ (Fig. 3b, Supplementary Fig. 8). The results indicate higher dispersions of Ru on R–TiO₂, agreeing with TEM results.

Fig. 3. Characterizing the chemical states of Ru species on TiO₂ supports.

a H₂-TPR of RuO₂/A–TiO₂ and RuO₂/R–TiO₂. b XRD patterns and c XPS spectra of Ru/TiO₂ catalysts.

We use X-ray photoelectron spectroscopy (XPS) to probe the chemical states of Ru. Figure 3c presents Ru-3d and C-1s lines of 2% Ru/TiO₂ catalysts after catalytic reactions, in which C-1s is set at 284.8 eV. Ru-3d_3/2 spectra show both metallic (Ru⁰) and oxidized (Ru^δ+) states, in which oxidized states mainly locate at Ru–TiO₂ interfaces. Ru^δ+/Ru⁰ ratio is 1.3 for Ru/R–TiO₂–H₂ and 1.5 for Ru/A–TiO₂–H₂ (Table 1). The approximately identical Ru^δ+/Ru⁰ ratios suggest that directly-reduced Ru NPs show similar interactions with R–TiO₂ and A–TiO₂, agreeing with their catalytic performances. However, pre-annealing in air can dramatically alter the states of Ru. Ru^δ+/Ru⁰ ratio increases to 2.2 for Ru/R–TiO₂–air–H₂, indicating increased interfacial contacts. This trend can be reproduced on other Ru/R–TiO₂ catalysts of varied R–TiO₂ supports and loading ratios of Ru (Supplementary Figs. 9 and 10, Supplementary Table 1). In contrast, Ru^δ+/Ru⁰ ratio decreases to 0.2 for Ru/A–TiO₂–air–H₂, and Ru^δ+ state is very weak. The opposite trends of Ru^δ+/Ru⁰ ratio also agree with their opposite catalytic performances.

Table 1.

XPS results of 2% Ru/TiO₂ catalysts.

Sample	Binding energy (eV)		Ruδ+/Ru0 radio
	Ruδ+	Ru0
Ru/R–TiO2–air–H2	281.0	280.0	2.2
Ru/R–TiO2–H2	281.4	280.3	1.3
Ru/A–TiO2–air–H2	280.3	279.4	0.2
Ru/A–TiO2–H2	281.5	280.5	1.5

Another feature of Ru/A–TiO₂–air–H₂ lies in the shift of Ru⁰-3d_5/2 from 280.5 to 279.4 eV, a binding energy even lower than that of Ru foil (280.1 eV, Supplementary Fig. 11). This phenomenon is usually ascribed to the occurrence of SMSI effect⁴⁶. This can also be supported by the catalytic performances. For CO₂ hydrogenation, SMSI and catalyst sizes can highly affect activity and selectivity^5,21,22,45. Both SMSI effect and size reduction can convert the product from CH₄ to CO^43,47,48. For example, Paraskevi reported that 3-nm Ru NPs showed the highest turnover frequency (TOF) on TiO₂, and bigger NPs favor CH₄⁵. In our system, pre-annealing increases the size of Ru NPs on A–TiO₂ from 2.1 nm to 2.8 nm, but the activity and CH₄ selectivity both dramatically decrease. The results indicate that SMSI effect should account for this change, because SMSI can generally decrease the activity and selectivity of CO₂ methanation¹⁸. For example, for Rh/TiO₂, adsorbates can induce SMSI to decrease the activity and selectivity for CO₂ methanation; Ru/TiO₂ also shows SMSI effect in reduction reactions^18,49,50. Therefore, varied interfacial compatibilities can vary MSI modes of Ru/TiO₂–air–H₂ catalysts: Ru/R–TiO₂–air–H₂ shows enhanced interfacial coupling bridged by RuO_x layers, while Ru/A–TiO₂–air–H₂ shows SMSI effect.

Surface atomic states of Ru NPs probed with CO-DRIFTS

Different MSI modes can vary the exposed surface atomic states of Ru NPs. We characterize surface Ru sites using diffuse-reflectance infrared Fourier transform spectroscopy (DRIFTS) at 25 °C with CO as the probe (Fig. 4). Figure 4a schemes the possible adsorption configurations of CO on Ru/TiO₂. For Ru/A–TiO₂–H₂, 2140 and 2080 cm⁻¹ result from multi-carbonyl-adsorption modes of CO on Ru sites with low coordination numbers (Ru(CO)_x, x = 2, 3), while the modes appear at 2138 cm⁻¹ and 2075 cm⁻¹ for Ru/R–TiO₂–H₂⁵¹. The broad peaks from 1900 to 2070 cm⁻¹ result from top-absorption modes of CO on Ru NPs (Ru–CO) and at the interface between Ru and TiO₂ (Ru_if–CO)^41,42. For Ru/R–TiO₂–air–H₂, the proportion of 1950 derived from Ru_if–CO was higher than directly reduced, agreeing with the XPS results. The bands at 2030 and 2015 cm⁻¹ belong to CO linearly adsorbed on Ru surfaces with A–TiO₂ and R–TiO₂ (Ru–CO), while the peak at 2105 cm⁻¹ on A–TiO₂ corresponds to Ru nanoclusters (Ruⁿ⁺–CO). These Ru nanoclusters catalyze the formation of CO (<2%) above 240 °C (Fig. 1c).

Fig. 4. Probing surface atomic states of Ru NPs with CO-DRIFTS at 25 °C.

a Adsorption configurations of CO on Ru/TiO₂ catalysts. b and c CO-DRIFTS of Ru/TiO₂ catalysts.

After pre-annealing, the total intensities decrease, and the relative intensities of the peaks also change. In particular, Ru(CO)_x mode on low coordinated Ru (Ru_LC) sites decreases for Ru/R–TiO₂–air–H₂, while it almost disappears for Ru/A–TiO₂–air–H₂. These changes mean that the ratios of Ru_LC sites decrease after annealing. Usually, the ratio of low coordinated surface atoms decreases with increased particle sizes. The result of Ru/A–TiO₂–air–H₂ is consistent with this trend, as shown by increased sizes of Ru NPs from 2.1 nm to 2.8 nm (Supplementary Fig. 7). The projected sizes of Ru NPs for Ru/R–TiO₂–air–H₂ and Ru/R–TiO₂–H₂ are 2.4 ± 0.4 nm and 2.5 ± 0.5 nm (Supplementary Fig. 7), which are almost the same. This suggests that the decreased ratios of Ru_LC sites did not result from increased particle sizes. TEM images show that Ru NPs display flatter shapes for Ru/R–TiO₂–air–H₂, owing to the stronger affinity of Ru with rutile supports bridged by interfacial RuO_x species (Fig. 2d–f). The greater curvature radius of flatter particles can lead to more ordered arrangement of surface atoms and increase the number of surrounding atoms. We also characterize the metal dispersion of Ru NPs with CO pulse adsorption. Metal dispersions of Ru (D_co) are 33.6% and 31.1% for Ru/R–TiO₂–air–H₂ and Ru/R–TiO₂–H₂, respectively, which are almost the same (Supplementary Table 2). The results thus indicate that size and surface area does not play critical roles in enhancing the catalytic performances of Ru NPs on R–TiO₂ supports. Therefore, such different surface atomic configurations of Ru nanoparticles directly modify their catalytic performances in CO₂ methanation.

Reaction mechanisms probed with operando FTIR

We further use operando Fourier transform-infrared spectroscopy (operando FTIR) to reveal CO₂ hydrogenation mechanisms on these Ru/TiO₂ catalysts (Fig. 5). The measurements were performed from 25 to 300 °C in 20-sccm gas flows of 60 vol%H₂/15 vol%CO₂/25 vol% Ar. For Ru/R–TiO₂ (Fig. 5a), the absorptions at 3017 and 1303 cm⁻¹ result from C to H bonds of CH₄, 1436 and 1560 cm⁻¹ from adsorbed formate species (*HCO₂), 1950, 2075 from adsorbed CO species (*CO), and 1360 cm⁻¹ from adsorbed carbonate (*HCO₃)^5,41,42,52.

Fig. 5. Revealing reaction mechanisms of CO₂ hydrogenation on Ru/TiO₂ catalysts with operando FTIR from 25 °C to 300 °C.

a Reaction mechanisms on Ru/R–TiO₂–H₂ and Ru/R–TiO₂–air–H₂. b Reaction mechanisms on Ru/A–TiO₂–H₂ and Ru/A–TiO₂–air–H₂.

At 25 °C, both Ru/R–TiO₂–air–H₂ and Ru/R–TiO₂–H₂ show weak adsorption peaks, but the reactions can obviously occur above 160 °C. At 160 °C, *HCO₃, *HCO₂, and *CO appear on both catalysts, meaning CO₂ is first activated as *HCO₃ and *HCO₂, and further reduced into *CO^8,41. The difference lies in 2075 and 3017 cm⁻¹. At 160 °C, CH₄ appears on Ru/R–TiO₂–air–H₂ but not on Ru/R–TiO₂–H₂, indicating that Ru/R–TiO₂–air–H₂ is more active for CO₂ methanation. This agrees with the enhanced catalytic results. Many researches have figured out that *CO and formate are two possible intermediates in thermal CO₂ hydrogenation reactions. In our results, stepwise-increasing reaction temperatures can lead to similar changes of *HCO₂ on the two samples (Fig. 5a). This suggests that formate is not likely the intermediate, or at least not linked to the distinctly different activity on rutile. Some previous reports have also concluded that *CO hydrogenation is the rate-determining step in the CO₂ hydrogenation on Ru/TiO₂⁴¹. Two catalysts show different adsorption modes of ^*CO, more obvious multicarbonyl Ru(CO)_n species at 2070 cm⁻¹ on Ru/R–TiO₂–H₂, which adsorbed on Ru_LC of the surfaces of Ru nanoparticles. It is inactive at low temperatures because H₂ cannot effectively reduce it at low temperatures⁵¹.

For Ru/A–TiO₂ catalysts, the modes at 3014 cm⁻¹ result from C to H bonds of CH₄, 2008 cm⁻¹ from *CO, and 1352 cm⁻¹ from *HCO₃ (Fig. 5b). This suggests that the reaction routes are the same with Ru/R–TiO₂, agreeing with our catalytic results in fixed-bed reactors. While the reversal peaks at 1644 and 1344 cm⁻¹ originate from the desorption of *H₂O, *OH, and *H on catalyst surfaces during the reaction⁵³. For Ru/A–TiO₂–air–H₂, the intensity at 2008 cm⁻¹ sharply drops compared with Ru/A–TiO₂–air–H₂, in line with CO-DRIFTS results in Fig. 4c. The highly decreased Ru–CO mode at 2008 cm⁻¹ indicates reduced exposure of surface Ru sites, which further supports the occurrence of SMSI, and agrees with the low catalytic activity (Fig. 1b). While SMSI effect can effectively convert the product from CH₄ into CO^12,17,18,48.

Discussion

Our results show that rutile and anatase TiO₂ supports can dramatically modify the morphology, surface atomic configuration, MSI mode, and catalytic performances of Ru catalysts for CO₂ hydrogenation reaction, although they share the same chemical compositions. Ru NPs adhere stronger with R–TiO₂ than A–TiO₂, which disagrees with the trend that more active surface-oxygen atoms lead to stronger interfacial adhesion^28,29. Instead, this confirms that interfacial compatibility plays critical roles in controlling the metal-support adhesion strength and MSI modes of Ru/TiO₂ catalysts.

For Ru/TiO₂ catalysts, RuO₂ shares the same lattice structure with R–TiO₂, thus, annealing RuCl₃–R–TiO₂ precursor can incorporate Ru atoms into the surface lattices of R–TiO₂ to form epitaxial RuO_x species (Fig. 6a, b). Such interfacial RuO_x species can act as anchoring layers to strongly bind Ru nanoparticles onto R–TiO₂ supports, which yields flat shapes with low contact angles and larger curvatures. This morphology can decrease the ratio of undercoordinated surface sites (Fig. 6a), and further modifies CO adsorptions and reaction routes. At the atomic scale, Ru^δ+ atoms can adequately occupy Ti sites, thus bonding to R–TiO₂ substrate with maximized bonding strength and density. While Ru⁰ atoms of Ru NPs can further bond to such Ru^δ+ sites through Ru–Ru metallic bonds (Fig. 6b). This kind of binding features can minimize interfacial defects of strain, dislocation, and vacancies, thus can further highly enhance metal-support adhesion and suppress interfacial phase separation.

Fig. 6. Atomic schemes showing the effects of interfacial compatibility on Ru/TiO₂ adhesion strengths.

a Scheme of epitaxial dispersion of Ru nanoparticle on rutile TiO₂(110) surface. b Interfacial Ru–O and Ru–Ru bonds in anchoring RuO_x layer on R–TiO₂. c Scheme of Ru nanoparticle on anatase TiO₂ (101) surface. d Scheme of interfacial of bonds and defects of Ru with A–TiO₂.

The lattice type of A–TiO₂ (I4₁/amd) is different from that of RuO₂ (P4₂/mnm), thus their interfacial atomic configurations do not match. Such misfit interfaces can form defects like edge dislocation and vacancy, which reduces interfacial adhesion and stability. Figure 6c–d schemes atomic interfacial contact of Ru nanoparticle on A–TiO₂ (101) surface. Some Ru atoms bond to surface oxygen atoms through Ru–O bonds, but their bonding lengths and strengths vary, depending on their atomic positions. Moreover, some atoms cannot effectively bond to surface-oxygen atoms due to misfit positions, and these sites form dislocations (Fig. 6d). Therefore, Ru NPs weakly adhere A–TiO₂ surfaces, and appear as spherical particles. This morphology yields more undercoordinated Ru sites. While the higher reducibility of surface-oxygen atoms can further drive the occurrence of SMSI effect, which further modifies CO adsorption, catalytic activity, and selectivity.

In summary, we have demonstrated that interfacial compatibility can critically control the interfacial coupling strength, surface atomic configurations, MSI modes, and catalytic performances of Ru/TiO₂ catalysts by varying interfacial adhesion strengths. For CO₂ methanation, enhanced interfacial coupling of Ru/R–TiO₂ can increase catalytic activity and CH₄ selectivity, while SMSI effect on Ru/A–TiO₂ can highly decrease catalytic activity and convert the product from CH₄ into CO. This is because Ru NPs can strongly adhere to R–TiO₂ supports and form flat particles with larger curvatures, in which interfacial RuO_x species act as anchoring layers with R–TiO₂; while Ru/A–TiO₂ show classic SMSI effect due to lattice misfit and higher reducibility of surface oxygen atoms. Therefore, interfacial compatibility is a critical structural feature that can intrinsically modulate MSI modes and catalytic performances, which might be realized through designing interfacial atomic configurations, introducing anchoring layers, thermal annealing, and oxidation treatments. This work paves the way to improve catalytic performances through engineering interfacial compatibilities between metal NPs and supports.

Methods

Chemicals

Titanium tetrachloride (TiCl₄, 99.5%), titanium trichloride (15.0~20% TiCl₃ basis in 30% HCl), TiO₂ (anatase, 25 nm), sodium hydroxide (NaOH, >98%), and sodium chloride (NaCl, 99.6%) were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. Ethanol (>99.7%), hydrochloric acid (HCl, 36–38%), and nitric acid (HNO₃, 68%) were purchased from Sinopharm Chemical Reagent Beijing Co., Ltd. Ruthenium trichloride hydrate (RuCl₃·xH₂O, 37.5–41% Ru) and titanium nitride (TiN, 99%) were purchased from Beijing Innochem Science & Technology Co., LTD.

Syntheses of rutile-type TiO₂ nanorods

R–TiO₂ nanorods were prepared using three methods:

1. TiCl₄ (2.5 mL) was slowly added into ice water (30 mL) under vigorous stirring. After being stirred for 15 min, the solution was transferred into a 40-mL Teflon-lined stainless-steel autoclave. Then the solution was heated at 170 °C for 14 h in an oven. After cooling down to room temperature, the products were collected by centrifugation and washed 4 times with deionized water, and dried at 60 °C for 12 h⁵⁴.

2. TiCl₃ (5 mL) was first mixed with 30 mL of 1.0 M NaCl solution. After being transferred into a 40 mL Teflon-lined stainless-steel autoclave, the solution was heated at 200 °C for 6 h.

3. TiN (0.5 g) was dispersed into 30 mL of 4.0 M HNO₃ solution, then was transferred into a 40 mL Teflon-lined stainless-steel autoclave. The mixture was heated at 180 °C for 24 h⁵⁵.

Syntheses of Ru/TiO₂ catalysts

All Ru/TiO₂ catalysts were prepared following the same procedure, in which only TiO₂ supports were changed. (1) TiO₂ supports were annealed in air at 500 °C for 10 h to stabilize their surfaces. (2) Then 1.0 g of the annealed TiO₂ supports and a certain amount of RuCl₃ were mixed in 10 mL of H₂O under sonication for 30 min. (3) The mixtures were rapidly frozen with liquid nitrogen, and further dried in a freeze drier. (4) As-obtained powders were calcined at 400 °C for 4 h in air or directly reduced with H₂ to prepare Ru/TiO₂ catalysts.

Characterization

Power X-ray diffraction (XRD) data were collected on a Bruker D8 diffractometer using Cu Kα radiation (1.5418 Å), which was operated at 40 kV and 40 mA with a scanning rate of 6 degree/min. High-resolution transmission electron microscopy (HRTEM), high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), and element mapping were performed on FEI Tecnai F30 transmission electron microscope (TEM) under an acceleration voltage of 300 kV. Size distributions of Ru nanoparticles were obtained through measuring at least 100 particles. X-ray photoelectron spectroscopy (XPS) data were collected on Thermo Scientific ESCALAB 250Xi system using Al Kα line as the X-ray source. The spectra were calibrated by with C1s peak at 284.8 eV.

H₂-TPR

H₂ temperature-programmed reduction (H₂-TPR) and CO pulse adsorption were performed on Autochem1 II 2920 instrument. Before TPR measurement, the samples were annealed in Ar at 300 °C for 60 min, then cooled down to 50 °C. The signals were recorded online with a thermal conductivity detector (TCD), as the reactor was heated to 800 °C at a heating rate of 10 °C /min under 10% H₂–90% Ar flows.

DRIFT

In situ diffuse-reflectance infrared Fourier transform (DRIFT) spectra of CO adsorption were performed on Thermo Fisher Nicolet iS50 with a resolution of 4 cm⁻¹ at 25 °C. Prior to CO adsorption, the sample was treated in H₂ flow at 300 °C for 1 h and then cooled down. Prior to collecting the background spectrum, the sample was purged with Ar for 30 min. Then 5% CO/Ar flow (20 sccm) was introduced into the reactor until saturated adsorptions. DRIFT spectra were collected until no gas-phase CO could be detected with Ar purging.

Operando FT-IR study of reaction mechanisms

The measurements were performed on Thermo Fisher Nicolet iS50. Prior to collecting the spectra, the samples were treated in H₂ flow at 300 °C for 1 h and then purging with Ar for 1 h. The background spectrum was collected, until the reactor cooling down to 25 °C. Subsequently, the feed gas (n(H₂):n(CO₂):n(Ar)=15:60:25) was introduced, then the IR spectra were collected at certain temperatures after being stabilized for 30 min.

Catalytic test of CO₂ hydrogenation

CO₂ hydrogenation reaction was carried out in a fixed-bed reactor made of stainless steel. First, the catalysts were in situ reduced at 300 °C for 1 h with 25-sccm pure H₂ before catalytic testing. After cooling down to 100 °C, the gas was switched to the reaction gas with molar ratio of n(H₂):n(CO₂):n(Ar) = 60:15:25. The reaction pressure was controlled at 1 atm, and the gas hourly space velocities (GHSVs) were 12,000 mL·g⁻¹·h⁻¹. The outgassing gas compositions were detected using an online gas chromatography (GC) equipped with a TCD detector after the reactions were stabilized for 25 min at specific temperatures.

The CO₂ conversion (XCO₂) was calculated according to the following equation:

$$X_{{\mathrm{CO}}_2}=\frac{n_{\mathrm{in}}\left({\mathrm{CO}}_2\right)-n_{\mathrm{out}}\left({\mathrm{CO}}_2\right)}{n_{\mathrm{in}}\left({\mathrm{CO}}_2\right)}=1-\frac{A_{\mathrm{out}}\left({\mathrm{CO}}_2\right)/A_{\mathrm{out}}\left(Ar\right)}{A_{\mathrm{in}}\left({\mathrm{CO}}_2\right)/A_{\mathrm{in}}\left(Ar\right)}$$ 2

where n_in(CO₂) and n_out(CO₂) refer to the molar number of the CO₂ before or after the reaction, respectively, the A_in(CO₂) and A_in(Ar) refer to the chromatographic peak areas of the CO₂ and Ar in the reaction gas, and the A_out(CO₂) and A_out(Ar) refer to the GC area after reactions.

The reaction rate (v) was calculated following:

$$v=\frac{\mathrm{GHSVs}\times X_{{\mathrm{CO}}_2}\times 15\%}{22400\times {\mathrm{m}}_{\mathrm{Ru}}}$$ 3

where GHSVs refer to the gas hourly space velocities and m_Ru refers to the mass of the Ru. Under this condition, the products are CO and CH₄, so the selectivity of CO and CH₄ (S_CO and SCH₄) meets

$$S_{\mathrm{CO}}+S_{{\mathrm{CH}}_4}=1$$ 4

$$S_{\mathrm{CO}}={\mathrm{f}}_{\mathrm{CO}/{\mathrm{CH}}_4}\cdot \frac{A_{\mathrm{CO}}}{A_{{\mathrm{CH}}_4}}\cdot S_{{\mathrm{CH}}_4}$$ 5

where the f_CO/CH4 refers to the relative correction factors of CO–CH₄ obtained by the calibrating gas; the A_CO and A_CH4 refer to the chromatographic peak areas of CO and CH₄, respectively.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Source data

Source Data ^{(1.3MB, xlsx)}

Author contributions

G.X. initiated, designed, and supervised this work, and wrote the manuscript. J.Z. prepared the materials and performed catalytic tests. Z.G. carried out DRIFTS experiments. T.Z., Z.L. and W.Z. helped with preparing the materials and catalytic experiments. G.X., X.L., and L.W. supported this research. All the authors discussed the results and commented on the paper.

Peer review

Peer review information

Nature Communications thanks Leon Lefferts and the other, anonymous, reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.

Data availability

The data supporting the findings of the study are available within the paper and its Supplementary Information. Source data are provided with this paper in excel format. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-021-27910-4.