Skull Melting Growth and Characterization of (ZrO2)0.89(Sc2O3)0.1(CeO2)0.01 Crystals

(ZrO2)0.89(Sc2O3)0.1(CeO2)0.01 crystals have been grown by directional melt crystallization in a cold crucible. The chemical and phase compositions of the crystals have been characterized using energy dispersion X-ray spectroscopy (EDX), Raman scattering spectroscopy and transmission electron microscopy (TEM). The X-ray photoelectron emission method has been used for determining the valence state of the Ce ions. We show that directional melt crystallization produces an inhomogeneous ceria distribution along the crystal length. The as-grown crystals are mixtures of cubic and rhombohedral zirconia modifications. The rhombohedral phase has an inhomogeneous distribution along crystal length. Melt crystallization does not produce single-phase cubic (ZrO2)0.89(Sc2O3)0.1(CeO2)0.01 crystals. The formation of the phase structure in the crystals for different synthesis methods has been discussed.


Introduction
Scandia-stabilized zirconia solid solutions have high oxygen ionic conductivity and show good promise for the production of solid-state electrolytes for electrochemical devices, e.g., solid oxide fuel cells (SOFCs). The highest conductivity for the ZrO 2 -Sc 2 O 3 system was observed in compositions containing 8-12 mol.% of Sc 2 O 3 which had a stable cubic structure at 700-800 • C. With a decrease in temperature, the high-conductivity cubic phase transforms to the rhombohedral β-phase, and the ionic conductivity of these materials decreases dramatically. Earlier data [1][2][3] showed that the formation of the rhombohedral phase in the ZrO 2 -Sc 2 O 3 system solid solutions occurs by ordering of the oxygen vacancies. However, there are indications of cubic phase stabilization in ZrO 2 ceramics containing 12 mol.% of Sc 2 O 3 [4,5]. The formation of the rhombohedral phase in these specimens was attributed to a homogeneous distribution of the Sc 3+ cations in the ZrO 2 matrix, which is achieved due to a low heating rate (~2 • C/min) during solid state synthesis and the presence of a small quantity of impurities [5]. This may destroy the long-range order of the oxygen vacancies and favor the stabilization of the cubic c-phase. The formation of the rhombohedral β-phase in the low-temperature range can also be avoided by co-doping with small quantities of some oxides (Gd 2 O 3 [6], CeO 2 [6], Y 2 O 3 [4], Yb 2 O 3 [7]). For example, introduction of CеО 2 into the ZrO 2 -Sc 2 O 3 system stabilizes the cubic phase since the larger Ce 4+ ions (ionic radius of Ce(VIII) 4+ = 0.97 Å) compared with that of the Zr 4+ ions (0.84 Å) favor the localization of the vacancies in the vicinity of the Zr 4+ cations, thus suppressing the formation of the low-symmetry rhombohedral phase [8]. The most promising composition in the ZrO 2 -Sc 2 O 3 -CеО 2 system is the one containing 10 mol.% of Sc 2 O 3 and 1 mol.% of CеО 2 (10Sc1CeSZ), which has the highest conductivity and highly stable properties under long-term exposure to SOFC working temperatures [9,10].
In our previous work [11], we studied the structural and transport properties of (ZrO 2 ) 1-x-y (Sc 2 O 3 ) x (CeO 2 ) y solid solution crystals (x = 0.08-0.10; y = 0.005-0.015) grown by directional melt crystallization. We found that this synthesis method does not provide single-phase cubic crystals of the studied composition range at room temperature, which is contradictory to the phase analysis data for specimens synthesized using various ceramic technologies. We studied, in detail, the phase composition of the (ZrO 2 ) 0.89 (Sc 2 O 3 ) 0.1 (CeO 2 ) 0.01 (10Sc1CeSZ) specimen in this work and investigated the factors affecting the formation of the crystal structure in this synthesis method. Also, we report data on the behavior of Ce ions during the melt synthesis.

Materials and Methods
(ZrO 2 ) 0.89 (Sc 2 O 3 ) 0.1 (CeO 2 ) 0.01 solid solution crystals were grown by directional melt crystallization in a cold crucible [12]. The crystals were grown by directed crystallization of the melt in a water-cooled copper container using direct high-frequency (5.28 MHz) material heating. The power of the high-frequency generator was 60 kW. The charge was prepared from min. 99.99% purity ZrO 2 , Sc 2 O 3 , and CeO 2 powders. The starting oxides, with a total weight of 6 kg, were pre-mixed mechanically and placed in a container. Metallic zirconium in the charge was used for starting heating. The melt crystallization rate was 10 mm/h. High-frequency heating was ceased after the end of the crystallization process. Further spontaneous cooling of the crystallized ingot occurred. The temperature of the ingot surface was recorded with a Gulton 900-1999 radiation pyrometer in a 2000-1000 • C range and a Pt/Pt-Rh thermocouple in the 1000-500 • C range; the ingot cooling rates in these temperature ranges were~200 • C/min and~30 • C/min, respectively.
The chemical composition of the as-grown crystals was examined with a JEOL 5910 LV scanning electron microscope (JEOL, Tokyo, Kanto, Japan) with an energy-dispersive AZtecENERGY analytical system equipped (Oxford Instruments, X-Max-80, Abingdon, UK). The references were molten ZrO 2 , Sc 2 O 3 and CeO 2 .
Phase analysis was performed by X-ray diffraction on a Bruker D8 diffractometer (Bruker, Billerica, MA, USA) using CuKα radiation. The specimens were cut in the form of plates perpendicular to the <100> direction. For phase analysis, planes with different indexes and different phases were set to the reflecting position. Raman spectra of the specimens were recorded on a Renishaw inVia Raman microscopic spectrometer (Renishaw, Wotton-under-Edge, Gloucestershire, UK) (λexc. = 532 nm) and with a home-made Raman setup that was described in previous works [13,14]. The crystal structure was studied using transmission electron microscopy at a 200 kV accelerating voltage with a JEM-2100 microscope (JEOL, Tokyo, Kanto, Japan). The specimens were prepared by ion etching on a PIPS II (GATAN, Milton Abingdon, UK) setup.
XPS analyses were performed on PHI 5000 VersaProbe II equipment (Physical Electronics, Chanhassen, MN, USA) using monochromatic AlKα radiation with a focused beam diameter of 200 µm. High-resolution (HR) spectra were acquired at an analyzer pass energy of 29.35 eV in increments of 0.25 eV. The binding energy (BE) of adsorbed hydrocarbons peak C1s (285.0 эB) was used as a reference value for adjusting the BE scale. Samples for analysis were obtained by cleaving crystals immediately before loading into the spectrometer. The electrostatic charge caused by photoelectron emission of the dielectric samples was compensated by a charge neutralization system.

Results and Discussion
Literary data described several types of ZrO 2 -Sc 2 O 3 system phase diagrams distinguished by the presence of a number of metastable phases in the solidus region [15][16][17]. It should be noted that the greatest differences between these phase diagrams are observed at temperatures below 1400 • C. However, their common feature, just like for the ZrO 2 -Y 2 O 3 phase diagram, is a high-temperature range of cubic solid solutions (at a stabilizing oxide concentration of 10 mol.%). Therefore, melt-grown 10Sc1CeSZ crystals have a cubic structure, which is retained during cooling until the bottom temperature limit of cubic phase existence. The high-temperature cubic phase is stable up to a temperature of~600 • C for a given composition [17]. Further cooling to room temperature may entail a phase transition to one of the room-temperature stable solid solution modifications (monoclinic, tetragonal or rhombohedral).
The ingot grown by directional melt crystallization contained separate crystals, which had a pillar shape typical of this synthesis method. The lateral crystal dimensions increased with a decrease in the directional crystallization rate from 10 to 3 mm/h, the maximum crystal size being 50 mm in length and a cross-section of 25 mm. Figure 1a shows the typical as-grown crystal appearance. The as-grown crystals had a dark-red color.

Results and Discussion
Literary data described several types of ZrO2-Sc2O3 system phase diagrams distinguished by the presence of a number of metastable phases in the solidus region [15][16][17]. It should be noted that the greatest differences between these phase diagrams are observed at temperatures below 1400 °C. However, their common feature, just like for the ZrO2-Y2O3 phase diagram, is a high-temperature range of cubic solid solutions (at a stabilizing oxide concentration of 10 mol.%). Therefore, meltgrown 10Sc1CeSZ crystals have a cubic structure, which is retained during cooling until the bottom temperature limit of cubic phase existence. The high-temperature cubic phase is stable up to a temperature of ~600 °C for a given composition [17]. Further cooling to room temperature may entail a phase transition to one of the room-temperature stable solid solution modifications (monoclinic, tetragonal or rhombohedral).
The ingot grown by directional melt crystallization contained separate crystals, which had a pillar shape typical of this synthesis method. The lateral crystal dimensions increased with a decrease in the directional crystallization rate from 10 to 3 mm/h, the maximum crystal size being 50 mm in length and a cross-section of 25 mm. Figure 1a shows the typical as-grown crystal appearance. The as-grown crystals had a dark-red color.  Figure 1b shows a plate cut along the crystal growth axis, which clearly shows a non-uniform color distribution in the crystal bulk. The color intensity is lower at the crystal periphery, resulting in colorless regions on the crystal surface, testifying to a decrease in the concentration of the Се 3+ cations, which have absorption bands in the visible spectral region and change the color of the crystals to orange-red [18]. Ingot air cooling causes Се 3+ ions to oxidize to Се 4+ , this process occurs at the crystal periphery. Since an ingot consisting of separate crystals cools down quite rapidly and the oxygen diffusion coefficient decreases dramatically with decreasing temperature, oxygen does not have  Figure 1b shows a plate cut along the crystal growth axis, which clearly shows a non-uniform color distribution in the crystal bulk. The color intensity is lower at the crystal periphery, resulting in colorless regions on the crystal surface, testifying to a decrease in the concentration of the Cе 3+ cations, which have absorption bands in the visible spectral region and change the color of the crystals to orange-red [18]. Ingot air cooling causes Cе 3+ ions to oxidize to Cе 4+ , this process occurs at the crystal periphery. Since an ingot consisting of separate crystals cools down quite rapidly and the oxygen diffusion coefficient decreases dramatically with decreasing temperature, oxygen does not have sufficient time to diffuse to the entire crystal bulk. The darker color of the crystal top part suggests a higher Cе 3+ ion concentration. Figure 2 shows data on component distribution along the 10Sc1CeSZ crystal length suggesting cerium oxide displacement toward the crystal top part during directional crystallization. A similar behavior was observed in cerium-doped cubic ZrO 2 -Y 2 O 3 crystals where the effective cerium distribution coefficient was~0.5 [16,18]. An increase in the impurity concentration at the end of the growth may violate normal crystallization conditions due to constitutional undercooling at the crystallization front causing growth defect formation in the form of growth bands with an elevated concentration of the displaced impurity, cellular structure, inclusions, etc. These defects are most typical of crystals growing in the central part of solidified melt ingots since the axial temperature gradients in the central part of the melt are lower than those at the periphery. This results from specific features of direct high-frequency heating of dielectrics in a cold crucible [16,18].
Crystals 2020, 10, x FOR PEER REVIEW 4 of 10 sufficient time to diffuse to the entire crystal bulk. The darker color of the crystal top part suggests a higher Се 3+ ion concentration. Figure 2 shows data on component distribution along the 10Sc1CeSZ crystal length suggesting cerium oxide displacement toward the crystal top part during directional crystallization. A similar behavior was observed in cerium-doped cubic ZrO2-Y2O3 crystals where the effective cerium distribution coefficient was ~0.5 [16,18]. An increase in the impurity concentration at the end of the growth may violate normal crystallization conditions due to constitutional undercooling at the crystallization front causing growth defect formation in the form of growth bands with an elevated concentration of the displaced impurity, cellular structure, inclusions, etc. These defects are most typical of crystals growing in the central part of solidified melt ingots since the axial temperature gradients in the central part of the melt are lower than those at the periphery. This results from specific features of direct high-frequency heating of dielectrics in a cold crucible [16,18]. It can be seen clearly that, at an early growth stage, the ceria concentration in the solid solution is lower than 1 mol.%; in the central crystal part, it is almost equal to 1 mol %; by the end of the growth, it increases to 1.5-2 mol.%. In the meantime, the scandia concentration decreases along the crystal.
Qualitative assessment of the Се 3+ ion distribution in the crystal length was also carried out using Raman spectroscopy based on the intensity of the peak near 2200 cm −1 corresponding to the 2 F7/2-2 F5/2 electron transition in the Се 3+ ions [19]. Figure 3 shows Raman spectra for different 10Sc1CeSZ crystal regions. It can be seen that the intensity of the peak characterizing the concentration of the Се 3+ ions increases in the top part of the crystal (Figure 3b). This is in agreement with the experimentally observed change in the crystal color along the length of the crystal. It can be seen clearly that, at an early growth stage, the ceria concentration in the solid solution is lower than 1 mol.%; in the central crystal part, it is almost equal to 1 mol %; by the end of the growth, it increases to 1.5-2 mol.%. In the meantime, the scandia concentration decreases along the crystal.
Qualitative assessment of the Cе 3+ ion distribution in the crystal length was also carried out using Raman spectroscopy based on the intensity of the peak near 2200 cm −1 corresponding to the 2 F 7/2 -2 F 5/2 electron transition in the Cе 3+ ions [19]. Figure 3 shows Raman spectra for different 10Sc1CeSZ crystal regions. It can be seen that the intensity of the peak characterizing the concentration of the Cе 3+ ions increases in the top part of the crystal (Figure 3b). This is in agreement with the experimentally observed change in the crystal color along the length of the crystal.
Thus, directional melt crystallization is accompanied not only by an increase in the overall ceria concentration but also by an increase in the Cе 3+ ion concentration in the crystal. In the meantime, the scandia concentration decreases along the crystal.
We used high-resolution X-ray photoelectron spectroscopy for quantitative evaluation of the Ce ion concentration for different valence states. Figure 4 shows the Ce3d high-resolution X-ray photoelectron spectrum for the as-grown 10Sc1CеSZ crystal which was obtained in the same analysis area, with a diameter of 200 µm and at the following time intervals: 0.5, 1 and 48 h after loading and preliminary pumping in the chamber at 5 × 10 −6 Pa. The approximation of the Ce3d doublet spectra (spin-orbit splitting into 3d 5/2 and 3d 3/2 peaks), which have a complex shape due to the presence of satellites structure due to multi-electron interactions, revealed the presence of VU, V"-U" and V "doublets -U "'from Ce 4+ ions and also Vo-Uo doublets, V'-U' from Ce 3+ ions. Thus, directional melt crystallization is accompanied not only by an increase in the overall ceria concentration but also by an increase in the Се 3+ ion concentration in the crystal. In the meantime, the scandia concentration decreases along the crystal.
We used high-resolution X-ray photoelectron spectroscopy for quantitative evaluation of the Ce ion concentration for different valence states. Figure 4 shows the Ce3d high-resolution X-ray photoelectron spectrum for the as-grown 10Sc1СеSZ crystal which was obtained in the same analysis area, with a diameter of 200 μm and at the following time intervals: 0.5, 1 and 48 h after loading and preliminary pumping in the chamber at 5 × 10 −6 Pa. The approximation of the Ce3d doublet spectra (spin-orbit splitting into 3d5/2 and 3d3/2 peaks), which have a complex shape due to the presence of satellites structure due to multi-electron interactions, revealed the presence of VU, V"-U" and V "doublets -U "'from Ce 4+ ions and also Vo-Uo doublets, V'-U' from Ce 3+ ions. The fraction of Ce 3+ ions was determined according to the widely used quantitative method [20,21], as the sum of the relative intensities of the Vo-Uo, V'-U 'peaks in the full Ce3d spectrum. However, the effect of cerium reduction was clearly manifested in this work during research under high-vacuum conditions of 5 × 10 −8 Pa and irradiation by focused X-ray radiation with a diameter of 200 µm and a power of 50 W. A noticeable decrease in the intensity of the peaks corresponding to Ce 4+ was observed after 1 and 48 h of exposure of the sample in vacuum, as can be seen from Figure 4a-c. Estimation of the fraction of Ce 3+ ions showed values of 40%, 50%, and 60%, respectively, depending on the exposure time of the sample in the chamber (0.5, 1, and 48 h). The effect of cerium reduction is well known in studies using this method. As was shown in [22], the exposure of CeO 2 powders in vacuum, the effect of X-ray radiation, and the surface charge effect of the sample affect the charge state of Ce ions. However, the reduction effect of cerium ions manifests itself more noticeably in our case when studying the surface of the cleaved crystal.
X-ray diffraction study of the phase composition of the crystals showed that the crystals consist of two phases, i.e., the cubic and the rhombohedral ones. Raman spectroscopy study of the crystal structure confirms the presence of these two phases (Figure 3a). The general pattern of the Raman spectra is similar to that observed earlier [23,24] for ZrO 2 -Sc 2 O 3 system specimens and contains cubic and rhombohedral phase bands. The intensity of the bands corresponding to the rhombohedral phase increases at the end of the crystal, with the bands becoming more pronounced. Thus, the rhombohedral phase concentration increases along the 10Sc1CeSZ crystal length during directional crystallization. Figure 5 shows TEM images of two regions in the bottom and top parts of the crystal. The bottom part of the crystal contains inclusions of large rhombohedral phase twins in the cubic matrix. The top part of the crystal consists mainly of rhombohedral phase twins. No twin-free rhombohedral phase regions are observed.
Thus, the Raman spectroscopy and TEM results suggest an inhomogeneous rhombohedral phase distribution along the crystal length: the rhombohedral phase content in the top part of the crystal is higher than that in its bottom part. The observed inhomogeneous distribution of the rhombohedral phase along the crystal length may be due to changes in the chemical composition along the crystal length, namely, changes in the total content of сeria, changes in the ratio of Cе 3+ /Cе 4+ , and changes in the scandia content. The fraction of Ce 3+ ions was determined according to the widely used quantitative method [20,21], as the sum of the relative intensities of the Vo-Uo, V'-U 'peaks in the full Ce3d spectrum. However, the effect of cerium reduction was clearly manifested in this work during research under high-vacuum conditions of 5 × 10 −8 Pa and irradiation by focused X-ray radiation with a diameter of 200 μm and a power of 50 W. A noticeable decrease in the intensity of the peaks corresponding to Ce 4+ was observed after 1 and 48 h of exposure of the sample in vacuum, as can be seen from Figure  4a-c. Estimation of the fraction of Ce 3+ ions showed values of 40%, 50%, and 60%, respectively, depending on the exposure time of the sample in the chamber (0.5, 1, and 48 h). The effect of cerium reduction is well known in studies using this method. As was shown in [22], the exposure of CeO2 powders in vacuum, the effect of X-ray radiation, and the surface charge effect of the sample affect the charge state of Ce ions. However, the reduction effect of cerium ions manifests itself more noticeably in our case when studying the surface of the cleaved crystal.
X-ray diffraction study of the phase composition of the crystals showed that the crystals consist of two phases, i.e., the cubic and the rhombohedral ones. Raman spectroscopy study of the crystal structure confirms the presence of these two phases (Figure 3a). The general pattern of the Raman spectra is similar to that observed earlier [23,24] for ZrO2-Sc2O3 system specimens and contains cubic  Thus, the Raman spectroscopy and TEM results suggest an inhomogeneous rhombohedral phase distribution along the crystal length: the rhombohedral phase content in the top part of the crystal is higher than that in its bottom part. The observed inhomogeneous distribution of the rhombohedral phase along the crystal length may be due to changes in the chemical composition along the crystal length, namely, changes in the total content of сeriа, changes in the ratio of Се 3+ /Се 4+ , Unlike the single-phase cubic 10Sc1CeSZ specimens synthesized using the ceramic technology, 10Sc1CeSZ crystallization from melt did not provide for cubic phase stabilization in the entire crystal bulk.
We can specify several substantial differences which may affect cubic phase stabilization in the melt-grown and the ceramic 10Sc1CeSZ specimens. First, cubic phase stabilization in the ceramic specimens occurs during sintering in a specific temperature range. If Daiichi Kigenso Kagaku Kogyo (DKKK) and Praxair powders were used as the source material, single-phase cubic specimens were only obtained at sintering temperatures of >1300 • C [25]. For coprecipitated powders, the respective temperature was 1500 • C [26]. The thermal conditions of the synthesis for the melt-crystallization-grown specimens differed considerably from those for the synthesis of ceramics. For directional crystallization from melt, the crystals start to grow at the crystallization front at the solidus temperature (~2800 • C). Then, as the crystals grow, the crystalline ingot first cools down gradually to 1200-1400 • C, and further cooling to room temperature occurs spontaneous.
Secondly, ceramic technologies typically use source materials containing considerable (to 2 mol.%) quantities of impurities such as Si, Na, Ti and Al [24,26]. The materials used for crystal growth are of much higher purity and, furthermore, directional crystallization causes additional crystal purification due to impurity displacement or evaporation.
Thirdly, the presence of grain boundaries in ceramic specimens can stabilize the high-temperature phases due to impurity segregation and inhomogeneous distribution of the stabilizing oxides in the grain volume.
Furthermore, it is believed that cerium ions exist in the solid solution of 10Sc1CeSZ ceramic specimens in the form of Cе 4+ ions [27]. One can assume that the combination of these factors determines the final phase composition of the crystals. It should be, however, noted that in spite of the abovementioned differences between the specimens synthesized using the ceramic technology and those grown by crystallization from melt, single phase cubic crystals were obtained in the ZrO 2 -Sc 2 O 3 -Y 2 O 3 systems [28].
Thus, the predominant factor in the growth of 10Sc1CeSZ crystals determining their phase composition is the presence of cerium ions having variable valence (Cе 3+ or Cе 4+ ).

Concluding Remarks
(ZrO 2 ) 0.89 (Sc 2 O 3 ) 0.1 (CeO 2 ) 0.01 crystals up to 50 mm in length and with a cross-section of up to 25 mm were grown using directional melt crystallization. Cerium ions exist in the as-grown crystals in the form of Cе 3+ and Cе 4+ ions. Study of the crystals showed that directional melt crystallization increases not only the overall ceria concentration but also the Cе 3+ ion content along the length of the crystals. X-ray diffraction, Raman spectroscopy and transmission electron microscopy studies of the phase composition of the crystals showed that the as-grown crystals are mixtures of cubic and rhombohedral zirconia modifications. The rhombohedral phase has an inhomogeneous distribution along the crystals: the rhombohedral phase content is higher in the top part of the crystal than in the bottom part. Rhombohedral phase contains twins, which have a large size. Thus, unlike single-phase cubic 10Sc1CeSZ specimens synthesized using the ceramic technology, 10Sc1CeSZ crystallization from the melt did not promote cubic phase stabilization in the entire crystal bulk.

Conflicts of Interest:
The authors declare no conflict of interest.