Nanometric Metal Overlayer Catalysts: Fundamental Aspects and Applications

Abstract

Nanometric metal overlayer catalysts have been developed as novel catalyst structures with high three-way catalytic performance for practical applications. The metal overlayer is prepared via a dry process using pulsed arc plasma deposition, unlike wet coating processes for conventional powder catalysts containing nanoparticles. The key point to achieving high performance is the extremely high turnover frequencies for specific chemical reactions catalyzed by a large two-dimensional surface compared with metal nanoparticles. This overview study presents the preparation, structure, performance, reaction mechanism, and thermal stability of metal overlayer catalysts, focusing on rhodium (Rh) and catalytic conversion of nitric oxide (NO). Rh plays a pivotal role in the reduction of NO to N₂ in automotive three-way catalytic converters. The potential benefit of the overlayer structure is the minimum loading of precious metal, such as Rh, which is limited and expensive.

Fig. 2 Schematic illustrations showing pulsed AP deposition of Rh onto a metal foil substrate and shaping process into a honeycomb structure. A part of this figure is reproduced from Ref. 4).

1. Background

Most conventional solid catalysts are prepared in the form of powders comprising porous supports and active metal nanoparticles dispersed thereon. On the outermost surface of the metal nanoparticle, the number of reacted molecules per unit of time and per metal atom can be defined in terms of the turnover frequency (TOF). The apparent catalytic activity equals the TOF multiplied by the number of active metal atoms. The number of metal atoms should be increased by decreasing the particle size to enhance the catalytic performance. However, this catalyst design strategy tends to fail because of easy deactivation caused by thermal instability and catalyst poisoning. Consequently, there are unavoidable problems because of trade-off relationships between the catalyst performance and life. New concepts for designing catalytic materials may be necessary to overcome these problems.

The catalyst design strategy described herein derives its origin from the finding of Dr. Oh et al.^1,2) They found in the catalytic CO–NO reaction that the NO dissociation occurs much more slowly on supported Rh nanoparticles than on the single crystal Rh.¹⁾ This finding indicates that the single crystal has a higher TOF than nanoparticles. They also evaluated the particle size effect of Rh in supported powder catalysts and found that the TOF of the CO–NO reaction increased significantly as the Rh particle size increased.²⁾ To verify this interesting phenomenon, we conducted the kinetic analysis for the CO–NO reaction and obtained a positive dependence of TOF on the Rh particle size (Fig. 1).³⁾ The TOF occurs in two orders of magnitude when the Rh particle size increases from 1 to 10,000 nm. The largest possible particle size corresponds to the single crystal of Rh, but we utilized a large two-dimensional (2D) thin film with a few nanometers of thickness, which is expected to achieve the TOF as high as a single crystal. To demonstrate this hypothesis, we have prepared a nanometric Rh thin film (hereinafter denoted as the Rh overlayer) that fully covers the metal foil substrate.^4–6) The Rh overlayer exceeds the TOF of Rh nanoparticles by two orders of magnitude at the largest. As-obtained metal foils can be shaped into a honeycomb structure, exhibiting a high catalytic performance comparable with conventional ceramic honeycomb catalysts containing Rh nanoparticles. Notably, another point is that the overlayer uses fewer amounts of Rh than the nanoparticle catalysts. The advance will be widely recognized as a future research trend because of the limited Rh supply and increasing demands in other applications.

Fig. 1

TOFs for CO–NO reaction over Rh catalysts of different particle sizes. Reproduced from Ref. 3).

This review article describes the fundamentals of the nanometric Rh overlayer catalysts and outlines their advances compared with Rh nanoparticle catalysts. The preparation, structure, catalytic activity, and reaction mechanism are described to clarify the reasons behind the high TOF for the catalytic reduction of NO. The reaction mechanism is complementarily studied via first-principles molecular dynamics simulations within the density functional theory (DFT) framework to analyze the NO surface adsorption and surface reactions. We also describe that a metal honeycomb comprising the Rh overlayer has several practical points of importance for current and future automotive emission control technologies, including compactness, back pressure, catalytic performance, and thermal stability. Finally, future opportunities and challenges are also addressed.

2. Process and Materials

The catalyst preparation herein follows the dry process; that is, the active metal is supported onto metal foils via pulsed cathodic arc plasma (AP) deposition under vacuum (Fig. 2).^7–11) The experimental setup comprises a vacuum chamber connected to a turbomolecular pumping system, an arc discharge source (ARL-300, Ulvac, Japan) fitted with an Rh metal cathode (ϕ10 mm, 99.9%, Furuya Metals, Co., Ltd., Japan), and a stage to place substrates. At a pressure value of <10⁻³ Pa, 1,000 shots of AP pulses were achieved in 0.2-ms with a current amplitude of 2 kA to generate on an Rh metal target with a 1-Hz frequency.

Fig. 2

Schematic illustrations showing pulsed AP deposition of Rh onto a metal foil substrate and shaping process into a honeycomb structure. A part of this figure is reproduced from Ref. 4).

By igniting pulsed arc discharges between a cathode and an anode, Rh is immediately vaporized and ionized, resulting in plasma ejection. A large arc discharge current through the cathode induces a strong circumferential magnetic field and accelerates the ejected plasma downward to make a collision with a Fe–Cr–Al metal foil substrate comprising 75, 20, and 5 atom% of Fe, Cr, and Al (Nippon Steel & Sumikin Materials, Japan). The metal foil material is widely used as a honeycomb substrate for automotive exhaust aftertreatment catalysts.^12–14) The Rh deposition mass on the metal foil was monitored using a quartz crystal microbalance (STM-2, Inficon, USA). After the Rh ions collide with the foil surface, they lose kinetic energy as they diffuse on the surface and finally deposit as Rh metal nanoparticles.⁴⁾ Initially, highly dispersed, but well-crystallized, Rh nanoparticles with a very narrow size distribution were deposited. The number of nanoparticles increased with an increasing number of pulses, leading to coalescence and lateral grain growth. Finally, a 2D Rh overlayer was formed on the foil. The surface coverage increased rapidly and reached almost 100% after 1,000 pulses, where the thickness was calculated to be ∼3–4 nm. The overlayer thickness can be controlled by the number of AP pulses.

3. Structure and Microstructure

Figure 3(a) shows the scanning electron microscope (SEM) and X-ray images of the cross-sectional surface of the Rh overlayer formed on a metal foil.⁴⁾ The as-deposited Rh is distributed uniformly on the top surface of the metal foil, showing intimate contact and strong adhesion at the interface. The X-ray photoelectron spectroscopy (XPS) analysis revealed that the surface of the metal foil is completely covered by a nanometric overlayer of Rh.⁴⁾ The Rh3d XPS combined with the depth profile analysis provides more information. In Fig. 3(b), the frontal spectrum corresponds to the outermost surface, whereas the other spectra are taken after etching by Ar⁺ bombardment. The as-prepared metal foil catalyst clearly shows Rh at the outermost surface. Two peaks are assigned to metallic Rh⁰, whereas peaks due to Rh³⁺ are not observed. The peaks remain unchanged during the first several etching periods, whereas they steeply disappear as the etching times increase. These results suggest the formation of a nanometric overlayer structure on the surface of the metal foils. Figure 4 shows the X-ray diffraction (XRD) pattern of the as-prepared Rh overlayer on the metal foil and polycrystalline Rh as a reference. Three peaks are assignable to the 111, 200, and 220 reflections of the face-centered cubic lattice, respectively. However, the 200 and 220 peaks are much less intense than the 111 peaks compared with the relative peak intensities of the polycrystalline Rh reference. This trend indicates the (111) orientation of the as-deposited Rh overlayer. Besides, a similar (111) orientation was observed for the Pt and Ir overlayers when they were prepared by the AP technique.^15,16)

Fig. 3

(a) SEM and elemental mapping images of the fracture surface of an Rh overlayer formed on a metal foil. (b) Rh3d XPS depth profiles of an Rh overlayer formed on the metal foil. Reproduced from Ref. 5).

Fig. 4

XRD pattern of an Rh overlayer as deposited on a Fe–Cr–Al metal foil and polycrystalline Rh as a reference. Peaks with an asterisk are reflections from the metal foil.

Using flat and corrugated metal foils after Rh deposition with both sides, a monolithic honeycomb can be prepared for catalytic tests (bottom of Fig. 2). As a reference catalyst, a commercially available ceramic honeycomb (cordierite, 2MgO·2Al₂O₃·5SiO₂, NGK Insulators, Japan) was coated with a slurry containing a powder catalyst (0.4 mass% Rh/ZrO₂). Figure 5 shows photographs of these two types of honeycomb prototypes with miniature sizes. The structural difference between these honeycomb catalysts is summarized in Table 1. Notably, despite the similar cell density and wall thickness (50 and 80 µm), the total wall thickness, including the catalyst layer, is larger for cordierite (105 µm) than metal (50 µm) honeycombs. This is apparently due to a very thin Rh overlayer (3-nm thick) compared with a washcoated Rh/ZrO₂ catalyst layer (several 10 µm in thickness). Consequently, the metal honeycomb coated with an Rh overlayer is characterized by a larger open frontal area, which is beneficial for a larger geometric surface area at a comparable pressure drop.

Fig. 5

Photographs of (a) metal- and (b) cordierite honeycombs with a miniature size (10 mm in diameter).

Table 1 Honeycomb structures and TOFs for Rh overlayer and Rh nanoparticle catalysts.

4. Catalytic Performance

The honeycomb comprising the Rh overlayer showed high catalytic performance for simulated three-way catalytic (TWC) reactions. Figure 6 shows a typical example performed at a gaseous hourly space velocity of 1.2 × 10⁵ h⁻¹,⁴⁾ which is close to that in the exhaust of a real engine. The Rh overlayer catalyst provides a steep rise in conversion >230°C, which is lower than that observed on the Rh nanoparticle (powder-coated honeycomb) catalyst. This outcome indicates that the former is superior to or comparable with the latter in its catalytic performance. This result was much more than expected, considering a less amount of loaded Rh and, thus, a small number of surface Rh sites in the overlayer structure (Table 1). Another interesting feature is the shape of the light-off curves, which represent the temperature dependence of conversion efficiencies (Fig. 6). Notably, the NO reaction on the Rh overlayer starts at the lowest temperature, followed by CO and subsequently C₃H₆. In contrast, the light-off of CO occurs first on the Rh nanoparticles. These results suggest that the Rh overlayer exhibits a high TOF, especially for the NO conversion, as discussed below.

Fig. 6

Light-off curves for a simulated stoichiometric TWC reaction (0.05% NO, 0.50% CO, 0.05% C₃H₆, 0.53% O₂, 0.17% H₂, 10% CO₂, 10% H₂O, and N₂ balance, 1.0 L min⁻¹, GHSV = 1.2 × 10⁵ h⁻¹) at atmospheric pressure over honeycomb catalysts comprising (a) Rh overlayer and (b) nanoparticles.

Since simulated TWC reactions involve several parallel reactions, including CO–NO, CO–O₂, and C₃H₆–O₂, the detailed catalytic behavior was investigated for each reaction to compare the Rh overlayer honeycomb and the reference cordierite honeycomb coated with powder catalyst containing Rh nanoparticles.³⁾ Figure 7(a) shows that the catalytic performance for the CO–NO reaction slightly exceeds that for the Rh overlayer compared with the Rh nanoparticle. Using the number of surface Rh sites and the NO conversion rate, the TOF was calculated to be more than tenfold greater for the Rh overlayer than that of the Rh nanoparticle (Table 1). This result suggests that the nanoparticle catalyst requires a large surface area (i.e., many active sites) to achieve high performance, whereas the Rh overlayer achieves such performance by the high TOF despite its minimal number of active sites. We also studied the catalytic performance and TOFs for CO–O₂ and C₃H₆–O₂ reactions (Fig. 7(b) and (c)). Unlike the CO–NO reaction, the Rh overlayer was less active for these reactions. The calculated TOFs for the CO–O₂ reaction is much lower for the overlayer, and those for the C₃H₆–O₂ reaction are small and nearly comparable (Table 1).

Fig. 7

Light-off curves for (a) CO–NO (0.1% NO, 0.1% CO, N₂ balance), (b) CO–O₂ (0.5% CO, 0.25% O₂, and N₂ balance), and (c) C₃H₆–O₂ (0.4% C₃H₆, 0.18% O₂, and N₂ balance) reactions over honeycomb catalysts comprising Rh overlayers and the Rh nanoparticles. Reprinted with permission from Ref. 3). Copyright 2019 American Chemical Society.

Also, we studied various metal overlayers other than Rh. In contrast to Rh, the Pd overlayer is less active for the CO–NO reaction than Pd nanoparticles (Pd/Al₂O₃), and their TOFs are nearly equal. In contrast, the Pt overlayer shows the worst light-off performance characteristics, but its TOF clearly exceeds that of the Pt nanoparticles (Pt/Al₂O₃). Interestingly, the Pt overlayer is most active among precious metals for low-temperature NH₃ oxidation and achieves more than 180-fold greater TOF than Pt/Al₂O₃ at 200°C.¹⁵⁾ A similar trend was also confirmed for the Ir overlayer, which exhibited more than 70-fold greater TOF than Ir/Al₂O₃ at 200°C.¹⁶⁾ Therefore, the superiority of metal overlayers over nanoparticles in terms of high TOFs strongly depends on the types of metal elements and catalytic reactions adapted.

In TWC reactions, the catalyst is exposed to dynamic oxidation–reduction perturbation atmosphere. Thermodynamically, the active metallic Rh is formed under a reducing (fuel-rich) condition, whereas it is oxidized to less active Rh₂O₃ under an oxidizing (fuel-lean) condition. Thus, the higher resistance of Rh to reoxidation is favored to preserve the high catalytic activity under practical reaction atmospheres,^17–19) which is another advance of the overlayer catalyst. Also, it was found that the Rh overlayer significantly improved the NO conversion to N₂ under a lean condition than the Rh nanoparticles (Rh/ZrO₂).²⁰⁾ At 400°C, metallic Rh formed on the Rh overlayer was gradually oxidized on exposure to excess O₂, whereas the Rh nanoparticles on ZrO₂ were readily oxidized to less active Rh₂O₃. The resistance of Rh against oxidation is a possible reason for the enhanced NO conversion efficiency under lean conditions.

5. Reaction Mechanism

Table 1 shows that the higher TOF of the Rh overlayer catalyst is most obvious in the CO–NO reaction. To elucidate the reason for the high TOF, experimental and theoretical combined mechanistic analyses were conducted. According to previous studies,^21–25) the following elementary steps are included in the CO–NO reaction over the metal surface:   

\begin{equation} \text{NO}_{\text{g}}\to \text{NO}_{\text{s}} \end{equation} (1)

  

\begin{equation} \text{CO}_{\text{g}}\to \text{CO}_{\text{s}} \end{equation} (2)

  

\begin{equation} \text{NO}_{\text{s}}\to \text{N}_{\text{s}} + \text{O}_{\text{s}} \end{equation} (3)

  

\begin{equation} \text{2N}_{\text{s}}\to \text{N}_{\text{2g}} \end{equation} (4)

  

\begin{equation} \text{CO}_{\text{s}} + \text{O}_{\text{s}}\to \text{CO}_{\text{2g}} \end{equation} (5)

  

\begin{equation} \text{NO}_{\text{s}} + \text{N}_{\text{s}}\to \text{N$_{2}$O$_{\text{g}}$} \end{equation} (6)

  

\begin{equation} \text{NO}_{\text{s}} + \text{N}_{\text{s}}\to \text{N}_{\text{2g}} + \text{O}_{\text{s}} \end{equation} (7)

where subscripts “g” and “s” denote the gas and adsorbed species on the metal surface, respectively. The reaction over the Rh overlayer and nanoparticles were studied using first-principles molecular dynamics simulations, which calculated energies for the adsorption steps of NO molecules (1), the subsequent dissociation (3), and the recombination (4) because the adsorption and reaction of the CO molecules (2) and (5), respectively, are very fast. The other reactions, (6) and (7), are unlikely to contribute as they can be judged experimentally and theoretically.

Assuming the Rh (111) surface as a model for the overlayer catalyst, the adsorbed NO molecule was found to be more stable on a threefold hollow site than on-top or bridge-type adsorption sites (Fig. 8).³⁾ In contrast, a cuboctahedral Rh₅₅ cluster was used as a model nanoparticle with three adsorption sites for the NO molecule, but the adsorption energy was similar regardless of the adsorption site.³⁾ The stable hollow-site NO adsorption onto the Rh overlayer was confirmed experimentally using in situ FTIR spectroscopy.³⁾ In contrast, the on-top and bridge NO species were dominant, but the hollow-site bands were completely absent on the supported Rh nanoparticles. Figure 8 represents the reaction steps of the adsorbed NO and estimated energies for NO dissociation and subsequent N–N recombination on the Rh overlayer and on the Rh nanoparticle. Starting with the NO molecule adsorbed on the hollow site of the Rh (111) overlayer, the dissociation of NO and subsequent N–N recombination required energy barrier levels of 1.65 and 1.76 eV, respectively. On the nanoparticles, the dissociation of the bridge NO showed a relatively smaller barrier (1.00 eV) compared with the overlayer. In contrast, the N atoms thus yielded were stabilized in the hollow site, requiring a large energy barrier of 2.18 eV to form N₂. Therefore, the N₂ formation through the recombination of N should be slower on the Rh nanoparticles than on the Rh overlayer. These DFT-based simulations conclude that the high-TOF characteristics of the overlayer catalyst can be explained by a lower energy barrier for N–N recombination. The proposed mechanism was rationalized by comparing the results with the empirical kinetics of CO–NO reactions.

Fig. 8

Estimated energy profiles for intermediates in NO reduction on (a) Rh (111) and (b) Rh₅₅ cluster. Steps (i) and (ii) show NO dissociation to form N_s atom, and steps (iii)–(v) show N_s–N_s recombination. Blue, red, and pink spheres represent N, O, and Rh atoms, respectively. Reprinted with permission from Ref. 3). Copyright 2019 American Chemical Society.

6. Thermal Stability

When the Rh overlayer was prepared via different techniques (i.e., the AP deposition and RF magnetron sputtering), their catalytic activities in terms of TOF were almost the same, but their thermal stability characteristics differed.⁶⁾ When repeating catalytic test cycles with temperature ramping from room temperature to 500°C, the Rh overlayer formed by AP deposition was rather stable because of the strong adhesion to the metal foil surface compared with that prepared by sputtering. However, the Rh overlayer must be thermally stable up to a temperature exceeding 900°C given the practical use in a TWC converter. When thermally aged in the air at 900°C, the surface of the overlayer is oxidized to less active Rh oxides, but the active metallic surface is recovered under the TWC reaction atmosphere.⁵⁾ The more serious problem is that the Fe–Cr–Al foil substrate develops a highly crystalline α-Al₂O₃ passivation layer, which completely covers the surface of the Rh overlayer (Fig. 9(a)).²⁶⁾ This structural deterioration markedly decreases the Rh concentration on the surface and deactivates the catalyst. The lost activity of the thermally aged Rh overlayer was unrecovered under the reaction atmosphere because this structural deterioration was irreversible. Hence, a 500-nm thick Zr buffer layer was inserted between the Rh overlayer and the metal foil to stabilize the overlayer structure to solve this problem.²⁶⁾ The Zr buffer layer preserves the outermost surface of the Rh overlayer by blocking the interaction between the Rh and Al₂O₃ layers (Fig. 9(b)). After thermal aging at 900°C, the Rh–Zr overlayer preserved more than 80% of the Rh surface coverage, whereas Rh was lost almost completely from the surface of sample without the Zr buffer layer. Furthermore, the Zr buffer layer negligibly affects the high TOF of the Rh overlayer. Consequently, the Rh–Zr overlayer successfully mitigates the thermal deactivation caused by the thermal aging at 900°C. A similar thermal stabilizing effect of the Zr buffer layer was confirmed for other metal overlayer structures comprising Pt and Ir.

Fig. 9

SEM and elemental mapping cross-section images of thermally aged Rh overlayers (a) without and (b) with a Zr buffer layer. Rh and Zr were deposited with 2,000 and 20,000 AP pulses, respectively. Reproduced from Ref. 26) with permission from The Royal Society of Chemistry.

7. Conclusions and Perspectives

This study demonstrates an example of an unusual, but possible, catalyst structure based on the metal overlayer, which can achieve a high catalytic performance comparable with or superior to nanoparticles because of extremely high TOFs. This improvement may enable the simple design of nonporous and low-surface area catalysts comprising thin metal films. However, this phenomenon strongly depends on the types of metals and catalytic reactions. Furthermore, there is insufficient knowledge of which reactions can satisfy the criteria. According to the literature,^23,27) several reactions, such as CO–NO, are known to be “structure sensitive” and their TOFs depend positively on the particle size of certain metals. These are typical candidates that can be applied to metal overlayer catalysts. To date, we are working toward developing other high-TOF reactions, especially in energy production and environmental protection. Furthermore, the design of structured and composite overlayers to further enhance the catalytic activity and stability is another study that should be conducted.^28–30) Fundamentally, elucidating the catalytic reaction mechanisms for high-TOF reactions is an essential issue. Because the mechanism differs from that of nanoparticle catalysts, new approaches based on surface science, DFT simulation, and material informatics are necessary.

Acknowledgments

This work was supported by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) program, “Element Strategy Initiative to Form Core Research Center (JPMXP0112101003)” which is run by MEXT, Japan, since 2012 and JSPS KAKENHI (Grant Number 22H00277). In addition, the authors acknowledge the contribution of Dr. Kenichi Koizumi and Prof. Masahiro Ehara of the Institute for Molecular Science in the DFT studies.

REFERENCES

• 1)   S.H.  Oh,  G.B.  Fisher,  J.E.  Carpenter and  D.W.  Goodman: J. Catal. 100 (1986) 360–376. doi:10.1016/0021-9517(86)90103-X
• 2)   S.H.  Oh and  C.C.  Eickel: J. Catal. 128 (1991) 526–536. doi:10.1016/0021-9517(91)90310-Z
• 3)   H.  Yoshida,  K.  Koizumi,  M.  Boero,  M.  Ehara,  S.  Misumi,  A.  Matsumoto,  Y.  Kuzuhara,  T.  Sato,  J.  Ohyama and  M.  Machida: J. Phys. Chem. C 123 (2019) 6080–6089. doi:10.1021/acs.jpcc.9b00383
• 4)   S.  Misumi,  H.  Yoshida,  S.  Hinokuma,  T.  Sato and  M.  Machida: Sci. Rep. 6 (2016) 29737. doi:10.1038/srep29737
• 5)   S.  Misumi,  H.  Yoshida,  A.  Matsumoto,  S.  Hinokuma,  T.  Sato and  M.  Machida: Top. Catal. 70 (2017) 955–961. doi:10.1007/s11244-017-0760-x
• 6)   S.  Misumi,  A.  Matsumoto,  H.  Yoshida,  T.  Sato and  M.  Machida: Appl. Surf. Sci. 429 (2018) 81–86. doi:10.1016/j.apsusc.2017.06.310
• 7)   S.  Hinokuma,  S.  Misumi,  H.  Yoshida and  M.  Machida: Catal. Sci. Technol. 5 (2015) 4249–4257. doi:10.1039/C5CY00636H
• 8)   S.  Hinokuma,  K.  Murakami,  K.  Uemura,  M.  Matsuda,  K.  Ikeue,  N.  Tsukahara and  M.  Machida: Top. Catal. 52 (2009) 2108–2111. doi:10.1007/s11244-009-9387-x
• 9)   S.  Hinokuma,  Y.  Katsuhara,  E.  Ando,  K.  Ikeue and  M.  Machida: Catal. Today 201 (2013) 92–97. doi:10.1016/j.cattod.2012.03.063
• 10)   S.  Hinokuma,  H.  Fujii,  Y.  Katsuhara,  K.  Ikeue and  M.  Machida: Catal. Sci. Technol. 4 (2014) 2990–2996. doi:10.1039/C4CY00291A
• 11)   H.  Yoshida,  N.  Yamashita,  S.  Ijichi,  Y.  Okabe,  S.  Misumi,  S.  Hinokuma and  M.  Machida: ACS Catal. 5 (2015) 6738–6747. doi:10.1021/acscatal.5b01847
• 12)  M. Maattanen and R. Lylykangas: SAE Paper (1990) 900505.
• 13)  M. Nonnenmann: SAE Paper (1990) 900270.
• 14)  R.M. Heck, R.J. Farrauto and S.T. Gulati: Catalytic Air Pollution Control: Commercial Technology, (Wiley, 1995).
• 15)   M.  Machida,  Y.  Tokudome,  A.  Maeda,  Y.  Kuzuhara,  T.  Hirakawa,  T.  Sato,  H.  Yoshida,  J.  Ohyama,  K.  Fujii and  N.  Ishikawa: ACS Catal. 10 (2020) 4677–4685. doi:10.1021/acscatal.0c00542
• 16)   M.  Machida,  Y.  Tokudome,  A.  Maeda,  T.  Koide,  T.  Hirakawa,  T.  Sato,  M.  Tsushida,  H.  Yoshida,  J.  Ohyama,  K.  Fujii and  N.  Ishikawa: ACS Omega 5 (2020) 32814–32822. doi:10.1021/acsomega.0c05443
• 17)   Y.  Nagao,  Y.  Nakahara,  T.  Sato,  H.  Iwakura,  S.  Takeshita,  S.  Minami,  H.  Yoshida and  M.  Machida: ACS Catal. 5 (2015) 1986–1994. doi:10.1021/cs5020157
• 18)   M.  Machida,  S.  Minami,  H.  Yoshida,  S.  Hinokuma,  Y.  Nagao,  T.  Sato and  Y.  Nakahara: J. Phys. Chem. C 119 (2015) 373–380. doi:10.1021/jp509649r
• 19)   H.  Asakura et al.: J. Am. Chem. Soc. 140 (2018) 176–184. doi:10.1021/jacs.7b07114
• 20)   H.  Yoshida,  R.  Kakei,  Y.  Kuzuhara,  S.  Misumi and  M.  Machida: Catal. Today 332 (2019) 245–250. doi:10.1016/j.cattod.2019.01.020
• 21)   M.  Shelef and  G.W.  Graham: Catal. Rev., Sci. Eng. 36 (1994) 433–457. doi:10.1080/01614949408009468
• 22)   K.Y.S.  Ng,  D.N.  Belton,  S.J.  Schmieg and  G.B.  Fisher: J. Catal. 146 (1994) 394–406. doi:10.1006/jcat.1994.1077
• 23)   C.H.F.  Peden,  D.N.  Belton and  S.J.  Schmieg: J. Catal. 155 (1995) 204–218. doi:10.1006/jcat.1995.1204
• 24)   H.  Permana,  K.Y.  Simon Ng,  C.H.F.  Peden,  S.J.  Schmieg and  D.N.  Belton: J. Phys. Chem. 99 (1995) 16344–16350. doi:10.1021/j100044a022
• 25)   V.P.  Zhdanov and  B.  Kasemo: Surf. Sci. Rep. 29 (1997) 31–90. doi:10.1016/S0167-5729(97)00009-5
• 26)   H.  Yoshida,  S.  Misumi,  A.  Matsumoto,  Y.  Kuzuhara,  T.  Sato,  J.  Ohyama and  M.  Machida: Catal. Sci. Technol. 9 (2019) 2111–2117. doi:10.1039/C9CY00348G
• 27)   R.  Si,  J.  Raitano,  N.  Yi,  L.  Zhang,  S.W.  Chan and  M.  Flytzani-Stephanopoulos: Catal. Today 180 (2012) 68–80. doi:10.1016/j.cattod.2011.09.008
• 28)   H.  Yoshida,  T.  Koide,  T.  Uemura,  Y.  Kuzuhara,  J.  Ohyama and  M.  Machida: Catal. Today 410 (2023) 102–108. doi:10.1016/j.cattod.2022.04.028
• 29)   H.  Yoshida,  Y.  Kuzuhara,  T.  Koide,  J.  Ohyama and  M.  Machida: Catal. Today 375 (2021) 258–263. doi:10.1016/j.cattod.2020.04.040
• 30)   K.  Hayashi,  S.  Kameoka and  A.-P.  Tsai: Mater. Trans. JIM 62 (2021) 1089–1096. doi:10.2320/matertrans.MT-M2021050