Introduction

Epitaxial coatings and interfaces of complex superconducting materials often exhibit properties far superior to those of their bulk counterparts1,2,3. Novel material properties are expected to arise from structures grown on engineered substrates. The substrate surface catalytic activity is defined as the ability of the surface to reduce the nucleation energy barrier, thus promoting the formation of desirable phases and suppressing secondary reactions4,5. The activity of a substrate is becoming increasingly important. As the materials become more complex, competition from secondary phases becomes an acute problem. The traditional approach has utilized an empirical search for compatible substrate materials6. Purposeful engineering of the substrate surface may allow for high performance films using robust and practical simple oxide templates.

CeO2, ceria is a promising simple oxide7. Its high melting temperature and cubic fcc (fluorite) structure make it an ideal substrate. A bulk (001) termination of ceria is charged (polar), which means that it undergoes a stabilization process by reconstruction and/or relaxation8. Stable non-polar surfaces of cubic oxides, such as (001) for the rock salt-type (MgO, NiO) and (111) for fluorite-type (CeO2) received much attention over the last several decades, being the subjects of intense experimental studies by a variety of techniques9. Polar surfaces of epitaxial films, such as (111) of MgO and (001) of CeO2 remain poorly explored.

In this work we manipulate the stabilization process in an effort to enhance the catalytic activity of the (001) ceria surface. We employ low-energy electron diffraction (LEED) and X-ray surface diffraction in order to relate the structure of the (001) ceria surface to its ability to nucleate high-temperature superconducting YBa2Cu3O7 (YBCO) during an ex-situ growth process10. The ex-situ growth process involves a chemical reaction, a reduction of the fluorinated precursor11. The precursor reduction is a rate-limiting step with requires a catalytically active substrate at the YBCO nucleation stage. We show that the nucleation process is strongly affected by the symmetry and periodicity of the surface structure of the (001) ceria termination.

Results

The 70 nm thick (001) ceria films were deposited on (001) yttrium-stabilized zirconia (YSZ) substrates by pulsed laser deposition, followed by annealing for 30 min at 900°C in various environments as follows: i) Sample Ce(H2-H2O): Mixture of 3.5% H2, balance Argon saturated with water vapor; the mixture included H2 and H2O partial pressures at approximately equal values, pH2pH2O ≈ 25 Torr; ii) Sample Ce(H2O): 20 Torr of water vapor; iii) Sample Ce(O2): 100% Oxygen. The samples were cooled down in the processing atmosphere and exposed to the ambient during the transfer for the LEED measurement.

Fig. 1 compares LEED reciprocal space maps (RSM) and AFM height maps of an as-grown sample (panels (a) and (c)) to that of a sample processed for 30 min at 900°C in 3.5% H2-Ar mix saturated with water vapor (panels (b) and (d)). The Atomic Force Microscopy (AFM) height demonstrates a typical transformation of a rough as-deposited surface of CeO2 into a conglomeration of flat-top grains. The corresponding LEED RSMs are shown in Fig. 1a,b. The RSM of the as-deposited sample is comprised of diverging streaks (Fig. 1a) which become vertical lattice rods after annealing (Fig. 1b) indicating formation of a structurally-ordered flat surface. The X-ray reflectivity detected a change of surface roughness from 2.5 nm in the as-deposited sample to 0.5 − 0.3 nm after annealing, thus confirming that post-annealing produces dense films with less than unit cell roughness (Supplementary Fig. 1Sa).

Figure 1
figure 1

Surface morphology transformation of the as-deposited CeO2 buffer after post-annealing.

Reciprocal space LEED maps (a,b) and AFM height maps (c,d) demonstrate the transition from a small-grain surface of the as-deposited sample (a,c) to flat-top, large grain structure after the annealing in H2 –H2O mixture for 30 min at 900°C, (b,d).

We determined the principal surface lattice parameter of all the three samples using LEED patterns, the Ag (001) surface was utilized as the calibration standard. These lattice parameters are 0.37 ± 0.02 nm, which is within the experimental error of the lattice constant of a bulk surface, 0.38 nm. The in-plane and out-of-plane bulk lattice parameters, measured by X-ray diffraction, were observed to be 0.542 ± 0.03 nm, with the sample-to-sample variation within the experimental error. The bulk structure of the film was shown to be consistent with cubic cerium oxide CeOx with x ≈ 2. The X-ray photoemission spectroscopy (XPS) analysis of samples Ce(H2-H2O), Ce(O2) and Ce(H2O) detected ≈ 10% Ce3+, independently confirming predominantly Ce4+ surface oxidation state (see Supplementary Fig. 1Sb,c,d). The XPS experiment was repeated after the LEED measurement, which also included annealing the sample at 600°C for 30 min in a UHV environment prior to the XPS spectra acquisition. The XPS spectra remained unchanged, confirming that the UHV exposure at 600°C did not affect the oxidation state.

Fig. 2 is a side-by-side presentation of the LEED patterns (averaged over fixed energy patterns acquired in the 10–45 eV range) for the samples a) Ce(H2-H2O), b) Ce(H2O) and c) Ce(O2). In the following discussion we use the 2D surface cell of the unreconstructed (001) Ceria surface (lattice constant 0.38 nm) as the reference structure, the reconstructions being defined as modifications of the surface cell. Boundaries of the first Brillouin zone (BZ) of the unreconstructed surface cell are indicated by dashed lines. The sections of the first (BZ) along [01] and [11] directions are shown in panels Fig. 2 d,e and Fig. 2f,g, respectively. The diffraction pattern of the Ce(H2-H2O) sample indicates an unreconstructed surface. The RSMs of this sample (Fig. 2d,f) reveal only the first order (01) and (11) reflections. In contrast, the LEED patterns from samples Ce(H2O) and Ce(O2) contain well-defined fractional order reflections, some indicated by arrows in Fig. 2b,c. The corresponding RSMs in Fig. 2e,g demonstrate that positions of the fractional reflections do not depend on the electron beam energy. This rules out a contribution by surface faceting.

Figure 2
figure 2

Reconstructions of (001) ceria surface.

LEED diffraction patterns of samples (a) Ce(H2-H2O); (b) Ce(H2O) and (c) Ce(O2). The patterns show distinct reconstruction reflections, some pointed out by arrows. (d–f) Panels (d) and (f) compare the reciprocal space sections of the apparently unreconstructed surface of the sample Ce(H2-H2O) with that of samples Ce(H2O) (panel e) and Ce(O2) (panel g). The arrows indicate fractional reflections. The dashed lines in the LEED patterns outline the first BZ.

With the exception of the Ce(H2-H2O) sample, the LEED patterns in Fig. 2 did not change as the sample was cooled from 600°C. Fig. 3 shows the change of the LEED pattern of the Ce(H2-H2O) sample upon cooling from (a) 600°C to (b) 300°C. The pattern at 600°C contains the primary reflections of the first BZ. The emergence of multiple fractional reflections was observed during the cooling, ½ order being the most prominent. Additionally, we observe several fractional reflections around the principal (01) peak.

Figure 3
figure 3

Ordering of Ce(H2-H2O) sample at 300°C.

LEED diffraction patterns of Ce(H2-H2O) sample (a) at 600°C and (b) at 300°C. The fractional reflections are labeled with arrows. The dashed lines outline the first BZ.

We conducted several experiments aimed at assessing the stability of the surface reconstructions to oxidation. Supplementary Fig. 2S compares LEED patterns of sample Ce(H2O) before and after the oxidation at 870°C in 300 mTorr of oxygen for 30 min. The pattern retained the characteristic fractional reflections, confirming the oxidation resistance of the surface structure. Complementary surface characterization using differentially pumped Reflection High-Energy Electron Diffraction (RHEED) documented the effect of oxygen exposure on the surface structure. Supplementary Fig. 3S compares LEED and RHEED patterns of Ce(H2O) sample. The RHEED pattern was acquired at 5 × 10−6 Torr of oxygen partial pressure, 600°C, which is well within Ce4+ stability range9. The presence of well-defined fractional reflection in the RHEED pattern suggests that the surface structure is comprised of Ce4+ ions.

Fig. 4a–c compares the nucleation density of 0.8 μm thick YBCO superconductor films deposited on the three reconstructed substrates used in the study. The YBCO grains can be seen as dark circles on a light background; a secondary Ba-Cu-O phase conglomerate is shown as grayish areas. The phases were identified by cross-referencing the optical images with the Raman spectra. We observe that the Ce(H2O) sample has the highest activity, which corresponds to the highest density of YBCO grains. Fig. 4d,e shows a comparison of the field dependences of the critical current density at 77 and 30 K correspondingly. The film deposited on Ce(H2O) substrates demonstrates approximately twice the critical current density of Ce(H2-H2O) or Ce(O2) substrates, both at 77 and 30 K.

Figure 4
figure 4

Effect of the surface reconstruction on the substrate activity.

Optical micrograph of YBCO grains (dark disks) and Ba-Cu-O secondary phase conglomerates (grayish areas) embedded in a precursor film deposited on (a) Ce(H2-H2O), (b) Ce(H2O) and (c) Ce(O2) buffers. Panels (d) and (e) compare field dependences of the critical current density at 77 K and 30 K correspondingly.

Discussion

Cerium oxide is a compound with high melting temperature (~2500°C). This explains the very limited mobility of Ce atoms at practical deposition temperatures (<1000°C). Typical as-deposited Ceria films can be described as Zone 2 films using Thornton's classification scheme12. Zone 2 films are comprised of dense arrays of columnar grains with faceted surfaces and voided grain boundaries. Since the (111) surface has the lowest energy in a wide range of oxidation potentials (1.41 J/m2 for (111) vs. 0.68 J/m2 for (100))13, columnar (001) CeO2 films are expected to be terminated by pyramid-like (111) terminations. Such surface structures were indeed observed by electron and tunneling microscopy methods14,15. A faceted surface is identified in a LEED experiment via the dependence of the diffraction peak positions in the lateral reciprocal space coordinates, kx−y, on the incident beam energy E16.

Fig. 1 shows the juxtaposition of an AFM height map of an as-deposited sample with its kx−y - E RSM. One can clearly see that cusp-like terminations of the Ceria grains in as-deposited film produce energy-dependent reflections which appear as tilted rods on the map. Post-annealing causes lateral grain growth with the formation of flat (001) terminations which replace cusp-like (111) facets17. This change in morphology is reflected in significant smoothing of the surface roughness and development of vertical lattice rods on the reciprocal space map (Fig. 1b). Ceria films on YSZ exhibit classical normal grain growth18. The lateral grain growth is checked by the thermal grooving effect19 when the grain size becomes approximately equal to the film thickness. The final film morphology is the result of a balance between grain boundary energy and energy of the free surface. As a result, (001) terminations, which are practically absent in a free standing Ceria nano-crystal20, become dominant features of the film surface (Fig. 1d)21,22.

Fig. 2 exhibits the central result of the paper. The origin of the surface reconstruction variety comes from the polar nature of the (001) surface of ceria23. An unreconstructed polar surface is intractable due to the infinite energy associated with the non-zero electric field23. The infinite energy problem can be resolved by removing 50% of the anions from one of the terminations (effectively placing them on the other side of the slab)24, which eliminates the internal electric field and renders the surface non-polar25. Indeed, a study of the (001) Ceria surface by the angle-resolved mass spectroscopy of recoiled ions has shown that half of the oxygen ions are missing from the (001) termination26. Partial reduction of the surface Cerium to 3+ would require more than 50% oxygen desorption in order to maintain the charge neutrality. The oxidation levels used to heat treat the sample are well within the Ce4+ stability range27, which is corroborated by XPS, X-ray diffraction and recent density functional theory (DFT) calculations9,28. We estimate that up to 25% of Ce3+ ions in the top layer would not be detected by XPS, which sets the lower range of the surface oxygen coverage, θ, at 0.45.

Once ≈ 50% of the oxygen atoms are desorbed, ordering of the remaining atoms can be described by lattice gas theory29,30 as an absorbate on a square lattice with the average absorption site occupancy θ ≈ 0.5. Lattice gas theory has been successful in predicting the ordering of H, O, S on low-index metal surfaces31. An important difference is that the coverage of a metal surface by a gas depends on the chemical potential of the absorption reaction. Coverage of a reconstructed oxide surface is fixed by the requirement of charge neutrality. In our model we assume that the surface stays oxygen terminated, which is supported by DFT calculations13,32. The total interaction energy of the lattice gas is:

Where ci is the local occupation variable, ci = 1 if the site i is occupied and zero otherwise, φij is equal to either φnn for the next neighbor pair or φnnn for a next nearest pair and φt is a three atom interaction term. One can easily show that Eq. 1 is equivalent to a spin-spin interaction term of an Ising magnet33.

The simplest case is a purely repulsive interaction between the nearest neighbors: φnn < 0, φt = φnnn = 0. The model predicts ordering of the gas into a ×2 lattice (similar to the Ising antiferromagnetic state) below a critical temperature, Tc, where Tc ~ φnn. The ordered state exists in the range of coverage 0.35 < θ < 0.64, with the maximum Tc at θ = 0.5. Fig. 5c shows the structure model of the ×2 (001) ceria surface that describes the reconstruction of the Ce(O2) sample. The model diffraction pattern, which assumes equal distribution of 90° c(2 × 1) rotational domains, contains characteristic fractional (½ ½) reflections, consistent with the experimentally-observed LEED pattern of the Ce(O2) sample, as shown in Fig. 2c. The oxygen vacancy rows in Fig. 5c are parallel to <100> direction which differs from the checkerboard pattern of the low-energy reconstructed (001) surface proposed by Conesa25. Even though electrostatic calculations25 showed that the the checkerboard pattern has relatively low energy, the diffraction produced by this surface does not contain the additional (½ ½) reflections, observed in the experiment (see Fig. 2c). The I-V curves calculated for this structural model (solid lines) are compared with the experimental ones (open symbols) in Fig. 5h. The results of the LEED intensity analysis suggest that the proposed model is compatible with the experimental data, thus identifying the surface of the Ce(O2) sample as an ordered lattice gas phase with purely repulsive nearest neighbor interaction.

Figure 5
figure 5

Possible surface O2− ion arrangements of reconstructed CeO2 (001) terminations.

The p(1 × 1) structure, (a), is modeled as random distribution of oxygen vacancies. The bottom row, (d)–(f) presents corresponding diffraction patterns calculated within the kinematic approximation; patterns (e) and (f) are calculated taking into account two orthogonal orientations of domains shown in (b) and (c). The dashed square outlines the first BZ of the unreconstructed surface. The arrows point out the extra reflections due to oxygen ordering. The shaded square in panels (a)–(c) represents 0.38 × 0.38 nm attachment site of a c-axis oriented YBCO nucleus. In panels (b) and (c) the unit cells are outline by a bold line. Note that p(4 × 1) domain of the Ce(H2O) sample offers a structurally perfect attachment site, while c(2 × 1) and p(1 × 1) have, on average, two vacancies per site. Panels g)–h) present results of quantitative LEED beam intensity analysis (solid line) of (00) and (01) beams of p(1 × 1) and c(2 × 1) arrangements.

The absence of super-lattice reflections in the LEED pattern from the Ce(H2-H2O) sample, Fig. 2a, is then explained in the context of the low ordering temperature, Tc of this surface. At 600°C the top oxygen is in a disordered gas phase. Therefore, the corresponding LEED pattern can be modeled assuming a 50% top oxygen occupancy of an unreconstructed surface, or a p(1 × 1) reconstruction, see Fig. 5a. The calculated I-V curves for (00) and (01) beams (solid lines) are compared with the experiment (open symbols) in Fig. 5g. The best fit, Rp ≈ 0.4, is obtained for the oxygen vacancy model with the first interlayer equal, within the error bar of the analysis, to the bulk value. Thus, on the basis of these results we can conclude that the model with randomly-distributed oxygen atoms, Fig. 5a, is valid.

Fig. 3b shows that the disordered oxygen vacancy states of the Ce(H2-H2O) samples order into a p(2 × 2) phase at ≈ 300°C. The satellite spots surrounding the principal (01) reflections are known to originate from incommensurate overlayers34. This strongly suggests the presence of an intermediate incommensurate phase, as predicted by the lattice gas theory for certain combinations of φnn and φnnn33.

Finally, the sample Ce(H2O) presents the most interesting case of a long-period structure. According to the lattice gas theory, ×4 reconstruction requires strong long range coupling φt in addition to the next nearest neighbor attraction29,33. We base our structure model on limited scanning tunneling microscopy (STM) data of heavily reduced CeO2 single crystal (001) termination with ×3 reconstruction (Norenberg et al.35). The STM images show that the surface oxygen orders in row-like arrangements with 90° rotational domains. A plausible structure that is consistent with the existence of ¼ fractional diffraction spots of the Ce(H2O) sample is shown in Fig. 5b. Here rows of the top layer oxygen order with ×4 periodicity. The arrangement can be described as interleaving Ce and O terminated p(4 × 1) domains. The calculated diffraction pattern (Fig. 5b) which includes reflections from two kinds of rotational domains, is in qualitative agreement with the experiment (Fig. 2b). Currently, a quantitative analysis of the structure is not feasible due to computational limitations.

Our model explains the variety of surface structures of the (001) ceria termination by different couplings between the surface oxygen. The coupling strength is determined by the subsurface cerium atom structure which forms during transformation of (111)- terminated grains into much larger flat-top grains (see Fig. 1). This explains the stability of the reconstructions: the oxygen surface coverage is fixed by the charge neutrality and the coupling strength is determined by the practically immobile cerium lattice. The origin of the cerium atom ordering lies in the still unknown mechanism of surface reconstruction and simultaneous grain growth during the annealing process. In this work the oxidation state was measured after exposing the samples to air. Therefore, we cannot exclude a potential role of the surface Ce3+ ions in the initial reconstruction process. The in-situ XPS measurements of Ceria surfaces show a substantial surface reduction under conditions that correspond to the Ce4+ state of bulk ceria36. It is likely that the difference between the three types of surface studied in this work can be traced to a different concentration of surface Ce3+ ions that existed during the annealing: one can reasonably expect the highest concentration of Ce3+ for the Ce(H2-H2O) sample and practically zero concentration for the Ce(O2) sample.

Most intriguing is the role of water vapor in enabling the long-range coupling and formation of long-period surface order as seen in Fig. 2b. A tendency of water vapor towards the surface reduction has indeed been observed experimentally by XPS37. This effect cannot be easily explained by a strictly thermodynamic argument, since the Gibbs energy of bulk ceria reduction by water is positive. It has been argued that the absorbed water or hydroxyl molecules promote diffusion of bulk oxygen vacancies towards the surface, thus effectively generating Ce3+ ions37. This argument is indirectly supported by the recent work of Azimi et al.38, who reported an unusual hydrophobic character of a rare-earth oxide due to shielding of 4f orbitals of rare-earth atoms by the outer 5s2p6 electrons. The observation implies that hydroxyl groups interact weakly with the surface cations and more strongly with the surface oxygen. This may explain the reducing action of water vapor and emergence of long-period ordering. A high-temperature DFT calculation, similar to the one performed for the non-polar (111) surfaces39, might provide a more quantitative description of the role of water vapor and Ce3+ ions in the high-temperature reconstruction.

We use a simple geometrical argument in order to visualize the effect of surface reconstruction on surface performance as a catalyst for YBCO nucleation. The experimental data shown in Fig. 4 suggest that the surface with the long-period reconstruction has the highest activity and yields epitaxial layers with best performance. The minimum size c-axis oriented YBCO nuclei would require a 0.38 × 0.38 nm square attachment site, which is superimposed on the structural models in Fig. 5 a–c. Both c(2 × 1) and p(1 × 1) domains have, on average, two oxygen vacancies per attachment site. In contrast, the long period p(4 × 1) domain offers a site free of desorbed oxygen, Fig. 5b. This model attributes the superior performance of the Ce(H2O) sample to the lower energy of the YBCO nuclei, which results in smaller YBCO grain size. Reduced YBCO grain size has a positive effect on the intra-grain Jc40 and the grain coupling41.

Summary

In summary, we show that post-annealing of a polar (001) Ceria surface can produce stable surface reconstructions, likely as a result of surface oxygen ordering. The reconstructions have markedly different surface activities, as demonstrated by the ability of the substrates to nucleate the high-temperature superconductor YBCO. The formation of desirable long-period reconstruction is related to a poorly understood high-temperature interaction of the (001) Ceria surface with water vapor. We suggest that in-situ surface X-ray diffraction could contribute to a better understanding of this effect and open new avenues for designing a highly active (001) surface.

Methods

Ceria film growth

The (001) CeO2 films were deposited on 10 × 10 mm (001) YSZ substrates (MTI Corporation) by pulsed laser deposition (PLD). The substrates were attached to a heater block by silver paste. Thereafter, the temperature of the substrate was raised to 600°C following the evacuation of the deposition chamber to 10−6 Torr. The deposition was carried out at 120 mTorr oxygen partial pressure, using a 5 Hz repetition rate of ~3 J pulses, which corresponded to an average deposition rate of 0.2 nm/s. The film average film thickness was 70 nm as confirmed by scanning electron microscopy and X-ray reflectivity, see corresponding profiles in Fig. 1aS in the Supplementary Section.

Surface analysis

The bulk structure of the samples was characterized using the single crystal X-ray diffraction beamline X-18A at NSLS. Both in-plane and out-of-plane reflections were used to determine the corresponding lattice constants. The surface morphology of the samples was inspected by Atomic Force Microscopy (AFM) prior to and after the heat treatments. X-ray photoelectron spectroscopy (XPS) was performed in an ultrahigh vacuum (UHV) chamber equipped with a SPECS Phoibos 100 MCD hemispherical analyzer. Al-Kα X-ray radiation was used for the XPS studies. The background pressure of 5.0 × 10−10 Torr was maintained throughout the XPS spectra acquisition. The X-ray reflectivity profiles were measured in a Rigaku Ultima III diffractometer in the parallel beam configuration. The Low Energy Electron Diffraction (LEED) experiment was performed using an Elmitec Low Energy Diffraction Microscope (LEEM) attached to the NSLS U5A beamline. We have determined that anneal at 600°C in UHV for 30 min removes the surface contaminants and permits acquisition of a usable LEED pattern. The 600°C sample temperature was maintained during the LEED experiment in order to avoid surface charging. The LEED patterns were acquired using a 5 micron aperture at the beam energies ranging from 10 to 45 eV with 0.1 eV intervals. A pattern from an unreconstructed Ag (001) surface was used as a calibration reference for the lattice parameter calculations. The Reflective High-Energy Electron Diffraction (RHEED) was performed using K-Space Associates differentially pumped RHEED system operating at 18 KV. The Raman microscopy was performed using an Alpha 300 R Raman microscope (WITec).

Structure modeling

The diffraction patterns from the model structures were generated using the kinematic approximation. The structure models were further refined using the dynamic approach. Intensity versus electron energy (I-V) curves were calculated using the Tensor LEED (TLEED) model42. The phase shifts for O and Ce were evaluated from the muffin-tin potential calculated for CeO2 by means of the Barbieri-van Hove program package (http://www.icts.hkbu.edu.hk/surfstructinfo/SurfStrucInfo_files/leed/). The LEED I-V curves of the (00) and (01) beams were used in the comparison with the experimental data in the 20–50 eV energy range. The imaginary part of the inner potential was set to 3 eV, whereas the real part was initially set to 10 eV and then optimized in the search of the best fit with the experimental curves. A Debye temperature of 630 K was used to describe the vibrational effects. The Pendry reliability factor (Rp)43 was used in the analysis to judge the agreement between experimental and calculated I-V curves. The computational approach at these low energies was validated on I-V curves of an Ag (001) recorded on the same instrument. The I-V curves calculated for the Ag (001) surface yielded Rp < 0.2.

Epitaxial growth on reconstructed surfaces

The effect of the surface reconstruction on the activity of the substrate was determined by growing epitaxial YBCO film via a fluoro-acetate method10. Fluorinated precursor layers were deposited on the substrate by spin-coating, which yielded 0.8 μm thick YBCO film. The films were processed at 800°C in an atmosphere comprised of 300 mTorr oxygen and 300 mTorr water vapor in order to convert the precursor into an epitaxial YBCO layer44. The YBCO nucleation density was inferred from the optical and the Raman microscopy (Vitec Raman microscope) maps of mechanically polished film samples45. The critical current densities of the samples were measured by the magnetization method in external magnetic fields up to 3 Tesla using a Quantum Design magnetic properties measurement system.