The controlled synthesis of oxide materials exposing only desired facets has been of general interest in the fields of solid state chemistry, material science, and heterogeneous catalysis [1, 2]. Cerium oxides, as the most abundant rare earth materials, have triggered tremendous interest in heterogeneous catalysis in particular, because of their unique capability to store and release oxygen and to facilely switch between Ce⁴⁺ and Ce³⁺ oxidation states [3-6]. As ceria-based materials are increasingly exploited for catalytic applications, molecular-level understanding of the surface chemistry of ceria has become an urgent issue.

Conventionally, single crystals of ceria cut out from the minerals have been used as models for surface chemistry and catalysis studies [7-10]. However, recent studies [11, 12] have pointed to the influence of impurities in ceria single crystals, mainly fluorine, which could cause misinterpretation of the surface chemistry of ceria. Meanwhile, the growth of ceria thin films on planar metal or oxide surfaces offers the advantage of high crystalline quality and purity, and has been used as an alternative method to synthesize model systems for catalytic studies [13-19]. Particularly, the growth of ceria on metal surfaces could be used to prepare (1) sub-monolayer (ML) ceria nanostructures (NSs) that allow the study of the metal-oxide interface [20-23], or (2) ceria thin films exhibiting different surface structures tuned by the interfacial interaction [24-26]. Thus, the versatility of oxide growth on metal single crystals has made it an attractive approach for the preparation of model catalytic systems. However, the growth of ceria thin films exposing high-surface-energy facets has been challenging [27].

Among the low-index facets of ceria, the surface structure of CeO₂(100) has been the least definitive and its catalytic chemistry has thus remained most puzzling [27]. As a polar surface, CeO₂(100) is expected to be highly active for catalytic reactions [28]. However, in practice, ceria nanocubes, which dominantly expose the (100) facets, have also not been found to be as active as ceria polyhedra, which expose mainly the (111) facets [29, 30]. To elucidate the surface chemistry of CeO₂(100), model CeO₂(100) catalysts with well-defined surface structures have been highly desired but are not easy to achieve on metal surfaces. Although there have been a few studies reporting the growth of CeO₂(100) nanoparticles on Ru(0001) [31, 32] and the pre-oxidized surface of Cu(111) [33-35], their surface atomic structures and the origin of formation have not been clear. Moreover, the growth of CeO₂(100) thin films remains to be explored.

Here, we report the growth of well-defined CeO₂(111) and CeO₂(100) on Pt(111) and demonstrate that the interfacial interaction between Pt and ceria could be utilized to selectively synthesize CeO₂(111) or CeO₂(100) thin films. The Pt/CeO_x system has been widely used for catalytic reactions, including CO oxidation [36], the water-gas-shift reaction [4], and the steam reforming reaction [37]. A strong metal-oxide interaction between Pt and ceria [38] has been observed to influence or determine the catalytic performance of Pt/CeO_x catalysts. However, the strong interaction between Pt and ceria has not been well-understood. Particularly, how the interfacial interaction could affect the structure of supported oxides remains largely unexplored. We have thus studied the growth and structures of CeO_x NSs and ultrathin films on Pt(111) to understand the nature of the strong metal-oxide interaction in the Pt/CeO_x system.

The experiment was carried out in a combined UHV system. The scanning tunneling microscope (STM) chamber is equipped with a Createc low-temperature scanning tunneling microscope (LT-STM, base pressure < 4×10^-11 mbar). The preparation chamber is equipped with an X-ray photoelectron spectroscope (XPS), cleaning facilities, and an e-beam evaporator (base pressure < 5×10^-11 mbar). STM measurements were conducted at 78 K, with an electrochemically etched tungsten tip. STM images were processed with SPIP software from Image Metrology, Denmark. XPS spectra were recorded with a SPECS PHOIBOS 100 analyzer using an Al K_α X-ray source (hν = 1486.6 eV) and a Mg K_α X-ray source (hν = 1253.6 eV) at the normal emission angle.

The Pt(111) single crystal (MaTecK) was cleaned by cycles of 1.5 keV Ar⁺ ion sputtering and annealing at 1000 K. Ce atoms were evaporated from a Ce foil (Alfa Assar, > 99.95%) in the e-beam evaporator and the evaporation flux was monitored by a grid at the exit of the evaporator. CeO_x NSs or thin films were prepared on Pt(111) by evaporating Ce atoms in 10^-7 mbar O₂ at 600 K. The deposition rate of ceria was calibrated by STM measurements on the surface coverages of CeO_x NSs and layers. Because as-deposited CeO_x NSs and layers could appear in the form of multilayers, monolayer equivalent (MLE) was used to represent the surface coverage, which was determined from the STM-measured surface area times the layer number of CeO_x NSs or layers. To prepare a ceria thin film that fully covers the metal substrate, a kinetically limited growth process was employed, i.e., the growth of the interfacial layers of ceria at 200 K, which was followed by the deposition of ceria at 600 K.

Fig. 1a shows a 0.4 MLE CeO_x/Pt(111) prepared by evaporating Ce atoms in 1.3×10^-7 mbar O₂ at 600 K. As-grown CeO_x NSs are found both at the step edges and on the terraces of Pt(111) and exhibit sizes ranging from 2 nm to 10 nm. These CeO_x NSs consist of both monolayer (ML) and bilayer (BL) oxide islands. For ML CeO_x NSs, two types of ceria islands are observed, with type-Ⅰ exhibiting the ordered crystalline structure and type-Ⅱ displaying irregular shape and surface structure (Fig. 1b and 1c). Type-Ⅰ ML CeO_x NSs predominantly exhibit the hexagonal shape with diameter < 5 nm. Their surfaces display the hexagonal CeO₂(111) lattice, with a lattice spacing of 3.8 Å (Fig. 1b). In contrast, type-Ⅱ ML CeO_x NSs are more popular on and usually appear larger than type-Ⅰ ML CeO_x NSs on Pt(111). The apparent height of type-Ⅱ ML CeO_x NSs (~2.9 Å) is lower than that of type-Ⅰ ML CeO_x NSs (~3.3 Å). Thus, type-Ⅱ ML CeO_x NSs can be assigned as sub-stoichiometric CeO_x NSs, while type-Ⅰ ML CeO_x NSs expose a stoichiometric CeO₂ surface. Meanwhile, BL CeO_x NSs have an apparent height (~6.0 Å) approximately twice that of ML CeO_x NSs and display mostly the hexagonal shape. The surfaces of BL CeO_x NSs always show the hexagonal CeO₂(111) lattice, with a lattice spacing of 3.8 Å (Fig. 1d).

When the surface coverage of ceria increases, CeO_x NSs start to coalesce with each other (Fig. 2a). Both ML and BL CeO_x NSs can still be observed on the surface. ML CeO_x NSs usually appear sub-stoichiometric, i.e., in the form of type-Ⅱ CeO_x NSs. While type-Ⅰ ML CeO_x NSs retain the truncated triangle shape with ordered CeO₂(111) lattice (Fig. 2b), type-Ⅱ ML CeO_x NSs appear more ordered than in Fig. 1 and display straight step edges. Bright protrusion lines can also be observed on the surface of type-Ⅱ ML CeO_x NSs (Fig. 2c), whereas the apparent height and the atomic lattice suggest these CeO_x NSs are sub-stoichiometric. Meanwhile, the shape of BL CeO_x NSs becomes more triangular (Fig. 2d).

Compared to Fig. 1, the growth of 0.8 MLE CeO_x NSs on Pt(111) led to ML and BL CeO_x NSs with a more triangle-like shape, which could be understood by the preferred growth of CeO_x. When deposited, cerium atoms would likely diffuse in the form of cerium oxide clusters on Pt(111) and attach to the step edges of existing CeO_x NSs. As a hexagonal ML CeO_x NS usually exposes two types of < 110 > -oriented edge structures alternating along the steps (Fig. 2e), we found that type-A edges are usually longer than type-B edges, in consistence with a previous report [39]. During growth, type-A edges might exhibit a lower barrier for the attachment of cerium oxide clusters, which causes the asymmetric growth of CeO_x NS and eventually the formation of triangle-like CeO_x NSs. The hexagonal BL CeO_x NSs also feature two types of edge structures, i.e., < 100 > -oriented edge and < 111 > -oriented edge [40]. Apparently, the < 111 > -oriented edges are more thermodynamically stable and thus favored during the growth of BL CeO_x NSs, leading to the formation of a triangular shape.

When the surface of the sample in Fig. 2a was annealed to 800 K in UHV, the morphology of CeO_x NSs changed drastically (Fig. 2f). ML CeO_x NSs disappeared and all CeO_x NSs became BL CeO_x NSs, indicating that BL CeO_x NSs are more thermodynamically stable. This also implied that the interaction between ceria layers is stronger than the interaction between ceria and Pt(111). Since the surface of BL CeO_x NSs always exhibits the stoichiometric CeO₂(111) lattice (Fig. 2g), the interfacial ceria layer of BL CeO_x NSs is expected to be sub-stoichiometric upon the transformation from ML to BL CeO_x NSs, which is further supported by the Ce 3d spectra from XPS measurements (Fig. 2h). Note that, after annealing at 800 K, the CeO_x NSs become hexagonal, indicating that the two types of step edges exhibit a similar stability in UHV.

As for the surface of 1.6 MLE CeO_x/Pt(111), the annealing at 850 K in UHV resulted in the Pt surface being covered by CeO_x NSs with straight step edges and of BL or tri-layer (TL) thickness (Fig. 3a). The atomic-resolution STM image of the BL CeO_x NSs reveals a (3×3)-CeO₂(111) super-lattice (Fig. 3b), which has been regularly observed in the growth of ceria on Pt(111). The (3×3) super-structure of CeO₂(111) can be ascribed to the formation of a coincidence lattice, i.e., the (3×3)-CeO₂(111) lattice in coincidence with the (4×4)-Pt(111) lattice. On the surface of TL CeO_x NSs, the (3×3) super-structure of CeO₂(111) is less obvious, but the electronic modulation by the Pt substrate can still be observed (Fig. 3c).

Interestingly, there are also numerous small CeO_x clusters on top of TL CeO_x NSs, but not on BL CeO_x NSs, even after annealing at 850 K (Fig. 3d). These CeO_x clusters typically exhibit a size ranging from 1.5 nm to 5 nm. Atomic-resolution STM images of these supported CeO_x clusters show that their surfaces expose the hexagonal CeO₂(111) lattice (Fig. 3e). Note that, ML-, BL- and TL-CeO_x NSs below 5 nm in size are usually not stable on Pt(111) after annealing at 850 K, and coalesce to form larger NSs. The presence of these small and crystalline CeO_x clusters indeed suggests a much stronger interlayer interaction for ceria NSs above TL thickness. The electronic interaction between Pt(111) and ceria could strongly affect the diffusivity and stability of supported CeO_x clusters. However, when the layer thickness of CeO_x NSs exceeds the TL thickness, electron transfer from Pt becomes rather weak, thereby exerting a negligible effect on the diffusion of top-layer CeO_x clusters.

From the above, although ceria NSs tend to wet the Pt(111) surface much better than the Au(111) substrate [41, 42], the growth of ceria NSs on Pt(111) does not follow exactly the layer-by-layer mode. Rather, the Pt(111) surface remains partially uncovered when Ce is deposited onto Pt(111) at elevated temperatures in the O₂ atmosphere. Even for 2 MLE ceria deposition at 600 K, which was followed by annealing at 800 K in 4×10^-7 mbar O₂, pits with a depth of ~0.6 nm (BL thickness) can be easily observed on the surface of ceria thin films on Pt(111) (Fig. 4a–4c). Moreover, when an additional ~2 MLE ceria was deposited at 600 K, the Pt(111) surface is still not fully covered, as evidenced by the number of pits with a depth of ~1.2 nm on the ceria surface (Fig. 4d–4f). Thus, even the deposition of 4 MLE ceria films cannot fully cover the Pt(111) substrate. This could be attributed to the much stronger interaction between ceria layers than between ceria and Pt(111).

To synthesize a well-defined ceria film that fully covers the Pt(111) substrate, a kinetically limited growth procedure was employed [13]. First, Ce atoms were deposited onto Pt(111) at 200 K in 1×10^-7 mbar O₂. After deposition, the Pt(111) surface was covered by small CeO_x clusters, which dispersed homogeneously on the surface and formed nucleation centers for further growth of ceria films on Pt(111) (Fig. 5a and 5b). Then, Ce atoms were deposited onto Pt(111) at 600 K in 1×10^-7 mbar O₂. The as-grown ceria films were further annealed to 900 K in 2×10^-7 mbar O₂ to improve the crystalline quality. This procedure could lead to the formation of a well-ordered CeO₂(111) film, covering fully the Pt(111) surface (Fig. 5c and 5d). On the surface of the flat CeO₂(111) film, both surface and sub-surface O vacancies can be observed, while the (3×3) super-structure of CeO₂(111) is not visible in the STM image (Fig. 5e), indicating that electronic transfer from the Pt(111) substrate imposes little effect on the surface of CeO₂(111) with thickness above four layers. Meanwhile, Ce 3d spectra from XPS measurements show that the CeO₂(111) thin film is indeed dominated by Ce⁴⁺ ions on the film surface (Fig. 5f).

In Section 3.1, we have shown that the first three layers of ceria could be affected by electronic interaction with the Pt(111) substrate. Using the kinetically limited growth procedure described above, a 3.5 MLE ceria thin film was deposited on Pt(111). Instead of annealing in O₂ at 900 K, the ceria film was annealed in UHV at 1000 K, which also resulted in the ceria film that fully covers the Pt(111) substrate. Surprisingly, a drastic change was observed in the morphology of the ceria film (Fig. 6a and 6b). Rather than displaying the hexagonal surface lattice of CeO₂(111), the surface of UHV-annealed ceria film exhibits square or rectangular lattices, which can also be inferred from the shape of the ceria islands. The rectangular ceria islands are typically 10–60 nm in diameter. XPS measurements show that UHV annealing at 1000 K causes significant reduction of the ceria film, such that Ce 3d spectra (Fig. 6c) show the surface to be highly reduced and dominated by Ce³⁺ ions. Thus, the symmetry change of the surface lattice is caused by the reduction of the ceria film. The reduction process is facilitated by the strong metal-oxide interaction between cerium oxide and Pt(111), consistent with the findings from CeO₂-on-Ag catalysts. Chang et al. [43] have reported surface-enhanced Raman spectroscopy (SERS) and catalytic studies of ceria thin layers on Ag particles and showed that Ce(Ⅲ) species could be stabilized by the Ag-CeO₂ interaction.

At the lower layer of the film surface (Fig. 7a), which corresponds to the ceria thin film of TL thickness, an ordered square lattice structure can be easily observed on the film surface. The atomic structure of the square lattice is magnified in Fig. 7a, and shows a lattice spacing of ~1.1 nm, which can thus be attributed to the reconstruction of a (100) surface. Three rotational domains with such a square lattice can be identified on the surface of the TL ceria film, and exhibit a rotational angle of 120° against each other. Previous studies have explored the CeO₂↔Ce₂O₃ transition of the CeO₂ (100) phase on Cu(111) induced by the deposition of Ce atoms [35]. The initial Ce deposition can cause surface reconstruction, to transform p(2×2)-CeO₂(100) into (1×1)-CeO₂(100). Low-energy electron diffraction (LEED) patterns show that further deposition of Ce atoms can lead to a surface reduction to form a c(4×4) structure, which is attributed to the c-Ce₂O₃(100) phase. Likewise, the observed square lattice in Fig. 7a can be assigned as c-Ce₂O₃(100).

The symmetry of thin oxide films grown on planar metal surfaces was usually dictated by the symmetry of the metal substrate. That being said, the growth of ceria on the (111) surface of metals should result in the CeO₂(111) surface, as described in Section 3.1. However, with careful control of the growth parameters, the (100) surface of cerium oxide could also be synthesized on the (111) surface of metal single crystals. Yang et al. [33] have shown that CeO₂(100) could be grown on Cu(111) by pre-oxidizing the Cu(111) substrate to form a surface oxide with rectangular symmetry. Flege et al. [32] showed that the growth of ceria at 830 ℃ or above on Ru(0001) could result in rectangular ceria nanoparticles owing to the formation of a coincidence lattice at the oxide–substrate interface. Höcker et al. [26] showed also that ceria nanoparticles on Ru(0001) could be reduced from CeO₂(111) to cubic Ce₂O₃(111) at 700 ℃ in UHV.

In this study, we showed that the surface symmetry of ceria could be altered simply by its reduction at 1000 K. The electronic interaction between ceria layers and Pt(111) could facilitate the reduction of ceria significantly to induce the formation of Ce₂O₃. Meanwhile, the Pt(111) substrate also provides an excellent substrate to form a coincidence lattice with c-Ce₂O₃(100), which is briefly illustrated in Fig. 7c. The near-perfect matching provides an explanation for the observation of c-Ce₂O₃(100), indicating that the Pt(111) substrate not only facilitates the reduction of cerium oxide via electron transfer but also provides a proper template to stabilize the highly reduced Ce₂O₃ phase.

The top layer (the fourth layer) of the ceria film after UHV annealing exhibits a row-like (3×n) surface structure (Fig. 8a). The atomic-resolution STM image (Fig. 8b) shows that the surface has a square lattice with a lattice spacing of ~3.9 Å, which corresponds to the surface lattice of CeO₂(100). Nörenberg et al. [44] have studied the surface structures of CeO₂(100) single crystals using STM and LEED. After annealing to 900 ℃, the CeO₂(100) surface exhibited the (3×1) and (3×2) reconstruction, which shows a similar structure to the (3×n) rows observed in our study. Thus, the (3×n) reconstruction observed on the ceria film on Pt(111) could be attributed to the surface reconstruction of CeO₂(100), which lowers the charge accumulation on the polar surface of CeO₂(100). In addition, bubble-like square protrusions could also be observed on the surface of the fourth ceria layer and appear ~0.6 Å higher than the neighboring surface plane. These protrusions display the same CeO₂(100) lattice as that of the neighboring surface plane. The height of these protrusions could be due to the inhomogeneity of the ceria layer underneath, which could be inferred from the morphology of the uncovered TL ceria film. Point defects could also be found on the top-layer surface of ceria and as dark holes in the square lattice, which could be attributed to a missing O atom (Fig. 8b).

On top of the ceria film surface, there were also ceria nanoislands, as illustrated in the lower part of Fig. 8a. The surface structure of ceria nanoislands is the same as the top layer (the fourth layer) of the ceria film. However, no (3×n) modulation is observed on the surfaces of ceria islands (Fig. 8c). Rather, these islands exhibit the (1×1)-CeO₂(100) surface lattice. Similar to the growth of CeO₂(111) thin films on Pt(111), the thicker the ceria film is, the less electronic modulation that surface ceria layer receives. Density functional theory (DFT) calculations have proposed a structural model for the reconstructed surface of CeO₂(100), that could compensate for the extra charge of the polar (100) surface while preserving the primitive (1×1) cell [45]. Fig. 8d shows that, by removing half of the outermost O^2- ions, the surface polarity could be compensated and the remaining O^2- ions could be arranged in a checkboard-like pattern, such that the surface remains a (1×1) structure. This polar-compensated (100) structure has also been confirmed by an angle-resolved mass spectroscopy study on the CeO₂(100) film [46]. Note that, previous studies on the growth of CeO₂(100) NSs on metal substrates have not been able to synthesize CeO₂(100) with a (1×1) structure. Our study suggested that the formation of a c-Ce₂O₃(100) buffer layer could serve as a template for the overlayer growth of the CeO₂(100) film and NSs.

On ceria nanoislands, two types of vacancy sites could be identified. Fig. 8e shows a missing atom in the square lattice, which is most likely a surface oxygen vacancy. In contrast, Fig. 8f shows the surface character of a sub-surface vacancy or vacancies, which appear as a cross-like dark area, and no missing atoms can be observed in the surface lattice.

When the (100) surface of the ceria film was annealed further at 900 K in 2×10^-7 mbar O₂, the squared-shaped islands disappeared, along with the squared surface lattice of CeO₂(100). The ceria film transformed again into the CeO₂(111) surface (Fig. 9a). The hexagonal surface lattice of CeO₂(111) could be clearly resolved by the atomic-resolution STM image in Fig. 9b. Note that, the ceria thin film of TL thickness on Pt(111) and exhibiting the CeO₂(111) surface structure could be transformed into the CeO₂(100) surface by annealing the film in UHV. However, such a transformation cannot be observed on thicker ceria films. Once the CeO₂(111) thin film of four-layer thickness is formed, high-temperature annealing or reduction cannot reverse the structure of CeO₂(111) back to CeO₂(100). This is consistent with the growth study observation in Section 3.1 that the fourth layer of CeO₂(111) is not or much less affected by the electronic interaction with the Pt(111) substrate, and thus much less flexible than thinner ceria layers on Pt(111).

While the strong metal-support interaction (SMSI) has often been invoked to explain their unique catalytic properties, the structural dynamics demonstrated above could provide an atomic-scale understanding of catalysis over ceria-supported Pt group metal catalysts. For instance, during the activation of three-way catalysts in a reducing atmosphere, Golunski et al. [47] have observed the migration of partially reduced ceria layers over the surface of Pt particles, which led to the loss of metal surface areas, but could still enhance the three-way catalytic activities of Pt/CeO₂ catalysts, even at low temperatures. While SMSI has been suggested to promote the formation of oxygen vacancies on ceria, we provided atomic-level evidence for the formation of interfacial Ce₂O₃ layer upon reduction, which could be responsible for the enhanced activities of Pt/CeO₂ catalysts [48]. Indeed, Pt nanoparticles with ceria layers have been employed in catalyst synthesis in recent practices and shown remarkable activities for CO oxidation [49] and the water-gas shift reaction [50]. Similar enhancement by ceria overlayers could also be observed on other metal particles, such as Pd [51], Ag [43], and Ir [52]. Thus, our observation of the structural dynamics of ceria thin layers on Pt(111) could be of general relevance to the catalytic interface formed between precious metals and ceria, and provide a profound understanding of the metal-oxide interfacial interaction.

We studied the growth and structures of CeO_x NSs and thin films on Pt(111). At low ceria coverage (0.4 MLE), both ML and BL CeO_x NSs were present on the Pt surface upon the deposition of ceria at 600 K. Two types of ML CeO_x NSs were found on Pt(111), with one type showing an ordered and stoichiometric surface of CeO₂(111) and the other appearing sub-stoichiometric with a disordered surface structure. An increase in the surface coverage of ceria induces a change in the shape of CeO_x NSs into a triangle-like shape and the formation of TL CeO_x NSs. After annealing at 800 K, ML CeO_x NSs transform into the more stable form of BL CeO_x NSs. Both BL and TL CeO_x NSs show that the stoichiometric surface of CeO₂(111) and their interfacial CeO_x layers were easily reduced due to the electronic interaction with the Pt substrate. Small CeO_x clusters supported on TL CeO_x NSs could be observed on the 1.6 MLE CeO_x/Pt(111) surface. These CeO_x clusters also displayed the surface lattice of CeO₂(111) and exhibited a thermal stability much more superior than that of ML, BL or TL CeO_x NSs of similar size on Pt(111). This implied that the electronic interaction with the Pt substrate affects mainly CeO_x NSs of TL thickness or thinner.

The growth of a well-defined ceria film fully covering the Pt(111) substrate cannot be achieved simply by reactive deposition of evaporating Ce atoms in O₂ at elevated temperatures. Rather, a two-step kinetically limited growth procedure needs to be employed in order to obtain a full CeO₂(111) layer covering the Pt(111) substrate. Interestingly, a surface of CeO₂(100) could be achieved when ~3–4 MLE CeO_x was deposited onto Pt(111) and the surface was annealed in UHV at 1000 K. At the CeO_x film surface, the lower layer (the third layer) displays the c-Ce₂O₃(100) phase, which serves as a buffer layer for the growth of CeO₂(100) at the top layer (the fourth layer). The c-Ce₂O₃(100) buffer layer is formed via high-temperature reduction facilitated by the strong electronic interaction between CeO_x and Pt(111). A (3×n) reconstruction could be observed on the surface of the fourth layer of CeO_x. On top of the CeO_x film surface, there were also ceria nanoislands, which exhibit the (1×1)-CeO₂(100) surface lattice. After annealing at 900 K in 2×10^-7 mbar O₂, the surface of the CeO₂(100) thin film is transformed into the hexagonal CeO₂(111) surface. Therefore, the structural dynamics and transition observed on CeO_x NSs and thin films on Pt(111) are determined by the competition between the interlayer interaction of CeO_x and the Pt-CeO_x interaction. Our study provides model systems for further catalytic studies of ceria catalysts and enables molecular-level studies of facet-dependent catalysis of ceria and Pt/CeO_x catalysts.